Compromising the phosphodependent regulation of the GABA_AR β3 subunit reproduces the core phenotypes of autism spectrum disorders

Significance

Alterations in the efficacy of neuronal inhibition mediated by GABA_A receptors (GABA_ARs) containing β3 subunits are continually implicated in autism spectrum disorders (ASDs). In vitro, the plasma membrane stability of GABA_ARs is potentiated via phosphorylation of serine residues 408 and 409 (S408/9) in the β3 subunit, an effect that is mimicked by their mutation to alanines. Here, we created a mouse in which S408/9 have been mutated to alanines (S408/9A). S408/9A mice exhibited altered dendritic spine structure, increased repetitive behavior, decreased social interaction, and an epileptic phenotype. Thus, mutation of S408/9 reproduces the core deficits seen in humans with ASDs. Collectively, our results suggest that alterations in phosphorylation and/or activity of β3-containing GABA_ARs may directly contribute to the pathophysiology of ASDs.

Abstract

Alterations in the efficacy of neuronal inhibition mediated by GABA_A receptors (GABA_ARs) containing β3 subunits are continually implicated in autism spectrum disorders (ASDs). In vitro, the plasma membrane stability of GABA_ARs is potentiated via phosphorylation of serine residues 408 and 409 (S408/9) in the β3 subunit, an effect that is mimicked by their mutation to alanines. To assess if modifications in β3 subunit expression contribute to ASDs, we have created a mouse in which S408/9 have been mutated to alanines (S408/9A). S408/9A homozygotes exhibited increased phasic, but decreased tonic, inhibition, events that correlated with alterations in the membrane stability and synaptic accumulation of the receptor subtypes that mediate these distinct forms of inhibition. S408/9A mice exhibited alterations in dendritic spine structure, increased repetitive behavior, and decreased social interaction, hallmarks of ASDs. ASDs are frequently comorbid with epilepsy, and consistent with this comorbidity, S408/9A mice exhibited a marked increase in sensitivity to seizures induced by the convulsant kainic acid. To assess the relevance of our studies using S408/9A mice for the pathophysiology of ASDs, we measured S408/9 phosphorylation in Fmr1 KO mice, a model of fragile X syndrome, the most common monogenetic cause of ASDs. Phosphorylation of S408/9 was selectively and significantly enhanced in Fmr1 KO mice. Collectively, our results suggest that alterations in phosphorylation and/or activity of β3-containing GABA_ARs may directly contribute to the pathophysiology of ASDs.

GABA_A receptors (GABA_ARs) are Cl⁻ selective ligand-gated ion channels that mediate phasic and tonic inhibition in the adult brain. Consistent with their roles in limiting neuronal excitability, benzodiazepines, barbiturates, general anesthetics, and neurosteroids exert their anxiolytic, anticonvulsant, hypnotic, and sedative effects via potentiating GABA_AR activity (1). GABA_ARs are heteropentamers constructed from α1–6, β1–3, γ1–3, δ, ε, θ, and π subunits. Phasic inhibition is principally mediated by receptors assembled from α1–3, β1–3, and γ2 subunits, whereas those receptors that mediate tonic inhibition contain α4–6, β1–3, and δ subunits (2). Studies using KO mice have shown that the β3 subunit is an essential component of receptor subtypes that mediate phasic and tonic inhibition (3). Together with the Fmr1 gene (Fragile X mental retardation), mutations to the 15q11–13 locus, where the GABA_AR β3 gene resides, are the leading monogenetic causes of autism spectrum disorders (ASDs) (4). Moreover, β3 subunit mutations have been described in seizure disorders, and alterations in subunit expression levels have also been reported in ASDs (3, 5).

In vitro studies have revealed that the β3 subunit plays a critical role in regulating the plasma membrane accumulation and synaptic targeting of GABA_ARs via phosphorylation of the intracellular serine residues 408 and 409 (S408/9) (6, 7). S408/9 are substrates of cAMP-dependent PKA, PKC, Ca²⁺-calmodulin type 2-dependent protein kinases (Cam KIIs), and cGMP-dependent protein kinase, and they are principally dephosphorylated by protein phosphatase 2A (8). S408/9 are the principal mediators of high-affinity binding to the clathrin adaptor molecule AP2 within the β3 subunit, and thereby facilitate GABA_AR endocytosis (9). Phosphorylation of S408/9 reduces the affinity of the β3 subunit for AP2 by 100-fold, and mutation of S408/9 to alanine residues (S408/9A) has been shown to mimic the effects of phosphorylation on AP2 binding to the β3 subunit (9, 10). Accordingly, overexpression of the mutant β3 S408/9A subunit in cultured hippocampal neurons leads to an increase in the number and size of inhibitory synapses (7).

Studies in animal models of ASDs have reported modifications in the expression levels of some GABA_AR mRNAs and proteins (11, 12). However, the mechanisms underlying these alterations in subunit expression and if they contribute to ASDs remain to be addressed. Therefore, in this study, we have analyzed the role that modified β3 subunit phosphorylation may play in the pathophysiology of ASDs. To test this role, we created a mouse in which the principal sites of phosphodependent regulation within the receptor β3 subunit, S408/9, have been mutated to S408/9A, a mutation that mimics the effects of their phosphorylation. S408/9A mice exhibited increased phasic but decreased tonic inhibition events, which correlated with alterations in the membrane stability of the receptor subtypes that mediate these distinct forms of inhibition. S408/9A mice exhibited alterations in dendritic spine structure, increased repetitive-like behavior, and decreased social interaction, which are hallmarks of ASDs. Therefore, our results provide evidence that alterations in the activity of GABA_ARs containing β3 subunits directly contribute to ASDs.

Results

Creation of a S408/9A Knock-In Mouse.

Alterations in the efficacy of GABAergic inhibition mediated by β3-containing GABA_ARs are strongly linked to ASDs. Phosphorylation of GABA_ARs regulates their exocytosis and endocytosis, and thereby their residence time on the neuronal plasma membrane and accumulation at inhibitory synapses (8). Central to these regulatory processes is the phosphorylation of S408/9 in the β3 subunit, which reduces the affinity of GABA_ARs for the clathrin adaptor protein AP2, as measured by using synthetic peptides corresponding to the β3 subunit and purified AP2 (9, 13). The significance of these findings has recently been questioned by studies that suggest that the arginine residues flanking S408/9 are the principal determinants of AP2 binding in the β3 subunit (14). To examine the significance of S408/9 further, we expressed the intracellular domain of the β3 subunit as a GST fusion protein (GSTβ3) or a fusion protein in which the respective residues were mutated to alanines (GSTβ3S408/9A) in Escherichia coli. The respective fusion proteins were then phosphorylated in vitro using purified PKC to final stoichiometries of ∼0.35 and ∼0.03 mol/mol, respectively, and then exposed to the μ2 subunit of AP2 (15) (Fig. S1A). Phosphorylation of GSTβ3 significantly reduced μ2 binding, whereas phosphorylation of GSTβ3S408/9A was without effect. Likewise GSTβ3S408/9A bound significantly lower levels of μ2 compared with GSTβ3. Collectively, these results suggest a key role for S408/9 and their phosphorylation in determining the affinity of GABA_ARs for AP2 (Fig. S1 A and B).

