Therapeutic Administration of Plasminogen Activator Inhibitor-1 Prevents Hypoxic–Ischemic Brain Injury in Newborns

Disruption of the integrity of the blood–brain barrier (BBB) is an important mechanism of cerebrovascular diseases, including neonatal cerebral hypoxia–ischemia (HI). Although both tissue-type plasminogen activator (tPA) and matrix metalloproteinase-9 (MMP-9) can produce BBB damage, their relationship in neonatal cerebral HI is unclear. Here we use a rodent model to test whether the plasminogen activator (PA) system is critical for MMP-9 activation and HI-induced brain injury in newborns. To test this hypothesis, we examined the therapeutic effect of intracerebroventricular injection of plasminogen activator inhibitor-1 (PAI-1) in rat pups subjected to unilateral carotid artery occlusion and systemic hypoxia. We found that the injection of PAI-1 greatly reduced the activity of both tPA and urokinase-type plasminogen activator after HI. It also blocked HI-induced MMP-9 activation and BBB permeability at 24 h of recovery. Furthermore, magnetic resonance imaging and histological analysis showed the PAI-1 treatment reduced brain edema, axonal degeneration, and cortical cell death at 24–48 h of recovery. Finally, the PAI-1 therapy provided a dose-dependent decrease of brain tissue loss at 7 d of recovery, with the therapeutic window at 4 h after the HI insult. Together, these results suggest that the brain PA system plays a pivotal role in neonatal cerebral HI and may be a promising therapeutic target in infants suffering hypoxic–ischemic encephalopathy.


Introduction
Recent studies indicate that neurovascular proteases, including matrix metalloproteases (MMPs) and secreted serine proteases, play a critical role in cerebrovascular diseases (Mun-Bryce and Rosenberg, 1998;Lo et al., 2004). This is because dysregulation of neurovascular proteases degrades the extracellular matrix (ECM) and blood-brain barrier (BBB), leading to brain edema, leukocyte infiltration, and the neuron-matrix detachment. In adult cerebral ischemia, MMP-9 has been suggested to be a promising therapeutic target, because its activity arises early after ischemia and either genetic or pharmacological inhibition of MMP-9 offers brain protection in animal models (Heo et al., 1999;Asahi et al., 2001;Gu et al., 2005). In contrast, the role of MMP-9 in neonatal cerebral hypoxia-ischemia (HI) is unclear. This is because the MMP-9 activity was only detectable at 24 h of recovery, when some irreversible brain damage has already occurred (Svedin et al., 2007). Thus, it seems unlikely that MMP-9 could be a key initiator of HI brain injury in newborns.
Tissue-type plasminogen activator (tPA) is another important neurovascular protease that may contribute to neonatal HI brain injury. tPA mainly circulates in the blood but also exists at a low level in the brain parenchyma, which can be induced by excitotoxins (Sappino et al., 1993;Tsirka et al., 1995). tPA directly triggers the opening of BBB through activation of the latent platelet-derived growth factor C and elevates the MMP-9 levels after stroke (Yepes et al., 2000;Sumii and Lo, 2002;Wang et al., 2003;Su et al., 2008). Furthermore, we recently reported that cerebral HI produces rapid (Ͻ1 h) and persistent (up to 24 h) induction of tPA surrounding the blood vessels and lateral ventricles in newborn brains (Adhami et al., 2008). The early induction of tPA suggests that it may be an initiator of HI brain injury in newborns.
To test this hypothesis, we used intracerebroventricular (ICV) injection of plasminogen activator inhibitor-1 (PAI-1) to block the parenchymal tPA and urokinase-type plasminogen activator (uPA) activity in the Vannucci model of cerebral HI in rat pups and examined its effects on MMP-9 activation and brain damage. Results of these experiments supported our hypothesis and suggested that plasminogen activators (PAs) are potential therapeutic targets in neonatal HI brain injury.

