Bidirectional homeostatic regulation of a depression-related brain state by GABAergic deficits and ketamine treatment

Abstract

Background

Major depressive disorder (MDD) is increasingly recognized to involve functional deficits in both GABAergic and glutamatergic synaptic transmission. To elucidate the relationship between these phenotypes we made use of GABA_A receptor γ2 subunit heterozygous (γ2^+/−) mice, which we previously characterized as a model animal with construct, face and predictive validity for MDD.

Methods

To assess possible consequences of GABAergic deficits on glutamatergic transmission we quantitated the cell surface expression of NMDA- and AMPA-type glutamate receptors and the function of synapses in the hippocampus and medial prefrontal cortex of γ2^+/− mice. In addition, we analyzed the effects of an acute dose of the experimental antidepressant ketamine on all these parameters in γ2^+/− vs. wild-type mice.

Results

Modest defects in GABAergic synaptic transmission of γ2^+/− mice resulted in a strikingly prominent homeostatic-like reduction in the cell surface expression of NMDA- and AMPA-type glutamate receptors, along with prominent functional impairment of glutamatergic synapses in the hippocampus and medial prefrontal cortex (mPFC). A single subanesthetic dose of ketamine lastingly normalized the glutamate receptor expression and synaptic function of γ2^+/− mice to wild-type levels, along with antidepressant-like behavioral consequences selectively in γ2^+/− mice. GABAergic synapses of γ2^+/− mice were potentiated by ketamine in parallel but only in mPFC.

Conclusions

Depressive-like brain states that are caused by GABAergic deficits involve a homeostatic-like reduction of glutamatergic transmission that is reversible by an acute, subanesthetic dose of ketamine, along with regionally selective potentiation of GABAergic synapses. The data merge the GABAergic and glutamatergic deficit hypothesis of MDD.

Introduction

Major depressive disorder (MDD) is a leading cause of total disability with limited treatment options that are often ineffective (1, 2). The etiology of MDD is poorly understood but thought to involve combinations of endogenous vulnerability factors and precipitating life events such as uncontrollable stress (3). Candidate vulnerability factors include diverse defects in γ-aminobutyric acid (GABA)ergic inhibitory neurotransmission, such as reduced concentrations of GABA (4–9), reduced expression of glutamic acid decarboxylase (GAD) (10, 11), reduced expression of GABA_A receptors (GABA_ARs) (12), and impaired function of GABAergic interneurons (10, 13, 14) (reviewed in 15, 16).

GABAergic deficit-induced changes in neural excitability (17) and reduced GAD-mediated conversion of glutamate to GABA may conceivably be causal for increased glutamate concentrations found in brain of MDD patients (18, 19). However, some studies of MDD point to reduced rather than increased brain content of glutamate (reviewed by 20), thereby suggesting a dynamic relationship between changes in GABAergic and glutamatergic transmission. Additional evidence suggestive of altered glutamatergic neurotransmission in MDD includes reduced expression and altered function of NMDARs (21, 22) and rapid therapeutic efficacy of subanesthetic doses of ketamine (23, 24). Ketamine exerts antidepressant activity as a non-competitive antagonist of NMDARs by ultimately enhancing glutamatergic synaptic transmission (25, 26). Importantly, ketamine is especially effective in otherwise treatment resistant forms of MDD associated with high anxiety (27). However, it is unclear how alterations in glutamatergic transmission and antidepressant efficacy of ketamine are functionally related to GABAergic deficits associated with MDD.

Stable functioning of neural networks in the face of rapid changes in neural excitability is critically dependent on homeostatic self-tuning mechanisms that, on a slower time course, preserve the balance of excitation and inhibition (E/I) and the average firing rates of neurons (28). Homeostatic mechanisms have most extensively been studied in cultured neurons. Of particular interest in the context of the present work is a slow form of homeostatic synaptic plasticity whereby pharmacologically induced hyperexcitability of cultured neurons is compensated by global scaling-down of glutamatergic synapses and scaling-up of inhibitory synapses (29–32).

Mice rendered hemizygous for the γ2 subunit gene (gabrg2) of GABA_ARs (γ2^+/− mice) have been extensively characterized as a model with construct, face and predictive validity for anxious MDD (reviewed in 15, 33). They exhibit a modest impairment of GABAergic transmission characterized by loss of the γ2 subunit in approximately15% of GABA_ARs averaged across brain regions, with the most prominent reductions in neocortex and hippocampus (−25 to −35%, 34). The γ2-lacking GABA_ARs are functionally impaired as indicated by their reduced channel conductance (12 vs. 28 pS) and failure to accumulate at synapses (34–36). Behaviorally, γ2^+/− mice exhibit signs of heightened anxiety, despair, anhedonia, and constitutive stress axis activation, and all these phenotypes are normalized by chronic treatment with the antidepressant desipramine (34, 37, 38). Cognitive alterations of γ2^+/− mice in ambiguous cue conditioning tests mimic emotional pattern separation defects associated with MDD (34, 39, 40). Moreover, a phospho-site mutation that increases the cell surface expression of γ2-GABA_ARs has antidepressant-like behavioral consequences (41). Thus, the γ2^+/− model lends support for a causative role of GABAergic deficits in the etiology of anxious MDD (15, 33).

