Inherited and de novo SHANK2 variants associated with autism spectrum disorder impair neuronal morphogenesis and physiology

SHANK2 variants are misplaced and functionally altered

To uncover the functional activity of rare SHANK2 variants, we investigated the effects of an R462X stop mutation, an L1008_P1009 duplication and a T1127M missense variant that were all detected in patients with ASD (6). The R462X mutation occurred de novo while the duplication and the missense variants were both inherited from apparently healthy mothers. We selected these two inherited variants because they localize at highly conserved positions in SHANK2; the T1127M mutation resides in the highly conserved Dynamin-2-binding motif and the L1008_P1009dup mutation in direct vicinity to the Homer-binding motif, suggesting that this may alter SHANK2 protein function. Mothers with the T1127M and L1008_P1009dup variants had a family history of learning difficulties or developmental delay.

To study the effects of the SHANK2 variants on the morphology of primary neurons, we overexpressed all three variants and the SHANK2 wild-type (WT) in primary hippocampal neurons as mCherry fusion proteins (Fig. 1A). Western analysis in HEK293 cells indicated an identical size of SHANK2 WT and variant L1008_P1009dup and T1127M using a SHANK2 primary antibody, whereas the stop mutation (R462X) resulted in the truncation of the protein (Fig. 1B).

In primary hippocampal neurons, SHANK2-WT exclusively accumulated in dendritic spines and localized postsynaptically to synapsin 1 (Fig. 1C and D). In contrast, the truncated SHANK2 (R462X) was evenly distributed throughout the dendrites and accumulated in the cell soma. The R462X variant did not co-localize with synapsin 1, suggesting that it is not efficiently targeted to functional synapses (Fig. 1C and D). The T1127M and the L1008_P1009dup variant still localized to synapses (Fig. 1C and E) but displayed smaller protein accumulations in the spines, compared with WT (Fig. 1F).

To analyze the effect of the variants on dendritic spine volume, we transfected primary hippocampal neurons with GFP and mCherry-SHANK2 fusion constructs (Supplementary Material, Fig. S1A). High-resolution z-stack images of GFP-positive neurons were taken with a confocal laser scanning microscope to measure the fluorescence intensity of GFP within the three-dimensional projections, which in turn allowed us to calculate the relative spine volumes. Overexpressing SHANK2-WT, we detected a significant increase in dendritic spine volume, similar to previously described findings (11) (Supplementary Material, Fig. S1A and B). For the SHANK2 mutants, we found in three independent experiments, analyzing more than 3600 spines from 18 neurons for each condition, that the SHANK2-R462X stop mutant entirely lacks the potential to increase spine volume (Supplementary Material, Fig. S1A and B). Similarly, also the SHANK2-T1127M and the L1008_P1009dup variants had no significant effect on the increase in spine volume (Supplementary Material, Fig. S1C). The lack of SHANK2 protein function of the stop mutant (R462X) might be caused by its misplaced localization due to the loss of the C-terminal SAM domain and the adjacent binding motifs for the interaction with Homer, Cortactin and Dynamin 2. In contrast, SHANK2-T1127M and the L1008_P1009dup might affect SHANK2 protein function rather than localization because they both still localize to dendritic spines but display significantly smaller clusters that might result in an altered PSD organization. These results provide first evidence that all three variants are impaired in function, with the most striking defect seen for the displaced R462X mutant.

