Recent advances of liquid biopsy: Interdisciplinary strategies toward clinical decision-making

3.1.1 Size based CTC isolation

Traditionally, the primary principle for CTC isolation is based on the size difference between CTCs and blood cells. Consequently, various filtration-based strategies have been developed and continuously improved (Figure 2A).⁶¹ These strategies offer simple fabrication processes and easy operation procedures, accelerating their clinical implementation. However, challenges such as low CTC recovery and clogging of the porous filter arise due to high filtration pressure and elevated hematocrit levels.⁶⁰ It is worth noting that the future of size-based isolation can serve as an accessible and preliminary method for CTC isolation. By sidestepping complex channel designs, they may enable rapid CTC isolation and enumeration within 10 min, while maintaining clinically acceptable sensitivity and specificity. Besides, further categorization based on cancer types is also required to achieve commercial and public acceptance.

3.1.2 Deformability based CTC isolation

In addition to the disparity in size, the less deformability of CTCs compared to red blood cells (RBCs) is another distinctive feature. Leveraging these differences, the deterministic lateral displacement (DLD) method has emerged as a powerful tool for the separation, purification, and enrichment of cells, encompassing aspects ranging from channel design to clinical applications (Figure 2B).⁶² By modifying pillar shapes, DLD chip with topology-optimized pillar shapes and a wider DLD channel can be developed to enhance the lateral displacement of cells with different sizes, achieving separation recovery rate of 97.1% for ECA cells with purity of 92.5%.⁶³

3.1.3 Hydrodynamics based CTC isolation

Microfluidic-based techniques leverage CTC hydrodynamics properties within the microchannels to drive the migration of cells, enabling the isolation of CTCs.⁶⁴ Cell experiences two internal forces, pushing them either toward or away from the channel walls.^{75, 76} The disparity in these forces anchored CTCs in the center of the microchannel while pushing blood cells toward the walls. Building upon this principle, inertial spiral combining with DLD sorting has been proposed for high-accuracy CTC separation from diluted blood samples (Figure 2C).⁶⁴ This microchannel achieved a high throughput of 400 μL/min and demonstrated the removal of 99% of blood cells while capturing 75% of tumor cells by adjusting the diameter, flow rate, and cell concentration. However, it is crucial to acknowledge as channel structures become more complex to enable more efficient CTC isolation, the operational complexity increases, posing challenges to their clinical applicability. Therefore, striking a balance between complexity and usability remains an ongoing challenge in the development of microfluidic devices for CTC isolation.

3.1.4 Viscoelastic based CTC isolation

To eliminate complex channel structures, the intrinsic properties of viscoelastic fluids have sparked considerable interest in viscoelastic non-Newtonian microfluidics offering rapid and label-free CTC isolation. By utilizing the combined forces of inertial lift and Dean drag inside simple microchannel, these devices enable efficient size-dependent CTC isolation.^{77, 78} By modulating the rheological properties of these viscoelastic fluid (mainly biological-friendly polymer solution), cell isolation can be achieved over a wide range of flow rates, which facilitates streamlined and effective CTC isolation. A microfluidic method establishes a stable flow interface between blood and viscoelastic fluid within a simple microchannel (Figure 2D),⁶⁵ by employing a concentration gradient of polyethylene oxide (PEG) for size-based and label-free CTC isolation. This microfluidic chip processed 1 mL of blood sample in just 30 min, achieving isolation efficiency exceeding 90%. The flow characteristics of viscoelastic fluids were also investigated by altering fluid concentrations for evaluating the behavior of particles with different sizes.⁶⁶ In a 0.2% hyaluronic acid solution, small particles were precisely focused at the center of the microchannel, while large particles were patterned into two distinct fluorescent streams resulting an impressive isolation efficiency of 94.8% and a purity of 98% for MCF-7 cells.

3.1.5 Light manipulation based CTC isolation

In addition to passive methods that rely on fluid design, external forces can be applied to selectively manipulate and separate CTCs from blood cells. Unlike passive isolation techniques that rely on fixed parameters of microfluidic structures, external forces can be customized and adjusted in real-time for individual patients with heterogeneous CTCs. Optical force is an appealing approach for CTC isolation based on their unique optical constants, but it faces challenges due to the similarity in optical constants between tumor cells and white blood cells (WBCs). To overcome this, Yang et al. introduced an innovative method that binds CTCs with RBCs, creating a noticeable difference in optical constants and achieving precise isolation without compromising specificity (Figure 2E).⁶⁷ By utilizing lasers in an optofluidic system, they attained high purity (>92%) and cell recovery rates (>90%) in CTC isolation. Other techniques like optically-induced dielectrophoresis and channel design have also demonstrated potential for utilizing optical forces in CTC isolation.^{68, 79} However, the small disparity in optical constants between CTCs and WBCs adds complexity to the pre-treatment process and device fabrication, despite the high purity and recovery rates.

