Neuron-derived extracellular vesicles reflect adaptive neuronal responses to cortical spreading depolarization: a biomarker study for migraine

Abstract

Background

Small extracellular vesicles (EVs) are nano-sized membranous particles transporting bioactive cargo, including proteins. In the central nervous system (CNS), neuron-derived EVs (nEVs) are thought to play roles in synaptic plasticity, metabolic regulation, and neuroinflammation. While their relevance in neurodegenerative and neuroinflammatory disorders is increasingly recognized, their role in migraine pathophysiology remains underexplored.

Objective

This study aimed to investigate the proteomic signature of nEVs isolated from the cortex of mice subjected to cortical spreading depolarization (CSD), a neurobiological event underlying migraine aura. We sought to identify molecular pathways activated in neurons during CSD and evaluate the potential of nEVs as biomarkers for aura-related brain activity.

Methods

CSD was induced either by pinprick in wild type mice or optogenetically in Thy-ChR2-YFP mice. Following brain perfusion and cortical tissue dissociation, total cortical EVs were isolated by ultracentrifugation whereas nEVs were isolated via immunoaffinity capture targeting neuronal L1 cell adhesion molecule (L1CAM) following nickel-based precipitation of total EVs. nEV proteome was analyzed using label-free quantitative mass spectrometry. Identified proteins were subjected to functional enrichment analysis to uncover relevant biological processes.

Results

Unbiased proteomic profiling revealed CSD-associated changes in pathways involved in transcriptional/translational regulation, cytoskeletal dynamics, stress response and metabolism. These exploratory and descriptive findings suggest that neuronal responses to CSD involve adaptive structural and metabolic alterations and are not limited to inflammatory signaling.

Conclusion

Our results highlight the potential of nEVs as dynamic reporters of cortical neuronal activity in a migraine model. Significant changes in nEV proteome suggest that the neuronal response to CSD extends beyond inflammatory signaling and encompasses adaptive mechanisms aimed at maintaining cellular homeostasis and synaptic integrity. Given their accessibility through peripheral fluids and potential capacity to reflect dynamic changes in neurons, nEVs emerge as promising candidates for investigating pathophysiology and biomarker identification in migraine.

Supplementary Information

The online version contains supplementary material available at 10.1186/s10194-025-02213-x.

Background

Cortical spreading depolarization (CSD), first described by Aristides Leão in 1944, is a wave of electrophysiological silence that propagates slowly across the cortex at approximately 3 mm/min [1, 2]. Building on this discovery, Milner [3] proposed that CSD could underlie the visual aura in migraine, based on the similar propagating rates of visual disturbances and CSD. Decades later, neuroimaging studies using magnetic resonance imaging (MRI) convincingly demonstrated slowly propagating blood oxygen level changes in the occipital cortex during migraine aura, lending support to CSD as the physiological basis of aura [4]. Despite this, direct electrophysiological evidence of CSD during aura in humans remained elusive until it was documented in neurosurgical intensive care patients with coincidental migraine aura using epidural electrodes [5]. However, these findings raised concerns about confounding effects of brain injury. More recently, a landmark case study demonstrated spreading EEG suppression consistent with CSD during migraine aura in a patient with an uninjured brain monitored for epilepsy surgery, thus providing the most direct and uncontaminated evidence to date [6]. Nevertheless, such invasive recordings are limited to rare clinical scenarios, and non-invasive detection of CSD still remains a major challenge. Small extracellular vesicles (EVs), which can cross the blood-brain barrier (BBB) and be isolated from peripheral blood, offer a promising avenue for non-invasive biomarker development [7]. These vesicles may not only help distinguish migraine aura from transient ischemic attacks (TIAs)—a diagnostic dilemma in emergency settings [8, 9]—but also reflect the intracellular biochemical state of neurons and glia during CSD, thereby contributing to our understanding of migraine pathophysiology.

EVs are membrane-enclosed nanoscale particles released by virtually all cell types and may play roles in proteostasis and intercellular communication [10]. They are broadly classified into exosomes (30–150 nm), microvesicles (100–1,000 nm), and apoptotic bodies based on their biogenesis and size [11]. Exosomes, which arise from the endosomal pathway via multivesicular bodies, carry a complex cargo of proteins, lipids, and nucleic acids reflecting the physiological or pathological status of the cell of origin [12, 13]. Their lipid bilayer confers protection to their cargo, enhancing stability in various biofluids and making them particularly appealing as biomarkers that can be obtained non-invasively from peripheral blood [14, 15].

In the central nervous system (CNS), EVs are secreted by all cell types under both homeostatic and pathological conditions. They are thought to contribute to physiological processes such as synaptic transmission, axonal guidance, neurogenesis, and immune modulation in addition to maintaining homeostatic protein levels in a cell [16, 17]. EVs also participate in neuropathological processes, facilitating the spread of misfolded proteins or inflammatory mediators in neurodegenerative diseases and traumatic brain injury [18, 19]. Importantly, their ability to cross the BBB bidirectionally provides a unique opportunity to access molecular signatures of CNS activity via peripheral blood sampling [20, 21]. Recent techniques allow for the immuno-isolation of neuron-derived EVs (nEVs), offering selective insight into neuronal alterations from systemic samples [22].

Despite this growing interest, the utility of EVs in unraveling migraine pathophysiology remains largely unexplored. Migraine is a highly prevalent and disabling neurological disorder, with an estimated global prevalence of 14% and ranking as the second most disabling condition worldwide [23]. Recent clinical work has begun to highlight the potential of circulating exosomal miRNAs and proteins as biomarkers in migraine diagnosis and treatment response. For instance, differential expression of exosomal miRNAs has been observed in patients with migraine without aura, with specific profiles correlating with treatment response to acupuncture [24]. Similarly, plasma-derived exosomes from chronic migraine patients have been proposed to harbor disease-specific proteomic signatures, including alterations in cytoskeletal, metabolic, and signaling proteins [25]. These findings corroborate the hypothesis that EV cargo may reflect and contribute to adaptive processes. However, it should be noted that these studies analyzed total plasma exosomes without neuron-specific enrichment, and current isolation techniques from blood have not been validated to unequivocally determine the cellular origin of circulating EVs.

