Penumbra-derived small extracellular vesicles carry pathogenic RNA driving remote damage after cortical infarction

Abstract

Small extracellular vesicles efficiently transport pathological RNAs in ischaemic stroke, but whether penumbral hypoxia-induced extracellular vesicles deliver pathogenic cargo to remote regions, impacting acute ischaemic stroke outcomes, remains unclear.

Using a focal ischaemia mouse model, we isolated brain-derived extracellular vesicles (BDEVs) from the cortical penumbra and determined their pathological impact on synaptically connected remote regions. In vitro, penumbral BDEVs (PEVs) exacerbated recipient neuronal damage, intensifying apoptosis and dendritic injury during oxygen-glucose deprivation/reoxygenation. In vivo tracing confirmed PEV transport via anatomical projections to remote thalamic neurons, where internalization triggered synaptic loss and apoptosis in distal thalamic nuclei.

Mechanistically, PEVs delivered the penumbra-specific circular RNA CircOGDH through trans-synaptic delivery to connected thalamic neurons, directly inducing synaptic and neuronal injury. Critically, cortical CircOGDH knockdown abolished thalamic CircOGDH accumulation and reversed neuronal loss and synaptic impairment, confirming its causal role in secondary damage. Transcriptomics further revealed PEV enrichment in RNAs dysregulating synaptic and axonal pathways.

Thus, we open a new view PEVs initiated a thalamus pathogenic pathway where BDEVs deliver CircOGDH and functional RNAs to drive non-infarcted region degeneration. This redefines remote post-stroke injury as an active RNA trafficking process, highlighting therapeutic opportunities for intercepting extracellular vesicle-mediated trans-synaptic pathology.

In a focal ischaemia mouse model, Deng et al. show that penumbra-derived brain extracellular vesicles travel trans-synaptically to distant thalamic neurons after cortical infarction. By delivering CircOGDH and other pathogenic RNAs, they trigger apoptosis and synaptic loss, driving secondary degeneration beyond the infarct.

Introduction

Ischaemic stroke exhibits high incidence and disability rates,^1,2 with patient prognosis critically dependent on the status of the ischaemic penumbra, a potentially salvageable brain region representing a key neuroprotective target.³ However, without timely reperfusion, ischaemia triggers deleterious molecular cascades that not only accelerate penumbra progression to the infarct core but also mediate secondary damage to remote areas. In both clinical middle cerebral artery (MCA) infarction and animal models of cortical infarction, the thalamus serves as a primary affected site.^4-7 Anterograde axonal degeneration, neuroinflammatory cascades and apoptosis-autophagy imbalance drive progressive neuronal damage in the thalamus, culminating in delayed neuronal death. Although clinical and animal studies confirm that disrupted corticothalamic neurotransmitter exchange induces cross-regional pathologies, and thalamic degeneration directly correlates with worsened functional outcomes, current mechanistic research predominantly focuses on axonal degeneration and inflammatory pathways. However, the role of small extracellular vesicles (SEVs) in mediating pathogenic molecule trafficking from the cortical ischaemic penumbra to thalamic remote damage remains unexplored.

Typically 30–150 nm in diameter, SEVs are secreted by nearly all cell types and serve as critical mediators of intercellular communication.⁸ Within the CNS, accumulating evidence demonstrates that SEVs facilitate not only local effects but also the long-distance propagation of pathogenic molecules. For instance, in Parkinson's disease, microinjection of plasma-derived SEVs from patients into striatum promotes the transfer of α-synuclein cargo to the anatomically connected substantia nigra in mice.⁹ Similarly, in Alzheimer’s disease, neuron-derived SEVs mediate trans-synaptic tau transmission through synaptically linked pathways, and brain-derived EVs (BDEVs) isolated directly from patient tissues propagate tau along hippocampal dentate gyrus projection.¹⁰ These findings collectively underscore the role of SEVs in pathology dependent on anatomical connectivity. However, in ischaemic stroke, research on SEVs has predominantly focused on local intercellular communication, often relying on cell culture supernatant-derived models.¹¹ Therefore, whether SEVs released in vivo from the ischaemic penumbra propagate along connected pathways to promote secondary damage in remote regions like the thalamus remains largely unexplored.

Small EVs are enriched in diverse RNA species, including miRNAs, circRNAs and mRNAs, which orchestrate post-stroke pathological cascades.^12-15 SEV RNAs derived from injured neurons or other brain cells modulate key pathological processes including neuroinflammation,¹⁶ blood–brain barrier disruption¹⁷ and neuronal apoptosis.¹⁸ These RNAs use distinct molecular mechanisms where miRNAs silence target genes,¹⁹ circRNAs sponge regulatory factors²⁰ and mRNAs template protein synthesis.²¹ For instance, our prior work identified CircOGDH derived from the ischaemic penumbra as a critical mediator, where knockdown of CircOGDH attenuated neuronal apoptosis in the penumbra, revealing circRNA-mediated dysregulation of neuronal function after ischaemia.^18,22 However, the role of SEV-derived RNAs specifically from ischaemic penumbral tissue is poorly understood.

In this study, we used an optimized purification method to isolate BDEVs from ischaemic penumbral brain tissue and completed their characterization. Subsequently, we demonstrated the anterograde transmission of penumbra BDEVs (PEVs) from the ischaemic cortex to synaptically connected thalamic region. Importantly, CircOGDH was found to disrupt the corticothalamic pathway by promoting non-ischaemia neuronal apoptosis. Furthermore, transcriptomic profiling of PEVs revealed potential target genes including suppressing axon regeneration-related genes [e.g. Sema3a (semaphorin 3a) and Robo2 (roundabout guidance receptor 2)] contributing to secondary damage.²³ Collectively, our findings suggest that pathological propagation mediated by PEVs is an available therapeutic target, and early intervention to block harmful vesicular cargo may help prevent degeneration of remote brain regions and promote recovery after ischaemic stroke.

Materials and methods

Animals

A total of 253 adult male and female C57BL/6J mice (22.0–25.5 g, 6 to 8 weeks old) were purchased from Zhuhai Bes Test Bio-Tech Co, Ltd. Mice were assigned to specific experimental groups without bias using a random number generator (GraphPad QuickCalcs), and no mice were excluded from the study. Mice were housed in a quiet, well-ventilated facility with ad libitum access to food and water. Environmental conditions were maintained at 25.0°C ± 1.0°C, 65.0% ± 5.0% humidity, and a 12-h light/dark cycle. All procedures were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals (NIH Publication No. 8023, revised 1978) and approved by the Experimental Animal Ethics Committee of Jinan University (approval ID: 20220303-56).

Animal model

Adult C57BL/6J mice were anaesthetized with isoflurane (4% for induction, 1.5% for maintenance; RWD Life Science, Cat. No. R510-22-16). Permanent occlusion of the distal branch of the left MCA was induced by monopolar electrocoagulation to establish a focal cortical infarction model. Briefly, a small temporal craniotomy (∼2 × 2 mm) was made above the zygomatic arch to expose the distal MCA. Under a surgical microscope, the main trunk and distal bifurcation of the MCA were identified, and the artery was coagulated both proximal and distal to the bifurcation using a monopolar electrocoagulator (0.5-mm tip; 2–3 W; 2–3 s per application) until vessel blanching and complete cessation of blood flow were observed. Sham-operated mice underwent identical procedures without vessel occlusion. Cerebral blood flow was monitored using laser speckle contrast imaging (PeriCam PSI system, Perimed AB) according to the manufacturer’s instructions, and successful MCA occlusion (MCAO) was confirmed by whitening of the cortical territory supplied by the distal MCA available, loss of arterial pulsation without reperfusion and a reduction in cortical blood flow. Postoperative mice were allowed to recover on a 37°C heating pad before being returned to their home cages. In a subset of animals, MRI was performed 3 h after MCAO, and the cortical ischaemic penumbra was defined by the diffusion-weighted imaging–FLAIR mismatch.¹⁸

Primary neuron culture and treatment

Primary cortical neurons were cultured as described previously.²⁴ Cortical tissues from embryonic Day 18–19 (E18–19) mouse embryos were digested with trypsin and dissociated into single-cell suspensions. Neurons were initially plated in DMEM/F12 containing 10% foetal bovine serum for 4 h and then maintained in a neurobasal medium (Gibco, Cat. No. 21103049) supplemented with 2% B27 (Gibco, Cat. No. 177504044) at 37°C in a 5% CO₂ incubator. Neurons were used for subsequent experiments after 5 days in vitro.

