Linking NMDA Receptor Synaptic Retention to Synaptic Plasticity and Cognition.

NMDA receptor (NMDAR) subunit composition plays a pivotal role in synaptic plasticity at excitatory synapses. Still, the mechanisms responsible for the synaptic retention of NMDARs following induction of plasticity need to be fully elucidated. Rabphilin3A (Rph3A) is involved in the stabilization of NMDARs at synapses through the formation of a complex with GluN2A and PSD-95. Here we used different protocols to induce synaptic plasticity in the presence or absence of agents modulating Rph3A function. The use of Forskolin/Rolipram/Picrotoxin cocktail to induce chemical LTP led to synaptic accumulation of Rph3A and formation of synaptic GluN2A/Rph3A complex. Notably, Rph3A silencing or use of peptides interfering with the GluN2A/Rph3A complex blocked LTP induction. Moreover, in vivo disruption of GluN2A/Rph3A complex led to a profound alteration of spatial memory. Overall, our results demonstrate a molecular mechanism needed for NMDAR stabilization at synapses after plasticity induction and to trigger downstream signaling events necessary for cognitive behavior.

Introduction

NMDA-type glutamate receptors (NMDARs) are key mediators of excitatory synaptic transmission in the brain (Swanger and Traynelis, 2018) contributing to synaptic plasticity and relevant for many forms of learning and memory (Morris et al., 1986, Lynch, 2004, Kullmann and Lamsa, 2007). However, the precise molecular mechanism by which postsynaptic NMDARs are retained at the synapse to allow for activity-dependent plasticity and expression of cognitive functions is still mostly unexplored.

The functional and pharmacological properties, the interacting proteins, and the subcellular localization of NMDARs strictly depend on their subunit composition, namely, the combination of the obligatory subunit GluN1 with the regulatory subunits GluN2 (A–D) and GluN3 (Paoletti et al., 2013). NMDAR subunit composition not only changes during neuronal development (Bellone and Nicoll, 2007) but at mature synapses can be modulated by synaptic activity and sensory experiences (Philpot et al., 2001, Philpot et al., 2003, Philpot et al., 2007, Sawtell et al., 2003, Yashiro et al., 2005) and can profoundly modify neuronal circuits and behavior. Recently, the unbalance in NMDAR subunit composition was recognized as a pivotal feature of several common neurological disorders (Sanz-Clemente et al., 2013, Lai et al., 2014, Shohami and Biegon, 2014, Gardoni and Bellone, 2015, Mellone et al., 2015).

GluN2A-containing NMDARs are rather stable at synapses, and several mechanisms have been called upon to regulate their synaptic retention. The C-terminal domain (CTD) of GluN2A enables interaction with the PDZ domain of scaffolding proteins, anchoring the receptor at the postsynaptic membrane as well as expressing correct downstream events (Horak et al., 2014, Shipton and Paulsen, 2013; Lussier et al., 2015, Stanic et al., 2015). Interestingly, mice lacking the CTD of GluN2A display impaired hippocampal long term potentiation (LTP) (Sprengel et al., 1998). Our group has previously reported Rabphilin3A (Rph3A) as a novel GluN2A synaptic partner needed to stabilize GluN2A/PSD-95 complex at the postsynaptic density (PSD). Disruption of the Rph3A/GluN2A/PSD-95 complex reduced GluN2A synaptic retention associated with increased GluN2A-containing NMDARs endocytosis (Stanic et al., 2015). Notably, this event is also associated with reduced dendritic spine density both in vitro and in vivo (Stanic et al., 2015).

Many studies assigned specific roles for synaptic GluN2A-containing NMDARs in the induction of LTP and Long-Term Depression (LTD) (Liu et al., 2004, Foster et al., 2010, Kellermayer et al., 2018). Deletion of GluN2A leads to reduced hippocampal LTP and impaired spatial learning (Sakimura et al., 1995, Kiyama et al., 1998, Kannangara et al., 2015). The use of GluN2A-specific antagonists prevented LTP but not LTD (Liu et al., 2004). Similarly, selective inhibition of GluN2A-containing receptors with low Zn2+ concentrations impaired LTP but not LTD (Papouin et al., 2012). Finally, different studies demonstrated an increase in GluN2A-containing NMDARs at the PSD following LTP induction (Barria and Malinow, 2002, Bellone and Nicoll, 2007, Peng et al., 2010, Baez et al., 2013), suggesting that NMDAR trafficking at synapses could play a relevant role in these events. However, many open questions remain about (1) the role of the complex of proteins responsible for synaptic retention of GluN2A-containing NMDARs in synaptic plasticity and (2) the link between GluN2A-containing NMDARs and cognitive behavior. Starting from these questions here we analyzed the role of Rph3A and Rph3A/GluN2A complex in the functional and morphological modifications of excitatory synapses following induction of LTP as well as in hippocampal NMDAR-dependent behaviors such as spatial learning.