In the present study, we aimed to characterize the proteomic content of nEVs isolated from the cortical tissue of mice subjected to a well-established migraine model based on CSD induction. Although EVs represent only a selective subset of the cellular proteome, our exploratory and descriptive nEV proteomic profiling indicates that CSD impacts the protein content of neuron-derived EVs, implicating changes associated with cytoskeletal dynamics and transcriptional and translational regulation, albeit indirectly. These data suggest that in addition to the inflammatory signaling [26], neuronal responses to CSD encompass an adaptive program involving structural reorganization and transcriptional reprogramming. Overall, our research provides a novel perspective on how neurons respond to and communicate cellular stress. In future studies the identified CSD-reactive nEV proteome can guide search of nEVs in peripheral blood.

Methods

Animals

Animal housing, care, and application of experimental procedures were all performed in accordance with institutional regulations as approved by the Hacettepe University Animal Experiments Local Ethics Committee (Approval numbers: 2021/24). The experiments were carried out according to the Guide for the Care and Use of Laboratory Animals and reported in accordance with the ARRIVE guidelines. The animals were housed under a 12-h light–12-h dark cycle at a temperature of 22 ± 3 °C and 40–60% humidity and allowed free access to food and water.

Adult male Swiss mice (n = 4 for CSD group, n = 4 for control group) and adult female Thy1-ChR2-YFP mice (n = 6 for CSD group, n = 6 for Sham group), which express the light-activated ion channel channelrhodopsin-2 fused to yellow fluorescent protein under the control of the mouse thymus cell antigen 1 (Thy1) promoter (Jackson Laboratories), were used. Mice were 8–10 weeks old and weighted 25–30 g in both CSD and control groups. One Swiss mouse that was subjected to CSD and one naïve Swiss mouse were used for preliminary mass spectrometry analyses of total EVs derived from the cortex to identify the EV pool that would be the subject of our subsequent experiments. All Thy1-ChR2-YFP mice were utilized for comprehensive mass spectrometry analyses of cortical nEVs isolated after CSD or sham surgery. Prompted by the proteomic results, cortices of three additional Swiss mice from CSD and sham groups were used to investigate ribosomal activity. Swiss mice were preferred over Thy1-ChR2-YFP mice due to the uneven expression of Thy in neurons across cortical layers. Consequently, CSD-induced stress could have differential impacts on neurons, which are individually assessed, one by one, in Nissl-stained cortical sections. Since neurons of wild-type Swiss mice are insensitive to light, pinprick was utilized to induce CSD in these mice.

Mice were randomly selected from breeding cages and assigned to either the CSD or sham condition without a predetermined order, but with the aim of equal distribution between groups. As label-free mass spectrometry and proteomic analyses are inherently unsupervised and unbiased, blinding was not required. The marked loss of nuclear HMGB1 immunoreactivity in neurons after CSD [27–31] interfered with complete blinding for the histological evaluation of Nissl-stained brain sections. Nevertheless, these sections were evaluated by two researchers, with cell counting conducted based on mutual agreement. If CSD could not be reliably induced or verified, the brains of mice were not processed for EV isolation. No EV lysates successfully isolated from any animal, nor any data points obtained from analyses, were excluded from the study.

Mice were anesthetized with urethane (1.25 g/kg, i.p.) for CSD induction through pinprick or with isoflurane (1.5–2%) under continuous oxygen delivery (2 l/min) for optogenetic stimulation and were placed in a stereotaxic frame (Digital Lab Standard Stereotaxic Frame, Stoelting). Body temperature was monitored with a rectal probe and maintained at 37.0 ± 0.2 °C by a homeothermic blanket control unit (Kent Scientific). Pulse rate and oxygen saturation were monitored by an oximeter using a mini Y-clip hind paw probe (The LifeSense^® VET pulse oximeter, Nonin Medical Inc.).

Induction of CSD

In Swiss albino mice, the parietal bone was thinned using a microdrill (Fine Science Tools, USA), and a 1.5-mm burr hole was opened over the frontal region of the right hemisphere (1 mm anterior and 1 mm lateral to bregma). The skull was irrigated with cold saline to prevent complications due to heating caused by the drilling procedure. The exposed dura was kept intact and maintained moist by repeated applications of saline until the experiment started. EEG gel was applied to the electrode tip to enhance electrical contact with bone. A reference electrode was placed between the layers of the neck muscles. Direct Current (DC) potential changes, heart rate, and tissue oxygen saturation were recorded using the Lab Chart data acquisition system (AD Instruments). A single CSD was induced by pinpricking the frontal cortex and verified with the DC potential shift observed over the parietal cortex. Following the induction of CSD, cortex tissues were harvested 1 h later and used for EV isolation.

For optogenetic experiments in Thy1-ChR2-YFP mice, the CSD waves were triggered by optogenetic stimulation (450 nm) delivered by a fiberoptic probe, which was positioned and secured over the skull (the same location where the cortex is pinpricked) by a cable holder as previously described by Houben et al. [32]. No burr hole drilling or thinning was performed in the skull for optogenetic stimulation to ensure minimal invasiveness. The optical fiber, 200 μm in diameter, with a numerical aperture of 0.53 was in full contact with the skull. A suprathreshold light stimulus of 110 mA was continuously applied for 10 s to trigger CSD. The laser light was turned off upon completion of the 10-s stimulus. This stimulation protocol has previously been optimized for our laboratory and reliably results in a CSD for every application. The ignition of a single CSD was confirmed in each mouse by the characteristic triphasic cerebral blood flow response (hypo-, hyper-, and secondary hypoperfusion) observed through the intact skull using laser speckle contrast imaging. Each event lasted approximately one minute, reflecting successful CSD induction in a non-invasive, controlled manner. Following the induction of CSD, cortex tissues were harvested 3 h later and used for nEV isolation (see Fig. 1).

Fig. 1.

Overview of the experimental workflow. Created in BioRender. Sever-Bahcekapili, M. (2025) https://BioRender.com/hntbxxd

Tissue homogenization

Following deep anesthesia, mice were transcardially perfused using a gravity-driven system with ice-cold saline containing 0.4% heparin for 2.5 min, followed by ice-cold PBS for an additional 2.5 min. Under sterile conditions and on ice, cortical tissue was rapidly dissected from the brain and homogenized using gentleMACS tubes and the gentleMACS dissociator in 500 µl of PBS supplemented with a 2x Halt protease and phosphatase inhibitor cocktail (Thermo Scientific™, Cat. No. 78440). The homogenate was transferred to microcentrifuge tube and the remaining tissue residue in the gentleMACS tube was rinsed with an additional 500 µl of PBS, and the wash was pooled with the initial homogenate. The combined sample was centrifuged at 1,500 × g for 10 min at 4 °C to remove debris, and the resulting supernatant was transferred to a clean microcentrifuge tube. To further eliminate cellular contaminants, the supernatant was subjected to a second centrifugation step at 10,000 × g for 30 min at 4 °C, after which the final supernatant—containing EVs—was collected for downstream isolation.

