Selective inhibition of GluN2D-containing N-methyl-D-aspartate receptors prevents tissue plasminogen activator-promoted neurotoxicity both in vitro and in vivo

Background Tissue plasminogen activator (tPA) exerts multiple functions in the central nervous system, depending on the partner with which it interacts. In particular, tPA acts as a positive neuromodulator of N-methyl-D-aspartate glutamatergic receptors (NMDAR). At the molecular level, it has been proposed that the pro-neurotoxicity mediated by tPA might occur through extrasynaptic NMDAR containing the GluN2D subunit. Thus, selective antagonists targeting tPA/GluN2D-containing NMDAR signaling would be of interest to prevent noxious effects of tPA. Results Here, we compared three putative antagonists of GluN2D-containing NMDAR and we showed that the new compound UBP145 ((2R*,3S*)-1-(9-bromophenan-threne-3-carbonyl)piperazine-2,3-dicarboxylic acid) is far more selective for GluN2D subunits than memantine and PPDA (phenanthrene derivative (2S*, 3R*)-1-(phenanthrene-2-carbonyl)piperazine-2,3-dicarboxylic acid). Indeed, in vitro, in contrast to the two other compounds, UBP145 prevented NMDA toxicity only in neurons expressing GluN2D (ie, in cortical but not hippocampal neurons). Furthermore, in cultured cortical neurons, UBP145 fully prevented the pro-excitotoxic effect of tPA. In vivo, we showed that UBP145 potently prevented the noxious action of exogenous tPA on excitotoxic damages. Moreover, in a thrombotic stroke model in mice, administration of UBP145 prevented the deleterious effect of late thrombolysis by tPA. Conclusions In conclusion, tPA exerts noxious effects on neurons by acting on GluN2D-containing NMDAR and pharmacological antagonists of GluN2D-containing NMDAR could be used to prevent the ability of tPA to promote neurotoxicity.


Background
Tissue plasminogen activator (tPA) is a serine protease initially described to promote vascular fibrinolysis by converting plasminogen into active plasmin [1]. But tPA is not just a vascular protease, since it is also expressed by nerve cells and exerts important functions in the brain parenchyma [2,3]. For instance, during perinatal development, tPA influences neuronal migration and synaptic outgrowth [4,5], while in the adult brain, tPA controls synaptic plasticity [6], long-term potentiation [7] and long-term depression [8]. In preclinical models of brain disorders, it has been shown that tPA can differentially influence the fate of nerve cells, due to its pleiotropic actions and targets that include, excitotoxicity, apoptosis, neuroinflammation or myelination [9][10][11][12][13].
The pleiotropic functions of tPA in the central nervous system can be explained by several mechanisms. Among them, there is strong evidence that the direct or indirect interaction of tPA with the N-methyl-D-aspartate receptor (NMDAR) exerts a positive neuromodulatory action [10,14,15]. tPA can thus increase NMDAevoked calcium entry, which can be trophic under physiological conditions, or alternatively noxious under excitotoxic conditions. NMDARs are ion-gated glutamate receptors, mainly described as heterotetrameric channels composed of two GluN1 subunits and two GluN2 subunits (in some cases, NMDARs may also contain GluN3 subunits [16,17]). GluN1 and GluN2 subunits exist as isoforms, explaining the varied electrophysiological properties of NMDARs [18]. NMDARs differentially control neuronal outcome (trophism versus neurotoxicity) depending on their neuronal localization (synaptic and extrasynaptic) [19,20], their subunit composition [21] and/or their anchorage with cytoplasmic or membrane-associated proteins [22].
In the present study, we investigated in vitro and in vivo, whether putative antagonists of GluN2D-containing NMDAR (memantine, PPDA and UBP145) could prevent the aggravating effect of tPA on NMDARmediated neurotoxicity. In vitro, we show that UBP145 is more selective than memantine or PPDA, as during excitotoxic challenges, it acted only on neurons expressing GluN2D. Accordingly, in vivo, UBP145 prevents the deleterious effects of exogenous tPA against excitotoxicity and ischemic stroke.

