The current study was approved by Hamad Medical Corporation IRB (MRC-03-19-020). The patients were admitted to the hospital within 24 h of acute ischemic stroke. After obtaining informed consent, the patient’s demographics were recorded in an electronic case record form. The initial clinical examination of the participants was completed within 24–48 h from the onset of symptoms. All patients had a brain MRI to confirm the location of the stroke. The clinical diagnosis, risk factors, comorbidities, and treatment course were recorded. 27 stroke patients were included in this study (median age = 48 years; gender = 25 males and two females). The stroke patients were divided into two clinical subgroups based on their HbA1c status DS group; HbA1c ≥ 6; n = 13) and nDS group; HbA1c ˂6; n = 14). In addition, samples were also obtained from 12 healthy controls (median age 46 years; gender = 8 males and four females) with no previous history of stroke. Blood samples were collected in plain tubes and allowed to clot for 30 min at room temperature and later centrifuged at 1,500 g for 10 min to separate serum. All serum samples were stored at −80°C until the EV isolation and mass spectrometry analysis were performed.

Equal volumes of serum (200 µL) diluted in PBS containing protease inhibitors from all the subjects were differentially centrifuged at 3,000 g for 15 min to remove cell debris and larger vesicles. The supernatants were transferred to sterile microcentrifuge tubes and were centrifuged at 20,000 g for 20 min at 4°C (Eppendorf Centrifuge 5424R, Hamburg, Germany) to isolate the microparticles. The microparticle-free serum (MPS) was diluted in PBS (pH 7.4). The MPS was used for the isolation of EVs by ultracentrifugation (Hitachi CP100NX, Tokyo, Japan) at 100,000 g for 120 min at 4°C. The supernatants were transferred to separate vials and pellets were resuspended in 1 mL PBS. The ultracentrifugation procedure was again repeated in similar conditions. The final pellets obtained at 100,000 g were resuspended in 30 µL of PBS and stored at −20°C.

Western blot was performed to characterize CD63 and CD81 EV markers in the isolated samples. Equal volumes (7 µL) of EVs from each sample were lysed in 4x Laemmli buffer (BioRad, Hercules, CA, United States) under reducing conditions with 5 mM of 2-mercapto-ethanol, followed by sample denaturation at 95°C for 10 min. Western blot was performed with 0.2 µm polyvinylidene difluoride (PVDF) membranes. The membrane was blocked with clear milk-blocking buffer prepared in tris buffer saline with tween (TBST; 0.1%) on a rocker in the cold room for 1 h. The blot was then exposed to CD63 and CD81 primary antibodies (Abcam, Cambridge, MA, United States) at 4°C overnight. On the following day, the blots were incubated with the corresponding secondary antibodies (goat anti-mouse-HRP; 1:1,000) for 1 h. After washing, the membranes were visualized by Clarity Western ECL substrate (Bio-Rad) and images were acquired using the Bio-Rad Chemidoc MP Imaging system.

The morphology of EVs was characterized by Transmission Electron Microscopy (TEM) and Dynamic Light Scattering (DLS) following protocols described in our previous studies (Akbar et al., 2022). For TEM, 5–10 µL of EV samples were placed on copper grids with 300 mesh holey-carbon film (TEM-CF300CU50; Sigma-Aldrich) and incubated at room temperature for 15 min. The adsorbed EVs on the TEM grid were fixed by adding a drop (5–10 µL) of 4% of formaldehyde and 2% glutaraldehyde solutions. The grid was washed three times with PBS. Following washing, 5 µL 0.1 M sodium cacodylate buffer was added for 5 min as a negative stain and washed three times again using ultrapure deionized water. Later, 0.2% uranyl acetate (filtered by a 0.2 µm syringe filter) was added to the TEM grid and incubated for 5 min. The grid was washed 5 times with ultrapure distilled water. Sample dehydration was performed with gradient ethanol series twice, incubation starting from 50%, 70%, 80%, and 90% for 5 min each and 100% anhydrous ethanol for 10 min. Finally, the TEM grids were air-dried for 5 min and stored in a vacuum desiccator. Next, the EVs on TEM grids were analyzed under TEM (FEI, Talos F200X).

