Evaluating Therapeutic Targets for Reperfusion-Related Brain Hemorrhage Carolina M. Maier, PhD, Lily Hsieh, BS, Trisha Crandall, BS, Purnima Narasimhan, PhD, and Pak H. Chan, PhD Objective: Early reperfusion after an ischemic stroke can cause blood–brain barrier injury with subsequent cerebral edema and devastating brain hemorrhage. These complications of early reperfusion, which result from excess production of reactive oxygen species, significantly limit the benefits of stroke therapies. In this article, we use a novel animal model that facilitates identification of specific components of the reperfusion injury process, including vascular injury and secondary brain damage, and allows assessment of therapeutic interventions. Methods: Knock-out (KO) mice containing 50% manganese-superoxide dismutase activity (SOD2-KO) and transgenic mice overexpressing SOD2 undergo transient focal ischemia and reperfusion followed by assessment of infarct, edema, hemorrhage rates, metalloproteinase activation, and microvascular injury. Results: SOD2-KO mice demonstrate delayed (⬎24h) blood–brain barrier breakdown associated with activation of matrix metalloproteinases, inflammation, and high brain hemorrhage rates. These adverse consequences are absent in wild-type littermates and minocycline-treated SOD2-KO animals. Increased hemorrhage rates also are absent in SOD2 overexpressors, which have reduced vascular endothelial cell death. Finally, we show that the tight junction membrane protein, occludin, is an early and specific target in oxidative stress-induced microvascular injury. Interpretation: This model is ideal for studying ischemia/reperfusion-induced vascular injury and secondary brain hemorrhage and offers a unique opportunity to evaluate antioxidant-based neurovascular protective strategies as potential adjunct treatments to currently approved stroke therapies such as thrombolysis and endovascular clot retrieval. Ann Neurol 2006;59:929 –938 The World Health Organization (www.who.int) estimates that 15 million people suffer a stroke each year, resulting in an annual mortality rate of 5 million and an additional 5 million people suffering permanent disability. Nearly 80% of strokes are due to occlusion of a cerebral artery by a thrombus. Early restoration of cerebral blood flow (reperfusion) can salvage hypoperfused brain tissue, thus limiting neurological disability. Reperfusion strategies have proved to be the most effective therapies for stroke treatment. In fact, the only two stroke therapies approved by the US Food and Drug Administration are a thrombolytic agent that can dissolve occlusive thrombi (tissue-plasminogen activator) and a mechanical device that is capable of direct retrieval of thrombi from within cerebral vessels (Merci Concentric Retriever). One of the principal limitations of these treatment strategies is that early reperfusion of ischemic brain tissue can result in harmful consequences, including breakdown of the blood–brain barrier, which can lead to cerebral edema, brain hemorrhage, or both. Hemorrhages after reperfusion are particularly devastating and associated with extremely high morbidity and mortality. Fear of reperfusionrelated hemorrhage substantially limits the use of stroke therapies; it is estimated that only 2 to 3% of stroke patients in the United States receive acute reperfusion therapy. The deleterious consequences of early restoration of cerebral blood flow after stroke are attributed to reperfusion injury, a process that further damages brain cells, the ischemic arterial wall, and the microvasculature.1–3 Yet, little is known about the pathophysiological mechanisms of reperfusion injury. This knowledge gap stems, in part, from the lack of animal models to facilitate identification of specific components of the injury process and to allow evaluation of therapeutic interventions. Free radicals, specifically reactive oxygen species (ROS) that are generated soon after vessel occlusion, as well as in later stages of ischemic reperfusion (eg, by inflammatory cells), are the fundamental mediators of reperfusion injury. Here, we use an innovative animal model to investigate the cellular and molecular events From the Departments of Neurosurgery and Neurology and Neurological Sciences and Program in Neurosciences, Stanford University School of Medicine, Stanford, CA. Published online May 3 2006 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/ana.20850 Received Nov 8, 2005, and in revised form Feb 27, 2006. Accepted for publication Mar 4, 2006. Address correspondence to Dr Chan, Neurosurgical Laboratories, Stanford University, 1201 Welch Road, MSLS P314, Stanford, CA 94305-5487. E-mail: email@example.com Published 2006 by Wiley-Liss, Inc., through Wiley Subscription Services 929 associated with the production of ROS and implicated in ischemia/reperfusion-induced blood–brain barrier disruption and intracerebral hemorrhage. Knock-out (KO) mice with targeted disruption of the inducible mitochondrial manganese-containing superoxide dismutase (SOD2) are subjected to a mild stroke followed by early reperfusion and up to 3-day survival. These heterozygous SOD2-KO mice (SOD2⫺/⫹) are more susceptible to ischemic damage than their wild-type (WT) counterparts and exhibit a significant increase in matrix metalloproteinase-9 (MMP9) expression after ischemia/reperfusion injury. The increase in MMP9 expression, which is particularly evident in blood vessels and inflammatory cells, precedes and is spatially associated with a substantial increase in brain hemorrhage rates. The tight junction transmembrane protein occludin is particularly susceptible to ROS- and MMP9-induced damage in SOD2-KO mice and may explain, in part, the increased microvascular injury observed in these animals. In contrast, animals overexpressing SOD2 (SOD2 transgenics [SOD2⫹/⫺]) and SOD2-KO mice treated with minocycline, a broadspectrum antibiotic with antiinflammatory properties, have significantly diminished hemorrhage rates, reduced infarct size, and improved survival. Materials and Methods The Stanford Institutional Administrative Panel on Laboratory Animal Care approved all animal protocols, and their guidelines were followed for all procedures. Focal Cerebral Ischemia and Drug Regimen Heterozygous SOD2-KO mice4 (CD1/SV129 background, backcrossed with CD1 for more than 10 generations), heterozygous transgenic mice overexpressing SOD25 (C57BL/6J background, backcrossed with C57BL/6 for more than 10 generations), and their respective WT littermates were used. Mice (32–35gm) anesthetized with 2% isoflurane (30% oxygen/70% nitrous oxide) and maintained at a surgical plane of anesthesia (1–1.5% isoflurane) underwent middle cerebral artery occlusion by the intraluminal suture method for 30 minutes (body temperature maintained at 37 ⫾ 0.5°C). Reperfusion was allowed to occur in animals for 24 or 72 hours, and they had free access to food and water; then they were killed with an isoflurane overdose at the desired end point. For the drug study, half the mice (SOD2-KO animals and WT littermates) received minocycline (day 1: at 1 and 4 hours after insult, 45mg/kg intraperitoneally; days 2 and 3: single dose of 22.5mg/kg), whereas the other half received equal amounts of saline. Histopathology, Perl’s Iron Staining, Immunohistochemistry, and Terminal Deoxynucleotidyl Transferase Deoxyuridine Triphosphate Nick End Labeling Staining Annals of Neurology Vol 59 No 6 Evans Blue Extravasation Six hours before death, 2.5ml/kg of 4% Evans blue (Sigma) in 0.9% saline was injected into every animal. Animals were anesthetized and transcardially perfused with 200ml heparinized saline followed by 3.7% paraformaldehyde. Brains were removed, fixed overnight in 3.7% paraformaldehyde, and sectioned with a vibratome at 20m. Sections were incubated (10 minutes) with 2g/ml Hoechst 33258 (Molecular Probes, Carlsbad, CA) for nuclear counterstaining and mounted, and blood–brain barrier integrity was assessed by fluoromicroscopic detection of Evans blue extravasation. Western Blot Analysis After transcardiac perfusion with heparinized saline (10U/ml) followed by 3.7% paraformaldehyde, the brains were removed, 930 postfixed (3 days), paraffin-embedded, sectioned (6m-thick coronal sections), and stained with hematoxylin and eosin. Infarct was evaluated in the hemisphere, cortex, and striatum; histological criteria included areas of pan-necrosis with shrunken dark neurons and glial pallor. Infarct areas (seven coronal levels6) were determined using an image-analysis system (MCID; Imaging Research, Linton, United Kingdom) and corrected for edema.7 Diaminobenzidine-enhanced Perl’s iron staining was conducted in deparaffinized tissue by incubation in 1% KFeCN/1% HCl (15 minutes) followed by methyl green counterstain. For immunofluorescence, frozen brain sections treated for endogenous peroxidases and blocked for avidin-biotin were used. Sections were incubated with primary antibody (MMP9, anti-rabbit and anti-human, 1:100 [Chemicon, Temecula, CA]; NeuN [Chemicon]; glial fibrillary acidic protein [GFAP; Sigma, St. Louis, MO]; factor VIII [Research Lab, South Bend, IN]; myeloperoxidase [Dako, Denmark]; CD11b, 1:200; 4°C, 1 hour [Accurate, Westbury, NY]) followed by secondary antibody (1:200; 1-hour paraffin sections; 10-minute frozen sections; 25°C). In the paraffinembedded sections, MMP9 incubation lasted 72 hours; antibodies were detected using a Vector-ABC kit and colorized with Vector-VIP (Vector Laboratories, Burlingame, CA). Vascular structures and microglia were identified with biotinylated isolectin-B4 (IB4; 3-h incubation; Vector); reaction products were detected with Vector-ABC kit and colorized with diaminobenzidine. Both colorizing agents were also used for identification of different cell types or for colocalizing protein expression through double labeling. Fluorescein avidin-DCS, Texas-Red Avidin-D, or AMCA Avidin-D was used for frozen sections; tissue was mounted with 4⬘,6 diamidino-2-phenylindole–containing medium. Negative control samples were run in parallel using adjacent sections incubated without a primary antibody. Terminal deoxynucleotidyl transferase deoxyuridine triphosphate nick end labeling staining was performed using the ApopTag peroxidase system (Intergen, Burlington, MA). In brief, sections were deparaffinized, treated with proteinase K (20g/ml for 15 minutes) and for endogenous peroxidases, and subjected to incubation with terminal deoxynucleotidyl transferase for 1 hour (37°C). Sections were washed, then incubated with antidigoxigenin peroxidase conjugate for 30 minutes, followed by peroxidase substrate, additional washes, and counterstaining with methyl green. June 2006 Each hemisphere was homogenized in suspension buffer (1M KOH, 1M KCl, 1M MgCl2, 0.5 EDTA, 0.5M EGTA, su- crose, protease cocktail) and centrifuged (14,000g, 10 minutes, 4°C), and supernatants were transferred into equal volumes of sodium dodecyl sulfate buffer with ␤-mercaptoethanol and stored (⫺20°C). Aliquots were subjected to sodium dodecyl sulfate polyacrylamide gel electrophoresis (10 –20% Trisglycine gel; Invitrogen, Carlsbad, CA). Polyvinylidene difluoride membranes were incubated in methanol, then in transfer buffer (5 minutes each). Transfer (25V, 1.5 hours) was followed by incubation in 5% milk in phosphate-buffered salineTween (60 minutes). Membranes were incubated overnight (4°C) in primary antibody (listed earlier plus hemoglobin [Research Plus, Manasquan, NJ], aquaporin-4 [AQP4; Chemicon], occludin, claudin-5, zonula occludens-1 [Zymed, San Francisco, CA]; 1:100 dilution, 1% milk in phosphatebuffered saline-Tween) and then in peroxidase-conjugated secondary antibody (1:200; 1 hour, 25°C). Signals were detected with a chemiluminescence kit enzyme chemiluminescence (ECL)-plus Western Blotting Detection System (GE Healthcare, Piscataway, NJ). To confirm consistent protein loading for each lane, we stained membranes for ␤-actin. Films were scanned with a GS-700 imaging densitometer (Bio-Rad, Hercules, CA); quantitative analyses were performed using MultiAnalyst software (Bio-Rad). Microvessel Isolation Brains (n ⫽ 10 per isolation) were obtained from mice subjected to ischemia/reperfusion as described earlier. Meninges, large surface vessels, and corpus callosum were removed and discarded. The tissue was then minced and digested with collagenase/dispase (Roche, Basel, Switzerland) for 1 hour and centrifuged at 1,000g for 5 minutes. The pellet was resuspended in 20% BSA and centrifuged at 3,000g for 10 minutes and digested in collagenase/dispase/DNAse containing M199 medium for 1 hour. This was then washed, resuspended in Earle’s Balanced Salt Solution (without phenol red) and centrifuged in a 50% Percoll gradient at 3,000g for 10 minutes. The band containing microvessels was collected, resuspended in 6M urea lysis buffer with a protease inhibitor cocktail overnight, then centrifuged, and protein concentrations were determined8,9 (with the above modifications). Mouse Cerebral Endothelial Cell Culture Cerebral cortices of 3-week-old WT and SOD2⫺/⫹ mice were isolated and then treated as described earlier with some modifications. The band containing microvessels was collected and plated onto collagen-coated dishes in endothelial cell medium with supplements (BD Biosciences, San Jose, CA) and grown to confluence before use. Treatment of Cultures with Oxygen-Glucose Deprivation The endothelial cells from both WT and SOD2⫺/⫹ mice were subjected to oxygen-glucose deprivation by replacing the medium in the cell culture plates with buffered salt solution without glucose and placing the cell culture plates in an anaerobic chamber (Plas Labs, Lansing, MI) at 37°C for 8 hours. The oxygen-glucose deprivation period was terminated by replacing the buffered salt solution with endothelial cell medium and placing the cell culture plates in the 5%CO2/95% air incubator for different reoxygenation periods. Two wells were examined per condition. Analysis of Brain Vasculature The integrity of the circle of Willis of SOD2⫺/⫹ and SOD2⫹/⫺ mice and their respective littermates was examined by carbon black injection. The animals were anesthetized with isoflurane and perfused through the left ventricle with heparinized saline and subsequently with a mixture of carbon black and 10% gelatin. The brain was removed and fixed in 3.7% paraformaldehyde for 24 hours. The cerebral vasculature was photographed with a dissecting microscope. Statistical Analyses Statistical analyses were done with one-way analysis of variance followed by two-tailed Student’s t test (two-group comparisons) and with the nonparametric Wilcoxon signed rank test for noncontinuous data. All analyses were made with Stat-View 4.0 software (SAS Institute, Cary, NC). Data in the text and figures are expressed as mean ⫾ standard error of the mean; p less than 0.05 was considered significant. Results Superoxide Dismutase-2 Knock-out Mice Are Highly Susceptible to Ischemic Brain Injury, Edema, and Hemorrhage after Ischemia/Reperfusion We subjected SOD2-KO mice and their WT counterparts to 1 hour of middle cerebral artery occlusion followed by reperfusion and established that the mortality rate for the SOD2-KO animals increased from 10% at 24 hours to nearly 90% at the 72-hour end point. In contrast, the 72-hour mortality rate in the WT animals was only 10%. Examination of the brain specimens showed a clear correlation between the mortality rate and the occurrence of large intracerebral hemorrhages in SOD2-KO animals. This was in sharp contrast with heterozygous SOD1-KO mice, which had a mortality rate (63% at 96 hours) that was associated with severe cerebral edema but not hemorrhage. We then reduced the ischemic period to 30 minutes and demonstrated