A Glutamate N-Methyl-d-Aspartate (NMDA) Receptor Subunit 2B–Selective Inhibitor of NMDA Receptor Function with Enhanced Potency at Acidic pH and Oral Bioavailability for Clinical Use

Abstract

We describe a clinical candidate molecule from a new series of glutamate N-methyl-d-aspartate receptor subunit 2B–selective inhibitors that shows enhanced inhibition at extracellular acidic pH values relative to physiologic pH. This property should render these compounds more effective inhibitors of N-methyl-d-aspartate receptors at synapses responding to a high frequency of action potentials, since glutamate-containing vesicles are acidic within their lumen. In addition, acidification of penumbral regions around ischemic tissue should also enhance selective drug action for improved neuroprotection. The aryl piperazine we describe here shows strong neuroprotective actions with minimal side effects in preclinical studies. The clinical candidate molecule NP10679 has high oral bioavailability with good brain penetration and is suitable for both intravenous and oral dosing for therapeutic use in humans.

SIGNIFICANCE STATEMENT

This study identifies a new series of glutamate N-methyl-d-aspartate (NMDA) receptor subunit 2B–selective negative allosteric modulators with properties appropriate for clinical advancement. The compounds are more potent at acidic pH, associated with ischemic tissue, and this property should increase the therapeutic safety of this class by improving efficacy in affected tissue while sparing NMDA receptor block in healthy brain.

Introduction

Cerebral ischemia, stroke, subarachnoid hemorrhage (SAH), and traumatic brain injury (TBI) all produce substantial neuronal death that, if not fatal, can create lasting disabilities with significant societal impact. Few therapeutic options are currently available for stroke apart from dissolution of the vessel clot in a subset of patients (Turner et al., 2020) or clot retrieval when blockages occur in large arteries (Awad et al., 2020). SAH can be treated with Ca²⁺ channel blockers (Carlson et al., 2020); however, there remains considerable opportunity for improved therapies, as a significant fraction of patients progress to subsequent ischemic episodes and death (Longstreth et al., 1993). No pharmacological strategy for neuroprotection in TBI has been approved yet (Crupi et al., 2020).

Extracellular glutamate concentrations increase in injured central nervous system tissue in both animal models and human patients with acute injury (Supplemental Table 1 in Yuan et al., 2015). One consequence of increasing extracellular glutamate is overactivation of NMDA receptors (NMDARs) which can be neurotoxic (Olney, 1969; Choi et al., 1988). It logically follows that inhibition of NMDARs during insults that raise glutamate should be neuroprotective, and the efficacy of several NMDAR antagonists has been confirmed in animal models of injury (Supplemental Table 2 in Yuan et al., 2015). However, promising preclinical results have not yet translated to clinical success, as multiple clinical trials in stroke or TBI using NMDAR antagonists either failed to improve patient outcome or were associated with unacceptable side effects (Supplemental Table 3 in Yuan et al., 2015). Evaluation of these clinical data and trial designs identified multiple potential reasons for these failures (Morris et al., 1999; Albers et al., 2001; Sacco et al., 2001; Gladstone et al., 2002; Farin and Marshall, 2004). Moreover, the heterogeneous nature of acute brain injury requires careful management of the inclusion criteria, and improved image-based analysis of collateral blood flow and hypoperfusion with comprehensive postinjury analysis could diminish variability (Kidwell et al., 2001; Narayan et al., 2002; Guenego et al., 2020). Treatment of patients within hours by first responders is now possible and may prove essential (Saver, 2013; Shkirkova et al., 2018). Thrombectomy and clot retrieval have also created new opportunities to administer neuroprotectants directly to the site of the injury despite arterial blockage (Patel et al., 2020).

Since the discovery of GluN2B antagonists, numerous scaffolds of highly selective GluN2B NMDA receptor antagonists have been reported (reviewed by Layton et al., 2006; Mony et al., 2009; Koller and Urwyler, 2010; Ruppa et al., 2012). These include phenethanolamines such as ifenprodil (Williams, 1993), eliprodil (Carter et al., 1988), radiprodil (Michel et al., 2014; Auvin et al., 2020), traxoprodil (Chenard et al., 1995), Ro 25-6981 (Fischer et al., 1997), MK-0657 (Addy et al., 2009), plus additional scaffolds including propanolamines (Yuan et al., 2015), benzimidazoles (McCauley et al., 2004; Davies et al., 2012), cyclic benzamidines (Nguyen et al., 2007), amino cyclopentanes (Layton et al., 2011), piperidinyl pyrrolidinones (Marcin et al., 2018), and other compounds (McIntyre et al., 2009; Mosley et al., 2009; Brown et al., 2011; Buemi et al., 2014; Dey et al., 2018; Thum et al., 2018). Whereas crystallographic evaluation of ifenprodil at the GluN1/GluN2B heterodimer amino terminal domain interface (Karakas and Furukawa; 2014) is the likely pose for many GluN2B-selective inhibitors, recent structural data revealed a unique region within the binding cavity that can be occupied by compounds with a more compact scaffold (Kemp and Tasker, 2009; Stroebel et al., 2016). The propanolamine EU93-31 (Tahirovic et al., 2008), which exhibits pH dependence for inhibition of NMDA receptors (Yuan et al., 2015), occupies both portions of the binding pocket (Regan et al., 2019).

GluN2B-selective compounds have been tested in preclinical and clinical studies for use in cerebral stroke (Supplemental Table 3 in Yuan et al., 2015), TBI (Bullock et al., 1999; Merchant et al., 1999; Yurkewicz et al., 2005), Parkinson’s disease (Steece-Collier et al., 2000; Liverton et al., 2007; Nutt et al., 2008; Addy et al., 2009; Michel et al., 2014, 2015), depression (Preskorn et al., 2008; Li et al., 2010; Bristow et al., 2017), and pain (Taniguchi et al., 1997; Sang et al., 2003; Abe et al., 2005; Liverton et al., 2007; Mony et al., 2009; Labas et al., 2011; Swartjes et al., 2011). However, despite clear achievement of preclinical efficacy, no GluN2B-selective inhibitor has been approved for clinical use.

Here, we introduce a new scaffold for GluN2B-selective inhibitors and identify NP10679 with high pH sensitivity, efficacy in a murine model of transient ischemia, high oral bioavailability, brain penetration, and a preclinical profile that supports NP10679 as a clinical candidate.