SHANK2 variants fail to rescue Shank2 knock-down in primary hippocampal neurons

To investigate the effect of reduced SHANK2 protein levels as predicted for individuals with ASD with a de novo microdeletion of the SHANK2 gene (6), we performed RNAi-mediated knock-down of Shank2 in primary hippocampal neurons. We used two shRNAs (#1, #2), which target different regions of Shank2 efficiently. HEK293 cells were co-transfected with either of the two shRNAs and Shank2 from Rattus norvegicus or human SHANK2. Rat Shank2 expression was specifically reduced but the expression of rat Shank1, rat Shank3 and human SHANK2 was not altered, indicating specificity (Supplementary Material, Fig. S2A). In shRNA#2-transfected cortical neurons, the endogenous Shank2 protein level was almost 50% reduced as indicated by western blot analysis (Supplementary Material, Fig. S2B). Similarly, in cultured hippocampal neurons, endogenous Shank2 protein levels were reduced by approximately half with both shRNAs, as measured by counting the number of Shank2 puncta along the dendrites (Fig. 2A and D). The reduced Shank2 expression significantly decreased spine volume (Fig. 2A and E), but had no effect on spine density (Fig. 2G). Reduced spine volume was accompanied by a significant decrease in the number of functional synapses along the dendrites indicated by co-localization of the postsynaptic marker PSD95 and the presynaptic marker Synapsin 1 (Fig. 2C and F). Under control conditions (control-shRNA), the majority of dendritic spines show a mature mushroom-shaped morphology, whereas the knock-down of Shank2 presented thin filopodia-like immature spines in the majority of cells (Fig. 2A). Hippocampal neurons with reduced Shank2 expression furthermore showed an increase of dendritic complexity (Fig. 2B). We performed Sholl analysis by counting the intersections of dendrites with overlay of concentric rings at 25 µm intervals centered on the cell soma, to describe the amount and distribution of dendrites. The total number of dendritic branches was significantly increased in Shank2 knock-down neurons compared with the GFP control conditions (Fig. 2H), with an accelerated growth of dendrites close to the cell soma (Fig. 2I).

We next asked whether SHANK2 WT and mutants are able to rescue the Shank2 knock-down phenotype in primary hippocampal neurons by co-transfection of shRNA#2 (directed against rShank2) with either the human SHANK2-WT or one of the three variants, respectively. Human SHANK2-WT and the duplication variant (L1008_P1009dup) were able to rescue the knock-down-mediated phenotype, normalizing both the dendritic spine volume and dendritic arbor development (Fig. 3A and C). In contrast, the R462X and T1127M variant could not rescue the reduced spine volume (Fig. 3A). As shown in Figure 3B, expression of SHANK2-WT reduces the total number of intersections below GFP control levels. In contrast to these findings, overexpression of SHANK2 has no effect on dendritic arbor development. In the Sholl analysis, the distribution of dendritic material throughout the neuron and the total number of intersections was not altered (Supplementary Material, Fig. S3). Comparing the degree of rescue obtained with either human SHANK2-WT or the three variants, we can demonstrate that only the R462X variant fails to rescue the excessive branching to a normal level, an effect that fades out 75 µm away from the cell body (Fig. 3B and C). In this respect, the missense and duplication variants behave like WT normalizing the number of intersections. The R462X mutant has the most severe effect on the neuronal morphology, whereas the duplication shows no functional loss in the rescue experiment. The loss-of-function effect of the T1127M variant is restricted to the dendritic spines. Given the disparate effects of different Shank2 mutants on dendrite and spine morphology, this suggests that either Shank2 is involved in the regulation of two separate pathways controlling dendrite or spine morphology, respectively, or that spine growth is more sensitive to a reduction in Shank2 activity compared with dendrite arborization.

Miniature EPSCs are specifically reduced by SHANK2-R462X overexpression

We next studied the physiological effects of R462X in vivo by rAAV-mediated gene delivery of Venus-tagged R462X into the hippocampus of adult mice. Three weeks after rAAV infections, at postnatal day P69 ± 1, the ex vivo recordings of isolated AMPAR-mediated miniature excitatory postsynaptic currents (mEPSCs, see Materials and Methods) of Venus-positive CA1 pyramidal cells showed that the average mEPSC amplitude was significantly reduced by ∼26% in SHANK2-R462X-infected neurons compared with SHANK2-WT overexpressing and WT neurons (Fig. 4A, Supplementary Material, Table S1). Since in age-matched mice the mEPSC amplitude between control and SHANK2-WT expressing cells is not significantly different, we conclude that R462X has a direct pathophysiological effect by reducing the mEPSC amplitude. The mEPSC frequency was similar for all groups, indicating that the SHANK2-R462X expression mainly affects AMPA-receptor-mediated postsynaptic excitatory transmission. No changes in mEPSC decay times between groups were detected. Analysis of inhibitory transmission (12) into the same cells (see Materials and Methods) showed that the amplitude and frequency distributions of spontaneous inhibitory currents (mIPSCs) were similar across experimental groups. Thus, neither the inhibitory drive in principal neurons nor the root mean square (RMS) current noise levels, input resistance and other neuronal control properties were affected by the overexpression of SHANK2-R462X (Fig. 4A, Supplementary Material, Table S1).

