University of Groningen
Isolation of extracellular vesicles with combined enrichment methods
Stam, Janine; Bartel, Sabine; Bischoff, Rainer; Wolters, Justina C
Published in:
Journal of Chromatography B-Analytical Technologies in the Biomedical and Life Sciences
DOI:
10.1016/j.jchromb.2021.122604
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2021
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Stam, J., Bartel, S., Bischoff, R., & Wolters, J. C. (2021). Isolation of extracellular vesicles with combined
enrichment methods. Journal of Chromatography B-Analytical Technologies in the Biomedical and Life
Sciences, 1169, [122604]. https://doi.org/10.1016/j.jchromb.2021.122604
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Journal of Chromatography B 1169 (2021) 122604
Available online 27 February 2021
1570-0232/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Review
Isolation of extracellular vesicles with combined enrichment methods
Janine Stam
a, Sabine Bartel
b, Rainer Bischoff
a, Justina C. Wolters
c,*aDepartment of Analytical Biochemistry, Groningen Research Institute of Pharmacy, University of Groningen, Groningen, the Netherlands
bDepartment of Pathology and Medical Biology, GRIAC Research Institute, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands cDepartment of Pediatrics, University of Groningen, University Medical Center Groningen, Groningen, the Netherlands
A R T I C L E I N F O Keywords: Exosomes Extracellular vesicles Isolation Method comparison Microvesicles
Systematic literature analysis
A B S T R A C T
Extracellular vesicles (EVs) are currently of tremendous interest in many research disciplines and EVs have potential for development of EV diagnostics or therapeutics. Most well-known single EV isolation methods have their particular advantages and disadvantages in terms of EV purity and EV yield. Combining EV isolation methods provides additional potential to improve the efficacy of both purity and yield.
This review assesses the contribution and efficacy of using combined EV isolation methods by performing a two-step systematic literature analysis from all papers applying EV isolation in the year 2019. This resulted in an overview of the various methods being applied for EV isolations. A second database was generated for all studies within the first database that fairly compared multiple EV isolation methods by determining both EV purity and EV yield after isolation.
From these databases it is shown that the most used EV isolation methods are not per definition the best methods based on EV purity or EV yield, indicating that more factors play a role in the choice which EV isolation method to choose than only the efficacy of the method. From the included studies it is shown that ~60% of all the included EV isolations were performed with combined EV isolation methods. The majority of EV isolations were performed with differential ultracentrifugation alone or in combination with differential ultrafiltration. When efficacy of EV isolation methods was determined in terms of EV purity and EV yield, combined EV isolation methods clearly outperformed single EV isolation methods, regardless of the type of starting material used. A recommended starting point would be the use of size-exclusion chromatography since this method, especially when combined with low-speed centrifugation, resulted in the highest EV purity, while still providing a reasonable EV yield.
1. Introduction
In the last decade, extracellular vesicles (EVs) have been attracting a large scientific interest as indicated by the exponential growth in the number of publications per year (see Fig. 1). EVs are of interest because almost all cells from all organisms can secrete EVs into the extracellular space with a cargo that consists of cell-specific proteins, RNA’s and lipids, although their composition is not necessarily entirely similar to the parental cell, since some components are enriched while others are absent [1].
The term ’EVs’ should be seen as an umbrella term for “all nano-particles that are naturally secreted by cells, encapsulated by a lipid bilayer and unable to replicate by themselves” as defined in 2018 in the minimal information for studies of extracellular vesicles (MISEV)
guidelines [2]. This definition was implemented because many different names have been attributed to EVs after their discovery in the previous century, which has caused a lot of confusion. In the beginning, EVs were often called ’virus-like particles’ because they seemed to behave simi-larly to viruses [3]. In 1981, the term “exosome” was coined and in 1987, this term was used for the first time to address “all cell-secreted vesicles that are derived from multivesicular bodies” [4,5]. Since then, the term ’exosome’ has been heavily used in literature to address EVs. However, the term exosome was not only used for EVs that fulfilled the original criteria for exosomes but allegedly also for other EV subtypes because many papers that used the term ’exosome’ did not show a clear connection between the studied EVs and their biogenesis via multi-vesicular bodies. While other EV subtypes such as microvesicles (ecto-somes) and apoptotic bodies may overlap with exosomes concerning
* Corresponding author at: Department of Pediatrics, University of Groningen, University Medical Center Groningen, Antonius Deusinglaan 1, 9713 AV Groningen, the Netherlands.
E-mail address: j.c.wolters@umcg.nl (J.C. Wolters).
Contents lists available at ScienceDirect
Journal of Chromatography B
journal homepage: www.elsevier.com/locate/jchromb
https://doi.org/10.1016/j.jchromb.2021.122604
Journal of Chromatography B 1169 (2021) 122604
their size (50–150 nm for exosomes versus 50 nm–5 µm for micro-vesicles/ apoptotic bodies) they are considered to be the direct result of an outward budding of the plasma membrane without the intervention of multivesicular bodies. Recent studies also suggest the presence of other smaller EV types, called ’exomeres’ (30–50 nm), the biogenesis of which is unknown so far [6]. To complicate things further, other clas-sification systems have been introduced to address specific EV subtypes. For instance, EVs can be named after their parental cell type (e.g. oncosomes for tumor-derived EVs, prostasomes for prostate epithelial cell-derived EVs), or they can be named after the organism that pro-duced them (e.g. ’outer membrane vesicles’ for gram-negative bacteria- derived EVs, ’membrane vesicles’ for archaea- and gram-positive bac-teria-derived EVs) [7]. Also, EVs may be named after their specific cargo because not all EVs contain the same cargo [8].
The MISEV 2018 guidelines state that EVs can only be addressed by their biogenesis, size, density, biochemical composition or descriptions of conditions or cell of origin when this is clearly demonstrated in the paper. In cases where this is not clear, the term “EVs” should be used [2]. In line with the MISEV 2018 guidelines, this review will be using the term ’EVs’. Other terms will only be used when referring to specific EV subpopulations.
