Multiplex Isolation and Profiling of Extracellular Vesicles using a Microfluidic DICE Device

Abstract

Profiling of extracellular vesicles (EVs) is an emerging area in the field of liquid biopsies because of their innate significance in diseases and abundant information reflecting disease status. However, unbiased enrichment of EVs and thorough profiling of EVs is challenging. In this paper, we present a simple strategy to immobilize and analyze EVs for multiple markers on a single microfluidic device and perform differentiated immunostaining-based characterization of extracellular vesicles (DICE). This device, composed of four quadrants with a single inlet, captures biotinylated EVs efficiently and facilitates multiplexed immunostaining to profile their extracellular proteins, allowing for a multiplexed approach for non-invasive cancer diagnostics in the future. From controlled sample experiments using cancer cell line derived EVs and specific fluorescence staining with lipophilic dyes, we identified that the DICE device is capable of isolating biotinylated EVs with 84.4% immobilization efficiency. We extended our study to profile EVs of 9 clinical samples from non-small cell lung cancer (NSCLC) patients and healthy donors and found that the DICE device successfully facilitates immunofluorescent staining for both the NSCLC patients and the healthy control. This versatile and simple method to profile EVs could be extended to EVs of any biological origin, promoting discoveries of the role of EVs in disease diagnostics and monitoring.

Graphical Abstract

Introduction

Extracellular vesicles (EVs) have emerged in recent years as a critical biomarker in liquid biopsies for both the diagnosis and monitoring of cancer. This is largely because of the cargo packaged inside of extracellular vesicles. It has been shown that EVs are used within the body for intercellular communication, making them useful for their diagnostic and prognostic potential.^1,2 The information carried by EVs has been shown to correlate with cancer aggressiveness, angiogenesis and metastatic sites.^3,4 These discoveries support opinions that EVs play an active role in cancer progression. Additionally, it has been shown that EVs isolated from cancer patients frequently show higher cancer-associated antigen expression and have increased EV concentration during circulation compared to healthy donors. This implies that EV number and protein expression could be a window to evaluate a patient’s cancer status in a non-invasive manner.^5,6 Previous studies have asserted the role of exosomal protein in diagnostic and prognostic work, however, more efficient profiling methods are urgently needed for diagnostic use of EVs.⁷

Thus far, conventional methods for cellular protein evaluation have been applied to EVs, despite the low protein loading of EVs. These methods include western blot⁸, ELISA⁹, and fluorescence activated cell sorting (FACS).¹⁰ Although western blot and ELISA are relatively easy and widely used techniques, they require large amounts of sample and are labor intensive. Automated FACS-based protein profiling is capable of multiplexed analysis of cells, but its detection threshold and size of analyte prevent their use for nano-size EVs. In order to gain better detection sensitivity with conventional FACS, Sharma et al. showed exosomal expression of melanoma associated antigens by using micro sized magnetic beads conjugated with melanoma antibody and a subsequent FACS method.¹¹ Jin et al. used aptamer nanoprobe conjugated microsphere to immobilize exosomes thus facilitating flow cytometry analysis.¹² While these techniques showed meaningful progress in EV protein analysis, they still caused inevitable sample loss and required extensive sample prep. The recently developed nanoFACS has the capabilities to work with EV samples, however, its high cost and bulky size make it ineffective for clinical use.¹³

Recent advancements in microfluidic technologies allow us to isolate and profile EVs using minimal resources and preprocessing. Microfluidic isolation facilitates high recovery of EVs, allowing for subsequent downstream analysis.^14–17 Immunoaffinity-based microfluidic capture devices functionalized with antibodies specific to EV surface markers allows a platform to profile the captured EVs by immunofluorescent staining, surface plasmon resonance (SPR), and surface enhanced Raman scattering (SERS).^18–20 On-chip isolation and profiling significantly lowers the necessary sample volume while achieving higher sensitivity. Im et al. used a 150 μL of preprocessed exosomal sample to evaluate ovarian exosome’s surface expression.¹⁹ However, this approach requires advanced knowledge of the target EVs in order to functionalize a device to capture the specific EVs within the sample. Due to innate limitations when using antibody capture, the recovered EVs may be enriched for one subset. This enrichment may lead to biased view of EVs and controversial clinical outcomes.