3.1.6 Dielectrophoresis based CTC isolation

The dielectrophoresis (DEP) technique has been proven as an effective method in isolating CTCs due to the distinct dielectric properties of CTCs compared to blood cells. When exposed to a non-uniform electric field, cells become polarized, generating a dipole moment across their structure. The polarized cells move based on the region of maximum field intensity, depending on factors such as conductivity, permeability, electric field magnitude and frequency, and liquid polarizability.⁸⁰ DEP have been integrated with flow fractionation to achieve high-rate and highly efficient CTC isolation (Figure 2F),⁶⁹ enabling continuous flow-through separation at a flow rate of 126 μL/min, removing RBCs and WBCs with efficiencies of 99.24% and 94.23%, respectively. Another study introduced a DEP-driven microfluidic device with a V-shape electrode, enabling the continuous isolation of CTCs from 2 mL peripheral blood sample within 1 h, resulting in a 96.5% depletion of blood cells and an 18.6-fold enrichment of cancer cells.⁷⁰

3.1.7 Acoustophoresis based CTC isolation

Acoustophoresis, a gentle and label-free method for trapping, separating, and concentrating particles or cells, has proven to be successful due to its ease of implementation and biocompatibility. This technique utilizes acoustic streaming and acoustic radiation force (ARF), based on acoustic waves generated by a piezoelectric actuator, to manipulate particles across a wide range of sizes.^{72, 81} By designing the channel width to create an optimized acoustic standing resonant cavity, the acoustofluidic-based particle separation technique achieves precise control over fluids and particles by efficiently coupling acoustic energy to the liquid.⁸² A simple and transparent acoustofluidic device have been demonstrated for effective sorting of particles and cancer cell lines under the standing acoustic field with separation efficiencies of approximately 97 ± 1.0% and 91.5 ± 4.5%, respectively (Figure 2G).⁷¹ A two-step acoustophoresis method uses an initial acoustofluidic to pre-separate cells based on their acoustic mobility for removing the RBCs.⁷² Subsequent purging step was employed to remove contaminating WBCs through negative selection acoustophoresis with anti-CD45-functionalized negative acoustic contrast particles. This method provides extensive elimination of WBCs combined with the gentle recovery of viable cancer cells suitable for downstream functional analyses and in vitro culture. The multi-stage acoustofluidic devices was integrated with different wavelengths driven by modulated signals to address the challenge of separating more than two different sizes with high efficiency and accuracy.⁸¹ The slanted angle, acoustic pressure, and resonant frequency of the device were systematically investigated, revealing that the multistage acoustofluidic devices achieved a separation efficiency of 99% for the three different size particles in theorical analyses.

3.1.8 Magnetophoresis based CTC isolation

Magnetophoresis-based CTC isolation techniques can be categorized into positive magnetophoresis and negative magnetophoresis. Positive magnetophoresis involves attracting CTCs toward a magnetic source, requiring the labeling of CTCs with magnetic particles.⁸³ Conversely, negative magnetophoresis repels CTCs away from the magnetic source, necessitating a higher susceptibility magnetic fluid.⁵⁶ Typically, a permanent magnet generates a magnetic field gradient that pushes the CTCs (without magnetic properties) away by attracting the magnetic medium. By adjusting the sample flow rate, negative magnetophoresis based CTC isolation was proposed by manipulated volumetric concentration of magnetic materials in the ferrofluid, and the magnetic intensity gradient (Figure 2H).⁷³ These optimized parameters resulted in high CTC recovery and separation efficacy in both spiked samples and clinical samples.

While label-free CTC isolation techniques preserve cell integrity and facilitate downstream analysis, active methods for CTC isolation have been limited by their low throughput, as they require a substantial amount of time for the samples to be influenced by external force fields. In contrast, passive methods do not rely on external force fields and can achieve relatively higher throughput. However, passive methods often involve intricate channel structures or complex fabrication processes that may largely elevate the threshold of large-scale clinical application. It is important to note that relying solely on physical features for CTC isolation may encounter problems like cellular heterogeneity in clinical trials, leading to reduced sensitivity and specificity for a specific group of patients. Addressing these challenges through future research by integrating several strategies of passive and active isolation will provide valuable insights for clinicians in their decision-making processes.

3.2.1 Specific molecules

After the successful utilization of antibody-labeled MNPs for CTC isolation,¹⁵ researchers have explored various specific antibodies to establish chemical bonds with CTCs. EpCAM protein has been widely targeted for CTC isolation due to its high expression in CTCs derived from epithelial tissues. For instance, Tseng et al. and other researchers have demonstrated the feasibility and high efficiency of anti-EpCAM-based CTC isolation across different cancer types.^{40, 45} However, the dynamic nature of cancer cells, particularly during EMT, can lead to changes in surface protein. During EMT, cancer cells exhibit reduced EpCAM but increased expression of other antigens such as N-cadherin.⁹³ This provides opportunities to use co-antibodies targeting both EpCAM and N-cadherin for CTC capture. Pei et al. developed a uniform poly-(lactic-co-glycolic acid) (PLGA) nanofiber substrate modified with co-antibodies.⁷⁴ Compared to single antibody approaches, the co-antibodies demonstrated improved capture efficiency for both epithelial (e.g., MCF-7, up to 66.5%) and mesenchymal (e.g., GIST882, up to 76.4%) cells. In addition to EMT, CTCs originating from different tissues may exhibit different proteins. Wang et al. addressed this challenge by demonstrating a reduced graphene oxide (rGO)-based CTC isolation platform modified with a mixture of anti-EpCAM and anti-PSMA (anti-prostate-specific membrane antigen).⁷ By investigating the ratio of anti-PSMA, the roughness of the rGO, and the capture time, they achieved a capture efficiency of up to 90.3% for a PCa cell line within 45 min.

