Proteomic analysis of cerebrospinal fluid extracellular vesicles reveals synaptic injury, inflammation, and stress response markers in HIV patients with cognitive impairment

CSF samples from 20 HIV+ subjects were collected between 1998 and 2013 by the National NeuroAIDS Tissue Consortium (NNTC) (Manhattan HIV Brain Bank, National Neurological AIDS Bank, California NeuroAIDS Tissue Network, and Texas NeuroAIDS Research Center) and stored at − 80 °C. All subjects were enrolled with written informed consent and institutional review board (IRB) approval at each study site. HAND clinical diagnoses were determined using established criteria [51] based on formal neurocognitive testing and neurological evaluation. Inclusion criteria were use of combination ART, age > 40 years, and undetectable CSF viral load (VL) (< 50 HIV RNA copies/ml). We also included one subject with detectable CSF VL (328 copies/ml).

Human glioblastoma U87 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% exosome-free fetal bovine serum (FBS), 100 U/ml penicillin, and 100 mg/ml streptomycin. The cells were maintained at 37 °C in a humidified atmosphere of 5% CO₂. To induce oxidative stress, U87 cells were treated with increasing concentrations (10, 25, 50, 100, 250, 500, 1000 μM) H₂O₂ for 4 h in serum-free medium_. After 4 h treatment, cells were washed with phosphate-buffered saline (PBS) and cultured in exosome-depleted medium for 72 h. Control and H₂O₂-treated U87 cells and culture supernatants were harvested for immunoblotting and EV isolation.

CSF samples (400 μl) were centrifuged at 3000×g for 15 min to remove cellular debris. Supernatants were immunoglobulin (Ig) depleted by incubation with protein A/G agarose beads and protein-L beads at room temperature for 1 h. Twelve common abundant proteins (α1-Acid glycoprotein, fibrinogen, α1-antitrypsin, haptoglobin, α2-macroglubulin, IgA, IgG, IgM, albumin, apolipoprotein A-I, apolipoprotein A-II, and transferrin) were depleted using Proteome Purify-12 immunodepletion resin (R & D Systems) at room temperature for 30 min. Abundant protein-depleted CSF samples were then incubated overnight at 4 °C with ExoQuick exosome precipitating reagent (System Biosciences, Inc., Mountain View, CA). The mixture was centrifuged at 1500×g for 30 min and EV pellets were resuspended in 20 μl PBS. EV-depleted CSF was concentrated 5-fold by passing through a 10K Amicon filter and stored at − 80 °C until further processing. EVs from control and H₂O₂-treated U87 cells were isolated from 150 ml cell culture media by ultracentrifugation. In brief, cell culture media was centrifuged at 300×g at 4 °C for 10 min to remove floating cells. Supernatants were passed through a 0.2 μm filter to remove contaminating apoptotic bodies, larger microvesicles, and residual cell debris. The flow through was centrifuged at 150,000×g at 4 °C for 90 min to pellet exosomes and smaller vesicles. The supernatant was removed and EV pellet was resuspended in 35 ml cold PBS and then centrifuged at 150,000×g for 90 min. The resulting EV pellet was resuspended in 100 μl PBS.

CSF and U87 EV concentrations and sizes were measured by nanoparticle tracking analysis (NTA) on a ZetaView instrument (Particle Metrix, Germany). For electron microscopy, CSF EVs isolated from 300 μl CSF were suspended in 10 μl PBS containing 1% paraformaldehyde. EV samples were adsorbed for 1 min to a formvar/carbon coated grid and fixed for 5 min in 1% glutaraldehyde. EVs were washed on a drop of water and stained with 1% uranyl acetate for 30 s. EV morphology was analyzed with a Tecnai G2 Spirit BioTWIN transmission electron microscope (TEM) equipped with an AMT 2 k CCD camera at the Harvard University TEM core.