Materials and Methods

Two-Electrode Voltage-Clamp Recordings from Xenopus laevis Oocytes

Stage V to VI Xenopus laevis unfertilized oocytes were purchased from Ecocyte (Austin, Texas) and injected with 5 ng of GluN1 and 10 ng of GluN2B ribonucleic acid made from cDNA. The cDNAs for human GluN1 and GluN2B, encoding National Center for Biotechnology Information reference sequences NM_007327.3 and NM_000834.3, respectively, were linearized, and ribonucleic acid made from cDNA was made as previously described (Traynelis et al., 1998). For activity measurements at GluN2A (NM_000833), GluN2C (NM_000835), and GluN2D (NM_000836) NMDA receptors cRNAs were prepared similarly. After injection, oocytes were incubated in Barth’s culture solution (88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO₃, 10 mM HEPES, 0.82 mM MgSO₄, 0.33 mM Ca(NO₃)₂, 0.41 mM CaCl₂, 10 U/ml Penicillin-Streptomycin, and 0.1 mg/ml gentamycin, at pH 7.4) at 18°C. Two-electrode voltage-clamp recordings were made at 22 to 23°C 2–7 days after injection using Warner OC725C amplifiers (V_HOLD –40 mV). Briefly, oocytes were perfused in recording solution (90 mM NaCl, 1 mM KCl, 10 mM HEPES, 0.01 mM EDTA, and 0.5 mM BaCl₂) adjusted to either pH 7.6 or 6.9 by addition of NaOH or HCl, respectively (pH 6.9 solutions were prepared by addition of HCl to pH 7.6 solutions to maintain an equal concentration of Na⁺ ions in both solutions). Drug concentration-response curves were obtained by application of increasing concentrations of the GluN2B-selective inhibitors in the presence of 100 μM glutamate and 30 μM glycine until steady-state conditions were obtained. Oocyte recordings were made from 4–10 oocytes per experiment (i.e., oocyte injection cycle) from ≥2 experiments. The concentration-response relationship for each oocyte was fit by eq. 1:

where minimum is the residual response in saturating concentration of the experimental compounds (constrained to be ≥0), IC₅₀ is the concentration of inhibitor that causes half-maximal inhibition, and n_H is the Hill slope.

In Vivo Model of Transient Focal Ischemia

All protocols involving animals were approved by the Georgia State University Institutional Animal Care and Use Committee, an Association for Assessment and Accreditation of Laboratory Animal Care–accredited program, and was under the supervision of a licensed veterinarian. Mice were group-housed and provided nestlets and shelters with access to food pellets and water ad libitum under a 12-hour light/dark cycle. Mice were brought to a separate room and housed for at least 30 minutes prior to initiation of the surgery.

Mice (C57Bl6, >90 days old; Jackson Laboratories) were subjected to transient (60-minute) middle cerebral artery (MCA) occlusion (MCAo), and the infarct volume was measured 24 hours postreperfusion, similar as previously described (Yuan et al., 2015). Male mice were used for this experiment to reduce potential confound by progesterone variation through estrous cycles, which can have neuroprotective actions. Briefly, transient ischemia was induced in anesthetized (2% isoflurane/98% O₂) mice by insertion of an intraluminal suture into the MCA for 60 minutes (Junge et al., 2003). The body temperature of each mouse was monitored with a rectal thermometer and maintained at 37°C through use of a homeothermic blanket. Changes in local cerebral blood flow were monitored with a laser Doppler flowmeter probe (Perimed) secured via glue to the skull 4–6 mm lateral and 2 mm posterior of bregma. An 11-mm 5-0 Dermalon or Look (SP185) black nylon nonabsorbable suture with the tip flame-rounded was introduced into the left internal carotid artery through the external carotid artery stump up to 10.5–11 mm of suture insertion. Only mice with a reduction in blood flow to <20% for 60 minutes and with recovery of blood flow to >90% after removal of the suture were included in the study. After the occlusion period, mice were placed back in their cages on a warming blanket (37°C) for several hours and monitored for righting reflex and ability to ambulate upon a gentle touch. At 24 hours postocclusion, mice were euthanized by isoflurane overdose, and the brain was quickly removed and cut into 2-mm sections and incubated in 2% 2,3,5-triphenyltetrazolium chloride (TTC) in phosphate-buffered saline (pH 7.4) at 37°C for 20 minutes and then placed at 4°C for imaging. The infarct area was then measured using National Institutes of Health IMAGE software (Scion Corporation, Beta 4.0.2 release). The lesioned area of each section was determined by digital threshold reductions in TTC staining to ≤20% lower intensity than that observed in the contralateral cortex. The infarct region was then manually outlined with a cursor, and the cubic volume of the infarct was determined for each slice and then summed across all four slices from each animal to obtain total infarct volume. A ratio of the contralateral to ipsilateral hemisphere volume was multiplied by the corresponding infarct section volume to correct for edema. Drug was administered by intraperitoneal injection 5 minutes prior to initiation of surgery (approximately 15 minutes prior to vessel occlusion). All drug doses were randomized, and investigator(s) were blinded throughout the study from the surgical procedure through analysis of stained sections to measure infarct volume. Based on historical variability and an anticipated effect size of 45%–50%, we estimated that n = 12 per group (four groups per study) was adequate to detect significant effects (α = 0.05) with sufficient power, (β = 0.90) (G*Power 3.1). The infarct volume after administration of a drug dose was compared with vehicle by one-way ANOVA and Dunnett’s tests (P < 0.05). A priori, mice were only removed from the study if the MCA was ruptured, resulting in a subarachnoid hemorrhage, or mice did not survive 24 hours postsurgery.

Measurement of Locomotor Activity and Rotarod Performance

Mice (C57Bl6, >90 days old; Jackson Laboratories) were placed in a closed (with light on) activity monitoring box for 1 hour to habituate prior to drug testing. After 1 hour, animals were removed and injected intraperitoneally with drug and then returned to the activity monitoring box, where total locomotor activity was monitored for 2 hours. The total number of light beam breaks in the cage (horizontal) was determined by computer, and results averaged for each drug. Results were analyzed by ANOVA and Dunnett’s post hoc test to compare horizontal activity of drug-treated groups to vehicle-treated controls. Male animals were used for these behavioral tests because only male mice were used in MCAo transient ischemia studies.

For Rotarod experiments, male C57BL/6 mice (>90 days old) were tested using a Rotamax 4/8 Rotarod (Columbus Instruments, Columbus, Ohio). Prior to training and testing, mice were brought to the testing room and allowed to acclimate for 2 hours prior to any further handling. Mice were placed on a rotating rod (5 rpm) that was 3.8 cm in diameter and 8 cm wide and elevated 30 cm from the floor of a chamber. After 10 seconds of rotation at a fixed velocity, the rotation was slowly accelerated from 5 to 35 rpm over a 5-minute period. The duration of time that the mouse could stay on the Rotarod, without hanging on for a full rotation or without falling, was recorded. Mice were trained four times at 25-minute intertrial intervals on each day for 2 days. On day 3, mice were randomly assigned to treatment groups and administered test drug or vehicle 25 minutes prior to testing (four trials with intertrial interval of 25 minutes) by intraperitoneal administration. Individuals performing the experiment were blinded to the identity of each treatment group. Results were analyzed by ANOVA and Dunnett’s post hoc test to compare duration of time on the rotating rod of drug-treated groups to vehicle-treated controls.

The locomotor and Rotarod studies were approved by the Georgia State University Institutional Animal Care and Use Committee, an Association for Assessment and Accreditation of Laboratory Animal Care–accredited institution, under the supervision of licensed veterinarians. Mice were group-housed and provided nestlets and shelters with access to food pellets and water ad libitum under a 12-hour light/dark cycle.

