Size-based separations for EVs include chromatography and filtration (Figure 2). Size exclusion chromatography (SEC) is a chromatography-based method that has been used to fractionate molecules based on their hydrodynamic size or molecular size (Sorci and Belfort, 2014). SEC has a wide range of applications and can isolate EVs from various biofluid samples (Williams et al., 2023). It is a high-recovery method for isolating EVs based on size separation, although it has some limitations related to lack of specificity due to the presence of contaminants with similar size profiles (Lozano-Ramos et al., 2015). In addition, it can successfully remove serum albumin from serum samples, reducing the probability of protein contamination. However, other proteins in a similar size range to EVs, such as lipoproteins, will inevitably be co-isolated, which is a limitation of this method (Koster et al., 2021). Ultrafiltration (UF) is another popular separation technique based on size. It is a promising method for isolating EVs due to its fast and simple operation, although it has some challenges associated with scaling up, particularly related to molecular weight and size considerations (Kim et al., 2016). This easy-to-use technique with relatively short separation times is often used to separate EVs from relatively dilute biological samples (Zhang et al., 2021). However, it has not addressed the problems of protein contamination and EVs membrane loss. There are also methods such as tangential flow filtration (TFF) and flow field-flow fractionation (FFF), all of which have limitations, such as the limited volume of samples to be processed. Kim et al. developed a novel dual-cyclic tangential flow filtration (dcTFF) chip to enable the efficient and selective isolation of highly purified, concentrated EVs (Figure 2C) (Kim et al., 2021). The proposed dcTFF system utilizes two membranes connected to a peristaltic pump to facilitate continuous circulation of the sample stream flowing to the membrane. The dual cyclic filtration approach used in the dcTFF system can isolate the specific EVs size range of 30–200 nm in one step, which is superior to direct filtration (DF) and single cyclic TFF (scTFF) methods. FFF is a versatile method for isolating EVs that offers a wide variety of eluent options and broad separation capabilities, but it requires specialized fractionation equipment and long duration due to considerations related to molecular weight and size (Sitar et al., 2015).

Ion exchange describes a specific chemical process in which EVs in a biofluid to be tested, like a cell culture medium, is exchanged with other ions having a similar charge. Its application is limited by its use in matrices containing complex charged biomolecules (Wang et al., 2023), so it is often used to optimize other separation techniques and further improve the purity of EVs. Another method is to use electrophoresis and dielectrophoresis (DEP) to separate EVs and their subgroups based on their electrophoretic mobility (Morani et al., 2020). For instance, a microfluidic-based on-chip immunoelectrophoresis technique demonstrated in Figure 3A can be applied for effective differential protein expression profiling of individual EVs (Akagi et al., 2015). This method involves electrophoretic experiments on EVs derived from MDA-MB-231 human breast cancer cell culture supernatant using a microcapillary electrophoresis chip and a laser dark-field microimaging system. In Figure 3B, Chen et al. have introduced a straightforward and user-friendly DEP technique for isolating EVs that offers higher recovery efficiency (>83%) and purity compared to the traditional method of ultracentrifugation (UC) (Davies et al., 2012). The DEP chip can realize isolation within 30 min, much faster than the 8 h duration required for the UC procedure. There are several different categories of electrophoresis, each with its advantages and limitations. Some of the most common types of electrophoresis are gel electrophoresis and capillary electrophoresis. This method can be implemented by imposing a microfluidic chip with a significantly high recovery rate (Davies et al., 2012).

On the surface of EVs, proteins, lipids, and polysaccharides are exposed as potential ligands available for selection by antibodies, lipid-binding proteins, and lectin for molecule-specific interactions with molecules that can isolate EVs. Immunoaffinity-based methods involve antibodies targeting specific EVs surface markers such as CD63, CD9, and CD81. In direct methods, EVs are captured directly by immobilized antibodies, while in indirect methods, EVs are first labeled with antibodies and then captured by immobilized secondary antibodies (Duijvesz et al., 2015). Immunoaffinity-based methods offer high specificity for EVs, and different subpopulations of EVs can be isolated by targeting different surface markers. However, these methods can be costly due to the high price of antibodies, and the isolation efficiency may alter depending on the quality of the antibody used. Immunoaffinity assays for EVs isolation can use submicron-sized immuno-MBs to improve specificity, sensitivity, and yield. This enhancement results from a larger surface area, no sample volume constraints, and a near-homogeneous capturing process (Liangsupree et al., 2021).

