An Integrated Digital Microfluidic Device for the Extraction and Detection of Extracellular Vesicle‐Based Molecules

Abstract

Extracellular vesicles (EV) are phospholipid‐encapsulated nanoparticles secreted by cells into their surrounding environment. EVs can transfer a variety of biomolecules that mediate intercellular communication and play a key role in physiological and pathological processes. Therefore, EVs are emerging as new biomarkers for diseases, therapeutic targets, and drug delivery vehicles. The isolation and detection of EVs requires time‐consuming and labor‐intensive processes, first to extract EVs from biological and physiological fluids and then to detect EV‐associated molecules with high sensitivity. The methodologies and instruments commonly used for EV analysis are not widely accessible outside of dedicated research laboratories, creating practical barriers for the study of EV‐associated molecules in biology research and clinical applications. To bridge this gap, we developed a proof‐of‐concept digital microfluidic device that can automatedly extract EVs from 20 µL of culture and human plasma samples within 25 min and detect EV‐bound proteins (e.g PD‐L1) on‐chip using an electrochemical sensor. This work serves as a framework for the development of streamlined EV analysis in both research and diagnostics.

This work presents a proof‐of‐concept digital microfluidic device that can automatedly extract EVs from microliters of biological and physiological fluids within 25 min and detect EV‐bound protein markers on‐chip using electrochemical sensor. This work serves as a framework for the development of streamlined EV analysis in both medicine research and clinical diagnostics.

1. Introduction

Extracellular vesicles (EV) are phospholipid‐encapsulated nanoparticles secreted by cells into their surrounding environment. These nanoscale membraned structures encompass a plethora of subtypes according to varied factors including origin, content, function, and biogenesis.^{[ 1 ]} On one hand, EVs can transfer a variety of bioactive molecules (e.g., protein, nucleic acids, lipids) that mediate, for example, intercellular communication between microbial and immune cells, processes of inflammation, antigen presentation, T and B cell development, and modulate systemic innate and adaptive immunity.^{[ 1} , ^{2 ]} On the other hand, EVs are also found in various biofluids (e.g., blood, saliva, urine) and they are emerging as a new class of non‐invasive biomarkers for diseases ranging from infections, immunological disorders, cancer, and sepsis.^{[ 3} , ⁴ , ^{5 ]}

Analysis of EV subpopulations from complex biofluids relies on extraction and detection using surface markers. Standard EV extraction relies on differential and density gradient ultracentrifugation, ultrafiltration size exclusion chromatography and precipitation‐based methods, and EV detection often uses surface markers against specific subpopulations. These processes are time‐consuming and labor‐intensive and are typically performed using bulky equipment in dedicated laboratories.^{[ 6} , ⁷ , ^{8 ]} More recently, nanoscale flow cytometry offers high resolution in detecting specific subsets of these EVs,^{[ 5} , ^{9 ]} however, the costly instrument is often found in core facilities, and data analysis often requires trained specialists. This creates a practical barrier for studies that require using these instruments to frequently monitor and evaluate EV‐based targets, whether those that are, for example, secreted by cells cultivated in controlled conditions, or in patients who receive cancer immunotherapies.^{[ 10} , ^{11 ]}

Therefore, recent attention has focused on microfluidic technologies to extract EVs due to their ability to handle microscale volumes and automate sophisticated processes in a small device. Passive particle separation methods are common in microfluidic devices to isolate EVs based on their sizes, such as through deterministic lateral displacement, inertial forces, and flow fractionation.^{[ 12} , ¹³ , ^{14 ]} Active methods, for example, ultrasound waves, have also been used in these devices to trap, separate, focus, and transport EVs by inducing differential forces on EVs based on their sizes.^{[ 15} , ¹⁶ , ^{17 ]} Yet, size‐based EV isolation often leads to lower purity because contaminants that have similar sizes can also be isolated undesirably, meanwhile, clogging can be another challenge. To overcome this limitation, electrokinetic methods have been explored as an alternative, because electrokinetic forces of different magnitudes can act on these EVs based on not only their sizes, but also the dielectric constant and the charge density of the EVs and their ambient medium.^{[ 18} , ^{19 ]} However, these methods often require integrating microscale electrodes at precise locations in microfluidic devices and modulating high alternating electric fields on these electrodes in a highly controlled manner, in addition, the use of un‐insulated electrodes in aqueous fluids risks electrolysis, which can fail the devices.

To overcome these limitations, we developed a digital microfluidic (DMF) device that can support the automated EV extraction from varied biofluids and the detection of specific subsets based on their surface markers. Distinct from many microfluidic platforms, DMF devices are composed of insulated electrode arrays that handle fluids based on electrowetting principles.^{[ 20} , ^{21 ]} This unique construct allows for the programmed handling of droplets (e.g., split, mix, move) in a timed manner without the need for physical microstructure (e.g., microvalves, microchannels).^{[ 22} , ^{23 ]} Therefore, these devices can enable users to perform multi‐step assays that require, for example, target extraction, wash, and incubation, in an automated manner with user‐defined parameters.^{[ 24} , ²⁵ , ^{26 ]} In this device, we used immunomagnetic beads to extract EVs on‐chip as immunoaffinity offers higher specificity for EV subsets,^{[ 27} , ^{28 ]} and then use electrochemical sensors to detect their surface markers.

In this proof‐of‐concept, we chose the immune checkpoint molecule PD‐L1 as EV‐bound target because its circulation in blood has been linked to resistance to PD1/PD‐L1 blockade^{[ 29} , ^{30 ]} – a therapy at the forefront of immunotherapies.^{[ 31} , ^{32 ]} Here, this DMF device can automatedly extract EVs from 20 µL of cell culture and human plasma samples within 25 min and detect PD‐L1⁺ EVs on‐chip using an electrochemical sensor, with a detection limit of 1 ×10⁴ EVs/mL. This work shows the feasibility of streamlining EV extraction from varied biofluids and the detection of specific EV subpopulations for biology research and clinical applications.

2. Results and Discussion

2.1. DMF Platform Overview

Overall, the DMF device consisted of two glass substrates, with a gap in‐between, where droplets were handled (Figure 1 ). The bottom substrate was patterned with arrays of actuation electrodes, reservoir electrodes, and electrical contact pads on the sides, the top substrate was coated with indium tin‐oxide (ITO) – a transparent conductive material that functions as the ground electrode of the device, while fluids can be seen from the top. The DMF device was controlled by a Dropbot system (Sci‐bots, Toronto, ON) that can program the electrodes to turn on/off to handle the droplets in a timed manner. Samples and reagents (e.g., culture medium, plasma, magnetic beads) can be pipetted onto different reservoir electrodes and then split into smaller droplets and handled on the actuation electrodes based on electrowetting principles. In this work, we performed varied functionalities in a single device, which include functionalizing magnetic beads, capturing EVs from biofluids, sample washing, and EV elution. We then inserted an Au wire‐based electrochemical sensor into the device to detect PD‐L1⁺ EVs.

