Research ArticleNeuroscience

Endothelium-targeted deletion of the miR-15a/16-1 cluster ameliorates blood-brain barrier dysfunction in ischemic stroke

See allHide authors and affiliations

Science Signaling  07 Apr 2020:
Vol. 13, Issue 626, eaay5686
DOI: 10.1126/scisignal.aay5686

Breaking down barriers

Stroke decreases the function of the microvascular endothelial cells in the blood-brain barrier. The resulting immune cell infiltration and solute leak exacerbate the neuronal death caused by ischemic injury. Ma et al. found that endothelial cell–specific ablation of the miR-15a/16-1 microRNA cluster protected mice from ischemic brain damage. Analysis of cultured microvascular endothelial cells showed that the miR-15a/16-1 cluster targeted the gene encoding the tight junction protein claudin-5, which is critical for the barrier function of these cells. These results suggest that antagonizing this microRNA cluster may prevent the breakdown of the blood-brain barrier that occurs after stroke.

Abstract

The blood-brain barrier (BBB) maintains a stable brain microenvironment. Breakdown of BBB integrity during cerebral ischemia initiates a devastating cascade of events that eventually leads to neuronal loss. MicroRNAs are small noncoding RNAs that suppress protein expression, and we previously showed that the miR-15a/16-1 cluster is involved in the pathogenesis of ischemic brain injury. Here, we demonstrated that when subjected to experimentally induced stroke, mice with an endothelial cell (EC)–selective deletion of miR-15a/16-1 had smaller brain infarcts, reduced BBB leakage, and decreased infiltration of peripheral immune cells. These mice also showed reduced infiltration of proinflammatory M1-type microglia/macrophage in the peri-infarct area without changes in the number of resolving M2-type cells. Stroke decreases claudin-5 abundance, and we found that EC-selective miR-15a/16-1 deletion enhanced claudin-5 mRNA and protein abundance in ischemic mouse brains. In cultured mouse brain microvascular ECs (mBMECs), the miR-15a/16-1 cluster directly bound to the 3′ untranslated region (3′UTR) of Claudin-5, and lentivirus-mediated ablation of miR-15a/16-1 diminished oxygen-glucose deprivation (OGD)–induced down-regulation of claudin-5 mRNA and protein abundance and endothelial barrier dysfunction. These findings suggest that genetic deletion of endothelial miR-15a/16-1 suppresses BBB pathologies after ischemic stroke. Elucidating the molecular mechanisms of miR-15a/16-1–mediated BBB dysfunction may enable the discovery of new therapies for ischemic stroke.

INTRODUCTION

According to the American Stroke Association, stroke remains one of the leading causes of death and long-term disability in the United States, and more than 87% of all strokes are ischemic (1). Untimely reperfusion after ischemic stroke induces secondary neuronal injury and harmful effects such as hemorrhagic transformation and blood-brain barrier (BBB) dysfunction in most of the stroke patients.

The BBB is a selectively permeable barrier that regulates proteins, metabolites, and ion transport crossing the central nervous system (CNS). Some water-soluble compounds and polar ions that barely cross the BBB are allowed to permeate because of BBB dysfunction (2). In addition, BBB leakage and dysfunction result in infiltration of circulating immune cells and extravasation of plasma components, which is considered a common mechanism in the pathogenesis of many neurological diseases, including ischemic stroke (3). Thus, preserving BBB integrity and normal function should be a valuable therapeutic approach in the treatment of ischemic stroke (4).

Many factors contribute to the physical barrier of the BBB, including cerebrovascular endothelial cells (ECs) and complex tight junctions. Cerebrovascular ECs are the major components of brain microvasculature, playing an essential role in the maintenance of BBB integrity and cerebral homeostasis (5). Tight junctions connect neighboring individual ECs to reduce the intercellular distance, thus preventing free diffusion of water-soluble molecules. The damage or loss of tight junctions results in free paracellular transport through gaps among ECs (6). Ischemia-induced endothelial tight junction injury increases cerebrovascular permeability or BBB leakage, contributing to primary/secondary ischemic brain injury and neurological deficits (710). Claudins are a class of more than 25 different family members that form paracellular barriers and pores and determine tight junction permeability (11). Claudin-5, as the predominant claudin expressed in ECs, plays a critical role in the induction and maintenance of TJ tightness. The human Claudin-5 gene is a 4522–base pair (bp) linear DNA located in chromosome 22q11.21, which comprises two exons and one intron (12). The transcript variant 1 of the corresponding exon is a 2332-bp linear mRNA, whereas the transcript variant 2 is a 1720-bp linear mRNA.

MicroRNAs are small endogenous noncoding RNA molecules that suppress gene expression by binding to the 3′ untranslated regions (3′UTRs) of one or more mRNAs in a sequence-specific manner. Cerebrovascular ECs are enriched with various microRNAs that regulate complex endothelial physiological functions (13). Although microRNAs are involved in regulating BBB integrity (1418), there is little direct evidence or mechanisms of microRNA regulation of BBB dysfunction. miR-15a and miR-16-1 are two highly conserved microRNAs located within a cluster of 54 bp on mouse chromosome 14. They form a structural and functional cluster (the miR-15a/16-1 cluster) by binding to their common mRNA targets. Our previous study has demonstrated that the miR-15a/16-1 cluster regulates mouse brain microvascular EC (mBMEC) apoptosis after oxygen-glucose deprivation (OGD) (19), and vascular EC-selective transgenic overexpression of the miR-15a/16-1 cluster inhibits angiogenesis after hindlimb ischemia (19, 20). Sun et al. demonstrated that the expression of endothelial miR-15a was elevated at transient ischemia or OGD-reoxygenation treatment. EC-selective deletion of the miR-15a/16-1 cluster enhances angiogenesis and improves long-term functional recovery after transient focal cerebral ischemia in mice by directly up-regulating the expression of several proangiogenic factors (21). However, whether endothelial miR-15a/16-1 can directly regulate BBB structure and function is still unclear.

Here, we investigated the effects and molecular mechanisms of endothelial miR-15a/16-1 on BBB dysfunction by using a clinically relevant transient focal ischemia and reperfusion model, middle cerebral artery occlusion (MCAo), in mice and a coculture in vitro BBB model. Our results suggest that endothelial miR-15a/16-1 aggravates BBB pathologies possibly through direct translational repression of Claudin-5 in cerebral microvascular ECs.

