|Home | About | Journals | Submit | Contact Us | Français|
NMDA receptors are composed of two GluN1 (N1) and two GluN2 (N2) subunits. Constituent N2 subunits control the pharmacological and kinetic characteristics of the receptor. NMDA receptors in hippocampal or cortical neurons are often thought of as diheteromeric, i.e., containing only one type of N2 subunit. However, triheteromeric receptors with more than one type of N2 subunit also have been reported and the relative contribution of di- and triheteromeric NMDA receptors at synapses has been difficult to assess. Because wild-type hippocampal principal neurons express N1, N2A and N2B, we used cultured hippocampal principal neurons from N2A and N2B-knockout mice as templates for diheteromeric synaptic receptors. Summation of N1/N2B and N1/N2A excitatory postsynaptic currents could not account for the deactivation kinetics of wild-type excitatory postsynaptic currents (EPSCs) however. To make a quantitative estimate of NMDA receptor subtypes at wild-type synapses, we used the deactivation kinetics, as well as the effects of the competitive antagonist NVP-AAM077. Our results indicate that three types of NMDA receptors contribute to the wild-type EPSC, with at least two-thirds being triheteromeric receptors. Functional isolation of synaptic triheteromeric receptors revealed deactivation kinetics and pharmacology distinct from either diheteromeric receptor subtype. Because of differences in open probability, synaptic triheteromeric receptors outnumbered N1/N2A receptors by 5.8 to 1 and N1/N2B receptors by 3.2 to 1. Our results suggest that triheteromeric NMDA receptors must be either preferentially assembled or preferentially localized at synapses.
NMDA receptors are tetramers of N1 and N2 subunits, with 2 of each subunit type per receptor. One gene with several splice variants encodes N1 and several genes encode N2 subunits (N2A - N2D). Each type of N2 subunit confers a distinct kinetic and pharmacological profile to the receptor. Expression of N2 subunits is developmentally and anatomically regulated and results in extensive heterogeneity in NMDA receptor properties. Studying the molecular composition of synaptic NMDA receptors is challenging however because neurons can make di- and triheteromeric receptors (Sheng et al., 1994). The roles of N2A or N2B subunits in synaptic plasticity and neurodevelopment have been studied using drugs that can distinguish between diheteromeric receptors. However, these drugs poorly discriminate between diheteromeric and triheteromeric receptors (Neyton and Paoletti, 2006). Thus, the possible contribution of triheteromeric receptors to the synaptic response has often been ignored. Yet, recent evidence in hippocampal neurons where N2A and N2B are highly expressed (Gray et al., 2011; Rauner and Köhr, 2011) suggests that triheteromeric receptors contribute to excitatory postsynaptic currents (EPSCs).
Recombinant diheteromeric receptor subtypes differ in their pharmacological sensitivity to modulatory ligands like zinc and ifenprodil, as determined by the constituent N2 receptor subunit. Receptor properties like deactivation kinetics and open probability can also be indicative of the diheteromeric receptor subtype (Vicini et al., 1998; Chen et al., 1999). For example, the decay of NMDA-receptor mediated EPSCs can differ several fold depending on the diheteromeric receptor subtype (Tovar and Westbrook, 2012; Logan et al., 2007; Cathala et al., 2000). In this context it becomes critical to know how synaptic NMDA receptor properties reflect the expression of more than one N2 subunit type, as is typical of mammalian central nervous system neurons. The pharmacological characteristics of recombinant triheteromeric receptors have been studied in heterologous cells (Hatton and Paoletti, 2005), but comparatively little is known about the behavior of triheteromeric NMDA receptors at synapses, particularly during phasic agonist presentation characteristic of neurotransmitter release.
We developed a method to quantify the contributions of NMDA receptor subtypes to EPSCs in principal excitatory neurons in the mouse hippocampus that express N1, N2A and N2B (Monyer et al., 1994). We used neurons from N2A or N2B knockout mice to define the properties of diheteromeric N1/N2B (B-type) or N1/N2A (A-type) receptors, respectively. The properties of wild-type EPSCs indicated that A-type or B-type receptors were present, but were not sufficient to account for the entire wild-type synaptic response. We conclude that three types of NMDA receptors contribute to EPSCs in wild-type neurons with the predominant contribution from triheteromeric receptors containing N2A and N2B subunits, along with N1.
All recordings were done on mouse hippocampal neurons, cultured on glial micro-islands, as described in Tovar and Westbrook (2012). Neurons from wild-type, N2A knockout (Sakimura et al., 1995), N2B knockout (Kutsuwada et al., 1996) and N2A/N2B double knockout mice were taken at P0 or P1, owing to early lethality of N2B knockout mice (Kutsuwada et al., 1996). All mice were in a C57BL/6 genetic background. Male mouse pups were used for all cell cultures. Mouse pups were decapitated, the brains were removed from the skulls and hippocampi were dissected from the brain. Glial micro-islands were prepared prior to culturing neurons, by plating hippocampal cells at 125,000 cells/35 mm dish onto coverslips that had been sprayed with a collagen and poly-D-lysine mixture (Tovar and Westbrook, 2012). After one week, glial micro-islands were treated with 200 μM glutamate to kill any remaining neurons and then neurons were added at 25,000 cells/35 mm dish. All glia were from wild-type mice. Cultures were grown in a tissue culture incubator (37°C, 5% CO2), in a medium made with Minimum Essential Media with 2 mM Glutamax (Invitrogen), 5% heat-inactivated fetal calf serum (Lonza), 1 ml/L of Mito+ Serum Extender (BD Bioscience) and supplemented with glucose to an added concentration of 21 mM. Genotypes of N2A and N2B knockout mice and double knockout (lacking N2A and N2B) mice were verified with polymerase chain reaction, using previously described primers (Tovar et al., 2000; Thomas et al., 2005). Double knockout mice were generated by crossing adults that were heterozygous at both alleles. All animals were treated in accord with OHSU and NIH policies on animal care and use.