Results
UBP145 selectively prevents tPA-promoted neurotoxicity mediated by GluN2D-containing NMDA receptor In a paradigm of rapidly triggered excitotoxicity (exposure to a high dose of NMDA for 1 hour and measurement of neuronal death 24 hours later), we have previously shown that tPA potentiates neuronal death of cortical neurons but not of hippocampal neurons. This differential sensitivity of cortical versus hippocampal neurons was respectively attributed to the presence and absence of GluN2D subunits in NMDAR at both the mRNA and protein levels [23]. Here, we first extended our previous observations by investigating the sensitivity to tPA of cortical and hippocampal neurons in vitro, in a paradigm of slowly triggered excitotoxicity (exposure to a low dose of NMDA for 24 h). As reported for rapidly triggered excitotoxicity [23], slowly triggered excitotoxic cell death was equivalent in both types of neurons, but tPA (20 μg/ml) increased excitotoxic neuronal death in cortical cultures but not in hippocampal cultures (see MAP2 immunostaining in Figure 1A and the corresponding quantification of neuronal death in Figure 1B; +30%, n = 12, p < 0.05). Our next step was to compare three pharmacological agents reported to be more or less selective antagonists of GluN2D, for their ability to prevent tPA-mediated pro-neurotoxicity: memantine [27,28], PPDA (phenanthrene derivative (2S*, 3R*)-1-(phenanthrene-2-carbonyl)piperazine-2,3-dicarboxylic acid) [24]and UBP145 ((2R*,3S*)-1-(9-bromophenan-threne-3-carbonyl)piperazine-2,3-dicarboxylic acid) [26]. We took advantage of the lack of expression of the GluN2D in cultured hippocampal neurons when compared to cortical neurons and we postulated that a molecule preferentially targeting the pro-neurotoxicity effects of tPA: i/should display a dose-dependent inhibitory effect on NMDA-mediated neurotoxicity in cultured cortical neurons, ii/should not influence NMDAmediated neurotoxicity in cultured hippocampal neurons; iii/should prevent the aggravating effect of tPA on NMDA-mediated neurotoxicity in cultured cortical neurons, at a concentration not influencing NMDA alonemediated neurotoxicity. The doses of memantine, PPDA and UBP145 were chosen according to their published affinity values [24,26,27].
Memantine was tested for its ability to influence NMDA-mediated neurotoxicity in both cultured cortical ( Figure 2A) and hippocampal neurons ( Figure 2B) at 1 and 10 μM. The lowest dose of memantine (1 μM) had no effect on the extent of NMDA-induced neuronal loss in cortical (n = 12, p < 0.05, Figure 2A) and hippocampal (n = 12, p < 0.05, Figure 2B) neurons. By contrast, at the highest dose (10 μM), memantine nearly fully prevented NMDA-mediated neuronal death in both cortical (n = 12, p < 0.05, Figure 2A) and hippocampal (n = 12, p < 0.05, Figure 2B) neurons. Since memantine displayed the same profile of inhibition of NMDAmediated neuronal death in cortical and hippocampal neurons, memantine did not meet our second selection criterion.
PPDA (from 0.02 to 2 μM) was tested in the same paradigm and like memantine, PPDA showed a comparable dose-dependent profile of inhibition of NMDAmediated neurotoxicity in both cultured cortical ( Figure  2C) and hippocampal neurons ( Figure 2D). Thus, the pharmacological profile of PPDA (similar potency to prevent NMDA-mediated excitotoxicity in cortical and hippocampal cultures) did not fit with our criteria.
In an in vivo excitotoxic challenge, the inhibition of GluN2D-containing NMDAR can fully prevent the proexcitotoxic effect of intravenously administered tPA Based on our in vitro data, the ability of UBP145 to prevent the pro-neurotoxicity of tPA was tested in mice, in a model of NMDA-induced excitotoxicity in the cortex. The dose of tPA was chosen according to our previous works [10]. For its intravenous administration, the dose of UBP145 was chosen to respect the molar ratio UBP145/ tPA shown to be efficient in our in vitro studies (Figure 4), while for its intra-structure administration, the dose of UBP145 was similar to that of PAI-1 (an inhibitor of tPA) known to be effective in preventing the noxious effect of intravenous tPA in the in vivo excitotoxic model [29]. NMDA alone induced a lesion restricted to the cortical area (n = 6). When injected intravenously, tPA (10 mg/kg) doubled the NMDA-induced cortical lesion (n = 6, p < 0.01) ( Figure 5A,B). In agreement with our in vitro observations, the cortical co-injection of UBP145 completely prevented the noxious effect of tPA administration on NMDA-mediated excitotoxicity (n = 6, p < 0.01) ( Figure  5A). In contrast to the neuroprotective effect of UBP145 when administered inside the lesion site, the intravenous administration of UBP145 (6.67 mg/kg) failed to prevent the pro-excitotoxic effect of exogenous tPA in our paradigm of cortical excitotoxicity ( Figure 5C,D). Our observations suggest that despite being a potent and selective blocker of the deleterious effect of tPA on GluN2D containing NMDAR, UBP145 does not cross the intact blood-brain barrier (BBB).
In an in vivo model of thrombotic stroke, the deleterious effect of late thrombolysis by tPA is reversed by UBP145, an antagonist of the GluN2D-containing NMDAR We have developed and fully characterized a model of thrombotic stroke in mice by injection of thrombin into the middle cerebral artery, leading to the formation of a local thrombus [30][31][32]. In this model, as in the clinical setting, early thrombolysis by intravenous injection of tPA is beneficial, while late thrombolysis by tPA is deleterious and increases BBB leakage [32]. We thus tested the effect of intravenous UBP145 (6.67 mg/kg) in this model of thrombotic stroke coupled or not to a late thrombolysis by tPA (10 mg/kg). In these experimental conditions, we have previously demonstrated an opening of the BBB [32]. Accordingly, the lesion volumes in mice receiving late thrombolysis by tPA were higher than the ones measured in mice with no tPA treatment ( Figure 6, +70%, n = 9, p < 0.05). When injected intravenously 4 hours after clot formation, although UBP145 did not influence the ischemic lesion in the absence of tPA, the deleterious effect of late thrombolysis (tPA iv, 4 hours after clot formation) was completely prevented by the co-injection of UBP145 (n = 7, p < 0.05).

