Systematic Evaluation of Human Plasma Extracellular Vesicle Isolation and Physical Stability by Comparing Ultracentrifugation and Magnetic Bead-Based Methods

Abstract

Introduction:

Extracellular vesicles (EVs) are lipid bilayer particles released by all cell types, carrying cargos that reflect the cellular states of their origin. Recently, EVs are increasingly recognized as valuable biomarkers and therapeutic vectors in oncology, but their clinical translation is limited by variability in isolation methods and uncertainty regarding long-term storage physical stability.

Methods:

We systematically compared human plasma EVs isolated by ultracentrifugation (UC) or magnetic bead (MB)-based methods under immediate analysis, stable freezing storage, and repeated freeze–thaw conditions. The morphology and protein profiling of EVs were characterized by transmission electron microscopy (TEM) and Western blotting (WB), respectively. EV concentration, particle size, and zeta potential were quantified by particle size analyzer.

Results:

TEM and WB analyses of human plasma EVs confirmed the efficacy of both the UC and MB isolation methods. UC-isolated EVs are of high yield but low physical stability, featuring size reduction and a shift toward more negative zeta potential values after freeze–thaw cycles. Fresh UC-EVs displayed heterogeneous size profiles, whereas freeze–thawed samples shifted to a dominant peak, consisting of small particles with increased counts. Although lower in yield, MB-isolated EVs retained their physical stability across all conditions.

Conclusion:

MB-based EV isolation offers physical stability for standardized diagnostic workflows, whereas UC-based EV isolation provides high yield for discovery studies but vulnerable to freeze–thaw stress. These findings provide an evidence-based framework for selecting EV isolation and storage methods to match downstream applications, guiding the standardization of EVs workflows for future precision oncology and personalized medicine.

Keywords

biobank storage plasma liquid biopsies

Introduction

Extracellular vesicles (EVs) are lipid bilayer nanoparticles released by nearly all cell types, carrying proteins, nucleic acids, and lipids that reflect the physiological or pathological state of their origin. In oncology, EVs contribute not only to biomarker discovery in cancer diagnosis but also to disease progression, as they can remodel the tumor microenvironment, facilitate immune evasion, and promote metastasis.^1,2 Especially valued for their accessibility and specificity, plasma-derived EVs are considered as ideal noninvasive liquid biopsy biomarkers for real-time disease monitoring and patient stratification, potentially guiding precision medicine approaches.^3–5 Recent advances further underscore their translational potential in cancer. EVs have been engineered to deliver therapeutic cargos across biological barriers, including exosomes carrying drugs across the blood–brain barrier for glioblastoma and inhalable CAR-T EVs for lung cancer immunotherapy.^6,7 Thus, EVs have transitioned from ancillary research materials to key biobanking assets with a direct clinical impact.

However, reproducibility in EV-based studies remains a significant challenge and hinders the translation of EVs into reliable diagnostic and therapeutic tools.⁸ Pre-analytical variability, including sample collection, processing, isolation, and storage, can substantially impact EV yield, integrity, and downstream assay performance.^9,10 The MIBlood-EV initiative has recently addressed this issue by proposing a comprehensive, flexible reporting framework for pre-analytical variables and quality control of plasma and serum samples, aiming to harmonize EV research protocols and improve translational reliability.^11,12 Despite such guidelines, the influence of isolation methods and storage conditions under biobank-relevant scenarios remains insufficiently characterized. Among the available techniques, ultracentrifugation (UC) is the current gold standard. As a classical separation approach, UC offers significant cost-effectiveness and experimental flexibility. It can achieve the high-throughput processing of multiple samples without relying on specific affinity labels. Although UC is technically challenging and prone to variability in yield, it is still the preferred separation method in basic research.¹³ In contrast, the commercial magnetic bead (MB) method provides a simplified and rapid workflow, with high specificity for EV isolation, but its reagent and consumable costs are significantly higher compared with UC.¹⁴ The lack of systematic evaluation between these two approaches leads to inconsistencies and impedes the regulatory standardization of EV biobanking and translational application. Besides, EV storage physical stability is also critical for clinical and research reproducibility. Studies have shown that key markers can be preserved for months to years at −80°C; yet, repeated freeze–thaw cycles cause particle aggregation and the loss of functional cargo.^9,15 Therefore, the challenge extends beyond EV isolation efficiency from the perspective of biobanking.¹⁶

To bridge the knowledge gaps and enhance reproducibility, we addressed these essential questions by comparing EV isolation yield and physical stability isolated from UC and MB methods. Additionally, we explored the physical stability of EVs under varying storage conditions, including stable freezing storage and repeated freeze–thaw cycles. By addressing these foundational questions, our work aims to establish robust methodologies that could standardize EV isolation and storage workflows, paving the way for more reliable cancer biobanking and translational applications.