Fig. S1.

Characterization of β3 phosphorylation and generation of S408/9A mice. (A) GST fusion proteins encoding the major intracellular domain of the β3 subunit (GST-β3) and a mutant fusion protein in which residues S408/9 were mutated to alanines (GSTβ3S408/9A) were purified from E. coli. Fusion proteins were subjected to in vitro phosphorylation using purified PKC. GSTβ3 was phosphorylated to 0.35 mol/mol, and GSTβ3S4089A was purified to 0.03 mol/mol. Phosphorylated fusion proteins were then exposed to brain lysates and immunoblotted with μ2 antibodies. Coomassie staining of the respective samples and an autoradiogram of samples exposed to ³²P–γ-ATP and PKC are shown. The level of μ2 binding was normalized to fusion proteins that had been incubated with heat-inactivated PKC. (B) AP2 binding to GSTβ3 and GSTβ3S408/9A was directly compared as outlined above. (C) Schematic diagram of the targeting construct for generating β3 S408/9A knock-in mice. The β3 S408/9A mice were generated by homologous recombination in ES cells. A targeting plasmid was constructed that contained a floxed neomycin (Neo) selection cassette in intron 8 of β3 genomic DNA and a flanking thymidine kinase-negative selection marker (TK). (D) Following Cre-mediated excision of the selection cassette in vivo, the presence of the mutant codons in exon 9 was confirmed by DNA sequencing. (E) PCR product of WT and β3 S408/9A DNA samples detecting the presence or absence of the remaining loxP site. (F) No gross morphological changes were observed in the hippocampus between WT and S408/9A mice as determined by Nissl staining. (Scale bar: 100 μm.)

To assess the significance of β3 subunit phosphorylation in determining the efficacy of GABAergic inhibition, we created a knock-in mouse in which S408/9 were mutated to S408/9A using homologous recombination in ES cells (Fig. S1 C and D). Mutation of the respective codons in exon 9 of the β3 subunit was confirmed by DNA sequencing (Fig. S1 C and D), and the respective mice were backcrossed on the C57BL/6J background in excess of 10 generations. S408/9A homozygotes were viable and bred normally and did not exhibit any overt phenotypes. Likewise, there did not appear to be any gross anatomical changes in the structure of the hippocampus (Fig. S1E). In hippocampal slices from WT mice, treatment with the PKC activator phorbol 12,13-dibutyrate produced a large increase in pS408/9 immunoreactivity, an effect not replicated in S408/9A mice (Fig. 1A). To control for possible global changes in GABA_AR phosphorylation mice, we analyzed phosphorylation of Y367 in the γ2 subunit, an accepted substrate of Src family kinases (16). Treatment of slices with vanadate, an inhibitor of tyrosine phosphatases, induced similar increases in Y367 phosphorylation in WT and S408/9A mice (Fig. 1A).

Fig. 1.

GABA_AR expression in S408/9A mice. (A) Treatment of hippocampal slices from WT mice with 100 nM phorbol 12,13-dibutyrate (PDBU) increased pS408/9 immunoreactivity consistent with published studies demonstrating that both residues are substrates of PKC (8). In contrast, pS408/9 immunoreactivity was not detected in S408/9A mice. (B) Biotinylation revealed that the surface levels of the β3 subunits, along with the synaptic α2 subunits, were increased in the S408/9A mice, whereas surface accumulation of GABA_ARs containing α4 subunits was decreased. Hippocampal slices were subjected to biotinylation, followed by immunoblotting with anti-α2, anti-α4, and anti-β3 subunit antibodies. The ratio of surface/total (S/T%WT) was measured in S408/9A mice and WT control mice, and values were then normalized to WT controls (100%, red line). Total expression levels were also compared between genotypes (T%WT). *Significantly different from control (P ˂ 0.05, t test; n = 5–6 mice of each genotype). (C) Forty-micron hippocampal slices were stained with antibodies against the α2 and α4 subunits and gephyrin fluorescent-conjugated secondary antibodies, followed by confocal microscopy. The number of α2/gephyrin-positive puncta and α4 puncta was then compared within the dentate gyrus between genotypes. Preventing phosphorylation of the β3 subunit increases the number of inhibitory synapses in the dentate gyrus of the hippocampus but reduces α4 puncta immunoreactivity. *Significantly different from control (P < 0.001, t test; n = 3 mice of each genotype). (Scale bars: 10 μm.)

Characterization of GABA_AR expression levels and synaptic targeting in S408/9A mice.

Next, we examined the effects of the S408/9A mutation on the cell surface accumulation of the GABA_ARs using biotinylation. The S408/9A mutation increased the cell surface expression levels of the β3 subunit to 145 ± 8% of WT (P < 0.05, n = 4 mice of each genotype; Fig. 1B) without modifying total subunit expression levels (P > 0.05, n = 4 mice of each genotype; Fig. 1C). In the dentate gyrus, phasic inhibition is mediated by GABA_AR assembled from of α1/2, β2, β3, and γ2 subunits, whereas subtypes containing α4–6, β2, β3, and δ subunits are responsible for tonic current. Therefore, we assessed the effects of the S408/9 mutation on the plasma membrane accumulation of the receptor α2 and α4 subunits. The plasma membrane levels of the α2 subunit were increased in S408/9A mice to 119.0 ± 5.2% of control (P < 0.05, n = 5–6 mice of each genotype; Fig. 1B). In contrast, the α4 subunit was decreased to 81.0 ± 4.5% of levels seen in WT (P < 0.05, n = 6 mice of each genotype; Fig. 1B). These changes in cell surface accumulation were mirrored by parallel changes in increased total expression levels of the α2 subunits and decreased total expression levels of the α4 subunits to 120.7 ± 4.3% and 85.3 ± 7.0%, respectively, compared with WT controls (P < 0.05, n = 4–8 mice of each genotype; Fig. 1B).