Figure 1.

Overview of the DMF device that can handle droplets based on the electrowetting principle. The platform was used for sample preparation: first, microliters of biofluids (e.g., culture medium, plasma) can be loaded into the device, and magnetic beads can then be functionalized on‐chip to capture EVs in these biofluids. The captured EVs can then be released, and gold wire‐based electrochemical sensors can be inserted into the device to detect EV surface molecules.

2.2. On‐Chip TIM‐4 Magnetic Beads Preparation

In this work, we used immunomagnetic beads to extract EVs on‐chip because the immunoaffinity offers higher specificity in target purification. Overall, immunoaffinity‐based EV extraction often relies on antibodies to target a specific surface marker. Because there lacks a surface marker that is universal to all EVs, immunoaffinity‐based methodology often targets at several common surface proteins, namely, CD9, CD63, and CD81, as these proteins belong to the tetraspanin family that is involved in the formation of the multivesicular bodies from which EVs originate.^{[ 33} , ^{34 ]} However, detaching these EVs that are captured on antibody‐functionalized beads for downstream analysis remains a challenge. This is because breaking the antigen‐antibody interaction (e.g., hydrogen bonds, electrostatic bonds, Van der Waals forces) typically requires extreme pH (e.g., pH 2.5 – 3.5) and high salt conditions (e.g., 1M NaCl), which often induce conformational changes in the antibodies and the surface antigens of the EVs and thus their functions.^{[ 35} , ^{36 ]}

Therefore, we used a phosphatidylserine (PS) – T cell immunoglobulin mucin protein‐4 (TIM‐4)‐affinity method to enable the extraction of EVs at high purity, which offers an intrinsic way to detach these EVs from further analysis. This is because PS is a negatively charged phospholipid often expressed on the inner layer of the lipid membrane of EVs and can flip to the outer membrane, and it is increasingly recognized as a new class of biomarker for EV detection and analysis.^{[ 37} , ^{38 ]} TIM‐4 can bind to PS on EVs in a calcium ion (Ca²⁺) dependent manner, therefore, a key merit is that the extracted EVs can be easily eluted from TIM‐4 protein in the presence of Ca²⁺ chelating agents for varied downstream analysis (Figure 2a).^{[ 39 ]}

Figure 2.

On‐chip EV extraction. a) An illustration of on‐chip magnetic beads functionalization for EV capture, and elution of EVs for downstream analysis. b) Optimization of on‐chip incubation with different lengths of incubation time (N = 3), c) Comparison of in‐tube and on‐chip EV extraction efficiency (N = 4).

First, we functionalized TIM‐4 on immunomagetic beads in the DMF device. Briefly, a 10 µL droplet of streptavidin‐coated magnetic beads (FUJIFILM Wako, Osaka, Japan)^{[ 40} , ^{41 ]} was dispensed from a reservoir electrode on‐chip, and the beads were retained by a magnet while the supernatant was split and removed. A second droplet that contained biotin‐labeled TIM‐4 (FUJIFILM Wako, Osaka, Japan) suspended in wash buffer (FUJIFILM Wako, Osaka, Japan) was then dispensed and mixed with the magnetic beads to immobilize TIM‐4 through biotin‐streptavidin conjugation. Multiple 20 µL droplets of wash buffer were then used in the same manner to remove the unbound molecules. The beads were resuspended in a 20 µL droplet of wash buffer until the start of EV isolation process. Note that many microfluidic processes use functionalized beads prepared off‐chip beforehand; this work streamlined the bead functionalization process with EV extraction, showing the feasibility of functionalizing different biorecognition agents onto beads for specific tests on demand.

2.3. On‐Chip EV Extraction from Cell Culture Media

First, we tested the feasibility of using our DMF device to extract EVs secreted by cells cultivated in medium with controlled conditions. In this proof‐of‐concept, we cultivated human breast cancer cell line (MDA‐MB‐231, System Biosciences, Palo Alto, CA, United States) for EV extraction, because these cells are highly secretory and often produce a wide variety of molecules (e.g., interleukins, growth factors) that can interfere with the on‐chip EV isolation process. Briefly, a total 18 µL sample was mixed with binding buffer (FUJIFILM Wako, Osaka, Japan) and Tetronic 904 (Florham Park, NJ, USA), which adjusted the final volume to 20 µL. The sample was then loaded into the DMF device with a pipette. The loaded sample was mixed with a 10 µL droplet of TIM‐4‐magnetic beads prepared on‐chip, and the mixture was incubated for 20 min. The magnetically retained beads were then washed with drops of wash buffer (FUJIFILM Wako, Osaka, Japan) for 3 times. After that, a 20 µL droplet of elution buffer (FUJIFILM Wako, Osaka, Japan) was then dispensed to mix with the beads and incubated for 5 min. In the same manner, the beads were aggregated with a magnet while the liquid was separated from the beads and moved onto a reservoir electrode for collection. Experimental images were shown in Figure S1 (Supporting Information) to illustrate this process.

In a unique way, DMF devices can handle droplets by turning the electrodes on/off automatedly. This offers an intrinsic merit to constantly agitate the droplets, such as during target capture, sample wash, and elution to enhance the mixing and binding efficiency. Our previous work showed that the turbulence created in droplets can enhance target binding efficiency by up to 3X.^{[ 42} , ^{43 ]} To optimize the incubation process, we performed on‐chip incubation at different time durations (5, 10, 20, 40 min). The beads were subsequently collected, and the captured EVs were lysed to release their protein content, which was then quantified using Micro BCA™ Protein Assay Kit (Thermo Scientific™, Waltham, MA, United States). Each test was repeated for 3 times, and results showed that constantly mixing samples and beads for 20 min has led to an average of 0.514 µg mL⁻¹ total exosomal protein, further increasing the incubation time did not lead to significant increase in protein quantity (Figure 2b). Therefore, in the bead‐based EV capture process on‐chip, we incubated the sample for 20 min. In these experiments, we used 10 µL of magnetic beads (≈100 µg) in each EV isolation process. This amount was selected based on the affinity between TIM4 and PS as recommended by the manufacturer.^{[ 44 ]} To best ensure the outcome, we tested 10 µL (≈100 µg) and 60 µL (≈600 µg) of beads under the same conditions. Results showed no significant increase in the measured total EV protein amount (Figure S2, Supporting Information), in fact, further increased concentration of beads can lead to overloading and thus lower the capture efficiency. Therefore, we used 10 µL of magnetic beads (≈100 µg) throughout the study.