Pharmacokinetic Studies

Pharmacokinetic studies on NP10679 were outsourced to Anthem Biosciences (Bangalore, India) and were performed after obtaining the Institutional Animal Ethics Committee permission in accordance with the Committee for the Purpose of Control and Supervision of Experiments on Animals guidelines. Evaluation of NP10679 properties was performed in male BALB/c mice (8–10 weeks old, 20–30 g). Briefly, mice were administered a 2 mg/kg or a 5 mg/kg dose (n = 3 each) by intraperitoneal injection (10 ml/kg dose volume), and blood samples were collected at 0.08, 0.25, 0.5, 1, 2, 4, 8, and 24 hours postdose in tubes containing sodium heparin on ice (study 4). In total, 100 μl of plasma was combined with 50 μl of internal standard (haloperidol, 10 μg/ml), and tubes were then spun at 4000g for 10 minutes (4°C), and plasma was transferred to clean tubes and stored at −80°C until analysis. The analyte NP10679 was quantified with an API 3200 Q-trap LC-MS/MS and compared with standards, and data were analyzed by WinNonlin 6.3 (Pharsight). In a separate study (study 3), BALB/c mice were administered either oral (10 mg/kg) or an intravenous (3 mg/kg) dose of NP10679 (injection volume of 10 ml/kg). Blood samples were collected on ice in sodium heparin tubes at 0.08, 0.25, 0.5, 1, 2, 4, 8, and 24 hours postdose. Samples were prepared and analyzed as described above, except here the internal standard was fluconazole (10 μg/ml). We also measured NP10679 in the brain compartment compared with plasma at 0.25 and 1 hour postdose with 3 mg/kg i.v. dosing in two separate studies with blood samples collected and prepared as described above. Here, brain samples were first washed in deionized water to remove blood, the weight was recorded, and then they were transferred into fresh water (1 ml), homogenized, and stored at −80°C until analysis. The ratio of compound in brain (grams) compared with plasma (milliliters) was then calculated. Study 1 used fluconazole as internal standard, whereas study 2A and 2B used haloperidol as the internal standard.

Formulation and Drug Dosing

For MCAo, locomotor, and Rotarod studies, NP10679, MK-801, and ifenprodil were formulated in 2% or 10% N′,N′-dimethylacetamide (DMA), 10% propylene glycol (PG), and 30% 2-hydroxypropyl-β-cyclodextrin in water (HPBCD), with a dose volume of 10 ml/kg, and administered intraperitoneally. Formulation for pharmacokinetic studies used either 2% or 10% N′,N′- dimethylacetamide/10% propylene glycol/30% 2-hydroxypropyl-β-cyclodextrin in water and a dose volume of 10 ml/kg (all routes of administration).

Liver Microsome Stability, Cytochrome P450 Inhibition, and Plasma Protein Binding

Metabolic stability was assessed using human and mouse liver microsomes (Xenotech). The final composition of the assay included 1 µM of test compounds or reference standards (imipramine and diclofenac sodium) prepared from DMSO stock, so that the final concentration of DMSO and acetonitrile was 0.2% and 0.8%, respectively. Compound was incubated with 0.5 mg/ml microsomal protein without (100 mM potassium phosphate buffer alone, pH 7.4) or with cofactors (5.0 mM glucose-6-phosphate, 0.06 U glucose-6-phosphate dehydrogenase, 2.0 mM MgCl₂, 1.0 mM NADP+/NADPH). Test compounds and standards were incubated at 37°C with human and mouse liver microsomes; aliquots of the reaction mixture (100 µl) were removed at 0, 5, 15, 30, 60, and 120 minutes, and the reaction was stopped by the addition of 2.5 ml tertiary butyl methyl ether and shaken for 15 minutes. The samples were then spun at 4000 rpm for 15 minutes at 10°C, the organic phase was evaporated to dryness, and then samples were reconstituted with solvent for LC-MS/MS analysis. The percentage of the compound remaining after specified incubation period was calculated with respect to the peak areas of compound at time 0 minutes.

Inhibition of CYP2D6 and CYP3A4 was accomplished using recombinant human isoforms and a Vivid cytochrome P450 blue screening kit (Invitrogen) by incubating 2-fold serial dilutions (nine dilutions) of test compounds (highest concentrations were 1, 5, or 10 μM) with kit reagents and reaction buffer according to the manufacturer’s methods in a 96-well plate. Plates were then incubated at room temperature for 30 minutes before fluorescence was measured with a plate reader. For these studies, reference standards ketoconazole (CYP3A4) and quinidine (CYP2D6) were used as controls.

Plasma protein binding was performed with a rapid equilibrium dialysis device containing dialysis membrane with a molecular weight cutoff of 8000 Da (ThermoFisher). The plasma samples (pH 7.4) and test article (1 or 5 µM) or reference standards (Warfarin and Propranolol, 10 µM) were combined (DMSO final conc 0.1%) with 300 µl of plasma sample and added to the sample chamber, and 500 µl of buffer was added into the buffer chamber. The rapid equilibrium dialysis device was sealed with adhesive film and then incubated at 37°C with shaking at 300 rpm for 4 hours. After incubation, an aliquot of 50 μl was removed from each well (plasma and buffer side) and diluted with equal volume of opposite matrix (plasma or buffer alone) to nullify the matrix effect and then extracted for analysis by LC-MS/MS. The amount of free material was determined by:

Lead Profiling Off-Target Screening

The in vitro effects of NP10679 on the human ether-à-go-go-related gene (hERG) potassium channel current (a surrogate for IKr , the rapidly activating, delayed rectifier cardiac potassium current) stably expressed in HEK mammalian cells were evaluated at room temperature using the QPatch HT (Sophion Bioscience A/S, Denmark), an automatic parallel patch-clamp system (ChanTest, Cleveland, OH). NP10679 was evaluated at 0.1, 0.3, 1, and 3 μM diluted in solution composed of (in mM) NaCl, 137; KCl, 4.0; CaCl₂, 1.8; MgCl₂, 1; HEPES, 10; glucose, 10; pH adjusted to 7.4 with NaCl. Each test concentration was tested in two or more cells (n ≥ 2). Duration of exposure to each test article concentration was 3 minutes. A positive control (0.5 μM E-4301) was used to confirm the sensitivity of the cells to an hERG inhibitor.

Off-target radioligand binding displacement studies of NP10679 were conducted at the National Institutes of Mental Health Psychoactive Drug Screening Program at the University of North Carolina at Chapel Hill (https://pdsp.unc.edu/). The NIMH PDSP is directed by Bryan L. Roth, MD, PhD, at the University of North Carolina at Chapel Hill and Project Officer Jamie Driscoll at NIMH, Bethesda, MD. (contract number HHSN-271-2008-00025-C; NIMH PDSP). Briefly, compounds were submitted to the NIMH PDSP program and screened at a single concentration (10 μM) of test article under equilibrium conditions for ability to displace specific radioligands from binding to their targets expressed in mammalian cell membranes in vitro. Each receptor target was assayed in quadruplicate, and the percent inhibition of radioligand binding at each target was determined at pH 7.4. If the percent inhibition was >50%, a full competition displacement binding study was conducted to determine an IC₅₀, and from this, a binding K_i using the Cheng-Prusoff equation (K_i = IC₅₀/[1 + (L/K_d)]), in which L is the radioligand concentration used in the competition binding assay and K_d is the radioligand equilibrium binding affinity determined in the above saturation binding assays.

