Capture and elution fidelity
As with all forms of biomolecule/particle isolation, a successful SPE spin-down methodology for exosome/EV isolation and recovery must provide not only for separation, but must do so without compromising the physical and biological attributes of the EVs. In this case, EVs must be isolated with respect to the components of the sample matrix, including salts, small molecules, such as amino acids, sugars, proteins, and genetic material. Previous reports have illustrated this capability via HIC separation of exosomes from diverse media [40,41,42]. In the case of the spin-down tip processing, the integrity of the physical and biological attributes of the exosomes was evaluated via SEM and immunofluorescence, respectively, following the elution steps to remove salts and adventitious proteins. In Fig. 2a (for the case of the commercial exosomes dispersed in water), the surface of the C-CP fibers at this stage is pristine, as indicated by the presence of globular vesicles without any remnants of salt crystals or the like.
To further illustrate the integrity of the captured exosomes, super-resolution confocal microscopy imaging was performed. Exosomes captured on C-CP fiber surfaces were immunolabeled using a primary antibody to the tetraspanin surface marker protein, CD81, and a fluorescent secondary antibody (AlexaFluor 647 goat anti-mouse). As seen in Fig. 2b, there are dispersed nanobodies (of the size range expected for the target exosomes) within the ~ 25 × 25 μm2 viewing region. Due to the resolution limits of the confocal microscope (~ 140 nm), it is important to note that fluorescent particles observed here are not necessarily individual exosomes, but perhaps small aggregates producing a more intense fluorescent response. Nevertheless, with regard to capture, the target exosomes are well dispersed on the fiber surface (without substantial debris), while maintaining their basic physical morphology and surface protein makeup. Indeed, the characteristics depicted in Fig. 2 are the first steps towards affecting a practical exosome diagnostic platform.
In those cases where further exosome characterization is required, such as in the search for surface biomarkers or genetic analysis (e.g., RNA-Seq) of the vesicular cargo, the organelles must be recovered (eluted) while maintaining their physical integrity and biological function. The most common method for assessment of the morphology of individual exomes is transmission electron microscopy, where both the size and vesicular structure are revealed. The TEM micrograph of an HIC-eluted EV (Fig. 3a) illustrates the successful maintenance of the physical structure through the isolation process. The biological fidelity of exosome populations is readily assessed through the use of dot blot assays (Fig. 3b), wherein a positive immune-response is obtained for the CD9, CD81, and CD63 antibodies in the post-tip isolation eluates. As seen in the various exposures, the recovered exosomes do indeed retain the surface markers of the three tetraspanin proteins confirming the presence and viability of the exosomes. While the dot blots do not reflect the retention of the encapsulated genetic materials, they suggest that the expected membrane-bound proteins remain intact.
Dynamic binding capacity (DBC)
The ability to effectively isolate and purify EVs is only relevant to the extent that it yields the required density of EVs necessary to provide meaningful sample data. As a general rule, most RNA-sequencing analyses require 109–10 exosomes for accurate profiling, while LC/MS proteomic studies require on the order of 1010–11 exosomes [53,54,55,56,57,58,59] To this end, the dynamic binding capacity of the 1 cm C-CP fiber spin-down tips was determined (Fig. 4). Unlike in the case of continuous processes , a frontal analysis was required. This was performed using discrete 50 μL aliquots of test solutions (exosomes in 25% glycerol/1 M (NH4)2 SO4), with the pass-through exosome content used to assess breakthrough/overload. Figure 4 shows the determined absorbance values, obtained by diluting 3 μL of the eluate in DI-H2O in a 1.5-mL cuvette. The absorbance response is not significant until aliquot #14, wherein the pass-through content increases rapidly, and a plateau is reached beyond aliquot #16, suggestive of surface saturation. Based on the general response, the eluate absorbance reaches one half of the steady maximum value (a measure of reaching DBC) with aliquot #15. At this point, based on per-aliquot particle densities of 4.65 × 1010, a DBC value on the order of ~ 7 × 1011 is achieved for a total volume of 750 μL. Though there is no consensus regarding a “healthy range” for exosome concentration, this value is in line with that expected in many native biofluids, including urine, milk, serum, and plasma. The capacity demonstrated at this early stage is on-par for what would be desired in the clinical and biochemical laboratory arenas.
In previous EV separations employing PET C-CP fiber columns in a Dionex Ultimate 3000 HPLC system (ThermoFisher Scientific, Waltham, MA, USA), UV-Vis absorbance at 203, 216, and 280 nm was used as a method for EV detection . Even with the well-known optical absorbance of some buffer/matrix components at these wavelengths, a successful method of exosome isolation should alleviate their contributions and allow ready quantification. The absorbance response observed in this instance is not due to the molecular absorption of an innate biomolecule, but rather corresponds to the light scattering due to the presence of the particles. Ultimately, the absorbance response was found to be directly proportional to the exosome content, for particles of different sources. As most methods of EV isolation carry along remnant proteins, there is a potential that the absorbance (scattering)-based measurement could be affected by their presence.