The hydrodynamic size of EVs was estimated using a Zeta-sizer nano ZSP (Malvern instruments, UK). Briefly, EVs were diluted in 1 X PBS (1:1 ratio), and 10 µL of the samples was transferred to a Quartz microcuvette (ZEN00400). DLS measurements were conducted at 24°C operating at 633 nm, and the back-scattered light was recorded at an angle of 175°. The sample was allowed to equilibrate for 2 min before each measurement. The light scattering was recorded for 200 s with three replicate measurements. Using the Malvern Instruments software DLS signal intensity was transformed to volume distribution, assuming a spherical shape of the EVs.

Invitrogen NuPAGE 4%–12% bis-tris polyacrylamide precast gels (Invitrogen, United States) were used for optimal separation of the proteins and sample quality check using standard protocols (Looze et al., 2009). Samples were prepared by dissolving 10 μL of EVs, 5 μL of NuPAGE LDS sample buffer (4x), 1 μL of dithiothreitol (DTT) (10X), and 6.5 μL of deionized water. The running buffer was composed of 50 mL of NuPAGE 20X MES dissolved in 950 mL of deionized water to prepare 1X SDS.

10 μg of protein were used for in-gel digest. Samples were run in NuPAGE 4%–12% Bis-Tris polyacrylamide for 10 min to get one big band. Gels were washed with deionized water for 15 min on the shaker. The big band per lane was cut into small equivalent blocks and transferred to 1.5 mL tubes. For digestion, 12.5 ng/μL of trypsin in 25 mM ammonium bicarbonate (ABC) was used for overnight sample incubated at 37°C. The next day, gel blocks were centrifuged for 1 min at 15,000 g and supernatants were transferred into fresh tubes. Later, gel blocks were covered with 30% acetonitrile (ACN) (Thermo Fisher Scientific, United States) and 3% trifluoroacetic acid (TFA) (Sigma-Aldrich, United States) for 10 min, and 100% ACN for 10 min. Finally, the supernatants were concentrated using speedvac at 37°C.

LC-MS/MS analysis was conducted on a Q-Exactive HFX mass spectrometer with a nano-flow chromatographic system (Easy nLC-II 1200, Thermo Scientific). Samples were injected into Acclaim™ PepMap™ 100 Nano-Trap column (particle size 5 μm, 100Å, I.D.100, length 2 cm, Thermo Scientific). Samples were eluted into EASY-spray C18 LC column (particle size 3 μm, 100Å, I.D.50, length 15 cm, Thermo ScientificTM) and separated by a 120-min gradient from buffer A (0.5% formic acid in distilled water to buffer B (80% ACN, 0.5% formic acid (FA). Ionized peptides were introduced into the mass spectrometer by EASY-Spray™ Source (Thermo ScientificTM) at 40°C column temperature. Full scans were acquired at a resolution of 70,000 (m/z 300). Data-dependent analysis (DDA) was performed on the 15 most intense ions detected in the full scan. Each LC-MS/MS sample was analyzed in triplicates.

Raw data were processed with the Genedata Refiner MS software (v13.0.1, Genedata, Basel, Switzerland). MS1 data were aligned between the runs, and MS2 data identifications were transferred to missing values. Processed MS2 spectra were used for identification. The FASTA database search was performed with the MASCOT search algorithm v2.6 against a UniProt/SwissProt database (2019) and limited to human entries while ensuring an adjusted FDR of <1%. The precursor tolerance was set to 20 ppm and MS2 tolerance to 0.05 Da. Only proteins detected with two unique peptides were considered for further quantitative analysis.