Peptides have also emerged as promising alternatives to antibodies in CTC isolation due to their advantages over antibodies, such as lower cost, improved preservation, stable consistency, and ease of chemical synthesis production. Hu et al. have been actively researching CTC enrichment platforms using peptides and have developed the TumorFisher® platform for highly efficient CTC isolation.⁹⁴ They have designed a variety of peptides for isolating CTCs from different types of cancers and have successfully translated these strategies into clinical practice. For CTCs undergoing EMT, Hu et al. screened and identified a novel peptide targeting N-cadherin,⁴¹ achieving high-efficiency capture of mesenchymal CTCs, with a capture rate of approximate 85%. In addition, López and colleagues formulated peptide-modified nano-emulsions specifically targeting cell membrane proteins EpCAM and EGFR on breast cancer CTCs.⁹⁵

Aptamers, synthetic oligonucleotides with specific affinity for proteins on the cancer cell membrane, have gained attention as another interesting molecule for CTC isolation.⁹⁶ Similar to peptides, aptamers are easily synthesized and modified, thereby enhancing their performance in CTC isolation. Song et al. developed a microfluidic chip modified with three-dimensional DNA structures (TDNs) as frameworks,⁹⁷ enhancing capture efficiency to around 87% due to the highly precise dimensions and rigid framework of TDNs. By functionalizing double-tetrahedral DNA probes on nanorough Ag nano-biointerfaces, the synergetic effects of molecules and topographies improved the capture efficiency to 90.2% for SGC-7901 cells in PBS.⁹⁸ Pei et al. utilized dual aptamers targeting EpCAM and N-cadherin proteins to enhance the capture sensitivity of both epithelial and mesenchymal CTCs,⁸⁹ achieving capture efficiencies of 91% for A2780 cells (high N-cadherin and low EpCAM) and 89% for OVCAR-3 cells (low N-cadherin and high EpCAM).

The basic principles underlying CTC isolation methods are similar, involving the identification of highly expressed proteins, screening specific molecules, modifying them on substrates, and isolating CTCs. However, an intriguing alternative approach utilizes a substrate covered with biological membrane, which might arouse broad interests. This method aims to address the limitations associated with recognizing EpCAM-negative CTCs and the non-specific adsorption of leukocytes. Wang et al. developed biomimetic magnetic beads (MBs) by cloaking them with a hybrid membrane derived from cancer cells and leukocytes.⁹⁰ These biomimetic MBs retained the CTC binding capability of cancer cell membranes while showing reduced affinity for background cells due to the leukocyte membrane.

3.2.2 Topological structure

The specific isolation also requires a substrate to load specific recognition molecules. Generally, three strategies have been frequently proposed as the substrates of CTC isolation, which are MBs,⁹⁹ microfluids,⁴⁰ and capture chip.⁹¹ Usually, the larger the specific surface area will load more recognizing molecules and generate topological match with cells, increasing the affinity between cells and substrate and thereby achieving higher capture efficiency.

Among the three strategies mentioned, nano-/micro-scale MBs are the most extensively utilized for CTC isolation.^{100, 101} By incorporating specific recognition molecules onto the surface of MBs, they can rapidly bind to target cells and be easily separated from other cells using magnetic fields. Depending on the choice of recognition molecules, MBs can be employed for positive CTC capture⁹⁹ or negative isolation of blood cells.¹⁰² Numerous reviews have been published focusing on MBs, covering aspects such as synthesis,¹⁰³ specific molecules,¹⁰⁴ CTC isolation,¹⁰⁵ and clinical applications.¹⁰⁶ However, most MBs currently in use have smooth surfaces without distinct structures, which can lead to challenges of low sensitivity, low specificity, and poor reproducibility, despite their widespread clinical utilization. Recently, Wang et al. have introduced a novel approach to synthesize nanofractal microparticles using electrostatic interaction-regulated emulsion interfacial polymerization.¹⁰⁷ These nanofractals exhibit exceptional capabilities for rapid protein capture, release, and separation. The topological features can well-matched with pseudopodia on cancer cells, suggesting a promising and versatile method for the synthesis of microparticles that enable efficient and rapid CTC capture. In addition to spherical MBs, Shi et al. developed DNA aptamer-functionalized magnetic short nanofibers for highly efficient capture and release of CTCs.¹⁰⁸ Through a blended electrospinning process, magnetic nanoparticles (MNPs) were encapsulated within nanofibers, resulting in structures with a mean diameter of 350 nm and an average length of 9.6 μm. These nanofibers exhibited specific capture capability for cancer cells, achieving an efficiency of 87%, presenting an alternative pathway for fabricating high-performance CTC isolation units, expanding the possibilities in this field.

