All the experimental procedures followed the guidelines established by the Italian Council on Animal Care and were approved by the Italian Government Decree No. 27/2010 (see Supplementary Information) and the Italian Legislation (L.D. no. 26/2014). All efforts were made to minimize the number of animals used and their sufferings. All animals were housed with 12/12-h light/dark cycle with food and water available ad libitum. Embryonic (E18) primary cultures of mouse hippocampal neurons were established from Eps8 KO¹ and wild type (WT) mice obtained from breeding settled up in our animal facility as described by Menna et al. (2013). Briefly, hippocampi were dissociated by treatment with trypsin (0.125% for 15 min at 371°C), followed by trituration with a polished Pasteur pipette. The dissociated cells were plated onto glass coverslips coated with poly-L-lysine at density of 400 cells/mm². For low density neuronal cultures used in paired electrophysiological experiments, dissociated cells were plated onto glass coverslips coated with poly-L-lysine at density of 150–200 cells/mm². The cells were maintained in Neurobasal (Invitrogen) with B27 supplement and antibiotics, 2 mM glutamine and 12.5 mM glutamate (neuronal medium).

Western blot analysis was performed in hippocampal tissue of Eps8 KO and WT adult (5 months of age) littermates mice. To obtain a preparation that contains selectively proteins of the excitatory PSD, subcellular fractionation was performed as reported previously with minor modifications (Gardoni et al., 2001b). Hippocampi were homogenized in 0.32 M ice-cold sucrose containing the following (in mM): 1 HEPES, 1 MgCl₂, 1 EDTA, 1 NaHCO₃ and 0.1 PMSF at pH 7.4, in the presence of a complete set of proteases inhibitors (Complete; Roche Diagnostics) and phosSTOP Phosphatase Inhibitor (Roche Diagnostics). The homogenized tissue was centrifuged at 13,000× g for 10 min (P2 fraction). The pellet was resuspended in buffer containing 75 mM KCl and 1% Triton X-100 and centrifuged at 100,000× g for 1 h. The supernatant was stored and referred as Triton X-100-soluble fraction (TSF). The final pellet (triton insoluble fractions, TIF) was homogenized in a glass–glass potter in 20 mM HEPES. TIF was used instead of the classical PSD because the amount of the starting material was very limited. The protein composition of this preparation was, however, carefully tested for the absence of presynaptic markers (i.e., synaptophysin; Gardoni et al., 2009). Similar protein yield was obtained in TIF purified from hippocampal tissue of all experimental groups.

WB analysis was performed in homogenate and TIF fractions. Protein samples were separated onto an acrylamide/bisacrylamide gel at the appropriate concentration and transferred to a nitrocellulose membrane. Nitrocellulose articles were blocked with 10% albumin in tris-buffered saline (TBS) and then incubated overnight at 4° with the primary antibodies. After extensive rinsing in TBS/0.1% Tween 20, the nitrocellulose articles were then incubated with horseradish peroxidase-conjugated secondary antibodies (goat anti-rabbit, for polyclonal antibodies, diluted 1:10,000 (Pierce); goat anti-mouse, for monoclonal antibodies, diluted 1:10,000 (Pierce). Membrane development was performed with the reagent Clarity Western ECL Substrate (Bio-Rad) or LiteAblot TURBO (Euroclone) and labeling was visualized by Chemidoc Imaging System and ImageLab software (Bio-Rad). For quantification, each protein was normalized against the corresponding tubulin band. The following unconjugated primary antibodies were used: GluN2A (diluted 1:1,000; Sigma-Aldrich M264), GluN2B (diluted 1:1,000; Neuromab 75–101), PSD-95 (diluted 1:2,000; Neuromab 75–028), tubulin (diluted 1:10,000; Sigma-Aldrich T9026), Phospho-p44/42 MAPK (Erk1/2; Thr202/Tyr204; diluted 1:2,000; Cell signaling 9101), p44/42 MAPK (Erk1/2; diluted 1:2,000; Cell signaling 9102), total GluN2B (Mellone et al., 2015) p1472 (diluted 1:1,000; Calbiochem 454583).