Total EV isolation using ultracentrifugation

For the isolation of EVs, cell debris and cellular contaminant-free brain homogenates were subjected to sequential high-speed ultracentrifugation. The 10,000 × g supernatant was first passed through a 0.22 μm syringe filter to remove residual large particles. Filtered samples were then transferred into ultracentrifuge tubes, carefully balanced with sterile PBS, and centrifuged at 100,000 × g for 3 h at 4 °C using a fixed-angle rotor. After completion, the supernatant was gently removed, leaving a minimal volume to avoid disturbing the EV pellet. Remaining liquid was eliminated by brief inversion and careful aspiration. The pellets were resuspended in 100 µl of filtered PBS. The resulting EV preparations were aliquoted and stored at -80 °C until further use.

Nickel-based isolation of total EVs

EVs were also isolated using the nickel-based isolation (NBI) protocol described by Notarangelo et al. [33], which enables rapid and efficient recovery of EVs from biological samples through electrostatic interactions between positively charged nickel beads and EVs. Briefly, agarose beads were functionalized with nickel cations and incubated with pre-cleared samples (final supernatant—enriched in EVs—following tissue homogenization and clarification by low-speed centrifugation) to capture negatively charged EVs. Following a 50-minute incubation at room temperature under gentle agitation, beads were collected by centrifugation and after washing steps with PBS, EVs were eluted using a physiological pH buffer composed of PBS supplemented with EDTA and citric acid in the presence of NaCl, forming a synergistic chelating solution. This method preserves EV integrity and minimizes contamination by co-isolated proteins, allowing for downstream molecular analyses such as protein and RNA extraction. The isolated EV suspension was supplemented with 1× Halt protease and phosphatase inhibitor cocktail and stored at -80 °C until further use.

Immunocapture of nEVs

The protein concentration of total EV preparations was quantified using the Pierce™ BCA Protein Assay Kit (Thermo Scientific, Cat. No. 23225). For immunocapture of nEVs, carboxylated latex beads (3 μm, 4% w/v) were first conjugated with an anti-L1CAM primary antibody (Miltenyi Biotec, Cat. No. 130-115-812). Beads were washed with Dulbecco’s Phosphate-Buffered Saline (DPBS), incubated with the antibody under gentle rotation at room temperature, and subsequently blocked in 5% Bovine Serum Albumin (BSA) prepared in DPBS. After blocking, beads were resuspended in 1% BSA and the appropriate volume of antibody-conjugated beads was mixed with EV preparations at a ratio of 1 µl beads per 3 µg of EV protein. The mixture was incubated with rotation at room temperature to allow binding of L1CAM-positive nEVs, followed by an overnight incubation in DPBS under gentle agitation. Beads were then pelleted by centrifugation, and nEVs were lysed directly on the beads using RIPA buffer. The lysates were sonicated on ice and incubated further on ice to ensure efficient lysis. After centrifugation, the supernatant containing nEV proteins was collected, and total protein concentration was measured using the Micro BCA™ Protein Assay Kit (Thermo Scientific, Cat. No. 23235). The lysates were stored at − 80 °C for subsequent mass spectrometry analysis.

Sample preparation for proteomics studies

Protein extracts from each sample were transferred to low-binding microcentrifuge tubes. Disulfide bonds were reduced with dithiothreitol (Sigma-Aldrich) and subsequently alkylated with iodoacetamide (Sigma-Aldrich) to prevent reformation. Proteins were then precipitated using a standard methanol/chloroform protocol. The resulting pellets were resuspended in a denaturing buffer containing 8 M urea and 50 mM Tris (pH 8.5). Prior to enzymatic digestion, the urea concentration was diluted to 1 M using 50 mM Tris buffer (pH 8.5). Protein digestion was carried out overnight at room temperature using a Trypsin/Lys-C mix (Promega) at an enzyme-to-substrate ratio of 1:100 (w/w). The digestion was terminated by addition of trifluoroacetic acid to a final concentration of 0.5% (v/v). Peptide mixtures were desalted using Sep-Pak C18 cartridges (Waters) according to the manufacturer’s instructions, dried in a vacuum concentrator (SpeedVac), and reconstituted in liquid chromatography mass spectrometry (LC-MS) grade water (Sigma-Aldrich) to a final concentration of 1 µg/µl.

Mass spectrometry

Peptide samples were analyzed on a Q Exactive Plus mass spectrometer (Thermo Scientific) equipped with an EASY-Spray nano-electrospray ionization source and coupled to an Ultimate 3000 RSLCnano HPLC system (Dionex). Peptides were loaded onto an EASY-Spray C18 analytical column (50 cm × 75 μm ID, 2 μm particle size, 100 Å pore size, Thermo Scientific) and separated using a 150-minute linear gradient from 5% to 95% buffer B (95% acetonitrile, 0.1% formic acid) at a flow rate of 250 nL/min, with buffer A composed of 0.1% formic acid in water. The mass spectrometer was operated in data-dependent acquisition mode. Full MS scans were acquired in the m/z range of 500–2500 at a resolution of 70,000, followed by MS/MS scans of the top most intense ions using higher-energy collisional dissociation with a normalized collision energy of 29 and resolution of 17,500.

Pooling strategy for nEV proteomic analysis

Due to the intrinsically low protein yield obtained from nEVs isolated from a single mouse cortex, sample pooling was implemented prior to LC–MS/MS to ensure sufficient protein input and minimize detection bias toward high-abundance proteins. For each experimental group (CSD or sham), samples with the lowest protein content were randomly paired and combined. Pooling was performed exclusively among samples belonging to the same condition, with no prior selection based on protein profile or experimental outcome. Following pooling, the CSD group comprised three composite samples (each from two biological replicates), whereas the sham group comprised four individual samples that already met the required input threshold and a composite sample from two sham biological replicates. Each pooled sample was processed independently through protein digestion, peptide cleanup, and MS acquisition, and data were analyzed as separate biological replicates within their respective conditions.