The following targets (with radioligand in parentheses) were tested: serotonin receptor-1A ([³H]8-hydroxy-2-dipropylaminotetrain), serotonin receptor-1B ([³H]5-carboxamidotryptamine), serotonin receptor-1D ([³H]5-carboxamidotryptamine), serotonin receptor-1E ([³H]5HT), 5-Hydroxytryptamine_2A ([³H]ketanserin), serotonin receptor-2B ([³H]lysergic acid diethylamide ), serotonin receptor-2C ([³H]mesulergine), serotonin receptor-3 ([³H]LY278584), serotonin receptor-5A ([³H]lysergic acid diethylamide), serotonin receptor-6 ([³H]lysergic acid diethylamide), serotonin receptor-7 ([³H]lysergic acid diethylamide), α adrenergic receptor-1A ([³H]prazosin), α adrenergic receptor-1B ([³H]Prazosin), α adrenergic receptor-1D ([³H]Prazosin), α adrenergic receptor-2A ([³H]rauwolscine), α adrenergic receptor-2B ([³H]rauwolscine), α adrenergic receptor-2C ([³H]rauwolscine), β adrenergic receptor-1 ([¹²⁵I]pindolol), β adrenergic receptor-2 ([³H]CGP12177), β adrenergic receptor-3 ([³H]CGP12177), benzodiazepine rat brain receptor rat brain site ([³H]flunitrazepam), dopamine receptor-1 ([³H]SCH23390), dopamine receptor-2 ([³H]N-methylspiperone), dopamine receptor-3 ([³H]N-methylspiperone), dopamine receptor-4 ([³H]N-methylspiperone), dopamine receptor-5 ([³H]SCH23390), dopamine transporter ([³H]WIN35428), δ-opioid receptor ([³H]D-Ala(2),D-Leu(5) enkephalin), GABAA ([³H]muscimol), H₁ ([³H]pyrilamine), histamine receptor-2 ([³H]tiotidine), histamine receptor-3 ([³H]α-methylhistamine), histamine receptor-4 ([³H]histamine), κ-opioid receptor ([³H]U69593), muscarinic receptor-1 ([³H]quinuclidinyl benzilate ), muscarinic receptor-2 ([³H]quinuclidinyl benzilate), muscarinic receptor-3 ([³H]quinuclidinyl benzilate), muscarinic receptor-4 ([³H]quinuclidinyl benzilate), muscarinic receptor-5 ([³H]quinuclidinyl benzilate), μ-opioid receptor ([³H]D-Ala(2),NMe-Phe(4), Gly-ol(5)enkephalin), norepinephrine transporter ([³H]nisoxetine), peripheral-type benzodiazepine receptor ([³H]PK11195), SERT ([³H]citalopram), σ1 ([³H]pentazocine(+)), and σ2 ([³H]1,3-di-o-tolylguanidine).

Some receptor targets were also tested in functional studies to establish whether NP10679 acted as an agonist or an antagonist of 5-HT_2A receptor function at pH 7.4 (Porter et al., 1999; CEREP laboratories, France). To evaluate agonism, HEK293 cells transfected with human 5-HT_2A receptors were incubated with increasing concentrations of NP10679 (duplicate wells/concentration) at 37°C for 30 minutes. Activation of the receptor was determined by changes in inositol phosphate-1 levels detected by an homogenous time resolved fluorescence method. Separate wells stimulated with 10 μM serotonin served as a positive control. To determine antagonism by NP10679, cells were incubated with increasing concentrations of the compound (duplicate wells/concentration) at 37°C for 30 minutes. Cells were stimulated with 100 nM serotonin. Activation of the receptor was determined by changes in inositol phosphate-1 levels detected by an homogenous time resolved fluorescence method. A control inhibitor, ketanserine, was run separately to confirm the accuracy and reliability of the assay data.

Similar studies were performed to evaluate agonism and antagonism in CHO cells transfected with human α_1A-adrenergic receptors incubated with increasing concentrations of NP10679 (duplicate wells/concentration) at room temperature at pH 7.4 (Vicentic et al., 2002). Activation of the receptor was determined by changes in intracellular [Ca²⁺] by a fura-2 fluorimetry detection method. Separate wells were stimulated with 30 nM epinephrine as a positive control. To evaluate antagonism in CHO cells expressing human α_1A-adrenergic receptors, wells were incubated with increasing concentrations of NP10679 (duplicate wells/concentration) at room temperature and then the cells were stimulated with 3 nM epinephrine. Activation of the receptor was determined by changes in intracellular [Ca²⁺] by a fura-2 fluorimetry detection method (CEREP, France).

To evaluate agonism and antagonism at human H₁-histamine receptors, HEK293 cells transfected with H₁ receptors were incubated with increasing concentrations of NP10679 (duplicate wells/concentration) at pH 7.4 at room temperature (Miller et al., 1999). Activation of the receptor was determined by changes in intracellular [Ca²⁺] by a fura-2 fluorimetry detection method. Separate wells were stimulated with 10 μM histamine as a positive control. To evaluate antagonism by NP10679, cells were incubated with increasing concentrations of the compound (duplicate wells/concentration) at room temperature, and cells were stimulated with 300 nM histamine. Activation of the receptor was determined by changes in intracellular [Ca²⁺] by a fura-2 fluorimetry detection method. A control inhibitor, pyrilamine, was run separately to confirm the accuracy and reliability of the assay data (CEREP, France).

Chemistry

Synthesis of EU-93-94 ((S)-N-(4-(3-(4-(3,4-dichlorophenyl)piperazin-1-yl)-2-hydroxypropoxy)phenyl)methanesulfonamide)

N-[4-[[(2S)-Oxiran-2-yl]methoxy]phenyl]methanesulfonamide (Tahirovic et al., 2008) (66 mg, 0.27 mmol) and 1-(3,4-dichlorophenyl) piperazine (62 mg, 0.27 mmol) were dissolved in ethanol (5 ml) and heated at reflux for 3 to 4 hours. The reaction mixture was then cooled to room temperature, and the solvent was evaporated in vacuo. The remaining residue was then purified via column chromatography on silica gel using 0%–30% 90%:10%:0.5% dichloromethane:methanol:NH₃ in dichloromethane to yield N-[4-[(2S)-3-[4-(3,4-dichlorophenyl)piperazin-1-yl]-2-hydroxy-propoxy]phenyl]methanesulfonamide (110 mg, 0.23 mmol, 85.4% yield).

¹H NMR (500 MHz, CDCl₃): δ 7.23 (d d, J = 8.9, 0.7 Hz, 1H), 7.16 – 7.10 (m, 2H), 6.91 (dd, J = 2.9, 0.7 Hz, 1H), 6.87 – 6.82 (m, 2H), 6.70 (ddd, J = 9.0, 2.9, 0.7 Hz, 1H), 4.09 (ddt, J = 9.3, 5.2, 4.1 Hz, 1H), 3.98 – 3.87 (m, 2H), 3.20 – 3.07 (m, 4H), 2.86 (d, J = 0.7 Hz, 3H), 2.79 – 2.71 (m, 2H), 2.65 – 2.48 (m, 4H).

¹³C NMR (126 MHz, CDCl₃): δ 156.77, 150.44, 132.70, 130.39, 130.03, 124.15, 122.31, 117.27, 115.37, 109.99, 70.51, 65.87, 60.49, 53.05, 48.65, 38.55.

HRMS calculated for C20H26N3O4Cl2F 474.10156; found 474.10258 [M + H].

