GRIP1 regulates synaptic plasticity and learning and memory

Significance

AMPA receptors (AMPARs) are the principle postsynaptic glutamate receptors mediating fast excitatory synaptic transmission in the brain. Regulation of synaptic AMPAR expression is required for the expression of synaptic plasticity and normal brain function. The turnover of AMPARs within synapses is highly dynamic, and the molecular mechanisms underlying AMPAR trafficking remain unclear. Here we report that GRIP1, an AMPAR-binding protein, plays an essential role in delivering AMPAR into synapses during synaptic plasticity, particularly in long-term potentiation. In addition, the deletion of Grip1 causes synaptic plasticity deficits and impaired learning and memory. Our study reveals a mechanism through which GRIP1 regulates AMPAR trafficking and impacts activity-dependent synaptic strengthening, as well as learning and memory.

Abstract

Hebbian plasticity is a key mechanism for higher brain functions, such as learning and memory. This form of synaptic plasticity primarily involves the regulation of synaptic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) abundance and properties, whereby AMPARs are inserted into synapses during long-term potentiation (LTP) or removed during long-term depression (LTD). The molecular mechanisms underlying AMPAR trafficking remain elusive, however. Here we show that glutamate receptor interacting protein 1 (GRIP1), an AMPAR-binding protein shown to regulate the trafficking and synaptic targeting of AMPARs, is required for LTP and learning and memory. GRIP1 is recruited into synapses during LTP, and deletion of Grip1 in neurons blocks synaptic AMPAR accumulation induced by glycine-mediated depolarization. In addition, Grip1 knockout mice exhibit impaired hippocampal LTP, as well as deficits in learning and memory. Mechanistically, we find that phosphorylation of serine-880 of the GluA2 AMPAR subunit (GluA2-S880) is decreased while phosphorylation of tyrosine-876 on GluA2 (GluA2-Y876) is elevated during chemically induced LTP. This enhances the strength of the GRIP1–AMPAR association and, subsequently, the insertion of AMPARs into the postsynaptic membrane. Together, these results demonstrate an essential role of GRIP1 in regulating AMPAR trafficking during synaptic plasticity and learning and memory.

The ability of the brain to learn, remember, and adapt requires changes in synaptic connectivity (1, 2). Synapses are dynamic and subject to cellular mechanisms that strengthen and weaken these neural connections throughout the lifespan of an organism. Associative, or Hebbian, synaptic plasticity is widely thought to be a key cellular mechanism underlying information storage (3, 4). In Hebbian plasticity, correlated action potential firing between presynaptic and postsynaptic neurons causes long-term potentiation (LTP) of synaptic strength. Conversely, uncorrelated spiking between presynaptic and postsynaptic neurons induces a long-term depression (LTD) at shared synapses (5). The molecular mechanisms of Hebbian plasticity are highly complex and currently under intense scrutiny, but the detailed picture remains incomplete.

Both LTP and LTD can be mediated by presynaptic mechanisms, such as enhanced or reduced neurotransmitter release probability, as well as by postsynaptic mechanisms, including changes in the sensitivity, properties, or abundance of postsynaptic receptors (6, 7). The α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type glutamate receptors (AMPARs) are the principle glutamate receptors that mediate the majority of fast excitatory synaptic transmission in the mammalian central nervous system, and the postsynaptic abundance of AMPARs is directly proportional to synaptic strength and modulated by a dynamic turnover of these receptors in response to synaptic activity (8, 9). Shuttling of AMPARs into and out of the postsynaptic membrane are the major mechanisms of NMDA receptor (NMDAR)-dependent LTP and LTD, respectively, at excitatory synapses (9). AMPARs are tetrameric assemblies composed of GluA1-4 subunits, which are subjected to specialized posttranslational modification and protein interactions that regulate AMPAR conductance and localization (9, 10). Many AMPAR-interacting proteins play critical roles in synaptic plasticity. For example, synapse-associated protein (SAP97) directly binds the GluA1 subunit and may promote AMPAR trafficking and LTP (11–13). Protein interacting with C-kinase 1 (PICK1), a GluA2 subunit-binding protein, functions to remove AMPAR from synapses and causes internalization of synaptic AMPARs, and deficits in Hebbian plasticity have been reported in PICK1 mutant mice (14, 15).