RESULTS

EC-selective deletion of the miR-15a/16-1 cluster ameliorates I/R-induced brain BBB damage

Ischemic stroke rapidly leads to a chain of neurochemical events including brain edema and BBB leakage of some water-soluble molecules into brain parenchyma (22). Laser spackle images indicated that wild-type (WT) and conditional knockout (cKO) mice subjected to the same ischemic insult did not show differences in cerebral blood flow (CBF) during and 15 min after ischemia/reperfusion (I/R) (fig. S1, A and B). Water content was measured 24 hours after 1 hour of transient focal ischemia in both MCAo-induced ischemic hemispheres (MCAo) and nonischemic hemispheres (sham) (Fig. 1A). MCAo significantly increased brain water content in both genotypes compared to sham operation; however, brain water content in the ischemic hemispheres of EC-miR-15a/16-1 cKO mice was significantly lower than that of WT controls (Fig. 1B).

Fig. 1 EC-selective deletion of the miR-15a/16-1 cluster decreases brain water content and BBB disruption in ischemic mice.

(A) Experimental design to measure BBB leakage in ischemic mice. (B) WT and EC-miR-15a/16-1 cKO mice were subjected to 1-hour MCAo and then 24-hour reperfusion. Water content was measured in dissected brains. Data are presented as means ± SEM (n = 14 to 15 mice per group). **P < 0.01 compared to sham group, Wilcoxon signed-rank test; ##P < 0.01 compared to WT + MCAo group, Mann-Whitney U test. (C) Representative images demonstrate brain leakage of EB (scale bar, 1 mm). (D) WT and EC-miR-15a/16-1 cKO mice were injected with EB 1 hour before being sacrificed. EB extravasation was measured in dissected brains, and data are presented as means ± SEM (n = 4 mice per group). **P < 0.01 compared to sham group, Wilcoxon signed-rank test; ##P < 0.01 compared to WT + MCAo group, Mann-Whitney U test. (E) Representative images demonstrate the extravasation of Alexa 555 cadaverine (0.95 kDa, red) or endogenous plasma IgG (150 kDa, green) into the brain parenchyma. At 24 hours after 1-hour ischemia, the area with microtubule-associated protein 2 (MAP2) immunofluorescence illustrates the ischemic area in adjacent sections from the same brains. Dashed line, leakage area or ischemic area (scale bar, 1 mm). (F and G) Quantitative analysis of volume of leakage of cadaverine or IgG at 24 hours after cerebral ischemia. Data are means ± SEM (n = 5 to 6 mice per group). *P < 0.05, **P < 0.01 compared to WT group, independent t test.

To assess the extravasation of blood components into brain parenchyma after BBB injury, Evans Blue (EB) was intravenously injected 1 hour before sacrifice (Fig. 1A). Cerebral ischemia induced robust BBB leakage of EB in both EC-miR-15a/16-1 cKO mice and WT controls (Fig. 1, C and D); however, EC-selective deletion of the miR-15a/16-1 cluster significantly reduced I/R-induced EB extravasation. Next, we determined the I/R-induced BBB leakage area by examining the extravasation of injected fluorescent tracer cadaverine (0.95 kDa) and endogenous plasma immunoglobulin G (IgG) (150 kDa) in mouse brains. As expected, I/R induced progressive leakage of these molecules at 24 hours after transient ischemia (Fig. 1, E to G), an effect that was significantly reduced by EC-selective deletion of miR-15a/16-1. The extravasation of plasma IgG reached a similar leakage volume as that of cadaverine at 24 hours after transient ischemia (Fig. 1, F and G), which is consistent with a previous study (10). However, brain regions with robust BBB breakdown at 24 hours after I/R evolved into infarct zones (Fig. 1E, dashed line area). These results suggested that EC-selective deletion of miR-15a/16-1 ameliorated MCAo-induced BBB disruption at 24 hours after transient ischemia.

EC-selective deletion of the miR-15a/16-1 cluster mitigates I/R-induced brain injury and improves post-stroke functional outcomes

Early BBB disruption in ischemic stroke is a key cause of subsequent parenchymal injury and long-term neurological deficits (10). Therefore, we investigated whether endothelial miR-15a/16-1 affected I/R-induced brain infarct size and post-stroke functional outcome. At 24 hours after ischemia, mice with EC-selective deletion of the miR-15a/16-1 cluster had significantly smaller infarct volumes compared with WT littermates (Fig. 2, A and B). According to a previously described five-point scale (23), EC-miR-15a/16-1 cKO mice had lower neurological scores, indicating a better post-stroke functional outcome (Fig. 2C). Quantification of the brain ischemic zones that evolved from BBB breakdown showed that EC-selective deletion of the miR-15a/16-1 cluster significantly reduced the MAP2 (microtubule-associated protein 2) negative-stained ischemic areas compared with WT controls (Fig. 2, D and E). This correlation was confirmed by Pearson coefficient analysis of ischemic volume and the leakage volume of both cadaverine and IgG 24 hours after transient ischemia (Fig. 2, F and G).

Fig. 2 EC-selective deletion of the miR-15a/16-1 cluster reduces brain infarct size and improves neurological outcomes.

(A) 2,3,5-Triphenyltetrazolium (TTC) (2%)–stained coronal sections (1 mm) from WT and EC-miR-15a/16-1 cKO ischemic mice at 24 hours after 1-hour ischemia. (B) Quantification of infarct volume was performed on TTC-stained coronal sections. Data are presented as means ± SEM (n = 4 to 5 mice per group). *P < 0.05, independent t test. (C) Neurological deficits were assessed at 24 hours after 1-hour ischemia. Data are presented as means ± SEM (n = 4 mice per group). *P < 0.05, independent t test. (D) Representative images demonstrate MAP2-immunostained brain sections used to calculate the ischemic volume (MAP2-negative area; scale bar, 1 mm). (E) Quantitative analysis of brain infarct volume on MAP2-stained coronal sections. Data are means ± SEM (n = 6 mice per group). *P < 0.05, independent t test. (F and G) Pearson correlation analysis of ischemic volume and BBB leakage of cadaverine or IgG.

EC-targeted deletion of the miR-15a/16-1 cluster reduces I/R-induced peripheral immune cell infiltration and regulates brain microglial polarization

BBB disruption results in robust infiltration of peripheral immune cells into brain parenchyma and microglial polarization, which facilitates subsequent irreversible BBB damage (24). We hypothesized that deletion of endothelial miR-15a/16-1 may reduce BBB disruption–induced infiltration of blood macrophages and neutrophils and regulate post-stroke microglia polarization. At 48 hours after ischemia, F4/80+ macrophages and Ly6B+ neutrophils were increased in peri-infarcted brain regions compared with sham controls (Fig. 3, A to D), indicating robust infiltration of peripheral macrophages and neutrophils into ischemic brain parenchyma. Compared with WT controls, I/R-induced infiltration of macrophages and neutrophils was statistically significantly lower in EC-miR-15a/16-1 cKO mice.

