Overview on exosomes

Exosomes are the smallest subtype of extracellular vesicles (EVs), ranging from 30 to 150 nm in diameter. These nanovesicles of endocytic origin are formed by plasma membrane budding, resulting in early endosomes that mature into multivesicular bodies (MVBs) which bud inward, creating intraluminal vesicles. In the final step, the MVBs can either intermediate the intracellular protein degradation process by fusion with lysosomes, or can fuse with the plasma membrane leading to exosomes release [1, 2]. These nanovesicles carry a variety of molecular cargo including proteins, lipids, and RNA [2]. This ability to pack material was initially thought to be only related to their function as cellular waste disposal systems. However, exosomes emerged as key players in numerous pathways, by mediating cellular communication and signalling events, including inflammation [3] and apoptosis [4]. Indeed, these nanovesicles have been widely related to cancer development [5, 6], and more recently with the pathogenesis of neurodegenerative diseases as Alzheimer’s disease (AD) and Parkinson’s disease [1, 7].

Exosomes can also be released from numerous types of cells including fibroblasts, intestinal epithelial cells, and neurons, being found in many biofluids as blood [8, 9], cerebrospinal fluid (CSF) [10], saliva [11], urine [12], and breast milk [11, 13]. Hence, these nanovesicles are likewise being addressed as ideal sources for biomarker discovery representing potential tools in disease diagnosis [1, 14]. Nonetheless, exosome isolation methods still require optimization and characterization to define the best procedures to be employed in clinical practice. Therefore, many research focused on these topics and initiatives promoted by the blood EVs work group are being directed to increase standardization and reproducibility in blood-derived EVs research [15].

Exosome isolation methods

Considering the increasing interest in this field, several techniques for exosome isolation and characterization have emerged (Fig. 1) [2, 14, 16]. Regarding the methods of isolation, ultracentrifugation (UC) is one of the most employed, consisting in the application of a high centrifugal force to a fluid to allow the deposition of particles according to their size (Fig. 1a). At first, samples need to be centrifuged at low speed to remove dead cells and cell debris and then, centrifuged at increasing speeds (~ 10,000 g) to sediment larger EVs, as microvesicles. This is followed by a high-speed ultracentrifugation step at (~ 100,000 g) to pellet small EVs such as exosomes. Usually, an additional UC step is performed to wash the exosome pellet in phosphate-buffered saline (PBS) and decrease protein contaminants [17]. Although UC is the standard method for exosome isolation, it is a time-consuming process that requires an expensive equipment (ultracentrifuge) and renders in low exosome yield. Further, biofluids with higher viscosity require longer UC steps and higher centrifugation periods which can compromise exosome integrity. UC can also be combined with a density gradient (dgUC) of sucrose, Nycodenz (iohexol), or iodixanol to separate particles according to their density, yielding exosome preparations with higher purity. In this method, samples are placed in the top of a density gradient medium (higher densities in the top) and exosomes move through density gradient medium during the UC step (Fig. 1b). The principle behind is that under centrifugation, particles with different sedimentation coefficients settle in distinct layers that can be further collected. Exosomes float until get equilibrium density, ranging from 1.10 to 1.21 g/mL on a sucrose gradient, forming a fraction zone that can be easily recovered [18,19,20].

Fig. 1
figure 1

Schematic representation of the most common exosome isolation methods. (a) Ultracentrifugation. (b) Density gradient ultracentrifugation step. (c) Ultrafiltration. (d) Size exclusion chromatography. (e) Membrane affinity. (f) Precipitation-based methods. (g) Immunoaffinity. (h) Microfluidics (acoustic)

Full size image

Other exosome isolation methods available that separate vesicles according to size are ultrafiltration (UF), size exclusion chromatography (SEC), and field-flow fractionation (FFF). In UF, one or several filters with distinct molecular weight cut-offs or size are used to isolate exosomes, separating them from large particles (Fig. 1c). Although UF does not require specialized equipment and is easy to perform, it presents low efficiency due to the clogging of filter units. To solve this problem, tangential flow filtration (TFF) had arisen. In contrast with the typical filtration, where pressure is applied in the same direction of the flow, in TFF the pressure is applied perpendicularly. This avoids the clogging and, by manipulating hydrodynamic flow forces, it allows a more efficient separation of vesicles from small protein contaminants, starting from high fluid volumes [21]. Another advantage of this procedure is that filters retentate can be recirculated and filtered repeatedly, increasing the protocol efficiency, and TFF use provides isolation of biological active EVs [22], while in typical UF the pressure applied frequently leads to vesicle deformation or lysis. In TFF applied to sequential filtration, at first, the biofluid is prefiltered to remove cells, cell debris, and large rigid vesicles. Next, TFF can be performed to exclude free proteins, using 500-kDa molecular weight cut-off (MWCO) hollow fibers and then, the biofluid can be filtered using 100- to 200-nm filters to separate exosomes from large flexible vesicles [23]. In SEC, vesicles in biofluids are separated according to their size when passing through a porous polymeric phase with beads, multiple tunnels, and pores (Fig. 1d). Vesicles with small hydrodynamic diameters usually ranging from 30 to 200 nm are trapped into pores and lately eluted. Particles with sizes higher than 1 μm do not enter the porous phase and particles smaller than exosomes enter in the porous zone but are not retained. The eluate is collected in several fractions, and, after column void volume, the first fractions contain large vesicles whereas the following contain small EVs (e.g., exosomes), and the last fractions have smaller particles or proteins. SEC is also a time-consuming method but provides pure exosome isolations, preserving exosome integrity, and thus, SEC can be combined with other techniques to increase EVs purity and yield [18,19,20, 24]. Membrane affinity–based spin column is another isolation technique based on chromatography (Fig. 1e). In this method developed by Qiagen (ExoEasy), the binding of EVs to a column membrane is promoted based on a vesicle-specific biochemical feature, but this interaction does not distinguish between exosomes and apoptotic bodies, cells, or cell debris, requiring a previous centrifugation step or filtration. Other larger particles or protein aggregates are removed through column washes and then EVs are eluted intact in an aqueous buffer with inorganic salts. As an option, EVs can be lysed and eluted with QIAzol and then, total RNA extracted upon addition of chloroform and centrifugation (ExoRNeasy) [25, 26]. In the recent field-flow fractionation method, fluids are injected into a channel with a permeable membrane that works as an accumulation wall. Then, these are subjected simultaneously to a longitudinal parabolic fluid that carries EVs along the channel and to a perpendicular gradient or force field that accumulate these vesicles at the bottom wall, according to their size. The asymmetric FFF is the most common gradient applied but others can be used as electrostatic forces or centrifugal force. In asymmetric FFF, smaller vesicles became positioned further from the wall while larger vesicles are pushed to it and, thus, small vesicles are eluted first than the larger [27, 28]. Although this is a rapid and efficient procedure, it requires trained individuals.