Glutamate receptor interacting protein (GRIP1) is a scaffolding protein that has seven postsynaptic density 95/discs large/zona occludens (PDZ) domains (16, 17). It interacts directly with the C terminus of both GluA2 and GluA3 through the fourth and fifth PDZ domains. GRIP1 has been shown to regulate the surface expression and synaptic stabilization of AMPARs (18, 19). Our previous studies, as well as the work of others, have suggested that there might be distinct pools of GRIP1 that differentially regulate AMPAR trafficking, although the predominate role of GRIP1 is to deliver AMPAR to the surface and stabilize them at synapses (20–23). Moreover, this regulation of AMPAR trafficking by GRIP1 has been shown to be essential for certain forms of synaptic plasticity, such as cerebellar LTD and homeostatic scaling (20, 24–26). Nevertheless, the need for GRIP1 in the expression of Hebbian LTP is unknown.

Here we investigated the function of GRIP1 in LTP and its role in learning and memory. We found that GRIP1 is recruited into synapses with AMPARs, and that the GRIP1–AMPAR interaction is enhanced during LTP. Moreover, the loss of GRIP1 blocks the activity-induced accumulation of synaptic AMPARs. Finally, Grip1 knockout (KO) mice exhibit learning and memory deficits, likely due to the compromised plasticity at active synapses in these animals. Taken together, our findings reveal an essential role of GRIP1 in synaptic plasticity and cognitive functions.

Results

GRIP1 Is Recruited into Synapses in Response to Chemically Induced LTP.

To investigate the role of GRIP1 in regulating AMPAR trafficking during LTP, we first examined GRIP1 expression and subcellular localization before and after LTP induction. Cultured rat cortical neurons were treated with glycine to induce chemically mediated LTP (cLTP), a well-established stimulation protocol mimicking NMDAR-dependent LTP (NMDAR-LTP) (27), and postsynaptic densities (PSDs) were isolated. Consistent with previous results, we saw significant increases in synaptic GluA1, GluA2, and GluA3 AMPAR subunits after cLTP, while total AMPAR subunit expression did not change (Fig. 1 A–C). Intriguingly, we observed a synaptic accumulation of GRIP1 protein following cLTP even though the total GRIP1 level remained unchanged (Fig. 1 A–C), suggesting that GRIP1 is recruited into synapses with AMPARs during LTP.

Fig. 1.

GRIP1 is recruited into synapses during cLTP. (A) Representative Western blots of proteins from PSD and total cell lysates (total) isolated from rat cortical neurons treated with (+) or without (−) glycine. (B) Quantification of protein levels in PSD (n = 5; Mann–Whitney U test). (C) Quantification of protein levels in total cell lysates. (n = 8; Student’s t test). (D) Representative Western blots of GRIP1 in P2 and S2 fractions isolated from rat cortical neurons treated with (+) or without (−) glycine. (E) Quantification of GRIP1 protein level in each fraction following cLTP (n = 15 to 16; Mann–Whitney U test). (F) Model of GRIP1 translocation during cLTP. Data are presented as mean ± SEM. n.s., not significant. **P < 0.01.

Our previous study suggests that there might be two membrane-associated pools of GRIP1 with distinct functions: the intracellular GRIP1 retains AMPARs intracellularly, while the plasma membrane-associated GRIP1 anchors AMPARs at the cell surface (20). In addition, there is abundant non–membrane-associated GRIP1 in the cytosol. To examine how these pools change during LTP, we performed subcellular fractionation to determine GRIP1 subcellular distribution on cLTP. Our data show that cLTP did not significantly change GRIP1 levels in either cytosol (S2) or membrane (P2) fractions (Fig. 1 D and E). Together, these results indicate that synaptic enrichment of GRIP1 after cLTP is due to translocation from the intracellular membrane pool, likely from the endosomes, rather than from the cytosolic non–membrane-associated pool (Fig. 1F).

Activity-Dependent Up-Regulation of Synaptic AMPARs Requires GRIP1 Expression.

