Cross talk between synaptic receptors mediates NMDA-induced suppression of inhibition

Mariangela Chisari, Charles F. Zorumski, Steven Mennerick


Past research has shown that calcium influx through NMDA receptors (NMDARs) depresses GABAA currents. We examined upstream triggers of this suppression, including involvement of target synaptic GABAA receptors and the NMDARs triggering suppression. In hippocampal neurons, conditioning with 20 μM NMDA for 20 s caused 50% suppression of GABA responses. The suppression was delayed by ∼60 s following NMDA application and persisted for at least 5 min following conditioning. Pharmacology experiments suggested a shift in both the sensitivity to GABA and a loss of functional receptors. NMDA conditioning strongly suppressed inhibitory postsynaptic currents and speeded decay kinetics. Synaptic NMDAR conditioning was necessary to suppress GABA current in pyramidal neurons; extrasynaptic NMDAR activation did not suppress, even when matched to synaptic activation. We found no evidence that specific synaptic NMDAR subunits mediate depression of GABA responses. Although physical colocalization of glutamate and GABAA receptors is mostly likely in extrasynaptic regions, our evidence suggests that NMDAR-induced suppression of GABA responsiveness prominently affects precise, moment-to-moment signaling from synaptic receptors to synaptic receptors.

  • synaptic plasticity
  • γ-aminobutyric acid
  • N-methyl-d-aspartate
  • hippocampal neurons
  • epilepsy

plasticity of fast synaptic transmission is thought to be an important substrate for behavioral change, including lasting memory formation. The two major neurotransmitters mediating fast transmission in the brain are glutamate and GABA, acting at their respective ionotropic receptors. Forms of short-term and long-term plasticity are well described for both transmitter systems (Malenka and Bear 2004; McBain and Kauer 2009). In addition, cross-transmitter receptor plasticity mediates changes in GABA transmission following glutamate signaling. Although either potentiation or suppression can occur, under many conditions Ca2+ influx through NMDA-type glutamate receptors (NMDARs) suppresses subsequent responsiveness of GABAA receptors (Stelzer and Shi 1994). This effect requires calcineurin recruitment and probably a subsequent change in the phosphorylation status of target GABAA receptors (Chen and Wong 1995; Huang and Dillon 1998; Vithlani and Moss 2009). NMDAR-induced GABAA receptor suppression is expected to be important for the balance between excitation and inhibition. This form of disinhibition could facilitate long-term potentiation but could also participate in runaway excitation and excitotoxicity. Despite its potential importance, questions remain about the upstream triggers and the functional effects on GABAA receptors targeted by this form of plasticity.

Heteropentameric GABAA receptors are found clustered at synapses, but most previous studies have examined responses to exogenous GABA (Stelzer and Shi 1994); therefore the impact on synaptic inhibition and inhibitory postsynaptic currents (IPSCs) is unclear. Because NMDA and GABAA receptors are most likely to colocalize in extrasynaptic domains (although see Bekkers 2005), synaptic receptors may not be prominently targeted for modulation. Furthermore, some evidence suggests that GABA EC50 (i.e., receptor sensitivity to GABA) is primarily affected by NMDA-induced suppression (Inoue et al. 1986; Mozrzymas and Cherubini 1998). A change in GABA EC50 may not have much impact on IPSCs, since IPSCs are thought to be generated by locally saturating GABA concentration (Jones and Westbrook 1995; Maconochie et al. 1994). On the other hand, some studies have suggested that a decrease in the number of functional receptors underlies suppression (Houston et al. 2009; Muir et al. 2010; Wang et al. 2003). This mechanism might be expected to have more impact on IPSCs. Suppression of synaptic GABAA receptors would impart potentially important temporal and spatial precision to plasticity.

NMDARs are clustered at excitatory synaptic junctions, but extrasynaptic receptors also mediate persistent excitation in some situations (Sah et al. 1989). There are differences in NR2A and NR2B subunit presence between synaptic and extrasynaptic NMDARs, but these differences are quantitative and not dichotomous (Harris and Pettit 2007; Petralia et al. 2010; Thomas et al. 2006). Selective subunit contributions notwithstanding, the location of NMDARs responsible for triggering suppression of GABAA currents is currently unknown.

Despite the expected colocalization of extrasynaptic NMDA and GABAA receptors, we find evidence that in pyramidal neurons, synaptic receptors are prominently involved in both induction and expression of cross-receptor suppression. We find evidence for both a change in sensitivity and maximum responses of target GABAA receptors. In pyramidal neurons, synaptic NMDARs appear to be exclusively involved in induction of suppression, although we find no evidence for differential subunit involvement. The results suggest that this form of receptor plasticity is adapted to affect synaptic transmission. In part, this specificity may help keep excitation-induced GABAA receptor suppression from exacerbating runaway excitation.