CSF EV and corresponding EV-depleted CSF samples were mixed with an equal volume of radioimmunoprecipitation assay (RIPA) lysis buffer (Triton X-100 1%, NaCl 150 mM, sodium deoxycholate 0.5%, Tris-HCL 50 mM, SDS 0.1%, pH 7.4). U87 cells and EVs were also lysed with RIPA buffer and protein concentration was estimated by BCA assay. Modified Laemmli 4X sample buffer was added to lysed EVs and EV-depleted CSF samples and boiled for 5 min. Equal volumes of CSF EVs and corresponding EV-depleted CSF samples were loaded on 4–12% gradient polyacrylamide gels. Forty micrograms of U87 cell and EV lysates were loaded in each lane. After electrophoresis, protein bands were transferred onto PVDF membranes for 1.5 h at room temperature. The membranes were blocked with 4% non-fat milk for 1 h and probed at 4 °C for human IgG (Sigma-Aldrich; Merck) CD9 (Santa Cruz Biotechnology), CD81 (System Biosciences), heat-shock protein 70 (HSP70) (System Biosciences), flotillin-1 (FLOT-1) (BD Biosciences), glial fibrillary acidic protein (GFAP) (Abcam), glutamine synthase (GLUL) (Abcam), parkinsonism associated deglycase (PARK7) (Abcam), and C-reactive protein (CRP) (Abcam). After secondary antibody treatment, blots were developed with enhanced chemiluminescence (ECL). Images were captured using the Bio-Rad ChemiDoc™ Imaging System. Densitometric quantification was carried out with ImageJ software.

To isolate CSF EVs for proteomics, CSF samples were pre-cleared using Protein A/G PLUS-Agarose beads (Santa Cruz Biotechnology) and Proteome Purify™ -12 kit (R&D systems). Immunodepleted CSF samples were characterized by SDS-PAGE and silver staining before EV isolation (Additional file 6: Figure S1a). CSF EVs were precipitated using ExoQuick reagent, digested in 0.5% RapiGest, and boiled at 100 °C for 5 min. Protein bands in EV fractions and EV-depleted CSF samples were separated on polyacrylamide gels and visualized by silver staining (Pierce) to estimate protein content. IgG depletion was confirmed by immunoblotting (Additional file 6: Figure S1b). EV and EV-depleted CSF proteins from corresponding CSF samples (400 μl) were analyzed by the Taplin Biological Mass Core Facility using an ABSciex 4800Plus MALDI-TOF/TOF mass spectrometer. In brief, proteins were reduced using 1 mM DTT (in 50 mM ammonium bicarbonate) for 30 min at 60 °C. Samples were then cooled to room temperature and iodoacetamide (stock in 50 mM ammonium bicarbonate) was added to a concentration of 5 mM for 15 min in the dark at room temperature. DTT was then added to a 5 mM concentration to quench the reaction. EV and EV-depleted CSF proteins were digested overnight using 5 ng/μl sequence grade trypsin (Promega, Madison, WI) at 37 °C. Samples were then desalted by an in-house desalting column. Peptides were extracted by removing the ammonium bicarbonate solution, followed by one wash with a solution containing 50% acetonitrile and 1% formic acid. Extracts were then dried in a speed-vac (~ 1 h) and stored at 4 °C until analysis. On the day of analysis, samples were reconstituted in 5–10 μl of HPLC solvent A (2.5% acetonitrile, 0.1% formic acid). A nano-scale reverse-phase HPLC capillary column was created by packing 2.6 μm C18 spherical silica beads into a fused silica capillary (100 μm inner diameter × ~ 30 cm length) with a flame-drawn tip [52]. After equilibrating the column, each sample was loaded via a Famos auto sampler (LC Packings, San Francisco CA). A gradient was formed, and peptides eluted with increasing concentrations of solvent B (97.5% acetonitrile, 0.1% formic acid). As peptides eluted, they were subjected to electrospray ionization and then entered into an LTQ Orbitrap Velos Pro ion-trap mass spectrometer (Thermo Fisher Scientific, Waltham, MA). Peptides were detected, isolated, and fragmented to produce a tandem mass spectrum of specific fragment ions for each peptide. Peptide sequences (and hence protein identity) were determined by matching protein databases with the acquired fragmentation pattern using Sequest software (Thermo Fisher Scientific, Waltham, MA) [53]. All databases included a reversed version of all sequences, and the data was filtered to 1–2% peptide false discovery rate (FDR). Common high abundance proteins were excluded from downstream analysis because the majority are likely to be blood-derived (Additional file 1).