Because contaminating concentrations of zinc prolongs the NMDA receptor-mediated EPSC deactivation (Tovar and Westbrook, 2012), we used tricine in all extracellular solutions to buffer zinc. During recordings, neurons were continuously perfused through gravity-fed flow pipes. Flow pipes were placed within 100 microns of the recorded neuron. The extracellular solution contained (in mM) 158 NaCl; 10 tricine; 2.4 KCl; 10 HEPES; 10 D-glucose; 1.3 CaCl2; 0.02 – 0.05 glycine and 2.5 μM NBQX. High purity salts and HPLC water were used for external solutions. Pipette solution contained (in mM): 140 potassium gluconate; 4 CaCl2; 8 NaCl; 2 MgCl2; 10 EGTA; 2 Na2ATP and 0.2 Na2GTP. The pH was adjusted to 7.4 with KOH. The osmolality of external and internal solutions was adjusted to 320 mosmol. For zinc application experiments, the free zinc concentration was calculated by the method in Fayyazuddin et al., (2000). Strychnine, NMDA, glycine, NVP-AAM077 and salts were obtained from Sigma-Aldrich or Fluka; NBQX, ifenprodil and (+)-MK-801 were obtained from Ascent Scientific.
Whole-cell voltage clamp recordings were done on neurons in micro-island culture, and used either single neurons (that made autapses) or pairs of neurons (that made reciprocal synaptic connections). To evoke synaptic currents, the command voltage was briefly stepped to +20 mV (for 0.5 ms) at the soma, producing an unclamped action potential, followed by the EPSC. Axopatch 1C amplifiers and AxoGraph X acquisition software were used for data acquisition. The series resistance was always less than 8 mΩ and was compensated by greater than 80% by the amplifier circuitry. Data were low-pass filtered at 5 kHz and acquired at 10 kHz. Postsynaptic currents were evoked at low frequency (0.1 – 0.2 Hz). All recordings were done at room temperature. All recordings were done on neurons at 10–16 days in vitro (DIV). For this work, we only analyzed EPSCs that were on excitatory neurons to reduce variability in postsynaptic neuron phenotype.
To estimate the NMDA receptor channel latency to first opening, we used high concentrations of MK-801 (25 μM) to block channels immediately upon opening. This is based on well-tested assumptions (Jahr, 1992), specifically that (1) once channels open, the vast majority are blocked very quickly, leading to (2) very few channels opening more than once because (3) once a channel is blocked, MK-801 does not unbind during the EPSC (Huettner and Bean, 1988). The time course of charge transfer in 25 μM MK-801 thus provides an approximation of the latency to first opening of NMDA receptor channels following neurotransmitter binding. Furthermore, in MK-801, the ratio of the charge at the time of the control EPSC peak to the total charge is the probability that channels have opened by the time of the control EPSC peak (Po*).
We used the competitive antagonist NVP-AAM077 (Auberson et al., 2002) to characterize A- and B-type receptors and to separate NMDA receptor subtypes. Even though NVP-AAM077 is a competitive antagonist, and thus sensitive to agonist equilibrium affinity differences between receptor subtypes, using NVP to differentiate between receptor subtypes was possible because (1) cells were equilibrated in NVP-AAM077, (2) the neurotransmitter glutamate is present very briefly (Clements et al., 1992) and (3) the glutamate binding rate of A-type and B-type receptors is comparable (Chen et al., 2001). The 10-fold difference in the NVP IC50 we measured between A-type and B-type receptor EPSCs confirmed this approach.
To estimate the fractional contributions of A-type, B-type and AB-type receptors to the wild-type EPSC, we fit the deactivation kinetics of wild-type EPSCs at each of 5 NVP-AAM077 concentrations with a sum of three double-exponential functions using nonlinear least-squares curve fitting in Matlab (Mathworks). Six to ten random initial conditions were used for each fit to assure that the global minimum had been achieved. The fractional contribution of each double-exponential to the EPSC, but not the kinetics of the individual double-exponential functions, were allowed to change with NVP concentration. One double exponential function had kinetic parameters and dose-response properties constrained to be within 1 or 2 standard deviations of the data from A-type EPSCs. A second double exponential function was similarly constrained to match B-type EPSCs. The third double exponential had unconstrained kinetics and NVP dose response characteristics. This third component provided an estimate of the fractional contribution to WT EPSCs, the triheterometeric receptor kinetics and the NVP dose-response profile. Values generated from fits with ± 1 S.D. and ± 2 S.D. constraints were very similar, so only ± 1 S.D. results are shown.