Materials and Methods

Collection of human plasma samples

The study methodologies conformed with the standards of the Declaration of Helsinki, and all aspects of this study were approved by the Ethics Committee of the First Teaching Hospital of Tianjin University of Traditional Chinese Medicine (TYLL2024[ZI]076). After an overnight fast, peripheral blood samples (5 mL) were collected using EDTA tubes from 54 consenting healthy adult donors (age: 23–49 years; female = 27), who constituted the study population. Plasma was separated by centrifugation within 2 hours of collection. Informed consent was obtained from all participants.

Isolation and storage of EVs

EVs were isolated from all 54 samples using both the UC and MB methods. EVs from each method were divided equally into immediate detection (ID), stable freezing storage (SFS), and cyclic freeze–thaw (CFT) groups (n = 18 per condition). For UC-based isolation, plasma samples underwent successive centrifugation steps: first at 3000×g for 10 minutes to remove cells and subsequently at 8000×g for 30 minutes to eliminate cellular debris. The supernatant was carefully collected after each step. The final clarification was ultracentrifuged at 110,000×g for 80 minutes using an Optima™ XPN-100 ultracentrifuge (Beckman Coulter). The resulting pellet was resuspended in 100 μL of PBS (pH = 7.4, G4202-500ML, Servicebio) to obtain the EV fraction. MB-based isolation was performed using the universal MB EV isolation kit (Tianjin Qiwu Technology Co., Ltd.) as per the manufacturer’s instruction. Briefly, MBs were incubated with plasma at a 1:1 ratio, and purified EVs were separated magnetically. EVs isolated by the two methods were either analyzed immediately after isolation or stored at −80°C in ultra-low-temperature freezers (ThermoFisher Scientific, Inc.) at our BioResource Research Institute. Samples from the SFS group were stored continuously at −80°C for 1 month. In contrast, samples from the CFT group underwent three freeze–thaw cycles over the same period, with temperature cycling between −80°C and 25°C. Subsequent measurements will use separate aliquots to avoid additional freeze–thaw cycles.

Transmission electron microscopy

To evaluate the size and shape of the EV particles, transmission electron microscopy (TEM) analysis was selected. The prepared EVs were fixed with 2% paraformaldehyde (Merck, Germany) and then placed on carbon-coated Formvar electron microscopy grids. As a part of the TEM study, EVs were fixed on a 300-mesh copper grid for 5 minutes, and negative staining was performed with a 2% (w/v) uranyl acetate solution. The examination and photography of the grids (Ted, Pella, Inc.) were conducted using a JEOL JEM 1011 transmission electron microscope (JEOL Ltd., Tokyo, Japan) equipped with an Orius SC200 camera (Gatan, Evry, France).

Protein concentration assay

The EV protein content was determined using the Pierce BCA Protein Assay Kit (Wuhan Doctoral Biotechnology Co., Ltd.) as per the manufacturer’s instructions. Absorption at 562 nm was measured with the Varioskan™ LUX Multimode Microplate Reader (ThermoFisher Scientific, Inc.), and protein concentrations were calculated according to the BSA standard.

Western blotting

Twenty micrograms protein of each sample was loaded for electrophoresis, followed by transfer to PVDF membranes (Millipore, ISEQ00010). Primary antibodies were as follows: CD9 (AF5139, Affinity Biosciences), CD63 (A19023, ABclonal), and TSG101 (DF8427, Affinity Biosciences). Protein bands were visualized on ChemiScope 6000 Chemiluminescence and a fluorescence imaging system (Clinx Science Instruments Co., Ltd.), and grayscale values were quantified using ChemStudio V3.1 software.

EV characterization by particle size analyzer

The EV concentration, size distribution, and zeta potential were analyzed using a Nanocoulter G analyzer (Resun Technology Co., Ltd.). UC- and MB-isolated EVs were diluted (1:200 and 1:20, respectively) in electrolyte buffer as per the manufacturer’s instruction. For the size/concentration, 200 μL of the sample was loaded into a nanopore chip, equilibrated for 5 minutes, and measured at 1.2 V (120 seconds, 10 kHz), excluding particles outside 60–200 nm. Polystyrene standards (60–200 nm, R² > 0.998) calibrated the system, with triplicate-derived geometric means reported (×10⁹ particles/mL). For the zeta potential, electrophoretic mobility was measured under gradient fields (0–15 V/cm), and potentials (mV) were calculated via the Smoluchowski equation (NanoSuite 2.1 software, hysteresis-corrected). Signal thresholds (100–1000 microseconds) excluded artifacts, with triplicate arithmetic means reported. Both protocols enabled the direct comparison of the isolation methods.