Scientific interest in EVs increased after the publication of two in-dependent studies in 2006 and 2007 that demonstrated that EVs can be specifically taken up by recipient cells and functionally transfer their content to initiate different intracellular effects in the target cell [9,10]. Current EV research can roughly be divided into three different sub-domains; 1) diagnostic EV research, which focuses mainly on specific proteins or RNAs present in EVs that could indicate a certain disease. For example, several EV-derived miRNAs have been proposed as biomarkers for various cancers [11]. Using EVs as a sample matrix for clinical biomarker research may significantly improve clinical care, since EVs can be derived from a patient’s body fluids, while many currently applied diagnostic methods require tissue biopsies obtained by invasive
procedures that cause substantial discomfort to patients. 2) biological
EV research is primarily concerned with expanding knowledge
regarding EV biogenesis, EV behavior in extracellular fluids, molecular mechanisms of sorting molecules into EVs, EV uptake mechanisms by recipient cells and the specific roles of EVs in physiological and patho-logical processes [12,13]. 3) therapeutic EV research is centralized on the use of EVs as part of a therapeutic regimen. A therapeutic effect is pursued by administering natural or slightly modified EVs to patients since EVs can be targeted to specific cell types. Especially in the combat against cancer and in regenerative medicine, EVs seem to have value as therapeutics, even though the field is still in its infancy [11,14].
Despite the attractiveness of EV research and its clear potential, the field still struggles with serious challenges that need to be overcome before EVs can be widely implemented for (clinical) applications. One of the more fundamental challenges is the lack of consensus regarding the isolation steps that should be taken before EVs can be used as proper diagnostics, therapeutics, or biological study material. To improve the comparability of results and to increase reproducibility across different laboratories and applications, it is necessary to define consistent criteria of EV characterization and to standardize the isolation procedures. This has also been recognized by the members of the International Society for Extracellular Vesicles (ISEV), which state that more attention should be paid to the development of appropriate EV isolation methods that have the potential for standardization [15]. However, standardization re-quires consistent analytical criteria, which can be challenging to develop for different biological samples that contain heterogeneous EVs comprised of many different biomolecules in a vesicular form that is inherently not very stable and thereby distinctive from the development of many other diagnostics or therapeutics.
Many different EV isolation methods are available, but until now, none of these methods is able to isolate EVs without the presence of contaminants that may interfere in the analysis. New and more advanced methods are still being developed but lack a wide acceptance
Fig. 1. Bar chart of the number of publications per year after searching for “extracellular vesicles” in the search engine PubMed (search was performed on 27- 05-2020).
and implementation due to specific expertise and high costs. Therefore, this review aims to clarify whether there is scientific evidence based on the current literature that combining several commonly used EV isola-tion methods is more favorable than single EV isolaisola-tion methods.
2. Perspective on EV isolation
2.1. Rationale behind EV isolation
EV isolation is crucial for most downstream applications because most biological samples are quite complex and contain supramolecular assemblies with similar biophysical properties such as lipoproteins and protein aggregates that are often more abundantly present than EVs. Without prior isolation of EVs from those matrices, these contaminants can be easily mistaken for EVs during commonly used EV character-ization methods, resulting in biased conclusions (see 2.2–2.5 for more information regarding common contaminants in various matrices). For instance, when determining EV yield, often, the total particle count is assessed using nanoparticle tracking analysis (NTA), which is a method that tracks individual nanoparticles and derives their size and concentration in suspension with the aid of the Stokes-Einstein equation [16,17]. However, a clear distinction between EVs and similar sized extracellular particles can often not be made. The same is true for other characterization methods such as tunable resistive pulse sensing (TRPS), a technique based on a temporary increase in resistance when a particle passes a submicrometer/micrometer-sized pore embedded in a membrane over which an electric potential is applied. By calibrating the system with particles of a known concentration, a given particle con-centration can be determined by summing up all peaks over a given time period [18]. Other characterization methods often applied to EVs are focused on detecting EV-specific markers and non-EV markers using
bulk immunoassays such as Western blots and/or enzyme-linked
immunosorbent assays (ELISAs) [19]. A way to determine total parti-cle concentration simultaneously with the presence of EV-specific markers is the use of high-resolution flow cytometry/nano-flow
cytometry, which is a method that analyses single particles in a flow
and determines several particle properties simultaneously [20,21]. High-resolution is required since conventional flow cytometry in-struments are often not suitable because the size of EVs falls below the lower detection limit. However, also here, EV isolation before EV characterization is preferable because they need to be analyzed on a single particle basis, which is challenging in a complex and compound- rich biological sample. EVs are also commonly characterized by
elec-tron-microscopy (EM) methods such as transmission EM, scanning EM
or cryo-EM that are able to reveal EV morphology, but can also provide indications of EV size, concentration and purity. The methods are based on electron beams that scan (scanning EM) or cross (transmission EM & cryo-EM) the samples in order to create images [22]. Nevertheless, EV isolation is still desirable, even though EVs can be discriminated from contaminants to some extent, but the workload is reduced with EV enrichment because fewer images have to be taken [23].
2.2. Common contaminants; Lipoproteins
Common contaminants in EV samples are lipoproteins. Lipoproteins (e.g. HDL/LDL) are produced by the liver and intestine and secreted into the bloodstream. They are therefore likely to contaminate blood-derived EV preparations, but cell culture supernatant may also not be free of lipoproteins, especially when culturing liver or intestinal cells/tissue or when fetal bovine serum is part of the cell culture medium. There are different lipoproteins known, and together, their serum concentration is ~1012 particles/mL, which is much higher than the reported concen-tration of EVs in serum (107–109 particles/mL) [24,25]. Chylomicrons are the largest lipoproteins (75–1200 nm) and are produced by the in-testine during a meal and released into the bloodstream via the lymphatic system to transport dietary triglycerides and cholesterol to the liver to be metabolized. Since EVs have a similar size as small chy-lomicrons, it may be challenging to separate them from each other [26] based on size alone. However, the density of EVs (1,10–1,19 g/cm3) is higher than the density of chylomicrons (<0,93 g/cm3) [27], which
Table 1
Overview of the advantages and disadvantages of most common EV isolation methods.
Method Separation principle Advantages Disadvantages
Ultracentrifugation *Density
*Floatation speed (in case of DG)
*Most widely used. *Robust method.
*Possibility to identify subpopulations when DG is applied.