In order to analyze EVs without any pre-enrichment bias, efforts have been made to develop new isolation strategies. Several groups have been successful at conjugating biotin to the surface of EVs. Biotinylation is a widely used strategy for the isolation and engineering of both the surface of cells and vesicles.^21,22,23 For use in EV isolation, the biotinylation reaction targets either lipids^24,25 or surface proteins.^22,23 Wan et al. developed a biotin conjugated lipid nanoprobe with a lipid tail for EV membrane insertion and immobilized this complex to a magnetic bead with NeutrAvidin.²⁵ Saunderson et al. used biotin conjugated linkers targeting exosomal surface proteins to study the role of EVs in vivo using immunofluorescence staining.²³ Similarly, Lee et al. recently suggested a multiple EV staining device that isolates EVs via biotinylation and profiles the isolated EVs using immunostaining.²⁶ They showed that ultracentrifugation, the gold standard of EV isolation, followed by biotinylation and on-chip protein profiling showed considerable promise without necessitating prior knowledge of the origin of the EVs in the sample.²⁶ At the same time, microfluidic devices incorporated with spatially patterned antibody barcodes has been developed to study EV’s heterogeneity.²⁷

Here, we present a novel way to capture and analyze EVs for multiple protein markers using a single microfluidic device, DICE. By using ultracentrifugation followed by biotinylation, we isolate and profile EVs on chip with minimal sample volume and resources. This is accomplished by utilizing precise deep-reactive ion etching microfabrication methods and highly preferential avidin/biotin binding affinity. EV capture is then performed using a microfluidic device functionalized with a NeutrAvidin binding surface. Once captured, the profiling and identification of EVs can be accomplished using multiplexed immunostaining, allowing for multiple conjugated fluorescent dyes to be washed across the immobilized EV surface to identify specific surface markers. As shown in Fig. 1c the device is designed with four separate quadrants, designed for each quadrant to be immunostained for a unique marker, thus achieving multiplexed analysis of EVs from one sample. The device’s small working volume (>100 μL) is suitable for EV applications, limiting the need for large sample volumes. Using this platform, we analyzed clinical blood samples from five patients with non-small cell lung cancer (NSCLC) aiming to verify the clinical potential of our platform. We performed multiplexed protein profiling using four different antibodies to demonstrate the future potential for clinical applications. Liquid biopsy studies in lung cancer have largely been limited to CTCs and ctDNA, many of which still require clinical validation as diagnostic and prognostic markers. More recently these studies have included cancer EVs, where groups have described important correlations between EV number and tumor progression. Lung cancer patients with adenocarcinoma have greater numbers of EVs in the blood compared with healthy controls.²⁸ At the same time, studies regarding lung cancer associated protein expression on EVs have shown that EVs from lung cancer express EGFR^29,30, PD-L1³¹ and vimentin.³² To perform multiplexed analysis of the EVs from lung cancer patients, we used three cancer-associated markers (EGFR^29,30, PD-L1³¹, and vimentin³²) and one extracellular vesicle-associated marker (CD9). The expression of these markers on EVs serve as example biomarkers to demonstrate future potential utility in lung cancer diagnosis and monitoring.

Fig. 1.

Differentiated immunostaining characterization of Extracellular vesicle (DICE) chip for extracellular vesicle profiling: (a) procedure of the DICE; (b) fabricated DICE device (left) and immobilized extracellular vesicles in DICE device (right); (c) design of the DICE device.

Materials and methods

Device Design and Fabrication

The DICE device has four quadrants, with each quadrant having 13×13 circular chambers connected by junction channels. Each circular chamber has a diameter of 100 μm and the pitch is 500μm (Fig.1c). The circular chamber design facilitates efficient EV isolation and easy analysis based on our previous simulation and experimental results.¹⁴ The height of the device is 50 μm and to achieve this high-aspect ratio design, we fabricate the mold using deep reactive-ion etching (DRIE) Fig. 2. Briefly, SPR220 was spin coated onto a four-inch silicon wafer at a thickness of 3 μm before exposure and post-baking. DRIE was then performed to achieve a height of 56.1±0.79 μm. After mold fabrication, the DICE device was fabricated by combining PDMS and PDMS curing agent mix (1:10) before baking overnight at 70°C. The prepared PDMS layer was bonded to an untreated glass slide by O2 plasma treatment.