Oxygen-glucose deprivation/reoxygenation (OGD/R) was induced by replacing the culture medium with glucose-free DMEM following two PBS washes and incubating cells in a hypoxia chamber (95% N₂, 5% CO₂, 37°C) for 3 h. For OGD/R injury, cultures were replenished with Neurobasal medium and incubated under normoxic conditions for 3 h. Cells were divided into four groups: Control, OGD/R+PBS, OGD/R+PEVs and OGD/R+SEVs. At reperfusion onset, 40 μg each of PEVs or sEVs were added to the respective groups.

Neurons under normoxic culture conditions were maintained in complete Neurobasal medium in a standard humidified incubator (37°C, 5% CO₂). Primary cortical neurons were incubated with BDEVs for 48 h. Cells were assigned to three groups: PBS, SEVs and PEVs. For EV treatment, 40 µg of sEVs or PEVs were added directly to the culture medium, while the PBS group received an equivalent volume of vehicle.

Isolation of BDEVs

A focal cortical infarction model was established as described above. At 3 h post-infarction, mice were euthanized and perfused transcardially with saline. Brains were quickly dissected and placed in Hibernate-E buffer (Thermo Fisher Scientific, Cat. No. A1247601) at 4°C. Cortical penumbra, sham and thalamic tissues were microdissected on ice using fine forceps and micro-scissors under a stereomicroscope (Leica MZ16). Under aseptic conditions, these tissues were sectioned into 2 × 2 × 2 mm cubes following standard sterile techniques.

As shown in Supplementary Fig. 1A, BDEVs were isolated via enzymatic digestion and sucrose density gradient ultracentrifugation (SDGU). Tissue blocks were weighed and incubated in Hibernate-E buffer containing 0.5 mg/ml collagenase type 3 (Worthington, Cat. No. LS004180) and 20 U/ml papain (Worthington, Cat. No. LK003178) at 37°C in a shaking water bath for 20 min. During digestion, tissue slices were inverted every 10 min before continuing incubation for another 10 min. The reaction was terminated by adding ice-cold inhibition buffer prepared from Hibernate-E medium supplemented with 1× phosphatase inhibitor cocktails (Beyotime, Cat. No. P1081/P1086) and 1× protease inhibitor cocktail (Sigma-Aldrich, Cat. No. P2714).²⁵

Crude BDEVs were isolated by sequential centrifugation at 300g and 2000g to remove debris and apoptotic bodies, followed by 10 000g to pellet large EVs. The supernatant was filtered (0.22 μm) and ultracentrifuged at 160 000g to obtain the crude BDEVs pellet. For purification, the crude BDEVs pellet was subjected to SDGU at 160 000g for 3 h using a discontinuous gradient of 0.88 M (density 1.13 g/ml) 30% sucrose and 1.18 M (density 1.18 g/ml) 40% sucrose layers. Post-centrifugation, five discrete fractions were sequentially collected: Fraction 1 (PBS layer), Fraction 2 (30% sucrose layer), Fraction 3 (30%/40% sucrose interface), Fraction 4 (40% sucrose layer) and Fraction 5 (pellet), each of which was subsequently washed via centrifugation in ice-cold Dulbecco’s PBS at 160 000g for 3 h. Transmission electron microscopy (TEM) and western blot analyses showed peak BDEVs enrichment in Fraction 2 (Supplementary Fig. 1B and C), leading to a simplified method using a single-step 30% sucrose cushion.

In the optimized protocol, digested brain suspensions underwent centrifugation at 300g, 2000g and 10 000g (4°C), followed by 0.22 μm filtration. The supernatant was ultracentrifuged at 160 000g (4°C, 3 h) to pellet EVs. This pellet was resuspended in PBS, overlaid on a 30% sucrose cushion, and ultracentrifuged again at 160 000g for 2 h. The final BDEV pellet was resuspended in PBS (Fig. 1B). All isolates met the MISEV 2023 minimal criteria for EV characterization.

Figure 1.

Penumbra-derived small extracellular vesicles (PEVs) exacerbate neuronal injury in vitro. (A) Schematic diagram of ultracentrifugation with the one-step sucrose cushion for brain-derived small extracellular vesicle (BDEV) isolation. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/fx5jkn6. (B) Western blot analysis of BDEV markers TSG101, HSP70 and CD81, and small extracellular vesicles (SEVs) negative marker calnexin in BDEVs isolated from cell lysates and brain tissues. (C) Typical transmission electron microscopy images of SEVs and PEVs. Scale bar = 200 nm. (D) Nanoparticle tracking analysis (NTA) showing size distribution of BDEVs isolated from Sham and permanent occlusion of the distal branch of the left middle cerebral artery (MCAO) 3 h mice. (E) NTA quantification of BDEVs concentration (particles/mg), with PEVs having significantly higher concentration than Sham-BDEVs. Data are presented as mean ± standard deviation (SD), n = 6, Student’s two-sample t-test. (F) Experimental scheme for co-culturing primary cortical neurons with BDEVs. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/5otio9p. (G) Oxygen-glucose deprivation/reoxygenation (OGD/R) induced primary cortical neurons treated with PBS, PEVs, or BDEVs derived from sham mice brain tissue (SEVs) in vitro. Top row: TUNEL (red) staining for apoptosis. Scale bar = 50 μm. Bottom row: Microtubule-associated protein 2 (MAP2) staining (red) for neuronal morphology (scale bar = 50 μm); insets show single neuron tracing using the NeuronJ plugin. Scale bar = 50 μm. (H) Quantification of the dendritic complexity by Sholl analysis. Intersections were used to quantify the MAP2-stained neural complexity in G, bottom row. Data are expressed as mean ± SD, n = 18 neurons per group from three experiments, one-way ANOVA test. (I) Quantitative analysis of TUNEL⁺/DAPI in G, top row. Data are presented as mean ± SD, n = 6 neurons per group, one-way ANOVA test. (J) Quantification of total neurite length in G, bottom row. Data are expressed as mean ± SD, n = 18 neurons per group from three experiments, one-way ANOVA test. P < 0.05 was considered significant. N.S., no significant.

BDEV characterization

For TEM, 10 μl of BDEV suspension was placed on carbon-coated copper grids and air-dried for 10 min. Grids were fixed in 2.5% glutaraldehyde for 5 min, washed in PBS, and negatively stained with 2% phosphotungstic acid (Solarbio, Cat. No. G1871) for 2 min. After drying, BDEVs were visualized using a Hitachi HT-7800 TEM at 80–120 kV.

Nanoparticle tracking analysis (NTA) was performed using the NanoSight NS300 system (NanoSight Technology) to quantify purified vesicles. Three 60-s videos were recorded at 30 frames per second (fps) per sample.²⁶

Western blot analysis was conducted to characterize EVs: harvested BDEVs were denatured in SDS loading buffer, separated by electrophoresis, and probed with antibodies against EV markers: CD81 (Proteintech, Cat. No. 27855-1-AP), HSP70 (Proteintech, Cat. No. 10995-1-AP), and TSG101 (Proteintech, Cat. No. 28283-1-AP). Calnexin (Proteintech, Cat. No. 10427-2-AP) was used as a negative control.

Extracellular vesicle labelling and uptake

BDEVs were incubated in PBS containing CellMask™ (1×; Invitrogen/Thermo Fisher, Cat. No. C10046) and DIO (1 µm; MCE, Cat. No. HY-D0969) at 37°C for 10 min. Excess dye was removed by ultracentrifugation at 160 000g (4°C, 70 min), and the pellet was resuspended in PBS. Immediately after MCAO, CellMask-labelled BDEVs were microinjected stereotaxically into the ipsilateral cortex or thalamus, DIO-labelled BDEVs were microinjected stereotaxically into the ipsilateral thalamus. At 3 h post-injection, mice were euthanized and brains were collected for cryosectioning. Distribution of EVs was visualized using Leica STELLARIS 5 confocal microscope.