*Heavy workload. *Might lead to ruptured EVs. *Low EV yield.
*Results are difficult to compare between different studies due to different rotor types and centrifugation speeds.
*Large volume of starting material required. Size-exclusion
chromatography *Hydrodynamic radius *Fast and simple. *EV integrity is maintained during isolation.
*May be used in upscaling processes.
*Potentially significant contamination with non-EV particles (e.g. lipoproteins).
Precipitation *Solubility *User-friendly method. *High EV yield.
*Many kits commercially available. *Relatively cheap method. *Possibility for clinical application.
*High risk of co-isolating non-vesicular contaminants.
*Polymers may be problematic for some downstream applications. *Might not be suitable for performing functional studies. Immunocapture *Surface markers *Very pure EV isolations.
*Can be combined with light scattering flow cytometry.
*Capable of isolating subpopulations.
*Currently not suitable for a complete isolation of all EVs. *Expensive.
Ultrafiltration *Hydrodynamic radius *Can be very well combined with other methods.
* Easy to perform in every lab.
*Filter type and ultrafiltration method influence EV yield and composition considerably.
*Rather limited resolution. Field-flow fractionation *Hydrodynamic radius
*Electrophoretic mobility * Can identify smaller EV subsets (e.g. exomeres). *Specific equipment required and the associated expertise. *Cannot be used as a stand-alone method. Microfluidics *Surface markers
*Hydrodynamic radius *Density
*EV isolation can be combined with EV characterization.
*Subpopulations can be isolated. *Small sample volumes can be used for EV isolation.
*Requires a specific level of expertise.
*Not suitable for preparative purposes (e.g. therapeutic applications).
Membrane affinity
Journal of Chromatography B 1169 (2021) 122604
allows separation. Very low-density lipoproteins (VLDL) are pro-duced by the liver and released in the bloodstream to transport tri-glycerides to different organs. They are characterized by a smaller size (30–80 nm) but a higher density (0,930–1,006 g/cm3) than chylomi-crons [26]. During the release of triglycerides, VLDL particles convert to
intermediate-density lipoproteins (IDL) (25–35 nm; 1,006-1,029 g/
cm3) and further to low-density lipoproteins (LDL) (18–25 nm; 1,019–1,063 g/cm3) [26]. The further they become transformed, the smaller their size and the higher their density due to the increase in cholesterol. High-density lipoproteins (HDL) are the smallest (5–12 nm) among the lipoproteins but are also the most heterogeneous and have the highest density (1,063–1,210 g/cm3), which partly overlaps with the density of EVs [26]. Their function is to transport cholesterol back to the liver or tissue lipoproteins [28]. VLDL, IDL and LDL are marked by the abundance of apolipoprotein B100, whereas chylomi-crons are highly enriched with the marker apolipoprotein B48 and HDL can be detected based on the apolipoprotein A1 marker [29,30].
2.3. Common contaminants; Tamm Horsfall protein
Tamm-Horsfall protein (80–90 kD), also called uromodulin, is a glycan-rich glycoprotein abundantly found in the urinary tract where it may reach urine concentrations of 1,5 mg/mL [31,32]. Tamm-Horsfall protein can also be found in serum, but in much lower concentrations (~200 ng/mL) and is, therefore, less problematic in blood-derived EV preparations [33]. It is produced by the ascending loop of Henle in the kidney and it has a plethora of different functions mainly related to the immune system and the counteracting of kidney stone formation. Due to all these features, the Tamm-Horsfall protein may interfere during the EV isolation process, because it may entrap EVs in its filamentous
network [32].
2.4. Common contaminants; Albumin
Physiological serum albumin (~65,5 kD) concentrations range be-tween 35 and 50 g/L [34]. Albumin aggregates may interfere with the measurement of total EV particle concentrations, which means that this protein is an important potential contaminant in blood-derived EV preparations.
2.5. Common contaminants; Casein
In milk, 80% of all soluble proteins are caseins (1,07–2,7 g/L in human milk) [35,36]. This mineral-binding protein forms micelles with a diameter of ~200 nm and a density of 1,078 g/cm3, which partly overlaps with EVs. Therefore, casein should be considered an important contaminant when studying milk-derived EVs [37,38].
3. Overview of standard EV isolation techniques
Through the years, many different techniques have been introduced for EV isolation. This section aims to point out the most-recognized EV isolation methods in detail and touch shortly upon some more recently developed, less frequently applied EV isolation methods. A summary of these methods is provided in table 1.
3.1. Ultracentrifugation
Since the start of EV research, most studies have applied
(differen-tial) ultracentrifugation ((diff)UC) as the isolation method for EVs,
Fig. 2. Methodology of the applied systematic literature analysis for the screening of all papers from 2019 containing the terms ‘extracellular vesicles’ and ‘isolation’. Database 1 was created after application of the first filtering and exclusion steps to investigate the current contribution of combined EV isolation methods. A second database was created selecting only the papers containing a comparison of combined EV isolation methods to single EV isolation method. A distinction was made between the total number of papers and the total number of isolations because some papers described several EV isolation approaches. Detailed criteria regarding inclusion/exclusion of a paper can be found in supplementary material 1. An overview of all the included papers can be found in supplementary material 2. J. Stam et al.
