N,N-dimethyltryptamine mitigates experimental stroke by stabilizing the blood-brain barrier and reducing neuroinflammation

Abstract

N,N-dimethyltryptamine (DMT) is a psychoactive molecule present in the human brain. DMT is under clinical evaluation as a neuroprotective agent in poststroke recovery. Yet, its mechanism of action remains poorly understood. In a rat transient middle cerebral artery occlusion stroke model, we previously showed that DMT reduces infarct volume. Here, we demonstrate that this effect is accompanied by reduction of cerebral edema, attenuated astrocyte dysfunction, and a shift in serum protein composition toward an anti-inflammatory, neuroprotective state. DMT restored tight junction integrity and blood-brain barrier (BBB) function in vitro and in vivo. DMT suppressed the release of proinflammatory cytokines and chemokines in brain endothelial cells and peripheral immune cells and reduced microglial activation via the sigma-1 receptor. Our findings prove that DMT mitigates a poststroke effect by stabilizing the BBB and reducing neuroinflammation. Such interactions of DMT with the vascular and immune systems can be leveraged to complement current, insufficient, stroke therapy.

The protective effects of N,N-dimethyltryptamine on the blood-brain barrier and microglia can be helpful in stroke therapy.

INTRODUCTION

Stroke is one of the most devastating diseases, imposing a serious long-term burden on patients, their families, and society, with an inherent high direct and indirect health care cost (1). Despite the development of numerous promising neuroprotective agents, none have successfully transitioned into routine clinical use (2). Stroke pathogenesis involves a complex network of interacting cells, pathways, and feedback loops. The resulting symptoms are frequently associated with long-term structural and functional impairment in the brain (1). Attempting to improve stroke outcomes by targeting a single pathogenic component or cell type has proven difficult and insufficient (3).

Stroke-induced damage is a dynamic process that directly affects elements of the neurovascular unit. Upon the stroke event, the activation of microglia, astrocyte, pericyte, and brain microvascular endothelial cells leads to the production of various proinflammatory cytokines and chemokines initiating neuroinflammation (4, 5). As a consequence, disruption of the blood-brain barrier (BBB) occurs with the consequent extravasation of blood-borne molecules into brain parenchyma (6, 7). For example, the wave of infiltrating peripheral leukocytes causes further BBB dysfunction and creates an inflammatory milieu of cytokines and chemokines that ultimately results in the exacerbation of neuroinflammation, edema, and secondary injury (8).

Current stroke therapy involves intravenous thrombolysis and endovascular thrombectomy, but these interventions are time restricted, not suitable for all patients and may result in serious complications (1). Treatment of poststroke brain edema and neuronal death is insufficient with a low number of clinical trials (1–3).

N,N-dimethyltryptamine (DMT) is found in various plants and animals, including humans, serving different biological functions (9, 10). DMT acts as a natural endogenous agonist of the sigma-1 receptor (11), through which it displays substantial versatility in modulating multiple physiological systems in mammals, such as mitochondrial function, cell survival, and proliferation (12). When administered exogenously, it has a complex and profound impact on human consciousness due to its interaction with serotonin, glutamate, and sigma receptors, each with varying affinities (13). Previously, we demonstrated that DMT administration reduces infarct volume and improves functional outcomes in rats with transient focal cerebral ischemia (14). These results have led to a phase 1 clinical trial (NCT05559931) investigating single and repeated doses of intravenously administered DMT in healthy individuals and the design of a phase 2 trial aimed at studying DMT as an enhancer of neuroplasticity and its effect on long-term recovery of patients with acute stroke.

Despite the advancement of DMT into a clinical trial to help the poststroke recovery of patients, there is still a knowledge gap in understanding the mechanism by which it mitigates the poststroke effects. Since BBB dysfunction is a central element in poststroke brain damage, coculture BBB models that allow cell-cell interactions are valuable to reveal the mechanism of potentially protective agents (15). In this work, we aim to investigate the effects and mode of action of DMT treatment on the BBB under and following ischemic conditions. We used both a rat model of transient focal brain ischemia and a complex in vitro coculture BBB model based on rat primary cells. We hypothesize that, as a pleiotropic psychoactive molecule due to its large interactome including the BBB and the peripheral immune system, DMT has a great therapeutic potential to treat neurological diseases with neuroinflammation. Here, we show that DMT reduces stroke infarct volume which is accompanied by reduction of cerebral edema, attenuated astrocyte dysfunction, and a shift in serum protein composition toward an anti-inflammatory, neuroprotective state. DMT restored tight junction integrity and BBB function both in vitro and in vivo. Last, DMT suppressed the release of proinflammatory cytokines and chemokines from brain endothelial cells and peripheral immune cells and reduced microglial activation, all in a sigma-1 receptor–dependent manner.

RESULTS

DMT reduces BBB disruption in a rat stroke model

In our well-established transient filament occlusion stroke model in rats, used in series of earlier experiments (14, 16), we apply 1 hour of transient middle cerebral artery occlusion (tMCAO) followed by 24 hours of reperfusion. Animals with a blood perfusion drop of at least 40% were randomized into treatment arms (fig. S1A). This setting is characterized by a cerebral lesion size adequate for investigations, minimizing unnecessary animal mortality associated with longer occlusion periods. It closely replicates the realistic scenario of an ischemic stroke in humans but comes with relatively high SDs in infarct volumes due to anatomical variations in brain vasculature collateralization among the animals, necessitating a higher number of animals for accurate results compared to alternative experimental models (12). To evaluate the impact of DMT treatment on ischemic stroke, we measured the volume of infarct lesions and assessed the extent of edema using Nissl-stained histology slices (Fig. 1A). Treatment with DMT resulted in a significant reduction in infarct volume and the associated edema compared to the control stroke group (Fig. 1, B and C).

Fig. 1. Effect of DMT on infarct size and barrier integrity in a rat model of stroke.

(A) Representative images of coronal brain sections from a DMT-treated rat, stained with cresyl violet, are shown sequentially from cranial to caudal sections. The infarct area is delineated by a yellow dotted line. (B and C) Comparison of brain infarct volumes and edema between the stroke group (n = 10) and the DMT-treated stroke group (n = 9). DMT treatment reduced both infarct volume (**P = 0.0327) and associated edema (**P = 0.0135); mean ± SD, unpaired t test. (D) Treatment and procedure timeline. (E) Representative magnetic resonance imaging (MRI) images of rats from the stroke group, DMT-treated stroke group, and DMT + BD1063 cotreated group (n = 10 rats per group) are shown in sagittal, axial, and coronal planes. (F) Representative images show green fluorescence, indicating FITC-albumin extravasation into the brain parenchyma. (G) Quantification of FITC-albumin extravasation in the control and injured brain hemispheres in the stroke, DMT-treated stroke, and DMT + BD1063–treated stroke groups (n = 7 rats per group). DMT reduced FITC intensity compared to the stroke group (***P = 0.0093). Cotreatment with DMT + BD1063 also decreased intensity relative to stroke (*P = 0.011) and showed no difference from the DMT treatment group (P = 0.4254); mean ± SD, two-way analysis of variance (ANOVA) with Bonferroni’s post hoc test. No change was observed in the noninjured hemispheres. AU, arbitrary units. (H) Quantification of FITC-albumin extravasation to the CSF by fluorescence spectroscopy (n = 5 rats per group). FITC intensity was lower in the DMT-treated group (*P = 0.0454) but not in the DMT + BD1063 cotreatment group compared to the stroke group (P = 0.3821).

On the basis of the histology results, small animal magnetic resonance imaging (MRI) was performed. To study the mechanism of action of DMT, a group of animals received a specific sigma-1 receptor antagonist, BD1063 treatment alongside DMT (Fig. 1D). T2 MRI lesion volumes (Fig. 1E) were significantly lower in the DMT-treated group compared to the control stroke group, confirming previous results reported earlier (14). BD1063, added to the DMT treatment, antagonized the effect of DMT suggesting a sigma-1 receptor–dependent neuroprotective effect (Fig. 1E).