Synthesis of NP10679 ((R)-6-(2-Hydroxy-3-(4-(4-(trifluoromethyl)phenyl)piperazin-1-yl)propoxy)-3,4-dihydroquinolin-2(1H)-one)

A suspension of (R)-6-(oxiran-2-ylmethoxy)-3,4-dihydroquinolin-2(1H)-one (100 g, 0.456 mol) and 1-(4-(trifluoromethyl)phenyl)piperazine (105 g, 0.456 mol) in ethanol (1 L) was stirred at 75°C for 21 hours with monitoring by HPLC. The reaction became a clear solution within 15 minutes at 75°C. The reaction mixture was cooled to 50°C, and the precipitated solid was filtered and washed with ethanol (200 ml). The collected solid was dried under vacuum to afford 175 g (85.3%) of the title compound. This material was analyzed by high performance liquid chromatography and found to be >99% pure, and with chiral high performance liquid chromotography to have > 98% enantiomeric purity. Recrystallization was performed by placing 260 g of NP10679 obtained from the previous steps in a 5-L round-bottom flask, to which was added a premixed solution of 1:1 methanol:acetone with constant stirring. The suspension was heated to 50°C with stirring until it became clear (approximately 30 minutes) and was then filtered through a 2-μM filter. The clear solution was cooled to 30°C over 15 minutes and added to water (13 L) under vigorous stirring over a 10-minute period. The precipitated solid NP10679 was stirred for 30 minutes at 30°C, filtered, washed with water (7.8 L) and dried in a vacuum tray drier at 70°C for 48 hours. The yield was 255 g of a white solid (98% yield).

¹H NMR (400 MHz, DMSO-d6): δ 9.90 (brs , 1H), δ 7.50 (d, J = 16 Hz, 1H), 7.05 (d, J = 16 Hz, 1H), 6.85 – 6.70 (m, 3H), 4.90 (brd, 1H), 4.00 – 3.80 (m, 3H), 3.30 – 3.20 (m, 4H), 2.90 – 2.75 (m, 2H), 2.70 – 2.30 (m, 8H).

¹³C NMR (75 MHz, CDCl₃): δ 171.82, 154.63, 153.09, 131.21, 130.04, 126.36, 124.94, 122.87, 121.22, 120.79, 120.36, 119.92, 116.30, 114.52, 113.95, 113.24, 70.70, 65.80, 60.48, 53.02, 47.98, 30.51, 25.58.

Mass spectrometry calculated for C23H26F3N3O3 500.47; found 500.30 [M+H].

Unless otherwise specified, all NMR spectra were obtained in deuterated chloroform (CDCl3) or deuterated dimethylsulfoxide (DMSO-d6) and referenced to the residual solvent peak; chemical shifts are reported in parts per million and coupling constants in hertz (Hz), s = singlet, d = doublet, t = triplet, dd = doublet of doublets, dt = doublet of triplets, ddd = doublet of doublets of doublets, etc. br = broad, m = multiplet.

Statistics

Values presented are means ± S.E.M. and are compared using Student’s paired or unpaired t test, or ANOVA followed by Dunnett’s test when appropriate. The number of observations was selected to yield a power of 0.9 for an α of 0.05.

Results

pH-Sensitive GluN2B-Selective Piperazines

We synthesized analogs of the 93-series pH-sensitive inhibitor EU-93-31, which was previously described and supported neuroprotection in the MCAo model of transient ischemia (Yuan et al., 2015). We built upon analogs of EU-93-31 by converting the four-carbon N-coupled aliphatic chain of EU-93-31 to a piperazine linker core (Table 1). Compound EU-93-94 retained good pH sensitivity and was evaluated in vivo and, after a 3 mg/kg intravenous dose in rat, revealed a plasma half-life of 2.6 hours, which was considerably longer than that for EU-93-31 (0.92 hours; Yuan et al., 2015). Given this advantageous property of the piperazine scaffold, we evaluated the pH dependence of numerous analogs on human GluN1-1a/GluN2B receptors (hereafter GluN1/GluN2B) expressed in X. laevis oocytes by measuring the IC₅₀ for each compound at pH 7.6 and pH 6.9, resulting in the selection of NP10679 for further evaluation (Table 1). NP10679 exhibits both a potent IC₅₀ at pH 6.9 of 23 nM with a ratio of IC₅₀ at pH 7.6 to that at pH 6.9 (referred to as pH boost) of 6.2-fold. Furthermore, the selectivity for inhibition of GluN2B receptors over GluN2A, GluN2C, or GluN2D receptors was maintained, as there was no considerable off-target inhibition at 3 μM for NP10679 (Fig. 1A; Table 1). We also determined the pH boost effect of other known GluN2B selective inhibitors—some tested clinically (radiprodil, traxoprodil) or well studied preclinically (ifenprodil, Ro 25-6981)—as reference standards (Fig. 1B). None of these four reference compounds possess high pH sensitivity for GluN2B inhibition, as all show <2-fold potency shifts in their IC₅₀ values at pH 6.9.

TABLE 1.

Fig. 1.

Pharmacology of NP10679 at NMDA receptors in oocytes. (A) NP10679 concentration-response curves for inhibition of GluN2B NMDA receptors expressed in Xenopus oocytes at pH 6.9 (blue symbols) and pH 7.6 (black symbols). Each fitted line and associated symbols represent the result from one experiment (i.e., frog injection, five total). The symbols are means from 12–22 oocytes from each experiment. The NP10679 effect on GluN2A (circles), GluN2C (squares), and GluN2D (triangles) NMDA receptors is shown at the indicated concentrations; mean ± S.E.M., n = 4 for GluN2A, 2C, and 2D results. The inset shows a representative current trace for NP10679 inhibition of GluN2B receptors at 0.1 μM at the two indicated pH values. Max G/G is the receptor current obtained in 100 μM glutamate and 30 μM glycine. (B) The pH potency boost for NP10679 (blue bar, n = 5 experiments) is compared with the pH boost for known preclinical and clinical GluN2B-selective inhibitors (n = 3 experiments each). The pH Boost is the ratio of IC₅₀ at pH 7.6/IC₅₀ at pH 6.9.

In Vitro Drug Profiling of NP10679

Because this piperazine class of selective GluN2B inhibitors delivered relatively potent molecules with high pH dependence (Table 1), we chose to further evaluate compounds in this class for preclinical in vivo efficacy in a rodent model of cerebral ischemia for potential in vivo side effects, in vitro off-target binding profiles, and plasma pharmacokinetics after both intravenous and oral dosing in rodents. We selected NP10679 as an analog with a number of advantages, including good potency at pH 6.9 and a high degree of pH sensitivity. We proceeded to more fully characterize this compound as a prototype clinical candidate.

Metabolic stability was carried out using human and mouse liver microsomes with 1 µM NP10679 prepared from DMSO stock (DMSO 0.2% final). The compound and standards were incubated with human and mouse liver microsomes with or without cofactors, and the samples were extracted and analyzed using LC-MS/MS. NP10679 exhibited moderate to good stability in both human and mouse liver microsomes such that 72% of NP10679 remained in incubations with human microsomes and 54% remained in incubations with mouse liver microsomes in the presence of cofactors after a 1-hour incubation at 37°C. NP10679 at 1 µM also did not inhibit human recombinant cytochrome 450 enzyme isoforms CYP3A4 or CYP2D6. NP10679 bound to human, mouse, and dog plasma proteins at 97.7% (n = 2), 98.2% (n = 2), and 98.2% (n = 1), respectively.