Living neurons were incubated for 10 min with rabbit antibodies directed against extracellular epitopes of GluN2B (diluted 1:100, Alomone Labs AGC-003), washed and fixed with 4% paraformaldehyde and 4% sucrose as described (Joshi et al., 2017). The guinea pig anti-Bassoon (diluted 1:500, Synaptic System 141004) was used to double stain the cultures. Secondary antibodies were conjugated with Alexa-488 and Alexa-555 (Alexa-Invitrogen). Images were acquired using a Zeiss LSM800 confocal microscope equipped with a Plan-Apochromat 63×/1.40 oil objective. Acquisition parameters (i.e., laser power, gain and offset) were kept constant among different experimental settings. Surface synaptic GluN2B staining was quantified as follows. GluN2B-, bassoon-positive puncta and GluN2B puncta colocalizing with bassoon (both number and size) were measured using ImageJ 1.46r software in GluN2B and bassoon confocal images after setting a fixed threshold using the “analyze particle” and “image processing” functions. The “synaptic surface GluN2B” is calculated by the ratio of the “area of surface GluN2B colocalizing with bsn/area of surface GluN2B.” Approximately 10–15 fields were acquired per independent experiments. All data are results of at least three independent experiments.

For the staining of F actin, vehicle- and Latrunculin B (LatB)-treated cultures were fixed with 4% paraformaldehyde and 4% sucrose as described (Bedogni et al., 2016). AlexaFluor555 conjugated–phalloidin (diluted 1:200, Molecular Probes A34055) and the mouse monoclonal PSD-95 (diluted 1:400, Neuromab 75–028) were used.

Whole cell voltage-clamp recordings were performed on WT and transgenic E18 hippocampal neurons maintained in culture for 13–15 DIV. During recordings cells were bathed in a standard external solution containing (in mM): 125 NaCl, 5 KCl, 1.2 MgSO₄, 1.2 KH₂PO₄, 2 CaCl₂, 6 glucose and 25 HEPES-NaOH, pH 7.4. Recording pipettes were fabricated from borosilicate glass capillary using an horizontal puller (Sutter Instruments) inducing tip resistances of 3–5 MΩ and filled with a standard intracellular solution containing (in mM): 130 Cs-gluconate, 8 CsCl, 2 NaCl, 4 EGTA, 10 HEPES-NaOH, 2 MgCl₂, 4 MgATP and 0.3 Tris-GTP.

For miniature AMPA-NMDA EPSC recordings, cells were bathed in a free Mg²⁺ solution containing (in mM) 125 NaCl, 5 KCl, 1.2 KH₂PO₄, 2 CaCl₂, 6 glucose and 25 HEPES-NaOH, tetrodotoxin (TTX) 0.001, Strychnine 0.001 and bicuculline methiodide 0.02, glycine 0.01 (pH 7.4; Tocris). To record the pure NMDA currents 6-cyano-7-dinitroquinoxaline-2,3-dione (CNQX) 20 μM (Tocris) were added to standard extracellular solution to block the AMPA component of the current. The patch pipette electrode contained the following (in mM): 130 CsGluconate, 8 CsCl, 2 NaCl, 10 HEPES, 4 EGTA, 4 MgATP and 0.3 Tris-GTP.

AMPA and NMDA evoked currents were recorded in isolated pairs of neurons in low-density cultures. Neurons were bathed in a standard external solution containing (in mM): 125 NaCl, 5 KCl, 1.2 MgSO₄, 1.2 KH₂PO₄, 2 CaCl₂, 6 glucose and 25 HEPES-NaOH, pH 7.4. Neurons were held at −70 mV or at +40 and eEPSC evoked by a 100-mV depolarization pulse in the presynaptic cell lasting 1 ms. The AMPA/NMDA ratio was calculated by estimating the respective AMPA and NMDAR current on the traces at +40 mV based on their different time courses (NMDA component was calculated 40 ms after stimulus artifact).

