Human melanoma cell line SK-MEL-147 was cultured in DMEM supplemented with 10% heat-inactivated FBS, penicillin (100 U/mL) and streptomycin (100 μg/mL), at 37°C in a 5% CO₂ atmosphere. For EV production, 1.6 × 10⁶ cells were grown in p150 plates for 6 days in DMEM supplemented with 5% EV-depleted FBS until they reached confluence. EV depletion from FBS was performed by ultracentrifugation at 120,000 g for 16 h.

Conditioned media was centrifuged for 5 min at 500 g and 30 min at 3,200 g to eliminate cells, cellular debris and apoptotic bodies. CM from 1 p150 plate (20 mL) was then concentrated (concentrated conditioned media: CCM) using Amicon Ultra-15 filters (100K, Millipore, Billerica MA) to a final volume of 400 μL, which was subsequently used for EV isolation.

Bradykinin-derived peptide (BP) previously employed in (Gori et al., 2020) was modified adding a 6 His-tag (6xHis-RPPGFSPFR) to allow divalent cation binding, enabling stable anchoring to resin supports. Tandem (BPt) or branched (BPb) molecular constructs were used as indicated, which are composed of 2 or 4 BP repetitions, respectively (see section below and Figure 4A). Agarose resins employed were customized by Agarose Bead Technologies (ABT) to improve the extern interaction of cations in the surface of the bead. The information of the different agarose resins employed are included in Table 1.

For each experiment, volumes, concentrations and incubation times are indicated in results and figure captions. In brief, for most experiments 0.5 mg of BP were coupled to 10 µL of agarose by incubation at RT in 0.2 mL tubes. After peptide binding, agaroses were washed twice with 100 µL of PBS to eliminate unbounded peptide and incubated with 100 µL of concentrated condition media for EVs binding over night at 4°C, under rotation. Unbound material was discarded, and resins were washed 5 times with 100 µL of PBS. All washes were performed pelleting the agarose at 1,000 rpm for 30 s.

For downstream dot blot analysis, resins were lysed with Laemmli buffer in a total volume of 30 µL (including resin volume). For EVs characterization by NTA, MET, western blot and for SiMoA assay EVs were eluted with 100 µL of filtered PBS containing 0.5 M imidazole (Sigma) (EVs characterization by NTA, MET, western blot). For uptake analysis by flow cytometry, EVs were eluted with 100 µL of filtered PBS containing 0.5 M imidazole or 100 mM EDTA, as indicated.

Empty columns with both upper and under cellulose frits were packed with 10 mL of 2% BCL agarose (Agarose Bead Technologies). Columns were washed with 2 volumes of filtered PBS before SEC. Concentrated condition media from 1 p150 (400 µL) was loaded and 25 fractions of 500 µL were collected by gravity elution using PBS as elution buffer.

For dot blot analysis, agarose-peptide-EVs resins were lysed with Laemmli buffer in a total volume of 30 µL (including resin volume) and heated for 10 min at 65°C. 1 μL from the different samples were directly spotted on a nitrocellulose membrane (GE Healthcare) and let to dry.

For SK-MEL-147 whole cell lysates, confluent cells were washed with PBS and lysed in TBS/1% Triton supplemented with protease inhibitors (Roche). Protein concentration was measured with Pierce™ BCA protein assay kit (Thermo Scientific, 23225).

For western blot analysis, EVs isolated by affinity chromatography were eluted from the resin with 100 µL of filtered PBS containing 0.5 M imidazole, lysed with Laemmli buffer and heated for 5 min at 96°C. 2·10⁹ isolated EVs or the indicated protein content of total lysates from SK-MEL-147 cell line were loaded in polyacrylamide SDS-PAGE gels. Electro-transference of proteins to nitrocellulose membrane (GE Healthcare) was carried out using wet electroblotting in TRIS-Glycine buffer with 20% of methanol, during 1.5 h at 300 mA. Both dot blot and western blot membranes were blocked with 5% skimmed-milk in TBS, 0.1% Tween-20 for 30 min.

Antibodies anti-CD81 (5A6; kindly provided by Dr S Levy, Stanford, USA), anti-histidine (H1029, Sigma-Aldrich, diluted 1:500) and anti-Apolipoprotein B (Calbiochen, 178467, diluted 1:500) were used for EV, peptide or LPP detection in dot blot, respectively. Other antibodies against EVs markers were also used in western blot analyses for EV detection: anti-CD9 (VJ/1.20; hybridoma supernatant, not diluted), anti-CD63 (Tea3.10; hybridoma supernatant, not diluted) (Yáñez-Mó et al., 1998), anti-TSG101 (Genetex, GTX118736, diluted 1:1,000), anti-Syntenin-1 (Synaptic systems, 133003, diluted 1:1,000), anti-ARF-6 (Sigma, A5230, diluted 1:1,000) and anti-flotillin (BD Biosciences, 610821, diluted 1:1,000). For non-EV markers detection, we employed anti-VDAC1 (Abcam, ab154856) and anti-calnexin (Enzo, ADI-SPA-865-F), both diluted 1:1,000. As secondary antibodies conjugated with HRP, we used anti-mouse (Thermo Fisher Scientific, 31430), anti-rabbit (Thermo Fisher Scientific, 31460) and anti-goat (Sigma, SAB3700316, diluted 1:10,000).

