Differential Effects of Ischemia and Inflammation on Plasma-Derived Extracellular Vesicle Characteristics and Function in a Mouse Model

TL;DR

The study examines how EVs from stroke and inflammation affect brain tissue differently, highlighting their potential roles in neuroinflammation.

Contribution

The paper reveals distinct EV profiles from stroke and inflammation models and their impact on cytokine expression in the brain.

Findings

Post-stroke EVs show different markers compared to inflammation-derived EVs.

Both stroke and inflammation-derived EVs increase pro-inflammatory cytokine gene expression in the brain.

No significant differences in microglial or astrocytic reactivity were observed.

Abstract

Extracellular vesicles (EVs) have long been understood to be important mediators of cell-to-cell communication and may lead to the molecular aftermath and exacerbation of brain injuries such as stroke. This study explored how the source of the EVs influenced their characteristics and the effect these differences had on naïve brain tissue. EVs were isolated from mice post-stroke in the acute or chronic stages of recovery in animals with and without reperfusion (transient and permanent middle cerebral artery occlusion) and from a model of systemic inflammation (i.p. lipopolysaccharide). The data show that neither stroke nor inflammation significantly increases EV numbers compared to sham or naïve animals. Post-stroke EVs exhibited a panel of different platelet and inflammatory markers when compared to EVs derived from a model of inflammation, reflecting differences between stroke and systemic immune activation. When injected into the brain, both stroke-derived and inflammation-derived EVs induced pro-inflammatory cytokine gene expression (IL-1β and CXCL1), suggesting a potential role in neuroinflammation. However, no clear group-level differences in microglial or astrocytic reactivity were detected at the level of regional histological assessment, despite consistent increases in ICAM-1 reactivity. The findings here underscore the complexity of EVs’ roles in pathophysiology and highlight the need for improved EV isolation methods. With further longitudinal studies, we may be able to more accurately determine how the context of the injury (reperfusion vs. no reperfusion vs. inflammation) might contribute to the EV populations and their function. Understanding more about EVs in different contexts will improve our ability to use EVs as biomarkers but also our capacity to interfere with EV biology as a novel therapeutic approach.

Figures

Full text

1. Introduction

Ischemia within the central nervous system (CNS) results in the rapid death of neurons and the activation of other key cells within the neurovascular unit and brain parenchyma. These include not only microglia and astrocytes but also cells such as endothelia and pericytes. Pathologically, cerebral ischemia has significant long-term consequences for brain health. Studies of the inflammatory status of the brain in the chronic phases post-stroke have suggested that activation of microglia and astrocytes persists far beyond what might be considered the acute ‘clean up’ phase after injury [1,2]. This chronic CNS inflammation is likely to contribute to the development of neurodegeneration and cognitive decline.

In neurodegenerative diseases such as Alzheimer’s and Parkinson’s, chronic inflammation is thought to contribute to the proteinaceous nature of the pathology [3,4]. In vascular dementia, where there is chronic vascular pathology within the CNS, inflammation may also play a crucial role [5,6]. In stroke, we have a vascular event, which has the capacity to precipitate inflammation and long-term vascular changes, but the nature of how this event is communicated throughout the brain is only just beginning to be understood [7,8].

Extracellular vesicles (EVs) are lipid-bound vesicles that are an important novel means of cell-to-cell communication [9,10]. They have traditionally been categorized according to their biogenesis, with, for example, ‘exosomes’ released from multivesicular bodies and ‘microvesicles’ blebbed from the cell surface [11]. However, the inherent challenges associated with isolating these individual populations of vesicles have necessitated the use of the term extracellular vesicle to cover the majority of the subtypes currently understood [11,12]. EVs have been known to be functional since the late 1980s, and more recently, they have been found to contribute to the exacerbation of inflammation after acute CNS injuries [13,14,15,16]. This promising preliminary data suggests that EVs could act as a signature for the injury and could be used as biomarkers for long-term outcomes [17]. Furthermore, the injury signature is likely to dictate the downstream function of the EV. However, because the study of EVs in CNS diseases is still a relatively new field, a large number of the published studies to date have used differential centrifugation to isolate EV populations, something which is known to increase protein contamination [12,18]. Here, we wished to determine the degree to which a ‘purer’ population of particles contributed to CNS inflammation and the degree to which the precipitating factors in the release of those particles, i.e., the nature of the injury, contributed to their functional effects. Understanding these aspects of EV biology will enable us to make more effective decisions on the use of EVs as biomarkers [17] and on the function of EVs in cell-to-cell communication.

2. Results

2.1. Neither Stroke nor Systemic Inflammation Induces Significant Increases in Circulating EV Numbers

EVs were isolated from the supernatant of animals after middle cerebral artery occlusion (MCAO) where the filament was removed after 20 min (transient occlusion—tMCAO) or where the filament remained in place (permanent occlusion—pMCAO). Animals were allowed to recover for 2 h (acute) or 7-day before plasma was collected for EV isolation. These animals were compared to those who had experienced an acute inflammatory challenge (6 h i.p. LPS) to determine the degree to which the circulating EV population might reflect an inflammatory response. After size exclusion chromatography (SEC), the first four fractions, where contaminating free proteins were excluded (Appendix A, Figure A1), were pooled and analyzed by nanoparticle tracking. There was no significant difference in average particle concentration between the groups (Figure 1A). Distribution suggests that EVs lie within the normal range expected (Figure 1B,C) and that reperfusion of the brain does not appear to significantly influence the distribution pattern of EVs found in the plasma. Electron microscopy shows the standard ‘cup-shaped’ EVs lying between 50 and 200 nm (Figure 1B,C—insets), demonstrating that SEC of cell culture supernatant is an effective way to isolate EVs. However, electron microscopy images also show some ‘EV-like’ structures and an abundance of miscellaneous particles (Figure 1B,C—inset). These are likely lipoprotein contaminants, and their removal was beyond the scope of the current study.

2.2. Mouse Plasma EVs Express High Levels of Integrins and Adhesion Molecules as Well as CD9 and CD81 but Low Levels of CD63, Irrespective of Treatment

Given the small starting sample volumes for the mouse plasma experiments, the use of traditional techniques such as Western blotting [12,19] for confirming the presence of EVs was challenging due to low protein yield. As such, more specialist techniques such as dedicated flow cytometry panels were used. Here, the EV-specific mouse immune-oncology panel includes the tetraspanins CD9, CD81 and CD63 and should be considered a surrogate for blotting. All samples, including naïve, showed high expression levels of CD9 and CD81, suggesting that EV isolation was successful and that the fractions chosen from early SEC did indeed contain EVs (Figure 2). No samples showed significant expression levels of CD63 (Figure 2). In addition to standard markers of EVs, early SEC fractions also showed increased expression levels of integrins, such as CD62P, CD29 and CD61, as well as inflammation-associated molecules, such as MHCII and CD44. As expected, given the high prevalence of platelet EVs in plasma samples, all samples also expressed high levels of CD41 (Figure 2). Although individual one-way ANOVAs identified some nominal differences in specific markers, these were not emphasized due to the risk of inflated Type I error and absence of correction for multiple comparisons (full panel graph found in Appendix A, Figure A2; raw data available in Supplementary File S1).

2.3. Reperfusion of the Brain Results in EVs in the Circulation with a More Pro-Inflammatory Protein Profile

The preservation and amplification of signal on EVs is a known mechanism of organ-to-organ communication [20,21]. Here, we wished to investigate the potential for the protein profile of EVs in the circulation to be affected by the original pathological stimulus. Given that inflammation is a known post-stroke outcome, we used a prototypical inflammatory stimulus (LPS) to generate plasma EVs with a stereotypical inflammatory profile. The data show that pooled EVs from the plasma of animals undergoing inflammation show upregulation of inflammatory markers on their EVs (Figure 3). These include C-reactive protein (Figure 3, highlighted box 1), E-selectin (Figure 3, highlighted box 2) and pentraxin-3 (Figure 3, highlighted box 3). These proteins are also present at high levels in EVs from animals two hours post-MCAO when the brain has been reperfused but appear lower in those animals where the brain has not been reperfused. E-selectin, for example, is particularly high at two hours post-tMCAO but remains low post-pMCAO until the seven-day timepoint, where it is possible chronic inflammation may be starting. Other molecules, such as serpin-E1 (Figure 3, highlighted box 4), appear only in prototypical inflammation but are not as highly expressed in any of the post-stroke EV samples (raw data available in Supplementary File S2).

2.4. EVs with Different Pathological Origins Induce Different Inflammatory Responses

Given the differences in protein profile on the EVs from different sources, our assumption was that they would have different physiological effects. One way in which EVs might be important post-stroke is through the communication of inflammatory signals to the circulation but also to distant parts of the brain. Here, we used a glass microcapillary to inject EVs into the naïve brain to determine their effect on the local immune cell population. In particular, we were interested in the degree to which pro-inflammatory genes were locally upregulated in response to exogenous EVs. We found that EVs derived from animals post-stroke with and without reperfusion of the brain induced different inflammatory profiles depending on when the EVs were harvested post-injury. Interestingly, despite the increased pro-inflammatory profile from transient EVs, they did not induce significant expression of either IL-1β or CXCL1 after injection into the naïve brain (Figure 4A–D). EVs derived from animals with a permanent occlusion of the MCA did induce significant upregulation of IL-1β and CXCL1 at 2 h post-occlusion (Figure 4A, B; p < 0.05, p < 0.001 respectively) but only of CXCL1 at 7-day post-occlusion (Figure 4D; p < 0.0001). We also found that EVs derived from the circulation of inflammatory animals induced inflammatory gene expression in the brain. However, they, like the EVs from transient occlusion animals, did not upregulate IL-1β (Figure 4E) but did upregulate CXCL1 mRNA at the site of injection (Figure 4F; p < 0.01).

Given the presence of markers of inflammation, such as CRP, on the surface of our EVs (Figure 3) and the fact that these inflammatory molecules are known to upregulate endothelial adhesion molecules [22], we hypothesized that our EVs would likely upregulate ICAM-1 and VCAM-1 in naïve animals. In fact, irrespective of which animals or timepoints our EVs were derived from, they induced no significant upregulation of VCAM-1 expression in the brain (Figure 5A,C,E). However, EVs from animals experiencing a transient MCA occlusion upregulated ICAM-1 mRNA expression in the brain when taken at both two hours and seven days post-occlusion (Figure 5B,D; p < 0.05 and p < 0.01, respectively). Similarly, EVs derived from animals after acute inflammation (LPS) also upregulated ICAM-1 mRNA expression when injected into a naïve brain (Figure 5F; p < 0.05).

2.5. EVs with Different Pathological Origins Elicit Differential Regional Inflammatory Responses in the Brain

In addition to mRNA expression profiles indicating localized inflammatory signaling, histopathological assessment was performed to evaluate regional, semi-quantitative changes in inflammatory marker reactivity. Tissue was stained for ICAM-1, GFAP, and Iba-1, and analysis was restricted to the ipsilateral hemisphere (see Appendix A, Figure A3 for representative images and hemispheric context). ICAM-1 reactivity was significantly increased at the regional tissue level in animals injected with EVs derived from 2 h transient (Figure 6A,D; p < 0.01) and 2 h permanent (p < 0.05) stroke models as well as in animals receiving EVs derived from systemic inflammation (Figure 6A,D; p < 0.01). In contrast, GFAP reactivity within the striatal region of interest did not differ significantly between treatment groups (Figure 6B,E). However, a significant increase in GFAP reactivity was observed in the adjacent corpus callosum (Appendix A, Figure A5A,B). Iba-1 reactivity did not differ significantly between groups at the level of regional histological analysis (Figure 6C,F), despite detectable staining in some specimens, indicating no measurable group-level differences in this marker under the exposure conditions used.

3. Discussion

EVs are known to be a means of cell-to-cell communication, but their role in the potentiation of injuries such as stroke is currently unclear. Here, we demonstrate that the original injury contributes to the physical make-up of the EV and that these different characteristics result in distinct but constrained inflammatory responses within the brain. Given the increasing interest in using EVs as biomarkers in diseases such as stroke [23], this paper demonstrates the need to approach these studies with caution, making sure that a diverse patient population with differing pathological etiologies is represented in order to extrapolate effective biomarkers of disease progression [17]. It also demonstrates the paucity of data currently available on the characterization of circulating EVs in mice. Throughout this discussion, findings are interpreted cautiously in light of well-recognized limitations (see Limitations) in EV isolation and characterization from biofluids, and the majority of results are discussed with initial confidence framing.

