Porous Membrane Adsorption for Capturing Small Extracellular Vesicles Enables Accessible Single-Particle Analysis

Open figure viewer Quantitative analysis of small extracellular vesicles (sEVs) at single-particle resolution remains challenging due to their nanoscale dimensions and compositional heterogeneity. Existing methods often rely on specialized and costly instrumentation, limiting accessibility for many researchers and necessitating extensive optimization. Here, we present a porous membrane adsorption for capturing small vesicles (PMACS), which employs a porous anodized aluminum oxide (AAO) membrane as nanoscale compartments to spatially isolate individual vesicles. This allows accessible single-particle discrimination using standard wide-field fluorescence microscopy without either advanced equipment or expert skills. Single-particle discrimination is achieved based on the principle that AAO membranes accommodate vesicles at most one particle per pore according to Poisson distribution and that occupied pores are sufficiently separated beyond the optical resolution limit, which ensures reliable attribution of fluorescent spots to individual vesicles. PMACS achieves reliable quantification of culture cell-derived sEVs across a broad dynamic range (10 3 to 10 6 particles/μL). This performance exceeds conventional methods without the need for either advanced instrumentation or expert operation. We demonstrate that PMACS allows colocalization analysis of sEV cargos by assessing the tetraspanin (CD9, CD63, and CD81) population of cultured cell-derived sEVs (Hela, A549, and HEK293T cell lines). The heterogeneous protein expressions on sEVs are visualized by means of a simple facile setup. Owing to its simplicity, sensitivity, and accessibility, PMACS provides a practical platform to advance studies of sEV heterogeneity and broaden participation in sEV science. Overview of the sEV detection by PMACS (porous membrane adsorption for capturing small vesicles). The general PMACS workflow is as follows: First, the fluorescently labeled sEVs are incubated with the AAO membranes under gentle agitation. In this step, sEVs are trapped in the pores of the AAO membranes (adsorption), which spatially isolate individual sEVs while accumulating sEV concentrations compared with bulk concentrations. Next, AAO membranes are washed with buffer (typically, PBS) to remove untrapped sEVs and surface-bound sEVs. Subsequently, fluorescence images of sEV-trapped AAO membranes are acquired with wide-field fluorescence microscopy without any advanced equipment such as super-resolution modules. Fluorescent spots can be regarded as individual sEVs due to spatial isolation of individual sEVs by nanoscale compartments of AAO pores. For the detailed principle, see the Principle of Single-Particle Resolution subsection. Overview of the sEV detection by PMACS (porous membrane adsorption for capturing small vesicles). The general PMACS workflow is as follows: First, the fluorescently labeled sEVs are incubated with the AAO membranes under gentle agitation. In this step, sEVs are trapped in the pores of the AAO membranes (adsorption), which spatially isolate individual sEVs while accumulating sEV concentrations compared with bulk concentrations. Next, AAO membranes are washed with buffer (typically, PBS) to remove untrapped sEVs and surface-bound sEVs. Subsequently, fluorescence images of sEV-trapped AAO membranes are acquired with wide-field fluorescence microscopy without any advanced equipment such as super-resolution modules. Fluorescent spots can be regarded as individual sEVs due to spatial isolation of individual sEVs by nanoscale compartments of AAO pores. For the detailed principle, see the Principle of Single-Particle Resolution subsection. To overcome these technical barriers, we explored a structurally engineered material that could spatially isolate individual vesicles for optical analysis. In this work, we employed a porous anodized aluminum oxide (AAO) membrane to capture sEVs with spatially controlled confinement. AAO is a well-characterized material that possesses highly ordered, uniform, and linear pores that can be precisely controlled by the anodization process. ( 25 ) This unique structural feature renders AAO utilized for confining biomolecules such as proteins, ( 26 , 27 ) nucleic acids, ( 28 , 29 ) and even lipid vesicles ( 30 ) to develop analytical platforms. ( 31 ) Our previous work revealed that AAO pores serve as scaffolds for immobilizing sEV-mimicking liposomes while maintaining their integrity, which implies the possibility of capturing sEVs within the pores. ( 30 ) Building upon this concept, we developed and validated a novel method for