Results

Rph3A-Positive Dendritic Spines Are Characterized by an Increased Spine Head Area and PSD Size

Rph3A is a known vesicle-associated presynaptic protein (Li et al., 1994, Burns et al., 1998) also highly enriched in dendritic spines at the lateral domain of the PSD (Stanic et al., 2015). Pre-embedding immunohistochemistry for Rph3A in rat stratum radiatum of the CA1 region of hippocampus revealed that 42.428 ± 2.301% of presynaptic terminals (n = 3632) and 48.275 ± 2.331% of dendritic spines (n = 3632) display Rph3A labeling (Figures 1A and 1B), thus indicating a similar enrichment of the protein at pre- and postsynaptic sites. We analyzed possible morphological differences between Rph3A positive (Rph3A+) and negative (Rph3A−) dendritic spines. Rph3A+ spines exhibit a highly significant increased spine head area (***p < 0.001; Figures 1C and 1D and Table 1) and PSD length (***p < 0.001; Figures 1E and 1F and Table 1) and also an augmented PSD thickness (*p < 0.05; Figures 1G and 1H and Table 1) compared to Rph3A− ones, thus defining a selective accumulation of the protein in more “mature” synapses.

Figure 1

Morphological Analysis of Rph3A-Positive and Rph3A-Negative Dendritic Spines

(A and B) Representative electron micrographs of stratum radiatum CA1 region of Rph3A negative (A, left image; Rph3A−) and positive (B, right image; Rph3A+) spinous synapses, respectively. Scale bar, 125 nm.

(C and D) Shifted distribution of spine head area toward bigger values in Rph3A+ spines (blue; n = 689/1,500, 3 rats) compared with Rph3A− spines (red; n = 811/1,500, 3 rats; p < 0.001, Mann-Whitney Rank Sum Test).

(E and F) Shifted distribution of PSD length toward bigger values in Rph3A+ spines (blue; n = 689/1,500, 3 rats) compared with Rph3A− spines (red; n = 811/1,500, 3 rats; p < 0.001, Mann-Whitney Rank Sum Test).

(G and H) Shifted distribution of PSD thickness toward bigger values in Rph3A+ spines (blue; n = 689/1,500, 3 rats) compared with Rph3A− spines (red; n = 811/1,500, 3 rats; p < 0.05; Mann-Whitney Rank Sum Test).

Table 1

Morphological Analysis of Rph3a+ and Rph3A− Dendritic Spines (n = 3 Rats, 500 Spines/Rat)

Labeling	Rph3A− Spines (n = 811/1,500)	Rph3A+ Spines (n = 689/1,500)	p Value (Rph3A+ versus Rph3A−)
PSD length (nm)	215 ± 2.38	232 ± 2.87	<0.001
PSD thickness (nm)	54.5 ± 0.96	56.7 ± 1.06	0.049
Spine Head Area (nm2)	99.9 ± 2.40	143 ± 3.54	<0.001

Modulation of Rph3A/GluN2A Complex by Long-Term Potentiation

It is well known that induction of LTP leads to an accumulation of GluN2A-containing NMDARs at the excitatory PSDs (Barria and Malinow, 2002, Grosshans et al., 2002, Bellone and Nicoll, 2007). By using a previously validated protocol, here we treated rat primary hippocampal neurons with Forskolin/Rolipram/Picrotoxin cocktail to induce chemical LTP (cLTP; Otmakhov et al., 2004, Dinamarca et al., 2016). Quantification of GluA1 Ser845 phosphorylation showed a significant increase of AMPA subunit phosphorylation levels after cLTP (Esteban et al., 2003, Oh et al., 2006, Hu et al., 2007, Makino et al., 2011; Figure S1). We observed that cLTP was sufficient to increase the levels not only of GluN2A (**p < 0.01; Figure 2A) but also of Rph3A (*p < 0.05; Figure 2A) in postsynaptic membrane fractions (Triton insoluble fractions, TIF; Gardoni et al., 2006) as measured 15 min after cLTP induction. Conversely, induction of cLTD (Oh et al., 2006, Marcello et al., 2013) did not induce any modification of GluN2A and Rph3A synaptic localization (Figure S2).