Extracellular vesicle depletion

Systemic inhibition of SEV secretion was achieved by intraperitoneal injection of GW4869 (1.25 mg/kg; MCE, Cat. No. HY-19363) or vehicle 1 h before model induction. For cortical-specific suppression, 2 μl of GW4869 (20 nM; MCE, Cat. No. HY-19363) or vehicle was injected stereotaxically into the left cortex of C57BL/6J mice [coordinates: anterior-posterior (AP) −0.2 mm, medial-lateral (ML) +2.5 mm, dorsal-ventral (DL) −1.5 mm] 1 h before MCAO. Brains were harvested 3 h after modelling, and EV secretion was quantified using NTA.

For thalamic-specific suppression, 300 nl of GW4869 (20 nM; MCE, Cat. No. HY-19363) or vehicle was injected stereotaxically into the ipsilateral thalamus of C57BL/6J mice (coordinates: AP −1.7 mm, ML +1.6 mm, DV −3.6 mm) 1 h before MCAO.

Staining

Tissue processing

For immunofluorescence, mice were perfused transcardially with 0.9% saline followed by 4% paraformaldehyde (PFA). Brains were dissected, post-fixed in PFA for 24 h, then cryoprotected sequentially in 20% and 30% sucrose (24 h each) at 4°C.

Frozen sections were prepared at different thicknesses depending on experimental needs: 40 μm for recombinant AAV (rAAV)-transfected corticothalamic tract analysis and 10 μm for other immunofluorescence applications.

Nissl staining and TUNEL staining

For Nissl staining, sections were fixed in 4% PFA (15 min), washed with PBS (3 × 5 min), permeabilized with 0.1% Triton X-100 (10 min), and stained with Nissl solution (1:200; Thermo Fisher, Cat. No. N21483) at room temperature for 30 min. After PBS rinses, sections were counterstained with DAPI (Beyotime, Cat. No. C1006), mounted with antifade mounting medium (Beyotime, Cat. No. P0126).

TUNEL staining for apoptosis detection was performed using a commercial kit (Beyotime, Cat. No. C1089) according to the manufacturer's protocol. Sections were fixed, permeabilized, incubated with TUNEL solution at 37°C in the dark for 1 h, then counterstained with DAPI.

Immunohistochemistry

Cells or frozen brain sections were fixed with 4% PFA for 15 min, permeabilized in PBS containing 0.1% Triton X-100 for 10 min, and subsequently blocked with 5% bovine serum albumin (BSA; Biofroxx, Cat. No. 9048-46-8) for 60 min at room temperature. Without intermediate washing, the samples were incubated overnight at 4°C with primary antibodies against MAP2 (Abcam, Cat. No. ab32454), PSD95 (postsynaptic density protein 95; Proteintech, Cat. No. 30255-1-AP), and synaptophysin (SYP; Proteintech, Cat. No. 17785-1-AP). The next day, samples were incubated with a mixture of fluorescent secondary antibodies for 1 h at room temperature. Fluorescence images were acquired using a Leica STELLARIS 5 confocal microscope.

Golgi staining

Golgi staining was performed using a modified Golgi-Cox impregnation method. Mice were transcardially perfused with 4% PFA, after which the brains were removed and post-fixed in 4% PFA for 24 h. The fixed brains were then sectioned into 2–3-mm-thick tissue blocks, immersed in Golgi dye solution (Servicebio) and stored in the dark for 14 days with the dye replaced every 3 days. Following staining, the tissue blocks were dehydrated in sucrose and softened in 75% glacial acetic acid. Coronal brain sections (100-μm thick) were prepared using a cryostat, then air-dried overnight. Sections were rinsed in distilled water, incubated in concentrated ammonia water for 10 min, washed again with distilled water and finally mounted using glycerine gelatine for imaging.

Neuronal structure complexity

Neurons labelled with MAP2 or visualized through Golgi staining were traced using the NeuronJ plugin in ImageJ. These tracings were used to quantify the number and length of neuronal processes. Neuronal complexity was assessed using Sholl plugin on the tracings.²⁷

Synaptic density

Fluorescence images were converted to 8-bit greyscale using ImageJ. The regions corresponding to PSD95 and SYP expression were identified using the Threshold function and regions of interest were added to the ROI Manager. The overlapping area between PSD95 and SYP signals was calculated to quantify synaptic co-localization.

Fluorescence in situ hybridization for brain tissues

RNA fluorescence in situ hybridization (FISH) was conducted on brain tissue using the RNA FISH Kit for frozen sections (GenePharma). Sections were treated with proteinase K at 37°C, dehydrated through an ethanol gradient, and denatured using a preheated solution at 78°C. After re-dehydration, hybridization buffer containing a CY3-labelled CircOGDH probe (5′–3′: AAUGU+CCCAUGACAAGU+UUGUCUGU+GGAUGA) was applied, and hybridization was performed at 37°C for 12–16 h in a humidified chamber. Post-hybridization washes were performed using preheated washing solution at 43°C, followed by DAPI nuclear staining. Slides were mounted with either glycerine or antifade reagent and imaged using a Leica STELLARIS 5 confocal microscope.

Intracranial injection

Mice were anaesthetized as previously described and positioned on a stereotaxic frame (RWD). Erythromycin eye ointment was applied to prevent corneal dryness. A small craniotomy was made using a dental drill, and a microinjection pump (RWD) was used to deliver injections at a slow flow rate to minimize local tissue damage.

To label the corticothalamic tract, 300 nl of rAAV2/9-hSyn-mCherry (2 × 10¹² viral genomes/μl; BrainVTA Technology) was injected into the left cortex of C57BL/6J mice using the following stereotaxic coordinates: AP −0.2 mm, lateral +2.5 mm, and ventral −1.5 mm.

To knockdown CircOGDH in the left peri-infarct cortex with shRNA constructs, adult male C57BL/6J mice (22.0–25.5 g, 6 to 8 weeks of age) were microinjected, respectively, with ADV-GFP-shRNA-CircCtrl (3 µl of 1 × 10¹⁰ viral genomes/µl, Gene Pharma) or ADV-GFP-shRNA-CircOGDH virus (3 µl of 1 × 10¹⁰ viral genomes/µl, Gene Pharma) at coordinates: anteroposterior −0.2 mm, lateral +2.0 (or +3.0) mm, and ventral −1.5 (or −2.0) mm. Six days after the microinjection of adenovirus, mice were divided into three groups: Sham+Sh_CircCtrl (n = 6); MCAO+Sh_CircCtrl (n = 6); MCAO+sh_CircOGDH (n = 6).

Stereotaxic injection of GW4869 is described earlier in the ‘Extracellular vesicle depletion’ section. To assess the effect of exogenous PEN-BDEVs, 300 nl of BDEV solution (1 μg/μl) was injected stereotaxically into the ipsilateral thalamus (AP −1.7 mm, ML +1.6 mm, DV −3.6 mm) immediately following MCAO induction. Mice were divided into four groups: (i) Sham (n = 6); (ii) MCAO (n = 6); (iii) MCAO+GW4869 (n = 6); and (iv) MCAO+GW4869+PEVs (n = 6). Brains were collected 14 days after injection.

To determine whether PEVs alone are sufficient to induce thalamic injury in the absence of ischaemia, we performed additional studies in sham mice. A total volume of 3 μl BDEVs (1 μg/μl) was injected stereotaxically into the left cortex (AP −0.2 mm, ML +2.5 mm, DV −1.5 mm). Mice were divided into three groups: (i) PBS (n = 6); (ii) SEVs (n = 6); and (iii) PEVs (n = 6). Tissue was collected 14 days after injection.

RNA extraction and real-time quantitative PCR

Total RNA was isolated from cells and brain tissues using TRIzol Reagent (Thermo Fisher Scientific, Cat. No.15596-026). Reverse transcription was carried out using PrimeScript RT Master Mix (Takara, Cat. No. #RR047A), and quantitative PCR was performed using LightCycler® 480 SYBR Green I Master (Roche, Cat. No. 04887352001). Primer sequences are provided in Supplementary Table 1. Relative expression levels were calculated using the 2^−ΔΔCT method.²⁸

RNA sequencing

Brain tissues from the infarct core and ischaemic penumbra were harvested 3 h after MCAO surgery, while corresponding tissues from age-matched sham-operated mice served as controls(n = 3 per group). SEVs and PEVs were isolated using ultracentrifugation combined with a one-step sucrose gradient method. Total RNA was extracted from the SEVs and PEVs, and RNA integrity and concentration were evaluated using an Agilent Bioanalyzer. For mRNA analysis, cDNA libraries were constructed following mRNA enrichment with oligo(dT) magnetic beads. For miRNA analysis, small RNA fragments (18–30 nt) were isolated and used to generate small RNA libraries. High-throughput sequencing was performed on the Illumina NovaSeq platform (paired-end, 150 bp read length) (Echo Biotech). Raw reads underwent quality control using FastQC, followed by alignment to the mouse reference genome (mm10) using HISAT2 for mRNA or Bowtie for miRNA. Gene expression levels were quantified using HTSeq-count, with mRNA expressed as fragments per kilobase per million (FPKM) and miRNA as transcripts per million (TPM).

