Subject recruitment and characteristics
This study included 25 patients with AR and 25 healthy subjects matched for age, gender and BMI, recruited between February and June 2019 at the Allergology Unit, Fondazione IRCCS Ca’ Granda -Ospedale Maggiore Policlinico, Milan, Italy. These subjects underwent a standard battery of skin prick test and/or eosinophil counts, in order to confirm or rebut AR diagnosis. Exclusion criteria included diabetes, hypertension, autoimmune diseases, cancer or other major chronic health condition, pregnancy, history of illicit drug use. After signing a detailed informed consent form, all participants were asked to donate blood and nasal swab samples. In addition, each participant gave his/her written informed consent to participate to the study and filled in a standardized questionnaire about demographics and lifestyle information (e.g. smoking habits, alcohol consumption, and diet). Clinical history was also collected for each of the AR patient.
Sample collection and processing
Nasal swabs were collected from each of the participant through nasal swab following WHO guidelines (https://goo.gl/pMzSrT) and stored at -80°. DNA was extracted using QIAamp® UCP Pathogen Mini (QIAGen, Hilden, Germany) following the manufacturers guidelines. The extracted DNA was stored at -20°C and later shipped to the sequencing service facility Personal Genomics Srl (Verona, Italy) to perform qualitative and quantitative checks, PCR amplification and second-generation sequencing analysis. Bacterial communities were investigated through amplicon sequencing analysis of the 16S rRNA gene hypervariable regions V3-V4, amplified with the primer pair Pro341F (5′-CCTACGGGNBGCASCAG-3′) and Pro805R (5′-GACTACNVGGGTATCTAATCC-3′), and then sequenced through the Illumina MiSeq platform using a paired-end library of 300bp insert size. Blood samples were collected in ethylenediamine tetra-acetic acid (EDTA) tubes and processed within 2-hours from the phlebotomy. The Minimal information for studies of extracellular vesicles (MISEV) 2018 guidelines were followed performing EV purification, isolation and characterization  and summarized in supplemental Table S1. Briefly, EDTA‐treated blood was centrifuged at 1200 × g for 15 min at room temperature to obtain platelet‐free blood plasma and then further centrifuged following a three‐step centrifugation protocol (1000, 2000, and 3000 × g for 15 min at 4 °C), and finally ultracentrifuged to obtain an EV‐rich pellet (110,000 × g for 75 min at 4 °C) as previously described .
Upstream analyses and operational taxonomic units (OTUs) clustering
Raw reads quality and statistics were checked using FastQC v0.11.2 and then trimmed at the 3’ end using Trimmomatic v0.32 to improve the following read joining. Fastq-join.py tool were applied to join the raw reads and quality filtered using a minimum base quality 20 (Phred-scale) over 5 bases sliding window and then analysed using the default settings for QIIME 1.9.1. After chimeric reads removal, the resulting reads were clustered using 97% of similarity applying the open-reference OTUs pipeline using USEARCH61. Taxonomic assignment was carried out with the RDP classifier  through the comparison of representative reads against the Greengenes v13.8 database using standard options. PyNast method and default settings suggested in QIIME pipeline were applied to align sequences [43, 44]. The resulting OTU table was successively filtered, removing singleton and low abundance OTUs to performer downstream and statistical analysis.
Downstream analyses were carried out using QIIME v1.9.1 parsing the above-described OTUs table. Taxonomic values within each sample and group was assigned to each OTU from the phylum to the genus level. OTUs, which fails genus attribution, were tagged as “Unassigned” followed the specific family label. To assess significant differences between the OTU abundance in the AR and HS groups a nonparametric t-test (999 Monte Carlo permutations), was applied. Before diversity analysis, all samples were rarefied to 10,000 sequences with a seed of 10, in order to avoid the influence of different sequencing depths. Alpha-diversity richness, evenness and genetic distance were calculated using observed OTUs, Shannon and PD_whole_tree indices. In addition, a nonparametric two sample t-test was applied to assess differences between groups of samples, with Monte Carlo permutations (999). Beta-diversity was examined applying the Weighted_Normalized UniFrac distance measure. To compare the tightness of clustering distance within all samples in a group comparing the state of disease, a two-sample t-test was performed (999 MonteCarlo permutations). Furthermore, to visualize and interpret the result of the applied distance measure, Principal Coordinate Analysis (PCoA) were performed on the produced distance matrix and plotted using Emperor and the adonis function in the R Vegan package was used to test significance in dissimilarity matrices between AR and HS groups.
Extracellular vesicles analysis: distribution and immunophenotyping
The EV enriched pellet obtained after the ultracentrifugation step was resuspended in 500-μL triple‐membrane filtered phosphate‐buffered saline PBS to perform flow-cytometry and nanoparticle tracking analyses (NTA).
NTA analysis was carried out with the Malvern NanoSight NS300 system (Malvern Panalytical Ltd., Malvern, UK), used to visualize the EVs by laser light scattering. For each sample five 30-sec records were registered. NTA output was then analysed with integrated NTA software (Malvern Panalytical Ltd.), providing high-resolution particle size distribution profiles and EV concentration measurements. NTA EV data were expressed as 106 for 1 ml of plasma.
To determine EV cellular origins, immunophenotyping was achieved with the MACSQuant Analyser flow cytometer (Miltenyi Biotec, Bergish Gladbach, Germany) following the customer-provided protocol (https://bit.ly/2z2s69i). The Fluoresbrite Carboxylate Size Range Kit I (0.2, 0.5, 0.75, and 1 µm) was used to set the calibration gate on the MACSQuant Analyser system. To evaluate integrity and to highlight the hEV subset, 60-µL sample aliquots were stained with 0.02 µM 5(6)-carboxyfluorescein diacetate N-succinimidyl ester (CFSE) at 37 °C for 20 min in the dark. CFSE is a vital dye non-fluorescent molecule, able to enter into MV, where intracellular esterase enzymes remove the acetate group and convert the molecule into the fluorescent ester form. To characterize and count hEVs, the following panel of antibodies was used: AbCD177 (neutrophils), AbCD14 (monocytes), AbCD61 (platelets), EpCAM (epithelium), AbCD62E (activated epithelium), AbCD203C (mast cells), AbCD294 (eosinophils). In addition, two different 60-µL sample aliquots for each sample were also incubated with specific antibodies to recognize the bEVs: LPS and a primary unconjugated Lipoteichoic acid combined with a secondary IgG-antimouse to distinguish EVs derived from gram-negative or gram-positive bacteria from the whole subset. Before use, each antibody was centrifuged at 17,000×g for 30 min at 4°C to eliminate aggregates. A stained PBS blank sample was used to detect auto-fluorescence of the antibody. Quantitative multiparameter analysis of flow cytometry data (expressed as 103 for 1ml of plasma) was run using FlowJo software (Tree Star, Inc., Ashland, OR, USA).
Descriptive statistics of the study participant characteristics were performed on all variables. Categorical data are presented as frequencies and percentages. Continuous variables are expressed as the mean ± standard deviation (SD). Multivariable negative binomial regression models were applied to evaluate marginal means of EVs and cell origins in AR and HS participants. Each model was adjusted for age, gender and smoking habits. Statistical analyses were carried out with SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA).