Rapid and efficient isolation platform for plasma extracellular vesicles: EV‐FISHER

Abstract

Extracellular vesicles (EVs) have found diverse applications in clinical theranostics. However, the current techniques to isolate plasma EVs suffer from burdensome procedures and limited yield. Herein, we report a rapid and efficient EV isolation platform, namely, EV‐FISHER, constructed from the metal‐organic framework featuring cleavable lipid probes (PO₄ ³⁻‐spacer‐DNA‐cholesterol, PSDC). The EV‐FISHER baits EVs from plasma by cholesterol and separates them with an ordinary centrifuge. The captured EVs could be released and collected upon subsequent cleavage of PSDC by deoxyribonuclease I. We conclude that EV‐FISHER dramatically outperforms the ultracentrifugation (UC) in terms of time (∼40 min vs. 240 min), isolation efficiency (74.2% vs. 18.1%), and isolation requirement (12,800 g vs. 135,000 g). In addition to the stable performance in plasma, EV‐FISHER also exhibited excellent compatibility with downstream single‐EV flow cytometry, enabling the identification of glypican‐1 (GPC‐1) EVs for early diagnosis, clinical stages differentiation, and therapeutic efficacy evaluation in breast cancer cohorts. This work portrays an efficient strategy to isolate EVs from complicated biological fluids with promising potential to facilitate EVs‐based theranostics.

Abbreviations

Acronym

Full English description

AFM

Atomic force microscopy

BC

Breast cancer patients

BC‐T

Breast cancer patients after treatments

CHO

Cholesterol

Cy5

Cyanine 5

DLS

Dynamic light scattering

DMEM

Dulbecco's Modified Eagle Medium

DMF

N, N‐Dimethylformamide

DNase I

Deoxyribonuclease I

EDX

Energy dispersive X‐ray

EV‐FISHER

Extracellular Vesicles Fisher

EVs

Extracellular vesicles

FBS

Fetal bovine serum

FITC

Fluorescein isothiocyanate

FT‐IR

Fourier transform infrared

GPC‐1

Glypican‐1

HD

Healthy donors

HEPES

4‐(2‐Hydroxyethyl)‐1‐piperazine ethane sulfonic acid

ICP‐MS

Inductively coupled plasma mass spectrometry

LMCT

Ligand to metal charge transfer

MEA

Cysteamine

MISEV

Minimal Information for Studies of Extracellular Vesicles

MOFs

Metal‐organic frameworks

NaCl

Sodium chloride

nFCM

Nano flow cytometer

NH₂‐BDC

2‐Amino‐1,4‐benzendicarboxylic acid

NTA

Nanoparticle tracking analysis

PBS

Phosphate buffer saline

PFP

Platelet‐free plasma

PSDC

PO₄ ³⁻‐spacer‐DNA‐cholesterol

PXRD

Powder X‐ray diffraction

RPMI 1640

Roswell Park Memorial Institute 1640

SDS‐PAGE

Sodium dodecyl sulfate‐polyacrylamide gel electrophoresis

SEC

Size exclusion chromatography

TEM

Transmission electron microscopy

UC

Ultracentrifugation

UV‐Vis

Ultraviolet‐visible

XPS

X‐ray photoelectron spectroscopy

ZrOCl₂·8H₂O

Zirconium(IV) chloride oxide octahydrate

1. INTRODUCTION

Extracellular vesicles (EVs) are phospholipid‐bilayer‐surrounded particles ranging from 30 to 1000 nm released by all types of cells in an evolutionally conserved manner (Liang et al., 2021). Increasing evidence has suggested that EVs play vital roles in intercellular communication, disease development and progression (Wiklander et al., 2019; Xie et al., 2019; Xu et al., 2018, ). In particular, EVs carry abundant biomolecules such as proteins, nucleic acids, lipids etc., which have excellent potential as next‐generation biomarkers (Vasconcelos et al., 2019; Wang et al., 2018). For example, glypican‐1 (GPC‐1) positive EVs were overexpressed in breast cancer, exhibiting excellent performance in early diagnosis and therapeutic efficacy monitoring (Melo et al., 2015). However, EVs with heterogeneous sizes are present in biological fluids, for example, plasma, serum, urine, and saliva, wherein non‐EV biomolecules are dominant. It is thus a formidable challenge to develop an efficient method to isolate EVs from complex fluids, such as plasma.

At the current stage, differential ultracentrifugation (UC) is the most commonly used method for isolating EVs from plasma. However, UC is time‐consuming and co‐isolates contaminants such as protein aggregates (Boriachek et al., 2018). Therefore, alternative methods such as asymmetric flow field‐flow fractionation (Zhang & Lyden, 2019), capillary‐channelled polymer fibre‐based chromatographic method (Jackson et al., 2021), exosome detection via the ultrafast‐isolation system: EXODUS (Chen et al., 2021), size exclusion chromatography (SEC) (Oeyen et al., 2018), ultrafiltration (Woo et al., 2017), polymeric precipitation (Doyle & Wang, 2019), and immunoaffinity (Choi et al., 2022; Zhu et al., 2021), have also been proposed to isolate EVs for varied downstream applications with different merits and limitations, which summarized in Table S1. Nevertheless, it remains a challenge to isolate plasma EVs with preferable yield in a clinical‐friendly manner.

Recently, a lipid nanoprobe has been developed to target the phospholipid bilayer of EVs to facilitate their isolation (Wan et al., 2017). Compared to the traditional EVs recognition elements like antibodies, peptides, and aptamers, the lipid nanoprobe captures EVs by targeting EV's phospholipid bilayer rather than the specific surface biomacromolecules (CD63, CD9, phosphatidylserine, etc.) (Koliha et al., 2016; Zhang et al., 2021; Zhao et al., 2016). In addition, the strong hydrophobic interaction of the lipid probe and EVs membrane promises an efficient binding. Therefore, this lipid‐based strategy proposes its superiority in isolating the total EV population and thus provides comprehensive sample information for clinical judgment.

