Subunit-specific mechanisms and proton sensitivity of NMDA receptor channel block


Corresponding author S. Dravid: Department of Pharmacology, Emory University School of Medicine, Rollins Research Centre, 1510 Clifton Road, Atlanta, GA 30322-3090, USA. Email:


We have compared the potencies of structurally distinct channel blockers at recombinant NR1/NR2A, NR1/NR2B, NR1/NR2C and NR1/NR2D receptors. The IC50 values varied with stereochemistry and subunit composition, suggesting that it may be possible to design subunit-selective channel blockers. For dizocilpine (MK-801), the differential potency of MK-801 stereoisomers determined at recombinant NMDA receptors was confirmed at native receptors in vitro and in vivo. Since the proton sensor is tightly linked both structurally and functionally to channel gating, we examined whether blocking molecules that interact in the channel pore with the gating machinery can differentially sense protonation of the receptor. Blockers capable of remaining trapped in the pore during agonist unbinding showed the strongest dependence on extracellular pH, appearing more potent at acidic pH values that promote channel closure. Determination of pKa values for channel blockers suggests that the ionization of ketamine but not of other blockers can influence its pH-dependent potency. Kinetic modelling and single channel studies suggest that the pH-dependent block of NR1/NR2A by (−)MK-801 but not (+)MK-801 reflects an increase in the MK-801 association rate even though protons reduce channel open probability and thus MK-801 access to its binding site. Allosteric modulators that alter pH sensitivity alter the potency of MK-801, supporting the interpretation that the pH sensitivity of MK-801 binding reflects the changes at the proton sensor rather than a secondary effect of pH. These data suggest a tight coupling between the proton sensor and the ion channel gate as well as unique subunit-specific mechanisms of channel block.

N-Methyl-d-aspartate (NMDA) receptors are involved in key physiological processes, such as synaptic plasticity and development (Dingledine et al. 1999). In addition, excessive glutamate release and overactivation of NMDA receptors can induce excitotoxic neuronal death during cerebral ischaemia, traumatic brain injury and status epilepticus (Obrenovitch & Urenjak, 1997; Dirnagl et al. 1999; Lee et al. 1999; Meldrum et al. 1999). A large number of studies have shown that NMDA receptor antagonists can reduce neurotoxicity in animal models of ischaemia (Gill et al. 1987; Church et al. 1988; Park et al. 1988; Albers et al. 1989; Wang & Shuaib, 2005). The transient ischaemia during occlusive stroke triggers changes in the nature of the extracellular milieu, including strong acidification of the ischaemic core with more modest acidification of penumbral regions (Giffard et al. 1990; Nedergaard et al. 1991; Chesler & Kaila, 1992). NMDA receptors are inhibited by protons with an IC50 value in the range of pH 6.9–7.3 (Giffard et al. 1990; Tang et al. 1990; Traynelis & Cull-Candy, 1990, 1991; Vyklicky et al. 1990), suggesting that even modest acidification could reduce or perhaps delay the contribution of NMDA receptors to neuronal death until the pH gradients surrounding the ischaemic insult dissipate (Tombaugh & Sapolsky, 1993).

The physical location within the ion channel complex of the proton sensor responsible for this inhibition remains unknown. However, mutagenesis studies of NMDA receptor subunits suggest that residues that reside in the linker regions connecting the agonist binding domain to the transmembrane pore-forming elements influence proton-sensitive gating (Zheng et al. 2001; Low et al. 2003). Mutation of residues deep in the channel pore also has strong effects on proton sensitivity (Kashiwagi et al. 1997; Traynelis et al. 1998), suggesting that there is tight coupling between the proton sensor and gating. Several channel blockers have been suggested to bind close to the gating machinery and directly interact with the process of gating (Huettner & Bean, 1988; Mori et al. 1992; Wollmuth et al. 1998; Yuan et al. 2005; Johnson & Kotermanski, 2006). We hypothesized that small organic molecules that interact with the gating machinery to block the channel should exhibit pH-dependent actions. Here we determine the proton sensitivity at all NR1/NR2 combinations of a wide range of NMDA receptor channel blockers, some of which are clinically relevant with regard to neuroprotection. We found that the proton-dependent effects on several channel blockers, such as MK-801, show stereo-selectivity for both recombinant and neuronal NMDA receptors. Kinetic modelling and single channel experiments further suggest that acidic extracellular pH, which reduces receptor open probability, increases the association rate of (−)MK-801 but not (+)MK-801. These findings have broad implications for understanding the relation of NMDA receptor structure to function of the channel pore as well as clinical potential for development of channel blockers that are pH dependent and subunit selective.


Expression of NMDA receptors in Xenopus oocytes

Cyclic RNA was synthesized from linearized template cDNA according to manufacturer's specifications (Ambion) as previously described (Traynelis et al. 1998). The quality and quantity of the synthesized cRNA was assessed by gel electrophoresis and spectroscopy. Briefly, stage V and VI oocytes were surgically removed from the ovaries of Xenopus laevis anaesthetized with 3-amino-benzoic acid ethylester (1 g l−1). Clusters of oocytes were incubated with 292 U ml−1 type IV collagenase Worthington (Freehold, NJ, USA) or 1.3 mg ml−1 type IV collagenase (Life Technologies, Gaithersburg, MD, USA; 17018-029) for 2 h in Ca2+-free solution comprised of (mm): 115 NaCl, 2.5 KCl and 10 Hepes, pH 7.5 (adjusted by 5N NaOH), with slow agitation to remove the follicular cell layer. Oocytes were then washed extensively in the same solution supplemented with 1.8 mm CaCl2 and maintained in Barth's solution comprised of (mm): 88 NaCl, 1 KCl, 2.4 NaHCO3, 10 Hepes, 0.82 MgSO4, 0.33 Ca(NO3)2 and 0.91 CaCl2 and supplemented with 100 μg ml−1 gentamycin, 40 μg ml−1 streptomycin and 50 μg ml−1 penicillin. Oocytes were injected within 24 h of isolation with 5 ng of NR1-1a subunit cRNA (hereafter NR1; GenBank U11418) and 5–10 ng of NR2A, NR2B, NR2C or NR2D subunit cRNA (GenBank D13211, U11419, M91563 or L31611, respectively) in a 50 nl volume, and incubated in Barth's solution at 18°C for 3–7 days; some oocytes were stored at 4°C after 3–5 days. All laboratory practices and animal care were consistent with current NIH guidelines and all experimental protocols were approved by the Emory University Institutional Animal Care and Use Committee (IACUC).

Voltage-clamp recordings from Xenopus oocytes

Two-electrode voltage-clamp recordings were made 2–4 days postinjection. Oocytes were placed in a dual-track recording chamber with a single perfusion line that split to perfuse two oocytes. Dual recordings were made using Warner OC725B two-electrode voltage clamps configured as recommended by the manufacturer. The bath clamps communicated across silver chloride wires placed into each side of the recording chamber, both of which were assumed to be at a reference potential of 0 mV. Oocytes were perfused with a solution comprised of (mm): 90 NaCl, 1 KCl, 10 Hepes and 0.5 BaCl2, pH 6.9 or 7.6, and recorded at a holding potential between −20 and −80 mV. All experiments were performed at room temperature (23°C).

Measurement of locomotor activity

Locomotor activity was measured using eight Accuscan Digiscan Activity Monitors (AccuScan Instruments, Inc., Columbus, OH, USA), with the aid of the VersaMax(r) software (version 1.30, AccuScan Instruments, Inc.). Sprague–Dawley rats (100–150 g) were placed individually into clear acrylic chambers (40 cm × 40 cm × 30 cm), each inside a ventilated, sound-attenuating cabinet illuminated by candescent light (approximately 45 lx). Following 30–60 min habituation in the test room, rats were monitored for a 1 h habituation period in the test chamber prior to drug testing. After 1 h, the animals were injected intraperitoneally (i.p.) with various doses of MK-801 or saline, and returned to the testing chamber for 2 h. During testing, total movement in the horizontal plane (horizontal activity) was monitored by an array of infrared beams surrounding the chambers. Movements were determined by breaks in photobeams and converted into locomotor activity counts with the aid of the software VersaDat(r). All test compounds were dissolved in physiological saline (0.9%) to provide the necessary drug concentrations for an injection volume of 1 ml kg−1.

Phencyclidine discrimination

Seven adult male Sprague–Dawley rats (COBS CD) were trained to discriminate injections of 2.0 mg kg−1 phencyclidine (PCP) from saline, as previously described (Willetts & Balster, 1988). They were individually housed with free access to water under a 12 h light–12 h dark cycle. Food (Harlan Teklad Rodent Diet, Williamston, IL, USA) access was restricted in order to increase lever-pressing for food. The subjects were trained daily (Monday–Friday) in 30 min sessions in standard two-lever operant conditioning chambers, that require the animal to make a decision (Coulbourn Instruments, Lehigh Valley, PA, USA). Completion of a fixed ratio of 32 on the correct lever (i.e. 32 consecutive correct lever presses) resulted in delivery of a 45 mg food pellet (P.J. Noyes Company, Inc., Lancaster, NH, USA). Saline and PCP were given i.p. under a double alternation schedule, 15 min prior to the beginning of the session. Incorrect responding reset the fixed ratio for correct-lever responding. Test sessions were conducted on Tuesday and Friday when the subjects met the following criteria on the four preceding training sessions (two PCP and two saline): (1) first fixed ratio completed on the correct lever; and (2) greater than 85% correct-lever responding over the entire session. During test sessions, completion of a fixed ratio on either lever resulted in the delivery of food.

