Plasma samples from COVID-19 patients were collected by the Ospedale Maggiore Policlinico (Milan, Italy) between March and June 2020 in the acute phase of infection from venous blood samples using Ethylenediaminetetraacetic acid (EDTA) as anticoagulant. The clinical classification of MILD and SEVERE patients was done following the recommended WHO clinical progression scale score (MILD patient scores ranged between 1 and 4, whereas SEVERE patients had scores between 5 and 10) (9). Detailed information is shown in Supplementary Table S1 . For the SEVERE group, samples were collected during patients’ hospitalization in COVID-19 high dependency units. COVID-19 patients were diagnosed by RT-PCR nasopharyngeal swab. HDs were chosen to be representative of both sexes and of varied ages (26–65 years). While gender was matched between COVID-19 patients and healthy controls in this cohort, the average age of the SEVERE and MILD group was not, the first being much higher than the second, following an intrinsic characteristic of COVID-19 epidemology that shows a worse prognosis in elderly individuals.

In this study, 100 µl of plasma samples were prepared by 3 centrifugation steps (Step 1: 10′ at 500 g; Step 2: 20′ at 2,000 g; Step 3: 30′ at 14,000 g) to eliminate red blood cells and cellular debris. After each step, the supernatant was transferred to a new tube and the pellet was discarded. After the last centrifugation, the supernatant was diluted in Phosphate-buffered saline (PBS). To exclude cell-free virions in the exosome preparations, we always used immune affinity purification method using anti-tetraspanin antibody-conjugated magnetic or latex beads.

Circulating exosomes from plasma of MILD and SEVERE patients or HDs were captured using the ExoTEST Kit (Hansa BioMed) according to the manufacturer’s instructions. The ELISA plate with 100-µl test samples loaded per well was incubated at room temperature with shaking for 30 min. After washing 3 times with washing buffer, 100 μl of biotinylated anti-SARS-CoV-2-S RBD primary antibody (1:1,000 ratio) were added to each well and incubated at room temperature while shaking for 2 h (2–3 rotations per second). The plate was washed again with the washing buffer, and 100 μl of diluted streptavidin HRP-conjugated were added to each well. The plate was incubated at room temperature while shaking for 1 h (2–3 rotations per second). Then, 100 μl of Substrate Chromogenic Solution was added to each well and incubated uncovered at room temperature in the dark for 5–10 min. The plate was monitored until a blue color was visible. At this point, the reaction was stopped by adding 100 μl of stop solution to each well. The absorbance was recorded at 450 nm within 10 min with an Infinite F200 (Thermo Fisher Scientific, USA).

For immunoisolation of exosomes recovered from plasma, 10 μl of 4-μm-diameter latex beads were incubated with 20 µl purified anti-CD63 mAb for 30 min at room temperature in a final volume of 50 μl. After 15 min, the volume was made up to 200 μl with PBS, and samples were incubated overnight at 4°C with gentle agitation. For Fluorescence activated cell sorting (FACS) analysis, exosomes recovered from blood plasma were incubated in 60 μl for 30 min at 4°C with anti-CD63-latex beads. The volume was made up to 400 μl with PBS and incubated for 2 h at 4°C. To eliminate the unspecific antibody binding, beads resulting after plasma incubation were blocked with 5% bovine serum albumin (BSA) solution for 30 min at room temperature. Vesicle-coated beads were washed twice in washing buffer (1% BSA in PBS) and resuspended in 400 μl of washing buffer, stained with the indicated fluorescent antibodies, and analyzed on a FACSCantoI flow cytometer (BD Biosciences) and FACSDiva software. Flow cytometry analysis was done interchangeably using an isotype antibody or a secondary antibody control.

For exosome integrity analysis, we also assessed syntenin compartmentalization in the exosome cargo. The Intracellular Fixation and Permeabilization Buffer Set (eBioscience) was used to fix and permeabilize exosomes for intravesicular staining. Exosomes were fixed in 100 μl of IC Fixation Buffer and incubated in the dark at room temperature for 20 min. After incubation, the samples were washed twice with 500 μl of permeabilization buffer and then resuspended in 100 μl of the same buffer. The samples were then incubated with PE anti-human syntenin for 30 min at room temperature. After incubation, the samples were washed twice with 500 μl of permeabilization buffer and resuspended in 200 μl of 1% BSA-PBS and analyzed.