To directly determine the functional roles of GRIP1 in LTP, we next used neurons derived from GRIP1 conditional KO mice. As has been shown previously (20, 25), GRIP1 protein was completely ablated in neurons transduced with lentiviruses expressing Cre recombinase (EGFP-IRES-Cre) (Fig. 2A). In unstimulated basal conditions, synaptic GluA1, GluA2, and GluA3 expression in Grip1-deleted neurons was comparable to that in control EGFP-expressing wild-type (WT) neurons (Fig. 2 A and B). Following cLTP induction, as expected, in control WT neurons, we observed a significant up-regulation of synaptic AMPARs. However, in Grip1 KO neurons, glycine treatment failed to induce synaptic enrichment of AMPARs (Fig. 2 A and C). Therefore, we concluded that GRIP1 is essential for AMPAR delivery to synapses during LTP.

Fig. 2.

Activity-dependent up-regulation of synaptic AMPARs requires GRIP1 expression. (A) Representative Western blots of proteins from PSD and total cell lysates from WT or Grip1 KO mouse neurons treated with (+) or without (−) glycine. (B) Quantification of protein levels in PSD under basal conditions in WT and Grip1 KO mouse neurons (n = 9 to 11; Student’s t test). (C) Quantification of protein levels in PSD from WT or Grip1 KO mouse neurons treated with (+) or without (−) glycine (n = 15; Mann–Whitney U test). Data are presented as mean ± SEM. n.s., not significant. *P < 0.05; **P < 0.01; ***P < 0.001.

Grip1 KO Mice Exhibit Impaired NMDAR-Dependent LTP.

To further confirm the function of GRIP1 in LTP, we examined the requirement for GRIP1 in LTP by performing intracellular whole-cell electrophysiological recordings. We crossed floxed Grip1 (Grip1^fl/fl) mice with the pan-neuronal Nestin-Cre line to delete Grip1 in neurons at embryonic stages (28). GRIP1 protein was undetectable in the hippocampus of Nestin-Grip1^fl/fl mice at postnatal day 21 (Fig. 3A); thus, we performed whole-cell recordings of LTP in acute hippocampus slices prepared from Nestin-Grip1^fl/fl mice and control Grip1^fl/fl littermates at 3 to 4 wk of age. We recorded from CA1 pyramidal neurons because LTP at Schaffer collateral-CA1 synapses is well known to be dependent on NMDARs and expressed primarily by increased synaptic AMPAR abundance (6). Furthermore, LTP and AMPAR trafficking in CA1 neurons have been functionally linked to animal learning and memory (29, 30). Our data reveal an ∼50% decrease in LTP expression in CA1 pyramidal neurons of Nestin-Grip1^fl/fl compared with control Grip1^fl/fl littermates (Fig. 3 B–D). This reduced potentiation lasted at least 50 min in neurons of Grip1 ^fl/fl; Nestin-Cre⁺ mice compared with control neurons from Grip1^fl/fl; Nestin-Cre-mice (Fig. 3D). The impaired LTP in Nestin-Grip1^fl/fl mice is not caused by inefficient induction, because excitatory postsynaptic current (EPSC) amplitudes at baseline and during LTP induction were comparable in Nestin-Grip1^fl/fl mice and control Grip1^fl/fl littermates under our induction protocol (Fig. 3E). Together, these data show that GRIP1 is required for induction and maintenance of LTP.

Fig. 3.

Grip1 KO mice exhibit impaired NMDAR-dependent LTP. (A) Representative Western blots of total cell lysates from hippocampus in Nestin-Grip1^fl/fl mice (Grip1 KO) and control Grip1^fl/fl littermates (WT). (B) Representative evoked EPSCs obtained from CA1 pyramidal neurons before and after LTP induction in response to 0.1-Hz stimulation of Schaffer collaterals. Dash lines represent baseline EPSCs. Solid lines represent EPSCs 40 min after LTP induction. (C) Averaged EPSC amplitudes normalized to baseline responses. Arrow indicates the pairing induction (200 pulses at 2 Hz paired with 0 mV depolarization). (D) Statistics of LTP at 30–50 min (n = 16 cells from 8 control Grip1^fl/fl littermate group; n = 17 cells from 8 Nestin-Grip1^fl/fl mice; Mann–Whitney test). (E) Averaged EPSC amplitudes at baseline and during induction. Data are presented as mean ± SEM *P < 0.05.