All other data was analyzed with Axograph X or Igor software. Measurements were done on at least 8 EPSCs that had been averaged. The exception was for traces in MK-801 which, owing to the use-dependence of this drug, required that at most 3 EPSCs be averaged.
EPSC deactivations were fit using the function,
where If and Is are the relative amplitudes of the fast and slow components, τf and τs are the fast and slow deactivation time constants and c is a constant that represents any residual current that was not fit (typically much less than 0.5% of the total amplitude). Fits to wild-type EPSC deactivations were started from the peak and extended to 3.5 seconds after the peak, typically allowing for at least 8 time constants of the larger decay component. Fits of the deactivations from mutant neuron EPSCs were done using shorter (1.5 seconds for N2B knockout neurons) or longer (4.5 seconds for N2A knockout neurons) time windows.
Weighted time constants were calculated as follows,
where τw is the weighted time constant, If and Is are the relative amplitudes of the fast and slow components and τf and τs are the fast and slow time constants. Dose-inhibition curves were fit with,
where IEPSC is the peak EPSC amplitude, IC50 is the concentration at which half the current is inhibited, [ligand] is the ligand concentration and nH is the Hill coefficient.
To enrich the wild-type EPSC with triheteromeric receptors, we first used NVP/MK-801 to progressively block the EPSC. This was designed to block all contribution from B-type receptors (see Figure 6). Following NVP/MK-801 treatment, we assumed only A-type and AB-type receptors should contribute to the residual EPSC. This assumption is based on the Po of B-type receptors (0.24), the fraction of receptors not blocked by 100 nM NVP (0.5) and an estimate of presynaptic release probability (0.2). At the concentration of (+)-MK-801 we used, any open NMDA receptor channel will become blocked. Thus, after 50 episodes in (+)-MK-801, the EPSC contribution from B-type receptors should be eliminated.
To estimated the contribution of A-type receptors in the enriched triheteromeric receptor fraction (in 50 nM NVP) we used the following equation:
where IEPSC is the EPSC amplitude, in this case 30.0 ± 1.4 % of control (n = 14) in 50 nM NVP-AAM077, [NVP] is the NVP-AAM077 concentration, IC50A and IC50AB are the IC50 values for NVP-AAM077 on A-type EPSCs and AB-type EPSCs, and A is the fractional contribution from A-type receptors to the EPSC. We estimated a mean IC50AB from 5-point dose-inhibition curves (30.0 ± 1.4 nM; n = 10), which in this case represents an upper limit of the NVP-AAM077 IC50AB. This fractional contribution (A), coupled with the known reduction of A-type receptors by 50 nM NVP-AAM077 (83%) and the reduction of the EPSC following treatment with NVP-AAM077 and (+)-MK-801 gives the maximal contaminating amount of A-type EPSC in the AB-type enriched EPSC.
The fractional contribution of each NMDA receptor subtype to the wild-type EPSC and the probability of a channel having opened by the time of the EPSC peak (Po*) were used to estimate the relative amounts of NMDA receptor subtypes expressed at synapses, using this relationship for any receptor type n,
where Rn is the fraction of receptor type n at synapses, F is the fractional contribution of that receptor type to the EPSC and P is the Po* of that receptor type.
For all 2-axis plots, the y-axis (ordinate) label is placed above the plot rather than to the left of the axis. Statistical significance was determined using Student’s unpaired or paired t-test, as appropriate. The threshold for statistical significance was set at 0.05. The IC50s and Hill coefficient data are reported as means ± standard deviations. All other data are reported as mean ± standard error of the mean, unless otherwise specified.
To examine the contribution of di- and triheteromeric receptors to wild-type EPSCs we used cultured hippocampal neurons from mice that lacked expression of N2A or N2B due to homologous recombination (Sakimura et al., 1995; Kutsuwada et al., 1996). Our experimental approach assumes that NMDA receptor-mediated EPSCs in knockout neurons result from a homogenous diheteromeric receptor population. To validate this approach, we examined the kinetic and pharmacological characteristics of EPSCs in excitatory neurons. EPSC deactivations from N2A and N2B knockout neurons were fit with the sum of two exponentials (τfast and τslow) and an average weighted time constant (τw) was calculated to allow easy comparison of kinetic parameters. As shown in Figures 1A and 1B, the EPSC deactivations from N2A and N2B knockout neurons differed by more than an order of magnitude (N2B knockout: 22.7 ± 2.2 ms; N2A knockout: 314.9 ± 73.3; n = 35 for each). The τw values of EPSC deactivations in the knockout neurons match the deactivation in response to brief application of glutamate (1 mM for 1 ms) at recombinant N1/N2A and N1/N2B receptors, respectively (Vicini et al., 1998). For brevity, we will refer to N1/N2A and N1/N2B diheteromeric NMDA receptors as A-type and B-type receptors, respectively, and to triheteromeric NMDA receptors containing N2A and N2B as AB-type receptors.