Bioinformatics analysis

Differential expression analysis was performed using DESeq2 with thresholds of |log₂FC| > 1 and FDR < 0.05. Target genes of differentially expressed miRNAs were predicted by intersecting results from miRDB (v6.0; prediction score ≥80) and TargetScan, retaining only consensus targets. To elucidate biological processes and pathways associated with differentially expressed mRNAs (DEmRNAs) and miRNA target genes (from DEmiRNAs) in penumbra versus Sham BDEVs, functional enrichment analysis was performed. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses were conducted using the clusterProfiler R package, with gene identifier conversion facilitated by the org.Mm.eg.db package. Statistical significance was defined as an adjusted P-value < 0.05. Results were visualized through bubble plots and pathway networks generated with ggplot2. Protein-protein interaction (PPI) network analysis was performed using the STRING database (v12.0; https://string-db.org/). Resultant networks were visualized and analyzed with Cytoscape software (v3.10.0).

Statistical analysis

Statistical analyses were conducted using GraphPad Prism 8.0 (GraphPad Software, Inc., La Jolla, CA, USA) and SPSS (Windows version 27.0; SPSS Inc., Chicago, IL, USA). Student’s two-sample t-test was performed to compare values between two groups. Multiple group comparisons were analysed using one-way ANOVA test. Data are presented as mean ± standard deviation (SD), and statistical significance was defined as P < 0.05. Representative images shown in the figures were selected to illustrate characteristic experimental results in an unbiased manner.

Results

BDEVs derived from the cortical penumbra exacerbate neuronal injury in vitro

To observe the contribution of PEVs to stroke progression, we isolated BDEVs from the penumbra of mice in MCAO of focal cortical infarction (Fig. 1A and Supplementary Figs 2 and 3). As a control, BDEVs were also isolated from the cortical brain tissue of sham-operated mice (SEVs). Initially, we used the traditional sucrose gradient method to isolate BDEVs, a procedure exceeding 30 h. BDEVs were mainly concentrated in the 30% sucrose fraction (Supplementary Fig. 1). Based on this observation, we optimized the protocol using a single sucrose cushion and reduced the centrifugation time, which enhanced the efficiency of BDEVs purification (Fig. 1A). The BDEVs conformed to established SEV criteria, exhibiting enrichment of canonical protein markers (Fig. 1B), a characteristic cup-shaped structure (Fig. 1C), and particle sizes ranging from 50 to 200 nm (Fig. 1D). Additionally, the secretion of PEVs was significantly more upregulated under hypoxic conditions compared to that of SEVs (Fig. 1D and E). To identify the primary cellular source of EVs in the ischaemic penumbra 3 h after MCAO, we first analysed cortical composition. Immunofluorescence revealed neurons (NeuN⁺) as the overwhelmingly dominant cell type, with astrocytes and microglia cells present at much lower densities. Subsequent western blot analysis of EVs isolated from the penumbra using cell-type-specific markers showed a marked enrichment of neuronal markers (SYN1, SNAP25). In contrast, astrocytic (EAAT2) and microglial (P2Y12, TMEM119) markers were either undetectable or unchanged. These consistent results demonstrate that neurons are the predominant source of PEVs (Supplementary Figs 4 and 5).

To evaluate their functional impact on recipient penumbra neurons, we used an in vitro model of OGD/R, which mimics ischaemic stroke conditions. Neurons subjected to OGD 3 h and were treated with BDEVs during the reoxygenation phase for 3 h (Fig. 1F). Intriguingly, PEV treatment significantly increased neuronal apoptosis, as determined by TUNEL assay (Fig. 1G and I), but showed no increase in the normoxia group (Supplementary Fig. 6A and B). Consistently, morphological analysis revealed that PEVs also aggravated structural damage, evidenced by reduced dendritic length and complexity (Fig. 1G, H and J), but showed no significant changes in the normoxia group (Supplementary Fig. 6A, C and D). SEVs did not exacerbate apoptosis or impair dendritic growth and branching relative to the OGD/R group without BDEVs, suggesting that BDEVs originating from normal tissue do not exert detrimental effects on ischaemic neurons. These results demonstrate that PEVs directly exacerbate neuronal injury under hypoxic conditions.

PEVs from cortical penumbra propagate synaptic injury of the thalamus after cortical infarction

BDEVs, which act as functional carriers of bioactive molecules, contribute to intercellular communication and play a crucial role in various pathological process of ischaemic stroke. To investigate whether PEVs originating from the cortex may be transported to brain regions, thereby mediating remote injury, electrocoagulation of the MCA was performed to induce a localized cortical infarction without direct thalamic damage (Supplementary Fig. 2A and B). To trace axonal projections to distant regions, we used anterograde non-trans-synaptic viral tracers to label downstream neuronal targets.

To determine whether remote non-ischaemic regions connected to the peri-infarct cortex were affected following ischaemic injury, we injected the anterograde tracer rAAV2/9-hSyn-mCherry-WPRE-hGH-polyA (rAAV) into the left cortex 21 days before MCAO and subsequently injected magenta-labelled PEVs into the penumbra region. The transport and distribution of PEVs to remote areas were assessed 3 h after infarction (Fig. 2A). As shown in Fig. 2B and C, rAAV2/9-mCherry fluorescence was detected in both the peri-infarct cortex and the ipsilateral non-ischaemic thalamus 3 h after MCAO, indicating the presence of structural connectivity between the peri-infarct cortex and the ipsilateral thalamus, consistent with previous studies. We next investigated whether PEVs could be transported along corticothalamic projections from the cortex to the thalamus. Immediately after MCAO, magenta-labelled PEVs were injected stereotaxically into the same cortical site previously infused with rAAV. Co-localization was observed between the PEVs and rAAV-labelled corticothalamic tracts (Fig. 2D), and uptake of PEVs by thalamic neurons was confirmed (Fig. 2E). These results suggest that PEVs can propagate along corticothalamic pathways and be internalized by thalamic neurons.

Figure 2.

Penumbra-derived small extracellular vesicles (PEVs) are transported from peri-infarct cortex to thalamus. (A) Schematic of rAAV2/9-hSyn-mCherry-WPRE-hGH-polyA (rAAV, stereotaxic coordinates: AP −0.2 mm, ML +2.5 mm, DV −1.5 mm; 2 × 10¹² vg/ml, 0.3 μl) and brain-derived small extracellular vesicles (BDEVs) (stereotaxic coordinates: AP −0.2 mm, ML +2.5 mm, DV −1.5 mm; 1 μg/μl, 2 μl) injection into cortex penumbra. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/1tikh0r. (B) Representative images showing rAAV (red) injection into the cortex of the left hemisphere. Right: Injection site with higher magnifications of the boxed areas. Scale bars = 500 µm (overview), 100 µm (magnified). (C) Confocal images of rAAV-labelled corticothalamic axons (red) in the thalamus (originating from peri-infarct cortex). Right: Images of the projection site with higher magnifications of the boxed area. Scale bars = 500 µm (overview), 100 µm (magnified). (D) Immunofluorescence co-staining of rAAV-labelled corticothalamic axons (grey) and BDEVs derived from penumbra (PEVs) (CellMask, magenta) in the thalamus at 3 h post-MCAO. Orange arrowheads indicate internalized PEVs. Scale bar = 25 μm.(E) Immunofluorescence co-staining of PEVs (CellMask, magenta) and neurons (grey) in thalamus at 3 h post-MCAO. Orange arrowheads indicate internalized PEVs. Scale bar = 25 μm. (F) Left: Representative confocal image of thalamic tissue collected 3 h after stereotaxic injection of PEVs (CellMask, magenta) into the ipsilateral thalamus immediately following MCAO. Scale bars = 500 μm. Right: Representative confocal image of successful uptake of PEVs (CellMask, magenta) by thalamic neurons (Nissl, green) 3 h after stereotaxic injection into the ipsilateral thalamus performed immediately after MCAO. Orange arrowheads indicate internalized PEVs. Scale bars = 10 μm. MCAO = middle cerebral artery occlusion; rAAV = recominant AAV.