Metal‐organic frameworks (MOFs), supported by principally unlimited permutations of metal ions and organic ligands, have emerged as promising materials for diverse applications (Hu et al., 2018; Wang et al., 2019; Zhao et al., 2018). Concerning nanomedicine, MOFs have been widely used as carriers or substrates for theranostic applications (Deng et al., 2019; Wu et al., 2018; Zhang et al., 2016). The adjustable pore size, extraordinary surface area, and abundant diversity of structure entitle MOFs to separate small molecules (e.g., gas, peptides, ions), yet the development of MOFs for macromolecule separation is still challenging (Lin et al., 2019; Wu et al., 2021; Zhang et al., 2017). Recently, an antibodies‐assisted MOF platform has been developed for EV isolation, named Tim4@ILI‐01 (Zhang et al., 2021). Yet this strategy still requires laborious modification and a long isolation time (> 2 h). Therefore, a rapid and efficient EV isolation platform based on MOFs remained largely unexplored.

Herein, we report a MOF platform termed EV‐FISHER for the rapid enrichment and isolation of EVs. EV‐FISHER is constructed from defected Zirconium (Zr)‐based MOF of UiO‐66‐NH₂ hybridized with cleavable lipid probes PO₄ ³⁻‐spacer‐DNA‐cholesterol (PSDC). As demonstrated in Figure 1(a), the UiO‐66‐NH₂ is first activated to acquire the defect structure, which provides it with more abundant modification sites. Next, the defected UiO‐66‐NH₂ is readily associated with PSDC through PO₄ ³⁻‐Zr(IV) (Ji et al., 2018) to form EV‐FISHER (PSDC@MOF), ensuring a simple and high‐affinity connection (Liu et al., 2014). Consequently, MOF UiO‐66‐NH₂ acts as the ‘handle’ to bind with the ‘fishing line’ PSDC.

FIGURE 1.

Schematic illustration of (a) the synthesis of EV‐FISHER and (b) the workflow of plasma EVs isolation

As illustrated in Figure 1(b) and Table S2, the EV‐FISHER ‘baits’ EVs from plasma with cholesterol and enriches them via low‐speed centrifugation. Subsequently, deoxyribonuclease I (DNase I) is added to hydrolyse the DNA in PSDC to detach EVs. Finally, purified EVs can be obtained after removing the MOF segment. The high affinity of cholesterol toward EVs’ membrane coupled with the advantages of MOF entitled EV‐FISHER to outperform the UC method in terms of time (∼40 min vs. 240 min), isolation efficiency (74.2% vs. 18.1%), and isolation requirement (12 800 g vs. 135 000 g). This work presents a promising strategy for fast and efficient plasma EVs isolation in general clinical laboratories for downstream analysis.

2. MATERIALS AND METHODS

2.1. Chemicals and materials

Zirconium(IV) chloride oxide octahydrate (ZrOCl₂·8H₂O, #Z820694), 2‐Amino‐1,4‐benzendicarboxylic acid (NH₂‐BDC, #10312‐55‐7), Hoechst 33258 (#B860511), and poly‐L‐lysine (#P856789) were purchased from Macklin. N, N‐Dimethylformamide (DMF, AR, 99.5%), ethanol absolute (AR, 99.7%), and acetic acid (AR, 99.5%) were obtained from Guangdong Guanghua Sci‐Tech. Glucose oxidase (#G2133), catalase (#C40), cysteamine (MEA, #30070), Tris base (#T1503), D‐(+)‐Glucose (#G7528), DNase I (#D5025), cholesterol (CHO, #C8667) and PKH67 Green Fluorescent Cell Linker Midi Kit (PKH67, #MIDI67) were supplied by Sigma‐Aldrich. Sodium chloride (NaCl, #C111533) was bought from Aladdin. 4‐(2‐Hydroxyethyl)‐1‐piperazine ethane sulfonic acid (HEPES, #H1095) was purchased from Solarbio. Hydrogen chloride (#C06808852) was purchased from Nanjing Reagent. Dulbecco's Modified Eagle Medium (DMEM), Roswell Park Memorial Institute 1640 (RPMI 1640), foetal bovine serum (FBS), penicillin‐streptomycin, and phosphate buffer saline (PBS, 1×, pH 7.4) were purchased from Gibco. Antibodies CD63 (#510953), CD9 (#380441), glypican‐1 (GPC‐1, #R22799) and TSG101 (#R25999) were obtained from ZEN BIO. Goat anti‐rabbit IgG/FITC antibody (#bs‐0295G‐FITC), goat anti‐rabbit IgG/Alexa Fluor 647 antibody (#bs‐0295G‐AF647) and goat anti‐rabbit IgG/gold antibody (#bs‐0295G‐Gold) were brought from Bioss. DNA sequences used in the paper were obtained from Sangon Biotech and listed in Table S3. All the materials were used as received without further purification. Ultrapure water was obtained from a PURELAB Classic system (ELGA LabWater).

2.2. Synthesis of UiO‐66‐NH2

UiO‐66‐NH₂ was synthesized according to the previous reports (Chen et al., 2019; Wang et al., 2017) with slight modification. The solution of ZrOCl₂·8H₂O (20 mg, 62.1 μmol) in DMF (5 ml) was added to a vial containing DMF (7 ml) and acetic acid (1.5 ml), after which the solution of NH₂‐BDC (11.7 mg, 64.6 μmol) in DMF (3 ml) was injected into the vial. Then the mixture was sonicated for 20 min and stayed at 120°C for 12 h to produce UiO‐66‐NH₂ nanoparticles. The resulting powder was collected by centrifugation (12 800 g, 10 min), and washed with ethanol three times before drying at 80°C under reduced pressure for activation.

2.3. Synthesis of EV‐FISHER

The activated UiO‐66‐NH₂ powder (10 mg) was dispersed in ultrapure water (25 ml) containing PSDC (2.8 μM). The mixture was left on a shaker for 12 h incubation at room temperature, and NaCl solution was slowly added to reach a concentration of 0.5 M. The formed EV‐FISHER (PSDC@MOF) was finally collected by centrifugation (12 800 g, 10 min) and washed with ultrapure water five times to remove unbound PSDC. Finally, the EV‐FISHER was re‐dispersed in HEPES buffer (25 mM HEPES, 100 mM NaCl) and stored at 4°C for further use.