To characterize the subcellular distribution of GABA_ARs in S408/9A mice, hippocampal sections were subjected to immunohistochemistry with antibodies against GABA_AR α2 or α4 subunits. In some experiments, sections were also stained with antibodies against the inhibitory scaffold protein gephyrin. Sections were then visualized using confocal microscopy, and subunit expression levels were quantified within the dentate gyrus. The number of inhibitory synapses defined as α2/gephyrin puncta were increased in S408/9A mice compared with WT littermates (23.4 ± 2.5 vs. 32.0 ± 0.6 for WT and S408/9A mice, respectively; P < 0.05, n = 3 mice of each genotype; Fig. 1C). In contrast, the number of puncta of α4 subunit immunoreactivity was decreased (13.1 ± 1.2 vs. to 37.4 ± 0.8 puncta per 2,500 μm² for S408/9A and WT mice, respectively; P < 0.001; Fig. 1C). Additionally, in contrast to the effects seen with GABA_ARs, the levels of gephyrin, postsynaptic density protein-95, and the AMPA receptor subunit GluA1 were unaffected by the S408/9A mutation (Fig. S2).

Fig. S2.

Analyzing glutamate receptor expression in S408/9A mice. (A) Postsynaptic protein expression is unaltered in S408/9A mice. To determine whether the observed changes in inhibitory synapses were associated with alterations in postsynaptic proteins, the total expression of the inhibitory synaptic marker gephyrin, excitatory synaptic marker PSD-95, and AMPA receptor GluA1 were measured. Hippocampal lysates from S408/9A and WT control mice were subjected to SDS/PAGE and immunoblotted with gephyrin, PSD-95, and GluA1 antibodies. Protein expression in S408/9A mice was normalized to WT controls (100%; red line). No significant differences in postsynaptic protein expression were observed (P > 0.05, t test; n = 8 mice of each genotype). (B) Forty-micron hippocampal slices were stained with antibodies against the postsynaptic protein PSD-95, followed by confocal microscopy. The pixel intensity of PSD-95 staining was then compared within the dentate gyrus between genotypes. Preventing phosphorylation of the β3 subunit does not alter PSD-95 immunoreactivity (P > 0.05, t test; n = 8 mice of each genotype). (Scale bar: 10 μm.)

Collectively, these results reveal that S408/9A mice have deficits in the cell surface accumulation of GABA_ARs that mediate tonic inhibition but an increased level of those GABA_ARs that mediate phasic inhibition.

S408/9A exhibited an increase in phasic inhibition and decreased tonic current.

To examine the functional significance of the alterations in GABA_AR expression seen in S408/9A mice, we used patch-clamp recording to analyze GABAergic inhibition in dentate gyrus granule cells (DGGCs). In DGGCs from S408/9A mice, tonic current was significantly decreased (24.7 ± 4.1 vs. 11.9 ± 2.2 pA for WT and S408/9A mice, respectively; P < 0.005, n = 10–11 cells, three mice of each genotype; Fig. 2A). In contrast, the amplitude of spontaneous inhibitory postsynaptic currents (sIPSCs) was significantly increased in S408/9A mice compared with WT controls (28.4 ± 2.4 vs. 42.5 ± 3.3 pA for WT and S408/9A mice, respectively; P < 0.005, n = 17–18 cells, three mice of each genotype; Fig. 2B). However, the frequency (4.5 ± 0.2 vs. 4.4 ± 0.1 Hz for WT and S408/9A mice, respectively) and decay time (44.7 ± 12.8 vs. 24.8 ± 2.5 ms for WT and S408/9A mice, respectively; P > 0.05, n = 17–18 cells, three mice of each genotype) of sIPSCs was comparable between genotypes.

Fig. 2.

Characterization of phasic and tonic inhibition in S408/9A mice. (A) Tonic conduction is decreased in S408/9A mice. Tonic current in DGGCs was compared between genotypes in the presence and absence of gabazine. The tonic current amplitude of S408/9A mice was significantly smaller than tonic current amplitude of WT mice. *Significantly different from control (P < 0.01, t test; n = 10–11 cells, three mice of each genotype). (B) sIPSCs were recorded from DGGCs of WT mice (black) and S408/9A mice (gray) *Significantly different from WT control (P < 0.001, t test; n = 17–18 cells, three mice of each genotype). (C) Examples of EPSCs recorded from DGGCs of WT mice (black) and S408/9A mice (gray) *Significantly different from control (P < 0.05, t test; n = 14–16 cells, three mice of each genotype). (D) Representative recordings of action potential firing in DGGCs from WT mice (black) and S408/9A mice (gray) in response to 0.5-s current injections from 20 to 300 pA in 20-pA increments. Average input/output curves from DGGCs from WT mice (black) and S408/9A mice (gray) are shown as a Boltzmann function generated by the averages of the fitted parameters. (Inset) No difference in the resting membrane potential (RMP) was observed between genotypes. Values are mean ± SEM (n = 19–21 cells, three mice of each genotype).

We also assessed the impact of the modifications in GABAergic inhibition on neuronal excitability. First, we compared the properties of excitatory postsynaptic currents (EPSCs) in DGGCs. In S408/9A mice, EPSC amplitude was significantly increased (14.6 ± 1.05 vs. 18 ± 1.02 pA for WT and S408/9A mice, respectively; P < 0.05; Fig. 2C) with no difference in frequency (1.4 ± 0.3 vs. 1.3 ± 0.2 Hz for WT and S408/9A mice, respectively). Moreover, their decay was also prolonged (2.1 ± 0.1 vs. 2.4 ± 0.2 ms for WT and S408/9A mice, respectively; P < 0.05, n = 14–16 cells, three mice of each genotype). However, net excitability of DGGCs and their resting membrane potentials were not modified by the S408/9A mutation (n = 19–20 cells, three mice of each genotype; Fig. 2D and Table S1).

Table S1.

Comparing the excitability of DGGCs between genotypes

No. of action potentials fired in DGGCs	WT	S408/9A
Current injection amplitude, pA
20	0.4 ± 0.4	0.0 ± 0.0
40	0.6 ± 0.5	1.0 ± 0.8
60	4.3 ± 1.8	6.9 ± 2.5
80	10.9 ± 2.9	17.1 ± 4.2
100	21.5 ± 3.5	24.5 ± 4.4
120	30.9 ± 3.6	34.0 ± 4.7
140	39.1 ± 3.6	42.6 ± 5.1
160	46.8 ± 3.8	49.7 ± 5.4
180	55.6 ± 3.8	57.9 ± 5.6
200	63.5 ± 4.3	65.0 ± 5.8
220	71.5 ± 4.6	72.2 ± 5.9
240	80.7 ± 4.7	77.5 ± 6.1
260	87.8 ± 4.9	83.2 ± 6.1
280	94.4 ± 5.0	89.5 ± 6.5
300	100.0 ± 4.9	93.6 ± 6.6
RMP, mV	−81.8 ± 1.2	−81.3 ± 1.2
I50, pA	169.0	164.2
k	48.0	51.9
n	19	20

Values are mean ± SE. I₅₀, amplitude of current injection eliciting 50% of maximum response; k, slope factor; n, number of mice; RMP, resting membrane potential.