Fig. 3 EC-selective deletion of the miR-15a/16-1 cluster reduces I/R-induced infiltration of peripheral macrophages and neutrophils and reduces M1-type microglia/macrophage numbers without affecting M2-type numbers.

(A) Representative images from the peri-infarcted areas demonstrate the infiltration of F4/80+ macrophages (green) in sham control and ischemic brain regions (MCAo). Mice were sacrificed at 48 hours after 1-hour transient ischemia. The dotted square indicates the region that is magnified in the images below (scale bar, 50 μm). DAPI, 4′,6-diamidino-2-phenylindole. (B) Representative images from the peri-infarcted area demonstrate the infiltration of Ly6B+ neutrophils (green) in sham and ischemic brain regions at 48 hours after 1-hour transient ischemia. The square indicates the region that is magnified in the images below (scale bar, 50 μm). (C) Quantitative analysis of infiltrated macrophages. Data are presented as means ± SEM (n = 5 to 6 mice per group). **P < 0.01 compared to sham group, Wilcoxon signed-rank test; #P < 0.05 compared to WT + MCAo group, Mann-Whitney U test. (D) Quantitative analysis of infiltrated neutrophils. Data are presented as means ± SEM (n = 5 mice per group). **P < 0.01 compared to sham group, paired t test; ##P < 0.01 compared to WT + MCAo group, independent t test. (E and F) Representative images from the peri-infarcted area demonstrate Iba1- and CD16/32-positive M1-type microglia/macrophages (E) and Iba1- and CD206-positive M2-type microglia/macrophages (F) in sham and ischemic brain regions at 48 hours after 1-hour transient ischemia. The dotted square indicates the region that is magnified in the images below (scale bar, 50 μm). (G and H) Quantitative analysis of M1-type microglia/macrophage number and M2-type microglia/macrophage number (n = 5 mice per group). **P < 0.01 compared to sham group, Wilcoxon signed-rank test; ##P < 0.01 compared to WT + MCAo group, Mann-Whitney U test.

To test whether endothelial miR-15a/16-1 affects microglial polarization, double immunolabeling with Iba1 and either CD16/32 or CD206 was performed to detect M1- or M2-type macrophage/microglia, respectively. At 48 hours after MCAo, increased numbers of M1- and M2-type macrophage/microglia were found in peri-infarcted brain regions (Fig. 3, E to H). Genetic deletion of the EC-miR-15a/16-1 cluster reduced the number of M1-type macrophage/microglia without affecting M2-type cell numbers (Fig. 3, G and H), suggesting that endothelial miR-15a/16-1 may promote M2- to M1-type microglial polarization after transient ischemia.

EC-selective deletion of the miR-15a/16-1 cluster enhances brain claudin-5 expression after transient ischemia

As a predominant component of the BBB, claudin-5 regulates paracellular permeability together with other tight junction proteins (12). As expected, Claudin-5 mRNA expression was significantly reduced at 24 hours after cerebral ischemia (Fig. 4A). This reduction was statistically significantly eliminated in mice with EC-targeted deletion of the miR-15a/16-1 cluster, in which Claudin-5 reached a similar or higher mRNA expression level compared to the sham group. Consistent with the elevated mRNA level, claudin-5 protein abundance was also significantly increased in the ischemic cortex of cKO mice (Fig. 4B). Immunofluorescence analysis confirmed that I/R significantly decreased the expression of claudin-5 in brain ECs, which was attenuated by genetic deletion of endothelial miR-15a/16-1 (Fig. 4, C and D).

Fig. 4 EC-selective deletion of the miR-15a/16-1 cluster enhances ischemic brain claudin-5 expression.

(A) Quantitative PCR analysis of Claudin-5 mRNA expression in cortex and striatum at 24 hours after 1-hour ischemia. Data were normalized to WT sham group and are presented as means ± SEM of five to six animals per group. **P < 0.01 compared to sham group, Wilcoxon signed-rank test; ##P < 0.01 compared to WT + MCAo group, Mann-Whitney U test. (B) Claudin-5 protein abundance was detected by Western blotting and normalized to WT sham group. Data from n = 4 independent experiments were analyzed; ##P < 0.01 compared to WT + MCAo group, Mann-Whitney U test. (C) Representative images demonstrate immunostaining of claudin-5 (green) and CD31 (red) in brain sections at 24 hours after 1-hour transient ischemia (scale bar, 50 μm or 20 μm). (D) Quantitative analysis of claudin-5–positive vessels in mouse brains at 24 hours after 1-hour transient ischemia (n = 5 mice per group). Data are presented as means ± SEM. *P < 0.05, **P < 0.01 compared to sham group, paired t test; ##P < 0.01 compared to WT + MCAo group, independent t test.

miR-15a/16-1 regulates OGD-induced BBB disruption in vitro

To investigate the role of the miR-15a/16-1 cluster in the regulation of BBB functional integrity, we used an in vitro BBB model in which a monolayer of mouse brain microvascular ECs (mBMECs) was cocultured with mouse primary astrocytes (Fig. 5A). mBMECs were subjected for 16 hours to OGD, an ischemia-like insult, and reoxygenated for different periods. Sixteen hours of OGD resulted in a significant decrease of the transendothelial electrical resistance (TEER) of mBMECs at 4 hours after OGD, which was further enhanced by lentivirus-mediated miR-15a/16-1 overexpression (Fig. 5B). Compared with non-OGD cultures, OGD and reperfusion immediately induced an increase in barrier permeability as assessed by paracellular permeability of dextran (3 kDa) from the luminal-to-abluminal side that progressively increased with time (fig. S2, A and B). Lentiviral overexpression of miR-15a/16-1 in mBMECs exacerbated dextran leakage, especially from 2 hours after OGD (Fig. 5C). Conversely, when miR-15a/16-1 was suppressed by lentivirally delivered small hairpin RNAs (miRZip 15a), the loss of TEER in mBMECs was significantly mitigated, but not prevented, and OGD-induced paracellular permeability of dextran in mBMECs was attenuated from 1 hour after OGD (Fig. 5, D and E). Thus, our data demonstrated that lentiviral gain- or loss-of-miR-15a/16-1 function can statistically significantly diminish or improve BBB functional integrity in mBMECs under in vitro ischemic conditions.

Fig. 5 Genetic manipulation of the miR-15a/16-1 cluster changes BBB functional integrity in vitro.