Another affinity method that is commonly used for EVs isolation is lectin-based capture. Lectins are carbohydrate-binding proteins that can bind to specific sugar moieties on the surface of EVs. This method involves using lectin-coated beads or columns to capture the EVs. Plant-derived lectins such as concanavalin A (ConA) and wheat germ agglutinin (WGA) have been used to isolate EVs. ConA is a lectin that binds to the -D-mannose and -D-glucose residues found on the surface of EVs (Krafft et al., 2017; Lennon et al., 2019). Conversely, WGA binds to the N-acetylglucosamine and sialic acid residues found on the surface of EVs. The advantage of lectin-based capture is that it can capture a broader range of exosomes, as lectins can bind to various sugar moieties on the EVs surface. However, the disadvantage of lectin-based capture is that it may also capture non-specific particles that express similar sugar moieties.

Chemical affinity-based methods involve using chemicals that interact with specific EVs components, such as phosphatidylserine and cholesterol. Phosphatidylserine is a phospholipid exposed on the outer membrane of many EVs, making it a common target for chemical affinity-based isolation methods. One example of a chemical that binds to phosphatidylserine is lactadherin, a protein with a high affinity for this phospholipid. Cholesterol is another component abundant in the membranes of many types of EVs. Cyclodextrins are cyclic sugars that have been shown to bind to cholesterol and can be used for EVs isolation. Lactadherin and cyclodextrin can be immobilized on a solid support to isolate EVs from biological samples (Hansen et al., 2020; Zidovetzki and Levitan, 2007). These methods are simple and effective, and many EVs can be isolated. However, the specificity can be limited, leading to the isolation of non-specific contaminants.

Besides the conventional isolation methods, there are also commercial kits available for EVs isolation, catering to different research needs and sample types (e.g., plasma, serum, cell culture media). These kits vary in their underlying isolation principles, including size exclusion, ultracentrifugation alternatives, precipitation, and immunoaffinity. Table 2 lists some commonly used kits for EVs isolation.

MBs are small magnetic particles composed of a core usually made of a metal compound and functional surface molecules. MBs have been applied in many biomedical engineering applications such as drug delivery, real-time monitoring, and cell separation due to their properties like high surface-to-volume ratio, high specificity, high sensitivity, and so on. MBs can be divided into different classifications according to different criteria. For example, based on the surface functionalization, the MBs can be classified into antibody-coated beads, streptavidin-coated beads, and aptamer-coated beads. Based on the size, the MBs can be divided into micro-sized beads and nano-sized beads. According to the main composition, the MBs can be classified into metal-based cores and polymer-based cores. In the following part, the synthesis methods, surface modification strategies are introduced as the basic preparation work before using MBs to separate EVs.

Based on synthesis process mechanisms, the synthesis method can be divided into three categories: physical methods, chemical methods, and biological methods. From the view of the change in size, MBs synthesis methods can be classified into top-down methods and bottom-up methods. Commonly, the physical methods are top-down methods since the synthesis process is to reduce the precursor material size through mechanical action, and biological and chemical synthesis are included in bottom-up method since these synthesis methods are to gather the smaller particles together to achieve the desired size (Nicule et al., 2022). Table 3 summarizes some common synthesis methods for MB.

Surface modification is a crucial process that enhances the functionality of MBs by altering their surface properties to improve anti-fouling, stability, specificity, and performance in various applications. It involves the introduction of chemical or biological moieties onto the bead surface, enabling selective interactions with target molecules such as EVs. There are several approaches to surface modification, broadly classified into physical adsorption, covalent bonding, and bioaffinity interactions. Physical adsorption relies on electrostatic or hydrophobic interactions to attach biomolecules, but may suffer from weak stability. Covalent bonding creates strong and durable linkages between the bead surface and functional molecules, such as antibodies or polymers, ensuring long-term stability. Bioaffinity interactions, such as biotin-streptavidin or protein A/G conjugation, provide highly specific and reversible binding, making them useful for controlled capture and release applications.

