Extracellular vesicles from long COVID patients increase cell stress through disrupted miR-204 and p53 pathways controlled by RUNX2

The aim of this study was to investigate the role of extracellular vesicles (EVs) in the pathogenesis of Long COVID (LC), focusing on their molecular composition and their functional effects on multiple cell types potentially involved in LC. We hypothesized that LC-EVs carry bioactive molecules associated with cellular stress contributing to the persistence of symptoms such as fatigue and breathlessness.

This was a comparative, in vitro experimental study involving EVs isolated from peripheral blood samples of 32 participants: 16 individuals diagnosed with Long COVID and 16 age- and sex-matched healthy controls EVs were characterized using nanoparticle tracking analysis (NTA), atomic force microscopy (AFM), transmission electron microscope (TEM) and flow cytometry to assess size, morphology and surface marker expression.

Functional assays were conducted by exposing various human cell lines—including lung epithelial cells (A549), primary human lung fibroblast (IMR90), human umbilical vein endothelial cells (HUVECs), aortic smooth muscle cells (ASMCs) and mesenchymal stem cells (MSCs)—to patient-derived EVs. We measured activation of stress-related pathways (SESN1/2, p53, p21), alterations in RUNX2 expression, and mitochondrial respiration capacity under metabolic stress. Transfection experiments were also employed to explore post-transcriptional regulatory mechanisms, particularly involving miR-204.

Samples from 16 long COVID patients (8 males and 8 females) and 16 healthy (8 male and 8 female) were analyzed to evaluate the role of EVs. The healthy subjects (controls) had previously been infected with SARS-CoV-2 and had fully recovered, exhibiting no acute symptoms and no evidence of long COVID—defined, according to WHO criteria, as new or persistent symptoms lasting three months or more post-infection. Inclusion and exclusion criteria are described elsewhere (respicovid Crisafulli [26]): briefly, we evaluated outpatients previously hospitalized for SARS-CoV-2 infection. Male and female long COVID patients were BMI and age matched. All outpatients met the World Health Organization (WHO) criteria for the definition of long COVID (https://www.ncbi.nlm.nih.gov/books/NBK605675/↗). Each participant gave written informed consent. All subjects underwent a comprehensive clinical evaluation and a medical history interview to exclude any comorbidities. Written informed consent was obtained from all participants, and the study was approved by the Ethics Committee of the Azienda Ospedaliera Universitaria Integrata of Verona, Italy (no. 2785CESC). The study design and procedures adhered to the principles of the Declaration of Helsinki.

The 6MWT was conducted on a straight 30 m rout. The start and endpoints were indicated by two Skittles, and the long COVID patients were instructed to walk as far as possible for six minutes. The protocol followed the American Thoracic Society (ATS) [27] guidelines. The distance walked (meters) was recorded. Predicted values were calculated via the equation by Enright et al. [28]. Heart rate (bpm) and pulse oximetry (sPO2) were monitored at rest and every minute (Pulsox-5, Konica Minolta Inc). Basal, peak and first-minute recovery blood pressure were also measured. 6MWT reference values for healthy subjects, obtained post-pandemic by De Soomer et al.show that healthy individuals perceive significantly less effort in terms of both dyspnea and muscle fatigue [29]. Taking these normative data into account, we did not perform the 6MWT in the control group.

Lung function tests were conducted in accordance with international guidelines [30] and were expressed as percentages of predicted values (Official Statement of the European Respiratory Society [31, 32]). Measurements were obtained using a flow-sensing spirometer (Jaeger MasterScreen PFT System) connected to a computer for data analysis. The parameters recorded included forced vital capacity (FVC), forced expiratory volume in the first second (FEV1), and total lung capacity (TLC). Diffusing capacity (DLCO) and transfer coefficient (KCO) were assessed using the single breath method.

Extracellular vesicles (EVs) were isolated using the qEVoriginal-70 nm Gen 2 column (Izon Science Limited ©), as previously reported [33]. Specifically, fractions collected between 0 and 2 mL were selected to reduce contamination from serum proteins [34]. The detection of proteins using Amido Black staining showed no traces of albumin in the selected EV isolation method. (Figure S1). In addition, the efficiency of this purification step was verified by Bradford colorimetric assay. Notably, the EV-containing fractions did not exhibit any reaction with the Bradford reagent (Figure S2), and no color change was observed. In contrast, fractions collected between 3.2 and 8 mL displayed a strong blue coloration, confirming the presence of serum proteins (Figure S2). These results indicate that the biological effects observed in downstream experiments can be solely attributed to the proteins and molecular components contained within the extracellular vesicles, which were successfully purified from serum protein contamination.