(A) Illustration of in vitro BBB coculture model consisting of mBMECs and mouse primary astrocytes. Monolayers of mBMECs were infected with nonfunctional lentiviral GFP controls (Lenti GFP or miRZip GFP), lentiviral vectors carrying mouse pre-miR-15a (Lenti miR-15a), or lentiviral vectors carrying small hairpin RNAs targeting the mouse pre-miR-15a (miRZip 15a). After 72 hours, mBMECs were subjected to 16 hours of OGD followed by different reoxygenation periods. Transepithelial/transendothelial electrical resistance (TEER) and paracellular permeability of fluorescent dextran (3 kDa) were measured at the designated time points. (B) Quantitative analysis of TEER values in BMECs infected by Lenti GFP or Lenti miR-15a at 4 hours after 16-hour OGD treatment. Data are presented as means ± SEM of three independent experiments (n = 3 independent wells per experiment). **P < 0.01 compared to non-OGD, Wilcoxon signed-rank test; #P < 0.05 compared to Lenti GFP, Mann-Whitney U test. (C) Paracellular permeability was evaluated by measuring the abluminal fluorescence intensity of Alexa Fluor 488 dextran at 0, 0.5, 1, 2, 4, and 24 hours after 16-hour OGD or without OGD. Relative fluorescence unit (RFU) was reported. Data are presented as means ± SEM of three independent experiments (n = 3 independent wells per experiment). *P < 0.05, **P < 0.01 compared to Lenti GFP, independent t test. (D) Quantitative analysis of TEER values in mBMECs infected by miRZip GFP or miRZip 15a at 4 hours after 16-hour OGD treatment. Data are presented as means ± SEM of three independent experiments (n = 3 independent wells per experiment). **P < 0.01 compared to non-OGD, Wilcoxon signed-rank test; #P < 0.05 compared to miRZip GFP, Mann-Whitney U test. (E) Paracellular permeability measurement of mBMECs infected by lentiviral vectors carrying small hairpin RNAs. Data are presented as means ± SEM of three independent experiments (n = 3 independent wells per experiment). **P < 0.01 compared to miRZip GFP, independent t test.

miR-15a/16-1 decreases endothelial claudin-5 expression after OGD

In our animal studies, we showed that the I/R-induced decreased brain claudin-5 was significantly attenuated in miR-15a/16-1 cKO mice. We therefore performed quantitative real-time polymerase chain reaction (PCR) in mBMECs to determine whether genetic manipulation of the miR-15a/16-1 cluster correlated with suppression of its downstream gene Claudin-5 under OGD. Compared with control cells, Claudin-5 mRNA expression was decreased in cells transduced with a nonfunctional control [green fluorescent protein (GFP)–expressing lentivirus] or miR-15a/16-1 overexpressing lentivirus immediately after 16-hour OGD treatment (Fig. 6A). At 24 hours after OGD, Claudin-5 was decreased to a greater extent in mBMECs with miR-15a/16-1 overexpression than in GFP-expressing cells. Consistent with the repressed mRNA level, claudin-5 protein expression in mBMECs was also significantly inhibited after 16-hour OGD treatment, and miR-15a/16-1 overexpression aggravated the loss of claudin-5 protein at 24 hours after OGD (Fig. 6B). Conversely, lentivirus-mediated knockdown of miR-15a/16-1 (miRZip 15a) led to increased claudin-5 abundance at 24 hours after OGD treatment (Fig. 6, C and D).

Fig. 6 Genetic manipulation of the miR-15a/16-1 cluster changes claudin-5 expression in vitro.

(A) Real-time PCR was used to measure Claudin-5 mRNA levels in mBMECs with lentiviral-mediated overexpression of miR-15a/16-1, which were normalized to Cyclophilin amounts and then to the control (Lenti GFP non-OGD) group. Data are presented as means ± SEM of three independent biological replicates. **P < 0.01 compared to non-OGD groups, Mann-Whitney U test; #P < 0.05 compared to Lenti GFP at 24 hours after 16-hour OGD, Mann-Whitney U test. (B) Claudin-5 protein abundance was measured in lentiviral miR-15a/16-1–overexpressing mBMECs by Western blotting and normalized to Lenti GFP non-OGD level. Data are presented as means ± SEM of four independent biological replicates. **P < 0.01 compared to non-OGD groups, Mann-Whitney U test; #P < 0.05 compared to Lenti GFP at 24 hours after 16-hour OGD, Mann-Whitney U test. (C) Real-time PCR was used to measure Claudin-5 mRNA levels in mBMECs with lentiviral-mediated down-regulation of miR-15a/16-1, which were normalized to control group (miRZip GFP non-OGD). Data are presented as means ± SEM of three independent biological replicates. **P < 0.01 compared to non-OGD groups, Mann-Whitney U test; #P < 0.05 compared to Lenti GFP or miRZip GFP at 24 hours after 16-hour OGD, Mann-Whitney U test. (D) Claudin-5 protein abundance was measured in mBMECs with lentiviral miR-15a/16-1 down-regulation and normalized to miRZip GFP non-OGD group. Data are presented as means ± SEM of four independent biological replicates. **P < 0.01 compared to non-OGD groups, Mann-Whitney U test; #P < 0.05 compared to miRZip GFP at 24 hours after 16-hour OGD, Mann-Whitney U test. (E) Representative immunostaining images demonstrate claudin-5 (red) and DAPI (blue) in cultured mBMECs with lentivirus-mediated miR-15a/16-1 overexpression or down-regulation. Scale bar, 20 μm. (F and G) Claudin-5 signal was assessed by quantitative analysis of fluorescence intensity value from two regions per sample (n = 5 to 6 independent samples per group). **P < 0.01 compared to non-OGD groups, Mann-Whitney U test; ##P < 0.01 compared to Lenti GFP or miRZip GFP at 24 hours after 16-hour OGD, Mann-Whitney U test.