At the start of the field of EV research, it was assumed that stimulation of any kind, be it in cell culture or as part of a disease state, resulted in increased EV production in the supernatant or bodily fluid they were being measured in [24]. However, it is now more widely accepted that absolute particle measurement from biofluids is challenging given the degree of lipoprotein contamination and the challenges associated with isolating a ‘pure’ population of EVs [12,17,18,19]. Indeed, studies in stroke show widely varying correlations between infarct volumes, outcomes and EV numbers, with studies showing a positive correlation [25] and no correlation at all [20,26]. These differences are likely to reflect differences in isolation and analysis techniques that are ubiquitous throughout the EV field and are the main reason behind the lack of correlational analysis between lesion volume and EV numbers in this study, which are in contrast to our previous findings [13]. The findings relating to EV number should, therefore, be interpreted with confidence at a comparative level between experimental groups but not as absolute measures of circulating EV abundance.

The EVs isolated here express high levels of CD9 and CD81 but not CD63. This has been found elsewhere in cell lines [21], and studies on plasma and serum have found that plasma often has significantly lower levels of CD63+ EVs than serum [27]. Given that there may also be a number of differences in EV marker levels between species [28,29], it is possible that CD63 might not be the ideal marker to identify EV populations in mouse plasma [30].

In addition to standard surface markers, EVs from all animals expressed high levels of CD41, suggesting a strong platelet origin [31]. The data also demonstrate the inherent noise in the stroke models used, with significant variability present in all post-stroke animals but absent in naïve and LPS-treated animals. The surface markers on EVs from a model of inflammation differ from EVs from post-stroke animals, suggesting that post-stroke inflammation and TLR-driven inflammation do not generate the same EV populations. For example, CD41 expression is variable and high in the acute transient stroke animals but much tighter, although still elevated, in the LPS animals. Given the important differences in immune-thrombosis and thrombo-inflammation [32], these differences may be key to understanding the origins and purpose of the inflammatory response post-stroke, but it is important to remember that these observations should be interpreted as qualitative differences in EV population composition rather than precise quantitative distinctions in individual marker abundance.

This difference is further backed up by the absence of some markers of inflammation on EVs derived from post-stroke animals. PAI-1 (serpin-E1), for example, is induced by circulating cytokines and is expected to be significantly upregulated after LPS stimulation [33]. Indeed, the work here shows it to be increased on inflammation-derived EVs but in relatively low abundance in most post-stroke EVs. In contrast, C-reactive protein (CRP), which we have previously shown to be increased on stroke-derived EVs [13], is present on both inflammatory EVs and on stroke-derived EVs, with higher abundance in the earlier phases post-stroke. This temporal pattern mirrors established plasma CRP dynamics following brain injury [34] and suggests that EV-associated CRP reflects injury context rather than generic inflammatory activation. These marker-specific patterns should be interpreted as qualitative differences in EV composition between pathological contexts rather than as exhaustive profiling of inflammatory signaling.

The inflammatory response to ischemia with reperfusion versus without reperfusion has been well documented [35,36], and differences in vascular reactivity between these models are likely to contribute to the distinct EV populations observed here. Reperfusion is associated with altered expression of vascular adhesion molecules [37], and differences in shear stress between transient and permanent occlusion models are known to influence EV release from endothelial cells [38]. In addition to model-specific effects, temporal differences in EV profiles are biologically plausible given the dynamic nature of post-stroke inflammation. Accordingly, the acute (2 h) and chronic (7-day timepoints used here reflect established phases of stroke pathophysiology and allow assessment of whether EV-associated inflammatory signatures persist beyond the immediate injury phase [39].

Given these model- and time-dependent differences in EV composition, an important next question is whether such distinctions have functional consequences within the CNS [9,10]. Here, introduction of EVs into the CNS had a variety of effects. Regionally increased ICAM-1 reactivity was observed, alongside increased GFAP reactivity within the corpus callosum, mostly in response to EVs derived from plasma taken during the acute stages of reperfusion or from inflammation. These largely negative findings are interpreted here as biologically informative within the resolution of the regional, semi-quantitative histological approach used, indicating that EV-induced vascular activation can occur in the absence of robust microglial or VCAM-1 responses under these conditions. These functional findings should be interpreted as evidence of selective, low-level vascular and astrocytic responsiveness to circulating EVs rather than as evidence of broad neuroinflammatory activation.