Fig. 2.

Fabrication of the DICE microfluidic device using deep reactive-ion etching (DRIE)

Device functionalization

For the surface modification of the device, we used standard avidin-biotin chemistry with optimization.¹⁴ Following initial device fabrication, ethanol with silane solution was injected into each device and incubated for an hour. Each device was then injected with GMBS solution (5 mL Ethanol + 14 μL GMBS) and incubated for 30 minutes. Each device was once again washed out with ethanol before injection with a NeutrAvidin solution (1 mL filtered PBS + 100 μL NeutrAvidin). Each device was then stored in a parafilm sealed Petri dish containing wet paper and stored at 4°C for future use. Devices used as no surface modification controls were not treated by the modification procedure. Control devices were injected with 3% BSA solution (0.03 g/1mL filtered PBS) to prevent nonspecific binding and incubated for at least 30 minutes. This 3% BSA solution was washed out with PBS buffer before storage and use.

Extracellular Vesicle Purification and Sample Preparation

The sample collection and experiments were approved by University of Michigan Institutional Review Board (IRB). Informed consents were obtained from all participants of this clinical study and blood samples were obtained after approval of the institutional review board at the University of Michigan. All experiments were performed in accordance with the approved guidelines and regulations by the ethics committee at the University of Michigan. EV purification from the blood sample was followed by recent consents for general EV research.^33,34 Whole blood samples were collected in EDTA tubes and were subsequently centrifuged at 2000 × g for 15 minutes to isolate plasma. The isolated plasma samples from cancer and healthy donors were kept at −80°C before use and used within 6 months. From the separated plasma layer, 100 μL aliquots were placed into ultracentrifugation tubes (Beckman Coulter, United State) along with 100 μL of 0.2 μm filtered PBS. The tubed samples then underwent Airfuge ultracentrifugation (Beckman Coulter, United States) using an A-100/30 angle rotor for 40 minutes at 100,000×g. After the 1^st ultracentrifugation, 150 μL of supernatant was removed from each tube and replaced with 150 μL of filtered PBS, followed by an additional round of ultracentrifugation at 100,000×g for 40 minutes. Following ultracentrifugation, 150 μL of supernatant from each tube was discarded and the remaining contents containing the purified EV pellet were ready for biotinylation. For the performance verification of the DICE device, purified EVs from A549 lung adenocarcinoma were purchased and used as a model sample. This EV sample was directly biotinylated without any purification steps.

Biotinylation of Extracellular Vesicles

Purified EVs were biotinylated for non-biased immobilization following the EZ-LINK protocol with optimization. EZ link biotin powder, EZ-Link Sulfo-NHS-LC-Biotin (ThermoScientific, United States) was utilized for EV biotinylation during these experiments. A 10× concentration of 300 μM biotin solution was prepared by dissolving 1.6 mg of E-Z link biotin in 10 mL of filtered PBS. This 10× solution was then diluted to 1× by incorporating 100 μL of the 10× solution with an additional 900 μL of filtered PBS. 100 μL of 1× 300 μM biotin solution was then added to 50 μL of purified EVs, followed by a 1-hour incubation. Refrigerated desalt spin columns were prepared using repeated PBS washing and centrifugation steps following the manufacturer stated procedure. After the hour-long EV incubation, 150 μL of the biotinylated EV sample was injected into the desalt spin column and centrifuged at 1500 × g for 2 minutes. The sample was then passed through a 0.2 μm filter and stored in vials at–80°C for future use. Biotinylation of purified A549 cell-derived EVs for the EV immobilization on chip experiment followed the same procedure as that of clinical samples. In order to evaluate the effect of biotinylation on EV’s surface marker binding sites, we prepared two different EV samples using a plasma sample from a healthy control: one before and one after biotinylation. EVs were isolated using an Airfuge ultracentrifugation using the previously described procedures. One 150 μL sample was then biotinylated using the previously outline biotinylation procedure, while another 150 μL sample did not undergo biotinylation. Both biotinylated and non-biotinylated samples were injected with 5 μL of CD9 anti-rabbit antibody in ultracentrifugation tubes and allowed to incubate for 1 hour. Each sample underwent a second Airfuge ultracentrifugation for 20 minutes at 100,000×g, followed by extraction of 100 μL of supernatant to remove excess antibody. 50 μL of filtered PBS was then added to the remaining sample in each UC tube, along with 3 μL of Alexafluor 647 goat anti-rabbit (ThermoScientific, United States). This secondary antibody was incubated for 1 hour followed by 20-minute ultracentrifugation at 100,000×g. After ultracentrifugation, 50μl of sample was extracted from each UC vial and the pellet was discarded. The fluorescence intensities from these samples were compared by using a fluorescence microscope.