2.4. Preparation of model EVs

Breast cancer cells MDA‐MD‐231 were cultured in DMEM containing 10% FBS and 1% antibiotics (penicillin‐streptomycin, 10 000 U/ml) with 5% CO₂ in air at 37°C. The cells were washed with PBS when they reached 70% coverage and further cultured in DMEM for 12 h. After that, cells were rewashed with PBS and cultured for 48 h in DMEM containing 1% EVs depleted FBS. Cell supernatant was then harvested and centrifuged at 300 g, 2000 g, and 10 000 g, respectively, for the removal of dead cells and cell debris. The resulting cell medium was centrifuged under 135 000 g for 70 min (SW 32 Ti Swinging‐Bucket Rotor, Beckman), and the precipitate was resuspended with HEPES buffer. This ultracentrifugation procedure was repeated. The resulting pellet (model EVs) was collected and stored at ‐80°C for further use.

2.5. Characterization of lipid probe on MDA‐MB‐231 cells by confocal microscopy

Breast cancer cells MDA‐MB‐231 were seeded in glass culture dishes with proper cell density. The cells were washed with PBS, and then DMEM containing Cy5‐DNA‐CHO (10 nM) or Cy5‐DNA (10 nM) were added for 2 h incubation with 5% CO₂ in air at 37°C. After that, cells were washed with PBS followed by stained with nuclear dye Hoechst 33258 for 10 min before observation (LSM 880, Carl Zeiss). The samples were observed with the excitation of 405 nm for Hoechst 33258 and 633 nm for Cy5.

2.6. Characterization of lipid probe on model EVs by super‐resolution microscopy

The glass culture dishes were first activated by poly‐L‐lysine (1 mg/ml, 100 μl) for 30 min at room temperature. Then, model EVs from MDA‐MB‐231 cells (10⁹ particles, 250 μl) were seeded in activated culture dishes for 30 min of absorption. Excess EVs were removed by HEPES buffer, after which Cy5‐DNA‐CHO (10 nM) was added for 30 min incubation. Unbound Cy5‐DNA‐CHO was washed away by HEPES buffer, and EVs@Cy5‐DNA‐CHO complexes were further stained with membrane dye PKH67 (0.5 μl of the PKH67 ethanolic dye solution + 250 μl of the Diluent C) for 3 min in the dark. Next, the dye was removed, and the sample was washed with HEPES buffer five times. Finally, the dishes were covered with the imaging buffer (see more details in Supporting Information) for observation (Nikon N‐STORM). Lasers with 488 and 647 nm wavelengths were selected for dye PKH67 and Cy5, respectively. Each channel was set to collect 10 000 images for analysis.

2.7. Isolation of model EVs by EV‐FISHER

2.7.1. Enrichment of model EVs

The model EVs with 8 × 10⁸/ml were incubated with EV‐FISHER (0.4 mg/ml, 500 μl) for 10 min on a HulaMixer® Sample Mixer at room temperature. Next, the samples were centrifuged (12 800 g, 10 min) and suspended in HEPES buffer to obtain EVs@EV‐FISHER complexes.

For transmission electron microscopy (TEM, Hitachi H‐7500) imaging, the EVs@EV‐FISHER solution was added to a copper grid for fixation and stained with 2% phosphotungstic acid for 5 min. The excess solution was removed with filter paper carefully. Finally, the samples were dried at room temperature naturally before observation.

For atomic force microscopy (AFM, Bruker) scanning, the EVs@EV‐FISHER solution was dropped on a mica sheet for fixation, followed by carefully drying the mica sheet before scanning.

2.7.2. Release of model EVs

DNase I (25 Units) dissolved in cleavage buffer (HEPES buffer, 2.5 mM Mg²⁺, 0.5 mM Ca²⁺) was added to EVs@EV‐FISHER complexes to release the captured EVs. The released EVs in the supernatant was collected after centrifugation (12 800 g, 10 min).

2.8. Calculation of isolation efficiency by nanoparticle tracking analysis (NTA)

The isolation efficiency (recovery rate) was defined as R = Nr/N, in which Nr represented the particle number recovered by EV‐FISHER and N denoted the particle number of model EVs used as the source material.

2.9. Characterization of model EVs membrane proteins by super‐resolution microscopy

The isolated model EVs by EV‐FISHER were absorbed on a poly‐L‐lysine (1 mg/ml, 100 μl) activated culture glass dish for 30 min, and excess EVs were washed away by HEPES buffer. Then EVs were stained with dye PKH67 (0.5 μl of the PKH67 ethanolic dye solution + 250 μl of the Diluent C) for 3 min in the dark. Next, the dye was washed away with HEPES buffer five times. Further, the EVs were labelled with antibodies CD63, CD9, and GPC‐1 (1:100), respectively, followed by Alexa Fluor 647 second antibody (1:100). The labelled EVs were covered by the imaging buffer (see more details in Supporting Information) for observation (Nikon N‐STORM). Samples were excited with 488 nm for dye PKH67 and 647 nm for Alexa Fluor 647 fluorescent antibody, respectively. Each channel was set to collect 10 000 images for analysis.

2.10. Plasma collection and storage

The blood samples were collected by anti‐coagulant tubes with the addition of dipotassium ethylene diamine tetraacetate (EDTA‐2K, AOSAITE Medical Devices Co., Ltd.). After the blood was drawn from patients or healthy donors, they were centrifuged (1200 g, 5 min) to obtain the plasma within 2 h. Afterwards, the plasma samples were centrifuged (2500 g, 15 min) twice to remove the platelet and collect the platelet‐free plasma (PFP) in the supernatant. The PFP samples were stored at ‐80°C for further use.