Since exosomes present negative charged membrane components (e.g., phosphatidylserine), several charge-based isolation methodologies have arisen to explore this property, including ion-exchange techniques or electrophoresis. In ion-exchange methods, negative-charged membranes interact with anionic positive charged surfaces (e.g., cationic polymer-coated beads) and, after, EVs are eluted by the addition of buffers with high salt concentration [29, 30]. In electrophoresis, EVs are separated according to their ability to move when an electric field is applied [31].

Precipitation is a method employed to isolate exosomes from body fluids by altering EVs solubility. This involves a low-speed centrifugation step to remove cells and debris followed by mixing and incubation with water-excluding polymers, like polyethylene glycol (PEG), that link to water molecules and enhance the precipitation of less soluble components, as exosomes. Subsequently, the mixture is centrifuged at low speed, pelleting EVs with exosome-like characteristics (Fig. 1f). Polymer-based isolation renders high exosome yields, preserving EVs integrity. However, it can also co-precipitate protein aggregates and other materials as lipoproteins, and the solely polymer reagent can interfere with downstream analysis as mass spectrometry, decreasing its sensitivity [18,19,20]. Nonetheless, the pellet resuspension in water or PBS promotes the disassembly of the polymer network. Polymers are biochemically inert and if needed the remaining residual polymers can be easily removed using gel filtration resin columns [32].

In addition to these methods, the discovery of exosome surface proteins and receptors allowed the development of immunoaffinity-based methods. These are commonly based on interactions between antigens, as surface exosome markers, and antibody coupled to magnetic beads (Fig. 1g). Further, the beads-exosomes complexes formed are characterized by several techniques as flow cytometry and electron microscopy. The main disadvantage of exosome immunocapture methods is that only subpopulations of exosomes are isolated, depending on the target exosome marker and, to note, sometimes it is difficult to elute exosomes from beads, decreasing its yield [18,19,20]. An innovative technology proposes the use of noncovalent interactions to allow a better disassembly of exosomes from beads. Superparamagnetic beads coupled with anti-CD63 (an exosomal marker) antibodies, based in the interaction between β-cyclodextrin and 4-aminoazobenzene, were constructed to capture exosomes that were then eluted by the addition of the competitive host molecule α-cyclodextrin [33]. This new methodology was efficient in exosome elution and render in pure exosome isolations [33], encouraging the use of immunoaffinity-based methods.

Microfluidic techniques have recently been developed to isolate exosomes from very small amounts of fluids in a rapid, automated, and cost-effective manner, even with high-throughput capacity [18,19,20]. These small platforms also allow both the isolation and the analysis of exosome cargo which is a great advantage for biomarker discovery and application in clinics. Several isolation principles have been used in microfluidics as immunoaffinity, size-based separation, and contact-free microfluidics. In immunoaffinity-based microfluidics, exosomes are captured by antibodies immobilized in the surface of the platform. As in typical immunoaffinity, these antibodies recognize specific exosome markers and, then, exosomes are eluted after buffer addition. Efforts have been made to improve the antibody immobilization surface area (e.g., using nanoporous structures) and to reduce the non-specific antibody binding to other non-exosome vesicles. To note, beyond exosome isolation, affinity-based microfluidics also allow simultaneous RNA extraction [34]. Another example is the ExoChip commercial device, with surfaces fabricated in polydimethylsiloxane and containing anti-CD63 antibodies immobilized. This device allowed exosome isolation and quantification from serum, after a fluorescent labelling, electron microscopy analysis, and the isolation of intact RNA from immobilized vesicles [35]. An evolved version of this method for isolation and characterization of cancer-specific exosomes has been published, and it is based on lipid-affinity microfluidics exosome isolation instead of immunoaffinity. Annexin V was immobilized on the chip surface and specifically interacted with phosphatidylserine, externalized only in the outer surface of cancer cells and exosome membrane [36]. It is expected that other affinity-based devices will be developed to meet the needs of specific exosome subpopulation isolation. Size-based microfluidics uses several strategies to separate the exosomes from fluids as nanoporous filters and porous silicon nanowire-on-micropillar structures. ExoTIC was the first size-based microfluidics device developed and consisted on the filtration of small volumes of biological fluids through a nanoporous membrane, separating exosomes with higher yield than typical UC [37]. Nanowire-based microfluidic devices were composed of several nanowires fixed in micropillars. These devices allow the isolation of specific size vesicles subpopulations by adjusting the space between nanowires, leading to vesicles trapping and filtering out cells, large vesicles, and proteins. Finally, exosomes were simply eluted with PBS by dissolving the nanowires [38]. In addition, microfluidics deterministic lateral displacement had been used to separate vesicles according to their size in pillar arrays [39, 40]. Recently, ZnO nanowire arrays with a bifunctional peptide were developed to capture cancer exosomes [41]. Regarding contact-free microfluidics, viscoelastic flow, acoustic, and electrophoresis are the most common. In viscoelastic flow microfluidics, distinct size vesicles migrate according to the manipulation of elastic lift forces applied in a viscoelastic medium [42]. In acoustic-based microfluidics, the application of ultrasound waves exerts differential forces into particles and causes the separation according to their physical properties (Fig. 1h). These devices also require a very small sample volume, which is an important advantage in clinical research. A combination of microfluidics technology and acoustics was employed in exosome isolation from blood, providing high yield and purity. Additionally, this method allowed automation of the EVs isolation process and the resulting vesicles were intact and retained biological activity [43]. Despite the promising results, these recent microfluidics-based methodologies need additional validation and run of large-scale tests [18,19,20]. Exosomes can be isolated through electrophoresis and microfluidic devices had been developed, taking advantage of this principle. Electroactive strategies can be used to drive exosomes along the devices and, for example, through membranes [44].