which is still widely used [4,39]. Therefore, (diff)UC has also been considered as the Gold Standard for EV isolation for a long time. In most cases, the protocol proposed by Th´ery et al. is applied, which includes five different centrifugation steps to separate particles based on differ-ences in density [27]. These five steps are: 1) 10 min at 300 g to remove cells, 2) 10 min at 2000 g to remove dead cells, 3) 30 min at 10 000 g to remove large cell debris, 4) 60 min at 100 000 g to pellet the EVs and 5) resuspension and repetition of step 4 to remove contaminating extra-cellular proteins [27]. Sometimes only the first three steps are per-formed (low speed centrifugation (LSC)), in case only large EVs are of
interest [40]. Nevertheless, (diff)UC is a very labor-intensive approach requiring a large amount of starting material compared to other isola-tion methods [41]. A possible reason for this could be that not all EVs sediment during step 5 of the general protocol as demonstrated by Baranyai et al., who repeated the last step in the protocol several times and found that even after four repetitions, EV markers could still be detected in the supernatant [42]. Another disadvantage is that the ob-tained results depend on the type of biological material, the specific type of rotor and the centrifugation time, hampering comparisons between different studies when these characteristics are not properly docu-mented [43,44]. EVs might also be damaged during (diff)UC due to high g-forces, resulting in EV aggregation or EV rupture [45,46]. This last issue has been addressed by using a cushion or a density gradient of iodixanol or sucrose [47]. With the cushioned-ultracentrifugation (CUC), EVs can be separated from proteins based on differences in density, since proteins have a higher density (1,35–1,41 g/cm3) while maintaining EV integrity [27]. With density gradient ultracentrifugation (DG), EVs can be separated based on their floatation speed and equi-librium density, enabling researchers to isolate even specific EV sub-types [48]. However, even though aggregation of EVs and protein contamination can be prevented by using a density cushion or gradient, the workload is considerably augmented, as is the total cost. Therefore, it is implausible that (diff)UC is favored as isolation method in the development of EV diagnostics and EV therapeutics in future.
3.2. Size exclusion chromatography
Although not as widely used as ultracentrifugation, size exclusion
chromatography (SEC), sometimes also called gel filtration
chroma-tography, is gaining popularity according to a recent review of the EV isolation field [49]. The main principle behind SEC is the separation of particles in solution based on their hydrodynamic radius by transporting the initial sample through a column filled with a porous material (polymer- or silica-based), representing the stationary phase. Larger molecules will have less access to the inner pore volume than smaller molecules and, therefore, elute earlier. Several protocols have been developed to apply SEC for EV isolation [50] and some commercial kits are available (e.g. qEV column; (Izon Science). A considerable advan-tage of SEC compared to (diff)UC is that it is easier and faster, especially when commercial kits are used. An additional benefit is that SEC does not affect EV integrity as often occurs with (diff)UC [45]. This method has also potential to be used for upscaling processes when it is applied in the form of size-exclusion liquid chromatography (SELC) [51]. However, some papers show that (diff)UC methods are superior regarding the purity of EVs when using plasma or serum as starting material because more lipoprotein and albumin contamination was observed compared to (diff)UC [39,52]. However, another study that tested plasma and cell culture supernatant concluded that SEC is at least as good as density gradient ultracentrifugation when it comes to EV yield and protein contamination [53].
3.3. Precipitation
Precipitation techniques can also be used for EV isolation because
they separate EVs from other compounds based on solubility. The addition of a reagent such as polyethylene glycol (PEG) with an average molecular weight of 10 kDa is often used for this purpose. PEG increases the number of hydrophobic interactions with EVs and between EVs, which leads to water exclusion and EV pellet formation after incubation and a single low-speed centrifugation step [54]. A plethora of com-mercial kits are currently available, such as the Total Exosome isolation kit (Invitrogen), ExoquickTM (System Biosciences), miRCURRY (QIA-GEN), Exoprep (HansaBioMed), Pure Exo (101 Bio), ExoGAG (Nasa-Biotech), Exosome precipitation solution (Immunostep) and the Total exosome isolation reagent (Thermo Fisher Scientific). Also, commercial kits are available that combine the precipitation principle with other
Table 2
Overview of the different starting materials for EV isolation as used in the 700 included papers in database 1 (see supplementary material 2). The number of papers that described a certain starting material for EV isolation is not neces-sarily equal to the total number of EV isolation approaches for this EV starting material because some papers used the same starting material for multiple EV isolation approaches.
Starting material for EV isolation Number of papers
Cell culture supernatant 422
Serum 125 Urine 34 Tissue 17 Semen 12 Milk 10 Cerebrospinal fluid 9
Bronchoalveolar lavage fluid 7
Parasite conditioned culture medium 5
Ascites 4 Saliva 4 Bile 3 Follicular fluid 3 Mucus 3 Oviductal fluids 3 Pleural exudate 3 Worm-derived supernatant 3 Feces 2 Tears 2
Bone marrow plasma 2
Dried plant material 2
Apoplastic fluid 1
Aqueous humor 1
Cod mucus 1
Cyst fluid 1
Fruit juice 1
Gingival crevicular fluid 1
Ginseng roots 1
Helminth cell culture supernatant 1
House dust mite supernatant 1
Hydatid cyst fluid 1
Micro-organisms supernatant 1
Oral swirls 1
Outgrowth embryo conditioned medium 1
Pancreatic juice 1
Peritoneal dialysis effluent 1
Placenta extract 1 Placental perfusate 1 Platelet concentrate 1 Platelet supernatant 1 Synovial fluid 1 Uterine flushings 1
Vaginal luminal fluid 1
Yeast suspension 1
Amniotic fluid 1
Cavitron ultrasonic surgical aspirator derived tissue 1 Fetal central nervous system sources 1 Flushed bone marrow interstitial fluid from femur 1
Gastric juice 1
Lung perfusate 1
Peritoneal lavage 1
Protoscolex culture supernatant 1
Uterine flushings 1
Cervical cytobrushes 1
Donor kidney perfusion fluid 1
Journal of Chromatography B 1169 (2021) 122604
methods like SEC; EXO-spin (CELL guidance systems). Precipitation techniques are considered user-friendly, relatively cheap and not very time-consuming and they seem to result in a higher EV yield than (diff) UC. These features make them more suitable for upscaling processes than (diff)UC in the development of EV diagnostics and EV therapeutics in the future [55]. However, it appears to be quite challenging to separate the PEG polymers from the EVs afterward, which is problematic for functional or therapeutic down-stream applications. Also, many other proteins pellet as well [56]. Therefore, it could be argued that many studies have overestimated their EV yield with precipitation methods based on the total protein content. Further, several studies showed that PEG-isolated EVs resulted in reduced cell viability when performing functional EV studies in contrast to EVs isolated by (diff)UC,
indicating that some co-isolated components may be toxic [57,58].
3.4. Immunocapturing
Immunocapture approaches result in the highest EV purity
compared to other isolation methods such as (diff)UC. Although indi-vidual procedures may differ, the main principle behind these methods is the binding between an EV membrane protein marker and a specific antibody. After washing the other sample components away, only EVs containing the target protein will remain. However, success depends on the specificity of the antibody and the absence of non-specific in-teractions. There are currently several ways to perform immunocapture, such as using microtiter plate-based techniques (only for minimal
Fig. 3. Distribution between single EV isolation methods and combined EV isolation methods for all of the 892 different EV isolation approaches described in the 700 papers included in database 1. The sum of the number of different EV isolation approaches (between brackets) subdivided per starting material is not necessarily equal to the total number of different EV isolation approaches because some papers used the same isolation approach for several different biological materials.
Abbreviations: BALF: bronchoalveolar lavage fluid, CCS: conditioned cell culture medium, CSF: cerebrospinal fluid, Other: the sum of other biological materials that were used in fewer than seven papers.
Fig. 4. Overview of EV isolation methods that were Top10 most frequently applied in the year 2019 for cell culture supernatant (A) and plasma/serum (B).
Ab-breviations: (diff)UC: differential ultracentrifugation, CCS: cell culture supernatant, DG: density gradient (ultra)centrifugation, EQ: Exoquick (precipitation), EXE: exoEASY (membrane affinity), LSC: low-speed centrifugation (<80 000 g), MC: miRCURY (precipitation), PEG: polyethylene glycol (precipitation), SEC: size-exclusion chromatog-raphy, TEI: Total exosome isolation kit (precipitation), UF: ultrafiltration.
amounts) affinity columns or magnetic beads [59,60]. An example of a relatively advanced method among the immunocapture approaches is
fluorescence-activated cell sorting (FACS) [61]. Using this method, EVs can be isolated based on light scattering characteristics and fluo-rescent tags bound to specific EV markers via labeled antibodies. Immunocapture is particularly suitable to isolate subpopulations of EVs based on specific surface markers. As there is no generic EV surface marker, immunocapture cannot isolate all EVs present in a sample, resulting in a lower overall yield than other isolation methods that isolate EVs based on size or density [48]. For this reason, immuno-capture approaches are most often used when a specific EV subpopula-tion needs to be studied and not so much when the study focus is on the general EV population. A commercial kit that applies immunocapture based on the tetraspanins CD9, CD63 and CD81 is currently already available; Exosome Isolation Kit Pan (Miltenyi Biotek). If more universal EV markers would become available, immunocapture approaches might have the potential to become the new Gold Standard for EV isolation, provided that the costs will decrease.
3.5. Ultrafiltration
A relatively fast and straightforward technique is ultrafiltration
(UF). This technique is also based on the size of EVs but has a lower
resolution compared to SEC. Ultrafiltration is particularly suitable to concentrate EVs that have been isolated by other approaches such as (diff)UC or SEC but can also be used as a primary EV isolation/
Table 3
Overview of the 21 papers included in the second database. Studies in bold are studies that compare single EV isolation methods with combinatory EV isolation methods. Bold studies represent studies that compared at least one single EV isolation method with at least one combined EV isolation method. Bold and underlined numbers represent the ranking values of the superior EV isolation methods. Abbreviations: (diff)UC: differential ultracentrifugation, CUC: cushioned-
density (ultra)centrifugation, DGUC: density gradient (ultra)centrifugation, EIK: Exosome isolation kit Pan (immunoaffinity based on CD9, CD63 & CD81), EP: Exoprep (precipitation method), EQ: Exoquick (precipitation method), EXE: exoEASY, GAG: ExoGag (glycosaminoglycan precipitation method), HPM: Heparin/ polymer coated microspheres, IBMF: Immunity-based microfluidics (antigen: annexin), LSC: low-speed centrifugation (<80 000g regardless of time), MC: miR-CURY (precipitation method), PEG: polyethylene glycol (precipitation method), PEX: Pure Exo isolation kit, PK: proteinase K treatment, SEC: size-exclusion chromatog-raphy, SELC: size-exclusion liquid chromatogchromatog-raphy, Spin: EXOspin (SEC + precipi-tation method), TEI: Total exosome isolation kit, UEP: Urine exosome precipiprecipi-tation (and RNA isolation) kit, UF: ultrafiltration.
Pubmed
ID Nr. Biological material Method comparison Ranking based on yield Ranking based on purity 30973950 1 CCS LSC + UF + DG 4 1 LSC + UF + (diff)UC + DG 1 4 LSC + UF + CUC + DG 3 3 LSC + UF + PEG + DG 2 2 31069026 2 CCS (diff)UC + UF +DG + (diff) UC 1 2 (diff)UC + UF 2 1 31236201 3 CCS (diff)UC + UF 2 2 LSC + UF + TEI 1 1 (diff)UC + UF +DG + UF 4 2 (diff)UC + SEC + UF 3 2 30728924 4 CCS (diff)UC + UF 1 1 LSC + UF + SEC 1 2 30728924 5 CCS (diff)UC + UF 1 1 LSC + UF + SEC 1 2 31588683 6 CCS (diff)UC 1 3 TEI 2 2 LSC þ IBMF 3 1 31842290 7 CCS GAG 2 2 (diff)UC 1 1 31680809 8 CCS LSC + UF + (diff)UC 2 1 LSC + UF + SELC 1 2 31596073 9 CCS (diff)UC + UF 2 1 (diff)UC + UF +HPM 1 2 30604646 10 Serum (diff)UC + UF (1) 3 1 (diff)UC + UF (2) 1 1 LSC + CUC + (diff)UC 2 2 31588683 11 Serum (diff)UC 1 1 LSC þ IBMF 2 2 31731761 12 Serum LSC + EQ 2 1 LSC + TEI 1 1 LSC + MC 3 2 29887978 13 Serum MC 4 2 EXE 1 3 SEC 2 5 Spin 3 4 (diff)UC 1 1 31671920 14 Serum LSC + TEI 2 1 1 2 Table 3 (continued) Pubmed
ID Nr. Biological material Method comparison Ranking based on yield Ranking based on purity LSC + UF + EXE + (diff) UC 31561474 15 Plasma (diff)UC 1 – LSC þ PEG 3 1 LSC þ PEG þ (diff)UC 2 2 30658418 16 Plasma (diff)UC 1 1 LSC þ EQ 3 1 LSC þ TEI 2 1 31839906 17 Plasma (diff)UC 1 6 LSC þ EQ 6 2 LSC þ UF 4 3 LSC þ SEC 3 5 LSC þ UF þ EXE 2 1 LSC þ PK þ TEI 5 4 30871557 18 Plasma (diff)UC 3 2 EQ 2 1 PEG 4 1 LSC þ PEX 1 3 30755672 19 Amniotic
fluid (diff)UC þ UF TEI 4 2 2 1
EP 5 1 EQ 3 3 SEC 1 4 30811394 20 Media from cultured helminths LSC þ UF þ SEC 2 2 (diff)UC 1 1 31194192 21 Urine LSC + UF + UEP 1 1 LSC + UF + EIK 4 3 LSC + UF + EXE 2 2 LSC + UF + MC 3 2 (diff)UC + UF +DG + (diff) UC 1 2
Journal of Chromatography B 1169 (2021) 122604
concentration technique [41]. The membranes are often made of cel-lulose, polyethersulfone or hydrosart and usually have a pore size of 10–100 kD [62]. Commercial kits are also available that are based on UF (e.g. Urine Exosome Purification kit (NORGEN BIOTEK CORP.)). When using ultrafiltration, a point of consideration is that the specific filtering device may have a considerable impact on the final result, with centrifuge-driven ultrafiltration being superior to pressure-driven ul-trafiltration regarding particle yield [53,62]. An advanced form of UF is
tangential flow filtration (TFF), which is similar to UF, albeit the
sample is flowing along the membrane instead of towards the membrane as is the case with conventional UF. TFF has been shown to be superior to (diff)UC in terms of EV yield, EV aggregation and batch-to-batch consistency [63].