The roles of ICAM-1 and VCAM-1 in the brain are separated by sensitivity and expression. ICAM-1 is much more ubiquitous, being expressed on astrocytes and microglia as well as endothelial cells, and is one of the first adhesion molecules to be activated during ischemia [40]. ICAM-1 is an early warning system and much more sensitive to low levels of inflammation and changes in shear stress [41,42]. It is not unsurprising that ICAM-1 protein expression followed a similar pattern to ICAM-1 mRNA, and the response may reflect the low-level inflammatory potential of EVs from LPS-challenged animals and from those deriving from the plasma of acute stroke animals. It is possible that higher concentrations of EVs or more prolonged exposure may result in increases in VCAM-1 expression and resultant system immune cell migration. Accordingly, the ICAM-1 responses observed here are most consistent with an early, sensitive vascular response detectable at the regional tissue level rather than with downstream leukocyte recruitment or, as mentioned above, widespread immune activation These questions remain the focus of our ongoing research.

The absence of detectable microglial activation should be interpreted as a constrained biological response under the exposure conditions used rather than as evidence that EVs lack immunomodulatory capacity. Astrocytes express lipid receptors that many microglial cells do not. They are sensitive, for example, to levels of LDL, and lipoproteins are known contaminants of EV preps from plasma, even when using SEC [12,19]. It is understood that cholesterol may play a role in astrocyte-mediated neuroinflammation [43], which may not occur in microglia. Together with the data on the lack of VCAM-1 signal, this suggests that our EVs may simply not be inducing a significant inflammatory response. Whilst an obvious next step would be to increase the dose of EVs, this might not represent what occurs in normal pathology, where EVs are likely to be released over long periods of time. As such, an osmotic mini-pump approach, which continuously infuses low levels of EVs into the CNS, might be more appropriate and is the current focus of our ongoing research in this area.

Limitations: There remains much that is poorly understood about EV biology, particularly functional work in vivo, and the limitations of this study largely reflect current constraints within the EV field rather than study-specific shortcomings. Although the isolation approach used here enriches for small EVs, it does not fully exclude other vesicle subtypes or non-vesicular particles. Accordingly, particle measurements derived from scatter-mode nanoparticle tracking analysis are interpreted comparatively rather than as absolute EV counts or, indeed, a metric of EV purity, given the known sensitivity of this approach to background particles and buffer conditions. While size-exclusion chromatography removes the majority of soluble protein contaminants, residual lipoprotein contamination remains a recognized limitation [44], particularly when working with the small plasma volumes obtainable from mouse models. The relative scarcity of in vivo mouse plasma EV datasets further limits the availability of robust positive and negative markers, as reflected by the low CD63 expression observed here. Although multiplex bead-based profiling enables high-throughput comparison across groups, resolution for individual markers is limited, and subtle changes may be missed, particularly after correction for multiple comparisons.

A further limitation is the inability to definitively assign cellular or tissue origin to circulating EVs using the approaches employed. The predominance of platelet-associated markers suggests a substantial non-CNS contribution, consistent with the acute phase response and systemic immune activation that occur post-stroke [45,46]. While an inflammatory EV group was included for comparison, EVs induced by pathogen-associated and damage-associated molecular patterns are likely to differ in origin and function. In addition, this study did not directly assess EV localization or uptake within the CNS following injection, and it, therefore, remains unclear whether observed effects reflect parenchymal entry, endothelial interaction, or perivascular activation. Although labeling strategies exist, they remain subject to significant technical limitations [47]. Defining EV cellular origin and CNS uptake with greater precision will require orthogonal labeling and single-cell approaches, which remain an important focus for future work. Finally, histological analyses in this study were intentionally designed as semi-quantitative, region-level assessments of inflammatory reactivity and are not intended to resolve cell-specific activation states or microanatomical heterogeneity.

4. Materials and Methods

Animals: Male CD1 mice, 8–10 weeks of age, were housed under standard diurnal lighting conditions (12 h) with ad libitum access to food and water. All procedures were carried out in accord with the UK Animals (Scientific Procedures) Act (1986), and licensed protocols were approved by local committees (LERP and ACER, University of Oxford) and carried out under license number P6CE80C21D (granted 17.12.22). Throughout the procedures, all efforts were made to adhere to the ARRIVE and IMPROVE guidelines [48,49]. Briefly, animals were randomly assigned to experimental groups prior to EV injection to ensure unbiased treatment allocation. Animal groups were: naïve, sham surgery, 2 h survival transient MCAO (2 h tMCAO), 2 h survival permanent MCAO (2 h pMCAO), 7-day survival transient MCAO (7-day tMCAO), 7-day survival permanent MCAO (7-day pMCAO) and inflammation (i.p. LPS 6 h survival). EVs were isolated, labeled, and coded by an independent researcher not involved in the injections or downstream analyses. Group identities were recorded in a central spreadsheet and kept blinded from the experimenter performing the injections and analyses. Blinding was maintained throughout data collection and analysis. For histological assessments, microscope slides were anonymized by covering sample identifiers with opaque tape. Quantification was performed without access to treatment information, which was only unblinded after all data had been collected and analyzed.