As discussed, several methods can be employed to isolate exosomes and novel approaches are still arising. All techniques exhibit advantages and limitations, and these have been summarized in Table 1, for the most common exosome isolation methods.

Table 1 Advantages and disadvantages of exosome isolation methods
Full size table

Exosome quantification and characterization techniques

According to MISEV2014 and MISEV2018 recommendations, the characterization of EV preparations must include the quantitative description of the EVs source and isolation procedure details, including the starting volumes of body fluids, cell culture media, or mass volume, when isolating EVs from tissues, and report of EVs size and concentrations. Characterization of single vesicles must be carried out through techniques that provide high-resolution images of EVs, based on electron microscopy, and using single-particle analyzers to detail EVs biophysical features. Total protein and lipid content can also be quantified and the purity ratios (proteins to particles, lipids to particles, or lipids to proteins) should be determined. In addition, at least three positive protein markers (transmembrane and cytosolic proteins) and one negative marker should be assessed in EVs preparations. Non-vesicular, co-isolated contaminants can also be addressed [45, 46].

A critical step after EVs isolation with exosome-like characteristics is its characterization in terms of particle size, morphology, concentration, and markers. Therefore, advanced methodologies have been developed to this end.

Nanoparticle tracking analysis (NTA) is used to measure particle size and concentration. To do so, a laser beam hits nanoparticles in liquid suspension and the light scattered is captured by a camera, generating several frames. According to Brownian motion, the smaller particles move faster than the larger. Using video analysis software, the movement of each particle is tracked and analyzed individually, and the diffusion coefficient determined. Then, this coefficient is used to calculate particle hydrodynamic diameter through the Stokes–Einstein equation. Particle concentrations are determined according to the number of particles in suspension, tracked individually. NTA provides a direct quantification of small particles with sizes between 40 and 1000 nm and concentrations between 107 and 109 particles per mL, requiring a minimum of 500 µL of particle suspension for the analysis. NTA results include particle diameter, concentration, size distributions, and 3D plots that combine size, particle number, and intensity of light scattered, facilitating the visualization of distinct nanoparticle subpopulations. Since NTA is based on Brownian motion, it is sensitive to the presence of aggregates or higher particles which can constitute a disadvantage of the technique. Despite that, NTA allows direct quantification of EVs in a quick and easy manner and it can also detect fluorescently labeled EVs [47, 48].

An alternative technique to measure EVs size is dynamic light scattering (DLS) and, like NTA, it is based on the Brownian motion principle. However, while NTA determines the individual particle size based on the diffusion coefficient, DLS calculates sizes according to intensity changes of scattered light, measuring a bulk of nanoparticles [49]. Although DLS is a very sensitive method, it presents limitations when distinguishing heterogeneous vesicle populations. In addition, the presence of larger particles or aggregates makes the detection of small nanoparticles difficult and influences the size determinations by DLS in a greater extent than by NTA [47, 50].

More recently, tunable resistive pulse sensing (TRPS) has been applied to measure the number and diameter of EVs by passing particles through the pores of a membrane with applied voltage, using small sample volumes, around 30 µL. As particles are forced to pass pores, the current flowing decreases. These changes in the electric current are proportional to the volume of each individual nanoparticle, and the nanoparticle flow rate allows to calculate the concentration of EVs. TRPS disadvantages include its lower sensitivity to detect small vesicles and the possibility of pore blockage by particles, resulting in measurement changes [2].

The preferred techniques employed to determine vesicle morphology are transmission electron microscopy (TEM) or cryo-electron microscopy (cryo-EM) [51, 52]. TEM is a commonly used technique; however, it must be considered that the dehydrating conditions used in the fixation of samples, the use of metal contrasting agents, and drying can affect the shape of EVs, generating the cup-shaped structure as an artifact in exosome preparations [52, 53]. In cryo-EM, samples are vitrified and, thus preserved in their native hydrated state, allowing to observe exosomes’ spherical shape and, clearly, the lipid bilayer [52, 54].

For confirmation of exosomal nature of preparations, transmembrane or GPI-anchored proteins and cytosolic proteins must be analyzed. Among the most commonly assessed are tetraspanins (e.g., CD63, CD81) and endosomal sorting complex required for transport (ESCRT)-I/II/III components and its accessory proteins (e.g., TSG101, ALIX, Flotillin’s 1 or 2, heat shock proteins 70 or 90). The presence of exosome-negative markers as calnexin, albumin, or apolipoproteins must be determined to assess the purity of EVs preparations [45]. The most used techniques to evaluate exosome markers include Western blotting and flow cytometry (that can also measure particle size and number) [51]. Enzyme-linked immunosorbent assay (ELISA) is another commonly used method that employs immunoaffinity technology to characterize and quantify exosomes [55,56,57]. Recently, ultrasensitive single-molecule array (SIMOA) assays were developed to measure the levels of tetraspanins CD9, CD63, and CD81 in CSF- and plasma-derived EVs [58].