In the light of our earlier results (14) and the lesion volume measurements above, DMT appears to be neuroprotective in stroke. To directly assess the BBB integrity in relation to edema and infarct volume reduction, fluorescein isothiocyanate (FITC)–albumin was injected to specific groups of animals after 23 hours of reperfusion followed by 1 hour of dye circulation before collecting the specimens for analysis (Fig. 1D). We measured the intensity of FITC-albumin in the brain parenchyma using histological slices (Fig. 1F and fig. S1B). The results demonstrated significantly reduced FITC-albumin extravasation in the DMT-treated group compared to the control stroke group (P = 0.0093). However, cotreatment with BD1063 did not mitigate the effect of DMT (Fig. 1G).

We also measured FITC-albumin in cerebrospinal fluid (CSF) samples by fluorescence spectroscopy. A significantly reduced FITC-albumin extravasation was found into the CSF in the DMT-treated group compared to the stroke group. There was a trend for BD1063 to counteract the effect of DMT, but it was statistically not significant (Fig. 1H and fig. S1C). These findings suggest that DMT has the capability to protect the BBB in stroke by improving its integrity. BD1063 could attenuate but not completely offset the effect of DMT, which suggests that the sigma-1 receptor is not the only pathway through which DMT exerts its beneficial effects.

DMT attenuates oxygen-glucose deprivation–induced barrier opening in a rat coculture model

To investigate the individual cell response to DMT rat primary brain endothelial cell, pericyte and glial cell cultures were studied. The effect of DMT was tested first in normoxia on brain endothelial cells in the concentration range of 1 to 300 μM (fig. S2, A and B). The change in the impedance of brain endothelial monolayers was not influenced by DMT, and only the 300 μM concentration showed a 5% decrease compared to the control. The oxygen-glucose deprivation and reoxygenation (OGDR) model was optimized on the cell types of the BBB in monoculture. Real-time impedance-based analysis was used to monitor the kinetics of the cell response to OGDR (Fig. 2A). After 6-hour OGD, the cells had a 24-hour reoxygenation period in the presence of DMT or DMT + BD1063.

Fig. 2. Effect of OGDR and DMT on a rat primary BBB coculture model.

(A) Experimental setup for impedance measurements. Brain endothelial cells: green, brain pericytes: orange, glial cells: blue. After 6-hour OGD, 24-hour reoxygenation was administered in the presence of DMT or DMT + BD1063. (B) Real-time impedance kinetics of rat primary brain endothelial cell, brain pericyte, and glial cell cultures and area under the curve values of cell index (0 to 30 hours); n = 5 to 13; mean ± SD, one-way ANOVA and Bonferroni’s post hoc test, ****P < 0.0001. (C) Experimental setup for barrier integrity measurements on the coculture model of the BBB in normoxia and OGDR. (D) TEER measurements after 24-hour reoxygenation. Mean ± SEM, n = 8. (E) Permeability for sodium fluorescein (SF) and (F) Evans blue–labeled albumin (EBA) across the BBB model after 24-hour reoxygenation. P_e, endothelial permeability coefficient. Mean ± SD, n = 4; one-way ANOVA with Bonferroni’s post hoc test, #P < 0.05, ##P < 0.01, and ###P < 0.0003 compared to the normoxia group. (A and C) Created with Biorender.com. Vigh, J. (2025) https://BioRender.com/9atsvr6https://BioRender.com/9atsvr6.

The cellular response for OGDR was different in each cell type (Fig. 2B). Oxygen-glucose deprivation (OGD) decreased the impedance of brain endothelial cells to 40%, but the impedance returned close to the baseline values (90%) during the 24-hour reoxygenation period. Glial cells were also sensitive to the OGD, and impedance values dropped to 63%, but after a quick peak, this cell type could not fully recover as indicated by the 71% value at the end of the monitoring period of 30 hours. Brain pericytes were the most sensitive to the OGD. The impedance of this cell type declined to 14% and, after the reoxygenation values, increased to 70%. The area under the curve values of the impedance expressed as cell index also supported that pericytes were the most sensitive cell type to OGDR, followed by brain endothelial cells and astrocytes which showed a much better recovery after OGD. The effect of DMT during the reoxygenation was also examined, but there was no change in the impedance of cell monolayers after the 24-hour reoxygenation (fig. S2, C to H).

OGDR decreased the metabolic activity measured by 3-(4,5-dimethyltiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay in brain endothelial cells and pericytes (fig S3). This effect was rescued by DMT at 100 and 200 μM concentrations (fig. S3A). For further experiments, the 200 μM concentration of DMT was selected. In brain endothelial cells, the decrease in metabolic activity after OGDR was blocked by DMT (fig. S3B) but not in pericytes (fig. S3C).

To examine the barrier integrity of the BBB model, a triple coculture was used (Fig. 2C). Transendothelial electrical resistance (TEER) was significantly decreased by OGDR (Fig. 2D), while the permeability for the small fluorescein increased by 2.4-fold (Fig. 2E) and albumin by 2.5-fold (Fig. 2F) indicating a barrier impairment. DMT significantly inhibited the increase in the transcellular albumin permeability but did not influence the changes in the paracellular pathway, namely, resistance and fluorescein penetration values (Fig. 2, D to F). BD1063, however, could not reverse the barrier protective effect of DMT (Fig. 2F).

DMT protects BBB integrity via tight junction protein claudin-5

A significant decrease (P < 0.0001) in tight junction protein claudin-5 (CLDN5) immunofluorescence intensity was observed in brain capillaries on the injured side compared to the intact side following induction of focal cerebral ischemia (Fig. 3A). DMT treatment prevented the loss of CLDN5 staining, as there was no significant intensity reduction between the injured and the intact hemispheres in the DMT-treated group. When BD1063 was administered alongside DMT, a decrease in CLDN5 immunofluorescence intensity was observed on the injured side compared to the intact side (P < 0.0001). It was similar to that seen in the vehicle-treated group, indicating that BD1063 had an antagonistic effect on DMT action via the sigma-1 receptor (Fig. 3B).

Fig. 3. DMT protects BBB integrity.

(A) CLDN5 immunofluorescent staining in brain sections. (B) Quantification of CLDN5 immunostaining. Mean ± SD, one-way ANOVA with Bonferroni’s post hoc test, ####P < 0.0001 compared to the control, uninjured side group, ****P < 0.0001 compared to the stroke group; n = 5 to 15. (C) CLDN5 immunofluorescent staining in cell culture. (D) Quantification of CLDN5 immunostaining. Mean ± SD, one-way ANOVA with Bonferroni’s post hoc test, #P = 0.0399 and ##P = 0.0058; compared to the normoxia group, n = 13 to 17. (E) AQP4 and GFAP immunofluorescent double staining in brain sections. (F) Quantification of AQP4 immunostaining. Mean ± SD, one-way ANOVA with Bonferroni’s post hoc test, ##P < 0.005 compared to the control side group, n = 5 to 15. (G) Quantification of GFAP immunostaining. Serum ELISA of (H) CLDN5, (I) MMP9, and (J) GFAP proteins. n = 9 rats per group, mean ± SD, one-way ANOVA with Bonferroni’s post hoc test, ##P < 0.01, ###P < 0.001, and ####P < 0.0001 compared to the sham group, *P < 0.05, **P = 0.0076, and ****P < 0.0001 compared to the stroke group.

A similar effect was seen in cultured brain endothelial cells (Fig. 3C), and the intensity of the CLDN5 staining was decreased after OGDR. There was no change between the control and the OGDR + DMT treatment group, while the staining intensity in the OGDR + DMT + BD1063 group was reduced (Fig. 3D). Next, we measured the serum levels of circulating CLDN5 and found a significant increase in stroke (Fig. 3H) indicating the enzymatic cleavage of brain endothelial junctions, which were protected by DMT treatment (P < 0.0001) but not antagonized by BD1063. The serum levels of the matrix metalloproteinase-9 (MMP9) enzyme increased in the stroke group (Fig. 3I). In the DMT group, serum MMP9 levels were significantly reduced to the level of the vehicle-treated stroke group (P = 0.0076). The effect of DMT on serum MMP9 level was not changed by BD1063 treatment (Fig. 3I).