Figure 2

Effect of Long-Term Potentiation on Rph3A/GluN2A Complex

(A) Bar chart (left) and representative western blots (right) for GluN2A and Rph3A protein levels in TIF from hippocampal primary cultures after cLTP. cLTP induction was performed in artificial cerebrospinal fluid (ACSF) without MgCl2, plus 50 μM Forskolin, 0.1 μM Rolipram, and 100 μM Picrotoxin for 16 min. Control groups were kept in normal ACSF. After that, cells were incubated back in ACSF with MgCl2 for 15 min (n = 9, t test; *p < 0.05; **p < 0.01; data are expressed as mean ± SEM). Molecular weight markers are indicated on the right.

(B) Dendritic spine positivity for Rph3A after cLTP in primary cultures (t test) and representative images (scale bar, 5 μm). **p < 0.01; data are expressed as mean ± SEM.

(C) Morphological analysis of Rph3A-positive spines before and after cLTP treatment (t test). *p < 0.05; data are expressed as mean ± SEM.

(D) Bar chart and representative images of PLA for Rph3A:PSD-95 (red bar) and Rph3A:GluN2A (green bar) in control cultures (CTR) and after cLTP (scale bar, 5 μm). Merge panels are shown in the right. *p < 0.05; **p < 0.01; data are expressed as mean ± SEM.

(E) Bar graph and representative blots of GluN2A and Rph3A protein levels in TIF from hippocampal cultures after cLTP, with (cLTP + Anis) or without (cLTP) protein synthesis inhibitor Anisomycin (40 μM) (RM one-way ANOVA, n = 8); *p < 0.05; data are expressed as mean ± SEM. Molecular weight markers are indicated on the right.

(F) Column graph and representative images of Puro-PLA analysis for Rph3A newly synthetized (Rph3A:puromycin) in dendritic and somatic compartments after cLTP in primary cultures (n = 39–43, scale bar, 10 μm).

(G) Bar graph and representative images of GluN2A and Rph3A protein levels in hippocampal TIF from young mice housed in enriched environment (EE) compared with standard cages (CTRL). *p < 0.05; data are expressed as mean ± SEM. Molecular weight markers are indicated on the right.

The confocal microscopy analysis in GFP-transfected primary hippocampal neurons revealed that cLTP significantly increased the percentage of Rph3A+ spines compared with basal conditions (Figure 2B; **p < 0.01). A careful morphological analysis of dendritic spines showed that Rph3A was selectively enhanced in mushroom spines after cLTP (Figure 2C; *p < 0.05); no alteration of Rph3A localization was observed in stubby or thin spines (Figure 2C).

We performed Proximity Ligation Assay (PLA) to address whether Rph3A accumulation in spines leads also to its interaction with the GluN2A/PSD-95 complex (Stanic et al., 2015). Analysis of PLA clusters after cLTP in GFP-transfected neurons (Figure 2D) showed a statistically significant increase in Rph3A/PSD-95 (**p < 0.01) and GluN2A/Rph3A (*p < 0.05) interactions within dendritic spines.

The observed increase of Rph3A levels at postsynaptic sites after cLTP induction could be the result of a modulation of protein trafficking or novel protein synthesis. To clarify this point, we blocked protein synthesis through anisomycin (40 μM) during cLTP. Protein levels of GluN2A and Rph3A at synapses did not differ in the presence or absence of anisomycin (Figure 2E), thus suggesting that novel protein synthesis does not play a key role in these events. To confirm these results, we performed analysis of Rph3A local synthesis after cLTP through a PLA of puromycin and Rph3A (puro-PLA assay; see Transparent Methods). This assay allows one to monitor the synthesis of novel Rph3A molecules in the different subcellular compartments (Dieck et al., 2015). As shown in Figure 2F, we did not observe any difference in the density of puromycin/Rph3A clusters between control and cLTP-treated neurons both in the soma and along dendrites. Overall, these in vitro data indicate that the cLTP-triggered accumulation of GluN2A/Rph3A/PSD-95 complexes at synapse cannot be ascribed to Rph3A de novo synthesis.