Here we explored the consequences of GABAergic deficits on glutamatergic synapses. We found that γ2^+/− mice exhibit reduced cell surface expression and function of NMDARs and AMPARs, along with reduced expression of the synaptic adhesion molecule neuroligin 1 (NL1) and defects in the density and function of glutamatergic synapses in the hippocampus and medial prefrontal cortex (mPFC). Similar defects were observed in γ2^+/− cultured neurons. Moreover, treatment of γ2^+/− mice (or cultures) with a subanesthetic dose of ketamine resulted in lasting (≥ 3 day) enhancement and normalization of glutamate receptor (GluR) expression and glutamatergic synapse function. Thus, depression-related brain states of γ2^+/− mice involve a homeostatic-like reduction of glutamatergic transmission that can be normalized lastingly by the rapidly acting antidepressant ketamine. Notably, ketamine also potentiated the function of GABAergic synapses but only in anterior cingulate cortex (ACC). These data unite the GABAergic and glutamatergic deficit hypotheses of MDD, suggest that MDD may be caused by aberrant homeostatic plasticity, and provide novel insights into the synaptic mechanisms underlying antidepressant efficacy of ketamine.

Materials and Methods

For a more detailed and complete version of Material and Methods see Supplement 1.

Production and husbandry of mice

Two different GABA_AR γ2^+/− mouse lines were used as part of this study, with virtually identical germ line deletions of exon 8 of the gabrg2 locus. A first line of γ2^+/− mice was maintained on a 129X1/SvJ background as previously described (34, 38, 42). A second line was generated on the C57BL/6J background by mating γ2^f/f mice (43) with an oocyte-specific Cre line followed by outcrossing of the Cre transgene. Mice used for experimentation were littermates produced by crossing of γ2^+/− mice and WT mice. The 129X1/SvJ line was used for preparation and analyses of cortical cultures, as well as biochemical and electrophysiological analyses of brain slices. The C57BL/6J line was used for biochemical and behavioral experimentation involving ketamine treatment.

Drug treatments

For treatment of cultures the drugs were diluted or dissolved in culture media to the following final concentrations: ketamine (10 μM, Ketaject^®, Phoenix Pharmaceutical, Inc., St. Joseph, MO); 2-amino-5-phosphonovaleric acid (APV, 100 μM, Sigma-Aldrich, St. Louis, MO); bicuculline (20 μM, R&D Systems^™, Minneapolis, MN); Ro25-6981 (10 μM, Sigma-Aldrich). For treatment of mice (8–9 weeks old), ketamine (Ketaject^® diluted to 1 mg/ml in 0.9% saline) was administered at 10 mg/kg (biochemical and electrophysiological analyses) or 3 mg/kg (behavioral analyses) (i. p.) as previously described (26, 44).

Cell surface biotinylation

Cortical cultures from γ2^+/− and WT mice were generated from embryonic day 14–15 embryos (129X1/SvJ line), subject to cell surface biotinylation at 21 days in vitro (DIV) and purification using NeutrAvidin agarose beads (Thermo, Rockford, IL) as described (45). For biotinylation of brain slices we adapted the protocol of Terunuma et al. (46). The biotinylated proteins were quantitated by SDS/PAGE/western blot using an Odyssey® CLx infrared imager (LI-COR, Lincoln, NE). Amounts of cell surface biotinylated proteins were normalized to amounts of β-tubulin in total extracts quantitated on parallel gels.

Immunofluorescent staining of cortical cultures

Immunofluorescent staining of neurons employed glia-free cortical cultures prepared from embryonic day 14–15 embryos as previously described (47). The cells were fixed, permeabilized and stained at 21 DIV as described (47) using rabbit anti MAP2 (1:1000, Ab5622), guinea pig anti VGluT1 (1:500, LV1439669), mouse anti PSD95 (1:1500, #28879, all from Millipore), mouse anti gephyrin (1:500, #147111), and rabbit anti VGAT (1:1000, #131002, both from Synaptic Systems). Synaptic immunoreactivities were developed and quantified as described (47).

Electrophysiology

Coronal slices (350 μm) containing the dorsal hippocampus or anterior cingulate cortex were prepared using a vibratome (Leica VT1200S), from 7- to 13-week-old 129X1/SvJ mice (of either sex) in a solution containing (in mM): 210 sucrose, 7 D-glucose, 25 NaHCO₃, 1.25 NaH₂PO₄, 2.5 KCl, 1.3 Na-ascorbate, 3 Na-pyruvate, 0.5 CaCl₂, 7 MgCl₂, saturated with 95% O₂/5% CO₂. During recordings slices were perfused with (in mM): 50 sucrose, 119 NaCl, 26.2 NaHCO₃, 11 glucose, 2.5 KCl, 1 NaH₂PO₄, 2.5 CaCl₂, and 1.3 MgCl₂, saturated with 95% O₂/5% CO₂, and 0.5 μM TTX, 10 μM NBQX, 25 μM D-APV for mIPSCs, or 50 μM PTX for EPSCs, or 10 μM NBQX and 50 μM PTX were added to perfusate lacking MgCl₂ for NMDAR-mediated EPSCs. Internal solutions consisted of (in mM): 127 CsCl, 8 NaCl, 1 CaCl₂, 10 HEPES, 10 EGTA, 0.3 Na₃-GTP, and 2 Mg-ATP, pH 7.2 for mIPSCs, or 127 CsMeSO₄, 8 NaCl, 1 CaCl₂, 10 HEPES, 10 EGTA, 0.3 Na₃-GTP, and 2 Mg-ATP, 0.1 spermine, 5 QX-314, pH 7.2 for EPSCs. All recordings were obtained at V_h = 70 mV, unless otherwise indicated. EPSCs were evoked using borosilicate glass pipette stimulators positioned 50–200 μm away from the primary apical dendrite of the recorded cell. Recordings and analyses were performed using pCLAMP 10 software (Molecular Devices). Only cells with a stable access resistance throughout the recording period were included in the analysis.

Statistics

Statistical comparisons were performed using two-tailed student’s T-tests or ANOVAs followed by posthoc analyses as detailed in the text and Figure legends.