On the basis of its prominent role in the cytotoxic edema of astrocytes and disruption of the BBB, we performed aquaporin-4 (AQP4) immunohistochemistry (Fig. 3E). Stroke led to a significant increase in the immunofluorescence intensity of AQP4 in the peri-infarct regions of the ipsilateral hemisphere compared to the contralateral side (P = 0.0033). In the DMT-treated group, there was no significant difference in AQP4 immunofluorescence intensity between the injured and the control sides indicating a protective effect. Following coadministration of BD1063 and DMT, an increase in AQP4 immunofluorescence intensity was detected on the injured side (P = 0.0011), which was similar to the changes seen in the vehicle-treated stroke group (Fig. 3F) implying that the effects of DMT on AQP4 expression can be mediated by the sigma-1 receptor.

In patients with acute stroke, there is a strong positive correlation between astrocyte-derived serum glial fibrillary acidic protein (GFAP) levels and various clinical parameters, such as infarct size (17–19), severity of neurology deficits in the acute phase (19), as well as long-term outcomes at 3 months (18) and 1 year (20). In our immunohistochemical experiments, we focused on astrocytes surrounding brain capillaries (Fig. 3E). In contrast to the glial end-feet marker AQP4, GFAP, a cytoskeletal protein labeling the entire astrocytic cell, showed no significant difference in immunofluorescence intensity between the injured and the control hemispheres in either experimental group (Fig. 3G). In cultured glial cells, OGDR increased the staining intensity of GFAP that was not significantly modified by DMT (fig. S4, A and B).

In addition, serum GFAP levels were measured to examine astrocytic and BBB dysfunction (Fig. 3J). In the stroke group, serum GFAP levels were elevated fourfold that was inhibited by DMT treatment (P < 0.0001). The effect of DMT was antagonized by the addition of BD1063 (DMT + BD1063 versus DMT: P = 0.0179), and no significant difference was seen in serum GFAP levels compared to the vehicle-treated stroke group (DMT + BD1063 versus vehicle: P = 0.0586). Thus, DMT treatment may alleviate the consequences of focal cerebral ischemia in a sigma-1 receptor–dependent manner.

DMT induces genes related to an anti-inflammatory effect and suppresses cytokine release in brain endothelial cells after OGDR

To investigate gene expressional changes in primary brain endothelial cells from cocultures upon DMT treatment, we performed massive analysis of cDNA ends (MACE-seq). Principal components analysis showed a clear separation of all OGDR-receiving samples and the normoxia group along principal component 2, indicating profound transcriptomic changes upon OGDR (Fig. 4A). Nine hundred forty-two genes were down-regulated, and 1071 genes were up-regulated by OGDR treatment alone compared to the normoxia group (Fig. 4B). On a pathway level, OGDR induced a transcriptomic signature associated with cellular stress, extracellular matrix remodeling, and cell proliferation (Fig. 4, C and D).

Fig. 4. The anti-inflammatory effects of DMT in cultured brain endothelial cells after OGDR.

(A) Principal components analysis (PCA) of normoxia and OGDR groups. (B) Volcano plot of the OGDR versus normoxia comparison. Log₂FC, log₂(fold change); FDR, false discovery rate. (C) Pathway analysis showing down-regulated and (D) up-regulated Gene Ontology (GO) terms and transcriptions factor (TF) binding sites (TF) in the OGDR group compared to normoxia. HIF-1α, hypoxia-inducible factor-1α; BP, biological process; MF, molecular function; CC, cellular component. (E) Volcano plot showing the effect of DMT and (F) the addition of BD1063 in OGDR on the brain endothelial transcriptome. The schematic figure above (G) shows the BBB coculture model and the luminal and abluminal fluid compartments. (G to N) Levels of TNFα, IL-1β, IL-6, and IL-10 cytokines in supernatant samples from the luminal and abluminal compartments of the rat triple coculture BBB model. n/d, not detected; n.s., not significant. n = 3 to 6, mean ± SD, one-way ANOVA with Bonferroni’s post hoc test, #P < 0.05, ##P < 0.01, ###P < 0.001, and ####P < 0.0001 compared to the normoxia group; ***P < 0.001 and ****P < 0.0001 compared between the OGDR groups.

Only 19 transcripts were significantly altered by the addition of DMT (OGDR + DMT) compared to OGDR alone. Differentially expressed genes include the down-regulation of Gbp2 and Cuedc2, which are involved in sensing inflammation, and the up-regulation of genes associated with an anti-inflammatory effect (Lrrc14 and Gdf15) and glucose metabolism (Flcn, Uap1l1, and Insig1; Fig. 4E). These genes were not affected by the addition of BD1063 (OGDR + DMT + BD1063) compared to the OGDR + DMT group (Fig. 4F). Together, these results suggest that while DMT might have an anti-inflammatory effect in brain endothelial cells, it has little effect on brain endothelial gene expression overall.

To investigate the inflammatory response to OGDR in the BBB model, we measured the cytokine profile of the cell supernatant from both the luminal and abluminal compartments (Fig. 4, G to N) representing the blood and brain sides of the BBB. In the normoxia group, the selected four cytokines could not be detected, except for interleukin-6 (IL-6) in the luminal compartment at a very low level (Fig. 4K). OGDR resulted in significantly higher levels of the proinflammatory cytokines tumor necrosis factor–α (TNFα), IL-1β, IL-6 (Fig. 4, G to L), and the anti-inflammatory IL-10 cytokine (Fig. 4, M and N) in supernatants from both compartments compared to the normoxia group. In the samples from the abluminal compartments, higher levels of proinflammatory cytokines were detected (Fig. 4, G to N) that can be explained by cytokine release from all three types of cells including astrocytes and microglia cells. DMT treatment showed a protective effect against the elevated cytokine levels, except for IL-10 (Fig. 4, G to N). The effect of DMT was antagonized by BD1063 in the case of proinflammatory TNFα, IL-1β, and IL-6 (Fig. 4, G to L), suggesting that these effects were mediated by the sigma-1 receptor.

All these results indicate that DMT exerts an anti-inflammatory effect in the BBB coculture model, including brain endothelial cells, but this cell type is not the main target based on the MACE-seq data. To further investigate the mechanism of the DMT-induced protection in the stroke model, we measured cytokine and chemokine levels in plasma, the expression of cytokine and chemokine genes in peripheral blood mononuclear cells (PBMCs), and the morphology of microglia cells in the brain.

DMT suppresses proinflammatory cytokine levels in stroke

After ischemic brain damage, brain-generated danger-associated molecular patterns (DAMPs) and cytokines enter the circulation through a dysfunctional BBB provoking peripheral innate immune responses (21). This peripheral immune activation aims to eliminate the potential threat through a massive and indiscriminate humoral and cellular inflammatory reaction (22).

We measured serum cytokine levels associated with BBB disruption to evaluate the peripheral innate immune responses related to stroke and DMT treatment. To investigate the inflammatory response to stroke, we included a sham-operated control group. Stroke resulted in significantly higher serum levels of the proinflammatory TNFα, IL-6, CXCL-1, and CXCL-10 (Fig. 5, A, C, E, and F) and significantly lower level of the anti-inflammatory IL-10 (Fig. 5D) compared to the sham group. DMT administration resulted in significantly lower serum levels of TNFα, IL-1β, IL-6, CXCL-1, CXCL-10 and significantly higher serum level of IL-10 compared to the control group, indicating a protective effect (Fig. 5, A to F). BD1063 treatment, when administered parallel to the DMT, could antagonize the effect of DMT on the levels of TNFα, IL-1β, IL-6, and IL-10 (Fig. 5, A to D), but it could not reverse the effects on the levels of circulating chemokines (Fig. 5, E and F). Further, DMT treatment resulted in significantly higher levels of the neuroprotective brain-derived neurotrophic factor (BDNF) in serum compared to the control group, and this effect was also neutralized by BD1063 treatment (Fig. 5G).

Fig. 5. The anti-inflammatory effects of DMT in a rat unilateral stroke model.

(A to G) Serum levels of cytokines TNFα, IL-1β, IL-6, IL-10, CXCL-1, CXCL-10, and BDNF measured by ELISA. (H to M) Expression of genes of cytokines TNFα, IL-1β, IL-6, IL-10, and chemokines CXCL-1 and CXCL-10 in PBMCs measured by quantitative polymerase chain reaction (qPCR). n = 9 rats per group, mean ± SD, one-way ANOVA with Bonferroni’s post hoc test. #P < 0.05, ##P < 0.01, ###P < 0.001, and ####P < 0.0001 compared to the sham group; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 compared between stroke groups.

To reveal the potential source of proinflammatory mediators in the blood, we also analyzed the PBMC gene expression profiles for selected cytokines and chemokines. The stroke group showed significantly elevated levels of Tnfα, Il-6, Cxcl-1, and Cxcl-10 gene expression levels compared to the sham group (Fig. 5, H, J, L, and M). DMT treatment reversed this process for Il-6, Tnfα, Cxcl-1, and Cxcl-10 gene expression levels, making them comparable to the expression levels observed in the sham group, suggesting an anti-inflammatory effect on PBMCs. DMT treatment resulted in significantly lower Il-10 gene expression levels compared to both the sham-operated and the stroke groups (Fig. 5K). PBMC Il-1β gene expression levels were similar in all treatment groups (Fig. 5I) hinting that PBMCs are not the source of the elevated serum IL-1β levels after stroke. BD1063 treatment parallel to the DMT could not antagonize the effect of DMT on PBMCs (Fig. 5, H to M) implying that these effects are not mediated by the sigma-1 receptor.

DMT suppresses microglia activation in stroke

Microglial response to stroke can also be tracked by distinct morphological changes. There is a decline in branching, accompanied by a retraction of branches and an increase in cell body size. Regarding microglia function, in the first week after focal cerebral ischemia, microglia are proinflammatory, while later, these cells usually exhibit prorepair activities. However, the microglia population is heterogeneous, i.e., there are microglia cells with distinct morphology and function at all points of the timeline (23, 24).

To assess microglial activity in the context of DMT treatment for stroke, we used ionized calcium-binding adaptor molecule 1 (IBA1) immunohistochemistry analyzing both immunostaining intensity (Fig. 6A) and microglia morphology (Fig. 6C). Upon evaluating the IBA1 immunofluorescence intensity, a significant increase was detected on the injured side in each experimental group, when compared to the intact side. DMT treatment had no effect on IBA1 immunofluorescence intensity in the stroke group that was the highest in the DMT + BD1063 cotreated group (Fig. 6B). In cultured glial cells, OGDR also increased the staining intensity of IBA1 in the microglial cells; however, that was significantly decreased by DMT and counteracted by BD1063 (fig. S4C).

Fig. 6. DMT helps to preserve microglia morphology in a rat unilateral stroke model.

(A) IBA1 immunofluorescent microglia cells in brain sections in all experimental groups. (B) Quantification of IBA1 immunofluorescence intensity. n = 5 to 15 images per group, mean ± SD, one-way ANOVA with Bonferroni’s post hoc test, ##P < 0.01 and ####P < 0.0001 compared to the control side, **P < 0.01 compared to the stroke + DMT group. (C) Morphology of IBA1 immunolabeled microglia cells in brain sections in all experimental groups. (D) Ratio of extensively ramified, homeostatic (cluster 0, gray), moderately ramified, intermediate (cluster 1, green), and less ramified, highly responsive (cluster 2, black) microglia detected in brain sections in all experimental groups. n = 5 animals per group; n = 10 images per side per animal; n = 553 cells per group.

For the morphological analysis of individual microglia cells from the peri-infarct regions, we focused on features such as fractal dimension, lacunarity, density, span ratio, maximum span across hull, area, perimeter, circularity, maximum/minimum radii, and mean radius. On the basis of the analysis of these parameters, the following three clusters were formed: cluster 0, comprising highly ramified, homeostatic microglia that survey their close environment; cluster 1, containing intermediate, less ramified microglia that respond to changes in their microenvironment; and cluster 2, consisting of the least ramified microglia which respond to the injury (Fig. 6C). Cluster 0 microglia were prevalent on the intact side in every experimental group, and some cluster 1 microglia were also observed (fig. S5). Following the induction of focal cerebral ischemia, a robust morphological change was seen: Cluster 2 microglia appeared in great numbers on the injured side, followed by some cluster 1 microglia, while cluster 0 microglia were barely present. In the DMT-treated group, an increase in the ratio of cluster 1 and cluster 0 microglia and a decrease in the ratio of cluster 2 microglia were seen on the injured side, compared to the stroke group. In the third experimental group, where animals were cotreated with BD1063, the distribution of microglia clusters on the injured side resembled that of the stroke group, indicating an inhibition of the protective effect via the sigma-1 receptor (Fig. 6D).

DISCUSSION

Despite decades of extensive research into the pathomechanisms of stroke, achieving meaningful positive clinical outcomes in the field of neuroprotection continues to pose a major challenge. In ischemic stroke, beyond direct neuronal damage, there is a spatially and temporally well-characterized cascade of events that result in BBB disruption. The initial step is the activation of brain endothelial cells, pericytes, and glial cells, followed by the release of proinflammatory mediators, the opening of the BBB, and brain extravasation of albumin and other toxic plasma proteins (6, 7). These steps further increase BBB disruption together with the infiltrating white blood cells and further promote neuroinflammation, neuronal injury, and lesion expansion (4, 25, 26).

Finding treatments that can protect not only neurons but also preserve BBB integrity is critical for improving stroke outcomes. The BBB is considered as a therapeutic target in stroke, and BBB protective molecules and strategies are widely investigated (27). Sigma-1 receptor agonists, such as PRE-084 (28), as well as antagonists, like E-52862 (29), are neuroprotective in ischemic stroke, although further studies are needed to dissect the complex interplay and overlapping effects of activators and inhibitors in stroke pathology (30). Sigma-1 receptor agonists act by protecting against BBB dysfunction, a key step for reducing neuroinflammation and preserving neurocognitive functions (30). Among the endogenous agonists of the sigma-1 receptor, DMT was shown to decrease spreading depolarization and support neuronal and astroglial survival in a rat model of acute global forebrain ischemia (31), while in our previous study, it reduced infarct volume and improved functional outcomes in a rat model of transient focal cerebral ischemia (14).

The present results strengthen the observations that DMT exerts a neuroprotective effect in rats with focal cerebral infarction, as evidenced by decreased infarct size measured by both histological analysis and MRI. Moreover, it extends these data by proving the BBB protective and anti-inflammatory effects of DMT in the brain and immune-competent cells (Fig. 7). DMT significantly attenuated stroke-associated edema and decreased BBB permeability based on a reduction in albumin extravasation into both brain parenchyma and CSF. The fact that the impact of DMT on decreasing edema was very pronounced suggests that its protective effects may be linked more closely to maintaining BBB integrity than solely preventing neuronal damage. This is particularly important, as BBB disruption is a key driver of secondary damage following ischemic stroke.

Fig. 7. BBB changes in stroke and the protective effects of DMT.

Schematic representation of the neurovascular unit (NVU) in physiological condition and in stroke. Arrows and inhibition signs (cyan) show the effect of the DMT in pathological condition. In stroke, tight junctions (red line between brain endothelial cells) are disrupted, and CLDN5 tight junction protein is decreased; in the blood soluble neurovascular unit, proteins appear, and the level of the inflammatory cytokines increases and that of anti-inflammatory cytokines decreases. In the brain, the morphology of microglia changes, the ratio of less ramified microglia cells increases, the AQP4 distribution changes, and the level of inflammatory cytokines increases. DMT decreases the level of soluble CLDN5, MMP9, and GFAP proteins and inhibits inflammatory cytokines in the blood, while it increases the level of BDNF and the anti-inflammatory cytokine IL-10. In the brain, DMT inhibits the morphological changes in microglia and the redistribution of AQP4 in astroglia and decreases proinflammatory cytokine production. Created with Biorender.com. Vigh, J. (2025) https://BioRender.com/o95e437.