Housing animals in enriched environment conditions induces neuronal plasticity events and accumulation of GluN2A-containing NMDARs at synapses (Philpot et al., 2001, Philpot et al., 2003, Philpot et al., 2007, Sawtell et al., 2003, Yashiro et al., 2005, Grilli et al., 2009, Summa et al., 2011, Bonfiglio et al., 2018). To evaluate whether Rph3A localizes at the postsynapse also after this in vivo plasticity, we purified the postsynaptic fraction from hippocampi of mice housed in enriched environment for 3 months or in standard cages. As previously reported, enriched environment promoted the increase of GluN2A-containing NMDARs at synapses (Figure 2G; *p < 0.05; Philpot et al., 2001, Philpot et al., 2003, Philpot et al., 2007, Sawtell et al., 2003, Yashiro et al., 2005). Similarly, the enriched environment induced also a significant increase of Rph3A levels in the postsynaptic fraction (Figure 2G; *p < 0.05).

Phospholipase C Activation Promotes Rph3A/GluN2A Interaction in the Postsynaptic Density

Different putative molecular mechanisms could be envisaged for the increased formation of the synaptic Rph3A/GluN2A complex induced by LTP. It is well known that LTP induces activation of tyrosine kinases (src/fyn) in the postsynaptic compartment leading to phosphorylation of NMDAR subunits (Nakazawa et al., 2001, Liu et al., 2004). In particular, GluN2A phosphorylation at Tyr1387, within the GluN2A domain involved in the interaction with Rph3A (Stanic et al., 2015), has been put forward for consideration (Yang and Leonard, 2001). However, as shown in Figure S3A, cLTP did not change the phosphorylation of GluN2A Tyr1387 in cultured cells. Moreover, co-localization analysis in COS-7 cells transfected with Rph3A and GluN2Awt/GluN2A-Y1387E (mimicking phosphorylation) showed that GluN2A phosphorylation in this tyrosine residue did not alter the capability of the subunit to interact with Rph3A (Figure S3B). Overall, these experiments indicate that GluN2A phosphorylation in Tyr1387 does not represent the molecular event regulating GluN2A/Rph3A complex formation following LTP induction.

Rph3A, through both its C2A and C2B domains, binds inositol triphosphate (IP3) in a Ca2+-dependent manner (Montaville et al., 2008, Ferrer-Orta et al., 2017). Interestingly, IP3 and Ca2+ binding to the C2A domain are reciprocally modulated in a positive manner. In particular, Ca2+ induces a conformational rearrangement of a specific Rph3A loop (namely, CBL3), which is involved in IP3 binding (Coudevylle et al., 2008, Guillén et al., 2013). Notably, IP3 and Ca2+ regulate also the formation of Rph3A complex with GluN2A (Stanic et al., 2015). Phospholipase C (PLC) cleaves phosphatidylinositol 4,5-bisphosphate (PIP2) into IP3 and diacylglycerol; afterward, IP3 releases Ca2+ from the ER, suggesting that activation could modulate the formation of Rph3A/GluN2A. PLCγ and PLCβ isoforms are both localized at the excitatory synapse and functionally associated to TrkB (Gottschalk et al., 1999) and group I metabotropic receptors (mGluR1/mGluR5; Chuang et al., 2001, Hannan et al., 2001), respectively. Activation of group I of metabotropic glutamate receptors (mGluRs) through DHPG (50 μM) increased postsynaptic levels of both Rph3A (Figure 3A, *p < 0.05) and GluN2A (Figure 3A, *p < 0.05) but not GluN2B (Figure 3A). Under the same experimental conditions, DHPG augmented Rph3A-GluN2A interaction as evaluated by the co-immunoprecipitation assay (Figure 3B, ***p < 0.001). Analysis of pERK phosphorylation was performed as a positive control of DHPG treatment (Gallagher et al., 2004; Figure S4, **p < 0.01).

Figure 3

mGluR/PLC Pathway Modulates Rph3A Synaptic Localization and Interaction with NMDAR

(A) Column chart and representative blots in TIF from primary cultures treated with 3,5-R,S-DHPG (DHPG, 50 μM for 15 min). DHPG increased GluN2A and Rph3A significantly, whereas no difference could be inferred for GluN2B (t test, n = 5). *p < 0.05; data are expressed as mean ± SEM. Molecular weight markers are indicated on the right.

