Multimode chromatography‐based techniques for high purity isolation of extracellular vesicles from human blood plasma

Abstract

Extracellular vesicles (EVs) play a pivotal role in various biological pathways, such as immune responses and the progression of diseases, including cancer. However, it is challenging to isolate EVs at high purity from blood plasma and other biofluids due to their low abundance compared to more predominant biomolecular species such as lipoprotein particles and free protein complexes. Ultracentrifugation‐based EV isolation, the current gold standard technique, cannot overcome this challenge due to the similar biophysical characteristics of such species. We developed several novel approaches to enrich EVs from plasma while depleting contaminating molecular species using multimode chromatography‐based strategies. On average, we identified 716 ± 68 and 1054 ± 35 protein groups in EV isolates from 100 µL of plasma using multimode chromatography‐ and ultracentrifugation‐based techniques, respectively. The developed methods resulted in similar EV isolates purity, providing significant advantages in simplicity, throughput, scalability, and applicability for various downstream analytical and potential clinical applications.

1. INTRODUCTION

Extracellular vesicles (EVs) are nanoscale membrane‐enclosed spherical structures that are shed from most cells within biological systems. EVs range in size, shape, density, and molecular composition, presumably reflecting the current state of the cell of origin when released into the extracellular space. Currently, EVs are classified into five major subpopulations: exomeres (<50 nm) (Zhang et al., 2018), exosomes (∼30–150 nm), microvesicles (∼100–1000 nm) (Harding et al., 1984), apoptotic bodies (>1000 nm) (Thery et al., 2018), and the newly termed supermeres (<50 nm, typically exhibiting smaller diameters than exomeres) (Zhang et al., 2021). Due to both exomeres and supermeres not having a lipid membrane, they should be referred to as non‐vesicular extracellular particles (NVEPs); however, they are still considered of interest (Zhang et al., 2018, 2021). The Minimal Information for Studies of Extracellular Vesicles 2018 (MISEV2018) is the current assembly of guidelines and considerations for researchers studying EVs that we have taken into consideration throughout this study (Thery et al., 2018). EVs consist of various biological molecules, including proteins, metabolites, and nucleic acids that are specific to their cell of origin as well as their biogenesis (Shah et al., 2018). Due to their critical roles in various biological processes such as immune responses and tumour progression while in circulation in the body, there is growing interest in exploiting EVs for liquid biopsy diagnostics as well as using their specificity for targeted therapeutics (Brown et al., 2020; Fais et al., 2016; Lu et al., 2019; Shah et al., 2018; Urabe et al., 2020). However, it is a challenge to isolate pure EVs from other biomolecular species, such as free proteins, protein complexes, lipoprotein particles, and other macromolecular components of complex biological matrices like blood plasma in a high throughput, time‐ and cost‐efficient process. Over the past decade, separation technology such as size exclusion chromatography (SEC) (Blans et al., 2017; Boing et al., 2014), tangential flow filtration (TFF) (Cheruvanky et al., 2007; Watson et al., 2018), precipitation kits (Helwa et al., 2017), and asymmetric flow field flow fractionation (AF4) (Petersen et al., 2014; Zhang et al., 2018) have made it possible to exploit the differences in EVs and to separate them from other highly abundant molecular and macromolecular species of complex biological specimens. These techniques appeared advantageous, increasing the effectiveness of isolation, compared to the traditional isolation methods based on ultracentrifugation (UC) and density gradient centrifugation, which both generate protein and EV aggregates (Boing et al., 2014) and are often low throughput. Additionally, each of these separation techniques has its own drawbacks.

UC‐based techniques, the current gold standard (Théry et al., 2006), uses differential centrifugation and UC at or above 100,000×g to isolate EVs based on density, which is a time‐consuming and expensive process, which does not appear to be easily scalable to process many samples at once (Lasser et al., 2012). However, due to the density and size overlap with several types of lipoprotein particles, UC‐based techniques are not adequate for sensitive molecular profiling, often resulting in contaminated samples (Karimi et al., 2018; Sodar et al., 2016). This is problematic for downstream EV characterization and proteomic analysis, as lipoprotein particles and their constituents can overload most analytical signals, making it nearly impossible to analyze intact EVs and their cargo with accuracy (Karimi et al., 2018). Particularly, apolipoprotein A‐I, the main protein constituent in high‐density lipoprotein (HDL), and apolipoprotein B100, the main protein in low‐density lipoprotein (LDL) (Mahley et al., 1984; Orlova et al., 1999) are of concern while working with EVs from plasma. Some reports have shown a higher purity EV sample can be obtained post‐UC processing by washing with PBS several times, but this will result in sample losses (Momen‐Heravi et al., 2013, 2012). Free plasma proteins are also of concern (serum albumin, immunoglobulins, alpha‐2‐macroglobulin, etc.) having concentrations > 1.5 orders of magnitude than apolipoproteins in plasma (Anderson & Anderson, 2002), which vastly limits the detection of lower abundant species such as EVs and their associated proteins. To alleviate this issue, we have exploited the properties of SEC, which is a widely used technique for isolation of EVs from plasma (Takov et al., 2019; Welton et al., 2015), to separate various biomolecules and macromolecular protein complexes based on their hydrodynamic diameter (Boing et al., 2014). There are multiple advantages of SEC, including higher throughput, cost‐effectiveness, ease of use, scalability, tolerance to various mobile phases, and the effectiveness of EV enrichment (Boing et al., 2014). While SEC is generally considered a low‐resolution technique, it typically provides sufficient separation power to resolve chromatographic peaks that correspond to EVs and other large Mda‐scale macromolecular assembles from smaller molecular species, such as high abundance proteins and protein complexes that elute later. However, the chromatographic peaks are usually not baseline resolved by the most common gravity flow‐ or pressure drop‐driven separations, using cartridge columns packed with only Sepharose CL‐2B or similar SEC beads (Boing et al., 2014; Takov et al., 2019).