Data processing statistical and bioinformatic analysis

Raw mass spectrometry data files were processed using Proteome Discoverer software version 2.2 (Thermo Fisher Scientific, Bremen, Germany) for peptide identification and protein inference. Unless otherwise specified, default search parameters were applied. Briefly, raw spectral data were imported into the software and subjected to database searching against the UniProt mouse reference proteome based on GRCm39 FASTA (UniProtKB mouse sequence databases, Proteome ID UP000000589, 54,791 entries). Peptide precursor mass tolerance was set at 20 ppm, and fragment mass tolerance was set at 0.8 Da. Protein identification was based on the detection of either a minimum of two unique peptides or a single peptide with a false discovery rate (FDR) below 1%.

To ensure data quality and reproducibility, predefined quality control metrics were used to assess the number and reproducibility of identified peptides across samples. The predefined quality control (QC) metrics included (i) peptide and protein identification rates, (ii) the percentage of peptides with a false discovery rate (FDR) < 1%, (iii) biological replicate correlation coefficients, and (iv) label-free quantification (LFQ) intensity reproducibility across samples. These parameters were used to ensure consistent data quality and were based on established proteomic QC standards [34, 35]. No samples were excluded from downstream analyses due to failure to meet the minimum peptide count threshold or insufficient coverage of neuronal peptides.

For protein quantification, label-free quantification was performed using the Data-Dependent Acquisition mode. Peak intensities were extracted and normalized within Proteome Discoverer, and protein-level abundance was determined based on the sum of associated peptide intensities. Peak list files were further mapped to UniProt accession numbers, and UniProt ID mapping was used to integrate and compare identified proteins with previously published proteomic datasets.

Statistical comparisons between groups were conducted to identify differentially expressed proteins. Principal component analysis (PCA) plots were generated using MetaboAnalyst 5.0 (for preliminary data set) or R (v4.4.2) using the prcomp R function (for nEV data set) on normalized protein intensity data from CSD and Sham groups. PCA was mean-centered without additional scaling, and the first two components (PC1 and PC2) were visualized in two dimensions using ggplot2. Data analysis was performed primarily using Perseus v2.0.11 (released September 2023) [36]. Proteins were filtered to retain only those with at least two valid intensity values in both experimental and control groups. Zero values were treated as missing data. Missing values were imputed within Perseus using a random normal distribution (width = 0.3; downshift = 1.8) relative to the minimum observed intensity of each sample. However, imputation was only applied to proteins with measurable intensity values in a sufficient number of replicates within each group. Proteins with missing values across most replicates, for which imputation would introduce artificial signals, were retained as NaN to avoid biasing fold-change estimates or downstream statistical analyses. Intensities were subsequently log₂-transformed and median normalized to correct for sample loading differences. For visualization, the data were standardized by row-wise Z-score transformation (per protein mean centering and scaling).

Our differential expression analysis was treated as exploratory and descriptive, focusing on relative fold-change trends rather than strict inferential testing. Differential protein abundance was assessed by Welch’s t-test between experimental and control groups in Perseus. Significance thresholds were set at |log₂ fold change| ≥ 1 and p < 0.05. Unsupervised hierarchical clustering was performed in R (v4.3.1) using the ComplexHeatmap package. Euclidean distance was used as the similarity metric, and clusters were generated with average linkage. Both protein rows and sample columns were clustered. Color scaling was based on standardized Z-scores, ranging from red (up-regulation) to blue (down-regulation). Heatmaps were exported at 300 dpi in JPEG format for publication quality. Functional enrichment analyses were performed using the web-based platforms PANTHERdb (version 19.0, released 2024-06-20) and EnrichR [37–39] (updated 2023-06-08). For PANTHER, the Statistical Overrepresentation Test was applied to identify enriched Gene Ontology (GO)/PANTHER terms. For EnrichR, enrichment analyses were performed using the ‘GO Biological Process 2023’ and ‘WikiPathways 2024 mouse’ libraries. In all analyses, the background dataset was defined as the total proteome identified in our mass spectrometry dataset, compared against the Mus musculus reference proteome in the web-based platforms. Of note, there are no web-based tools using the neuronal EV proteome as reference database for enrichment analyses. Nevertheless, available tools have been instrumental in disclosing the related protein classes despite the partial representation of the cellular proteome in EVs.

The statistical test used for enrichment was the hypergeometric test as implemented by each platform. P-values were adjusted for multiple testing using the Benjamini–Hochberg false discovery rate (FDR) correction. Enriched terms were considered significant at adjusted p < 0.05 with a minimum gene set size of 4. Analyses were performed separately for each condition (CSD and control) based on the list of proteins significantly altered (up or down regulated) in each group.

Immunofluorescent labelling

Mice were deeply anesthetized and transcardially perfused with 0.04% heparinized saline followed by 4% paraformaldehyde (PFA). Brains were rapidly dissected, post-fixed in 4% PFA overnight at 4 °C, and cryoprotected in 30% sucrose for 48 h. Coronal brain Sections (20 μm thick) were prepared using a cryostat (CM1100, Leica GmbH) and stored at -20 °C until use.

To visualize Nissl substance, NeuroTrace fluorescent Nissl stain (Invitrogen) was prepared by warming the stock solution to room temperature and centrifuging briefly to collect the contents. The stain was diluted 1:200 in Phosphate-Buffered Saline (PBS) and applied to the tissue sections, followed by a 20-minute incubation at room temperature. After removal of the stain, sections were washed for 10 min in wash buffer and then washed twice in PBS for 5 min each. An additional extended wash step was performed in PBS for 2 h at room temperature. Nuclear counterstaining was performed by adding Hoechst dye (~ 60 µL per slide) before mounting coverslips. For immunofluorescence staining, sections from sham and CSD (1-hour post-induction) brains were initially equilibrated at 4 °C for 5 min and then washed with TRIS-Buffered Saline (TBS) at room temperature for 5 min. Permeabilization was performed using TBS containing 0.25% Triton X-100 for 10 min at room temperature, followed by blocking in 10% normal goat serum in TBS for 1 h. Sections were then incubated overnight at 4 °C with a rabbit polyclonal anti-HMGB1 primary antibody (Abcam, ab18256) diluted 1:100 in blocking buffer. After four washes (4 min each) in wash buffer, slides were incubated with Cy3-conjugated goat anti-rabbit IgG secondary antibody (Jackson ImmunoResearch) diluted 1:200 in blocking buffer for 1 h at room temperature, followed by another series of four washes. Fluorescent images were acquired using a laser scanning confocal microscope (SP8, Leica GmbH) equipped with appropriate filter sets.

