Dissociation and Protection of the Neurovascular Unit after Thrombolysis and Reperfusion in Ischemic Rat Brain

Abstract

In the ischemic brain, reperfusion with tissue plasminogen activator (tPA) sometimes causes catastrophic hemorrhagic transformation (HT); however, the mechanism remains elusive. Here, we show that the basement membrane, and not the endothelial cells, is vulnerable to ischemic/reperfusion injury with tPA treatment. We treated a spontaneously hypertensive rat model of middle cerebral artery occlusion (MCAO) with vehicle alone, tPA alone, or a free radical scavenger, edaravone, plus tPA. Light and electron microscopic analyses of each microvascular component revealed that the basement membrane disintegrated and became detached from the astrocyte endfeet in tPA-treated animals that showed HT. On the other hand, edaravone prevented the dissociation of the neurovascular unit, dramatically decreased the HT, and improved the neurologic score and survival rate of the tPA-treated rats. These results suggest that the basement membrane that underlies the endothelial cells is a key structure for maintaining the integrity of the neurovascular unit, and a free-radical scavenger can be a viable agent for inhibiting tPA-induced HT.

Introduction

Ischemic brain damage can be ameliorated, if cerebral blood flow is restored by thrombolytic agents, for example, tissue plasminogen activator (tPA), within a short time (The National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group, 1995), but delayed reperfusion with tPA can cause hemorrhagic transformation (HT) (The NINDS t-PA Stroke Study Group, 1997). Cerebral microvascular integrity depends largely on three components, namely the vascular wall, formed by endothelial cells; the blood–brain barrier (BBB), provided by endothelial tight junctions; and the basal membrane, which lines the endothelial cells (del Zoppo, 2006). Free radicals, the fundamental mediators of reperfusion injury (Chan, 2001) and reperfusion-related HT (Maier et al, 2006), are generated soon after vessel occlusion, with explosive propagation after reperfusion. Various proinflammatory mediators (matrix metalloproteinases (MMPs), thrombin, vascular endothelial growth factor, and bradykinin) (Aschner et al, 1997; Kamiya et al, 1993; Rosenberg, 2002; Suarez and Ballmer-Hofer, 2001) increase in the ischemic brain, accompanied by brain edema, endothelial cell death (Maier et al, 2006), disruption of tight junctions, and basal membrane/extracellular matrix (collagenIV, laminin-1, and fibronectin) loss (del Zoppo and Mabuchi, 2003). Any of these changes could be involved in HT associated with tPA treatment. To elucidate the mechanism underlying tPA-associated HT, here we treated a spontaneously hypertensive rat model of middle cerebral artery occlusion (MCAO) with vehicle or tPA, and observed the neurovascular unit components, with and without a free radical scavenger, edaravone (Figure 1). We found that the basement membrane is the major structure disrupted in tPA-induced HT, and edaravone could prevent the HT-associated damage.

Figure 1

Experimental groups included the V+V group (n=6), V+tPA group (n=16), and E+tPA group (n=11). Rats were killed 24 h after the reperfusion. E, edaravone; MCAO, middle cerebral artery occlusion; V, vehicle.

Materials and methods

Animals

Spontaneously hypertensive rats (male, 11 weeks old, 250 to 280 g, total n = 33; Disease Model Cooperative Research Association, Kyoto, Japan) were used, because we had observed that these rats presented more apparent HT after MCAO compared with the Wistar rat in our preliminary examination. They were fed ad libitum with 1% NaCl water for 14 days, and used for this study when they became 13 weeks old. All experimental procedures were approved by the Animal Committee of the Okayama University Graduate School of Medicine.

Experimental Groups and Drug Treatment

Three groups of rats were studied. The V+V control group (n = 6) received four injections of vehicle (physiologic saline, intravenously, 0.5 mL) every 1.5 h during 4.5 h of cerebral ischemia, followed by the same vehicle (intravenously) at reperfusion. The V+ tPA group (n = 16) received vehicle treatment during cerebral ischemia as above, followed by tPA treatment (Grtpa, Mitsubishi Tanabe Pharma Corporation, Osaka, Japan, intravenously, 10 mg/kg) at reperfusion. The E + tPA group (n = 11) received four injections of edaravone (Mitsubishi Tanabe Pharma, intravenously, 3 mg/kg) during cerebral ischemia, followed by tPA treatment as above at reperfusion (Figure 1). This relatively high dose of tPA was chosen based on the report of an ∼10-fold difference in fibrin-specific activity between humans and rodents (Korninger and Collen, 1981). Edaravone was injected repeatedly because it has a very short half-life (T_1/2 = 5.4 mins) (Watanabe et al, 1994) and free radicals are generated not only soon after vessel occlusion but also after reperfusion (Zhang et al, 2004).