In addition, Western blot analysis was performed to assess the presence of standard EV markers (Figure S3). EV samples were prepared using two distinct protocols depending on the markers to be analyzed. For the detection of TSG101, Syntenin-1, ALIX, and Calnexin, EVs were boiled for 10 min at 95 °C in 6× Laemmli SDS sample buffer (VWR International, Dietikon). For the analysis of CD63 and CD81, EVs were instead incubated in 6× Laemmli sample buffer without SDS. Protein separation was performed on 4–20% Mini-PROTEAN^® TGX™ precast gels (Bio-Rad Europe), followed by transfer onto PVDF membranes using a semi-dry transfer system (Bio-Rad). Membranes were incubated with the following primary antibodies: anti-TSG101 (1:1000, Abcam, ab125011), anti-Syntenin-1 (1:1000, Abcam, ab19903), anti-ALIX (1:1000, Abcam, ab186429), anti-Calnexin (1:2000, Abcam, ab133615), anti-CD63 (1:1000, Invitrogen, 10628D), and anti-CD81 (1:1000, BD Biosciences, 555675). Thus, treatment of A549 cells with the later fractions (protein-rich fractions) of EVs isolationfrom either control or Long COVID samples did not induce significant changes in RUNX2 expression (Figure S4).

To minimize intra-group variability and obtain representative profiles, EVs were isolated from pooled plasma samples within each group (Long COVID patients and healthy controls), and the pooled samples were used for all downstream analyses, including NTA, TEM and flow cytometry.

NTA is based on the dynamic light scattering (DLS) phenomenon where the diffusion coefficient of each tracked particle is calculated in an aqueous environment by the Stokes–Einstein equation [35]. Thus, NTA was performed by using Malvern NanoSight NS300 instrument (Worcestershire, UK). NTA was also used to estimate EVs concentration before starting subsequent experiments.

AFM was performed using an NT-MDT Solver Pro atomic force microscope (Moscow, Russia) with NT-MDT NSG01 golden-coated silicon tip in semi-contact mode, applying a scanning frequency of 1 Hz in order to produce optimized AFM images. Furthermore, the microscope was calibrated by a calibration grating (TGQ1 from NT-MDT) to avoid nonlinearity and hysteresis in the measurements. EVs were first loaded on a bracket covered by an inert mica surface, and after 30 min of solvent evaporation, the analysis was performed. Finally, images were processed with the Scanning Probe Image Processor (SPIP™) program (Friis Jan, J. Scanning Probe Image Processor 4.8) [36].

One drop of the sample (about 25 µl) was placed on a 400 mesh holey film grid; after staining with 2% uranyl acetate (for 1 min) the sample was observed with a Tecnai G2 (FEI) transmission electron microscope operating at 100 kV. At least 20 images for each condition were captured with a Veleta (Olympus Soft Imaging System) digital camera.

EVs samples were subjected to surface epitopes bead-based flow cytometry analysis (MACSPlex Exosome kit Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer protocol. In brief, 4.62 × 10⁷ EVs from each condition (Control- and LC-EVs) were diluted in MACSPlex buffer to a total volume of 120 µl and incubated with MACSPlex exosome capture beads overnight. APC-labeled CD6, CD63 and CD81 antibodies were used for counterstaining. APC-A values were acquired with a BD LSRFortessa X-20 cytometer (BD biosciences, Heidelberg, Germany) and the FlowJo software (Ashland, Oregon, United States) was used to analyze data. For each sample, the median signal intensity of the signals detected for the CD9, CD63 and CD81 capture beads was calculated, and their geometrical mean was used as the normalization factor for each sample.