U87 cells were seeded in 96-well plates at a density of 5 × 10⁴ cells/well. Cells were treated in triplicate with 10, 25, 50, 100, 250, 500, or 1000 μM H₂O₂ for 4 h. The media was removed, cells were washed with PBS and cultured in fresh DMEM supplemented with 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.5 mg/ml MTT reagent (Sigma-Aldrich; Merck) for 4 h in the dark at 37 °C. The resultant blue formazan crystals in cells were dissolved in dimethyl sulfoxide. Optical density in each well was determined at 570 nm using a microplate reader.

Intracellular ROS was measured using dichlorodihydrofluorescein diacetate (H₂DCFDA) reagent. Cells were treated with 50, 100, or 250 μM H₂O₂ for 4 h. Mock and H₂O₂-treated cells were loaded with 20 μM H₂DCFDA for 45 min at 37 °C. Following two washes with pre-warmed PBS, fluorescence was measured at 485ex/535em nm or analyzed with a fluorescence microscope. Data were normalized to corresponding protein concentrations.

Functional annotation was performed by gene ontology (GO) mapping and expression annotation using Biobase (genexplain.​com/​transfac) and PANTHER (pantherdb.​org). Proteins identified by ≥ 2 peptide count were included for analyses. Heatmaps were generated using log₂ transformed peptide intensities from LC-MS analyses using R (version 3.5.1). Immunoblot bands were quantitated using ImageJ software and graphical representations plotted in GraphPad prism version 7.0. Densitometric values represent mean + SEM. Differences in CSF EV concentrations and number of proteins detected in CSF EV between groups were analyzed by Mann-Whitney U test. Differences between HAND and control subjects, or between control and H₂O₂-treated U87 cell lysates and EVs, were analyzed by one-way ANOVA. Differences in ROS induction and cell viability between control and H₂O₂-treated U87 cells were analyzed by t test.

Demographic and clinical characteristics of the study cohort are summarized in Table 1. The cohort consisted of 20 HIV+ subjects with advanced disease (90% with nadir CD4 count < 200 cells/μl and median nadir CD4 count 31 cells/μl) and therefore at high risk of CNS injury. Subjects were male, 75% white, with median age 52.2 years (interquartile range [IQR], 46.9–58.7 years). Median duration of HIV infection was 13.3 years (IQR, 10–21). Ten HIV+ subjects were cognitively impaired (9 had HAND diagnoses, of which 2, 3, and 4 subjects had ANI, MND, and HAD, respectively, while 1 had neuropsychological impairment due to other causes [NPI-O]). HAD subjects were included to evaluate changes in the CSF EV proteome in severe as well as milder forms of HAND. The majority (90%) were on ART (80% on protease inhibitors), with 65% and 94% having undetectable plasma and CSF VL, respectively. Subjects with HAND were mostly viremic (60%) and had low median CD4 count of 157 cells/μl (IQR, 75–182), while those without HAND were mostly aviremic (90%) and had a higher median CD4 count of 262 cells/μl (IQR, 146–411). With regard to substance use and comorbidities, the prevalence of smoking, alcohol use, HCV seropositivity, and depression was 75%, 15%, 45%, and 44% among all HIV+ subjects, respectively, with no significant differences by HAND status, while cocaine use and cerebrovascular disease were more frequent in HAND subjects (40% and 30%, respectively). Seven HAND and 3 non-HAND subjects died within 6 years following the sample date; causes of death included cirrhosis, end stage renal disease, pneumonia, lymphoma, and non-AIDS-defining cancers including metastatic adenocarcinoma, metastatic anal squamous cell carcinoma, lung cancer, and renal cell carcinoma. Only one of these 10 autopsied cases had HIV encephalitis.

CSF EVs were isolated and evaluated for morphological and molecular characteristics. The workflow is shown in Fig. 1a. TEM revealed both round and cup-shaped vesicles with diameter between 50 and 150 nm, corresponding to size of exosomes (Fig. 1b). The average diameter of CSF EVs was 126 nm with a peak at 100–150 nm (Fig. 1c). However, some smaller (< 50 nm) as well as larger vesicles (> 150 nm) were also detected; hence, particles were termed EVs rather than exosomes. EV concentrations ranged between 10⁹ and 10¹² particles/ml with a mean value of 1.9 × 10¹¹ particles/ml. Isolated EVs were further characterized by immunoblotting for exosome markers CD9, CD81, FLOT-1, and HSP70. Concentrated EV-depleted CSF from the corresponding samples were used as controls. CD9, CD81, FLOT-1, and HSP70 were enriched in EV fractions compared to EV-depleted CSF samples (Fig 1d). CSF EV proteins were characterized by silver staining, which indicated depletion of abundant proteins and distinct patterns of protein bands in EV fractions compared to EV-depleted CSF (Fig. 1e) (Additional file 6: Figure S1). The endoplasmic reticulum (ER) membrane markers calnexin and ERP72 were not detected in CSF EV fractions by western blotting, suggesting EV preparations were free of ER contamination (Additional file 6: Figure S2).