Apart from MBs, structured substrates offer unique surfaces for interacting with cells and which have been already intensively summarized by excellent researchers. We aim to categorize perspectives based on the principles of designing structured substrates. Along with the development of photolithography, structures like nanowires demonstrated their potential in providing higher specific surface area. Tseng et al. have optimized this technique and expanded its application to isolate not only individual CTC but also CTC clusters, broadening the scope of cancer monitoring beyond simply CTC enumeration.^{40, 55, 109} Wang et al. have provided insights into bio-inspired interfaces such as nanosheets, nanofractals, and nanoflowers.^{7, 46, 57, 91, 92} These nanostructured interfaces can achieve topological matching with cancer cells and their pseudopodia, and specifically interact with molecule on the cell membrane. Additionally, PLGA nanofibers have also been utilized to entangle and trap cells with the help of specific recognition molecules.¹¹⁰

To enhance CTC capture efficiency, the principles can be summarized as follows: (i) increasing the specific surface area to accommodate more specific recognition molecules, (ii) providing a topological match with the cell structure by exhibiting similar sizes (80–100 nm for cell pseudopodia and 5–15 μm for the cell body) or complementary structures (nanowires for pseudopodia and micro-bowls for the cell body), and (iii) trapping cells based on their size differences. In addition, traditional specific isolation remains reliant on passive interaction with cells, signifying that the capture substrate waits for the cells' actions, instead of actively applying structure onto cells, as immune cells do. Future developments in active CTC capture could significantly reduce the time and process, making it easier for clinicians to access cancer information and facilitating large-scale physical examinations for early cancer screening.

In conjunction with CTC isolation methods based on physical characteristics, superior capture efficiency and specificity demand increasingly intricate substrate architectures and more potent, selective chemical bonds. These tactics, despite showcasing exceptional capture performances and offering novel CTC isolation methods, often rigidly confine cells to the capture substrates.^{7, 91, 92} This can pose challenges when trying to extract the wealth of data contained within the CTCs. Contemporary advancements in chemically based isolation strategies have paved the way for multi-analytical capabilities, incorporating techniques such as photoelectrochemical biosensors,¹¹¹ surface plasmon resonance (SPR),¹¹² and sensors employing quenching effects.¹¹³ These strategies, however, continue to grapple with obstacles related to expensive reagents, laborious procedures, and non-intuitive readouts - all key considerations for clinical implementation in cancer patients.

3.3.1 Thermo-response

A prevalent strategy for the release of CTCs from capture substrates employs thermo-responsive materials, which primarily invalidate the link between recognition molecules and the substrate in response to temperature changes. Among thermo-responsive polymers, poly(N-isopropylacrylamide) is frequently used for this purpose.¹¹⁶ This polymer undergoes structural conformation changes below the lower critical solution temperature (around 20°C), transitioning from hydrophobicity to hydrophilicity thereby eliminating the hydrophobic interaction with recognition molecules. When engineering switchable SA-recognition sites into thermo-responsive hydrogel, the bingeing can by reversed from strong binding at 37°C and weak binding at 25°C, thus capturing cancer cells and non-invasively releasing them by reducing the temperature (Figure 4A).¹¹⁷

In addition to thermo-response bond invalidation, Xu et al. developed a near-infrared (NIR) light-responsive microfluidic chip to invalidate the micro-area of a gelatin substrate, facilitating biocompatible single-cell manipulation.¹²² The captured single CTC can be recovered in large quantities or individually released at a specific location using NIR light, thereby precisely acquiring information about CTC heterogeneity at the single-cell level. For the advancement of thermo-response CTC release, it is important to balance the temperature change range and the thermo-response time to preserve cell integrity and viability.¹²³

3.3.2 Light-response

Light-responsive CTC release can be achieved by designing bonds that are sensitive to light, like those commonly used in drug release applications. For example, Soper et al. synthesized a 7-(diethylamino)coumaryl-4-methyl photo-release agent that undergoes rapid cleavage upon exposure to visible light (400–450 nm) (Figure 4B).¹¹⁸ This light-responsive bond minimizes nucleic acid damage and exhibits minimal side reactions. In their study, SKBR3 cells can be rapid and efficient released, with 94% release efficiency and 94% cell viability after only 2 min of visible light LED exposure. The use of visible light for release also prevents damage to nucleic acids caused by photo-absorption and oxidation typically associated with UV irradiation. Thus, the overall design of light-responsive CTC release aims to identify specific bonds that are sensitive to visible light and easily modifiable for efficient release.

3.3.3 Electro-response

By leveraging electrochemical manipulation, the conductive substrate delivers the electro-signal to the location of electrically-responsive units, consequently instigating chemical reactions. Electrochemical reductive desorption presents several advantages, encompassing the preservation of cell viability, strong release potential, and time efficiency. A cyto-sensor was designed using gold nanoparticles-loaded two-dimensional bimetallic PdMo (2D Au@PdMo) nanozymes, coupled with thiol-anti-EpCAM for the CTC isolation (Figure 4C).¹¹⁹ Subsequently, the captured CTCs could be efficiently and non-destructively released from the modified electrodes through the reductive desorption of the Au-S bond with release rate between 93.7% and 97.4% and high cell viability. By incorporating antibody-modified with electrochemically cleavable semiconducting silicon surface, specific individual cells can be released of interest,¹²⁴ when both light and electrochemical potential were simultaneously applied.