The size distribution and concentration of plasma-derived exosomes were analyzed by nanoparticle tracking analysis (NTA) using NanoSight model NS300 equipped with a Blue488 laser and a sCMOS camera (Malvern Instruments, Malvern, UK). Here, 2 μl of sample was diluted in 1 ml of PBS and then injected into the laser chamber. The following settings were used for data acquisition: camera level, 13; acquisition time, 60 s; and detection threshold, 5. Data were analyzed using the NTA 3.4 software (Malvern Instruments). Three recordings were performed for each sample. The evaluation of the particle size distribution (PSD) was performed through the following parameters: Mean, Mode, SD, D10, D50 (Median), and D90, which indicate, respectively, the average, most frequent particle class size, standard deviation, and the 10%, 50%, and 90% percentiles of the analyzed particles.

In this study, 5 μg of anti-CD63, anti-CD9, and anti-CD81 mAbs, diluted in 200 μl PBS, were added to 1.5 mg (50 μl) of the protein G-coated Dynabeads^® suspension and incubated with rotation for 20 min at room temperature. The Abs-coated beads were separated from the non-bound antibodies by placing the tube on a magnet for 1 min and removing the supernatant. Beads–Ab complexes were resuspended in 200 μl PBS with Tween^®-20 and washed by gentle pipetting. Plasma samples (100 μl) were added to the Dynabeads^®–Ab complexes, and the beads were incubated with rotation Overnight (O/N) at 4°C. The Dynabeads^®–Ab–antigen complexes were washed 3 times using 200 μl PBS and were resuspended in 100 μl PBS. The bead suspension was then transferred to a clean tube to avoid co-elution of proteins bound to the tube wall. The tube was then placed on a magnet for 1 min, and the beads were recovered. Next, 50 μl of premixed NuPAGE^® LDS Sample Buffer and NuPAGE Sample Reducing Agent (mixed as per manufacturer’s instructions) were added to the Dynabeads^®–Ab–antigen complex and heated for 10 min at 70°C. Then, the tube was placed on a magnet for 1 min, and the supernatant/sample was loaded onto a NuPAGE 4%–12% Bis–Tris gel (Invitrogen, Waltham, MA, USA) and transferred onto a Polyvinylidene fluoride (PVDF) membrane (Merck Millipore, Burlington, MA, USA). Membranes were blocked in 5% non-fat powdered milk in TBS-T (0.5% Tween-20) and probed with the indicated antibodies for 2 h at room temperature. Membranes were washed, and HRP-conjugated antibodies were added for 1 h at room temperature. Detection was performed using enhanced chemiluminescence (ECL) reagents (Invitrogen) according to the manufacturer’s guidelines.

Exosomes recovered from plasma of COVID-19 patients and HDs were analyzed using ExoView Tetraspanin chips (NanoView Biosciences, Boston, MA, USA) arrayed with antibodies against the CD81, CD63, CD9, and CD41a proteins. Mouse IgG1 was used as a negative control. In brief, 35 µl of each sample were dropped onto the chip surface (placed in a sealed 24-well plate) and incubated for 16 h at room temperature. Each chip was then washed once on an orbital shaker with PBST (PBS supplemented with 0.05% Tween-20) for 3 min, then washed three additional times with PBS for 3 min. After washing, the chips were incubated with anti-SARS-CoV-2-S RBD, conjugated with Alexa Fluor 647, in PBST supplemented with 2% BSA in a volume of 250 μl for 2 h at room temperature without shaking. Next, each chip was washed once with PBST, 3 times with PBS, once in filtered deionized water, and then dried at room temperature for 1 h. The chips were then imaged with the ExoView R100 reader (ExoView) and analyzed using the ExoViewer software with a sizing threshold set to 50–200-nm diameter. The resulting size and fluorescence intensity information for each individual exosome was exported to Excel for statistical analyses. Fluorescence values are reported in arbitrary units.