The rats were tested at (+)MK-801 doses of 0.01–0.25 mg kg−1 and (−)MK-801 doses of 0.1–1.7 mg kg−1, administered i.p. 30 min prior to each session. A dose–effect curve for PCP was obtained in the same animals for purposes of comparison. Data from test sessions were analysed by determining the mean percentage of responses on the PCP-associated lever and the overall mean response rate for all subjects. Data from sessions with less than 0.05 responses s−1 were excluded from determination of the mean percentage PCP-lever responding. Full substitution for PCP required greater than 80% PCP-lever responding, while part substitution was defined as producing between 20 and 80% PCP-lever responding. To compare the relative potencies of the test drugs, the ED50 values for generalization from PCP (the dose resulting in 50% substitution for the PCP training dose) and for response rate effects that reflect intoxication (the dose required to cause a 50% decrease in rates of responding relative to saline control test rates) were calculated using regression analysis of the linear portions of the dose–effect curves. Saline control rates were calculated for each subject as the average of the rates of responding on saline control tests, which preceded and followed each dose–response curve determination. Significant differences in potency were established when the 95% confidence limits for the ED50 values did not overlap. After the initial training for PCP discrimination, the rats were tested with different drugs for a period of 12–18 months. After training, animals were rarely tested for only one drug, since there is no evidence to suggest that previous testing alters subsequent results. The animals were only killed owing to age or health issues, using a high dose of pentobarbitone or CO2 inhalation followed by bilateral pneumothoracotomy.

Flurothyl seizures

Male C57Bl/6 mice (42 days old) were obtained from Jackson Laboratories. Mice were injected i.p. with saline (0.9%), (+)MK-801 (0.01–3 mg kg−1 in saline), or (−)MK-801 (0.1–3 mg kg−1 in saline), 15 min before inhalation exposure to flurothyl (2,2,2-trifluroethyl ether, Sigma). Seizures were induced by placing mice individually in a 2.8 l closed plastic chamber into which flurothyl was introduced. Flurothyl was administered by infusion (20 μl min−1) using a 1 ml syringe driven by a pump (KD Scientific Model 200) onto filter paper suspended at the top of the chamber. The latencies to onset of clonic seizures, tonic–clonic seizures and tonic hindlimb extension were recorded. Tonic hindlimb extension was defined as bilateral tonic extension of the forelimbs followed by the hindlimbs. At the onset of tonic hindlimb extension, flurothyl infusion was ceased and the chamber was opened to room air.

Intracellular Ca2+ monitoring

Primary cultures of neocortical neurons were obtained from embryonic day 16–17 Swiss-Webster mice. Briefly, pregnant mice were killed by CO2 asphyxiation, and embryos were removed under sterile conditions. Neocortices were collected, stripped of meninges, minced by trituration with a Pasteur pipette and treated with trypsin for 25 min at 37°C. The cells were then dissociated by two successive trituration and sedimentation steps in soybean trypsin inhibitor and DNase containing isolation buffer, centrifuged, and resuspended in Eagle's minimal essential medium with Earle's salt and supplemented with 2 mm l-glutamine, 10% fetal bovine serum, 10% horse serum, 100 IU ml−1 penicillin and 0.10 mg ml−1 streptomycin, pH 7.4.

Intracellular Ca2+ was monitored as previously described using the Ca2+-sensitive fluorescent dye fluo-3 AM (Dravid et al. 2004). Briefly, mouse neocortical neurons grown in 96-well plates were washed four times with Locke's buffer containing (mm): 154 NaCl, 5.6 KCl, 1 MgCl2, 2.3 CaCl2, 8.6 Hepes, 5.6 glucose and 0.1 glycine, pH 7.4. Cells were washed with an automated cell washer (Labsystems, Helsinki, Finland), and subsequently incubated for 1 h at 37°C with dye loading buffer (100 μl per well) containing 4 μm fluo-3 AM and 0.04% pluronic acid in Locke's buffer. After 1 h incubation in dye loading medium, cells were washed four times with Locke's buffer. The final volume of Locke's buffer in each well was 100 μl. Solutions of MK-801 and NMDA were made at four times the final desired concentration, and 50 μl was added sequentially to the culture wells to give a final well volume of 200 μl. Calcium ion fluorescence was monitored using FLIPR (Fluorometric Imaging Plate Reader, Molecular Devices). Neurons were excited by the 488 nm line of the argon laser, and Ca2+-bound fluo-3 AM emission in the 500–560 nm range was recorded with the CCD camera shutter duration set at 0.4 s. Prior to each experiment. average baseline fluorescence was set between 10 000 and 20 000 fluorescence units by adjusting the power output of the laser. Fluorescence readings were taken every 1–6 s.

Cytotoxicity assay

Lactate dehydrogenase (LDH) was measured in medium from neocortical neurons grown in 12-well plates as previously described (Dravid et al. 2004). Briefly, conditioned medium was collected, and neocortical neurons were washed three times with Locke's buffer. The neurons were pre-incubated for 5 min with the indicated concentrations of MK-801, followed by exposure to 30 μm NMDA plus MK-801 for 2 h at 22°C. At the termination of NMDA exposure, the incubation medium was collected for later analysis of LDH activity, and the neurons were washed twice in Locke's buffer followed by replacement with 0.5 ml of the previously collected conditioned medium. The cell cultures were then returned to a 37°C incubator and, 24 h after NMDA exposure, growth medium was collected and saved for analysis of lactate dehydrogenase activity according to the method of Koh & Choi (1987).

Whole-cell patch clamp recording

Whole-cell patch recording was performed in human embryonic kidney cells-293 (HEK 293 cells; ATCC CRC-1573) as previously described (Traynelis & Wahl, 1997). Briefly, HEK 293 cells were plated onto 12 mm glass coverslips that were coated with poly d-lysine (5 μg ml−1). Cyclic DNAs encoding a fusion protein of NR1–GFP (constructed by replacing the C-terminal of NR1-1a after Glu925 with GFP) and NR2A (GenBank D13211) were transiently transfected into cells using the calcium-phosphate precipitation method. The ratio of cDNAs encoding NR1–GFP:NR2A was 1:1 with a final concentration of 0.5 μg ml−1 of cDNA. After 6 h of incubation with cDNA, the medium was replaced and supplemented with 2–3 mm Mg2+ and 200 μmdl-2-amino-5-phosphonopentanoic acid. Recordings were typically made over the next 24 h. External recording solution consisted of (mm): 150 NaCl, 10 Hepes, 3 KCl, 0.5 CaCl2 and 0.01 EDTA; pH 7.4, 0.31–0.33 osM l−1. The internal solution consisted of (mm): 110 gluconic acid, 30 CsCl, 4 NaCl, 5 Hepes, 5 BAPTA, 0.5 CaCl2 and 2 MgCl2; pH 7.3 (adjusted with CsOH), 0.29–0.30 osM l−1. Whole-cell current recordings were digitized at 2 kHz using Clampex version 8.0 (Axon Instruments, Union City, CA, USA). Rapid solution exchange was achieved with a two-barrel theta glass pipette controlled by piezoelectric translator (Burleigh). Fitting of macroscopic currents was performed using Channelab (Synaptosoft, Decatur, GA, USA).

Single channel recordings

HEK 293 cells were transiently transfected by the calcium phosphate method or using Fugene transfection reagent (Roche Diagnostics) with wild-type NR1, NR2A and GFP at a ratio of 1:2:1. Outside-out membrane patches were obtained from transiently transfected HEK 293 cells as previously described (Erreger et al. 2005a), and NR1/NR2A current responses were recorded under voltage clamp at −60 mV using an Axopatch 200B amplifier. Current recordings were filtered at 5 kHz using an eight-pole Bessel filter (−3 dB; Frequency Devices) and digitized at 20–40 kHz (Clampex version 9). For outside-out patches, the external and internal solutions were the same as for whole-cell currents.

For cell-attached patches, the pipette solution consisted of (mm): 150 NaCl, 10 Hepes, 0.5 CaCl2, 3 KCl, 1 glutamate, 0.05 glycine and 0.01 EDTA (pH 7.3 unless otherwise stated, 23°C) and the pipette potential was set to +60 mV.

Single channel records were segmented and idealized using the QuB software ( Data were segmented into 50 ms segments containing only apparently a single active channel. Segments with multiple channels simultaneously open were not analysed. A resolution of 50 μs was imposed on the idealized data record. Dwell-time histograms were fitted with multiple exponential components using maximum likelihood methods (ChanneLab).