Grip1 KO Mice Display Impaired Learning and Memory.

Synaptic plasticity is a key molecular mechanism underlying learning and memory. In light of the impaired synaptic plasticity observed in Grip1 KO neurons, we next evaluated the cognitive functions of Grip1 KO mice. As Nestin-Cre mice exhibit cognitive behavioral abnormalities, including reduced contextual and cued conditioned fear responses (31, 32), we used calmodulin-dependent kinase II (CaMKII)-Grip1^fl/fl mice to perform behavior experiments, with Grip1^fl/fl littermates serving as controls. We chose to use an inhibitory avoidance (IA) task, as IA learning induces LTP and synaptic AMPAR incorporation in the hippocampus and thus is dependent on hippocampal function (30, 33–35). As described previously (36), adult mice were habituated to a chamber divided into two separate compartments: light and dark. When placed in the light side, both CaMKII-Grip1^fl/fl mice and control Grip1^fl/fl littermates entered the dark side after a short time (Fig. 4 A and B). At 24 h after training, during which a mild foot shock was delivered after the mice entered the dark side, control mice showed a significant longer step-through latency, indicating a clear IA memory (Fig. 4 A and B). However, CaMKII-Grip1^fl/fl mice failed to learn the IA task and showed no significant difference between before and after the IA training (Fig. 4 A and B).

Fig. 4.

Grip1 KO mice display impaired learning and memory. (A) Cartoon illustration of the IA task. (B) Quantifications of the latency to cross over to the dark chamber at training and 24 h later in CaMKII-Grip1^fl/fl (Grip1 KO) and control Grip1^fl/fl littermates (WT) (n = 15 WT; n = 13 Grip1 KO; Mann–Whitney U test). (C) Quantification of total, central, and peripheral ambulatory activities in open-field chambers (n = 15 WT; n = 13 Grip1 KO; Mann–Whitney U test). (D) Quantification of the anxiety index, calculated as the activity in the peripheral divided by the activity in the center for each mouse (n = 15 WT; n = 13 Grip1 KO; Student’s t test). Data are presented as mean ± SEM, n.s., not significant; ***P < 0.001.

To rule out potential behavioral interference, such as general motor activity and anxiety, we also performed open-field test with the CaMKII-Grip1^fl/fl mice and their control Grip1^fl/fl littermates. No significant differences in locomotion and time spent in the center vs. the periphery of the chamber were observed in these mice (Fig. 4 C and D), supporting our conclusion that GRIP1 is specific and essential for hippocampal-dependent learning and memory.

GRIP1-GluA2 Association Is Enhanced during LTP.

Finally, we examined the underlying molecular mechanism through which GRIP1 regulates AMPAR trafficking during LTP. GRIP1 directly binds with the C-termini of GluA2/3 AMPAR subunits, and the interaction is highly regulated by activity and plays critical roles in AMPAR trafficking (10). For example, during cerebellar LTD, GluA2 S880 is phosphorylated, which disrupts GRIP1 binding to GluA2 and in turn increases GluA2–PICK1 interaction to promote AMPAR endocytosis (37–39). In addition, another study showed that regulation of GluA2 Y876 phosphorylation could gate GluA2 S880 phosphorylation (40). During cerebellar LTD, dephosphorylation of GluA2 Y876 occurs, which increases GluA2 phospho-880, decreases GRIP1 interactions, and thus accelerates GluA2 internalization (40). We recently showed that phosphorylation of GluA2 Y876 directly increases GRIP1 binding to GluA2, and that this regulation is necessary for synaptic upscaling (41).