Results

GABAergic deficits of γ2-deficient neurons result in homeostatic downregulation of glutamate receptors

Chronic blockade of GABA_ARs in cultured neurons with bicuculline (BIC) results in homeostatic downregulation of NMDA and AMPAR function (29) (Fig. S1A in Supplement 1). Therefore, to begin to test whether similar changes in glutamate receptor (GluR) expression might occur in γ2^+/− mice we began our analyses in γ2^−/− and WT cultured cortical neurons (21 days in vitro). As a proxy for NMDARs we quantitated the cell surface expression of the obligatory subunit GluN1, as well as the GluN2B subunit, which is part of an NMDAR subtype that is implicated in mediating the detrimental effects of excess glutamate (48) and antidepressant effects of ketamine (49). In addition, we quantitated the cell surface expression of neuroligin 1 (NL1), a postsynaptic cell adhesion molecule that controls the accumulation of NMDARs at synapses (50). Lastly, we surveyed the expression of AMPARs using antibodies for GluA2/3 subunits. Using cell surface biotinylation assays we found that the plasma membrane accumulation of all four proteins was drastically reduced in γ2^−/− vs. WT cultures (Fig. 1A). Remarkably, the cell surface expression of the same four proteins was reduced also in γ2^+/− vs. WT cultures (Fig. 1B), with effect sizes comparable to those in γ2^−/− vs. WT cultures (Fig. 1A) and BIC treated WT cultures (Fig. S1A in Supplement 1). Total expression of GluN1 and GluA2/3 was unaffected in γ2^−/− vs. WT cultures and BIC-treated WT cultures, suggesting that the changes at the cell surface were due to impaired trafficking of receptors (Fig. S1A, B in Supplement 1). Therefore, a modest defect in GABA_AR function in γ2^+/− cultures (36, 51) leads to a profound downregulation of GluRs that is reminiscent of homeostatic scaling down of synapses induced by complete pharmacological blockade of GABA_ARs.

Figure 1. Analyses of cell surface expression of synaptic proteins in cortical cultures.

A γ2^−/− cultures showed significantly reduced surface levels of NL1, GluN1, GluN2B and GluA2/3, compared to WT neurons (γ2^−/− vs. WT: NL1, P < 0.001, n = 5–6; GluN1: P < 0.01, n = 5–7 cultures; GluN2B, P < 0.01, n = 12–13; GluA2/3, P < 0.01, n = 3–4, t-tests). B. The same markers were also reduced in γ2^+/− cultures (γ2^+/− vs. WT: NL1, P < 0.01, n = 4–5; GluN1, P < 0.001, n = 12; GluN2B, P < 0.01, n = 10–11; GluA2/3, P < 0.05, n = 14–15; t-tests). C. A time course of ketamine treatment of γ2^+/− cultures revealed increased GluN1 at 3 h of ketamine treatment (γ2^+/− vs. γ2^+/− Ket, P < 0.01, n = 9) and remained elevated thereafter (P < 0.05, n = 6–13 for both the 4.5 and 6 h time points, ANOVA, Dunnett’s tests). GluA2/3 was significantly increased first at 4.5 h and then remained elevated (γ2^+/− vs. γ2^+/− Ket, 4h, P < 0.05, 6h, P < 0.001, n = 11–12, ANOVA, Dunnett’s tests). Cell surface GluN1 and GluA2/3 levels showed a greater 3h ketamine response for GluN1 than GluA2/3 [F_(1,48) = 4.78, P < 0.05, ANOVA]. Data represent means ± SE. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Ketamine-induced reversal of glutamate receptor deficits

We hypothesized that down-regulation of GluRs was related to the depressive-like brain state of γ2^+/− mice (37, 38) and hence that it should be reversible by antidepressant concentrations of ketamine. We treated γ2^+/− cultures with ketamine (10 μM) for variable amounts of time and found that GluN1 cell surface expression was increased significantly within 3 h of treatment (Fig. 1C). Cell surface AMPARs were increased similarly but with about a 1.5 h delay. Ketamine also increased cell surface expression of NL1, with a time course similar to that of GluN1 (Fig. S2A in Supplement 1), which is consistent with a role of NL1 in cell surface trafficking of NMDARs (50). The effect of 10 μM ketamine on cell surface GluRs of γ2^+/− cultures was reproduced by the GluN2B specific antagonist Ro25-6981 (10 μM) and the competitive NMDAR antagonist APV (100 μM) (Fig. S2B, C, in Supplement 1).

Glutamatergic synapse density is reduced by GABA_AR deficits and normalized by ketamine treatment

To address whether GABA_AR deficits and ketamine impact the density of glutamatergic synapses we immunostained cultured cortical neurons for the pre- and postsynaptic markers vGluT1 and PSD95. The density of punctate immunoreactivity for both markers and their colocalization along dendrites was reduced in γ2^+/− vs. WT neurons (Fig. 2A–E). Treatment of γ2^+/− cultures with ketamine (10 μM, 6 h) normalized the expression and colocalization of vGluT1 and PSD95 to WT levels (Fig. 2A–E), while the size of immunoreactive puncta was unaffected by genotype and drug treatment (Fig. 2F). Thus, GABA_AR deficit-induced reductions in GluR and NL1 cell surface expression and their normalization by ketamine correlate with changes in the number of synapses, rather than a change in protein accumulation at individual synapses.

Figure 2. Analyses of glutamatergic synapses of γ2^+/− cortical cultures by immuno-staining.

