Extracellular vesicles in systemic juvenile idiopathic arthritis

Abstract

Systemic juvenile idiopathic arthritis is a chronic pediatric inflammatory disease of unknown etiology, characterized by fever, rash, hepatosplenomegaly, serositis, and arthritis. We hypothesized that intercellular communication, mediated by extracellular vesicles, contributes to systemic juvenile idiopathic arthritis pathogenesis and that the number and cellular sources of extracellular vesicles would differ between inactive and active states of systemic juvenile idiopathic arthritis and healthy controls. We evaluated plasma from healthy pediatric controls and patients with systemic juvenile idiopathic arthritis with active systemic flare or inactive disease. We isolated extracellular vesicles by size exclusion chromatography and determined total extracellular vesicle abundance and size distribution using microfluidic resistive pulse sensing. Cell-specific extracellular vesicle subpopulations were measured by nanoscale flow cytometry. Isolated extracellular vesicles were validated using a variety of ways, including nanotracking and cryo-electron microscopy. Extracellular vesicle protein content was analyzed in pooled samples using mass spectrometry. Total extracellular vesicle concentration did not significantly differ between controls and patients with systemic juvenile idiopathic arthritis. Extracellular vesicles with diameters <200 nm were the most abundant, including the majority of cell-specific extracellular vesicle subpopulations. Patients with systemic juvenile idiopathic arthritis had significantly higher levels of extracellular vesicles from activated platelets, intermediate monocytes, and chronically activated endothelial cells, with the latter significantly more elevated in active systemic juvenile idiopathic arthritis relative to inactive disease and controls. Protein analysis of isolated extracellular vesicles from active patients showed a proinflammatory profile, uniquely expressing heat shock protein 47, a stress-inducible protein. Our findings indicate that multiple cell types contribute to altered extracellular vesicle profiles in systemic juvenile idiopathic arthritis. The extracellular vesicle differences between systemic juvenile idiopathic arthritis disease states and healthy controls implicate extracellular vesicle–mediated cellular crosstalk as a potential driver of systemic juvenile idiopathic arthritis disease activity.

Increased levels of extracellular vesicles from multiple cell types during active systemic juvenile idiopathic arthritis suggest that extracellular vesicles may be a potential disease driver.

1. Introduction

Systemic juvenile idiopathic arthritis (sJIA) is a severe form of chronic pediatric arthritis distinguished from other juvenile idiopathic arthritis subtypes by significant systemic inflammation, manifesting as daily fevers, lymphadenopathy, hepatosplenomegaly, serositis (e.g. pericarditis, pleuritis), and rash. The presence of these symptoms characterizes active disease, and patients fluctuate between episodes of active disease (“flares”) and periods of inactive disease (quiescence).¹ Adult-onset Still disease is the adult equivalent of sJIA.² Much is still unknown about sJIA immunopathogenesis, but existing data suggest that innate immune responses drive disease onset and flares, whereas the adaptive immune response likely contributes to the persistent arthritis that affects ∼30% of patients.¹ During sJIA flares, levels of circulating platelets, monocytes, and neutrophils are elevated. Proinflammatory cytokines, notably IL-1, IL-6, and IL-18, are thought to be key to perpetuating the inflammatory cascade in sJIA.¹ Evidence for endothelial activation in sJIA includes the presence of elevated D-dimer levels in the majority of patients with active disease.^3,4 In addition, biopsy specimens of skin in active sJIA show elevated expression of endothelial adhesion receptors, likely induced by calcium-binding S100A8/S100A9 complexes (calprotectin), which are expressed by infiltrating monocytes, neutrophils, and activated keratinocytes.^5,6 These adhesion receptors enhance leukocyte extravasation. However, overall information on processes driving endothelial cell activation is limited.

Extracellular vesicles (EVs) are submicron-sized, phospholipid membrane-enclosed circulating particles produced by every cell type. EVs range in size from 50 to 1,000 nm in diameter and include exosomes, activation- or apoptosis-induced microvesicles/microparticles (MVs), and apoptotic bodies.⁷ Once dismissed as cellular debris, EVs are now known to contain biologically active material and act as vectors for intercellular communication. EV surface markers reflect their cellular origin and influence their targeting. EV cargo, which includes microRNA (miRNA), cytokines, and other mediators, also is related to cellular origin, potentially providing a retrospective snapshot of cell state at the time of EV release. EVs are highly abundant in biofluids and, as opposed to single-molecule circulating biomarkers, may carry molecular signatures associated with specific phenotypes.⁸

Due to the small size of EVs, phenotyping methods have remained limited in sensitivity and scale, and despite significant progress in the past few years, EV methodology remains a work in progress. Furthermore, until recently, there have been relatively few studies of EVs in human biofluids, because EV isolation methods have traditionally required large sample volumes. Flow cytometry has been increasingly used for EV analysis, with varied degrees of success. Recent advances now allow EV characterization using small amounts of blood,⁹ making it more feasible to investigate patient samples, including those from children.

EVs are increasingly recognized as critical mediators of systemic inflammation. Higher levels of EVs from particular cell types are observed in various autoimmune and inflammatory conditions and correlate with disease activity.^10–12 In sJIA, differences in the abundance and cargo content of EVs between quiescent and active disease states could provide insight into sJIA pathogenesis and serve as biomarkers of disease activity and predictors of treatment response. Using a methodology that enables some of the most unbiased and mechanically gentle approaches to study EVs, we characterized EVs during different clinical states of sJIA.

2. Patients and methods

2.1. Patient population and controls

The study was conducted in accordance with the Stanford Institutional Review Board (protocols #13932 and 41402). Patients were recruited from the Pediatric Rheumatology Clinic of Stanford University and the Pediatric Rheumatology Clinic of the University of California, San Francisco. Patients were enrolled for blood donation and collection of associated clinical information at routine visits when a blood draw was required for clinical care. Blood from patients in the Stanford Pediatric Endocrinology Clinic, who presented for evaluation of either short stature or precocious puberty, served as our immunologically healthy control population.

Enrolled subjects contributed to 1 of the following 3 groups based on their clinical status: patients with sJIA with active disease (these included treatment-naive and treatment-refractory subjects, as defined in recent clinical trials in patients with sJIA¹³), patients with sJIA with inactive disease (patients with sJIA responsive to therapy based on recent assessments indicating clinically inactive disease¹⁴), and age-matched healthy control subjects. Key inclusion criteria were (i) patients meeting the International League of Associations for Rheumatology (ILAR) classification criteria for sJIA¹⁵ or meeting modified ILAR criteria with quotidian fever lasting <2 wk¹⁵ and (ii) age >6 mo and <19 yr. Key exclusion criteria were (i) enrollment in blinded clinical trials, (ii) pregnancy, and (iii) nonambulation.

Definition of disease state: As previously described,¹⁶ comprehensive clinical information was collected at each sJIA patient visit, including history, physical exam (including presence of fever, rash, and joint count), and clinical laboratory values (erythrocyte sedimentation rate [ESR], D-dimers, C-reactive protein [CRP], ferritin, complete blood count [CBC], and platelet count), as well as medications at the time of the visit. Based on these parameters, each sJIA sample was classified as active disease (flare, F) or inactive disease (quiescence with or without medication, Q). Demographic and clinical information for all patients and controls included in the study are presented in Supplementary Tables 1 and 2, respectively. Due to limitation in plasma volume, different samples were used for different experiments.