3.6. Field-flow fractionation
Field-flow fractionation (FFF) is a relatively new method within
the EV isolation field. Particles in the injected sample are separated in a
channel with a semi-permeable membrane under the influence of a longitudinal parabolic flow, while a perpendicular gradient or force field is applied simultaneously. This gradient or field can be provoked using an electrostatic force, thermal energy gradient, centrifugal force, or an additional flow (asymmetric flow field-flow fractionation: AF4). EVs were shown to be successfully isolated based on their electrophoretic mobility by electric FFF [64]. However, in most cases, AF4 is used, probably also because the exomere subpopulation was originally iden-tified with this type of FFF [6]. A detailed protocol has been developed for exomere isolation using AF4 [65]. However, also, this technique has its limitations. For instance, it is not easy to apply in a conventional lab since specific instrumentation and trained personnel are required to optimize the various parameters (e.g. cross-flow velocity, channel height, membrane type). Besides, samples need to be pre-concentrated first, which means that AF4 cannot be applied as a stand-alone technique.
Fig. 5. Overview of the performance of EV isolation methods from database 2 in terms of EV purity and EV yield, as indicated by their summed ranking values corrected by their frequency of use and sorted by the EV purity. A) Comparison between all EV isolation methods. B) Comparison between the EV isolation methods of all studies that used cell culture supernatant as starting material. C) Comparison between the EV isolation methods of all studies that used plasma/serum as starting material. D) Comparison between the EV isolation methods of all studies that tested combined EV isolation methods as well as single EV isolation methods.
Ab-breviations: (diff)UC: differential ultracentrifugation, CUC: cushioned-density (ultra)centrifugation, CCS: cell culture supernatant, DGUC: density gradient (ultra)centrifuga-tion, EIK: Exosome isolation kit Pan (immunoaffinity based on CD9, CD63 & CD81), EP: Exoprep (precipitation method), EQ: Exoquick (precipitation method), EXE: exoEASY, GAG: ExoGag (glycosaminoglycan precipitation method), HPM: Heparin/polymer coated microspheres, IBMF: Immunity-based microfluidics (antigen: annexin), LSC: low- speed centrifugation (<80 000g regardless of time), MC: miRCURY (precipitation method), PEG: polyethylene glycol (precipitation method), PEX: Pure Exo isolation kit, PK: proteinase K treatment, SEC: size-exclusion chromatography, SELC: size-exclusion liquid chromatography, Spin: EXOspin (SEC + precipitation method), TEI: Total exosome isolation kit, UEP: Urine exosome precipitation (and RNA isolation) kit, UF: ultrafiltration.
3.7. Microfluidics
Another example of an innovative EV isolation method is
micro-fluidics. A significant advantage of microfluidics is that EV isolation and
characterization can be combined on a single chip, reducing the total workload [66]. EVs are either isolated based on specific surface markers (microfluidics-based immunoaffinity capture) or their size and density (acoustofluidics, nano-sized deterministic lateral displacement, membrane-filtration microfluidics, viscoelastic flows or nanowire traps) similarly to other commonly used techniques [67]. Due to these sepa-ration principles, it could be argued that microfluidics is not mainly a distinctive isolation method but rather an extension to other commonly used methods (e.g. immunocapture approaches or ultrafiltration). However, the use of the microfluidic chip allows EV isolation from minimal sample volumes and specific parameter modulation allows isolation of EV subpopulations and prevents EV damage (e.g. with acoustic waves), which can usually not be obtained with more standard techniques [68,69]. Nevertheless, this method also fails to fully separate EVs from lipoproteins [69]. Further, the fabrication of a microfluidic chip takes time and requires a certain level of expertise to develop.
3.8. Other less commonly applied strategies to isolate and purify EVs
EVs can also be isolated based on membrane affinity methods that isolate EVs based on their membrane. An example is the use of annexin V that has high affinity for membrane compounds such as phosphati-dylserine membrane lipids [70]. However, before membrane affinity methods can be applied, all cells and cell fragments need to be separated from the sample in order to prevent contamination. A commercial kit is
also already available based on membrane affinity; exoEasy Maxi kit (QIAGEN). Another example is the use of heparin/ poly
(dia-llyldimethylammonium chloride) (polyDDA)-coated microspheres (HPM), which can enrich EVs from biological materials by electrostatic
interactions, hydrogen bonding, Vander Waals interactions and hydro-phobic effects between the positively-charged amino acids and the negatively-charged HPM [71]. The level of protein contaminants can also be reduced by the addition of proteinase K to the sample, which breaks down all extracellular protein. A disadvantage, however, is that EV-associated membrane proteins may also be damaged, which alters EV functions [72].