Our results showed that neither PEVs nor SEVs caused detectable neuronal or synaptic injury in the ipsilateral thalamus in sham mice. Quantitative analyses revealed no significant changes in synapse density (SYP/PSD95), neuronal integrity (Nissl⁺/DAPI) or apoptosis (TUNEL⁺ cells) (Supplementary Fig. 7B–D). Subsequently, to investigate the impact of PEVs delivered to thalamic neurons on thalamic integrity, we intraperitoneally administered GW4869, a neutral sphingomyelinase inhibitor that blocks SEV biogenesis, 1 h before MCAO to systemically inhibit endogenous SEV production.^29,30 PEVs were then injected stereotaxically into the ipsilateral thalamus following MCAO. Three hours after thalamic injection of PEVs, uptake by thalamic neurons was confirmed (Fig. 2F). To evaluate long-term neural effects in non-infarcted regions, brain sections were collected 14 days post-MCAO and subjected to Nissl and TUNEL staining to assess neuronal damage, as well as immunofluorescence co-localization of SYP and PSD95 suggesting presynaptic and postsynaptic structural integrity (Fig. 3A), respectively. Compared with the MCAO group at 14 days, mice treated with GW4869 exhibited greater preservation of intact neurons (Fig. 3B and D), reduced apoptosis (Fig. 3B and E), and increased synaptic density, as reflected by enhanced SYP/PSD95 co-localization (Fig. 3B and C). However, supplementary stereotaxic injection of exogenous PEVs into the thalamus following GW4869 treatment reversed these protective effects. Specifically, the GW4869+PEVs group showed fewer intact neurons (Fig. 3B and D), significantly increased apoptotic activity (Fig. 3B and E), and markedly reduced synaptic density (Fig. 3B and C) relative to the GW4869 group.

Figure 3.

Penumbra-derived small extracellular vesicles (PEVs) propagate synaptic damage and neuronal death after focal cortical infarction. (A) Schematic timeline: Intraperitoneal injection of GW4869 1 h before MCAO, followed by intraperitoneal injection of PEVs or PBS into the ipsilateral thalamus 0.5 h post-MCAO, with tissue harvest at MCAO 14 days. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/y8utbb2. (B) The procedure of inhibiting BDEVs secretion in the peri-infarct cortex of MCAO mice via intraperitoneal injection of GW4869, followed by stereotaxic administration of PEVs (CellMask, magenta) in the ipsilateral thalamus. Top two rows: Representative images showing synaptic staining for synaptophysin (SYP) (green) and postsynaptic density protein 95 (PSD95) (red) in the ipsilateral thalamus of sham mice or mice with focal cortical infarction at 14 days, treated in vivo with PBS, GW4869 or GW4869+PEVs. Scale bars = 50 μm. Second row: The white outlines indicate the co-localized regions of SYP and PSD95; magnified views of the dotted outlined areas in the first row, respectively. Scale bars = 10 μm. Third and fourth row: Representative images showing NISSL staining (green) and TUNEL staining (red) in the ipsilateral thalamus of sham mice or mice with focal cortical infarction at 14 days, treated in vivo with PBS, GW4869, or GW4869+PEVs. Orange arrowheads indicate TUNEL-positive apoptotic cells. Scale bars = 50 μm. (C) Quantitative analysis of the relative level of synaptic density in the ipsilateral thalamus in G. Data are expressed as mean ± standard deviation (SD), n = 6, one-way ANOVA test. (D and E) Quantitative analysis of intact TUNEL⁺ cells/DAPI (%) and intact neurons (NISSL⁺)/DAPI (%) in B. Data are expressed as mean ± SD, n = 6, one-way ANOVA test. P < 0.05 was considered significant. N.S. = not significant. MCAO = middle cerebral artery occlusion.

Further, to suppress the secretion of PEVs in the ischaemic penumbra, GW4869 was also administered via stereotaxic injection into the cortical penumbra 1 h prior to MCAO model induction (Fig. 4A). Using an NTA assay to quantify extracted BEDVs (per milligram), we found that GW4869 significantly reduced BDEV levels in both the penumbra and the thalamus (Fig. 4B and C). Consistent with previous findings, treatment with GW4869, markedly attenuated neuronal loss and apoptosis in the thalamus, and increased the synaptic density. Notably, under GW4869 treatment, subsequent injection of PEVs into the thalamus significantly reinstated neuronal apoptosis and loss (Fig. 4G and H). These data indicate that PEVs alone are insufficient to trigger thalamic neuronal degeneration under non-ischaemic conditions, indicating that PEVs alone are not sufficient to elicit neuronal damage inhealthy tissue, and that additional context-dependent factors likely contribute to their pathogenicity instroke.

Figure 4.

Suppression of cortical penumbra-derived small extracellular vesicle (PEV) secretion attenuates synaptic damage and neuronal death in the thalamus following focal cortical infarction. (A) Stereotaxic injection of GW4869 into peri-infarct cortex at 1 h pre-MCAO reduced BDEVs production in ipsilateral cortex (B) and thalamus (C), as shown by nanoparticle tracking analysis of BDEVs isolated from brain tissues harvested at 3 h post-MCAO. Data are presented as mean ± standard deviation (SD), n = 6, Student’s two-sample t-test. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/h63r4n4. (D) Schematic timeline: Stereotaxic injection of GW4869 into the peri-infarct cortex 1 h before MCAO, followed by stereotaxic injection of PEVs or PBS into the ipsilateral thalamus 0.5 h post-MCAO, with tissue harvest at MCAO 14 days. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/ayft99o. (E) Top two rows: Synaptophysin (SYP) (green) and postsynaptic density protein 95 (PSD95) (red) staining in ipsilateral thalamus at 14 days post-infarction (Sham, MCAO 14d+PBS, MCAO 14d+GW4869, MCAO 14d+GW4869+PEVs groups). Boxed area in the top row is magnified in the panel below. Second row: White outline indicates the SYP/PSD95 co-localized area. Scale bars = 50 μm, 10 μm (magnified). Bottom two rows: Nissl (green) and TUNEL (red) staining in the same groups; nuclei (DAPI, blue). Scale bar = 50 μm. (F) Quantitative analysis of the relative level of synaptic density in the ipsilateral thalamus in E. Data are presented as mean ± SD, n = 6, one-way ANOVA test. (G and H) Quantitative analysis of TUNEL⁺ cells/DAPI (%) and intact neurons (NISSL⁺)/DAPI (%) in the ipsilateral thalamus in E. Data are presented as mean ± SD, n = 6, one-way ANOVA test. P < 0.05 was considered significant; N.S. = not significant. BDEV = brain-derived small extracellular vesicles; MCAO = middle cerebral artery occlusion.

PEVs mediate the delivery ofCircOGDH from cortical to thalamic neurons via trans-synaptic transport

To investigate the anterograde trans-synaptic transfer mechanism of PEVs, we conducted a series of experiments. First, CellMask-labelled PEVs were injected stereotaxically into the cortex, and their distribution was examined in the caudal putamen—a well-established third-order projection site of thalamic relay neurons.³¹ Cortically delivered PEVs were robustly detected in Nissl⁺ neurons of the caudal putamen, whereas PEVs injected directly into the putamen showed minimal thalamic uptake, supporting a predominantly anterograde rather than retrograde mode of spread (Supplementary Fig. 8). Second, triple-label confocal imaging of thalamic sections for presynaptic (SYP), postsynaptic (PSD95) and PEV markers revealed that cortical PEVs strongly co-localized with SYP⁺ and PSD95⁺ synaptic compartments, while striatal-injected PEVs were predominantly extrasynaptic (Supplementary Fig. 9). Third, local inhibition of EV secretion by thalamic injection of GW4869 prior to stroke effectively suppressed locally derived EV signals but did not reduce the accumulation of cortex-derived PEVs in the thalamus, indicating that thalamic EV secretion does not substantially influence the uptake of cortical PEVs (Supplementary Fig. 10). Together, these findings demonstrate that cortex-derived PEVs propagate anterogradely along connected pathways, associate specifically with synaptic sites in the thalamus, and do not rely on retrograde transport by thalamocortical neurons.