Device Operation and Performance Verification

The prepared model samples or patient biotinylated EV samples were processed using a Harvard syringe pump. Filtered PBS was first injected through the inlet of each device to remove leftover NeutrAvidin. A total of 100 μL of pre-enriched EV sample was then pumped through the inlet of each device, followed by another PBS wash step to clear any non-bonded sample from the PDMS chamber. Multiplexed immunofluorescence staining was performed by using each of four different outlets to inject either specific antibodies or lipophilic dyes.

Field Emission Scanning Electron Microscopy (FE-SEM) Analysis

EVs captured within the device were fixed in 2.5% glutaraldehyde for one hour to retain their morphology. After rinsing with PBS, the samples were dehydrated in a graded concentration of ethanol (50%, 70%, 90%, 95%, and 100%) for 10 min at each step (two times for 100%). The samples were dried using hexamethyldisilane, followed by overnight air drying in the hood. The dehydrated samples were then mounted on aluminum stubs and sputter coated with gold to create a conductive layer. The samples were observed by FEI Nova 200 Nanolab Dualbeam FIB scanning electron microscope at the Electron Microscopy Analysis Lab (MC2) at University of Michigan.

Nanoparticle Tracking Analysis

For the size profile of the biotinylated EVs, nanoparticle tracking analysis was performed using NanoSight NS300 (Malvern Instruments, UK). The resulting EVs after biotinylation and DICE processing were loaded on the NanoSight and profiled for total concentration and size. NTA visualizes the scattered light from the movements of individual vesicles simultaneously in the field of view, and they are monitored through a video sequence acquired over 20 seconds. Each experiment was carried out in triplicate, and data acquisition and processing were performed using NanoSight NS300 Control Software.

On-chip Immunostaining Analysis

For immunofluorescence staining, each device was first blocked with 5% BSA solution. Next, inlet tubing was removed from each device and 3” tubing was attached to each of the four outlets. Through each of these outlets, 50 μL of either PD-L1, Vimentin, CD9, or EGFR (1:20 dilution in 1%BSA solution) were applied and incubated overnight at 4°C. After overnight incubation, a PBS wash was run through each outlet to remove excess primaries. The outlet tubing was then removed, and a single tube was attached to the inlet. A secondary solution of 2.5μl Anti-mouse IgG2b 488 and 2.5μl Anti-rabbit IgG 647 in 100μl of 1%BSA was pumped through the device inlet and incubated for 35 minutes. A final PBS wash was applied through the inlet to clear unbound secondaries. Each device was then imaged at 4×, 10×, 20×, and 40× magnification (Eclipse TI2, Nikon, Japan.) Devices were scanned in FITC for PD-L1 confirmation and CY5 for confirmation of Vimentin, CD9, and EGFR. Analysis was performed by measuring the fluorescent intensity of each circular chamber in each quadrant. The average intensity across all chambers per quadrant was taken by dividing total intensity by the area of the chambers summed. The standard deviation is then representative of the variations in fluorescence intensity across all the chambers in one quadrant.