Results

Changes in the proteome of cortex-derived total EVs caused by CSD

In a preliminary experiment before selectively focusing on nEVs, total EVs were isolated from the cortex of a CSD-induced (by pinprick) and naïve control mouse using ultracentrifugation. Preliminary experiments were conducted on total EVs to establish an initial understanding of the cell groups and proteins primarily impacted by CSD via changes in EV proteome. Although the non-invasive induction of CSD in channel rhodopsin-expressing transgenic mice was the primary approach of our study, we considered the possibility that extra cation channel expression in cortical neurons of transgenic mice might subject these neurons to additional osmotic stress during CSD-induced depolarization. Therefore, we opted for wild-type Swiss mice. Consequently, we utilized pinprick as the trigger, as these mice do not express a light-sensitive channel, and the cortical total EVs were isolated using a straightforward ultracentrifuge-based method one hour after CSD, as no cell-specific separation was required in preliminary experiments. Our decision to adopt a one-hour release schedule was influenced by our previous experience, particularly the early release of EVs after CSD [26, 30].

For label-free quantitative proteomics raw LC-MS/MS data were analyzed in three technical replicates in preliminary experiments using Proteome Discoverer and processed through MetaboAnalyst 5.0 and Perseus. Data were normalized using the interquartile range method and scaled using the Pareto scaling algorithm. To assess overall variation between samples, principal component analysis (PCA) was performed. The PCA plot revealed clear separation between the CSD and control replicates, indicating consistent grouping based on experimental condition (Figure S1). Differential protein expression (DEP) analysis revealed a total of 187 proteins that were significantly altered (22 up-, 165 down-regulated; log2FC ≥ 1, p < 0.05) between the CSD and control groups.

Pathway analysis conducted using PANTHER (Protein ANalysis THrough Evolutionary Relationships) revealed that the DEPs enriched in protein classes associated with translation, RNA metabolism/transcription, and cytoskeleton. However, the false discovery rate (FDR) for these classes was marginally greater than 0.05, suggesting the need for further refinement. Given the exploratory nature of this preliminary data from Swiss mice, we expanded the pool of DEPs by lowering the thresholds to |log₂ fold change| ≥ 0.58 and p < 0.1 to broaden the input protein list and identify emerging functional patterns. For confirmatory analyses and interpretation (Table 2), stringent thresholds of |log₂ fold change| ≥ 1 and p < 0.05 were applied. All DEPs identified were individually listed within the identified PANTHER Protein Classes. Of note, the number of DEPs meeting the thresholds of fold change (FC ≥ 2) and p-value < 0.05 was less than the number of proteins in the “#” column within the corresponding PANTHER Protein Classes. Table 2 also presents a categorized list of 50 DEPs based on their primary functional roles annotated in the UniProt and Human Protein Atlas databases.

Table 2.

List of 50 DEPs identified in cortex-derived EVs 1 h after CSD

Transcription	mRNA processing		Translation	Cytoskeleton
T2FB	FXR2	SSB	RPS26	DC1I1	TBAL3
RALY	TARS3	ATXN2	RPS14	CNTP2	DYNC1LI1
RPC6	DHX30	HNRNPA1	RM38	SPTA1	DYNC1I2
ZZEF1	DDX6	RBM26	FKBP2	MYL6	TUBGCP2
	DDX23	RBM7	EEF1D	SEPTIN8	KIF2A
	CELF2	RPL23	EIF3J1	ARPC3	CKAP5
	REXO1	ALYREF2	EIF2S1	DBN1	PFN1
	SNRPD3	HTATSF1	ABCF2	ACTL6B	MYO1B
	NARS1	TNRC6B	ADARB1
			RPL31
			THOC1
			COPS6

All DEPs (p < 0.05) are up- or downregulated by twofold or greater

In our previous studies, we identified vesicular HMGB1 release from neurons using techniques such as immunohistochemistry, Western blotting, and FACS [26–30]. In the present study, we confirmed the presence of HMGB1 in nEVs isolated from both the CSD group and naïve mouse cortex with LC–MS/MS. Notably, no other cytokines, chemokines, or interferons were detected in more than 3000 unique EV protein identified by LC-MS/MS. This suggests that inflammatory mediators other than HMGB1 are not exported in appreciable amounts with EVs in the cortex after a brief perturbation such as a single CSD unlike several neuroinflammatory conditions [40, 41].

Transcriptomic expression data for the top six upregulated proteins (FXR2, TARS, RPS26, ZZEF1, DHX30, CELF2) were retrieved from mousebrain.org and DropViz databases. These proteins exhibited significantly higher expression in neurons compared to astrocytes, microglia (including activated subsets), and oligodendrocytes, implying that predominantly nEVs serve as the source of these proteins (Figure S2). Additionally, UniProt annotations indicated that these proteins are primarily associated with ribosomal components and protein synthesis functions in line with enrichment analysis (Tables 1 and 2).

Table 1.

PANTHER protein class analysis

PANTHER protein class	Mus musculus (REF) #	#	Expected	Fold enrichment	+/-	raw P value	FDR
Translation Factor	124	8	1.40	5.73	+	8.29E-05	3.26E-03
Translational Protein	344	14	3.88	3.61	+	3.90E-05	2.56E-03
Actin or Actin-binding Cytoskeletal Protein	267	10	3.01	3.32	+	9.42E-04	2.32E-02
Non-receptor Serine/Threonine Protein Kinase	359	12	4.04	2.97	+	8.18E-04	2.30E-02
RNA Metabolism Protein	800	26	9.01	2.88	+	1.37E-06	1.35E-04
Cytoskeletal Protein	617	19	6.95	2.73	+	7.88E-05	3.88E-03

Mus musculus (REF) #: Number of proteins in each protein class in the Mus musculus reference dataset. #: Number of differentially expressed proteins (DEPs) from the present dataset assigned to each class. Expected: Number of DEPs expected in each class by chance, based on the reference proteome distribution. Fold Enrichment: Ratio of observed to expected counts, indicating the magnitude of enrichment. +/–: Direction of enrichment (“+” = overrepresented; “–” = underrepresented) in the CSD group relative to sham control. Raw P value: Unadjusted probability from the PANTHER Statistical Overrepresentation Test. FDR: Benjamini–Hochberg corrected p-value for multiple testing