Focal Cerebral Ischemia

To mimic a clinical situation reperfused by tPA, we designed the present experiment whereby tPA was administrated just before a reperfusion by pulling out a nylon thread from the origin of MCA, so that the tPA-associated damage could be assessed under similar conditions of reperfusion (Supplementary Table 1). However, in this model, it was not addressed whether edaravone could affect the clot bursting effect of tPA. In brief, the rats were anesthetized with a nitrous oxide/oxygen/isoflurane mixture (69/30/1%) administered through an inhalation mask. The left carotid bifurcation was exposed, and the external carotid artery was coagulated distal to the bifurcation. A silicone-coated 4–0 nylon thread was then inserted through the stump of the external cerebral artery and gently advanced (∼18mm) to occlude the origin of the MCA. After occlusion for 4.5 h, which could cause hemorrhagic infarction after tPA treatment (Zhang et al, 2004), the nylon was gently withdrawn to restore the blood flow in the MCA territory and the incision was closed. The caudal tail artery and left femoral vein were cannulated for continuous monitoring of the mean arterial blood pressure, analysis of arterial blood gases, blood glucose level, and serum cytokines/chemokines/MMPs, and for intravenous drug administration. During surgery, the rectal and left temporal muscle temperatures were maintained at 37.0°C±0.3°C by placing the animals on a heating bed (Model BWT-100; Bio Research Center, Nagoya, Japan). The cortical cerebral blood flow in the left MCA territory was measured during ischemia by laser-Doppler flowmetry using an MBF3D (Moor Instruments Ltd, Axminster, UK).

Table 1

Physiologic parameters in the three experimental groups

Animal group	pH	pO2 (mmHg)	pCO2 (mmHg)	MABP (mmHg)	TT (°C)	BG (mg/dL)
Baseline
V+V (n = 6)	7.47±0.09	95.5±11.8	37.8±2.3	117.3±29.2	37.1±0.1	129.8±14.2
V+tPA (n = 16)	7.42±0.03	103.4±18.4	43.8±5.9	122.9±29.6	37.0±0.2	96.9±16.7
E+tPA (n = 11)	7.43±0.04	104.5±17.3	42.2±5.1	117.2±29.3	37.0±0.3	99.3±18.7
1.5 h after the induction of ischemia
V+V (n = 6)	7.37±0.02	82.7±7.1	46.3±3.7	95.5±15.8	37.2±0.2	127.0±10.2
V+tPA (n = 16)	7.39±0.06	91.9±19.4	43.8±5.9	99.6±24.2	37.1±0.1	92.4±16.2
E+tPA (n = 11)	7.37±0.06	92.8±20.4	47.9±5.4	95.0±34.3	37.1±0.1	95.2±8.2
4.5 h after the induction of ischemia
V+V (n = 6)	7.36±0.05	85.3±14.6	50.8±6.8	103.8±15.4	37.2±0.2	129.6±12.9
V+tPA (n = 16)	7.42±0.03	101.4±22.4	46.1±6.0	104.4±30.2	37.1±0.2	102.7±11.6
E+tPA (n = 11)	7.42±0.03	95.5±21.5	48.6±5.7	102.1±34.3	37.1±0.1	107.6±13.8

BG, blood glucose level; E, edaravone; MABP, mean arterial blood pressure; tPA, tissue plasminogen activator; TT, temporal muscle temperature; V, vehicle. Data are the mean±s.d.

Multiplex ELISA

We used an ELISA (enzyme-linked immunosorbent assay) micro-array kit (Pierce Endogen, Woburn, MA, USA) to assess the levels of secreted cytokines/chemokines/MMPs (IL-1β, IL-6, tumor necrosis factor-α, macrophage inflammatory protein-1α, Fractalkine, and MMP-9). Blood was drawn from the left femoral vein at 0 and 3 h after the induction of ischemia and at 24 h after reperfusion, and the serum level of each cytokine, chemokine, and MMP was measured according to the manufacturer's specifications, as described earlier in detail by Lee et al (2007).