2.2. Sample collection and generation of platelet-poor plasma

Plasma was obtained from venous blood collected in heparinized Vacutainer (BD) tubes. The blood was kept at room temperature (RT) and, within a maximum of 2 h postdraw, was spun at 25 °C at 550 × g for 5 min to separate plasma from cells. The collected plasma was then spun at 4 °C at 1,000 × g for 5 min, transferred to a new tube, and subsequently spun at 1,000 × g for 15 min to remove platelets. Platelet-poor plasma samples were stored at −80 °C until analysis; hereafter “plasma” refers to platelet-poor plasma. Prior to analysis, thawed plasma was spun at 2,500 × g for 15 min to remove any previously remaining platelets, and supernatant was transferred to DNA LoBind Microcentrifuge Tubes (Eppendorf) to optimize plasma EV yield, except as otherwise noted.

2.3. EV enrichment/isolation from plasma by size exclusion chromatography

Total plasma EVs were isolated using an input of 100 µL plasma and a qEV size exclusion chromatography (SEC) column (Izon, product code SP2) for each sample per the manufacturer's protocol. Briefly, 250-µL fractions were eluted using 0.1-µm filter-sterilized phosphate-buffered saline (PBS) (Hyclone). The first 1 mL of eluate (i.e. the void volume) was discarded. Fractions 6 to 10, known to contain the highest yield of EVs, were collected together into a DNA LoBind Microcentrifuge Tube (Eppendorf) and stored at −80 °C until analysis.

2.4. EV size and concentration measurement using microfluidic resistive pulse sensing

Particle diameters and concentrations of SEC-generated EV isolates were determined using a microfluidic resistive pulse sensing (MRPS) particle analyzer, the nCS1 (Spectradyne), per the manufacturer's protocol and as previously described.⁹ A TS-400 and TS-2000 cartridge (Spectradyne) enabled detection and measurement of EVs ranging from 75 to 2,000 nm. Then, 5 µL of EV isolate was loaded into each single-use cartridge for analysis. Results were consistently within 1% to 2% of duplicate measurements (performed at least twice for each sample) and within 4% to 6% of replicate measurements (performed twice for several control samples).

2.5. EV analysis by flow cytometry with wash step

Initial EV analysis was performed by staining plasma EVs as described in Inglis et al.^17,18 with modifications. Antibodies for these experiments were pretitrated and combined into a cocktail for staining (Supplementary Table 3) onto a 0.2-µm centrifugal filter column (Nanosep; Pall Corporation); cocktails were then filtered by centrifugation at approximately 750 × g for 2 min at RT prior to use. Bodipy FL Maleimide (Invitrogen, ThermoFisher), with similar spectra as FITC, was also used to label EVs;¹⁹ the dye was diluted 1:400 in PBS (Gibco) and then filtered by centrifugation at 14,000 × g for 2 min at RT prior to use. Plasma samples were rapidly thawed at 37 °C, mixed by inversion, and centrifuged at 1,000 × g for 2 min, and 100 µL of plasma was added to a well of a 96-well round-bottom plate. Diluted Bodipy dye (6.5 µL), followed by the filtered antibody cocktails, was added to the plasma and mixed by pipetting. The plate was then incubated at 4 °C for 30 min. Samples from each well were subsequently transferred to a clean 0.2-µm centrifugal filter column. Each well was washed with 200 µL PBS, and this volume was added to the same filter column with the corresponding stained plasma. Samples were then centrifuged at 850 × g for 3 min at RT; if residual sample remained on top of the filters, an additional spin of 1 to 2 min was performed. The dry top of the filter was resuspended with 300 µL fresh PBS by pipetting gently up and down 8 times. The resuspended material was transferred to a 5-mL polypropylene tube for flow cytometry analysis, just prior to which TruCount BD beads were added to the samples to standardize test volumes for EV counting per stained plasma aliquot.

Samples were analyzed on a BD Influx Flow Cytometer & Cell Sorter (Stanford Shared FACS Facility, SSFF). A 0.1-nm filter for the sheath fluid was added to minimize background noise. Light side-scatter (SSC) relative size range of detection was standardized in each experiment by using FITC-conjugated beads of known sizes from 100 to 500 nm (Apogee Flow Systems). PBS (0.1 nm filtered) was used to set background threshold. Compensation beads were used to set voltage for each fluorochrome (Molecules of Equivalent Soluble Fluorochrome) per the International Society for Extracellular Vesicles guidelines.²⁰ Samples were run for 1 min; then, 20 μL 10% Igepal (Sigma) was added to the sample and mixed by gentle pipetting, and the tube was run again for 1 min. FlowJo software (Becton, Dickinson & Company) was used to analyze the samples. Compensation was performed using single-stained beads (BD Compbeads; BD Biosciences). Bodipy-positive vesicles were selected, based on a Bodipy-negative sample. The number of events positive before and after detergent lysis was determined using the TruCount beads. The final number of EVs was determined as the difference between the positive events before and after detergent lysis.

2.6. EV analysis by high-resolution flow cytometry

For scientific rigor, plasma EVs were also analyzed using fluorescent antibody–labeled samples prepared without a wash step as described above, to ensure that particles smaller than 200 nm were included in our analyses.

Prior to sample staining, individual antibodies were filtered to remove aggregates using 0.2-µm centrifugal filter columns and spun at 800 × g for 2 min at RT. Then, 0.1-µm filtered 1× PBS (Hyclone) was used to dilute antibodies to optimal concentration (Supplementary Table 3) to a final volume of 1 µL per antibody per 15 µL sample stained. Staining was performed either with a single antibody or with a combination of a maximum of 2 antibodies in order to minimize fluorescence spectral overlap. Single or dual staining was performed with 15 µL plasma in a DNA LoBind Microcentrifuge Tube (Eppendorf), or a well in a 96-well round-bottom plate, and incubated at 4 °C overnight (minimum 8 h), protected from light. Unstained plasma (15 µL) was also included for each stained sample (incubated overnight with 2 µL PBS) to serve as one of the negative controls. As above, all mixing steps related to high-resolution flow cytometry (HRFC) were performed by gentle pipetting rather than vortexing, which is known to increase the likelihood of protein aggregation that could lead to false-positive event detection.¹⁷ Just prior to flow cytometry, 1 µL of each sample was diluted with 200 µL freshly prepared dilution buffer containing 2 mM EDTA in 1× PBS and 200-nm FITC-conjugated polystyrene beads (Fluoresbrite YG Microspheres, 200 nm; Polysciences) at a concentration of 500 beads/µL, included in the dilution buffer for every sample to ensure equal volumes were measured.