Proteomic profiling of cortical neuronal EVs after optogenetically induced CSD

Considering the above preliminary findings from total cortical EVs suggesting that neurons might be the primary site of main proteomic alterations, we selectively investigated the changes in the proteome of nEVs induced by CSD. To achieve this, we non-invasively induced CSD in transgenic Thy1-ChR2-YFP mice using optogenetic stimulation, thereby mitigating the adverse effects of experimental injury. Three hours after CSD induction, we isolated nEVs from the cortices. We selected a 3-hour interval following CSD, considering that proteomic alterations may become more prominent at later stages of the post-CSD response compared to those observed 1 h after CSD, as reported in previous studies [31, 42].

Following the precipitation of the total cortical EV population using the NBI method, protein concentrations from all experimental groups were sufficient for L1CAM-based immunoaffinity capture of nEVs. However, after immunocapture, the amount of nEV protein in some samples was insufficient to yield an LC–MS/MS readout with adequate proteomic depth, i.e., a sufficient number of distinct protein identifications. Consequently, the low-protein samples from the same group were combined, resulting in a total of three CSD and five sham samples for LC-MS/MS analysis. From these eight samples, LC-MS/MS identified an average of 901 ± 81 (mean ± SE) unique proteins. Of these, 707 ± 63 proteins (78 ± 0%) were annotated as mouse orthologs of human brain neuronal proteins based on the Ensembl Genome Database, using the biomaRt R package [43, 44]. These results confirm the effectiveness of L1CAM-based immunocapture in selectively isolating nEVs.

Proteomic profiling of nEVs identified a total of 65 differentially expressed proteins (59 up- and 6 down-regulated) compared to sham controls (log2FC ≥ 1, p < 0.05). Pathway enrichment analysis conducted on the upregulated subset of these DEPs using PANTHER Protein Class, GO Biological Process, and the ‘WikiPathways 2024 mouse’ database consistently demonstrated a significant overrepresentation of pathways associated with transcription, mRNA processing, translation, and regulation of the cytoskeleton, complementing the preliminary total EV analysis, which primarily identified downregulated proteins, suggesting related, yet generally opposing, regulatory trends (Table 3). Although the nEV content does not directly correspond to the intracellular protein levels, and may also reflect changes in vesicle release dynamics, cargo sorting or turnover, this observation could indicate the depletion of the relevant proteins at 1 h and their subsequent increased synthesis for replenishment.

Table 3.

PANTHER protein class analysis

PANTHER protein class	Mus musculus (REF) #	#	Expected	Fold Enrichment	+/-	raw P value	FDR
Translation Initiation Factor	92	4	0.28	14.38	+	1.74E-04	1.14E-02
Translation Factor	124	5	0.37	13.34	+	3.69E-05	7.27E-03
Translational Protein	344	7	1.04	6.73	+	7.95E-05	7.83E-03
Cytoskeletal Protein	617	8	1.86	4.29	+	5.27E-04	2.60E-02

Mus musculus (REF) #: Number of proteins in each protein class in the Mus musculus reference dataset. #: Number of differentially expressed proteins (DEPs) from the present dataset assigned to each class. Expected: Number of DEPs expected in each class by chance, based on the reference proteome distribution. Fold Enrichment: Ratio of observed to expected counts, indicating the magnitude of enrichment. +/–: Direction of enrichment (“+” = overrepresented; “–” = underrepresented) in the CSD group relative to sham control. Raw P value: Unadjusted probability from the PANTHER Statistical Overrepresentation Test. FDR: Benjamini–Hochberg corrected p-value for multiple testing

Furthermore, the PANTHER Overrepresentation Test, using the GO Biological Process Complete Annotation Data Set, identified 13 differentially expressed proteins (DEPs) (log2FC ≥ 1, p < 0.05) associated with the cellular stress response and ubiquitination. Table 4 enumerates 50 DEPs grouped into the protein classes identified. The remaining DEPs were dispersed across loosely defined categories, including enzymes (NMT1, MDP1, PRKCSH, CNDP2), proteins involved in endocytosis (CLTC, EPN1), cell adhesion (CHL1, NCAN), acidification of intracellular compartments (ATP6V1C1, ATP6V1A), and others such as gephyrin (Table S1).

Table 4.

List of 50 DEPs identified in nEVs 3 h after optogenetically induced CSD

Transcription	mRNA processing	Translation		Cytoskeleton		Stress response	Ubiquitination
		Ribosomal	ER-Golgi transport
SSBP3	SNRPC	RPL22	TRAPPC3	BAIAP2	GMFB	CLU	RNF181
COMMD10	LUC7L3	EIF3K	TRAPPC4	BRK1	DYNLL1	DNAJA1	SKP1
GTF2F2	U2AF2	EIF2S2*	SAR1B	DCTN3	PAFAH1B1	VSNL1	COMMD9
SRSF5	SNRPE	EIF2B4		DCTN2	RCC2	PARK7	COMMD1
SUB1		RPL23		DCTN1	SEPTIN3	BOLA1	COMMD2
EEF1D				SPTAN1	ADD3	EIF2S2*	COPS8
SALL2				WIPF3	PACSIN1		SGTA
				PAK1	ARPC3
				WASL	NHERF1
				LIN7C

All DEPs (p < 0.05) are up- or downregulated by twofold or greater. *EIF2S2 is a translation initiation protein induced in response to cellular stress

To assess whether DEPs could distinguish CSD from sham samples, we performed unsupervised hierarchical clustering of the 65 proteins identified in nEVs. The heatmap revealed a partial segregation of CSD and sham samples into distinct clusters, demonstrating consistent group-specific proteomic signatures (Fig. 2). Proteins upregulated in CSD clustered together, prominently including cytoskeletal regulators (e.g., ARPC3, WASL, DCTN) and translational initiation factors (EIF2S2, EIF3K, EEF1D), while sham samples grouped with proteins exhibiting lower abundance of these factors. These exploratory findings with a limited number of biological replicates indicate that CSD triggers detectable proteomic shifts in nEV cargo, primarily involving pathways of transcription/translation, cytoskeletal dynamics, and stress adaptation.