Behavioral Analysis

At 24 h after reperfusion, the surviving rats were tested for behavioral changes and scored as described by Bederson et al (1986), with minor modifications, as follows: 0, no observable neurologic deficits; 1, failure to extend the right forepaw; 2, circling to the contralateral side; 3, falling to the right; and 4, unable to walk spontaneously.

Histochemistry

We focused on the changes at and around blood vessels, and analyzed the infarcted brain tissue mainly through histochemistry in this study. For light microscopy, the surviving rats (V +V group, n = 6; V+ tPA group, n = 8; and E + tPA group, n = 8) were anesthetized by intraperitoneal injection of ketamine hydrochloride, and then perfused with chilled phosphate-buffered saline, followed by 4% paraformaldehyde in 0.1 mol/L of phosphate buffer. After postfixation overnight, 50-µm-thick sections were cut with a vibrating blade microtome (VT1000S; Leica, Heidelberg, Germany). To analyze the brain hemorrhage, we performed iron staining using an enhanced Perl's reaction. The brain sections were incubated with Perl's solution (5% potassium ferrocyanide and 5% HCl, 1:1) for 45 mins, washed in distilled water, and incubated again in 0.5% diamine benzidine tetrahydrochloride with nickel for 60 mins, as described by Wu et al (2003). For immunohistochemistry, the following primary antibodies were used: mouse anti-4-hydroxyl-2-nonenal (4-HNE) antibody (1:200; JaICA, Shizuoka, Japan); mouse anti-N-(hexanonyl) lysine (HEL) antibody (1:200; JaICA); rabbit anti-MMP-9 antibody (1:200; Chemicon, Temecula, CA, USA); rabbit antioccludin antibody (1:100; Zymed, South San Francisco, CA, USA); rabbit anti-collagenIV antibody (1:100; Novotec Lyon, France); mouse anti-glial fibrillary acidic protein (GFAP) antibody (1:200; Chemicon); and goat antiaquaporin4 (1:100; Santa Cruz Biotechnology, Santa Cruz, CA, USA). Matrix metalloproteinase-9, collagenIV, GFAP, and aquaporin4 were detected with secondary antibodies conjugated with Alexa Fluor™ (Molecular Probes, Eugene, OR, USA). To estimate the expressions of 4-HNE, HEL, MMP-9, occludin, or collagenIV, the sections were incubated with the respective first antibody and then with an appropriate biotin-labeled secondary antibody (1:500). To estimate the rat IgG, which has been reported to leak from blood vessels into the brain parenchyma in areas where the BBB has broken down (Richmon et al, 1998), sections were incubated with biotin-labeled goat anti-rat IgG antibody (1:500). After incubation with an ABC Elite complex (Vector Laboratories, Burlingame, CA, USA), the signal was visualized with diaminobenzidine tetrahydrochloride. To detect vascular endothelial cells, we used N-acetylglucosamine oligomer (NAGO) as the specific endothelial cell marker (Augustin et al, 1995). The sections were incubated overnight with biotinylated lycopersicon esculentum lectin (1:200; Vector Laboratories), which binds NAGO. The sections were then incubated with the ABC Elite complex, and visualized with diaminobenzidine tetrahydrochloride or a tyramide signal amplification kit (TSA Tetramethyl rhodamine system; PerkinElmer Life Science, Boston, MA, USA).

Electron Microscopic Analysis

For electron microscopy (V + tPA group, n=3; E+ tPA group, n = 3), the brains were perfused with 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 mol/L of phosphate buffer and, after postfixation overnight, 1.0-mm-thick coronal sections were cut with a vibrating blade microtome. We identified the section located 1.2 to 0.2mm rostral to bregma (Paxinos and Watson, 1982), and dissected the peri-infarct area of the ipsilateral cortex into small pieces (3.5 × 1.0 × 1.0mm) for thin sectioning. The brain sections were postfixed in 2% osmium for 1.5 h, dehydrated, and embedded in Epon 812 Resin (TAAB, Berks, UK). Ultrathin (0.06 µm) sections were cut with a diamond knife, stained with uranyl acetate and lead citrate, and examined with an electron microscope (H-7100S; Hitachi, Tokyo, Japan).