Determination of pKa

A PCA200 pKa Titrator (Sirius Analytical Instruments, East Sussex, UK) was used to determine pKa values for some of the channel blockers, with the assistance of Refinement Pro software (Sirius Analytical Instruments, East Sussex, UK). MK-801 (maleate salt) was dissolved at 1.315 mm in 80% MeOH in water with 0.15 m KCl. Three titrations were performed and a pKa determined using the Yasuda–Shedlovsky extrapolation implemented as described by the manufacturer.


Figure 1 shows the structures for all compounds tested. MK-801 ((5R,10S)-5-methyl-10,11-dihydro-5H-dibenzo[a,d]cyclohepten-5,10-imine) stereoisomers, norketamine, memantine and remacimide were obtained from Tocris Bioscience (Ellisville, MO, USA). Ketamine isomers, CNS1102, pentamidine, 9-aminoacridine, amantadine and dextromethorphan were obtained from Sigma Chemical Co. (St Louis, MO, USA). 1-Phenylcyclohexylamine (PCA), N-methyl PCA, N-ethyl PCA, N-propyl PCA, dextrorphan, levorphanol and levomethorphan were a gift from Dr Stephen Holtzman.

Figure 1.

Structures of NMDA antagonists tested for their pH sensitivity

Statistical analysis

Student's unpaired t test was used for comparisons unless otherwise stated. ANOVA with Newman–Keuls post hoc test was performed for multiple comparisons. Data are expressed as means ±s.e.m.;P < 0.05 was considered statistically significant.


Anticonvuslant activity and side-effect profile of MK-801 stereoisomers in vivo

The stereoisomers of MK-801 are potent NMDA antagonists that induce a number of behavioural effects when administered in vivo. The relative potency of the two stereoisomers of MK-801 was evaluated in a series of in vivo assays (Table 1 and Fig. 2). The laevorotatory isomer, (−)MK-801, is less potent at blocking native NMDA receptors and is therefore considered the biologically less active isomer (Raffa et al. 1989; Khanna et al. 1991; Kochhar et al. 1991). First, we tested the ability of the MK-801 stereoisomers to block generalized seizures in a murine model of epilepsy in which male C57Bl/6 mice inhaled convulsant doses of the anaesthetic flurothyl (see Methods). We estimated the anticonvulsant potency as the concentration of compound that increases the time to tonic hindlimb extension by 3 min (Fig. 2A and Table 1). By this measure, we calculated a threefold potency difference between (+)MK-801 (0.28 mg kg−1) and (−)MK-801 (0.85 mg kg−1), consistent with previous studies (Wong et al. 1986).

Table 1.  ED50 for in vivo tests of MK-801 effects
 (+)MK-801 ED50 (mg kg−1)(−)MK-801 ED50 (mg kg−1)(+)MK-801 ED50/(−)MK-801 ED50
  1. a ED50 to increase the latency to tonic hindlimb extension by 3 min. b ED50 to increase horizontal locomotor activity threefold over saline baseline. c ED50 for substitution for the PCP training dose (95% confidence interval 0.02–0.04 for (+)MK-801 and 0.18–0.44 for (−)MK-801); ED50 for PCP was 1.0 mg kg−1. d ED50 for decreases in rates of responding relative to saline control test sessions (95% confidence intervals 0.17–0.25 for (+)MK-801 and 0.45–17.7 for (−)MK-801); ED50 for PCP was 6.7 mg kg−1. The decrease in response rates probably reflected intoxication.

Seizurea0.280.85 3.0
Locomotor activityb0.122.7122.6
PCP substitutionc0.030.28 9.3
PCP response rated0.212.8113.4
Figure 2.

(+)MK-801 isomer is more potent isomer than (−)MK-801
A, the delay to the initiation of the tonic hindlimb extension (THE) is shown for C57Bl/6 mice in a chamber infused with the convulsant anaesthetic flurothyl (see Methods). The grey lines indicate the dose required to delay progression to hindlimb extension by 3 min (180 s). The filled in area at the bottom of the graph shows saline control. Data points are means ±s.e.m. of 8–11 rats. B, administration of MK-801 was performed after a 1 h habituation, and locomotor activity monitored over a 2 h period. The total horizontal activity from different rats was averaged and expressed as a function of dose. The biphasic nature of the response to (+)MK-801 reflected the severe ataxia that occurred with the highest doses. Data points are means ±s.e.m. of 3–11 rats. Ca shows the percentage of the correct PCP-lever responding and Cb the response rates. Values above PCP and saline are the results of control tests conducted before each dose–response curve. Mean percentage PCP-lever responding was based on 6 of 7 rats for the (−)MK-801 doses of 0.17 and 0.25 mg kg−1 and 3 of 7 rats for the PCP dose of 8 mg kg−1. All other data points are the means ±s.e.m. of 7 rats. Concentration values are plotted on a logarithmic scale.

Despite well-described neuroprotective and anticonvulsant actions of MK-801, a number of serious side-effects, including ataxia and psychoses, have prevented further therapeutic consideration (Ellison, 1995; Thornberg & Saklad, 1996; Andine et al. 1999). Both of these effects are considered to arise from strong MK-801 block at all NMDA receptor subunits in a non-selective fashion. We therefore evaluated the efficacy and dose dependence of the two MK-801 stereoisomers in animal models of ataxia and psychomimetic effects. MK-801 is well known to increase locomotor activity, followed at higher doses by motor impairment that progresses to ataxia (Tricklebank et al. 1989; Hiramatsu et al. 1989). Locomotor activity of rats in units of light beam breaks was recorded from optical monitoring boxes. Animals spent 1 h habituating, after which they were injected i.p. with increasing doses of MK-801 stereoisomers or saline. Both isomers caused a dose-dependent increase in locomotor activity (Fig. 2B and Table 1), which reached a plateau for (−)MK-801 but decreased for (+)MK-801 as the dominant behaviour became motor impairment or ataxia. The potency for (+)MK-801 doubling of activity level was ∼22-fold higher than for (−)MK-801.

Drug discrimination procedures have been useful for comparing the acute effects of psychoactive drugs and presumed analogues, and are considered to produce results predictive of the drugs' subjective effects in humans (Holtzman, 1990; Balster, 1991). Typically, channel blocking agents with high affinity, such as PCP, dextrorphan and (+)MK-801, produce full substitution for one another (Balster & Willetts, 1988; Koek et al. 1990; Nicholson et al. 1999). Several previous studies that have tested both MK-801 isomers present a complex set of results (Genovese & Lu, 1991; France et al. 1991; Geter-Douglass & Witkin, 1997). We evaluated both MK-801 isomers for production of PCP-like discriminative stimulus effects in rats trained to discriminate PCP from saline in a standard two-lever operant conditioning model as summarized in Figs 2C1 and 2C2 and Table 1. Phencyclidine produced a dose-dependent increase in PCP-lever responding, with one or more doses producing full substitution in all subjects. Both (+)MK-801 and (−)MK-801 also produced a dose-dependent substitution for the PCP training dose. While (+)MK-801 produced full substitution in all subjects at one or more doses that did not suppress rates of responding, only six of seven rats receiving (−)MK-801 fully generalized from PCP. The remaining rat selected only the saline lever across the doses tested. As shown in Table 1, (+)MK-801 was ∼10-fold more potent than its stereoisomer and over 30-fold more potent than PCP for both behavioural measures. In contrast, (−)MK-801 was only two- to threefold more potent than PCP in the production of response rate suppression and PCP-like discriminative stimulus effects.

Antagonist stereoselectiviy and proton sensitivity in recombinant NMDA receptors

We determined MK-801 stereoisomer potency at pH 7.6 using two-electrode voltage-clamp recordings of Xenopus oocytes expressing recombinant NR1 in combination with each of the four different NR2 subunits. By ‘potency’, we refer to the concentration–effect relationship between channel blocker and receptor, which is typically quantified by determining a half-maximally effective concentration for inhibition of the current (IC50). The potency for (+)MK-801 inhibition is similar among all four NR2 subunits at pH 7.6 (IC50 range 9–38 nm; Fig. 3 and Table 2; see also Monaghan & Larsen, 1997). However, (−)MK-801 exhibits more variability in potency among NR2 subunits (IC50 range 32–354 nm). Interestingly, the relative difference in potency between the two stereoisomers of MK-801 at pH 7.6 is strongly dependent on the identity of the NR2 subunit. NR1/NR2A receptors showed the greatest difference in potency between the two stereoisomers, with (+)MK-801 being 23-fold more potent than (−)MK-801 (Fig. 3A). (+)MK-801 is 3.6-fold, 1.6-fold and 4.4-fold more potent than (−)MK-801 at NR1/NR2B, NR1/NR2C and NR1/NR2D receptors, respectively (Fig. 3 and Table 2).

Figure 3.

NR2 subunit dependence of MK-801 potency
Recombinant NMDA receptors expressed in oocytes were recorded using two-electrode voltage clamp and activated by coapplication of 50 μm glutamate and 30 μm glycine plus increasing concentrations of (+)MK-801 or (−)MK-801 at pH 7.6 to establish the concentration–response curve for MK-801 inhibition of all heterodimeric NR1/NR2 subunit combinations (A, NR2A or NR2B; B, NR2C or NR2D). MK-801 was applied for 1–2 min at each concentration. The oocytes were held at −40 mV. Each point represents measurements from 6–40 oocyte recordings.