For SARS-CoV-2-S immunolabeling and stimulated emission depletion (STED) microscopy evaluation, latex bead-isolated exosomes were used, similarly to what was previously described for FACS immunostaining. Exosomes prepared from blood plasma of HD and MILD and SEVERE COVID patients were incubated with anti-CD63-latex beads, and beads were then blocked with 5% BSA solution to eliminate unspecific antibody binding. Vesicle-coated beads were washed and seeded onto glass coverslips (n. 1.5 thickness; electron microscopy), then stained with anti-SARS-CoV-2-S RBD (1:500), followed by hybridization with secondary antibodies conjugated with the Star*RED fluorophore (Abberior; 1:200), and finally mounted onto glass slides with ProlongGlass mounting reagent for super-resolution (Molecular Probes, Thermo Fisher Scientific). Samples were acquired using a demo version of the Abberior STEDYCON microscope for simultaneous confocal and STED microscopy (kind collaboration with Abberior Instrument) with 4 excitation laser lines and a 775-nm depletion STED laser. Star*RED fluorophore excitation was kept at 10% power of the 640-nm laser power, whereas Star*RED fluorophore depletion was obtained at 100% laser power in order to achieve 30-nm resolution, at pixel size 15 nm, with 7 lines of STED acquisition over different Z-plan, spanning throughout the whole z-dimension of the latex beads. A total of n = 9 beads were acquired in STED imaging for all conditions (HD, MILD, SEVERE) with duplicate independent biological replicates, reaching a total of n = 18 for each condition. STED-resolved microscopy images were quantified upon precise photon counting. Data were normalized with subtraction of photon counts of background signal, as evaluated by anti-SARS-CoV-2-S RBD–Star*RED in HD samples, with mean photon count in raw images =44, whereas mean photon count in raw images from MILD and SEVERE samples were =152 and =80, respectively.

Transmission electron microscopy evaluation of RBD immunogold labeling was performed on HD, MILD COVID-19, and SEVERE COVID-19 plasma-derived exosomes purified via lattice-bead immunocapture as described in the previous sections and stained in suspension with rabbit anti-SARS-CoV-2-S RBD (1:500) followed by 5-nm Gold-anti-Rabbit Fabs (BBI International) (1:200). After staining, samples were fixed in 1% glutaraldehyde and finally kept in PBS pending TEM observation. TEM sample preparation was performed similarly to what has been recently described (10). Here, 5 μl of the pooled EV sample was layered onto a formvar/carbon-coated 200 mesh grids and allowed to settle for 20 min. No further negative stain was performed in order to preserve a proper contrast for immunogold labeling. The grids were blotted and allowed to air-dry at room temperature. The observations were carried out with a JEOLJEM-1011 (Jeol Jem, Peabody, MA, USA) transmission electron microscope operated at 100 kV. Images were captured using a Morada G2 TEM digital camera (Olympus Soft Imaging Solutions, Münster, Germany) and iTEM software. Negative controls (not shown) were performed in the absence of primary antibodies. Here, 10–15 images were captured from each of three randomly selected areas of each grid at ×50,000 and ×100,000 lens magnification. The camera magnifications were calibrated using a grid with a grating replica (EMS cata #80050) with line spacing of 463 nm (2,160 lines/μm). Scale bars reflect the magnification at the camera. TEM micrographs were analyzed manually. Rounded or “cup-shaped” particles with high-contrast edges were considered exosomes and measured using Fiji-ImageJ (with Java 1.8.0_172, 64-bit).

Single-molecule super-resolution microscopy for RBD fluorescent immunolabeling on MILD COVID plasma-derived exosomes was performed with direct stochastic optical reconstruction microscopy (D-STORM) modality (i.e., spontaneous particle blinking with only excitation in fluorophore reporter wavelength). Exosomes were purified via lattice-bead immunocapture as previously described and stained in suspension with rabbit anti-SARS-CoV-2-S RBD (1:500) followed by AlexaFluor-647 Goat anti-Rabbit IgGs (Molecular Probes, Thermo Fisher) at 1:200. After staining, samples were fixed in 1% Paraformaldehyde (PFA) for fluorophore preservation and finally kept in PBS pending STORM assay. For acquisition, 5 μl of the labeled exosome pool were carefully layered at the center of a glass-inserted 35-mm petri dish (CellView) and air-dried under the laminar hood for 30 min to achieve a correct glass-surface deposition of labeled exosomes before adding 250 μl of freshly prepared STORM buffer (Abbelight) and topping the sample with a 22-mm square glass. The achieved volume of buffer of acquisition surface was optimized for correct evanescence wave formation within total internal reflection fluorescent (TIRF) illumination. Acquisitions were performed on an N-STORM instrument (Nikon Instruments) mounted on a Nikon Ti widefield with DU-897 EM-CCD camera (Andor Technology) with TIRF illuminator and ×100 TIRF (NA 1.49) objective (Nikon Instrument), coupled with 10-mW 647 excitation/reported laser (CrystaLaser) used at 70% power for D-STORM for 10,000 frames/acquisition over a constant TIRF-plane angle, with 1 frame exposure detection (at 10–20 ms range) at 17 Mhz, for both 2D and 3D STORM. After acquisitions, particle-detection data reconstruction was obtained with the ad hoc STORM analysis module in NIS-Elements v.5.31 (Lim-Nikon Instruments) with the following parameters for blinking molecule detection over acquired frames: standard Gaussian fitting for the localization of specific molecules characterized by minimum bit Height 500, maximum bit Height 65535, over a CCD bit baseline 100, for specific counting of molecules with minimum width 20 nm, maximum width 200 nm, initial fit width 100 nm. For 2D STORM, the x and y coordinates from each blinking molecule were retrieved for molecule localization; for 3D STORM, the 3 coordinates (x, y, z) were retrieved and 3D volume viewing was performed.