Analysis was performed using a BD FACSAria II (SSFF). Fluidics and instrument settings were established to minimize background noise while maximizing detection of submicron-sized EVs: using a 70-µm nozzle with PBS sheath fluid at a pressure of 70 psi and drops generated at a frequency of 87.0 mHz, threshold was changed from the default forward scatter (FSC) setting of 5,000 to the SSC set at 200 for particle detection. Flow rate was set at 1.0. A 50-mW 488-nm blue laser was used with an SSC photomultiplier (PMT) and 488/10 BP filter, as well as a 505 splitter with a 525/50 BP filter for the detection of FITC. SSC and FITC PMT voltages were adjusted to enable visualization of 7 distinct populations of Megamix-Plus SSC and Megamix-Plus FSC beads (BioCytex), a commercially available set of FITC-conjugated polystyrene beads ranging from 100 to 900 nm in diameter; background noise was kept separate and below a rate of 1,000 events/s (Fig. 1A). The whole volume of 200 µL of immunolabeled, diluted plasma samples was run at these settings, while ensuring that threshold rate remained below 40,000 events/s with an electronic abort rate below 10%. A consistent volume of 10 µL of each sample was analyzed by designating a gate around the 200-nm FITC-labeled beads and setting a stop time of 5,000 events (Fig. 1B). Immediately after each sample was run, 1.6 µL Triton X-100 (10% aqueous solution; Sigma-Aldrich) was added to the remaining sample (∼160 µL) for a final detergent concentration of 0.1%, gently mixed by pipetting, and run again using the same settings (Fig. 1C).

Fig. 1.

HRFC-based visualization of EVs and distinction between fluorescently labeled EVs and nonvesicle signals by HRFC using detergent lysis. (A) EV gating strategy: SSC and FITC PMT voltages adjusted to enable visualization of 7 distinct sets of FITC-conjugated Megamix polystyrene beads ranging from 100 to 900 nm in diameter. These settings were used to visualize EVs from plasma in the same size range. (B) Representative dot plot of plasma stained with a FITC-conjugated anti-CD9 antibody. Double (white) boxes represent gates to measure the fluorescent-positive EV population. Smaller (yellow) box indicates the gate set around the 200-nm FITC-conjugated beads included in the dilution buffer for every sample to ensure equal volumes were measured. (C) Left: FITC-positive signal from unstained plasma (showing the 200-nm beads used to ensure that equal volumes were measured). Middle: FITC-positive events from plasma stained with FITC-conjugated anti-CD41 antibody. Right: remaining FITC-positive events following detergent lysis of the same sample, representing nonvesicle signal. The amount of fluorescent-positive EVs can be determined by subtracting remaining fluorescent-positive events from the fluorescent-positive events measured before detergent lysis.

Surface marker–positive populations were determined using uniform gates set around fluorescent events detected in stained samples, with gate positioning optimized by ensuring that no positive events were included in the corresponding unstained sample upon applying the same gate (Fig. 1C). EV counts were measured using FlowJo analysis software. Final EV concentration was determined by subtracting the number of positive events in the detergent-lysed sample from the positive events in the prelysed sample and then multiplying by 20 to extrapolate EVs/µL, accounting for the 200-fold dilution and the 10-µL diluted sample processed per run. In general, the majority of positive events were no longer detectable following detergent lysis, indicating that prelysis signals represented membrane-bound vesicles; as expected, the population of 200-nm polystyrene beads remained intact following addition of detergent (Fig. 1C). Detergent lysis–based method was favored over isotype antibodies as negative controls due to its ability to more appropriately mimic background fluorescence in the setting of EV range detection.¹⁷

Except for Fig. 3, all other flow cytometry experiments were performed using this methodology without additional filtration of the sample.

Fig. 3.

Patients with sJIA at flare show higher concentration of EVs/µL shed by various cell types in plasma. EV cellular origin is indicated by the presence of associated surface markers detected by flow cytometry as follows: (A) platelets (CD41⁺), (B) endothelial cells (CD146⁺), and (C) PECAM-1⁺ (CD31⁺) EVs represent those shed by endothelial cells undergoing apoptosis, a consequence of chronic activation, and (D) erythrocytes (CD235⁺). F indicates samples from patients with sJIA with disease flare (n = 12), Q from patients with sJIA with quiescent disease (n = 7, except for (B), n = 6), and healthy pediatric controls, n = 11. Plasma samples were analyzed by flow cytometry as described in “EV analysis by flow cytometry with wash step.” Boxes extend from the minimal to maximum values, with line at median. All groups were compared using Kruskal–Wallis test and Dunn's multiple comparisons test; *P < 0.05, **P ≤ 0.01. Lines between the sJIA Q and sJIA F groups represent the same patient at different stages of the disease. Summary demographic and clinical information for subjects are in Supplementary Tables 1 and 2.

2.7. Validation of platelet EVs

We isolated and validated platelet EVs from plasma as described.⁹ Sorted Megamix polystyrene bead controls and sorted cell- and size-specific platelet EV populations were evaluated by a wide variety of approaches, including nanoparticle tracking analysis (NTA; NanoSight Ltd, using the manufacturer's instructions) and cryo-electron microscopy (Cryo-EM) using copper Quantifoil holey carbon support grids (Ted Pella 658-300-CU) vitrified on liquid ethane using a Mark IV Vitrobot (Thermo Fisher). The conditions utilized for the cryopreservation were 100% humidity, blot force of 1, and blotting time of 3 s. Low-dose conditions were used to acquire images on a FEI Krios-Titan operated at 300 kV using a Falcon II direct electron detector. Negative controls included grids with no EVs. For platelet activation electron microscopy studies, we isolated platelets from plasma and activated them by putting the tube on ice for 20 min. Platelets and nanoscale flow cytometry isolated platelet EVs were fixed in 2.5% glutaraldehyde, rinsed with water, stained with uranyl acetate, and imaged by transmission electron microscopy (TEM). Whole-platelet TEM imaging employed mitochondria cristae morphology as an internal control for platelet integrity. Sorted EV TEM and Cryo-EM required identification of plasma membrane–bound vesicles.

2.8. Measurement of plasma-based and EV-associated cytokines by Luminex

Cytokines were detected from plasma samples and corresponding SEC-generated EV isolates using a Luminex-based platform measuring a total of 76 human cytokines. Prior to Luminex analysis, plasma and EV isolates were treated with 0.1% Triton X-100 to facilitate release of membrane-encapsulated cytokines. Then, 100 µL plasma and 200 µL EV isolate were analyzed for each sample/subject. Kits were purchased from EMD Millipore Corporation and used according to the manufacturer's recommendations with modifications as follows: the assay included 3 panels; panel 1 was Milliplex HCYTMAG60PMX41BK with IL-18 and IL-22 added to generate a 43-plex panel. Panel 2 was Milliplex HCP2MAG62KPX23BK with MIG/CXCL9 added to generate a 24-plex panel. Panel 3 included the Milliplex HSP1MAG-63 K with resistin, leptin, and hepatocyte growth factor added to generate a 9-plex panel. Before mixing plasma with antibody-inked magnetic beads, plasma samples were diluted 2-fold for panels 1 and 2 and 10-fold for panel 3. As SEC-generated EV isolates were already more dilute at baseline compared to plasma samples, no further dilution for EV isolates was performed before Luminex analysis.

The setup of the assay was as recommended by the manufacturer. Briefly, samples were mixed with antibody-linked magnetic beads on a 96-well plate and incubated overnight at 4 °C with shaking. Cold and RT incubation steps were performed on an orbital shaker at 500 to 600 rpm. Plates were washed twice with Wash Buffer in a Biotek ELx405 washer. Following 1-h incubation at RT with biotinylated detection antibody, streptavidin-PE was added for 30 min with shaking. Plates were washed as above and PBS added to wells for reading in the Luminex FlexMap3D Instrument with a lower bound of 50 beads per sample per cytokine. Each sample was measured in duplicate. Custom Assay Chex control beads were purchased from Radix Biosolutions and added to all wells. For each analyte, average sample medium fluorescence intensity (MFI) was divided by the MFI corresponding to the appropriate background control (Stanford Human Immune Monitoring Center–based control for plasma and PBS containing 0.1% Triton X-100 for EV isolates) to determine the relative MFI.