Standard in‐tube EV extraction requires up to 500 µL of sample and reagents, yet the on‐chip EV extraction method reduced sample and reagent volumes by up to 25X. Given the significantly reduced sample and reagent volume used on‐chip, we performed a systematic comparison between the on‐chip and the in‐tube EV isolation methods under matched experimental conditions. To ensure comparability, both methods used the same amount of magnetic beads, incubation time, buffer composition, and input sample volume. In this experiment, cell culture medium was collected from MDA‐MB‐231 cells, filtered, and then concentrated for subsequent processes. The initial sample (prior to EV extraction) and the flowthrough (supernatant after EV extraction process) were both analyzed using nanoparticle tracking analysis (NTA), and EV capture efficiency was calculated as (Total particles in the initial sample − Total particles in the flowthrough) / Total particles in the initial sample × 100%. Results showed that, on average, the on‐chip EV capture efficiency (50.6%) was comparable to that of the in‐tube process (45.6%) (Figure 2c).

To ensure the successful extraction of EVs from a complex biological sample that contains a wide range of cells, proteins, extracellular vesicles, and soluble factors, we used NTA, transmission electron microscopy (TEM), and Western blotting to verify the concentration, size distribution, and surface markers of the extracted EVs. First, we used NTA to quantify the size distribution and the concentration of the EVs, and results showed that 93.2 ± 4.0% of the EVs isolated on‐chip and 95.3 ± 2.8% of the EVs isolated in‐tube were ≤200 nm. (Figure  3a,b). The on‐chip extraction led to an EV concentration of (1.29 ± 0.23) × 10¹⁰ EVs/mL, while the in‐tube extraction led to an EV concentration of (1.29 ± 0.19) × 10¹⁰ EVs/mL (Figure 3c). Other microfluidic‐based methods reported the extraction of 10⁷–10⁹/mL EVs,^{[ 45} , ^{46 ]} however, the number of EVs depends on varied factors, including specific approaches as well as samples. As previously reported, 90% of the total EV population is represented by small EVs (<200 nm) and 10% by large EVs (>200 nm).^{[ 47 ]} Therefore, the size range of extracted EVs conformed with the size distribution of typical EVs. Furthermore, TEM imaging was used to visually confirm the morphology of the isolated EVs, which exhibited the expected spherical shape and lipid bilayer structure, which is consistent with previous reports (Figure 3d).^{[ 48 ]}

Figure 3.

Characterization of EVs. a,b) NTA results show the number of EVs extracted on‐chip and in‐tube, and their size distribution (different colors indicate different experiments), c) Comparison between on‐chip and in‐tube EV extraction experiments (N = 4), d) TEM images of extracted EVs, e,f) Western Blot to confirm the purity of the extracted EVs, g) Comparison of EVs extracted (in‐tube and on‐chip) from MDA‐MB‐231 wildtype (WT) and with PD‐L1 knocked out (KO).

In addition, we examined the purity of the extracted EVs and assessed the presence of contaminants from the source samples. Briefly, we performed Western Blot on the extracted EVs to examine a set of markers, including calnexin, CD9, and flotillin‐1. Calnexin is an endoplasmic reticulum protein often used as a marker for assessing cellular contaminants in EV samples ^{[ 49} , ^{50 ]} CD9 is a classical surface marker on EVs, and flotillin‐1 is a protein associated with lipid rafts that is often enriched in EVs and minimally present in cellular components. GAPDH is often used as a loading control. Compared with EV extraction using standard methodology (e.g., ultracentrifugation) with regular sample volumes (e.g., 1–5 mL), we extracted EVs from 50 – 250X lowered volumes. This can lead to practical challenges in performing Western Blot analyses because of the significantly reduced protein amount (by ≈100X) compared with standard input (10–50  µg), especially when multiple protein markers are probed. First, to minimize cross‐reactivity and ensure specificity on the same blot, it is ideal to probe each target protein separately on the same membrane through sequential immunoblotting. However, this process involves repeated washing, re‐blocking, and re‐probing, which can compromise the integrity of the membrane and thus the quality of the signal. To best ensure the outcome of the Western Blot analysis with EVs isolated from a very small volume of sample, we probed the panel of protein markers (calnexin, CD9, flotillin‐1) in two separate Western Blot experiments.

First, calnexin and CD9 were probed in the EVs isolated on‐chip (Figure 3e), and peripheral blood mononuclear cell (PBMC) lysate was used as a control experiment and elution buffer as a blank control. This is because, while certain subsets of PBMCs (e.g., 10%) can express CD9, this surface marker is more often enriched in EVs, meanwhile, calnexin is often found in cellular components and minimally present in EVs, offering an effective means of assessing EV purity. Results confirmed minimal cellular contaminants and the presence of CD9 in EVs, verifying the purity of the extracted EVs.

Then, calnexin and flotillin‐1 were probed in a second experiment, in which breast cancer cell lysate (MDA‐MB‐231) was used as a control, of which the EVs were derived from. Flotillin‐1 is often highly expressed in EVs, and these molecules can also be found in the cell lysate. This is because flotillin‐1 is primarily located in lipid rafts, plasma membrane microdomains, and endosomes, therefore the disruption of the cellular compartments can lead to the release of these molecules into the lysate,^{[ 51} , ^{52 ]} Meanwhile, to assess the level of non‐specific binding, biotinylated BSA (Thermo Fisher Scientific, USA) was used to functionalize magnetic beads (instead of biotinylated TIM‐4) as a negative control, with the same MDA‐MB‐231 culture medium used as the input sample. In the EVs extracted from these cells, the presence of flotillin‐1 confirmed the enrichment of EVs, the absence of calnexin indicated minimal presence of cellular contaminants, and the negative control showed minimal non‐specific capture. Taken together, these results further verified the purity of the extracted EVs (Figure 3f).