The A549 cell line was obtained from the American Type Culture Collection (CCL-185-ATCC ATCC, Philadelphia, PA, United States). Cells were maintained at 70% of confluency in DMEM (Cat. Number 11965092, Gibco, Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% FBS (Cat.Number 10270106, Gibco, Thermo Fisher Scientific, Waltham, MA, USA) and 1% penicillin-streptomycin (PSA) (Cat.Number 17-745E, Lonza, Basel, Switzerland). IMR90 cell line was obtained from the American Type Culture Collection (CCL-186-ATCC, Philadelphia, PA, United States). Cells were maintained at 70% of confluency in DMEM, supplemented with 10% FBS and 1% penicillin-streptomycin. Human Aortic Smooth Muscle Cells (HAOSMCs) (HAOSMCs) (Cat. Number C-12533, PromoCell, Heidelberg, Germany) were maintained at 70% of confluency in Smooth Muscle Cell Growth Medium 2 (Cat.Number C-22062, PromoCell, GmbH Sickingenstr. 63/65 69126 Heidelberg, Germany. Cat.No. C-22062), enriched with the proprietary SupplementMix (Cat.Number C-39267, PromoCell, GmbH Sickingenstr. 63/65 69126 Heidelberg, Germany Cat.No. C-39267) containing FCS, Epidermal Growth Factor, Basic Fibroblast Growth Factor and Insulin; mediums were enriched with 1% PSA. Human bone marrow-derived mesenchymal stem cells (hBM-MSCs) (Cat.Number C-12974, PromoCell, Heidelberg, Germany) were maintained at 70% of confluency in the MesenPRO RS Basal Medium (12747-010, Gibco, Life Technologies Corporation, 3175 Staley Rd., Grand Island, NE, USA) supplemented with 10% supplement mix and 1% penicillin-streptomycin-amphotericin (Cat.Number 17-745E, Lonza, Basel, Switzerland). HUVEC cells were obtained from Promocell (Cat. No: C-12200, Heidelberg, Germany) and maintained at 70% of confluency in presence of M200 medium (Cat.Number M200500, Thermo Fisher Scientific, Waltham, MA, USA), supplemented with 10% low serum growth supplement (LSGS) (Cat. Number S00310 Thermo Fisher Scientific, Waltham, MA, USA) and 1% penicillin-streptomycin amphotericin.

All cells were cultured in a humidified CO₂ incubator at 37 °C. Cells at 80% confluence were treated with EVs extracted from serum pools belonging to long covid patients or controls. EVs were added directly to fresh medium, with a final concentration of 9.8 × 10⁸ EVs/mL. Treatment with the conditioned medium lasted for 18 h before collection. The cells were collected using 1% trypsin (Cat. Number 59427 C, Merck, Rahway, New Jersey (USA). The EV concentration and exposure time (9.8 × 10⁸ EVs/mL for 18 h) were selected based on preliminary experiments conducted in our laboratory confirmed that this dose did not induce cytotoxic effects and ensured sufficient EV uptake in cells.

A549, HUVEC, HAOSMC, and MSCs cells were separately seeded onto chamber slides, each in their respective basal media. Cells were then stained with a vital fluorescent dye, the Vybrant Cell Labeling Solution (Invitrogen, Waltham, MA, USA), using 3,3′-dioctadecyloxacarbocyanine perchlorates (DiO), which emits green fluorescence at 501 nm. The cell nuclei were stained with Hoechst dye (Thermo Fisher Corporation, Waltham, MA, USA). EVs were stained with 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI), a red fluorescent dye with emission at 565 nm, and added to the cells for conditioning.

Following an 18-hour incubation, cells were fixed with 4% paraformaldehyde (PFA) and visualized using the EVOS M5000™ Core Imaging System (Invitrogen, ThermoFisher Scientific, Waltham, MA, USA).

Immunofluorescence analysis was conducted as previously described by Deiana et al. [37] on A549 cells conditioned with EVs. Briefly, cells were fixed with 4% paraformaldehyde (PFA), permeabilized with 0.5% Triton X-100, and processed as previously reported [37]. The primary antibodies CK19 (Ref: PA0468, Leica, Wetzlar, Hesse) and Vimentin (Ref: PA0640, Leica, Wetzlar, Hesse) were applied as specified in the respective datasheets and incubated overnight at 4 °C.

The slides were then incubated with Alexa-Fluor 488-conjugated anti-goat secondary antibodies in mice (Cat. Number A25618, Invitrogen, ThermoFisher Scientific, Waltham, MA, USA), diluted 1:300 as recommended by the datasheet. Nuclear staining was performed using ProLong™ Gold Antifade Mountant with DAPI (Cat. Number P36941, Invitrogen, ThermoFisher Scientific, Waltham, MA, USA). Negative controls were performed by using only secondary antibodies (Figure S5). Images were captured using the EVOS M5000™ Core Imaging System (Invitrogen, ThermoFisher Scientific, Waltham, MA, USA).

The RUNX2 gene was cloned into the pcDNA3 expression vector, as we previously reported [37]. Briefly, the amplified RUNX2 gene sequence (NM_001024630.4) was initially inserted into the pCR™ 2.1 vector (Thermo Fischer Scientific, Waltham, MA, USA), then excised using EcoRV/XhoI digestion and cloned into the pcDNA3-Flag-HA vector (Addgene plasmid #10792, Watertown, MA, USA) [37]. Transfection of the RUNX2-expressing plasmid was carried out using RUNX2 Lenti ORF particles from Origene Technologies (Rockville, MD, USA), following the manufacturer's instructions. A549 cells were grown to 70% confluence and subsequently transfected with 10 µg/mL RUNX2-expressing pcDNA3 vector, using Lipofectamine 3000 (Thermo Fisher Scientific, Waltham, MA, USA; reference: L3000-008), as we previously reported [38].