2.9. Mass spectrometry

SEC-isolated EVs in PBS from each group (flare, quiescence, and control) were pooled and concentrated using 3-kDa filters (Millipore). Protein concentration was determined using the BCA assay (Biorad). Approximately 1 µg of each EV pool was run on 4% to 15% Mini Protean TGX gradient gel and silver stained (Thermo Pierce Silver Staining kit) for evaluation prior to mass spectrometry.

For mass spectrometry, each pooled sample was divided into 5 replicate aliquots each, and all 15 samples were subsequently processed in parallel. From the EV-containing samples, approximately 5 µg protein was combined 1 to 1 with 2× lysis buffer (10% sodium dodecyl sulfate, 100 mM triethylammonium bicarbonate, pH 7.5), and proteins were reduced with 5 mM dithiothreitol, alkylated with 15 mM iodoacetamide, and digested with trypsin using the S-Trap system according to the manufacturer’s instructions (PROTIFI, C02-micro-40, NC1588050). Resulting peptides were dried down, resuspended, and acidified in 1% trifluoroacetic acid (TFA). EV peptides were desalted using in-house constructed (SDB-RPS) SPE Stagetips.²¹ Briefly, Stagetips were activated with methanol, conditioned with 80% acetonitrile (ACN)/0.1% TFA, and equilibrated with 0.2% TFA. Peptides were then added, washed with 99% isopropanol/0.1% TFA, and washed twice with 0.2% TFA, once with 0.1% formic acid and eluted with 60% ACN/0.5% ammonium hydroxide. The eluted peptides were flash frozen and dried down.

Approximately 1 µg desalted peptides was analyzed on a Fusion Lumos mass spectrometer (Thermo Fisher Scientific) equipped with a Thermo EASY-nLC 1200 LC system (Thermo Fisher Scientific). Peptides were separated by capillary reverse-phase chromatography on a 25-cm column (75-μm inner diameter, packed with 1.6 μm C18 resin, AUR2-25075C18A; Ionopticks). Peptides were introduced into the Fusion Lumos mass spectrometer using a gradient with 3% to 27% buffer B (0.1% (v/v) formic acid in acetonitrile) for 52.5 min followed by 27% to 40% buffer B for 14.5 min at a flow rate of 300 nL/min. Data were acquired in a top-speed data-dependent mode with a duty cycle time of 1 s. Full MS1 scans were acquired with an m/z scan range of 375 to 1,500 m/z in the orbitrap mass analyzer with a resolution of 120,000 and 1e6 automatic gain control (AGC) target with the maximum injection time set to Auto. Selected precursor ions from MS1 were subjected to fragmentation using higher-energy collisional dissociation (HCD) with a quadrupole isolation window of 0.7 m/z and normalized collision energy of 31%. The resulting MS2 scan of the HCD fragments was sent to the ion trap mass analyzer set to Turbo scan speed and 1e4 AGC target with the maximum injection time set to Auto. Target precursors were then dynamically excluded for 60 s.

Resulting mass spectrometry raw data files were searched against a human Uniprot database (downloaded on 14 July 2021) using a reverse-decoy method and quantified using MaxQuant software (version 2.0.3.0) with default settings along with the maxLFQ quantification algorithm activated. Briefly, a 1% false discovery rate was set at the peptide spectrum and protein matching with a minimum peptide length of 7, 2 missed cleavages were allowed, and enzyme was set to trypsin. Oxidation of methionine and protein N-terminal acetylation were set as variable modifications along with carbamidomethylation of cysteines set as a fixed modification.

2.10. Statistical analysis

GraphPad Prism (version 9.3.1) was used for all statistical tests of the EV size range and flow cytometry data. Following tests for normal distribution, all 3 groups (sJIA F, sJIA Q, and control) were analyzed using Kruskal–Wallis, with posttesting using Dunn's multiple comparisons test. For EV level comparisons, P values were based on 2-tailed calculations, and P < 0.05 was the cutoff for significance. Correlations were analyzed using Spearman's correlation.

Heatmaps were generated using the open-access software available through https://software.broadinstitute.org/morpheus, which also enabled hierarchical clustering analysis using the Pearson correlation metric with an average linkage method following log₂ transformation of EV concentration measurements.

For the mass spectrometry analysis, the resulting mass spectrometry protein groups data table from MaxQuant was uploaded and analyzed using Perseus software (version 1.6.6.0). Proteins identified by reverse sequences, only identified by site and potential contaminants, were filtered out. Protein group abundance results were log₂ transformed and filtered to select for proteins quantified from at least 3 of 5 replicates in at least 1 condition and for at least 1 unique peptide per identified protein. Proteins with missing values were imputed with a width of 0.3 and downshift of 1.8. A 2-sample t-test was then performed with an s0 set to 0.1 between groups. The resulting t-test data were then uploaded to GraphPad Prism (version 9.4.1) to generate volcano plots. Gene ontologies of identified proteins were analyzed using the FunRich software (http://www.funrich.org/), version 3.1.4, and with the use of QIAGEN IPA (https://digitalinsights.qiagen.com/IPA) (version 81348237). The networks and functional analyses were generated through the use of QIAGEN IPA (https://digitalinsights.qiagen.com/IPA).²²

3. Results

3.1. Total plasma EV size distribution and concentration

To capture the full range of EVs in sJIA, we used SEC to isolate EVs and MRPS to measure particle diameter and concentration. Figure 2A shows a representative graph of SEC-isolated EVs measured by MRPS using 2 different cartridges to capture a broad size range. We observed that EV concentration decreased with increasing diameter. The majority of EVs were <200 nm in diameter, and no significant difference in plasma EV concentration was seen between controls and patients with sJIA (Fig. 2B). Furthermore, there was no significant difference in plasma EV concentrations between sJIA F and Q samples (Fig. 2A and B).

Fig. 2.

Size distribution and concentration of total plasma EVs are similar in patients with sJIA and healthy controls. EVs were isolated from plasma by SEC; size range and concentration were measured by MRPS using the nCS1 particle analyzer. (A) Representative plot of EV concentration (particles/mL, y-axis) and EV diameter (nm, x-axis) measured from EV isolate. A TS-400 cartridge was used to measure concentration of EVs ranging from 75 to 350 nm in diameter and a TS-2000 cartridge for EVs ranging from 250 to 2,000 nm in diameter. Similar concentrations measured by both cartridges in overlapping size ranges (250–350 nm) indicate accuracy and reliability across cartridge measurements. (B) Total plasma EVs/µL from healthy pediatric controls (n = 10), patients with sJIA with quiescent disease (sJIA Q, n = 10), and patients with sJIA with systemic flare (sJIA F, n = 7). Boxes extend from the minimal to maximum values, with line at median. All groups were compared using Kruskal–Wallis test, and in all comparisons, P > 0.05. Lines between the sJIA Q and sJIA F groups represent the same patient at different stages of the disease. Summary demographic and clinical information for subjects are in Supplementary Tables 1 and 2.