Next, we investigated whether viral SHANK2-R462X transduction by injection of rAAV into the brains of newborn mice (13) was long lasting. We injected Venus-tagged SHANK2-WT and Venus-tagged SHANK2-R462X at P0 and recorded mEPSCs and mIPSCs 3 months later in brain slices (Fig. 4B). Venus-positive pyramidal cells located either in the hippocampal CA1 region or in layer 2/3 of the entorhinal cortex (EC) were selected for recordings (Fig. 4B). The average mIPSC amplitude and frequency obtained from Venus-positive CA1 and from layer 2/3 pyramidal cells remained similar between rAAV SHANK2-WT and rAAV SHANK2-R462X mutant infected mice, as in the CA1 cells of adult injected mice (Fig. 4B). In contrast, however, the average mEPSC amplitudes from SHANK2-R462X-expressing CA1 and layer 2/3 pyramidal cells (Fig. 4B) were decreased by ∼33 and ∼17%, respectively, compared with corresponding SHANK2-WT-expressing cells. The mEPSC frequency and other control neuronal parameters were similar across groups (Supplementary Material, Table S1), demonstrating that the persistent expression of SHANK2-R462X leads to a postsynaptically localized reduction of AMPA-receptor-mediated currents.

However, the high-resolution analysis of the AMPA receptor distribution at CA1 synapses in the stratum radiatum revealed an increased size and number of AMPA receptor clusters in adult SHANK2-R462X-expressing mice compared with GFP-expressing controls (Fig. 4C and D). The Golgi stain showed densely packed filopodia structures instead of mature spines at CA1 dendrites of SHANK2-R462X-expressing mice (Fig. 4E), which might indicate that AMPA receptor clusters in the neuropil of SHANK2-R462X mice are not localized at active synapses. This hypothesis is supported by the reduced surface expression of AMPA receptors in primary hippocampal neurons overexpressing SHANK2-R462X. By immunostaining, we observed a subtle but significant decrease in AMPA receptor cluster sizes at spine heads of neurons overexpressing SHANK2-R462X in comparison with GFP/mCherry control conditions (Supplementary Material, Fig. S4). A reduced amount of AMPA receptors at the spines of neurons is consistent with the decrease in mEPSC currents observed in mice.

In summary, we conclude that the expression of SHANK2-R462X for 2–3 weeks in the brain of adult mice has a dominant negative effect. SHANK2-R462X expression is sufficient to reduce the amplitude, but not the kinetics of excitatory AMPAR-mediated transmission. It does not affect the inhibitory drive converging into SHANK2-R462X-expressing cells, and it leads to changes in spine structures and AMPA receptor distribution.

Postnatally expressed SHANK2-R462X mutant alters mouse cognitive behavior

The strong and persistent expression of SHANK2-R462X in the brain of P0 injected mice might permit a first behavioral analysis of mice overexpressing SHANK2-R462X in the neocortex. We used well-established tests for locomotor activity, novel object exploration and executive functions that have been previously shown to be sensitive to altered AMPA receptor signaling (14,15).

The postmortem analysis of the mouse brains showed that the GFP signal in rAAV-injected mice (P0) was 5-fold stronger than the signal for Venus of rAAV-SHANK2-R462X injected mice 7 months after injection (Fig. 5A and C). In both mouse cohorts, the GFP as well as the Venus-SHANK2-R462X expression was prominent in the cortex and hippocampus (Fig. 5B and D). By quantitative immunoblotting, the rAAV-injected mice could be divided into high (H) and low (L) expressing mice (Fig. 5D), whereas the endogenous AMPA receptor subunit levels for GluA1 and GluA2 and the amount of phosphorylated GluA1 isoforms GluA1-S831 and GluA1-S845 were unchanged in both groups (Fig. 5E). Since the viral transduced genes are not expressed ubiquitously, alteration in AMPA receptor expression or phosphorylation cannot be excluded at the cellular level.