Quantitative Analysis

For the semiquantitative evaluation of rat IgG staining, sections every 300 µm were captured through a microscope (Axioplan 2; Carl Zeiss, Jena, Germany). The staining intensity was measured with an image processing software (Scion Image, Scion Corporation, Frederick, MD, USA). For the semiquantitative evaluation of histochemical stainings, such as for iron, 4-HNE, HEL, MMP-9, NAGO, occludin, or collagenIV-positive blood vessels, the stained sections were selected from three levels of the caudate putamen (1.2, 0.7, and 0.2mm rostral to the bregma; Paxinos and Watson, 1982) of each animal, and three areas in the periinfarct cortex in each section were chosen randomly and captured at −200 magnification with a microscope (BX51; Olympus, Tokyo, Japan). We confirmed the border between the ischemic core and peri-infarct lesion through cresyl violet staining of adjacent sections as reported earlier (Omori et al, 2002). Blood vessels were identified by their morphology, and the staining intensity (iron, 4-HNE, HEL, MMP-9, NAGO, and collagenIV) was measured using the software (Scion Image). To quantify the degradation of occludin protein surrounding the vascular wall, we scored the occludin expression on each blood vessel (3, more than 75% of the vascular wall was positive for occludin; 2, 50% to 75%; 1, 25% to 49%; 0, > 25%). To assess the detachment of astrocyte endfeet from the basement membrane in the GFAP/collagenIV double-labeled sections, three levels of sections were selected as described above, and four areas in the ipsilateral peri-infarcted cortex in each section were chosen randomly and captured at −100 magnification with a confocal laser microscope (LSM510; Carl Zeiss). We measured the area between the astrocyte endfeet and basement membrane of each blood vessel, and the length of each blood vessel. We then calculated the ratio of the area to the length, which we called the ‘vascular dissociation index’ in this study. This index was very useful for analyzing the spatial dissociation of the vessels from the astrocyte endfeet.

Statistical Analysis

Values are expressed as means±s.d. The frequencies of death (survival rates) were compared by an m × n χ² test. To evaluate the differences in infarct volume, physiologic parameters, and cytokine/chemokine/MMP expression, we used the nonrepeated measures such as analysis of variance and the Student-Newman-Keuls (SNK) test. The differences in behavior, the intensity of rat IgG, iron, 4-HNE, HEL, MMP-9, NAGO, occludin, and collagenIV staining, and the vascular dissociation index were evaluated for statistical significance by the Kruskal–Wallis Htest and the Mann–Whitney U-test with Bonferroni's correction. In all statistical analyses, significance was accepted at P > 0.05.

Results

We carefully monitored several physiologic parameters, including regional cerebral blood flow during the experiments, and found no significant differences among the three experimental groups (Table 1 and Supplementary Table 1). We then evaluated the survival rate, motor function, brain hemorrhage, and infarct volume in the three groups. Despite 4.5 h of MCAO, which caused severe brain infarction, no vehicle-treated (V+ V) rats had died 24 h after the reperfusion. However, 34% of the tPA-treated (V + tPA) rats had died, and all the dead animals showed obvious and massive intracerebral hemorrhage. In contrast, none of the rats treated with edaravone plus tPA (E + tPA) had died (Figure 2A). We also compared the rats' behavior using the protocol described by Bederson et al (1986), and found that motor function was significantly better in the E + tPA group than in the V+ tPA group (*P > 0.05; Figure 2B). Intracerebral hemorrhage transformation, indicated by iron release from the breakdown of hemoglobin, was observed in the surviving rats (Figure 2C, arrowheads). Macroscopic HT was observed in 33.3% of the V+V group (n = 6), 81.3% of the V+ tPA group (n = 16), and 27.3% of the E + tPA group (n = 11). We evaluated the intracerebral hemorrhage volume by the intensity of iron histochemistry. Brain sections from rats treated with edaravone (E + tPA) showed significantly less iron staining (*P > 0.05; Figure 2D). Next, we studied the functional integrity of the BBB by staining endogenous rat IgG, which has been reported to leak from blood vessels into the brain parenchyma in areas where the BBB has broken down (Richmon et al, 1998). Edaravone had a weak but nonsignificant effect, not only on the leaking of rat IgG but also on the infarct size (Figures 2E and 2F). These results indicated that tPA administration to the ischemic brain decreased the survival rate, adversely affected motor function, and also indicated increased the severity of brain hemorrhage, and that edaravone was protective.