Therefore, S408/9A mice have reduced tonic but enhanced phasic inhibition, but these changes do not lead to any gross changes in neuronal excitability, presumably because S408/9A mice also exhibit elevations in EPSC amplitude.

S408/9A Mice Exhibit Deficits in Social Interaction and Increased Repetitive Behavior.

ASDs have a common core of behavioral deficits, including reduced social interaction and increased repetitive behavior. Therefore, we assessed if any of these parameters are altered in S408/9A mice. First, we assessed if the respective mutation modifies motor coordination as measured using the rotarod, a critical control for data interpretation in rodent behavioral experiments. The latency to fall in the rotarod test was not different at 16–32 rpm in S408/9A mice compared with WT mice (Fig. S3A). Likewise, there were no differences between genotypes for the total distance traveled or dwell time in the center of the open field (Fig. S3 B and C). We also compared anxiety-like behavior using the light/dark test. WT and S408/9A mice spent equivalent time in the light chamber (46.9 ± 2.2% vs. 41.8 ± 2.5% for WT and S408/9A mice, respectively; P > 0.05, n = 12 mice of each genotype; Fig. S3D).

Fig. S3.

Characterization of motor activity and anxiety-like behavior in S408/9A mice. (A) Locomotor activity of WT and S408/9A mice was examined using the rotarod test, and the latency to fall was determined. There were no significant differences in motor activity (P > 0.05, t test; n = 10–11 mice of each genotype). RPM, revolutions per minute. (B) Time spent in the center of the arena in an open field was compared for WT and S408/9A mice over a 10-min time course. (C) In the open-field assay, there were no significant differences in the percentage of time spent in the middle of the open-field arena or in the distance traveled between genotypes (P > 0.05, t test; n = 12 mice of each genotype). (D) Percentage of time spent in the light arena of the light/dark box assay was determined for WT and S408/9A mice. There were no significant differences in the percentage of time spent in the light arena between genotypes (P > 0.05, t test; n = 12 mice of each genotype).

Social behavior was measured using the two-choice and three-chamber social interaction test. In the two-choice assay, WT controls spent more time in the chamber containing the novel mouse than in the chamber containing the empty cage, whereas S408/9A mice had no preference for the novel mouse over the empty cage (time spent interacting with empty cage: 49.8 ± 7.4 s (WT), 86.6 ± 17.6 s (S408/9A); time spent interacting with novel mouse: 119.4 ± 19.0 s (WT), 92.7 ± 17.7 s (S408/9A); n = 12 mice of each genotype, P < 0.01; Fig. 3A). Consistent with the profound deficits in social interaction observed in the two-choice assay, S408/9A mice exhibited no preference for the novel mouse in the three-chamber social interaction assay. WT controls spent more time in the novel mouse sniffing zone than in the empty cage sniffing zone (time spent interacting with empty cage: 55.4 ± 5.8 s (WT), 71.9 ± 8.4 s (S408/9A); time spent interacting with novel mouse: 123.8 ± 5.1 s (WT), 85.7 ± 6.9 s (S408/9A); n = 12–16 mice of each genotype, P < 0.01; Fig. 3B). In contrast, S408/9A mice showed no preference for the novel mouse over an empty cage, consistent with ASD-like deficits in social interaction. As a control for our measurements for social preference, we examined the ability of S408/9A mice to discriminate between a novel object and a familiar object. Both WT controls and S408/9A mice spent more time interacting with the novel object (66 ± 0.9% vs. 64 ± 7.7% of total time for WT and S408/9A mice, respectively; P < 0.05, n = 10–11 mice of each genotype; Fig. 3C).

Fig. 3.

Characterization of social interaction and repetitive-like behavior in S408/9A mice. (A) Preference for an empty cage or novel mouse was determined in a two-choice social interaction setup. *Significantly different from control (P < 0.01, ANOVA with Tukey’s post hoc comparison; n = 12 mice of each genotype). (B) Three-chamber social interaction assay was used to measure the preference for a novel mouse or an empty cage. The amount of time mice spent in a predefined sniffing zone around the empty cage or novel mouse was digitally tracked. *Significantly different from control (P < 0.001, ANOVA with Tukey’s post hoc comparison; n = 12–16 mice of each genotype). (C) Preference for a familiar object or a novel object was determined for the genotypes. *Significantly different from control (P < 0.05, ANOVA with Tukey’s post hoc comparison; n = 10–12 mice of each genotype). (D) Repetitive-like behavior was assessed using the marble burying assay. (P < 0.01, t test; n = 18–19 mice of each genotype). (E) Learning and memory were assessed using the fear-conditioning assay. The percentage of freezing per minute was assessed during training and for context-dependent and cued-dependent learning. *Significantly different from control (P < 0.05, t test; n = 12–13 mice of each genotype).

To measure repetitive-like behavior in S408/9A mice, we used the marble burying assay. S408/9A mice displayed an increase in repetitive-like behavior in the marble burying test (6.0 ± 0.6 marbles in 30 min compared with 3.8 ± 0.5 marbles in WT controls; P < 0.001, n = 18–19 mice of each genotype; Fig. 3D). Therefore, S408/9A mice exhibited an increase in repetitive-like behavior and deficits in social interaction, which are behavioral characteristics commonly observed in ASDs.

Finally, we assessed if the S408/9A mutation has an impact on cognition by comparing contextual and cued fear conditioning between genotypes. S408/9A mice exhibited enhanced context-dependent learning (training: 11.8 ± 3.0% freezing per minute (WT), 32.1 ± 4.3% freezing per minute (S408/9A); P > 0.05; context: 8.9 ± 3.4% freezing per minute (WT), 21.4 ± 4.2% freezing per minute (S408/9A); P < 0.05; Fig. 3E) compared with WT controls but no statistical differences in cue-dependent learning (cued: 45.1 ± 6.6% freezing per minute (WT), 60.8 ± 6.4% freezing per minute (S408/9A); P > 0.05; Fig. 3E).

S408/9A Mice Have Increased Seizure Susceptibility.