For a closer behavioral analysis, we divided the group of rAAV-SHANK2-R462X-expressing mice in high and low expressing animals (Fig. 5) by a post hoc median split. At the age of 3–4 months, all mice showed comparable body weight and regular locomotor behavior in the rotarod and in the open field test (Fig. 6A and data not shown). In the open-field test, however, when a novel object was introduced into the center of the open arena, only AAV-GFP control injected mice showed the expected high exploration rate of the novel object (P< 0.01), whereas both SHANK2-R462X-expressing groups did not exhibit significant increase in object exploration (Fig. 6B). Moreover, we subjected the mice to an escape task testing executive functions [the so-called puzzle box test (15)] in which the animals have to escape from a lit compartment into a dark compartment, with increasing difficulty of escape strategies from trial 1 to 11. In this test, high-level SHANK2-R462X-expressing mice showed a significant impairment (P< 0.01) compared with low level and GFP-expressing mice, respectively, in the more difficult trials requiring elaborated escape strategies (Fig. 6D). This phenotype was not confounded by alterations in anxiety behavior, since all groups showed similar emotional behavior in a standardized light–dark box test with simple escape possibilities (Fig. 6C). In summary, our data demonstrate that the postnatal expression of the SHANK2-R462X in mice can lead to impaired cognitive functions.

Ethics statement

All experimental procedures were performed according to the animal welfare guidelines of the Max Planck Society (Charles River). Animal experiments were performed under the license at the Regierungspräsidium Karlsruhe 35-915.81/G-71/10, 35-915.81/G-56/02 and 35-915.81/G-49/09.

Plasmid constructs

Cloning of SHANK2, with the partial coding sequence from KIAA1022fg00613 clone (accession number AB028945, Kazusa DNA Research Institute, Japan), into the pENTR2B vector (Invitrogen) was via SalI and EcoRV restriction sites, and SalI and PshAI in the KIAA construct. The N-terminal part of SHANK2 was PCR-amplified from human brain cDNA, introducing a SalI restriction site at the 5′end. The PCR product was inserted into pENTR2B-KIAA1022 using SalI and SacI, resulting in pENTR2B-SHANK2-WT. Mutagenesis was performed using the QuikChange Lightning Site-directed Mutagenesis Kit (Stratagene) on 70 ng of SHANK2 WT plasmid to generate the L1008_P1009dup, T1127M and R462X mutants. Each clone was purified with the Pure Yield Plasmid Midiprep System (Promega) and sequenced to rule out additional mutations. For fluorescence image analysis, the SHANK2 WT and mutant constructs were transferred to pcDNA3.1-mCherry-DEST vector using the Gateway recombinant technology (Invitrogen). The cloning of Shank2 from Rattus norvegicus was performed using cDNA from the rat brain P18 prefrontal cortex as a template for PCR, with the primers amplifying the smallest neuronal isoform of rat-Shank2, introducing an N-terminal BamHI and a C-terminal NotI restriction site. The PCR fragment was cloned into pENTR3C (Invitrogen) and then transferred into pcDNA3.1-FLAG-DEST vector (generated in our laboratory). The pSUPER RNAi expression system was used for siRNA-mediated knock-down of SHANK2. Two independent pSUPER constructs were generated with BglI and HindIII restriction sites (see primer sequences). The co-shRNA is described elsewhere (28). The pGW1-HA-Shank1 and Shank3 expression constructs were used as described (29). For the AAV plasmids, the SHANK2-WT and SHANK2-R462X sequences were sub-cloned into the pAAV-Syn-Venus vector (30), generating pAAV-Syn-Venus-SHANK2-WT and pAAV-Syn-Venus-SHANK2-R462X plasmids.

Cell culture and transfection

Dissociated primary hippocampal neurons from embryonic day 18 (E18) Sprague–Dawley rats (Charles River Laboratories, Sulzfeld, Germany) were prepared and cultured as described elsewhere (28). Neurons were transfected after 10 days in culture (10 DIV) with Lipofectamine 2000 (Invitrogen), using a total of 1 µg of DNA (500 ng SHANK2 construct or mutant, 100 ng of GFP-plasmid, 400 ng of pcDNA3-vector) for each well of a 24-well plate. HEK293 cells were maintained in DMEM (GIBCO) plus 10% FCS, 1 mm glutamine, 100 U/ml penicillin and streptomycin. Transfections of the different constructs were performed with Lipofectamine 2000. For shRNA experiments, pSUPER plasmids were used either at 1.5 µg/ml (HEK293 cells, together with 0.5 µg/ml of the Shank constructs) or at 5 ng/ml (primary neurons). Rescue experiments in primary hippocampal neurons were performed with 5 ng/ml pSUPER-shRNA #2, 125 ng/ml of SHANK2 WT or mutant constructs and 50 ng/ml of GFP plasmid DNA. For the GFP control, we added 125 ng/ml of pcDNA3.1-mCherry-DEST vector and 50 ng/ml of GFP plasmid DNA. pcDNA3.1 vector was calculated to reach a final concentration of 500 ng/ml of total DNA for each reaction.