All statistical analyses were performed using R 4.2.2. Group comparisons were made, and volcano plots were constructed using a general linear model (glmQLFit) in the edgeR package (version 3.14.0). Pairwise contrasts (log2 foldchange (log2FC)) and Benjamini–Hochberg (BH) p = 0.05 were used for the analysis. Clinical parameters, including age, gender, BMI, statin usage, and anticoagulant usage, were utilized as covariates for glmQLFit analysis. Volcano plots were constructed for nDS vs HC (Figure 3), DS vs HC (Figure 4), and DS vs nDS (Figure 5) comparisons. PCA plot was constructed using the PCA tools package in R. Glucose, insulin, HbA1c, blood pressure, and lipid profile were compared among clinical parameters using either T-tests or ANOVA.

The observed proteins were analysed using the g:Profiler, a web server for functional enrichment analysis. p-values were calculated and corrected with BH. Only hits from GO:CC and REAC with a cut-off p < 0.01 were allowed.

Differentially expressed proteins (nDS vs HC = 4; DS vs HC = 123, and DS vs nDS = 149) were further processed using the Ingenuity Pathway Analysis (IPA) core analysis for functional pathway annotation. The IPA canonical pathway predicted activation (z-score >2) and inhibition (z-score ≤ −2) of specific pathways using the submitted protein list and log2FC values.

13 DS and 14 nDS, along with 12 HC subjects, were included in this study. Stroke risk factors in the DS group were hypertension (85%) and dyslipidemia (100%). 15% of DS patients were taking anticoagulants, 77% were taking antihypertension medications, and 100% were taking antiplatelets and statins. As per TOAST classification (Adams et al., 1993), 77% of the DS patients had small vessel disease (SVD), 15% had a stroke of other determined etiology (SDO), and 7% had intracranial hemorrhage (ICH). In the nDS group, stroke risk factors included hypertension (71%), dyslipidemia (91%), and atrial fibrillation (AF) (7%). 14% of nDS patients took anticoagulants, 71% antihypertension, 85% antiplatelets, and 93% statins. According to the TOAST classification, 50% of the nDS patients had SVD, 28.5% had LVD, 7% had a cardioembolic stroke, 7% had SDO, and 7% had ICH. Table 1 summarizes clinical characteristics of the patients.

The characterization of EVs by TEM showed that EVs were spherical or oval with an average diameter of 90 nm and sizes ranging from 30 nm to 200 nm. Figure 1A depicts a representative TEM image from a healthy control patient. No differences in EV morphology or sizes were found between the HC, DS, and nDS groups.

Figure 1B depicts the hydrodynamic sizes of EVs from all three groups. DLS size measurements revealed a wide range of EV sizes. The z-average (d.nm) hydrodynamic diameters of EVs from the HC, DS, and nDS were 110 nm, 118.6 nm, and 122 nm, respectively.

Western blot analysis of the endosome-specific tetraspanin CD63 and CD81 was performed to determine the endosomal origin of the EVs. The CD63 and CD81 were seen as band between 50–70 KDa and 22–25 KDa respectively. Figure 1C highlights that CD63 and CD81 were expressed in our samples, inferring that the EV isolation procedure was successful.

In the next step, we aimed to characterize and compare EVs proteome with the serum and EV-free serum samples. Mass spectrometry was employed to analyze four random samples each of EV, serum, and EV-free serum. The PCoA plot analysis demonstrated distinct clustering of EV samples compared to serum and EV-free samples. EV samples formed a distinct cluster, separate from the clustered grouping of serum and EV-free serum samples (Supplementary Figure S1A). This suggests that the proteome of EVs differs from that of serum. Heatmap analysis similarly indicated that proteins elevated in EVs exhibit lower levels in serum samples and vice versa (Supplementary Figure S1B). The proteins that were found to be higher in EVs were cross-referenced with the ExoCarta database (http://www.exocarta.org/), an exosome database that provides information on the contents identified in exosomes across multiple organisms. Few of the proteins that exhibited comparatively higher levels in EVs are enlisted in Supplementary Figure S1C. Additionally, proteins highlighted in red correspond to entries documented in the ExoCarta database. Finally, it was also observed that enlisted proteins are interconnected as revealed by protein-protein interaction networks functional enrichment analysis (Supplementary Figure S1D) (https://string-db.org/).