Table 2.  pH dependence of channel blocker potency
NMDA receptor antagonistspKa
pKaTotal protonated (%)NR1/2 A IC50m)NR1/2B IC50m)NR1/2C IC50m)NR1/2D IC50m)
ExptPredpH 7.6pH 6.9pH 7.6pH 6.9RatiopH 7.6 7.6/6.9pH 6.9RatiopH 7.6 7.6/6.9pH 6.9RatiopH 7.6 7.6/6.9pH 6.9Ratio 7.6/6.9
  1. PCA is 1-phenylcyclohexylamine. Pred is the predicted pKa and Expt is the experimentally determined pKa value (see below) reported by: a MacDonald et al. (1991); b Moffat et al. (2004); c Baselt (2004); d Mozayani (2003); e Freudenthaler et al. (1998); f Spector (1988); and g Casadio & Melandri (1977). * Indicates pKa measurements determined in this study (see Methods). Values of IC50 were determined from two-electrode voltage-clamp recordings from between 6 and 42 oocytes at each pH value (holding potential =−40 mV). NMDA receptors were activated by coapplication of 50 μm glutamate and 30 μm glycine. Averaged concentration–effect data were fitted (least-squares criterion) by: Percentage response = 100/(1 + ([blocker]/IC50)n), where IC50 is the concentration of blocker at which the response is 50% inhibited and n is the Hill slope, which ranged between 0.6 and 2. Potency boost (i.e. ratio 7.6/6.9) is the IC50 at pH 7.6 divided by IC50 at pH 6.9 for the average concentration–response curve where every oocyte is equally weighted. Values of pKa were calculated using ACD (Advanced Structural Development) Structure Design Suite software (v9.0, ACD Inc., Toronto, ON, Canada; Slater et al. 1994). Structures were drawn in Chembridge Soft's ChemDraw Ultra 8.0 and exported to Structure Design Suite. The calculations for percentage unprotonated species are based on predicted pKa when reported pKa is not available.

N-Methyl PCA10.199.799.94.921.563.23.530.675.3
N-Ethyl PCA10.299.81001.490.3664.10.440.231.9
N-Propyl PCA10.299.81001.060.3523.00.630.361.7

NMDA receptors are inhibited by extracellular protons (IC50 value near physiological pH, or 50 nm H+) in a manner that is proposed to reflect inhibition of conformational changes that are required for pore opening (Banke et al. 2005). MK-801 is an open channel blocker that binds deep within the channel pore and can become trapped after agonist unbinding. Several recent findings suggest that MK-801 and other so-called trapping channel blockers interact with the gating machinery to alter channel function (e.g. Yuan et al. 2005; Blanpied et al. 2005). Since both protons and trapping channel blockers interact with NMDA receptor gating, we examined whether protonated receptors might interact differently with MK-801 compared with receptors recorded at alkaline pH. To examine the nature of coupling between channel blockers and proton-sensitive gating of NMDA receptors, we studied inhibition of receptor-mediated currents in Xenopus laevis oocytes at pH 6.9 and 7.6 by both stereoisomers of MK-801 (Fig. 4 and Table 2). The potency of (−)MK-801 is 6.4-fold higher at pH 6.9 than at pH 7.6 for NR1/NR2A. This pH dependence of potency is paradoxical because the channel open probability is lower at low pH (Fig. 8 and Table 6; see also Banke et al. 2005), resulting in reduced access for the open channel MK-801 binding site. For NR1/NR2D, (−)MK-801 shows more modest pH selectivity (potency boost at low pH) of 2.8-fold. (+)MK-801 is 2.5-fold more potent at pH 6.9 than at pH 7.6 for NR1/NR2A. Other combinations of (+) and (−)MK-801 isomers and NR2 subunits did not show strong differences in potency between pH 6.9 and 7.6.

Figure 4.

Differential coupling of MK-801 stereoisomers to proton-sensitive gating of recombinant NMDA receptors
A, two-electrode voltage-clamp current recordings show the inhibition of NR1/NR2A NMDA receptors by (−)MK-801 at pH 7.6 and 6.9 in Xenopus oocytes. The NMDA receptors were activated by 50 μm glutamate and 30 μm glycine; increasing concentrations (nm) of (−)MK-801 were coapplied with agonists to establish the concentration–response curve. The oocytes were held at −40 mV. B, concentration–response curves for NMDA receptor current inhibition by MK-801 isomers at pH 7.6 and 6.9 at NR1/NR2A receptors. Each point represents the mean ±s.e.m. of 6–40 oocyte recordings. The data for pH 7.6 are the same as those shown in Fig. 3.

Figure 8.

Non-linear fitting kinetic scheme to response waveforms suggests that (−)MK-801 has a faster association rate at protonated NR1/NR2A receptors
A, unitary currents shown were recorded from cell-attached patches on HEK 293 cells expressing NR1/NR2A receptors (filtered at 5 kHz, −3 dB). The pipette contained 1 mm glutamate and 50 μm glycine (pipette potential +60 mV). Upward deflections reflect channel openings. Representative traces are shown for two different patches at pH values of 7.3 and 6.7. B, the distribution of open channel dwell times for the same patches shown in A is plotted as a histogram on a log-square root (SQRT) scale. The open time distribution was fitted with the sum of two exponential components for each patch (see Table 6 for mean open time constants). C, the arithmetic mean channel open duration is significantly shorter at lower pH (n= 6–9, ANOVA *P < 0.05 **P < 0.01; see Table 6 for fitted time constants). D, whole-cell recording from NR1/NR2A receptors at −60 mV (pH 6.9) in response to 50 μm glutamate and 50 μm glycine in the absence or presence of 100 nm (−)MK-801. The trace has been converted to open probability as described in the text. Data were fitted to the model depicted in Fig. 7 using a simplex algorithm to alter model parameters at each iteration; least-squares criteria were employed as determined from normalized waveforms. Fitted waveform is shown in terms of open probability predicted by the model (grey curve). The mean rate constants obtained by fitting each whole-cell recording are given in Table 7.

Table 6.  Open times for NR1/2A receptors
pHMean open time (ms)τ1 (μs)Area (%)τ2 (ms)Area (%) NP o
  1. Cell-attached patches from HEK 293 cells transfected with NR1/NR2A channels were activated by 1 mm glutamate and 50 μm glycine (n= 6–9). Mean open time is the arithmetic mean open duration independent of any fitting. The open time histograms were best fitted by two exponential components. The time constant and relative areas are given for these two components. The value of NPo quantifies the proportion of time spent in the open channel state during analysed segments.

8.03.2 ± 0.366 ± 5  33 ± 23.5 ± 0.467 ± 20.305 ± 0.035
7.32.3 ± 0.271 ± 4  26 ± 32.5 ± 0.374 ± 30.298 ± 0.035
6.71.6 ± 0.159 ± 0.333 ± 41.7 ± 0.167 ± 40.051 ± 0.012

A series of chiral channel blockers (see Fig. 1), including dextromethorphan/levomethorphan, dextrorphan/levorphanol, ketamine and norketamine were tested for potency of inhibition across NR2 subunits at pH 6.9 and 7.6 (Table 2). We also examined non-chiral blockers memantine, amantidine, 9-aminoacridine, pentamidine and several phenylcyclohexylamine derivatives (Bormann, 1989; Erdo & Schafer, 1991; Costa & Albuquerque, 1994; Kornhuber et al. 1994). These compounds vary widely in their structures, and have been suggested to have different binding sites within the pore and different kinetics of block and unblock at NMDA receptors (Sobolevsky et al. 1999; Sobolevskii & Khodorov, 2002; Blanpied et al. 2005; LePage et al. 2005; Yuan et al. 2005). Thus we expect them to show differential sensitivity to extracellular protons. With the exception of ketamine derivatives (MacDonald et al. 1991), none of the channel blockers changes its ionization in ways that might substantially influence the proton sensitivity.

The trapping blockers (−)MK-801, ketamine and norketamine all showed a potency boost greater than fivefold at pH 6.9 compared with pH 7.6 for NR2A receptors (Table 2). The pH-dependent potency boost for MK-801 isomers was dependent on the identity of the NR2 subunit, being pronounced for NR1/NR2A, but nearly absent for NR1/NR2B, NR1/NR2C or NR1/NR2D. Other compounds showed modest pH sensitivity, with no obvious trends (Table 2). Furthermore, the differential pH sensitivity of stereoisomers of MK-801 suggests that a unique structure–activity relationship may exist, describing how a channel blocker interacts with proton-sensitive gating. For ketamine, it has been shown previously that acidification increases the potency of ketamine, which could be accounted for by the enhanced fraction of the charged form at low pH if the charged form is more potent than the uncharged form (MacDonald et al. 1991). However, ionization cannot account for the pH sensitivity of MK-801 isomers, since the pKa of MK-801 (8.37) is greater than that of ketamine (7.5), rendering the ionization state of MK-801 less sensitive to changes from pH 7.6 to 6.9. Table 2 summarizes the effect of pH on the ionization of channel blockers.