single-particle analysis of sEVs using commercially available AAO with a pore diameter of approximately 200 nm, which closely matches the sEV diameter ( Figure 1 ). AAO pores trap individual sEVs at spatially separated intervals as if nanoscale compartments accommodating individual vesicles. This avoids the difficulties in single-particle recognition caused by the signal bleeding and overlapping from neighboring vesicles. This approach, which we call PMACS (porous membrane adsorption for capturing small vesicles), enables the optical single-particle discrimination of sEVs using standard wide-field fluorescence microscopy. This readily available instrument requires no specialized equipment or expert skills, thereby providing an accessible platform for investigating sEV heterogeneity. To address this limitation, single-particle analysis has emerged as a powerful strategy to visualize the heterogeneous population of sEVs. However, analyzing sEVs at single-particle resolution is generally challenging because their tiny size falls below the optical resolution limit. ( 17 ) Standard optical techniques, like wide-field fluorescence microscopy, are thus limited by signal bleeding and overlapping, which prevents the distinction of individual particles. Several techniques have recently been developed to enable single-particle analysis. Commercially available technologies include nanoparticle tracking analysis (NTA), nanoflow cytometry (nFCM), and single-particle interferometric reflectance imaging sensor (SP-IRIS). NTA tracks Brownian motion of each particle through time-sequential imaging of scattered light to calculate size distributions and can also analyze cargo colocalization by mounting a fluorescence pass filter. ( 18 ) nFCM is an advanced flow cytometry technique equipped with a high-performance detector capable of sensing particles smaller than 100 nm, beyond the size limit of conventional flow cytometry. ( 19 ) SP-IRIS detects sEVs immobilized on antibody-coated chips with interferometric reflectance imaging. ( 20 , 21 ) Recent studies have also proposed other techniques for analyzing sEVs' heterogeneity, including proximity barcoding, ( 22 ) digital ELISA, ( 23 ) and super-resolution microscopy. ( 24 ) These techniques are highly valuable for studying sEVs' heterogeneity because they enable the precise characterization of individual particles. However, they require highly specialized instrumentations for analyzing sEVs that are often expensive and/or demand expert knowledge and skills for operation, making them inaccessible to researchers outside highly dedicated laboratories to sEV science. Moreover, these techniques provide reliable analytical results only within a narrow concentration range (vide infra), necessitating tedious optimization procedures. Such a situation has restricted the widespread adoption of single-particle analysis in general biological research. Therefore, an easy-to-use and accessible method for single-particle analysis is urgently needed. The cargos of sEVs, such as miRNAs and proteins, reflect the metabolic states of the secreting cells. Thus, analytical techniques for cargos are crucial both for understanding the fundamental mechanisms underlying their pathological roles and for developing clinical diagnostic tools. Basic techniques for sEV cargo analysis rely on means traditionally used for proteins and nucleic acids, such as ELISA, Western blotting, and RNA-seq. These techniques generally display the average value of all particles in the sample but fail to capture the heterogeneous population of sEVs, hampering a precise understanding of their biological roles and clinical potential. sEVs, mainly composed of exosomes and ectosomes, have been intensively investigated since the discovery in 2007 that exosomes mediate intercellular communication by delivering their cargo. ( 9 ) Over the past two decades, it has been revealed that sEVs are involved in cancer progression, ( 10 , 11 ) infection processes, ( 12 ) neurodegeneration, ( 7 , 13 ) and wound healing. ( 14 ) Owing to these associations with key biological events, sEVs hold great potential as biomarkers for these diseases and as drugs and drug delivery systems. ( 15 , 16 ) While diagnostic, prognostic, and therapeutic applications have been widely pursued, fundamental knowledge including their origins, metabolic fates, and relationships between sEVs and their cargo remains limited, despite its importance for establishing practical applications. Almost all eukaryotic cells secrete lipid bilayer-enclosed particles called extracellular vesicles (EVs), which package proteins, nucleic acids, and other metabolites within them. ( 1−3 ) Based on their biogenesis, EVs