NP10679 was also tested at 10 μM for binding to 41 neurotransmitter receptors, enzymes, and channels via displacement of a radioligand in multiple competitive receptor binding assays (Supplemental Figs. 4 and 5; Supplemental Table 1). Targets for which 10 μM NP10679 displaced >50% of the radioligand were followed up with full dose-effect displacement studies, which identified sub-micromolar K_i values for five of these targets, the 5-HT_2A serotonin receptor (0.638 μM), the α_1A (0.603 μM) and α_1D (0.495 μM) adrenergic receptors, the H₁ histamine receptor (0.040 μM), and the serotonin transporter SERT (0.135 μM) (Table 2). Three receptors (5-HT_2A, α_1A adrenergic, and H₁ histamine) were also tested for functional agonism and antagonism; in all cases, the compound behaved as an antagonist (Supplemental Figs. 1–5; Table 2). Inhibition of the human delayed rectifier cardiac potassium current channel (hERG channel) was measured in mammalian HEK cells transfected with the hERG potassium channel cDNA via patch-clamp electrophysiology across four concentrations of NP10679, which revealed an IC₅₀ for inhibition of 0.617 μM (Supplemental Fig. 6; Table 2).

TABLE 2.

K_i and functional inhibition at 12 off-target sites. The K_i values for 11 off targets that were first identified as hits (>50% radioligand displacement by 10 μM NP10679) in a screen of 41 targets are given. K_i values were determined from full binding displacement curves. The functional IC₅₀ values for several targets were also determined. All assays were conducted at pH 7.4.

Target	K i	Functional IC50
	μM
5-HT1B	>10	nd
5-HT1D	2.29	nd
5-HT2A	0.638	1.71
5-HT2B	1.92	nd
α1A	0.603	0.154
α1B	1.92	nd
α1D	0.495	nd
α2C	3.09	nd
H1	0.040	0.073
SERT	0.135	nd
σ2	1.98	nd
hERG	nd	0.617

α1A, α adrenergic receptor-1A; α1B, α adrenergic receptor-1B; α2C, α adrenergic receptor-2C; α1D, α adrenergic receptor-1D; 5-HT_1B, serotonin receptor-1B; 5-HT2A, serotonin receptor-2A; 5-HT_2B, serotonin receptor-2B; 5-HT_1D, serotonin receptor-1D; nd, not done.

In Vivo Efficacy and Side Effect Studies

MCAo Neuroprotection and Pharmacokinetics

Prior generations of nonselective NMDAR inhibitors that blocked all NMDARs regardless of subunit composition produce both off- and on-target adverse effects, which complicated or aborted clinical development. The most prominent side effects reported included motor dysfunction, cognitive impairment, and psychotomimetic effects such as hallucinations and disorganized thought (Lees et al., 2000; Sacco et al., 2001; Diener et al., 2002; Rowland, 2005; Wood, 2005; Muir, 2006; Blagrove et al., 2009). Although GluN2B-selective NMDAR negative allosteric modulators appear to be tolerated better than competitive antagonists or channel blockers, they still exhibit some side effects (Chaperon et al., 2003; DeVry and Jentzsch 2003; Yurkewicz et al., 2005; Nicholson et al., 2007; Preskorn et al., 2008; Nutt et al., 2008). Our working hypothesis is that the pH sensitivity of the piperazine-containing compounds should increase the therapeutic ratio such that neuroprotection is achieved in ischemic penumbral regions surrounding the infarct (pH 6.9), with reduced inhibition of GluN2B-containing NMDARs in healthy brain tissue at pH 7.4.

In MCAo experiments, all drug doses were randomized, and the investigator(s) were blinded throughout the study from surgery through analysis (see Materials and Methods). NP10679 was administered prior to transient ischemia induced by occlusion of the middle cerebral artery. Vehicle-treated mice exhibited substantial neuronal cell death with a 101 ± 8.7 mm³ infarct volume after 60 minutes of transient ischemia. By comparison, infarct volume was reduced in a dose-dependent manner by NP10679, with an ED₅₀ of 1 mg/kg i.p. dose and a maximum infarct volume reduction of 52% (Fig. 2). Both the 5 mg/kg (56 ± 6.6 mm³) and 10 mg/kg (49 ± 3.0 mm³) NP10679 doses significantly reduced infarct volumes from vehicle control.

Fig. 2.

Neuroprotection by NP10679 reduces infarct volume in the MCAo model of transient ischemia in mice. (A) NP10679 reduces infarct volumes measured 24 hours post-MCAo with an ED₅₀ of 1 mg/kg, i.p. NP10679 was administered 15 minutes prior to initiation of transient MCAo for 60 minutes. In all experiments, the drug doses were randomized and the investigators were blinded from surgical procedure through analysis of stained sections to determine infarct volume. The plot is pooled data across three independent experiments. Data are means ± S.E.M. for n = 9 (0.2 mg/kg), 13 (0.5 mg/kg), 21 (1 mg/kg), 12 (2 mg/kg), 12 (5 mg/kg), 24 (10 mg/kg), and 34 (Veh) mice. **P < 0.01 from vehicle (ANOVA, Dunnett’s). (B and C) Representative images of TTC-stained sections of vehicle (B) and drug-treated brain. (C). A representative image is shown for each dose, with each brain cut into four sections (top to bottom), and both sides of the section are shown (left and right images). The red portion of the image is TTC stain of undamaged tissue. Vehicle (Veh) is the formulation alone as described in Methods.

In pharmacokinetic studies, mice were dosed with a solution orally (10 mg/kg) and i.v. (3 mg/kg) to determine both oral bioavailability and plasma pharmacokinetics for NP10679 (Fig. 3A; Table 3). The plasma terminal half-life for the oral route was 7.06 hours, and for intravenous administration it was 8.56 hours, with a high volume of distribution of 1.59 L/kg and clearance of 2.44 ml/min per kilogram and high oral bioavailability (75.7%) (Table 3). In a separate study, mice were dosed i.p. with 2 and 5 mg/kg of NP10679 to provide drug disposition information in mice after the same dose and route of administration as used in the MCAo neuroprotection studies. Here, NP10679 displayed an expected dose dependence with peak levels of 581 and 1431 ng/ml of NP10679 in plasma 30 minutes postdosing, respectively, and with plasma half-lives of 7.5–9.9 hours (Table 3). Thus, a single intraperitoneal administration of NP10679 provided ample exposures to drive neuroprotection over a large fraction of the 24-hour postischemia period. In Fig. 3B, we present the free plasma levels (unbound drug) calculated at both 2 and 5 mg/kg i.p. doses, showing that free drug levels after the 5 mg/kg dose were above the IC₅₀ for inhibition of GluN2B-containing NMDA receptors at pH 6.9 (Fig. 3B). Further, separate studies show that NP10679 exhibits high brain penetration as measured in two studies reporting a range of 1.3- to 2.6-fold higher levels found in the brain compartment compared with plasma levels in mice 1 hour after intravenous dosing (Table 3). Based on these brain:plasma ratios, we estimate that, after a 5 mg/kg i.p. dose used in MCAo studies, free drug concentration in brain may reach 60–134 nM, 51–103 nM, 33–66 nM, and 28–56 nM at 1, 2, 4, and 8 hours postdosing, respectively. Given the potency for NP10679 at pH 6.9 is 23 nM, we anticipate that occupancy of GluN2B receptors at pH 6.9 is sufficiently high to drive significant GluN2B inhibition.