To ensure that the target subpopulation of EVs (PD‐L1⁺ EVs) was successfully extracted using this approach, we used our device to extract EVs from the supernatant of MDA‐MB‐231 cell culture and the cells with PD‐L1 knocked out (Figure 3g). Results showed that overall, the number of total EVs extracted from both samples in our device was comparable with standard in‐tube extraction. Results showed that among EVs secreted by wildtype cells, 7.12% expressed PD‐L1, while only <0.1% PD‐L1⁺ EVs were present in those that were secreted by cells with PD‐L1 knocked out. In the post‐extraction analysis, results showed that 3.18% of the EVs were PD‐L1⁺ in the wildtype sample, and <0.1% were PD‐L1⁺ in the knocked‐out sample. Nanoscale flow cytometry results (forward scatter plots) were shown in Figure S3 (Supporting Information).

2.4. Characterization of the Gold Nanoparticle Functionalized Working Electrode

Gold electrodes have been widely used in electrochemical sensors due to their excellent electrical conductivity and biocompatibility. To further enhance the sensitivity of our electrochemical sensor for detecting specific EV subsets, we modified the surface of the gold working electrodes with gold nanoparticles to amplify the electrical signals and enhance sensitivity in electrochemical detection. Briefly, gold nanoparticles were prepared through the reduction of gold precursor HAuCl₄ using trisodium citrate, as detailed in the methods section. To ensure the quality of the synthesized gold nanoparticles, we used TEM to verify the morphology of the nanoparticles, and their diameter averaged ≈15 nm (Figure 2b; Figure S4a, Supporting Information). The gold nanoparticles displayed a characteristic absorption peak at a wavelength of ≈ 520 nm in the UV–vis spectrum owing to the surface plasmon vibration of gold particles (Figure S4c, Supporting Information).

The bare gold wire electrode was then functionalized with the synthesized gold nanoparticles via the interaction of thiol group and gold atoms. We then characterized the electrical performance of the gold nanoparticle‐modified working electrode by cyclic voltammetry (CV) measurements using the electrochemical active probe KFe₃(CN)₆ mixed with supporting electrolyte KCl solution. The anodic/cathodic peak current ratio of the redox probe was close to 1, which indicated the reversible electron transfer reaction (Figure S5a, Supporting Information). Compared with a bare gold working electrode, the current response of the gold nanoparticle‐modified working electrode was enhanced by ≈2X (Figure S5b, Supporting Information), indicating a boost in the electron transfer capability of the gold working electrode. This enhanced electrical property of gold nanoparticle‐modified working electrode was mainly attributed to the excellent catalytic activity of the gold nanoparticles, because these nanoparticles can facilitate the transfer of electrons from the electrolyte to the surface of the working electrode, which leads to increased electrical current. Furthermore, these nanoparticles can roughen the surface of the working electrode, which can provide an increased number of binding sites for biorecognition molecules (e.g., antibodies, aptamers).

To detect a subset of the EVs eluted from magnetic beads, CD9 antibodies were on the working electrode to re‐capture EVs on the working electrode. The binding of antibodies on the working electrode led to a decrease in electrical current in CV measurement (Figure S5, Supporting Information). This was because the antibodies had poor electrical conductivity, which hindered the electron transfer during the reaction of the electrochemical probe. These outcomes verified that CD9 antibodies were immobilized on the working electrode.

To investigate the time‐longitudinal stability of the modified working electrode, we performed CV measurements in a mixture of KFe₃(CN)₆ and KCl. Over 40 days, the working electrode maintained ≈98% of the initial electrochemical response with a relative standard deviation (RSD) of 2.03% (Figure S6, Supporting Information). This is promising, because the stability of these working electrodes indicated superior reproducibility of electrochemical sensors over time, and there holds potential for mass‐producing these sensors and storing them until use in DMF devices.

Using these wire‐based electrodes as plug‐in electrochemical sensors in DMF devices has several advantages. First, this can bypass the need for integrating microscale sensing electrodes within the square actuation electrodes in DMF devices, which often requires laborious and time‐consuming microfabrication in cleanroom settings.^{[ 43 ]} Instead, these gold wire‐based electrochemical sensors can be mass‐produced in a relatively simple manner. Second, DMF devices and electrochemical sensors can be independently built, enabling the use of universal DMF device layouts. Meanwhile, varied biorecognition elements (e.g., antibodies, aptamers, peptides) can be immobilized on the surface of the working electrodes on demand and insert into a DMF device to detect specific targets, which links upstream on‐chip sample preparation and downstream detection in a smooth manner.

2.5. Constructing an Electrochemical Immunosensor for PD‐L1⁺ EVs

Among the extracted EVs, we chose PD‐L1⁺ EVs as the target subset for detection in this proof‐concept platform because of their essential roles in biology research and clinical applications. For example, tumor‐derived PD‐L1⁺ EVs in host macrophages can impede antibacterial immunity, which warrants further research, meanwhile, these EVs also indicate one's responsiveness to PD1/PD‐L1 cancer immunotherapy.^{[ 29} , ^{30 ]}

Therefore, based on the surface‐modified working electrode, we constructed an electrochemical immunosensor to detect PD‐L1 expression level on EVs extracted on‐chip (Figure 4a). Briefly, anti‐CD9 was immobilized on the surface of gold nanoparticles functionalized gold wire via EDC/NHS coupling chemistry to re‐capture the EVs eluted from the magnetic beads. Then, biotin‐modified anti‐human PD‐L1 antibody (detection antibody) was added to label PD‐L1⁺ EVs to form sandwich‐structured biocomplexes, which were then bound to streptavidin‐poly‐HRP. TMB, the electrochemical probe, was then added so that the oxidation could occur for electrochemical measurement. Here, poly‐HRP was used to amplify the electrochemical signals due to its high catalytic activity. Meanwhile, the gold nanoparticles that were modified on the wire‐based electrode served as another means for signal amplification, further enhancing the sensitivity of the sensor.

Figure 4.

a) Schematic illustration of PD‐L1⁺ EV detection using the designed electrochemical sensor. Calibration plots and selectivity of PD‐L1⁺ EV. (b) Amperometry plots (i∼t) of PD‐L1⁺ EV detection at different concentrations. c) Current signal versus concentration plots of PD‐L1⁺ EV at various concentrations. d) Current intensity ratio caused by soluble PD‐L1 (sPD‐L1), EVs with PD‐L1 knocked out (PD‐L1⁻ EV), and PD‐L1⁺ EV. (N = 3)

We then tested the feasibility of using the gold wire‐based electrochemical immunosensor for on‐chip EV detection. Briefly, we inserted the gold wire‐based immunosensor from the side of the DMF device to reach the electrochemical detection zone. Electrochemical measurements were then performed to detect PD‐L1⁺ EVs among EVs that were extracted on‐chip (Figure S7a, Supporting Information). Compared with off‐chip electrochemical detection, performing electrochemical measurement on‐chip significantly reduced the required reagent volumes by 10X. Results show that the electrical current signal from the sample on‐chip was significantly higher from the blank control, indicating the feasibility of detecting PD‐L1⁺ EV on chip (Figure S7b, Supporting Information).