In order to obtain higher resolution by SEC, we chose to evaluate the performance of two types of CaptoCore resins (Cytiva), a multimode fully porous stationary phase, which demonstrates both SEC and anion exchange (AEX) chromatography modes of separation at the same time. While CaptoCore resin allows negatively charged molecular species that are less than the molecular weight cut‐off (MWCO) of the two CaptoCore resins (i.e., 400 kDa or 700 kDa) become trapped on the AEX surface inside the beads by interacting with the positively charged octylamine functional group (James et al., 2016; Zhang & Liu, 2016). Meanwhile, the SEC mode enables the separation of EVs, which are generally larger than the MWCOs of the beads (i.e., the exclusion limit), from other smaller molecular and macromolecular species (James et al., 2016). CaptoCore has been used in a number of ways to purify complex biological samples. For example, CaptoCore has been used to isolate virus particles as an alternative to UC for a more time‐efficient isolation method (James et al., 2016; Zhao et al., 2015). There are multiple characteristics shared between EVs and virus particles, such as size, shape, and chemical composition, which all influence cellular uptake, and hence, the possibility for particular isolation approaches (Gould et al., 2003). CaptoCore has also been employed to remove unlabelled dye from EV samples intended to be analyzed by flow cytometry, named EV‐Clean (Welsh et al., 2021). Moreover, a recent work has been published to determine higher resolution separation of EVs from lipoprotein particles utilizing various forms of Sepharose (i.e., SEC only) resin and CaptoCore at the bottom of a SEC column (Ter‐Ovanesyan et al., 2023). Given the wide array of CaptoCore uses, we hypothesized that CaptoCore stationary phases could be efficiently used for EV purification from complex biological matrices in a simple, time‐efficient, cost‐effective, robust, and high throughput method compared to UC‐based isolation as well as providing higher or similar EV enrichment to already established SEC‐based methods, giving researchers and clinicians more options for effective and efficient EV isolation.

In this study, we evaluated the EV isolation performance of conventional SEC columns packed with Sepharose CL‐2B beads, a common SEC stationary phase for EV isolation (Boing et al., 2014; Monguio‐Tortajada et al., 2019), and determined the optimal column geometry for EV enrichment. Then, the two types of CaptoCore resins were investigated alone for their performance in plasma‐derived EV enrichment, which we named Suspension mode (Sus). Additionally, to further improve upon a traditional SEC method, which uses Sepharose CL‐2B, we employed the use of CaptoCore by layering the resin on top of CL‐2B beads in a biphasic chromatographic form we named a Sandwich (SW) column to allow for better resolution between EVs and free plasma proteins. The top CaptoCore layer also acts as a pretreatment to decrease the dynamic range between EVs and plasma proteins, prior to separation, and expectedly, the CaptoCore layer does not significantly contribute to additional band‐broadening as opposed to placing it at the bottom of the column (Figure 1a). The purity, size distribution, morphology, concentration, and proteomic molecular profiles of the resulting EV isolates were characterized using transmission electron microscopy (TEM), tunable resistive pulse sensing (TRPS), immunoaffinity‐based interferometry, gel electrophoresis, western blotting, nanoflow cytometry (nFCM), and proteomic analysis by nanoflow liquid chromatography coupled with tandem mass spectrometry (nLC‐MS/MS). Results show that successful EV isolation was achieved with substantial depletion of free plasma proteins using both Sus and SW modes of separation, saving overall sample processing time, with similar EV enrichment to the traditional Sepharose CL‐2B SEC‐ and UC‐based methods. The impact of the different pore sizes of the two types of CaptoCore beads on the isolation of EVs was also investigated. We determined that the EV recovery rate was higher using CaptoCore 400 beads, specifically Sus 400, without the use of an SEC column. However, this resin was less efficient in depleting free plasma proteins due to the smaller pore size, which did not trap as many free plasma proteins and protein complexes. In contrast, the CaptoCore 700 resin resulted in a lower EV recovery rate but in better efficiency of free plasma protein depletion, which ultimately led to a higher purity EV isolate, specifically from the developed SW 700 column. Our goal was to develop an accessible, cost‐effective, high‐throughput, and time‐efficient method that results in EV isolation and protein depletion without a requirement for high‐end instrumentation or specialized technical skills of a researcher.

FIGURE 1.

Initial assessment of CaptoCore for the use of EV isolation. (a) General workflow for isolation and characterization methods of blood plasma‐derived EVs. (b–i) CaptoCore 400 and 700 pore size comparison by SEM imaging. The size distribution varies slightly between individual beads within CaptoCore 400 and 700 (b and f, respectively) with a median bead diameter of ∼90 µm. A single bead of CaptoCore 400 and 700 was individually imaged (c and g). Inset images reveal the pore size differences between the two types of beads (d and h) and is more exemplified by zooming in to the nanoscale (e and i). (j) Protein binding capacity of CaptoCore 400 and 700 measured using a BCA assay. (k) EV recovery rate measured by TRPS, starting material is equivalent to 0.1 µL undiluted plasma (j and k). CaptoCore 700 has a lower EV recovery rate, but a higher protein binding capacity. CaptoCore 400 has a higher EV recovery rate, but a lower protein binding capacity. (l–n) Elution profiles of EVs (black) and free plasma proteins (red) in CL‐2B (l), SW 400 (m), and SW 700 (n). Fractions were collected in 200 µL increments.

2. METHODS

2.1. Plasma pre‐processing

Cell‐ and platelet‐poor anticoagulated with EDTA plasma was collected from self‐declared healthy male donors and pooled together (IRB#2001P000591 (BIDMC) and IRB #17‐12‐14 (NU)). One mL plasma aliquots were stored at −80°C until use. Prior to EV isolation, the plasma was thawed at room temperature and spun at 14,000×g for 10 min (Centrifuge 5424 R, Eppendorf; Hamburg, Germany), and the lower plasma layer was collected; this is referred to as platelet‐poor plasma (PPP) (Figure 1a).

2.2. Size exclusion chromatography column geometry optimization

An empty affinity chromatography (AC) column (Cat. #004212, Biocomma; Shenzhen, China) with a stopper was packed with 2, 3, or 4 mL of Sepharose CL‐2B (CL‐2B) to a bed height of 5, 6, or 7 cm, respectively. A hydrophilic filter, 8.3 mm diameter, 1.6 mm length, 50 µm pore size (Cat. #ACF‐083‐16‐50‐1, Biocomma) was placed into the column before loading CL‐2B, and placed on top of the CL‐2B in some column formats. The column was rinsed with two column volumes (4, 6, or 8 mL, respectively) of 0.22 µm filtered 1× PBS. 100 µL of the sample (50 µL PPP mixed with 50 µL PBS) was loaded onto the column and eluted with 1× PBS. Using an Automatic Fraction Collector (AFC, iZon Science Ltd.; Christchurch, New Zealand), 200 µL fraction collection began (30 fractions total) once the sample was loaded. Fractions were used for BCA and TRPS analysis. EV‐containing fractions were pooled together and subsequently analyzed as one sample, unless stated fractions were analyzed.