To evaluate the increase in hsa-miR-204-5p expression, A549 were transfected with a mirVana™ hsa-miR-204-5p mimic (Catalog number: 4464066 ID: MC11116, Invitrogen by Thermo Fisher Scientific) at different concentrations (5 nM, 10 nM, 15 nM, 20 nM, 25 nM) or silencer-negative control siRNA (Cat#: AM14611, Ambion by Thermo Fisher Scientific, Waltham, MA, USA). The A549 were plated at a density of 3.5 × 10⁵ cells in T25 flasks and cultured in the presence of DMEM. At a cell confluency of 70%, transfection was carried out using Lipofectamine 3000 (L3000-008, Invitrogen by Thermo Fisher Scientific Baltics UAB, Vilnius, Lithuania). Cells were incubated for 24 h hours with the transfection complex; after incubation cells were collected and samples stored until use.

For RUNX2 silencer, transfection was carried out according to the manufacturer's instructions using Lipofectamine 3000. A549 were transfected with Silencer SelectPre-designed siRNA RUNX2 (REF: 4392420, siRNA ID: s2456 Invitrogen by Thermo Fisher Scientific) and silencer-negative control siRNA. The cells were transfected for 24 h and treated with Long Covid EVs for 18 h. After transfection and treatment, cells were harvested using trypsin 1%, and the cell pellet was processed for protein extraction.

Protein concentration was determined using the BCA Protein Assay Kit (Cat. Number eMP014500, Quantum Protein, Bicinchoninic Protein Assay, Euroclone S.p.A. Società a Socio Unico Via Figino, 20/22 20016 Pero, MI) and measured with a VICTOR Microplate Reader (PerkinElmer, Waltham X3-2030-0030, MA, USA). Subsequently, protein samples were normalized with Pierce™ RIPA buffer (Cat.Number: 78501 Thermo Fisher Corporation, Waltham, MA, USA) 1X and Loading Buffer. After incubating at 99 °C for 7 min, samples were loaded onto precast Mini-PROTEAN^® TGX gels (Bio-Rad, 2000 Alfred Nobel Dr, Hercules, CA, United States, California) for SDS-PAGE. SDS-PAGE was conducted in Running Buffer containing Tris (Cat.Number T1503, Merck, Rahway, New Jersey (USA)), Glycine (Cat.Number ECB2072, Euroclone S.p.A. Società a Socio Unico Via Figino, 20/22 20016 Pero, MI) and SDS (Cat. Number 11667289001, Merck, Rahway, New Jersey (USA)at 125 V for 15 min, then at 145 V for 1 h. Proteins were transferred to a PVDF membrane (Cat.Number: 88518 Invitrogen- Thermo Fisher Corporation, Waltham, MA, USA), activated with ethyl alcohol and water, and immersed in Transfer Buffer 1 × (1 L of 10X solution, 30.3 g of Tris (Cat.Number T1503, Merck, Rahway, New Jersey (USA)), 144.0 g of Glycine (Cat.Number ECB2072, Euroclone S.p.A. Società a Socio Unico Via Figino, 20/22 20016 Pero, MI, Italy) for 15 min. The transfer, using a Mini-PROTEAN^® Tetra Cell apparatus (Bio-Rad, 2000 Alfred Nobel Dr, Hercules, CA, United States, California) at 100 V in ice, lasted 1 h. After, the membrane underwent Amido Black staining (Cat. Number A8181- Sigma-Aldrich-St.Louis, Missouri, USA) and verification, with bands observed after destaining in MilliQ water for 10 min (Cat. Number: P7170- Sigma-Aldrich-St.Louis, Missouri, USA) the membrane was blocked with 5% fat-dry milk (Cat. Number 42590.02, SERVA Electrophoresis GmbH, Carl-Benz-Str., 69115 Heidelber) in TBS-Tween (Cat.Number 1706435, Bio-Rad, 2000 Alfred Nobel Dr, Hercules, CA, United States, California) for 1 h, followed by overnight incubation with the primary antibody at 4 °C. Simultaneously, the following antibodies were used: RUNX2 1:1000 55–62 kDa, (Cat.Number 8486 Cell Signaling- 3 Trask Lane Danvers, MA 01923), p53 1:1000 53KDa.(Cat.Number: FNab06083, FineTest), p21 1:1000 21KDa (Cat. Number: GTX629543, GeneTex), SESN1 1:500 57KDa, (Cat.Number: PA5-98142- Invitrogen Thermo Fisher Corporation, Waltham, MA, USA), SESN2 1:1000 54KDa (Cat Number: ab-178518 Abcam), Phospho-p53 (Ser15) (16G8) (Cat. N°9286, Cell signaling technologies, Danvers, Massachusetts, USA). β-actin 1:10000 42–55 kDa (Cat. Number: MA1-140-Invitrogen Thermo Fisher Corporation, Waltham, MA, USA) served as a housekeeping protein.