Figure 2

Tissue plasminogen activator administration into ischemic rat brain resulted in a lower survival rate, worsening motor function, and obvious intracerebral hemorrhage. Edaravone, a free radical scavenger, controlled these tPA-mediated exacerbations. (A) Survival rate of the three experimental groups, determined 24 h after reperfusion. Tissue plasminogen activator injection without edaravone significantly lowered the survival rate (*P<0.05, m × n χ² test). Experimental groups included V+V, MCAO treated with vehicle (n=6); V+tPA, MCAO with tPA (n=16); and E+tPA, MCAO with edaravone and tPA (n=11) (Figure 1). (B) Motor functional analysis performed 24 h after reperfusion. Compared with the group treated with tPA alone, the edaravone- and tPA-treated groups showed significant improvement (*P<0.05, Mann–Whitney U-test). Data represent the mean±s.d. (C) Antihemorrhagic effects of edaravone on tPA-induced postischemic brain hemorrhage 24 h after MCAO. Arrowheads indicate the area where brain hemorrhage was evident, on the surface of the infarcted brain. Bar=5 mm. (D) Relative intensity of iron histochemistry determined 24 h after reperfusion. Edaravone decreased the amount of iron released in the ischemic brain treated with tPA (*P<0.05, Mann–Whitney U-test). Data are the mean±s.d. (E and F) Endogenous rat IgG leaks from blood vessels into the brain parenchyma through blood–brain barrier broken down. There was no statistically significant difference in both the relative intensity of rat IgG (E) and the infarct volume (F) at 24 h after MCAO (nonrepeated ANOVA). Data are the mean±s.d.

We next evaluated the effects of the tPA and E + tPA treatments on the level of a spectrum of factors, including lipid peroxide, cytokines, chemokines, and MMPs. We frequently observed blood vessels positive for 4-HNE, an end product of lipid peroxidation, in the peri-infarct lesions of the V+ tPA group (Figure 3A, black arrowheads), consistent with the theory that reactive oxygen species (ROS) are mainly produced in the peri-infarct area (Peters et al, 1998). Intracerebral hemorrhage was also observed in the peri-infarct lesions (Figure 3A, white arrowheads), indicating the association of an activated radical system with intracerebral hemorrhage. We did not detect the expression of 4-HNE or HEL, a marker of early lipid peroxidation, in the contralateral side of the three experimental groups (data not shown). Tissue plasminogen activator treatment increased the 4-HNE expression of blood vessels in the peri-infarct lesions of the ipsilateral side (**P > 0.01; Figures 3B and 3C), whereas edaravone decreased the levels of 4-HNE and HEL (Figures 3B to 3E). As lipid peroxidation might be part of a larger response arising after ischemia/reperfusion with tPA treatment, we also assessed the levels of selected cytokines, chemokines, and MMPs in rat serum, using a multiplex ELISA consisting of a 6-plex array (IL-1β, IL-6, tumor necrosis factor-α, MIP-1α, Fractalkine, and MMP-9). The level of IL-1β or IL-6 was high even at the baseline, probably because we used spontaneously hypertensive rats fed with 1% NaCl water for 14 days in this study (Sanz-Rosa et al, 2005; Demougeot et al, 2007). There was a significant difference between the V+ tPA and E + tPA groups only in MMP-9 expression (pro- and active forms) (Table 2). Therefore, we examined histologic sections stained with a pan anti-MMP-9 antibody (i.e., that labeled both the pro- and active forms). Matrix metalloproteinase-9 was expressed inside the vascular endothelium 5 mins after reperfusion, but it had extravasated into the perivascular area at 24 h after reperfusion (Figure 3F, arrowheads). Semiquantitative analysis of the MMP-9 staining showed that edaravone significantly decreased the MMP-9 expression of blood vessels in the peri-infarct area (*P > 0.05; Figures 3G and 3H). These data suggest that edaravone suppressed the tPA-induced lipid peroxidation and MMP-9 expression at and around cerebral vessels in the rat brain after ischemia/reperfusion.