Cell cultures.

Hippocampal neurons were prepared from postnatal day 1–3 rat pups anesthetized with isoflurane, under protocols consistent with National Institutes of Health guidelines and approved by the Washington University Animal Studies Committee. Protocols were adapted from previously published accounts (Mennerick et al. 1995). Hippocampi were sliced (500-μm thickness) and digested with 1 mg/ml papain in oxygenated Leibovitz L-15 medium (Invitrogen, Gaithersburg, MD). Tissue dissociation was made by mechanical trituration in modified Eagle's medium (Invitrogen) containing 5% horse serum, 5% fetal calf serum, 17 mM d-glucose, 400 μM glutamine, 50 U/ml penicillin, and 50 μg/ml streptomycin. Cells were plated in modified Eagle's medium at a density of ∼650 cells/mm2 as mass cultures (onto 25-mm cover glasses coated with 5 mg/ml collagen) or 100 cells/mm2 as “microisland” cultures (onto 35-mm plastic culture dishes coated with collagen microdroplets on a layer of 0.15% agarose). Neuronal cultures were maintained at 37°C in a humidified incubator with 5% CO2-95% air. The antimitotic cytosine arabinoside (6.7 μM) was added 3–4 days after plating to inhibit glial proliferation. The following day, one-half of the culture medium was replaced with Neurobasal medium (Invitrogen) plus B27 supplement (Invitrogen).


Whole cell recordings were performed at room temperature from neurons cultured for 10–13 days using an Axopatch 200B amplifier (Molecular Devices, Sunnyvale, CA). For recordings, cells were transferred to an extracellular solution containing (in mM) 138 NaCl, 4 KCl, 2 CaCl2, 10 glucose, 10 HEPES, 0.02 glycine, 0.001 2,3-dihydroxy-6-nitro-7-sulfonyl-benzo[f]quinoxaline (NBQX), and 0.01 d-2-amino-5-phosphonovalerate (d-APV) at pH 7.25. NMDA was applied in a d-APV-free solution. The tip resistance of patch pipettes was 3–6 MΩ when filled with an internal solution containing (in mM) 130 CsCl, 2 NaCl, 0.1 EGTA, and 10 HEPES at pH 7.25, adjusted with NaOH. For experiments using 1 mM GABA, the chloride driving force was altered to elicit currents in the range of 1–2 nA and patch pipettes were filled with the following internal solution containing (in mM) 124 cesium methanesulfonate, 6 CsCl, 2 NaCl, 0.1 EGTA, and 10 HEPES at pH 7.25, adjusted with NaOH. Holding voltage was set at −70 mV. Patch-pipette tip resistance for autaptic recordings was 2.5–3.5 MΩ when filled with an internal solution containing (in mM) 140 KCl, 4 NaCl, 0.1 EGTA, and 10 HEPES at pH 7.25, adjusted with KOH. Access resistance (8–10 MΩ) was compensated 80–100% for synaptic recordings. For autaptic responses, cells were stimulated with 1.5-ms pulses to 0 mV from −70 mV to evoke transmitter release. Drugs were applied with a multibarrel, gravity-driven local perfusion system. The estimated solution exchange times were 96 ± 3 ms (10–90% rise), estimated from junction current rises at the tip of an open patch pipette. For exogenous GABA applications, three concentrations of GABA were used, as indicated: 10 μM GABA was used as a concentration near the EC50 and that exhibits strong desensitization (Shu et al. 2004); 1 mM GABA was used as a saturating concentration (Shu et al. 2004); and finally, to avoid complications from fast receptor desensitization relative to drug application times, 0.5 μM GABA, a low, nondesensitizing concentration, was used for many experiments.

Data analysis.

Data acquisition and analysis were performed primarily using pCLAMP 9 software (Molecular Devices). Data were processed with Microsoft Excel and presented as means ± S.E. Statistical significance was determined using a Student's two-tailed t-test. Data plotting, statistical analysis, and figure preparations were completed with GraphPad Prism (GraphPad Software, La Jolla, CA) and Adobe Photoshop (San Jose, CA). Offset times were measured using standard exponential fitting functions.


All drugs were obtained from Sigma (St. Louis, MO) except for d-APV and NBQX, which were obtained from Tocris (Ellisville, MO).


Properties of target GABAA receptors.