3.2. Measurement of cell-specific populations of EVs by flow cytometry with wash step

As various studies demonstrate that EV production is increased in pathologic states,^23,24 we sought to determine if the concentration of plasma EVs from specific cell types differs between patients with sJIA with inactive or active disease and healthy controls (Supplementary Tables 1 and 2 for patients and controls information and Supplementary Table 3 for information on antibodies used as cell markers). Using a filtration-based approach to prepare stained plasma EVs for flow cytometry,¹⁸ we found that EVs from platelets (CD41⁺) and endothelial cells (CD146⁺) were elevated during systemic flare relative to levels in quiescent sJIA as well as healthy controls (Fig. 3A and B). CD31⁺ EVs were also more abundant during flare relative to quiescence and healthy controls, which may indicate chronic endothelial cell activation (Fig. 3C). Erythrocyte-associated (CD235a⁺) EVs were not found to be elevated in sJIA compared to controls (Fig. 3D). The data in Fig. 3 were generated using a method that included a wash step requiring a 0.2-µm centrifugal filter, which may have eliminated a significant portion of particles <200 nm; as such, it is likely that mostly larger EVs (>200 nm) were analyzed with this approach.

3.3. Visualization and measurement of EVs by nanoscale flow cytometry (HRFC)

As the majority of plasma EVs are <200 nm in diameter, we used HRFC to evaluate the cellular origin of the EVs and determine changes between clinical disease states and in relation to healthy controls. We utilized a recently described approach⁹ that allows visualization of fluorescently labeled EVs as small as 100 nm (a size approximation relative to polystyrene beads) without requiring any filtration of stained plasma; filtration eliminates EVs below a particular size range prior to flow cytometry analysis. The gating strategy is described in Fig. 1. Supplementary Tables 1 and 2 contain patients and controls information, and Supplementary Table 3 has information on antibodies used as cell markers.

In single-stained EVs, we found that, relative to control samples, patients with sJIA with systemic flare had significantly higher levels of EVs marked by CD144, CD31, CD14, and CD16, while no differences between the groups were observed for single-stained CD41⁺ and CD62P⁺ EVs (Fig. 4). These results indicate increased EV production from endothelial cells (CD144⁺) and possibly chronically activated ECs based on CD31⁺, monocytes (CD14⁺), and neutrophils/natural killer (NK) cells (CD16⁺). In double-stained EVs from the same subjects, we found that CD41⁺/CD62P⁺ EVs (indicating activated platelets) and double-stained CD14⁺/CD16⁺ EVs (associated with intermediate monocytes) were significantly elevated in both sJIA F and Q in comparison to healthy controls (Fig. 5A). CD144⁺/CD31⁺ EVs (representing chronically activated endothelial cells) were also significantly elevated in sJIA F in comparison to healthy controls (Fig. 5A). Hierarchical clustering analysis of double-positive EV subpopulations showed clustering of disease state–specific groups, indicating an EV signature for active sJIA (Fig. 5B), suggesting that multiplex EV labeling may significantly increase cell specificity.

Fig. 4.

Single-stained surface marker–specific EV subpopulations measured by HRFC show higher EV concentrations in patients with sJIA at flare. Plasma EVs/µL from healthy pediatric controls (n = 12), patients with sJIA with quiescent disease (sJIA Q, n = 11), and patients with sJIA with systemic flare (sJIA F, n = 9). Boxes extend from the minimal to maximum values, with line at median. All groups were compared using Kruskal–Wallis test and Dunn's multiple comparisons test; *P < 0.05, **P ≤ 0.01. Lines between the sJIA Q and sJIA F groups represent the same patient at different stages of the disease. Summary demographic and clinical information for subjects are in Supplementary Tables 1 and 2.

Fig. 5.

Double-stained surface marker–specific EV subpopulations measured by HRFC discriminate between patients with sJIA and controls. (A) Double-stained plasma EVs/µL from the same subjects in Fig. 4. Boxes extend from the minimal to maximum values, with line at median. All groups were compared using Kruskal–Wallis test and Dunn's multiple comparisons test; *P < 0.05, **P ≤ 0.01. Lines between the sJIA Q and sJIA F groups represent the same patient at different stages of the disease. (B) Heatmap representation of double-positive EV populations analyzed by unsupervised hierarchical clustering. The rows show concentration for the double-positive EVs (positive for both CD41 and CD62P, CD14 and CD16, or CD144 and CD31), and the columns represent the samples (C, control; Q, sJIA quiescence; and F, sJIA flare). Red color represents high concentrations, whereas blue shade represents minimal levels, of the indicated EV populations. Log₂-transformed values with hierarchical clustering performed using Pearson correlation metric with average linkage method (https://software.broadinstitute.org/morpheus/).

We also analyzed EVs stained individually with CD15 (from neutrophil and monocyte populations), CD56 (expressed in NK cells, T cells, Dendritic Cells, and monocytes upon activation²⁵), CD105 (endoglin, expressed by angiogenic endothelial cells), or CD235, expressed by erythrocytes. Only CD235 EVs were significantly higher in sJIA F compared to the control group (Fig. 6).

Fig. 6.

Elevated EVs derived from CD235⁺ cells measured by HRFC. Boxes extend from the minimal to maximum values, with line at median. (A) CD15⁺ and CD56⁺ EVs/µL based on single staining of plasma from healthy pediatric controls (n = 5), sJIA Q (n = 6), and sJIA F (n = 5). (B) CD105⁺ EVs/µL based on single staining of plasma from healthy pediatric controls (n = 7), sJIA Q (n = 8), and sJIA F (n = 6). (C) CD235a⁺ EVs/µL based on single staining of plasma from healthy pediatric controls (n = 6), sJIA Q (n = 5), and sJIA F (n = 4). All groups were compared using the Kruskal–Wallis test and Dunn's multiple comparisons test; *P < 0.05. Lines between the sJIA Q and sJIA F groups represent the same patient at different stages of the disease. Summary demographic and clinical information for subjects are in Supplementary Tables 1 and 2.

For all single-marker profiles, we also analyzed EVs in relation to their size (small: <200 nm; large: >200 nm relative to polystyrene beads). Consistent with data above, the majority of EVs were <200 nm. More pronounced disease- and activity-related differences were observed in EVs <200 nm (Supplementary Fig. 1).

3.4. Validation of platelet EVs

We performed EV validation in a variety of ways using EV platelets from plasma and from in vitro activated platelets. We observed that even fractions from a large SEC preparation were composed of a mixture of small and large EVs, with the relative proportion changing from earlier to later fractions, as seen by nanoscale flow cytometry and TEM (Supplementary Fig. 2A–F). More specific validation comes from imaging isolated platelets and their EVs relative to polystyrene beads. The analysis showed that while platelet-poor plasma retains a few platelets, the majority of CD41⁺ signals come from EVs between 300 and 400 nm in this example, similar to previous work⁹ (Supplementary Fig. 2G and H). Nanoscale flow cytometry–sorted EVs from activated platelets (CD41⁺/CD62P⁺) were analyzed by NTA, showing a heterogeneous population with a peak EV concentration around 400 nm. MVs as well as small EVs (exosomes) were identified using scanning and Cryo-EM analysis (Supplementary Fig. 2I–N).