Zinc and ifenprodil are canonical ligands that have been used to distinguish between diheteromeric receptor subtypes (Paoletti and Neyton, 2007). Zinc (100 nM free; see Methods) and ifenprodil (3 μM) reduced the EPSC peak from N2B and N2A knockout neurons (Figure 1C and 1D) to 44.4 ± 1.4 (n = 9) and 18.2 ± 1.5 % of control (n = 16), respectively, comparable to values reported for A-type or B-type recombinant receptors (Rachline et al., 2005; Perin-Dureau at el., 2002). Thus the kinetic and pharmacological properties of synaptic receptors from N2A or N2B knockout neurons are consistent with B-type and A-type NMDA receptors, respectively.
To investigate whether excitatory hippocampal neurons expressed other NMDA receptor subunits, we crossed the N2A and N2B knockout mouse lines to generate double knockout (DKO) mice that lacked N2A and N2B. Neurons from these mice showed no NMDA receptor-mediated EPSCs (0/18), but had normal AMPA receptor-mediated EPSCs (Figure 1E). Additionally, most DKO neurons (38/44) lacked any current in response to exogenously applied NMDA (100 μM, Figure 1F). In the 6 DKO neurons with NMDA-evoked currents, the mean peak amplitude of this current (Figure 1G) was less than 0.5% (0.039 ± 0.013 nA) of the mean wild-type controls (8.2 ± 0.6 nA, n = 18). We were able to determine that 2 of these 6 were GABAergic interneurons; however we did not determine the neurotransmitter phenotype of the other 4 neurons. Our data demonstrate that genetic deletion of both N2A and N2B results in excitatory hippocampal neurons with no NMDA receptor-mediated EPSCs, and no NMDA-evoked currents.
In the absence of N2 subunits, NMDA receptors composed of N1 and N3 can form but these receptors are gated by glycine exclusively rather than glutamate and glycine (Chatterton et al., 2002). Glycine application (100 μM; 500 ms) in the presence of strychnine (2 – 10 μM, Vhold = −70 mV) resulted in small inward currents in DKO neurons (−4.6 ± 1.0 pA; n = 19). However, when the holding potential was shifted to −35 mV, glycine application produced small outward currents (32.4 ± 9.5 pA; n = 4), consistent with a glycine receptor-mediated chloride conductance. These results confirm that the predominant NMDA receptor subunits expressed in excitatory hippocampal neurons are N1, N2A and N2B and validate our use of neurons from N2A and N2B knockout mice as models for diheteromeric B-type and A-type NMDA receptors.
To estimate the contribution of A- and B-type receptors to EPSCs in wild-type neurons, we measured the NMDA receptor-mediated EPSC deactivations from N2A and N2B knockout neurons. The EPSC deactivations from wild-type, N2A and N2B knockout neurons were well fit with the sum of two exponentials. As expected for A-type receptors, EPSC deactivations from N2B knockout neurons were much faster than B-type EPSC deactivations from N2A knockout neurons. Wild-type EPSCs had intermediate deactivation kinetics (Figure 2A and Table 1). The behavior of wild-type EPSCs could result from the combination of two diheteromeric receptor subtypes or it could include a contribution from AB-type receptors. In A-type receptor EPSCs (from N2B knockout neurons) τslow contributes minimally to the deactivation because of the small value of Islow. Therefore the wild-type EPSC amplitude at long latency after the peak (>1000 ms) will lack any contribution from A-type receptors. The amplitude of the wild-type EPSC at this long latency then provides an estimate of the maximal contribution from B-type NMDA receptors to the wild-type EPSC deactivation, if A-and B-type receptors are the only NMDA receptor subtypes present.
To make this estimate, we peak-scaled wild-type EPSCs with a two exponential function made from the mean B-type EPSC deactivation parameters (from N2A knockout neurons) and measured the ratio of the wild-type current amplitude to that of B-type at 1000 ms after the EPSC peak (greater than 8 times τslow for A-type receptors; Figure 2B). The amplitude of the wild-type EPSC measured in the time window (200 ms) corresponded to a maximal contribution from B-type receptors of 23.9 ± 2.3% (n = 51; Figure 2C). This estimate assumes there is no long latency contribution from triheteromeric receptors at 1000 ms after the peak. To determine whether linear summation of A-type and B-type receptors was sufficient to approximate the deactivation of wild-type EPSCs, we added the mean waveforms from A-type and B-type EPSCs, with our estimate of the maximal contribution to the EPSC from B-type receptors ± 1 SD (23.9 ± 16.6%) added to a balance from A-type receptors (76.1 ± 16.6%). The summed waveform did not match the wild-type deactivation waveform and was not well fit using a two exponential function constrained by the mean wild-type deactivation time constants (red dashed line; Figure 2D). Thus, estimates based on combinations of diheteromeric EPSC waveforms fail to recapitulate the wild-type EPSC deactivation.