CD14 monocytes purified from peripheral blood mononuclear cells (PBMCs) by positive immunomagnetic sorting were loaded with 10 × 10⁷ of purified MILD patient-recovered exosomes for 6 h with rotation at 37°C. After incubation, monocytes were washed with RPMI containing 10% serum (exosome free), irradiated, cocultured with autologous negative immunomagnetic purified CD4⁺ T cells, and labeled with cell trace. Interleukin (IL)-2 was added after 48 h of culture. After 4 days, cells were stained with anti-CD4/BUV395 and anti-HLA-DR/BUV 650 and analyzed by flow cytometry (BD Symphony).

Exosomes were purified from the plasma of 3 HDs, 4 patients presenting MILD COVID-19 symptoms, and 4 patients presenting SEVERE COVID-19 symptoms using qEV SMART Size Exclusion Chromatography (SEC) columns following manufacturer’s instructions. Briefly, 150 ul of plasma were centrifuged at 1,500 g for 10 min in order to remove cells and large contaminant particles. After a subsequent centrifugation at 10,000 g for 10 min, supernatants were loaded on IZON qEV/35nm columns, previously washed and equilibrated with 3 volumes of PBS. After column loading, samples were eluted with PBS using a void volume of 1 ml and an elution volume of 600 µl. Purified exosomes were then concentrated and processed for mass spectrometry analysis.

Quantitative proteome profiling from exosomes was achieved by a library-based approach combined with MS boxcar acquisition method in a label-free experiment as reported before (11). We constructed three different libraries by pooling proportional quantities of different exosome preparations from each sample type (MILD, SEVERE, and HD) assessed by Bicinchoninic acid (BCA) protein quantitation assay. Samples were run independently and then aligned to their respective libraries. In brief, proteins were denatured, reduced, alkylated, and digested, and peptides were purified on StageTips (12) using reagents from the PreOmics “iST” Kit (P.O. 00001, PreOmics GmbH). Only samples for the library were fractionated using a commercial High pH Reversed-Phase Peptide Fractionation Kit (Thermo Scientific). In all cases, dried peptides were reconstituted in 5 µl of LC-LOAD buffer (P.O. 00001, PreOmics GmbH). Samples were measured using liquid chromatography–mass spectrometry (LC-MS) instrumentation consisting of an EASY-nLC 1200 system coupled to a nano-electrospray ion source and a Q Exactive HF Orbitrap (all Thermo Fisher Scientific). Purified peptides were separated on an EasySpray PEPMAP RSLC C18 (Thermo Fisher Scientific) kept at 45°C constant to reduce column back pressure. Solvent A was 0.1% formic acid (FA), and solvent B was 0.1% FA in 80% Acetonitrile (ACN). Samples were loaded in aqueous 0.1% (FA) solution at constant pressure of 980 Bar. Peptides were separated with a gradient of 3%–30% solvent B over 39 min followed by a gradient of 30%–60% for 5 min and 60%–95% over 1 min at a flow rate of 300 nl/min. For the library, standard DDA experiments were performed with a data-dependent top15 method. In brief, MS spectra (from m/z 375–1,550) were analyzed in the Orbitrap detector with resolution R = 60,000 at m/z 200. The 10 most intense peptide ions with charge states ≥2 were sequentially isolated to a target value of 3e6 and fragmented by Higher Energy Collision Dissociation (HCD) with a normalized collision energy setting of 28%. The maximum allowed ion accumulation times were 20 ms for full scans and 80 ms for MSMS, and the target value for MSMS was set to 1e5. The dynamic exclusion time was set to 20 s. Unfractionated peptides corresponding to the study samples were injected in single-shot analysis (number of replicates equal to 4) with a BoxCar scan method where instrument acquisition was controlled by MaxQuant Live software (version 1.2) keeping default scan protocol parameters (13).