A–C. Representative micrographs of neurons cultured form WT (A) and γ2^+/− (B, C) embryos (DIV 20–21) subjected to treatment with ketamine (Ket) (C) and immunostained for the dendritic marker MAP2 (A1, B1, C1, blue), PSD95 (A2, B2, C2, green) and vGluT1 (A3, B3, C3, red). Boxed dendritic segments including merged images are shown enlarged to the right of the panels (A4, B4, C4). Scale bar, 16.7 μm. D. The density of punctate dendritic immunoreactivity of tear-drop-shaped cells for PSD and VGluT1 was significantly reduced in γ2^+/− vs. WT cultures [PSD95, F_(2,33) = 4.11, VGluT1, F_(2,33) = 5.17, P < 0.05 for both proteins, ANOVA; WT vs. γ2^+/−, P < 0.05, n = 21–23 cells for both PSD95 and VGluT1, Tukey’s test]. Puncta densities γ2^+/− cultures treated with ketamine were reversed to WT levels (γ2^+/− vs. γ2^+/− Ket, P < 0.05, n = 21–23, for PSD95 and VGluT1, Tukey’s test). E. The fraction of VGluT1 puncta colocalized with PSD95 was reduced in γ2^+/− vs. WT cultures and restored by ketamine treatment of γ2^+/− cultures [F_(2,33) = 8.71, P < 0.01, ANOVA. γ2^+/− vs. WT and γ2^+/− vs. γ2^+/− Ket, P < 0.01, n = 21–23 for both comparisons, Tukey’s test). F. Puncta sizes were unaltered across conditions [PSD95, F_(2,33) = 5.05; VGluT1, F_(2,33) = 1.392, ANOVA, P, n.s.]. Data represent means ± SE. *, P < 0.05; **, P < 0.01, Tukey’s tests.

The GABA_AR γ2^+/− model shows increased sensitivity to anxiolytic- and antidepressant-like behavioral effects of ketamine

Assuming that ketamine induced surface expression of GluRs was related to its antidepressant activity we predicted that ketamine exerts increased antidepressant behavioral effects in γ2^+/− compared to WT mice. However, preliminary experiments designed to address behavioral effects of ketamine revealed, inexplicably, that γ2^+/− and WT mice maintained on a 129X1/SvJ strain background failed to show antidepressant-like behavioral responses to ketamine (not shown), reminiscent of other mouse strains that are insensitive to ketamine (52, 53). We therefore re-derived γ2^+/− mice on a C57BL/6J genetic background (Materials and Methods), which has been widely used for studies of ketamine. Interestingly, eight hours after a single dose of ketamine (3 mg/kg, i.p.) the anxiety-like phenotype of γ2^+/− mice in the Elevated Plus Maze was fully normalized to WT levels, without effects in WT mice (Fig. 3A). Moreover, ketamine had antidepressant-like consequences in the Forced Swim Test selectively in γ2^+/− but not WT mice (Fig. 3B). Thus, γ2^+/− mice are more sensitive than WT to the anxiolytic- and antidepressant-like behavioral effects of ketamine.

Figure 3. GABA_AR γ2^+/− mice exhibit increased behavioral sensitivity to anxiolytic- and antidepressant-like effects of ketamine.

Separate groups of GABA_AR γ2^+/− mice and WT littermates were injected with vehicle (saline) or ketamine (3 mg/kg, i.p.) and subjected to behavioral testing 8 h after treatment. A-C. In the Elevated Plus Maze, the % time spent on open arms (A) showed a significant overall treatment effect [F(1,43) = 8.17, P = 0.006, Two-way ANOVA] and a strong trend towards a genotype x treatment interaction [F(1,43) = 3.81, P = 0.057]. Posthoc analyses revealed an anxiety-like reduction in the % time spent on open arms in vehicle treated γ2^+/− vs. WT mice (P < 0.05, n = 12-16) consistent with the phenotype previously reported for γ2^+/− mice on the 129X1/SvJ background (34, 37). Ketamine treatment resulted in an anxiolytic-like increase in the % time on open arms in γ2^+/− [γ2^+/− Ket vs. Veh, P < 0.001, n = 12-14) but not WT mice (WT Ket vs. Veh, P > 0.05, n = 13-16, Fisher’s test). Similarly, analyses of % open arm entries (B) showed a significant overall treatment effect [F_(1,51) = 7.71, P = 0.008 Two-way ANOVA] and a significant anxiolytic-like effect of ketamine selectively in γ2^+/− (γ2^+/− Ket vs. Veh, P < 0.01, n = 12-14) but not WT mice (WT Ket vs. Veh. P > 0.05, n = 14-16, Fisher tests). The four groups were indistinguishable with respect to closed arm entries (C) [genotype: F_(1,51) = 0.52, P = 0.473, treatment: F_(1,51) = 0.71, P = 0.404, Two-way ANOVA]. D, E. In the Forced Swim Test, the latency to first immobility (D) showed an overall genotype effect [F(1,90) = 19.78, P < 0.001, n = 18-27, Two-way ANOVA] and a genotype x treatment interaction [F(1,90) = 4.45, P = 0.038]. Posthoc group comparisons showed a reduced Latency to First Immobility specifically in vehicle-treated γ2^+/− vs. WT mice (γ2^+/− Veh vs. WT Veh. P < 0.001, n = 18-25; WT Ket vs. Veh, P > 0.05, n = 20-21, t-tests) and an increased Latency to First Immobility in ketamine-treated γ2^+/− vs. vehicle-treated γ2^+/− mice (γ2^+/− Veh vs. γ2^+/− Ket. P < 0.01, n = 18-23, t-test), consistent with the depressive-like phenotype previously reported for γ2^+/− mice on the 129X1/SvJ background (37, 38) and with ketamine-induced changes in GluR expression and synapse function that were limited to or greater in γ2^+/− vs. WT mice. Measurements of total time spent immobile (E) showed an overall treatment effect [F(1,90) = 9.32, P = 0.003, n = 18-27, Two-way ANOVA] and an antidepressant-like drug effect specifically in ketamine vs. vehicle treated γ2^+/− but not WT mice (γ2^+/− Ket vs. Veh, P < 0.01, n = 18-25; WT Ket vs. Veh, P > 0.05, n = 20-21, Fisher’s tests).