Middle Cerebral Artery Occlusion (MCAO) induction: For surgery, animals were anesthetized in 1.5–2% isoflurane carried in an oxygen:nitrous oxide mix (60:40, 2 L/min). Animals were initially placed prone on a nose cone, and a small incision was made over the area of the cortex supplied by the MCA. The skull was slightly thinned, and a laser doppler probe was placed to ensure that a minimum 70% drop in baseline blood flow occurred at the time of MCA occlusion. At this point, animals were placed supine on the nose cone, and a midline incision was made in the neck. The common carotid artery was isolated and temporarily ligated. The MCA was occluded using the Longa method [50]. Briefly, the external carotid was isolated and ligated at the distal end. The internal carotid artery was temporarily clamped, and an incision was made in the external carotid. A 6-0 nylon filament (Duccol, Massachusetts, US) was inserted into the external and held in using a temporary suture around the external. The filament was then manipulated into the internal carotid, after removing the clamp, and proceeded along the middle cerebral artery. Laser doppler blood flow was monitored for a significant drop, and at this point, the temporary suture was tightened to keep the filament in place during occlusion. For animals undergoing temporary MCAO, the filament remained in place for 25 min prior to removal and cauterization of the hole in the external carotid artery and removal of the ligature around the common carotid. For animals undergoing permanent MCAO, the filament remained in place, and the animals were recovered. Survival times were hyperacute (2 h) or chronic (7-day). For those animals surviving for 7-day, gel-based food and additional s.c. saline were provided throughout the recovery period.

Modeling inflammation: Animals received a single dose of LPS (0.5 mg/kg in 0.9% saline; from E. coli O111:B4; Sigma, Welwyn Garden, UK) i.p. to induce a typical systemic inflammatory response [51]. Animals were allowed to survive for 6 h prior to culling, when blood was collected for EV isolation. Saline (vehicle) controls were used.

Extracellular vesicle (EV) isolation: For all animals, blood was collected by cardiac puncture using a heparinized needle and taken into heparinized tubes. Whole plasma was centrifuged for 15 min at 2500 g, and the supernatant was then centrifuged for a further 15 min at 2500 g. Platelet-free plasma (PFP) was then aliquoted into 100 µL aliquots and stored at −80 °C until EV isolation. EVs were isolated fresh prior to each experiment. Briefly, 100 µL of PFP was applied to single-use size exclusion columns (70 nm Legacy—iZON, UK), and 12 × 200 μL fractions were collected using an iZON automated fraction collector, with vesicles being pooled from fractions 1–4 (Appendix A, Figure A1). Size exclusion was used to enrich for small extracellular vesicles (EVs), which are likely to predominantly include exosomes and small microvesicles based on size. However, this isolation method may also capture other vesicle subtypes and does not allow complete discrimination between EV populations - hence, the use throughout the manuscript of the umbrella term ‘extracellular vesicle’ [11].

Extracellular vesicle analysis: Counts were performed using the Zetaview platform (Particle Metrix, Cambridge, UK). EV samples were diluted in PBS (1:1000) and were analyzed for size and concentration using NTA scatter-mode video capture for 0.5 s at 11 positions with the following acquisition settings: shutter speed  =  100, sensitivity  =  80, cycle number  =  2, frame rate  =  30, minimum brightness  =  25, maximum size  =  1000, minimum size  =  5, trace length  =  15; and were analysed using the integral ZetaView software (version 8.05.12 SP2, Particle Metrix, Cambridge, UK). After counting aliquots of EVs, they were stored at −80 °C until further use.