Total RNA was extracted using the RNeasy^® protect mini kit (Qiagen, Hilden, Germany) following the manufacturer's protocol. The quantity and quality of RNA samples were assessed using the 'Qubit™ RNA HS assay kit' (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) and a Qubit 3 Fluorometer (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA -REF Q3321). RNA quality was assessed by spectrophotometric analysis by using NanoDrop ND-100 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA). Both A260/280 and A260/230 ratios were measured to evaluate purity, and our samples showed values close to ~ 2.0 (A260/280) and 2.0–2.2 (A260/230), which are indicative of good RNA quality.

For total RNA, reverse transcription was performed with the first-strand cDNA synthesis kit (GE, Healthcare Bio-sciences AB, Uppsala) following the manufacturer's protocol. The thermocycler steps were performed using a GeneExplorerTM Thermal Cycler.

The Real-Time PCR was performed using the TaqMan Universal PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) with the following commercially pre-designed probes (Table S1) and the iTAQ Universal SybrGreen SuperMix (Bio-Rad, 2000 Alfred Nobel Dr, Hercules, CA, United States, California) with the primers reported (Table S2).

Next, The MicroAmp Optical 96-Well Reaction Plate (applied biosystems, Thermo Fisher Scientific, Waltham, MA, USA) was used to load the 19 µL of PCR mix into designed wells. Next, 1 µL (equivalent to 10 ng, after concentration adjustment) of cDNA was added. Three copies of each sample were loaded. Following a brief centrifugation, the plate was covered and sealed using the proper membrane (Opti-seal, AB analitica). For this procedure, the LineGene 9620 Real-Time PCR System (Aurogene, Hangzou Bioer Technology) was used. The PCR stage was repeated for 45 cycles following the different steps and lengths using the Taqman Master Mix or the SybrGreen SuperMix as previously reported [39]. At least three independents' analyses were performed, and each analysis was performed in triplicate with Ct values averaged.

MiRNAs were extracted from EVs using the qEV RNA Extraction Kit (Izon Science Limited) according to the manufacturer's instructions. The concentration and purity of the extracted miRNAs were assessed using a Nanodrop ND-1000 (Thermo Scientific). Specific miRNA 204-5p was reverse-transcribed into cDNA using the TaqMan™ MicroRNA Reverse Transcription Kit (4366596, Applied Biosystems by Thermo Fisher Scientific, Baltics UAB, Vilnius, Lithuania) with the miRNA-specific primer, following the manufacturer's protocol. PCR amplification was performed in a total volume of 20 µL, including 2 µL of cDNA and TaqMan Universal PCR Master Mix. TaqMan miRNA-specific probes for hsa-miR-204-5p-5p (000508, 20X, FAM; Thermo Fisher Scientific, Life Technologies Corporation, Carlsbad, CA, USA) were obtained from Applied Biosystems (Thermo Fisher Scientific, Life Technologies Corporation, Carlsbad, CA, USA). U6 snRNA (001973, 20X, FAM; Thermo Fisher Scientific, Life Technologies Corporation, Carlsbad, CA, USA) was used as housekeeping control, showing stable expression across samples. The PCR was conducted for 45 cycles. Results were normalized to the housekeeping U6 snRNA, and relative fold expression differences were calculated using the ΔΔCt method as previously reported [40].