3.3.4 pH-response

The release of captured CTCs can be facilitated by manipulating the pH value, primarily to invalidate the link between the recognition molecule and the substrate. Han and colleagues integrated pH-responsive poly-L-lysine functionalized carbon nanotube into a microfluidic chip for the proficient capture and release of cancer cells (Figure 4D).⁵⁸ Using anti-EpCAM, they achieved a capture efficiency of 86.7 ± 9.3%. Subsequently, by raising the pH value to 11.0 to induce a conformational change, they accomplished a release efficiency of 84.7 ± 12.2%. In addition, with a biological viability of 84.6 ± 5.5%, the liberated cells could be cultured for up to 7 days without alterations in morphology and EpCAM expression. However, a pH-responsive approach demands either a pH alteration beyond the standard cell culture environment range, or a prolonged treatment duration that might impair cell viability and integrity.

3.3.5 Enzyme-response

The detachment of CTCs from substrates through enzyme responses is a common approach in cell culture, typically involving trypsin to invalidate the membrane-coherent molecules. While this method has been utilized for CTC release,¹²⁵ the enzyme's action can harm the surface markers and membrane structures, potentially disrupting subsequent CTC analysis. To mitigate this issue, enzymes designed specifically to invalidate the links between specific molecules and substrates have been developed to retain the information encapsulated in cancer cells. A multi-marker microfluidic chip modified with aptamer leverages the synergistic effect of multivalent aptamers and DLD arrays (Figure 4E).¹²⁰ The aptamer nanostructures enhance the accessibility of nucleases, enabling cell release with an efficiency of nearly 80% and a viability of 90.42%, while also maintaining molecular integrity. Similarly, Wang and colleagues utilized DNA probes for capture and nondestructive release of CTCs.⁹⁸ After capturing 90.2% of SGC-7901 cells in PBS, 93.4% of cells were released via Zn²⁺-assisted DNAzyme cleavage, with the viability of post-released CTCs reaching approximately 98.0%. These specialized aptamers, specifically targeting the designed site, offer a unique perspective on efficient CTC isolation and low-damage release, by simultaneously constructing the structure and composition of the aptamer.

3.3.6 Substrate breakdown

One of the most straightforward strategies for detaching CTCs from the capturing substrate involves simply degrading the substrate itself. This can be achieved through various methods, each tailored to the intrinsic properties of the substrate.¹²⁶ For instance, Guo and colleagues utilized TiO₂ nanopillar arrays coated with a gelatin film for the efficient capture and damage-free release of CTCs (Figure 4F).¹²¹ The interaction between the cell membrane and the nanostructure substrate contributed to an impressive CTC capture rate (94.98%). The gelatin layer, with outstanding biocompatibility, can be swiftly digested by matrix metalloproteinase. This allows for the non-destructive release of CTCs, with nearly 100% release efficiency and 100% cell activity.

In general, there are multiple strategies for efficient CTC release based on the properties of recognition molecules, which should be chosen based on the materials and cells with various capture substrates or strategies with the principles to ensure the viability of cells.

4.1.1 Size based EV isolation

Size-selective filtration has been a classical method for differentiating between blood cells and EVs. Previous approaches utilized porous membranes with specific pore sizes to block blood cells (200–600 nm) while enriching EVs (30–50 nm).¹³³ However, these methods faced challenges such as channel blockage, integrity damage, and impurities, limiting their broader application. Recently, a centrifugal microfluidic device with two nanofilters enabled automated enrichment of EVs smaller than 600 nm from urine within 30 min 134 Despite its efficiency, this method exerted a high shear force that could potentially damage the EVs. In a novel approach, Sun et al. leveraged the concept of electric-hydraulic analogy to design cascaded microfluidic circuits for pulsatile filtration of EVs (Figure 5A).¹³⁵ This design incorporated a cell-removal circuit and an EV-isolation circuit. At the core of this system was the fluidic capacitor, which comprised an elastomeric membrane sandwiched between two microchambers. The membrane could deform under applied pressure, storing and releasing fluid accordingly. Upon applying pressure, the membrane deformed, allowing particles to flow, while reducing the pressure caused the membrane to recover, collecting the particles in another channel. This microfluidic device enabled clog-free, gentle, high-yield, and high-purity isolation of EVs directly from blood within a short timeframe of 30 min.

4.1.2 Viscoelastic based EV isolation

The burgeoning field of viscoelastic non-Newtonian microfluidics offers an avenue for the rapid, label-free isolation of EVs from various body fluids. Using a similar principle to the viscoelastic-based isolation of CTCs, this methodology harnesses the power of elastic forces to guide larger EVs toward the center of the microfluidic channel. Sun et al. presented a microfluidic system leveraging viscoelasticity for the direct, size-dependent, and label-free separation of EVs from either cell culture media or serum (Figure 5B).¹³⁶ The system adeptly isolated EVs from intricate media housing a multitude of EVs of varying sizes. By introducing a minor quantity of biocompatible polymer as an additive in the media, they were able to manipulate the viscoelastic forces acting on EVs. This approach demonstrated high separation purity (>90%) and recovery (>80%) rates, promisingly supporting the streamline of EVs analysis of biochemical applications.