Human biofluids as sources of exosomes: challenges and considerations

The isolation of EVs from biofluids can hold a huge potential to advance biomarker discovery, either in the diagnostic or therapeutic fields. Nevertheless, exosome isolation from biofluids is challenging since each biofluid has its specific composition and physical properties, requiring comparison of the various isolation methods to identify the best approaches to be used in each situation. In this review, some exosome isolation challenges and considerations are addressed for human plasma, serum, CSF, saliva, and urine.

In the case of blood, both plasma and serum are viscous fluids, highly concentrated in proteins (e.g., as albumin) which turn difficult the isolation of pure exosome preparations. It was reported that plasma is more viscous than serum, and thus a trend for lower sedimentation efficiency and higher microvesicles size was observed in the former biofluid, after vesicle isolation through UC [59]. Moreover, other considerations should be taken, as suggested by the ISEV position paper [60]. Although most studies use plasma for EVs isolation, evidence supports that EVs released from platelets soon after blood collection account for a huge percentage of EVs in serum, which might be an advantage for some type of studies [60]. Hence, it would be relevant to conduct additional studies focusing on plasma and serum differences during EVs preparation.

Further, several pre-analytical factors must be taken into consideration prior EVs isolation, since these can change EVs yield and cargo or interfere with downstream applications, namely the time between blood collection and EVs isolation; the type of blood collection tubes; the transport conditions; the temperature and storage periods; the centrifugation; the fasting status; and the physical exercise.

Focusing on some of these factors, the time and type of blood collection tubes are pre-analytical variables that can influence the content in EVs. A study comparing serum collection tubes and ethylenediaminetetraacetic acid (EDTA), heparin, and citrate phosphate dextrose adenine (CPDA) reported that after collection with CPDA and EDTA tubes, the EVs concentration significantly decreased about 1 h after the blood collection, increased after 8 h and, then, returned to the initial levels after 24 h, while for serum or heparin tubes, no changes were observed [61]. Heparin and EDTA tubes were associated with a higher number of EVs while citrate tubes with lowest EVs concentration, although not statistically significant [62, 63]. Another study reported increased concentrations of serum EVs when compared with EVs from EDTA-plasma, citrate, or acid citrate dextrose [64]. No changes in EVs size or morphology had been reported in the use of either citrate, sodium citrate theophylline adenosine dipyridamole, EDTA, or heparin collection tubes [63]. Furthermore, other authors reported that heparin interacted with downstream polymerase chain reaction analysis and directly with EVs, by decreasing their binding and blocking EVs transfer to recipient cells [65].

The impact of transportation was also addressed comparing blood samples placed in an orbital shaker for 1 h, with samples without shaking. The agitation prompted an increase in several EV markers in blood collected in EDTA tubes, which leads to the speculation that the EVs adhered to platelets could be released with shaking [61]. When comparing short (days) and long (months) storage periods, and storage at room temperature and 4 °C, variations were likewise observed. For short-term, storage at low temperatures (−20 °C to −160 °C) showed relatively low signal intensities as detected by EV markers when compared to storage at room temperature or 4 °C. By contrast, for long-term storage, a tendency towards increased signal intensities with lower temperatures was observed [61]. When freezing the samples, heparin tubes provided more stable samples and EDTA tubes presented the highest variations [61]. In particular, long-term storage of plasma EVs at −80 °C or storage of urinary-derived exosomes at −80 °C for 1 week to 7 months did not affect EVs stability [25, 66]. It is mandatory to address the ideal storage conditions for exosomes since it can vary depending on the type of biofluid source [67]. Moreover, successive freeze and thaw cycles are discouraged [68].

The impact of the fasting status at the time of blood collection for EVs isolation is still unclear. Plasma concentrations in EVs, measured with NTA, did not change significantly after the ingestion of a high-fat meal but the plasma concentration of EVs strongly correlated with plasma very-low-density lipoproteins (VLDL) and serum triglyceride concentrations after meal ingestion suggest that NTA particle count may be influenced by the presence of these particles [63]. Another study that evaluated the number of particles through TRPS reported a higher number of particles in the postprandial state, when compared with fasting, and numerous lipoproteins, including low-density lipoproteins (LDL) in both fasting and after meal, although no differences were observed in particle sizes [69]. Chylomicrons, the largest lipoproteins, typically increase in circulation after meals [70] and are metabolized further in VLDL, intermediate-density lipoproteins, LDL, and high-density lipoproteins. All of these lipoproteins are highly abundant in blood and represent relevant contaminants of EVs preparations derived from this biofluid [69], being difficult to distinguish these from exosomes due to size overlap. The literature suggests that blood collection in fasting state is preferable although interpreting conclusions should be cautious, and efforts are still being made to improve EVs isolation and obtain purest preparations under these conditions. The combination of distinct isolation techniques was proven to reduce lipoprotein contamination. Lower amounts of APOE and APOB were observed in serum-derived EVs obtained through cushion UC when compared with UC. Decreased lipoprotein contamination was also obtained in EVs preparations from density gradient cushion UC or from the combination of qEV and dgUC when compared to EVs obtained only through dgUC [71]. In addition, the combination of UC, an iodixanol density cushion, and SEC decreased lipoprotein contamination in plasma-derived EVs. After UC, EVs pellet was placed on the top of a density cushion and the obtained high-density band was then collected and placed in a SEC column. EVs were mainly eluted in fractions 8 and 9 as evidenced by high levels of flotillin-1 and low levels of ApoA1 [72]. A three-step protocol was also stablished to isolate EVs from blood or cell culture media. At first, EVs were isolated using PEG or UC, and then, floated in an iohexol density gradient during 16 h of UC. Finally, EVs were applied to SEC-based columns. ApoA1 and ApoB100 previously identified in density gradient fractions containing EVs were highly reduced after the SEC step [73]. Additional methods have also been developed to separate exosomes from lipoproteins. Some of these are the acoustofluidic-based separation or agarose gel electrophoresis. As referred above, acoustofluidics is based on the different particle’s behavior under acoustic wave pressure. This allowed to distinguish between IDL, VLDL, and chylomicrons, which moved to acoustic pressure antinodes, whereas EVs and HDL particles moved to acoustic pressure nodes [74]. Since EVs, HDL, VLDL, and LDL particles are negatively charged, these can move when an electric field is applied. Hence, lipoproteins and EVs can be separated in 1% agarose gel electrophoresis with tris–acetate-EDTA, according to their size and zeta potential properties. Also, lipoproteins are dyed blue with Sudan Black B which detects triglycerides. EVs are not enriched in triglycerides and thus are not stained. In electrophoresis, the first blue leading band contained HDL and the final leading band contained a mixture of LDL and VDLD. EVs were located between these two bands. Fractions were collected and analyzed by DLS and Western blot [75]. Although the method combination can reduce contaminants, this increases the costs and complexity of the protocols and, unfortunately, leads to the loss of exosomes along the procedures. Beyond lipoproteins or protein aggregates, viruses are frequent contaminants in blood-derived EVs, which prompted the development of new isolation methods such as the nanoscale flow cytometry for EVs separation from viruses [76]. Finally, physical exercise was reported to change the EVs concentration and cargo. A general increase of EVs concentration in circulation after exercise was observed, in both humans and mice [77,78,79,80].