To gain insight into the mechanisms of these actions, we first examined the phosphorylation levels of GluA2 S880 and Y876 during cLTP. Following glycine treatment, we observed a significant increase in GluA2 phospho-Y876 level while phospho-S880 level was decreased (Fig. 5 A and B). Given that phosphorylation of GluA2 Y876 increases GRIP1 binding but phospho-S880 inhibits it, these results imply that the GRIP1–GluA2 association might be enhanced during LTP. To confirm this, we performed coimmunoprecipitation experiments to directly examine GRIP1-GluA2 binding. Since GluA2 is a transmembrane protein, we used the membrane fraction (P2) to avoid the artificial binding of cytoplasmic GRIP1 with GluA2. As expected, we observed increased coimmunoprecipitation between GRIP1 and GluA2 following cLTP treatment (Fig. 5 C and D), indicating a stronger GRIP1–GluA2 association during LTP. Taken together, these results show that GRIP1–GluA2 interaction is enhanced during LTP, involving coordination of GluA2-Y876 and GluA2-S880 phosphorylation.

Fig. 5.

The GRIP1–GluA2 association is enhanced during LTP. (A) Representative Western blots of proteins from P2 in rat cortical neurons treated with (+) or without (−) glycine. (B) Quantification of phospho-Y876 and phospho-S880 levels following cLTP (n = 7 to 12; Student’s t test). (C) GluA2 was immunoprecipitated with specific GluA2 antibody from P2 from rat cortical neurons treated with (+) or without (−) glycine, followed by Western blot analysis of GRIP1 and GluA2. (D) Quantification of relative GRIP1–GluA2 interactions in P2 during cLTP (n = 10; Mann–Whitney U test). (E) Model of GRIP1 regulation of AMPAR trafficking during LTP. Data are presented as mean ± SEM. n.s., not significant; *P < 0.05; ***P < 0.001.

Discussion

In the present study, we have demonstrated an essential role of GRIP1 in synaptic plasticity and learning and memory. We found that GRIP1 traffics with AMPARs from intracellular membranes into synapses during LTP, and that the interaction between GRIP1 and AMPAR is also strengthened as a consequence of increased GluA2-Y876 phosphorylation and decreased GluA2-S880 phosphorylation. Grip1 KO neurons have impaired LTP, and Grip1 KO mice exhibit learning and memory deficits.

The regulation of GRIP1 in AMPAR trafficking has been a controversial subject. Some studies have shown that GRIP1 delivers AMPARs to the cell surface and stabilizes AMPARs at synapses (18, 42), while other studies have suggested that GRIP1 retains AMPAR intracellularly (43, 44). We recently provided evidence suggesting the presence of two membrane-associated pools of GRIP1, one pool associated with the plasma membrane to anchor AMPARs on the cell surface and the other pool in the cytoplasm to retain AMPARs within intracellular compartments (20). These two distinct pools function cooperatively to regulate AMPAR trafficking. For example, in homeostatic up-scaling, in which surface AMPARs are greatly up-regulated, synaptic GRIP1 is increased and the association between GRIP1 and synaptic AMPARs is strengthened, while intracellular GRIP1–AMPAR interaction is reduced. Furthermore, the increase in membrane-associated GRIP1 is accompanied by a decrease in cytosolic non–membrane-associated GRIP1, indicating a translocation of GRIP1 from the cytosol to membrane. Intriguingly, we did not observe any changes in cytosolic GRIP1 during LTP despite an increase in synaptic GRIP1 level, suggesting that the synaptic accumulation of GRIP1 during LTP is caused by translocation of GRIP1 from intracellular membrane compartments, likely from endosomes, rather than from the cytosolic non–membrane-associated pool (Fig. 5E). Although GRIP1 is required for both LTP and tetrodotoxin (TTX)-induced up-scaling, the mechanisms through which GRIP1 regulates these processes differ. These data, together with the findings of previous studies, indicate that there are distinct but overlapping characteristic features of Hebbian and homeostatic synaptic plasticity (45, 46).

