Extracellular vesicles hold immense promise for biomedical applications, but current isolation methods are time-consuming and impractical for clinical use. In this study, we present a microfluidic device that enables the direct isolation of extracellular vesicles from large volumes of biofluids in a continuous manner with minimal steps.
Extracellular vesicles (EVs) hold immense potential for various biomedical applications, including diagnostics, drug delivery, and regenerative medicine. Nevertheless, the current methodologies for isolating EVs present significant challenges, such as complexity, time consumption, and the need for bulky equipment, which hinders their clinical translation. To address these limitations, we aimed to develop an innovative microfluidic system based on cyclic olefin copolymer-off-stoichiometry thiol-ene (COC-OSTE) for the efficient isolation of EVs from large-volume samples in a continuous manner. By utilizing size and buoyancy-based separation, the technology used in this study achieved a significantly narrower size distribution compared to existing approaches from urine and cell media samples, enabling the targeting of specific EV size fractions in future applications. Our innovative COC-OSTE microfluidic device design, utilizing bifurcated asymmetric flow field-flow fractionation technology, offers a straightforward and continuous EV isolation approach for large-volume samples. Furthermore, the potential for mass manufacturing of this microfluidic device offers scalability and consistency, making it feasible to integrate EV isolation into routine clinical diagnostics and industrial processes, where high consistency and throughput are essential requirements.
Extracellular vesicles (EVs) are cell-derived membrane-bound particles comprising two main types: exosomes (30-200 nm) and microvesicles (200-1000 nm)1. Exosomes form through inward budding of the endosomal membrane within a multivesicular body (MVB), releasing intraluminal vesicles (ILVs) into the extracellular space upon fusion with the plasma membrane1. In contrast, microvesicles are generated by outward budding and fission of the cell membrane2. EVs play a crucial role in intercellular communication by transporting proteins, nucleic acids, lipids, and metabolites, reflecting the physiological state of the cell, including growth, angiogenesis, metastasis, proliferation, and therapy resistance3. As a result, they have emerged as promising biomarkers and therapeutic targets for diseases, including cancer, highlighting their potential in diagnostics and drug delivery systems4.
To fully utilize EVs in disease diagnostics and therapeutics, efficient isolation from various biofluids is crucial5. Common methods include ultracentrifugation (UC), density gradient centrifugation, size exclusion chromatography (SEC), filtration, and immunoisolation6. UC is a widely used technique but may yield particles of similar density that are not EVs and can generate EV aggregates7. SEC has gained popularity due to its ability to provide higher purity samples by excluding particles based on size rather than density8. However, careful selection of the appropriate pore size for the SEC column and optimization of chromatography conditions are essential to minimize co-isolation of unwanted particles like chylomicrons and low-density lipoproteins8. Despite their effectiveness, both methods are time-consuming and challenging to automate, especially for larger volume samples like cell media or urine, limiting their scalability for industrial applications9.
In recent years, asymmetric field flow field fractionation (A4F) has evolved as a powerful separation technique for size and buoyancy-based micro- and nanometer-sized particle separation10. The operational principle of A4F relies on a microfluidic channel endowed with a porous membrane at its base, generating a force exerted towards the membrane called cross-flow10. When combined with Brownian motion and Poiseuille flow inherent to the system, cross-flow facilitates efficient particle separation due to varying particle position within the flow dynamics11. Despite the benefits, this method is limited to sample volumes within the microliter range12 and requires an additional focusing step, extending the duration of the process10.
Over the last decade, microfluidics has gained prominence as a tool for rapid, efficient, and clinically reliable EV separation13. However, most microfluidic methods designed for EV separation are optimized for small-volume, high-concentration EV samples or depend on complex separation procedures14. Furthermore, within the field of microfluidics, polydimethylsiloxane (PDMS) is recognized as the golden standard material owing to its optical transparency, biocompatibility, and ease of use15. Nevertheless, its known propensity to absorb small lipophilic molecules, including EVs, can be problematic for its application in the EV field13.