Hippocampal neurons were challenged with test applications of 10 μM GABA (the approximate EC50 concentration; Shu et al. 2004) before and after a conditioning application of NMDA (20 μM; Fig. 1A). Immediately after NMDA conditioning, peak GABA current was typically only slightly suppressed, but suppression gradually increased to a stable level within 2 min (Fig. 1A). Recovery was not explicitly investigated, but responses exhibited only slight recovery during the ensuing 5 min of recording. Because we set a fixed intracellular Cl concentration using whole cell recordings, NMDA-induced alterations in the Cl gradient (Lee et al. 2011) are unlikely to participate in the observed depression. Conditioned 10 μM GABA responses exhibited weaker desensitization than baseline responses (Fig. 1, A and B), and kinetics of macroscopic desensitization were slower in conditioned GABA responses (Fig. 1B), although quantification of this effect may be obscured by drug delivery that underestimates the peak of rapidly desensitizing currents (see methods). The reduction in GABA peak current we observed was consistent with several reports from the literature which have shown that Ca2+ influx and calcineurin activation depress GABAA receptor function, likely by altering the receptor phosphorylation status (Chen and Wong 1995; Stelzer and Shi 1994).

Fig. 1.

GABAA current (IGABA) is suppressed by NMDA receptor (NMDAR) activation. A: representative traces of GABA and NMDA currents elicited from the same hippocampal neuron. GABAA receptors were activated several times with 10 μM GABA (25 s between traces) until stable current amplitudes were obtained (traces before NMDA application). After a 20-s application of 20 μM NMDA, GABA elicited a smaller current (1st trace after NMDA), which decreased over time with sequential GABA applications, and it did not recover in the following 3 min after NMDA conditioning. Inset shows a time course of peak IGABA for 5 cells. Time 0 is the time of NMDA conditioning. B: summary of NMDA conditioning effects on GABA response: peak current, desensitization ratio, and desensitization decay time constant, as indicated. Comparisons were made between GABA current recorded 1–2 min after conditioning and the response immediately preceding NMDA conditioning. Values are means ± SE (n = 5). *P <0.05; **P <0.01; ***P <0.0001 (paired t-test).

On the basis of the change in macroscopic desensitization, we hypothesized that reduced GABA current might result from a shift in the receptor sensitivity for agonist (Inoue et al. 1986). First, we examined tests of allosteric modulators on responses to GABA. We used the benzodiazepine lorazepam as a probe of GABA sensitivity. Because benzodiazepines shift the GABA EC50 without altering the maximum attainable current, potentiation depends on the effective agonist concentration relative to the maximum response. For instance, a response 10% of maximum will be more strongly potentiated than a response that is 50% of maximum because of the ceiling of 100%. We coapplied lorazepam (1 μM) with an approximate EC50 GABA concentration of 10 μM (Shu et al. 2004) before and after NMDA conditioning. We observed that lorazepam potentiation reliably increased with NMDA-induced suppression (Fig. 2, A and C).

Fig. 2.

Allosteric modulation is altered after NMDA conditioning. A and B: representative traces of hippocampal neurons challenged with 10 μM GABA and the positive modulator lorazepam (1 μM; A) and 3α5αP (0.01 μM; B) before and after 20-s NMDA (20 μM) conditioning. Note that 3α5αP was preapplied for 15 s before coapplication with GABA. Lorazepam and 3α5αP barely potentiated GABA current before conditioning, but the effect of potentiation was stronger after NMDAR activation. C and D: summary of potentiated current by lorazepam (C) and 3α5αP (D) before and after NMDA conditioning. Values are means ± SE (n = 6 for lorazepam; n = 11 for 3α5αP). **P <0.01; ***P <0.0001 (paired t-test).

The increased lorazepam potentiation might result from a shift in GABA sensitivity of γ2-containing GABAA receptors. Alternatively, the result could be caused by fewer available lorazepam-insensitive receptors following conditioning. We distinguished these possibilities using a different allosteric modulator, the neurosteroid 3α5αP (10 nM). 3α5αP (allopregnanolone) is a broad-spectrum GABAA receptor modulator that, like lorazepam, exhibited stronger effects on NMDA-suppressed GABA responses (Fig. 2, B and D). Because δ-containing receptors are thought to be particularly sensitive to neurosteroids (Brown et al. 2002; Wohlfarth et al. 2002), we also specifically determined the contribution of δ-containing GABAA receptors to responses under our conditions. Neurons were challenged with the δ-selective agonist 4,5,6,7-tetrahydroisoxazolo(5,4-c)pyridin-3-ol (THIP; 0.5 μM) (Meera et al. 2011) alone and coapplied with the δ-selective allosteric modulator DS2 (1 μM) (Wafford et al. 2009). We failed to observe measurable currents in response to this δ-selective concentration of THIP, and currents were not detectably enhanced by the coapplication of DS2 (data not shown). We conclude that our cultures contain mostly pyramidal neurons and interneurons lacking δ-containing receptors. Together, the lorazepam and 3α5αP results are consistent with the hypothesis that NMDA-induced suppression includes a decrease in receptor agonist sensitivity.