For whole-cell total currents recording were bathed in a free Mg²⁺ solution containing (in mM) 125 NaCl, 5 KCl, 1.2 KH₂PO₄, 2 CaCl₂, 6 glucose and 25 HEPES-NaOH, TTX 0.001, Strychnine 0.001 and bicuculline methiodide 0.02, glycine 0.01 (pH 7.4). NMDA whole-cell total currents were elicited by fast local application of saturating NMDA concentrations, which activate both synaptic and extrasynaptic receptors. NMDA (200 μM) will be applied with or without inhibitors (ifenprodil 3 μM or Zinc 10 nM in zinc-free tricine solution to block GluN2B and GluN2A respectively; Paoletti et al., 1997).

To measure rectification, spontaneous AMPA-mediated currents were measured at −60 and at +50 mV, under conditions in which spontaneous AMPA currents could be isolated. For these experiments the ACSF contained 0.01 μM TTX (to raise spike threshold), APV (100 μM) and MK801 (50 μM) to block NMDA currents, and bicuculline (20 μM) to block inhibitory potentials. The internal solution contained spermine (120 μM) was Cs-based to facilitate voltage-clamping at +50 mV. To compute the rectification index the average peak amplitude at +50 was divided by the average peak amplitude at −60; the smaller this value, the more rectification. To induce synaptic scaling, neurons were treated with 1 μM TTX for 48 h and then mEPSC currents were recorded.

For the analysis of the tonic, inward, noisy current recordings were carried out in [0 Mg²⁺]_e, 20 μM CNQX, bicuculline (50 μM) and TTX (1 μM), before and after blocking NMDAR function with APV (50 μM) or ifenprodil (3 μM). Two alternative approaches were used to quantify tonic NMDAR current. First, it was assessed as the shift in holding current resulting from NMDA blocker (AP5) application, second, AP-5-associated change in a baseline noise. Quantification of background noise was obtained plotting all values of recordings traces and comparing the fluctuation of values distribution.

Recordings were performed at room temperature in voltage clamp mode using a Multiclamp 700B amplifier (Molecular Devices) and pClamp-10 software (Axon Instruments). Series resistance ranged from 10 MΩ to 20 MΩ and was monitored for consistency during recordings. Cells in culture with leak currents >100 pA were excluded from the analysis. Signals were amplified, sampled at 10 kHz, filtered to 2 or 3 KHz and analyzed using pClamp 10 data acquisition and analysis program.

Statistical analysis was performed using Prism6 (GraphPad), data are presented as mean ± SEM from the indicated number of experiments. After testing whether data were normally distributed or not, the appropriate statistical test, followed by specific multiple comparison post hoc tests, has been used as indicated in figure legends. Kolmogorov–Smirnov test was used to determine significance in cumulative distributions of mEPSC amplitudes. Differences were considered to be significant if P < 0.05 and are indicated by one asterisk; those at P < 0.01 are indicated by double asterisks; those at P < 0.001 are indicated by triple asterisks, those at P < 0.0001 are indicated by four asterisks.

To test whether the lack of Eps8 affects NMDAR function, NMDA-mediated miniature excitatory currents (mEPSCs) were recorded in mature WT and Eps8 KO primary hippocampal neurons (DIV 14–15) in Mg²⁺ free solution in the presence of TTX (1 μM) to block action potential-mediated release, the GABA-A receptor antagonist bicuculline (20 μM) and the selective antagonist of AMPA/kainate glutamate receptors CNQX (20 μM). A significantly lower frequency of NMDA-mediated mEPSCs was recorded in mutant neurons compared to WT (Hz, WT = 2.153 ± 0.162, n = 11 cells; EPS8 KO = 1.491 ± 0.218, n = 17 cells, Unpaired t-test *P = 0.0374 data are expressed and mean ± SEM; Figures 1A,B), suggesting a reduced NMDA synaptic component in mature Eps8 KO hippocampal cultures.