Materials and Methods
Animal surgery and quantification of tissue loss. Seven-day-old Wistar rat pups were used for the cerebral ischemia-hypoxia model and ICV injection as described previously (Rice et al., 1981;Adhami et al., 2008). The procedures were approved by the Institutional Animal Care and Use Committee. The percentage of tissue loss in the cerebral cortex, hippocampus, and striatum was quantified by comparison with individual counterparts on the contralateral hemisphere.
Reagents and biochemistry. Plasminogen/casein zymogram, MMP zymogram, immunoblot, and immunocytochemistry were performed as described previously (Adhami et al., 2008). The procedure of quantifying Evans blue dye extravasation has been described previously (Su et al., 2008). A stable mutant form of human PAI-1 (Berkenpas et al., 1995) was purchased from Molecular Innovations.
Magnetic resonance imaging. All data were acquired on a Bruker BioSpec 7T system with 40 G/cm gradients using a custom-built 25 mm single-turn transmit/receive solenoid coil. Animals were scanned in two cohorts, with two each saline-treated or PAI-1-treated animals in the first cohort and three each saline-or PAI-1-treated animals in the second cohort. Animals were brought to the scanner 24 h after HI induction. They were anesthetized and maintained with 1% isoflurane in air and kept warm with heated air circulating through the magnet bore. T2weighted anatomical images were acquired using a two-dimensional RARE (rapid acquisition and relaxation enhancement) sequence with an effective echo time of 76.96 ms, repetition time of 1000 ms, field-of-view (FOV) of 19.2 ϫ 19.2 mm 2 , and a 256 ϫ 192 matrix size. Diffusion tensor images were acquired with a spin echo sequence using an echo time of 21 ms, repetition time of 1100 ms, b-value of 800, six diffusion directions, FOV of 19.2 ϫ 19.2 mm 2 , and a matrix of 128 ϫ 128. For the second cohort, T2 maps were calculated from data acquired with a spin echo sequence using echo times of 20, 40, 60, 80, and 100 ms at a repetition time of 1800 ms with the same FOV and matrix as used for the diffusion scan.
Diffusion data were processed using the Bruker online processing software to calculate apparent diffusion coefficient (ADC) maps and directionally color-encoded (DEC) maps of the fractional anisotropy (FA). T2 maps were calculated using the Bruker online data processing software.
Statistical analysis. Values are represented as mean Ϯ SD or SEM as indicated. Quantitative data were compared between different groups using Microsoft Excel two-sample (unpaired) t test assuming equal variance.

Inhibition of plasminogen activators blocks MMP-9 induction after HI
To examine the relationship between PAs and MMP-9 in neonatal cerebral HI, we performed ICV injection of saline or recombinant human PAI-1 (1.9 g per pup) ipsilateral to carotid ligation at the end of a 90 min hypoxic insult (10% O 2 ) in postnatal day 7 rat pups and collected the brains at 4 or 24 h recovery for biochemical analysis.
The immunoblot detected the ICV-injected PAI-1 on the HIchallenged side of brain (R) at 4 but not 24 h of recovery (n ϭ 4) ( Fig. 1 A, B). The plasminogen/casein zymogram showed a low level of basal tPA activity in the brain, which was increased on the HI-stressed side of brain at both 4 h (1.28-fold increase; p Ͻ 0.05) and 24 h (2.23-fold; p Ͻ 0.005; n ϭ 6) of recovery after salineinjection ( Fig. 1 A-C). Furthermore, the tPA induction was accompanied by an increase of the uPA activity at both time points after HI. In contrast, the induction of MMP-9 and MMP-2 was only detectable at 24 h but not 4 h of recovery. The observed timings of MMP-9 induction were consistent with a previous report (Svedin et al., 2007).