2.11. Isolation of plasma EVs by EV‐FISHER

The PFP (1 ml) was diluted to 3 ml with HEPES buffer before incubated with EV‐FISHER (0.4 mg/ml, 1 ml) on a HulaMixer® Sample Mixer at room temperature. Next, the samples were centrifuged (12 800 g, 10 min) and washed with HEPES buffer to obtain EVs@EV‐FISHER complexes. Then, DNase I (50 Units) dissolved in cleavage buffer (HEPES buffer, 2.5 mM Mg²⁺, 0.5 mM Ca²⁺) was added to EVs@EV‐FISHER complexes to release the captured EVs. Finally, the released EVs were collected in the supernatant after centrifugation (12 800 g, 10 min). All the plasma samples were obtained from Nanfang Hospital, Southern Medical University, with approval (NFEC‐201512‐K2).

2.12. Isolation of plasma EVs by UC

The PFP (1 ml) was diluted to 4 ml with PBS, followed by centrifugation (135 000 g, 70 min). Next, the obtained precipitation was suspended by PBS and subjected to the second round of centrifugation (135 000 g, 70 min). Finally, the sediment (EVs) was collected with PBS.

2.13. Isolation of plasma EVs by SEC

The isolation of plasma EVs by SEC was achieved using the EV isolation kit (qEVoriginal Size Exclusion Columns, Izon) according to the instructions. Briefly, the column was attached upright and flushed with 10 ml PBS. Subsequently, the PFP (500 μl) was loaded onto the loading frit and run into the column, followed by elution with PBS. Finally, the EV component was collected in a fraction from 2.5 to 4 ml of elution.

2.14. GPC‐1 EVs detection by nano flow cytometer (nFCM)

The PFP (150 μl) were incubated with EV‐FISHER (0.4 mg/ml, 500 μl) for 10 min on a HulaMixer® Sample Mixer at room temperature. Next, the samples were centrifuged (12 800 g, 10 min) and washed with HEPES buffer to obtain EVs@EV‐FISHER complexes. Antibody GPC‐1 (1:500) was introduced in the EVs@EV‐FISHER for 30 min to label the captured EVs. Then the GPC‐1@EVs@EV‐FISHER complexes were washed and further labelled with a second antibody (FITC) for 30 min. Next, the samples were washed and treated with DNase I (25 Units) to release the labelled EVs. Finally, the labelled EVs were collected by centrifugation (12 800 g, 10 min) for nFCM analysis.

The nano flow cytometer was purchased from Xiamen NANOFCM. Before the measurement, the flow cytometer system was purged with company‐equipped wash buffer and ultrapure water and calibrated according to instructions (Tian et al., 2018). Then, the EV samples (100 μl) were loaded on the flow cytometer for analysis (FITC fluorescent channel).

2.15. Statistical analysis

Statistical analysis was performed using Student's two‐tailed t‐tests in Graph Prism 8 (GraphPad Software, CA) to determine significant differences. The data are presented as mean ± standard error of the mean of at least three independent experiments. P < 0.05 was considered statistically significant.

3. RESULTS

3.1. Synthesis and characterization of EV‐FISHER

The UiO‐66‐NH₂ nanoparticles were synthesized according to the previous reports with optimization (Chen et al., 2019; Wang et al., 2017). Specifically, reactions of ZrOCl₂·8H₂O with NH₂‐BDC in the presence of acetic acid gave rise to the UiO‐66‐NH₂ nanoparticles. The obtained particles were heated at 80°C under reduced pressure for activation, followed by incubation with PSDC to obtain EV‐FISHER.

Transmission electron microscopy (TEM) indicated that the morphologies of UiO‐66‐NH₂ and EV‐FISHER were comparable (Figure 2a–b), with the octahedral shapes of UiO‐66‐NH₂ replicated in EV‐FISHER, suggesting that PSDC modification has a negligible impact on the MOF structure. Dynamic light scattering (DLS) analysis indicated that the particle size increased from 301 nm for UiO‐66‐NH₂ to 342 nm for EV‐FISHER (Figure 2c). Upon decorating with PSDC, the zeta potential also shifted from positive (4.6 mV for UiO‐66‐NH₂) to negative (−21.7 mV for EV‐FISHER) due to the negatively charged PO₄ ³⁻ and DNA (Figure 2d).

FIGURE 2.

Characterization of EV‐FISHER platform. TEM images of (a) UiO‐66‐NH₂ and (b) EV‐FISHER. (c) The size distribution and (d) zeta potential analysis of UiO‐66‐NH₂ and EV‐FISHER, respectively. (e) The FTIR spectra of phosphate (blue), UiO‐66‐NH₂ (black), and EV‐FISHER (red). (f) The EDX analysis of EV‐FISHER. (g) The XPS patterns Zr 3d peaks in UiO‐66‐NH₂ (I) and EV‐FISHER (II). (h) The UV‐Vis spectrum for EV‐FISHER modified with different concentration of PSDC; (A‐N) in the image indicated the concentration of PSDC was 0, 0.1, 0.3, 0.5, 0.7, 0.9, 1.1, 1.4, 1.7, 2.0, 2.5, 3.5, 5.0, 7.0 μM, respectively. (i) The PXRD patterns of UiO‐66‐NH₂ soaked in various solvents

In the Fourier transform infrared (FT‐IR) spectroscopy of EV‐FISHER, there is a new peak of the PO₄ ³⁻ group at 1049 cm⁻¹ as compared to UiO‐66‐NH₂ (Figure 2e). The energy dispersive X‐ray (EDX) analysis (Figure 2f) shows the presence of the P element in EV‐FISHER. The P 2p peak (134.08 eV) also appeared in X‐ray photoelectron spectroscopy (XPS) (Figure S1). The Zr 3d peaks with the binding energy of 185.25 eV and 182.86 eV in UiO‐66‐NH₂ slightly changed to 185.10 eV and 182.74 eV in EV‐FISHER (Figure 2g), accountable for its coordination to PO₄ ³⁻. The absorption peak at 278 nm in the ultraviolet–visible (UV‐Vis) spectrum of UiO‐66‐NH₂ slightly blue‐shifted to 275 nm with the increased concentrations of PSDC to form EV‐FISHER (Figure 2h). In addition, the binding of PSDC could break the ligand to metal charge transfer (LMCT) effect on UiO‐66‐NH₂, thus recovering its fluorescence (Figure S2) (Yang et al., 2015). The above results collectively authenticated the successful coating of PSDC on the surface of UiO‐66‐NH₂ via the Zr‐PO₄ ³⁻ coordination and thus the reliable construction of EV‐FISHER.