4. Systematic literature analysis
Given the many different EV isolation methods with their different working principles and their associated advantages and disadvantages, it is worth considering combining different methods to enhance EV purity. Methods that isolate EVs based on size could for instance be combined with methods that isolate EVs based on density. To investigate to what extent combining different EV isolation methods is already being done and whether this approach is superior to single EV isolation methods in terms of EV purity and EV yield, we have performed a sys-tematic literature analysis using PubMed. Our study was focused solely on the year 2019, which allowed us to thoroughly investigate the details of a publication while still providing a representative view of the current state-of-the-art. An overview of the methodology of our two-step liter-ature analysis can be found in Fig. 2, and detailed inclusion criteria for papers can be found in supplementary material 1.
Journal of Chromatography B 1169 (2021) 122604
4.1. Current distribution between single and combined EV isolation methods
The initial search resulted in 896 hits in PubMed. After a preliminary exclusion of papers that were not accessible to us or that did not include a thoroughly described EV isolation procedure, 700 papers containing information about 892 EV isolation approaches (single as well as com-bined methods) were included in database 1 of the analysis (see sup-plementary material 2). In about half of all EV isolations, cell culture supernatant was used as starting material and in only a fifth of all EV isolations plasma/serum was used as the starting material. In total, 56 different starting materials were used for isolation of EVs (see table 2). Fig. 3 was created with the aid of database 1 showing the distribution between single and combined EV isolation methods, also separated by starting material. The results show that about 60% of EV isolations were performed with combined EV isolation methods. Interestingly, the proportion of combined EV isolation methods for plasma/serum (52%) was lower than the proportion of combined EV isolation methods for CCS (59%), while plasma/serum is known to be a more complex biofluid than CCS. Urine had a similar distribution between single and combined EV isolation methods as CCS, while with milk or tissue as starting ma-terial, this distribution was largely shifted toward combined EV isolation methods (80% and 70%, respectively).
When looking into more detail to the various applied methods, among the ten most frequently applied EV isolation approaches (see Fig. 4), six combined EV isolation methods were present for both CCS and plasma, which reflects the total distributions between single and combined EV isolation methods as depicted in Fig. 3. Interestingly, for
CCS, there is a clear preference for two methods, while the distributions are more gradual for the various methods applied for plasma/serum. For both CCS and plasma/serum, (diff)UC was the most favored method of all twenty-six different single EV isolation methods used in database 1 (see Fig. 4). This is not unexpected since (diff)UC was the first method that was used to isolate EVs [4]. After (diff)UC, LSC was most often preferred as a single EV isolation method for CCS, while the commercial ExoQuick® precipitation kit (EQ) was the second-most chosen single EV isolation method for plasma/serum. Regarding the specific combina-tions of methods, (diff)UC was most often combined with UF for CCS and plasma/serum. Other methods that were regularly included in combined approaches were LSC and several precipitation kits. While comparing CCS and plasma/serum, it can also be noticed that no combined EV isolation approaches were present in the top ten for plasma/serum that combined more than two different EV isolation methods, while multiple of those combinations rank within the top ten for CCS.
For the other starting materials (apart from CCS and plasma/serum), (diff)UC and (diff)UC + UF were also the most popular EV isolation methods for most other starting materials, while the second most frequently used single and combined EV isolation methods were heavily dependent on the type of starting material (see supplementary material 2).
4.2. Efficacy of combined EV isolation methods
For the next step in our literature analysis, we selected all those studies in database 1 that compared two or more EV isolation methods and screened for the effects of combined EV isolation methods in terms
Fig. 5. (continued).
of EV purity and EV yield. We first checked and selected each study for comparability (i.e. same starting material for EV isolation and same characterization methods) and second that EVs were characterized ac-cording to the MISEV 2018 guidelines (i.e. assessment for positive as well as negative EV markers, EV yield determination) (see supplemen-tary material 1). Each included publication was subsequently analyzed in-depth by one reviewer for the different EV isolation approaches described for EV purity and EV yield. A second reviewer checked again based on the fore-mentioned criteria whether each paper should be included or excluded in database 2.
After establishing database 2, the EV isolation approaches were ranked in each included study based on their EV yield and on their EV purity for comparison within studies (see table 3). For determining the EV yield rankings, NTA data was most often used. Only when these data were not available, total protein content data was used. For determining EV purity rankings, particle/protein ratios and Western blot data with specific EV markers and non-EV markers were used (see supplementary material 3). The higher the ranking value, the better the method per-formed within the study. For comparison between the included studies, for each EV isolation method, the ranking values from the individual studies were summed up and subsequently corrected for the frequency of use for both EV purity and EV yield (see supplementary material 3). We have chosen this ranking approach since different studies cannot always be compared entirely in a consistent manner due to differences in experimental details that are not always clearly stated.
Eighteen papers from the first database that described twenty-one different comparisons (some papers contain a comparison with multi-ple starting materials for EV isolation and are there included separately for each sample type) between EV isolation approaches were included in
database 2 (see table 3). For CCS, nine studies were selected, of which one compared single and combined EV isolation methods and the others compared either various combined methods or single methods. For plasma/serum, nine studies were selected, five of which made a com-parison between single and combined EV isolation methods and the others compared either various combined methods or single methods. For other starting materials, one study on amniotic fluid and one on medium from cultured helminths that made a comparison between single and combined EV isolation methods matched all inclusion criteria. The last included study compared various combined isolation methods using urine as starting material.
Regarding the comparison within studies, we found eight studies in our database that compared single with combined EV isolation methods. In six of these studies, combined methods outperformed the single methods regarding EV yield and in four of these studies, combined methods outperformed single methods regarding EV purity. In case the combined EV isolation method was superior to the single EV isolation method, LSC was always part of the combination.
Our next step was to determine which EV isolation methods were best when a comparison was made between the studies present in data-base 2 (see supplementary material 3). When the type of starting ma-terial was not taken into account, the precipitation kit Exoprep (EP) and the EV isolation approach that combined LSC with proteinase K
(PK) and the total exosome isolation kit (TEI) performed equally best
regarding EV yield (see Fig. 5a). For EV purity, LSC combined with SEC performed best, followed by SEC as single EV isolation method. Clearly, the trend can be seen regarding methods that score higher on EV purity, very often score lower on EV yield and vice versa.