Membranes were revealed with Super Signal West Femto HRP substrate (Thermo Scientific). A 4,000-mini system (General Electrics) was used to acquired images, which were lately processed with Fiji ImageJ.

EV from conditioned media (CM) were isolated by BPt agarose beads as described above. For SEC by IZON kit, 150 µL of concentrated CM (45x) were loaded onto a SEC column (IZON qEV 70 Legacy). Fractions of 200 µL each were collected; the largest vesicles amount was found in fraction 7, according to manufacturer protocols suggestion. Polymer precipitation was performed by adding ExoQuick Reagent (ExoQuick) to 5 mL of 1x CM and incubating overnight at 4°C according to manufacturer instructions. Then precipitate vesicles were recollected by centrifugation and resuspended in fresh PBS. Bead immunocapture was performed by using commercially available streptavidin-beads (Invitrogen, Exosome-Streptavidin Isolation/Detection Reagent) functionalized with pan-tetraspanin biotinylated antibodies (CD9, Ancell, clone SN4/C3-3A2, CD63, Ancell, clone AHN16.1/46-4-5, and CD81, Ancell, clone 1.3.3.22). One test was performed using immune-capturing only, according to manufacturer suggestion (0.2 mL of 1X CM), another test was performed coupling the bead based immune-capture sample to 0.2 mL of EVs previously concentrated by polymer precipitation (ExoQuick) obtained from a starting volume of 9.5 mL of 1x CM. All incubations were performed according to manufacturer instructions. Vesicles were released from beads in 0.1 mL glycine 0.2 M, pH 2.4. For ExoSpin kit 1 mL of 1x CM sample were treated according to manufactures instructions. For MagCapture™ Exosome Isolation Kit PS (Fujifilm) 9 mL of 1x CM were incubated on beads and vesicles were released according to manufacturer instructions.

SiMoA assay was run according to previously devised protocols (Frigerio et al., 2022). Briefly, beads were conjugated according to Quanterix Homebrew kit instructions using the recommended buffers to pan-tetraspanin antibodies (anti-CD9, Ancell, clone SN4/C3-3A2, anti-CD63 antibody, Ancell, clone AHN16.1/46-4-5 and anti-CD81 antibody Ancell, clone 1.3.3.22). Samples were analyzed with three-steps assay according to manufacturer instructions. 0.1 mL of each sample was diluted 1:4 in a diluent and placed in a well; after that a capture step with conjugated paramagnetic beads was performed. To detect captured EVs a mixture of biotinylated anti-tetraspanin antibodies was used (anti-CD9, Ancell, clone SN4/C3-3A2, anti-CD63 antibody, Ancell, clone AHN16.1/46-4-5 and anti-CD81 antibody Ancell, clone 1.3.3.22). Detection was allowed by streptavidin-β-galactosidase (SBG), which acts on the fluorogenic substrate resorufin β-D-galactopyranoside (RGP). First, a calibration curve was produced by analyzing serial dilutions of the 45x CM sample as quantified for particle concentration by NTA analysis. For the recovery yield calculation, the AEB (Average Enzyme per Bead) raw data provided by the instrument for each sample were first interpolated in the calibration curve to provide a particle concentration, which was then normalized by the volume of the starting material.

The number of particles and the size distribution of EVs after SEC or affinity isolation were analyzed with NanoSight NS300 (Malvern Instruments Ltd., Malvern, United Kingdom), equipped with a 532 nm laser. EV-enriched fractions were diluted 100 times for the analyses. Results were analyzed with NTA 3.0 software.

EVs isolated by SEC or affinity were diluted by half in filtered PBS, adsorbed on carbon-coated nickel grids and contrasted with uranyl acetate. Sample visualization and image acquisition were performed in a transmission electron microscopy JEM1400 Flash (Jeol). The size and morphology of isolated EVs were analyzed with TEM Exosome Analyzer software (Kotrbová et al., 2019).