GABA_AR deficits of γ2^+/− mice result in downregulation of glutamate receptors in vivo

To begin to assess the fate of glutamatergic synapses in vivo we quantified the cell surface expression of GluRs in acute brain slices of the hippocampus and the anterior cingulate/prelimbic cortex (ACC/PLC), which are among the brain areas with the greatest GABA_AR deficit in γ2^+/− mice (34) and essential for antidepressant drug-induced behavioral effects in rodents (54, 55). The cell surface AMPARs and NMDARs were drastically reduced in γ2^+/− vs. WT mice in both hippocampus and ACC/PLC (Fig. 4A).

Figure 4. Downregulation of glutamate receptors in γ2^+/− vs. WT mice is reversed by ketamine treatment.

A. The cell surface expression of GluN1 and GluA2/3 was reduced in γ2^+/− vs. WT mice both in hippocampus (GluN1, P < 0.001; GluA2/3, P < 0.05, n = 12–13 mice for both comparisons) and ACC/PLC (P < 0.05, n = 12–13, for both comparisons). GluN2B cell surface expression was reduced in hippocampus of γ2^+/− mice (γ2^+/− vs. WT, P < 0.05, n = 6–7) and trended lower in ACC/PLC (P = 0.15, n = 6–7). B. In γ2^+/− animals analyzed 24h after an acute dose of ketamine (10 mg/kg, i.p.) the cell surface expression of NMDARs and AMPARs in hippocampus was increased compared to vehicle-treated γ2^+/− controls (P < 0.05, n = 7–11, for the GluN1, GluN2B and GluA2/3 comparisons) to levels comparable to those of WT mice in (A) (γ2^+/−+Ket vs. WT, P, n.s., n = 9–12). In ACC/PLC, ketamine treatment of γ2^+/− mice resulted in increased expression of NMDARs (γ2^+/− Ket vs. Veh: GluN1, P < 0.05) but not AMPARs (GluA2/3, P, n.s., n = 8–9 for both comparisons). Cell surface GluN2B levels trended in the same direction as GluN1 but the effect was not significant (P = 0.12, n = 3). C. In hippocampus of γ2^+/− animals analyzed three days after ketamine treatment the cell surface expression of NMDARs and AMPARs remained increased compared to controls (P < 0.05, n = 10–12, for both GluN1 and GluA2/3). By contrast, in ACC/PLC, the GluN1 cell surface level had returned to base line (P, n.s., n = 8–9) but the GluA2/3 cell surface expression showed a strong trend of an increase (P = 0.08, n = 6/group) that was not yet evident at one day after treatment (B). D. In WT mice analyzed 24 h after ketamine treatment the expression of NMDARs and AMPARs was unchanged compared to vehicle controls in both hippocampus and ACC/PLC (WT Veh vs. Ket: P, n.s., n = 5–6, for all four comparisons). Data are from mice maintained on a C57BL/6J background. The genotype differences were reproduced with 129X1/SvJ mice (Fig. S1C in Supplement 1). Data represent means ± SE. *, P < 0.05; ***, P < 0.001, t-tests.

A subanesthetic dose of ketamine given to γ2^+/− mice normalizes NMDAR cell surface expression in vivo

To examine whether ketamine affected expression of GluR in vivo we treated γ2^+/− mice with ketamine (C57BL/6J, strain, 10 mg/kg, i.p.) and 24 h later harvested hippocampal and ACC/PLC brain slices for quantification of cell surface proteins. Ketamine treatment of γ2^+/− mice resulted in significant upregulation and normalization of cell surface NMDARs in both hippocampus and ACC/PLC (Fig. 4B). Expression of the GluN2B subunit appeared increased to similar levels as GluN1, although the effect was more variable and significant only in hippocampus. Total expression of NMDARs remained unaltered, indicating that ketamine acted posttranslationally to normalize impaired cell surface accumulation of NMDARs. Unlike NMDARs, cell surface AMPARs of γ2^+/− mice were upregulated by ketamine only in hippocampus and unaffected in ACC/PLC. Moreover, this drug effect in hippocampus involved increased total expression of AMPARs, rather than merely a change in cell surface accumulation (Fig. 4B). Notably, comparing the hippocampal cell surface expression of GluRs of ketamine treated γ2^+/− and drug naïve WT mice (Fig. 4A and B, data normalized to values of drug naïve γ2^+/− mice) indicated that ketamine fully restored the expression of GluRs from γ2^+/− to WT levels [γ2^+/−+Ket vs. WT: P, n.s., n = 9–12, for both GluN1 and GluA2/3, ANOVA, Tukey test).

Given the lasting therapeutic effects of ketamine in patients (23) we further assessed whether the drug effects on GluR expression seen one day after treatment remained measurable three days after treatment. Interestingly, in hippocampus, both GluN1 and GluA2/3 remained significantly elevated compared to vehicle-treated γ2^+/− mice three days after treatment (Fig. 4C). By contrast, in PFC, the effects of ketamine on GluN1 were no longer detectable. Instead, there was a strong trend for upregulation of GluA2/3 that was not observed one day after treatment (Fig. 4 B, C). Thus, a subanesthetic dose of ketamine can lastingly reverse homeostatic downregulation of glutamatergic synapses in the hippocampus, induced by impaired GABAergic transmission for at least three days post treatment. Moreover, ketamine induced augmentation of AMPAR expression in PFC may be delayed relative to that of NMDAR and slower than in hippocampus.