are broadly categorized into several subgroups, including exosomes, which originate from multivesicular bodies, and ectosomes (microvesicles), which bud directly from the plasma membrane. Other subtypes, such as apoptotic bodies and recently discovered migrasomes ( 3−5 ) and mitovesicles, ( 6 ) as well as nonvesicular particles like exomeres ( 3 , 7 ) have also been reported. Although these vesicles differ in origin and composition, no unique molecular marker has been identified that definitively distinguishes one subtype from another. Consequently, a size-dependent nomenclature has been used unless the biogenesis pathway is specifically characterized: small EVs (sEVs; <200 nm) and large EVs (lEVs; >200 nm). ( 3 , 8 ) Conventional optical sEV detection methods, such as imaging of sEVs immobilized on antibody-coated glass slides, frequently suffer from poor sensitivity and limited quantitativity due to variability in surface preparation. In contrast, the highly ordered and intrinsically less variable nanostructure of AAO contributes to accurate and reproducible detection. Consequently, PMACS exhibits high sensitivity and wide quantitativity without requiring specialized instrumentation or advanced technical skills, offering a user-friendly platform that outperforms conventional methods in both sensitivity and quantitativity. This high performance of PMACS for the model vesicles appears to be based on the highly ordered nanostructure of AAO allowing efficient accumulation of vesicles within the pores, along with the strong correlation with input liposome concentrations. The approximately 70-fold accumulation of vesicles within AAO pores relative to the bulk suspension was estimated from adsorption isotherm measurement ( Figure S5 and Table S3 ), which likely underlies the high sensitivity achieved by the PMACS method. Proof of concept of PMACS using model vesicles (Rhd-liposomes). (a) Schematic diagram displaying Rhd-liposome detection procedure by PMACS using AAO membranes with 200 nm pores. Rhd-liposomes with a diameter of 100 nm were used as a model vesicle mimicking sEVs. (b) Representative fluorescence images of Rhd-liposomes at different concentrations. (c) Quantification of fluorescent spots on images vs the input concentrations of Rhd-liposomes, fitted with y = ax b (R 2 = 0.9996). The liposome diameter and the AAO pore diameter were approximately 100 and 200 nm, respectively. Error bars indicate standard deviations of five replicates. Proof of concept of PMACS using model vesicles (Rhd-liposomes). (a) Schematic diagram displaying Rhd-liposome detection procedure by PMACS using AAO membranes with 200 nm pores. Rhd-liposomes with a diameter of 100 nm were used as a model vesicle mimicking sEVs. (b) Representative fluorescence images of Rhd-liposomes at different concentrations. (c) Quantification of fluorescent spots on images vs the input concentrations of Rhd-liposomes, fitted with y = ax b (R 2 = 0.9996). The liposome diameter and the AAO pore diameter were approximately 100 and 200 nm, respectively. Error bars indicate standard deviations of five replicates. As shown in Figure 2 a, the PMACS workflow composes of three facile steps: (1) incubation of vesicles with AAO membranes to trap vesicles within the pores, (2) a washing step to remove untrapped and surface-bound vesicles, and (3) fluorescence imaging with a wide-field fluorescence microscope. Following this procedure, the AAO membranes were incubated with serial dilutions of Rhd-liposomes, and the resulting fluorescence images were then acquired. A representative fluorescence image of liposome-adsorbed AAO displays numerous fluorescent spots ( Figure 2 b). Significantly, the density of the observed spots correlates with the input liposome concentration. The detected fluorescent spots showed a quantitative relationship on a logarithmic scale ranging from 10 2 to 10 5 particles/μL ( Figure 2 c). It is noted that above this range, excessive fluorescent spots hindered quantitative counting due to signal overlapping. Under our experimental conditions, PMACS was able to detect concentrations as little as 8.91 × 10 1 particles/μL of Rhd-liposomes. Concentrations more diluted than this did not differ significantly from the blank sample ( Figure S4 ). These results clearly indicate that PMACS holds potential for sEV quantification with notably high sensitivity. As a reference, NTA, one of the most widely used sEV quantification techniques, limits its detection range from 10 4 to 10 6 particles/μL even when detecting ideal nanoparticles like polystyrene beads (comparison with other conventional methods will be discussed later). ( 18 , 32 ) To experimentally demonstrate the proof of concept