that the SOD2-KO mortality rate was reduced to 30% at 72 hours. Histopathology showed that these animals had a statistically significant increase in infarct size at all levels examined (Fig 1A) at both 24 and 72 hours compared with the WT mice, as well as an increase in the rate of hemorrhagic transformations (SOD2-KO: 8/9; WT: 3/8) at 72 hours, which was not observed at 24 hours (see Fig 1B). These hemorrhagic transformations were generally widespread throughout the ischemic tissue and often quite large in the SOD2-KO mice (see Figs 1B, C), and many were observed in the borderzone between infarct and surviving tissue. In contrast, the few (and typically small) hemorrhagic transformations observed in the WT littermates were found predominantly in the ventral portion of the ischemic hemisphere, near the entry point of the middle Maier et al: Reperfusion-Related Hemorrhage 931 Fig 1. Reperfusion-induced brain injury and hemorrhage and the effects of minocycline treatment. After 30 minutes of middle cerebral artery occlusion, superoxide dismutase-2 knock-out (SOD2-KO; black bars) animals had a statistically significant increase in infarct size at all levels examined (A; *p ⬍ 0.05, analysis of variance) at both 24 and 72 hours of reperfusion compared with wild-type (WT; white bars) littermates. The use of 2,3,5-triphenyltetrazolium chloride to delineate infarcts showed that SOD2-KO animals had increased intracerebral hemorrhage rates at 72 hours of reperfusion, which were absent at 24 hours, as well as in most of the WT littermates (B). These results were confirmed by Western blot analysis (B, bottom), which showed increased brain hemoglobin values, indicating intracerebral hemorrhage, in the SOD2-KO mice compared with WT littermates at 72 hours of reperfusion (optical density measurements for SOD2-KO: 174 ⫾ 34; WT: 29 ⫾ 18; p ⫽ 0.009, Student’s t test). Additional confirmation came from Perl’s iron staining (brown stain with methyl green counterstain), which showed that hemorrhagic transformations, although generally widespread throughout the ischemic tissue (C), were often observed in the borderzone between infarct and surviving tissue (dashed line delineates the infarct). In addition, several SOD2-KO mice (5/9) had large hemorrhagic transformations (D) associated with microvascular leakage (left inset) and activated macrophages (right inset). SOD2-KO mice also had a significant increase in cerebral edema compared with the WT littermates at the 72-hour end point (E; *p ⫽ 0.018, Student’s t test); which was associated with a significant increase in aquaporin-4 (AQP4) expression (F; *p ⫽ 0.006, Student’s t test). Treatment with minocycline reduced infarct size at the 72-hour end point in the SOD2-KO mice and WT littermates at all levels examined compared with saline treatment (G; *p ⬍ 0.05, analysis of variance). Perl’s iron staining also demonstrated a reduction in hemorrhagic transformations in the minocycline-treated SOD2-KO mice (H). Ctx ⫽ cortex; Hem ⫽ hemisphere; St ⫽ striatum. Scale bar ⫽ 20m. cerebral artery, suggesting that these lesions resulted from mild mechanical damage to the occluded vessel. Brain edema, which plays a critical role in the pathophysiology and morbidity of stroke, was increased by 55% in SOD2-KO mice compared with the WT littermates at the 72-hour end point ( p ⫽ 0.018; see Fig 1E). The increase in cerebral edema in these animals was associated with a significant increase in AQP4 expression ( p ⫽ 0.006; see Fig 1F), a water-selective transporting protein that is abundantly expressed at the blood–brain barrier (in perivascular and endothelial cells) and at the interface of brain tissue and cerebrospinal fluid.10 –12 We previously have shown that AQP4 deletion in mice reduces brain edema after an ischemic stroke.13 Whether the increase in AQP4 expression after ROS-induced reperfusion injury is involved in exacerbating edema formation or occurs as the brain’s attempt to facilitate clearance of excess water remains to be determined. 932 Annals of Neurology Vol 59 No 6 June 2006 We previously have shown that oxidative stress is involved in mediating blood–brain barrier disruption during the first 3 to 7 hours of reperfusion after an ischemic event through MMP activation.14 In addition, by comparing SOD2-KO with WT control animals, we demonstrated that after 1 hour of middle cerebral artery occlusion, SOD2 reduces apoptotic cell death.15 Moreover, SOD2 attenuates cytochrome c release from the mitochondria to the cytosol, a critical step in the intrinsic mitochondrial-dependent signaling pathway of the cell death program.16 These findings, in conjunction with the remarkably increased 72-hour hemorrhage rate observed in the SOD2-KO animals, led us to hypothesize that ischemia/reperfusion caused increased ROS production in mitochondria, leading to MMP activation, blood– brain barrier breakdown, vasogenic edema, and delayed ischemic neuronal damage. Minocycline Reduces Infarct Size, Hemorrhage Rate, Mortality, and Matrix Metalloproteinase-9 Expression To test our hypothesis and validate our model, we treated SOD2-KO mice and their WT littermates with minocycline, a commonly used tetracycline derivative that crosses the blood–brain barrier and reduces cytochrome