2.3. Suspension mode optimization

2.3.1. Binding kinetic testing

Three hundred microliters of CaptoCore 700 resin were washed three times with 1 mL of 0.22 µm filtered (Cat. #16534, Sartorius; Goettingen, Germany) 1× PBS by spinning at 300×g for 2 min. The supernatant was removed after each wash. Three sets of protein standards (1:1 mixtures of IgG and BSA at 2, 6, and 10 mg of total protein. e.g., 1 mg IgG:1 mg BSA, etc.) were incubated with 300 µL of CaptoCore 700 resin for up to 60 min, while the supernatant was collected every 5 min after spinning at 300×g for 2 min, for protein concentration measurements (Figure S1A–S1B). Additionally, 1 mL of plasma sample (100 µL platelet‐poor plasma, 900 µL 1× PBS) was added to the washed CaptoCore 700 beads, and placed on a rotating mixer for 60 min at room temperature. A 10 µL sample was removed after every 5 min of mixing after spinning at 300×g for 2 min and used for protein quantification by BCA and separation by SDS‐PAGE (Figures S1C–S1E).

2.4. Binding capacity in suspension mode

0.1, 0.3, 0.5, and 0.75 mL of CaptoCore 400 (Figure S2) and 0.1, 0.3, and 0.5 mL of CaptoCore 700 (Figure S3) beads were washed, as stated previously. Protein standards (IgG and BSA at a ratio of 1:1 IgG to BSA for total protein amounts of 2, 6, and 10 mg) were mixed and incubated with each volume of both CaptoCore resins. After incubation, the suspension was spun down, and the residual protein concentration in the supernatant was measured by BCA. Additionally, 1 mL of plasma sample (100 µL PPP, 900 µL 1× PBS) was added to each, and tubes were placed on a rotating mixer. After 30 min, samples were spun at 300×g for 2 min, and the supernatant was removed for BCA and SDS‐PAGE.

2.5. Optimized EV isolation by suspension mode

One hundred microliters PPP was diluted by 1:10 with 1× PBS before incubating with either 0.75 mL CaptoCore 400 or 0.5 mL CaptoCore 700 resin on a mix‐rotator at room temperature for 30 min. The mixture was then spun down at 300×g for 2 min, and the supernatant was carefully collected. The supernatant was concentrated down to ∼30 µL using a centrifugal filter (MWCO 30 kDa, Amicon, United States).

2.6. Sandwich column mode optimization

An empty AC column (Biocomma; Shenzhen, China) was packed with Sepharose CL‐2B resin with either 0.75 mL of CaptoCore 400 or 0.5 mL of CaptoCore 700 on top of the Sepharose CL‐2B. A filter (Biocomma; Shenzhen, China) was placed between the two resin layers in some column formats. The column was rinsed with two column volumes of 0.22 µm filtered 1× PBS. After rinsing, a stopper was placed onto the column outlet, and 100 µL of the sample (50 µL PPP mixed with 50 µL 1× PBS) was loaded on the top of the column. The PPP was allowed to incubate for 10 min with the CaptoCore 400 or CaptoCore 700 beads, the stopper was opened, and fractions were collected in the same manner as the SEC column optimization.

2.7. Optimized EV isolation by sandwich mode

One hundred microliters PPP was diluted by 1:1.5 with 1× PBS and then loaded onto a SW column with 7 cm CL‐2B and an additional 0.75 mL CaptoCore 400 or 0.5 mL CaptoCore 700 resin on the top. Elution started after 10 min of incubation was allowed. After the void volume, EV‐containing fractions of 200 µL each were collected by an AFC. EV‐containing fractions were pooled together and subsequently analyzed as one sample, unless stated that fractions were analyzed separately.

2.8. EV isolation by two‐step ultracentrifugation

Platelet‐poor plasma was diluted in 1× PBS and ultracentrifuged using a SW 41Ti swinging bucket rotor at 100,000×g for 2 h at 4°C on a Beckman Optima XL‐90 ultracentrifuge. The pellet was resuspended in 1× PBS. A subsequent dilution was performed in the ultracentrifuge tubes by adding 1× PBS, and it was centrifuged again at 100,000×g for 2 h at 4°C (Thery et al., 2018). The resulting pellet was resuspended in 200 mL of 1× PBS and then stored at −80°C for further experiments.

3. RESULTS

3.1. CaptoCore stationary phases deemed suitable for EV enrichment

First, we tested the CaptoCore beads for their binding kinetics and binding capacity properties using both protein standards and platelet‐poor plasma to understand what the optimal conditions for protein depletion and EV enrichment would be. The volume of beads in suspension and the geometry of the bed packed into cartridge‐format columns, as well as other operational conditions (i.e., washes, equilibration, EV isolation, incubation time), were optimized as discussed in the supporting information and shown in Figures S1–S3). The morphology of the beads was assessed by SEM imaging to understand bead diameter, shape, and pore size; CaptoCore 400 shown in Figures 1(b)–(e), and CaptoCore 700 shown in Figures 1(f)–(i). Images of the beads demonstrate consistent bead and pore size distributions across all images with no visible mechanical damage or aberrations. The larger pore size of the CaptoCore 700 (Figure 1i) media compared to CaptoCore 400 (Figure 1e) is evident from the larger pore networks. To understand the effect pore size would have on EV isolation, we evaluated how plasma treatment with CaptoCore beads alone in a tube (suspension mode; Sus) compared to plasma treatment with the CaptoCore stationary phase packed in a cartridge‐type column format (without any other stationary phases) to comparatively assess the performance of each CaptoCore‐based mode in EV purification from plasma proteins and protein complexes. We found that operation in suspension mode, compared to operation in a cartridge‐type column format, resulted in a higher binding capacity and continued to use suspension mode for further experiments (Figure S4).

3.2. Suspension mode configuration for EV isolation

As we determined, CaptoCore resin is a promising dual‐mode chromatographic medium for EV enrichment and protein depletion. In parallel, 0.75 mL CaptoCore 400 and 0.50 mL of CaptoCore 700 were incubated in suspension mode with platelet‐poor plasma. The volume ratios for the bead suspension to platelet‐poor plasma were optimized prior to further experiments (Figures S2 and S3). We could not detect leftover free plasma proteins by BCA in the supernatants of both Sus 400‐ and Sus 700‐treated samples (Figure 1j), indicating the protein concentration was below the limit of detection, with the use of untreated plasma as a control. For a more thorough analysis of free plasma proteins, other analytical techniques (western blot and proteomic analysis) were employed. By TRPS measurements, the EV‐like particle recovery rate in Sus 400 demonstrated a recovery rate of 80.1 ± 4% EVs recovered from the resin, whereas Sus 700 had an EV‐like particle recovery rate of 23.3 ± 1%, where more than 70% were presumably trapped inside the pores of the CaptoCore beads (Figure 1k). To understand the residual protein in each set of Sus‐based EV isolation experiments, we used the more sensitive method of SDS‐PAGE and visualized by silver stain (Figure S5). The EV isolates prepared using CaptoCore 400 resin demonstrated a higher amount of residual plasma protein when compared to the EV‐enriched samples prepared using the CaptoCore 700 resin, indicated with a red oval (Figure S5). These initial results acquired using protein concentration measurements, TRPS and SDS‐PAGE indicate that based on the difference in the pore size, the CaptoCore 400 and CaptoCore 700 stationary phases will provide a different yield and purity in EVs and NVEPs; however, both can be used in suspension mode for EV enrichment from platelet‐poor plasma.