for PMACS (porous membrane adsorption for capturing small vesicles), we first used fluorophore (Lissamine rhodamine B)-labeled liposomes with an average diameter of approximately 100 nm as model vesicles (Rhd-liposomes). The liposomes were prepared by manual extrusion, and the diameter and concentration were determined by NTA ( Figure S1 and Table S1 ). The AAO membranes with a pore diameter of approximately 200 nm purchased from Cytiva (Tokyo, Japan) were characterized by scanning electron microscopy (SEM), indicating the average pore diameter of 212 nm ( Figure S2 and Table S2 ). The liposomes were loaded into the AAO pores by incubation with AAO membranes under gentle agitation at room temperature. The fluorescence intensity of Rhd-liposome suspension decreased over the incubation time, which indicates that the liposomes were adsorbed within the AAO pores ( Figure S3 ). According to our previous work, ( 30 ) liposome adsorption is an electrostatically unfavorable process because of the negative surface charge of both liposomes and AAO pores, and a relatively weak interaction is indicated to contribute to the adsorption (due to a Gibbs energy change (ΔG°) of approximately −25 kJ/mol). Although we have not achieved a definitive conclusion, we suppose that liposomes are trapped via physical adsorption mainly driven by van der Waals interaction. In Figure S3 , it is noted that the time course of the fluorescence change showed that the adsorption stabilized after 10 h of incubation, suggesting that equilibrium had been reached. To ensure complete equilibration, all subsequent measurements were performed after at least 18 h of incubation. Subsequently, the size distribution of fluorescent spots on AAO was measured. The average fluorescent spot size (225 ± 9.6 nm) was larger than the average vesicle size determined by NTA (108 ± 1.0 nm) but was comparable to the pore size of AAO as measured by SEM imaging (212 ± 37 nm). The overlaid distributions of fluorescent spot size and AAO pore diameter showed strong agreement ( Figure S7 ), indicating that each fluorescent spot on AAO corresponds to a particle-occupied pore rather than reflecting the particle itself. This likely results from multiple reflections of emitted light within the internal surface of the pores, causing each pore to appear illuminated. Taken together, these observations suggest that individual vesicles can be distinguished by standard wide-field fluorescence microscopy as spatially separated illuminated pores. To verify these theoretical considerations, the nearest-neighbor distance (NND) was measured for each fluorescent spot by using super-resolution microscopy (AX-R with NSPARC, Nikon, Tokyo, Japan) that achieves a resolution of approximately 100 nm. ( 34 ) By compiling NNDs for a total of 160 fluorescent spot coordinates ( Figure 3 d), the mean and the minimum distances were found to be 5.63 and 1.04 μm, respectively; in line with the theory, the distances between the neighboring spots were far greater than the resolution limit of microscopy (≈350 nm), elucidating that spatial separation of vesicles was successfully achieved. It is noted that NNDs were calculated from the center-to-center distance among fluorescent spots, thereby making them slightly larger than the distances for the edge-to-edge. Despite this overevaluation of NNDs, the size of fluorescent spots remained approximately 225 nm (see the next paragraph), which clearly guarantees that the edge-to-edge distances are greater than 350 nm. Additionally, to elaborately validate the experimental NND analysis against our theoretical consideration, we simulated the NND distribution based on the pore occupancy statistics. The probability distribution of NNDs was obtained by the procedure described in Supporting Discussion 1 (cf. the Supporting Information file) derived from a simple mathematical premise without considering any interference like steric hindrance or electrostatic repulsion. Briefly, assuming random particle capture following Poisson statistics, the probability of particles entering kth neighboring layer pores was first calculated ( Figure 3 e), where k represents the concentric honeycomb-like pore layer counted from a pore occupied by a particle (cf. inset, Figure 3 e). Each kth layer corresponds to a specific radial distance from the reference pore, which was then converted into the nearest-neighbor distance (NND) distribution. As shown in Figure 3 f, the simulated distribution of NNDs displayed closely matched the experimental distribution obtained by super-resolution microscopy ( Figure 3 d). This agreement strongly supports the validity of the theoretical model and confirms that the observed particle