The differences in the pH sensitivity of the two MK-801 stereoisomers implies that structural determinants of binding must also dictate how pH affects the potency of channel blockers. Interestingly, a number of channel blockers showed distinct potencies at each of the NR2 subunits (see Table 2). The dependence of potency on the NR2 subunit suggests that either the kinetics or structural determinants of block are influenced by the NR2 subunits. Although conservation of proposed membrane-lining helices and re-entrant loops is high, structural determinants of the channel block site probably vary among the NR2 subunits.

Voltage dependence of MK-801 channel block in recombinant NMDA receptors

Use-dependent open channel block of NMDA receptors by dissociative anaesthetics such as PCP and ketamine is usually characterized by voltage dependence, with blockers showing a greater potency at hyperpolarized potentials than at depolarized potentials (Huettner & Bean, 1988; MacDonald & Nowak, 1990; MacDonald et al. 1991; Bolshakov et al. 2003). However, blockade of NMDA receptors by higher affinity antagonists, such as (+)MK-801, is less dependent on membrane potential (Halliwell et al. 1989). Voltage dependence of open channel blockers could be due to interaction with a permeant ion binding site that resides deep within the transmembrane electric field (Zarei & Dani, 1995). In order to examine whether the proton-mediated enhancement of potency reflects a change in the manner by which MK-801 interacts with the transmembrane electric field, we tested the voltage dependence of the potency of both MK-801 stereoisomers at pH 6.9 and 7.6 over a range of holding potentials (−20 to −80 mV). Concentration–response curves were established in oocytes expressing recombinant NMDA receptors, and recorded over a range of holding potentials. MK-801 was applied for up to 1–2 min at increasing doses. For NR1/NR2A, 1–2 min of application was sufficient for reaching steady state. For NR1/NR2C and NR1/NR2D, a duration of 2 min was used. However, this duration was not sufficient to reach steady state for concentrations less than the IC50 of MK-801, which may have led to an underestimation of potency. At pH 7.6, both isomers showed modest voltage dependence at NR1/NR2A receptors (Table 3), with (+)MK-801 and (−)MK-801 showing e-fold decrease in IC50 per −22 and −19 mV hyperpolarization in holding potential, respectively. The stereoisomers of MK-801 blocked NR1/NR2B receptors in a different fashion at pH 7.6 (Table 3), with (+)MK-801 somewhat less voltage dependent than (−)MK-801, showing e-fold change in IC50 per −57 and −27 mV for (+)MK-801 and (−)MK-801, respectively. Acidification had a strong effect on voltage dependence of both isomers. At pH 6.9, the voltage dependence of (−)MK-801 was reduced at both NR1/NR2A and NR1/NR2B receptors; voltage dependence could not be fitted by a Boltzmann equation for (+)MK-801 at pH 6.9. The loss of voltage dependence for the (+) isomer compared to the partial maintenance of voltage dependence for the (−) isomer at acidic pH might account for a portion of the differential proton sensitivity of the stereoisomers. These data further suggest that protonation of the NMDA receptor can modify the manner in which (+)MK-801 interacts with the channel pore, including its ability to sense the intrapore electric field.

Table 3.  Voltage dependence of MK-801 potency
 Potential (mV)(+)MK-801(−)MK-801
IC50 (nm) pH 7.6IC50 (nm) pH 6.9IC50 ratio 7.6/6.9IC50 (nm) pH 7.6IC50 (nm) pH 6.9IC50 ratio 7.6/6.9
  1. Composite concentration–response curves for MK-801 isomers were recorded from Xenopus oocytes expressing NR1 and different NR2 subunits (see Methods and Table 2) over a range of holding potentials (−20 to −80 mV). NMDA receptors were activated by coapplication of 50 μm glutamate and 30 μm glycine. Responses for each oocyte were averaged together over all experiments. The IC50 values were determined as described in Table 2. Six to 42 oocytes were recorded from for each condition; mV per e-fold change in IC50 was calculated by fitting a plot of the IC50versus voltage to the equation: EC50 (V) = Amplitude/{1 + exp[(VVo)/(dV)]}, where dV is the mV per e-fold change.

mV per e-fold change in IC50−22 mV−19 mV−63 mV
mV per e-fold change in IC50−57 mV−27 mV−52 mV

Proton sensitivity of MK-801 antagonist potency in native NMDA receptors

In order to test whether the ability of MK-801 isomers to sense the protonation state of recombinant receptors is relevant to native neuronal NMDA receptors, we performed calcium imaging and cytotoxicity assays in mature neocortical neurons. Neocortical neurons possess heteromeric NMDA receptors composed of NR1 subunit and predominantly NR2A and NR2B subunits (Li et al. 1998; Janssens & Lesage, 2001). Fluo-3 AM-loaded neurons were treated with 30 μm NMDA, and the inhibition of Ca2+ influx by MK-801 application was monitored. When the extracellular pH was changed from pH 7.6 to 6.9, (−)MK-801 showed a 5.8-fold increase in potency, whereas (+)MK-801 showed a smaller 3.6-fold increase in potency (Fig. 5A and B and Table 4). Thus, similar to the proton-induced shift in potency observed in recombinant receptors, (−)MK-801 isomer has a higher proton sensitivity compared with (+)MK-801 in native NMDA receptors in neocortical neurons.

Figure 5.

Differential coupling of MK-801 stereoisomers to proton-sensitive gating of native NMDA receptors
A, inhibition of NMDA (30 μm; 100 μm glycine)-mediated calcium influx in cerebral cortical neurons by MK-801 isomers was determined as a function of pH using the Ca2+-sensitive dye fluo-3 AM (see Methods). B, each point on the concentration–response curve for MK-801 inhibition is the mean ±s.e.m. of the area under the curve (n= 4–6). C, the potency of MK-801 isomers was determined against an excitotoxic insult by 30 μm NMDA plus 100 μm glycine in neocortical neurons (see Methods). Each point on the concentration–response curve represents the mean ±s.e.m. of measurements (n= 3).

Table 4.  pH dependence of MK-801 potency in native NMDA receptors
IC50 (nm) pH 7.6IC50 (nm) pH 6.9IC50 ratio 7.6/6.9IC50 (nm) pH 7.6IC50 (nm) pH 6.9IC50 ratio 7.6/6.9
  1. A pH-dependent shift in the potency of MK-801 isomers is evident in native mouse NMDA receptors. Intracellular calcium was monitored using the Ca2+-sensitive dye fluo-3 AM in neocortical neurons. Cell death was measured using the LDH efflux assay in neocortical neurons at two different pH values. Values of IC50 were obtained from non-linear least-squares fitting of the concentration–response curves shown in Fig. 5. Data are averaged from 3–6 measurements in 2 experiments. The NMDA receptors were activated by 50 μm glutamate and 50 μm glycine.

[Ca2+]i80 ± 1321 ± 133.6524 ± 15790 ± 135.8
LDH efflux35 ± 1539 ± 140.9580 ± 12375 ± 137.7

NMDA receptor activation and attendant elevation of intracellular divalent ion concentration leads to cell death (Koh & Choi, 1987; Choi, 1992; Sattler et al. 1998; Sattler & Tymianski, 2001). We therefore evaluated whether the pH-dependent shift in potency of MK-801 isomers to inhibit Ca2+ influx correlated with their neuroprotective ability. Cell death assays that monitored the release of a stable intracellular enzyme, lactate dehydrogenase (LDH), were performed at 24 h following a 2 h application of 30 μm NMDA in the absence or presence of MK-801. Remarkably, the neuroprotective potency of (−)MK-801 was augmented 7.7-fold when pH was reduced to 6.9, whereas no change in potency of (+)MK-801 isomer was seen (Fig. 5C and Table 4). Moreover, while (+)MK-801 is 17-fold more potent than (−)MK-801 as a neuroprotectant at physiological pH, the pH potency boost for (−)MK-801 renders it almost equipotent to (+)MK-801 as a neuroprotectant at pH 6.9, a pH value that is relevant for the ischaemic penumbra.