Cyclic olefin copolymer (COC) is a frequently used material in microfluidics due to biocompatibility, small absorption of molecules, and high chemical resistance15. However, the fabrication of COC devices often involves complex processes or specialized equipment16. Alternatively, off-stoichiometry thiol-ene (OSTE) is a promising alternative to PDMS due to decreased absorption of small molecules, superior chemical stability, ease of fabrication, and scalable fabrication process17,18. However, due to complex connections to tubing, devices can be prone to leaking19.
The aim of this study was to engineer and fabricate a microfluidic device combining OSTE and COC and bifurcated A4F principle for EV separation from large-volume samples such as urine or cell media.
Sample collection was approved by the Latvian University Life and Medical Science Research Ethics Committee (decision N0-71-35/54)
NOTE: The materials used in this study are included in the Table of Materials file.
1. Three-dimensional (3D) printed mold fabrication
2. Preparation of the PDMS molds
3. Preparation of OSTE-COC top channel
4. Preparation of OSTE-COC bottom channel and device assembly
5. Device evaluation
6. Device setup
7. Device testing with standardized latex beads
8. Device testing with urine samples
9. Device testing with conditioned media
10. Isolation of EVs using ultracentrifugation
11. Isolation of EVs using size exclusion chromatography (SEC)
12. EV characterization
13. NTA
14. dsELISA for EV markers
We fabricated a microfluidic device using a 3D printed double negative mold (Figure 1) via soft-lithography (Figure 2A) for high throughput EV separation based on the bifurcated A4F principle (Figure 2B,C). The setup requires a pump and a flow-through station, as can be seen in Figure 3, for the isolation of EVs in an automated manner. Firstly, to evaluate the proof of concept of the devices, a mixture of polystyrene beads with diameters of 100 nm and 1000 nm was prepared to represent vesicles and fine cell debris, respectively10. Experiments were conducted with varying flow rates for the bead mix, both with and without bifurcating flow, to investigate the effect of linear velocity on separation efficiency. Across all experiments, the recovery of small beads remained consistent and above 90%10,showing the potential of the device to recover EVs.
Then, we assessed and compared the potential of the COC-OSTE device in isolating EVs from large volumes (>1 mL) from complex biofluids with minimal pre-processing. As such, urine from 10 healthy donors (Figure 4) and cell media from two different prostate cell lines (Figure 5) were used as a template to simultaneously isolate EVs following three different procedures: ultracentrifugation (UC), size exclusion chromatography (SEC), and the A4F microfluidics-based OSTE-COC device. After isolation, the total number of particles and their size distribution were assessed using nanoparticle tracking analysis (NTA). On average, the OSTE-COC device showed better total particle recovery from biofluids compared to UC, but the statistical significance was only achieved when combining the particle numbers from both ports (Figure 4A). In order to compare the device performance with other systems, R & L outlet together should be taken into consideration. As shown in Figure 4, L & R outlet together performance on recovering EVs outperforms UC and SEC. Separately, the L-port was designed to capture the small EV fraction, while the R-port was designed to collect the bigger EV fraction with other molecules of similar size. Interestingly, the recovery using the R-PORT of the OSTE-COC device was slightly higher than SEC and UC alone (Figure 4A). CD63 expression showed a similar pattern (Figure 4C). This finding indicates that the OSTE-COC device was more effective in total EV recovery. Equivalent size distribution was found between the different methodologies, except for UC, which shows a bigger particle size distribution (Figure 4B).
Comparable results were observed in cell media cultures. In both scenarios, the total particle recovery from both device ports exhibited superior performance compared to SEC or UC methodologies (as depicted in Figure 5A,B). Notably, EVs derived from PC3 cells demonstrated a distinct size distribution, with greater homogeneity in the L-PORT distribution when contrasted with other experimental groups (Figure 5C,D). Furthermore, the analysis of CD63 expression confirmed the higher EV recovery rates achieved using the COC-OSTE device (as illustrated in Figure 5E,F). A summary comparing the isolation characteristics of the different methodologies examined in this study can be found in Table 1.
