Memantine Improves Safety of Thrombolysis for Stroke

Background and Purpose— Despite side effects including N-methyl-D-aspartate-mediated neurotoxicity, recombinant tissue-type plasminogen activator (rtPA) remains the only approved acute treatment for ischemic stroke. Memantine, used for treatment of Alzheimer disease, is an antagonist for N-methyl-D-aspartate receptors. We investigated whether memantine could be used as a neuroprotective adjunct therapy for rtPA-induced thrombolysis after stroke. Methods— In vitro N-methyl-D-aspartate exposure, oxygen and glucose deprivation, and N-methyl-D-aspartate-mediated calcium videomicroscopy experiments were performed on murine cortical neurons in the presence of rtPA and memantine. The therapeutic safety of rtPA and memantine coadministration was evaluated in mouse models of thrombotic stroke and intracerebral hemorrhage. Ischemic and hemorrhagic volumes were assessed by MRI and neurological evaluation was performed by the string test and automated gait analysis. Results— Our in vitro observations showed that memantine was able to prevent the proneurotoxic effects of rtPA in cultured cortical neurons. Although memantine did not alter the fibrinolytic activity of rtPA, our in vivo observations revealed that it blunted the noxious effects of delayed thrombolysis on lesion volumes and neurological deficits after ischemic stroke. In addition, memantine rescued rtPA-induced decrease in survival rate after intracerebral hemorrhage. Conclusions— Memantine could be used as an adjunct therapy to improve the safety of thrombolysis.

T he thrombolytic agent recombinant tissue-type plasminogen activator (rtPA) remains the only approved acute treatment for ischemic stroke. 1 Nevertheless, its use is limited to a short therapeutic window 2 due to a lack of efficiency and potential deleterious effects 3 when thrombolysis is delayed or performed in susceptible individuals. 4 tPA exerts several key functions at the levels of both the blood-brain barrier and brain parenchyma. 5 Vascular tPA promotes fibrinolysis and can cross the blood-brain barrier. 6 In the brain parenchyma, it can act in concert with tPA released by neuronal and endothelial cells. 5 Due to its interaction with several proteins/receptors, 7 tPA does not act solely as a plasminogen activator. For instance, tPA also binds to low-density lipoprotein receptor-related protein and thus activates metalloproteinase-dependent blood-brain barrier leakage. 8 tPA also promotes the conversion of platelet-derived growth factor-C to its active form, a mechanism suggested to impair blood-brain barrier integrity. 9 Interestingly, tPA is also a positive neuromodulator of N-methyl-D-aspartate receptors (NMDAR) leading to an increased sensitivity of neurons to excitotox-icity. 10,11 Thus, besides its proven beneficial fibrinolytic activity in patients with stroke, rtPA may promote hemorrhagic transformations 12 and neuronal death. 4 NMDAR-mediated excitotoxicity is thought to be a major cause of neuronal death after stroke. 13 Unfortunately, all strategies targeting NMDAR have failed in the clinical setting. 14 Previously, in a model of thrombotic stroke in mice, we evidenced that an experimental strategy designed to block the interaction of tPA with NMDAR persistently reduced brain lesions, neurological deficits, and extended the therapeutic window of rtPA. 15 The uncompetitive NMDAR antagonist memantine (1amino-3,5-dimethyladamantane) was approved for the symptomatic treatment of the moderate to severe forms of Alzheimer disease. Some preclinical studies in rodents have suggested the potential use of memantine as a neuroprotective agent for ischemic or hemorrhagic stroke. 16 -18 Remarkably, memantine acts preferentially on NMDAR containing the GluN-2D subunit, 19 which is actually the preferential target of the proneurotoxic action of tPA. 20,21 Thus, we postulated that memantine may be a good candidate in combination with rtPA treatment to improve stroke outcome.