3.3. Sandwich column configuration for multimode chromatography‐based EV isolation

The separation efficiency of the sandwich column format for both CaptoCore 400 (SW 400) and CaptoCore 700 (SW 700) fractions was assessed in the same way as for the SEC column fractions (Supporting Information) using TRPS, BCA assay, and silver‐stained SDS‐PAGE (Figure S6). For the initial optimization experiments, 3 mL or 4 mL Sepharose CL‐2B was packed into a column and then topped by a layer of either 0.75 mL CaptoCore 400 (Figure S6A–B) or 0.5 mL CaptoCore 700 (Figure S6C–D) resin. Based on the binding kinetic results shown in Figures S1C–D, 10 min incubation was used before elution and fraction collection. EVs were mainly enriched in fractions 6–9 in the SW 400 column (Figures S6A–B), while in the SW 700 column, the EV peak was mainly in fractions 6–8 (Figures S6C–D). The recovery rates of EVs operated in SW mode were calculated by TRPS measurements (Figures 1i–n), finding that the recovery of particles was 56.9 ± 8% in SW 400 and 17.7 ± 8% in SW 700. Interestingly, the protein peaks were lower in using CaptoCore 400, possibly due to CaptoCore 700 trapping more lipoprotein particles, shown by western blot and proteomic data, allowing for more free plasma proteins to be detected by BCA. When comparing the sandwich columns to a standard CL‐2B column (Figure 1l–n), the protein peaks in the sandwich columns resulted in lower protein concentrations with smaller areas of the protein peak, indicting higher rates of protein depletion. The addition of a CaptoCore layer also resulted in a higher resolution between the EV and protein peaks and, therefore, in an increased purity of EV isolates compared to the EV enrichment using Sepharose CL‐2B alone. The efficient purification of the EV‐rich fractions from free plasma protein was further confirmed for both SW 400 and SW 700 columns by SDS‐PAGE (Figure S6E–F).

3.4. Assessment of EV isolation methods by intact EV and protein analysis

To assess intact EVs, TEM imaging was used to visually confirm the presence of EVs in each of the EV isolates as well as to compare the apparent size and morphology of EVs isolated by each of the evaluated techniques (Figure 2a). Through TEM analysis, all examined techniques resulted in successful EV isolation, which was evident from the detection of classically shaped EV‐like structures of similar size and morphology that are characteristic of EVs. Based on the TEM imaging, the apparent EV sizes are similar for the evaluated isolation techniques, which can be seen by the inset images and by TRPS data (Figures 2a and S6). As a negative control, dPBS was assessed, with no EV‐like features detected (Figure S7).

FIGURE 2.

Intact EV morphology and size distribution assessed by TEM imaging and nFCM. (a) EVs from all isolation methods show cupping, a classic type of EV morphology from the sample preparation process. Their size can be approximated to be <200 nm in diameter. Scale bars at the top right of images correlate to the images and scale bars at the bottom left of images correlate to the inset panels. (b) nFCM was employed to determine particle counts across a 30‐s acquisition. Particle counts correlate to expected concentrations for each isolation method, with Sus 400 generating the most counts within the gated area over a 30 s acquisition time and a UC‐generated EV isolate displaying the least amount of counts within the same gated area over the same acquisition time. (c) Median particle size distributions were determined for the same gated area across the same acquisition time, showing similar median sizes across all methods. CaptoCore 700‐generated samples have a slight increase in their size distributions, presumably due to smaller particles becoming trapped inside the larger pores. UC also has a slight increase in determined size, presumable to due the loss of smaller particles during the isolation process. (d) A histogram of all isolation methods illustrating how comparable each method is to each other.

Furthermore, we characterized the performance of EV enrichment using side scatter measurements from nanoflow cytometry (Figure 2b). Gating strategies are shown in Figure S8. Particle counts per acquisition time correlate to the approximate concentration of EVs and other nanoparticles within the samples (Figure 2b). An expected trend is observed where Sus 400 has the highest counts out of the CaptoCore‐based samples, followed by Sus 700, and the sandwich columns. Interestingly, the SW 400 and SW 700 columns are similar in their counts per acquisition time, within error. The measured size distributions show most EVs being isolated are <200 nm in diameter, which was also observed with TRPS measurements (Figure S9). On average, SEC shows the lowest median size distribution with 125.3 ± 0.6 nm measured between replicates. The use of CaptoCore in suspension mode shifts the population slightly higher, with the median size of Sus 400 and Sus 700 measured to be 126.0 ± 0.0 nm and 127.7 ± 1.5 nm, respectively. Using CaptoCore in sandwich mode shifts the population higher (compared to SEC) as well, with the median size of SW 400 and SW 700 measured to be 127.3 ± 0.6 nm and 131.3 ± 4.0 nm, respectively. The UC‐derived isolate showed a lower EV signal (counts per acquisition) with larger median particle sizes of 131.7 ± 7.8 nm. However, an overlaid histogram of all isolation methods shows the overall similarity of each technique, each isolating EVs in the exosome size range (Figure 2d). By comparing the nFCM to TRPS size histograms (Figure S9), they are similar and complementary to each other, with the majority of particles ranging between 100 and 200 nm in diameter, indicating CaptoCore is suitable for EV enrichment. Scattering cross sections are shown in Figure S10.