A change in apparent GABA sensitivity could result from a change in GABA binding or from a change in receptor gating. To begin to distinguish these possibilities, we tested the effects of NMDA conditioning on an agonist, pentobarbital, which binds to a site distinct from the GABA binding site and directly opens the channel at sufficiently high concentration (Amin and Weiss 1993). In the same cell, we sequentially applied 0.5 μM GABA and 100 μM pentobarbital, and we compared the current elicited before and after NMDA conditioning (Fig. 3A). A low concentration of GABA was used in this experiment for comparison with the low-efficacy, nondesensitizing responses gated by pentobarbital. NMDA conditioning suppressed responses to both agonists similarly (Fig. 3A; normalized responses of 0.29 ± 0.05 for GABA and 0.23 ± 0.03 for pentobarbital), suggesting that GABA binding is not primarily affected by NMDAR stimulation and that alterations in gating more likely underlie the sensitivity change.

Fig. 3.

Suppressed GABA current partly results from loss of functional GABAA receptors. A: representative traces of current elicited by 0.5 μM GABA and 100 μM pentobarbital (Pb) in the same neuron before and after NMDAR stimulation. B: representative traces of 10 μM and 1 mM GABA responses before (left) and after (right) NMDA conditioning on a representative hippocampal neuron. Bar graph shows that responses to saturating 1 mM GABA were less strongly suppressed (shaded bar) than responses to a subsaturating concentration of GABA (solid bar). Values are means ± SE (n = 5). *P < 0.05.

Results of Figs. 13A do not exclude the possibility that a loss of surface GABA receptors also participates in NMDA-induced suppression (Wang et al. 2003). To test this possibility, we challenged cells with a subsaturating (10 μM) and a saturating (1 mM) concentration of GABA before and after NMDA conditioning (Fig. 3B). To decrease current amplitudes and associated access-resistance errors, these experiments were performed with reduced intracellular Cl concentration (see methods). Conditioned GABA responses at both concentrations were suppressed, but the extent of current reduction was stronger for the subsaturating GABA response than for the saturating response (Fig. 3B, n = 5), demonstrating that the fraction of maximum current represented by 10 μM GABA significantly decreased after conditioning (0.73 ± 0.07 before vs. 0.51 ± 0.07 after, P < 0.01). This experiment directly supports the idea that both sensitivity (GABA EC50) (Inoue et al. 1986) and functional receptor number (Wang et al. 2003) are depressed by NMDA conditioning.

To test synaptic GABAA receptor involvement, we examined IPSCs from hippocampal neurons. We evoked IPSCs from isolated hippocampal neurons before and after NMDA conditioning. Consistent with the idea that synaptic GABAA receptors are prominent targets of modulation, NMDA conditioning depressed IPSC peak current (Fig. 4, A and B) and also accelerated offset decay times of evoked IPSCs (Fig. 4A, inset, and Fig. 4B). These two changes are consistent with the alterations in both sensitivity (gating) and functional receptor number observed with exogenous application of GABA.

Fig. 4.

NMDA conditioning affects peak current and decay time of inhibitory postsynaptic currents (IPSCs). A: traces of evoked IPSCs from the same isolated hippocampal neuron before and after NMDA conditioning. Inset shows scaled IPSCs from the main panel to highlight the effect of NMDAR activation on decay times. For this and subsequent displays of recurrent (autaptic) postsynaptic currents, transient currents associated with stimulation have been partially removed for clarity. B: summary of changes in peak current (left) and 10–90% decay time (right) of IPSCs recorded as in A. Values are means ± SE (n = 15). *P < 0.05; **P < 0.01 (paired t-test).

Because autaptic IPSCs under normal divalent conditions exhibit a large quantal content, spillover onto extrasynaptic receptors could participate in the IPSCs shown in Fig. 4. To address this, we performed additional experiments in low Ca2+ and elevated Mg2+ to reduce quantal content. IPSCs under these conditions also exhibited depression following NMDA conditioning in normal extracellular Ca2+ (Fig. 5A, 27 ± 12% depression vs. 51 ± 8% depression in regular quantal content, n = 6). Furthermore, conditioning with NMDA did not affect the paired-pulse ratio of IPSCs (Fig. 5B), consistent with an effect on postsynaptic GABAA receptors but not presynaptic release.

Fig. 5.