Neuronal Cultures
Cultured cortical neurons were prepared from Swiss mouse embryos (E14 -15) provided by CURB (Caen, France), as previously described. 21 To inhibit glial proliferation, ␤-D-arabinoside (10 mol/L) was added after either 3 days in vitro or 7 days in vitro for neuronal death experiments and calcium videomicroscopy, respectively.

Excitotoxic Neuronal Death
As previously described, 21 excitotoxicity was induced at 12 to 13 days in vitro by exposure to NMDA (10 mol/L) in serum-free Dulbecco modified Eagle medium supplemented with 10 mol/L of glycine for 24 hours. NMDA was applied alone or together with rtPA (0.3 mol/L) and/or memantine (1, 5, or 10 mol/L). After 24 hours, neuronal death was quantified by measurement of lactate dehydrogenase released from damaged cells (Roche Diagnostics, Mannheim, Germany).

Oxygen and Glucose Deprivation on Neuronal Cultures
Oxygen and glucose deprivation was performed in a hypoxic chamber (IN VIVO 2 500; Ruskinn) programmed at 1% O 2 , 5% CO 2 , and 37°C. At 11 to 12 days in vitro, cortical neurons were submitted to an oxygen and glucose deprivation or not for 30 minutes. In the chamber, neurons were switched to glucose/serum free deoxygenated Dulbecco modified Eagle medium and then were treated with rtPA (0.3 mol/L) and/or memantine (1 or 10 mol/L). After 30 minutes, cells were removed from the hypoxic chamber and the medium was replaced with oxygenated Dulbecco modified Eagle medium (4.5 g/L glucose). rtPA and/or memantine was again added during reoxygenation. Neuronal death was assessed 24 hours later by lactate dehydrogenase measurement.

Calcium Videomicroscopy
As previously described, 3 neurons (14 days in vitro) were loaded with Fura2/AM (10 mol/L) for 45 minutes at 37°C. NMDA stimulations (50 mol/L) were performed twice to show the reproducibility of responses. Then, neurons were treated for 45 minutes with memantine (1 and 10 mol/L) and/or rtPA (0.3 mol/L), and another NMDA exposure was then applied. Fura-2 (excitation: 340/380 nm, emission: 510 nm) ratio images were acquired with a CCD camera (Princeton Instrument, Trenton, NJ) and digitized (256ϫ512 pixels) using Metafluor 6.2r6 software (Universal Imaging Corporation, Chester, PA). The area under the curve is represented as a percentage of the first response to NMDA exposure.

Animals
Male Swiss mice (39Ϯ3 g) provided by CURB (Caen, France) were housed with a 12-hour light/12-hour dark cycle in standard polypropylene cages (37ϫ23.5ϫ18 cm; Charles River, France) with free access to water and food (SDS Dietex, France). Experiments were performed in accordance with the French (Decree 87/848) and the European Communities Council (Directive 86/609) guidelines. During surgery, mice were deeply anesthetized and maintained with 2% isoflurane in a 70%/30% gas mixture (N 2 O/O 2 ) and the rectal temperature was maintained at 37Ϯ0.5°C. For in vivo experiments, a cluster randomized design was established thanks to the Excel randomization function with 6 animals per day including a minimum of one animal per group each day.

Thrombotic Stroke Model and Thrombolysis
Anesthetized mice were placed in a stereotaxic frame. As previously described, 22 thrombotic stroke was induced by injecting thrombin (1.0 IU in 1 L) directly in the middle cerebral artery. To induce thrombolysis, rtPA was intravenously injected (tail vein, 200 L, 10 mg/kg, 10% bolus and 90% infusion over 40 minutes) either at 4 hours or 20 minutes postclot onset 15 with or without an intravenous bolus of memantine (200 L, 20 mg/kg) just before rtPA infusion. Control groups received the same volume of saline with or without memantine (nϭ7-8 per group). Cerebral blood velocity was continuously monitored by laser Doppler (Oxford Optronix) up to 1 hour after clot formation and also during drug injection.