Virus infection

Recombinant AAVs were generated as described (30), and purified by affinity chromatography (31). Virus titers were determined in hippocampal primary cultures (1.0 × 10⁷–10⁸ particles/ml). For adult infections, viruses were delivered into the brain of deeply anesthetized (ketamine/xylazine mix) 6-week-old C57BL/6N mice (Charles River) by stereotactic injections (30). Stereotactic coordinates relative to the bragma were: anteroposterior (AP) 2.1 mm, lateral ±1.6 mm, depth 0.3–1.6 mm for hippocampal and cortical injections (one purified virus/injection). For P0 injection, mice from two litters of C57BL6/N mice (Charles River) were randomly injected with either rAAV-Syn-GFP or rAAV-Syn-Venus-SHANK2-R462X virus at postnatal day 0. The rAAV viruses were injected in both hemispheres targeted to the hippocampus. Each mouse received two injections on each hemisphere: AP 1.0 mm, lateral ±1.0 mm, depth 1.6 mm; AP 2.0 mm, lateral ±1.0 mm, depth 1.5 mm; 1 (microlitres purified virus/injection site).

After injection, the pups were put back to the home cage with their mothers. The identification of the injected virus in the animal was done postmortem after the behavior analysis by immunoblotting.

Brain sections

Mice were anesthetized with isofluran and either decapitated directly or perfused intracardially with warm 1× PBS (137 mm NaCl, 2.7 mm KCl, 4.3 mm Na₂HPO₄/2H₂O, 1.4 mm KH₂PO₄) and 4% paraformaldehyde (PFA) in PBS prior to decapitation. Brains were removed and fixed in ice-cold 4% PFA for 2 h, embedded in 2.5% agarose/PBS and sliced on a vibratome into 70–100 μm sections.

Immunocytochemistry

Hippocampal primary neurons were fixed on 18 DIV in 4% PFA for 15 min, permeabilized with PBS plus 0.2% Triton X-100 and blocked in PBS containing 10% normal goat serum. We used a rabbit polyclonal anti-synapsin-1 antibody (1:500, Chemicon), a mouse anti-Shank2 antibody (1:500; NeuroMab, Antibodies, Inc.) or a mouse anti-PSD95 antibody (1:500; NeuroMab, Antibodies, Inc.) together with Alexa Fluor 647 goat anti-rabbit (1:1000; Invitrogen) or Alexa Fluor 568 rabbit anti-mouse (1:1000, Invitrogen). Virus-injected mice were fixed with 4% PFA/PBS, and 70 μm sections were obtained from the brain. Staining of brain sections was done with rabbit anti-GFP antibody (1:5000, Abcam) and FITC-coupled anti-rabbit secondary antibody (1:200, Jackson ImmunoResearch). For the GluA1 cluster staining, the brain sections were treated at 37°C for 10 min with 0.3 mg/ml Pepsin (DAKO Carpinteria) in 0.2 N HCl and washed briefly in PBS before immunostaining. The sections were incubated with rabbit anti-GluA1 (1:1000, Abcam) and followed with Cy3-coupled anti-rabbit secondary antibody (1:500, Jackson ImmunoResearch). The Golgi staining was performed with the commercial FD Rapid GolgiStain^TM Kit (FD NeuroTechnologies).