2.6. Characterization of the Electrochemical Immunosensor

We characterized the performance of the electrochemical immunosensor and performed PD‐L1⁺ EV detection. In cyclic voltammogram measurement, two pairs of reduction peak were observed due to the poly‐HRP‐catalyzed TMB oxidation (Figure S8, Supporting Information), which were typical characteristics of two‐electron reduction and oxidation of TMB. The reduction peak current of oxidized TMB significantly increased at ‐0.1 V, leading to an asymmetric reduction peak, which confirmed the occurrence of TMB electrocatalysis. The voltage of the oxidized TMB was fixed at ‐0.1 V to reduce oxidized TMB when quantifying the PD‐L1⁺ EV. In these tests, the electrochemical response gradually increased with the concentration of PD‐L1⁺ EV (Figure 4b), and there was a good linear range between 1 × 10⁴ to 1 × 10⁸ EVs/mL (Figure 4c). Based on these results, the limit of detection was calculated to be 3.84 ×10³ EVs/mL, which was consistent with the outcome of the measurement. Further lowered concentration of PD‐L1⁺ EV (as low as 1 × 10³ EVs were tested) remained outside of the detection limit. To examine the reproducibility of the immunosensor, we performed parallel measurements of five samples that contained PD‐L1⁺ EVs (10⁶ EVs/mL). Results show that the variation in current response across these samples remained insignificant (Figure S9a, Supporting Information), with an RSD of 5.05% (Figure S9b, Supporting Information).

While PD‐L1 molecules can be expressed on EVs, these molecules also exist in free form as soluble protein molecules. These two forms of PD‐L1 molecules have distinct indications in EV biology and disease diagnosis, and it is therefore crucial to detect PD‐L1⁺ EV while limiting interference from soluble PD‐L1 molecules. To ensure the specificity of the detection, we used our electrochemical immunosensors to measure soluble PD‐L1 in the extracted EVs. In most cases, the level of soluble PD‐L1 in human plasma is estimated to be <100 pg mL⁻¹,^{[ 53 ];} therefore, we evaluated the interference of 100 pg mL⁻¹ soluble PD‐L1 when detecting PD‐L1⁺ EVs (Figure S10, Supporting Information). Compared with PD‐L1⁺ EV, the current response from soluble PD‐L1 was ≈7X less (Figure 4d). It is mainly because this strategy limits the undesired capture of soluble PD‐L1 molecules, as we first extracted EVs from human plasma in a non‐specific manner (using PS surface marker), and then used CD9 antibody to re‐capture these EVs on the working electrode to detect those that expressed PD‐L1. Meanwhile, we further validated the experiments by using EVs with PD‐L1 knocked out (PD‐L1⁻ EVs) as another control (Figure S10, Supporting Information). Results showed that the signals from PD‐L1⁺ EV were ≈6X higher than those of EVs with PD‐L1⁻ EVs (Figure 4d), thus confirming the selectivity of the sensor.

2.7. On‐Chip EV Extraction from Human Plasma

To test the feasibility of extracting EVs from body fluids that consist of complex cellular and molecular components using this approach, we used the DMF device to extract EVs from human plasma samples. Overall, the on‐chip process remained the same as described earlier. A key aspect in processing plasma samples is to minimize bead aggregation, as this can reduce the surface area for EV binding. Therefore, plasma samples were first centrifuged at 8,000 xg for 10 min and then filtered (0.22 µm) to remove debris present in the sample. A total of 18 µL of processed plasma sample was used in each experiment. As described earlier, the sample was mixed with binding buffer (FUJIFILM Wako, Osaka, Japan) and Tetronic 904 (Florham Park, NJ, USA), which adjusted the final volume to 20 µL. Plasma samples were collected from five immunotherapy‐refractory melanoma patients and three healthy controls. The level of PD‐L1⁺ EVs in each sample was first measured using flow cytometry.

To evaluate the performance of the microfluidic device for extracting EVs in plasma samples, we first compared the efficiency of on‐chip and in‐tube EV extraction methods using commercially obtained plasma samples from healthy donors, following the same analysis as described earlier. Results showed that the on‐chip extraction method led to (1.84 ± 0.44) × 10¹⁰ EVs/mL, and the in‐tube method led to (1.13 ± 0.17) × 10¹⁰ EVs/mL (Figure 5a), and the EV capture efficiency was 27.7% on‐chip and 20.8% in‐tube (Figure 5b). The substantially lowered particle counts observed in the control group using biotinylated BSA in place of biotinylated TIM4 under otherwise identical in‐tube conditions demonstrate the specificity of EV capture. Compared with culture medium, lowered EV recovery was observed in plasma samples, this can be attributed to the high abundance of non‐EV particles in plasma samples (e.g., lipoproteins and protein aggregates).^{[ 54 ]} These particles are very often challenging to distinguish from EVs in NTA characterization, leading to an overestimation of the initial concentration of particles loaded into the instrument. As a result, the calculated EV recovery efficiency may appear lower than the actual yield of extracted EVs.

Figure 5.

a) Feasibility of extracting EVs from 20 µL of human plasma samples (N = 4), b) EV capture efficiency of on‐chip and in‐tube approach (N = 4), c–e) Western Blot results showed the purity of the EVs extracted from human plasma samples, EV size distribution in f) 5 patients and g) 3 healthy controls (colored lines show different patients), Comparison of PD‐L1⁺ EV levels between 5 patients and 3 healthy controls measured by h) NTA, i) the electrochemical sensor, and j) nanoscale flow cytometry.

Western Blot was then used to confirm the presence of common EV surface markers. Three separate sets of Western Blot experiments were performed on the panel of protein markers in the same manner as described earlier to ensure the integrity of the blot membrane and the quality of the outcome. Briefly, first, calnexin and CD9 were probed in the EVs isolated on‐chip, and PBMC lysate was used as a control. While a varied fraction of PBMCs in human plasma (e.g., 10–30%) can express CD9, this surface marker is more often enriched in EVs, which was confirmed in the outcome. (Figure 5c). Meanwhile, in plasma samples, sometimes, a small fraction of PBMCs can exist, such as left from the upstream plasma separation process, however, the presence of CD9 marker and the absence of calnexin in EV samples confirmed the purity of the EVs extracted from human plasma samples. (Elution buffer was used as a negative control.)