Electron microscopy: For negative staining, freshly glow discharged (15 mA, 25 s) 300 mesh carbon-coated EM grids (C267, TAAB Laboratories Equipment Ltd., Berkshire, UK) were inverted carbon-side down onto a 10 µL droplet of sample and incubated for 2 min. Grids were then gently blotted with filter paper and transferred onto a 20 µL droplet of 2% (w/v aq.) uranyl acetate solution. Samples were incubated for ten seconds, blotted and air-dried. After removal of excess uranyl acetate, the samples were dried for 10 minutes and analyzed using a Tecnai 12 TEM (120 kV; FEI/Thermo Fisher, Birmingham, UK) using a Gatan OneView CMOS camera (×29,000 magnification; Gatan Inc., Oxfordshire, UK). Negative staining was kindly performed by Dr. Errin Johnson (Dunn School of Pathology, University of Oxford).

Flow cytometry: The MACSPlex EV kit (IO—immuno-oncology; Miltenyi, Slough, UK) was used to not only characterize the surface profile of the EVs but also to ensure the expression of ‘standard’ markers of EVs (CD63, CD9, CD81), as per MISEV guidelines [12,19]. EVs were isolated from plasma, as above, and 10^12^ EVs per sample were run according to the manufacturer’s instructions, with an overnight incubation at 4 °C to increase specificity. Samples were run on the CytoFLEX flow cytometer (Beckman Coulter, Brea, CA, USA), with lasers adjusted for use with this kit (advice kindly provided by Dr. Josh Welsh and Dr. Jamie Cooper). Samples were set to record at least 5000 single-bead events per sample. Bead populations were identified based on their fluorescence in the PE and FITC channels. For analysis, samples were normalized to background fluorescence units acquired for buffer and isotype controls. The values were normalized to the mean MFI of the REA control to determine the relative expression of each marker.

Dot blots: EVs were isolated from plasma, as above, 10^12^ EVs were applied to Proteome Profiler XL Array (Mouse; Biotechne, Abingdon, Oxfordshire, UK), and samples were run according to manufacturer’s instructions, with an overnight incubation at 4 °C to increase specificity. Image analysis was performed using the circle tool in the BioRad ImageLab software (version 6.1), where a circle of a specific size was drawn and then placed over each spot on the blot (Appendix A, Figure A4). Reference spots for each blot were set as ‘100’, and all dots within that blot were calculated as relative to those in order to reduce inter-blot variability that might be introduced by comparing blots directly to naïve samples. Relative levels for each blot were then plotted as a heat-map using R (version 3.6.0 and RStudio version 2024.04.1 + 748).

Stereotaxic surgery: Animals were anesthetized (using the same parameters as for MCAO surgery) prior to being placed on a stereotaxic frame. A midline incision was made on the scalp, and a burr hole was drilled in the skull above the coordinates for the dorsal striatum (mm from Bregma +0.5 A/P, −1.5 M/L and −2.5 D/V). A total of 1 µL EVs (1 × 10^10^/µL) in PBS was injected over five minutes using a glass microcapillary. Animals were allowed to survive for 24 h prior to culling. The EV dose used for stereotactic injection was selected as a pragmatic and conservative quantity that could be reliably delivered within the legal maximum 1 µL injection volume for intracranial injections. As this study was designed to assess comparative effects of EVs from different pathological contexts rather than to establish dose–response relationships, the same EV dose was applied across all groups to enable relative interpretation. Accordingly, the absence or presence of effects at this dose should be interpreted as indicative of differential EV bioactivity under matched exposure conditions rather than as evidence of a threshold or maximal response.

Tissue collection: For fresh tissue, animals were perfused with ice-cold 0.9% saline containing 5 U/mL heparin (Sigma, Welwyn Garden, UK). Liver and brain tissues were snap frozen and stored at −80 °C until further use. For fixed tissue, animals were perfused with room temperature 0.9% saline until tissues were cleared, followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer. Tissue was post-fixed overnight in 4% PFA and then cryoprotected in 30% sucrose until the tissue had sunk.

Tissue processing: Fresh and fixed brains were cryosectioned at 12 μm using a Leica cryostat ±500 μm A/P of the injection location. For fixed sections, tissue was placed on gelatin-coated slides, air-dried and stored at −80 °C. For fresh sections for qPCR, sections were split into ipsilateral and contralateral hemispheres and stored at −80 °C.