The use of blood-derived EVs in the neurodegeneration field rise with the appearance of a two-step neuronal-enrichment exosome isolation methodology [81,82,83]. It combined EVs isolation from a biofluid through ExoQuick (ExoQ), a precipitation-based method, with a subsequent immunoprecipitation step with antibodies against specific neuronal surface markers (NCAM, L1CAM) [84], astrocyte (GLAST, GFAP, GS) [85], or oligodendrocyte-specific markers (PLP, CNP) [86]. The main purpose of this workflow was the isolation of subpopulations of EVs which could be neuronally derived and, thus, hold important biomarkers for neurodegenerative disease diagnostics. The characterization of L1CAM-positive EVs revealed the enrichment in neuronal-specific proteins [81] and several studies found reproducible altered protein levels between controls and individuals with neurodegenerative diseases, as Alzheimer’s or Parkinson’s diseases, when testing these L1CAM-enriched blood-derived exosomes. However, concerns regarding the L1CAM enrichment neuronal specificity arise as this protein was found expressed in other tissues than brain (e.g., kidneys). It was likewise observed that L1CAM can exist in both transmembrane and soluble forms and, that the former can be cleaved by metalloproteases, generating a soluble ectodomain which is released extracellularly, plus a cytosolic domain bound to the plasma membrane [87,88,89]. In addition, another L1CAM soluble form was recently found, generated through alternative splicing [90]. Since most of the antibodies used in the neuronal EVs isolation, particularly in the immunoprecipitation step, were raised against the L1CAM ectodomain, the specificity of the bound, only to transmembrane L1CAM forms present on EVs, cannot be ensured. Besides it is expected that these antibodies also bind to soluble L1CAM forms. Importantly, in a recent study that used SEC and dgUC to separate EVs from soluble proteins, in both plasma and CSF, it was shown that the majority of L1CAM found in EVs fractions was not bound to EVs, but instead was mainly present as a soluble form [91]. Furthermore, it was observed that the majority of exosomal markers CD9, CD63, and CD81 measured through SIMOA assays were not eluted in the same fractions as L1CAM and albumin, for both SEC and density gradient methods. Complementary Western blot analysis of L1CAM present in CSF and plasma was also carried out, using two antibodies, one directed to the external and the other to the internal L1CAM domain. For CSF, only a 200-kDa band corresponding to the binding of the external domain antibody was observed, whereas in plasma two bands at approximately 220 kDa were detected, reflecting the bound to both internal and external domains. In addition, a mass spectrometry analysis of plasma L1CAM did not detect the transmembrane domain [91]. At this point, it is not clear which is the proportion of EVs that bound L1CAM in biofluids and the exact nature of exosomes obtained by this enrichment procedure.

Hence, further validation of analytical tools used in EVs isolation and biomarker discovery are still needed. The identification of new neuron-specific targets may improve the use of EVs in the clinical research in a wide range of brain disorders, either as a source of biomarkers or as new avenues to the use of EVs as drug delivery vehicles.

In the context of neurological disorders and biomarker discovery, exosomes isolated from distinct biofluids are indeed the focus of intensive research. Cerebrospinal fluid–derived exosomes represent important tools at this level. This biofluid is present in the brain ventricles and the subarachnoid spaces, being collected through lumbar puncture, an invasive procedure [92]. CSF-derived exosomes are expected to carry mainly neuronal-derived proteins and nucleic acids; however, EVs concentrations are lower in CSF when compared with blood-derived exosomes [24]. EVs isolation from CSF requires higher starting sample volumes, which is often difficult to obtain, mainly in the case of healthy individuals. The collection, storage, and processing of CSF are susceptible to a wide range of variability sources which strongly affect the downstream molecular analysis, as is the case for Aβ, total-Tau, and P-Tau 181 (the biomarker triplet measured in AD neurochemical-based diagnosis), making it difficult to establish universal cut-offs. In an attempt to standardize procedures, reduce these sources of intra- and inter-laboratory variabilities, and encourage the use of CSF biomarkers, the international quality control (QC) program was implemented by the Alzheimer’s Association, and headed by the Clinical Neurochemistry Laboratory at the University of Gothenburg, Sweden [93, 94]. Several sources of variability comprising pre-analytical, analytical, and post-analytical factors were identified, including biological interindividual variability and CSF collection procedures, sample storage and shipment, not excluding CSF contamination with blood during lumbar puncture procedure (pre-analytical); training and equipment, kits lot-to-lot variability in the biomarker triplet analysis (analytical), and data handling (post-analytical) [93, 94]. Sources of variabilities identified for this CSF biomarker triplet can be extended to CSF-derived EVs analysis, but studies of this nature are still lacking for these and other targets or human biofluids.