3.5. Correlations between cell-specific populations of EVs, circulating blood counts, and markers of inflammation

We next determined whether the concentrations of EVs derived from different cell sources correlated with the number of the related cells in the blood, as assessed by CBC (Table 1). The concentration of neutrophils was positively correlated with the concentration of CD16⁺ EVs. Neutrophil concentration also correlated with CD14⁺ EV concentration, but the correlation was not significant for double-stained CD14⁺CD16⁺ EVs, which were likely produced by monocytes. There was also a statistically significant positive correlation with double- and single-stained CD144⁺CD31⁺ EVs and neutrophils. Total monocyte concentration was negatively correlated with the level of double-stained CD14⁺CD16⁺ EVs. On the other hand, there was no correlation between the concentration of platelets and the abundance of CD41⁺ and CD62P⁺ EVs, either as single- or double-stained EVs.

Table 1.

EV subpopulations correlate with circulating cell types. Spearman correlation values for concentration of neutrophils and monocytes and EV subpopulations in sJIA subjects; numbers represent r and P values, respectively; significant correlations are in bold.

Cell Number (CBC)	CD144+ EVs	CD31+ EVs	CD144+CD31+ EVs	CD14+ EVs	CD16+ EVs	CD14+CD16+ EVs	CD15+ EVs	CD41 EVs	CD41+CD62P+ EVs
Neutrophilsa	0.69; 0.003	0.54; 0.026	0.59; 0.014	0.53; 0.029	0.69; 0.003	0.41; 0.101	0.45; 0.23	−0.02; 0.9	−0.38; 0.13
Monocytesb	−0.22; 0.41	−0.07; 0.81	−0.09; 0.75	−0.41; 0.114	−0.25; 0.34	−0.49; 0.023	0.07; 0.87	−0.19; 0.47	−0.01; 0.96
Plateletsc	0.37; 0.12	0.48; 0.04	0.37; 0.12	0.13; 0.6	0.22; 0.37	0.17; 0.48	0.44; 0.2	−0.06; 0.79	−0.28; 0.25

n = 17 for neutrophil correlations, except with CD15⁺ EVs, n = 9.

n = 16 for monocyte correlations, except with CD15⁺ EVs, n = 9.

n = 19 for platelet correlations, except with CD15⁺ EVs, n = 10.

ESR is a marker of inflammation that, although not specific to sJIA, usually increases during flares of systemic symptoms. We have shown ESR to be potentially informative to probe disease biology,²⁶ and ESR was available for most samples tested. We observed that ESR was correlated with levels of CD31⁺ EVs and CD144⁺ EVs as single-stained EVs, but the correlation decreased when double-stained EVs were analyzed (Table 2). ESR was positively correlated with CD16⁺ EVs. Levels of ferritin were positively correlated with CD31⁺, CD14⁺, and CD14⁺/CD16⁺ EVs (Table 2).

Table 2.

EV subpopulations correlate with measures of inflammation. Spearman correlation values for ESR and ferritin and concentration of EV subpopulations in sJIA subjects; numbers represent r and P values, respectively; significant correlations are in bold.

	CD31+ EVs	CD144+ EVs	CD31+CD144+ EVs	CD14+ EVs	CD16+ EVs	CD14+CD16+ EVs	CD15+ EVs	CD41+ EVs	CD62P+ EVs	CD41+CD62P+ EVs
ESRa	0.54; 0.02	0.5; 0.03	0.45; 0.053	0.39; 0.1	0.48; 0.038	0.42; 0.08	0.64; 0.05	−0.20; 0.4	0.1; 0.69	−0.23; 0.35
Ferritinb	0.84; 0.002	0.54; 0.094	0.58; 0.064	0.78; 0.006	0.62; 0.048	0.67; 0.028	Ndc	−0.17; 0.61	0.13; 0.71	−0.24; 0.49

n = 19 for ESR correlations, except with CD15⁺ EVs, n = 10.

n = 11 for ferritin correlations.

Nd, not done.

3.6. Measurement of tetraspanin marker–positive EV populations

There is not one universal EV marker; however, CD9, CD63, and CD81 are tetraspanin proteins commonly found on subsets of exosomes. No significant difference in total EV subpopulation for any of these 3 EV subpopulations was found between the groups, although median abundance was relatively higher in sJIA F samples (Supplementary Fig. 3A). This trend is more pronounced in EVs <200 nm, where CD63⁺ EVs <200 nm were significantly more abundant in sJIA F compared to sJIA Q but not for EVs >200 nm (Supplementary Fig. 3B and C).

3.7. Measurement of EV protein content

EVs carry bioactive protein cargo, and we hypothesized that the content of EVs would differ between sJIA and controls. We initially performed an analysis of EV content using a Luminex platform. Most analytes in EV isolates were below the limit of detection, due to diluted samples (a result of the SEC process). Nevertheless, we detected increased resistin in sJIA F relative to control, both in EVs as well as in plasma, even after adjusting for total EV concentration. Similar MFI distribution for resistin in plasma and EVs supports the idea that the majority of resistin is associated with EVs, rather than being in soluble form (Supplementary Fig. 4).

Next, we pooled and concentrated the EV isolates into 3 groups (control, sJIA flare, and sJIA quiescence) and carried out mass spectrometry–based protein analysis. Of nearly 1,300 identified proteins (Supplementary Table 4), 1,145 (90.1%) were found to be shared among all 3 groups (Fig. 7A). We used the FunRich analysis tool to analyze the gene ontology of identified proteins, finding that at least 87% (383) of the identified proteins were previously identified in EVs, as defined by the Vesiclepedia database accessed on 8 August 2022 (Fig. 7B). Further analysis showed that these proteins are mainly of extracellular, exosomal, and lysosomal origin (Fig. 7C). Proteins related to immunoglobulins were abundantly identified in the overall data set (Supplementary Table 4).

Fig. 7.

The majority of EV proteins identified by mass spectrometry are shared among the control, flare, and quiescent groups and are mostly of extracellular and exosomal origin. (A) Venn diagram showing the relationship between the proteins identified by mass spectrometry in isolated EVs from the 3 pooled groups (control, flare, and quiescent). (B) Venn diagram showing the relationship between the number of proteins identified in the EVs from the 3 pooled (flare, quiescent, and control) groups (Common) and proteins identified in the Vesiclepedia database. (C). Bar graph showing cellular origin of EV proteins identified in the EVs from the 3 pooled (flare, quiescent, and control) groups. (B) and (C) analysis using the software FunRich.