Allosteric regulators of proton sensitivity regulate MK-801 potency

Although the pH changes described here will only modestly influence the ionization state of MK-801 (pKa 8.37; see Table 2), protons can have a wide range of actions at proteins, owing to ionization of a multitude of amino acid residues, protein-associated lipids and carbohydrates. In addition, pH changes can alter surface charge through changes in binding of cations to polar or ionized lipid headgroups. Therefore, it is unclear whether the pH-dependent changes we detect in MK-801 potency at NR1/NR2A receptors reflect the extracellular voltage-independent inhibitory site for protons (Traynelis & Cull-Candy, 1990, 1991), or other unrelated actions of protons. To test this idea, we exploited the mechanism of Zn2+ inhibition at NR2A-containing receptors. Zinc ions bind with high affinity to the amino terminal domain of the NR2A subunit. Occupancy of its high-affinity binding site by Zn2+ increases the potency for proton binding to their inhibitory site (Choi & Lipton, 1999; Low et al. 2000; Erreger & Traynelis 2005), which increases the level of tonic proton inhibition exerted at physiological pH. If the pH-dependent potency shift of MK-801 reflects actions through protonation of the receptor at the previously described proton sensor, we predicted that MK-801 potency at NR1/NR2A receptors subjected to saturating Zn2+ (1 μm) should also be altered. Specifically, we predicted that extracellular Zn2+ should induce a similar effect on potency of MK-801 isomers to that produced by an increase in proton concentration, since Zn2+ increases the proton affinity, which leads to greater occupancy of proton binding sites at physiological pH. At pH 7.6 in the presence of 1 μm extracellular Zn2+, which will induce inhibition comparable to that produced by a change in pH to 6.9 (Low et al. 2000), the potency of (−)MK-801 increased sevenfold at NR1/NR2A receptors (IC50 without Zn2+ 264 nm; IC50 with 1 μm Zn2+ 37 nm; n= 6–9). This change in (−)MK-801 IC50 is similar to that produced at pH 6.9. Again consistent with effects of reduced pH, (+)MK-801 showed a smaller threefold shift in potency in the presence of Zn2+ (IC50 without Zn2+ 11 nm; IC50 with 1 μm Zn2+ 4.2 nm; n= 6–9). Similar experiments with (+/−)ketamine were performed. A modest twofold increase in potency of (+/−)ketamine was observed in the presence of 1 μm Zn2+ (n= 4, data not shown), suggesting that for (+/−)ketamine most of the pH-dependent shift in potency is due to change in ionization state of (+/−)ketamine. These results strengthen the idea that the shift in potency of MK-801 isomers observed at lower pH is likely to involve the ability of MK-801 to sense the ionization state of the proton sensor.

Proton effects on MK-801 binding and unbinding

The proton-induced potency shifts for MK-801 that we have measured in Xenopus oocytes reflect changes in the association rate for MK-801, changes in the dissociation rate, changes in the behaviour of the MK-801-bound channel, or some combination thereof. Since NMDA channels can close with MK-801 still bound, MK-801 can become trapped in the ion pore following dissociation of agonist. Thus, the dissociation of MK-801 from the blocking site requires continuous agonist application in order for the channel to be in the open state, which allows the unbinding of MK-801 (Huettner & Bean, 1988; Halliwell et al. 1989; Parsons et al. 1993). In order to distinguish which of the hypothetical steps describing the actions of MK-801 were pH dependent, we studied the effect of increasing extracellular proton concentration (low pH) on association and dissociation kinetics of MK-801 at NR1/NR2A receptors (Fig. 6).

Figure 6.

Low pH accelerates macroscopically derived (−)MK–801 association rate with NR1/NR2A receptors
A, the time course of MK-801 (30–100 nm) block of whole-cell currents in HEK 293 cells expressing NR1/NR2A receptors in response to rapid application of 50 μm glutamate and 50 μm glycine is shown at different pH values. HEK 293 cells were held under voltage clamp at −60 mV. B, values for τON were determined by fitting the decay in current in the presence of MK-801 with an exponential function using a simplex algorithm (least-squares criteria): Response = Amplitude exp(−time/τ) + steady state. The reciprocal of τON was plotted against concentration of MK-801 to estimate macroscopically derived kON and kOFF by: 1/τON=kON[MK-801]+kOFF. Each point represents measurements from 3–5 cells.

Whole-cell patch clamp recordings were obtained from HEK 293 cells transiently transfected with NR1/NR2A. A rapid drug application system with open tip exchange times of 1–2 ms was used for these experiments; we have previously estimated that this system exchanges solution around an HEK 293 cell with a time constant (τ) of 7 ms (Mott et al. 2001). MK-801 was applied in the continuous presence of 50 μm glutamate and 50 μm glycine after the agonist-induced current had reached steady state (Fig. 6A). The onset of MK-801 inhibition could be approximated by a single exponential function, and individual current traces were fitted to estimate the τON for onset of MK-801 block (simplex algorithm, least-squares criteria). The reciprocal of τON was linearly related to the concentration of MK-801 (Fig. 6B). Thus, MK-801 association rate (kON) and dissociation rate (kOFF) could be estimated from the linear regression, assuming that MK-801 binds to a single site in a reversible fashion and that MK-801 binding and unbinding are rate-limiting steps. Block by both the isomers of MK-801 was reversed within 20–30 s in the presence of agonists. However, the τOFF rates showed cell-to-cell variability for the same conditions, and no clear trend was observed for dissociation time course in different pH conditions. Previous studies have found that recovery from MK-801 block is slow and may take several minutes (Huettner & Bean, 1988; MacDonald et al. 1991). However, the faster recovery rate observed in our study compared with the previous studies may reflect use of heterologously expressed NR1/NR2A receptors versus neurons, use of supramaximal concentration of full agonist favouring high open probability, and use of a lower concentration of MK-801 facilitating full removal of drug during the switch to wash solution in our piezo-driven theta-tube perfusion system. Figure 6B shows that the linear regressions for (+)MK-801 τON at pH 6.9 and 7.6 are superimposable, whereas there is a significant difference in the linear regressions of (−)MK-801 τON at the two pH conditions. The macroscopically-derived association rate, kON, of (−)MK-801 increased 10-fold with the change in pH from 0.36 × 106m−1 s−1at pH 7.6 to 3.6 × 106m−1 s−1at pH 6.9 (Table 5). In contrast, no change in the association rate was observed for (+)MK-801. We subsequently calculated the affinity constant, Kd, from these estimations of kON and kOFF derived from macroscopic current time course. For (−)MK-801, Kd decreased sixfold from 416 nm at pH 7.6 to 74 nm at pH 6.9. By contrast, the estimated Kd for (+)MK-801 (32 nm) was unaffected by pH (Table 5).

Table 5.  MK-801 rate constants determined from macroscopic response relaxation
Fitted rates(+)MK-801(−)MK-801
pH 7.6pH 6.9pH 7.6pH 6.9
  1. Values of τON were determined by fitting the current relaxation after stepping from 50 μm glutamate and 50 μm glycine into glutamate–glycine plus MK-801 to an exponential function (Fig. 6 legend) as shown in Fig. 6A using simplex algorithm (least-squares criteria). The reciprocal of τON was plotted against the concentration of MK-801 to estimate kON and kOFF by: 1/τON=kON[MK-801]+kOFF.

MK-801 kON (m−1s−1)3.9 × 1063.8 × 1060.36 × 1063.6 × 106
MK-801 kOFF(s−1)
MK-801 Kd (nm)333241674

Typically, manipulations that reduce open probability, such as acidic pH (Banke et al. 2005), should reduce the macroscopic (or apparent) association rate for channel blockers, since the reduced open probability provides less opportunity for the blocking molecule to access its binding sites. Paradoxically, however, it appears that reducing pH increases the apparent association rate of (−)MK-801 for NR1/NR2A receptors, which we interpret as the underlying basis for pH-dependent potency boost.

Microscopic analysis of proton effects on MK-801 binding

The calculations described in Table 5 provide a first approximation of macroscopic association and dissociation rates from whole-cell current waveforms, assuming a simple reversible interaction within the channel pore without provision for blocker trapping or modification of channel gating by MK-801. In order to examine in more detail the potential mechanisms underlying proton-induced acceleration of MK-801 association, we evaluated a simple kinetic scheme (Fig. 7) that has previously been proposed to account for the actions of trapping blockers (Blanpied et al. 1997; Dilmore & Johnson, 1998). The essence of this scheme is the ability of receptors to close and agonist to unbind while retaining MK-801 binding in the pore.

Figure 7.

Hypothetical scheme for MK-801 binding to NR1/NR2A receptors
A model was adapted from Blanpied et al. (1997) to explain the mechanism of trapping blockers such as MK-801. R, receptor; A, agonist; B, channel blocker; D, desensitized state.

In order to evaluate this scheme, we first sought to place as many constraints on the model as possible. As a first step towards doing this, we evaluated the effect of protons on NR1/NR2A channel properties to get an estimate of closing rates to be used for the model (Fig. 8). Single channel currents were recorded from HEK 293 cells transiently transfected with NR1/NR2A receptors in cell-attached patch configuration, a single channel recording configuration that should maintain channel function closest to that in the whole-cell configuration. The effect of pH on the open times of the channel was measured in unpaired recordings at different pH values (Fig. 8AC). A change from pH 8.0 to 6.7 reduced the mean open time of NR1/NR2A from 3.29 ± 0.36 to 1.62 ± 0.07 ms (n= 7–9; Fig. 8C and Table 6); there was no significant difference in single channel current amplitude (data not shown). Linear interpolation of these results leads to a predicted change in mean channel open time between pH 7.6 and 6.9 from 2.79 to 1.93 ms. These data allowed us to fix the channel closing rate in the absence of MK-801 to the reciprocal of mean open time, 358 (pH 7.6) and 518 s−1 (pH 6.9). The rates for glutamate binding and unbinding steps were adapted from Erreger et al. (2005b), and assumed to be the same for both MK-801-bound or MK-801-unbound receptor. Receptors were assumed to be saturated by glycine. Based on the previously published open probability (Po) of NR1/2A receptors (Erreger et al. 2005a) and inhibition of NR1/NR2A currents by protons in oocytes (Low et al. 2003), we estimated the maximal Po to be approximately 0.56 and 0.30 at pH 7.6 and 6.9, respectively. The opening rate (β) at the two pH values was fixed based on simulations of receptor activation in the absence of MK-801 so that predicted maximal Po for the receptor was 0.56 (pH 7.6) or 0.30 (pH 6.9).