ASDs are frequently comorbid with epilepsy. Therefore, we analyzed the impact of the S408/9A mutation on seizure susceptibility. Mice were implanted with EEG electrodes, and seizures were induced with 20 mg/kg of kainic acid and monitored to ensure seizures reached stage 3–4 on the Racine scale (forelimb clonus and rearing with forelimb clonus). Representative EEG traces during epileptiform events captured 30 min after kainic acid injection illustrate a marked increase in epileptiform activity in S408/9A mice over WT controls (Fig. 4A). Strikingly, S408/9A mice exhibited a faster onset to the first epileptiform event compared with WT mice (4.6 ± 0.7 vs. 11.2 ± 1.1 min of latency for S408/9A and WT mice, respectively; P < 0.001, n = 8–10 mice of each genotype; Fig. 4B). Additionally, S408/9A mice entered status epilepticus at an earlier time point than WT controls (41.8 ± 4.3 vs. 61.5 ± 8.6 min for S408/9A and WT mice, respectively; P < 0.05, n = 8–10 mice of each genotype; Fig. 4B). Furthermore, S408/9A mice spent a greater percentage of the total time experiencing epileptiform activity compared with WT controls (82.5 ± 2.1% vs. 65.4 ± 6.4% for S408/9A and WT mice, respectively; P < 0.05, n = 8–10 mice of each genotype; Fig. 4B). Therefore, the S408/9A mutation significantly increases seizure susceptibility.

Fig. 4.

Characterization of seizure susceptibility in S408/9A mice. (A) Representative EEG traces for WT and S408/9A mice at baseline and 30 min after kainic acid injection. (B) Latency to the first epileptiform activity, onset of status epilepticus (SE), and total time experiencing epileptiform activity were digitally tracked and measured. S408/9A mice had a shorter latency to the first epileptiform activity, faster onset of status epilepticus, and spent a great percentage of time experiencing epileptiform activity compared with WT controls. *Significantly different from control (P < 0.05, t test; n = 8–10 mice of each genotype).

S408/9A Mice Exhibited Increased Hippocampal Dendritic Spine Density.

Alterations in spine structure are a common feature in ASDs (17). Therefore, we assessed if the alterations in GABAergic inhibition have an impact on dendritic structure. Brains from both genotypes were subjected to Golgi staining and visualized using stereology. We quantified the density of spines along 30-μm-long sections of dendrites in the molecular layer above the dentate gyrus of the hippocampus. In S408/9A mice, a pronounced increase in spine density was seen (28.6 ± 0.7 vs. 34.8 ± 1.0 per 30-μm dendrite for WT and S408/9A mice, respectively; P < 0.0001, n = 12 neurons in three independent experiments; Fig. 5A). Similarly, the spines in S408/9A mice appeared to be more filopodia-like in structure compared with their equivalents in WT mice, and some exhibited pronounced branching. Therefore, S408/9A mice displayed similar abnormalities in spine structure to those abnormalities reported in Fmr1 KO mice and patients with ASDs.

Fig. 5.

Characterization of dendritic spine density in DGGCs from S408/9A mice and β3 phosphorylation in Fmr1 KO (FXS) mice. (A) Representative dendrites from DGGCs of two different mice. (Scale bar: 2 μm.) The number of spines per 30 μm of dendrite was then determined and compared between genotypes. *Significantly different from control (n = 12 neurons from three independent experiments, three mice of each genotype; P < 0.0001, t test). (B) Hippocampal lysates from WT and Fmr1 KO mice on a FVB/N (Friend leukemia virus B strain) background were subjected to SDS/PAGE and immunoblotted with pS408/9, pS383, and β3 subunit antibodies. The pS408/9/β3 or pS383/β3 ratios in Fmr1 KO mice (FXS) were normalized to WT controls (100%; red line). *Significantly different from control (P < 0.01, t test; n = 3 mice of each genotype).

S408/9A Phosphorylation Is Increased in Fmr1 KO (Fragile X Syndrome) Mice.

To assess the relevance of our studies further, we measured S408/9 phosphorylation in Fmr1 KO mice, a model of Fragile X syndrome (FXS). S408/9 phosphorylation was significantly increased to 285.6 ± 24.3% in Fmr1 KO mice compared with WT controls (P < 0.005; n = 3 mice of each genotype; Fig. 5B). In contrast to S408/9, phosphorylation of S383 was similar in Fmr1 KO mice compared with WT controls (P > 0.05, n = 3 mice of each genotype; Fig. 5B).

Discussion

Deficits in the efficacy of neuronal inhibition mediated by GABA_ARs containing the β3 subunit are widely believed to contribute to the pathophysiology of ASDs. Accordingly, modifications in GABA_AR β3 subunit gene structure and/or protein expression levels, together with both deletions and duplications of the 15q11–13 locus, are leading causes of ASDs (5, 18).

In vitro studies have shown that phosphorylation of S408/9 regulates the cell surface stability and synaptic accumulation of β3-containing GABA_ARs (7), and, here, we have analyzed the significance of this putative regulatory process for the pathophysiology of ASDs. To do so, we created a S408/9A knock-in mouse using homologous recombination. S408/9A mice were viable and did not exhibit any overt phenotypes or gross alterations in the structure of the hippocampus. Consistent with the lowered affinity for AP2 in the S408/9A mutation in vitro, the cell surface expression levels of the β3 subunit in the hippocampus were increased in S408/9A mice compared with WT controls. Within the dentate gyrus, the β3 subunit assembles with the α2/γ2 or α4/δ subunits to form GABA_AR subtypes that mediate phasic and tonic inhibition, respectively (2). Strikingly, in the dentate gyrus of S408/9A mice, there was a significant increase in the expression levels of the α2 subunit and the number of inhibitory synapses. In contrast, deficits in the expression levels of the α4 subunit were seen. Thus, in addition to membrane trafficking, S408/9 and/or their phosphorylation may play a role in regulating the assembly of individual GABA_AR subtypes. Significantly, our findings of alterations in α4 and β3 subunit expression are consistent with GABAergic dysfunction in Mecp2 and Fmr1 KO mice, which are accepted models of ASDs (11, 19, 20).

Consistent with this result, the amplitude of sIPSCs was increased and their decay was slowed in DGGCs from S408/9A mice. In contrast, a reciprocal decrease in the cell surface accumulation of the α4 subunit was seen in S408/9A mice, which paralleled a reduction in tonic current. The gross excitability of DGGCs as measured by current injection was not altered between strains, reflecting the increase in EPSC amplitude in S408/9A mice. Therefore, preventing the phosphodependent modulation of the GABA_ARs by mutating S408/9 in the β3 subunit to alanines alters the equilibrium between phasic and tonic inhibition in the dentate gyrus.