(B) Graph and blots for co-immunoprecipitation of Rph3A and GluN2A from P2 fraction of primary cultures treated with DHPG. The analysis revealed increased binding of Rph3A with GluN2A (n = 4). ***p < 0.001; data are expressed as mean ± SEM. Molecular weight markers are indicated on the right.

(C) Bar graph and representative blots for Rph3A and tubulin in TIF from primary hippocampal neurons treated with BDNF (50 ng/mL, 3 h; n = 5; t test). *p < 0.05; data are expressed as mean ± SEM. Molecular weight markers are indicated on the right.

Finally, as further demonstration of the role of PLC in the modulation of Rph3A retention at the excitatory synapse, we showed that also the treatment of hippocampal primary cultures with the TrkB agonist Brain Derived Neurotrophic Factor (BDNF, 3 h, 50 ng/mL), leading to activation of the PLCγ pathway, increased Rph3A protein levels in the postsynaptic fraction (Figure 3C, *p < 0.05).

Modulation of Rph3A/GluN2A Complex Governs Plasticity at Molecular and Structural Level

To test the direct role of Rph3A in the membrane localization of GluN2A-containing NMDARs following induction of cLTP, we used a small hairpin RNA for Rph3A (tGFP-shRph3A) to downregulate Rph3A in the presence or absence of cLTP induction, and we evaluated the surface localization of the GluN2A subunit (Figure 4A). As expected (Baez et al., 2018), induction of cLTP promoted an accumulation of GluN2A at the cell surface (Figure 4A, ***p < 0.001). Notably, Rph3A silencing prevented GluN2A accumulation at the cell surface following induction of cLTP (Figure 4A, ***p < 0.001 shSCR-cLTP versus shRph3A-cLTP), thus suggesting that formation of GluN2A/Rph3A complex is required for the stabilization of GluN2A at synapses following LTP.

Figure 4

Effect of Modulation of Rph3A/PLC Pathway on GluN2A Surface Staining, GluA1 Synaptic Localization and Spine Morphology

(A) Bar graph and representative images of GluN2A surface expression before (CTRL) and after cLTP (cLTP) in tGFP-shScramble (shSCR) or tGFP-shRph3A transfected neurons. cLTP induction was performed in ACSF without MgCl2, plus 50 μM Forskolin, 0.1 μM Rolipram and 100 μM Picrotoxin for 16 min. Control groups were kept in normal ACSF. After that, cells were incubated back in ACSF with MgCl2 for 15 min (one-way ANOVA Tukey post hoc n = 41–46, scale bar, 4 μm); ***p < 0.001.

(B) Graph chart and representative images of GluA1 cluster width before and after cLTP in hippocampal neurons transfected with shSCR or shRph3A (one-way ANOVA Tukey post hoc, n = 9–11, scale bar, 4 μm); *p < 0.05, ***p < 0.001.

(C) Column graphs and representative images of shSCR or shRph3A transfected neurons before and after cLTP. cLTP induction was performed in ACSF without MgCl2, plus 50 μM Forskolin, 0.1 μM Rolipram, and 100 μM Picrotoxin for 16 min. Control groups were kept in normal ACSF. After that, cells were incubated back in ACSF with MgCl2 for 2 h (one-way ANOVA Tukey post hoc, n = 8, scale bar, 4 μm); **p < 0.01, *p < 0.05.

(D) Inhibition of Phospholipase C with U73122 (1 μM) during cLTP on hippocampal primary cultures recapitulates control levels of Rph3A and GluN2A in TIF (one-way ANOVA Bonferroni post hoc, n = 5–7); *p < 0.05, ***p < 0.001. Molecular weight markers are indicated on the right.

(E) Bar chart and representative images of spine morphology analysis after cLTP induction in the presence (cLTP + U73122) or absence (cLTP) of U73122 (1 h after cLTP; one-way ANOVA Bonferroni post hoc, n = 25, scale bar, 4 μm); **p < 0.01. ***p < 0.001.