We subsequently fitted four to seven raw traces of whole-cell recording in response to 30, 100 or 300 nm MK-801 (as shown in Fig. 6A), normalized to peak open probability (Po) values in each condition, with the model shown in Fig. 7, maintaining the above-mentioned constraints (Fig. 8D). Although solution exchange around cells will be slower than that at outside-out patches used to determine peak open probability (Erreger et al. 2005), the use of a maximally effective concentration of glutamate (50 μm) leads to similar rise times for both recordings, suggesting that peak open probability determined in patches is a reasonable estimate of peak open probability in whole-cell recordings. Table 7 summarizes the averaged fitted rate constants. The MK-801-bound channel closure (α′) and opening rates (β′), desensitization rates and the MK-801 association and dissociation rates were the free parameters during the least-squares minimization protocol. As seen in Table 7, the fitted kON for (−)MK-801 was increased eightfold at pH 6.9 compared with pH 7.6, whereas there was no significant shift in the kON for (+)MK-801 (P > 0.05, Student's unpaired t test). Fitted kOFF values were faster than those determined from relaxation analysis because the model in Fig. 7 incorporates other steps that can slow dissociation of MK-801. We used our fitted parameters from Table 7 to simulate macroscopic responses like those shown in Fig. 6. Analysis of the relaxation time constants for these simulated responses provided an additional check on the validity of our analysis. This model adequately predicted the pH-dependent changes in τON as well as a similar trend in potency boost for the two isomers. Simulated inhibition by (−)MK-801 was threefold more potent at pH 6.9 versus 7.6, whereas (+)MK-801 simulated potency was pH insensitive. From these experiments, we conclude that protonation of NR1/NR2A receptors leads to an increase in the association rate of (−)MK-801 and a higher affinity for (−)MK-801 compared with unprotonated receptors. The modelling also provided additional insight into the mechanism of action of trapping blockers. The rates controlling the gating equilibrium in the model were different in the absence (β/α) or the presence (β′/α′) of the blocker. The channel gate is, in fact, more likely to be closed in the presence of the blocker for both (+) and (−)MK-801 (Table 7). A similar mechanism has been proposed for amantadine (Blanpied et al. 2005).

Table 7.  MK-801 rate constants determined from non-linear fitting of trapped block mechanism to macroscopic response time course
Fitted parameters(+)MK-801(−)MK-801
pH 7.6pH 6.9pH 7.6pH 6.9
  1. The model shown in Fig. 7 was fitted to the NMDA receptor current response time course in response to 50 μm glutamate and 50 μm glycine. The average response from each cell was normalized to the open probability determined from Erreger et al. (2005a). Data were fitted using a simplex algorithm to alter model rate constants at each iteration (least-squares criteria). Each rate is the mean of the model fitted to response waveforms from 3–5 cells. The following parameters were fixed: glutamate association, k+= k+′= 1.04 × 107m−1 s−1 and glutamate dissociation, k=k′= 73 s−1 (Erreger et al. 2005a); for pH 7.6, open probability (Po) = 0.56, α= 359 s−1, β= 488 s−1; and for pH 6.9, Po= 0.30, α= 518 s−1, β= 223 s−1. The Po values were estimated from a peak open probability of 0.5 for NR1/2A receptors (Erreger et al. 2005a) and proton IC50 of 119 nm for NR1/2A receptors (Low et al. 2003). The value of α was the reciprocal of the measured open time, and β was chosen to give the desired Po. The degree of desensitization varied between individual traces and was independent of pH or the response to MK-801. Thus, the onset and recovery rates for desensitization were fitted for each cell, and on average ranged between 2.8 and 8.7 s−1 and between 0.5 and 2.9 s−1, respectively, for all conditions.

MK-801 kON (m−1 s−1)3.2 × 1075.6 × 1070.6 × 1074.9 × 107
MK-801 kOFF (s−1)
MK-801 Kd (nm)21512864864.7
β′ (s−1)220250250180
α′ (s−1)56032004900140

If this interpretation of a pH-dependent association rate for (−)MK-801 is correct, we should be able to observe changes in the concentration dependence of channel open duration in the presence of (−)MK-801. This is because the accelerated association rate should increase the likelihood that (−)MK-801 will bind during an opening, and thereby decrease the mean open duration since open periods will be prematurely terminated by MK-801 binding and subsequent channel block (Fig. 9A). To test this prediction at the single channel level, we evaluated the mean open duration of NR1/NR2A receptors in excised membrane patches during application of varying concentrations of (−)MK-801 at −60 mV at pH 6.9 and 7.6. This method has been used previously to estimate the blocking rate of (+)MK-801 (Jahr, 1992). Change in pH did not change the conductance of the channel, which at pH 7.6 was 66.8 ± 1.7 pS (n= 5) and at pH 6.9 was 64.5 ± 1.0 pS (n= 5). Figure 9B shows that the mean open duration in the presence of 30 μm (−)MK-801 was significantly lower at pH 6.9 compared with pH 7.6 (P < 0.05, Student's unpaired t test). In addition, the least-squares fitting of the reciprocal of mean open duration showed that the blocking rate was fivefold higher at pH 6.9 (8.6 × 106m−1 s−1) compared to pH 7.6 (1.6 × 106m−1 s−1). Thus, similar to the increase in the blocking rate at pH 6.9 observed in whole-cell experiments and kinetic modelling, single channel data also confirm that the blocking rate of (−)MK-801 increases at lower pH.

Figure 9.

Single channel analysis suggests that (−)MK-801 has a faster association rate at protonated NR1/NR2A receptors
A, unitary currents recorded from outside-out patches excised from HEK 293 cells expressing NR1/NR2A receptors are shown (holding potential was −60 mV). Traces were filtered at 2 kHz for the purpose of illustration. Dotted line indicates closed state (C); open state is indicated by O. (−)MK-801 at 10 μm reduces the arithmetic mean channel open duration at pH 6.9. B, mean open duration in the presence of 30 μm (−)MK-801 at two different pH conditions is expressed as a percentage of mean open duration of control. Reduction in mean open duration at pH 6.9 is significantly different from at pH 7.6 (*P < 0.05 Student's unpaired t test, n= 4–5 for each condition). C, reciprocal of mean open time is plotted as a function of (−)MK-801 concentration. Each point is the mean ±s.e.m. determined from 4–7 patches. The slope of this relationship was determined by linear regression and represents the blocking rate of (−)MK-801. Blocking rate at pH 7.6 is 1.6 × 106m−1 s−1 and at pH 6.9 is 8.6 × 106m−1 s−1.


In this study, we provide the first comprehensive evaluation of both MK-801 stereoisomers across the full family of NMDA receptor subtypes. The combination of in vitro and in vivo approaches suggests possible NR2 subunit-specific mechanisms for the behavioural effects of MK-801. The modest stereoselective potency of (+)MK-801 over (−)MK-801 for anticonvulsant activity (Fig. 2A) is consistent with the involvement of NR2A,B,D-containing receptors in seizure initiation by flurothyl, since these subunits all show MK-801 stereoselectivity that is equal to or greater than stereoselectivity for anticonvulsant activity (Table 2). However, the larger relative potency differences for (+)MK-801 over (−)MK-801 for doubling of locomotor activity (Fig. 2B) and PCP-like discriminative stimulus effects (Fig. 2C) suggest that these effects may involve receptors containing the NR2A subunit, which shows a larger difference in stereoselective potency. This is consistent with the modest effect of NR2B antagonists on locomotor activity (Higgins et al. 2003; Altas Y, Geballe M, Gruszecka-Kowalik E, Liotta D, Washburn M, Myers S, Lyuboslavsky P, Le P, French A, Irier H, Choi W-B, Easterling K, MacNamara JO, Dingledine R, Traynelis SF & Snyder JP, unpublished observations) and PCP-discrimination (Wiley et al. 1997).

Additionally, we describe the potency of a wide range of structurally dissimilar channel blockers at all NR2 subunits. A number of single channel and macroscopic properties have suggested that NR2A- and NR2B-containing receptors are functionally distinct from NR2C- and NR2D-containing receptors (Monaghan et al. 1998; Erreger et al. 2004; Clarke & Johnson, 2006). Our data go beyond this subdivision to suggest that structural and functional differences exist within the pores of all subtypes, including NR2A and NR2B. The pH sensitivity of (−)MK-801 potency reflects fundamental changes in receptor–drug interaction, not changes to the ionization of the drug (pKa 8.37). Indeed, we provide data suggesting that the effects of protons on (−)MK-801 but not (+)MK-801 potency reflect actions at a well-described extracellular inhibitory proton site on the receptor. Binding of Zn2+ to the amino terminal domain of NR2A enhances the pH sensitivity of this site, and therefore its protonation at physiological pH. As predicted, Zn2+ binding and enhancement of protonation have the same differential effect on MK-801 stereoisomers as lowering pH in the absence of Zn2+. In addition, the potency of (−)MK-801 was strikingly similar to that of (+)MK-801 at acidic pH in the neuroprotective assay. Our data provide a straightforward mechanistic explanation for this increase in (−)MK-801 potency, since the apparent macroscopic MK-801 association rate increases at low pH. Further analysis of macroscopic waveforms as well as single channel currents confirms that protons can increase the association rate for (−)MK-801 with its intrapore binding site. These data further suggest that tight coupling exists between an ionizable proton sensor and gating elements, such that voltage-dependent binding within the channel pore by trapping blockers can sense the protonation state of the receptor either within the pore at their binding site or en route to the binding site.