Image analysis

To determine the effect on spine volume in primary hippocampal neurons, the SHANK2 constructs were co-transfected with GFP and processed for confocal microscopy at 18 DIV. High-resolution z-stack images of GFP-positive neurons were taken with the 63× objective of the confocal laser scanning microscope (Zeiss). Neurons displaying pyramidal morphology were chosen randomly and spine volumes were subsequently analyzed with the ImageJ software. Signal intensity levels of each construct were similar, excluding the possibility of effects of transfection efficiencies. The clustering properties of SHANK2 constructs were quantified by calculating the synaptic clustering ratio (SCR), the ratio of the average pixel intensity in dendrites versus spine heads. Transfected neurons were randomly chosen for each construct from two independent transfections (12 cells per construct). The average pixel intensity was measured in dendrites (excluding the spine heads) and in the spine heads by tracing their individual perimeters. The size of SHANK2 construct clusters and the endogeneous Shank2 clusters in dendritic spines as well as the synaptic density measures were determined with the ‘analyze particle’ function of the ImageJ 1.41f software, using threshold images. Particles <0.1 µm² were excluded from the analysis. For the brain sections, evaluation of the immunostaining and Golgi staining was done with confocal microscopy (SP2, Leica) using either 5× for overview or 63× for high-magnification images. Two mice from each genotype were used for the GluA1 cluster staining. The laser power of confocal microscopy was kept the same during all recordings. In the stratum radiatum region of the hippocampus, 4 stacks containing 40 images each with an area of 29.7 × 29.7 μm² and an interval of 81 nm were taken for the analysis. All stacks were first restored with the deconvolution software HuygensPro (SVI Systems) applying CMLE with 300 iterations, 0.0001% quality threshold and a theoretical point spread function. Deconvoluted images were then imported to ImageJ for particle counting analysis using a fixed threshold algorithm.

Western blot analysis

Infected hippocampal tissue of mouse brains was collected and homogenized in ice-cold buffer (25 mm HEPES, pH 7.4) containing a protease inhibitor cocktail (Complete, Roche, Pharma AG). After 5 min of 2000 r.p.m. low centrifugation, the supernatnant containing total proteins was collected. For membrane proteins, they were obtained by 13 000 r.p.m. high centrifugation of the pellets. Of total and membrane proteins, 5 μg was separated by SDS–PAGE (10% separating and 4% stacking gels) and transferred to nitrocellulose membranes. Western blots were probed with monoclonal mouse anti-GFP (1:10000, Abcam), polyclonal rabbit anti-GluA1 (1:5000, Abcam), monoclonal mouse anti-GluA2, monoclonal mouse anti-SHANK2 (1:2000), monoclonal rabbit anti-pospho-Ser831-GluA1 (1:8000, Millipore), monoclonal rabbit anti-phospho-Ser845-GluA1 (1:2000, Millipore), polyclonal rabbit anti-P38 (1:2000, Abcam) and monoclonal mouse anti-GAPDH (1:5000, Abcam). Fluorescent coupled secondary antibodies anti-rabbit IgG (IRDye800DX, 1:10 000, Rockland) and anti-mouse IgG (IRDye700DX, 1:5000, Rockland) were used. Fluorescence was detected by the fluorescence imaging scanner (Starion/FLA9000, GE-Healthcare). Quantification of western blots was done using Multi-Gauge (Fuji-Film) and Image J (NIH).