Based on the fact that the extent of scattering would be (nominally) inversely related to the incident wavelength and that proteins (being composed of aromatic amino acids) absorb at 280 nm, response functions were prepared at 203, 216, and 280 nm. Lyophilized exosome standards from the urine of reportedly healthy donors (previously shown to have latent proteins present) were used to create standard curves. Here, 1–35 μL of the exosome standards (2.3 × 1012 particles mL−1) were diluted to 1.0 mL in DI-H2O, presenting a concentration range of ~ 2.3 × 109–8.0 × 1010 particles mL−1 (Table 1). The slope of the 280 nm function is approximately 40% higher than the lower wavelengths. The stronger absorbance at 280 nm reflects the inevitable presence of proteins (which contain aromatic amino acids) in the commercial exosome material. Indeed, the characteristics for the lower wavelengths are virtually identical, with much better regression statistics than at the higher wavelength. Based on these figures of merit, and fewer contributions from background proteins, the shorter wavelengths are preferred. While the limits of detection and quantification are not as low as with other methods (e.g., immunoassays) [60,61,62,63,64,65], the values are relevant for most biological and clinical systems of interest, particularly in consideration of the total sample volume required (< 50 μL) and ease of determination.
Isolation and quantification of EVs/exosomes in diverse media
As proof of concept towards the efficacy of the HIC spin-down tip approach to exosome isolation and quantification, the commercial exosome standards (2.73 × 1012 particles mL−1) were spiked into DI-H2O, mock urine, reconstituted non-fat milk, and exosome-depleted fetal bovine serum (FBS) matrices. Two dilution factors were employed (1/100; 1/1000), as a quantitative test of the response, as well as tolerance towards the challenges of the matrices themselves. The matrices were mixed (50:50 v/v) with the HIC loading solvent (2 M (NH4)2SO4) in PBS. While diH2O presents a pristine environment, the mock urine matrix presents high salinity and is small molecule-heavy, the milk has high protein content, and the FBS contains fat and high-protein content. These model biofluids are obvious target matrices from which exosome/EVs may be extracted for diagnostic purposes. In terms of loading and elution, the procedure involved a spin-down under high-salt conditions, followed by elution of proteins with 50 μL of 25% glycerol and 1 M (NH4)2SO4 in PBS. This fraction was collected for absorbance measurements of protein/exosome content. Finally, the EV fraction was eluted in 50 μL of 50% glycerol in PBS and collected for the determination of vesicle content. Though glycerol has been used as a biological preservative , it is not ideal for all downstream analyses (i.e., proteomic analysis) where necessary vesicle lysing may be hindered. In these cases, acetonitrile may be used as a substitute elution phase, as previously reported [42, 43].
Essential to the quantification process of EVs in different matrices is the assumption that EVs may be quantitatively immobilized and recovered from the fiber surfaces. The latter point has been evaluated in recent studies using the chromatographic (column) platform, wherein recoveries of adsorbed EVs were greater than 80% . Parallel evaluation of the recoveries was performed here via UV absorbance (using the previously generated aqueous matrix calibration functions) and an ELISA assay. The determined numbers of EV particles for the two dilution factors, as determined via optical absorbance (203 nm), are presented in Fig. 5. Aliquots (50 μL) for both the protein and exosome elution fractions were diluted to 1 mL for the absorbance measurements. Starting with the lowest (1/100) dilution factor, no absorbance response is seen in the protein fractions for aqueous and mock urine phases, but there is a measurable absorbance, equivalent to 5.3 × 109 and 2.4 × 1010 EVs, for the milk and exosome-depleted FBS matrices, respectively. These respective responses are not surprising, as the latter matrices have appreciable protein content and corresponding appreciable absorbance, while the aqueous and urine matrices do not. On the other hand, absorbance measurements taken of the presumed EV fraction yield statistically identical values for the aqueous, mock urine, and non-fat milk matrices, as they would be expected. Interestingly, a much higher (~ 2×) recovery of EVs was observed in the exosome-depleted FBS exosome elution fraction. The precision of triplicate measurements for each of the matrices was better than 8.4% RSD.