However, in contrast to saline injection, the PAI-1 injection greatly reduced the tPA activity at both 4 h (0.55-fold of the basal level) and 24 h (0.79-fold of the basal level) of recovery ( p Ͻ 0.005 compared with saline-injected animals; n ϭ 6 for each time point) ( Fig. 1 A-C). The HI-induced uPA activity was also diminished to 57-78% of those in saline-injected animals. Furthermore, the PAI-1 treatment almost completely blocked the HIinduced MMP-9 and MMP-2 activation at 24 h of recovery ( Fig.  1 B). These results indicated that activation of the PA system precedes and is required for MMP-9 induction in neonatal cerebral HI. zymography showed an increase of both tPA and uPA activities on the HI-challenged side of brain at 4 h (R4) and 24 h (R24) of recovery compared with the unchallenged (UN) brain in saline-injected animals. PAI-1 injection greatly dampened the induction of PAs and MMPs at both 4 and 24 h of recovery. Second panel, Immunoblot showed retention of the ICV-injected human PAI-1 at 4 h of recovery, but this was greatly reduced at 24 h after HI. Third panel, Gelatin zymography showed that the post-HI activation of MMP-9 and MMP-2 mainly occurred at 24 h of recovery. Fourth panel, Immunoblot against ␣-actin verified equal loading of a comparable amount of protein in all examined samples. Shown are representative in four to six sets of samples. C, Quantification of tPA activity as fold change to the baseline level. D, A large area of Ig extravasation was detected on the HI-challenged side of brain in saline-injected animals but not on the contralateral side or in PAI-1-injected animals (n ϭ 4). Ctx, Cerebral cortex; Hip, hippocampus. E, Quantification of Evans blue extravasation at 24 h recovery (n ϭ 3). Shown in E and G are mean and SE. p value is determined by t test. L, Left; R, right.

PAI-1 therapy protects against HI-induced BBB permeability and brain edema
Because tPA and MMP are both potent neurovascular proteases, we next examined the effect of PAI-1 therapy on HI-induced BBB permeability at 24 h of recovery. We found that saline-injected animals consistently displayed Ig extravasation over a larger area on the HI-stressed side of brain than PAI-1-injected animals (n ϭ 4) (Fig. 1 D). In addition, the PAI-1 therapy decreased the extent of HI-induced Evans blue dye extravasation from 19 to 3% over the contralateral side of brain ( p Ͻ 0.05; n ϭ 3) (Fig. 1 E). These results suggest more severe BBB damage in saline-injected animals.
Next, we used a 7 tesla magnetic resonance imaging (MRI) system to examine the effect of PAI-1 therapy in preserving the BBB integrity at 24 h after HI. The MRI study showed a large area of increased T2 signal-an indication of water accumulation-on the ipsilateral hemisphere in saline-injected but not PAI-treated animals (Fig. 2 A, B). Similarly, a greater increase in T2 relaxation time over the contralateral side (68 to 95 ms; 37.5% increase) was found in saline-treated animals than in PAI-1 treated animals (71 to 82 ms; 15.1% increase) ( p Ͻ 0.05 comparing the percentage increase from the contralateral side; n ϭ 3) (Fig. 2C).
The saline-injected animals also exhibited a large area of reduced ADC-a sign of diffusion-restricted intracellular space or tortuous extracellular pathway and generally used as an indicator of cytotoxic edema (Moseley et al., 1990)-on the ipsilateral hemisphere (Fig. 2 D). Quantification showed that ADC dropped from 1.005 Ϯ 0.047 m 2 /ms on the contralateral side to 0.755 Ϯ 0.064 on the HI-challenged side in saline-injected animals ( p Ͻ 0.005; n ϭ 3) (Fig. 2 F). In contrast, the reduction of ADC after HI was insignificant in PAI-1 animals (1.032 Ϯ 0.047 on the contralateral side and 1.026 Ϯ 0.051 on the ipsilateral side) (n ϭ 3) (Fig. 2 E, F ). Together, these results indicate that inhibition of plasminogen activators reduces HIinduced BBB damage and brain edema in newborns.