Statistical analysis

Statistical analyses were conducted using one-way ANOVA, followed by Tukey’s post hoc test for intergroup comparisons, with paired t-tests and correlation analyses performed where appropriate. Data processing and visualization were performed using GraphPad Prism 8, with statistical significance defined as p < 0.05. Further details are provided in the corresponding figure legends.

Results

Human plasma EV characterization

Human plasma EVs were isolated from 5 mL blood samples of the healthy donors using the UC and MB methods. Samples were categorized into three groups, namely ID, SFS, and CFT. TEM analysis revealed that both isolation methods using the ID, SFS, and CFT samples yielded vesicles with typical EV morphology—circular to elliptical or cup-shaped structures with bilayer membranes and low electron density. While most vesicles (30–150 nm) were individually dispersed, some aggregates were observed. Size distribution analysis demonstrated that UC-isolated vesicles showed greater heterogeneity (30–150 nm), whereas MB-isolated vesicles were more uniformly distributed within the 30–150-nm range (Fig. 1A and Supplementary Fig. S1). Western blot confirmation of EV markers (CD9, CD81, TSG101) validated the successful isolation of EVs by both the methods (Fig. 1B). These findings demonstrate that while both the UC and MB approaches effectively isolate EVs, they exhibit distinct size distribution profiles, with MB providing more consistent particle sizes within the classical EV range.

FIG. 1.

Characteristic analysis of human plasma EVs. (A) TEM of ID samples revealed the morphological characteristics of EVs isolated by the UC and MB methods. (B) Western blot analysis confirmed the expression of EV markers (CD9, CD63, and TSG101) in EV samples across all experimental groups (ID, SFS, and CFT groups). EV, extracellular vesicle; TEM, transmission electron microscopy; ID, immediate detection; UC, ultracentrifugation; MB, magnetic bead; SFS, stable freezing storage; CFT, cyclic freeze–thaw.

Concentration stability of EVs under different storage conditions

To explore the physical stability of EVs under different storage conditions, we systematically analyzed plasma EV concentrations from the healthy donors across different storage conditions employing a particle size analyzer. UC-isolated EVs exhibited significant concentration variations between the treatment groups; the ID group showed the lowest vesicle counts, while the SFS group demonstrated a 152% increase (p < 0.05), and the CFT group demonstrated a 334% increase (p < 0.001). CFT group processing further elevated concentrations by 72% versus SFS (p < 0.01) (Fig. 2A). MB-isolated vesicles followed similar trends, but the EV concentration differences among the three groups were not significant, reflecting that the storage conditions have a minimal impact on EVs obtained by the MB extraction method (Fig. 2B). Notably, UC consistently yielded higher absolute concentrations than MB across all groups: 16-fold in ID (p < 0.001), 34-fold in SFS (p < 0.001), and 68-fold in CFT (p < 0.001) (Fig. 2C). These findings demonstrate that UC-isolated vesicles exhibit greater sensitivity to freeze–thaw stress compared with MB-isolated counterparts.

FIG. 2.

Different extraction and preservation methods of human plasma EVs exhibit varying sensitivities to concentration. (A) UC-isolated plasma EVs exhibited high sensitivity to storage conditions, with the SFS group showing significantly increased concentrations versus the ID group. Further CFT amplified this effect. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, *p < 0.05; ID vs. CFT, ***p < 0.001; SFS vs. CFT, **p < 0.01). (B) In contrast, MB-isolated vesicles demonstrated greater physical stability, with no significant concentration changes in the ID, SFS, or CFT group. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, p > 0.05; ID vs. CFT, p > 0.05; SFS vs. CFT, p > 0.05). (C) UC consistently yielded higher EV concentrations than MB across all conditions, although MB isolation provided the superior preservation of vesicle integrity during storage. Statistical analysis was performed using one-way ANOVA with paired t-tests (ID, ***p < 0.001; SFS, ***p < 0.001; CFT, ***p < 0.001).