Since the stability of EV‐FISHER is of great importance during EVs isolation, the structural stability of both UiO‐66‐NH₂ and EV‐FISHER were assessed. The powder X‐ray diffraction (PXRD) patterns of UiO‐66‐NH₂ immersed in various solvents were in good agreement with the simulated pattern, implying its bulk phase stability (Figure 2i). Moreover, no free ligand NH₂‐BDC or free Zr(IV) ion was detected in the supernatant from EV‐FISHER by mass spectrum (Figure S3) and inductively coupled plasma mass spectrometry (ICP‐MS) (Table S4) neither. Additionally, the EV‐FISHER showed excellent stability in ultrapure water for 45 days (Figure S4) and exhibited better dispersivity than UiO‐66‐NH₂ in the HEPES buffer (Figure S5). In summary, the prepared EV‐FISHER shows sufficient stability to fulfil the requirements of EVs isolation in biological fluids.

3.2. Characterization of cholesterol‐decorated lipid probe

As a lipid probe, PSDC was designed to isolate EVs by inserting cholesterol into the EVs’ phospholipid‐bilayer membrane. We first investigated this approach at the cell level, as the membrane components of EVs are similar to that of their parent cells. Instead of the PSDC, the fluorophore Cy5 labelled lipid probe Cy5‐DNA‐cholesterol (denoted as Cy5‐DNA‐CHO) was employed and incubated with MDA‐MB‐231 cells for 2 h. As shown in Figure 3(a), the red fluorescence confirmed the successful insertion of the Cy5‐DNA‐CHO. In contrast, the red fluorescence was absent in the control groups treated with Cy5‐DNA, suggesting that cholesterol is a critical component in inducing lipid membrane binding.

FIGURE 3.

Characterization of cholesterol‐decorated lipid probe. (a) The confocal images of breast cancer cells MDA‐MB‐231 treated with Cy5‐DNA‐CHO and Cy5‐DNA, respectively. (b) The lipid probes Cy5‐DNA‐CHO (red) bind to model EVs (green) derived from MDA‐MB‐231 cells characterized by super‐resolution microscopy. (c) The TEM images of synthesized liposomes with good dispersion and united shape and the (d) Liposomes@EV‐FISHER complexes. (e) The SDS‐PAGE analysis of the combination between PSDC (1 μM) and EVs. Lane 1: marker, lane 2: PSDC only, lane 3: EVs (2.5 × 10⁹) only, lane 4: PSDC + 1 × 10⁹ EVs, lane 5: PSDC + 2.5 × 10⁹ EVs, lane 6: PSDC + 5 × 10⁹ EVs, lane 7: PSDC + 1 × 10¹⁰ EVs

The lipid probe Cy5‐DNA‐CHO and membrane binding were further verified on EVs. The model EVs collected from MDA‐MB‐231 cells were first incubated with Cy5‐DNA‐CHO (red fluorescent) and stained with membrane dye PKH67 (green fluorescent). Super‐resolution imaging pictures depicted excellent co‐localization of model EVs (green) and the lipid probes (red) (Figure 3b) with a co‐localization ratio of 87% (Figure S6), showcasing efficient recognition of cholesterol towards the EVs membrane. Additionally, EV‐like artificial liposomes with negatively charged (Figures 3c and S7) were utilized to mimic the standard EVs consisting of only phospholipid bilayers. The liposomes showed good binding with EV‐FISHER (Figure 3d and S8), which indicates the recognition between cholesterol and membrane only involves the phospholipid layer. As demonstrated in sodium dodecyl sulphate‐polyacrylamide gel electrophoresis (SDS‐PAGE) analysis (Figure 3e), the addition of an increasing concentration of EVs to PSDC (lanes 4–7) showed decreasing bands for the PSDC, and EVs adducts are too large to enter the gel. Taken together, our results confirmed that the lipid probe PSDC is capable of binding to EV membrane.

3.3. Fishing model EVs with EV‐FISHER

We subsequently characterized the feasibility of EV‐FISHER for fishing EVs, and the model EVs from MDA‐MB‐231 cells were used as the source material. Although there were reports on directly capturing peptides or exosomes with MOFs (Chen et al., 2015; Han et al., 2017; Zhang et al., 2020), our findings show that EVs do not attach to the UiO‐66‐NH₂ without decorating PSDC (Figure 4a). On the contrary, model EVs with typical membrane structure (red arrows) can be observed on EV‐FISHER (Figure 4b, S9), and similar results were obtained by atomic force microscopy (AFM) as indicated by green arrows (Figures 4c and S10), implying the successful fishing of EVs. Furthermore, the membrane proteins CD63, CD9, and GPC‐1 of the captured EVs were identified as active by immunogold labelling assay (Figure S11). These results demonstrated that EVs maintained structural integrity after being captured by EV‐FISHER.

FIGURE 4.

Enrichment and isolation of model EVs by EV‐FISHER. TEM images of (a) UiO‐66‐NH₂ and (b) EV‐FISHER after capturing EVs. The arrows in Figure 4b represent the captured EVs. (c) The AFM image of EVs@EV‐FISHER complexes. The arrows in the figure represent the captured EVs. (d) Characterization of the hydrolysis ability of DNase I towards PSDCs by SDS‐PAGE; lane 1: marker, lane 2: PSDC7, lane 3: PSDC7 + DNase I, lane 4: PSDC5, lane 5: PSDC5 + DNase I, lane 6: PSDC3, lane 7: PSDC3 + DNase I, lane 8: PSDC1, lane 9: PSDC1 + DNase I. The arrows in the image implied the hydrolysed DNA fragments. (e) EV concentrations after treated with DNase I (red) and Triton X‐100 (black) at different time points. (f) Particle counting of model EVs isolated by EV‐FISHER. Optimization of the EV‐FISHER with (g) PSDC with different base‐pair lengths, (h) PSDC8 concentration, and (i) incubation time with EVs (n = 3)

Following the successful EVs fishing, we studied the subsequent release of EVs. As shown in Figure 4(d), DNase I efficiently cleaved PSDC regardless of lengths (PSDC1: 36 bp, PSDC3: 28 bp, PSDC5: 22 bp, and PSDC7: 16 bp) and produced DNA fragments (red arrows). It is also suggested that the EV concentration was hardly affected by DNase I (Figure 4e), while it was distinctly reduced by Triton X‐100, indicating DNase I does not affect the integrity of EVs. Furthermore, the captured model EVs were released from the EVs@EV‐FISHER system after DNase I treatment (Figure 4f), while no EVs appeared in the control groups (Figure S12). These results confirmed that EV‐FISHER could successfully isolate model EVs.