Low quantal content IPSCs are depressed by NMDAR conditioning and paired-pulse ratios are unaffected. A: IPSCs obtained in 1 mM Ca2+ and 2 mM Mg2+ to limit transmitter release. Samples are IPSC pairs elicited before (left) and after (right) NMDA conditioning. B: summary of effects on peak IPSC (left) and paired-pulse ratio (right). Values are means ± SE (n = 6 cells). *P < 0.05 (paired t-test). C: sample spontaneous IPSCs from a conventional mass culture of hippocampal neurons before and after NMDA (20 μM) conditioning for 30 s (top). Histogram (bottom) shows peak amplitude distributions of all spontaneous IPSCs recorded before (open bars) and after (filled bars) conditioning. Example is representative of 7 cells recorded in 1 mM Ca2+, 0 mM Mg2+ in the presence of 1 μM 2,3-dihydroxy-6-nitro-7-sulfonyl-benzo[f]quinoxaline (NBQX) and 10 μM d-2-amino-5-phosphonovalerate (d-APV). d-APV was removed and 10 μM glycine was added during NMDA conditioning. See results for summary statistics.

We also examined spontaneous IPSCs in networks of neurons in conventional cultures where monosynaptic interactions are less exaggerated than in recurrent autaptic connections. We used a bath solution of 1 mM Ca2+ and no added Mg2+ to keep quantal content low. Under these conditions, average spontaneous IPSC amplitude in 7 neurons was −323 ± 106 pA and was suppressed following NMDA (20 μM) conditioning by 58 ± 10% (n = 7 neurons, P < 0.05). Sample records and amplitude histograms before and after NMDA are shown in Fig. 5C for one of these neurons. Together, results shown in Figs. 4 and 5 suggest a strong effect of NMDA conditioning on synaptic GABAA receptors.

Properties of the NMDARs inducing suppression.

In the remainder of this work, we characterized the triggers for NMDA-induced suppression. In preceding experiments, NMDA conditioning was achieved with a moderate concentration of exogenously applied NMDA, which tonically and indiscriminately activated all extrasynaptic and synaptic NMDAR populations. We asked specifically whether synaptic activation of NMDARs is sufficient to suppress GABAA responses. In isolated neurons we first evoked 5 control, false NMDA excitatory postsynaptic currents (EPSCs) at 20-s intervals in the presence of d-APV (50 μM). This treatment accounts for any effects of repeated depolarization and resulted in only mild suppression of responses to 0.5 μM GABA (Fig. 6, A and B), possibly through a small suppressing effect of breakthrough NMDA current combined with stimulus-dependent Ca2+ influx. In the same neuron, we removed the d-APV and repeated the protocol to evoke recurrent, autaptic EPSCs. Under these circumstances, conditioning by 5 NMDA EPSCs strongly suppressed GABA responses, similar to the effect of exogenous NMDA (Fig. 6, A and B). To ensure that GABA responses were indeed suppressed by NMDA EPSC conditioning and not by a time-dependent effect, we omitted the false EPSC portion of our protocol in a different set of cells, and we found that NMDA EPSC conditioning still strongly suppressed GABA responses. Conditioned GABA current was reduced to 0.47 ± 0.05 that of baseline current (P <0.0001, 2-tailed paired t-test). These results indicate that synaptic conditioning efficiently suppresses GABAA currents.

Fig. 6.

Evoked NMDA excitatory postsynaptic currents (EPSCs) suppress GABA current. A: traces from the same hippocampal neuron challenged with 0.5 μM GABA (as indicated) before and after 5 evoked “false” NMDA EPSCs (elicited in presence of 50 μM d-APV) and 5 evoked NMDA EPSCs. GABA current is suppressed only after NMDAR activation with effective EPSCs. B: summary of EPSC peak current (left) from 10 cells challenged with protocol shown in A and normalized GABA current (right). Values are means ± SE. *P < 0.05; **P < 0.01; ***P < 0.0001 (paired t-test).

GABAA receptor suppression after NMDA conditioning might be mediated by specific NMDAR subunits, since specific subunits are associated with other forms of plasticity (Izumi et al. 2005; Yashiro and Philpot 2008). NMDARs are tetrameric complexes formed by NR1 and NR2 subunits. NR2A vs. NR2B subunits impart pharmacological and kinetic variability to NMDARs and may be differentially modulated (van Zundert et al. 2004). NR2B subunits are present both synaptically and extrasynaptically, although their relative contributions change with development (Thomas et al. 2006; Tovar and Westbrook 1999). We used pharmacological subtraction to examine NR2 subunit contributions to synaptic conditioning effects. We first evaluated NMDA EPSCs in the presence of 10 μM ifenprodil to selectively block NR2B-containing NMDARs (Izumi et al. 2006; Izumi et al. 2005). Ifenprodil inhibited ∼60–70% of peak EPSC (Fig. 7A). We found that 3 μM d-APV, a nonselective NMDAR antagonist, matched the amount of ifenprodil inhibition (Fig. 7, A–C). GABA responses measured before and after five evoked, antagonist-depressed NMDA EPSCs were significantly suppressed after conditioning (Fig. 7D). The amount of GABA current suppression did not differ between d-APV and ifenprodil, although suppression was milder than observed with full-amplitude EPSCs (cf., Fig. 6). These results suggest that there is no particular subunit preference to NMDA-induced GABAA suppression when synaptic receptors drive suppression. However, NR2B subunits play a prominent role in EPSCs in neonatal tissue and cultures (Tovar and Westbrook 1999); it is possible that under conditions in which NR2B plays a less prominent overall role, selectivity of its contributions may emerge.

