Background
Excitatory synapses in the mammalian brain are specialized by dense thickenings of proteins, referred to as postsynaptic densities (PSDs), in which glutamate receptors, cell adhesion proteins, scaffold proteins, and signaling molecules tightly associate via protein-protein interactions[
1,
2]. Two types of ionotropic glutamate receptors, AMPA-type (AMPAR) and NMDA-type (NMDAR), mediate excitatory synaptic transmission. PSD-95-like membrane-associated guanylate kinases (PSD-MAGUKs), including PSD-95, SAP102, PSD-93, and SAP97, are major scaffold proteins in the PSD[
3]. These proteins comprise five protein-protein interacting domains, three PDZ domains in the N-terminus, followed by an src homology-3 (SH3) domain and a guanylate kinase (GK) domain in the C-terminus, and are thus thought to regulate synaptic transmission. The first and second PDZ (PDZ1/2) domains of PSD-95 bind to the extreme C-terminus of NMDAR subunit 2 (GluN2) and also interact with the C-termini of auxiliary subunits of AMPAR, transmembrane AMPAR regulatory proteins (TARPs), by which PSD-MAGUKs regulate AMPAR clustering at synaptic sites[
4‐
6].
PSD-95 is a core component of the PSD. Based on quantitative mass spectroscopy, PSD-95 is ~6-fold more abundant than PSD-93, ~8-fold more than SAP102, and ~40-fold more than SAP97 in PSDs of the adult rat forebrain[
7]. Therefore, PSD-95 is a key molecule in mature synapses. The results of a number of studies by means of acute knockdown of endogenous PSD-95 or knockout (KO) mice of PSD-95 suggest that PSD-95 has a critical role in regulating the subunit composition of NMDARs and the level of AMPARs and its activity-dependent change at synaptic sites via interactions with PDZ domains[
8‐
14]. Further, the SH3 and GK domains in the C-terminus are essential for targeting PSD-95 to synaptic sites and the induction of NMDAR-dependent long-term depression (LTD)[
15]. Due to functional redundancy among PSD-MAGUKs and their multiple protein-interacting domain structure, however, the specific roles of individual PSD-MAGUKs and each PDZ domain during development in vivo have remained unclear.
Here, we focused on the PDZ1/2 domains and generated mutant cDNA knockin (KI) mice in which the PDZ1/2 domains of PSD-95 were unable to bind ligands, but retained their overall structure, to assess its function in vivo with minimal uncontrollable compensatory effects by other PSD-MAGUKs[
16,
17]. We studied the developmental accumulation of PSD proteins, hippocampal synaptic transmission, and behavior of the KI mice. The KI mice showed decreased levels of mutant PSD-95, PSD-93, and AMPAR subunits, but increased levels of SAP102 in the PSD fraction; greatly enhanced hippocampal long-term potentiation (LTP); and abnormal anxiety-like behavior and deficits in spatial and conditioned fear memory. These findings suggest that PSD-95 plays a central role among PSD-MAGUKs in modulating synaptic functions.
Discussion
The present study, using mutant PSD-95 cDNA KI mice, showed that PSD-95 is essential for the normal assembly of PSD-MAGUKs and the assembly of at least ~50% of basal AMPAR levels in the PSD, normal induction of synaptic plasticity in the hippocampus, and normal hippocampus-dependent behavior, such as learning and memory.
The genetically modified KI mice presented here expressed 1d2d-PSD-95-EGFP at ~7% of the level of WT PSD-95 in hippocampal homogenates. PSD-95 and other PSD-MAGUKs also interact with non-PDZ ligands, such as SAPAP/GKAP proteins via its GK domains, which in turn interact with another family of scaffolding proteins, Shank, and microtubule-associating motor protein dynein[
26‐
28]. Altered PSD-MAGUK protein levels might also change protein association contexts at synapses via their interaction with these molecules. Therefore, the observed phenotypes could be attributable to the substantial reduction in the levels of PSD-95, which interact with PDZ ligands as well as non-PDZ ligands, the expression of PDZ1/2-domain ligand binding-deficient PSD-95, and/or compensation by other PSD-MAGUKs. A number of the properties of the KI mice, however, were surprisingly different from those of PSD-95 KO mice, e.g., reduced levels of PSD-93 and AMPAR subunits in PSDs during development, normal magnitude of LTD induction, noticeably small body weight, marked hypokinetics in an unfamiliar environment, and markedly abnormal anxiety-related behavior. These different properties are considered to be the result of the expression of 1d2d-PSD-95-EGFP in vivo, although the possibility of different compensatory effects in KI mice from those in KO mice could not be excluded. Thus, the targeted introduction of mutant PSD-95 into mice allowed us to have a better understanding of the specific roles of the ligand-binding of the PDZ1/2 domains of PSD-95 in synaptic functions.