Immunohistochemistry and analysis: To assess the naïve CNS response to exogenous EVs, brain sections were stained for intracellular adhesion molecule (ICAM-1)-1 (1:1000; Sigma-Aldrich/eBioscience 12-0549-42), glial associated fibrillary protein (GFAP; 1:500; AbCam ab7602, Cambridge, UK) and ionized calcium binding adaptor molecule 1 (Iba-1)+ microglia (Iba-1; 1:500; AbCam; ab178847) with a cresyl violet counterstain. Nonspecific binding was blocked using 10% serum from the species in which the secondary antibody was raised (diluted in PBS) for 1 h at room temperature (RT) and incubated with the primary antibody in PBS at 4 °C overnight. Secondary antibodies were biotinylated (1:100; VectorLabs, UK) and applied for 1 h at RT, followed by visualization using the avidin-biotin-peroxidase method (1:100, 1 h at RT, VectorLabs, Peterborough, UK) using 3,3′-diaminobenzidine to visualize the stain. Quantification of immunohistochemical staining was performed using a modified version of a previously published protocol [26]. Four representative coronal sections were selected per brain, and the striatum was imaged bilaterally (ipsilateral and contralateral) at 20× magnification using ImageScope software (version 12.4.6.5003) and a Leica Aperio Slide Scanner (Leica Biosystems, Milton Keynes, UK). Images were processed in ImageJ (FIJI version 2.14.0), where color deconvolution was applied to separate the DAB signal from cresyl violet (when both stains were present). A square region of interest (ROI) with a fixed diameter of 500 ImageJ units was placed within the striatal region of each image; the location of the quantified region is indicated in the inset panel of Figure 6A. The ‘Measure’ function was used to quantify average staining intensity and percentage area stained (reported in arbitrary units). This analysis was focused on a defined ROI within the striatum rather than the entire hemisphere or brain, and therefore, the percentage values reported reflect the area stained within that specific ROI. This technique was applied because ICAM-1 staining is not specific to the vasculature but can be present on microglia as well, and as such, traditional cell counting methods are not necessarily appropriate. We, therefore, used the same technique across all staining moieties.

qPCR: RNA was extracted using the RNEasy Mini Kit (Qiagen, Manchester, UK) with QiaShredders. cDNA was generated using the High-Capacity cDNA Kit (Applied Biosystems, Warrington, UK) and diluted to a final concentration of 5 ng/µL. A total of 25 ng of cDNA was used in the quantitative PCR reaction with PrecisionPLUS SYBR green master mix (Primer Design Ltd., Southampton, UK). Primer sequences (5′-3′) were as follows: GAPDH F: AACGACCCCTTCATTGAC R: TCCACGACATACTCAGCAC; IL-1β F: CAACCAACAAGTGATATTCTCCAT R: GGGTGTGCCGTCTTTCATTA; CXCL-1 F: GCAGACAGTGGCAGGGATT CXCL-1 R: GTGGCTATGACTTCGGTTTGG; ICAM-1-1 F: CAATTTCTCATGCCGCACAG R: AGCTGGAAGATCGAAAGTCCG; VCAM-1 F: TGAACCCAAACAGAGGCAGAGT R: GGTATCCCATCACTTGAGCAGG.

Statistical analysis: Statistical analyses were performed using GraphPad Prism 9.0 software. All data were tested for normality, and post hoc tests were adjusted accordingly. Results were considered statistically significant at p < 0.05.

5. Conclusions

The functionality of EVs, especially in vivo, is an under-researched area [9,10]. This is, in part, due to the technical challenges associated with isolating and measuring EVs and, in part, due to the nature of EV release under physiological conditions. In a stroke, for example, EVs are likely to be released into the circulation during the acute phase, but during the recovery phase, the pathophysiological processes in the CNS are going to be different, resulting in a different EV population. This means that the in vivo responses we get to isolated populations are, given current technological limitations, only ever going to present us with a snapshot of the role of EVs in ongoing pathological processes. In future experiments, using devices that administer EVs more slowly (such as osmotic mini-pumps) is likely to reflect the gradual release and uptake of EVs more accurately but may still fail to reflect the complex temporal dynamics of EVs in vivo. Despite these technical limitations, the data in this paper demonstrates clearly that the origins of EVs in pathology affect not only their physicochemical composition but also the downstream tissue-level inflammatory effects they have on other cells. Establishing the mechanisms that underlie these processes is vital to our long-term understanding of how pathological events, such as a stroke, can affect the brain.