distribution is governed by stochastic pore occupancy. Herewith, the consistency with experimental NND measurements and simulation elucidated validity of the theoretical consideration assuming the Poisson distribution aforementioned, thereby inductively supporting single-particle discrimination. The calculated value of P ii for each vesicle concentration is listed in Figure 3 c. Even at the highest concentration tested (10 6 particles/μL), the probability that two or more particles were closer than 350 nm was less than 0.5%. This indicates that over 99.5% of particles were separated by more than 350 nm at concentrations below 10 6 particles/μL. Therefore, wide-field microscopy is capable of distinguishing individual particles under these conditions. Principle of single-particle discrimination. (a) Schematic illustration indicating honeycomb-structured AAO pores containing 0, 1, and 2 particles. The probability that n particles enter in a single pore can be estimated by the Poisson distribution. Note that this illustration does not indicate the actual calculated distribution to visualize simply. (b) Schematic illustration of pores containing particles (magenta) and pores closer than 350 nm (yellow). For optical discrimination of neighboring fluorescent spots, particles need to be located apart by more than the optical resolution limit, d (≈350 nm), from others. In this figure, when the center pore (magenta) is occupied by a particle, neighboring particles should not enter the yellow region (excluded region). (c) Table summarizing the percent probabilities that n particles enter a single pore (P i (X = n)) and that neighboring particle-containing pores are located closer than 350 nm (P ii ). C 0 is the input concentration of vesicles [particles/μL], and Q ad(s) is the quantity of vesicles adsorbed in the surface region of the AAO membranes [particles]. (d) Distribution of nearest-neighbor distances [μm] (NNDs) of fluorescent spots observed by super-resolution microscopy. (e, f) Simulated probability that particles enter pores on the kth layer, where the layer is counted from the nearest particle-entered pore (inset) (e), and that plotted vs NNDs, by converting k into distance (f). Principle of single-particle discrimination. (a) Schematic illustration indicating honeycomb-structured AAO pores containing 0, 1, and 2 particles. The probability that n particles enter in a single pore can be estimated by the Poisson distribution. Note that this illustration does not indicate the actual calculated distribution to visualize simply. (b) Schematic illustration of pores containing particles (magenta) and pores closer than 350 nm (yellow). For optical discrimination of neighboring fluorescent spots, particles need to be located apart by more than the optical resolution limit, d (≈350 nm), from others. In this figure, when the center pore (magenta) is occupied by a particle, neighboring particles should not enter the yellow region (excluded region). (c) Table summarizing the percent probabilities that n particles enter a single pore (P i (X = n)) and that neighboring particle-containing pores are located closer than 350 nm (P ii ). C 0 is the input concentration of vesicles [particles/μL], and Q ad(s) is the quantity of vesicles adsorbed in the surface region of the AAO membranes [particles]. (d) Distribution of nearest-neighbor distances [μm] (NNDs) of fluorescent spots observed by super-resolution microscopy. (e, f) Simulated probability that particles enter pores on the kth layer, where the layer is counted from the nearest particle-entered pore (inset) (e), and that plotted vs NNDs, by converting k into distance (f). Based on the above results, sensitivity and quantitativity of PMACS demonstrated comparable to or exceed those of commonly used techniques for sEV detection and quantification. According to previously published literature, reliable quantification ranges of ELISA, NTA, and SP-IRIS are typically reported as 10 5 to 10 7 particles/μL, ( 35−38 ) 10 4 to 10 6 particles/μL, ( 18 , 32 ) and 10 4 to 10 6 particles/μL, ( 39 , 40 ) respectively ( Table S4 ). Although direct cross-validation is required to provide rigorous comparison in performance benchmarking, PMACS holds great advantages to both conventional single-particle analysis techniques and bulk-based techniques in terms of operational simplicity. Accordingly, PMACS can be a practical and versatile tool for sEV detection and quantification. Overall, PMACS has great potential for analyzing sEVs at single-particle resolution with high sensitivity and a wide dynamic range. Next, we validated the versatility of the method for analyzing sEVs by independently labeling other surface marker