Loss of the ligand-binding ability of the PDZ1/2 domains of PSD-95 led to altered levels of PSD-MAGUKs in the PSD fraction of the hippocampus, substantially low levels of mutant PSD-95 (~8% of that of WT at P30), significantly reduced levels of PSD-93 (~70% that of WT at P30), and increased accumulation of SAP102 (~280% of that of WT at P30) during development (Figure
2). This increased expression of SAP102 was also observed in hippocampal homogenates (Figure
3). Similarly, the increased expression of SAP102 in hippocampal extracts is reported in PSD-95 KO mice and PSD-95/PSD-93 double KO mice, suggesting a certain compensatory function of SAP102[
9,
29]. On the other hand, decreased levels of PSD-93 in the PSD fraction are not observed in the PSD-95 KO mice. Because PSD-93 can form heteromultimers with PSD-95 in vivo and 1d2d-PSD-95-EGFP is unstable in spines, a fraction of PSD-93 directly interacting with 1d2d-PSD-95-EGFP might be degraded[
30].
With regard to the assembly of glutamate receptors in hippocampal PSDs, a marked reduction (to ~50%) of AMPAR subunits, GluA1, GluA2, and TARPs, was observed from P20–P65 in KI mice, whereas the levels of the NMDAR subunits GluN2A and GluN2B in the PSD fraction and GluA1 and GluA2 in hippocampal homogenates from KI mice were not significantly affected (Figures
2,
3). These results suggest that the ligand-binding ability of the PDZ1/2 domains of PSD-95 or the PSD-95 protein level itself regulates synaptic targeting of at least ~50% of basal AMPARs via TARPs, but not of NMDARs in hippocampal neurons in vivo during development. In PSD-95 KO and PSD-95/PSD-93 double KO mice, the replacement of GluN2B-containing NMDARs with GluN2A-containing NMDARs is failed during synapse maturation[
10,
12]. The biochemical investigation of the hippocampal PSD fraction performed here may not have been sensitive enough to detect the replacement of GluN2B- with GluN2A-NMDARs.
The deficiency in PSD-95 also caused a reduction in basal AMPAR-mediated synaptic transmission, as determined by the reduced input/output ratios (Figure
4A,
5C), and enhanced the magnitude of LTP following high frequency stimulation in hippocampal Schaffer collaterals/commissural projections-CA1 and medial perforant path-DG synapses of KI mice. The reduced basal synaptic efficacy is consistent with the reduced levels of GluA1 and GluA2/3 in the PSD. Since PPR was increased in KI mice, presynaptic changes may also be involved in the reduced synaptic efficacy. In the PSD-95 KO mice, the enhanced LTP observed in the CA1 region was suggested to originate from an increase in the population of silent synapses, which are preferential sites of AMPAR insertion during LTP[
10,
11,
31]. The enhanced LTP observed in the KI mice could be caused by the same mechanism. KI mice older than 10 weeks, however, did not show enhanced LTP in the CA1 region (Figure
5A), whereas basal transmission was still decreased at older age. Accumulation of the PSD proteins that compensate for the defects may explain this developmental change of LTP, and a presynaptic effect may be predominantly involved in the reduced basal transmission in older age. Further, LTP was enhanced at DG synapses in older mice (Figure
5B). These findings suggest that the ligand-binding ability of the PDZ1/2 domains of PSD-95 is differentially involved in mechanisms underlying the induction of NMDAR-dependent LTP in CA1 and DG.
The absence of LTD was previously observed in hippocampal slices from PSD-95 KO mice but not PSD-93 KO mice, and the importance of PSD-95 in the induction of LTD has been also demonstrated[
8,
11,
15,
32,
33]. Moreover, studies using short hairpin RNA-mediated knockdown of endogenous PSD-95 or PSD-93 expression suggest considerable heterogeneity of excitatory synapses in CA1 pyramidal neurons with respect to the levels of PSD-95 and PSD-93[
9,
34]. Taken together, these observations imply that the induction of LTD leads to decreases in the levels of APMARs only from synapses that mainly express PSD-95. In contrast to the increased levels of LTP, however, a normal level of LTD was still induced in hippocampal slices from KI mice following low frequency stimulation, despite the low expression level of 1d2d-PSD-95-EGFP. One possible explanation for the normal induction of LTD is that the impaired basal transmission is a result of the reduced AMPARs to nearly the same extent in both types of synapses that mainly express PSD-95 or PSD-93. Indeed, our KI mice had reduced amounts of not only PSD-95 but also PSD-93 (Figure
2B), and similar reduction of basal transmission with normal LTD is reported in TARP γ-8 KO mice as the consequence of general retraction of AMPAR from hippocampal synapses[
35]. Another possibility is that a compensatory mechanism that was different from that of PSD-95 KO mice enabled the induction of LTD in the synapses that mainly express PSD-93. Thus, the normal induction of LTD observed here could be due to either reduced basal AMPAR transmission or an equal level of LTD induction at the synapses that mainly express PSD-93 and those that mainly express 1d2d-PSD-95-EGFP.