PAI-1 therapy protects against HIinduced white-matter damage and cortical degeneration
Next, we used diffusion tensor imaging (DTI) to test whether ICV injection of PAI-1 lessens the HI-induced whitematter injury in this model, which is commonly used as an experimental paradigm of periventricular leukomalacia in infants (Volpe, 2008). We found that salineinjected rats showed ϳ10% reduction of FA in the corpus callosum (CC) at 24 h of recovery ( p ϭ 0.005; n ϭ 5) (Fig. 2 I). In contrast, PAI-1-treated animals showed little reduction of FA in the CC (Fig. 2 I). Furthermore, DEC map of DTI-a method to highlight the orientation of anisotropic tissues (Chahboune et al., 2007)-frequently showed partial absence of the corpus callosum tract in salineinjected animals (Fig. 2G, arrows) but not in PAI-1-treated animals (Fig. 2 H). These results suggest that PAI-1 therapy ameliorates the white-matter injury in neonatal cerebral HI.
To confirm the efficacy of PAI-1 therapy against HI-induced axonal damage, we examined the histology of saline-or PAI-1treated animals at 48 h recovery (n ϭ 9 for each). This analysis showed that almost all saline-injected animals had brain damage, ranging from massive cystic degeneration to multiple columnar lesions in the cerebral cortex (Fig. 3B, arrows), associated with positive terminal deoxynucleotidyl transferase-mediated biotinylated UTP nick end labeling (TUNEL) stain (Fig. 3E) and a swollen defasciculated CC (outlined in Fig. 3B). In contrast, the PAI-1-treated animals exhibited milder CC swelling and only isolated spots of TUNEL stain (Fig. 3C,F ). No apparent lesion was found on the contralateral side of brain in either saline-or PAI-1-injected animals (Fig. 3 A, D).
In addition, the myelin basic protein (MBP)-positive oligodendrocyte processes in the CC were more fragmented and the soma less obvious in saline-injected animals (Fig. 3 H, K ) when compared with those on the contralateral side (Fig. 3G,J ) or the PAI-1-injected animals (Fig. 3 I, L). Furthermore, the salineinjected animals tend to have more OX42-positive macrophages in the CC on the HI-challenged side of the brain than the contralateral side or PAI-1-treated animals (Fig. 3M-O). These results suggest that PAI-1 therapy decreases HI-induced whitematter injury partially through blocking the brain infiltration of inflammatory cells.  A, B, T2-weighted image of HI-challenged and saline-treated (A) or PAI-1treated (B) brain at 24 h. Saline-treated animals show a large area of increased T2 signal intensity on the lesion (R) side. C, Quantification showed T2 signal increased from 68 ms on the contralateral side to 95 ms on the ipsilateral side in saline-injected animals. By comparison, T2 signal increased from 72 ms on contralateral side to 83 ms on ipsilateral side in PAI-1-injected animals. The difference of increase from the contralateral side was significant ( p ϭ 0.009 by t test; n ϭ 3). D, E, Diffusion-weighted image of saline-treated (D) or PAI-1-treated (E) brain at 24 h. Note the large area of decreased ADC on the lesion side of saline-injected rat brains. F, Quantification showed a significant reduction of ADC in saline-injected animals ( p Ͻ 0.005 by t test; n ϭ 3). G, H, DTI and quantitative DEC map at 24 h. f, Fibrim. Note the absence of fibrim and posterior CC (arrows) in saline-injected (G) but not in PAI-1-injected (H ) animal brain. I, Quantification showed FA dropped from 0.347 on the contralateral side to 0.314 on the ipsilateral side in saline-injected rats ( p ϭ 0.005 by t test; n ϭ 5). There was no obvious difference of FA on contralateral side (0.383 Ϯ 0.018, mean Ϯ SEM) and ipsilateral side (0.378 Ϯ 0.020) in PAI-1-treated animals (n ϭ 5). L, Left; R, right.