For the case of the higher dilution factor (1/1000), it would be expected that the recoveries would be proportionally (~ 10×) less, but potential matrix effects would be lessened as well. Here, the responses for the protein elution fractions for the aqueous, mock urine, and non-fat milk matrices fall below the level for accurate determination. The FBS protein elution still shows a measurable absorbance response, equivalent to 3.2 × 108 EVs. This is to be expected with the high concentration of total protein in the original matrix. The greater than expected decrease in apparent concentration is due to lessened amounts of protein aggregation in the more dilute solution. That noted, the determined concentrations in the respective EV fractions are indeed ~ 10× less than the more concentrated case for all matrices. Here again, a high level of precision in the EV recovery is obtained (< 6.9% RSD), with the determined particle numbers across the first three matrices being virtually the same, and a significantly higher exosome recovery again for the exosome-depleted FBS matrix. Thus, based on the absorbance-based quantification method, there is no significant difference in EV recoveries across the diverse aqueous, mock urine, and non-fat dry milk matrices. More importantly, the fractional recoveries for the two dilutions are approximately 75% versus the initial number of EVs applied to each tip for these matrices. This value reflects a significantly more efficient recovery of exosomes when compared to the fractional recoveries of other methods, such as ultracentrifugation, which results in equal or lesser concentrated recoveries of exosomes, though requiring nearly 90 times the starting sample volume. As previously mentioned, a significant increase in recovery was observed from the FBS matrix. Marketed as an “exosome-depleted” FBS source, the manufacturer claims the depletion of 90% or more of native endogenous exosomes. The increase in EV recovery for the FBS matrix may be due to remnant exosomes from the native FBS matrix (known to contain high concentrations of EVs).
To verify and quantify the presence of remnant (native) extracellular vesicles in the exosome-depleted FBS matrix, the tip isolation of exosomes was performed on an exosome-spiked aqueous solution, the exosome-depleted FBS, and the FBS spiked with exosome standard. In the spiked-solution cases, the primary stock solution was added at a 1:100-μL ratio to the matrix. The optical absorbance of the eluate was detected at the 203 nm wavelength and used to quantify the exosomes based on the previous aqueous solution calibration function. Figure 6a shows the resulting exosome concentrations, where approximately the same number of exosomes were quantified in the eluates from the aqueous and native exosome-depleted FBS solutions. Addition of the spike to the FBS yields an ~ 73% increase in the determined density, a value in line with a combination of the responses for the aqueous solution and the FBS sample, as would be expected as the spike values are the same for the first and third cases. Importantly, the levels of precision are very uniform ranging from 5.4–8.2% RSD for triplicate isolation and measurement sets. Based on the determinations performed here, the exosome concentration in the “depleted” FBS is approximately 1.5 × 1010 particles mL−1. This value is less than recently published values of 2.27–2.93 × 1011 particles mL−1 . Based on those values, the material employed here meets the stated 90% clearance target stated by the manufacturer, with ~ 6.6% remaining based on the published values.
The presence of exosomes in the depleted FBS was further confirmed physically using STEM and nanoparticle tracking analysis. Figures 6b–d are micrographs of the exosome eluted fractions for the same three cases, exosomes spiked (1:100) in aqueous solution, the native FBS, and exosomes spiked into FBS. In all three cases, the typical halo-structure objects are clearly revealed, having diameters on the order of 80–120 nm. NTA analysis was performed to analyze the size distribution of the eluted exosome populations. The graphical size distributions of the eluted exosomes are presented in ESM. Statistically, larger numbers of exosomes are observed in the case of the spiked-FBS (as suggested in the data of Fig. 5), though the means (~ 96 nm) and modes (74.3 and 77.7 nm) of the distributions are very similar. What are quite different are the broader distribution aspects, where the spiked-FBS displays a D90 (upper limit inclusive of 90% of the population) of 155.4 nm while the spiked aqueous population exhibits a D90 of 130.1 nm. This relationship is not surprising as the FBS is a far more diverse matrix than the human urine originating spike matrix.
As a complement to the use of absorbance spectrophotometry to perform quantification, spin-down tip recoveries were also assessed via a standard ELISA assay for the antibody response to the CD81 tetraspanin surface protein. Presented in Fig. 7 are the determined number of particles is reported for the same two dilution values (1/100 and 1/1000) as depicted in Fig. 5 aqueous, mock urine, and non-fat milk test matrices. (The FBS material was received after the University ELISA facilities were closed due to COVID-19 protocols and so were not part of this assay.) The determinations were made on the same collected EV elution fractions as used in the absorbance measurements. As reported for the 1/100 dilution samples, the numbers of collected EVs are statistically identical for the three different matrices. As expected, the level of precision of the bioassay is somewhat degraded from the absorbance measurements, but with a variability of < 9.1% RSD, the results are in line with what would be expected. With increased dilution, the number of particles is statistically lower, with similar repeatability, but not in the direct proportion seen in the absorbance case. Again, the recoveries across the matrices are similar, maintaining the same relative responses among each. The measured CD81 expression reflects the fact that the exosome biogenesis process, and therefore, surface protein expression is due to many stochastic processes. Though exosomes from identical cells may be produced via the same mechanisms, exosome populations are heterogeneous, and differences in protein expression are expected. Also, while glycerol in the elution buffer was used to increase exosome stability and prevent aggregation, the presence of glycerol may also have an effect on the conformation of exosome surface proteins in the eluate. Changes in protein conformation due to the presence of glycerol has been previously reported [68, 69], where proteins are altered to more compact states. This has been found to affect antigen–antibody binding interactions, specifically in ELISA applications . The observation of non-linear quantification of exosomes seen in Fig. 7, when compared to absorbance-based results in Fig. 5, is most likely due to these effects.