Discussion
We have previously demonstrated that tPA is pro-excitotoxic in cortical neurons but not in hippocampal Figure 3 UBP145 has no effect on NMDA-mediated neurotoxicity in hippocampal neurons but prevents tPA-induced potentiation of NMDA-mediated neurotoxicity in cortical neurons. Cortical (A, C) and hippocampal (B) neurons were exposed to NMDA (10 μM) with or without UBP145 and tPA (20 μg/mL) for 24 hours. Quantification of neuronal death by LDH release in the medium: (A, B) UBP145 2 μM protected cortical but not hippocampal neurons from neurotoxicity, (C) although UBP145 0.2 μM did not protect cortical neurons from excitotoxicity, it was able to prevent tPA-induced potentiation of excitotoxicity (n = 12; *p < 0.05; NS: not significant).
neurons. We also evidenced by using both pharmacological and molecular approaches that this discrepancy could be due to a selective action of tPA on extrasynaptic NMDAR containing the GluN2D subunit [23]. These observations support the ideas that there are several subpopulations of NMDAR and that these subpopulations exert differential and even sometimes opposite actions (such as trophic versus neurotoxic effects. These differential actions of NMDAR can be at least in part attributed to variations in their subunit composition and/or recruitment of different intracellular signalling pathways. To date, this idea had been mainly investigated for GluN2A and GluN2B [21,33,34], but little is known regarding other GluN2 variants. Thus, further investigating the relative contribution of different NMDAR populations is relevant for many physiological/pathological situations, including the ones involving tPA and/or GluN2D, such as learning/memory processes [35][36][37], food intake [38], stroke [3,23], Alzheimer's disease [39,40] or nociception [41]. We thus sought here to characterize among the currently available drugs suggested to preferentially act on GluN2D-containing NMDAR (ie, memantine, PPDA and UBP145), which one is the more efficient in targeting the potentiating effects of tPA on NMDARmediated neuronal death.
In the United States and Europe, memantine is indicated to treat patients with moderate to severe Alzheimer's disease. Recently, it has been reported that in the presence of 1 mM magnesium, memantine becomes 6-8 fold more selective for GluN2C-or GluN2D-containing receptors over those with either GluN2A or GluN2B subunits [28]. However, in our hands, memantine was as efficient in preventing NMDA neurotoxicity in We then used another weakly-selective GluN2D competitive antagonist, PPDA [24,25] and the more selective agent, UBP145 [26]. Much lower doses of PPDA than those of memantine were used, but like memantine, PPDA displayed a similar potency to prevent NMDAmediated excitotoxicity in cortical and hippocampal cultures. By contrast, we found that low doses of UBP145 could differentially affect NMDA-induced death of cortical and hippocampal neurons that respectively express and do not express GluN2D. UBP145 thus appears as a sufficiently-selective pharmacological antagonist of GluN2D as to selectively protect GluN2Dexpressing cortical neurons. At a very low dose -not influencing the extent of NMDA-induced excitotoxicity- , UBP145 prevented tPA from exacerbating neuronal death. Consistent with this, tPA was unable to increase NMDA-evoked calcium influx when co-administered with UBP145. These in vitro data argue again that GluN2D-containing NMDAR are the critical effectors of the pro-excitotoxic effects of tPA. From a molecular point of view, this is in agreement with the demonstrations that neurotoxicity is mediated by extra-synaptic NMDAR [19] and that GluN2D is exclusively distributed in extra-synaptic NMDAR [37].
We then used UBP145 in vivo in an attempt to prevent the deleterious action of tPA during excitotoxic challenges. First, we observed that UBP145 can block the pro-excitotoxic effect of exogenous tPA (administered intravenously) in a model of cortical excitotoxicity. However, this prevention requires that UBP145 is administered in the brain parenchyma, as no effect was observed for an intravenous injection. This suggests that in this model that does not induce blood-brain barrier leakage [29], UBP145 cannot extravasate into the brain parenchyma. Yet, in a model of thrombotic stroke highly relevant to the clinical setting [30][31][32], we found that UBP145 co-injected intravenously with tPA fully prevented the noxious action of late tPA-mediated thrombolysis on the extent of ischemic brain damage. This strongly suggests that, in conditions of ischemic brain injury, in which the integrity of the BBB is altered [30][31][32], the GluN2D-containing NMDAR antagonist UBP145, could be used as an adjunct therapy to increase the therapeutic window of tPA-mediated thrombolysis that today, is restricted to the first 4.5h after stroke onset [42].