Particle size stability of EVs under different storage conditions

To explore the changes in the particle size of EVs under different storage conditions, we analyzed the plasma EV particle size of the healthy donors in different storage states. Experimental data showed that there were significant differences in the particle size of EVs extracted by the UC method among the treatment groups; the average particle size of EVs in the ID group was 141.50 ± 7.45 nm; after storage at −80°C for 1 month, the average particle size of EVs in the SFS group decreased by 25% (106.17 ± 4.23 nm, p < 0.001). Due to repeated freeze–thaw cycles, the CFT group further induced a decrease in particle size of 14% (91.11 ± 7.56 nm, p < 0.001) compared with the SFS group. Compared with the ID group, the overall particle size of the CFT group decreased by 35% (p < 0.001) (Fig. 3A). In contrast, the particle size stability of EVs extracted by the MB method was significantly better: the average particle size of the ID group was 87.56 ± 4.23 nm, while the SFS group only slightly decreased by 0.4% (87.22 ± 4.02 nm, p > 0.05), and the CFT group decreased by 3% (84.39 ± 3.50 nm, p > 0.05) compared with the SFS group. Compared with the ID group, the overall particle size of the CFT group decreased by 4% (p < 0.05) (Fig. 3B). Comparative analysis revealed that UC-isolated vesicles were consistently larger than MB-isolated counterparts: 62% larger in ID (p < 0.001), 22% in SFS (p < 0.001), and 8% in CFT (p < 0.01) (Fig. 3C). These findings demonstrate that vesicle size stability is critically influenced by both the isolation methodology and storage conditions. The MB method exhibited markedly better preservation of particle size integrity during stable freezing storage and freeze–thaw cycles, suggesting its greater suitability for clinical applications requiring standardized vesicle preservation.

FIG. 3.

Different extraction and preservation methods of human plasma EVs exhibit varying sensitivities in terms of the particle size. (A) UC-isolated plasma EVs demonstrated marked sensitivity to storage conditions, with the SFS group exhibiting significantly reduced particle size versus the ID group. The CFT group further decreased the particle size. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, ***p < 0.001; ID vs. CFT, ***p < 0.001; SFS vs. CFT, ***p < 0.001). (B) In contrast, MB-isolated vesicles showed greater stability, with only marginal particle size changes following stable freezing storage or freeze-thaw cycles. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, p > 0.05; ID vs. CFT, *p < 0.05; SFS vs. CFT, p > 0.05). (C) UC-isolated vesicles consistently displayed larger particle sizes compared with MB-isolated counterparts. Statistical analysis was performed using one-way ANOVA with paired t-tests (ID, ***p < 0.001; SFS, ***p < 0.001; CFT, **p < 0.01).

Zeta potential analysis of EVs under different storage conditions

We systematically examined the effects of different storage conditions on the zeta potential of plasma-derived EVs from the healthy donors. For clarity, all EV zeta potential values are negative; therefore, changes are described by the absolute value of magnitude, with an “increase” indicating a shift toward more negative values. The results revealed significant variations in the zeta potential among UC-isolated EVs across different treatment groups. The ID group showed a zeta potential of −4.11 mV, while the SFS group increased the absolute value by 35% (−5.56 mV, p < 0.05). The CFT group further increased the absolute value by 6% compared with the SFS group (−5.92 mV, p > 0.05). Compared with the ID group, the overall zeta potential of the CFT group increased by 44% (p < 0.001) (Fig. 4A). As determined by UC, the samples from both the SFS and CFT groups maintained comparable stability. In contrast, MB-isolated vesicles demonstrated superior stability: the ID group measured −6.94 mV, with only 8% reduced in SFS (−6.37 mV, p > 0.05), while 0.4% increased in CFT compared with the SFS group (−6.39 mV, p > 0.05). Compared with the ID group, the overall zeta potential of the CFT group reduced by 8% (p > 0.05) (Fig. 4B). Comparative analysis showed that UC-isolated vesicles consistently exhibited lower zeta potentials than MB-isolated counterparts: 62% higher in ID (p < 0.001), 14% in SFS (p < 0.05), and 8% in CFT (p < 0.001) (Fig. 4C). These results establish the zeta potential as a sensitive indicator of vesicle stability during storage. The MB method’s ability to maintain both membrane integrity and surface charge stability suggests its superior suitability for clinical applications requiring long-term vesicle preservation.

FIG. 4.