Likewise, calnexin and flotillin‐1 were probed in the second experiment, in which breast cancer cell lysate (MDA‐MB‐231) was used as a control, of which the EVs were derived from. Biotinylated BSA was used as a negative control to assess non‐specific binding (Figure 5d). In the third experiment, apolipoprotein B (ApoB) was probed along with calnexin and flotillin‐1 (Figure 5e). This is because, ApoB is a key structural protein component of lipoproteins, and can be easily co‐isolated with EVs from plasma or serum, especially during ultracentrifugation. The presence of flotillin‐1, the absence of calnexin and ApoB as well as the clean outcome of negative controls (biotinylated BSA) further confirmed the purity of the EVs extracted from clinical plasma samples.

In this platform, we used clinical plasma samples (5 patients, 3 healthy controls) and performed on‐chip EV isolation and PD‐L1⁺ EV detection. To assess the concentration the EVs extracted from plasma samples, NTA results showed that the on‐chip approach yielded (1.97 ± 1.09) × 10¹⁰ EV/mL in patient samples and (5.53 ± 0.22) × 10⁹ EV/mL in healthy controls with comparable size distribution (Figure 5f–h). This is because in cancer patients, blood often contains a higher number of EVs, especially those that carry tumor‐derived molecules.^{[ 55 ]}

Electrochemical detection was performed in triplicate for each sample to measure the level of PD‐L1⁺ EVs (Figure S11, Supporting Information). Compared with healthy controls, the current signal from patient‐derived plasma samples was significantly higher (p < 0.01) (Figure 5i). To validate our approach, we compared the on‐chip results with data obtained from nanoscale flow cytometry (Figure 5j; Figure S12, Supporting Information). Compared with results from flow cytometry, the electrochemical sensor achieved comparable results in differentiating the levels of PD‐L1⁺ EVs in the plasma sample. Meanwhile, the on‐chip EV extraction significantly reduced the time needed as in standard processes. These results highlight the potential of this platform for rapid and efficient detection and monitoring of plasma‐derived EV markers.

While the platform holds promise for the extraction of EV from physiological fluids and detecting their surface markers in decentralized settings, the microfabrication of these microfluidic devices still relies on multiple specialized processes (e.g., photolithography, thin film coating, physical vapor deposition) in cleanroom facilities, limiting the scalable production of these devices with low cost. However, these devices can often be cleaned and reused for EV isolation as long as the electrodes and the dielectric layer remain intact. For example, DI water with surfactants (e.g., Tween 20, Pluronic F‐127) can be used to rinse the device, and re‐coating a hydrophobic layer (e.g., spin/dip coating) can restore the super hydrophobicity of the surface of the device to ensure the electrowetting functionality. Electrochemical sensors are more often used as disposables to ensure the detection outcome of specific EV markers, these sensors can typically be produced in batches and inserted into the devices at the time of detection. Our future efforts will be geared toward scaling the microfabrication of these devices, lowering the costs, including printing electrodes on flexible substrates (e.g., polyethylene terephthalate film, paper), exploring alternative dielectrics that can be easily coated in standard lab settings, and increasing the throughput of this system for parallel processing.

3. Conclusion

A key contribution of this work is the development and integration of microscale electrochemical sensors into a DMF device to streamline and automate EV extraction and detection on‐chip. The proof‐of‐concept platform can perform magnetic bead functionalization and EV extraction from 20 µL of cell culture media and clinical plasma samples in 25 min, and the electrochemical sensor can detect as low as 10⁴ PD‐L1⁺ EVs/mL in 5 min. The sensors showed the desired level of selectivity and stability over time. Compared with standard EV analysis, this portable system significantly reduced the need for infrastructure that is typically found in different core facilities and the multiple time‐consuming and labor‐intensive processes, which hold promise for EV‐based analysis in biology research and clinical applications. Ultimately, we envision that a platform such as this can be translated into a tool that can be used on a regular bench to enable the analysis of EV‐based biomarkers in life sciences labs as well as clinical settings.

4. Experimental Section

Materials

Hydrogen tetrachloroaurate trihydrate (HAuCl₄·3H₂O), trisodium citrate, ferric chloride (FeCl₃), potassium chloride, mercaptopropionic acid (3‐MPA), cysteamine, sulfuric acid (H₂SO₄), potassium ferricyanide (K₃Fe(CN)₆), and bovine serum albumin (BSA) were purchased from Sigma‐Aldrich. 3,3′,5,5′‐Tetramethylbenzidine (TMB) substrate solution, N‐hydroxysuccinimide (NHS), 1‐ethyl‐3‐[3‐dimethylaminopropyl]carbodiimide hydrochloride (EDC), poly‐HRP streptavidin, human PD‐L1 (B7‐H1) Fc chimera Recombinant protein, and biotin‐CD274 (PD‐L1) monoclonal antibody were purchased from Thermo Fisher Scientific. CD9 (D8O1A) Rabbit mAb was purchased from Cell Signaling Technology (Danvers, MA). Gold, platinum, and silver wires of 0.25 mm diameter were purchased from Goodfellow Advanced Materials (Cambridge, UK). Aluminum powder of 0.3 µm and 0.05 µm was purchased Burhler (Lake Bluff, IL).

Apparatus

The morphology of gold nanoparticles was characterized by transmission electron microscopy (TEM, JEOL 1400, USA). The size distribution of these nanoparticles was analyzed in ImageJ, and their UV–vis absorption spectrum was recorded using a UV–vis spectrometer (BMG LABTECH, Germany). Electrochemical measurements were performed in an electrochemical workstation (CHI660E) in a Picoamp Booster and Faraday Cage (CHI200B, CH Instruments, Inc., Austin, TX, USA). The concentration and size distribution of EVs were analyzed using nanoparticle tracking analysis (NTA) in NanoSight NS300 (Malvern Instruments, UK) and ZetaView® TWIN (Particle Metrix Inc. USA). Western blotting was performed, and proteins were transferred using the Bio‐Rad Trans‐Blot TURBO Transfer System (Bio‐Rad Laboratories, Hercules, CA, United States). The blot images were captured using the ChemiDoc™ MP Imaging System and Invitrogen™ iBright™ Imagers (iBright 1500, Thermo Fisher Scientific, USA).