Three significant factors that affect EV's capture and release, including the PSDC length, PSDC concentration, and incubation time, were assessed to improve the isolation efficiency of EV‐FISHER. According to the previous reports, the PSDC with long single‐stranded DNA might be adsorbed on the surface of UiO‐66‐NH₂, thus losing its mobility and decreasing capturing ability (Yang et al., 2015; Zhang et al., 2014; Zhao et al., 2016). Therefore, PSDC with different lengths (PSDC1, 36 bp) to (PSDC9, 10 bp) (Table S3) were modified on UiO‐66‐NH₂, and the resulting EV‐FISHER platforms were named PSDC1@MOF to PSDC9@MOF, respectively. As displayed in Figure 4(g) and S13, the isolation efficiencies were nearly inversely related to the length of PSDC, with the highest rate of 73.7% obtained for PSDC8 (13 bp).

The PSDC density on the UiO‐66‐NH₂ surface was subsequently evaluated using PSDC8. The isolation efficiency gradually improved with the increasing amount of PSDC8, and the best performance was achieved when the concentration of PSDC8 was 2.8 μM. However, the isolation efficiency dropped when the PSDC8 concentration was further increased to 4 μM, presumably due to the steric hindrance (Figure 4h). Therefore, the PSDC8 with a concentration of 2.8 μM was chosen as the optimal concentration to rule out the preferred incubation time. Figure 4(i) exhibited that 74.2% of isolation efficiency was obtained at 10 min incubation, and extending incubation time did not hamper or improve the efficiency, implying the EVs fishing was completed at that time.

According to the above results, the mechanism of the efficient isolation performance of EV‐FISHER could be elucidated. The surface of EVs is negatively charged, which exhibits weak interaction with the similarly charged molecules. However, the MOF UiO‐66‐NH₂ with positive charge (4.6 mV, Figure 2d) showed almost negligible isolation capacity on EVs (Figure 4a, S12), while EV‐FISHER (−21.7 mV, Figure 2d) exhibited high EVs isolation efficiency (Figure 4g). These findings might indicate that EV‐FISHER captures EVs mainly depend on the lipid probe PSDC rather than electric charge. Additionally, there were abundant non‐EV particles with negative charge within biofluids besides EVs. The negative charge surface of EV‐FISHER is expected to reduce the unspecific absorption, thus improving the purity and yield of EVs.

Furthermore, to confirm the universality of the EV‐FISHER, another two types of EVs derived from red blood cells (RBC) and fibroblast (L929) were tested. As demonstrated in Figure S14, EV‐FISHER can also isolate these two types of EVs with the efficiency of 81.4% ± 3.21% and 69.4% ± 2.88%, respectively (n = 5), implying the EV‐FISHER should be applicable for EVs from various cell origins.

3.4. Characterization of the model EVs isolated by EV‐FISHER

The model EVs isolated by EV‐FISHER showed the classic ‘cup shape’ membrane structure under TEM (Figures 5a, S15), indicating the isolation process does not damage the integrity. The NTA result revealed that 91.5% of the isolated EVs were within 300 nm with a peak at 123.0 nm (Figure 5b), which was in good accordance with the EVs definition. Furthermore, as expected, three EV markers of CD63, CD9, and GPC‐1 were present in the isolated EVs, but not calnexin (a negative control marker) (Figure 5c). Additionally, these proteins (red fluorescence) were identified as active by super‐resolution imaging (Figure 5d) with excellent EVs membrane (green fluorescence) colocalization. Besides, the isolated EVs could be effectively taken up and internalized by MDA‐MB‐231 cells (Figure S16). In summary, EV‐FISHER enables the isolation of EVs without damaging their structure and protein components.

FIGURE 5.

Characterization of isolated model EVs by EV‐FISHER. (a) TEM images of the isolated model EVs by EV‐FISHER. (b) Size distribution of the model EVs isolated by EV‐FISHER. (c) The western blot for protein analysis of the model EVs isolated by EV‐FISHER. (d) The super‐resolution microscopy images of the isolated model EVs (green) labelled with fluorescent antibodies (red) CD63, CD9, and GPC‐1, respectively (scale bar: 200 nm). (e) Comparison of isolation efficiency between UC and EV‐FISHER using model EVs as source material. (f) The recovery of model EVs isolated by EV‐FISHER with the addition of haemoglobin, bilirubin, and chyle, respectively

The isolation efficacy of EV‐FISHER and UC was compared through the model EVs recovery rates. The result showed that EV‐FISHER recovered 74.2 ± 7.1% of EVs within an hour, while UC yielded a recovery rate of only 18.1 ± 3.5% after 4 h of laborious processing (Figure 5e), demonstrating the superiority of EV‐FISHER in the isolation efficiency, time, and cost. Subsequently, a universal EV biomarker CD9 was employed to validate the quality of EVs isolated by two methods. It was found that a similar protein ratio was observed in two groups on nano flow cytometer (nFCM) analysis (Figure S17), indicating EV‐FISHER has the potential to be a substitute for UC in EV isolation for downstream analysis. In addition, the EV‐FISHER proposed a relatively stable isolation efficiency with the coefficient of variation of 7.07% ± 1.77% over 7 days (Figure S18).