Claudin-5 is located at endothelial cell-cell appositions under normal conditions. OGD results in a loss of endothelial claudin-5 including protein degradation and redistribution (25). As expected, our immunostaining data showed that OGD was associated with a decrease in claudin-5 immunofluorescent signal at extracellular cell-cell contact sites in all treatment groups (Fig. 6E) that did not differ between groups at 0 hours after 3 hours of OGD treatment (Fig. 6, E to G). However, at 24 hours after OGD, compared to the GFP-transduced mBMECs, those overexpressing the miR-15a/16-1 cluster showed a significant decrease in claudin-5 abundance, whereas those with knockdown of the miR-15a/16-1 cluster (by miRZip 15a) showed a significant increase in claudin-5 abundance. These results suggest that inhibition of the miR-15a/16-1 cluster may enhance claudin-5 protein abundance in the cerebrovascular endothelial barrier after OGD.

miR-15a/16-1 directly regulates claudin-5 expression in vitro

Claudin-5 regulates the integrity and permeability of the BBB (12). Using bioinformatics analysis, we found a conserved miR-15a/16-1 binding site within the 3′UTR of mouse Claudin-5 (Fig. 7A). To confirm that miR-15a/16-1 cluster can directly down-regulate Claudin-5, dual-luciferase assays were performed in a mouse brain EC line (bEnd.3 cells; Fig. 7B). The Firefly/Renilla dual-luciferase reporter plasmids, fused to either WT Claudin-5 3′UTR or a mutant Claudin-5 3′UTR (in which CUG was mutated to GCC), were transfected into bEnd.3 cells with lentivirus-mediated overexpression or knockdown of miR-15a/16-1. The luciferase activity of the reporter vector containing WT mouse Claudin-5 3′UTR mRNA, but not that of the reporter vector containing the mutated sequence, was significantly reduced by the overexpression of miR-15a/16-1 in bEnd.3 cells (Fig. 7C). Conversely, lentiviral loss of miR-15a/16-1 resulted in increased luciferase activity of vectors containing the WT mouse Claudin-5 3′UTR sequence but not that of vectors containing the mutated sequence (Fig. 7D). These data suggest that miR-15a/16-1 may repress mouse Claudin-5 translation by directly binding to the complementary sequences located in its 3′UTR region.

Fig. 7 miR-15a/16-1 translationally suppresses mouse Claudin-5 in mBMECs.

(A) Partial sequence of mature mouse miR-15a and its binding site in the 3′UTR region of mouse Claudin-5. The miR-15a reporter vector containing cytomegalovirus (CMV)–driven expression of luciferase complementary DNA (cDNA) fused to mouse Claudin-5 WT 3′UTR (Wt. Seq.) or to the miR-15a/16-1 binding site mutant 3′UTR (Mut. Seq.) was constructed and transfected into bEnd.3 cells. bEnd.3 cells were infected with a lentivirus containing pre-miR-15a (Lenti miR-15a) or small hairpin miR-15a (miRZip 15a) for 72 hours before performing luciferase reporter activity assays. (B) Experimental design of the luciferase assay with bEnd.3 cells. (C and D) Quantification of the luciferase activity of the reporter vectors described in (A) in cells with miR-15a overexpression or knockdown. Data are presented as the means ± SEM of four to six independent samples. *P < 0.05 compared to Lentiviral GFP or miRZip GFP group, Kruskal-Wallis H test followed by a Dunn post hoc test.

DISCUSSION

The present study investigated a role of endothelial miR-15a/16-1 in the regulation of BBB function after ischemic stroke in vivo and in vitro. In this study, we demonstrated that endothelium-targeted deletion of the miR-15a/16-1 cluster mitigated BBB leakage, decreased infiltration of peripheral macrophages and neutrophils, and reduced brain water content and brain infarction in stroke mice. Moreover, EC-selective deletion of the miR-15a/16-1 cluster inhibited M1-type microglia/macrophage polarization in the peri-infarct area. These effects might be achieved by up-regulation of Claudin-5. Our findings suggest that EC-selective deletion of the miR-15a/16-1 cluster functions to suppress BBB pathologies after ischemic stroke, which may be relevant for the development of microRNA-based therapies in ischemic stroke.

Stroke-induced breakdown of BBB is associated with angiogenesis, and EC-selective deletion of miR-15a/16-1 enhances the expression of some proangiogenic factors including vascular endothelial growth factor A (VEGFA) as early as 1 day after ischemia (21). Moreover, administration of VEGF within 24 hours from ischemia onset enhances post-stroke angiogenesis and promotes BBB breakdown in rats (26, 27). In the present study, the BBB breakdown at early phase of ischemic stroke (within 24 hours) was associated with infarct volume, which is in line with a previous study, showing that early BBB disruption accurately predicts final brain infarct size in ischemic mice (28). Early BBB damage is a key cause of parenchymal injury after stroke. Leakage of dextran (3 kDa) into the ischemic region can be detected within the first 30 min, and plasma IgG (150 kDa) is detected until 3 hours after ischemia (10). At 24 hours after transient ischemia, the differential distributions of these two molecules disappear, suggesting that the development of BBB disruption mainly happens within the first 24 hours. Consistent with previous findings, cadaverine (0.95 kDa) and IgG (150 kDa) in stroke mouse brains were similarly distributed at 24 hours after cerebral ischemia in the present study. In addition, BBB leakage positively correlated with brain infarction at 24 hours after transient ischemia. For this reason, protection of the BBB in the early phase of ischemic stroke is essential for maintaining brain function, and better understanding of the pathogenesis of BBB leakage is therefore critical to develop restorative therapeutics in ischemic stroke (29).

Although different mechanisms are reported to be involved in the pathogenesis of BBB disruption, down-regulation of endothelial claudin-5 is critical for post-stroke BBB disruption, especially for the extravasation of small macromolecules (≤3 kDa) (10, 12, 30). Brain claudins are directly responsible for determining the permeability of endothelial tight junctions. Claudin-5 is the predominant claudin expressed in ECs and has been considered to be a promising candidate target in the treatment of ischemic stroke (12, 31). The loss of claudin-5 alone is sufficient to cause functional changes in the BBB and induce behavioral changes in normal animals (3234). Mice deficient in Claudin-5 show extravasation of tracers with a molecular weight below 0.8 kDa into the brain, which is normally restricted in WT mice (25). Post-stroke treatment with an miR-150 antagonist, which promotes claudin-5 expression by targeting the mRNA encoding the angiopoietin receptor Tie-2, effectively protects BBB permeability (14). In addition, miR-181-a has been reported to decrease claudin-5 expression by targeting Krüppel-like factor 6 (35). In contrast, our data suggest that miR-15a/16-1 may regulate BBB function by directly targeting claudin-5.