Device Microfabrication and Assembly

The DMF device layout was designed in AutoCAD (Autodesk, Mill Valley, CA, USA). The bottom substrate contained 8 reservoir electrodes, 40 actuation electrodes (4 mm x 4 mm), and electrical contact lines and pads. Overall, the device was microfabricated using our established protocol.^{[ 42} , ⁴³ , ^{56 ]} Briefly, electrodes were photolithographically patterned on chrome (Cr) ‐coated glass slides (KLOE, Saint‐Mathieu‐de‐Tréviers, France), and 5 µm Parylene C was deposited onto the glass substrate to insulate the electrodes (Specialty Coating Systems, Clear Lake, WI, USA). The substrate was then spin‐coated with FluoroPel PFC1101V (Cytonix, Beltsville, MD, USA) as the hydrophobic layer and incubated on a hot plate (180 °C, 30 min). An ITO‐coated glass slide was then assembled onto the bottom substrate using an electrically conductive tape (127 µm) (3M, Ted Pella, Inc., Redding, CA, USA), so that the two substrates were connected in parallel with a space in between.

DMF Device Operation

Each actuation electrode can handle a 10 µL droplet. The device was controlled in a Dropbot system (Sci‐bot, Inc), and droplets were handled through MicroDrop software. Samples and reagents were pipetted onto the reservoir electrode, and droplets were dispensed from the reservoir and actuated across the electrode arrays at 100 V and 10 kHz, waste fluids were transferred to the waste reservoirs.

Preparation of Gold Nanoparticles

Gold nanoparticles were synthesized by reducing HAuCl₄·3H₂O with citrate. In brief, 25 mL aqueous solution of 1% HAuCl₄ (tetrachloroauric acid) was refluxed and stirred, and 2.5 mL 1% trisodium citrate solution was quickly added to the boiling HAuCl₄ solution. In this reaction, trisodium citrate donated electrons to Au³⁺ in the HAuCl₄ solution, reducing these ions to metallic gold element that aggregated to form nanoparticles. The reflux reaction was maintained at a boiling temperature for 20 min to facilitate the rapid reduction of Au³⁺ ions. After 20 min, the reaction mixture was left to cool at room temperature. The synthesized gold nanoparticles were stored at 4°C until use.

Preparation of Gold Nanoparticle Functionalized Working Electrode

In this work, platinum wires and gold wires were used as counter electrodes and working electrodes, respectively. These wires were polished with aluminum powder (0.3 and 0.05 µm) under ultrasonic treatment, and then thoroughly rinsed with ultrapure water. Ag/AgCl reference electrode was fabricated through the reaction between Ag wires and FeCl₃ solution at room temperature for 20 min. The performance of the self‐made Ag/AgCl reference electrode was evaluated by the open‐circuit potential measurement. In the absence of electrical current, only ≈8 mV of potential difference was observed between the reference electrode and a master electrode, indicating excellent functionality and stability of the Ag/AgCl reference electrode (Figure S13, Supporting Information).

Gold nanoparticle‐functionalized working electrode was prepared using established methods.^{[ 57} , ^{58 ]} First, the bare gold wire electrode was electrochemically activated by cycling between −0.2 and 1.5 V versus saturated calomel electrode (SCE) in 0.1 mol L⁻¹ H₂SO₄ solution to remove gold oxide, until reproducible cyclic voltammogram plots were obtained. Then, the activated bare gold electrode was immersed in 0.1 m mercaptopropionic acid (3‐MPA) solution for 12 h at room temperature, where 3‐MPA molecules were absorbed onto the gold surface and formed a self‐assembled monolayer. The 3‐MPA functionalized gold electrode was then immersed in 0.1m ethanolic solution of cysteamine overnight at 4°C. Next, the thiol‐modified gold electrode was dipped into the gold colloid solution at 4°C for 24 h, so that the gold nanoparticles could covalently bind to the thiol groups on the gold electrode surface. The unreacted gold nanoparticles were removed by washing with ultrapure water, and the gold nanoparticle‐functionalized working electrode was kept at 4°C until use.

The capture antibody was immobilized onto the gold nanoparticle‐functionalized working electrode through EDC/NHS coupling chemistry. Briefly, the working electrode was first dipped into 0.1m ethanolic solution of 3‐MPA for 12 h. The carboxylic group‐functionalized working electrode was activated in a mixture of 20 mm EDC and 20 mM NHS solution at room temperature for 1 h. The gold electrode was washed thoroughly with ultrapure water to remove unreacted EDC and NHS molecules. The working electrode was then incubated with 0.8 ug/mL anti‐CD9 at 4°C overnight. Next, the anti‐CD9 immobilized working electrode was reacted with 1% BSA for 30 min to minimize nonspecific absorption. The anti‐CD9 immobilized working electrode was stored at 4°C until use. The surface modification of the gold working electrode was characterized by cyclic voltammetry measurements.

In‐Tube EV Extraction

The MagCapture Exosome Isolation Kit PS (FUJIFILM Wako, Osaka, Japan) was used, and all reagents were prepared according to the manufacturer's protocol to extract EVs from cell culture and plasma samples. Unless otherwise specified, all incubation steps were performed on a rotating mixer. Briefly, the magnetic beads were washed three times with wash buffer. The Tim‐4 protein, tagged with biotin, was then attached to the streptavidin on the beads in a 20‐min incubation step, with the buffer refreshed twice. The sample containing 0.01% v/v Tetronic® 904 (BASF Canada Inc, Mississauga, Canada) was subsequently added to the magnetic beads, and the mixture was incubated for 20 min. After three washes, 20 µL of elution buffer was added to release the EVs from the beads. The isolated EVs were stored at −80 °C until further use.

Nanoscale Flow Cytometry

We used nanoscale flow cytometry (Apogee A60‐MicroPlus) to quantify the absolute concentrations (#/mL) of total EVs and PD‐L1⁺ EVs in the supernatant of cell culture medium, spiked platelet‐free plasma, and patient‐derived plasma as previously described.^{[ 9 ]} Briefly, the extracted EVs (20 µL) (both on‐chip and in‐tube) were incubated with a fluorescent antibody against PD‐L1 (H1A clone, from Dr. Haidong Dong).^{[ 59 ]} Each sample was run in triplicate, and antibody‐matched isotypes were used as negative controls. Flow cytometry data were obtained blindly and analyzed with FlowJo software to determine and apply gates, generate reports with scatter plots, EV concentrations for PD‐L1⁺ EVs, and data summaries. Additional methodological details can be found in the MIFlowCyt‐EV report in the Supplementary Information.