proteins, CD9 and CD81, with fluorophore-conjugated antibodies (AF488-CD9 and AF594-CD81). As expected, fluorescent spots were observed on the AAO membranes for both AF488-CD9 and AF594-CD81 ( Figure 4 d,f). This is consistent with the Western blotting analysis, which confirmed that the tested HEK293T-sEV samples contained CD63-, CD9-, and CD81-positive particles ( Figure S8 ). These results showed that PMACS can indeed detect positive signals of various sEV cargos. The quantitative analysis for each marker protein depicted a linear relationship on a log−log plot ( Figure 4 e,g), which strongly suggests that PMACS is versatilely applicable for immunostaining-based quantification of sEVs. In our experimental conditions, AF488-CD9 displayed a similar tendency to AF647-CD63, exhibiting comparably bright fluorescent spots and covering a wide dynamic range (3.71 × 10 3 to 3.80 × 10 6 particles/μL, R 2 = 0.993). Meanwhile, we found a relatively modest brightness and a narrower dynamic range (5.94 × 10 4 to 3.80 × 10 6 particles/μL, R 2 = 0.996) for AF594-CD81. This difference is likely attributable to inefficient fluorescent labeling with AF594-CD81, which could stem from poorer antibody recognition and/or lower CD81 expression on the sEV surface. Detection of HEK293T-derived sEVs (HEK293T-sEVs) by PMACS. (a) Schematic diagram illustrating the experimental procedure of antibody-labeled HEK293T-sEVs by PMACS. The sEVs were collected from the cell culture supernatant by PS affinity isolation. (b) Representative fluorescence images of HEK293T-sEVs labeled with AF647 anti-CD63, (d) AF488 anti-CD9, and (f) AF594 anti-CD81 antibodies. These images were obtained at the input concentration of 9.50 × 10 5 particles/μL. (c) Quantifications of fluorescent spots vs the input concentrations of sEVs labeled with AF647 anti-CD63, (e) AF488 anti-CD9, and (g) AF594 anti-CD81 antibodies. The number of fluorescent spots was fitted with y = ax b (R 2 = 0.96, 0.993, and 0.996 for AF647-CD63, AF488-CD9, and AF594-CD81, respectively). Each experiment was composed of four subruns, and these analysis results were integrated to form a single replicate. Error bars indicate standard deviations of five replicates. Detection of HEK293T-derived sEVs (HEK293T-sEVs) by PMACS. (a) Schematic diagram illustrating the experimental procedure of antibody-labeled HEK293T-sEVs by PMACS. The sEVs were collected from the cell culture supernatant by PS affinity isolation. (b) Representative fluorescence images of HEK293T-sEVs labeled with AF647 anti-CD63, (d) AF488 anti-CD9, and (f) AF594 anti-CD81 antibodies. These images were obtained at the input concentration of 9.50 × 10 5 particles/μL. (c) Quantifications of fluorescent spots vs the input concentrations of sEVs labeled with AF647 anti-CD63, (e) AF488 anti-CD9, and (g) AF594 anti-CD81 antibodies. The number of fluorescent spots was fitted with y = ax b (R 2 = 0.96, 0.993, and 0.996 for AF647-CD63, AF488-CD9, and AF594-CD81, respectively). Each experiment was composed of four subruns, and these analysis results were integrated to form a single replicate. Error bars indicate standard deviations of five replicates. After validating PMACS using model vesicles, we next applied the method to demonstrate its applicability for sEV analysis. First, we performed a titration of cell-derived sEVs by fluorescently labeling CD63, a representative marker protein on sEV surfaces, to evaluate the sensitivity and quantitativity of PMACS ( Figure 4 a). sEVs isolated from the HEK293T cell line (HEK293T-sEVs), which were confirmed to express CD63 ( Figure S8 ), were labeled with an Alexa Fluor 647 dye-conjugated anti-CD63 (AF647-CD63) antibody and subsequently incubated with the AAO membranes. Here, labeling prior to capture by AAO aims to avoid potential risk of steric hindrance to antibody-epitope recognition under nanoscale confinement in the pores. As shown in Figure 4 b, bright fluorescent spots were observed for HEK293T-sEVs stained with an AF647-CD63 antibody, as similar to the validation by Rhd-liposomes (cf. Figure 2 ). The number of detected fluorescent spots depicted a linear relationship against the serial dilution of sEVs on a log−log plot ( Figure 4 c), covering a wide dynamic range (3.71 × 10 3 to 3.80 × 10 6 particles/μL, R 2 = 0.96). Furthermore, a concentration as little as 3.71 × 10 3 particles/μL of HEK293T-sEVs gave a signal significantly distinguishable from the blank control samples that are without vesicles and bare AAO membranes (p < 0.05) ( Figure S9 ). It is noteworthy that no blocking reagent was used in this workflow, owing to the sufficiently low signal from blank controls indicating that nonspecific binding and background