Acquired raw data obtained by mass spectrometry analysis were analyzed using the MaxQuant (MQ) (14) version 1.6.10.43, and peptide lists were searched against the human Uniprot FASTA database (74470 Entries) with the Andromeda search engine (15). The main search was performed with an initial mass tolerance of 7 ppm. False discovery rate (FDR) for both protein and peptide identifications was set to a maximum of 1% with enzyme specificity set to Trypsin/P. A maximum of 2 missed cleavages was allowed, and the minimum peptide length was fixed at 7 amino acids and Carbamidomethylation of Cysteine was specified as a fixed modification. Peptides were identified with an initial precursor mass deviation of 7 ppm and a fragment mass deviation of 20 ppm. “Match between run algorithm” (MBR) in MaxQuant (16) was performed after constructing a matching library consisting of polled samples. For label-free protein quantitation (LFQ), we required a minimum ratio count of 2 (17). All proteins and peptides matching to the reversed database were filtered out. ProteinGroups.txt table from MQ output was analyzed using Perseus platform version (18). In brief, two-sample Student’s t-test was used to determine the significantly changed proteins between disease and control groups with a permutation-based FDR of 5%. For significant hits, minimal fold changes together with p values (controlled by the s0 parameter in Perseus) were used with a permutation-based FDR of 0.05 resulting from an s0 set to 0.1. Heatmap was generated using Heatmapper (19). Proteins were hierarchically clustered by setting Euclidean distance and average linkage method as parameters. Specific biological process enrichment and their plotting were generated in R Studio using R v4.0.3 and topGO (20) v2.42.0 passing the weight algorithm and Fisher test to the runTest function. A cutoff of 0.05 was applied, and the top 18 enriched terms were visualized with ggplot v3.3.3 (21). Protein–protein interaction (PPI) networks of unique upregulated and downregulated proteins are retrieved using the online version of Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database version 11.0, setting the maximum number of interactions for the first and second shell to not more than 10 interactors. Generated networks are imported in Cytoscape v3.8.2, where functional enrichment was performed using STRINGapp enrichment. Gene Ontology (GO) term sorting is performed by combining strength and number of input proteins retrieved for the term. Omics visualizer app (22) is used to show biologically relevant terms. For the other comparative analyses, two-tailed t-test and ANOVA test were used.

In order to characterize exosomes of different COVID-19 patients, plasma from 20 individuals who tested positive for SARS-CoV-2 infection by nasopharyngeal swab real-time PCR was collected in the acute phase of the disease (within 21 days from the diagnosis). Overall, we enrolled 11 males and 9 females with a median age of 57.5 years [interquartile range (IQR) 27.5–70] and a median time from diagnosis of 14 days (IQR 14–16). Patients with a maximal WHO clinical progression score [https://doi.org/10.1016/S1473–3099(20)30483–7] between 1 and 4 were classified as MILD, whereas those with a maximal score between 5 and 10 were classified as SEVERE (9) ( Supplementary Table S1 ). The selected MILD patients never transited to a SEVERE state during the course of the disease. As a control group, we enrolled 20 COVID-negative HDs. At first, we tested whether exosomes recovered from plasma of COVID-19 patients exposed any SARS-CoV-2-derived peptide. By using ExoTEST™, a platform for exosome quantification and characterization (23–25), and commercial anti-SARS-CoV-2-S antibodies for the detection of the S protein, we found that the SARS-CoV-2-S protein or derived fragments were clearly present in exosomes of COVID-19 patients but not in those of HDs, as expected ( Figure 1A ). Next, we subgrouped COVID-19 patient exosomes into MILD and SEVERE and analyzed whether SARS-CoV-2-S was more abundant in one of the two classes. Intriguingly, MILD patients had a higher amount of circulating SARS-CoV-2-S⁺ exosomes compared to patients with SEVERE symptoms ( Figure 1B ). To confirm this unexpected result, we recovered extracellular vesicles (EVs) from the plasma of either COVID-19 patients or HDs through sequential centrifugation and, after validating their size through NTA ( Figure S1A ), we used two alternative exosome isolation/characterization approaches: one based on anti-CD63-conjugated bead purification, followed by flow cytometry using anti-SARS-CoV-2-S-RBD antibodies for S protein detection ( Figure S1B ) and the other based on immunoprecipitation using anti-tetraspanins (CD63, CD9, CD81), followed by Western blotting using both anti-SARS-CoV-2-S RBD and anti-SARS-CoV-2-S1 antibodies (the latter specific for S protein subunit 1) in order to confirm the origin of the fragments detected. Flow cytometry confirmed that anti-SARS-CoV-2-S RBD antibodies bound exosomes of MILD COVID-19 patients with higher affinity compared to those of SEVERE patients, while they showed negligible binding to HD exosomes ( Figure 1C and Figure S1C ). Integrity of the analyzed exosomes was assessed by probing for cytosolic exosomal marker syntenin before and after permeabilization. Syntenin was detectable only after permeabilization, as expected for fully intact exosomes ( Figure S1D ). Western blotting revealed that both anti-SARS-COV-2-S RBD and anti-SARS-CoV-2-S1 antibodies recognized protein bands predominantly in exosomes of MILD patients. These bands had molecular weights (MWs) lower than 130 kDa ( Figure 1D ), thus corresponding to degradation fragments of the S protein, whose full-length monomer is 180 kDa. Purity of exosomal preparations was confirmed by positive detection of canonical exosomal markers, namely, CD63, CD9, TSG101, and syntenin, and lack of contaminant plasma proteins (i.e., ApoA) ( Figure 1E ).