Saliva is a more easily accessible fluid, also holding potential as a source of biomarkers. Nonetheless, until now few studies used saliva to address physiological or pathological conditions, particularly in the field of neurodegenerative diseases. Sample collection needs urgent standardization since it was reported that saliva collection and composition depend on variables as the location in the oral cavity, the productive salivary gland (parotid, submandibular, sublingual, or minor salivary glands), oral cavity diseases, smoking, drinking, and food intake. Even the time of saliva collection for EVs isolation could constitute a source of variability since saliva composition can change according to the circadian cycle [95, 96]. Complementary studies addressing the impact of non- and stimulated-saliva collection procedures on EVs isolation, as well as the effects of blood contamination either derived from teeth brush or oral cavity wounds, are still needed. Other important aspects that should be considered during EVs isolation from the saliva are the presence of cells and solid contaminants that should be removed by low-speed centrifugation or filtration; samples viscosity, which require the sample dilution in PBS; and the high abundance in immunoglobulins and amylase that can mask the presence of lower abundant proteins. Some strategies used to remove amylase were affinity adsorption [97] and affinity chromatography columns combined with filter systems [98].

Likewise, urine collection involves a non-invasive procedure, it is available in large amounts and, importantly, urine-derived EVs are highly stable. One of the major challenges in exosome isolation from this biofluid is removing the Tamm-Horsfall glycoprotein (THP) or uromodulin which is the most abundant urine protein. This protein can trap or bind EVs [99], leading to EVs co-precipitation at low-speed centrifugations that are usually carried out to remove cell debris and other contaminants. A recent study characterized the first pellet of the differential centrifugation protocol (low speed ≈ 21,000 g) obtained after the remotion of THP by Tris (2-carboxyethyl) phosphine hydrochloride. This pellet, frequently discarded, presented EVs within the size of 40–250 nm with round morphology, and proteomic analysis revealed the presence of exosomal markers. Data support that this pellet obtained at lower centrifugation speed can potentially constitute a source of EVs-related biomarkers [100]. Other methods were used to extract THP, such as combination with various chaotropic reagents aiming its denaturation [101], salt precipitation using NaCl [102], addition of ZnSO4 to promote THP oligomerization and easier sedimentation [103], salting-out CHAPS lysis buffer [99], sucrose gradient, or the single use of dithiothreitol (DTT) to disrupt the cysteine-cysteine interactions [104]. Nevertheless, some protocols were not completely efficient on THP remotion, and agents as DTT could promote protein remodulation with consequences for downstream analysis. Although less abundant than THP, other proteins can constitute important contaminants of urinary EVs preparations, as albumin [105], aquaporin-1 and aquaporin-2 [106] uroplakin, and prokaryotes [107].

Comparative studies on exosome isolation from biofluids

Considering all challenges in EVs isolation from biofluids, the question to be answered lies on the best method to apply in each case. An overview of the literature allowed to identify 55 comparative studies where EVs were isolated from one or more body fluids, using different methodologies (Table 2). The literature search was performed in the PubMed and included two keywords schemes: (1) “((Exosomes) OR (Extracellular vesicles)) AND ((Serum) OR (Plasma) OR (CSF) OR (Saliva) OR (Urine)) AND (Isolation methods comparison)” and (2) “((Exosomes) OR (Extracellular vesicles)) AND ((Serum AND Plasma) OR (Serum AND CSF) OR (Serum AND Saliva) OR (Serum AND Urine) OR (Plasma AND CSF) OR (Plasma AND Saliva) OR (Plasma AND Urine) OR (CSF and Saliva) OR (CSF and Urine) OR (Saliva and Urine)) AND (isolation methods). Review articles and non-comparative studies or comparative studies based on cell culture models or animals were excluded. The data collected will be subsequently discussed. For EVs isolation from CSF, four comparative studies evaluating the different EVs isolation methodologies performance were found, two of which isolated vesicles from CSF and the other two compared EVs isolation from CSF, plasma, and serum (Table 2). In these studies, the CSF starting volumes ranged between 500 µL and 8 mL and UC was compared with SEC, precipitation, ultrafiltration liquid chromatography, membrane affinity, and immunoaffinity methods. UC had the lowest particle yield, compared with precipitation and ultrafiltration methods, as determined by NTA [24, 108] and also by SIMOA, when evaluating the enrichment in exosome markers in UC, SEC, and precipitation methods [58]. Another study compared MagCapture and ExoIntact (two immunoaffinity bead–based methods), ExoEasy (based on membrane affinity), and EVSecond L70 (based on SEC). ExoEasy outperformed the other methods in particle yield and enrichment in CD63 and CD81 but its exosome preparations were enriched in albumin [109]. None of the four studies compared the exosome isolation methodologies regarding total RNA or miRNAs yield.