The performance of MBs in EV isolation is fundamentally dictated by their synthesis route and surface functionalization, which collectively influence capture efficiency, purity, and specificity. In this chapter, we discuss how synthesis and modification methods influence MB-based EVs isolation. Furthermore, we compare the effects of these various synthesis and modification methods on EVs isolation efficiency in Table 5 for clearer visualization (Jiang et al., 2020).

MBs are often chosen as the solid media when conjugate ligands like antibody to develop the affinity-based isolation technique. Take the antibody-based affinity method as an example. Briefly, the procedure of this isolation method is that the antibody-conjugated MBs are first incubated with the biological fluid. After the antibody captures the EVs, discard the supernatant and collect the EV-MB conjugates. The next step is to resuspend the EV-MB conjugates and incubation for EVs elution. Finally, the MBs are removed by magnetic separation, and the EVs are collected in the supernatant (Chen et al., 2020).

Xu’s research team introduced a MB-based method for EVs isolation, which achieves twice the yield and comparable purity of harvested EVs compared to ultracentrifugation (UC) (Fang et al., 2021). This method is called MB-mediated selective adsorption strategy. The brief procedure is shown below. First, incubate the mixture of MBs, stock solution, PBS, and biological fluid. After the incubation, discard the supernatant and disperse the MBs-EV conjugates for EVs elution. The last step is to remove the MBs by magnetic separation, and the harvested EVs remain in the supernatant. Although this isolation method generates a much higher yield than the UC method, there are still improvements that could be optimized to improve the yield.

Compared to solo MBs EVs isolation, the MB-based microfluidic isolation significantly improves the EVs recovery rate, reduces the reagent volume, and cuts the operation time. The basic principle is shown in Figure 5A. Microfluidic designs create controlled laminar flow and tunable shear forces, which enhance the probability of contact between EVs and functionalized MBs by minimizing turbulence and promoting uniform trajectories. Within confined microchannels, the spatial distribution of beads can be precisely optimized, thereby reducing aggregation and ensuring a higher effective surface area for vesicle capture. When coupled with external magnetic fields, the microfluidic environment enables dynamic manipulation of bead positioning and movement, allowing selective retention of target EVs while unbound impurities are washed away efficiently. This synergy between microfluidics and MBs not only improves capture efficiency and specificity but also enhances reproducibility and scalability by enabling continuous, automated, and high-throughput separation. Collectively, these mechanistic insights highlight how the convergence of fluid dynamics and magnetic control provides a powerful framework for advancing EV isolation technologies.

Ferraro’s team developed an automated droplet microfluidic platform (Figure 5B) that integrates on-chip droplet incubation with in-line magnetic extraction to enable efficient and reproducible immunomagnetic isolation of EVs from large-volume samples (up to 2 mL). The captured EVs were validated by protein assays, nanoparticle tracking, and miRNA analysis. This strategy demonstrated shorter incubation times and substantially higher capture efficiency than conventional batch protocols (Meggiolaro et al., 2024). We recently developed an integrated biochip platform (Figure 5C), which integrates in situ electrochemical protein detection with on-chip antifouling-immunoMBs modified with CD81 antibodies and zwitterion molecules, enabling efficient isolation and detection of neuronal EVs. The capability of the system to isolate common EVs and detect neuronal EVs associated with Parkinson’s disease in human serum is successfully demonstrated, using the transmembrane protein L1-cell adhesion molecule (L1CAM) as a target biomarker. For the first time, this study discovered that the level of L1CAM/neuronal EVs particles in serum could serve as a reliable indicator to distinguish Parkinson’s disease from control groups with AUC = 0.973. This approach has the potential to advance the diagnosis and biomarker discovery of various diseases (Li et al., 2024). Besides EVs isolation, microfluidic chip can also help with EVs signal output. Zheng’s team implemented a droplet digital ExoELISA (Figure 5D) in which individual exosomes were captured on antibody-functionalized MBs and labeled with an enzyme-tagged detection antibody. The capture exosomes on MBs were co-encapsulated with substrate into microdroplets to produce fluorescent readouts that were enumerated and converted via Poisson statistics to yield absolute concentrations of protein-specific exosomes (limit of detection ≈5 exosomes μL⁻¹). The clinical feasibility demonstrated by quantifying GPC-1–positive exosomes in breast cancer patient samples (Liu C. et al., 2018).