Different extraction and preservation methods of human plasma EVs exhibit varying sensitivities in terms of the zeta potential. (A) UC-isolated vesicles showed significant zeta potential reductions after stable freezing storage (SFS vs. ID) and cyclic freeze–thaw (CFT vs. ID) but no difference between the SFS and CFL groups. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, **p < 0.01; ID vs. CFT, ***p < 0.001; SFS vs. CFT, p > 0.05). (B) MB-isolated vesicles demonstrated superior stability with minimal changes. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, p > 0.05; ID vs. CFT, p > 0.05; SFS vs. CFT, p > 0.05). (C) UC-isolated vesicles consistently exhibited lower zeta potentials than MB-isolated counterparts. Statistical analysis was performed using one-way ANOVA with paired t-tests (ID, ***p < 0.001; SFS, *p < 0.05; CFT, **p < 0.01).

Particle size distribution of EVs under storage stress

To investigate the particle size dispersion of EVs under various storage conditions, we conducted a comprehensive analysis of plasma-derived EVs from the healthy donors. Our experimental results demonstrated significant variations in the zeta potential among UC-isolated EVs across the treatment groups. The ID group exhibited a D90-D10 size range of 88.22 nm, which increased by 5% after stable freezing storage (SFS: 92.67 nm, p > 0.05). The CFT group substantially reduced this range by 21% (73.33 nm, p < 0.001) compared with the SFS group. Compared with the ID group, the overall D90-D10 size range of the CFT group decreased by 17% (p < 0.001) (Fig. 5A). In contrast, MB-isolated vesicles demonstrated superior stability: the ID group averaged 70.11 nm, with an only 1% increase in SFS (71.17 nm, p > 0.05) and 16% decrease in CFT (59.44 nm, p < 0.05), compared with the SFS group. Compared with the ID group, the overall D90-D10 size range of the CFT group decreased by 15% (p < 0.05) (Fig. 5B). Comparative analysis revealed that UC-isolated vesicles consistently showed broader size distributions than MB-isolated counterparts: 20% larger in ID (p < 0.01), 32% in SFS (p < 0.001), and 3% in CFT (p < 0.001) (Fig. 5C). Size heterogeneity analysis (Fig. 5D–F) further demonstrated that MB isolation yielded more uniform vesicle populations across all conditions. These findings indicate that vesicle size dispersion is critically influenced by both the isolation methodology and storage conditions. While UC isolation produced more heterogeneous populations sensitive to freeze–thaw stress, MB isolation maintained superior size distribution stability during storage. The MB method’s ability to preserve vesicle homogeneity under various storage conditions suggests its greater suitability for clinical applications requiring standardized vesicle preparations.

FIG. 5.

Different extraction and preservation methods of human plasma EVs have varying sensitivities to particle size dispersion. (A) UC-isolated vesicles showed significant size distribution changes: the CFT group decreased the D90-D10 values versus the ID group and also decreased the D90-D10 values versus the SFS group. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, p > 0.05; ID vs. CFT, **p < 0.01; SFS vs. CFT, ***p < 0.001). (B) MB-isolated vesicles demonstrated greater stability, with marginal D90-D10 decreases after CFT treatment compared with the ID and SFS groups. Statistical analysis was performed using one-way ANOVA with Tukey’s post hoc test (ID vs. SFS, p > 0.05; ID vs. CFT, *p < 0.05; SFS vs. CFT, *p < 0.05). (C) UC-isolated vesicles consistently exhibited broader size distributions (higher D90-D10) than MB-isolated counterparts across all conditions. Statistical analysis was performed using one-way ANOVA with paired t-tests (ID, ***p < 0.001; SFS, ***p < 0.001; CFT, ***p < 0.001). (D–F) Heterogeneity analysis revealed that UC preparations maintained significantly greater size variability than MB isolates in all treatment groups (ID, SFS, and CFT). *p < 0.05, **p < 0.01,***p < 0.001.

Size–charge correlation in EVs across isolation and storage conditions

Comprehensive analysis of UC- and MB-isolated EVs from all treatment groups (ID, SFS, and CFT) revealed a significant correlation between the particle size and zeta potential in the ID group (R² = 0.6570, p < 0.0001). In the SFS group, there was a weak correlation between the vesicle size and potential (R² = 0.1103, p < 0.05), while in the CFT group, there was no strong correlation between the vesicle size and potential (R² = 0.0304, p > 0.05) (Fig. 6). These findings suggest that EV surface charge characteristics are intrinsically linked to the vesicle size, with smaller vesicles demonstrating greater charge stability. These results have important implications for clinical EV applications, as they suggest that the optimization of isolation protocols to yield smaller vesicle populations may enhance long-term storage stability through the maintenance of favorable surface charge characteristics.