Previous in vitro studies have suggested that for synaptic GABA_ARs containing α1/2β3γ2 subunits, mutation of S408/9A does not compromise receptor assembly, or transport to the plasma membrane, but selectively reduces their endocytosis (9). Therefore, the increased synaptic accumulation of the α2 subunit in S408/9A mice presumably results from their enhanced residence time on the plasma membrane. In contrast to synaptic receptors, the mechanisms that control the membrane trafficking of receptors containing α4 subunits are less well defined. However, it is evident from our results that phosphorylation of S408/9 may play opposing roles in regulating the assembly and membrane trafficking of α4 subunit containing GABA_ARs to those roles established for subtypes that mediate phasic inhibition. Clearly, further studies are required to identify the role that S408/9 and their phosphorylation plays in regulating the differential membrane trafficking of GABA_AR subtypes that mediate phasic and tonic inhibition.

In parallel with these modifications in GABAergic inhibition, the number of dendritic spines was strikingly increased in the mutant mice. Significantly, overexpressing S408/9A cDNA in cultured hippocampal neurons leads to similar changes in spine maturity to those changes seen in S408/9A mice (7). It is emerging that tonic inhibition plays a central role in reducing neuronal excitability, particularly at depolarizing membrane potentials (21). Given the emerging role for tonic inhibition in regulating neuronal output, the deficits in the efficacy of this process in S408/9A mice may lead to prolonged neuronal depolarization and enhanced intracellular Ca²⁺ signaling, which would be predicted to have significant effects on spine architecture.

In addition to modifications in dendritic structure, ASDs have a common core of behavioral symptoms, including increased repetitive behavior and anxiety, together with deficits in social interaction (17). Therefore, we compared these parameters between genotypes. Compared with WT controls, S408/9A mice exhibited decreased social interaction as measured in both the two-choice and three-chamber assays; however, S408/9A mice and WT controls shared a similar preference for a novel object over a familiar object. S408/9A mice also exhibited an increase in repetitive-like behavior as measured using the marble burying assay. Significantly, the mutation did not alter locomotor activity, behavior in the open field, or behavior in the light/dark test. S408/9A mice did not appear to exhibit any modifications in anxiety. ASDs are often comorbid with epilepsy, and consistent with this comorbidity, S408/9A mice exhibit increased sensitivity to kainate-induced seizures. Thus, although the changes in phasic and tonic inhibition seen in S408/9A mice do not appear to lead to changes in net neuronal excitability, they are important determinants of epileptogenesis. Finally, we noted that in the Fmr1 KO mouse model of FXS, specific alterations in S408/9 phosphorylation were seen. The mechanism linking deficits in FMRP expression and enhanced S408/9 phosphorylation requires further experimentation. However, it is interesting to note that FMRP is part of a signaling complex that contains protein phosphatase 2A (PP2A), the principle phosphatase that dephosphorylates S408/9 (10, 22, 23). Thus, deficits in FMRP expression levels may affect the stability and/or subcellular targeting of PP2A, leading to reduced rates of S408/9 dephosphorylation.

In conclusion, our results suggest that alterations in the phosphorylation status of GABA_AR β3 resulting in compromised GABAergic inhibition are central to the pathophysiology of FXS. Therefore, restoring the efficacy of tonic inhibition may be a useful therapeutic strategy to alleviate the burdens of ASDs.

Materials and Methods

All protocols were approved by Tufts University’s Institutional Animal Care and Use Committee and were conducted in accordance with the NIH Guide for the Care and Use of Laboratory Animals (24). More detailed information on materials and methods is provided in SI Materials and Methods.

Creation of the S408/9A Mice.

The S408/9A mice were created using homologous recombination as detailed previously (25).

Biochemical Measurements.

Antibodies used in this study have been described previously, as have the methods for immunoblotting and immunoprecipitation (22, 26).

Electrophysiology and EEG Recordings.

Phasic and tonic inhibition was measured using the patch-clamp technique (27). The Pinnacle Technology system and LabChart (ADInstruments) were used for EEG recordings and analysis as previously described (28).

Behavior.

Measurements using the rotarod and activity in the open field and anxiety were assessed as described previously (26).

SI Materials and Methods

Mice.

Ten- to 12-wk-old male S408/9A mice and the corresponding WT littermates were housed in a 12-h light/dark cycle with standard rodent food and water ad libitum. For behavioral testing, animals were acclimatized to the testing room for 1 h before the start of all procedures.

Gabrb3 (Gene That Encodes for the β3 GABA_AR Subunit Protein).

S408/9A mice were generated by gene targeting in murine ES cells using standard techniques. Briefly, a targeting vector was constructed containing 7.4 kb of homologous genomic sequence (from 3 kb upstream of the 5′ end of exon 9 to 4.4 kb downstream of the 5′ end of exon 9). The genomic DNA was subcloned from a substrain 129S5 BAC, obtained from the Wellcome Trust Sanger Institute. In exon 9 (the final exon), the codons for S408/9 of the GABA_AR β3 subunit were mutated to alanine codons (S408/9A). A loxP-flanked neomycin expression cassette (loxP-FRT-PGKneobpA-FRT-SA-4xSv40pA-SD-loxP) was inserted 0.3 kb upstream of the 5′ end of exon 9. An HSV-TK (herpes simplex virus-thymidine kinase) cassette for negative selection was positioned 3′ of the homology. The NotI-linearized targeting vector was electroporated into TC-1 ES cells (derived from substrain 129S6, a gift from Phil Leder, Department of Genetics, Harvard Medical School, Belmont, MA). Correctly targeted clones were identified by PCR and sequenced to verify the double-point mutation and the loxP sites, and they were injected into C57BL/6J mouse blastocysts. Chimeric male mice were bred with WT C57BL/6J females, and the entire neomycin expression cassette was excised by Cre/loxP-mediated recombination. Mice were backcrossed to C57BL/6J for a minimum of 10 generations. Experimental animals were derived from heterozygous breedings. The proposed allele symbol for the β3-S408/9A allele in the MGI database is Gabrb3^tm2Uru.

Antibodies.

Polyclonal rabbit antibodies against α2, α4, β2, and δ were gifts from Verena Tretter and Werner Sieghart (both from Medical University Vienna, Vienna). Anti-β3 and anti–phospho-β3 (phospho-S408/9 and phospho-S383) antibodies were generated and verified by the S.J.M. laboratory (9, 10, 12). The following antibodies were purchased from commercial vendors: monoclonal anti-α1 (clone N95/35; NeuroMab), monoclonal anti-gephyrin (catalog no. 147011; Synaptic Systems), and monoclonal anti-actin (catalog no. A2228; Sigma–Aldrich).

Western Blot.