Magnetic Resonance Imaging
Experiments were carried out on a Pharmascan 7T (Bruker, Germany). T2-weighted images were acquired using a multislice multiecho sequence: TE/TR 51.3 ms/2500 ms. Two-dimensional timeof-flight angiographies (TE/TR 12 ms/7 ms) were acquired at 6, 24, and 48 hours after stroke. Angiographic score was assessed with blinding of the evaluator and graded as follows: 0ϭcomplete disappearance of the occluded artery on angiography images, 1ϭpartial disappearance, and 2ϭnormal appearance. Gradient echo sequences with flow compensation T2*-weighted images (TE/TR 7.7 ms/500 ms) were acquired after stroke. In the ICH model, T2*weighted images were also acquired to evaluate hemorrhagic volumes. Ischemic and hemorrhagic volumes were quantified on MRI using ImageJ software. Volumetric measurements were performed by an experimenter "blind" to treatments.

String Test
Mice were suspended by their fore paws on a horizontally stretched wire (diameter 1.5 mm), and their ability to remain suspended on the wire was evaluated. After training (3 consecutive trials before surgery), a single 30-second trial was performed the day of test. A 0 to 4 rating system was used to evaluate the performance of each animal: 0ϭunable to remain on the string; 1ϭhangs by both fore paws; 2ϭattempts to climb onto string; 3ϭboth fore paws and one or both hind paws around the string; and 4ϭ4 paws and tail around the string with lateral movement.

CatWalk XT System
The apparatus (Noldus Information Technology, Wageningen, The Netherlands) consists of an enclosed walkway (130ϫ68ϫ152 cm) with a glass plate, a light source, and a high-speed video camera that accurately records footprints. The video capture is then processed by the CatWalk XT software (Version 9). After training (2 explorations of the walkway before surgery), mice were subjected to 3 consecutive runs of gait assessment at 24 hours poststroke.
All behavioral tests were conducted by 2 experimenters "blind" to treatments.

Statistical Analyses
Power analysis for thrombotic stroke and ICH models was determined a priori from data of the literature and preliminary results, respectively, using G*Power software.
Results obtained in vitro and in vivo are expressed as meanϮSD. Statistical analyses were performed using JMP software. All probability values are 2-tailed. Statistical significance was concluded for PϽ0.05.
Parametric tests were performed to analyze calcium videomicroscopy. Thus, data were analyzed by 2-way analyses of variance with repeated measurements followed when appropriate by supplemen-tary Tukey honestly significant difference (post hoc tests). The log-rank test was used to assess the statistical significance of the Kaplan-Meier survival curves.
For in vitro neuronal death measurements, ischemic and hemorrhagic volumes analyses, behavioral experiments, clot lysis assay, and angiographies analyses, data were analyzed with nonparametric tests. In these cases, Kruskal-Wallis tests were used followed, when appropriate, by Mann-Whitney U tests as post hoc tests. In addition, Pearson correlation test was used to examine the potential relationships between functional data (size of the right hind limb print) and the weight of mice.

Memantine Prevents the Proneurotoxic Effects of rtPA In Vitro
Memantine dose-dependently prevented NMDA-mediated neuronal death with near complete protection achieved at 10 mol/L ( Figure 1A; Ϫ91% compared with NMDA; PϽ0.05). Interestingly, at 1 mol/L, a dose that by itself does not protect neurons against basal NMDA-mediated neurotoxicity ( Figure 1A), memantine prevented the 42% rtPAdependent potentiation of NMDA-induced neuronal death ( Figure 1B; PϽ0.05). Memantine was also tested on cortical neurons subjected to oxygen and glucose deprivation. A dose of 1 mol/L memantine (a dose with no effect on oxygen and glucose deprivation-induced neurotoxicity by itself) totally prevented the rtPA-induced potentiation after a 30-minute oxygen and glucose deprivation ( Figure 1C; Ϫ64% compared with rtPA; PϽ0.05). Accordingly, calcium videoimaging performed as an index of NMDAR activity (Figure 2A