Transmission Electron Microscopy

3 µL of undiluted exosome suspension was deposited onto copper grids and allowed to dry for 5–7 min. The grids were then washed by placing them on top of a drop of Nanopure water for 30 s, followed by blotting, this step was repeated three times. For staining, the grids were placed on a drop of 1% phosphotungstic acid (PTA, pH 7.0) for 30 s, blotted, this step was repeated two times. After being placed on top a 3rd drop of 1% PTA for 1 min, the grids were allowed to air‐dry completely. Grids were then imaged under a JEOL JEM‐1400Plus transmission electron microscope.

Preparation of EVs from Cell Culture

Human MDA‐MB‐231 breast cancer cells (ATCC, Manassas, VA, USA) were maintained in a humidified incubator at 5% CO₂. MDA‐MB‐231 cells were cultured in high glucose DMEM (Corning, New York, NY, USA) with 10% fetal bovine serum (Thermo Fisher Scientific, Waltham, MA) and 5% penicillin and streptomycin (Thermo Fisher Scientific, Waltham, MA). Cells were cultured at 70–80% confluency in a 6‐well plate (Thermo Fisher Scientific, Waltham, MA) and transfected with transfected with a pcDNA3.1/PD‐L1vector that contained full‐length cDNA of human PD‐L1 as our target sample, as these transfected cells had an overexpression of PD‐L1 in the tumor cells that would secrete more PD‐L1⁺ EVs. Meanwhile, we used the same cell line with PD‐L1 knocked out by CRISPR/Cas9 technology. To collect cell‐derived EVs, cells were cultured in serum‐free media for 48 h. Conditioned medium was collected and centrifuged twice at 2500g (κ factor – 9153), 20 °C for 15 min with max break. EV‐containing supernatants were then concentrated using ultrafiltration columns with a MWCO cut‐off of 100 KDa (Millipore Sigma, UFC910008). The EV‐enriched concentrates were aliquoted in cryovials and stored at −80 °C until use.

PD‐L1 knock‐out: Human B7‐H1 was knocked‐out by CRISPR/Cas9 technology as previously described.^{[ 60 ]} The guide sequence (5′‐ATTTACTGTCACGGTTCCCA‐3′) specific to human PD‐L1 exon 3 (second coding exon), designed using CRISPR DESIGN tool (http://crispr.mit.edu) and cloned into px458 plasmid coexpressing GFP (Addgene, #52961). Thirty‐six h after transfection, cells were sorted for GFP and sub‐cloned using flow cytometry. Two weeks later, single cell subclones were genotyped by PCR and validated Western blotting for PD‐L1 protein depletion. PD‐L1 expression level was determined by flow cytometry and Western blotting.

Plasma Sample Preparation

The sample collection and experiments were approved by Mayo Clinic Institutional Review Board (IRB). Whole blood samples from healthy donors (IRB#21‐013474) and melanoma patients (IRB#20‐003367) were drawn into 10mL tubes and were subsequently centrifuged twice at 2,500 xg for 15 min to collect the cell‐free plasma. The isolated plasma samples were kept at −80°C until use.

Nanoparticle Tracking Analysis of EV

A total of 800 µL of the prepared solution was loaded into the sample jig of NanoSight NS300, and the laser module was mounted within the main instrument housing. NTA detected scattered light from particles of interest based on their Brownian motion, which was recorded through video sequences performed in triplicate. Data acquisition and processing were performed using NanoSight NS300 control software. The concentration of particles within the EV size range was used to calculate the EV capture and release efficiency. For ZetaView® TWIN, prior to each experimental session, the instrument was calibrated and auto‐aligned using polystyrene latex beads prepared at a standardized dilution. The same calibration procedure was applied across all sessions to ensure inter‐batch comparability. Samples were diluted in TBS to fall within the optimal concentration range recommended by the manufacturer. Data acquisition and analysis were performed using ZetaView software (version 8.05.16 SP3).

Western Blot

The eluted EVs were lysed with RIPA buffer supplemented with protease inhibitor tablets. The collected proteins were prepared in 6× Laemmli buffer with 2‐mercaptoethanol and heated to 95 °C for 10 min before loading onto the 4–15% Mini‐PROTEAN TGX Gels (Bio‐Rad Laboratories, Hercules, CA, United States). After denaturing, the proteins were transferred to nitrocellulose membranes (Amersham, St. Louis, MO, United States) and then analyzed by standard Western blotting procedures at 100 V for 70 min. The membrane was then blocked in 5% non‐fat milk (Bio‐Rad Laboratories, Hercules, CA, United States) in tris‐buffered saline with 0.1% Tween 20 detergent (TBST) (Sigma, St. Louis, MO, United States) for at least 60 min. Immunoblotting was performed using rabbit anti‐CD9 (Cell Signaling Technologies (CST), Danvers, MA, United States, cat# 13174), rabbit anti‐GAPDH (CST, cat#2118) or rabbit anti‐calnexin (CST, cat#2679) primary antibodies, followed by HRP‐linked anti‐rabbit IgG secondary antibody (CST, cat#7074). To minimize cross reactivity on a same blot, each target protein was probed separately on the same membrane by sequential immunoblotting. After imaging each blot, the membrane was thoroughly washed, re‐blocked and re‐incubated with the next primary antibody.

Detection of PD‐L1⁺ EV with Electrochemical Sensor

The anit‐CD9 modified working electrode of the electrochemical sensor was inserted into the 40 µL of EVs extracted in the DMF device (from PBS, cell culture and human plasma samples) at various concentrations for 1 h, and the working electrode was washed with PBS solution. Next, 40 µL of biotinylated anti‐human PD‐L1 antibody was added to the working electrode. After incubation for 1 h, working electrode was thoroughly washed with PBS solution to remove unbonded antibody. The working electrode was then incubated with 40 µL of streptavidin‐poly‐HRP diluted in PBS for 30 min. After that, the electrochemical detection was performed in 3,3′,5,5′‐Tetramethylbenzidine (TMB) substrate. The electrochemical analysis was performed on a three‐electrode system. Cyclic voltammetry was conducted in the range of −0.5–0.6 V, and the scan rate was 0.1 V. The detection signal was recorded through amperometry measurement with the potential fixed at ‐0.1 V. The electrochemical reduction current was measured at 200 s after the TMB oxidation reaction reached a steady state.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Supporting Information

Lo T.‐W., Liu J., Zhang Y., et al. “An Integrated Digital Microfluidic Device for the Extraction and Detection of Extracellular Vesicle‐Based Molecules.” Small 21, no. 38 (2025): e04335. 10.1002/smll.202504335

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.