Fig. 7.

GABA suppression is not governed by specific subunit composition of synaptic NMDARs. A: representative traces of evoked NMDA EPSCs in regular saline (solid traces) and after application of ifenprodil (10 μM; shaded trace), a selective antagonist of NR2B-containing receptors. B: NMDA EPSCs obtained from a sibling neuron were evoked in the presence of 3 μM d-APV to mimic the amount of block obtained in the presence of ifenprodil. C: concentrations used for the antagonists in A and B produced matched suppression of NMDAR EPSCs. Values in C and D are means ± SE (n = 10 for d-APV; n = 9 for ifenprodil). D: summary of normalized GABA current after NMDA EPSC conditioning in the presence of antagonists and the integrated current (shown as charge summed across 5 evoked EPSCs). In both antagonist protocols, GABA responses were suppressed. *P <0.05, significantly different from baseline GABA response indicated by dotted line (paired t-test).

Although results of Figs. 6 and 7 show that synaptic receptor activation is sufficient to suppress GABAA responses, we might expect that extrasynaptic NMDARs are most colocalized with GABAA receptors and are thus especially well positioned to suppress GABAA receptor responsiveness. Therefore, we designed an experiment to test the efficiency of extrasynaptic NMDAR activation in suppressing GABAA responses. Our strategy was to use the irreversible, activation-dependent blocker MK-801 during synaptic glutamate release to eliminate synaptic NMDARs, and then to condition unblocked extrasynaptic receptors with exogenous NMDA. To avoid Ca2+ influx during synaptic receptor activation and MK-801 block, we used the Ca2+ independent secretagogue hypertonic sucrose in the nominal absence of extracellular Ca2+ to evoke glutamate release (Rosenmund and Stevens 1996) (Fig. 8A). We repeatedly coapplied 0.5 M sucrose with 30 μM MK-801 until EPSCs were <10% of the initial amplitude (Fig. 8A, sucrose + MK-801). This protocol effectively blocked synaptic NMDARs without altering GABAA receptor responses. Taking advantage of long-lived MK-801 block of synaptic NMDARs, we next applied a high concentration of NMDA (300 μM in normal extracellular Ca2+ for 20 s) to activate unblocked extrasynaptic receptors. This strong extrasynaptic NMDAR activation was insufficient to suppress GABAA responses (Fig. 8, A and B).

Fig. 8.

Extrasynaptic NMDAR activation is not sufficient to suppress GABAA receptor function in excitatory neurons. A: sequence of conditioning and test applications is shown. Hypertonic sucrose (0.5 M) was applied in nominally Ca2+-free saline in the presence of 30 μM MK-801. After functional elimination of NMDA EPSCs with sequential sucrose + MK-801 applications (shaded trace), GABA current was unchanged (3rd panel). Subsequent conditioning of remaining, unblocked extrasynaptic receptors with a high NMDA concentration (300 μM) failed to suppress the GABA response. B: bar graph summarizing effects on GABA responses at various stages during the protocol. Values are means ± SE (n = 5 cells). The trend toward a potentiated GABA response with extrasynaptic receptor conditioning was not significant (P = 0.3). C: representative traces from a glutamatergic (GLU) cell conditioned with exogenous application of 5 μM NMDA, applied to all receptors. D: representative traces of a solitary GABAergic cell with the same protocol as in C. E: summary of NMDAR charge density (left) and normalized GABA responses (right) elicited before and after conditioning with 5 μM NMDA in glutamatergic and GABAergic cells. Values are means ± SE (n = 11 for GLU; n = 9 for GABA). GABA response amplitude is normalized to the GABA response before conditioning in the same cell. *P <0.05, significantly different from baseline GABA response indicated by dotted line (paired t-test).