Figure 3

The free radical scavenger, edaravone, decreased lipid peroxidation and MMP-9 expression in the peri-infarct area. (A) Coronal sections of the occluded hemisphere, including a raw brain section (left panel) and a section stained with an anti-4-HNE antibody (right panel), obtained from rats of the V+tPA group 24 h after reperfusion. Arrowheads point to borders between intact and infarct areas, in which not only cerebral hemorrhage (red area in left panel) but also 4-HNE-positive cells (brown area in right panel) are pointed out. Co, cerebral cortex; St, striatum. Bar=1 mm. (B to E) Anti-4-HNE (B) and anti-HEL (D) staining of the peri-infarcted lesion, and their respective relative intensity (C and E). The V+V group (n=6), V+tPA group (n=16), and E+tPA group (n=11). Bar=100 µm. (F) Double staining of the peri-infarct area with an anti-MMP-9 antibody (green) and lycopersicon esculentum lectin, which binds to N-acetylglucosamine oligomer (NAGO) on the surface of endothelial cells. Arrowheads point to MMP-9 expression in the perivascular area. Bar=20 µm. (G and H) Anti-MMP-9 staining (G) of the peri-infarcted lesion 24 h after reperfusion and (H) its relative intensity. Bar=100 µm. Statistical analysis: Mann–Whitney U-test (C, E, and H). Data represent the mean±s.d.

Table 2

Multiplex cytokines/chemokines/MMPs array analysis in rat serum

Cytokines/chemokines/MMPs	IL-1β (pg/mL)	IL-6 (pg/mL)	TNF-α (pg/mL)	MIP-1α (pg/mL)	Fractalkine (pg/mL)	MMP-9 (pg/mL)
Baseline
V+V (n = 6)	68.9±18.7	1,178±200	117.3±29.2	46.2±13.6	66.4±2.9	8.2±8.0
V+tPA(n = 16)	61.1±8.3	856±63	122.9±29.6	41.4±19.1	67.6±18.1	13.4±5.2
E+tPA (n = 11)	64.0±25.3	953±294	117.2±29.3	51.7±20.1	79.6±20.7	11.4±5.2
3 h after the induction of ischemia
V+V (n = 6)	48.8±22.0	1,231±130	95.5±15.8	62.6±10.4	69.0±20.1	12.8±6.6
V+tPA (n = 16)	55.5±11.0	721±99	99.6±24.2	64.8±36.1	73.6±13.4	12.6±6.0
E+tPA (n = 11)	55.4±5.4	679±69	95.0±34.3	46.6±26.4	78.4±8.7	12.0±11.8
24 h after reperfusion
V+V (n = 6)	44.2±26.0	655±84	103.8±15.4	21.5±8.6	63.3±25.7	15.2±6.4
V+tPA (n = 16)	53.3±16.9	965±324	104.4±30.2	23.6±4.8	72.2±17.4	29.0±14.4
E+tPA (n = 11)	74.0±17.5	1,190±285	102.1±34.3	25.8±3.2	86.2±6.7	15.2±6.2a

E, edaravone; IL, interleukin; MIP-1α, macrophage inflammatory protein-1α; MMP, matrix metalloproteinase; TNF-α, tumor necrosis factor-α, tPA, tissue plasminogen activator; V, vehicle.

Data are the mean±s.d.

a

The MMP-9 expression level in the E+tPA group was significantly lower than that in the V+tPA group (P>0.05).

Reperfusion-derived free radicals induce lipid peroxidation, especially in the membrane phospholipids of vascular endothelial cells (Abe et al, 1988; Phelan and Lange, 1991; Zhang et al, 2004). Moreover, lipid peroxide and MMP-9 are reported to cause the deterioration of tight junction proteins and the basement membrane (Usatyuk et al, 2006; Yang et al, 2007). To determine which component of the vascular unit was most disrupted in the three experimental models, we analyzed markers for vascular endothelial cells (NAGO), tight junctions (occludin), and basement membrane (collagenIV) located from the inner lumen side to the outer peri-vascular side of the cerebrovascular endothelium. To our surprise, there was no statistically significant difference in the expression of the endothelial cell marker, NAGO, in the three experimental models (Figures 4A and 4B). In contrast, edaravone treatment inhibited the disruption of the tight junction protein, occludin (*P<0.05; Figures 4C and 4D). The basement membrane protein, collagenIV, was obviously degraded by the tPA administration, and this tPA-induced degradation was inhibited by the edaravone treatment (*P>0.05; Figures 4E and 4F). These surprising results indicated that tPA treatment preferentially damages the outer-to-inner vascular components.