Several microRNAs, such as miR-155, miR-124, and let-7c, play an important role in the polarization of macrophages and microglia (3638). In ischemic stroke, activation of resident microglia is the first step of the inflammatory response in the brain, which is followed by infiltration of peripheral immune cells including neutrophils and macrophages. Infiltrating macrophages that express similar antigens and mediators as resident microglia during recruitment and activation may have a greater role in the disease pathology (39, 40). They activate various neurovascular cell types in the brain penumbra, including microglia, which reduces the expression of tight junction proteins (41, 42). Newly recruited macrophages/microglia at the site of injury mainly express M2 signature genes, whereas at 1 week after injury, M1-type macrophages/microglia dominate the areas surrounding the site of injury (43, 44). In the present study, only M1-type macrophages/microglia were significantly decreased by the deletion of endothelial miR-15a/16-1 at 48 hours after ischemia, suggesting that deletion of endothelial miR-15a/16-1 not only reduced the infiltration of peripheral macrophages but also inhibited this phenotype shift from an initial inflammation-resolving M2 to an proinflammatory M1 phenotype through a yet undefined mechanism. However, the main source of M1-type microglia is still controversial because there is no specific marker to identify the original resident microglia from the infiltrated macrophage–derived microglia (45). TMEM119 has been reported to distinguish resident microglia from the circulating blood-derived macrophages in mouse brain tissue after systemic inflammation and traumatic brain injury. However, a reliable marker to identify infiltrated monocyte–derived microglia/macrophages is still under debate (39, 46).

The successful application of in vitro models is essential to mimic the mechanisms of disease and reduce the use of animals. In this study, we used mBMECs and mouse astrocytes to mimic ischemia-reperfusion in vitro. The noncontact coculture BBB model with ECs and astrocytes has been used to study stroke due to several advantages described previously (9). mBMECs are a specialized type of ECs that can form a tighter endothelium compared to peripheral ECs. Communication between ECs and astrocytes can augment many BBB features, particularly in vitro, such as tighter tight junctions (physical barrier) (2). Moreover, astrocytes can secrete various signaling proteins, including glial-derived neurotrophic factor (GDNF), transforming growth factor–β (TGF-β), basic fibroblast growth factor (bFGF), and angiopoietin 1 (Ang1) (47). This coculture model allows the induction of ECs by diffusible factors from astrocytes to mimic the normal in vivo condition. However, other cells such as pericytes and neurons also play important regulatory roles in the induction and maintenance of the BBB (48).

Together, our findings provide evidence that EC-selective deletion of the miR-15a/16-1 cluster attenuates BBB pathologies after ischemic stroke. These results have important implications in CNS disorders other than stroke, because BBB dysfunction is a characteristic feature of many neurological diseases, including traumatic brain injury, multiple sclerosis, and neurodegenerative diseases. Future pharmacological inhibition of miR-15a/16-1 activity may be important for the development of microRNA-based therapies in ischemic stroke.

MATERIALS AND METHODS

Animals

EC-selective miR-15a/16-1 cKO (EC-miR-15a/16-1 cKO) mice were generated as described (8, 21). Briefly, miR-15a/16-1–floxed homozygous mice (miR-15a/16-1flox/flox) were crossed with VE-Cadherin (Cdh5)-Cre transgenic mice (Jackson Laboratory, Cdh5-Cre Tg, catalog no. 006137). Both Cdh5Cre Tg and miR-15a/16-1flox/flox mice are derived from C57BL/6J background. Offspring were genotyped by PCR as DNA was obtained from tail-snip biopsies using transgene-specific oligonucleotide primers for Cdh5-Cre and miR-15a/16-1-floxP (21). Two groups of mice were used for this study: (i) EC-selective miR-15a/16-1 cKO (EC-miR-15a/16-1 cKO) mice with a genotype of Cdh5-Cre–transgenic and miR-15a/16-1–floxed homozygous (Cdh5Cre Tg/miR-15a/16-1flox/flox) and (ii) littermate WT control mice with a genotype of Cdh5-Cre WT and miR-15a/16-1–floxed homozygous (Cdh5Cre WT/miR-15a/16-1flox/flox). Mice were housed in a temperature- and humidity-controlled animal facility with a 12-hour light-dark cycle. Food and water were freely available. A total of 96 male mice were used for this study. All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee of Health Guide for the Care and Use of Laboratory Animals. Animals were randomly assigned to various experimental groups, and all assessments were performed and calculated by an investigator blinded to genotype.

Mouse model of transient focal cerebral ischemia

Male mice (8 to 12 weeks, 23 to 28 g) were anesthetized with isoflurane (3% for induction and 1.5% for maintenance) in mixed O2 and N2O. Focal cerebral ischemia was induced by introducing a 7-0 monofilament suture (Doccol Corporation) in the left common carotid artery, advanced to the origin of the middle cerebral artery for 1 hour as described previously (19, 49). Body temperature was measured with a rectal thermometer and maintained at 37.0 ± 0.5°C by a temperature-controlled heating pad during the ischemic period. Only those mice with CBF reduced to less than 25% of baseline were considered to be successfully MCA occluded (MCAo). Reperfusion was performed by removing the suture after 1-hour occlusion. Animals that did not show reperfusion of CBF over 60% of baseline were excluded. Regional CBF was measured using a laser speckle imager (Perimed PeriCam PSI HR) at 15 min before MCAo surgery, 15 min after occlusion, and 15 min after the onset of reperfusion. Two identical elliptical regions of interest were selected. The relative CBF was calculated as the ratio of the ipsilateral and contralateral hemispheres, which was then normalized to the mean value of pre-surgery CBF baseline level for each animal. Animals were allowed 24- or 48-hour reperfusion before being sacrificed according to the design of the experiment.

Cell-based in vitro noncontact BBB model and measurement of BBB functional integrity

Mouse astrocytes were purchased from ScienCell (catalog no. M1800-57) and grown in regular 12-well plates at 37°C in a humidified incubator until they reached confluence. mBMECs were purchased from Cell Biologics (catalog no. C57-6023) and seeded onto collagen- and fibronectin-coated transwell inserts (Corning) at a density of 1 × 105 cells per insert. Only up to eight passages were used for experiments. The in vitro BBB model was established as shown in Fig. 5A, and cultures were maintained in endothelial growth medium supplemented with 10% fetal bovine serum (FBS) and a supplement kit (Cell Biologics, catalog no. M1168) at 37°C in a humidified incubator overnight. The inserts seeding mBMECs were separated from the coculture system, and mBMECs were infected by lentivirus carrying pre-miR-15a or small hairpin RNAs targeting the mouse pre-miR-15a or the nonfunctional controls, respectively (see below). After 48 hours of infection, the medium containing lentivirus was replaced with mBMEC growth medium, and mBMECs were moved back to the coculture system with astrocytes for 24 hours. To assess BBB functional integrity, both TEER and tracer flux of dextran Alexa Fluor 488 (3000 molecular weight, Invitrogen, catalog no. D34682) were measured at designed time points or conditions (50). The reading of total resistance (RTotal) was measured with an Epithelial Volt/Ohm Meter (WPI) at room temperature according to the manual. The value of each sample (RTEER) was corrected by the reading of a cell-free insert (RBlank) and calculated with the polyester membrane area (SMembrane) as follows: RTEER = (RTotalRBlank) × SMembrane. TEER values were reported in units of ohm·cm2. To assess the permeability of the barrier to the paracellular tracer compound, dextran Alexa Fluor 488 was added into the upper compartment (representing the luminal compartment) at a working concentration of 1 μg/ml medium. At 0.5, 1, 2, 4, and 24 hours after treatment, 50-μl medium was collected from the lower compartment (representing the abluminal compartment). Fluorescence intensity of dextran in the abluminal compartment that permeated from the luminal compartment at each time point was measured using the SpectraMax i3× Multi-Mode Detection Platform (Molecular Devices). The fluorescence intensity of abluminal dextran Alexa Fluor 488 was measured at each time point and corrected by the respective blank (same treatment without adding dextran in the luminal compartment). Relative fluorescence unit (RFU) was recorded.