Qualitative comparison of the proteins found in each sample set showed a number of proteins exclusively expressed in EVs from sJIA flare (Supplementary Table 5), including the inflammation-responsive serum amyloid A2 protein (SAA-2), fitting with the systemic character of the disease. Two other proteins, BCAR1 and HEATR5B, have no known association with arthritis pathologies. The heat shock protein 47 (HP47; encoded by the MFGE8 gene) was also identified only in EVs from flare. Another set of proteins was found only during quiescence, including soluble TNFR1B variant 1 (Supplementary Table 6), which is the TNF-inhibitor drug, etanercept. Considering that 4 of the 10 quiescent samples are from patients on etanercept at the time of blood draw, the protein might be a contaminant from the plasma, as previously described.²⁷

Quantitative analysis of the pooled samples showed that several proteins were differentially expressed between flare and quiescence as well as compared to the control group, with proteins associated with acute phase response showing the largest upregulation in flare (Fig. 8A and B and Supplementary Tables 7 and 8). We used Ingenuity pathway analysis (IPA) to evaluate gene ontologies of proteins differentially expressed in flare compared to quiescence; the cutoff for selection was a P value (False Discovery Rate corrected) of ≤0.05 and at least 1.5-fold change (up or down). The analysis confirmed the acute phase response pathway as the top canonical pathway associated with the differentially expressed proteins in EVs from the sJIA flare. IPA analysis also showed that pathways associated with coagulation and the complement system are decreased in flare in comparison to the quiescence EV profile (Fig. 9A and B, Supplementary Table 9). IPA upstream analysis, which predicts molecules that may modulate the molecules expressed in sJIA EVs at flare, showed IL-6 (z-score 2.5) and IL-1 (z-score 2.5) among the top activated regulators, consistent with the known upregulation of these cytokines during sJIA flare,¹ in addition to IL-22 (z-score 2.6), previously found at elevated levels in serum from patients with rheumatoid arthritis.²⁸ TNF was also found to be an activated regulator (z-score 2.4). IL-10RA (z-score −2.6) and TGFB1 (z-score −2.6) were identified as inhibited regulators (Supplementary Fig. 5). Comparison of flare and control proteomic EV profiles showed similar results, with proteins associated with acute phase response showing the largest differences and with similar pathways involved (Supplementary Table 10).

Fig. 8.

Differential expression of proteins in EVs among the 3 groups. (A) Flare × Quiescence, (B) Flare × Control, and (C) Quiescence × Control comparisons. Solid black points are statistically significant, with points to the right (red points) manually selected to represent upregulated proteins and points to the left (blue points) representing downregulated proteins. Points without labels are predominantly represented by immunoglobulin heavy and light chain. Proteins intensities were normalized by log₂ transformation, and Student’s t-tests were performed using permutations-based False Discovery Rate.

Fig. 9.

Proteins involved in acute phase reaction, coagulation, and complement are differentially expressed in EVs from sJIA flare in relation to sJIA quiescence. IPA analysis showing (A) the top pathways involved. (B) Stacked top pathways, showing percentage of proteins downregulated (blue) and upregulated (orange) in each pathway. Numbers at the end of bar represent total molecules in the data set.

Comparisons of quiescence EV content with the control group showed that the highest difference was expression of TNFR1B/etanercept (Fig. 8C). CRP was increased during quiescence compared to control, as well as several proteins associated with actin cytoskeleton signaling, namely, ACTN1, FN1, FLNA, and TLN1 (Fig. 8C).

4. Discussion

In this study, we used a combination of SEC/MRPS and high-resolution nanoscale flow cytometry to investigate the concentration and origin of EVs in sJIA, as well as mass spectrometry–based proteomics to analyze EV protein content. SEC has been shown to preserve EVs very well, including SEC columns designed to purify EVs from small plasma volumes. Although ultracentrifugation (UC) has long been the gold-standard method of isolating whole EV populations, the volume required for UC is prohibitive for the study of pediatric blood samples. Also, it is now appreciated that the mechanical forces imposed by UC significantly compromise the structural integrity of EVs and the ability to isolate accurate EV populations. Furthermore, compared to previously used approaches that rely on dynamic light scattering (such as nanoparticle tracking analysis), resistive pulse-sensing techniques provide superior resolution and detection sensitivity, in part due to their ability to only measure spherical particles.²⁹

Using flow cytometry, we identified a unique EV signature that distinguishes systemic sJIA flare from healthy controls, based on the cellular source of EVs as reflected in surface marker expression. Included in this signature are EVs carrying markers of chronically activated endothelial cells (CD144/CD31), intermediate monocytes (CD14/CD16), and activated platelets (CD41/CD62P). sJIA quiescent samples clustered together, away from the flare samples, and closer to but in a separate cluster from healthy controls. This result supports our previous proposal that quiescent sJIA is not a complete return to healthy status but a state of “compensated inflammation.”³⁰

Although prior work has shown more abundant levels of platelets, neutrophils, and monocytes during sJIA flare,¹⁶ the increases in EV subpopulations observed during sJIA flare do not appear to be simply related to higher levels of the circulating cell types producing such EVs. In fact, the level of total monocytes tended to negatively correlate with the level of monocyte-associated EVs. For neutrophils, however, neutrophil counts positively correlated with higher levels of CD14⁺ and CD16⁺ EVs but not with double-positive CD14⁺CD16⁺ EVs, which are likely derived from monocytes, suggesting a direct connection between elevated circulating neutrophils and the abundance of neutrophil-derived EVs. Furthermore, we also found statistically significant positive correlations between neutrophil counts and CD31⁺CD144⁺ EV levels. Direct interaction between neutrophils and endothelial cells, or between neutrophil-derived EVs, as discussed below, may affect production of EVs by endothelial cells.

Neutrophil-, monocyte-, and platelet-derived EVs may be involved in sJIA pathogenesis in several ways. Neutrophils have an activated phenotype in sJIA, both during active disease, including onset, and during inactive disease.^31,32 Neutrophil-derived EVs may have the potential to affect the endothelium, damaging or protecting the endothelial layer depending on their cargo.³³ Monocyte-derived EVs contain inflammasome components and mature IL-1β, a key cytokine in sJIA.³⁴ EVs from platelets also have been shown to carry mature, functional IL-1β.^35,36 Monocyte stimulation boosts EV release and increases activating effects on endothelial cells,³⁴ suggesting a novel mechanism of inflammatory potentiation. Similarly, platelets and monocytes selectively package miRNA into EVs upon exposure to various stimuli. The miRNA content within EVs is markedly more diverse than that within the parent cells. EV-associated miRNA can be released into endothelial cells and subsequently downregulate target gene expression, leading to sustained effects on endothelial cell phenotype and behavior.^37,38 Thus, both proinflammatory cytokines and miRNA carried by EVs can regulate the biologic function of recipient cells.

The increase in EVs from endothelial cells, including chronically activated endothelial cells (CD144/CD31), during sJIA flare is of special interest. This increase was observed using both our initial flow cytometry analysis of larger EVs and subsequent nanoscale flow cytometry analyses that also captured EVs smaller than 200 nm in diameter. Endothelial dysfunction is often seen, but not currently well understood, in sJIA and other rheumatic and inflammatory diseases complicated by vascular injury. Activation of the vascular endothelium is observed at sJIA onset and during flares, with upregulation of leukocyte adhesion molecules like E-selectin, vascular cell adhesion molecule 1, and intercellular adhesion molecule 1 on the cell surface. Endothelial activation leads to recruitment of neutrophils and monocytes into tissues.⁵ Rash and coagulopathy, the latter reflected by elevated circulating D-dimers, also indicate endothelial activation in sJIA.^3,4 Endothelial cells can be activated by EVs carrying IL-1β,³⁴ a cytokine strongly implicated in sJIA based on the striking efficacy of therapeutic IL-1 blockade.^13,39,40 We also found that levels of ESR, although a nonspecific measure of inflammation, correlate with CD31⁺ EVs and CD144⁺ EVs in sJIA, suggesting that these EVs may be directly involved in clinical manifestations in sJIA. These results also highlight that endothelial cells may play a role both as target cells and as a source of EVs in sJIA. Endothelial cell activation may thus contribute importantly to sJIA pathogenesis, and further studies of endothelial cell–associated EVs are warranted.