Standard Western blotting protocols were used as previously described (10). Briefly, hippocampi were rapidly dissected, flash-frozen, and lysed in lysis buffer composed of the following: 20 mM Tris⋅HCl (pH 8.0), 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 10 mM NaF, 2 mM Na₃VO₄, 10 mM pyrophosphate, 0.1% SDS, and 50 mM NaF. Total protein concentration was established, and 40 μg of hippocampal lysate was subjected to SDS/PAGE, transferred to nitrocellulose membranes, and blocked with 5% (wt/vol) BSA in Tris-buffered saline-Tween 20 for 1 h. Membranes were immunoblotted with the indicated primary antibodies, and following extensive rinsing, they were probed with HRP-conjugated secondary antibodies and detected with enhanced chemiluminescence. Blots were imaged, and data were normalized to actin and quantified with the CCD-based LAS 3000 system (FujiFilm).

Surface Biotinylation.

Three hundred fifty-micron-thick hippocampal slices from 10- to 12-wk-old male WT control (C57BL/6) mice or age-matched male S408/9A mice were prepared on a vibratome (Leica VT1000S) in cold, oxygenated artificial cerebral spinal fluid (ACSF) composed of 124 mM NaCl, 3 mM KCl, 25 mM NaHCO₃, 2 mM MgSO₄, 2 mM CaCl₂, 1.1 mM NaH₂PO₄, and 10 mM glucose (pH 7.4). After sectioning, slices were allowed a 1-h recovery period in oxygenated, ice-cold ACSF. Following the recovery period, slices were incubated for 30 min at 4 °C with 1 mg/mL NHS-SS-biotin (Pierce). Following three rinses in ice-cold ACSF to remove excess biotin, hippocampal samples were lysed. After adjusting for protein concentration, hippocampal lysates were incubated with streptavidin beads (Pierce) overnight at 4 °C. Bound substances were subjected to SDS/PAGE, followed by immunoblotting with the indicated antibodies. Immunoblots were quantified using the CCD-based LAS 3000 system. For phospho-specific antibodies, the ratio of phospho-β3/total-β3 was calculated and S408/9A values were normalized to WT control (100%).

EEG Recordings.

EEGs were performed as previously described (28). Age-matched male littermates, aged 10–12 wk, were used for recording EEGs in awake, behaving animals. Mice were deeply anesthetized and implanted with prefabricated head mounts containing six-pin connectors with two electromyogram reference electrodes (Pinnacle Technology) 2.0 mm posterior to the bregma, along the midsagittal suture, and 2 mm below the dura. Following a 1-wk recovery period, EEG activity was monitored using the Pinnacle Technology turnkey system with a 100× amplifier and were high-pass-filtered at 1.0 Hz (PowerLabs; ADInstruments). A 30-min baseline was obtained before an i.p. injection of 20 mg⋅kg⁻¹ kainic acid (Tocris), followed by continuous EEG monitoring for an additional 120 min. Epileptiform activity was defined as electrographic events with amplitudes greater than twofold the standard deviation of the averaged baseline that last a minimum of 10 s and are separated from another event by greater than 10 s. Epileptiform activity was additionally confirmed by an increase in the power and frequency of high-amplitude events. Latency to the first seizure was defined as the time from injection of kainic acid until the first electrographic seizure. Duration of epileptiform activity was calculated from the cumulative time in minutes of epileptiform activity divided by 120 min. The onset of status epilepticus was defined as epileptiform activity lasting longer than 5 min with no silent period greater than 10 s. Duration of epileptiform activity, latency to the first epileptiform event, and latency to onset of status epilepticus were quantified using LabChart, version 7 (ADInstruments).

Immunohistochemistry.

Ten- to 12-wk-old male mice were perfused intracardially with 4% (wt/vol) paraformaldehyde, brains were dissected and immersed in 30% (wt/vol) sucrose for 72 h, and 40-μm-thick sections were prepared from fresh-frozen tissue using a microtome. Free-floating sections were processed for immunohistochemistry. Tissue was extensively rinsed in BupH-PBS (Thermo Fisher Scientific) and blocked for 1 h in 10% (wt/vol) BSA with 0.3% Triton X-100 in BupH-PBS. Tissue was incubated overnight at room temperature with the indicated antibodies, extensively rinsed in BupH-PBS, and incubated for 2 h in Alexa Fluor 488 and/or Alexa Fluor 568 secondary antibodies (Life Technologies, Thermo Fisher Scientific). After extensive rinsing with BupH-PBS, tissues were mounted, dried, and coverslipped with ProLong Gold anti-fade reagent with DAPI (Invitrogen). Images were obtained with a Nikon Ti microscope and analyzed with ImageJ (NIH) software. Controls consisted of incubating slices in the absence of primary antibody.

Electrophysiology.

The brains of male mice were rapidly dissected and placed in oxygenated, ice-cold ACSF solution containing 225 mM sucrose, 2.95 mM KCl, 1.25 mM NaH₂PO₄, 26 mM NaHCO₃, 0.5 mM CaCl₂, 10 mM MgSO₄, 10 mM d-glucose, and 3 mM kynurenic acid and were bubbled with 95% O₂/5% CO₂ (330–340 mosmol⁻¹). Coronal (350-μm-thick) slices were cut using a vibratome. The slices were then incubated at −33 °C in ACSF containing 126 mM NaCl, 2.5 mM KCl, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, 2 mM CaCl₂, 10 mM d-glucose, and 2 mM MgCl₂ saturated with 95% O₂/5% CO₂ (pH 7.4, 300–310 mosmol⁻¹) for a minimum of 1 h before experimentation. Hippocampal slices were viewed on a fixed-stage upright microscope (Nikon FN-1) with a 40× water immersion objective equipped with differential interference contrast/infrared (DIC/IR) optics. Slices were maintained at 32 °C and gravity-superfused with ACSF solution throughout experimentation. For current-clamp recordings in S408/9A mice, slices were incubated in ACSF in the absence of kynurenic acid. Input/output curves were produced as previously described from DGGCs (28). Briefly, visually identified DGGCs were injected with 0.5-s square wave current injections from 20 to 300 pA in 20-pA increments.