Our earlier research showed that hypoxia-induced circular RNA, CircOGDH, is significantly upregulated in the ischaemic penumbra and in plasma exosomes of patients with acute ischaemic stroke, where it contributes to neuronal damage.¹⁸ Based on this finding, we hypothesized that PEVs mediate the delivery of CircOGDH to the thalamus via corticothalamic projections. To test this hypothesis, FISH staining of mouse brain sections revealed accumulation of CY3-labelled CircOGDH in the thalamus following MCAO (Fig. 5A). As the thalamus was not directly affected by ischaemia, these findings suggest a cortical origin of the detected CircOGDH. Furthermore, CircOGDH was predominantly localized in thalamic neurons, indicating potential transport along corticothalamic projections (Fig. 5B).

Figure 5.

Penumbra-derived small extracellular vesicles (PEVs) transport CircOGDH along corticothalamic pathways after focal cortical infarction. (A) Fluorescence in situ hybridization (FISH) experiments showing CY3-labelled CircOGDH (red) in mice with sham or MCAO 3 h after focal cerebral ischaemia. Scale bars = 50 µm and 20 µm (magnified view). (B) FISH experiments showed the localization of CircOGDH (red) in the ipsilateral thalamic neurons (green) of mouse with focal cortical infarction at 3 h. Scale bars = 50 µm and 20 µm (magnified view). (C) Schematic diagram showing real-time quantitative PCR (RT-qPCR) analysis performed on brain tissue or brain-derived small extracellular vesicles (BDEVs) isolated 3 h after MCAO. (D) RT-qPCR analysis showing the comparison of CircOGDH expression levels in the ipsilateral cortical and thalamic brain tissues between mice with focal cortical infarction at 3 h and Sham mice. Data are presented as mean ± standard deviation (SD), n = 6, Student’s two-sample t-test. (E) RT-qPCR analysis showing the comparison of CircOGDH expression levels in BDEVs from the ipsilateral cortical and thalamic brain tissues between mice with focal cortical infarction at 3 h and sham mice. Data are presented as mean ± SD, n = 6, Student’s two-sample t-test. (F) Schematic diagram showing the procedure of stereotaxic injection (s.i.) of GW4869 into the peri-infarct cortex 3 h before modelling, followed by stereotaxic injection of PEVs into the thalamus immediately after modelling, with sampling and observation performed at 3 h after MCAO. Yellow arrowhead indicates the ipsilateral thalamus. (G) As shown in the flow chart in F, FISH was performed in the ipsilateral thalamic tissue from mice in the PBS group, the GW4869 group and the GW4869+PEVs group at 3 h after focal cerebral infarction to observe the expression and distribution of CircOGDH. Scale bar = 20 µm. (H) Left: Schematic diagram showing the procedure of stereotaxic injection of sh_CircOGDH adenovirus (ADV) into the peri-infarct cortex 3 h before modelling, followed by stereotaxic injection of PEVs into the thalamus immediately after modelling, with sampling and observation performed at 3 h after MCAO. Right: Representative images showing the distribution of sh_CircOGDH ADV infected brain tissue (green) in the region of the peri-infarct cortex of MCAO mice. Scale bar = 500 μm. (I) RT-qPCR analysis showing the comparison of CircOGDH expression levels in brain tissue from the ipsilateral cortex and the ipsilateral thalamus between the MCAO 3 h+sh_CircOGDH group and the MCAO 3 h+sh_CircCtrl group. Data are presented as mean ± SD, n = 6, Student’s two-sample t-test. (J) RT-qPCR analysis showing the comparison of CircOGDH expression levels in PEVs from the ipsilateral cortex between the MCAO 3 h+sh_CircOGDH group and the MCAO 3 h+sh_CircCtrl group. Data are presented as mean ± SD, n = 6, Student’s two-sample t-test. (K) Schematic diagram showing the procedure of stereotaxic injection of GW4869 into the peri-infarct cortex 3 h before modelling, followed by stereotaxic injection of sh_CircOGDH adenovirus into the thalamus immediately after modelling, with sampling and observation performed at 3 h after MCAO. Yellow arrowhead indicates the ipsilateral thalamus. (L) As shown in the flow chart in K, FISH was performed in the ipsilateral thalamic tissue from mice in the MCAO 3 h+sh_CircCtrl group, the MCAO 3 h+sh_CircOGDH group and the MCAO 3 h+sh_CircOGDH+PEVs group to observe the expression of CircOGDH in the thalamus. Scale bar = 20 μm. Created in BioRender. Zhu, Z. (2026) https://BioRender.com/nh4g0o0. P < 0.05 was considered significant. N.S. = not significant. MCAO = middle cerebral artery occlusion.

We subsequently extracted cortical and thalamic tissues along with their respective BDEVs and quantified CircOGDH expression using real-time quantitative PCR (RT-qPCR) (Fig. 5C). The results showed significant upregulation of CircOGDH in both the peri-infarct cortex and the ipsilateral thalamus 3 h after cortical infarction (Fig. 5D). Moreover, CircOGDH levels were higher in the PEVs group than in the SEVs group, while BDEVs isolated from the thalamus of MCAO mice showed no significant increase in CircOGDH expression compared to the Sham group (Fig. 4E). We also found that stereotaxic injection of GW4869 into the cortex, which inhibits PEV release, significantly reduced thalamic CircOGDH expression in MCAO mice (Fig. 5F and G). However, exogenous administration of PEVs into the thalamus markedly restored CircOGDH expression (Fig. 5G), confirming vesicle-dependent transport.

To confirm that thalamic CircOGDH originates primarily from the cortical ischaemic penumbra, we stereotaxically injected sh_CircCtrl or sh_CircOGDH GFP-ADV into the cortex 6 days prior to MCAO to achieve specific knockdown of CircOGDH. As shown in Fig. 5H, GFP fluorescence indicated successful cortical adenovirus (ADV) expression 6 days post-injection. Three hours after MCAO, brain tissues from the penumbra and thalamus, as well as BDEVs from the penumbra, were isolated from MCAO+sh_CircCtrl and MCAO+sh_CircOGDH mice. Quantitative PCR confirmed that sh_CircOGDH significantly reduced CircOGDH expression in both the cortical penumbra and thalamic tissue (Fig. 5I). Similarly, CircOGDH expression in PEVs isolated from MCAO+sh_CircOGDH mice was significantly lower compared to the MCAO+sh_CircCtrl group (Fig. 5J). Meanwhile, FISH staining showed markedly reduced CircOGDH expression in the ipsilateral thalamus after ADV-sh_CircOGDH treatment (Fig. 5K and L), which was restored by thalamic injection of PEVs (Fig. 5L).

Collectively, these data demonstrate that CircOGDH, which is upregulated in the ischaemic penumbra, is transported via PEVs through corticothalamic synaptic pathways to thalamic neurons during acute focal cortical infarction.

PEV-transportedCircOGDH accelerates synaptic damage in the corticothalamic pathway

To investigate the role of CircOGDH derived from cortical PEVs in corticothalamic synaptic injury, we performed a targeted knockdown of cortical CircOGDH using stereotaxic injection of ADV-sh_CircOGDH prior to MCAO induction. As shown in Fig. 6A and D, Nissl staining revealed that CircOGDH knockdown significantly mitigated neuronal loss in both the peri-infarct cortex and the ipsilateral thalamus at 14 days post-infarction. Consistently, TUNEL assay results demonstrated reduced apoptosis in these regions in the sh_CircOGDH group compared with the sh_circCtrl group (Fig. 6E and F). To assess neuronal structural integrity, Golgi staining was performed. CircOGDH knockdown in MCAO mice restored dendritic architecture, as evidenced by improvements in the number, length and complexity of dendritic branches, along with increased spine density relative to the sh_CircCtrl group (Fig. 7A–D). Western blot analysis further confirmed that cortical CircOGDH knockdown more effectively restored expression levels of synaptic integrity and neuronal plasticity markers in both the peri-infarct cortex and ipsilateral thalamus compared to controls (Fig. 7E and F).

Figure 6.