To identify protein cargo in CSF EVs from HIV+ subjects, isolated EV fractions and corresponding EV-depleted CSF samples were subjected to untargeted LC/MS/MS analysis. Combined analyses from 20 HIV+ subjects identified a total of 2727 and 1626 proteins in EV and EV-depleted CSF samples, respectively, using the SEQUEST search algorithm with a 1% FDR threshold (Fig. 2a). Proteins detected in EV and EV-depleted CSF samples are shown in Additional files 2 and 3, respectively. Unique peptides per protein in EV fractions ranged from 1 to 172 (median 2) and in EV-depleted CSF from 1 to 50 (median 2), with 12.4% and 12.2% average sequence coverage, respectively. Previous studies have shown that albumin, antitrypsin, apolipoproteins, cystatin, haptoglobin, immunoglobulins, macroglobulins, prostaglandins, and transferrin are highly abundant in CSF irrespective of disease status [17, 18, 54‐57]. In EV fractions and EV-depleted CSF, 26% and 22.4% of proteins, respectively, were identified as high abundance CSF proteins and excluded from downstream analyses (Additional files 2 and 3). Among the remaining proteins, 1134 (56.2%) in EV fractions and 702 (55.6%) proteins in corresponding EV-depleted CSF were identified by two or more peptides in at least one subject.

GO analysis and expression annotation of proteins detected with ≥ 2 unique peptide counts in EV fractions and EV-depleted CSF mapped to biological processes including immune responses, inflammation, stress responses, metabolic processes, and mitochondrial functions, as well as cellular components including neurons, myeloid cells, astrocytes, oligodendrocytes, endothelial cell, epithelial cells, and vesicles (Fig. 2b, c). Given that HIV may infect choroid plexus (CP) cells [58, 59] and dysregulate BBB functions [60, 61], CP and BBB markers were manually curated [48, 62‐64]. Full lists of proteins in each biological process and cellular component category are shown in Tables 2 and 3 (CSF EV) and Additional files 4 and 5 (EV-depleted CSF). A subset of proteins (25 in EV fractions and 13 in EV-depleted CSF) identified in ≥ 6 subjects by single peptides with > 10% coverage including CD9, CD81, HLA-A, HLA-DRA, DEFA1, INA, MOG, MRC2, REG3A, TIMP1, TIMP2, and OCLN were also included. As expected, more vesicle-related proteins were identified in EV fractions (19%) compared to EV-depleted CSF (9%). Stress and immune responses were the most frequently represented biological processes in both EV (21% of proteins for both) and EV-depleted CSF (27% and 24% of proteins, respectively) (Fig. 2b, c), while epithelial cells and neurons were the most frequently represented cell types in EVs (24% of proteins) and EV-depleted CSF (27% of proteins), respectively.

Although total peptide counts and percentage of proteins mapped to individual biological process and cellular component categories in EV and EV-depleted CSF were similar, the majority of proteins were more abundant in EV fractions compared to EV-depleted CSF. Proteins related to exosome biogenesis (PDCD6IP, SDCBP, ARF1, ARF3, and ARF6), tetraspanins (CD81 and CD9), RAB proteins, and heat shock proteins were more abundant in EV fractions compared to EV-depleted CSF (Fig. 3a). Several markers of specific cell types, including neurons (L1CAM, NFASC, and synaptic proteins NPTN, NPTXs, NRXNs, SNAP91, SYN1), astrocytes (ALDH1L1, GFAP, GLUL, S100B), myeloid cells (MMP2, MPO, MRC1), and oligodendrocytes (CRYAB, EPHX2, MBP, PLP1, TNR) were identified mainly in EV fractions (Fig. 3b). Endothelial cell (ICAM, VCAM, VWF), epithelial cell (ATP1A, ATP2B, EZR, LAMA, LAMB, LAMP, CLIC6, GNAI), BBB (AGRN, AQP1, AQP4, DAG1, FBLNs, NIDs), and CP (ACO, ATP5, CALM, CAT, CLIC6, SEPT2) markers were also detected in EV fractions, while a subset of cell-type markers including B2M, CD14, CHI3L1, ENO2, NCAM1, NCAM2, NRCAM, PEA15, PEBP1, CA2, and LAMP2 were detected in both EV fractions and EV-depleted CSF. Many proteins related to inflammation (CRP, GAS6, LTA4H, MSN, PLTP, POSTN, TROVE2), immune responses (ACP1, ANXAs, HLAs, MPO, NCI, NTC5C2, PSMA7), stress responses (GSR, HSP70, HSP90, NAMPT, PRDXs, SNCA, SNCB), and mitochondrial functions (ACOT, DNM1L, DNPEP, GLUD1, RAN, VDAC) were more abundant in CSF EVs compared to EV-depleted CSF (Fig. 3c).