Elevated serum ferritin is often observed in sJIA and may be useful in diagnosis.^41,42 In this study, we found that ferritin levels in patients with sJIA were positively correlated with CD14⁺CD16⁺ monocyte-associated EVs. Interestingly, serum ferritin levels have been shown to correlate with levels of activated monocytes/macrophages, and high ferritin may be associated with endothelial cell dysfunction,⁴³ suggesting that cell activation state may affect production of EVs in sJIA.

Although we found that amounts of EVs with specific cell markers were altered in sJIA, we did not find a difference in total concentration and size distribution of EVs between sJIA and healthy controls. The precise size of EVs relative to polystyrene beads imaged by nanoscale flow cytometry is the subject of intense controversy due to potential differences in light refractive index and light scatter.⁴⁴ However, for the purposes of our study, the relative EV size serves as a biomarker, and our flow cytometry–based size estimates were further bolstered by our MRPS findings that the majority of SEC-isolated EVs were <200 nm in diameter. Similarly, we did not find a difference in the concentration of EVs bearing the exosome-associated tetraspanin markers CD9, CD63, or CD81 in sJIA compared to healthy controls, except for a significant increase in CD63⁺ EVs below 200 nm during sJIA flare. This result indicates that changes in concentrations of EVs from specific cell sources potentially predominate in sJIA, although further studies are needed to clarify the role of other EV types in sJIA. Validation of isolated EVs from platelets using various approaches showed a size heterogeneous population, and small (exosomes) and larger EVs (microvesicles) were identified.

We performed an initial analysis of EV protein content using EV isolates in a Luminex assay, which found resistin to be differently expressed in EVs from flare patients and controls, with most other analytes (except plasminogen activator inhibitor-1, and macrophage migration inhibitory factor) below the limit of detection for EV isolates. This result likely reflects the relatively dilute volume of EVs due to the SEC process. Nevertheless, the finding of elevated resistin in plasma during sJIA flare is in agreement with previous findings of elevated serum resistin in JIA, especially sJIA.⁴⁵ Our results further suggest that a significant portion of the plasma resistin is associated with EVs. Resistin in humans is produced mostly by peripheral blood mononuclear cells (PBMCs) and macrophages, suggesting its link to inflammation, whereas in mice, it is expressed by adipocytes. Resistin induces the production of several cytokines, including cytokines important in sJIA pathology such as IL-1, possibly by binding to toll-like receptor 4 and activation of intracellular signaling pathways, including NF-κB. Resistin may also affect endothelial cell function.⁴⁶

Our subsequent analysis of EV content was performed in pooled samples—flare, quiescence, and control—using mass spectrometry. We found that around 90% of proteins identified were common among the 3 groups, with a small number of proteins uniquely identified in flare or quiescence, as well as in controls. Pooling of samples may lead to a loss of sensitivity, with small but statistically significant changes occurring in a few samples being lost in pooled samples. Among the identified proteins, several are related to immunoglobulins. Immunoglobulins are considered a common contaminant of exosomes isolated by ultracentrifugation, polyethylene glycol–based approaches, or SEC-based methods,²⁷ and additional steps may be necessary to decrease the amount of immunoglobulins associated with isolated EVs. Although immunoglobulins are considered mostly a contaminant of EV isolates, Huang et al.⁴⁷ found that exosomes associated with immunoglobulins in plasma were able to activate the classical complement pathway and that mice lacking IgG due to a defect in B-cell maturation showed decreased retinal vascular damage in a model of diabetic retinopathy. The authors also showed that depletion of exosomes by ultracentrifugation led to reduction of immunoglobulins, suggesting that most immunoglobulins in circulation may be associated with exosomes. Immunoglobulins associated with exosomes, or at least a subset of these exosomes, may have some biological function, such as complement activation.

Among the unique proteins identified in flare, in addition to SAA-2,⁴⁸ we found BCAR1, part of the Cas (Crk-associated substrate) family of scaffold proteins;⁴⁹ the cytoskeletal protein HEATR5B, previously identified in EVs from cerebrospinal fluid of individuals with traumatic brain injury;⁵⁰ and the HP47 protein (encoded by the MFGE8 gene), which is found in multiple cell types, including endothelial cells, and has been associated with several inflammatory diseases⁵¹ but not previously with sJIA, to the best of our knowledge. Further studies will be needed to confirm the significance of these findings in relation to sJIA disease activity. In EVs from quiescent patients, the only unique protein besides immunoglobulins was TNFR1B variant 1, which is etanercept, and might have been isolated with the SEC EV preparations as previously observed.²⁷

Further analysis of the proteins of the pooled EVs from flare patients showed a largely proinflammatory profile, with increased expression of components from inflammatory-associated pathways, and with IL-1 and IL-6, major cytokines involved in sJIA,¹ as potential upstream regulators. The serine protease inhibitors, SERPINA1 and SERPINA3, as well as SERPINC1 (antithrombin III), are enhanced during flare, while several coagulation factors are decreased, suggesting that coagulation, as well as components of the complement cascade, are inhibited in this sample of EVs. This finding contrasts with an RNA expression study that analyzed PBMC samples from patients with sJIA with recent-onset disease, prior to treatment,⁵² where genes associated with coagulation and complement were upregulated. In our study, regulation of coagulation and complement pathway components in EVs might have been modified by treatment as well as duration of disease. Another possibility, not mutually exclusive, is a role for EVs in the control of inflammatory response in sJIA.⁵³

Our analysis of EVs in sJIA suggests that this approach may offer additional insight into sJIA disease pathology. However, our study has several limitations. sJIA is a rare pediatric disease, and available samples were limited in volume. We used an existing collection of samples collected in heparinized tubes, which could artificially increase the number of certain types of EVs.⁵⁴ Further studies will be necessary to confirm our findings, and a valuable extension of our work will be functional assessment of the various populations of EVs. Comparison to similar inflammatory diseases, such as Kawasaki disease and sepsis, will be helpful to determine the specificity of our findings for sJIA, and studies of clinical sJIA subtypes (systemic vs arthritis-only flare, monocyclic vs polycyclic vs persistent course) will be helpful to determine the potential of EVs as prognostic biomarkers. Longitudinal samples will be useful to determine the potential of EVs to serve as predictive biomarkers of impending flare or response to therapy.

In summary, we show that in sJIA, EVs from different cell sources are elevated, especially during flare, implicating activation of these cell types and their production of EVs in sJIA. Importantly, EVs found in plasma could be a response to, as well as a cause of, inflammation. The finding of elevated endothelial cell–associated EVs and the potential to stain EVs for tissue-specific markers suggest that plasma EV analysis may serve as a proxy to probe tissues that are difficult to access in patients with sJIA. Our findings were generated using relatively small volume samples (∼200 µL for HRFC), indicating a promising opportunity for the study of diseases of childhood.

Authorship

J.M. designed research studies, performed experiments, acquired and analyzed data, and wrote the first draft of the paper. T.M. designed experiments and critically reviewed the paper. M.M. and Y.J. performed experiments and data acquisition. K.J.S. designed research studies and critically reviewed the paper. F.M.C. performed experiments and data analysis. J.E.E. designed experiments and critically reviewed the paper. C.M. designed research studies, performed experiments, acquired and analyzed data, and wrote the paper. E.M. obtained funds, designed research studies, and wrote the paper.

Supplementary material

Supplementary materials are available at Journal of Leukocyte Biology online.