c release from mitochondria17 and also has been shown to attenuate stroke severity.18 Minocyclinetreated mice (SOD2-KO and WT littermates) had a significant reduction in infarct size at 72 hours of reperfusion compared with saline-treated animals (see Fig 1G). In addition, minocycline treatment reduced the hemorrhage rate in SOD2-KO mice (3/11 vs 7/9 treated with saline; see Fig 1H), as well as the mortality rate (2/13 vs 4/13 treated with saline). Because reactive microglia pro- duce MMP919 and minocycline has been shown to reduce microglial activation,20 we examined the effects of minocycline on MMP9 expression in our model. We found that minocycline significantly reduced MMP9 expression in SOD2-KO mice at 24 and 72 hours of reperfusion (Fig 2), with the relative reduction being particularly significant at the latter time point (75% relative reduction at 72 hours; see Fig 2D). Superoxide Dismutase-2 Knock-out Mice Have Increased Matrix Metalloproteinase-9 Expression in the Vascular Endothelium The ischemia/reperfusion-induced increase in blood– brain barrier disruption has been linked to inflammatory events that involve ROS production and Fig 2. Effects of minocycline on ischemia/reperfusion-induced matrix metalloproteinase-9 (MMP9) expression. Western blot analysis with a rabbit anti–mouse MMP9 antibody (105kDa band) in whole-brain homogenates showed that minocycline (Min) treatment significantly reduced MMP9 expression in superoxide dismutase-2 knock-out (SOD2-KO) mice at 24 (*p ⫽ 0.03) and 72 hours (†p ⫽ 0.05) of reperfusion compared with saline (Sal) treatment (A), although this effect was not significant in the wild-type (WT) littermates at 72 hours of reperfusion (B). We also tested a rabbit anti–human MMP9 antibody (C). Pro-MMP9 was evident as 92kDa characteristic bands in the saline- and minocycline-treated animals, whereas activated MMP9 was detected as 88kDa bands. Similar results were obtained from independent studies (n ⫽ 3) for A and B. (D) The overall relative reduction in MMP9 expression with minocycline treatment (from three separate experiments). Maier et al: Reperfusion-Related Hemorrhage 933 leukocyte-endothelial cell adhesion and migration. This is supported by studies that show that agents targeting either the generation of ROS by endothelial cells and/or the adhesion of leukocytes to vascular endothelium are generally effective in blunting the ischemia/ reperfusion-induced increase in microvascular permeability.21 We recently determined that leukocytes, specifically neutrophils that comprise more than 50% of circulating leukocytes, are not the primary source of MMP9 protein in the brain even at the peak of leukocyte infiltration (24 –72 hours after stroke onset)6 and, thus, are unlikely to be the key contributor to the blood–brain barrier breakdown observed in SOD2-KO mice. In contrast, MMP9 immunoreactivity in vascular structures (found throughout the infarcted hemisphere but most prominently in the borderzone between infarct and surviving tissue) was significantly stronger in SOD2-KO mice compared with WT littermates. To further study the relationship among ROS production, MMP9 expression, and blood–brain barrier disruption, we injected Evans blue into the jugular vein of the SOD2-KO and WT mice after ischemia/reperfusion. Extravasation of Evans blue from vessels, which indicates vascular permeability, was detected throughout the ischemic hemisphere (Fig 3, red fluorescence). MMP9 immunofluorescence (see Fig 3, green fluorescence) was easily discerned in vessels showing Evans blue extravasation. MMP9 expression was particularly prominent in small vessels in the infarct border (see Fig 3B), but could also be observed in vascular structures within severely damaged areas in the center of the infarct, the ischemic core (see Fig 3C). MMP9immunoreactive cells were also observed in the corpus callosum (see Figs 3D, E), as well as in larger vessels such as the middle cerebral artery (see Figs 3F, G). Reactive Oxygen Species Mediate Increases in Endothelial Matrix Metalloproteinase-9 and Damage to Microvascular Endothelial Tight Junctions The brain endothelial cells that form the blood–brain barrier have tight junctions that are critical for maintaining brain homeostasis and restricting permeability. Some of the major components of tight junction complexes are zonula occludens-1, occludin, and claudin-5, all of which are targets of MMP9 degradation. The volume of the brain capillary endothelium forming the blood–brain barrier is approximately 10⫺3 parts of the whole brain,22 thus making detection of small changes Fig 3. Spatial relationship between Evans blue extravasation and matrix metalloproteinase-9 (MMP9) expression. Representative photomicrographs show colocalization of Evans blue (red fluorescence) and MMP9 (green fluorescence) in a superoxide dismutase-2 knock-out (SOD2-KO) animal subjected to 30 minutes of focal cerebral ischemia and 72 hours of reperfusion. Extravasation of Evans blue from vessels was detected throughout the ischemic territory (A). MMP9 immunofluorescence was evident in small vessels showing Evans blue extravasation in the infarct border (B) where small hemorrhages could be detected (arrow and inset) and within severely damaged areas