autofluorescence are negligible. This contributes to maintaining the operational simplicity of PMACS. Despite the rapid emergence of clinical applications, comprehensive fundamental research is still required to fully understand the molecular mechanism and the biological behavior of sEVs, which is critical for developing diverse applications. The tetraspanin family, including CD9, CD63, and CD81, is enriched on the surface of sEVs and is frequently utilized as a marker protein to specify unknown samples as sEVs. However, state-of-the-art studies have revealed that these proteins are unevenly distributed across sEVs. That is, not all particles express the same combination of them, and sEVs can be classified into several subgroups based on their expression profiles. (40−44) It has been gradually elucidated that these tetraspanins are involved in molecular transport and the biogenesis of sEVs (45−49) and serve as potential markers for validating the origin of sEV secretion pathways. This may therefore allow for the specification of sEV subtypes. (41,50) Population analysis of tetraspanins is essential to unravel the complex biological dynamics of sEVs. Because the PMACS method provides single-particle resolution for sEV detection as discussed above, it allows quantitative colocalization analysis of multiple markers on individual vesicles. To demonstrate this capability, we assessed the colocalization of the tetraspanin family members CD9, CD63, and CD81 on sEVs derived from Hela, A549, and HEK293T cell lines (Figure 5a). Each sEV sample was labeled using a cocktail of AF488 anti-CD9, AF647 anti-CD63, and AF594 anti-CD81 antibodies. Fluorescent spots overlapping across the emission channels were identified as individual vesicles that coexpress the corresponding proteins. 2.92 × 105 particles/μL of Hela-sEVs, 3.50 × 105 particles/μL of A549-sEVs, and 9.50 × 105 particles/μL of HEK293T-sEVs were applied to the system, and a total of 3220, 5809, 13874 fluorescent spots were detected in fluorescence images, respectively (Figure 5b). The total numbers of detected fluorescent spots were slightly lower in multiple labeling compared to single labeling, but no statistical difference was observed between multiplex- and single-stained sEVs at the 5% significance level (Figure S10). Thus, the ratio of colocalization is unlikely to be significantly affected by either interference or aggregation among antibodies. Figure 5 View LargeDownload to Slide Figure 5. Colocalization analysis of tetraspanins (CD9, CD63, and CD81) on sEVs. (a) Schematic diagram illustrating the experimental workflow of colocalization analysis for multiplex-stained sEVs. (b) Representative fluorescence images of multiplex-stained HEK293T-sEVs. In the merged image, AF488-CD9, AF647-CD63, and AF594-CD81 fluorescence was shown in green, magenta, and yellow channels, respectively. (c) Tetraspanin expression populations of Hela-, A549-, and HEK293T-sEVs shown as pie charts (top) and as Venn diagrams (bottom). Five replicates were conducted as each replicate consists of four subruns. Figure 5 View LargeDownload to Slide Figure 5. Colocalization analysis of tetraspanins (CD9, CD63, and CD81) on sEVs. (a) Schematic diagram illustrating the experimental workflow of colocalization analysis for multiplex-stained sEVs. (b) Representative fluorescence images of multiplex-stained HEK293T-sEVs. In the merged image, AF488-CD9, AF647-CD63, and AF594-CD81 fluorescence was shown in green, magenta, and yellow channels, respectively. (c) Tetraspanin expression populations of Hela-, A549-, and HEK293T-sEVs shown as pie charts (top) and as Venn diagrams (bottom). Five replicates were conducted as each replicate consists of four subruns. Close Figure 5. The ratio of each combination of protein coexpression is shown in Figure 5c. For Hela-sEVs, CD63 was the most broadly expressed marker at 59.2%, followed by CD9 at 43.5%. In contrast, CD81 expression was the lowest at 15.3%. A549-sEVs showed that CD9 was the dominant marker (58.5%), followed by CD63 (43.2%). CD81 expression remained low at 17.9%. HEK293T-sEVs exhibited the highest CD9 expression among the three cell lines, at 71.4%, with CD63 at 38.0%. CD81 expression was the lowest among all samples, at 9.4%. sEVs derived from any source cell lines shared the trend that CD9 and CD63 were expressed much more broadly than CD81, but the levels of CD9 and CD63 were varied. Paying attention to the overlapping across the channels, a few fluorescent spots appeared at corresponding positions across the filter channels (indicating colocalization), while the vast majority of fluorescent spots were observed in a single channel. Thus, colocalization of any combination of these