In line with a reduced NMDA synaptic component, the NMDA evoked current appeared to be lower in Eps8 KO neurons relative to controls. Indeed, temporal separation of AMPA and NMDA evoked components (obtained by maintaining neurons in 1.2 mM Mg²⁺ to block NMDA currents and record the AMPA component at –70 mV, or by stimulating neurons at +30 mV to record the NMDAR component of the EPSC) revealed a significant increase of the AMPA/NMDA ratio in Eps8 KO neurons (WT = 2.7 ± 0.2, n = 9; KO = 4.12 ± 0.3, n = 5; Mann Whitney test *P = 0.0159; Figures 1E–G). On the contrary the AMPA evoked component, obtained by measuring the EPSC peak amplitude at −70 mV (see also “Materials and Methods” section), was not changed between WT and Eps8 KO neurons (pA, WT = 37.2 ± 11, n = 9; KO = 35.4 ± 5.5, n = 6; Mann Whitney test P = 0.9).

Notably, Eps8 KO miniature NMDA currents displayed a larger amplitude and a slower decay time (pA: WT = 8.527 ± 0.3878; EPS8 KO = 10.7 ± 0.7462; data shown as cumulative probability, Kolmogorov-Smirnov test, * = 0.0348, D = 0.4500. Decay time, ms: WT = 94.8 ± 4; EPS8 KO = 125 ± 6; Unpaired t-test **P = 0.0010; Figures 1C,D).

Altogether these results indicate that Eps8 KO hippocampal neurons display reduced NMDAR synaptic activity which is also characterized by higher amplitude and a slower decay time, raising the possibility that mutant synapses may be endowed with different subtypes of NMDAR subunits (Cull-Candy and Leszkiewicz, 2004; Erreger et al., 2005; Santucci and Raghavachari, 2008).

Given that Eps8 KO hippocampal neurons display slower decay of NMDA current (Figure 1D), we assessed the possible occurrence of GluN2-related differences in NMDAR stoichiometry at the mutant and WT synapses. To probe the subunit composition for GluN2B and GluN2A, we tested the effect of ifenprodil, a GluN2B subunit-specific antagonist of NMDARs (Tovar and Westbrook, 1999; Cull-Candy et al., 2001), in 14–15 DIV WT and Eps8 KO primary hippocampal cultures.

Three micromolar ifenprodil, a concentration which does not affect GluN2A-containing receptors (Tovar and Westbrook, 1999), blocked ~40% of the synaptic NMDA-mediated mEPSCs in Eps8 KO neurons vs. ~10% in WT neurons (Hz, WT ctrl: 2.130 ± 0.15, n = 8 cells, WT + ifenprodil: 1.99 ± 0.18 n = 9 cells; EPS8 KO Ctrl: 1.432 ± 0.11 n = 12 cells, EPS8 KO + ifenprodil: 1.044 ± 0.127, n = 9 cells; pA: WT Ctrl: 8.4 ± 0.3, WT + ifenprodil: 7.675 ± 0.6; EPS8 KO Ctrl: 10.8 ± 0.78, EPS8 KO + ifenprodil: 7.0 ± 0.7 Mann Whitney Test *P = 0.0270; Unpaired t-test **P = 0.0065; Figures 2A–C) indicating that mature EPS8 KO primary hippocampal cultures display an altered GluN2B/2A synaptic composition. To further investigate the contribution of specific receptor subtypes to NMDA currents in WT vs. Eps8 KO neurons, we recorded the total current produced by fast local application of saturating NMDA concentrations (200 μM, 5 s), which activate both synaptic and extrasynaptic NMDARs. NMDA application produced large responses, with a comparable amplitude between WT and EPS8 KO neurons (pA/pF: WT, 19.24 ± 1.5 n = 14 cells; EPS8 KO, 17.76 ± 2 n = 16 cells; Unpaired t-test P = 0.6200; Figures 2D,E). However, mutant neurons were more sensitive to the non-competitive GluN2B blocker, as shown by the significantly larger inhibition of NMDA-mediated total current in EPS8 KO hippocampal neurons exposed to 3 μM ifenprodil with respect to WT neurons (% ifenprodil inhibition: WT, 43.71 ± 6 n = 8 cells; EPS8 KO, 61.8 ± 3 n = 8 cells, Mann Whitney Test *P = 0.0127; Figure 2F). This evidence is consistent with a higher amount of GluN2B-containing receptors in mutant neurons. Notably, exposure of WT and KO neurons to tricine (10 mM), a chelator of Zn²⁺ ions which tonically inhibits GluN2A-contaning NMDARs by binding to the Zn²⁺ binding site on the GluN2A subunits (Paoletti et al., 1997), resulted in enhanced NMDA responses at a significantly higher extent in WT relative to KO neurons (Figure 2G; Mann Whitney Test, ***P < 0.001 and *P = 0.0220). These data reveal a reduced GluN2A- and an increased GluN2B-component of NMDAR activity in Eps8 KO hippocampal neurons with respect to age-matched WT primary hippocampal cultures suggesting that the lack of Eps8 affects the process of NMDAR complex maturation in neurons.