in the center of the infarct, the ischemic core (C). MMP9 immunoreactivity was also observed in the corpus callosum (D), although vessels in this region were not always immunoreactive (E; inset shows the negative control), as well as in larger vessels such as the middle cerebral artery (F, G). Scale bar ⫽ 20m. 934 Annals of Neurology Vol 59 No 6 June 2006 in tight junction components difficult in whole-brain homogenates. To address this issue, we adapted a microvessel isolation technique that allows us to isolate free radical–mediated blood–brain barrier–specific events after ischemia/reperfusion injury. To generate sufficiently large protein concentrations for the Western blots, we pooled a total of 10 brains for each isolation procedure (1 lane ⫽ 10 brains). The end product (microvessel pellet) was tested for factor VIII (an endothelial cell marker), NeuN (a neuronal marker), and glial fibrillary acidic protein, a glial marker, to ensure successful isolation (Figs 4A, B). As expected, we found progressive degradation of zonula occludens-1, occludin, and claudin-5 after ischemia/reperfusion injury. The SOD2-KO animals were particularly sensitive to occludin degradation and showed an associated increase in MMP9 expression (see Fig 4C) compared with WT littermates (see Fig 4C). Results from five separate experiments showed that MMP9 expression in SOD2-KO animals was, on average, 21% greater at 24 hours and 14% greater at 72 hours than in WT littermates. To confirm our MMP9 results from the microvessel isolations, we also tested the effects of oxygen-glucose deprivation on endothelial cell cultures from the SOD2-KO and WT mice. After 8 hours of oxygenglucose deprivation, which resulted in 50% cell death, and 0 to 8 hours of reoxygenation, endothelial cells from the SOD2-KO animals showed a marked increase in MMP9 expression compared with endothelial cell cultures from the WT littermates (see Fig 4E). By 24 hours of reoxygenation, there was a reduction in MMP9 levels in the SOD2-KO endothelial cell cultures that was inversely proportional to cell death in those cultures. WT endothelial cell cultures, in contrast, showed a substantial increase in MMP9 levels and only a small increase in cell death at the same time point. Superoxide Dismutase-2 Protects against Blood–Brain Barrier Damage after Ischemia/Reperfusion Injury As proof of principle, we subjected transgenic mice overexpressing SOD2 (which have a 2.5-fold increase in SOD2 activity) and their WT littermates to the same ischemia/reperfusion paradigm. Histopathology and Perl’s iron staining showed that the rate of hemorrhagic transformations at 72 hours was quite low in both groups (SOD2 transgenic: 3/9; WT: 3/9) and was restricted mostly to the entry point of the middle cerebral artery and, thus, likely due to mechanical damage by the occluding filament. Because genetic mutations can affect vessel anatomy and, hence, stroke outcome, we analyzed the brain vasculature of the mice by carbon black injections. All animals showed an intact circle of Willis (Fig 5), indicating that there were no differences in gross vascular anatomy among the SOD2-KO mice, transgenic animals overexpressing SOD2, and their respective WT counterparts. However, when we examined the microvasculature within the infarcted area, we observed qualitatively more cells stained positive for terminal deoxynucleotidyl transferase deoxyuridine triphosphate nick end labeling in microvessels of the SOD2-KO mice compared with their WT littermates, whereas such cells in the transgenic animals overexpressing SOD2 were rare. Discussion SOD2-KO mice and the newly developed transgenic mouse overexpressing SOD2 provide a unique opportunity to elucidate, in a molecular fashion, the oxidative mechanisms that mediate brain injury after stroke and mitochondria-related blood–brain barrier dysfunctions. Furthermore, SOD2-KO animals can facilitate the development and evaluation of treatment strategies that diminish the complications of cerebral reperfusion. This is particularly important given the approaches being investigated to improve stroke treatment. One strategy for increasing the number of patients eligible for thrombolysis consists of delaying the progression of stroke by delivering high-flow oxygen to patients before vessel recanalization.23 Oxygen therapy can enhance the formation of ROS and, thus, increase blood– brain barrier damage, but recent work suggests that this therapy may extend the reperfusion window when treatment is administered during the ischemic episode.24 The effects of combining oxygen therapy with thrombolytic agents or clot retrieval devices have not yet been determined. In this study, we observed that SOD2-KO mice had a high rate of reperfusion-related brain hemorrhage, larger infarct volumes, and cerebral edema. We hypothesized that these hemorrhages and the amplification of ischemic injury were caused by increased production of ROS in mitochondria, leading to activation of MMPs, blood–brain barrier breakdown, vasogenic edema, and delayed ischemic neuronal damage. To test our hypothesis, we treated SOD2-KO mice with minocycline, an agent with a primary mechanism of action that is a direct inhibition of cytochrome c release from mitochondria, followed