proteins was not dominant, given that single expression of either CD9, CD63, or CD81 single-positive vesicles accounted for more than 80% of the total detected particles. For instance, in Hela-sEVs, the proportions of double-positive vesicles were 11% for CD9/CD63, 1.8% for CD63/CD81, and 2.1% for CD81/CD9. The triple-positive population (CD9/CD63/CD81) was as low as 1.4%. Similarly, A549-sEVs and HEK293T-sEVs showed low coexpression levels for any combination compared with single-positive expression. Therefore, these results indicate that CD9, CD63, and CD81 are not universally shared markers across all sEVs. Instead, sEVs would be divided into subgroups based on their tetraspanin expression patterns. (40−44) In order to validate the PMACS results, we performed nanoflow cytometry (nFCM) for A549-sEVs labeled with AF488 anti-CD9 and AF647 anti-CD63 antibodies. Notably, the nFCM system, which can detect single particles smaller than 100 nm in diameter, (19) provided data consistent with the PMACS results (Figure S11). Specifically, the nFCM analysis showed that CD9 was expressed on more than half of the vesicles, CD63 on fewer, and that the colocalization of CD9 and CD63 accounted for approximately 10% of the total CD9- or CD63-positive vesicles. Hence, these results reveal that the values obtained by PMACS accurately reflect the population of tetraspanin expression. Although CD9, CD63, and CD81 have been regarded as exosomal markers for a long time, (51) recent works have unveiled that they are not ubiquitously expressed on all exosomes. (40−44,50) Indeed, other types of EVs such as ectosomes may contain some combinations of these tetraspanins more abundantly than exosomes. In this context, a recent study by the Mathieu group demonstrated that sEVs bearing CD9 and CD81 bud mainly from the plasma membrane based on tracking protein transports in Hela cells. In contrast, those bearing CD63 were shown to originate from the late endosome. These findings strongly imply that CD9 and CD81 are likely to be more enriched on ectosomes while CD63 is predominantly localized on exosomes. (50) Other colocalization analysis similarly supports that most particles contain only one of these proteins. (44,52,53) Among them, a TIRF imaging-based study reported the high frequency of single-positive expression of CD9, CD63, and CD81 and that CD9/CD81 double-positives occur more frequently than CD9/CD63 and CD63/CD81 combinations in HEK293- and MCF7-derived sEVs. (52) A relatively high frequency of CD9/CD81 colocalization was also reported by other groups, which was considered to originate from the plasma membrane, supporting Mathieu's findings. (54) It is noted that these expression trends are not exclusive; some vesicles indeed contain the combination of CD9/CD63 or CD63/CD81. Our analysis results seem to be consistent with these recent works, (52,53) as most particles show single-positive expression of CD9, CD63, or CD81. However, CD9/CD81 colocalization was less observed in our results, which indicates that our results do not show complete consistency with these reports. This difference may be partly ascribed to suppression of overestimated colocalization, possibly because aggregated sEV particles are excluded through the size-exclusion effect. Conventional single-particle analysis techniques hold potential risk of colocalization overestimation caused by particle aggregates. In contrast, because PMACS captures sEVs within pores, particles larger than the pore size would be physically excluded, thereby reducing the potential risk of overestimation. While this concept can be certainly expected in the context of porous material and colloid interface science, other factors, such as sEV innate heterogeneity derived from biogenesis and isolation bias, were not rigorously ruled out in this study, so that further comprehensive investigation is needed to elucidate this. Of course, the expression patterns vary depending on the cell type, culture conditions, and sEV isolation methods. (43,52) Some other literature, for instance, reported markedly different colocalization patterns (40,41,43) such as a high ratio of CD9/CD63 colocalization and high expression of CD81. (43) In addition, other latest works provided implications that could deny the specific dependence of CD63 transport to sEVs on endocytosis. (48,55) Therefore, the accurate profiling of these tetraspanins remains controversial, requiring further investigations to fully connect their molecular components with biological dynamics. In this context, PMACS should be an invaluable tool to evaluate molecular components on individual sEVs, which will assist in the further exploration of these tetraspanin dynamics.