Before investigating the practical application, it is necessary to evaluate the anti‐interference ability of EV‐FISHER. According to the guideline of Interference Testing in Clinical Chemistry (McEnroe et al., 2018) published by Clinical and Laboratory Standards Institute, the model EVs were isolated by EV‐FISHER with the addition of three common interferons in plasma, including haemoglobin, bilirubin, and chyle, respectively. Then, the universal biomarker CD9 of the isolated EVs was analysed in different groups by nFCM. As demonstrated in Figure 5(f), there is no difference of CD9 EVs ratio with the addition of haemoglobin (15.0 μM and 25.0 μM, respectively; clinical reference range: < 18.8 μM), bilirubin (10.3 μM and 17.1 μM, respectively; clinical reference range: 3.4−17.1 μM), and low concentration of chyle (2.6 mM; clinical reference range: 0−5.2 mM), while only a slight reduction in the group with a high concentration of chyle (5.2 mM). Therefore, EV‐FISHER could potentially isolate EVs in plasma.

3.5. Fishing plasma EVs with EV‐FISHER

Encouraged by the above results, EV‐FISHER was employed to isolate plasma EVs. It is shown that EV‐FISHER proposed excellent enrichment efficiency toward plasma EVs (Figure 6a). Moreover, the isolated plasma EVs distributed from 15 to 765 nm by NTA (Figure 6b) and the EV markers CD9, CD63 and TSG 101 were also identified (Figure S19). As illustrated in Figure 6(c), there were 9.04 ± 0.68  × 10¹⁰ EVs particles isolated from 1 ml of plasma by EV‐FISHER, which is 4.48 times more than by UC (2.02 ± 0.39 × 10¹⁰), suggesting a better yield of EV‐FISHER in practical plasma samples.

FIGURE 6.

Characterization of the plasma EVs isolated by EV‐FISHER. (a) TEM image of EV‐FISHER capturing EVs from plasma. (b) Size distribution of plasma EVs isolated by EV‐FISHER. (c) The concentration analysis of EVs isolated from 1 ml of plasma by UC and EV‐FISHER, respectively. (d) Gene Ontology analysis of the peptides from plasma EVs isolated by EV‐FISHER, UC, and SEC, respectively

Moreover, a 4D label‐free proteomic profiling was applied to characterise the plasma EVs isolated by EV‐FISHER, using EVs isolated by UC and size‐exclusion chromatography (SEC) as references. The results showed that 277, 311 and 309 types of proteins were detected from the EVs isolated by EV‐FISHER, UC and SEC, respectively (Supplemental File: ‘Protein quantitation’, Table S5). All the proteins were at least detected in two isolation methods, suggesting good consistency among EV‐FISHER and these two well‐accepted EV isolation approaches (Figure S20). Additionally, the Venn diagram also demonstrated that 247 proteins were identified in all three types of methods (Figure S20), with 79.42% and 79.94% of the proteins from UC and SEC overlapped with EV‐FISHER, respectively. It is deduced that the slight diversity of the proteomics results between the three isolation methods might be attributed to the different isolation principles. In addition, 21 out of the top 100 proteins from the Vesiclepedia database were identified by EV‐FISHER, where 100% and 90.5% were consistent with UC and SEC, respectively. According to the cellular distribution analysis of the identified peptides, EV‐FISHER demonstrated similar protein distribution patterns with UC and SEC among three functional regions, including biological process, cellular component, and molecular function (Figure 6d). In addition, according to our proteomic results, there were 15 types of apolipoproteins detected. As shown in Figure S21, the EV‐FISHER demonstrated obviously less apolipoprotein contamination than SEC in plasma EVs isolation and exhibited lower apolipoprotein concentration than UC in 10 of 15 apolipoproteins (Table S6 and Supplemental File: ‘Protein quantitation’). Therefore, the established EV‐FISHER platform displays a better performance over UC and SEC in apolipoprotein elimination during EV isolation from blood.

3.6. Breast cancer diagnosis and monitoring using plasma EVs isolated by EV‐FISHER

The clinical adaptability of EV‐FISHER was further studied in clinical cohorts. Firstly, the plasma samples from a breast cancer cohort (Figure 7a, Cohort 1, Table S7 and Table S8) consisting of breast cancer patients (BC) with different subtypes and clinical stages (n = 200; that is 84 for stage I, 75 for stage II, 41 for stage III&IV) and healthy donors (HD, n = 30) were collected. Next, the plasma EVs were isolated by EV‐FISHER, labelled with breast cancer biomarker GPC‐1 fluorescent antibody, and conducted single EV analysis on nFCM. As demonstrated in Figure 7(b), a significant analysis of the GPC‐1 EVs between healthy donors and breast cancer patients was performed, conforming to the previous report (Melo et al., 2015). It is also found that the GPC‐1 EVs in the bloodstream increased from stage I to stage IV, which might be able to utilize as an indicator of breast cancer development. In addition, the receiver operating characteristic (ROC) curves showed an effective classifier with an area under the curve (AUC) of 0.835 to differentiate healthy individuals and breast cancer patients (Figure 7c). Specifically, better diagnostic efficacy was shown for the cancer patients in stage I, stage II, and stage III&IV with the AUC of 0.724, 0.854, and 0.872, respectively.

FIGURE 7.

Detection of GPC‐1 EVs isolated from breast cancer patients and healthy donors. (a) The workflow of breast cancer diagnosis and therapeutic efficacy monitoring includes plasma collection, plasma EVs isolation, and GPC‐1 EVs detection. (b) The GPC‐1 EVs in healthy donors and breast cancer patients across clinical stages. (c) ROC curve analysis of breast cancer patients with different clinical stages and healthy donors. (d) Dynamic change of GPC‐1 EVs in breast cancer patients before and after treatments. Variation of GPC‐1 EVs in breast cancer patients with (e) stage I, (f) stage II, and (g) stage III&IV before and after treatments

Additionally, it is reported that the cancer‐derived EVs subpopulation of a patient keeps dynamic fluctuation, which could accurately reflect the therapeutic efficacy. Thus, we also collected plasma samples from breast cancer patients before (BC) and after treatments (BC‐T) (Figure 7a, Cohort 2, Table S9) for GPC‐1 EVs identification. Figure 7(d) depicts a significant reduction in GPC‐1 EVs after breast cancer patients underwent treatments, especially for the patients in stages I and II (Figure 7e‐f). However, the results showed no statistical difference (p = 0.0973) of GPC‐1 EVs in the patients in stages III&IV before and after treatments (Figure 7g), which may be accounted for the insufficient cases. Considering the above results, the EV‐FISHER is capable of isolating plasma EVs rapidly and efficiently, with good compatibility on downstream single EV subpopulation analysis for clinical cancer diagnostic and treatment efficacy evaluation.