EVs obtained by BPt affinity method (and SEC) were labelled with Alexa 633 C5 Maleimide (Invitrogen) at 4°C ON. Maleimide excess was removed by SEC in columns of 2 mL using 2BCL agarose. Positive EVs fractions were pooled, and the number of particles was determined by NTA. 25,000 EVs per cell were incubated with 100000 SKMEL-147 cells for 30 min, 1 h or 2 h (as indicated) in regular culture conditions to allow EV uptake. Cells were washed with PBS, detached with trypsin and washed again with PBS. EV uptake was measured with a FACSCantoII flow cytometer and results analyzed with FlowJo software.

In order to determine if membrane–sensing peptides can be employed for EVs isolation, we designed a strategy for affinity isolation with agarose beads of EVs from concentrated conditioned media. We employed a tandem version of a bradykinin-derived peptide (BPt) that has been previously reported to efficiently capture sEVs on a flat well format (Gori et al., 2020). We modified the peptide to include a 6His tag that could non-covalently bind to divalent cations-carrying agarose resins (Figure 1A).

First, we tested BPt binding to different cation agaroses which differ in the cation employed and cation density (see Table 1 under methods). For these experiments, we incubated 0.1 mg of peptide with 50 μL of the different resins for 1 h at RT, under rotation. Resins were then washed twice with PBS and boiled in a final volume of 150 μL (including resin volume) with Laemmli buffer. Peptide binding was analyzed by dot-blot of 1 μL of lysate sample using an anti-poli-His antibody. As can be clearly seen in Figure 1B, cobalt agarose was the most efficient in BPt binding and chosen for all subsequent experiments.

To assess whether this strategy was efficient in EV capture and to analyze the amount of peptide required, we incubated for 1 h different amounts of BPt with 10 μL of cobalt agarose. After washing, each sample was incubated ON with 100 μL of concentrated conditioned media from a SKMEL147 culture, washed again with PBS and boiled with Laemmli buffer in a final volume of 30 μL. 1 μL of each sample was analyzed by dot blot using anti-CD81 tetraspanin as EV marker. We observed dose response precipitation of EVs as we increased the amount of BPt, which demonstrated an effective EVs isolation capacity of bradykinin peptide (Figure 2A). As control, we incubated parallel dot blots with anti-ApoB antibodies to assess lipoprotein (LPP) contamination of the EV samples. As shown in Figure 2B, we could not observe binding of lipoproteins to the peptide-agarose, which confirms the specificity of this BPt for lipids on curved membranes as those in EVs, but not on LPPs.

Agarose resins are composed of porous beads of several hundreds of nm in diameter. EVs would not be able to access the pores, but the soluble peptide can well bind to those cations exposed inside the pores. Therefore, EV binding would only be efficient to the molecules of BPt exposed on the agarose resin beads surface but not to those bound to cations inside the pores (Figure 1A). To optimize peptide exposure on the surface of the resins, we reduced peptide-resin incubation times. In these experiments, we incubated 0.5 mg of peptide with 10 μL of agarose resins for different times, in a final volume of 100 μL. After washing the resin, we performed the incubation with 100 µL of the concentrated conditioned media ON at 4°C, in a final volume of 100 μL. As shown in Figure 2C, we could observe an increment in the EV capture when BPt was incubated with agarose only for 10 min. Again, no LPP binding to the resins could be observed (Figure 2D).

Then, we analyzed different incubation times of BPt-carrying resins with concentrated conditioned media for EV binding. We incubated 0.5 mg of peptide with 10 μL of cobalt agarose for 10 min and after washing, with 100 μL of concentrated conditioned media for different times at 4°C. These analyses revealed that overnight incubation was the most optimal timing for this step (Figure 2E).

Since peptide exposure on the surface seemed to be a limiting factor, we decided to compare two different Cobalt-borne resins that differ in the length of the arm that connects the agarose surface to the cobalt chelate, but also in the density of cobalt chelates (Figure 3A), having the resin with the long-arm (12 atoms LA-Co Ag) half of the cobalt chelates (Table 1). Two different amounts of BPt (500 or 250 μg) were incubated for 10 min with 10 μL of both resins (Cobalt agarose and Long Arm Cobalt agarose; Co Ag and LA-Co Ag, respectively). EVs binding was also carried out as in previous experiments. Although the number of binding sites for peptide in LA-Co agarose were lower than in simple cobalt agarose, an increment in the amount of BPt bounded to LA-Co Ag was observed, being even significantly higher using 500 µg of peptide (Figure 3B). When we analyzed and quantitated EV binding, we could also observe a better isolation yield for LA-Co agarose (Figure 3C). These results further confirm that peptide exposure on the agarose beads is a crucial factor for the efficient EV binding. Therefore, LA-Co Ag was employed in subsequent experiments.