PAI-1 therapy protects against HIinduced brain damage
Finally, we examined the therapeutic effect of PAI-1 injection at 7 d of recovery. The brains of saline-or PAI-1-treated animals in each experiment were photographed for records (examples in Fig. 4 A, B), serial sectioned, and Nissl stained (Fig. 4C,D). The extent of brain injury was quantified as the percentage of tissue loss, compared with their counterparts on the contralateral hemisphere, in the cerebral cortex, hippocampus, and striatum. For comparison, we have shown previously that ICV injection of ␣2-antiplasmin, an inhibitor of one of the many downstream effectors of PAs, produced a maximal 55% reduction of tissue loss if it was administered within 2 h after HI (Adhami et al., 2008).
For deriving the dose-response curve ( Fig. 4 E), saline or a varying dose of PAI-1 (0.95-3.8 g) was injected within 10 min after the HI insult. The extent of tissue loss in saline-injected animals was 67 Ϯ 3% (mean Ϯ SE) in the cerebral cortex, 64 Ϯ 2% in the hippocampus, and 53 Ϯ 3% in the striatum (n ϭ 29). ICV injection of PAI-1 in all doses provided a significant reduction of tissue loss in all three regions (n ϭ 11-29 for each dose; p Ͻ 0.001). PAI-1 at 1.9 g appeared to have the best protection, which remarkably decreased tissue loss to 4 Ϯ 1% in the cerebral cortex and 5 Ϯ 1% in the hippocampus or striatum when compared with contralateral counterparts (n ϭ 29).
To determine the therapeutic window ( Fig. 4 F), 1.9 g of PAI-1 was injected at 1, 2, or 4 h recovery (n ϭ 10 for each), and the effects were compared with those after immediate post-HI injection. We found that delayed injection of PAI-1 at 1 or 2 h still provided a significant protection in all three regions when compared with saline injection ( p Ͻ 0.001). Even 4 h delayed injection of PAI-1 decreased tissue loss significantly: 45 Ϯ 4% in the cerebral cortex ( p ϭ 0.01), 42 Ϯ 3% in the hippocampus ( p ϭ 0.02), and 34 Ϯ 3% in the striatum ( p ϭ 0.09). Together, these results suggest that inhibition of PAs is a powerful brain-protection strategy in neonatal HI.

PA is upstream of MMP in neonatal HI brain injury
Uninhibited extracellular protease activity in the "neurovascular unit"-a conceptual entity that comprises neurons, microvessels, and the supporting glial cells-has an important pathogenic role in cerebrovascular disorders (Mun-Bryce and Rosenberg, 1998;Iadecola, 2004;Lo et al., 2004). These "neurovascular proteases" include MMPs, plasmin, plasminogen activators, and thrombin. Among them, MMP-9 and tPA have been discussed as potential therapeutic targets in adult ischemic stroke, because they both show early induction after in-jury, and either genetic or pharmacological inhibition of their activities offers protection in animal models (Wang et al., 1998;Nagai et al., 1999;Yepes et al., 2000;Asahi et al., 2001;Cinelli et al., 2001;Gu et al., 2005). However, little is known about the relative roles of tPA and MMPs in neonatal cerebral HI.
In this context, the present study revealed a surprising, causal relationship between PA and MMP in neonatal cerebral HI. We have shown previously that HI induces rapid tPA and uPA activity within 4 h in the newborn brain (Adhami et al., 2008), whereas the induction of MMP-9 occurs at 24 h after HI (Svedin et al., 2007). Here, we further demonstrate that early inhibition of PAs after the HI insult is sufficient to prevent the subsequent MMP activation in newborn brains. These results suggest that PA is upstream of MMP in this experimental paradigm. In conclusion, we suggest that future studies are warranted to examine whether infants diagnosed with HIE or at a high risk of cerebral palsy have greater levels of tPA and plasmin in the brain or the CSF. If so, inhibition of the parenchymal PA system may be a promising new therapy to overcome this devastating perinatal disorder.