Results and discussion

Biotinylation of EVs for the DICE platform

From identical volumes of healthy plasma samples, we evaluated if the biotinylation procedure effects their total concentration, size, and membrane protein availability (Fig.3). We found a maximum loss of 29% of the total vesicles following the biotinylation procedure (Fig.3a), which includes desalting column and additional 200nm filtration steps. The purity, defined as the percentage of the vesicles that fall within a standard EV size range (30–150nm), remained similar. The size distribution (mean and mode) before and after biotinylation showed no difference (Fig.3b). Because our biotinylating agent targets EV surface proteins to anchor, we investigated the degree to which the biotinylation inhibited surface protein binding ability. Thus, we stained the EVs before and after biotinylation by using anti-CD9, and we found that our biotinylation does not significantly affect immunostaining of a commonly expressed surface protein (Fig.3c–d). From previous studies by Cole et al. using BALM-1 cells, there was no saturation point for cellular uptake of biotin even up to 50mM concentrations.³⁵ While this study was given for cells, the concentration we used for EVs was chosen to be low enough to remain effective while not critically limiting antibody binding sites. This is supported by additionally previous work that used a notably higher concentration of E-Z link biotin solution (2.2 mM) to biotinylate EVs while reporting significant EV capture by CD169 on cell surfaces.²³

Fig. 3.

Biotinylation of EVs for the DICE platform: (a) EV concentration and purity comparison before and after biotinylation; (b) EV size change comparison; (c-d) EV immunostaining results using exosomal marker CD9 in fluorescence intensity (c) and quantitative images after staining (d). (Scale bar=250μm).

Device Optimization and On-chip EV immobilization

To isolate and analyze EVs in a non-biased way, we designed an EV isolation device that was antigen independent. For that, we pre-conjugated biotin onto the surface of EVs for immobilization on our microfluidic device. This streamlined procedure uses only 20 μL of sample and is enough to analyze using immunostaining. After the biotinylation of EVs, we evaluated their size profiles by using NTA. Fig. S3 shows that the average vesicle size within tested samples falls within the standard 50–200 nm size distinction of EVs, which agrees with previously reported values.^36,37

The ability of the DICE device to utilize Avidin/Biotin affinity for EV capture was examined by NTA, fluorescence imaging and scanning electron microscope analysis. First, we prepared two different DICE devices with and without NeutrAvidin conjugation and processed an identical amount of biotin conjugated A549 EVs to evaluate the EV immobilization efficiency of our device. The concentration of initial and resulting (after EV capture) samples were compared using the NanoSight NS300 (Marven Instruments, UK) EV concentration within biotinylated samples measured both pre and post-capture using NTA.

As seen in Fig. 4a, NeutrAvidin conjugated DICE devices are nearly four times more effective than control devices at immobilizing biotinylated EVs. To confirm this immobilization qualitatively, we used immunofluorescent staining and SEM analysis. Non-conjugated and NeutrAvidin conjugated devices were injected with biotinylated EVs, followed by an incubation periods and PBS wash. Green PKH dye was then applied to each device, staining the lipid bilayer of EVs still present within the device. Fig. 4b shows the significant presence of EVs within NeutrAvidin conjugated devices, demonstrating the effectiveness of the Avidin/Biotin based EV capture within the small chambers of the DICE device. To further confirm the presence of EVs in the device, SEM images were taken showing the capture of EVs both in the chambers and along the channels of the device, Fig. 4c. The 25,000× image confirms the previously reported spherical morphology of EVs, further confirming the pure isolation of EVs from our samples.

Fig. 4.

Biotinylation and isolation of extracellular vesicles: (a) immobilization efficiency of A549 derived exosomes compared to a control device without streptavidin conjugation; (b) the immobilized extracellular vesicles stained with lipophilic dye, PKH-green; (c) scanning electron microscope (SEM) analysis of the on-chip immobilized extracellular vesicles from the lung cancer cell line, A549.

On-chip Multiplexed Staining

Simultaneous, multiplexed staining utilizing multiple dyes through such a small chambered device presents challenges in allowing smooth flow through several inlets and exiting a single outlet. The ability for the DICE device to handle this issue were tested using four distinctly dyed PBS solutions. As shown in Fig. 5 (left), all four colors flowed smoothly into the device through separate inlets without backpressure. The figure also shows all four fluids ran through their designated channels without mixing or cross contamination within channeling. From this study, we found that 10 μL/h is enough to capture EVs, while ensuring no cross flow and contamination between quadrants. The functionality of staining different quadrants for different EV markers simultaneously was then tested using two separate PKH dyes. Biotinylated EVs were captured on chip, followed by application of PKH red and PKH green dyes pumped into two diagonal quadrants each. Fig. 5 (right) shows both dyes were successful in staining the lipid bilayer of EVs in their designated channels. Aside from outlet mixing, no crossover was observed away from the outlet within these channels.