by inhibition of downstream cell death–related mechanisms.17 Minocycline treatment led to reduced hemorrhage rates, infarct volumes, and mortality. Although we cannot exclude the possibility that the reduction in microvascular injury was due, in part, to a reduction in infarct size, our findings demonstrate that pharmacological agents have the potential to attenuate reperfusion-related brain injury and that our model can be used to evaluate putative therapeutic agents. To clarify the mechanisms underlying the increased brain injury observed in the SOD2-KO mice, we developed a unique technique for microvessel isolation af- Maier et al: Reperfusion-Related Hemorrhage 935 Fig 4. Microvessel isolations and endothelial cell cultures after ischemia/reperfusion injury. Western blot analysis of two representative microvessel isolation experiments (A and B are the graphic representations of the Western blots shown in D). After 30 minutes of middle cerebral artery occlusion and 24 or 72 hours of reperfusion, microvessels were isolated from superoxide dismutase-2 knock-out (SOD2-KO) mice and wild-type (WT) littermates (n ⫽ 10 per group for each isolation). The microvessel pellets were tested for factor VIII, NeuN, and glial fibrillary acidic protein (GFAP) to ensure a clean isolation. Although some variability was observed, our results showed little contamination from either neuronal or glial cells. The samples then were tested for zonula occludens-1 (ZO-1), occludin, and claudin-5 expression, and our results showed progressive degradation of all three proteins during evolution of the infarct. The SOD2-KO animals were particularly sensitive to occludin degradation (A, B). (C) The optical density (OD) measurements of MMP9 expression from optimal microvessel isolation. The SOD2-KO mice showed an increase in MMP9 expression compared with WT littermates at 24 and 72 hours of reperfusion. (E) We obtained similar results when we subjected cerebral capillary endothelial cell cultures from the SOD2-KO and WT mice to 8 hours of oxygen-glucose deprivation, which resulted in 50% cell death, followed by reoxygenation. Endothelial cells from the SOD2-KO animals showed a marked increase in MMP9 expression at the end of the ischemic period and up to 8 hours of reoxygenation. By 24 hours, MMP9 levels declined in the SOD2-KO endothelial cells, which paralleled a concomitant increase in cell death in those cultures. MMP9 expression in the WT endothelial cell cultures did not increase until 24 hours of reoxygenation, at which point most cells were still viable. WB ⫽ whole brain. 936 Annals of Neurology Vol 59 No 6 June 2006 Fig 5. Vascular integrity as assessed by carbon black injections and terminal deoxynucleotidyl transferase deoxyuridine triphosphate nick end labeling (TUNEL) staining. Analysis of the brain vascular anatomy showed an intact circle of Willis in the superoxide dismutase-2 knock-out mice (SOD2⫺/⫹) and SOD2 overexpressors (SOD2⫹/⫺) and their respective WT littermates (A–D, top). Representative photomicrographs from animals subjected to 30 minutes of middle cerebral artery occlusion and 24 hours of reperfusion (middle and bottom rows). The SOD2-KO mice showed increased TUNEL staining (brown; methyl green counterstain) throughout the infarcted area compared with WT littermates. Arrowheads indicate TUNEL-positive endothelial cells in small vessels. TUNEL immunoreactivity was rarely seen in vascular structures from the SOD2 transgenic mice. Scale bar ⫽ 20m. ter in vivo ischemia/reperfusion. This technique allowed us to verify the existence of blood–brain barrier injury and demonstrate that this injury was likely mediated by an increase in MMP9 expression. Furthermore, this technique allowed us to establish that, in endothelial cells, one of the key consequences of mitochondrial oxidative stress and the ensuing overexpression of MMP9 is damage to the tight junction transmembrane protein occludin. This finding supports previous work demonstrating that occludin can be selectively cleaved by metalloproteinases,25 and that phosphorylation of occludin appears to be a key regulator of tight junction permeability.26 Finally, although reducing the capacity of mitochondria to scavenge superoxide radicals in endothelial cells increases susceptibility to damage and death, we were able to demonstrate that overexpression of SOD2 has a protective effect on the microvasculature. Whether this increase in mitochondrial antioxidant capacity can enhance cell-survival mechanisms is a crucial question that is currently under investigation. The availability of transgenic mice overexpressing SOD2 allows us to perform direct comparisons of the microvascular responses between these mice and SOD2-KO mice after ischemia and, thus, to examine the specific molecular and cellular mechanisms that underlie ischemia/reperfusion-related brain injury. The techniques described here will facilitate these discoveries and ultimately translate into new therapeutic interventions for stroke patients. This work was supported by American Heart Association Postdoctoral Fellowship (0120142Y, C.M.M.), the NIH (National Institute of Neurological Disorders and Stroke, NS 14534, NS 25372, NS 36147, NS 38653, P.H.C.), and an American Heart Association Bugher Foundation Award (P.H.C.). We thank R. Sobel for expert advice, T.