A549 cells were plated in a T75 flask. Cells were detached with 0.05% trypsin and counted by trypan blue exclusion. After centrifugation at 200 gmax for 5 min at 25 °C, medium was removed, and cells were suspended in PBS and aliquoted for respirometry. A total of 2 × 10⁶ A549 cells were resuspended directly in the 2 mL Oroboros Oxygraph-2k (Oroboros Instruments, Corp., Innsbruck, Austria) chambers, filled with Mir05 medium (0.5 mM EGTA, 3 mM MgCl2·6 H2O, 60 mM lactobionic acid, 20 mM taurine, 10 mM KH2PO4, 20 mM HEPES, 110 mM sucrose, 1 g/L BSA). The published substrate-uncoupler-inhibitor titration (SUIT)−008 O2 ce-pce D025 was used to measure O₂ Flux in pmol⋅s^–1⋅2 million^–1 cells. The assay was performed at 37 °C. Catalase was added to create MiR06, with a final concentration of 280 U/mL to induce hyperoxia. After measuring endogenous whole-cell respiration (ROUTINE state), cells were permeabilized by the addition of digitonin. In particular, we identified the optimal concentration of digitonin for cell permeabilization by performing titration experiments determining that a final concentration of 50 µg/mL was the most effective. Thus, substrates and inhibitors were added in the following order and final concentrations: pyruvate (5 mM) and malate (2 mM), ADP (2.5 mM), glutamate (10 mM), succinate (10 mM), carbonyl cyanide m-chlorophenylhydrazone (CCCP, multiple 0.5 μm titrations), rotenone (0.5 µM), and antimycin A (2.5 µM). Residual oxygen consumption was subtracted from the oxygen flux as a baseline for all respiratory states. Data analysis was performed using DatLab 5.2 software (Oroboros, Paar, Graz, Austria). Chemicals were purchased from Sigma Aldrich/Merck Life Science. This titration method was completed within 90 min.

In order to assess mitochondrial function, A549 cells were seeded at the density of 1.5 × 10⁴ cells/well or 2 × 10⁴ cells/well and IMR90 cells were seeded at density of 1 × 10⁴ cells/well in a V7 XFe24-well cell-culture microplate and contextually treated with Control (CTRL) or Long Covid EVs (LC EVs). After 18 h, the media was changed with XF-RPMI medium (Seahorse Bioscience, cat. No. 103576-100) supplemented with 10 mM glucose (Agilent Technologies, Milan, Italy 103577-100), 1 mM Sodium Pyruvate (Agilent Technologies, Milan, Italy 103578-100), and 2 mM glutamine (Agilent Technologies, Milan, Italy 103579-100), pH 7.4, and the cells were incubated at 37 °C in a non-CO₂ incubator for 1 h. Then, the Cell Mito Stress Test was performed by using the specific kit (Seahorse XF Cell Mito Stress Test Kit. 103015-100) in a Seahorse XFe24 Extracellular Flux Analyzer (Agilent Technologies, Milan, Italy), following manufacturer's instructions. The oxygen consumption rate (OCR), as a readout of mitochondrial function, was measured at the baseline and after sequential addition of 1 mM oligomycin A (port A), 2 mM of carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP, port B), and 1 mM each of Rotenone and Antimycin A (port C). Raw OCR data were normalized to the DNA content per well (µg DNA/well) that was quantified with the CyQUANT Cell proliferation assay kit (Thermo Fisher Scientific, cat. No. C35007), in accordance with manufacturer's instructions. The OCR profiles are shown in Figure S6.

The data are presented as mean ± standard deviation (SD) or mean ± standard error of the mean (SEM), as indicated. Statistical analyses included Student's paired t-test for two-group comparisons and one-way or two-way analysis of variance (ANOVA) for comparisons across multiple groups. Post hoc multiple comparisons were conducted using Dunnett's/Tukey's test to determine specific group differences. Adjustments for multiple comparisons were applied to control the family-wise error rate, ensuring the overall type I error rate remained at α = 0.05. Although normality and homoscedasticity were not formally tested, the use of parametric methods is supported in small to moderate samples [41].

For the statistical analysis of qPCR data, we used the log2 fold change values calculated via the ΔΔCt method. Analyses were based at least on two (for analyses performed on Oroboros) or three independent experiments, each performed in triplicate. Statistical evaluations were conducted using SPSS for Windows, version 22.0 (SPSS Inc., Chicago, IL, USA), or GraphPad Prism (version 9), with a p-value ≤ 0.05 considered statistically significant.

Sixteen patients were evaluated 4.5 ± 0.6 months after COVID19. Characteristics of long COVID patients are reported in Table 1. The severity of the initial infection required hospitalization in all patients. Ten of 16 needed O₂ therapy and 2 were admitted into the intensive care unit. The duration of the hospitalization was 9.7 ± 6.2 days. All Long COVID patients consistently reported signs of fatigue, often described as overwhelming and persistent, interfering significantly with daily activities and quality of life. In particular, persisting symptoms included fatigue in seven patients; one patient reported resting dyspnea, and four experienced exercise-induced dyspnea. Additionally, three Long COVID patients reported a persistent cough. Spirometry parameters of subjects are reported in Table 2. The 6-minute walking distance Test (6MWT) revealed a normal functional status in 15/16 patients, with only one male long COVID patient that covered a distance < 80% of the predicted value. Interestingly, at the modified CR10 Borg Scale [42] after the 6-minute walking, patients reported an exertional breathlessness rate of 2.5 ± 1.86 and a leg discomfort rate of 3.25 ± 1.94. Leg discomfort seemed to be more pronounced than perceived dyspnea (p < 0.05). Pulse oximetry values at rest (97.1 ± 1.1) and at peak exercise (96.5 ± 1.6) showed no significant differences (p = 0.120); however, 2 patients had a sPO2 < 95% after the 6MWT. Data of 6MWT are reported in Table 3.