In the culture model, brain pericytes were the most sensitive to OGD, and in contrast to brain endothelial cells which showed a good recovery by impedance measurement, they remained highly affected at the end of the reoxygenation period. This result is in line with previous works showing either brain pericyte constriction and death in ischemia in vivo (32) or that, in a BBB culture model, brain pericytes act as drivers of dysfunction in OGDR (15). While DMT did not influence the decreased impedance after OGDR in individual cell types, it rescued the drop in metabolic activity in brain endothelial cells but not in pericytes. As shown by resistance measurements, permeability assays, and CLDN5 immunostaining in the coculture BBB model, the barrier integrity was compromised under stroke-like conditions. The increase in albumin permeability, which showed the greatest change in the barrier integrity assays was inhibited by DMT significantly, in accordance with the results obtained on the in vivo stroke model. Although not significant, a trend was seen for the changes in the paracellular pathway, namely, resistance and fluorescein permeability.

In the in vivo stroke model, in parallel with functional measurements of BBB disruption, a decrease in the immunostaining of CLDN5, the dominant tight junction protein at the BBB, and an increase in AQP4 staining intensity as well as its redistribution were seen on brain sections, without change in the glial marker GFAP. AQP4 is a marker of astrocytic end-feet (33), and its expression increases following ischemia, leading to astrocyte cytotoxic swelling (34, 35). In the stroke group, an increase in the serum level of soluble neurovascular unit proteins CLDN5, GFAP, and MMP9 were measured which were blocked by DMT (Fig. 7). Clinical data demonstrate that, at the acute phase of stroke, an increase in serum markers associated with BBB disruption, like MMP9 (36), GFAP (18) and CLDN5 (37), determines the long-term clinical outcome and can also differentiate ischemic stroke in patients (38), which highlight the importance of the present results on BBB protection.

Since brain endothelial cells are key neurovascular unit components primarily responsible for BBB integrity, we analyzed gene expression in this cell type. OGDR induced a robust change in the transcriptomic signature that was associated with cellular stress, extracellular matrix remodeling, and cell proliferation. In the OGDR + DMT group, only a few differentially expressed genes were identified. Compared to the OGDR group, two genes involved in sensing inflammation were down-regulated, and four genes associated with an anti-inflammatory effect and glucose metabolism were up-regulated, in the OGDR + DMT group. Those were not changed by BD1063, meaning that this effect was not mediated by the sigma-1 receptor. These data indicate that DMT exerts an anti-inflammatory effect on brain endothelial cells without a major impact on the global expression of brain endothelium–related genes. The anti-inflammatory effect of DMT was verified by assessing the cytokine profile of cell supernatants from the luminal and abluminal compartments of the BBB model. OGDR robustly elevated the levels of proinflammatory cytokines TNFα, IL-1β, and IL-6 in supernatants from both compartments. Such effect was inhibited by DMT in a sigma-1 receptor–dependent manner. The higher levels of proinflammatory cytokines in the abluminal compartment of the BBB model can be explained by the contribution of brain pericytes, astrocytes, and microglia cells, as it was demonstrated in primary cell-based BBB coculture models by inflammatory stimulation (39). OGDR caused a modest elevation in the level of anti-inflammatory IL-10 cytokine which was not affected by DMT.

The anti-inflammatory effect of DMT was proven on the BBB coculture model; however, the MACE-seq data indicated that brain endothelial cells may not be the main target of DMT. To further investigate the mechanism of the DMT-induced neuroprotective phenomenon in our stroke model, we measured cytokine and chemokine levels in serum, the expression of their corresponding genes in PBMCs, and the morphology of microglia cells in the brain. In the pathomechanism of stroke, DAMPs, cytokines, and chemokines produced by the brain cross the compromised blood-brain and blood-CSF barriers interacting with the peripheral immune system and activating systemic innate immunity (40, 41). Cytokines and chemokines are also secreted by the cells of neurovascular unit, which is boosted in inflammatory conditions, regulated by cellular cross-talk and are actively transported from blood to brain (39). Peripheral innate immune cells are rapidly activated through leukocyte pattern recognition receptors (40), leading to profound changes in their gene expression profiles, particularly those involved in activation and differentiation (42, 43). This is followed by an increased production of inflammatory mediators and elevated serum cytokine levels within hours after ischemia (22, 44). Such cascade leads to a wave of peripheral immune cells invading the infarcted area, exacerbating neuroinflammation, and causing secondary lesion expansion. A series of clinical data supports that the extent of the activation of the peripheral immune system is related to brain infarct volume, stroke severity, and long-term outcomes (45–50).

In accordance with preclinical and clinical data from literature, levels of proinflammatory cytokines TNFα, IL-1β, and IL-6, as well as the chemokines CXCL-1 and CXCL-10 in the serum, measured 24 hours after stroke, were significantly increased, which were inhibited by DMT. Furthermore, DMT treatment resulted in a significantly elevated serum level of the anti-inflammatory cytokine IL-10 and the neuroprotective and proneuroplastic BDNF (Fig. 7). In patients with stroke, reduced serum IL-10 levels have been independently linked to neurological worsening (51) and poor functional outcomes (52), while elevated serum BDNF levels have been associated with better prognosis (53) and a lower risk of poststroke cognitive impairment (54) due to its primary neurotrophic activity. Associated to the reduced serum inflammatory cytokine levels after DMT treatment, we identified the reduced gene expression of Il-6, Tnfα, Cxcl-1, and Cxcl-10 in PBMCs, as a possible source. DMT also reduced the gene expression of IL-10.

Microglia cells which have cross-talk with all the cell types in brain parenchyma respond to early signals in stroke (6). These include the release of DAMPs and adenosine triphosphate (55, 56), secretion of inflammatory cytokines (57), and release of glutamate from neurons (58). Microglia start to produce proinflammatory cytokines and MMP9 within hours after stroke onset (59). Furthermore, the leakage of blood components through the compromised BBB serves as an additional proinflammatory signal, inducing a self-amplifying inflammatory response (60, 61). These events lead to morphology change in microglia cells, namely, less ramified cell shape linked to more reactive phenotype, that we could also demonstrate in the stroke model. DMT effectively blocked this change and helped to preserve the highly ramified morphology that is associated with homeostatic functions of microglia.

In the present study, we demonstrated that DMT exerts a protective effect on the BBB integrity in stroke through a series of consistent in vivo and in vitro results. It is shown that DMT exerted most of its protective effects via sigma-1 receptors and cells of the neurovascular unit, previously unknown therapeutic targets in stroke. The addition of the receptor antagonist BD1063 neutralized the effect of DMT on stroke lesion volume, in vivo and in vitro inflammatory cytokine levels, serum BDNF and GFAP levels, as well as on the intensity of CLDN5 and AQP4 immunostaining and microglial morphology. However, in vivo BBB permeability, serum levels of chemokines, CLDN5 and MMP9, and the gene expression profile of PBMCs were not affected. Regarding the potential target cell types of DMT, in addition to the cells of the BBB, we identified PBMCs and microglia which were strongly protected by DMT in the stroke model.

Together, our findings show that DMT, an endogenous ligand of the sigma-1 receptor, is not only neuroprotective but also has complementary antineuroinflammatory and BBB-protective effects. These outcomes of DMT effect on poststroke conditions manifest via smaller stroke lesion volume, less edema, and better-preserved BBB integrity in rats after tMCAO. Furthermore, DMT enhanced brain endothelial cell viability and barrier function in the coculture BBB model. Upon DMT treatment, rats with stroke showed better-maintained tight junctions, lower serum CLDN5, MMP9, and GFAP levels, reduced AQP4 overexpression, and elevated BDNF serum levels. DMT also decreased inflammatory cytokine and chemokine release, both in vivo and in vitro, enhanced anti-inflammatory gene induction in brain endothelial cells, suppressed inflammatory cytokine and chemokine gene expression in PBMCs, and reduced microglia activation. The inhibitory effect of DMT on neuroinflammation was mediated by the sigma-1 receptor. DMT and other sigma-1 receptor agonists via their interaction with the vascular and immune systems have the potential to complement current insufficient poststroke therapy due to their neuroprotective, central, and peripheral anti-inflammatory and BBB-protective effects.

MATERIALS AND METHODS

Animal model and treatment protocol

All procedures were conducted in accordance with the ARRIVE guidelines and the guidelines set by the European Communities Council Directive (86/609 EEC and 2010/63/EU) and approved by the Animal Care and Use Committee of the Semmelweis University (XIV-I-001/ 29-7/2012) and the Government Office of Pest County, Hungary, Department of Animal Health (PE/EA/01445-6/2022). All measurements were blinded to the treatment. Humane endpoints were established in accordance with the IMPROVE (Ischaemia Models: Procedural Refinements Of in Vivo Experiments) Guidelines (62). The personnel conducting the data analysis were blinded to the treatment groups.