Our kinetic analysis suggests the presence of a third type of synaptic NMDA receptor. We used a pharmacological approach to address this possibility. An ideal subtype-specific competitive antagonist should simply prevent agonist binding and otherwise not affect channel gating. Although zinc or ifenprodil differ in their affinity for different NMDA receptor subtypes by more than 2 orders of magnitude (Neyton and Paoletti, 2006), these modulatory ligands prolong NMDA receptor-mediated EPSC deactivations (Tovar and Westbrook, 2012), making them unsuitable for our analysis strategy. We used the competitive antagonist NVP-AAM077 (NVP) because A-type and B-type receptors differ in their IC50s for NVP by an order of magnitude (Neyton and Paoletti, 2006). As expected, A-type receptor EPSCs were more sensitive to NVP by a factor of 10 than B-type receptor EPSCs (A-type: IC50 = 10.01 ± 0.82 nM, nH = 1.05 ± 0.06; B-type: IC50 =119.41 ± 8.34 nM; nH = 1.14). Wild-type neuron EPSCs had an intermediate sensitivity to NVP (IC50 = 37.4 ± 6.2 nM; nH = 1.00 ± 0.10). The single-site isotherm fit to the wild-type dose-inhibition curve (in red) was much better than a two component isotherm using the IC50s for A-type and B-type receptor (in solid black; Figure 3A).
If wild-type EPSCs result solely from mixtures of A-type and B-type NMDA receptors, then the deactivation kinetics should become progressively slower with increasing NVP concentrations because low NVP concentrations will preferentially block the faster A-type receptors. Thus at high NVP concentrations (≥ 100 nM) the deactivation of the unblocked EPSC should be composed almost exclusively of B-type receptors. The weighted EPSC deactivation time constants of wild-type EPSCs became slower with increasing NVP concentrations (Figure 3B), consistent with block of A-type receptors. However, wild-type EPSCs never became as slow as those from B-type NMDA receptors. Even in an NVP concentration (300 nM) at which 97% of the A-type receptors are expected to be blocked, the wild-type EPSC deactivation (τw = 204.5 ± 16.7 ms; n = 16) was still faster than the deactivation from B-type receptors (τw = 325 ± 22.2 ms; n = 14; p < 0.0001; Figure 3C). NVP did not modulate receptor gating because increasing the NVP concentration had no effect on EPSC deactivations of diheteromeric receptors in knockout neurons (Figure 3B).
By similar reasoning, if wild-type EPSCs are composed exclusively of A-type and B-type receptors, we should be able to isolate an NVP-sensitive A-type component because low concentrations of NVP (≤10 nM) have little effect on B-type receptors. We therefore subtracted the wild-type EPSC in 10 nM NVP from the control EPSC (Figure 3D). The NVP-sensitive difference current deactivation in wild-type neurons was significantly slower (τw = 68.9 ± 12.6, n = 18) than the deactivation in N2B knockout EPSCs (τw = 22.1 ± 0.5 ms, n = 20; p < 0.0005). Overall the NVP data confirm the kinetic analysis that A-type and B-type receptors cannot be the sole receptor types contributing to wild-type EPSCs. Additionally, the wild-type EPSC dose-inhibition data was well fit with a single-site isotherm, raising the possibility that a third receptor type with intermediate NVP sensitivity is the major component.
The opening kinetics of A-type and B-type synaptic receptors are distinct and result in differences in channel open probability, as measured in knockout neurons (Tovar and Westbrook, 2012). We asked whether differences in the opening kinetics between receptor subtypes could be used to reveal a third NMDA receptor population. To examine the relative contribution of different synaptic NMDA receptor subtypes we measured the latency to first opening (following neurotransmitter binding) and the conditional channel open probability of knockout and wild-type neurons. We used the high affinity NMDA receptor open channel blocker MK-801 for this purpose (Huettner and Bean, 1988; Figure 4A) because only channels that open in response to synaptically-released glutamate will be blocked and because blocked channels do not become unblocked during the experiment. Block by MK-801 is limited by the channel opening rate following glutamate binding (Jahr, 1992; Tovar and Westbrook, 2012). Thus the EPSC in high concentrations of MK-801 (> 20 μM) is a distribution of how long it takes for channels to open following neurotransmitter binding (first latency), as shown in Figure 4A (top red trace). We used two measures to compare receptor types. We integrated the EPSC in time to measure the charge transfer and used the time at which 60% of charge transfer (Q60) in MK-801 has occurred (Figure 4A, lower trace). We also measured the probability of a channel having opened by the time of the control EPSC peak (Po*), which is the ratio of charge at control EPSC peak to the total charge.
The values of Q60 and Po* for wild-type EPSC differed significantly from either A-type or B-type receptors (Figure 4B and 4C, respectively). To expand this analysis, when we eliminated fast A-type receptors in wild-type neurons with increasing concentrations of NVP, the Q60 increased, as expected from an increasing contribution from B-type receptors (Figure 4D). However, even when we blocked 92% of A-type receptors with 100 nM NVP, the Q60 of the remaining wild-type EPSC (47.8 ± 2.2 ms; n = 7) was still much faster than for B-type receptors (109.7 ± 6.7 ms; n = 14; p < 10−6). Moreover, increasing NVP from 30 to 100 nM halved the EPSC peak, but the Q60 did not change (30 nM: 50.2 ± 8.2 ms; n = 7, p = 0.69). Similarly, the Po* of wild-type EPSCs changed only slightly in NVP (Figure 4E; control: 0.35 ± 0.01; 100 nM NVP: 0.30 ± 0.01; p < 0.05). Thus the opening kinetics of wild-type EPSCs in the absence of A-type receptors must reflect a significant population of receptors other than exclusively B-type receptors.