GRIP1 binds directly with GluA2 and GluA3 AMPAR subunits but not with the GluA1 subunit; however, the role of GRIP1 in activity-dependent AMPAR trafficking is not subunit-specific. The cLTP-induced increases in synaptic AMPAR levels, including levels of GluA1, GluA2 and GluA3 subunits, are blocked in Grip1 KO neurons. Synaptic incorporation of calcium-permeable AMPARs (CP-AMPARs) following LTP induction, mostly with GluA1 homomers, has been reported (47, 48); however, these findings have not been consistently replicated (49, 50). Furthermore, the presence of CP-AMPARs at synapses during LTP, if any, is very brief before their replacement by GluA2-containing AMPARs, which are thought to be essential for LTP maintenance (48, 51). Therefore, GluA2-containing AMPARs play a major role in LTP, and GRIP1 regulates GluA1 expression by controlling GluA1-GluA2 heteromers. Notably, LTP is partially impaired in hippocampus slices of Grip1 KO mice. Some compensatory involvement of other AMPAR-binding proteins, such as the transmembrane AMPAR regulatory proteins, GRIP2, a GRIP1 homolog, PICK1, or GluA1-interacting partners such as SAP-97 or protein 4.1 N, also may contribute to this phenotype (10, 52).

AMPARs are subject to posttranslational modifications, including phosphorylation, ubiquitination, and palmitoylation. These modifications have a significant impact on AMPAR trafficking, primarily by affecting the binding of other proteins with AMPARs. Phosphorylation is the most extensively studied modification, and numerous phosphorylation sites have been characterized (9). For example, GluA1-S567 and GluA1-S831 can be phosphorylated by Ca²⁺/CaMKII (53, 54), and protein kinase C (PKC) phosphorylates GluA2-S863 and GluA2-S880 (37, 55). Along with serine/threonine phosphorylation, tyrosine phosphorylation of AMPARs also has been reported. GluA2-Y876 and GluA3-Y881 can be phosphorylated by the Src family tyrosine kinases (41, 56). These various phosphorylation modifications have distinct but significant roles in modulating the properties, function, and trafficking of AMPARs (9, 10). The binding affinity of GRIP1 for GluA2 is largely affected by the phosphorylation of GluA2-S880 and GluA2-Y876. GluA2-S880 phosphorylation prevents GRIP1 binding, while GluA2-Y876 phosphorylation enhances GRIP1–GluA2 interaction (40, 41). During cLTP, we observed a concomitant decrease in GluA2-S880 phosphorylation and an increase in phosphorylated GluA2-Y876. In addition, we found a stronger association of GRIP1 with GluA2 during LTP. These data support a model in which more GRIP1 binds with AMPARs as a result of coordinated GluA2-S880 and GluA2-Y876 phosphorylation to deliver them into synapses during LTP (Fig. 5E). Interestingly, a single point mutation that prevents phosphorylation of Y876 alone is insufficient to block LTP (41). It is possible that these two sites may function cooperatively and redundantly.

Because GRIP1 is critical for synaptic plasticity, and synaptic plasticity is crucial for cognitive function, we examined the role of GRIP1 in learning and memory with the knowledge that GRIP1 is essential for LTP. We found that Grip1 KO mice have significant impairments in learning and memory, supporting the notion that regulation of AMPAR trafficking during synaptic plasticity is important for brain function (1). Indeed, various GRIP1 single nucleotide polymorphisms have been reported to strongly associate with autism (19), and these variants have altered interactions with GluA2/GluA3 and thus affect AMPAR trafficking. In addition, a recent study on the GRIP1-binding protein GRASP1 showed that intellectually disability-associated GRASP1 mutations have convergent impairments in their interactions with GRIP1, which in turn blocks endosomal delivery of AMPARs to the synapses and LTP (36). Together, our findings elucidate the function of GRIP1 in synaptic plasticity and may shed light on the mechanisms underlying neurodevelopmental disorders characterized by synaptic and cognitive abnormalities.

Materials and Methods

Neuronal Culture.

Cortical neurons from embryonic day 18 rat or mouse pups were plated on poly-l-lysine–coated tissue culture dishes at a density of 75,000 cells/cm² in 5% horse serum (Invitrogen) containing Neurobasal medium (Invitrogen) supplemented with 2% B-27, 2 mM Glutamax, and 50 U/mL pen-strep. Neurons were switched to 1% horse serum containing Neurobasal medium after neurons grew attached to the plate, and were then treated with FDU (5 mM 5-fluoro-2′-deoxyuridine and 5 mM uridine) to inhibit glia proliferation at day in vitro (DIV) 5, and fed twice per week with glia-conditioned 1% horse serum containing Neurobasal medium with supplements.