FIG. 6.

Correlation analysis between the zeta potential and particle size of EVs in human plasma extracellular vesicles. (A) The zeta potential of EVs is highly correlated with the particle size in the ID group. Statistical analysis was performed using correlation analyses. (ID, ****p < 0.0001). (B) The zeta potential of EVs is weakly correlated with the particle size in the SFS group. Statistical analysis was performed using correlation analyses (SFS, *p < 0.05). (C) There is no correlation between the zeta potential and particle size of EVs in the CFT group. Statistical analysis was performed using correlation analyses (CFT, p > 0.05).

Discussion

In this study, our results reveal a clear trade-off between EV yield and storage physical stability. Our primary objective was to perform a systematic and practical comparison of commonly used plasma EV isolation workflows under clinically relevant biobanking conditions. UC and MB methods were employed for EV isolation, while density gradient centrifugation was excluded to avoid the potential selective loss of certain EV populations and to preserve the key phenotypic differences caused by freeze–thaw. UC yielded significantly higher EV concentrations than MB isolation, providing good cost-effectiveness when large quantities are required. However, we observed that UC-isolated EVs were highly sensitive to freeze–thaw stress, leading to significant declines in the mean particle size and zeta potential. In contrast, MB-isolated EVs retained their size integrity and surface charge remarkably well, even after multiple freeze–thaw cycles. This suggests that despite the low yield, the immunoaffinity capture method provides clear advances in preserving the vesicle structure.¹⁷

To note, the EV size distribution is closely dependent on factors including cell origins, isolation methods, and storage conditions. One important observation in this study is that fresh UC-isolated EVs exhibited multiple size peaks across the full range, indicating the heterogeneous populations of plasma-derived vesicles. The variation likely reflects EVs originating from different cell types, biogenesis pathways, and functional states.¹⁸ The increased heterogeneity of EVs following UC primarily arises from the mechanical damage induced by high shear forces, the aggregation of vesicles under centrifugal stress, and the co-sedimentation of nonvesicular impurities, including protein aggregates. These findings align with emerging perspectives that EVs should be viewed not as discrete subtypes but as a continuum of vesicles with diverse sizes, membrane compositions, and cargos.¹⁹ In contrast, after repeated freeze–thaw cycles, these EVs showed a single dominant peak at the lower end, with an increased total particle count, suggesting the fragmentation of larger EVs into smaller particles due to storage stress.

However, a Nanocoulter G particle size analyzer cannot distinguish between intact vesicles and nonvesicular debris. This shift toward smaller particles raises concerns that these fragments may have lost structural integrity. In addition, different EV isolation methods also contribute to the size heterogeneity.²⁰ Using fresh plasma, UC-isolated EVs showed continuous size distribution, whereas MB-isolated EVs exhibited a single peak in size. Moreover, this instability of UC-isolated EVs further led to the lack of functions in certain cases, making this method less practical for EV-based therapies.²¹ Our results underscore the importance of tailoring EV isolation strategies to downstream applications. For biobanking and clinical assay development, the long-term preservation of EV membrane integrity may outweigh raw particle recovery.^22,23

The reported zeta potential values of EVs show considerable variability depending on the isolation method. Meanwhile, using the same instrument for detection, it was found that the negative zeta potential of serum EVs extracted from UC was −15mV, which is lower than the threshold of −20 mV generally associated with stable colloidal dispersions.²⁴ In our study, plasma EVs demonstrated a zeta potential ranging from −2 to −10 mV, which is also lower. This reduced surface charge indicates a diminished energy barrier against aggregation, potentially compromising the physical stability of the EV preparations. This phenomenon may be attributed to the formation of a protein corona or specific conditions inherent to our isolation protocol. The MB method effectively preserved the membrane surface charge of EVs. In contrast, while UC-isolated samples from the SFS and CFT groups retained considerable stability, their surface charge was notably elevated compared with the ID group. Overall, these findings suggest that the MB method offers superior performance in maintaining the native EV surface charge compared with UC. Besides size heterogeneity, our data demonstrated a direct relationship between the particle size and zeta potential, which reflects the surface charge of EVs. Larger particles tended to display higher zeta potentials, whereas smaller particles had more negative potentials, indicating better physical stability.