Lentivirus generation for overexpression or deletion of miR-15a/16-1

Lenti-miR precursor vector carrying mouse pre-miR-15a (Lenti miR-15a) or miR lentivector carrying small hairpin RNAs that target the mouse miR-15a gene (miRZip 15a) and nonfunctional controls (Lenti GFP and miRZip GFP) were generated as previously described (20). Recombinant lentiviral vectors were transfected into 293TN cells in the presence of PureFection Transfection Reagent [System Biosciences Inc. (SBI)] and pPACKH1 Packaging Plasmid Mix (mixture of three plasmids: pPACKH1-GAG, pPACKH1-REV, and pVSV-G) for 48 to 72 hours. Medium was collected, and lentiviral particles were further precipitated with PEG-it virus precipitation solution (SBI). Lentiviral particles were concentrated and aliquoted in cryogenic vials stored at −80°C. Cultured mBMECs and bEnd.3 cells were infected with an equal amount of miRZip GFP and miRZip 15a or Lenti GFP and Lenti miR-15a diluted in antibiotic-free growth medium for 48 to 72 hours. Levels of infection efficiency were evaluated with the EVOS FL Imaging System (Life Technologies), and cells were ready for experiments when 70 to 80% of lentivirus-infected cells exhibited green fluorescence.

Oxygen-glucose deprivation

To mimic ischemia in vitro, mBMECs were exposed to OGD for 16 hours (19, 51). Briefly, mBMEC growth medium was replaced with pre-flushed glucose-free Dulbecco’s modified Eagle’s medium (DMEM; Gibco, catalog no. 11966). The chamber was flushed for 20 min with 95% (v/v) nitrogen and 5% (v/v) carbon dioxide and incubated at 37°C in the incubator. After 16 hours, glucose-free DMEM was replaced with normal growth medium and mBMECs were incubated for a period indicated in each experiment.

Immunohistochemistry

For in vivo experiments, mice were deeply anesthetized and transcardially perfused with 0.9% NaCl and then with 4% paraformaldehyde (PFA). Brains were harvested, fixed in 4% PFA overnight at 4°C, and cryoprotected in 30% sucrose. Frozen serial coronal brain sections (25 μm) were prepared on a cryostat (Leica) and processed directly for Alexa Fluor 555–conjugated cadaverine detection. The rest of the sections were stored in a cryoprotectant at −20°C until use. Sections were blocked with 5% normal donkey serum for 1 hour and then incubated overnight in various primary antibodies. Immunohistochemistry was performed on brain sections using the following primary antibodies: mouse anti-F4/80 (BioLegend, catalog no. 123102), rat anti-Ly6B (Abcam, catalog no. ab53453), rabbit anti-NeuN (EMD Millipore, catalog no. ABN78), rabbit anti-Iba1 (Wako, catalog no. 019-19741), mouse anti-CD16/32 (BD Biosciences, catalog no. 553142), goat anti-CD206 (R&D Systems, catalog no. AF2535), rabbit anti-CD31 (Abcam, catalog no. ab28364), and mouse anti–Claudin-5 (Thermo Fisher Scientific, catalog no. 35-2500). After washing, sections were incubated with secondary antibodies conjugated with 488 or Cy3 (Jackson ImmunoResearch) for 1 hour at room temperature. Additional sections from each experiment condition were incubated with all the solutions except the primary antibodies to serve as negative controls to assess nonspecific signals. Images were acquired with an Olympus FluoView FV1000 confocal microscope (Olympus America) from five sections (from bregma 1.1 to −1.46) per animal. Cerebral neutrophil and macrophage infiltration were evaluated by counting the numbers of cells positively stained by F4/80+ or Ly6B+, respectively. Images were processed with ImageJ to blindly count automatically recognized cells. Cell number was calculated from five regions (three from the cerebral cortex and two from the striatum within the peri-infarct areas, which were identified as areas closed to the boundary of ischemic core) of each section and expressed as the mean number of cells per square millimeter. Polarization of M1- or M2-type microglia/macrophages was assessed by counting the number of cells that were positive for CD16/32 or CD206 and Iba1 (52). The expression of claudin-5 in ECs was evaluated by the quantification of the CD31+/Claudin-5+ area in three peri-infarct cortex regions per section.

For in vitro experiments, mBMECs were seeded in multiple culture sliders (Corning) and subjected to the indicated treatments. Cells were washed with phosphate-buffered saline (PBS), fixed with 4% PFA 15 min at room temperature, blocked with 5% normal donkey serum for 1 hour, and incubated overnight with rabbit anti–Claudin-5 antibody (Abcam, catalog no. ab15106). Cells were then washed with PBS and incubated with donkey anti-rabbit Cy3 for 1 hour at room temperature. Images were acquired from two random fields from each sample. The intensity of fluorescence was calculated by thresholding to the same percent of histogram, and mean arbitrary units (a.u.) were reported to evaluate the amount of claudin-5 expression.

Measurement of brain infarct and neurological deficit

Mice were sacrificed at 24 hours after 1-hour MCA occlusion. Mouse brains were cut into 1-mm-thick slices, which were stained with 2% 2,3,5-triphenyltetrazolium (TTC; Sigma-Aldrich) for 15 min at room temperature. Infarct volume was determined with ImageJ software. Infarct volume was also confirmed by immunostaining brain sections for MAP2 (Sigma-Aldrich, catalog no. AB5622). Seven sections per animal were quantified and calculated as the volume of brain infarct area. Neurological deficit was scored on a five-point scale as described previously (23): 0, no observable neurological deficits (normal); 1, failure to extend right forepaw (mild); 2, circling to the contralateral side (moderate); 3, falling to the right (severe); 4, mice could not walk spontaneously; 5, depressed level of consciousness (very severe).