Glycine-Induced LTP.

Cortical neurons at DIV 18 to 20 were first incubated in artificial cerebrospinal fluid (ACSF) containing 143 mM NaCl, 5 mM KCl, 10 mM Hepes, 10 mM glucose, and 2 mM CaCl₂ (pH 7.4) supplemented with 1 mM MgCl₂, 500 nM TTX, 20 μm bicuculine, and 1 μm strychnine at 37 °C for 30 min, and then treated with 200 μm glycine in ACSF with the same supplements except MgCl₂ for 10 min, followed by a 20-min recovery in ACSF supplemented with 1 mM MgCl₂, 500 nM TTX, 20 μm bicuculine, and 1 μm strychnine. The cells were then harvested for further experiments.

Lentivirus Generation.

The lentiviruses were generated based on a protocol provided by Carlos Lois of MIT, Cambridge, MA (57). Targeted cDNA for viral expression was first cloned into a pFUW vector containing a ubiquitin promoter and then cotransfected with Δ8.9 and VSVG packaging constructs into HEK293T cells when the cells were 90% confluent. At 1 d after transfection, the supernatant of HEK 293T cells containing released viruses was collected and concentrated by centrifugation at 250,000 × g for 2 h at 4 °C. The supernatant was discarded, and the pellet was resuspended in Neurobasal medium.

Subcellular Fractionation.

Rat or mouse cortical neurons were harvested in homogenate buffer (320 mM sucrose, 5 mM sodium pyrophosphate, 1 mM EDTA, 10 mM Hepes pH 7.4, 200 nM okadaic acid, 2.5 mM sodium orthovanadate, and protease inhibitor mixture [Roche]) and homogenized using a 26-gauge needle. The homogenate was then centrifuged at 800 × g for 10 min at 4 °C to yield P1 and S1. S1 was centrifuged at 17,000 × g for 20 min to yield P2 and S2. P2 was then resuspended in water adjusted to 4 mM Hepes pH 7.4, followed by 30 min of agitation at 4 °C. Suspended P2 was centrifuged at 25,000 × g for 20 min at 4 °C. The resulted pellet was resuspended in 50 mM Hepes pH 7.4, mixed with an equal volume of 1% Triton X-100, and agitated at 4 °C for 10 min. The PSD fraction was generated by centrifugation at 32, 000 × g for 20 min at 4 °C.

Coimmunoprecipitation.

The P2 membrane fraction was lysed in PBS containing 50 mM NaF, 5 mM sodium pyrophosphate, 1% Nonidet P-40, % sodium deoxycholate, 1 μM okadaic acid, 2.5 mM sodium orthovanadate, and protease inhibitor mixture (Roche). The anti-GluA2 antibody (032.19.9, made in- house) or control IgG antibody was precoupled to protein A Sepharose beads and incubated with 200 μg P2 protein in lysis buffer at 4 °C for 2 h. The beads were then washed in lysis buffer six times, followed by 2× sodium dodecyl sulfate (SDS) loading buffer elution. Bound proteins were resolved by SDS-polyacrylamide gel electrophoresis for Western blot analysis.

Antibodies.

The following antibodies were used: anti–beta-tubulin mAb (Sigma-Aldrich), anti-GluA1 N-terminal antibody mAb (4.9D, made in-house), anti-GluA2 N-terminal antibody mAb (032.19.9, made in-house), anti-GluA2 phospho-S880 specific mAb (02.22.4, made in-house), anti-GluA2 phospho-Y876 specific mAb (045.10.5, made in-house), anti-PSD95 mAb (NeuroMab), anti-GluA3 pAb (JH4300, made in-house), anti-GRIP1 mAb (BD Biosciences), anti-GRIP1 pAb (Chemicon), and anti-GRIP1 pAb (JH2260, made in-house).

Electrophysiology.