Importantly, ketamine had no effect on GluR cell surface expression in WT mice, independent of brain region analyzed (Fig. 4D, 1-day treatment), and consistent with the selective behavioral effects of ketamine in γ2^+/− but not WT mice described in Fig 3. Notably, downregulation of GluRs was evident in γ2^+/− mice of both genetic backgrounds [Fig. 4A (C57BL/6J) and Fig. S1C in Supplement 1 (129X1/SvJ)], suggesting that differential behavioral sensitivity of the two genetic backgrounds to the effects of ketamine was due to strain differences acting downstream of altered GluR expression.

GABA_AR deficits reduce the number of functional glutamatergic synapses in vivo and this defect is reversed by a single dose of ketamine

We next assessed functional defects in glutamatergic transmission using voltage clamp recordings of CA1 pyramidal cells in hippocampal slices. The frequency of spontaneous excitatory postsynaptic currents (sEPSCs) was drastically reduced in γ2^+/− vs. WT cells, while the amplitude was unaffected (Fig. 5A–C). Ketamine administered to mice 24 h prior to recordings normalized the sEPSC frequency of γ2^+/− mice to WT levels and had no effect in WT mice (Fig. 5C). Neither ketamine nor genotype affected the amplitude of sEPSCs (Fig. 5B). Consistent with results obtained by immunostaining of cultures (Fig. 2), these findings suggest that GABA_AR deficits and ketamine affect glutamatergic transmission through a change in the number of functional synapses rather than through changes in the abundance of AMPARs at synapses.

Figure 5. Downregulation of glutamatergic synaptic inputs to γ2^+/− CA1 pyramidal neurons is reversed by ketamine.

A–C. sEPSC recordings from CA1 pyramidal neurons of γ2^+/− and WT mice injected 24 h earlier with saline (Veh) or ketamine (Ket). Representative traces are shown in (A), with summary quantification of sEPSC amplitude in (B) and frequency in (C). Note the significant decrease in frequency of sEPSC recorded from vehicle-treated γ2^+/− vs. WT mice that was normalized by ketamine treatment (genotype x treatment interaction for sEPSC frequency, F_(1,86) = 4.473, P = 0.037, Two-way ANOVA; P < 0.01, n = 12–41 cells, for both group comparisons, Tukey’s test). D–G. AMPAR EPSCs recorded from CA1 pyramidal cells evoked by stimulation of the SC (D, E) or TA path (F, G). Schematics with sites of stimulation and recording and average traces at progressively larger stimulation intensities are shown in (D) and (F). AMPAR responses were reduced in γ2^+/− vs. WT mice at SC (E) and TA (G) synapses. Moreover, ketamine treatment restored synaptic responses of γ2^+/− mice to WT levels [(E) # WT Veh vs. γ2^+/− Veh F_(1,138) = 5.391, P = 0.022, Two-way ANOVA; P < 0.05, n = 9–17 cells, Tukey’s test; (G) ### WT Veh vs. γ2^+/− Veh, F_(1,102) = 23.78, P < 0.001, Two-way ANOVA, P < 0.05, Tukey’s test). H–K. NMDAR EPSCs recorded from CA1 pyramidal cells evoked by stimulation of the SC (H, I) or TA path (J, K). Schematics with sites of stimulation and recordings, along with average sample traces at progressively larger stimulation intensities are shown in (H) and (J), with summary quantification in (I) and (K). NMDAR responses in γ2^+/− mice were reduced compared to WT at both SC (I) and TA (K) synapses. Ketamine treatment restored the synaptic NMDAR responses of γ2^+/− mice to WT levels [(I) # WT Veh vs. γ2^+/− Veh, F_(1,89) = 4.952, P = 0.029, Two-way ANOVA; ### γ2^+/− Veh vs. γ2^+/− Ket, F_(1,95) = 12.31, P < 0.001, Two-way ANOVA; (K) ### WT Veh vs. γ2^+/− Veh F_(1,90) = 15.55, P < 0.001, Two-way ANOVA; ### γ2^+/− Veh vs. γ2^+/− Ket F_(1,96) = 12.55, P < 0.001, Two-way ANOVA; n = 8–9 cells for all groups in (I) and (K)]. Data represent means ± SE. *, P < 0.05; **, P < 0.01.

Chronic stress has been shown to preferentially affect temporoammonic (TA) synapses onto CA1 pyramidal neurons rather than Schaffer collateral (SC) synapses (56). To further characterize the synaptic deficits of γ2^+/− CA1 pyramidal cells we recorded AMPAR- and NMDAR EPSCs evoked by selective stimulation of either the SC or TA pathway. AMPAR EPSC amplitudes were significantly reduced in γ2^+/− vs. WT slices, with a more pronounced effect at TA synapses (Fig. 5D–G). Moreover, ketamine treatment of γ2^+/− mice normalized SC-evoked AMPAR EPSCs to WT levels, without corresponding effects in WT mice (Fig. 5E). At TA synapses, ketamine potentiated the AMPAR responses, with greater effects in γ2^+/− than WT mice (WT Ket as % WT Veh vs. γ2^+/− Ket as % γ2^+/− Veh, F_(1,163) = 17.73, P < 0.0001, Two-way ANOVA) (Fig. 5F, G). Similar to AMPAR EPSCs, NMDAR currents of SC-CA1 and TA-CA1 synapses of γ2^+/− mice were impaired at baseline, and restored to WT levels by ketamine pretreatment (Fig. 5H–K).