One possibility to explain the lack of effectiveness of extrasynaptic receptor activation is that global Ca2+ influx through extrasynaptic receptors is simply insufficient to drive GABAA suppression. To test this hypothesis, we exogenously activated NMDARs in glutamatergic cells using a lower concentration of NMDA (5 μM), applied to the full complement of synaptic and extrasynaptic receptors. This level of overall receptor activation was chosen to match the charge density produced by 300 μM NMDA applied selectively to extrasynaptic receptors (−84.3 ± 9.2 pC/pF, n = 11 excitatory neurons challenged with 5 μM NMDA, and −61.9 ± 23.0 pA/pF, n = 5 cells from Fig. 8B challenged with 300 μM NMDA at extrasynaptic receptors; P > 0.05). In this matched protocol, we found that low-level activation of all receptors, synaptic and extrasynaptic, significantly suppressed GABAA current (Fig. 8, C and E). Therefore, despite matching the Ca2+ load, only synaptic NMDARs effectively suppressed GABA responsiveness.

In solitary, autaptic GABAergic cells NMDARs exhibit synaptic-like clustering, despite the absence of glutamatergic presynaptic terminals (Rao and Craig 1997). Figure 5 predicts that this arrangement should suffice to induce robust GABAA receptor suppression, equivalent to principal, glutamatergic neurons. To test this, we applied 5 μM NMDA in solitary interneurons in sibling cultures of those in Fig. 8C. We found an NMDA charge density on interneurons comparable to that on glutamatergic cells (Fig. 8E), consistent with the idea that NMDAR distribution is similar (Rao and Craig 1997). GABAA responses were also depressed by 5 μM NMDA on solitary interneurons (Fig. 8, D and E), despite the lack of bona fide synaptic receptors. These results demonstrate that interneurons, even in the absence of presynaptic glutamate inputs, are competent to exhibit GABAA receptor suppression.

Figure 9 summarizes the effects of different modes and degrees of NMDAR activation on glutamatergic neurons from our studies, compiled from experiments in Figs. 68. All experiments evaluated the effect of conditioning on responses to nondesensitizing concentrations of GABA (0.5 μM). It is clear that both exogenous and synaptic NMDAR conditioning produce GABAA receptor suppression, but synaptic receptor activation is far more efficient. In excitatory neurons, extrasynaptic receptor activation with 300 μM NMDA for 20 s and synaptic receptor activation with endogenous glutamate produced similar charge influx, but extrasynaptic receptor activation failed to significantly suppress GABAA responses. Exogenous application of 20 μM NMDA to all receptors achieved a degree of suppression similar to that of synaptic receptor activation but required ∼6-fold more NMDA-induced charge.

Fig. 9.

Synaptic currents are most efficient for suppressing GABAA responses. Charge influx, obtained by integrating NMDAR currents from various protocols used in Figs. 68, are plotted (filled bars, left axis) with the extent of GABA suppression (open bars, responses to 0.5 μM GABA, right axis) obtained from the same protocol.


Previous studies have reported NMDA-induced suppression of GABAA receptor responsiveness and have elucidated the downstream pathways of this cross-receptor plasticity. However, the triggering receptors and target receptors underlying physiological induction of suppression have remained unclear. Our results suggest that synaptic receptors are prominently involved in both the induction and the expression of cross-receptor plasticity. Synapses are thought to cluster receptors for the relevant neurotransmitter to the exclusion of other receptors. Thus, from the perspective of receptor localization, our result may be somewhat surprising. On the other hand, the involvement of synaptic receptors ensures precise location and timing of induction and expression of cross-receptor plasticity.

Gating and surface expression of several GABAA receptor subunits are modulated by kinase-mediated phosphorylation and phosphatase-mediated dephosphorylation (Inoue et al. 1986; Lu et al. 2000). Among kinases, CaMKII acts on GABAA β-subunits, leading to potentiated GABA responses (Hinkle and Macdonald 2003; Houston et al. 2008, 2009; Marsden et al. 2007). Dephosphorylation by Ca2+-activated phosphatases is important for suppression of GABA responses. The phosphatase calcineurin appears particularly important and participates in a form of long-term depression of GABAA responsiveness (Chen and Wong 1995; Huang and Dillon 1998; Lu et al. 2000; Stelzer and Shi 1994). Under our conditions, suppression of GABAA receptor responsiveness was clearly most robust. The conditions governing NMDAR-induced potentiation of inhibition remain unclear, although NMDAR activation potentiation may also involve different messengers and, at synapses, may include both pre- and postsynaptic changes (McBain and Kauer 2009).