We used the NVP dose-inhibition data and the deactivation kinetics of A-type and B-type receptors to estimate the relative contribution of NMDA receptor subtypes to the wild-type EPSCs. We tested five NVP concentrations on each neuron because wild-type EPSC deactivations were considerably more variable than EPSCs from knockout neurons (Table 1), as expected if two or more receptor types contribute to wild-type EPSCs. Dose-inhibition data from several wild-type neurons is shown superimposed on fits to the dose-inhibition data from N2A and N2B knockout neurons (Figure 5A). The single-site isotherm fits for each individual neuron were distinctly steeper than the theoretical mixes of A- and B-type receptors. The mean Hill coefficient for the individual neurons was 1.04 ± 0.03 (n = 10), consistent with a predominant contribution from a novel receptor-antagonist interaction.
The pharmacological and kinetic properties of A- and B-type receptor EPSCs were used to estimate the contribution of each NMDA receptor subtype to wild-type EPSCs. This estimate was based on the known reduction of A- and B-type receptors by NVP and the fact that the deactivation of wild-type EPSCs became slower with increasing NVP concentrations, whereas the deactivations of pure A-type or B-type in knockout neurons did not. We used this data to constrain a series of two-exponential fits (A-type, B-type and unconstrained) of the wild-type EPSC deactivation and dose-inhibition data at each of the 5 NVP concentrations tested. The unconstrained variables provided the kinetic and dose-inhibition characteristics of AB-type receptors (see Methods). The contribution of diheteromeric receptors to the wild-type EPSC was surprisingly small (A-type: 20.1 ± 2.8%; B-type: 16.8 ± 3.0%) whereas the contribution of triheteromeric AB-type NMDA receptors predominated (Figure 5B, AB-type: 63.1 ± 4.1%). The model predicts that AB-type receptors have an NVP IC50 of 28.7 ± 1.2 nM with a Hill coefficient of 1.17 ± 0.08 (n = 9), comparable to the mean IC50 and Hill coefficient from fits of each of the 5-point dose-inhibition curves (30.0 ± 1.4 nM). Thus our data indicate that at least two thirds of the wild-type EPSC results from activation of triheteromeric AB-type receptors.
Although triheteromeric receptors have been implicated in synaptic transmission (Gray et al., 2011; Rauner and Köhr, 2011) it has not been possible to directly examine the properties of these receptors. To investigate triheteromeric receptors in relative isolation, we first evoked wild-type EPSCs in the presence of NVP (100 – 200 nM) and MK-801 (25 μM). At this concentration, NVP prevents opening of 92 – 97% of A-type receptors, thus masking their contribution to the EPSC. The remaining EPSC (resulting from B- and AB-type receptors) was repeatedly evoked in the presence of MK-801 (50 to 75 episodes; Figure 6A), resulting in a progressive and permanent block of the receptors not protected by NVP (Rosenmund, et al., 1993; see Methods for details). Because B-type receptors are less NVP-sensitive, they should become permanently blocked during stimulation in NVP/MK-801, thus enriching the contribution of A-type receptors to the remaining EPSC. Following removal of NVP and MK-801, the A- and AB-type receptors that had been masked by NVP constituted 36.4 ± 2.8% of the control EPSC (n = 17). Consistent with enrichment of A-type receptors, the τw of the remaining EPSC after removal of NVP/MK-801 was faster (44.2 ± 1.9 ms) than before NVP/MK-801 application (88.5 ± 5.8 ms; p < 10−6; Figure 6B and 6C). Finally, to isolate AB-type receptors, EPSCs were recorded in 50 nM NVP which blocks 86% of A-type receptors. With this enrichment method, coupled with the reduction of the EPSC by 50 nM NVP, we estimate the maximal contamination of A-type receptors in the triheteromeric-enriched EPSC to be less than 6% (see Methods).
The weighted deactivation time constant of the triheteromeric AB-type receptor-enriched EPSC was 78.7 ± 3.0 ms (n = 18). The deactivation of AB-type receptor EPSCs was similar to the EPSC kinetics preceding NVP/MK-801 application (Figure 6C; p = 0.31; paired comparison), but substantially faster than B-type receptors. Because removal of the B-type receptors did not affect the deactivation, this result provides further evidence that AB-type NMDA receptors constitute the major receptor type at wild-type hippocampal neuron synapses. Because τslow contributes significantly to the deactivation of enriched AB-type EPSCs (Table 1), the contribution of B-type receptors to the wild type EPSC may actually be much less than our previous maximal estimate based on EPSC deactivation (Figure 2C). Consistent with this idea, the estimate of the maximal contribution of B-type receptors in the wild-type EPSCs from these neurons (Figure 6D), done as in Figure 2C, was not affected by isolation of AB-type receptors (18.8 ± 1.9% before and 21.5 ± 1.7% after, n = 17, p = 0.23).