Figure 1: 3DP serpentine-shaped double negative mold dimensions and isometric view. Please click here to view a larger version of this figure.
Figure 2: COC-OSTE microfluidic device. (A) Scheme of the different main steps of fabricating the OSTE-COC device. (B) Device working principle. (C) Image of the finished device. Scale bar: 15 mm. This figure has been modified with permission from Priedols et al.10 and Bajo-Santos et al.20. Please click here to view a larger version of this figure.
Figure 3: Experimental configuration of the device. The syringe pump is on the left, the OSTE-COC device is in the middle, and the recovery station is on the right. This figure has been modified with permission from Priedols et al.10 and Bajo-Santos et al.20. Please click here to view a larger version of this figure.
Figure 4: Urinary EV size distribution and particle recovery from 10 donors using ultracentrifugation (UC), size-exclusion chromatography (SEC), and the COC-OSTE device. (A) Particle amount recovered from each isolation method by nanoparticle tracking analysis (NTA). Data represented as mean +/- standard deviation. Statistical significance denoted by * (p<0.05). (B) Boxplots display the average particle size distribution among all urine samples assessed by NTA, with whiskers indicating the minimum and maximum values. P-values were derived from comparisons to UC, with **** indicating high statistical significance (p<0.0001). (C) The median and range of the average CD63 amount, assessed using double-sandwich enzyme-linked immunosorbent assay (dsELISA), for each isolation method, were calculated across all samples. L-port: Left port; R-port: Right port. This figure has been modified with permission from Bajo-Santos et al.20. Please click here to view a larger version of this figure.
Figure 5: Characterization of EVs isolated from PC3 and LNCaP cells using different methods. (A) Particle amount recovered from PC3 media using nanoparticle tracking analysis (NTA) represented by mean+/- standard deviation. (B) Particle amount recovered from LNCaP cell media using NTA represented as mean +/- standard deviation. (C) Median particle size distribution of EVs from PC3 cultures, along with the range, was determined by NTA. (D) Median size distribution of EVs isolated from LNCaP cultures with range. (E) Median and range of CD63 expression for each isolation method for PC3 cell line-derived EV, using double-sandwich enzyme-linked immunosorbent assay (dsELISA). (F) Median and range of LNCaP-derived EV CD63 expression by dsELISA for each isolation method. UC: Ultracentrifugation; SEC: Size-exclusion chromatography; L-port: Left port; R-port: Right port. This figure has been modified with permission from Bajo-Santos et al.20. Please click here to view a larger version of this figure.
UC | SEC | COC-OSTE | |
Processing time/sample | ++/+++ | +++ | + |
Throughput | ++ | + | +++ |
Cost/sample | + | +++ | ++ |
Purity | + | +++ | ++ |
Automatization | ++ | + | +++ |
EV Yield | ++ | ++ | +++ |
Size selection | + | ++ | ++ |
Table 1. Comparison of isolation characteristics of EVs with the three methods UC SEC, and the COC-OSTE device. UC: Ultracentrifugation; SEC: Size-exclusion chromatography. COC-OSTE: cyclic olefin copolymer-off-stoichiometry thiol-ene. +: low; ++: medium; +++: high. * Device dependent.
The presented microfluidic device offers a promising method for the isolation and extraction of EVs from biological fluids, addressing some of the critical limitations of existing gold standard methods such as UC and SEC12. UC and SEC are known to be labor-intensive, time-consuming, and suffer from low yield, making them less suitable for high-throughput applications where large quantities of EVs are needed21,22. In contrast, the microfluidic device described can continuously process a substantial volume of up to 20 mL of biological fluid with minimal user input, making it a potential game-changer for industrial or clinical settings. One of the key advantages of this microfluidic device is its ability to standardize EV isolation and improve reproducibility compared to traditional methods that involve manual steps. By reducing the variability of EV isolation device manufacturing through mass manufacturing, the device's performance can be better controlled and consistent, which is vital for applications like EV-based therapeutics and diagnostics. Furthermore, the device's compatibility with large-volume manufacturing via reaction injection molding of suitable substrates further enhances its potential for scalability and broader adoption.