Assessment of brain water content

Brain water content was evaluated by measuring water content using a dry-wet method as described previously (49). Mice were sacrificed by CO2 exposure. The ipsilateral and contralateral hemispheres were recorded as the MCAo group and sham control, respectively. Wet weights were measured directly, and dry weights were obtained after the hemispheres were heated at 100°C in an oven for 24 hours. Both wet weight (WWet) and dry weight (WDry) were measured to quantify brain water as follows: Brain water content = (WWetWDry)/WWet × 100%. Results were reported as the percentage of increased weight.

Evaluation of BBB leakage

BBB permeability of large molecules was assessed by the classical EB extravasation assay (19) and a fluorescent tracer assessment. EB (4%) (Sigma-Aldrich) was injected through the femoral vein at 1 hour before mice were sacrificed. Mice were perfused with 0.9% NaCl, and the brains were collected and separated into ipsilateral (MCAo) and contralateral (sham) hemispheres. Mouse brain tissues were homogenized in N,N-dimethylformamide (Sigma-Aldrich) and centrifuged at 25,000 relative centrifugal field for 45 min. The supernatants were collected to detect the absorption of EB in each tissue. EB levels were quantified by the following formula: (A620nm − (A500nm + A740nm)/2)/mg WWet. The fluorescent tracer Alexa Fluor 555–conjugated cadaverine (Thermo Fisher Scientific, catalog no. A30677) was injected 30 min before mice were sacrificed. Brain sections were prepared as previously described, and images were acquired using an inverted Nikon Diaphot-300 fluorescence microscope equipped with a SPOT RT slider camera and Meta Series software 5.0 (Molecular Devices). Six sections encompassing the MCA territory were quantified for the cross-sectional area of cadaverine fluorescence. These sections were summarized and multiplied by the distance between sections (1 mm) to yield a leakage volume in cubic millimeters. To measure endogenous IgG leakage, adjacent sections from the same brains used for cadaverine leakage quantification were used. Sections were blocked with avidin followed by biotin and incubated with biotinylated anti-mouse IgG reagent (Vector Laboratories, catalog no. MKB-2225) overnight at 4°C. Sections were then incubated with streptavidin 488 (Jackson ImmunoResearch) at room temperature for 30 min. Images were acquired and quantified as described above.

Quantitative real-time PCR

Mice were sacrificed at 24 hours after 1-hour MCA occlusion and perfused with 0.9% ice-cold saline. Mouse brains were removed and stored at −80°C until use. Total RNA was isolated from cerebral cortex by using TRIzol reagent (Invitrogen) as described previously (51). To detect Claudin-5 in mBMEC cultures, quantitative real-time PCR was performed using the primers 5′-TAAGGCACGGGTAGCACTCA-3′ (forward) and 3′-GGACAACGATGTTGGCGAAC-5′ (reverse) using total RNA isolated with an RNeasy Mini kit (QIAGEN). Relative Claudin-5 expression was corrected by Cyclophilin expression and then normalized to WT sham or non-OGD controls.

Western blotting

Total protein from the cerebral cortex or mBMEC culture was homogenized in lysis buffer and isolated as described (19). The blots were blocked with 5% nonfat milk and incubated with mouse anti–Claudin-5 antibody (Thermo Fisher Scientific, catalog no. 35-2500) or β-actin (Sigma-Aldrich) for 1 hour at room temperature. Then, membranes were incubated with horseradish peroxidase–conjugated secondary antibodies (Cell Signaling) for 1 hour and developed using a Pierce ECL Western blotting detection kit (Thermo Fisher Scientific) and Amersham High-Performance Chemiluminescence Films (GE Healthcare). ImageJ software was used to quantify Western blot signals. The relative claudin-5 amount was corrected by β-actin and then normalized by WT sham or non-OGD controls.

Functional analysis of the miR-15a/16-1 interaction with 3′UTR of mouse Claudin-5 mRNA

A miTarget microRNA 3′UTR luciferase vector (pEZX-MT01) was purchased from GeneCopoeia (Rockville, USA), in which a 608-bp fragment of the 3′UTR of the mouse Claudin-5 gene contains the miR-15a/16-1 binding sequence. Mutant 3′UTR of the mouse Claudin-5 gene from the miR-15a/16-1 binding sites of perfect complementarity (pEZX-MT0101) was also purchased. mBMEC lines (bEnd.3 cells) were purchased from the American Type Culture Collection (ATCC) and cultured in DMEM (ATCC, catalog no. 20-2002) supplemented with non–heat-inactivated 10% FBS. bEnd.3 cells were seeded at a density of 0.5 × 105 cells per well in a 24-well plate and infected by the generated lentiviruses for 48 hours as described before to achieve miR-15a overexpression or knockdown. Cells were then cotransfected with a Claudin-5 3′UTR luciferase reporter construct or a construct with mutant putative miR-15a/16-1 binding site using Lipofectamine 2000 (Invitrogen) for 6 hours. Firefly and Renilla luciferase activities were determined at 48 hours after transfection using the Dual-Luciferase Assay Kit (Promega) with the SpectraMax i3× Multi-Mode Detection Platform (Molecular Devices). Individual relative luciferase activity was normalized to the corresponding Firefly-Renilla luciferase activity.

Statistics analysis

Data analysis was performed with IBM SPSS Statistics 24 at a confidence of 95%, and figures were generated from GraphPad Prism 6. Normality of continuous variables was assessed by the Kolmogorov-Smirnov test if N ≥ 30 and the Shapiro-Wilk test if N < 30. To assess the difference between two independent groups, an independent t test was used for normally distributed variables. The Mann-Whitney U test was used to explore differences between two nonnormally distributed variables; Kruskal-Wallis H test followed by a Dunn post hoc test was used for more than two groups. To assess the difference between two related groups, a paired t test was used for normally distributed variables and Wilcoxon signed-rank test was used for nonnormally distributed variables. Results were presented as means ± SEM. Correlations between normally distributed continuous variables were checked using Pearson coefficients. P < 0.05 was considered statistically significant.

SUPPLEMENTARY MATERIALS

stke.sciencemag.org/cgi/content/full/13/626/eaay5686/DC1

Fig. S1. EC-selective deletion of miR-15a/16-1 does not change the regional CBF during MCA occlusion and 15 min after reperfusion.

Fig. S2. OGD impairs BBB paracellular permeability in mBMECs.

REFERENCES AND NOTES

Funding: This work was supported by the NIH (grants NS091175, NS112181, and NS08682 to K.-J.Y.). Author contributions: F.M. and K.-J.Y. designed research; F.M., P.S., and X.Z. performed research; F.M. analyzed data; and F.M., K.-J.Y., and M.H.H. wrote the paper. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article