It is well known that LTP also increases GluA1 clusters at synapses (Malinverno et al., 2010). As shown in Figures 4B and 4C, induction of cLTP increased GluA1 cluster width (Figure 4B, ***p < 0.001), paralleled by augmented spine density (Figure 4C, **p < 0.01) and spine head width (Figure 4C, *p < 0.05). Interestingly, Rph3A silencing completely prevented any modification of GluA1 cluster size (Figure 4B, *p < 0.05 shSCR-cLTP versus shRph3A-cLTP) and spine density (Figure 4C, *p < 0.05 shSCR-cLTP versus shRph3A-cLTP). Moreover, Rph3A silencing prevented the enlargement of spine head width induced by LTP (Figure 4C, p > 0.05, shSCR-CTRL versus shRph3A-cLTP).

To evaluate the role of PLC in these events, we inhibited PLC activity through U73122 (1 μM) during cLTP. Co-incubation with PLC inhibitor recapitulated basal levels of both Rph3A (*p < 0.05, U73122 + cLTP versus cLTP) and GluN2A (*p < 0.05, U73122 + cLTP versus cLTP) in the postsynaptic fraction (Figure 4D). It is well known that LTP increases dendritic spine density in hippocampal neurons (Chidambaram et al., 2019). In addition, we previously reported that Rph3A silencing or disruption of its interaction with GluN2A is sufficient to reduce spine density in resting conditions (Stanic et al., 2015). Notably, we now observed that inhibition of PLC with U73122 during cLTP not only decreased synaptic localization of Rph3A/GluN2A complex (Figure 4D) but also completely blocked the increase in dendritic spine density produced by cLTP (Figure 4E; ***p < 0.001 cLTP versus control; **p < 0.01 cLTP versus cLTP + U73122). Overall, these results suggest that PLC activation is required to modulate Rph3A/GluN2A retention at synapses needed for structural modifications following induction of LTP in primary hippocampal neurons.

Overall, the above-described results indicate that induction of cLTP in primary hippocampal neurons promotes synaptic enrichment of the Rph3A/GluN2A complex, demonstrating also a key role for Rph3A in LTP-dependent molecular and morphological modifications of dendritic spines, namely, LTP-dependent trafficking of AMPARs and formation of novel dendritic spines.

GluN2A/Rph3A Complex Is Necessary for LTP Induction and Spatial Learning

We previously reported that perturbing GluN2A/Rph3A interaction in vivo with TAT-2A40 interfering peptide (containing the GluN2A1349-1389 domain involved in the interaction with Rph3A) decreases the amplitude of NMDAR-mediated currents and GluN2A levels at dendritic spines (Stanic et al., 2015, Stanic et al., 2017). Here, we acutely treated adult mice with TAT-2A40 or its control TAT-scramble (TAT-SCR) peptide (3 nmol/g, i.p., single injection). One hour after the treatment, animals were sacrificed for ex vivo molecular and electrophysiological analyses. As previously reported (Stanic et al., 2015, Stanic et al., 2017), treatment with TAT-2A40 leads to a specific reduction of GluN2A but not GluN2B subunits at synapses leading to an overall significant decrease of synaptic GluN2A/GluN2B ratio with no modification of GluN1 (see Figure S5). LTP was induced by stimulation of Schaffer collaterals in CA1 stratum radiatum (see Transparent Methods). As expected, in hippocampal slices from animals treated with TAT-SCR peptide we observed the induction and the maintenance of the LTP (Figures 5A–5C). On the contrary, in hippocampal slices from animals treated with TAT-2A40 peptide, LTP induction was completely impaired (Figures 5A–5C; **p < 0.01).

Figure 5

In Vivo Effect of Rph3A/GluN2A Interfering Peptide on LTP Induction and Spatial Memory

(A–C) Somatic whole-cell voltage-clamp recordings were made from CA1 pyramidal cells using 2–6 ΩM electrodes. The internal solution contained (in mM) 115 CsMeSO4, 20 CsCl2, 10 HEPES, 2.5 MgCl2, 4 NaATP, 0.4 NaGTP, 10 NaCreatine, and 0.6 EGTA (pH 7.2). Synaptic responses were collected with a Multiclamp 700B amplifier (Axon Instruments, Foster City, CA, USA), filtered at 2 kHz, digitized at 5 kHz, and analyzed online using Igor Pro Software (Wavemetrics, Lake Oswego, OR, USA). All data are expressed as mean ± SEM. Cells were held at −70 mV, and LTP protocol was induced by pairing the cell at 0 mV at a frequency of 2 Hz for 90 s. The amplitude of TAT-2A40 treated animals, as well as LTP kinetic, was completely impaired compared with TAT-SCR (AMPLITUDE, **p < 0.01, unpaired t test).