Despite the clustering of 1d2d-PSD-95-EGFP at postsynaptic sites, some presynaptic alterations were also suggested in the KI mice (Figsures
4B,
5D). Consistently, a similar observation was reported using PSD-95 KO or knockdown experiments[
8,
36]. Further, the relationships between PSD-MAGUKs and the presynaptic releaseprobability were suggested at other excitatory synapses[
37]. The increased PPR could be at least partly due to the reduced levels of 1d2d-PSD-95-EGFP and the subsequent decreased levels of 1d2d-PSD-95-EGFP-neuroligin complexes. Considering the intact PDZ3 domain of our construct and the effectiveness of 1d2d-PSD-95-EGFP protein judged by normal LTD induction, however, the significance of PDZ1/2 domains that do not bind to neuroligin should be taken into account.
We performed comprehensive behavioral analyses of KI mice. These mice exhibited abnormal anxiety-related behavior, significant hypoactivity in novel environments, and decreased social interactions in their home cages. These behavioral abnormalities may be related to neurodevelopmental disorders, such as autism, via synaptic dysfunction caused by the loss of PSD-95 itself and/or the ligand-binding ability of the PDZ1/2 domains of PSD-95[
38]. Autism is characterized by abnormal social interactions, deficits in communication, and high levels of repetitive behaviors. In the three-chambered social interaction tests, KI mice showed a significant lack of interest in others based on their stay time around the cages, as well as decreased locomotor activity (Additional file
1: Figure S5). In the Y-maze test, KI mice showed remarkably decreased exploration and impaired working memory, as well as repetitive exploration of the same arm with few alternations. These behaviors might comprise autism-like features. On the other hand, markedly decreased stereotypic counts in the open field tests suggest that increased self-grooming by the KI mice is unlikely in the novel environment (Figure
6H). In PSD-95 KO mice, significantly increased self-grooming behaviors were observed only in the home cage[
38]. Further, synaptic proteins, such as Shank and neuroligin, which interact with PSD-95, are also implicated in autism[
39]. The relation of PSD-95 with autism should be studied further.
In the fear conditioning test, while KI mice showed higher freezing during fear conditioning and in the first 1 to 2 min in the cued testing with an altered context at 24 h after conditioning (Figure
7H), they showed a significantly decreased freezing response after 7 days in the altered context (Figure
7J). This trend in the freezing level was also observed with context testing (Figure
7G,I) Therefore, although we could not exclude the possibility that hypoactivity and a high anxiety level of KI mice affected the freezing measurement, high freezing behaviors observed in the KI mice were associated with aversive experiences, suggesting that KI mice are impaired in fear memory retention or enhanced fear extinction.
The observed severe deficits in spatial, working, and fear memory are likely to be associated with abnormal synaptic transmission in the hippocampus, particularly in DG synapses, because the behavioral tests were conducted in adult KI mice older than 3 months of age. Further studies to investigate the roles of PSD-95 at the medial perforant path-DG synapses will provide a better understanding of the contribution of DG synapses to learning and memory formation.
Our study shows that PSD-95 and/or the ligand-binding ability of the PDZ1/2 domains of PSD-95 are crucial for normal synaptic clustering of the PSD-MAGUKs and AMPAR subunits; normal synaptic transmission in the hippocampus; and normal learning and memory formation, including acquisition, consolidation and retention. The results encourage future studies, such as a detailed comparative analysis of PSD-95 KO mice and 1d2d-PSD-95-EGFP KI mice and overexpression studies of 1d2d-PSD-95-EGFP in vivo that will clarify the specific roles of the ligand-binding ability of the PDZ1/2 domains of PSD-95.
Competing interest
The authors declare that they have no competing interests.
Authors’ contributions
HN, YI, KK, KT, HT, SS, HS, TM and TD designed research; HN, YI, KK, TT, KN, HT, and TD performed research; KT and TM analyzed data; HN, YI, KK, KT, HT, SS, TM, YF and TD wrote the paper. All authors read and approved the final manuscript.