To further examine the idea that the reduced sEPSC frequency of γ2^+/− neurons reflected a reduction in the density of synapses (Fig. 5A, B) we assessed the synaptic release probability by measuring paired pulse ratios (PPRs) from SC and TA path-stimulated pyramidal cells (Fig. S3 in Supplement 1). The PPRs were unaffected by genotype for both types of synapses and increased by ketamine selectively at TA-CA1 synapses. Thus, genotype- and treatment-dependent alterations of glutamatergic transmission primarily reflect changes in synapse number rather than release probability. Further, neither genotype nor ketamine treatment affected the rectification of AMPAR responses evoked by SC or TA pathway stimulation (Fig. S4 in Supplement 1), indicating that the calcium permeability of AMPARs does not play a significant role in the forms of plasticity examined here.

Ketamine enhances GABAergic synaptic inhibition in frontal cortex

Ketamine has potent seizure-suppressing effects in animal models of epilepsy and in patients (57–59). We therefore wondered whether ketamine-induced potentiation of glutamatergic transmission was accompanied by an increase in GABAergic inhibition. Comparison of γ2^+/− and WT cultured cortical neurons by immunofluorescent staining for the pre- and postsynaptic markers vesicular GABA transporter (VGAT) and gephyrin revealed a significant reduction in the size and density of punctate gephyrin immunoreactivity in γ2^+/− vs. WT cultures, with both of these defects fully reversed by 6 h treatment of γ2^+/− cultures with 10 μM ketamine (Fig. 6 A–G). A modest but significant increase in the density of gephyrin puncta was also observed in WT cultures (Fig. 6F). VGAT staining of γ2^+/− cultures was increased by ketamine in parallel with gephyrin (Fig. 6G), suggesting that ketamine potentiated the postsynaptic apparatus while also increasing the number of GABAergic synapses in γ2^+/− cultures. Consistent with this interpretation, the colocalization of VGAT and gephyrin was unaffected by genotype or ketamine treatment (Fig. 6H). As expected, the amplitude of miniature inhibitory postsynaptic currents (mIPSCs) recorded from CA1 and L2/3 ACC pyramidal cells of γ2^+/− brain slices was reduced compared to WT, while the frequency was unaltered (Fig. 6 I–L). Therefore, γ2^+/− neurons display a defect in GABAergic inhibition that is not compensated for by any process resembling homeostatic scaling up of inhibitory synaptic strength seen following prolonged treatment of cultured neurons with BIC. Curiously, we found that ketamine had no effect on mIPSCs recorded from CA1 neurons of γ2^+/− mice (Fig. 6 I, J). However, ketamine fully restored the amplitude of mIPSCs recorded from γ2^+/− L2/3 ACC pyramidal cells to WT levels, along with a prominent increase in the frequency of mIPSCs observed selectively in γ2^+/− but not WT mice (Fig. 6 K, L). Ketamine did not affect the amplitude or frequency of mIPSCs of WT mice, consistent with our other observations that showed that ketamine effects on GluR expression, synapse density and emotional behavior were enhanced or specific for γ2^+/− mice representing the pathological condition. Notably, glutamatergic synapses of L2/3 pyramidal cells were downregulated in γ2^+/− mice, and restored by ketamine administration to WT levels (Fig. S5A, B in Supplement 1), similar to results obtained with CA1 pyramidal cells. Thus, GABAergic synapses in ACC of γ2^+/− mice are potentiated by ketamine both pre- and post-synaptically, in concert with restoration of glutamatergic synapses and antidepressant behavioral effects.

Figure 6. Characterization of GABAergic synapse deficits and their reversal by ketamine in γ2^+/− cultured cortical neurons, and CA1 and L2/3 ACC of γ2^+/− mice.

A–H. Cortical cultures (DIV21) prepared from WT (A, C) and γ2^+/− (B, D) embryos were either untreated (A, B) or treated with ketamine (C, D) (10 μM, 6 h) and subjected to immunofluorescent staining for gephyrin (green, A1–D1) and VGAT (red, A2–D2). Colocalization in merged images is shown in yellow (A3–D3) with enlarged dendritic segments depicted to the right. Scale bar, 16.7 μm. E. Quantitation of punctate immunoreactivity in dendritic segments of pyramidal neurons showed that the size of gephyrin clusters was reduced in γ2^+/− vs. WT neurons and restored to WT levels by ketamine treatment [F_{(3, 73)} = 5.3, P < 0.01, ANOVA, γ2^+/− vs. WT, P = 0.05; γ2^+/− vs. γ2^+/− Ket, P < 0.01, WT vs. WT Ket, P, n.s., n = 18–21, Bonferroni]. F. The density of punctate gephyrin staining per 40 μm was significantly reduced in γ2^+/− vs. WT cultures [F_{(3, 73)} = 64.9, P < 0.001, ANOVA; γ2^+/− vs. WT, P < 0.05, n = 18–21, Tukey’s test] and increased by ketamine selectively in γ2^+/− cultures (γ2^+/− vs. γ2^+/− Ket, P < 0.001, WT vs. WT Ket, P, n.s., n = 18–21, both comparisons, Tukey’s test). G. The density of VGAT puncta along 40 μm segments of dendrite was increased by ketamine independent of genotype [F_(1/73) = 16.4, P < 0.001, ANOVA; WT vs. WT Ket, P < 0.01, γ2^+/− vs. γ2^+/− Ket, P < 0.05, n = 18–21, t-tests). H. The fraction of gephyrin puncta colocalized with VGAT was unaltered across conditions [F_(3,73) = 1.99, P = 0.12, ANOVA]. I–L. Representative traces of mIPSC recordings from CA1 (I) and L2/3 ACC pyramidal neurons (K) of WT and γ2^+/− mice injected with saline (Veh) or Ket 24 h prior to recording. (J) Quantification of mIPSC amplitude and frequency for CA1 neurons showed a significant decrease in amplitude in γ2^+/− vs. WT mice, regardless of treatment (mIPSC amplitude by genotype, F_(1,45) = 18.02, P = 0.001, Two-way ANOVA, P < 0.05, n = 10–13 cells; Tukey’s test). (L) Quantification of mIPSC amplitude and frequency for ACC neurons showed a significant decrease in amplitude in γ2^+/− vs. WT mice that was reversed by ketamine treatment of γ2^+/− mice (mIPSC amplitude, genotype x treatment interaction, F_(1,25) = 7.879, P < 0.01, n = 6–8 cells; Two-way ANOVA; treatment comparison F_(1,25) = 9.388, P = 0.005, Bonferroni). Data represent means ± SE. * P < 0.05, ** P < 0.01; *** P < 0.001.