Electrophysiology

The brain of mice was dissected out and combined EC-dorsal hippocampus slices (350 μm) were cut in the horizontal plane (HM 650 V, Microm International, Walldorf, Germany) in a dissection buffer (4°C) containing (in mm): 212.7 sucrose, 5 KCl, 1.25 NaH₂PO₄, 10 MgCl₂, 0.5 CaCl₂, 26 NaHCO₃, 10 dextrose, saturated with 95% O₂/5% CO₂ (pH 7.4). Individual slices were gently transferred to normal artificial cerebrospinal fluid (ACSF) for at least 1h prior to recording. Normal ACSF was similar to the dissection buffer except that sucrose was replaced by 119 mm NaCl, MgCl₂ lowered to 1 mm and CaCl₂ raised to 2 mm. Visualized whole-cell recordings (IR-DIC, Axioskop 2 FS mot, Zeiss International, Germany) were made from EC layer 2/3 (≥ 35% depth from the pia), and from hippocampal CA1 pyramidal cells. Cortical pyramidal cells were identified by their pyramidal shaped, prominent apical dendrites and regular spiking pattern in response to depolarizing current (0.1–0.5 nA, 1 s). A subset of 20 pyramidal-like cells was filled with 0.1% Lucifer yellow and we found that 95% were spine-containing pyramidal cells, validating the selection criteria. Borosillicate glass recording pipettes (4–6 MΩ) were filled with intracellular solution containing (in mm): 130 (Cs)gluconate, 5.5 KCl, 1 EGTA (pH 8), 10 HEPES, 4 (Mg)ATP, 0.5 (Na₃)GTP, 10 (Na₂)phosphocreatine and 5 QX-314 (lidocaine N-ethyl bromide) (pH 7.25, 280–290 mOsm), and a liquid junction potential of 16.3 mV was compensated (30°C). Only cells with holding currents ≤100 pA, series resistance ≤20 MΩ and input resistance ≥150 MΩ were studied. Cells were discarded if any of these control parameters, monitored with a 100 ms voltage pulse of −10 mV, changed ≥20% between two flanking epochs sampled at the beginning and end of the recordings. Data were filtered at 2 kHz and digitized at 5 kHz using an EPC-10 amplifier controlled by Patchmaster (HekaElektronik, Lambrecht, Germany). We quantified the net excitatory and inhibitory postsynaptic drive from the same recorded cells. Spontaneous miniature excitatory and inhibitory currents (mPSCs) were recorded in the presence of tetrodotoxin citrate (1 μm TTX, Biotrend), (dl)-2-amino-5-phosphonovaleric acid (50 μm d,l-APV, Biotrend) and CGP 55845 (2 μm, Biotrend), all diluted in the ASCF at 2 ml/min, 30 ± 1°C. To isolate AMPAR-mediated miniature currents (hereafter referred to as mEPSCs), neurons were held at V_h= −70 mV [E_AMPA= 6.21 ± 0.84 mV, n = 11, in (–)-bicucullinemethiodide; 20 μm, Sigma Aldrich], whereas for GABA_A-mediated inhibitory currents (hereafter referred to as mIPSCs), neurons were held at V_h= 0 mV [E_GABAA = −67.62 ± 3.10 mV, n = 15, estimated in the presence of 6-nitro-7-sulfamoyl-benzo(f)quinoxaline-2,3-dione, 10 μm, NBQX, Biotrend]. Spontaneous currents were detected and analyzed using the Mini Analysis Program (Synaptosoft, Inc.) with a threshold set at three and five times the RMS of membrane current noise, for mEPSCs and mIPSCs, respectively. Cells showing a negative correlation between mPSC amplitude and rise time were excluded from this analysis (i.e. dendritic filtering present). The first 300 events/cell with rise time ≤3 ms for mEPSCs (≤5 ms for mIPSCs) were used to estimate the cell's mPSC amplitude and frequency distributions. At least two cells per animal were recorded and average amplitude and frequency (per animal) were compared across interleaved experimental conditions.

Mouse behavior analysis

Mice were housed four per cage in an animal room with constant temperature (22 ± 1°C) and adapted to a new 12 h light/dark cycle with free access to food and water. Experiments were performed between 10 a.m. and 5 p.m. during the light phase. One week prior to starting the tests, mice were single-housed and handled extensively (15 min per mouse per day) to reduce their anxiety and get acquainted with the testing environment (32).

Statistical analysis

Data are presented as mean ± SD unless otherwise stated. Testing for normal distribution was made by the Kolmogorov–Smirnov or the Shapiro–Wilk test. Concerning the experiments in primary hippocampal neurons, the comparisons of mean differences between groups were made using the two-way ANOVA test including the different conditions and controlling for the different replicates followed by a Scheffé post hoc test with significance levels for both set at P< 0.05. ANOVA P-values are shown for the different conditions. mPSC amplitudes were statistically compared using an unbalanced one-way ANOVA test and their cumulative distributions with a Kolmogorov–Smirnov test. Average frequencies were compared by using an ANOVA test followed by Fisher's protected least significant difference (PLSD) post hoc test. mPSC kinetics were compared using an unpaired Student's t-test followed by a Scheffé' post hoc test. Significance levels were set at P < 0.05. Group plots are presented as average ± SEM. Tests were performed with programs written in MATLAB 7.8 (R2009a; MathWorks, Inc.) or SPSS 16.0 software for Windows. For the behavior tests, one-way ANOVA was performed for the open-field activity tests; two-way ANOVA followed by post hoc Bonferroni test was done for the open-field novel object exploration test; light/dark box test was compared with one-way ANOVA and both two-way ANOVA and repeated measurement ANOVA were applied for the puzzle box test, showing similar results. Mann–Whitney tests (nonparametric test) were performed for both GluA1 cluster number and size counting analysis.