Taken together, these results suggest that SARS-CoV-2-S-derived fragments are present in exosomes of all COVID-19 patients but are specifically and preferentially enriched in those of patients classified as MILD.

Next, we addressed whether the presence of SARS-CoV-2-S-derived peptides was associated to specific exosomal subpopulations by using the ExoView™ platform, which allows for single particle interferometric imaging measurements and analysis of differential coexpression of exosomal markers (26, 27). Exosomes were captured using antibodies against the canonical CD9, CD63, and CD81 exosome markers and platelet-specific CD41a ( Figure 1F ). CD41a was added to our strategy, since platelets function as an important exosomal source and are known to be hyperactivated in COVID-19 patients (28). By label-free detection, we observed that most of the exosomes were captured by anti-CD9 and anti-CD41a antibodies, and intriguingly, exosomes of MILD COVID-19 patients were remarkably more abundant compared to those of both SEVERE patients and HDs ( Figure 1G ). In addition, we found that exosomes from COVID-19 patients had an increased size distribution compared to HD exosomes, independently of the specific marker analyzed ( Figure S2 ). Interestingly, CD41a⁺ exosomes of MILD COVID-19 patients were larger (50–160 nm) than those of SEVERE COVID-19 patients (50–100 nm) ( Figure S2 , top), possibly indicating a different cargo for the two exosome groups. Next, we wanted to assess whether the SARS-CoV-2-S-derived fragments were evenly distributed in exosomes or enriched in one of the four exosome subpopulations. We found that SARS-CoV-2-S was detected in all the exosomes analyzed, except for the CD81⁺ subpopulation ( Figure 1H ). When comparing MILD and SEVERE COVID-19 patient-recovered exosomes, we found that SARS-CoV-2-S peptides were drastically more enriched in the CD41a⁺ and CD9⁺ subpopulations of those recovered from MILD patients ( Figure 1H ).

This result was confirmed using STED microscopy, a fast-performing super-resolution technique for resolving objects smaller than the light diffraction limit. By visualizing and quantifying the levels of SARS-CoV-2-S fragments on bead-purified CD9⁺ exosomes, we confirmed a significant enrichment of SARS-COV-2-S peptides in MILD COVID-19 patient compared to SEVERE COVID-19 patient exosomes ( Figure 2A and Figure S3 ).