We next decided to compare different architectures of the Bradykinin peptide, including the tandem repeats employed so far (BPt) or a tetra-branched (BPb) (Figure 4A). Multivalent probes were indeed showed to play a favorable role on EV binding efficiency (Saludes et al., 2013; Gori et al., 2020), likely due to cooperative effects that amplify peptide binding affinity (Gori et al., 2017). Using the same amount of peptide (500 μg) for binding to 10 µL of LA-Co agarose (Figure 4B), we observed a clearly higher EVs binding when using BPt in comparison to BPb version (Figure 4C). Thus, we continued using BPt for subsequent experiments.

Finally, we also tested if changes on buffer pH would affect both the steps of peptide binding and EVs isolation. Again, we employed 10 µL of LA-Co agarose, 0.5 mg of peptide and 100 µL of concentrated conditioned media diluted by half in PBS at the different pH for peptide and EV binding. When experiments were carried out in PBS at pH5, we could neither observe BPt binding to the resin nor EV isolation (Figure 5A). BPt peptide binding to the resins was slightly higher in PBS at pH7 compared to PBS at pH9. In contrast, although the amount of peptide that bound to the resin was lower at pH9, EV yield was in some experiments a bit higher although differences were not significant (Figure 5B), however, the combination of both factors (peptide binding and EV capture) made differences in results between pH7 and pH9 not significant. Subsequent experiments were performed at neutral pH, to facilitate downstream functional analyses.

Although for many characterization experiments such as proteomics, western blot or RNA analyses, we can directly assess the composition of resin-bound EVs, the His-Tag approach we developed allows for a simple and gentle separation of EVs from the column after purification by simple competition with imidazole or elution with EDTA. Thus, for a more detailed characterization of EVs isolated by BPt affinity method, we eluted EVs from the resins by incubating 40 µL of resin (LA-Co agarose carrying 1 mg of BPt and incubated with 200 µL of CCM) with 100 µL of PBS/0.5 M imidazole. For comparison we isolated EVs from 400 µL of CCM by standard size exclusion chromatography (SEC).

EVs size distribution of the isolated EVs was similar to those obtained by SEC from the same CCM, as revealed by both NTA analysis (Figure 6A) and TEM images analysis (Figure 6B). Moreover, we could observe a clear reduction in the amount of LPP particles in TEM images of EV samples obtained by BPt affinity method, in comparison to EV samples obtained by SEC (Figure 6B). We performed western blot analysis normalizing the samples to the number of particles as quantitated by NTA. We could also corroborate the isolation of EVs using BPt, by detection of EV-markers like TSG101, Syntenin-1 and Flotillin (Figure 6C). The absence of non-EV markers like VDAC and Calnexin (Figure 6D) also corroborated proper isolation of EVs.

BPt optimized protocol was compared to commercially available tools and kits: size exclusion chromatography by IZON qEV columns, polymer precipitation (Exoquick), bead-based immunocapture by Dynabeads conjugated to anti CD9/CD63/CD81 antibodies (immune-capture was tested alone or after concentration via polymer precipitation), ExoSpin, and bead-based phosphatidylserine affinity capture by MagCapture™ (Fujifilm). Although different starting sample volumes and concentrations were used to comply with manufacturer’s instructions and kit requirements, all different methods allowed to collect particle fractions within 100–200 µL of final volume (see Materials and Methods). Customized SiMoA (Single Molecule Assay) pan-tetraspanin EV immune-detection (Frigerio et al., 2022) was selected to provide high sensitivity estimation of EV yield following each isolation method. SiMoA assays were developed according to the Quanterix Homebrew assay instructions as detailed in the Materials and Methods Section. Serial dilutions of the starting EV sample (CM 45x) were analyzed, and a calibration curve establish according to NTA determination of particle concentration in each sample (Figure 7A). Then all EV samples obtained after isolation with BPt and the selected commercial kits were analyzed by SiMoA assay. The raw data obtained were first interpolated in the calibration curve for determination of concentration of pan-tetraspanin positive particles in each analyzed sample, which was then normalized by the volume of the starting sample material to calculate the isolation yield of each method (Figure 7B). EV recovery from IZON columns resulted in the highest yield whereas BPt protocol was superior to all other tested tools and kits.

As previously mentioned, one of the advantages of the approach here designed is the ability to gently elute EVs from the purification beads and its potential application in subsequent functional experiments. Thus, to assess the functionality of the eluted EVs, we evaluated the capacity of SKMEL-147 cells to uptake EVs isolated using BPt in comparison to EVs isolated by SEC. To this end, EVs were labelled with a fluorescent maleimide compound and then EVs were incubated with SKMEL147 cells before analysis by flow cytometry of EV capture. As observed in Figure 8, EVs isolated by BPt and posterior elution with imidazole (Figure 8A) or EDTA (Figure 8B) were efficiently captured by target cells, which validated their use in functional experiments.