Fig. 5.

Multiplex immunostaining of extracellular vesicles using DICE device. Scale bar=200μm.

Multiplexed profiling of the EVs from clinical samples

We analyzed our small clinical cohort (n=5 NSCLC patients, n=4 healthy control) in terms of nanoparticle tracking analysis and immunostaining results. Multiplexed profiling of EVs from patient samples was carried out using the DICE device. Lung cancer patient and healthy donor blood samples were processed to isolate and then biotinylate EVs for experimental use. EV sample concentrations were measured using NTA analysis before being applied to the DICE device. EV concentration and size results, given in Fig. 6a, demonstrate a 186% greater EV concentration within samples derived from lung cancer patients than in healthy patient samples. While the greater number of EVs in cancer has been previously reported^5,6, this result also indicates the utility of this EV isolation method for lung cancer samples. Recent studies with larger sample size reported no significant difference in total EV concentration in plasma between cancer patients and healthy donors.^38,39 Mean value of particles within both lung cancer patients and healthy donors were reasonably equivalent (116.92 vs. 117.25). Healthy donor and lung cancer patient samples were applied to NeutrAvidin conjugated DICE devices for EV immobilization. Multiplex profiling was then performed utilizing simultaneous application of PD-L1, Vimentin, CD9, and EGFR antibodies followed by secondary fluorescent antibodies. Fluorescent imaging showed considerable presence of all four antibodies on the surface of the captured EV.As demonstrated in Fig. 6b, which depicts the significant capture of EGFR-expressing EVs, three of the four quadrants of each device were positive in their specified secondary fluorescent channel (FITC Anti-mouse IgG2b 488 for PD-L1 and CY5 Anti-rabbit IgG 647 for Vimentin, CD9, and EGFR). Fluorescent intensity was determined by measuring the intensity of the entire area of each quadrant of the device and subtracting away the background intensity. This normalized the intensity normalizes for fluctuations in background light and small device defects. Additionally, the same volume of plasma was run through each device, to normalize for volume. The three quadrants that were positive (Vimentin, CD9 and EGFR) were all stained with the same secondary antibody (Anti-rabbit IgG 647) while the secondary antibody for PD-L1 was different because of the differing host of the PD-L1 antibody (rabbit vs mouse). It is possible that the reason there was no positive signal for PD-L1 was because of the differing imaging wavelength as well as the different secondary antibodies. Further testing would be needed to verify that this was a true negative PD-L1 staining. Other than HC1, EGFR intensities of lung cancer patients were found to be higher than that of healthy controls. Vimentin, which was found to be expressed on lung cancer EVs, is also notably expressed in EVs from healthy donors. Because vimentin is one of the most dominant cytoskeletal elements in leukocytes⁴⁰, one might want to interpret a high expression of vimentin on healthy EVs with care.

Fig. 6.

DICE-based multiplexing results of 5 different clinical samples: (a) EV concentration and size distribution analysis; (b) evaluation of EGFR expression on EVs from a lung cancer patient. (Scale bar=50μm); (c) immunofluorescence intensity analysis of clinical samples using three different protein markers including Vimentin, CD9 and EGFR. Expression in arbitrary units.

By multiplexing and immunostaining EVs immobilized on a microfluidic device, we have shown the feasibility of using such a device for clinical studies. We were able to successfully demonstrate the presence of all four markers in each sample, including the healthy control. While the healthy control sample showed lower fluorescent intensity compared to lung cancer patient samples, this would need to be validated in a larger cohort for confirmation.

Conclusions

As a proof of concept, we have showed the potential of the microfluidic DICE device as a multiplexed isolation and protein profiling tool through simultaneous studies using cancer cell derived EVs and plasma samples from cancer and healthy donors. These results demonstrate that ability of the DICE device to successfully capture and multiplex concentrated EV samples. As biotinylation of concentrated EV samples allows for the capture of EVs of unknown origin, the ability to derive the protein profile of EVs on the DICE device enables a rapid and thorough investigation into the vesicles source. The ability to separate and differentiate multiple primary and secondary antibodies within the four channels of the device demonstrates the DICE device’s potential as a platform for multiplex profiling involving antibodies other than the four tested here.