Conclusion
In conclusion, we provide here strong evidence that tPA influences neurotoxicity selectively through GluN2Dcontaining NMDAR and we provide support for the use of antagonists like UBP145 in order to selectively prevent the deleterious effects of tPA on NMDAR-mediated pathological functions.

Excitotoxic neuronal death
Excitotoxicity was induced at 12-13 DIV by exposure to NMDA (10 μM) in serum-free DMEM supplemented with 10 μM of glycine for 24 hours. NMDA was applied alone or together with rtPA (20 μg/ml) and/or antagonists. After 24 hours, neuronal death was estimated by Microtubule-associated protein 2 (MAP2) immunostaining and quantified by measurement of the activity of lactate dehydrogenase (LDH) released from damaged cells into the bathing medium with a cytotoxicity detection kit (Roche Diagnostics, Mannheim, Germany). The LDH level corresponding to the maximal neuronal death was determined in sister cultures exposed to 200 μM NMDA. Background LDH levels were determined in sister cultures subjected to control washes. Experimental values were measured after subtracting LDHmin and then normalized to LDHmax-LDHmin to express the results in percentage of neuronal death.

Calcium videomicroscopy
Experiments were performed at room temperature on the stage of a Nikon Eclipse inverted microscope equipped with a 75 W Xenon lamp and a Nikon 40x, 1.3 numerical aperture epifluorescence oil immersion objective. Cell cultures were transferred into a serum free medium (HBBSS) and loaded with fura-2/AM 10 μM (Invitrogen) for 45 minutes at 37°C. Neurons were washed and NMDA treatment (1 mL at 50 μM) was applied using a peristaltic pump (2 mL/minute). UBP145 (0.2 μM) and/or tPA (20 μg/mL) were directly applied in the medium and incubated for 45 minutes. Fura-2 (excitation: 340, 380 nm, emission: 510 nm) ratio images were acquired with a CCD camera (Princeton Instrument, Trenton, New Jersey), and digitized (256 × 512 pixels) using Metafluor 4.11 software (Universal Imaging Corporation, Chester, Pennsylvania).