4.1.3 Hydrodynamics based EV isolation

Microfluidic channels offer an innovative approach to EV isolation, allowing for the separation of particles based on their hydrodynamic properties. One of these techniques is asymmetric flow field-flow fractionation (AF4), which uses particle density and hydrodynamic properties to separate nanoscale soluble particles, potentially as small as a few nanometers.¹⁴³ AF4-based devices generate an orthogonal force to a porous membrane within the microfluidic channels by using a combination of field flow and Poiseuille flow. This force effectively separates particles based on their different lateral positions in the channel, driven by their respective diffusion rates (Figure 5C).¹³⁷ DLD-based isolation can also effectively separate EVs. In this method, particles smaller than a critical diameter follow the streamline through a pillar array without lateral displacement (depicted by the red line). In contrast, particles larger than the critical diameter flow away from a pillar and are laterally shifted to the next row (depicted by the blue line). Astier et al. scaled DLD to the nanoscale (nano-DLD) by using arrays of uniform gap sizes ranging from 25 to 235 nm (Figure 5D).¹³⁸ This nano-DLD array effectively separates particles ranging from 20 to 110 nm based on size, with excellent resolution, resulting in further size-based displacement of EVs. Morgan et al. applied an alternating current electric field perpendicular to the fluid flow, which led to an approximate ten-fold reduction in the intrinsic critical diameter of the device.¹⁴⁴ Pinched Flow Fractionation (PFF) is another promising strategy for particle separation based on size. In this method, particles are directed along the sidewall of a microchannel under a strong sheath fluid flow. Due to this flow, smaller particles tend to position closer to the sidewall, while larger particles are located slightly further away. This subtle position variation is then amplified by streamline divergence in a downstream section with sudden branching, resulting in effective size-based separation. Hong and colleagues demonstrated the use of PFF with multiple outlets with sudden bunching for separating particles of different sizes (Figure 5E) with potential in the isolation of specific EV subpopulations.¹³⁹

4.1.4 Light manipulation based EV isolation

Light manipulation has recently demonstrated their potential of manipulating the separation of EVs. Unlike the optical force-based CTCs isolation, manipulating smaller entities like EVs could prove more effective and straightforward. Sun et al. have explored the potential of light manipulation in EVs separation, demonstrating an innovative laser-induced thermophoretic aptasensor (Figure 5F).³⁴ They enabled enrichment of EVs conjugated with aptamers through size-dependent accumulation, relying on the interaction of thermophoresis, diffusion, and convection, all induced by localized laser heating. Their approach enabled a 1400-fold enrichment within just 10 min, with no thermal degradation of EV surface proteins and molecular cargos occurred. Furthermore, the team combined this technique with PEG to enhance the EVs' accumulation.¹⁴⁵ This integration allowed for the sensitive, selective, and in situ detection of mutated RNA within EVs, proving a significant stride in EVs research and potential diagnostic applications.

4.1.5 Dielectrophoresis based EV isolation

DEP is an electrokinetic phenomenon that allows for the manipulation of suspended particles within an electric field based on their size and electrical properties. Yobas et al. demonstrated the use of DEP in a microfluidic device for continuous-flow, label-free size fractionation of EVs (Figure 5G).¹⁴⁰ By combining electrothermal fluid rolls and DEP, they achieved the separation of larger EVs from smaller ones. By optimizing voltage activation and flow rate, the smaller EVs were concentrated around the channel centerline and exited through the main branch, effectively separated from the large EVs. Another approach involved combining alternating current electroosmosis and DEP to attract EVs toward a sensor chip surface, specifically concentrating them on plasmonic sensing areas. This method enabled the sensitive collection of tumor-derived EVs in a short time, revealing significant differences in yield signals between cancer and healthy control groups, demonstrating the potential of DEP-based techniques for both EV isolation and disease diagnostics.¹⁴⁶

4.1.6 Acoustophoresis based EV isolation

Acoustophoresis centrifugation have revealed its potential in separating particles based on size differences and become one of the popular high-precision and low-damage isolation strategies. Lee and colleagues developed an acoustic nanofilter system that separates EVs from biological samples,¹⁴⁷ employing surface acoustic waves to generate ARF. Larger EVs are pulled toward the pressure nodes near the channel sidewalls, while smaller EVs remain at the central flow as the acoustic force exerted on them is weaker. A unique approach was developed called Acoustic Nanoscale Separation via Wave-pillar Excitation Resonance for the fractionation of nanoscale bioparticles (Figure 5H).¹⁴¹ They used excitation resonance to form an array of virtual acoustic wave pillars in a microfluidic channel, where the pillars act as filters, exerting stronger influence on larger particles to separate EVs. Their strategy provides a rapid and efficient method for the high-purity fractionation EV subpopulations from biofluids in less than 10 min. Collins et al. expanded on these methodologies by developing a device that leverages focused traveling acoustic waves to manipulate micro and nanoparticles.¹⁴⁸ They observed different focus points for nanoparticles of 100 nm, 300 nm, and 500 nm diameters, regardless of their initial positions, demonstrating the flexibility and precision of this approach.

4.1.7 Magnetophoresis based EV isolation

Magnetophoresis-based methods for EV isolation employ magnetic fields to segregate EVs. Typically, this process utilizes a magnetic field gradient from a permanent magnet source to displace EVs by attracting a magnetic medium such as an aqueous suspension of MNPs. These nanoparticles have a magnetic susceptibility much greater than that of a paramagnetic solution, enabling the manipulation of bioparticles within a microchannel using regular permanent magnets. Yang's group proposed an improved separation platform that is both label-free and biocompatible, with a higher sample throughput (Figure 5I).¹⁴² To enhance the magnetic force and achieve nanoscale resolution, while also addressing the biocompatibility issues of the ferrofluid, they used arrays of magnetic poles. This approach successfully produced an ultra-high magnetic field gradient of 105 T/m within the microchannel, enabling the isolation of EVs from cell culture media with a recovery rate of 86% and a purity of 80% at a sample flow rate of 150 μL/h. This methodology demonstrates the potential of magnetophoretic techniques in the efficient and high-throughput isolation of EVs.