3.5. Immunoaffinity‐based assays confirm EV‐characteristic molecular features by interferometry and western blotting

Immunoaffinity‐based interferometry is a light‐based analytical technique that has been shown to be useful for nanoparticle analysis, based on a Single Particle Interferometric Reflectance Imaging Sensor to count and image fluorescently stained EVs on a chip (Daaboul et al., 2016). We employed this technique to qualitatively and quantitively assess EV enrichment and to comparatively phenotype EV isolates based on EV‐characteristic protein markers. Quantitively, EV isolates generated using the developed and conventional EV purification approaches were fluorescently tagged and counted for three separate commonly probed EV protein membrane markers: CD9, CD63, and CD81 (Figure 3a). Three fluorescent channels were used for the three EV proteins of interest: CD9 – blue, CD63 – red, CD81 – green. Each isolation method showed the detection of EVs, with a clear pattern emerging. The use of both Sus 400 and Sus 700 increases EV recovery compared to a CL‐2B SEC column, which is complementary to intact EV nFCM data, within error (Figure 2b). These immunoaffinity‐based interferometry data showed Sus 400 having the greatest yield with 82.5 ± 3.0% EV‐like particles recovered followed by 30.2 ± 17.1% EV‐like particles recovered in Sus 700 and with the SW 400 column, CL‐2B column, and SW 700 column following (Figure 3a). Similarly, the Sus‐based recovery rates obtained by TRPS measurements (Figure 1k) observed the same trend of Sus 400 having a higher yield than Sus 700. The use of SW columns decreases the EV recovery rate, but upon further analysis, increases the purity of the EV isolation relative to Sus‐based samples. Interestingly, Sepharose CL‐2B columns resulted in only 9.0 ± 18.7% recovery, and the UC‐derived EV isolate resulted in 42.6 ± 7.1% recovery, relative to total plasma, with dPBS used as a negative control. Qualitatively, interferometry was used to image the samples based on the detection of the above markers (Figure 3b). Due to the possibility that an EV can have more than one of these proteins on its surface, multiple tags can be observed in each channel (i.e., red and green tags can be seen in the blue CD9 channel, etc.). Interestingly, the gold standard technique for EV isolation based on UC appeared to consistently result in aggregation and clustering of EVs, and possibly proteins and lipoprotein particles, shown as large purple clumps. Representative replicate images are shown in Figure S11.

FIGURE 3.

Immunoaffinity‐based assays were employed to determine intact particle concentration as well as in the characterization of specific protein markers. (a) Immunoaffinity‐based interferometry was used to determine particle concentration, demonstrating a similar trend to nFCM, with the Sus 400‐generated sample yielding the highest concentration of tagged and counted particles. An inset insert illustrates that the SW 700‐generated EV isolate has a lower concentration by ∼1–2 orders of magnitude than other isolation methods; however, it has more detectable particles by one order of magnitude than the negative control, dPBS. (b) Qualitatively, immunoaffinity‐based interferometry also provided images of the fluorescently tagged particles. Noticeably, UC‐generated EVs aggregating together, whereas all other isolation methods do not. (c) Western blot against proteins of interest to determine if plasma depletion was successful while retaining EVs. Apolipoprotein A1 and IgG heavy chain show the greatest depletion, with little to no bands detected. Albumin and apolipoprotein B‐100 have detectable bands; however, CaptoCore‐based methods show weaker bands compared to a conventional SEC‐generated EV isolate. With CD9 being detectable in each isolation method and plasma proteins decreasing, this lowers the dynamic range between EV‐related proteins and plasma proteins. (d) Silver stain SDS‐PAGE demonstrates total protein amount in each isolation method, indicating CaptoCore‐based samples, especially CaptoCore 700‐based samples, have lower total protein amounts. Sample loading volumes for both c and d are as follows from left to right: 1 µL SEC, 10 µL SEC, 1 µL Sus 400, 1 µL Sus 700, 10 µL SW 400, 10 µL SW 700, 0.01 µL plasma.

Next, to evaluate the developed methods, the EV isolates were assessed by western blotting against CD9 and the highly abundant plasma proteins, serum albumin, IgG, apolipoprotein A1, and apolipoprotein B100, that are known to co‐isolate with EVs, especially by UC (Figure 3c). Western blotting showed successful depletion of these highly abundant plasma proteins across all isolation methods, relative to platelet‐poor plasma. One microliters plasma equivalent for Sus‐based samples was used to not overwhelm the signal, and 10 µL plasma equivalent for SW‐based samples was used to have enough detectable signal. For comparison, 1 and 10 µL plasma equivalents of an SEC‐based sample were used in the first two lanes of the western blot, respectively. Noticeably, a trend appears where all protein bands appear weaker in both CaptoCore 700‐based samples compared to CaptoCore 400 samples for each respective isolation mode. However, when comparing the 1 µL Sus‐based samples to 1 µL SEC sample, the CD9 band is more intense with similar intensities of plasma proteins which indicates a higher concentration of EVs in Sus‐based samples than in SEC, complimenting nFCM and TRPS data. Comparing the 10 µL SW‐based samples to the 10 µL SEC sample, the plasma proteins appear weaker in the SW‐based samples with CD9 to have a comparable intensity, with a slight loss in the SW 700 sample. This indicates lower free plasma protein contamination with similar EV concentrations. These trends can also be observed by SDS‐PAGE, where the total protein content is lower in CaptoCore 700 samples relative to their CaptoCore 400 counterparts (Figure 3d). Moreover, similar to the western blot, the CaptoCore‐based methods show better free plasma protein depletion efficiency compared to their equivalent volumes of the SEC‐based EV isolation method overall. However, due to CaptoCore 400 having a pore size of 19 ± 1 nm, its ability to deplete LDL (∼22 – 27 nm) (Kreimer et al., 2015) and subsequently apolipoprotein B‐100 is not as efficient in suspension mode (Figure 3c). When used in tandem with Sepharose CL‐2B as a SW 400 column, the efficiency of depleting apolipoprotein B‐100 increases.