The GABAA receptor subunits targeted by NMDA-induced suppression are not yet fully worked out. Our results suggest that synaptic GABAA receptors are strongly targeted, but they do not exclude effects on extrasynaptic GABAA receptors. In fact, because NMDAR EPSCs of solitary glutamatergic neurons suppress responses to exogenous GABA (Fig. 6), it is clear that at least some nonsynaptic GABAA receptors can be suppressed. Because extrasynaptic receptors are normally exposed to subsaturating GABA concentrations, extrasynaptic currents may be susceptible to both the alteration in sensitivity and the loss of functional receptors described herein. By contrast, peak IPSCs are probably mainly susceptible to the alteration in functional receptor number (Wang et al. 2003), as a result of the locally saturating cleft concentration of GABA (Maconochie et al. 1994). Quantitative IPSC suppression thus better mirrors the reduction of responses to 1 mM exogenous GABA (Figs. 35).

The strong involvement of synaptic receptors, especially synaptic NMDARs for triggering modulation, was somewhat surprising to us, because receptor cross talk would seem most likely in extrasynaptic regions where receptors would be most intermingled. Previous work has suggested that NMDARs are more efficient than voltage-gated Ca2+ channels at depressing GABAA currents (Stelzer and Shi 1994); our work extends this observation and demonstrates that synaptic NMDARs are more efficient than nonsynaptic NMDARs. There is no evidence for a significant difference in the Ca2+ permeability of synaptic vs. extrasynaptic receptors or of NMDARs of different subunit compositions (Garaschuk et al. 1996; Monyer et al. 1994). Therefore, to modulate GABAA receptors, synaptic NMDAR-induced intracellular signals must presumably travel some distance to target receptors. One possibility is that Ca2+ diffuses a sufficient distance to recruit an effector (e.g., calcineurin) near the GABAA receptor. Another possibility is that Ca2+ actions are local, near the site of NMDAR activation, but the effector acts over some distance to modulate the GABAA receptor.

Isolated GABAergic interneurons, functionally defined by autaptic IPSCs, exhibited NMDA-induced currents and charge influx comparable to solitary glutamatergic neurons (Fig. 8E). This observation could reflect known synaptic-like clustering of NMDARs in the absence of glutamatergic input (Rao and Craig 1997). Such clustering would be a substrate for localized sites of Ca2+ influx and effector recruitment, similar to those important at bona fide synaptic receptors in pyramidal neurons. An alternative possibility to explain the results in Fig. 8, D and E, is that unclustered extrasynaptic receptors at sufficiently high density can mediate GABAA receptor suppression similar to that induced by bona fide synaptic NMDARs. Regardless, this result suggests that interneurons are developmentally capable of assembling all of the downstream machinery for NMDAR-induced suppression of GABAA receptors in the absence of synaptic glutamate inputs. The results further suggest that NMDA-induced IPSC suppression could result in network inhibition (through disinhibition of interneurons) or excitation (through disinhibition of principal neurons).

Our approach was decidedly reductionist, with attendant caveats. We focused on the isolated autaptic preparation where all NMDA inputs onto a cell can be activated and blocked for clear isolation of extrasynaptic receptors (Fig. 8). Isolated cells also allowed us to help exclude potential contributions of NMDA-evoked neuromodulators (e.g., peptides, cannabinoids, adenosine), which could have effects that complicate interpretations in heterogenous networks. We found that several key aspects of suppression were also present in more conventional cultures (e.g., Figs. 1 and 5C), suggesting our results are not an artifact of the autaptic preparation. Because the fundamental phenomenon of NMDA-induced suppression of GABAA receptors has been observed in slice tissue (Lu et al. 2000) and other acute preparations (Chen and Wong 1995), we anticipate that our major conclusions regarding synaptic receptor involvement will apply to these more intact circuits, although direct demonstration of this awaits further work.

In summary, our results demonstrate that NMDAR-induced suppression of GABAA receptor responsiveness is a robust yet specific form of cross-receptor plasticity. At face value GABA suppression may augment Hebbian forms of plasticity such as NMDAR-dependent long-term potentiation of EPSPs (Wigstrom and Gustafsson 1983). Suppression could also exacerbate conditions of excitotoxicity. However, GABAA receptor responsiveness is suppressed in both glutamate neurons and GABA neurons. The relative strength of GABAA receptor suppression in the two cell types during physiological or pathophysiological activity may be important for the network effects of suppression. Furthermore, the preferential role of synaptic receptors for both inducing and expressing the suppression suggests that suppression is adapted for precise, localized changes in inhibition.


This work was supported by National Institutes of Health Grants GM47969, AA17413, and MH77791 (to C. F. Zorumski) and NS54174 and MH78823 (to S. Mennerick) and by the Bantly Foundation.


The authors declare that they have no potential conflict of interest in the studies herein.


We thank Ann Benz and Amanda Taylor for technical help with cultures and laboratory members for discussion.


View Abstract