Fig. 3.

Pharmacokinetics of NP10679 in mouse. (A) Total plasma levels of NP10679 after a 10 mg/kg oral dose (black symbols) and a 3 mg/kg i.v. dose (open symbols) in mice are plotted. (B) Free plasma levels of NP10679 after a 2 mg/kg (open symbols) and a 5 mg/kg i.p. (black symbols) dose in mice are given. For (A) and (B), data are means ± S.E.M. (n = 3 per data point). The gray dashed line indicates the IC₅₀ of NP10679 for pH 6.9 inhibition of GluN2B receptors, the green dashed line is the functional IC₅₀ at H₁ histamine receptors, and the blue dashed line is the functional IC₅₀ at hERG potassium channels.

TABLE 3.

Pharmacokinetic parameters determined in mice for NP10679.

	Oral Bioavailability		Intraperitoneal Dosing		Brain:Plasma Ratiosa
Species	Mouse BALB/c		Mouse BALB/c		Mouse BALB/c
Dose (mg/kg)	10	3	2	5	3
Route	Oral	i.v.	i.p.	i.p.	i.v.
Studyb	3	3	4	4	1	2A	2B
Plasma Cmax (ng/ml)	3600	2350	581	1470	1740	600	817
Plasma Tmax (h)	1.0	0.08	0.5	0.5	1.0	1.0	1.0
Brain conc. (ng/g)	—	—	—	—	2300	1420	2100
Brain:plasma ratio	—	—	—	—	1.3	2.4	2.6
Plasma AUClast (h*ng/ml) (0–24 h)	45,000	17,800	4750	11,800	—	—	—
Plasma AUCinf (h*ng/ml)	49,500	—	5690	13,100	—	—	—
Plasma AUCextrap (%)	9.18	—	16.6	9.8	—	—	—
Plasma T1/2 (h)	7.06	8.56	9.9	7.5	—	—	—
Plasma MRTlast	7.16	7.08	7.5	6.7	—	—	—
Vss (L/kg)	—	1.59	—	—	—	—	—
CL (ml/min per kilogram)	—	2.44	—	—	—	—	—
Bioavailability (F %)	75.7	—	—	—	—	—	—

^aBrain:plasma ratio studies reported only at 1 h.

^bAll studies used 10% DMA/10% PG/30% HPBCD/50% sterile water formulation, except study 2B, which used 2% DMA/10% PG/30% HPBCD/58% sterile water. T1/2, elimination half-life; AUClast, area under the plasma concentration-time curve from time zero to last measurement; AUCinf, area under the plasma concentration-time curve from time zero to infinity; AUCextrap, area under the plasma concentration-time curve extrapolated from time t to infinity as a percentage of total AUC; MRTlast, mean residence time from time zero to last measurement; Vss, apparent volume of distribution at steady state; CL, apparent total body clearance of drug from plasma. A dash ( - ) indicates that parameter not determined.

Rotarod and Locomotor Activity

We subsequently tested whether NP10679 perturbed motor coordination or function. Mice were tested in a Rotarod challenge study after dosing with NP10679. Here, mice were trained on 2 consecutive days for ability to stay on the rotating and accelerating bar with four trials each day (intertrial interval of 25 minutes). Mice demonstrate improved performance from day 1 to day 2 and across intraday trials as shown in Fig. 4A. On day 3, mice were randomly assigned to treatment groups, dosed with vehicle or drug, and then tested four times beginning 25 minutes postdose, and the mean latency to fall was established for each trial (Fig. 4A). There was no significant impairment by NP10679 when dosed at 2 mg/kg or 5 mg/kg across all four trials. The 10 mg/kg NP10679 dose group had a reduced latency to fall in the fourth trial (87 ± 13 seconds) compared with vehicle (168 ± 14 seconds) that was significant (P < 0.01, ANOVA, Dunnett). However, no significant change from vehicle was observed in trials 1, 2, or 3 for this dose group. Reduced performance in trial 4 after 10 mg/kg NP10679 may be attributable to off-target inhibition of H₁ histamine receptors (Table 2). By contrast, a 30 mg/kg dose of ifenprodil led to a significant reduction in the latency-to-fall score in all four trials tested, with a score in the fourth trial of 47 ± 12 seconds (mean ± S.E.M., n = 8; Fig. 4B). The higher dose for ifenprodil was selected because it is less potent than NP10679 for inhibition of GluN2B receptors, with the ifenprodil IC₅₀ value of 170 nM for inhibition of GluN2B receptors (Kew et al., 1996; Mott et al., 1998) and 263 nM for neuroprotection in vitro (Chenard et al., 1991). Deficits observed in rotarod by ifenprodil may be related to α1-adrenergic receptor binding (110 nM) reported previously (Chenard et al., 1991, 1995).

Fig. 4.

Side effects measured in Rotarod challenge and locomotor activity. (A) Rotarod. Mice were trained on the Rotarod on 2 consecutive days (days 1 and 2), with four trials per day and an intertrial interval of 25 minutes. On day 3, mice were randomized to groups and administered vehicle (open circle), ifenprodil (open triangle, 30 mg/kg), or NP10679 at 2 mg/kg (light gray circle), 5 mg/kg (medium gray circle), or 10 mg/kg (black circle). The mean ± S.E.M. (n = 8) latency to fall for each trial was calculated for each group. *P < 0.01 from vehicle, ANOVA, Dunnett’s on day 3 within trials was determined. (B) Locomotor activity. Mice were habituated for 1 hour in a closed locomotor activity box and then removed and administered either vehicle (Veh, n = 6), MK-801 (MK at 0.3 mg/kg, n = 4) or NP10679 (20 mg/kg, n = 6 mice) by intraperitoneal injection and then placed back in the boxes, and the horizontal locomotor activity was counted for 2 hours. *P < 0.01, from vehicle (ANOVA, Dunnett’s). Total number of light beam breaks during the sample time are reported on the abscissa, which is representative of horizontal movement. Vehicle (Veh) is the formulation alone as described in Methods.

We also measured the ability of a single dose of NP10679 to alter the locomotor activity of mice in a closed, lighted chamber (Fig. 4B). After a 1-hour habituation period, mice were administered a 20 mg/kg dose of NP10679 or 0.3 mg/kg MK-801 and were returned to the closed lighted chamber, and horizontal activity was measured for 2 hours. NP10679 at this dose reduced horizontal activity, which was 1456 ± 194 beam breaks per 2 hours in drug compared with 2104 ± 219 beam breaks in vehicle (n = 6 each), but was not significant. Administration of 0.3 mg/kg MK-801 led to a large and significant increase (12,135 ± 3386; P < 0.01) in activity (n = 4).