Discussion

We have shown here that the GABAergic deficit-induced depression-related brain state of γ2^+/− mice involves a homeostatic-like reduction of glutamatergic transmission. A modest reduction in GABAergic transmission in γ2^+/− mice, consisting of a ~20% reduction in the mIPSC amplitude of principal cells, leads to a robust ~30–50% reduction in cell surface NMDA and AMPARs and a ~50% reduction in sEPSC frequency. Hyperexcitability of cultured neurons induced by chronic BIC treatment is known to result in homeostatic scaling-down of glutamatergic synapses in concert with scaling-up of inhibitory synapses (29, 30) and increased expression of postsynaptic GABA_ARs (31). Compared to complete (but transient) blockade of GABA_ARs by BIC treatment of cultures, γ2^+/− mice and cultures display only a modest reduction of GABA_AR function. However, unlike in BIC-treated cultures, inhibitory synapses are not strengthened in γ2^+/− mice, possibly due to the limiting amounts of γ2-GABA_ARs. It is conceivable then that homeostatic mechanisms of γ2^+/− mice that serve to restore the balance of excitation and inhibition ended up exacerbating downregulation of glutamatergic transmission. Similar homeostatic mechanisms are likely to operate under conditions of chronic or repeated stress, which are used to model depression in rodents, involve downregulation of AMPARs and NMDARs (60), result in increased behavioral responsiveness to ketamine (61), and hence involve changes in glutamatergic transmission comparable to those observed in the γ2^+/− model. Importantly, chronic stress also increases the chloride reversal potential, which renders GABAergic inhibition ineffective or excitatory (62–64). Thus, stress-induced downregulation of GluRs may be a consequence of excessive excitatory drive and impaired upregulation of inhibition analogous to mechanisms observed in the γ2^+/− model. Our findings merge the GABAergic (15, 16) and glutamatergic (21, 60) deficit hypotheses of MDD and provide a novel mechanism for how these two neurotransmitter systems interact in the etiopathology of MDD.

Second, we showed that the GABAergic deficit-induced adaptations of glutamatergic synapses in γ2^+/− mice are reversed by an acute subanesthetic dose of ketamine. Normalization of GluR expression and synapse density was also observed in γ2^+/− cultures in the continuous presence of ketamine or APV or Ro25-6981, while WT cultures were unaffected by ketamine or Ro25-6981. These observations suggest that downregulation of NMDARs in γ2^+/− cultures and mice involves excessive or untimely activation of NMDARs, most likely due to chronically increased release of glutamate. Recent evidence indicates that the initial effect of a subanesthetic dose of ketamine is to block a subpopulation of NMDARs that are active at rest (44), followed by increased expression of AMPARs (26). Our data extend these findings and show that initial antagonistic effects of ketamine on NNMDARs are followed by restoration of previously compromised expression and function of postsynaptic NMDARs, in addition to increased expression and function of AMPARs. Moreover, the effects of ketamine were greatly enhanced or specific for γ2^+/− vs. WT mice, i.e. under conditions representing the pathological condition.

Third, we showed that in concert with enhanced glutamatergic transmission ketamine potentiates the pre- and postsynaptic function of GABAergic inhibitory synapses of L2/3 ACC pyramidal cells of γ2^+/− mice. Notably, this drug effect occurred despite the impairment of inhibitory synapses that is evident in untreated γ2^+/− mice. An enduring potentiation of synaptic inhibition is consistent with the known seizure suppressing effects of ketamine (57–59).

Fourth, downregulation of cell surface NMDARs in γ2^+/− neurons and its reversal by ketamine were paralleled by corresponding changes in the cell surface expression of NL1, a synaptogenic cell adhesion protein that contributes to activity-dependent balancing of glutamatergic and GABAergic synaptic transmission (65) and controls the synaptic accumulation of NMDARs (50). Thus, aberrant trafficking of NL1 may play a key role in maladaptations observed in γ2^+/− mice and their reversal by ketamine.

GABA_AR-deficit-induced downregulation of cell surface GluRs is reminiscent of chronic stress-induced downregulation of AMPARs and NMDARs in mPFC (66–68) and downregulation of AMPARs selectively at TA-CA1 synapses (56, 69). Similar to selective stress-induced impairment of TA-CA1 synapses, the functional deficit in γ2^+/− mice was greater at TA-CA1 than at SC-CA1 synapses (TA-CA1 γ2^+/− as % WT vs. SC-CA1 γ2^+/− as % WT, F_(1,120) = 5.714, P = 0.0184, Two-way ANOVA) (Fig. 5E vs. G). TA-CA1 synapses map to distal apical dendrites of CA1 pyramidal cells that are targets of somatostatin (SST)-positive oriens-lacunosum moleculare (O-LM) cells (70). SST interneurons are highly sensitive to stress and functionally impaired in MDD patients (71). Thus, endogenous or stress-induced GABAergic deficits may be key for stress-induced downregulation of CA1 pyramidal neuron glutamatergic synapses. Elucidating the signaling pathways underlying homeostatic plasticity of GABAergic and glutamatergic synaptic transmission should lead to novel approaches for the treatment of MDD.