EVs of both control and Long COVID subjects (Fig. 1A-N) were analyzed using Nanoparticle Tracking Analysis (NTA), Atomic Force Microscopy (AFM) and Transmission Electron Microscope (TEM). These techniques allowed us to accurately assess the size distribution and morphological characteristics of the EV populations, providing a comprehensive comparison between the two groups.

The differences observed (Table 4), both in the mean and mode values, are within the standard deviation range.

TEM provided images of vesicles after dehydration and negative staining. The extra cellular vesicles appear predominantly spherical and the bilayer structure can be distinguished, as indicated by the higher contrast observed along the vesicle edges (Fig. 1I-L). Therefore, based on this analysis, there are no substantial differences in the size distribution between the two populations.

Moreover, the quantification of EVs did not reveal any significant differences between the Control and Long COVID groups (Table 5).

No statistically significant difference was observed between the groups (p = 0.58) (Ten independent experiments).

We then analyzed the exosome surface markers using flow cytometry. As shown in Fig. 1L-M, while the levels of most markers were comparable, we observed significant differences for specific markers. Specifically, we identified a significant overexpression of markers associated with platelet hyperactivation in EVs derived from long COVID (LC) subjects (CD29, CD41b, CD42a, and CD69).

To investigate the effects of LC-EVs on lung cells, we exposed A549 cells to EVs isolated from either control subjects or LC patients for 18 h. As shown in Fig. 3A, EVs from healthy donors and Long Covid (LC) subjects could cross cellular membranes of lung cells and disperse within the cells in an identical manner (Fig. 2A).

Notably, vimentin and CK19 were significantly overexpressed in lung cells treated with LC-EVs when compared to control EVs (Fig. 2B). However, the migratory capacity of the cells, evaluated by scratch assay (Fig. 2C) as well as cell viability (Fig. 2D) of cells treated with control or LC-EVs revealed no significant differences. Moreover, our results showed a marked overexpression of RUNX2 in cells treated with LC-derived EVs compared to controls (Fig. 2D-E). Interestingly, also SESN2 levels were increased in cells treated with LC-EVs (Fig. 2E) while SESN1 levels were similar (Fig. 2E). Protein levels of p53 and p21 were higher in cells treated with LC-EVs than in controls too (Fig. 2E). The increased p53 phosphorylation (Fig. 2G), together with the upregulation of SESN2 and p21 (Fig. 2H), supports the activation of p53 and its functional transcriptional activity.

To explore whether the proteins modulated by LC-EVs are functionally connected, we performed an exploratory protein-protein interaction (PPI) analysis using the STRING database (https://string-db.org/↗; accessed on 10 September 2024). It is important to note that the list of proteins analyzed was pre-selected based on experimental modulation; therefore, this analysis is a descriptive support to identify potential shared pathways or biological functions. As shown in Fig. 3A-B, the predicted interactions within the functional network were statistically significant (p-value: 0.00658), indicating a strong functional relationship between the proteins involved. In addition, the functional enrichment analysis (protein-protein interaction enrichment) revealed the involvement of the modulated proteins in several biological processes such as stress induced senescence and mitochondrial DNA metabolic process (Fig. 3C).

To investigate mitochondrial function, we used high-resolution respirometry (Oroboros O2k) under hyperoxic conditions to maximize cellular respiratory capacity (Fig. 3D). After an 18-hour stimulation with EVs from control and long COVID (LC) subjects, oxygen consumption was significantly higher in LC-EV-treated cells Fig. 3E. To understand the mechanisms underlying this enhanced oxygen consumption, we conducted a MitoStress test in A549 cells using the Seahorse system. Under stress conditions caused by higher cell density—where cells compete for limited resources such as oxygen and nutrients—the metabolic capacity of LC-EV-treated cells declined. Specifically, as cell density increased, Maximal Respiration significantly decreased in these cells, whereas cells treated with control vesicles maintained stable Maximal Respiration Fig. 4A. This suggests that LC-EVs impair the cells' ability to adapt to resource-limited conditions. Accordingly, ATP production (Fig. 4B) and basal respiration (Fig. 4C) were also reduced at higher cell densities in the presence of LC-EVs, further supporting the idea that these vesicles negatively impact mitochondrial adaptability in response to metabolic stress.

Expanding the number of nodes in the analysis of protein interactions has allowed us to identify additional processes, including apoptotic cell death triggered by DNA damage (Figure S7).