For voltage-clamp recordings, the internal pipette solution contained 140 mM CsCl, 1 mM MgCl₂, 0.1 mM EGTA, 10 mM Hepes, 2 mM Mg-ATP, 4 mM NaCl, and 0.3 mM Na-GTP (pH 7.25, 280–290 mosM). Series resistance and whole-cell capacitance were continually monitored throughout the experiment. After these parameters had stabilized for a minimum of 5 min, synaptic currents were recorded using an Axopatch 200B amplifier (Molecular Devices), filtered at 2 kHz, sampled at 20 kHz, digitized (Digidata 1320A; Molecular Devices), and stored for off-line analysis (Minianalysis; Synaptosoft, Inc.). Access resistance of 10–20 MΩ (∼80% compensation) was monitored using a voltage step of −5 mV, and data from cells were discarded when >20% change occurred.

sIPSCs were analyzed using pClamp 9.2 (Clampfit; Axon Instruments) and MiniAnalysis software using pooled population data expressed as the mean ± SEM. The decay phase was fitted with a monoexponential function (τ decay), and rise time was analyzed by comparing the mean with 10–90% rise time. Tonic current was calculated as the difference between the holding current before and after application of the indicated GABA_AR antagonist (≥200 μM SR95531 or 100 μM picrotoxin). The mean holding current of a 10-ms epoch was sampled every 100 ms for a single data point, and a Gaussian fit was applied to the resulting all-points histogram. Epochs containing synaptic events or an unstable baseline were excluded from the analysis. Input/output curves were fitted with a Boltzmann function: f(W) = Max/(1 + exp [(I − I₅₀)/k]), where Max = maximum response, k = slope, and I₅₀ = amplitude of current that produces 50% of the maximum response.

Fear Conditioning.

Fear conditioning was carried out in a single chamber with foot-shock bars. For conditioning and context testing, the chamber walls were covered with patterned panels. The conditioning procedures were a 180-s baseline period, followed by two trials separated by 120-s intervals. For each trial, a 15-s 70-dB tone was played, which terminated with the application of a 1.5-mA foot shock during the last 1 s. Twenty-four hours later, context-dependent memory was assessed by placing the mice in the chambers containing the patterned panels (conditioning environment). The amount of time the mice spent freezing during the trials was digitally tracked and analyzed. Two hours after context testing, cue-dependent memory was assessed by placing the mouse in a chamber without the patterned panels and measuring the amount of time it spent freezing during a 120-s baseline period, followed by two 70-dB tones played for 15 s separated from one another by 60-s intervals.

Open-Field Test.

Following a thorough cleaning of the arena with 70% (vol/vol) ethanol, individual mice were placed in the center of a 40 × 40-cm arena and movement was digitally tracked (EthoVision XT 7.0; Noldus Technology). Time spent in a predefined center arena and total distance traveled were calculated.

Light/Dark Box Assay.

Following a thorough cleaning with 70% (vol/vol) ethanol of the two equally sized chambers (each chamber is 21 × 21 cm), individual mice were placed in the dark chamber and allowed to roam freely between the dark and light chambers. Movement and time spent in the different chambers were digitally tracked and calculated (EthoVision XT 7.0).

Marble Burying.

Mice were transferred into a new cage with 6-inch deep bedding, and 20 glass marbles were placed evenly spaced apart in a 4 × 5 pattern on its surface. Mice were individually placed in the cage, and the number of marbles buried to two-thirds of their depth after 30 min was counted.

Two-Choice Social Assay.

In the two-choice assay, the subject mouse is placed at the bottom of a T-shaped container made of Plexiglas with two perpendicular arms (the long arm is 63.5 × 10 cm, and each short arm is 55 × 10 cm, surrounded by 32-cm high walls). All chambers and wire cups were cleaned with 70% (vol/vol) ethanol between each subject mice. The first session is a habituation period, where the subject mouse was able to roam freely across the three arms. This session was followed by a testing session, where a novel object (empty cup) was placed in the left arm or a novel mouse (stranger mouse under the wire cup) was placed in the right arm. The subject mouse was able to move freely across all three arms, and the cumulative time spent in the left arm or right arm was digitally tracked and quantified (EthoVision XT 7.1; Noldus Technology).

Social Novelty.

Briefly, on the first day of training, mice were exposed to the open-field arena (40 × 40 cm) for 10 min. On the second day of training, mice were placed in an empty open-field arena for 5 min, followed by 5 min of exposure to two empty cages. On the third day, testing for social novelty was carried out by placing the subject mouse in an empty open-field arena for 5 min, followed by another 5 min when an empty cage (familiar object) or a novel object (empty salt shaker) was placed in opposite corners of the open-field arena. The amount of time the mice spent exploring in a predefined zone around the objects was digitally tracked (EthoVision XT 7.0).

Three-Chamber Social Test.

Briefly, we constructed a three-chamber apparatus with Plexiglas containing three equally sized chambers (30 × 30 cm); two doors that can be remotely opened separate the middle chamber from the left and right chambers. A wire cup (4-inch diameter; item no. 010591315704, organizeit.com) was used as the empty cage or housed a stranger mouse (novel mouse), and the sniff zone 5 cm from the center of the wire cup was predefined. All chambers and wire cups were cleaned with 70% (vol/vol) ethanol between subject mice. A 10-min habituation period where the subject mouse was placed in the center chamber with the doors to the other chambers closed was followed by another 10-min session where the subject mouse was able to roam freely across all three chambers. These sessions were followed by a third 10-min session where a novel object (empty cup) was placed in the left chamber or a novel mouse (stranger mouse under the wire cup) was placed in the right chamber. The subject mouse was freely able to move across all three chambers, and the amount of time it spent in the predefined sniff zone (5-cm radius proximal to each cage) was digitally monitored (EthoVision XT 7.1). A fourth 10-min session was used to test for social novelty, where a second stranger mouse was placed in the previously empty cage. The subject mouse was then able to roam freely across all three chambers, with the right chamber holding the previously encountered mouse (familiar mouse) and the left chamber holding the novel mouse (stranger mouse). Cumulative time spent in the left or right sniffing zone was digitally tracked and quantified. (EthoVision XT 7.1).

Golgi Staining.

For Golgi staining, we used the protocol recommended by the manufacturer of the FD Rapid GolgiStain Kit (FD NeuroTechnologies, Inc.). Briefly, mice were deeply anesthetized, and brains were rapidly extracted and rinsed in Milli-Q water generated at Tufts University within the S.J.M. laboratory. Brains were immersed in the provided solutions for 14 d at room temperature in the dark. Following Golgi–Cox immersion, 200-μm-thick coronal sections were prepared on a vibratome. Sections were then developed according to the manufacturer’s protocol and dehydrated in 50%, 75%, 95%, and absolute ethanol; cleared in xylene; and coverslipped with Permount (catalog no. SP15; Thermo Fisher Scientific). Dendritic spine analysis was performed on dendritic spines located immediately adjacent to the DGGC layer in the molecular layer of the hippocampus using images collected with a 60× objective (Nikon). Spine density in 30-μm-long sections was calculated.

Statistical Analysis.

All data are presented as the mean ± SEM and were analyzed with the Student’s t test, one-way ANOVA with Turkey’s multiple comparison test, and two-way ANOVA with Bonferroni’s post hoc comparison. Unless otherwise indicated, P < 0.05 was considered statistically significant. All analyses were done using Prism 6 (GraphPad).