The transient MCAO model was applied on male Wistar rats [Toxi-Coop Ltd., Budapest, Hungary, body weight (bw): 280 ± 20 g] under continuous isoflurane anesthesia (Harvard Apparatus, Holliston, MA, USA). Following the surgical exposure of the right internal carotid artery, we positioned the suture while monitoring the cerebral blood flow over the territory of the right middle cerebral artery with Laser-Doppler flowmetry (Perimed Inc., Stockholm, Sweden). Animals with a perfusion drop of at least 40% were randomized for the treatment arms. Ischemia was maintained for 60 min (16).

DMT (Sigma-Aldrich, St. Louis, MO) treatment started immediately before the removal of the filament by an intraperitoneal injection of DMT (1 mg/kg bw) that was dissolved by 0.1 ml of 70% ethanol and diluted to 1 ml with saline solution. Following this step, a continuous infusion of DMT (2 mg/kg bw per hour) was given using intraperitoneally placed osmotic pumps (Alzet, ALZA, Palo Alto, CA) for 24 hours. Rats in the control group received a vehicle bolus only without any DMT treatment. For the sigma receptor antagonist group, animals received 1-(3,4-dichlorophenethyl)-4-methyl-piperazine dihydrochloride (BD1063, 2 mg/kg bw per 24 hours; Axon Medchem BV, Groningen, the Netherlands) in parallel with the DMT via separate osmotic pumps, following an intraperitoneal bolus of DMT and BD1063 (1 mg/kg bw each). Dosing was determined on the basis of previous in vivo experiments (14). For the cytokine and PBMC tests, sham-operated groups were included with preparation of an incision over the right carotid artery and opening the abdominal cavity just like in treated animals.

To acquire blood and brain specimens, animals were anesthetized intraperitoneally with a mixture of ketamine (50 mg/kg bw) and xylazine (4 mg/kg bw). Blood was taken after cannulation of the aorta. For PBMC separation, mononuclear cell preparation tubes (BD Vacutainer CPT, product number: 362782) were used. Brains were fixed via transaortic perfusion with 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer. The fixed brains were immersed in a 30% glucose solution for 2 days, after which free-floating coronal sections with a thickness of 40 μm were prepared using a microtome (Leica SM2000R, Nussloch, Germany).

Lesion volume and edema measurement

Cresyl violet staining (Nissl staining) was performed according to the well-known method: (63). Briefly, the procedure involves immersing the sample in a solution containing cresyl violet (0.1 g), glacial acetic acid (0.2 ml), and distilled water (DW; 100 ml) with a pH of 3.2 for 90 min. Following this, the sample is rinsed in 96% ethanol, dehydrated, and cleared using isopropanol and xylene. Last, the sample is mounted and covered.

Volumes derived from histological slices were calculated using the following formula: sum of region of interest cross-sectional areas × number of slices × distance between the sections. Edema was calculated according to the following: (volume of the ipsilateral hemisphere − volume of the contralateral hemisphere)/volume of the contralateral hemisphere. Edema-corrected lesion volume was calculated as follows: infarct volume × (1 − volume of the edema) (64, 65). Slides were scanned on a Zeiss Axioscan 7 Microscope Slide Scanner (Karl Zeiss AG, Oberkochen Germany). QuPath v0.5.1 was used for image processing. For the MRI lesion volume measurement protocol, see (14).

BBB permeability in vivo

FITC-albumin visualization and quantification in CSF

To evaluate the integrity of the BBB, 21 animals (n = 7 per group) were receiving 1 ml of (20 mg in 1 ml of 0.9% NaCl) FITC-albumin (Sigma-Aldrich, Saint Louis, MO, USA, product number: A9771) via the lateral tail vein 1 hour before termination. After 1 hour of circulation, CSF was taken from the cisterna magna with the help of a stereotaxic instrument (Harvard Apparatus, Holliston, MA, USA and Hamilton high-performance liquid chromatography syringe 100 μl, Hamilton Company, Reno, NV, USA), and then brains were fixed via transcardiac perfusion with 4% PFA in 0.1 M phosphate buffer (66). From the fixed brains, free-floating sections (40 μm) were made. To detect FITC-albumin in brain sections, images were acquired on a 780LSM confocal laser-scanning microscope (Karl Zeiss AG, Oberkochen, Germany), and the fluorescence intensity of the images was expressed in arbitrary units. In the case of the control side, two images per animal (n = 14 images per group) were taken, and in the case of the injured side, 10 images per animal (n = 210) were taken. The amount of FITC-albumin in CSF samples (n = 5 per group) was quantified by fluorescence spectroscopy at 495-nm excitation and at 520-nm emission wavelengths using a Multiskan Sky microplate reader (Thermo Fisher Scientific, Waltham, MA). Fluorescence intensity values were normalized to the signal of the injected FITC-albumin stock solution, which was defined as 100%.

RNA isolation, cDNA synthesis, and quantitative PCR

Total RNA was isolated from PBMC pellets using TRIzol reagent (Invitrogen, Waltham, MA) according to the manufacturer’s protocol. The yield was measured by ultraviolet photometry on a NanoDrop One instrument (Thermo Fisher Scientific, Waltham, MA). RNA integrity was checked on an Agilent Bioanalyzer (Agilent Technologies, Santa Clara, CA), and samples with RNA integrity number > 8 were included in further analyses. One microgram of RNA from each sample was used to generate first-strand cDNA using the High Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). Targeted gene expression measurements were performed using specific TaqMan Real-Time PCR Assays (Thermo Fisher Scientific). The ΔCt method was used for normalization, and PPIA, the gene of cyclophilin A, was set as reference gene. Statistical analysis was performed using GraphPad Prism software, and a nonparametric Wilcoxon paired test was applied to compare the different conditions and identify significantly different expression patterns. Differences were considered significant at P < 0.05.

Assessment of circulating biomarkers

Cytokine, chemokine, and neurotrophin levels in rat serum and culture supernatant samples were measured using OptEIA kits (BD Biosciences, Franklin Lakes, NJ) and high-sensitivity enzyme-linked immunosorbent assay (ELISA) assays (for IP-10/CXCL10 and CINC-1/CXCL1; Thermo Fisher Scientific), following the manufacturer’s recommendations. The precision of the kits was the following: intra-assay variation: coefficient of variation (CV) < 10%; inter-assay variation: CV < 12% (CV% = SD/mean × 100).

Cell cultures

Primary rat brain endothelial cells (RBECs), rat brain pericytes (RPCs), and glial cells were isolated and cultured according to previous protocols (67, 68). Briefly, after isolation from 4-week-old outbred Wistar rats from both sexes, RBECs were seeded onto collagen type IV (100 μg/ml; Sigma-Aldrich, catalog no. C5533) and fibronectin (25 μg/ml; Sigma-Aldrich, catalog no. F1141)–coated culture dishes (Corning Costar, New York, NY, USA). Culture medium was based on a Dulbecco’s modified Eagle’s medium (DMEM)/F-12 medium (Gibco, Waltham, MA, USA), supplemented with plasma-derived bovine serum (15% in the first 3 days and then 10%, plasma derived serum, First Link, Wolverhampton, UK), 10 mM Hepes, heparin (100 μg/ml), insulin (5 μg/ml), transferrin (5 μg/ml), sodium selenite (5 ng/ml) (Insulin-Transferrin-Selenium, Gibco), basic fibroblast growth factor (1 ng/ml), and gentamicin (50 μg/ml) at 37°C in a humidified incubator with 5% CO₂. To eliminate the contaminating cell types which are P-glycoprotein negative, RBEC medium contained puromycin (3 μg/ml) during the first 3 days of culture (68). After isolation, pericytes were seeded onto culture dishes (VWR International, Radnor, PA, USA) coated with collagen type IV (100 μg/ml). Glial cells were isolated from newborn rats and were plated onto uncoated 75-cm² flasks (TPP, Trasadingen, Switzerland). The cultures contained ~80% astrocytes and ~20% of microglial cells as described in our previous study (69) and as shown in fig. S4A. Both pericytes and glial cells were cultured in DMEM medium (low glucose, Gibco, subsidiary of Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Pan-Biotech, Aidenbach, Germany) and gentamicin (50 μg/ml).