Discussion

The most important finding of this study is the discovery of a new scaffold for GluN2B-selective antagonists that incorporates a piperazine into the chain linking two aromatic groups. This compound series shows strong sensitivity to extracellular pH, which should enhance therapeutic efficacy for treating acute injury while mitigating on-target side effects by virtue of enhanced potency in ischemic tissue that possesses an acidic extracellular pH compared with lower potency in healthy tissue at normal pH. Our in vivo preclinical data bear out this prediction, showing potent neuroprotection in a preclinical transient focal ischemia model, with limited behavioral effects on locomotor activity or Rotarod performance. Off-target profiling supports strong selectivity, with some activity at IC₅₀ values for four biogenic amine receptors that were 3–74 times higher than the IC₅₀ of NP10679 at GluN2B at pH 6.9 (23 nM), similar to what has been described for other pH-dependent GluN2B-selective inhibitors (Yuan et al., 2015). Similarly, there was no detectable inhibition of cytochrome P450 enzymes tested, and NP10679 has good metabolic stability in human and mouse liver microsomes and an acceptable plasma free fraction. In addition to these properties, NP10679 showed a pharmacokinetic profile that was superior to the straight chain analog EU93-31 (Yuan et al., 2015). In particular, NP10679 showed excellent brain penetration, a long plasma half-life, and high oral bioavailability (Table 3). This combination of properties suggests that clinical development of the piperazine series of GluN2B inhibitors such as NP10679 may provide an opportunity to advance this compound toward clinical trials for a number of unmet clinical needs involving acute neuronal injury. The utility of this class may be supported by the concept that conditions with high-frequency firing that produce metabolic changes in pH and local acidification, such as in inflammatory pain, may also be benefited.

One unique feature of this series of piperazine compounds is their pH sensitivity, which is higher than conventional well characterized GluN2B inhibitors that have been previously described (e.g., Fig. 1B). We specifically developed a structure activity relationship for this series with elevated pH sensitivity in mind to improve the safety profile of this series for use in acute injury. The idea builds on decades of work showing that extracellular pH acidifies during acute ischemic injury and in TBI for multiple reasons (Mutch and Hansen, 1984; Smith et al., 1986; Nedergaard et al., 1991; Katsura et al., 1992; Katsura and Siesjo, 1998). We do not, however, anticipate under normal excitatory synaptic transmission that NP10679 will appreciably engage the pH-sensitive mechanisms, since normal synaptic transmission does not produce detectable acidification. Rather, excitatory synaptic transmission typically produces a brief alkalinization (Tong et al., 2006; Makani and Chesler, 2007). In addition, reduced extracellular pH is not expected at extrasynaptic NMDA receptors in normal brain, suggesting NP10679 will be less effective in inhibiting these GluN2B-containing NMDA receptors under normal conditions. By contrast, the acidification that occurs during ischemia, driven both by elevated CO₂-producing HCO₃⁻ and H⁺ and a shift to anaerobic metabolism with production of lactic acid, will broadly reduce pH throughout the extracellular space. We assume that these mechanisms, which are strong drivers of infarct and penumbral acidification during ischemia, will equally affect both synaptic and nonsynaptic GluN2B-NMDA receptors. Thus, piperazine-containing GluN2B inhibitors should be effective in a range of indications that may lead to local acidification in brain, including subarachnoid hemorrhage. Indeed, the pH-sensitive GluN2B inhibitor NP10075, which also contains a piperazine core, was recently shown to improve behavioral measures in a preclinical model of SAH (Wang et al., 2014).

The endogenous neurosteroid pregnenolone sulfate potentiates maximally activated GluN2A-containing NMDA receptors to a greater extent at pH 6.5 compared with pH 8.5, raising the possibility that neurosteroid modulators in general might exhibit pH-sensitive actions (Kostakis et al., 2011). Jang et al. (2004) show that although mutagenesis studies suggest pregnenolone sulfate (but not the inhibitory neurosteroid 3α5βS) shares structural determinants with proton inhibition on the GluN2B subunit, the potentiating actions of pregnenolone sulfate on these NMDA receptors are not dependent on extracellular pH. Multiple neurosteroid inhibitors, such as 3α-ol-5β-pregnan-20-one hemisuccinate, can inhibit NMDA receptors and reduce cell death in cultured neurons challenged with NMDA as well as in vivo after MCAo in rats (Weaver et al., 1997, 2000). The IC₅₀ values for neurosteroid inhibition of NMDA receptors have not been reported at acidic pH, and thus it remains unclear whether this class of negative allosteric modulators exhibits pH dependence similar to NP10679. However, there is evidence that neurosteroid inhibition is use-dependent, which may be favorable under conditions observed during ischemia (Petrovic et al., 2005). Furthermore, negatively charged neurosteroid inhibitors (Weaver et al., 1997, 2000) may interact with charged amino acids that might alter their ionization state in acidic conditions, and thus the pH dependence remains an open question.

Whereas our in vivo side effect studies in Rotarod and locomotor activity after acute treatment with NP10679 showed limited effects at high doses, we have not tested NP10679 for cognitive side effects. Studies in rodents in the Morris water maze and operant-delayed match-to-position tasks with other selective GluN2B inhibitors have shown no adverse events (Guscott et al., 2003; Higgins et al., 2003; 2005) or have shown impairments in working memory in the water maze or Y maze (Zhang et al., 2013). However, when tested in nonhuman primates in cognitive tasks and learning paradigms after acute administration, two GluN2B inhibitors, traxoprodil and BMT-108908, produced cognitive impairments in a dose-dependent manner (Weed et al., 2016). In this study, effects were transient in nature and correlated with elevated GluN2B receptor occupancy. With respect to traxoprodil, which does not possess a significant pH sensitivity for NMDAR inhibition, we suggest the ∼6-fold pH-dependent shift in NP10679 potency may provide an additional advantage to further separate side effects from desired on-target ischemic tissue neuroprotection.

We believe the high potency, advantageous pharmacokinetics, pH sensitivity, demonstrated preclinical efficacy, and reduced side effects for piperazine analogs like NP10679 increase the likelihood that NMDA receptor inhibitors with these properties will be clinically viable for neurologic situations that are accompanied by acidification of the extracellular environment. If the pH sensitivity of the GluN2B-selective inhibitors can improve the therapeutic safety profile of these treatments in clinical trials for stroke that allow rapid delivery of drug to the site of action in patients stratified by image analysis (Hill et al., 2012; Saver 2013), the neuroprotective potential of NMDA receptor block may positively improve patient symptoms after acute injury.

Abbreviations

GluN1

glutamate N-methyl-d-aspartate receptor subunit 1

GluN2

glutamate N-methyl-d-aspartate receptor subunit 2

H₁

histamine receptor-1

HEK

human embryonic kidney

hERG

human ether-à-go-go

HPLC

high-performance liquid chromatography

HT_2A

serotonin receptor-2A

MCA

middle cerebral artery

MCAo

middle cerebral artery occlusion

NIMH

National Institutes of Mental Health

NMDA

N-methyl-d-aspartate

NMDAR

N-methyl-d-aspartate receptor

PDSP

Psychoactive Drug Screening Program

SAH

subarachnoid hemorrhage

SERT

serotonin transporter

TBI

traumatic brain injury

TTC

2,3,5-triphenyltetrazolium chloride

Authorship Contributions

Participated in research design: Myers, Ruppa, Wilson, Tahirovic, Menaldino, Koszalka, Zazcek, Dingledine, Traynelis, Liotta.

Conducted experiments: Myers, Ruppa, Wilson, Tahirovic, Lyuboslavsky, Menaldino, Dentmon.

Performed data analysis: Myers, Ruppa, Wilson, Tahirovic, Lyuboslavsky, Menaldino.

Wrote or contributed to the writing of the manuscript: Myers, Ruppa, Wilson, Tahirovic, Lyuboslavsky, Menaldino, Dentmon, Koszalka, Zazcek, Dingledine, Traynelis, Liotta.