Next, to further validate this result at nanometer levels, we performed TEM using a suspension immunogold labeling protocol optimized to detect SARS-CoV-2-S peptides on immunocaptured exosomes ( Figure 2B ). No gold deposition (namely, anti-SARS-CoV-2-S RBD labeling) was detectable in HD micrographs ( Figure 2C upper panel), whereas an increased level of gold deposition was evident on the ringlike surface of exosomes of MILD compared to SEVERE patient samples ( Figure 2C middle and lower panel). To better visualize SARS-CoV-2-S-derived peptides at the single-molecule level and define their spatial localization on the surface of exosomes, we performed single-molecule localization microscopy via TIRF/Direct-STORM with stochastic blinking reconstruction, as previously reported (29). Given the high density of blinking fluorophores needed for this approach, the experiment was performed only on MILD patient exosomes. Single-molecule imaging of MILD patient exosomes revealed that SARS-CoV-2-S-derived fragments were widespread and highly abundant across the entire exosomal surface ( Figures 2D–F ). Moreover, we also performed 3D-STORM on best-emitting samples in order to retrieve within the TIRF plan a correct volumetric localization of blinking molecules. Such analysis further confirmed the localization of RBD at the level of the exosome surface, with no signal localizing within the inner vesicle compartment ( Figure 2G and Supplementary Video 1 ).

The data obtained so far suggested that the higher abundance of SARS-CoV-2-S⁺ exosomes could be somewhat beneficial to MILD patients and contribute to their better outcome during viral infection. Among various possibilities, we hypothesized that exosomes could contribute to the adaptive immune response against SARS-CoV-2 infection by possibly exposing SARS-CoV-2-S-derived particles. To verify this assumption, at first, we analyzed the origin of SARS-CoV-2-S⁺ exosomes. Since all exosomes contain proteins derived from the parent cell and can partly maintain parent cell functionality, we characterized exosome origin using antibodies against major immune cell surface receptors. We found that SARS-CoV-2-S⁺ exosomes of MILD COVID-19 patients were mostly of B cell, dendritic cell, and monocyte/macrophage origin, as they displayed B-cell marker CD19, integrin CD11b, costimulatory molecule CD86, and MHC-class II HLA-DR ( Figure 3A ). This result indicated that most MILD patient exosomes containing SARS-CoV-2-S fragments are derived from antigen-presenting cells (APCs) and could possibly maintain antigen-presenting functions. Since proper antigen-presenting capability requires the presence of both adhesion molecules and costimulatory factors, we quantified the levels of adhesion factor Intercellular Adhesion Molecule 1 (ICAM-1) and of costimulatory molecules Human Leukocyte Antigen-DR isotype (HLA-DR), B7-2, and B7-H1 (30–32) on the surface of patient-recovered exosomes. We found that all MILD patient exosomes were in general more enriched in all the markers analyzed ( Figure 3B ). These data suggested that SARS-CoV-2-S⁺ exosomes recovered from MILD COVID-19 patients could act per se as antigen-presenting vehicles and modulate antigen-specific T-cell responses. To verify this hypothesis, at first, we incubated in vitro exosomes of COVID-19 patients and HDs with autologous CD4⁺ T cells and analyzed cell proliferation. We found that COVID-19 patient-recovered exosomes were capable of stimulating CD4⁺ T-cell growth, while HD-recovered exosomes were not ( Figure 3C ). Most interestingly, exosomes of MILD COVID-19 patients were more efficient in stimulating CD4⁺ T-cell growth compared to those of SEVERE COVID-19 patients ( Figure 3C ). To assess whether exosomes of MILD COVID-19 patients induced T-cell growth by actual T-cell activation, we tested their capability to modulate the expression of middle, CD25, and late, HLA-DR, T-lymphocyte activation markers cocultured in matched CD4⁺ T cells. We found that both markers were actually drastically upregulated only in the samples cocultured with exosomes of MILD COVID-19 patients ( Figure 3D ), suggesting that T-cell proliferation was indeed a consequence of efficient T-cell activation triggered by exosomes. Comparable results were obtained when we repeated the assay using allogeneic T cells derived from HDs, thereby excluding the possibility that the lower immunostimulatory activity of SEVERE patient exosomes could be due to dysfunctional T cells (data not shown).

Next, we checked whether antigen presentation was driving exosome-dependent T-cell activation. We replicated the previous experiment but blocking MHC-II with specific antibodies. We found that MHC-II blockade significantly reduced exosome-dependent T-cell activation ( Figure 3E ), confirming that an antigen-presenting activity of MILD COVID-19 patient exosomes was the actual driving force of CD4⁺ T-cell activation. This evidence was further confirmed by the fact that MILD patient exosomes stimulated CD4⁺ T cells to produce high levels of IL-2, a cytokine whose secretion is known to be triggered early during antigen presentation ( Figure 3F ). Interestingly, two other cytokines secreted during antigen presentation, interferon (IFN)γ and tumor necrosis factor (TNF)α, were not enriched possibly due to their different induction kinetics ( Figure 3F ).