Penumbra-derived small extracellular vesicles (PEVs) transported CircOGDH accelerated neuronal damage in corticothalamic pathway after focal cortical infarction. (A and D) Representative images (A) and quantitative analysis of Nissl staining (D) showing neurons in the peri-infarct cortex and ipsilateral thalamic neurons along the corticothalamic pathway in mice transduced with sh_CircCtrl or sh_CircOGDH adenovirus at 14 days after focal cerebral ischaemia. (Nissl-positive neurons, magenta; DAPI, blue; cortex n = 8; thalamus n = 6; scale bar = 20 μm). Data are presented as mean ± standard deviation (SD), one-way ANOVA test. (B) Co-localization of contacts between recombinant AAV (rAAV)-transfected corticothalamic tracts and Nissl-positive neurons in the thalamic region. Corticothalamic tracts, red; Nissl⁺ neurons, magenta; arrowheads indicate contacted cells; scale bars = 10 μm (left); 5 μm (right). (C) Representative images of boutons of rAAV-transfected corticothalamic tracts per 20 μm. (corticothalamic tracts, red; arrowheads indicate boutons; scale bar = 10 μm). (G) Quantitative analysis of contacts between rAAV-transfected corticothalamic tracts and Nissl-positive neurons in the thalamic region. Data are presented as mean ± SD, n = 6, one-way ANOVA test. (H) Quantitative analysis of boutons of rAAV-transfected corticothalamic tracts per 20 μm. Data are expressed as means ± SD, n = 6, one-way ANOVA test.(E) and (F)Representative images and quantitative analysis of TUNEL staining showing apoptotic cells in the peri-infarct cortex and ipsilateral thalamus along the corticothalamic pathway in mice transduced with sh_CircCtrl or sh_CircOGDH adenovirus at 14 days after focal cerebral ischaemia (TUNEL, red; DAPI, blue; n = 6; bar =20 μm). Data were presented as mean ± standard error of the mean, Student’s two-sample t-test. P < 0.05 was considered significant. N.S. = not significant. MCAO = middle cerebral artery occlusion.

Figure 7.

Penumbra-derived small extracellular vesicle (PEV)-transported CircOGDH accelerates synaptic damage in corticothalamic pathway after focal cortical infarction. (A) Representative images of Golgi-Cox staining showing the peri-infarct cortex and ipsilateral thalamus along the corticothalamic pathway in mice transduced with sh_CircCtrl or sh_CircOGDH adenovirus at 14 days after focal cerebral ischaemia (scale bar = 100 μm), followed by analyses of intersections via Sholl analysis (C and D), dendritic length (B, left), and dendritic spine density (B, right). Data are expressed as means ± standard deviation (SD), n = 10–28, one-way ANOVA test. (E and F) Representative western blotting and quantification of postsynaptic density 95 (PSD95) and synapsin-1 in each group. Data are presented as mean ± SD, n = 6, one-way ANOVA test. P < 0.05 was considered significant. N.S. = not significant. MCAO = middle cerebral artery occlusion.

Given the critical role of corticothalamic tract remodelling in post-stroke neural plasticity, we next examined pathway-specific alterations using rAAV for anterograde tracing from the cortex to the thalamus. CircOGDH knockdown preserved axonal bouton density along corticothalamic projections (Fig. 6C and H) and enhanced synaptic contact formation between cortical axons and thalamic neurons relative to controls (Fig. 6B and G). These structural improvements were associated with restored neuronal connectivity.

Collectively, these findings identify CircOGDH as a key mediator of PEV-induced synaptic disruption in corticothalamic pathways following cortical infarction.

PEVs implicate a potential molecular basis for brain injury

To explore the molecular mechanisms by which PEVs contribute to ischaemic brain injury, we conducted transcriptomic profiling. Differential expression gene analysis using a threshold of |log₂FC| > 1 and FDR < 0.05 identified 2213 significantly altered mRNAs in PEVs, comprising 1399 upregulated and 814 downregulated transcripts (Fig. 8A, left). In parallel, miRNA profiling revealed 170 differentially expressed conserved miRNAs between PEVs and SEVs, including 64 upregulated and 106 downregulated miRNAs in PEVs (Fig. 8A, right). Target prediction using the TargetScan and miRDB databases yielded 6921 putative mRNA targets for these miRNAs.

Figure 8.

Penumbra-derived small extracellular vesicles (PEVs) are implicated in a potential molecular basis for brain injury. (A) Volcano plot of 2-fold differentially expressed mRNA (left) (DEmRNA) and miRNA (right) (DEmiRNA) (PEVs versus SEVs). (B) Gene ontology (GO) enrichment analysis of upregulated DEmRNAs, log₂FC > 1 (PEVs versus SEVs). Top: Enrichment results of biological process (BP) and (bottom) Kyoto Encyclopedia of Genes and Genomes (KEGG). (C) GO enrichment analysis of target-downregulated DEmiRNAs, log₂FC > 1 (PEVs versus SEVs). Top: Enrichment results of biological process (BP) and (bottom) Kyoto Encyclopedia of Genes and Genomes (KEGG). Red indicates pathways enriched in both upregulated DEmRNAs and target-downregulated DEmiRNAs in B and C. (D) Gene Ontology and KEGG Pathway Analysis visualized through an Enrichment Map (EMAP) of upregulated mRNAs and target-downregulated DEmiRNAs, log₂FC > 1 (PEVs versus SEVs). (E) Venn diagram showing the overlap between upregulated DEmRNA and target-downregulated DEmiRNAs, log₂FC > 2 (PEVs versus SEVs), based on score (degree) in the top 20 genes, highlighting Sema3a and Robo2 as shared genes. (F) Constructed interaction network featuring downregulated DEmiRNAs, their predicted target genes, and upregulated DEmRNAs. (G) Chord diagram of GO-BP and KEGG enrichment analysis for hub genes of upregulated mRNAs and target-downregulated DEmiRNAs (∣log₂FC∣ > 2, PEVs versus SEVs). Pathways in blue indicate key enriched terms; genes in red represent core genes. SEV = small extracellular vesicles.

To understand the functional significance of these DEmRNAs, we performed GO and KEGG enrichment analyses. Upregulated DEmRNAs were significantly enriched in biological processes related to synaptic and axonal functions, such as synapse organization, axonogenesis and regulation of synaptic activity (Fig. 8B). KEGG pathway analysis also showed enrichment in neurodegenerative disease pathways and axonal guidance mechanisms. Based on the enrichment of synaptic and axonal pathways among upregulated mRNAs, we analysed the target genes of downregulated miRNAs, which may lead to transcript upregulation by relieving translational repression. These target genes showed significant overlap with key upregulated pathways, including synapse organization and axonogenesis in GO, and axon guidance in KEGG (Fig. 8C). Functional linkage analysis using the Enrichment Map (EMAP) platform demonstrated strong interconnections among these pathways (Fig. 8D). PPI networks were constructed for both the upregulated DEmRNAs and target-downregulated DEmiRNAs, using a stringent threshold of |log₂FC| > 2.

We initially identified the top 20 hub genes based on degree centrality in the PPI network from both the upregulated DEmRNAs and the target gene sets of downregulated DEmiRNAs. Subsequently, to explore the potential mechanisms of these genes in synaptic and axonal-related pathways, we performed functional enrichment analysis using GO and KEGG. The results revealed that these genes were significantly enriched in pathways such as axonogenesis and synapse organization, and involved into several mechanisms including Rap1/Ras, mTOR, PI3K-Akt, MAPK signalling, and long-term potentiation (Fig. 8G and Supplementary Table 1). Furthermore, Venn diagram analysis of the intersection between the top 20 hub genes from upregulated DEmRNAs and target genes of downregulated DEmiRNAs, identified Sema3a and Robo2 as common to both groups (Fig. 6E), and closely associated with axonal inhibition.³² To investigate the regulatory relationships between these overlapping hub genes and their corresponding DEmiRNAs, we constructed an miRNA–hub gene interaction network using Cytoscape software (Fig. 8F). Notably, mmu-miR-33-5p was observed to simultaneously target both key genes, Sema3a and Robo2. RT-qPCR confirmed that both mRNAs were significantly elevated in PEVs compared with SEVs, consistent with the sequencing data. Consistent with the qPCR results, expression of both Sema3a and Robo2 was markedly increased in Nissl-labelled thalamic neurons at 3 h after cortical stroke. Their expression remained elevated relative to sham levels at 1 day post-MCAO, though a gradual decline was observed (Supplementary Figs 11 and 12).