OGDR model

To model hypoxia, 6-hour OGD and 24-hour reoxygenation were performed (Fig. 2C). The timing of the model was made similarly to the study of Sato et al. (15). A special O₂/CO₂ incubator (MCO-50M multigas incubator, PHC Europe, Breda, The Netherlands) was used to generate the 1% O₂ and 5% CO₂ level. The oxygen was bubbled out with N₂ gas from the serum and glucose-free, GlutaMAX (Gibco) supplemented cell culture media (DMEM/F-12, Biowest, catalog no. L0091) before the 6-hour OGD. Control cells were kept under control conditions in a humidified incubator with 5% CO₂ and the basic, previously described cell culture medium during the 6 + 24-hour treatment. The control group is referred to as “normoxia” (Fig. 2C). At the start of the 24-hour reoxygenation, 200 μM DMT with or without 2 μM BD1063 was added to the cell culture medium.

Impedance-based measurement of endothelial cells, pericytes, and glial cells

Real-time impedance measurement–based cell analysis was performed using special 96-well plates with integrated gold electrodes (E-plate 96; Agilent, Santa Clara, CA, USA, catalog no. 300600910) using an xCELLigence RTCA SP device (Agilent Technologies). Plates were coated with collagen type IV (100 μg/ml) and fibronectin (25 μg/ml) for the RBECs and collagen type IV (100 μg/ml) for the RPCs and glial cells. A cell-free background was measured for each well in culture medium. Then, 5 × 10³ RBECs per well, 6 × 10³ RPCs per well, and 6 × 10³ glial cells per well were seeded in the plates and grown until confluency. The impedance of the cell layers was measured at 10 kHz every 10 min. Confluent RBEC monolayers were treated with 550 nM hydrocortisone for 24 hours before the OGDR treatment to enhance barrier properties. The readout of impedance measurements using this device is the cell index, which correlates with the strength of cell-cell contacts and cell adhesion. Cell index is defined as Z_n − Z₀, where Z_n is the impedance at a certain time point and Z₀ is the impedance of the cell-free background in each well. Cell index values were normalized to the last time point before the OGDR treatment (70).

Barrier integrity tests

Triple coculture model

To examine the barrier integrity, 8.5 × 10⁴ RBECs and 1.5 × 10⁴ RPCs were cocultured on cell culture inserts (Corning, catalog no. 3460) with 2.5 × 10⁴ glial cells on the bottom of the plate for 5 days in control conditions, and then OGDR treatment was performed. At the start of the 24-hour reoxygenation, 200 μM DMT with or without 2 μM BD1063 was added to the normal cell culture medium which was used during the culture. After the OGDR treatment, the supernatant was collected and centrifuged at 700g 5 min to remove the cell debris. Until further use, the supernatants were stored at −20°C. The RBECs were fixed with ice-cold acetone:methanol solution, 1:1 ratio for 2 min, and were immediately rehydrated with 1% FBS containing phosphate-buffered saline (PBS). Glial cells were fixed with 4% PFA for 15 min at room temperature (RT) for immunocytochemistry.

TEER measurement

To examine the sodium ion permeability, TEER was measured by an EVOM Volt-ohm meter (World Precision Instruments Inc., Sarasota, FL, USA). STX-2 chopstick electrodes were used. The resistance of the inserts without cells was also measured as a background TEER which was subtracted from the measured values. Raw ohm data were expressed relative to the surface area of the endothelial monolayer (ohm × square centimeters) as described in our earlier publications (70).

Permeability measurement

After the 24-hour reoxygenation, barrier integrity was examined by the permeability measurement of two fluorescent dyes, sodium fluorescein [SF; 10 μg/ml, molecular weight (MW) = 376 Da] and Evans blue–labeled albumin (EBA; 170 μg/ml, MW = 67 kDa) in Ringer-Hepes buffer [118 mM NaCl, 4.8 mM KCl, 2.5 mM CaCl₂, 1.2 mM MgSO₄, 5.5 mM d-glucose, and 20 mM Hepes (pH 7.4)] as described previously (70). The concentration of the SF and EBA marker molecules in samples from the lower compartment was determined by a Fluorolog 3 spectrofluorometer (Horiba Jobin Yvon). Passage across cell-free inserts was also measured. Endothelial permeability coefficients (P_e) were calculated from clearance values of tracers as described in detail in our previous publication (67).

Cell culture immunocytochemistry

CLDN5 immunocytochemistry

The fixed RBECs were washed three times with PBS and blocked with 3% bovine serum albumin (BSA)–PBS for 1 hour at RT. The anti-CLDN5 primary antibody (for detailed antibody information, see table S1) was diluted in the blocking solution and was incubated overnight (ON) at 4°C. After the washing step (three times, PBS), the A488 anti-rabbit secondary antibody and Hoechst 33342 nucleus dye diluted in PBS were used for 1 hour at RT. After washing three times in PBS followed by rinsing in DW, the membranes were cut out and mounted on glass slides with Fluoromount-G (Southern Biotech, AL, USA, catalog no. 0100-01) and examined with a confocal microscope.

GFAP-IBA1 immunocytochemistry

The GFAP-IBA1 immunocytochemistry was performed similarly to the CLDN5 immunostaining. Because glial cells were fixed with PFA, 0.2% Triton X-100 (TX) was used to permeabilize the cell membranes for 10 min at 4°C. After the blocking step, anti-GFAP and anti-IBA1 primary antibodies were used ON, at 4°C. Anti-mouse A555 and anti-goat A488 secondary antibodies were added for 1 hour at RT. After the last washing step, cells were mounted with Fluoromount-G.

Cell culture immunocytochemistry image analysis

CLDN5 and GFAP-IBA1 fluorescent images were taken with the Leica TCS SP5 confocal laser scanning microscope (Leica Microsystems, Wetzlar, Germany). To analyze the intensity of the staining, images were taken with the same setup (n = 13 to 20 images per group). For image evaluation, the ImageJ software (National Institutes of Health, Bethesda, MD, USA) was used.

MTT cytotoxicity assay

RBECs and RPCs were cultured in 96-well plates (Corning). Cells were seeded at 5 × 10³ cells per well density. When the cell layers were confluent, OGDR treatment was performed. After the OGDR, the medium was removed completely, cells were treated with MTT (0.5 mg/ml; Sigma-Aldrich, M5655) diluted in serum and phenol red-free, DMEM/F-12 medium (Gibco) for 3 hours in the CO₂ incubator. After crystal formation, medium was flicked out from the plate, and the blue crystals were dissolved in dimethyl sulfoxide (Sigma-Aldrich/Merck; 100 μl per well). Absorbance was measured at a wavelength of 595 nm. MTT reduction of normoxia wells was used as 100% viability.

Total RNA isolation

RBECs in the triple coculture model were grown on Transwell inserts (Corning, catalog no. 3460) for 5 days. After OGDR (6-hour OGD + 24-hour R), RPCs were removed, and RBECs were lysed directly with buffer RLT Plus (QIAGEN, catalog no. 1053393) on the inserts liked in our previous article (70). RNA was isolated using the RNeasy Plus Micro Kit (QIAGEN, catalog no. 74034) containing an integrated genomic DNA eliminator spin column according to the manufacturer’s protocol. RNA integrity was analyzed using automated capillary electrophoresis (RNA Pico Sensitivity Assay, LabChip GX II Touch HT instrument, PerkinElmer). Samples were stored at −80°C until further analysis.

Library preparation and 3′ RNA sequencing

Genome-wide gene expression profiling was performed using RNA sequencing at GenXPro GmbH, Frankfurt, Germany. For library preparation, samples with 1 μg of purified RNA were used; a total of 20 libraries were constructed. Fragmented RNA was reverse transcribed into cDNA, using barcoded oligo(dT) primers containing TrueQuant unique molecular identifiers, followed by template switching. Library amplification was done using polymerase chain reaction (PCR), purified by solid-phase reversible immobilization beads (Agencourt AMPure XP, Beckman Coulter, catalog no. A63882). Sequencing was performed on an Illumina NextSeq 500 platform.