To understand the role of LC-EVs in affecting A549 lung cells, we analyzed the RUNX2 mRNA levels in the cargo of LC-EVs and controls. Thus, we observed increased RUNX2 gene expression in cargo LC EVs compared to controls (Fig. 5A). Therefore, to investigate the role of RUNX2 in modulating the increase in SESN2 and the activation of the p53/p21 pathway, we overexpressed RUNX2 in A549 cells. Forced expression of RUNX2 induced the activation of the p53/p21 pathway and resulted in increased SESN2 levels (Fig. 5B), as observed in lung cells treated with LC-EVs. In addition, to determine whether the effect observed after LC-EV treatment was mediated by RUNX2, we exposed RUNX2-silenced A549 cells to LC-EVs and subsequently analyzed the cellular markers as well metabolic stress response. As shown in the Fig. 5C, RUNX2 downregulation leads to a reduction in stress protein levels even in the presence of LC-EVs, when compared with scramble/control cells. With respect to metabolic stress, RUNX2 downregulation was able to restore EV-treated cells to scramble/control levels (Fig. 5D). Thus, we explored the role of molecules involved in the post-transcriptional regulation of RUNX2. Through an in silico analysis using TargetScan, we identified a putative hsa-miR-204-5p binding site in the 3′-UTR of RUNX2 mRNA. In particular, the binding site is in an evolutionarily conserved region (Fig. 5E). Thus, by analyzing the hsa-miR-204-5p content in EVs cargo, we observed reduced expression of hsa-miR-204-5p in LC-EVs compared to controls (Fig. 5F). Therefore, to test the role of hsa-miR-204-5p in modulating RUNX2 expression, hsa-miR-204-5p (mimic hsa-miR-204-5p) was delivered into A545 cells via Lipofectamine-mediated transfection at concentrations ranging from 5 to 25 nM. As shown in Fig. 5G, RUNX2 protein levels decreased in response to increasing concentrations of mimic hsa-miR-204-5p. Similarly, higher concentrations of mimic hsa-miR-204-5p (ranging from 5 to 15 nM) also reduced the protein levels of SESN2, P53, and p21. (Fig. 5H). However, in the presence of high levels of miR-204, the addition of LC-EVs induced only a slight increase in RUNX2, while the other stress markers remained unchanged (Fig. 5I).

In IMR90 fibroblast lung cells, exposure to LC-EVs (Fig. 6A), impaired cellular homeostasis, as evidenced by a pronounced increase in RUNX2 expression (Fig. 6B) together with the upregulation of stress-associated markers (Fig. 6C). At the metabolic level, LC-EVs detrimentally affected mitochondrial function by reducing maximal respiratory capacity (Fig. 6D), while ATP levels and basal respiration remained unchanged, suggesting a selective impairment of bioenergetic flexibility without compromising basal energy supply.

We also observed a significant increase in RUNX2 expression in both endothelial cells (EC) and aortic smooth muscle cells (ASMC). As shown in Fig. 7A, EVs from both control and long COVID patients were able to enter endothelial cells, and after 18 h, they were fully localized within the cytoplasm.

Both gene expression (Fig. 7B) and RUNX2 protein levels (Fig. 7C) were higher in cells treated with LC-EVs compared to controls. Additionally, SESN1 levels were similar, while p53, p21, and SESN2 levels were higher in endothelial cells exposed to LC-EVs compared to controls. Similarly, EVs from both control and long COVID subjects efficiently entered ASMCs, with complete internalization observed within 18 h (Fig. 7D). In ASMC we also observed an upregulation of the RUNX2 gene (Fig. 7E) and related protein levels (Fig. 7F) following treatment with LC-EVs. However, we observed higher levels of SESN1 in ASMC treated with LC-EVs compared to controls, while SESN2 levels were similar in both groups (Fig. 7F). As observed in lung cells and endothelial cells, we observed higher levels of p53 and p21 in cells treated with LC-EVscompared to controls (Fig. 7F).

We also investigated the impact of LC-EV on a human MSCs line. Our data demonstrated that similarly to the above-investigated cells, EVs from both control and long COVID sources entered the mesenchymal stem cells, with complete internalization observed within 18 h (Fig. 8A).

In hMSCs we also observed increased gene expression levels of RUNX2 following treatment with long COVID-derived EVs compared to controls (Fig. 8B). Additionally, we observed an increase in SOX9 gene expression and a decrease in PPARG expression (Fig. 8B). Additionally, we observed higher levels of RUNX2, SESN1, p53 and p21 proteins in MSCs treated with LC-derived EVs (Fig. 8C).