Table 2 Summary table of comparative studies focusing on exosome isolation methods in human biofluids. Several exosome isolation methodologies were compared using human serum, plasma, cerebrospinal fluid, saliva, and urine. The comparison addressed particle size, yield, exosome markers tested, and purity. *Contamination or negative exosome markers
Full size table

Regarding plasma, information from twenty-four comparative studies was collected, of which nine also included particle isolation from other biofluids (Table 2). Plasma volumes used for exosome isolation ranged from 10 µL to 25 mL, with UC-based methodologies requiring the highest plasma volume, as described for CSF. Ultracentrifugation with OptiPrep density gradient used 25 mL of plasma [121]; however, most of UC protocols used 500 µL or 1 mL of plasma. The lower plasma volume (10 µL) was applied to exosome total isolation chip, based on filtration [37]. Typically, all studies reported vesicles with cup-shaped morphology within the exosomes expected size. Likewise, precipitation and/or SEC methodologies presented higher particle yields when compared with UC [24, 37, 114, 115, 117, 147, 153]. Nonetheless, the combination of UC with the use of discontinuous iodixanol gradient was reported to be useful to increase particle yield and purity [121]. Recent methodologies as the microfluidic device [113], the clustering and scattering method [115], the Exodisc-B centrifugal device [116], and the electrokinetic devices outperformed UC in yield [152]. The recent column-based AppiEV method, using an anionic polysaccharide-modified filter, presented similar particle sizes to UC, but it isolated more EVs than UC, ExoQ, or ExoEasy [111]. To consider, some studies did not assess the purity of EVs isolated which is an important aspect since higher particle yields are no warranty of high-quality of EVs preparations. In general, for plasma-derived EVs isolation, the highest purity was reported for SEC-based methodologies when compared with UC, membrane affinity, and/or precipitation methodologies [24, 26, 58, 119, 120]. Nonetheless, others reported that UC rendered in purest preparations when compared with precipitation-based methods [114, 115].

As mentioned above, few information is depicted in the literature regarding plasma pre-processing variables, but EDTA tubes for blood collection were the most common used tubes [24,25,26, 37, 112, 115,116,117, 120, 148, 149]. Further, RNA yields obtained from plasma-derived exosomes isolated using UC, membrane affinity, and precipitation- and column-based methods were compared in several studies and, in general, similar RNA patterns were obtained for the distinct methodologies [26, 111, 114, 147,148,149, 151]. However, the clustering-and-scattering method was superior in terms of RNA yield and purity than ExoEasy, UC, and ExoQ [115] and exosomes isolated using Exodisc-B had five times more RNA than EVs isolated through UC [116]. Higher amounts of DNA were found in serum-derived exosomes when compared with plasma-derived exosomes [148]. In addition, the three most used markers to characterize plasma-derived EVs preparations were TSG101, CD63, and CD9 and, as negative markers, calnexin, ApoA1, and ApoB.

Serum-derived EVs isolation performance was addressed in twenty-five comparative studies that employ distinct methodologies in serum, or including other biofluids (Table 2). Biofluid starting volumes used vary between 50 µL and 5 mL. UC and precipitation-based methods were the most used isolation methodologies and required the highest starting sample amounts. EVs isolated were compatible with exosome morphology and size but there was a trend for larger vesicles isolated through UC when compared with precipitation and/or SEC-based methods [71, 122, 123, 126, 127, 132] and larger vesicles obtained with membrane affinity–based methods when compared with UC [127, 131], although within the expected size range. Overall, precipitation and/or SEC-based methods isolated more particles than UC [24, 71, 123,124,125,126,127, 130,131,132,133,134,135, 147, 154, 155] and vesicle preparations employing SEC or UC exhibited superior purity than precipitation methods [24, 126, 127, 133]. Interestingly, one study showed that plasma-derived EVs isolated through UC contained more exosomal proteins, identified by mass spectrometry, than serum-derived EVs, suggesting that EVs from plasma present higher purity [150]. Another study reported that despite ExoSpin isolated more particles than UC, this cannot be understood as a higher enrichment in exosomes because in TEM preparations only a few vesicles from ExoSpin immunostained for exosome markers. The enrichment in exosome markers for UC was also confirmed by ELISA assessment of CD9, CD63, and CD81 levels [123]. Recently, a new Extracellular Vesicle Capture by AnTibody of CHoice and Enzymatic Release (EV-CATCHER) method was developed to improve the immuno-based isolation of small EVs, using CD63, CD9, or CD81 antibodies coupled to beads. This selective EV purification technique was compared with other 7 EVs immuno- and 2 column-based methods, ExoQ and UC. EV-CATCHER provided high yield of small EVs and was suitable for high-throughput small-RNA sequencing [125]. Immunomagnetic sequential ultrafiltration (iSUF) is another new EVs purification methodology that combines tangential flow filtration, a centrifuge enrichment, and an immunocapture step. The iSUF was compared with TEI, qEV, and UC and rendered in high yields and pure EVs preparations [122]. This study shows that the combination of distinct methods can be useful in small EVs isolation. Concerning RNA profiling, precipitation-based methods, immunomagnetic sequential ultrafiltration, or nanochips provided higher exosomal miRNA yields than UC [122, 131, 134, 155]; nonetheless, exosome preparations obtained through UC were more pure than ExoQ and contain less free miRNAs than ExoQ [126]. Like for plasma, the most common exosomal markers tested in these preparations were TSG101, CD63, CD81, and CD9 and the negative markers were calnexin, ApoA1, and ApoB.

In sum, for plasma- and serum-derived exosome isolation, most of the studies compared UC, precipitation-based methods, and SEC, suggesting that these are the most commonly used thus far, although novel approaches are arising in the field. In terms of particle yield, precipitation and/or SEC outperformed UC but the purest blood-derived exosome preparations were reported to be obtained through UC and/or SEC. Some studies combined UC and SEC methods in an attempt to improve exosome isolation but consensus still need to be reached in terms of yield or purity.