A further strategic consideration is whether to store samples as isolated EVs or to freeze plasma before EV isolation in the context of cancer biobanking. Freezing plasma preserves EVs in their original biofluid environment, fitting the long-term biobanking purpose when research objectives are not yet clear. However, isolating EVs from thawed plasma can potentially introduce lipoprotein or protein contaminants, leading to reduced purity and increased noise, not properly reflecting the actual biologically relevant signals.²⁵ In contrast, storing purified EVs allows immediate downstream analysis and can preserve morphology and cargo stability. Moreover, emerging genetically engineered EVs usually require a substantial amount of time and effort to produce, thus, it is more efficient to prepare them in batches and store as isolated EVs rather than repeatedly manufacturing fresh EVs for each use from their original source.²⁶ However, repeated freeze–thaw cycles can lead to EV aggregation and membrane disruption.²⁷ Therefore, further investigations will be needed in optimizing storage conditions, including the freezing buffer, to preserve the integrity and function of EVs.¹⁵ Trehalose stabilizes lipid bilayers and prevents ice crystal formation by forming a protective glassy matrix during freezing and thawing, thus significantly reducing vesicle aggregation and preserving functional EV markers.²⁸ DMSO has membrane toxicity and is difficult to remove, while the high viscosity of glycerol hinders particle size analysis and molecular experiments. Therefore, the selection of protective agents needs to be carefully weighed against their potential impact on subsequent analysis. Interestingly, EVs themselves have been used as cryoprotective additives to preserve sperm cell viability during freezing.²⁹

One limitation of our current study is the use of plasma from healthy individuals instead of patients. While it is critical to establish the baseline, this does not fully represent clinical conditions in diagnostic scenarios. Now that the baseline has been set up, future work will extend to patient-derived samples under real-world conditions. In addition, this study primarily evaluates the physical stability of plasma-derived EVs, while the integrity of EV cargo remains to be addressed in future functional and molecular analyses.

In conclusion, our findings offer critical evidence that supports EV isolation and storage approaches under different conditions. MB-based EV isolation offers physical stability for standardized diagnostic workflows, whereas UC provides high yield for discovery studies. These findings provide an evidence-based framework for selecting EV isolation and storage methods to match downstream applications, guiding the standardization of EV workflows for future cancer diagnostics, therapeutics, and large-scale biobanking.

Footnotes

Author Disclosure Statement

No conflicting financial interests exist.

Funding Information

This work was supported by the National Natural Science Foundation of China (No. 81774059) and Tianjin Nankai District Traditional Chinese Medicine Inheritance and Innovation Development Demonstration Pilot Project (No. 20240204006).

Supplemental Material

References

Ljungstrom

, Oltra

. Methods for extracellular vesicle isolation: Relevance for encapsulated miRNAs in disease diagnosis and treatment. Genes (Basel), 2025; 16(3):330.

Huber

, Kors

, Schütt

, et al. The role of plasma-derived small extracellular vesicles in pre-metastatic niche formation through modulation of macrophages in head and neck squamous cell carcinoma. Br J Cancer, 2025; 133(1):121–130.

Rios de Los RiosResendiz

, Herrmann-Sim

, Wilkesmann

, et al. A translational protocol optimizes the isolation of plasma-derived extracellular vesicle proteomics. Sci Rep, 2025; 15(1):24292.

Trauner

, Bernath-Nagy

, Kalinyaprak

, et al. Isolation, characterization, and proteomic analysis of plasma-derived extracellular vesicles for cardiovascular biomarker discovery. J Vis Exp, 2025(215):e67083.

Sun

, Zhang

, et al. Exosomal CCT6A secreted by cancer-associated fibroblasts interacts with beta-catenin to enhance chemoresistance and tumorigenesis in gastric cancer. Adv Sci (Weinh), 2025; 12(38):e06674.

Han

, Huang

, Li

, et al. Engineering exosome membrane disguised thermal responsive system for targeted drug delivery and controlled release across the blood-brain barrier. Mater Today Bio, 2025; 32:101656.

Zheng

, Zhu

, Tang

, Li

, Jiang

, Huang

. Inhalable CAR-T cell-derived exosomes as paclitaxel carriers for treating lung cancer. J Transl Med, 2023; 21(1):383.

, Liu

, Yu

, et al. Revisiting the advances and challenges in the clinical applications of extracellular vesicles in cancer. Cancer Lett, 2024; 593:216960.

Sivanantham

, Jin

. Impact of storage conditions on EV integrity/surface markers and cargos. Life (Basel), 2022; 12(5):697.

10.