Without preselection of EV subpopulations, most physical feature-based isolation nanotechnologies retain the primary bulk properties and biological activity of EVs. However, their clinical applications may be largely restricted because of the laborious fabrication of the facilities and the robustness of these complex systems. Thus, the simplification of the design and process technology should be considered in the subsequent development of these EV capture devices.

4.2.1 Antibody based EV isolation

Antibody-based EV isolation employs a method similar to antibody-based CTC isolation. This technique utilizes surface proteins found in EV membranes, such as CD63, CD9, CD81, EpCAM, EGFR, and GPC-1, to bind and isolate specific EVs from various fluids.¹⁴⁹ For example, microvilli-like silicon nanowire arrays featured an increased surface area and high-efficiency EV isolation utilizing modification of anti-EpCAM (Figure 6A).¹⁵⁰ The combination of antibodies targeted EpCAM, ASGPR1, and CD147 on the EV membrane, to create a hepatocellular carcinoma (HCC)-specific EV purification system.²⁹ The antibodies were modified with click-on reaction agents trans-cyclooctene (TCO), and the capture substrate (silicon nanowires) was modified with tetrazine (Tz). After the slow interaction reaction between TCO-antibodies and EVs in fluids, the rapid click-on reaction of TCO with Tz will bond EVs to the capture substrate soon after TCO-EVs injection. Similarly, Stott et al. designed a sensitive microfluidic platform for tumor-specific EV-RNA isolation within 3 h¹⁵⁷ which was equipped with a nanostructured substrate functionalized with streptavidin-coated nanoparticles to bind with antibodies.

4.2.2 Aptamer based EV isolation

Aptamers, due to their high specificity and affinity like antibodies, are capable of binding to specific targets. This characteristic makes them valuable for developing affinity-based isolation strategies. For example, aptamer-coupled Au-decorated polymorphic carbon (CoMPC@Au-Apt) exhibited EV capture from urinary of gastric cancer patients. By utilizing a CD63-specific aptamer, they achieved efficient capture within a short period of 45 min (Figure 6B).¹⁵¹ In another study, Anastasia Malek and colleges developed an AuNP aptasensor for the analysis of CD30-positive EVs.¹⁵⁸ This approach facilitated the isolation of CD30-positive EVs from Hodgkin and Reed-Sternberg cells, providing insights into the activity of classical forms of Hodgkin lymphoma. Chen et al. utilized PD-L1-targeting-aptamer-anchored magnetic microspheres for the specific capture of PD-L1-positive EVs.¹⁵⁹ Their research revealed that PD-L1-positive EVs can directly bind and inhibit T cell proliferation, like their parent cells. Additionally, these PD-L1-positive EVs can transfer immunosuppressive proteins related to the “synapse” to antigen-presenting cells, inducing a T-cell-inhibitory phenotype. Compared to antibody-based capture techniques, aptamer-based methods hold great potential for EVs separation due to their high binding affinity for protein biomarkers present on the surface of tumor-derived EVs.

4.2.3 Peptides based EV isolation

Peptides with a specific affinity for EV surface proteins are used for EV isolation. Ouellette's group developed a unique class of peptides, exhibiting a specific affinity for canonical heat shock proteins, have been validated to capture these protein-containing EVs from sources like cell culture growth media, plasma, and urine.¹⁶⁰ Okochi et al. identified a peptide based on the amino-acid sequence of EWI-2 protein (Figure 6C),¹⁵² which specifically bonded with CD9 protein on cancer EVs. By combining this peptide with a ZnO-binding sequence, they realized reversible capture of EVs under a neutral salt condition.

4.2.4 TiO₂-phosphate affinity based EV isolation

Non-specific EV isolation strategies based on chemical affinity typically leverage the affinities between lipid and the capture substrate. Among these methods, the TiO₂-phosphate affinity demonstrated simple fabrication processes, straightforward isolation protocols, and compatibility with downstream analyses. Hu et al. employed this user-friendly strategy selectively isolating EVs and detecting their metabolic alternations (Figure 6D).¹⁵³ Zhang and co-workers reported a similar TiO₂-mediated EV isolation strategy from a small volume of plasma, where they simultaneously profiled proteins and metabolites.¹⁶¹ Another research group utilized this strategy for reversible EV isolation, achieving high recovery rates (93.4%) using an alkaline solution.⁴³ In addition, Deng et al. utilized dual-affinities MBs, integrating TiO₂ with CD63 specific aptamer, achieving high-efficient EV isolation up to 92.6% within 10 min⁴⁹ However, this strategy can isolate all particles with a phosphate group, thereby requiring a preliminary treatment to first remove cells.