Discussion

Beyond interventions targeting the acute ischaemic penumbra, remote damage significantly influences long-term functional impairment after stroke. Axonal degeneration primarily underlies remote damage by structurally connecting affected regions; however, studies investigating the specific mechanisms from the penumbra to non-infarcted regions remain limited.

This study observed that ischaemic stimulation elevates release of BDEVs from the penumbra. These BDEVs exacerbate local neuronal injury and propagate along the corticothalamic pathway, inducing secondary damage within remote thalamic regions. This effect is mediated by pathogenic RNAs, including CircOGDH, which are associated with axonal and synaptic dysfunction. Consequently, therapeutic strategies targeting BDEVs could protect the penumbra and prevent the spread of neurotoxic components to distant brain areas.

The clinical relevance of these acute-phase interventions is underscored by the fact that synaptic impairment is a fundamental pathological mechanism in the early stages of stroke, directly contributing to impaired synaptic transmission and neuronal death.^33-35 Similarly, animal studies have further demonstrated that timely intervention can lead to axonal regeneration and synaptic remodelling within 1–4 weeks post-MCAO.^36,37 Early intervention targeting deleterious molecular pathway and salvaging penumbral integrity have been shown to improve clinical outcomes, particularly at 3 months post-stroke.^38,39 While peri-infarct recovery and cortical reorganization are recognized as essential components of BDEVs in post-stroke rehabilitation.^40,41

Building on early synaptic vulnerability and timely intervention’s therapeutic value, elucidating the cellular origin of EVs is crucial. Ischaemic EVs derive from neurons, astrocytes and microglia in the CNS. For BDEVs isolated from MCAO mice, Brenna et al.⁴² found astrocyte EVs enriched at 24 h post-MCAO, whereas Bub et al.⁴³ observed an increased oligodendrocyte contribution at 72 h. Beyond tissue EVs, plasma-based analyses have identified microglia-associated EVs signatures in experimental stroke models,⁴⁴ Chen et al.⁴⁵ demonstrated that astrocytic metabolic stress can enhance exosome biogenesis through HK2-mediated regulation of nSMase1. Our present findings further extend this understanding by revealing that at 3 h post-MCAO, neurons constitute the predominant cellular source of EVs derived from the ischaemic penumbra. This observation aligns with our earlier work showing that ischaemic neurons within the penumbra release EVs involved in neuronal apoptosis¹⁸ and neurovascular unit regulation.¹⁷ Collectively, EV origins vary depending on cell type, brain region and time point, underscoring the importance of defining the cellular source for accurate interpretation of EV-mediated mechanisms.

PEVs deliver pathogenic molecules across synapses to anatomically connected thalamic nuclei, thereby inducing remote neuronal injury in non-infarcted regions. Notably, this trans-synaptic transport capacity mirrors sEV-mediated interneuronal propagation observed with misfolded proteins including tau and α-synuclein. Importantly, under pathological conditions, upregulated sEV secretion facilitates systemic trafficking of detrimental cargo from peripheral organs to the brain, subsequently driving neural circuitry damage that compromises synaptic integrity and white matter homeostasis. For example, foam cell-derived sEVs exacerbate post-ischaemic white matter injury via miR-101-3p delivery to microglia,⁴⁶ alongside adipose-derived sEVs inducing hippocampal synaptic damage and cognitive deficits through neuronal miR-9-3p uptake.⁴⁷ Future therapeutic approaches may target pathogenic EV biogenesis, intercellular transmission or recipient cell uptake to mitigate such remote damage.

Small EVs mediate intercellular communication by transferring cargo such as RNAs between source and recipient cells and consequently regulate biological function of recipient cells. We have reported that CircOGDH enrichment in plasma sEVs of patients with acute ischaemic stroke, where it correlates with penumbra size and promotes neuronal apoptosis. In addition, PEVs exhibited elevated CircOGDH expression compared to SEVs. Knockdown of CircOGDH or inhibition of cortical SEVs secretion in the penumbra of MCAO mice that reduced CircOGDH levels and reversed neuronal and synaptic damage in the ipsilateral thalamus. These findings suggest that reducing CircOGDH plays an important role in alleviating PEV-mediated remote thalamic damage after focal cortical infarction. Subsequently, through tissue-derived EVs from brain combined with transcriptomic sequencing, we characterized alterations in intercellular communication and the impact of other RNA cargo trafficking from an extracellular perspective. Bioinformatics analyses revealed that PEVs are enriched in RNAs involved in synaptic and axonal regulation. Sema3a and Robo2 are shared hub genes across both mRNA and miRNA-targeted gene sets. Proteins regulated by these genes function as intrinsically inhibitory factors within neurons, suppressing axonal growth. Rhoa, identified as the top-ranked hub gene among downregulated miRNA targets by connectivity analysis, functions as a downstream molecule of Sema3a.^32,48 Further work supports this link, as its activation has been associated with impaired axon regeneration⁴⁹ and ischaemic damage,^50,51 potentially via the Ras and PI3K-Akt pathways. In addition, beyond axonal associations, mRNAs identified as top-ranked hub genes were also linked to synaptic pathways. The upregulation of the synapse-related gene Grin1 is consistent with prior studies showing that increased expression of excitatory synaptic proteins contributes to excitotoxicity and neuronal damage in acute stroke.^52,53 These PEV targets, along with their upstream regulators such as downregulated miRNAs (e.g. miR-154-3p), have not been previously reported in the context of ischaemic stroke and may represent novel neuroprotective targets for future research.

Furthermore, our optimized method for BDEV isolation significantly reduces processing time compared with the conventional multi-layer sucrose gradient approach, decreasing it by at least 50%. This reduction minimizes the degradation and loss of sEVs commonly associated with prolonged ultracentrifugation. Electron microscopy confirmed the isolation of vesicles with high purity and intact morphology. This methodological advancement enabled the successful extraction of sufficient PEVs for downstream transcriptomic analysis. However, BDEVs contain numerous RNAs, and the molecular mechanisms underlying their role in stroke-induced synaptic dysfunction require further exploration.

In conclusion, hypoxic conditions markedly enhance SEV-mediated intercellular communication within the brain. By directly isolating and purifying BDEVs from ischaemic penumbral tissue and performing transcriptome sequencing, this study provides a comprehensive overview of dynamic extracellular signalling events during the acute phase of stroke at the tissue microenvironment level. Compared with SEVs derived from cell culture supernatants, tissue-derived SEVs more accurately reflect the in vivo pathological state.^54,55 Our findings clearly demonstrate that PEVs are capable of trans-synaptic transport along the corticothalamic pathway and contribute to remote neural injury, providing novel insights into the mechanisms of secondary neural circuit impairment following stroke.

Data availability

Data are available from the corresponding authors upon reasonable request.

Funding

This work was supported by grants from the National Natural Science Foundation of China (82571469, 82571471, 82271304, 82171316, 82503549, 82401561, 82501552, 82201437, 82401525 and 82503549), the Guangdong TeZhi Plan (0720240214), State Key Laboratory of Bioactive Molecules and Druggability Assessment. Guangdong Basic Research Center of Excellence for Natural Bioactive Molecules and Discovery of Innovative Drugs, the First Affiliated Hospital of Jinan University (SKLBMDA202504021), Science and Technology Project of Guangzhou (202102070001, 2025A04J6829, 2025A04J3475), Guangdong Basic and Applied Basic Research Foundation (2025B1515020086, 2025A1515010586 and 2023A1515111085), Sun Yat-Sen University Hundreds of Talent Program (1320324001), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0533700), China Postdoctoral Foundation General Program (2024M760457 and 2025M782022), Sun Yat-sen Memorial Hospital Medical Scientist Training Program (1320900067). National Natural Science Foundation of China (82503549), Guangdong Basic and Applied Basic Research Foundation (2025B1515020086, China), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2024ZD0533700, China), Sun Yat-sen University Hundreds of Talent Program (1320324001, China), Sun Yat-sen Memorial Hospital Medical Scientist Training Program (1320900067, China).

Competing interests

The authors report no competing interests.

Supplementarymaterial

Supplementary material is available at Brain online.