Paired littermates of Nestin-Grip1^fl/fl and Grip1^fl/fl mice (both males and females) at postnatal day 19 to 28 were anesthetized with the inhalation of isoflurane before decapitation. Then 300-μm-thick transverse hippocampal slices were prepared with a vibratome (Leica VT 1200s) in ice-cold oxygenated (95% O₂/5% CO₂) dissection buffer containing 210 mM sucrose, 7 mM glucose, 26.2 mM NaHCO₃, 2.5 mM KCl, 1 mM NaH₂PO₄, and 7 mM MgSO₄. Slices were recovered in a submersion chamber filled with oxygenated ACSF (119 mM NaCl, 26.2 mM NaHCO₃, 11 mM glucose, 2.5 mM KCl, 1 mM NaH₂PO₄, 2.5 mM CaCl₂, and 1.3 mM MgSO₄) at 36 °C for 30 min before recording.

For LTP recordings, slices were perfused in ACSF in the presence of 100 μM picrotoxin at room temperature. Hippocampal CA1 neurons were patched by glass pipettes (4 to 5 MΩ) which were filled with internal solution (115 mM Cs-MeSO₃, 0.4 mM EGTA, 5 mM TEA-Cl, 2.8 mM NaCl, 20 mM Hepes, 3 mM Mg-ATP, 0.5 mM Na₂-GTP, 10 mM Na phosphocreatine, and 5 mM QX-314, pH 7.2; osmolality 295 to 300 mOsm). Cells were held at −70 mV, and responses were evoked at 0.1 Hz by electrical stimulation (0.1 ms, 8 to 20 μA) via a bipolar electrode positioned at the midline of the Schaffer collateral. LTP was induced by a train of 200 pulses administered at 2 Hz paired with 0 mV depolarization. Signals were measured with MultiClamp 700B amplifier and digitized using a Digidata 1440A digitizer (Molecular Devices). Data acquisition were performed with pClamp 10.5 software and digitized at 10 kHz. Data are presented as responses averaged at 1-min intervals and then normalized to the average of baseline response. Access resistance (R_a) was monitored throughout the recording. Cells in which the R_a > 20 MΩ or the R_a varied by >20% were discarded.

Behavior Assays.

Adult male and female CaMKII-Grip1^fl/fl mice and Grip1^fl/fl littermates (4 to 5 mo old) were grouped and housed with both genotypes. For the open-field test, mice were placed in a photobeam- equipped plastic chamber (45 × 45 cm, PAS open-field system; San Diego Instruments) for free exploration for 30 min. The peripheral area (425 cm²) was defined by the two side-photobeams, 1-2 and 15-16, while the central area (1,600 cm²) was defined by photobeams 3 to 14 at each direction. Movements and rearing behavior were tracked using the SDI Photobeam Activity System (San Diego Instruments).

For the IA task, mice were handled for 3 min each day for 4 consecutive days before testing. The step-through IA apparatus (Gemini Avoidance System) consists of a rectangular chamber divided into two separate compartments (light and dark) connected by a guillotine-style door. The latency to crossover was recorded automatically. On day 1 (habituation), an individual mouse was placed in the light compartment for free exploration until it entered into the dark side and the door immediately closed. The mouse was promptly put back to the home cage after entering the dark side. On day 2 (training), the mouse was reintroduced to the light compartment, and a scrambled 0.7-mA, 2-s foot shock was delivered immediately after the mouse crossed to the dark compartment. The mouse was then put back into the home cage. On day 3 (test), at 24 h after the training, the mouse was reintroduced into the light compartment, and the latency to step through to the dark side was recorded as a measure of memory retention.

Statistical Analysis.

All statistical analyses were performed in GraphPad Prism 7. The Shapiro–Wilk test was first performed to determine whether the data were normally distributed, and then comparisons were made using parametric or nonparametric tests, as appropriate. For data that passed the normality test, statistical significance was determined by unpaired two-tailed Student’s t test as indicated in the figure legends. For data that did not pass the normality test, statistical significance was determined by the unpaired two-tailed Mann–Whitney test as indicated in the figure legends. All data are presented as mean ± SEM.

Data Availability.

All study data are included in the paper.