Memantine Reduces the Noxious Actions of Delayed Thrombolysis by rtPA
Memantine was then tested in a thrombotic stroke model in mice with rtPA-induced reperfusion. 22 In this model, early thrombolysis by rtPA (20 minutes postonset) is beneficial, whereas late thrombolysis (4 hours postonset) is deleterious. 15 The effect of intravenous memantine (20 mg/kg, a dose chosen based on the literature 24 ) as an adjunct treatment to late thrombolysis by rtPA was thus tested. T2-weighted MR images at 6, 24, and 48 hours after stroke ( Figure 3A) revealed that although memantine did not alter the extent of the ischemic damage in the absence of rtPA ( Figure 3B Although the deleterious effects of delayed thrombolysis appeared at 6 hours postclot onset and continued to progress between 24 and 48 hours, the lesion volumes of control, memantine, and memantine/rtPA-treated groups reached their maximum at 24 hours ( Figure 3A-B). T2-weighted MR images 15 days postischemia showed that animals with late rtPA-induced thrombolysis have a smaller cortical volume than those treated with memantine ( Figure 3A-C). Regardless of the treatment group, no bleeding complication was detected on high sensitive T2*-weighted images ( Figure 3D). Similar experiments were performed with both memantine and rtPA, either alone or in combination, injected 20 minutes after stroke onset at a time for which rtPA alone was reported to have benefit. 15 Memantine alone was not protective (online-only Data Supplement Figure I). Similarly, the combination of rtPA with memantine at an early time point (20 minutes postclot onset) did not show additive neuroprotective effect when compared with rtPA-induced thrombolysis alone (online-only Data Supplement Figure I).

Memantine Precludes the Deleterious Effects of Delayed Thrombolysis by rtPA on Sensorimotor Functions
Sensorimotor functions were evaluated by the string test 24 hours after stroke onset (online-only Data Supplement Figure Gait analyses were also performed 24 hours poststroke. The maximum contact area spatial parameter (ie, the maximum area of a paw in contact with the glass plate during the stance phase) was used as a measure of individual paw prints, as previously described. 25 The size of the right hind limb print was reduced (Ϫ27%) in delayed reperfused animals ( Figure  4A; saline versus late rtPA; PϽ0.05), whereas memantine alone did not improve the deficit measured in saline animals ( Figure 4A; saline versus late memantine; PϾ0.05). The coinjection of memantine with rtPA prevented the deleterious effects of late thrombolysis ( Figure 4A; late rtPA versus late memantine/rtPA; Pϭ0.05 and saline versus late memantine/ rtPA; PϾ0.05). No correlation was found between the size of the right hind limb print and the weight of mice ( Figure 4B; PϾ0.05).

Memantine Does Not Alter the Fibrinolytic Action of rtPA
This beneficial effect of memantine, when combined with rtPA, was not associated with an impaired fibrinolytic activity of rtPA, as evidenced by in vitro clot lysis assays ( Figure 5), similar rates of reperfusion (online-only Data Supplement

Memantine Rescues rtPA-Induced Hematoma Expansion and Survival Rate After ICH
We have investigated the effect of memantine in a model of brain hemorrhage induced by an intrastriatal injection of collagenase in mice. As shown on T2*-weighted images, although intravenous rtPA increased the hemorrhagic volumes when injected 30 minutes after ICH, this prohemorrhagic action of rtPA was prevented by the coinjection of memantine ( Figure 6A). Remarkably, the survival of rtPAtreated animals after ICH was increased by a coinjection of memantine ( Figure 6B; log-rank PϽ0.05).