(D and E) Mean discrimination index and mean exploration time evaluated in the Spatial Object Recognition test, 60 min after treatment; ***p < 0.001 versus TAT-SCR Student's t test; §§§p < 0.001 versus corresponding stationary object, TAT-SCR; two-way ANOVA followed by Bonferroni test.

(F and G) Cumulative mean of horizontal (F) and vertical (G) counts evaluated for 10 min in an automated activity cage. N = 10 animals for each group.

Changes in NMDAR synaptic levels are triggered by synaptic plasticity and by spatial memory formation (Baez et al., 2018). In particular, an increase in the synaptic GluN2A/GluN2B subunit ratio could act as a stabilizer of synaptic/circuital changes, hence leading to stabilization of memory consolidation, particularly spatial representations (Baez et al., 2018). Starting from these considerations, we performed a spatial object recognition behavioral task to assess the effect of disrupting Rph3A/GluN2A interaction on spatial learning, in the same experimental conditions used for electrophysiology (TAT-2A40 versus TAT-SCR, 3 nmol/g, i.p., single injection). In the Spatial Object Recognition test, locating the object to a novel configuration during the T2 phase induced a significant treatment effect in terms of mean discrimination index between TAT-SCR- and TAT-2A40-treated mice (t18 = 5.61, ***p < 0.0001, Figure 5D). During T1 phase, all groups of mice showed a similar mean exploration time for each object (TAT-SCR: Object 1 = 11.7 ± 1.0; Object 2 = 12.4 ± 0.96. TAT-2A40: Object 1 = 12.9 ± 0.82; Object 2 = 13.7 ± 0.75). During T2 phase two-way ANOVA revealed differences among groups (treatment effect: F(1,36) = 4.269, p < 0.05; object effect: F(1,36) = 8.79, p = 0.0053; interaction treatment x object: F(1,36) = 12.28, p = 0.001). Post hoc analysis revealed that the mean exploration time of the displaced object was significantly higher than that of the stationary object after treatment with TAT-SCR (Figure 5E; $$$p < 0.001 versus the corresponding stationary object). Conversely, no difference was shown in the mean exploration time between the two objects after treatment with TAT-2A40 (Figure 5E). The mean number of horizontal and vertical movements did not change between the two groups (Horizontal counts: t18 = 1.252, p = 0.23, Figure 5F; Vertical counts: t18 = 0.325, p = 0.74, Figure 5G).

Discussion

The mechanism by which GluN2A-containing NMDARs accumulate at the synapse following activity-dependent plasticity and how this relates to the expression of given cognitive functions has been approached. Here we indicate that the formation of the Rph3A/GluN2A complex is needed for molecular and structural modifications of dendritic spines induced by LTP. In vivo disruption of Rph3A/GluN2A interaction by an interfering peptide leads to both LTP and spatial memory impairment corroborating this finding.

Rph3A is a Rab effector protein involved in neurotransmitter release at the presynaptic terminal, and its conformation and activity are strictly modulated by the presence of Ca2+ ions and IP3 (Coudevylle et al., 2008, Montaville et al., 2008, Guillén et al., 2013). Recently, Rph3A has been detected also at dendritic spines, where it interacts with and promotes synaptic retention of GluN2A-containing NMDARs (Stanic et al., 2015). Here we explored the molecular mechanisms by which Rph3A binds the GluN2A subunits following different paradigms of activity-dependent plasticity induced both in vitro, as cLTP or mGluR5 activation, and in vivo through enriched environment. All these forms of plasticity converge on promoting Rph3A accumulation at dendritic spines and its Ca2+/IP3-dependent interaction with the NMDAR subunit.

Our electron microscopy data eventually clarify the pre- and postsynaptic enrichment of Rph3A. We show by pre-embedding immunohistochemistry that about half of presynaptic terminals as well as dendritic spines in hippocampus display Rph3A, thus suggesting a similar distribution of the protein at the two sides of the excitatory synapse. Importantly, we observed that spines in which we detect Rph3A have an increased spine head area and PSD length and thickness, suggesting a higher stability of neuronal transmission through these Rph3A-positive connections.