3.6. Proteomic analysis indicates depletion of highly abundant plasma proteins by use of CaptoCore

NanoLC‐MS/MS‐based bottom‐up proteomic profiling was performed across all samples to identify and quantify proteins recovered by each EV isolation method. The abundance levels of the most predominant plasma proteins (serum albumin, IgG, etc.) (Pieper et al., 2003) and selected EV‐characteristic proteins (CD9, annexin A2, etc.) using guidelines from MISEV2018 and other reports (Dovrat et al., 2014; Hoshino et al., 2020; Kugeratski et al., 2021; Thery et al., 2018) were compared for their relative abundances (Figures 4a–d). An export of all identifications along with their abundances is provided in Supporting Information. To understand the similarities between methods, unsupervised hierarchical clustering conducted on the quantitative proteomic data derived from the different modes of EV isolation demonstrated a substantial similarity between all generated proteomic EV profiles in general, where SW‐based samples cluster together and show a significant degree of similarity with the UC‐based isolate (Figure 4a). Sus‐based samples also cluster together and the SEC‐based isolate showed the highest degree of difference from the rest of the samples. In general, the modes of isolation perform similarly to each other, where there are slight differences in isolation of EVs and protein depletion depending on the resin and mode of isolation used. Zooming into the topmost clusters, the more abundant plasma proteins (apolipoprotein A‐1, haptoglobin, IgG, etc.) are reliably quantified without causing major interference with the detection and quantitation of lower abundance proteins (Figure 4ai). The UC‐ and SW 700‐based techniques resulted in the higher depletion efficiency for the highest abundance free plasma proteins according to the measured protein abundance levels. After sample processing by the assessed EV isolation techniques, the dynamic range of plasma and EV proteins’ abundance levels decreases, allowing lower abundance EV proteins to be reliably detected and quantitated in another cluster (Figure 4aii). EV‐related proteins such as 14‐3‐3 zeta/delta, CD9, and GAPDH naturally cluster together, indicating similarities among these proteins and, therefore, similarities among the EVs when they are isolated, as opposed to randomly being dispersed within the hierarchical clustering (Figure 4a).

FIGURE 4.

Proteomic analysis of human blood plasma‐derived EV‐related proteins and associated plasma proteins. (a) A heat map showing all proteins detected during bottom‐up mass spectrometry analysis for each isolation method. AI: An inset of the topmost cluster shows highly abundant plasma proteins clustered together. AII: An inset cluster of EV‐related proteins. 14‐3‐3 zeta/delta, CD9, and GAPDH all cluster together. (b) Extracted heat map of EV‐related proteins. (c) Extracted heat map of all apolipoproteins detected. (d) Extracted heat map of top plasma proteins. All scales are Log₂ Abundance. (e) Venn Diagram comparing the number of identified proteins of traditional‐based isolation methods (SEC and UC) to CaptoCore‐based Sus isolation methods. (f) Venn Diagram comparing the number of identified proteins of traditional‐based isolation methods (SEC and UC) to CaptoCore‐based SW isolation methods. (g) Venn Diagram comparing the number of identified proteins of each CaptoCore‐based method, Sus compared to SW. (h) A PCA plot illustrating that UC‐based EV isolates cluster further away from all other methods. (i) GO Term analysis of selected terms of interest showing the percentage of identified proteins within each method.

Several EV‐related proteins were comparatively examined using the same hierarchical clustering approach based on their measured abundances, as well as a negative EV‐marker, calnexin (Figure 4b) (Thery et al., 2018). Overall, all evaluated EV isolation techniques demonstrated largely similar patterns in protein abundances. However, each method showed specific differences in the abundance of the EV‐related proteins of interest. Specifically, the SEC and Sus 400 samples showed the highest correlation in clustering, closely followed by SW 700 and UC preparations thereafter. Sus 700 and SW 400 clustered further away, presumably due to their lower recovery of ALIX, a cytosolic protein involved in multivesicular body formation through the endosomal sorting complex required for transport (ESCRT) pathway (Henne et al., 2011). Additionally, each isolation method does not recover all EV‐related proteins equally. For example, UC‐based isolation has a ∼10‐fold decrease in abundance in annexin A2, a membrane‐binding protein involved in Ca²⁺ regulation, compared to Sus 400. While UC did recover 14‐3‐3 zeta/delta, CD9, and GAPDH at higher abundances compared to other methods, the CaptoCore‐based methods demonstrated the ability to also recover these proteins. ALIX, flotillin‐1, cofilin‐1, and syntenin‐1 exhibited the most different trends from other EV‐characteristic proteins between the tested methods, which may indicate that different EV isolation techniques have certain preferences and biases in the recovery of the specific EV and NVEP types depending on where the proteins are localized within the cell at the time of EV formation (Figure 4b). Overall, the UC‐, SEC‐, and Sus 400‐based isolation methods showed slightly higher recoveries of EV‐related proteins compared to the others, which agrees with our nFCM and TRPS data (Figure 2b and Figure S9, respectively). Additionally, calnexin, an endoplasmic reticulum protein that serves as a negative EV‐marker (Bonsergent et al., 2021), was found in highest abundance in the UC preparation relative to the SEC and CaptoCore preparations.

By assessing the quantitative profiles of apolipoproteins (Figure 4c), different recovery patterns were observed than EV‐related proteins. The SEC, Sus 400, and SW 400 methods demonstrated the lowest efficiency in lipoprotein depletion, on average, and particularly, for apolipoproteins A‐1 and B‐100. The SW 700 column‐ and UC‐based techniques showed the highest depletion efficiencies for apolipoproteins, specifically, this is exemplified for the apolipoproteins with lower molecular mass (i.e., apolipoproteins A‐II, C‐I, M, etc.), where more of these particles can fit inside the CaptoCore 700 pores. Furthermore, SW 700 has, on average, ∼60‐fold decrease in abundance in apolipoprotein A‐1 compared to the standard CL‐2B SEC column and overall ∼16‐fold lower levels of all apolipoproteins compared to CL‐2B SEC column, which indicates higher depletion efficiency for apolipoproteins with the combined use of CL‐2B SEC and CaptoCore 700 resin. Similarly, the recovery rates based on abundance levels of the representative most abundant plasma proteins were also examined (Figure 4d). Overall, similar to the trend observed for lipoprotein depletion, the UC and SW 700 methods demonstrated significantly higher depletion efficiency for top plasma proteins (e.g., serum albumin, IgG, IgA, haptoglobin, serotransferrin, and alpha‐1‐antitrypsin) compared to the standard SEC CL‐2B column and other CaptoCore‐based methods. Serotransferrin, for example, showed a 25‐fold decrease in abundance in the SW 700 column compared to the SEC‐based approach. Again, this can be attributed to the combined action of the multimode bilayer SW 700 columns, where the free plasma proteins and protein complexes are becoming efficiently trapped inside the CaptoCore beads, and the remaining species are better separated in the SEC dimension during processing because of the lower dynamic range of the non‐EV species. It was also observed that the use of CaptoCore alone (both Sus 400 and Sus 700) does not deplete apolipoproteins and free plasma proteins as well as in a column format (i.e., Sus 400 has ∼10‐fold higher serum albumin compared to its SW 400 counterpart and Sus 700 has ∼7‐fold higher serum albumin compared to SW 700) (Figure 4d). This trend is observed in almost all free plasma proteins as well as in apolipoproteins, where the use of CaptoCore in a column format has higher depletion rates than in suspension mode. These data reflect the difference in the pore size of CaptoCore beads and complement our hypothesis that the CaptoCore 700 depletes more free plasma proteins (e.g., serum albumin, IgG, serotransferrin) and highly abundant lipoprotein particles of various types compared to the other evaluated CaptoCore‐based formats. Specifically, the SW 700 column also enriches EVs, based on abundance levels of EV‐associated proteins (e.g., CD9, annexin A2, 14‐3‐3 zeta/delta) to a similar efficiency to UC‐based samples and at a higher efficiency to standard SEC columns (Figure 4b).