Figure 4

Tissue plasminogen activator injection disrupted the basement membrane around vessels, rather than the endothelial cells that are exposed to the blood flow. (A and B) Lectin staining for NAGO on endothelial cells (A), and its relative intensity (B) in the peri-infarct area 24 h after MCAO. Experimental groups: V+V group (n=6), V+tPA group (n=16), and E+tPA group (n=11). Bar=100 µm. (C) Immunohistochemical staining for occludin, a tight-junction protein in the peri-infarct area 24 h after MCAO. Occludin was expressed around blood vessels. Bar=50 µm. (D) Degradation score for estimating the occludin expression around blood vessels (details in Materials and methods). Boxes in panel D indicate data between the 25th and 75th percentiles, with the horizontal bar reflecting the median (▪, mean; –, 10th and 90th percentiles) (*P<0.05). (E and F) Immunohistochemical staining (E), and relative intensity (F) of collagenIV, a basal membrane protein, in the peri-infarct area 24 h after MCAO, indicating that reperfusion with tPA disrupted the basement membrane, and edaravone administration blocked this disruption (*P<0.05). Statistical analysis: Mann–Whitney U-test (B, D, and F). Data represent the mean±s.d.

This disruption of the basement membrane led us to hypothesize that the tPA-induced dissociation of the neurovascular unit involves detachment of the astrocyte endfeet surrounding the vessel from the basement membrane. We therefore examined whether dissociation occurred between the collagenIV-positive basement membrane and the GFAP-positive astrocyte foot processes. In the tPA-treated rat, no dissociation of the neurovascular unit was found on the contralateral, nonischemic side (Figure 5A); however, marked dissociation was observed on the ipsilateral side (**P > 0.01; Figure 5B), which was dramatically improved by edaravone (**P > 0.01; Figure 5C). In addition, we used an astrocyte endfeet marker, aquaporin4 (Nielsen et al, 1997), and confirmed that marked dissociation was observed on the ipsilateral side of the tPA-treated group (Figures 5E to 5G). Moreover, an electron microscopic examination revealed that the basement membrane had become thin and showed degradation, and that the edematous astrocyte endfeet had detached itself from the basement membrane on the ipsilateral side in the tPA-treated group (Figures 5I and 5L). In contrast, edaravone rescued both the basement-membrane degradation and the detachment of astrocyte endfeet (Figures 5J and 5M). These observations indicated that the mechanism of neurovascular unit dissociation by tPA involved the detachment of astrocyte endfeet from the basement membrane, which was prevented by edaravone.

Figure 5

Reperfusion with tPA induced detachment of the astrocyte endfeet from the basement membrane. (A to C) Double immunohistochemistry of the peri-infarct area with anti-GFAP (red) and anti-collagenIV (green) antibodies 24 h after reperfusion. Compared with the contralateral side of the V+tPA group (A), the space between the GFAP-positive astrocyte endfeet and the collagenIV-positive basement membrane was greater in the V+tPA group (B), and the neurovascular unit was greatly improved in the E+tPA group (C). Bar=20 µm. (D) The vascular dissociation index, which estimates the space between the astrocyte endfeet and the basement membrane (details in Materials and methods) (**P<0.01, n=5 per group, Mann–Whitney U-test, mean±s.d.). (E to G) Aquaporin4 (AQP4)/collagenIV-double immunohistochemistry of the peri-infarct area obtained from the contralateral side of V+tPA (E), the ipsilateral side of V+tPA (F), and the ipsilateral side of E+tPA (G) 24 h after reperfusion. Bar=20 µm. (H to M) Electron microscopic view of microvessels in a peri-infarct area obtained from the contralateral side of V+tPA (H and K), the ipsilateral side of V+tPA (I and L), and the ipsilateral side of E+tPA (J and M). The boxed areas in panels H, I, and J indicate the fields shown in panels K, L, and M, respectively. Arrowheads point to the basement membrane, which on the ipsilateral side of V+ tPA (I and L) was degraded and became thin. Asterisks indicate the space between the basement membrane and astrocyte endfeet. The tight junction between adjacent endothelial cells was structurally maintained (arrow in L). a, astrocyte endfeet; e, endothelial cell; L, lumen. Bar=0.5 µm.