Fig. 2.

Unsupervised hierarchical clustering of 65 DEPs in cortical nEVs after CSD. Heatmap depicts Z-score standardized log₂ protein intensities across CSD (n = 3) and sham (n = 5) groups. Both proteins (rows) and samples (columns) were clustered using Euclidean distance and average linkage. Red indicates higher relative abundance; blue indicates lower relative abundance. Dendrograms show partial separation of CSD and sham groups, reflecting consistent group-specific proteomic signatures. Annotated color bar indicates sample grouping (CSD vs. Sham)

As previously reported [30], nEVs released by CSD were isolated and characterized based on established criteria for neuronal extracellular vesicles, including the presence of canonical EV markers. Consistent with previous studies, HMGB1 was detected in these vesicles derived from mouse cortices, supporting their designation as nEVs. Notably, no other cytokines, chemokines, or interferons were identified in the nEVs, suggesting that inflammatory mediators, other than HMGB1, are not significantly exported in nEVs following a single CSD event. These proteomic changes align with a post-CSD adaptive response in neurons, likely reflecting the cellular effort to restore structural integrity and homeostasis following the acute ionic and osmotic disturbance caused by massive depolarization and swelling [45], which necessitate transcriptional activity and protein synthesis.

Histologically validation of the proteomic response

To histologically validate this fundamental proteomic response, the increased ribosomal/translational activity in neurons, coronal brain sections collected one hour after CSD were stained using NeuroTrace™ 500/525 green-fluorescent Nissl stain to label ribosomes [46] and with anti-HMGB1 antibodies to individually identify the stressed neurons that released the pro-inflammatory alarmin, HMGB1. Given consistent evidence that HMGB1 release occurs exclusively from neurons during CSD [27–31], and Nissl primarily stains ribosome/RNA-rich neurons with prominent nucleoli [47], we did not employ an additional label for identifying neurons. Neurons that exhibited translocation of HMGB1 from nucleus to cytoplasm, previously documented to be a trigger for CSD-induced inflammatory signaling [29, 30], exhibited intensified cytoplasmic Nissl staining relative to unaffected neurons with HMGB1-positive nuclei. To assess the association between HMGB1 nuclear immunoreactivity and the presence of cytoplasmic Nissl staining, Fisher’s Exact Test was performed on the pooled data from three mice (364 neurons in somatosensory cortex were counted from 3 mice brains subjected to a single pinprick induced CSD). The analysis revealed a statistically significant difference between the two groups (176 HMGB1-positive vs. 188 HMGB1-negative neuronal nuclei) in terms of cytoplasmic faint Nissl staining (odds ratio = 0.053, p < 0.0001). Specifically, only 20% of HMGB1-positive nuclei exhibited faint cytoplasmic Nissl staining, whereas 83% of HMGB1-negative nuclei showed strong staining. This pattern indicates an elevated ribosomal activity in HMGB1-releasing stressed neurons following CSD, consistent with the nEV proteomic data (Fig. 3).

Fig. 3.

A microscopic image of cortical deep layers rich in neurons from a mouse subjected to CSD and stained with anti-HMGB1 and NeuroTrace™ Nissl stain. Neurons lacking nuclear HMGB1 staining (stars) exhibit strong cytoplasmic Nissl staining in contrast to neurons with HMGB1-positive nuclei but faint cytoplasmic labeling (arrows). The insets show a neuron identified by Hoechst staining and exhibiting strong cytoplasmic Nissl staining. However, it is completely devoid of HMGB1 immunoreactivity. The graph illustrates the proportion of neurons with HMGB1-positive or negative nuclei that exhibit strong cytoplasmic Nissl staining

Discussion

This exploratory and descriptive study provides the first comprehensive proteomic analysis of nEVs after CSD. As hypothesized, but beyond our expectations, EVs do open a highly informative window to the post-CSD events happening in the brain. While bulk ionic changes and biochemical alterations during and after CSD have been well characterized [48], this study marks a concurrent and unbiased examination of the multifaceted cellular responses restoring homeostasis after CSD-induced osmotic and metabolic challenges. These findings reinforce the emerging concept that cellular responses to stress involve complex and concomitant reactions that cannot be reduced to individual molecules or pathways [49]. Notably, although the EV proteome represents the entire cellular changes only partially, a sufficient number of protein alterations that can be categorized within one of the protein classes were identified. For instance, about 38 proteins involved in transcription/translation and cytoskeletal dynamics were found to be significantly different in nEV proteome after CSD, suggesting an altered activity in these cellular pathways. The enrichment of proteins from the same pathways in both nEVs and cortex-derived, non-selected total EV population suggests that these changes are predominantly of neuronal origin, a finding further supported by histological evidence. However, this does not preclude the involvement of non-neuronal cells—particularly astrocytes, which are known to undergo swelling during CSD. Future studies are warranted to investigate the molecular responses to CSD and EV dynamics in non-neuronal cell populations [50].

Although the data obtained 1-hour post-CSD were preliminary, the persistence of changes in the same pathways from 1 to 3 h post-CSD suggests that the restorative processes are ongoing and that these observations are reproducible. Our study was not designed to assess proteomic changes over time, however, subtle alterations in proteins within a pathway from 1 to 3 h could be attributed to shifting protein subtype demand during the repair process. This suggests that proteomic time series may provide unprecedented insights into cellular processes in future studies. Certainly, cell-type-specific proteomics could provide more comprehensive insights than the EV proteome in experimentally investigating the situation. However, EVs offer the advantage of being exploited to observe these processes from the peripheral blood of patients. In this regard, brain-derived nEVs establish a framework for anticipating the characteristics of blood-derived nEVs for future studies although this remains a challenge in terms of selective and efficient isolation of nEVs of brain origin from the periphery. For instance, nEVs obtained from migraine with aura patients might not be informative about inflammatory mediators other than HMGB1 within the hours following a single aura, while they are instrumental for monitoring restorative processes and their variations between attacks, considering factors such as attack frequency, individual patients, and sex. Of note, we observed parallel proteomic changes in female and male mice even though different mouse strains were used.