Mai’s team has refined a MB-based method that does not rely on antibodies. They initially optimized it in batch processing and later adapted it for use in microfluidic droplet systems (Morani et al., 2022). They greatly enhanced the reproducibility of EVs recovery and eliminated positive false bias by incorporating a washing step and optimizing the protocol. This approach significantly increases the recovery rate (up to 50%) while reducing sample and reagent volumes (by more than 10 times) and operation time (by three times) compared to the traditional batch mode.

Several studies have demonstrated the successful application of integrated microfluidic chips to isolate and detect EVs in various diseases. One example is the successful implementation of an integrated microfluidic chip to isolate and detect prostate-specific membrane antigen (PSMA)-positive EVs in urine samples from prostate cancer patients (Zhao et al., 2016). The chip utilized a combination of PSMA antibody functionalized MBs for EVs capture, followed by RNA extraction and reverse transcription for PSMA mRNA detection, showing higher sensitivity than traditional ELISA methods in prostate cancer monitoring. Another study by Jeong et al. in Figure 6 shows that the chip integrated multiple functions into a single device, including EVs capture, washing, elution, and on-chip detection (Jeong et al., 2016). The chip consists of a microchannel coated with anti-CD63 antibody, which precisely captures EVs expressing CD63 on their surface. The captured EVs can be subsequently washed and eluted from the chip using a buffer solution. The eluted EVs can be directly analyzed on the chip using surface plasmon resonance (SPR) or fluorescence imaging. In addition to their diagnostic applications, integrated microfluidic chips have also been used in research on the role of EVs in diseases such as non-small cell lung cancer (NSCLC). Chen et al. developed a microfluidic chip for isolating and profiling EVs in NSCLC patients (Chen et al., 2023). The chip utilized a combination of immunoMB-based solation and surface-enhanced catalytic hairpin assembly imaging for EVs profiling. They identified a unique EVs signature in NSCLC patients that could serve as a biomarker for early-stage diagnosis.

Magnetic flow cytometry is an advanced method that leverages magnetic particles and sensors integrated into microfluidic structures for the dynamic analysis of fluid samples, which could be utilized to isolate EVs. It combines flow cytometry principles with magnetic separation to isolate specific cells like EVs. MBs are conjugated to antibodies targeting specific cell surface markers, so when these cells pass through a magnetic field, they are deflected and can be separated from the rest of the sample due to the interaction between the magnetic particles and the magnetic field, allowing for the separation, counting, and characterization without the need for optical detection methods (Bha et al., 2018).

Specific to the research mentioned above, Xu’s group discovered that in their experiment, when polystyrene-based MBs are used as adsorption carriers for isolating EVs, the yield and purity are significantly lower compared to using silica-based MBs. This discrepancy might be attributed to the stronger adsorption properties of the polystyrene surface on proteins and EVs, driven by hydrophobic interactions. Additionally, EVs adsorbed onto polystyrene-based MBs are more difficult to elute compared to those on silica-based MBs. Also, through a controlled experiment, they found that with the same functional group (hydroxyl group), MBs with a diameter of 1 μm have a higher purity, and with the same diameter of 1 μm, the types of functional group on the MBs’ surface have little effect on the isolation purity. However, increasing the number of beads to increase the binding surface area could increase the isolation efficiency (Fang et al., 2021).