How does protonation influence (−)MK-801 binding?

We have employed a model of NMDA receptor activation and MK-801 block adapted from Blanpied et al. (1997) and with some parameters imported from the model of Erreger et al. (2005b). Since some rates were derived from a model (Erreger et al. 2005b) with different connectivity of states, these rates should be considered imprecise parameters. However, the present model accurately predicts the amplitude and the time course of the current response in this study (Fig. 8D); both the present model and the model of Erreger et al. (2005b) predict similar values for glutamate EC50 and peak open probability (data not shown). Additionally, because the fits here were done on MK-801 blockade of current responses to maximally effective concentrations of glutamate after reaching steady state, the conclusions drawn from modelling results are largely insensitive to parameters such as the agonist binding/unbinding or desensitization rates.

The NMDA receptor activation process is highly dependent on the concentration of extracellular protons (e.g. Traynelis & Cull-Candy, 1990, 1991; Banke et al. 2005). Single channel data show that NR1/NR2A receptors have reduced mean channel open dwell time and reduced open probability (NPo) at low pH (Fig. 8 and Table 6). Previous data have shown that NR1/NR2B receptors have reduced open probability but no marked change in mean channel open dwell time at low pH (Banke et al. 2005). This suggests that the mechanism of proton inhibition of NR1/NR2A differs slightly from NR1/NR2B receptors. Our working hypothesis is that the pH dependence of channel open probability for NMDA receptors reflects specific hydrogen bonds made by ionizable amino acid residues within the receptor protein. In this model, protonated residues could reduce the likelihood that a channel can be activated by making new hydrogen bonds (e.g. see proton sensitivity of K channels; Schulte et al. 1999; Yang et al. 2000; Xu et al. 2000; Sackin et al. 2005). It follows that the activation energy for rearrangements necessary for channel opening could be increased by the energy of several hydrogen bonds. The energy required to change the equilibrium constant for channel opening 100-fold is 2.73 kcal mol−1, well within the range of energy supplied by one or a few hydrogen bonds. The location of the proton sensor that mediates this effect is likely to be close to (or integrated into) the part of the protein that forms the channel activation gate (Low et al. 2003). This close association of proton sensitivity and the inner portions of the pore that open and close raises the possibility, at least for NR1/NR2A, that molecules block the channel deep in the pore (like MK-801) in a manner that can either sense the protonation of the receptor or the biophysical results of receptor protonation (e.g. change in intrapore dielectric constant).

Several examples of pH-dependent ion channel antagonists exist. Perhaps the best-known example of a drug class whose actions are influenced by the tissue pH are the local anaesthetics (Hille, 1977; Schwarz et al. 1977). In contrast to our findings, the relative protonation state of the sodium channel is not sensed by the local anaesthetics; rather, it is the ionization state of local anaesthetic that is responsible for the increase in activity at alkaline pH. A similar example of a pH-dependent drug ionization involves the competitive NMDA receptor antagonists that contain a phosphono group, which possess a pKa in the physiological range. The potency of these compounds is decreased as extracellular pH becomes acidic (Benveniste & Mayer, 1992; Myers et al. 2006). Interestingly, the potency of NR2B-selective compounds, such as ifenprodil, is also pH dependent (Pahk & Williams, 1997; Mott et al. 1998). However, their pH sensitivity appears to reflect a reciprocal interaction between the ability of ifenprodil to enhance pH inhibition and the ability of acidic extracellular pH to enhance ifenprodil potency. MacDonald et al. (1991) and Yamakura et al. (2000) have described pH-sensitive potency of channel block of NMDA receptors by meperidine and ketamine, which may have clinically important ramifications. Our data are consistent with their conclusions, and suggest that at least part of the pH sensitivity of ketamine is dependent on the ionization state of ketamine (MacDonald et al. 1991); the same may be true for norketamine, since its pKa value is similar to ketamine and is in the range of physiological pH (Table 2).

Open channel blockers like MK-801 have been suggested to bind deep in the pore and close to the region of the NMDA receptor that controls the gate. We have found that (−)MK-801 competes for [3H](+)MK-801 (data not shown), suggesting that at least some of the structural determinants for binding are similar for the two stereoisomers. The finding that (−)MK-801 has a higher affinity and faster association rate for protonated NR1/NR2A receptors suggests that the protonated receptors undergo some structural changes that increase the ability of (−)MK-801 to reach its binding pocket deep within the pore and/or orientate itself in its pocket correctly. This finding is intriguing, since protonation will decrease the fraction of time that receptors are open, and therefore should decrease the apparent association rate by reducing opportunity for MK-801 binding. Several speculative ideas that might provide mechanistic interpretation can be offered. For example, perhaps protonation of the receptor leads to occlusion of the ion permeation pathway while still allowing partial dilatation of the outer vestibule. This would make the closed receptor more accessible to the blocker, and perhaps facilitate block with channel opening. Another potential effect of protons might be on a hypothesized permeant ion binding site, which could interact in some way with the MK-801 binding site to accelerate its apparent association (Antonov et al. 1998). This permeant ion site prevents the entry of some organic blockers as well as Mg2+ (Antonov & Johnson, 1999; Zhu & Auerbach, 2001) and could also be in the path of MK-801 entry. If protonation of NMDA receptor reduces the binding of a permeant ion to its site, it would lead to an increase in blocking rate, a decrease in IC50 and decrease in the voltage dependence of the channel blocker. We see similar effects of protonation on (−)MK-801. It is important to bear in mind that the decrease in mean channel open time by (−)MK-801 at pH 6.9 suggests that even at acidic pH (−)MK-801 still behaves as an open channel blocker, reaching its binding site within an open channel faster at low pH, which consequently reduces the duration during which a channel remains open. Furthermore, our data suggest that protons do not affect the chord conductance of NR1/NR2A receptors, both in cell-attached patches and in outside-out patches, suggesting that changes in ion conduction pathway that alter ion flux are not responsible for the pH-dependent potency differences of (−)MK-801. Clearly, more work remains to be done before the stereo-selective and subunit-selective pH-dependent acceleration of MK-801 binding can be understood.

Subunit specificity of pore blocking molecules

An important finding of this study is the variation of the potency of a structurally diverse group of channel blockers for different NR2-containing receptors. For instance, the NR2C subunit at pH 7.6 showed lower IC50 (> sevenfold higher potency) compared with NR2A for (−)MK-801, (+)ketamine, (+/−)norketamine, dextromethorphan, levomethorphan, dextrorphan and memantine (see Table 2). This is despite the possibility that NR1/NR2C receptors may have a lower open probability (S.M. Dravid & S.F. Traynelis, unpublished data) and briefer open duration than NR1/NR2A receptors (Stern et al. 1992). The potency of (−)MK-801 was markedly (> 10-fold) different for NR2A- and NR2B-containing receptors, further suggesting that intrapore structural differences exist even between these subtypes; (+)ketamine and (+/−)norketamine also showed > sixfold lower IC50 at NR2B. Despite the multiple potential mechanisms for subunit selectivity of channel blockers, three broad conclusions can be drawn from these results. First, the pore-forming elements for the different NR2-containing receptors are probably distinct enough to account for the variation in potency of channel blockers. Second, it may be possible to develop subunit-selective NMDA channel blockers that are better tolerated than non-selective channel blockers (Monaghan & Larsen, 1997). Third, the finding of pH-dependent, stereo-selective and subunit-specific actions of channel blockers suggests that previously unappreciated features of channel block exist that could be exploited to develop NMDA blockers with intriguing properties.



The research was supported by NIDA grant DA-07218 (T.F.M.), NIDA R01-01442 (K.N., R.B.), NINDS R01-NS36777 (S.F.T.), MCV grant support (K.N., R.B.) and Epilepsy foundation (S.M.D.). We thank Drs Shigetada Nakanishi, Stephen Heinemann and Tom Hughes for providing cDNAs. The technical assistance of Li Hua is greatly appreciated. We thank Dr Stephen Holtzman for sharing reagents and Dr Jon Johnson for critical comments on the manuscript. We thank Serdar Kurtkaya and Dr James P. Snyder for calculating pKa. We thank Dr Kris Bough for advice on the flurothyl seizure model.

Conflict of interest disclosure

Stephen Traynelis is a co-inventor on a pending patent involving the pH sensitivity of NMDA receptor antagonists.