Accumulation of GluN2A-containing NMDARs at synapse is a highly validated molecular event occurring after LTP induction (Barria and Malinow, 2002, Grosshans et al., 2002, Bellone and Nicoll, 2007). Overall, these studies indicate that the GluN2A-containing NMDARs move at the synapses thanks to mobilization of preassembled NMDARs from non-synaptic pools. However, the molecular mechanisms responsible for a selective accumulation of receptors containing the GluN2A subunit are far to be understood. We show that Rph3A represents a needed protein in these processes thanks to its selective binding to GluN2A but not to other GluN2-type regulatory NMDAR subunits (Stanic et al., 2015). Different experimental protocols can be used to induce cLTP in dissociated hippocampal neurons, the treatment with glycine being the more commonly used (Lu et al., 2001). Here we show that induction of cLTP by using the Forskolin/Rolipram/Picrotoxin cocktail in primary cultures acts by augmenting AMPAR surface insertion and phosphorylation at GluA1-Ser845 (Joiner et al., 2010) as well as NMDAR activity/synaptic stabilization. Notably, Rph3A interaction with GluN2A plays a key role in the increase of NMDAR activity at synapses. In addition, our results show now that induction of LTP promotes accumulation of Rph3A in mushroom-type dendritic spines, where it interacts with GluN2A-containing NMDARs thus leading to synaptic retention of the receptor (Stanic et al., 2015). This event is not associated with de novo Rph3A protein synthesis as indicated by cLTP experiments performed in the presence of anisomycin or by the puromycin-PLA assay. Moreover, activation of the mGluR/PLC pathway plays a fundamental role in these events also confirming that Ca2+/IP3 strictly modulate the capability of Rph3A to interact with protein partners (Coudevylle et al., 2008, Guillén et al., 2013).

A number of previous studies demonstrated an involvement of mGluRs/PLC pathway in both LTP and LTD. Even if the role of mGluR-dependent synaptic LTD in physiology and disease is well established (see for review Lüscher and Huber, 2010), activation of group I mGluRs through DHPG can facilitate also LTP through a PLC signaling cascade (Cohen et al., 1998, van Dam et al., 2004, Mellentin et al., 2007). Interestingly, inhibition of phospholipase C by U73122 abolished the priming of LTP induced by DHPG (Cohen et al., 1998). Moreover, the LTP induction protocol can increase the amount of GluN2A at CA1 synapses in a mGluR5 and NMDAR-dependent manner (Matta et al., 2011). In particular, in hippocampal CA1 pyramidal neurons, the developmental GluN2A/GluN2B switch driven acutely by activity requires activation of NMDARs and mGluR5 and it involves PLC activation (Matta et al., 2011). Here, we demonstrate that in vitro activation of PLC is essential for cLTP-associated biochemical and morphological plasticity, driving Rph3A/GluN2A summon in PSD. Furthermore, activation of different PLC-grouped metabotropic receptors increased Rph3A protein levels at synapses, indicating PLC as a key enzyme upstream of Rph3A/GluN2A complex formation.

Furthermore, Rph3A silencing or disruption of Rph3A/NMDAR complex by an interfering peptide not only blocks GluN2A accumulation at postsynaptic membranes but also prevents the induction of LTP and formation of new spines. Notably, treatment of animals with either Rph3A silencing or cell permeable peptide disrupting Rph3A/NMDAR complex impairs the acquisition of spatial memories.

Our data are in close agreement with previous reports showing that rising of hippocampal GluN1/GluN2A NMDARs at synapses appears to be a general feature after novel spatial memory acquisition (Cercato et al., 2017). As reviewed by Baez et al. (2018), the GluN1/GluN2A subunits increase at synapse starting from about 20 to 30 min after plasticity induction or memory acquisition could represent a check point or a synaptic tag for plasticity establishment or memory consolidation (Baez et al., 2018). Overall, our results demonstrate that GluN2A interaction with Rph3A is needed for NMDAR stabilization in hippocampal PSD after LTP induction and to trigger downstream signaling necessary for LTP synaptic adaptation and cognitive behavior.

Limitations of the Study

Even if our results strengthen the putative role of Rph3A as an attractive pharmacological target for several neurological conditions in which GluN2A-containing NMDARs are not correctly functioning (Sanz-Clemente et al., 2013, Lai et al., 2014, Shohami and Biegon, 2014, Gardoni and Bellone, 2015, Mellone et al., 2015), additional long-term studies in disease models are needed to confirm the involvement of Rph3A and Rph3A/GluN2A complex in these brain disorders.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.