Taha

. Plasma versus serum for extracellular vesicle (EV) isolation: A duel for reproducibility and accuracy for CNS-originating EVs biomarker analysis. J Neurosci Res, 2023; 101(11):1677–1686.

11.

Lucien

, Gustafson

, Lenassi

, et al. MIBlood-EV: Minimal information to enhance the quality and reproducibility of blood extracellular vesicle research. J Extracell Vesicles, 2023; 12(12):e12385.

12.

Gustafson

, Nieuwland

, Lucien

. MIBlood-EV: An online reporting tool to facilitate the standardized reporting of preanalytical variables and quality control of plasma and serum to enhance rigor and reproducibility in liquid biopsy research. Biopreserv Biobank, 2025; 23(1):62–64.

13.

Zhang

, Jeppesen

, Higginbotham

, Franklin

, Coffey

. Comprehensive isolation of extracellular vesicles and nanoparticles. Nat Protoc, 2023; 18(5):1462–1487.

14.

, Tsantilas

, Park

, et al. Enrichment of extracellular vesicles using Mag-Net for the analysis of the plasma proteome. Nat Commun, 2025; 16(1):5447.

15.

Görgens

, Corso

, Hagey

, et al. Identification of storage conditions stabilizing extracellular vesicles preparations. J Extracellular Vesicle, 2022; 11(6):e12238.

16.

Mora

, Alvarez-Cubela

, Oltra

. Biobanking of exosomes in the era of precision medicine: Are we there yet? Int J Mol Sci, 2015; 17(1):13.

17.

Fernandes

, Ruiz

, Bezemer

, Detmers

, Hermans

, Peixoto

. Targeted isolation of extracellular vesicles from cell culture supernatant using immuno-affinity chromatography. Sep Purif Technol, 2025; 358:130312.

18.

, Li

, Tong

, et al. Systematic evaluation of isolation techniques and freeze-thaw effects on plasma extracellular vesicle heterogeneity and subpopulation profiling. J Extracell Biol, 2025; 4(6):e70058.

19.

Manno

, Bongiovanni

, Margolis

, Bergese

, Arosio

. The physico-chemical landscape of extracellular vesicles. Nat Rev Bioeng, 2024; 3(1):68–82.

20.

Carney

, Mizenko

, Bozkurt

, et al. Harnessing extracellular vesicle heterogeneity for diagnostic and therapeutic applications. Nat Nanotechnol, 2025; 20(1):14–25.

21.

Fernandez-Saez

, Bravo

, Pena

, Garcia

. The proteome of circulating extracellular vesicles and their functional effect on platelets vary with the isolation method. Sci Rep, 2025; 15(1):20490.

22.

Weber

, Ritter

, Han

, et al. Development of a sampling and storage protocol of extracellular vesicles (EVs)—Establishment of the first EV biobank for polytraumatized patients. Int J Mol Sci, 2024; 25(11):5645.

23.

Torres

, Bernardo

, Sanchez

, Morato

, Solana

, Carrillo

. Comparing the proteomic profiles of extracellular vesicles isolated using different methods from long-term stored plasma samples. Biol Proced Online, 2024; 26(1):18.

24.

Yang

, Guo

, Fang

, et al. Quality and efficiency assessment of five extracellular vesicle isolation methods using the resistive pulse sensing strategy. Anal Methods, 2024; 16(32):5536–5544.

25.

Ahmadian

, Jafari

, Tamadon

, Ghaffarzadeh

, Rahbarghazi

, Mahdipour

. Different storage and freezing protocols for extracellular vesicles: A systematic review. Stem Cell Res Ther, 2024; 15(1):453.

26.

Ren

, Liu

, Abdollahi

, Tavasolian

. Genetically engineered exosomes as a potential regulator of th1 cells response in rheumatoid arthritis. Biopreserv Biobank, 2023; 21(4):355–366.

27.

Bosch

, de Beaurepaire

, Allard

, et al. Trehalose prevents aggregation of exosomes and cryodamage. Sci Rep, 2016; 6:36162.

28.

Murray

, Kilbride

, Gibson

. Trehalose in cryopreservation. Applications, mechanisms and intracellular delivery opportunities. RSC Med Chem, 2024; 15(9):2980–2995.

29.

Shamsi

, Jozani

, Asadpour

, Rahbar

, Taravat

. Seminal plasma-derived exosome preserves the quality parameters of the post-thaw semen of bulls with low freezeability. Biopreserv Biobank, 2025; 23(4):364–373.