Animals
Experiments were performed on male Swiss mice (32 ± 3 g, produced and provided by our local animal facilities, France) in accordance with European communities Council (Directives of November 24, 1986 (86/609/ EEC)) and French Legislation (act no. 87-848, Ministère de l'Agriculture et de la Forêt) on Animal Experimentation and were approved by the local ethical committee. Mice were deeply anesthetized with isoflurane 5% and maintained under anesthesia with 2% isoflurane in a 70%/30% gas mixture (N 2 O/O 2 ) during surgery. The rectal temperature was maintained at 37 ± 0.5°C throughout the surgical procedure using a feedbackregulated heating system.

In vivo excitotoxic lesions
A cortical unilateral injection (coordinates: 0.5 mm posterior, 3.0 mm lateral, -0.8 mm ventral to the bregma; stereotaxic atlas G. Paxinos & K.B.J. Franklin) of NMDA (2.5 nmoles; in 0.33 μl), co-injected or not with UBP145 (0.05 nmoles) was performed after placing the animals in a stereotaxic frame. Solutions were injected by the use of a micropipette made with hematologic micropipettes (calibrated at 15 mm/μl; assistant ref 555/5; Hecht, Sondheim-Rhoen, Germany). The micropipette was removed 3 minutes later. After the excitotoxic lesion, a 200 μl bolus intravenous injection of rtPA (10 mg/kg) or saline was performed through a catheter previously inserted into the tail vein.
In a second set of experiments, UBP145 was tested for its ability to cross the blood-brain barrier. In this case, cortical excitotoxic lesions were performed with NMDA, as above, followed by tail vein injections of rtPA alone (10 mg/kg) or in association with UBP145 (pH 7.4; 6.67 mg/kg).
After 24 hours, mice were deeply anesthetized and then transcardially perfused with cold heparinized saline (15 ml) followed by 150 ml of fixative (phosphate buffer 0.1 M. pH 7.4 containing 2% paraformaldehyde and 0.2% picric acid). Brains were post-fixed (18 hours; 4°C) and cryoprotected (sucrose 20% in phosphate buffer; 24 hours; 4°C) before freezing in Tissue-Tek (Miles Scientific, Naperville, IL, USA). Cryomicrotome-cut transversal sections (20 μm) were collected on poly-lysine slides and stained with Fluoro-Jade C [43]. Subsequently, images were captured with a Leica DM6000 microscope-coupled to a coolsnap camera. Cortical lesion surfaces were determined using sections covering 500 μm anterior and posterior to the injection site by using an image analysis system (MetaVue software, Molecular Devices, Downingtown, PA, USA).

Thrombotic model of stroke and thrombolysis
The same male Swiss mice were placed in a stereotaxic frame and maintained under anesthesia. After a skin incision between the right eye and ear, the master muscle was excised and a small craniotomy (Ø 1 mm) was performed on the parietal bone in order to expose the right middle cerebral artery (MCA). Then, 1 μl of murine α-thrombine (1UI. Sigma Aldrich) was injected in the MCA using hematological micropipettes to activate the coagulation cascade as previously described [30]. To induce thrombolysis, rtPA (10 mg/kg) was injected (200 μl, i.v., 10% bolus and 90% infusion over 40 minutes), 4 hours after MCA occlusion, with or without UBP145 (6.67 mg/kg). Finally, control groups received the same volume of saline with or without UBP145. Cerebral blood velocity was measured by laser-Doppler flowmetry using an optic fiber probe (Oxford Optronix, UK) placed on the skull in the MCA territory. After 24 hours, brains were collected and frozen in isopentane. Then, coronal brain sections (20 μm) were stained with thionine and analyzed with an image analysis software (Image J).

Statistical analyses
The results are expressed as the mean ± S.E.M. Statistical analyses were performed by the Kruskall-Wallis' test, followed by post hoc comparisons, with the Mann-Whitney's test. For calcium videomicroscopy, statistical analyses were performed using Student test. Statistical significances were concluded for p < 0.05.