Since Eps8 regulates actin dynamics in filopodia and spines, we aimed to assess whether the integrity of the actin cytoskeleton affects the subunit composition of NMDARs. To this aim, primary cultured hippocampal neurons were treated with LatB (300 nM from 10 DIV to 14 DIV) to depolymerize actin and disrupt its coupling to NMDARs during synapse development in vitro (Allison et al., 1998; Sattler et al., 1999). We then measured the total current produced by fast local application of NMDA (200 μM, 5 s), before and after bath application of ifenprodil (3 μM). Chronic treatment with LatB does not affect the current density of NMDA responses (pA/pF: vehicle, 23.34 ± 1.752 n = 18 cells; LatB chronic, 24.71 ± 2.09 n = 17 cells; Unpaired t-test P = 0.602; Figures 2H,I). However, LatB-treated cultures were more sensitive to the GluN2B blocker, ifenprodil, as shown by the increased percentage of inhibition of NMDA currents (% ifenprodil inhibition: vehicle, 55.73 ± 4.5 n = 19 cells; LatB chronic, 71.21 ± 1.9 n = 15 cells, Unpaired t-test **P = 0.0078; Figure 2J). Furthermore, we performed live staining of LatB-or vehicle-treated hippocampal cultures with antibodies directed against the extracellular epitopes of GluN2B. Increased exposure of GluN2B-containing receptors at the synaptic surface, identified by the presynaptic marker bassoon, was detected upon LatB treatment (Figures 2K,L; synaptic surface GluN2B: Vehicle = 1 ± 0.074 n = 13 fields; LatB = 1.531 ± 0.073 n = 13 fields in at least three independent experiments. Mann-Whitney Rank Sum Test, P ≤ 0.0001), suggesting that the integrity of F-actin is required to set the proper NMDAR composition at the synapse. The efficacy of LatB (300 nM from 10 DIV to 14 DIV) in inducing F-actin depletion in neuronal processes was assessed by Phalloidin staining (Supplementary Figure S2). These data indicate that actin depolymerization, similarly to Eps8 genetic depletion, results in enhanced exposure of GluN2B-containing receptors at the synaptic surface.

The role of Eps8 in regulating the NMDA receptor subunit composition at synapses was also examined by Western blotting (WB) analysis in homogenates and postsynaptic TIF (Gardoni et al., 2001a, 2009) prepared from hippocampi of adult WT or Eps8 KO mice (5 months of age). At first, we confirmed that TIF sample was actually enriched in postsynaptic proteins. As shown in Figure 3A, the protein levels of the GluN2A-subunit of NMDAR and of the postsynaptic scaffold protein, PSD-95, were strongly higher in TIF fraction compared to the total homogenate.