Next, we assessed the similarities and differences in the total number of proteins quantified between each isolation method using Venn diagrams (Figure 4e–g). While many of the quantified proteins are expectedly shared between isolation methods, the relative abundances were shown to be different, as discussed above (Figure 4a–d). Comparing conventional EV‐isolation techniques to the Sus‐generated samples, the majority of the proteins are shared between these methods (Figure 4e). Noticeably, 141 proteins are unique to UC, many of which are not used as EV‐markers, and some are plasma‐associated proteins (e.g., immunoglobulins). A similar outcome occurs when comparing the conventional techniques to the SW‐based EV isolation methods; the UC‐generated isolates resulted in the detection of 145 unique proteins, many of which are not EV‐markers (Thery et al., 2018) (Figure 4f). In both Figure 4(e, f), the CaptoCore samples have a small number of proteins that are unique and specific to them, indicating a significant similarity of the compared EV isolation techniques based on the proteomic profiles generated using both conventional (SEC and UC) and CaptoCore‐based techniques. There also appears to be a significant overlap in profiled proteins when comparing all CaptoCore‐based EV isolation techniques, with the majority of the proteins shared between methods (Figure 4g). Moreover, these trends in protein identifications and abundances were reproducible across experiments. Overall, taking into account the number of proteins identified by each method, what proteins they are, and their abundances, our comparative analysis indicates that there is a similar efficiency of plasma protein depletion and EV purity in the UC‐ and SEC‐generated samples compared to CaptoCore‐based EV isolates. Furthermore, a principal component analysis (PCA) plot highlights the differences and similarities between the evaluated EV isolation techniques based on the results of qualitative and quantitative proteomic profiling of the EV isolates (Figure 4h). While all CaptoCore isolates exhibit similarities by clustering together, UC‐derived isolates resulted in a distant cluster shifted away from the rest based on both PC1 and PC2, indicating the types of proteins and protein amounts that are being isolated by the UC‐based approach are significantly different from the isolates purified using the SEC‐ or CaptoCore‐based methods. Ovals indicate the 95% confidence interval.

Gene Ontology (GO) term enrichment analysis shows the percentage of enrichment for several cellular components of interest compared to the total proteins identified (Figure 4i). Every isolation method yields the highest percentage of proteins that are associated with EVs, followed by the plasma membrane, cytosol, and extracellular region. Many proteins associated with EVs originate from the cytosol or the plasma membrane (e.g., CD9 is a membrane protein as well as an EV‐related protein) and will contribute to both GO Terms. As a reference, all isolation methods were compared to the Human Plasma PeptideAtlas (Schwenk et al., 2017) (built in July 2021), which contains ∼4400 canonical proteins. Expectedly, the use of CaptoCore yields similar percentages of EV‐related proteins compared to the SEC‐ and UC‐generated samples, indicating the developed methods are optimized for EV isolation. Plasma protein terms (IgG complex, HDL, LDL, VLDL) are included to illustrate the percentage of plasma proteins identified in each sample is minimal (< 10%).

4. DISCUSSION

Several straightforward, accessible, reproducible, scalable, and efficient multimode chromatography‐based methods have been developed for the effective depletion of plasma proteins and lipoprotein particles to isolate EVs and NVEPs from plasma at high purity as an alternative to the current gold standard and conventional EV enrichment techniques. First we optimized a SEC column using Sepharose CL‐2B in at various lengths (Figure S12). Then, by using two types of CaptoCore resins in two different modes, alone in suspension (Sus) or packed in a multimodal column format (SW), we optimized parameters and procedures for the isolation of EVs from plasma. The proposed multimode chromatography‐based techniques offer cost‐efficiency, scalability, and higher throughput for EV isolation compared to the UC‐based approaches. This advantage is especially pertinent for potential clinical applications, where UC‐based EV isolation has a low throughput, requires expensive equipment, and can cause aggregation of EVs, proteins, and lipoprotein particles during the isolation process (Brennan et al., 2020; Momen‐Heravi et al., 2012). Moreover, the use of CaptoCore resin instead of a Sepharose CL‐2B SEC column can offer higher or similar EV recovery rates, depending on the mode of separation, measured by TRPS and interferometry, with higher free plasma protein depletion, assessed by western blot and bottom‐up proteomic data.

To inspect the CaptoCore resin, SEM imaging was employed, and it was confirmed that the beads are nearly spherical in shape, with no visible abreactions or damage (Figures 1b–i). The relative pore size difference between CaptoCore 400 and CaptoCore 700 was also confirmed, with the 700 media having larger pores. This agrees with previously reported work that CaptoCore 400 and 700 have an average bead size of 110 and 91 µm with pore sizes of 19 ± 1 and 50 ± 1 nm, respectively (Mi et al., 2021; Sanchez‐Trasvina et al., 2020). To understand how CaptoCore resin would interact with EVs and plasma proteins in Sus mode, the binding kinetics and binding capacity of the resin were tested (Figures S1–S3). This showed equilibrium was reached after 30 min of incubation and that the maximum binding capacity was determined to be 19.4 ± 0.4 mg plasma protein/mL CaptoCore 400 and 23.75 ± 0.5 mg plasma protein/mL CaptoCore 700. The resins were also tested to determine the depletion of free protein amount by BCA and intact EV recovery rate by TRPS (Figures 1j, k), respectively. This showed no detection of plasma protein in both resins, with EV recovery rates being comparable to the reported values from other commercially available size exclusion methods, having ∼60%–80% recovery and ∼40% recovery for UC‐based samples from plasma (Tian et al., 2020). The differences in EV recovery were expected, given the different MWCOs of the two resins; the larger pore size of CaptoCore 700 would have allowed smaller EVs and EV‐like particles (e.g., exomeres and lipoproteins) into the pores, while the smaller pore size of CaptoCore 400 would exclude larger proteins and protein complexes. However, smaller proteins, such as monomers of serum albumin and IgGs, could become trapped in the CaptoCore 400 pores, contributing to the lower levels of protein detected by BCA. We found that the use of CaptoCore in suspension mode provides the highest yield of EVs, with CaptoCore 400 having the highest EV recovery, followed by Sus 700, SW 400, and SW 700. Interestingly, the Sepharose CL‐2B columns did not show as great of a recovery, highlighting that the use of CaptoCore as a pretreatment prior to separation with Sepharose CL‐2B will increase EV recovery.