Functional isolation of AB-type receptors permitted us to directly examine their basic properties (Figure 7; Table 1). The Po* from these receptors, as measured using the MK-801 method, was 0.28 ± 0.02 (n = 9), closer to the value for B-type than A-type receptors (Figure 4C). The NMDA receptor ligands zinc and ifenprodil bind to homologous regions of the N2A and N2B amino-terminal domain, respectively (Paoletti, 2011) These ligands have been used extensively to examine A- and B-type receptors, but how these ligands directly affect AB-type receptors at synapses is unknown. As shown in Figure 7A, both ligands decreased the peak of the AB-type receptor-enriched EPSC but only zinc prolonged the deactivation. This result indicates that in AB-type receptors, occupancy of a single high affinity zinc binding site produces synaptic charge redistribution.
To compare the actions of these amino-terminal domain ligands, we plotted the ligand-induced change in total charge transfer as a function of the change in the EPSC peak amplitude (Figure 7B). Competitive antagonists like NVP are not expected to alter the EPSC deactivation kinetics. Thus for any dose of NVP, the reduction in EPSC peak and charge should fall on the unity line, as shown for NVP on A-type EPSCs. Ifenprodil (3 μM) reduced the EPSC peak and charge of AB-type receptors to similar degrees, consistent with its lack of effect on the deactivation, producing a result comparable to that of a competitive antagonist. In contrast, zinc (100 nM, free) greatly prolonged the EPSC deactivation of AB-receptors, resulting in a significant redistribution of charge attributed to the τslow component of deactivation (60.3 ± 1.3% in control; 72.1 ± 1.9% in 100 nM zinc; n = 9; p < 0.001). AB-type receptors thus showed a departure from unity in Figure 7B that was comparable to A-type receptors.
We used a kinetic and pharmacological approach to investigate the relative contribution of NMDA receptor subtypes to EPSCs in wild-type mouse hippocampal neurons. This approach revealed that the majority contribution to EPSCs was from receptors with a triheteromeric molecular composition, containing N2A and N2B along with N1; diheteromeric receptors together accounted for at most one-third of the wild-type EPSC. Isolation of the triheteromeric population revealed kinetic and pharmacological properties that were distinct from either A- or B-type receptors. Our results indicate that NMDA receptor assembly, targeting and function in synaptic plasticity could be largely determined by AB-type receptors.
Although it has often been assumed that cortical synapses contain predominately, or exclusively, diheteromeric NMDA receptors with either N2A or N2B subunits, multiple lines of evidence have suggested that this view is too simplistic. N2A and N2B can assemble in the same biochemical protein complex in rodent cortex, suggesting the existence of triheteromeric receptors (Sheng et al., 1994; Chazot 1994; Luo et al., 1997; Dunah and Standaert, 2003), although their relative abundance in postsynaptic density preparations from the hippocampus has been questioned (Al-Hallaq et al., 2007). From a functional perspective, triheteromeric receptors also have been identified in heterologous cells (Hatton and Paoletti, 2005). Recently, the fact that NMDA receptor-mediated EPSC properties cannot be explained based on diheteromeric receptor properties has implicated triheteromeric receptors at hippocampal synapses (Gray et al., 2011; Rauner and Köhr, 2011). Because hippocampal neurons in our experiments expressed only NR1, N2A and N2B, we were able to characterize three subtypes of synaptic NMDA receptors. The wild-type EPSC deactivation time course could not be accounted for by summation of the deactivation time courses of A- and B-type receptors. Our analysis assumes that A-type and B-type receptors in knockout and wild-type neurons have identical properties. We functionally isolated and characterized the third population, the AB-type receptor. These receptors had kinetic and pharmacological properties that reflected a contribution of both N2A and N2B. Neurons often express more than one type of N2 subunit. Thus triheteromeric NMDA receptors in other cell types may be more common that previously appreciated. For example, a triheteromeric receptor with N1, N2B and N2D has been suggested in cerebellar Golgi cells (Cathala et al., 2000). Splice variants of the NR1 subunit (Rumbaugh et al., 2000) could add further complexity to the functional properties of triheteromeric receptors.
Because single cells express multiple types of NMDA receptors, neurons likely possess mechanisms that control receptor assembly and localization at synapses. For example, synaptic inputs onto single neurons from different pathways activate synaptic NMDA receptors with different properties (Kumar and Huguenard, 2003; Arrigoni and Greene, 2004). By using the differences in Po* between receptor subtypes and their relative contribution to the wild-type EPSC amplitude in excitatory hippocampal neurons, we calculated that AB-type receptors are 5.8 and 3.2 times more abundant at these synapses than A- or B-type receptors, respectively. Whether this selectivity for AB-type receptors results from preferential assembly or preferential localization is yet to be determined. The initial step in receptor assembly is the dimerization of N1 and N2 subunits, followed by association of dimers into tetramers (Furukawa et al., 2005; Schuler et al., 2008). If we assume equal expression levels of N2A and N2B (Monyer et al., 1994) and comparable association rates between N1 and N2A or N2B, then AB-type receptors should only be twice as abundant at synapses as either A- or B-type receptors. Because over-expression of individual subunits does not affect the synaptic NMDA receptor complement (Prybylowski et al., 2002), the excess of AB-type receptors may result from preferential retention at synapses. The preponderance of AB-type receptors at synapses indicates that localization or anchoring of NMDA receptors at synapses cannot be determined simply by the association of either N2A or N2B with postsynaptic density proteins. The intracellular carboxyl-terminal domains of N2A and N2B and their binding partners confer distinct roles in synaptic signaling (Ryan et al., 2013). It is thus interesting that the majority of synaptic NMDA receptors have access to a broader than expected array of intracellular signaling pathways.