Next, we examined the relationship between the CSF EV proteome and HAND. CSF EV concentrations and number of proteins detected in EV fractions were higher in HIV+ subjects with HAND compared to without HAND (p = 0.004 and p = 0.007, respectively; Fig. 4a, b). In a sensitivity analysis excluding HAD subjects, CSF EV concentrations remained higher in subjects with milder forms of HAND (ANI + MND) compared to without HAND (Additional file 6: Figure S3a). Comparative size distribution histograms of CSF EVs of representative subjects from each HAND subgroup (Additional file 6: Figure S3b) provided further evidence that subjects with ANI, MND, or HAD had higher CSF EV concentrations compared to those without HAND. Although the number of abundant proteins was similar in subjects with and without HAND (range 160–292 and 129–301, respectively) (Fig. 4c, left), the number of proteins not classified as abundant was greater in HAND subjects (range 67–1098) (Fig. 4c, right). Furthermore, in EV fractions, 579 proteins were exclusively detected in subjects with versus without HAND, compared with 43 proteins detected uniquely in subjects without HAND (Fig. 4d). Among 507 proteins detected in both groups, 491 proteins were increased (77 proteins with ≥ 2 fold change and p < 0.05), while 16 were decreased in subjects with HAND compared to those without HAND (Fig. 4e). To evaluate CSF EV profiles in subjects with milder forms of HAND, we reanalyzed the data excluding subjects with HAD. Among 1010 proteins identified by ≥ 2 unique peptides in subjects with mild HAND (ANI + MND) and without HAND, 462 proteins were exclusively detected in subjects with ANI or MND compared to 55 proteins in subjects without HAND. Among 493 proteins detected in ANI + MND as well as non-HAND groups (Fig. 4f), 136 were increased (≥ 2 fold change and low stringency p value < 0.1) of which 20 were significantly altered (≥ 2 fold change and p < 0.05) in subjects with ANI or MND compared to those without HAND (Fig. 4g). Volcano plots with total number of proteins detected in all subjects with and without HAND (n = 1134) or in ANI + MND and no HAND (n = 1010) with ≥ 2 peptides are shown in Additional file 6: Figure S3c. While some proteins including ALDH1A1, ANXAs, CD14, LAMB1, LRP1, MAPK1, MFGE8, NFASC, NPTN, PDCD6IP, PRDXs, SELENBP1, and UBE2N were significantly increased or showed similar trends in subjects with HAND (ANI + MND + HAD) or ANI + MND compared to controls, a subset of proteins including CAPN1, CRMP1, LAMB2, RAB10, and CD163 were significantly increased only in the comparison of ANI + MND vs. controls.