To ensure reproducibility and consistent performance, an extensive and well-defined protocol is necessary, both to produce the chip and for its usage in experiments. The chip is designed with industrialization in mind. However, during the device fabrication in the laboratory environment, attention must be paid to avoiding the formation of bubbles, ensuring precise and robust bonding, and conducting thorough testing at higher flow speeds than those expected in real experiments. This precaution is especially important for COC-OSTE devices, as they can be prone to leaking due to unsuccessful bonding, especially at higher flow rates. Additionally, since OSTE is a photosensitive material, the device fabrication should be performed in yellow light to avoid unnecessary exposure to UV. As this technology is still in its initial stages, researchers should experiment with different flow speeds to identify the optimal setting for their specific application since standardized protocols for this novel method are yet to be established, and liquid viscosity can clearly affect EV isolation efficacy based on our results. By following these guidelines and addressing the challenges of standardization, the potential of this microfluidic device for EV isolation can be fully realized across various research areas and applications.
The method of EV isolation using the microfluidic device allows for extensive modification and troubleshooting to optimize performance. To improve particle separation, researchers can explore longer channels, different membrane porosities and pore densities, and altered dimensions. If there are a lot of large particles in the small particle outlet, testing a higher buffer inlet speed can be beneficial, while if there are many small particles in the large particle outlet, experimenting with a longer channel or a membrane with larger pores is recommended. Addressing particle absorption can involve adjusting flow speeds, surface modifications, and OSTE composition or changing the UV exposure time. For devices experiencing instability and leakage, decreasing flow speeds and ensuring proper gluing are crucial steps. To minimize EV damage, researchers can reduce flow speeds, wash with PBS more thoroughly after device sterilization, and consider using smaller membrane pores or alternative mold design techniques and materials. Through these adjustments, the microfluidic device can be fine-tuned, unleashing its potential for diverse applications in EV research and related fields. Additionally, to optimize the device's performance, upstream sample characterization, including density and viscosity measurements, could be incorporated to adjust flow parameters. Moreover, particular emphasis must be placed in the nature of the sample, given the established non-sterility of biofluids23. Consequently, the sterilization of samples before their administration in clinical applications should be duly considered.
Despite these challenges, the microfluidic device holds immense potential in various research areas. Its ability to continuously isolate EVs from large volumes of highly heterogeneous samples opens the door for automation, facilitating high-throughput applications and enabling more in-depth analysis of EVs from diverse sources. Integrating this microfluidic module into different devices, such as hollow fiber cell bioreactor cartridges or high-resolution flow cytometers, could make EV applications more industry-friendly and drive innovation in fields like drug delivery, diagnostics, and cosmetics4. Overall, this microfluidic device represents a significant step forward in the field of EV research, offering an efficient and standardized method for EV isolation with promising applications in a wide range of research an industrial settings.
The authors have nothing to disclose.
We thank all the donors who participated in this study, the staff of the Latvian Genome Database for providing the samples. The Institute of Solid-State Physics, University of Latvia as the Center of Excellence has received funding from the European Union's Horizon 2020 Framework Programme H2020-WIDESPREAD-01-2016-2017-TeamongPhase2 under grant agreement No. 739508, project CAMART2. This work was supported by The Latvian Council of Science Project No. lzp-2019/1-0142 and Project No: lzp-2022/1-0373.