4. CONCLUSION

In summary, a universal EV isolation platform of EV‐FISHER has been constructed based on the lipid probe functionalized UiO‐66‐NH₂. EV‐FISHER exhibited merits over UC in terms of its advantages: fast, efficient, and without requiring expensive equipment. The established platform demonstrates the sufficient capability to harvest EVs from plasma in an effective approach. Combining the downstream single EV analysis on nFCM, EV‐FISHER enables quick and accurate identification of breast cancer at early stages and provides information on therapeutic efficacy evaluation. Moreover, owing to the delicate design of this isolation strategy, the proposed MOF‐based EV isolation platform can be repeated in clinical laboratories with available commercial reagents. This study sheds insights into the design of the clinical applicable EV isolation strategy, promoting the transformation of EV research in liquid biopsy.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

Supporting information

Figure S1. XPS spectra of (a) UiO‐66‐NH₂ and (b) EV‐FISHER.

Figure S2. Fluorescence spectrum of UiO‐66‐NH₂ before and after the addition of PSDC under the excitation at 328 nm.

Figure S3. Mass spectrum analysis of ligand NH₂‐BDC, HEPES buffer, and supernatant from EV‐FISHER. The peak 179.86 represented ligand NH₂‐BDC, and peaks 236.94 and 236.92 represented HEPES.

Figure S4. Size distribution of UiO‐66‐NH₂ and EV‐FISHER in 45 days.

Figure S5. Image of (a) UiO‐66‐NH₂ and (b) EV‐FISHER in HEPES buffer under 2 h incubation.

Figure S6. The lipid probes Cy5‐DNA‐CHO (red) bind to model EVs (green) derived from MDA‐MB‐231 cells characterized by super‐resolution microscopy. (Scale bar: 2 μm).

Figure S7. (a) Size distribution and (a) zeta potential analysis of synthesized liposomes.

Figure S8. TEM pictures of (a) liposome and (b‐d) Liposomes@EV‐FISHER complexes.

Figure S9. (a,b) TEM images of EVs@EV‐FISHER complexes. Red arrows in the pictures indicated the captured EVs.

Figure S10. (a‐d) AFM images of EVs@EV‐FISHER complexes. Green arrows in the pictures indicated the captured EVs.

Figure S11. TEM images of immunogold@EVs@EV‐FISHER complexes. The captured EVs were labelled with (a) CD63, (b) CD9 and (c) GPC‐1 immunogold, respectively. Red arrows in the pictures represented the binding immunogold on the surface of EVs.

Figure S12. (a) Particles analysis of model EVs samples isolated by UiO‐66‐NH₂. Particles analysis of (b) HEPES buffer and (c) cleavage buffer.

Figure S13. Size distribution of the isolated model EVs by (a‐i) PSDC1@MOF to PSDC9@MOF.

Figure S14. TEM images of EVs derived from (a) RBC and (b) fibroblast cells L929. Nanoparticles tracking analysis of (c) RBC and (d) L929 derived EVs isolated by EV‐FISHER. (e) The isolation efficiency of EV‐FISHER using RBC and L929 cells derived EVs as source material, respectively (n = 5).

Figure S15. (a,b) TEM images of model EVs isolated by EV‐FISHER.

Figure S16. Confocal images of MDA‐MB‐231 cells treated with model EVs (stained with PKH67) isolated by EV‐FISHER (scale bar: 20 μm).

Figure S17. Flow cytometry analysis of CD9 EVs isolated by UC and EV‐FISHER, respectively (n = 3).

Figure S18. Repeatability of the EVs isolation efficiency on EV‐FISHER over seven days.

Figure S19. Western blot analysis of plasma EVs isolated by EV‐FISHER.

Figure S20. Comparative analysis of EV cargo proteins isolated by EV‐FISHER, UC and SEC, respectively. The Venn diagram shows the overlap among proteomic data from three types of EVs isolation methods.

Figure S21. The apolipoprotein analysis of EVs isolated by EV‐FISHER, UC, and SEC, respectively.

Table S1. The methods that applied for plasma EVs isolation [2‐12].

Table S2. The timetable of each step for EVs isolation by EV‐FISHER.

Table S3. DNA sequences used in the paper were referred to a previous report [13] with modification to fit this design.

Table S4. ICP‐MS analysis of H₂O, PBS, PSDC, supernatant from UiO‐66‐NH₂, EV‐FISHER, and UiO‐66‐NH₂ + PBS, respectively for Zirconium ion detection.

Table S5. The retrieve parameters for mass spectrometry.

Table S6. The apolipoproteins analysis of EVs isolated by EV‐FISHER, UC, and SEC, respectively.

Table S7. The plasma samples of the breast cancer patients (Cohort 1).

Table S8. The plasma samples of healthy donors (Cohort 1).

Table S9. The plasma samples used in the manuscript from patients with breast diseases before or after treatments (Cohort 2).

Pan, W.‐L. , Feng, J.‐J. , Luo, T.‐T. , Tan, Y. , Situ, B. , Nieuwland, R. , Guo, J.‐Y. , Liu, C.‐C. , Zhang, H. , Chen, J. , Zhang, W.‐H. , Chen, J. , Chen, X.‐H. , Chen, H.‐Y. , Zheng, L. , Chen, J.‐X. , & Li, B. (2022). Rapid and efficient isolation platform for plasma extracellular vesicles: EV‐FISHER. Journal of Extracellular Vesicles, 11, e12281. 10.1002/jev2.12281