Next, we investigated if MILD COVID-19 patient-recovered exosomes were capable of favoring CD4⁺ T-cell activation driven by other APC sources. We checked if MILD COVID-19 patient-recovered exosomes could enhance CD4⁺ T-cell activation. Coculturing MILD COVID-19 patient exosomes with CD4⁺ T cells and autologous monocytes as primary APC source caused a significant increase in CD4⁺ T-cell HLA-DR protein levels ( Figure 3G ), an effect that was totally abolished when we blocked MHC-II, confirming the requirement of cross-presentation for efficient CD4⁺ T-cell activation and indicating a peculiar antigen cross-presentation activity of MILD COVID-19 exosomes.

Our findings suggest that MILD COVID-19 patient-recovered exosomes are capable of favoring CD4⁺ T-cell activation by functioning as an antigen-presenting source and by promoting T-cell activation.

With the intent to corroborate our results and better define the functional roles of COVID-19 patient-recovered exosomes, we defined their proteomes. For each patient class (MILD and SEVERE), we selected the 4 exosome samples with the highest SARS-CoV-2-S positivity (indicated by an asterisk in Supplementary Table S1 ), while for HDs, exosome samples were selected randomly. Mass spectrometry was performed on whole exosome protein content. Comparing the proteomes of the three sets of samples, we identified 130 differentially abundant proteins ( Supplementary Table S2 ), 92% of which were previously annotated in public EV databases [EXOCARTA (33)] as EV/exosome components, confirming the efficacy of our exosome purification. As expected, and in line by previously reported studies regarding conventional exosome purification methods (34, 35), highly abundant apolipoproteins were detected in all three sample types analyzed ( Supplementary Table S2 ) but did not interfere with downstream analyses. Pairwise correlation analysis of protein contents clearly revealed that samples of the same class displayed a high internal correlation mirrored by lower correlation with samples of other classes ( Figure S4A ), indicating a high homogeneity among exosomes recovered from different donors. This result was further confirmed by hierarchical clustering that, by gathering the 130 proteins into 7 groups, clearly separated MILD COVID-19-, SEVERE COVID-19-, and HD-recovered exosomes ( Figure 4 , left). To assess the characteristics of each cluster found, we carried out a comparative analysis using FunRich (36). Our results revealed that the clusters linked to HD exosomes correlated with complement activity and immune response activation, those linked to SEVERE COVID-19 patient exosomes correlated with immune/inflammatory responses and protein metabolism, while those mainly related to MILD COVID-19 patient exosomes were related to immune response, cell growth, signal transduction, and MHC class II receptor functionality ( Figure 4 , right), well confirming our in vitro results that depicted MILD COVID-19 patient exosomes as strong stimulators of CD4⁺ T-cell activation and growth.

Next, we focused on differentially abundant proteins. We identified 102 proteins differentially present in MILD COVID-19- vs. HD-recovered exosomes and 87 proteins differentially abundant in SEVERE COVID-19- vs. HD-recovered exosomes. Here, 72 proteins were shared between the two groups, while a total of 45 proteins were specific for one group or the other ( Figure 5A and Figure S4B ). By focusing on the latter 45 proteins, we selected the 22 with the highest enrichment. These included 16 proteins for the MILD and 6 for the SEVERE patient group ( Figure 5B ). GO term analysis indicated that the proteins recovered from MILD COVID-19 patient exosomes were mainly involved in immune cell activation, while those recovered from SEVERE COVID-19 patient exosomes were mainly involved in stress and inflammatory responses ( Figure 5C ). Such results were also confirmed by interrogating the STRING (37) database. Exosomes recovered from MILD COVID-19 patients showed an enrichment in proteins involved in pathways related to antigen processing, presentation of exogenous peptides (FDR <1.56E-07), and myeloid cell activation (FDR <6.65E-08) ( Figure 5D , left), while those recovered from SEVERE COVID-19 patients showed an enrichment in proteins involved in pathways regulating acute inflammatory responses (FDR <3.58E-13) and complement activation (FDR <3.07E-14) ( Figure 5D , right), further strengthening the notion that MILD and SEVERE patient exosomes are drastically different and possess different immunomodulatory functions, the first mainly involved in immune cell activation and the second principally involved in inflammation.