Discussion

Together, our findings indicated that it was not the inner lumen of cerebral endothelial cells, but the outer, peri-vascular side of the basement membrane that was severely degraded in the rat ischemic brain treated with tPA. At the peri-vascular extracellular space in an animal model of cerebral ischemia, large quantities of ROS are generated after the onset of ischemia (Kontos et al, 1992). In addition, after MCAO, SOD2-KO mice exhibit a significant increase in MMP-9 and a higher rate of brain hemorrhage (Maier et al, 2006), indicating that the excess ROS activate MMP-9, which can degrade the major basal membrane component, collagenIV. Several groups have reported the decreased expression of adhesion molecules, such as laminin (Hamann et al, 1996) and integrin α1β6 (Wagner et al, 1997), in the perivascular extracellular space in the ischemic animal brain. In contrast, vascular endothelial cells, which are constantly exposed to blood flow, rarely show dUTP incorporation into their DNA in the ischemic animal brain, indicating a low level of injury (Tagaya et al, 1997). In addition, an in vitro study showed a correlation between endothelial cell death and extracellular matrix disruption by MMP activation (Lee and Lo, 2004). Moreover, we showed not only the loss of basement membrane protein but also a spatial dissociation between the vascular basement membrane and the astrocyte endfeet in the ischemic brain treated with tPA. These findings suggest that the basement membrane/extracellular matrix linking the endothelial cells might be essential for maintaining the integrity of the neurovascular unit, which is disrupted by tPA treatment.

The mechanism by which edaravone exerts its salutary effect on tPA-induced HT remains obscure, but several inferences can be drawn from this study. We observed that tPA treatment dramatically increased the 4-HNE expression of blood vessels in peri-infarct lesions (Figure 3B). Iron and thrombin, which can be released from blood clots after tPA treatment, have been reported to induce ROS (Nakamura et al, 2008). These results lead to speculation that exogenous tPA may, directly or indirectly, induce plenty of ROS in an infarcted brain. Ischemia/reperfusion injury also generates ROS, which can be associated with MMP-9 activation (Maier et al, 2006). We also showed that administration of edaravone resulted in the reduction of both 4-HNE and MMP-9 expressions on blood vessels of the peri-infarct area (Figures 3B and 3G). Therefore, edaravone may inhibit the activation of MMP-9, which can cause HT, by scavenging the tPA-induced ROS or by trapping the ROS derived from ischemia/reperfusion injury. However, other mechanisms are possible, and additional studies are needed to understand it in detail.

We reported earlier that edaravone ameliorates edema in the ischemic brain (Abe et al, 1988), and reduces the infarct size after 1.5 h of MCAO in rat brain reperfused with tPA, accompanied by reductions in the free radical end products of lipids, proteins, and DNA (Zhang et al, 2004). In a clinical trial, edaravone strongly attenuated the resulting disability in humans 90 days after acute ischemic stroke without serious adverse events (The Edaravone Acute Brain Infarction Study Group, 2003), and it has been used clinically in Japan as a neuroprotective agent for acute stroke patients since 2001. In a large clinical trial, another free radical scavenger, NXY-059, initially appeared to reduce disability after stroke (the first Stroke-Acute-Ischemic-NXY-Treatment trial, SAINT I) (Lees et al, 2006), but this effect could not be reproduced (SAINT II) (Shuaib et al, 2007). However, one mutual characteristic of NXY-059 was the potential to decrease symptomatic intracerebral hemorrhage after tPA treatment (Lees et al, 2006). NXY-059 is water soluble (octanol/water partition coefficients; cLog P =–2.09), and a study using a rat model showed it cannot easily pass through the BBB even after MCAO (Kuroda et al, 1999). In contrast, edaravone has a biphasic, water-soluble, and lipid-soluble nature (cLog P = 1.33), and when intravenously administered, can easily pass through the BBB to enter the brain parenchyma and cerebral fluid (Yamamoto et al, 1996). As the site most vulnerable to free radical damage after reperfusion with tPA is on the outer side of the vascular endothelium, an intravenously administered compound should have a biphasic soluble nature to reach the most vulnerable site of the brain. Thus, this unique chemical property of edaravone might be an advantage for its delivery to the basement membrane/extracellular matrix underlying the endothelial cells. Therefore, we propose that a combination therapy with edaravone and tPA could provide important therapeutic benefits for acute stroke patients, not only in reducing the infarct size but also in minimizing the catastrophic HT. On the basis of these unique effects of edaravone, a new clinical trial is now being undertaken in Japan to investigate the benefits of using it to protect the neurovascular unit in combination with tPA treatment.