We observed a notable enrichment of ribosomal proteins and transcription/translation-associated factors in brain-derived nEVs. This suggests that CSD causes an increased protein synthesis, consistent with the transcriptomic upregulation previously reported 1 h post-CSD [26]. Such parallel increases at both transcript and vesicle protein levels reinforce the notion of integrated, multi-layered cellular responses, which may represent a coordinated recovery as well as stress-reporting mechanism following CSD-induced osmotic and metabolic challenges. The cytoskeletal pathway also emerged prominently in our proteome datasets. Cytoskeletal dynamics are critical during CSD, as extensive ionic flux and subsequent neuronal swelling necessitate rapid cytoskeletal reorganization to maintain cell integrity and support spine repair [45, 51, 52]. The detection of numerous cytoskeleton-associated proteins in nEVs suggests that EV-mediated trafficking may help eliminate damaged cytoskeletal components or transmit structural cues to neighboring cells—a mechanism recently proposed in dendritic spine remodeling [53]. In neurons, membrane-associated periodic skeleton remodeling is a highly dynamic process that is driven by the phosphorylation of adducin and the degradation of spectrin [54]. Both of these proteins were altered in nEVs 3 h after CSD. Metabolic adaptation (e.g., NMT1, MDP1, PRKCSH, CNDP2) and oxidative stress response (e.g., DNAJA1, PARK7, BOLA1, EIF2S2) were also notable features within nEV proteome, marked by alterations in metabolic enzymes and stress-resilience proteins in line with investigations of brain tissue, CSF and plasma samples [55–60]. The current knowledge on the biology of EVs does not permit us to draw definitive conclusions. EV content may be impacted by changes in vesicle release dynamics, cargo sorting, or turnover in addition to direct shifts in the overall protein abundance. However, we presume that altered activity in these pathways may facilitate or decrease the export of proteins that can be sorted into EVs (leading to increased or decreased EV protein levels), thereby influencing intracellular protein homeostasis and potentially intercellular communication. These findings are in line with recent migraine patient studies. Zhang et al. [25] observed chronic migraine-associated EV proteomic changes, including metabolic and cytoskeletal proteins. However, the cellular origin of their observations remains unclear as they did not selectively isolate the brain-derived nEVs from patients’ plasma, which make only a small proportion of plasma EVs [61].

In conclusion, this study demonstrates that CSD induces a coordinated neuronal response at least partly reflected in the cargo of nEVs—encompassing not only inflammation, but also cytoskeletal maintenance, metabolic resilience, and regulation of protein synthesis (Fig. 4). By highlighting nEVs as both reporters and potential modulators of neuronal homeostasis, our findings pave the way for novel approaches in biomarker discovery, mechanistic investigation, and EV-based therapeutic strategies for migraine.

Fig. 4.

Pathways Implicated by CSD-induced Differential Protein Expression in Neuronal EVs. Created in BioRender. Sever-Bahcekapili, M. (2025) https://BioRender.com/134nfxu

Conclusions

In this exploratory and descriptive study, we present a comprehensive proteomic characterization of nEVs isolated from cortical tissue of mice subjected to CSD—a fundamental neurophysiological event implicated in migraine pathophysiology. Using high-resolution mass spectrometry and bioinformatic analysis, we identified CSD-associated alterations in the proteomic of nEV cargo, revealing significant changes in proteins involved in cytoskeletal organization, intracellular transport, and protein synthesis. These findings suggest that the neuronal response to CSD extends beyond inflammatory pathways and encompasses broader adaptive mechanisms aimed at maintaining cellular homeostasis and synaptic integrity.

Our results underscore the potential of nEVs as informative reporters of neuronal molecular activity. Given their accessibility through peripheral fluids and their promising capacity to reflect dynamic changes in the neuronal intracellular environment, nEVs emerge as promising candidates for biomarker discovery in migraine, with implications for understanding individual susceptibility to migraine, disease progression, and therapeutic response. These findings lay the groundwork for future efforts to leverage EV-based biomarkers for diagnostic, prognostic, and therapeutic applications in migraine, and underscore the broader relevance of EV biology in CNS disease mechanisms. However, it is important to acknowledge that nEVs offer only a limited window into cellular events. Additionally, findings from murine models may not always be directly applicable to human conditions. This issue is further compounded by the challenge of selectively and efficiently isolating nEVs of brain origin from peripheral blood in sufficient quantities for proteomic analysis. For example, determining whether observed changes in the EV proteome—such as inflammatory signatures in patients with TIA or stroke—originate from the CNS is challenging. This is further complicated by peripheral mechanisms (e.g., systemic inflammation, atherosclerosis), which can confound the interpretation of EV-associated biomarkers and hinder direct comparisons with other CNS-specific perturbations [62–64].

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BBB

Blood-Brain Barrier

BSA

Bovine Serum Albumin

CNS

Central Nervous System

CSD

Cortical Spreading Depolarization

DCL

Direct Current

DEP

Differential Protein Expression

DPBS

Dulbecco’s Phosphate-Buffered Saline

EVs

Extracellular Vesicles

FDR

False Discovery Rate

GO

Gene Ontology

L1CAM

Neuronal L1 cell adhesion molecule

LC-MS

Liquid Chromatography Mass Spectrometry

NBI

Nickel-Based Isolation

nEVs

Neuron-Derived Extracellular Vesicles

PANTHER

Protein ANalysis THrough Evolutionary Relationships

PBS

Phosphate-Buffered Saline

PCA

Principal Component Analysis

PFA

Paraformaldehyde

TBS

TRIS-Buffered Saline

Author contributions

Design and conceptualization of the study: TD, MSB, ŞEE; acquisition and analysis of data: TD, MSB, CCA, BS, ÜG, BO, MŞ, NB; scientific discussions and interpretation of data: TD, MSB, BO, ŞEE, BS; Preparing figures: MSB, TD; drafting the manuscript: MSB, TD. All authors have read and approved the manuscript.

Funding

This work was supported by the Scientific and Technological Research Council of Turkey (TUBITAK) under grant 120C122 and by an unrestricted grant to Turgay Dalkara by the Turkish Academy of Sciences. The research conducted received partial funding from the Ministry of Development-Republic of Türkiye, under project number 2016 K121230. Additionally, Bekir Salih acknowledges the Turkish Academy of Science (TUBA) for their support.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information file. The corresponding author can provide additional data upon reasonable request.

Declarations

Ethics approval and consent to participate

All animal experiments were performed under the guidance of Hacettepe University Animal Experiments Local Ethics Committee (Approval numbers: 2021/24).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.