To characterize heterogeneity of proteins mapped to selected cellular components (myeloid cells, astrocytes, neurons, BBB, CP) and biological processes (immune/inflammatory and stress responses) between and within individual HIV+ subjects, we generated supervised heatmaps using the sum of peptide intensities from 101 representative proteins (Fig. 5). Proteins included in this analysis had ≥ 2 peptide counts in CSF EV fractions from 6 or more HIV+ subjects. Eight proteins detected with ≥ 2 peptide counts in < 6 subjects (e.g., MIF, MMP9, NEFL, PEA15, S100B, AQP4, SNCA, SNCB) were also included due to their known importance as markers for respective cell types or biological processes. Most EV proteins including inflammatory/immune response (ANXA, CRP, DPYSL2, ENO1, EZR, HLA, ITGB1, TIMP1), stress response (PARK7, PRDX2, SNCA, SNCB, SNCG, VIM), neuronal (NFASC, NPTN, NRXNs), astrocyte, (ALDH1L1, GFAP, GLUL, PEA15, S100B, SLC1A3) and CP (ATP1A1, ATP1A2, ATP1B1, ATP5B, CLIC6) markers were more abundant in subjects with versus without HAND, a finding that correlated with higher CSF EV concentrations and plasma VL in HAND. However, a subset of neuronal (NCAMs, NRCAM), myeloid cell (CD14, CDH13, CDH2, CHI3L1), and BBB (FBLNs, LAMAs, LAMBs, LAMC1, NIDs) markers were detected in both HAND and non-HAND subjects, largely irrespective of EV concentrations and plasma VL. Unsupervised clustering of these same proteins segregated 9 of 10 HAND subjects (Additional file 6: Figure S4), while one non-HAND subject with high plasma VL and EV concentration clustered with HAND subjects. Several astrocyte markers (ALDH1L1, GFAP, GLUL, S100B, SLC1A3) clustered with stress response (HSP, SOD, PRDX, GSTM, GSTP, GPI, STIP) and inflammation (DPYSL2, ANXA, EZR, ENO1) markers, suggesting inter-relationships between astrocyte-derived EVs, stress responses, and inflammation.

To further investigate the role of astrocyte-derived EVs in stress responses and inflammation in HIV+ subjects with and without HAND, we selected a subset of markers for astrocytes (GFAP, GLUL), stress responses (PARK7, PRDX2, HSP70), and inflammation (CRP) for validation by immunoblotting. For this purpose, we used independent pools of 300 μl CSF from HIV-negative, HIV+ without HAND, and HIV+ with HAND (ANI, MND, and HAD) subjects. CSF EVs from HIV-negative subjects were included as an additional control. Mean EV concentrations from three independent CSF pools are shown in Fig. 6a. Exosome markers CD9 and FLOT-1 were included to detect presence of exosomes in the EV preparations (Fig. 6b). Bands in each lane were normalized to corresponding EV concentrations measured by NTA. Higher levels of CD9 and FLOT-1 were detected in HAND compared to non-HAND and HIV-negative subjects, suggesting higher abundance of exosomes. Increasing trends of PRDX2, PARK7, and HSP70 were detected in CSF EVs of mild (ANI and MND) as well as more severe HAND (HAD) compared to control subjects (Fig. 6b, c). GFAP and GLUL were also detected in CSF EVs, suggesting astrocytes are a potential cellular source of CSF EVs in HIV+ subjects. Furthermore, GFAP and GLUL levels were increased in CSF EVs of subjects with HAND compared to without HAND. These findings suggest that CSF EVs secreted from astrocytes and other cell types in HIV+ subjects carry cargo related to immune/inflammation and stress responses.

To test whether astrocytes are a potential cellular source of EVs carrying inflammation and stress response markers experimentally, mock and H₂O₂-treated astrocytic U87 cells and U87-derived EVs were used to further evaluate astrocyte, stress response, and inflammation markers detected in Fig. 6b and c (GFAP, GLUL, PRDX2, PARK7, CRP). U87 cells were treated with 10, 25, 50, 100, 250, 500, or 1000 μM H₂O₂ for 4 h, which showed a dose-dependent effect in three independent experiments and induced 3% to 58% cell death with increasing doses (Fig. 7a). Intracellular ROS was detected by H₂DCFDA staining of U87 cells following treatment with 50, 100, 250, or 500 μM H₂O₂ for 4 h. Higher ROS production was observed with increasing H₂O₂ concentrations from 50 to 250 μM (Fig. 7b, c). Due to increased cell death, we did not include H₂O₂ concentrations higher than 250 μM in further experiments. Control and H₂O₂-treated (50, 100, 250 μM) cells and EVs were characterized by immunoblotting for astrocyte (GFAP, GLUL), inflammation (CRP), and stress response (PRDX2, PARK7) markers (Fig. 7d). Bands in each lane were normalized to respective GAPDH bands. Increasing trends in expression of PRDX2 and PARK7 were detected in H₂O₂-treated U87 cells and EVs compared to respective controls (Fig. 7e), suggesting that astrocyte-derived EVs can carry these stress response markers under oxidative stress conditions.