Bioinformatic analysis of RNA sequencing data

Approximately 20 million single 75-bp reads were obtained per library. Unprocessed sequencing reads were adapter-trimmed and quality-trimmed using Cutadapt (version 3.4). FastQC was used to assess the quality of sequencing reads. Cleaned reads were mapped to the reference genome using STAR (version 2.7.10a), allowing for spliced alignments. Quantification of mapped reads to each gene was performed using STAR and RSEM (version 1.3.3) softwares, and quantification of transcript abundance was also performed using RSEM. MultiQC (version 1.12) was used to create a single report visualizing output from multiple tools across many samples, enabling global trends and biases to be quickly identified. Testing for differential gene expression was performed using the DESeq2 R/Bioconductor package. As a result, log₂FC [log₂(fold change)] and P values were obtained for each gene in the dataset. To account for multiple comparisons, false discovery rate (FDR) was calculated using the Benjamini-Hochberg method. Genes with an FDR < 0.05 and log₂FC > 0.415 or log₂FC < −0.415 were considered to be differentially expressed. To perform functional enrichment analysis, the g:GOSt tool in g:Profiler was used to identify overrepresented Gene Ontology terms.

Brain section immunohistochemistry

CLDN5 immunostaining

For the immunostainings, three brain sections per animal and five animals per group were used (free-floating sections, 40 μm, total of 15 sections per group). Brain sections were washed three times for 10 min with PBS on a horizontal shaker. For cell membrane permeabilization, 98°C water bath was used for 1 hour, while the samples were in PBS. After the permeabilization, brain sections were cooled down to RT. The blocking step was performed with 2% normal horse serum (NHS), 0.3% BSA, and 0.5% TX-PBS for 2 hours at RT. Anti-CLDN5 primary antibody was diluted in the blocking solution and was kept on the samples for ON, at 4°C. The next day, after the washing step (three times in PBS), the A594 anti-rabbit secondary antibody in PBS was incubated on the samples for 2 hours at RT. After this step, sections were washed three times with PBS, and the cell nuclei were stained with Hoechst 33342 for 15 min at RT. The last PBS washing was followed by a single washing step in DW. The sections were mounted on glass slides with Confocal Matrix (Micro Tech Lab, Graz, Austria) and examined with a confocal microscope.

GFAP-AQP4 immunostaining

The brain sections were washed between the steps the same as in the case of the CLDN5 immunostaining. Samples were permeabilized with 1% TX-PBS for 10 min. After the washing step, 2% NHS, 0.3% BSA, and 0.5% TX-PBS blocking solution was used, and the goat anti-GFAP was added ON at 4°C. In the case of the double staining, the primary antibodies were used in two steps. On the second day, anti-goat A488 was added for 2 hours at RT. After the next washing step, the rabbit anti-AQP4 primary antibody was used (ON, 4°C). On the third day, sections were washed, and the anti-rabbit A594 secondary was used for 2 hours at RT. Then, all samples were washed, and cell nucleus staining was performed for 15 min at RT. After the last washing step, samples were mounted similarly to the CLDN5 mounting step.

IBA1 immunostaining

The brain sections were washed between the steps as for the previous immunostaining. Samples were permeabilized in two steps. First, hot 0.01 M citrate buffer for 20 min in a 70°C water bath was used. Then, the brain sections were washed with PBS and permeabilized more with 0.5% TX-PBS for 30 min. After the washing step, 2% NHS, 0.3% BSA, and 0.5% TX-PBS blocking solution was used, and the goat anti-IBA1 was added ON at 4°C. On the second day, anti-goat A488 was added for 2 hours at RT. After the next washing step, the A488 anti-goat secondary was used for 2 hours at RT. Following the washing steps, nucleus staining was performed for 15 min at RT. After the last washing step, samples were mounted similarly to sections stained for CLDN5. All antibodies, dyes, and solutions are listed in table S1.

Confocal microscopy

Immunolabeling for CLDN5 was recorded in single slice images with a confocal laser scanning microscope (Olympus Fluoview FV1000, Olympus Life Science Europa GmbH, Germany), while GFAP, AQP4, and IBA1 immunolabeling was detected using a spinning disk confocal microscope (Zeiss, Germany) in stack images containing 15 to 20 slices. The confocal laser scanning microscope configuration was the following: objective lens: UPLSAPO, 60×; numeric aperture (NA), 1.35; sampling speed, 4 μs/pixel; scanning mode, sequential unidirectional; excitation, 405 nm (4′,6-diamidino-2-phenylindole) and 543 nm (Alexa Fluor 594). The configuration of the spinning disk confocal microscope was as follows: objective lens: PlanApoN, 60×; NA, 1.42; camera binning, 1; Z distance, 0.3 μm; excitation, 405 nm (Hoechst 33342), 488 nm (Alexa Fluor 488), and 543 nm (Alexa Fluor 594).

Analysis of microglia morphology

Individual microglia cell morphology was studied according to the work of Fernández-Arjona et al. (71). The IBA1 immunostained sections were screened using the Spinning Disk Confocal Microscope to identify the area affected by the experimental stroke. The corresponding contralateral brain area was used as control. Images (n = 10 in the control or stroke side, respectively, in each animal) were captured by an expert who was blind to the experimental groups. All images (n = 300 from three experimental groups, each containing five animals) were recorded. Together, 553 individual microglia cells were identified in the control and stroke sides in each experimental group. Single-cell images were extracted in Photoshop CS4 (Adobe, San Jose, CA, USA), and contrast/brightness was set to a level which resulted maximal cell structure visibility. Images were then converted to binary structures using the ImageJ program and further corrected in Photoshop CS4 to represent all microglia processes as continuous lines.

Various features describing microglia morphology, such as fractal dimension, lacunarity, density, span ratio, maximum span across hull, area, perimeter, circularity, max/min radii, and mean radius were extracted from the cell profiles with the use of the Box Counting Fractal Analysis method using the FracLac plugin in ImageJ (24). Data were normalized and then processed with Weka Explorer (72) data mining software package. As a final step, data were divided into three different clusters with the expectation-maximization clustering algorithm (73), and their distribution was examined in the experimental groups.

Brain section immunohistochemistry image analysis

The original images were segmented to create images containing one single microvessel, microglia, or astrocyte, respectively. The resulting images were analyzed with MATLAB (MathWorks, Natick, MA, USA). First, the contrast of each image was enhanced via histogram equalization. Then, the grayscale images were converted to binary, and the small objects (<4 pixels) were eliminated. The pixel number of the binary images represented the area of the labeled structures, while the total intensities were determined via masking the grayscale images with the corresponding binary images. Each total intensity was normalized to the corresponding area.

Serum CLDN5, MMP9, and GFAP ELISA

Concentrations of CLDN5, MMP9, and GFAP in rat blood serum were quantified using rat CLDN5 (Elabscience, catalog no. E-EL-R2502), MMP9 (Elabscience, catalog no. E-EL-R3021), and GFAP (Elabscience, catalog no. E-EL-R1428) ELISA kits. Briefly, blood serum was collected using EDTA-Na₂ as an anticoagulant. Samples were centrifuged for 15 min at 1000g at 2° to 8°C within 30 min of collection. Samples from nine animals per group were tested with two technical parallels, with 100-μl sample per well for each assay. After the addition of samples to the wells, the assay was carried out according to the manufacturer’s instructions. The Multiskan EX (Thermo Fisher Scientific, catalog no. 51118170) microplate reader was used to measure the absorbance at 450 nm.

Statistical analysis

Data were presented as mean ± SD. One-way or two-way analysis of variance (ANOVA) followed by Bonferroni’s post hoc test were used. Data analysis was performed using GraphPad Prism 8.1 Software (GraphPad Software Inc., La Jolla, CA). Differences were considered statistically significant at P < 0.05.

Supplementary Materials

The PDF file includes:

Figs. S1 to S5

Table S1

Legend for data S1

Other Supplementary Material for this manuscript includes the following:

Data S1