Discussion
Neuroprotective agents with high affinity for NMDAR such as MK-801 led to clinical noxious effects, including drowsiness and even coma. 26 Memantine is an uncompetitive, open-channel NMDAR antagonist with low to moderate affinity, which can reduce NMDAR activity. 17,26 Accordingly, preclinical data indicate that memantine may exert neuroprotective effects on dementia and global and focal cerebral ischemia. 24,27,28 Memantine is approved in Europe and in the United States for the symptomatic treatment of the moderate to severe forms of Alzheimer disease. Despite its beneficial fibrinolytic activity, rtPA may have noxious effects after stroke by promoting hemorrhagic transformations 12,29 and NMDAR-mediated neurotoxicity. 3,4,15 Our data reveal that memantine counteracts the deleterious effects of rtPA in cultured neurons subjected to a set of excitotoxic paradigms as well as in vivo under ischemic or hemorrhagic conditions in mice. Functional assessments also demonstrate the beneficial action of memantine. In a model of embolic stroke in rats, Back and coworkers 30 failed to reveal a benefit of memantine when injected intraperitonally. Here, we show that combination of memantine with rtPA is associated with better sensorimotor status (string test) and use of the plantar surface of the ipsilateral hind limb (CatWalk), a parameter also impaired in other cortical infarct models. 25 Memantine, as an individual treatment, is not effective in our thrombotic stroke model. These data are in agreement with the literature, because any beneficial effect of memantine alone was reported on a stroke model performed in healthy adult mice or rats when administrated after stroke onset. Memantine, injected poststroke onset, was only reported to display beneficial effects on a model of hypertensive rats 31 and on a model of embolic stroke in the rabbit. 17 This lack of efficacy of memantine alone could be related either to the time of rtPA treatment or the severity of the ischemic lesion. In our hands, memantine alone is not beneficial but rescues the benefit of rtPA treatment when injected late after stroke onset at a time when rtPA alone is deleterious. Thus, memantine increases the therapeutic window of rtPA-induced thrombolysis.
At the mechanistic level, some reports suggest that the neuroprotective properties of memantine could be mediated by an increased production of brain-derived neurotrophic factor in the brain. 32 Alternatively, it was also reported that the memantine is a 6-to 8-fold more selective antagonist for GluN-2C-or GluN-2D-containing NMDAR than for GluN-2A or GluN-2B subunits. 19 These observations and our present results are in agreement with previous demonstrations that the proneurotoxic effect of rtPA occurs preferentially through GluN-2D-containing extrasynaptic NMDAR. 20,21 Unfortunately, we have reported that the new selective  GluN-2D antagonist UBP145 ([2R*,3S*]-1-[9-bromophenanthrene-3-carbonyl]piperazine-2,3-dicarboxylic acid) was not effective when injected intravenously in a model of cortical excitotoxicity. It is why we decide to focus our interest on memantine, a previously reported NMDAR antagonist with a selectivity for GluN-2D-containing NMDAR, also approved for human use with no major side effect reported so far. 33 Our present data suggest that memantine specifically prevents the deleterious effects of rtPA on NMDAR signaling without affecting its beneficial vascular effects and the basal NMDAR functions. Our results on an ICH model are consistent with previous studies showing a beneficial effect of memantine in a similar ICH model 34 associated with reduced levels of active endogenous tPA and matrix metalloproteinase-9. 35 Similarly, MK-801, a NMDAR antagonist, was shown to counteract the neurotoxicity of rtPA in a model of ICH in pigs. 36

Summary/Conclusions
Although its safety in the acute stroke settings remains to be determined, both the proven safety of memantine in chronic poststroke aphasia 37 and the present study provide in vitro and in vivo evidence supporting the use of memantine as an adjunct therapy to improve safety of rtPA-induced thrombolysis.

Sources of Funding
This work was supported by grants from the "Institut National de la Santé Et de la Recherche Médicale" (INSERM). Axel Montagne is recipient of a PhD fellowship from the "Conseil Régional de Basse-Normandie" (CRBN) and Guerbet.