No difference in the expression levels of any tested NMDAR subunit was found in the homogenate prepared from hippocampi of Eps8 KO compared to WT adult mice (Figures 3B,C). Conversely, profound differences in NMDAR composition were detected in the postsynaptic compartment as indicated by the immunoblot and relative quantification of TIF fraction (Figures 3B,D). In line with electrophysiological data reported in Figures 2A–G, Eps8 KO tissue was characterized by a significantly higher GluN2B content in the synaptic TIF fraction (Kruskal-Wallis test followed by Dunn’s Multiple Comparison test **P < 0.01, n = 4) in the absence of any significant alteration in the GluN2A subunit (Figures 3B,D).

Previous studies reported that direct phosphorylation of the GluN2B subunit by tyrosine kinases in a specific phospho-site is crucial for the dynamic regulation of GluN2B trafficking/turnover (Dunah and Standaert, 2001). Of relevance, GluN2B phosphorylation is mainly restricted to a synaptic fraction (Dunah et al., 2000; Gardoni et al., 2006). We therefore monitored GluN2B Tyr phosphorylation using phospho-specific antibodies raised against the Tyr1472 phospho-site in the C-terminal domain of the receptor subunit (Y1472-GluN2B site). Interestingly, the staining pattern produced in TIF fractions by the GluN2B phospho-specific antibody paralleled the altered level of total GluN2B (Figures 3B–D), thus indicating a significant increase of the GluN2B phosphorylated form in Eps8 KO hippocampal neurons compared to WT (Kruskal-Wallis test followed by Dunn’s Multiple Comparison test *P < 0.05 4, n = 4). Furthermore GluN2B Y1472 phosphorylation disrupts the binding to the AP-2 clathrin-associated adaptor protein complex, which targets proteins for endocytosis (Lavezzari et al., 2003). Therefore, an increase in Y1472 GluN2B phosphorylation levels may be linked to decreased GluN2B endocytosis, thus representing a possible mechanism underlying the increase of GluN2B-containing NMDARs in Eps8 KO synapses.

To assess whether the increase of GluN2B-containing receptors is restricted only at synaptic regions, we evaluated the levels of phosphorylated ERK1/2 in hippocampi of mutant adult mice relative to WT. Indeed, it has been shown that synaptic NMDARs promote ERK phosphorylation, whereas extrasynaptic NMDARs lead to ERK dephosphorylation and subsequent inactivation (Hardingham and Bading, 2010). As shown in Figures 3B,C, no increase of phosphorylated ERK1/2 levels (Kruskal-Wallis test followed by Dunn’s Multiple Comparison test, n = 4) occurred in Eps8 KO homogenate. This result suggests that the ratio of synaptic vs. extrasynaptic GluN2B-containing NMDARs is not changed probably due to an increase GluN2B-containing NMDARs also at the extrasynaptic level. To further investigate this issue we measured the tonic inward noisy current in [0 Mg²⁺]e and 20 μM CNQX, which has been demonstrated to involve extrasynaptic NMDARs activation in cortical pyramidal neurons (LoTurco et al., 1990; Gottesman and Miller, 2003; Povysheva and Johnson, 2012), possibly due to ambient glutamate or glutamate spillover (Sah et al., 1989; Cavelier et al., 2005; Le Meur et al., 2007). Recordings of NMDA-mediated mEPSCs in [0 Mg²⁺]e at −70 mV from EPS8 KO primary hippocampal neurons displayed a higher noisy signal than WT (Figures 4A,B). Bath application of the NMDAR blocker, APV (50 μM), to WT and EPS8 KO primary cultures reduced noise and abolished the tonic inward current (Figures 4A,B). Moreover, Eps8 KO hippocampal neurons display enhanced background noise respect to WT, measured as the variance of the change of baseline noise at −70 mV before and after bath application of ifenprodil (3 μM; Unpaired t-test **P = 0.0044; Figures 4C,D). These data suggest that Eps8 KO neurons display increased content of GluN2B-containing receptors at both synaptic and extrasynaptic sites.

All together these data demonstrate that Eps8 KO hippocampal neurons display altered expression of GluN2A- and GluN2B-containing NMDARs and indicate that mutant neurons show a significantly higher amount of both synaptic and extrasynaptic GluN2B-containing NMDARs.