4.2.5 Lipid based EV isolation

Except form the head of phospholipid, the lipid molecules on the surfaces of EVs significantly influence EV isolation. In 2017, 1,2-distearoyl-sn-glycero-3-phosphoethanolamine (DSPE) emerged as a promising probe for non-specific EV isolation, since it can insert into the phospholipid membrane of EVs through its two hydrophobic fatty acid tails.⁴⁴ Wang et al. prepared DSPE-functionalized UiO-66 metal-organic framework,¹⁵⁴ by leveraging the synergistic affinity interaction of Zr⁴⁺ ion and DSPE with EVs for efficiently EV enrichment from urinary samples, achieving release efficiency up to 99% through alkaline elution (Figure 6E). Chen et al. intravenously injected DSPE-PEG-Biotin to directly label circulating EVs in vivo, aiming to minimize alterations in their behavior.¹⁶² Using this strategy, they revealed the distinct in vivo fate of circulating EV subpopulations originating from different cell sources. Notably, they discovered that erythrocyte-derived EVs exhibited the longest lifespan. Similar to pretreatment with antibodies, this approach allows for an interval between interaction and isolation, facilitating the enrichment of EVs.

4.2.6 Phosphatidylserine affinity based EV isolation

Due to the nano-size of EVs, the EV membrane contains a larger proportion of cone-shaped lipids such as phosphatidylserine (PS), which typically isn't present on the outer layer of normal cells. By exploiting specific affinities toward PS, EVs can be efficiently isolated. Annexin A5 and T-cell membrane protein 4 (TIM-4) have been proposed as efficient probes for EV isolation due to their high affinity for PS molecules. For example, Nagrath et al. conjugated Annexin A5 with a microfluidic chip to isolate cancer-associated EVs and release immobilized EVs using Ca²⁺ ion chelation (Figure 6F).¹⁵⁵ Deng et al. modified a herringbone microfluidic chip with TIM-4 to isolate tumor-derived PS-positive EVs,¹⁶³ demonstrating that cancer patients have a higher number of PS-positive EVs circulating in the bloodstream compared to healthy donors. Ye and colleagues presented a combination of TIM-4 based specific isolation and a signal transduction strategy for the high-sensitive detection of CD63 positive EVs, achieving an extremely low limit of detection around 4.39 × 10³ particles/mL, which underscore the potential of PS-specific approaches in enhancing the efficiency of EV isolation and detection.¹⁶⁴

4.2.7 Zwitterionic coordination based EV isolation

Zwitterionic coordination leverages the specific polyvalent interactions between the phosphatidylcholine (PC), a prevalent headgroup on EV membranes, and choline phosphate, an “inverse” of PC.¹⁶⁵ A method was introduced based on zwitterionic coordination that enabled the extraction of EVs from a variety of biofluids such as blood serum, urine, and saliva, in 30 min, yielding over 90% purity and recovery (Figure 6G).¹⁵⁶ Furthermore, the PC-CP coordination could be reversibly regulated by merely altering the surrounding temperature, facilitating the instant release of EVs. In a related study, zwitterionic polymeric coacervates were designed, composing sulfabetaine methacrylate and [2-(methacryloyloxy)ethyl]trimethylammonium.¹⁶⁶ The researchers used these coacervates to recruit and release EVs involving one-minute incubation followed by the separation through centrifugation and release approximately 86% of EVs after NaCl treatment.

Based on above-mentioned strategies, EVs can be rapidly and high-efficiently isolated from body fluids. While, EVs exist in near all body fluids making them promising biomarkers for various disease (Table 2). Other than blood samples, urine with large volume and easy sampling processes attracted large attention for guiding clinical decision-making in diseases associated with urinary system, such as focal segmental glomerulosclerosis,¹⁷¹ systemic lupus erythematosus,¹⁸³ PCa,^{172, 173} and bladder cancer.¹⁸⁴ Likewise, tear and saliva have also been applied for decision-making of diseases related to their origin, such as dry eye syndrome¹⁷⁵ for tear or periodontitis¹⁷⁸ and glioblastoma¹⁷⁹ for saliva. As another kinds of body fluids accumulated in body cavity, cerebrospinal fluid and ascites contain hormones, and other actively or passively secreted metabolites from cells and cell apoptosis. Extracellular vesicles isolated from cerebrospinal fluid can reflect the situation of diseases related to brain and central nervous system like aneurysmal subarachnoid hemorrhage¹⁷⁷ and cancer leptomeningeal metastasis.¹⁷⁶ EVs in ascites also demonstrated great performance in cancer diagnosis¹⁸¹ and therapy monitoring.¹⁸²

Even though EVs from various body fluids can reflect multi-dimensional situation of diseases, it can be noticed most of novel strategies focused on the universal-available and high-stable fluids for feasibility verification, due to several challenges caused by their heterogeneous physical and chemical characteristics. For example, the EV concentration of urine and saliva is easily affected by different physiological conditions, and the timing of sample collection largely affects the results of EV detection. In addition, the sampling processes are damageable for cerebrospinal fluid, or require precision operation for tears. This gap between advanced isolation strategies and real-world clinical applications hinders the implantation of EV related liquid biopsy into clinical scenarios, leading to the consequences that clinician trends to rely on commercial-available, mature but inefficient technologies (Table 2).

In summary, the next generation of EV isolation strategies should retain the primary bulk and properties biological activity of EVs, while simplifying laborious fabrication and operation processes in the meantime to broaden their clinical applications for various diseases using different body fluids.