0.1 µm carboxylate FluoSpheres | Invitrogen | #F8803 | Stock concentration: 3.6 x 1013 beads/mL (LOT dependent) |
0.5 mL microcentrifuge tubes | Starstedt | 72.704 | |
1 mL Luer cone syringe single use without needle | RAYS | TUB1ML | |
1.0 µm polystyrene FluoSpheres | Invitrogen | #F13083 | Stock concentration: 1 x 1010 beads/mL (LOT dependent) |
10 mL Serological pipettes | Sarstedt | 86.1254.001 | |
15 mL (100k) Amicon Ultra centrifugal filters | Merck Millipore | UFC910024 | |
2.0 mL Protein LoBind tubes | Eppendorf | 30108132 | |
20 mL syringes | BD PlastikPak | 10569215 | |
250 µm ID polyether ether ketone tubing | Darwin Microfluidics | CIL-1581 | |
3 kDa MWCO centrifugal filter units | Merck Millipore, | UFC200324 | |
5 mL Medical Syringe without Needle | Anhui Hongyu Wuzhou Medical | 159646 | |
50 mL conical tubes | Sarstedt | 62.547.254 | |
70 Ti fixed angle ultracentrifuge rotor | Beckman Coulter | 337922 | |
800 µm ID polytetrafluoroethylene tubing | Darwin Microfluidics | LVF-KTU-15 | |
96 well microplate, f-bottom, med. binding | Greiner Bio-One | 655001 | ELISA plate |
B-27 Supplement (50x), serum free | Thermo Fisher Scientific | 17504044 | |
Bovine serum albumin | SigmaAldrich | A7906-100G | |
COC Topas microscopy slide platform | Microfluidic Chipshop | 10000002 | |
COC Topas microscopy slide platform 2 x 16 Mini Luer | Microfluidic Chipshop | 10000387 | |
Elveflow OB1 pressure controller | Elvesys Group | ||
Luer connectors | Darwin Microfluidics | CS-10000095 | |
Mask aligner Suss MA/BA6 | SUSS MicroTec Group | ||
Mixer Thinky ARE-250 | Thinky Corporation | ||
NanoSight NS300 | Malvern Panalytical | NS300 | nanoparticle analyzer |
Optical microscope Nikon Eclipse LV150N | Nikon Metrology NV | ||
OSTE 322 Crystal Clear | Mercene Labs | ||
PBS TABLETS.Ca/Mg free. Fisher Bioreagents. 100 g | Fisher Scientific | BP2944-100 | |
PC membrane (50 nm pore diameter, 11.8% density) | it4ip S.A., Louvain-La Neuve, Belgium | ||
Petri dishes, sterile | Sarstedt | 82.1472.001 | |
Plasma Asher GIGAbatch 360 M | PVA TePla America, LLC | ||
qEVoriginal/35 nm column | Izon | SP5 | SEC column |
QSIL 216 Silicone Elastomer Kit | PP&S | ||
Resin Tough Black | Zortrax | ||
SW40 Ti swing ultracentrifuge rotor | Beckman Coulter | 331301 | |
Syringe pump | DK Infusetek | ISPLab002 | |
T175 suspension flask | Sarstedt | 83.3912.502 | |
TIM4-Fc protein | Adipogen LifeSciences | AG-40B-0180B-3010 | |
TMB (3,3',5,5'-tetramethylbenzidine) | SigmaAldrich | T0440-100ML | Horseradish peroxidase substrate |
Tween20 | SigmaAldrich | P1379-100ML | |
Ultracentrifuge Optima L100XP | Beckman Coulter | ||
Ultrasonic cleaning unit P 60 H | Elma Schmidbauer GmbH | ||
Universal Microplate Spectrophotometer | Bio-Tek instruments | 71777-1 | |
Urine collection cup, 150mL, sterile | APTACA | 2120_SG | |
Whatman Anotop 25 Syringe Filter | SigmaAldrich | 68092002 | |
Zetasizer Nano ZS | Malvern Panalytical | dynamic light scattering (DLS) system | |
Zortrax Inkspire | Zortrax |