The vast majority of NMDA receptor activation occurs at synapses following phasic release of neurotransmitter. The time course of NMDA receptor activation reflects a combination of the non-stationary nature of transmitter presentation and the constituent receptor subtypes that are activated. Within this context, the non-equilibrium behavior of triheteromeric synaptic receptors revealed several interesting molecular outcomes. The EPSC deactivation kinetics of AB-type receptors were more similar to those of A-type than B-type receptors, indicating a predominant influence of the N2A subunit on deactivation. In contrast, the Po* of AB-type receptors was more similar to B-type than A-type receptors, suggesting that the N2B subunit has a predominant influence on gating steps that lead to the open state. The N2 subunits contain the binding site for glutamate and gating of NMDA receptors requires binding of two glutamate molecules (Clements and Westbrook, 1992). Thus the distinct kinetics of AB-type receptors could simply arise from the molecular compromise of rapid glutamate unbinding from N2A producing the fast phase of deactivation, whereas the influence of N2B is rate-limiting for channel opening. However, more complex allosteric explanations are also possible. We did not measure the voltage-dependence or calcium permeability of AB-type receptors. However, these properties of NMDA receptors, unlike the deactivation kinetics and open probability, are quite similar between A- and B-type receptors (Monyer et al., 1994), and thus AB-type receptors are likely to be similar to diheteromeric receptors.
Ifenprodil, NVP and zinc have been used to classify native NMDA receptor subtypes as either “N2A-containing” vs “N2B-containing”, thus creating the perception that native receptors are predominantly diheteromeric. These antagonists have been used to study synaptic plasticity and in developmental and behavioral experiments to assign functional roles to particular NMDA subunits. However this pharmacological strategy is valid only if triheteromeric receptors are either not present or they can be similarly pharmacologically separated. Our results indicate that zinc or ifenprodil also directly reduce EPSCs from AB-type receptors, but less potently than for A- or B-type receptor EPSCs. This reduction is consistent with lower efficacy of these ligands for the AB-type receptor (Hatton and Paoletti, 2005). The intermediate effects of ifenprodil and zinc, as well as the competitive antagonist NVP on triheteromeric receptors indicates that these ligands act on multiple NMDA receptor subtypes. Thus many of the experimental findings that have resulted from the use of ifenprodil or NVP could have arisen simply from a reduction in the NMDA receptor-mediated EPSC (Berberich et al., 2005), rather than a subtype-specific action.
Allosteric modulators like ifenprodil and its derivatives have been extensively investigated for their therapeutic potential for neurological and psychiatric disorders (Mony et al., 2009; Popescu et al., 2010). Drugs that alter channel gating may have greater therapeutic potential than competitive antagonists that simply prevent channel gating. The effects of benzodiazepines or barbiturates on inhibitory postsynaptic currents provide one such example (Zhang et al., 1993). Assessment of such modulators requires consideration of the phasic nature of neurotransmitter release as well as non-stationary receptor activation, the latter being the predominant form of NMDA receptor activation. Zinc and ifenprodil prolong EPSCs in A-type and B-type synaptic receptors, respectively by increasing the total amount of time the fully-bound receptor spends in a closed state (Tovar and Westbrook, 2012). Although these effects were pronounced in homogenous synaptic diheteromeric receptor populations, in wild-type EPSCs, only zinc prolonged the deactivation. One reason for this is because zinc binding to AB-type receptors results in significant prolongation of the EPSC.
Drugs that target B-type receptors or the N2B amino-terminal domain have proven to be of limited clinical relevance (Villmann and Becker, 2007). Two observations from our work provide a potential explanation for the lack of efficacy of ifenprodil and its derivatives as therapeutics. First, B-type receptors, at least in the hippocampal neurons in our experiments, are a minor component of wild-type EPSCs. Secondly, the ifenprodil effect on AB-type receptors was functionally analogous to that of a competitive antagonist, whereas zinc prolonged EPSCs generated by triheteromeric AB-type receptors. This difference indicates that the amino-terminal domains of N2A and N2B contribute differentially to gating in the AB-type receptor complex (Geilen et al., 2009). If our result that AB-type receptors are predominant at synapses is generalizable to other neurons that express N2A and N2B, then drugs that target the amino-terminal domain of the N2A subunit may prove more promising as clinical therapeutics (Nozaki et al., 2011).
We are grateful to Dr. Masayoshi Mishina for the gift of the N2A and N2B knockout mice, to Dr. Yves Auberson (Novartis Pharma AG, Basel, Switzerland) for the gift of NVP-AAM077, and to Drs. Craig Jahr and Eric Schnell for their careful reading and comments on this manuscript and to AeSoon Bensen for technical assistance. This work was supported by NS 26494 and MH 46613 to GLW.
Contributions: KRT designed, performed and analyzed the experiments; MLM assisted with data analysis; GLW assisted with planning of the experiments and wrote the manuscript with KRT.