When the comparison between studies that used CCS as starting
Journal of Chromatography B 1169 (2021) 122604
material was made (see Fig. 5b), it turned out that (diff)UC combined
with DG and two forms of UF performed equally well to LSC þ UF þ DG in terms of EV yield. For EV purity, LSC þ UF þ (diff)UC þ DG
seemed to be superior to the other methods. When the comparison
be-tween studies that used serum/plasma as starting material was made (see
Fig. 5c), LSC þ PK þ TEI performed best in terms of EV yield, while LSC
þSEC and SEC alone performed equally best in terms of EV purity.
When the comparison between studies that tested single and combined EV isolation methods was made, EP and LSC þ PK þ TEI performed equally best regarding EV yield, while LSC þ SEC performed best regarding EV purity (see Fig. 5d).
5. Discussion
In this review, we performed a systematic literature analysis over the year 2019 to screen for the use and benefits of combined EV isolation methods in terms of EV purity and EV yield, subdivided per biological material.
The first part of our analyses looked at the current use of EV isolation methods and showed that about 60% of all EV isolations were based on combined methods. According to another literature analysis, this num-ber was only ~40% [73]. This difference could be explained by the fact that the analysis published by Royo et al. was not based on a certain number of publications but on the response of 357 highly active EV researchers to a questionnaire over the timeframe March till August 2019. The difference may be due to the fact that EV researchers from the Royo study used more often plasma/serum than the authors of the publications we included in our database, since we have found that combined EV isolation methods are more often applied with CCS than with plasma/serum as starting material.
The reason for this difference between CCS and plasma/serum is not clear because we anticipated that the use of combined EV isolation methods would be more beneficial for more complex biological mate-rials such as plasma/serum than for CCS. A possible reason might be that cell culture supernatant is usually easier to obtain in larger volumes than plasma/serum. Therefore, more experimenting regarding a suitable EV isolation method is possible with CCS compared to plasma/serum derived from patients, which would also be an explanation for our observation that the combination of more than two different EV isola-tion methods was more common for CCS than for plasma/serum. Another reason could be that if EVs should become implemented in clinical biomarker screening, this screening should be fast, which favors the choice for a single instead of a combined EV isolation method. CCS is also more often used to study specific biological mechanisms, and that, therefore, a higher purity is required to be able to attribute the observed effects to well-characterized highly purified EVs thus favoring combined isolation methods.
The second part of our analyses showed that combined EV isolation methods provided a clear benefit regardless of the type of starting ma-terial. The optimal combination of methods varies between different starting materials, as expected due to the differences in sample complexity and potential contaminations. In case of plasma/serum, a precipitation method should be part of the approach when the focus is merely on a high EV yield, while for CCS a combined (diff)UC approach could also be used for this purpose. Regarding EV purity, a method that should certainly deserve more attention according to our study is SEC. Preferably, when combined with a form of centrifugation, relatively pure EV fractions can be isolated, while still resulting in a reasonable EV yield. It was already demonstrated before that SEC should be combined with a form of (ultra)centrifugation in order to remove lipoproteins as efficiently as possible from plasma/serum based on the EV proteome [26]. The smaller lipoproteins with a similar density to EVs can then be removed by SEC, while the lipoproteins with a lower density, but a similar size to EVs can be separated with a form of centrifugation.
Surprisingly, the results of our first database and our second database are not entirely in agreement. Even though our analyses in database 2
clearly show the benefit of combined isolation methods, we do not see this pattern for the applications when screening all publications from 2019. Our first database showed that the single (diff)UC method and the (diff)UC + UF method were most often applied for EV isolation inde-pendently of the type of starting material. However, according to our second database, these methods were not the optimal choices regarding both EV purity and EV yield.
These results could be biased possibly by the limitations from our current study design, namely looking at a single publication year only and the risk for interpretation-dependent annotations and application of exclusion criteria for the generation of both databases. For the genera-tion of the second database, we therefore applied a double reviewer combination to minimize bias in the database creation. With the screening of all papers in 2019 the review creates a clear indication of the current EV field, showing a clear bias towards (diff)UC methods either as single method or combined with UF. Though our database does not provide a complete overview of previous years, it is likely to expect that this observation can be extended for earlier EV work, since these methods were at the basis of the EV studies.
Additionally, thorough characterization of these properties is very important to study EVs. It was surprising to see so many studies focusing on the function of EVs or even EV diagnostics and EV therapeutics without careful assessment of the purity of the EV preparation and thorough characterization of the isolation method. This may create a situation where results may not be reproducible and thus not compa-rable. The EV field can learn from other fields that have been (are) in a similar situation, notably the field of biomarker discovery and devel-opment [74]. In the field of clinical diagnostics and therapeutics, it is indispensable to work with well-characterized EV preparations and highly reproducible, standardized isolation procedures to translate re-sults from the research stage to the clinic and possibly commercialization.
Even though our work demonstrates that combined EV isolation methods perform better, we would still like to stress that there is currently not one EV isolation approach that is best for all EV study purposes. The type and sample volume of starting material matters, but also the downstream application. For instance, if a study is solely focused on EV RNA’s, some protein contamination is acceptable, also, if a high yield is required because of a limited sample volume, contami-nation of other compounds is unavoidable. More and larger comparative studies should still be encouraged so that in future a more complete image could be obtained regarding which isolation approaches can be recommended in terms of EV yield and EV purity per different starting material and per different type of common downstream application. Until then, all EV isolations should be documented in detail to ensure reproducibility and where possible, several EV isolation approaches should be tested before specific biological functions are attributed to EVs. Nevertheless, the authors recommend to include the SEC method in an EV isolation method, since this research has shown that SEC is often beneficial for EV purity, while still resulting in an acceptable EV yield. In summary, this review provides an overview of currently applied EV isolation methods and an overview of which combination of isolation methods provides the best results for EV purity and EV yield for various sample materials. It therefore does not only provide a summary of the current applications in the field, but is also suitable for selecting an optimal isolation strategy for the isolation of EVs from various biofluids.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
We acknowledge the support for this work that J. Stam received from