-T. Huang and A. Richardson for helping us set up the SOD2 transgenic colony, T.P. Davis and collaborators for their expert advice in microvessel isolation procedures, G.W. Albers for his expert advice and editorial assistance, P. Bracci for statistical analysis, C. Christensen for editorial assistance, and E. Hoyte for preparation of the figures. Maier et al: Reperfusion-Related Hemorrhage 937 References 1. Aronowski J, Strong R, Grotta JC. Reperfusion injury: demonstration of brain damage produced by reperfusion after transient focal ischemia in rats. J Cereb Blood Flow Metab 1997;17: 1048 –1056. 2. Brott T, Broderick JP, Kothari R. Thrombolysis for acute ischemic stroke. In: Welch KMA, Caplan LR, Reis DJ, et al., eds. Primer on cerebrovascular disease. San Diego: Academic Press, 1997:728 –731. 3. Hamann GF, Okada Y, Fitridge R, del Zoppo GJ. Microvascular basal lamina antigens disappear during cerebral ischemia and reperfusion. Stroke 1995;26:2120 –2126. 4. Li Y, Huang TT, Carlson EJ, et al. Dilated cardiomyopathy and neonatal lethality in mutant mice lacking manganese superoxide dismutase. Nat Genet 1995;11:376 –381. 5. Epperly MW, Travis EL, Whitsett JA, et al. Overexpression of manganese superoxide dismutase (MnSOD) in whole lung or alveolar type II cells of MnSOD transgenic mice does not provide intrinsic lung irradiation protection. Int J Cancer 2001;96: 11–21. 6. Maier CM, Hsieh L, Yu F, et al. Matrix metalloproteinase-9 and myeloperoxidase expression: quantitative analysis by antigen immunohistochemistry in a model of transient focal cerebral ischemia. Stroke 2004;35:1169 –1174. 7. Swanson RA, Morton MT, Tsao-Wu G, et al. A semiautomated method for measuring brain infarct volume. J Cereb Blood Flow Metab 1990;10:290 –293. 8. Pardridge WM, Eisenberg J, Yamada T. Rapid sequestration and degradation of somatostatin analogues by isolated brain microvessels. J Neurochem 1985;44:1178 –1184. 9. Zlokovic BV, Mackic JB, Wang L, et al. Differential expression of Na,K-ATPase alpha and beta subunit isoforms at the bloodbrain barrier and the choroid plexus. J Biol Chem 1993;268: 8019 – 8025. 10. Amiry-Moghaddam M, Ottersen OP. The molecular basis of water transport in the brain. Nat Rev Neurosci 2003;4: 991–1001. 11. Frigeri A, Gropper MA, Turck CW, Verkman AS. Immunolocalization of the mercurial-insensitive water channel and glycerol intrinsic protein in epithelial cell plasma membranes. Proc Natl Acad Sci U S A 1995;92:4328 – 4331. 12. Nielsen S, Nagelhus EA, Amiry-Moghaddam M, et al. Specialized membrane domains for water transport in glial cells: highresolution immunogold cytochemistry of aquaporin-4 in rat brain. J Neurosci 1997;17:171–180. 938 Annals of Neurology Vol 59 No 6 June 2006 13. Manley GT, Fujimura M, Ma T, et al. Aquaporin-4 deletion in mice reduces brain edema after acute water intoxication and ischemic stroke. Nat Med 2000;6:159 –163. 14. Gasche Y, Copin JC, Sugawara T, et al. Matrix metalloproteinase inhibition prevents oxidative stress-associated blood-brain barrier disruption after transient focal cerebral ischemia. J Cereb Blood Flow Metab 2001;21:1393–1400. 15. Noshita N, Sugawara T, Fujimura M, et al. Manganese superoxide dismutase affects cytochrome c release and caspase-9 activation after transient focal cerebral ischemia in mice. J Cereb Blood Flow Metab 2001;21:557–567. 16. Chan PH. Mitochondria and neuronal death/survival signaling pathways in cerebral ischemia. Neurochem Res 2004;29: 1943–1949. 17. Zhu S, Stavrovskaya IG, Drozda M, et al. Minocycline inhibits cytochrome c release and delays progression of amyotrophic lateral sclerosis in mice. Nature 2002;417:74 –78. 18. Yong VW, Wells J, Giuliani F, et al. The promise of minocycline in neurology. Lancet Neurol 2004;3:744 –751. 19. Rivera S, Ogier C, Jourquin J, et al. Gelatinase B and TIMP-1 are regulated in a cell- and time-dependent manner in association with neuronal death and glial reactivity after global forebrain ischemia. Eur J Neurosci 2002;15:19 –32. 20. Yrjanheikki J, Tikka T, Keinanen R, et al. A tetracycline derivative, minocycline, reduces inflammation and protects against focal cerebral ischemia with a wide therapeutic window. Proc Natl Acad Sci U S A 1999;96:13496 –13500. 21. Granger DN. Ischemia-reperfusion: mechanisms of microvascular dysfunction and the influence of risk factors for cardiovascular disease. Microcirculation 1999;6:167–178. 22. Li JY, Boado RJ, Pardridge WM. Blood-brain barrier genomics. J Cereb Blood Flow Metab 2001;21:61– 68. 23. Singhal AB, Benner T, Roccatagliata L, et al. A pilot study of normobaric oxygen therapy in acute ischemic stroke. Stroke 2005;36:797– 802. 24. Kim HY, Singhal AB, Lo EH. Normobaric hyperoxia extends the reperfusion window in focal cerebral ischemia. Ann Neurol 2005;57:571–575. 25. Wachtel M, Frei K, Ehler E, et al. Occludin proteolysis and increased permeability in endothelial cells through tyrosine phosphatase inhibition. J Cell Sci 1999;112(pt 23):4347– 4356. 26. Hirase T, Kawashima S, Wong EY, et al. Regulation of tight junction permeability and occludin phosphorylation by Rhoap160ROCK-dependent and -independent mechanisms. J Biol Chem 2001;276:10423–10431.