Five comparative studies focused on EVs isolation from saliva or saliva and other biofluids (Table 2). From these, only three studies detailed the saliva collection conditions, one of them collected unstimulated saliva [152]; other collected the saliva between 9 and 11 am to reduce circadian cycle variations and asked participants to not eat or drink 1 h before collection [136] or wait 30 s after water consumption and before saliva collection [153]. Saliva starting volumes vary between 50 µL and 5 mL and EVs preparations presented nanovesicles within the expected exosomal size range. Similar to the other biofluids, a trend for lower particle yields obtained through UC was observed when compared with precipitation methods or electrokinetic device [136, 137, 152, 154], except in comparison with chitosan, a naturally occurring polymer [153]. However, in terms of EVs purity, UC presented higher purity than the other methods [136, 137]. No differences in RNA yields were described for exosome preparations obtained with UC or PEG-based methods [154]. The characterization of saliva-derived EVs also included the assessment of CD63, CD81, and CD9.

Information was also collected from twelve comparative studies focusing on EVs isolation from urine, from which three also included nanovesicle isolation from other biofluids. The starting volumes ranged from 500 µL to 200 mL and the most used method to remove THP from urine was the addition of DTT [139, 140, 145]. EVs obtained through the distinct methods had exosomal compatible morphology but, in terms of particle yield, controversies arise. Some studies reported that UC isolated more urinary particles than SEC [140], precipitation methods [139, 140, 153], while others reported that SEC [139], but also nanoDLD (microfluidics) and membrane affinity–based methods [155] or EV-TRAP based on functionalized magnetic beads [141], isolated more nanovesicles than UC. Regarding purity, it was reported that EVs preparations obtained through UC had higher purity than precipitation-based methodologies [145] or hydrostatic filtration dialysis (HFD) [138], whereas other studies showed that ultrafiltration [144] or EV-TRAP [141] provides more pure EVs preparations than UC. Distinct results were also found regarding RNA yield. Total EVs RNA yields were similar between UC, HFD, or spin columns [138, 142] while superior miRNA yield was reported for UC when compared with SEC or ExoQ [139]. In addition, lower miRNA or mRNA was obtained for exosome preparations using UC when compared with precipitation-based methods, as Ymir [143] or ExoQ modified protocol [145]. The latter comprises incubation with a higher volume of precipitation reagent (3.3 mL) than recommended in datasheet (2 mL), and a centrifugation at 10,000 g instead of 1500 g. In the comparative studies collected, CD9, TSG101, and CD63 were the most used markers to characterize exosomal preparations nature.

Future comparative studies should focus on the comparison of new methodologies and on the identification of pre-analytical variability sources, thus contributing to more reproducible and standardized approaches.

EVs biobanking

In the era of precision medicine, biomarkers gain increased importance not only for disease diagnosis but also as potential contributors for a personalized therapy. Considering the exosomal potential as sources of biomarkers and therapeutic vehicles, it is of maximum importance to biobank exosomes. In this context, it became of extreme importance to define and standardize the best method for each biofluid and annotate pre-analytical variability sources, as the daily time of biofluid collection, sample collection tubes, time, and temperature storage conditions.

Standardization of EVs isolation workflows and implementation of QC metrics for EVs biobanking would be essential to ensure high quality of the exosomes preparations and increase data reproducibility. A new ISEV Subcommittee on scientific rigor and reproducibility was stablished and a survey carried out in 2019 revealed that only a small percentage of respondents had biobanking EVs, and that most of the biobank users did not perform any kind of sample QC [157]. It is important to implement QC measures, focusing on sample hemolysis, platelet counting, and other blood chemical parameters, and address EVs purity by monitoring the amount of the typical contaminants, as the content in albumin and lipoproteins. Biobanking exosomes and the implementation of good practices and QC measures would accelerate the translation of EVs discoveries to clinical practice.

Conclusion and future perspectives

The unique nature of each human body fluid represents a set of challenges when isolating EVs, among which are nanovesicle abundance, the distinct sample viscosity, and the presence of different co-contaminants. To overcome this, several exosome isolation methodologies have been developed and combined, also to ensure a better balance between EVs yield and purity. Further, diverse pre-analytical and analytical variability sources co-exist and must be properly identified and controlled. Moreover, in clinical settings, the human sample volumes are limited, as well as human and material resources, prompting the choice for methods that can provide high EVs yield, in a cost- and time-effective way. Comparative studies addressing the performance of distinct exosome methodologies, in one biofluid or more, constitute important tools to help the translation of EVs from bench to bedside, and these have been summarized in Table 2. For the body fluids included in this review, UC, precipitation-based, and SEC-based methods were the most commonly used, exhibiting a good performance in terms of balance between EVs yield and purity (Fig. 2).

Fig. 2
figure 2

Methods for exosome isolation in distinct body fluids. Several exosome isolation methodologies for human body fluids have been developed (left). These were addressed in comparative studies, as described in Table 2. For each biofluid, the volumes range used in these studies as well as the most used methods were represented on right. Exosomes obtained from human peripheral biofluids can constitute useful diagnostic and/or therapeutic tools. Abbreviations: CSF, cerebrospinal fluid; IA, immunoaffinity; MA, membrane affinity; MFI, microfluidics; PP, precipitation-based method; SEC, size exclusion chromatography; UC, ultracentrifugation; UF, ultrafiltration

Full size image

A unique exosome isolation method suitable for all biofluids would be the ideal solution, but it is unexpected. Instead, simplified, high yield, relatively pure, and high-throughput new techniques will certainly arise considering the distinct biofluid biochemical properties. Also, new EVs characterization approaches will contribute to better distinguish EVs preparations with exosome-like characteristics from contaminants. This will facilitate the choice of the most suitable exosome isolation method for each biofluid. In addition, establishment of standardized procedures will be fundamental to improve exosome preparation reproducibility and quality. The ideal exosome isolation method would comprise an automatized platform allowing the extraction, characterization, and analysis of EVs content in a simple way. Hence, efficient, reproducible, and standardized EVs isolation procedures will facilitate the use of exosomes as sources of biomarkers or as therapeutic vehicles, opening avenues for its potential application in translational and clinical medicine.