Furthermore, TEM imaging was used to visually confirm the presence of EVs by the different isolation types confirming that the EVs that were isolated exhibit visually similar morphology characteristics such as cupping, a common EV characteristic when samples are dried (Raposo et al., 1996; Raposo & Stoorvogel, 2013) (Figure 2a). Once CaptoCore was determined to be suitable for EV enrichment, further intact particle analysis was performed by nFCM and immunoaffinity‐based interferometry (Figures 2b and 3a,b), respectively. Utilizing nFCM, median particle size diameters were obtained, showing each tested technique isolates particles with overall similar diameters. Median particle diameter was obtained first by measuring the side scatter and refractive index (RI) of each PS bead and extrapolating to the measured side scatter and RI of each EV isolate, using methods shown in previous work (Welsh & Jones, 2020; Welsh et al., 2020a, 2020b) (Figure 2c). However, a trend was observed when using CaptoCore, which would lead to the larger median particle size becoming larger. This can be attributed to smaller EVs and NVEPs, apolipoproteins, and other nanoparticles becoming trapped inside the internal positively charged surface of the CaptoCore beads. This causes the shift in the median sizes to become higher in both cases of using CaptoCore 700 in suspension mode and sandwich mode compared to CaptoCore 400. However, the UC‐derived isolate indicated the largest median diameter, presumably from obtaining larger EVs and EV aggregates during the harsh isolation process, which agrees with previously reported size distributions post‐UC processing (Momen‐Heravi et al., 2012). It should be noted that the limit of detection is ∼100 nm for a biological structure in both analysis types of nFCM and TRPS (Figure 2d and Figure S9, respectively), which also hinders the detection of smaller EVs.

Immunoaffinity‐based interferometry demonstrated intact EV particle recovery, with both Sus‐based methods outperformed a conventional CL‐2B SEC column, with Sus 400 recovering more particles than UC‐based isolation (Figure 3a). The SW‐based columns showed a recovery rate similar to a conventional CL‐2B SEC column. Furthermore, fluorescent images revealed evidence of particle aggregates during UC isolation, with all applied techniques showing distinct individual particles (Figure 3b). These images agree with previously reported data that conventional UC‐based isolation techniques can cause EV loss and lipoprotein‐EV combined aggregation with each centrifugation step (Brennan et al., 2020). In contrast, all other evaluated multimode chromatography‐based EV isolation techniques did not show any signs of EV fusion or protein aggregation based on the immuno‐interferometry assays. Obtaining individual particles is needed for several applications where it is critical to preserve EV structure and morphology.

Protein recovery was assessed by qualitative western blotting, SDS‐PAGE, and mass spectrometry. All tested isolation methods indicated depletion by western blots against four of the top most abundant plasma proteins, compared to plasma without EV enrichment: serum albumin, IgG (H), apolipoprotein A1, and apolipoprotein B‐100 (Figure 3c). CD9 was used to probe for EV recovery, with enrichment in all developed techniques, with CaptoCore 400‐based samples yielding higher recoveries compared to their CaptoCore 700 counterparts. SDS‐PAGE complements this as well, with the total protein amount being higher in CaptoCore 400 samples, compared to CaptoCore 700‐based methods (Figure 3d). The mass spectrometry proteomic profiling experiments demonstrated that the abundance levels of all proteins in the EV isolates generated using the SEC CL‐2B‐ and UC‐based techniques were similar to that of the developed multimode chromatography techniques, indicating similar EV purification efficiencies, with slight differences (Figure 4a). Abundance levels were also assessed for specific EV‐related proteins, apolipoproteins, and top plasma proteins, with CaptoCore 400 samples having slightly higher protein abundances compared to CaptoCore 700 samples (Figure 4b–d). Upon further analysis, Venn diagrams confirmed that the developed methods are comparable to each other in the number of proteins quantified, having the majority of the proteins (<600 proteins in all methods) be shared among samples (Figure 4e–g), complementing the heat map of all quantified proteins in Figure 4(a). Using a PCA plot, clustering between samples can be observed using PC 1 and PC 2 (Figure 4h). Interestingly, the UC‐based sample clusters furthest away from all other samples due to the differences in the number of quantified proteins and their relative abundances. GO term analysis shows the majority of proteins identified in each sample are associated with EVs, acting as an indicator that the developed methods are suitable for EV enrichment (Figure 4i).

Altogether, the developed novel chromatography‐based techniques we describe in this work resulted in comparable intact EV enrichment and lipoprotein particle and free plasma protein depletion to the established methods with subtle differences. Through intact particle and protein analysis, we determined that CaptoCore is a suitable media for EV enrichment from plasma, with the methods being more time‐sensitive and assessable than UC‐based isolation and can offer more efficiently purified EV isolates than a standard CL‐2B column. For ease of use, the proposed techniques can be scaled up or down to accommodate various sample volumes of interest and the number of samples to process simultaneously. The choice of the stationary phase (CaptoCore 400 or 700) as well as the format (Sus or SW) used for isolation, can be based upon the requirements and capabilities of the downstream assays, where either high purity or high recovery of EV isolation needs to be prioritized. Moreover, based on the underlying separation mechanisms, the developed techniques can be potentially applicable for the fast and simple isolation of virus and AAV‐based therapeutic particles due to their chemical and structural similarities with EVs.

AUTHOR CONTRIBUTIONS

Alan J. Zimmerman: Data curation; formal analysis; methodology; software; writing—original draft. Getulio Pereira de Oliveira Jr.: Data curation; formal analysis; methodology; writing—review and editing. Xianyi Su: Data curation; investigation; methodology; writing—review and editing. Jacqueline Wood: Data curation; investigation; methodology; writing—review and editing. Zhengxin Fu: Data curation; formal analysis; investigation; methodology. Brandy Pinckney: Data curation; formal analysis. John Tigges: Data curation; formal analysis. Ionita Ghiran: Supervision; writing—review and editing. Alexander R. Ivanov: Conceptualization; funding acquisition; supervision; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Zimmerman, A. J. , de Oliveira, G. P. , Su, X. , Wood, J. , Fu, Z. , Pinckney, B. , Tigges, J. , Ghiran, I. , & Ivanov, A. R. (2024). Multimode chromatography‐based techniques for high purity isolation of extracellular vesicles from human blood plasma. Journal of Extracellular Biology, 3, e147. 10.1002/jex2.147