The raw data of LC‐MS/MS was processed in REFINER MS^® 7.5 software (Genedata, Switzerland). In total, 1,288 proteins were identified in the EVs. G:profiler enrichment analysis for the cellular compartment (GO:CC) revealed that many proteins belonged to extracellular vesicles. For example, blood microparticles (GO:0072562) and EVs (GO:1903561) enlisted most of the observed proteins at a significantly high p-value (P_adj ≤ 5.6 × 10⁻⁵⁵). Furthermore, the g:profiler REAC analysis mapped EV proteins associated with the complement system (REAC:R-HAS-166658 and REAC:R-HAS-977606). Figures 2A, B show the top six GO:CC and two GO:REAC hits with significant p-values for the annotated functions. The analysis confirmed the EV origin of the proteins.

In the next step, proteins (n = 387) with more than two unique peptides were used for statistical comparison. A PCA plot was constructed to compare the study groups (Figure 2C). Changes in the DS group proteome was highlighted in PCA, revealing a spatial pattern. The analysis observed that the DS group clustered distinctly from the HC and nDS groups at PC1. Although the nDS and HC groups were distinct, many samples overlapped. These findings suggest qualitative and quantitative differences between the EV proteomics of the DS group compared to the others.

Differential expression analysis of 387 proteins was performed to compare the nDS and HC groups using a general linear model. Herein nDS vs HC, only four proteins exhibited statistically significant changes (p ≤ 0.05 and log2FC ± 0.58). Comparing nDS with the HC group, only one protein (FIBG) was elevated in the nDS group, while three (FA5, PIGR, and TSP1) were higher in the HC group (Figure 3).

Of the DS vs HC, comparison of 387 proteins, 123 proteins exhibited significant changes (p ≤ 0.05 and log2FC ± 0.58). Twenty-eight proteins in the DS group showed increased levels compared to 95 proteins, which were higher in the HC group (Figure 4). The most elevated proteins in the DS group were TTC16, FIBG, IKKE, BD1L1, PR14L, and FIBB.

The DS vs. nDS comparison found 149 proteins that differed significantly (p ≤ 0.05) and log2FC ±0.58). In the DS group, 33 proteins were elevated, while 116 were higher in the nDS group (Figure 5). The most elevated proteins in the DS group relative to the nDS group were PR14L, BD1L1, ZC3H3, IKKE, TTC16, DYH8, K1C10, and F13A.

IPA core function analysis investigates statistically significant (p ≤ 0.05) canonical pathways. Differentially expressed proteins (DS vs HC = 123 proteins and DS vs nDS = 149 proteins) were used for IPA core function analysis. Due to the presence of only four proteins with an FDR p ≤ 0.05 in the nDS vs HC comparison, conducting IPA pathways analysis for this comparison was not feasible. For DS vs HC results, “LXR/RXR activation” was the most inhibited pathway, followed by “response to elevated platelet cytosolic Ca²⁺“, and “DHCR24 signaling pathway” (Figure 6A). Nevertheless, several pathways, such as “acute response signaling,” “coagulation system,” and “fibrin clot formation,” exhibited notable activity, although discerning a specific pattern indicating the activation or inhibition of these pathways proved challenging. The IPA core analysis conducted on 149 differentially expressed proteins in the DS vs nDS comparison (Figure 6B) revealed the assignment of several pathways similar to those identified in the DS vs HC comparison. These pathways included “acute phase response signaling,” “LXR/RXR activation,” “response to elevated platelet cytosolic calcium,” and “DHCR24 signaling pathway.” Additionally, two pathways, “complement system” and “complement cascade,” were found to be activated.