What is more, besides the optimization within one isolation strategy mentioned above, different isolation platforms also differ from others in efficiency. For instance, microfluidic isolation has advantages like lower volume, shorter incubation time, higher EVs recovery rate and purity over the conventional sole MBs in-batch isolation. For microfluidic isolation, Mai’s team conducted optimizations in several key aspects of their device. They explored the effects of varying MB concentrations (ranging from 0.5 to 2 mg/mL), the chemical surface of the beads (with carboxylic or silane groups), the incubation temperature for EVs capture (spanning from 4 °C to 25 °C), and PEG concentrations (between 5% and 15% w/v). They found that the MBs capture performance for the carboxylic group achieved EVs yield of 61%, which was significantly higher than silane group MBs with a yield of less than 5%, while the reaction temperature from 4 °C to 25 °C did not affect the EVs yield a lot. For the concentration of MBs, a too-high concentration would cause poor recirculation of the device since the beads would cluster within the channel. A high concentration of PEG would increase the viscosity, which limits the MBs’ EVs capture ability. Finally, the highest EVs recovery rate of 25% ± 8% was achieved with a PEG concentration of 5% and a carboxylic MBs concentration of 1.5 mg/mL with 1 h incubation at 25 °C. Besides, the moving pattern of droplets also affects the recirculation of beads in the microfluidic device. They proposed that the U-shaped move pattern could achieve the EVs recovery rate of 39% ± 3% (Morani et al., 2022). The development of integrated microfluidic chips for separating and detecting EVs has been an active area of research in recent years. Integrated microfluidic chips can combine multiple steps, such as EVs capture, washing, and detection, into one device. This allows for efficient isolation and analysis of EVs from biofluids with minimal sample and reagent consumption. In the meantime, the integrated microfluidic chip can provide an airtight space to avoid environmental contamination and achieve automatic operation.

To improve the MB-based EVs isolation efficiency, there are several factors to take into consideration such as incubation time, temperature, the level of surface marker expression, and the concentration of target vesicles are important considerations. Additionally, the nature and state of the target molecule or structure, the characteristics of the antibody–antigen interaction, the type and concentration of the sample, and the ratio of beads to target molecules can all significantly impact the success of magnetic separation (Sioud, 2015). The detailed optimization depends on the specific reagents and methods used in the isolation process.

One strategy to improve EV isolation efficiency is to integrate isolation with downstream elution and analysis. Although there are commercial kits for MB-based EVs isolation like ExoCAS-2, which is easy to operate, there are no downstream lysis and analysis, which do not provide an elution process to recover the intact EVs, and the lack of universal EVs markers is another disadvantage for these strategies to ensure total capture of all EVs (Morani et al., 2022).

Another strategy to improve the EVs isolation efficiency is to build a multiplex bead-based platform that could achieve simultaneous capture of different EVs subpopulations. Stefan Wild’s group proposed a new multiplex bead-based platform has been introduced to analyze up to 39 different surface markers in a single sample (Koliha et al., 2016). This platform integrates capture antibody beads with fluorescently labeled detection antibodies. It enables the analysis of EVs that express surface markers recognized by both types of antibodies. Finally, they demonstrate that MBs can separate EVs mixtures and analyze them with the multiplex platform. The multiplex platform generates 39 different bead populations, each population of beads is conjugated with a distinct capture antibody, which specifically recognizes and binds to EVs that carry the corresponding antigen. With the present strategies to improve the EVs isolation efficiency, there are more advances could be achieved in the future. For example, the MBs used in the isolation can be further functionalized, introducing new ligands and using multi-marker strategies can significantly increase the isolation efficiency. Also, developing automation and high-throughput platforms like microfluidics and flow cytometry can improve efficiency by combining the isolation process with the downstream analysis.

In the future, more hybrid techniques platforms should be developed to utilize MBs with other conventional isolation methods to create new possibilities for improving the EVs isolation efficiency. Another important demand is to standardize the protocol for MBs-based EVs isolation techniques, achieving reproducible and repeatable protocols for different isolation platforms to be applied in clinical applications.

One of the major challenges in EVs isolation using MBs is non-specific binding. Non-targeted biomolecules such as proteins, lipids, and other cellular debris would adhere to the beads ‘surface, leading to contamination, reducing the purity of the isolated EVs, and potentially interfering with downstream analyses. These non-specific bindings occur due to inadequate antifouling coatings or improper surface functionalization of the beads. Polymer-coated MBs, such as PEGylated or zwitterionic-coated beads, can overcome the challenge of nonspecific binding, but they may also alter the biological properties of EVs. For example, PEGylation can mask surface markers on EVs, making them less recognizable by target cells or tissues (Xing et al., 2017). Moreover, the high hydrophilicity of polymers could interfere with the binding of EVs to their natural targets in vivo, thus limiting their applications in drug delivery and therapeutics (Lowe and McCormick, 2002). While surface modifications with anti-fouling coatings can reduce these effects, achieving complete selectivity remains difficult. To address this, we propose several potential directions for resolution. Firstly, engineering surfaces with optimal spatial organization, controlling the density and chain length of the anti-fouling molecules to minimize steric interference, determine the optimal spatial arrangement and stoichiometric ratio between anti-fouling molecules and capture ligands. Secondly, developing stimuli-responsive coatings (e.g., pH-, temperature-sensitive) that can dynamically modulate the conformation of the surface molecules to facilitate capture and enhancing purification. For example, MBs are equipped with a dynamic coating of anti-fouling molecules in the buffer, and the coating will switch to a capture-active state when MBs are placed in sample media via a structural reconfiguration, which unmask the binding sites for high-efficiency EVs capture. In summary, MB-based EV isolation faces a trade-off between reducing nonspecific bindings and EVs isolation efficiency. Future efforts are needed to solve this situation by optimizing antifouling ligands spatial organization and developing stimuli-responsive coatings to dynamically balance capture efficiency with isolation purity.

Reproducibility is a critical concern in EVs research. Even slight variations in isolation protocols can lead to inconsistent results. Factors such as differences in bead surface chemistry, incubation times, buffer compositions, and sample processing conditions can affect the isolation results. Additionally, batch-to-batch variability of antibodies is also the key factor to affect reproducibility results. Currently, there is no universally accepted standard for MB-based EVs isolation due to the different beads recipes and isolation protocols, making it difficult to compare results across studies. Therefore, the development of standardized protocols, reference materials, and quality control measures is essential to improve reproducibility. Key criteria must include capture yield (e.g., % recovery of spiked EVs), purity (e.g., ratio of target EVs to co-isolated proteins), and specificity (e.g., enrichment of subpopulations in complex samples). Establishing standardized protocols and reference materials will enable objective comparison and reproducibility across studies and platforms. To conclude, reproducibility in EVs isolation is hindered by protocol variations and material inconsistencies. Establishing standardized methods and reference materials with defined metrics like yield, and purity.

Recovering intact EVs after MB isolation is another challenge, because the elution conditions (e.g., low pH, detergents or enzymatic treatments) can damage EVs membranes and their biological functionality. Ideally, elution methods should efficiently release EVs while maintain their structural integrity and composition. The beads would interfere with downstream analyses, such as RNA sequencing or protein analysis. If the MBs are not entirely removed, they are likely to co-purify with EVs and lead to false positives or other artifacts (He et al., 2016). Developing mild elution approaches, such as reversible binding chemistries or cleavable linkers, is crucial to overcoming this limitation. Here we propose some potential solutions. Firstly, design stimuli-responsive systems. For instance, engineer the MBs with photothermal or electrochemical materials to enable on-demand release via external triggers such as near-infrared light or applied potential, inducing localized phase transitions or electrostatic repulsion. Secondly, the integration of cleavable linkers, such as reducible disulfide bonds (-S-S-) or photolabile moieties (e.g., o-nitrobenzyl), between MBs and capture ligands enables controlled EV release. Disulfide cleavage requires reducing agents (e.g., DTT, TCEP), while photocleavage offers reagent-free elution via specific wavelength exposure. Both strategies facilitate efficient recovery but necessitate post-elution purification to eliminate chemical or light-induced artifacts. In conclusion, eluting intact EVs without damage remains challenging due to harsh conditions that compromise membrane integrity and functionality. Solutions include developing stimuli-responsive release systems (e.g., photothermal/electrochemical triggers) and incorporating cleavable linkers (e.g., disulfide or photocleavable bonds), though these require post-elution purification to avoid interference in downstream assays.