Molecular Architecture of Postsynaptic Interactomes

Abstract

The postsynaptic density (PSD) plays an essential role in the organization of the synaptic signaling machinery. It contains a set of core scaffolding proteins that provide the backbone to PSD protein-protein interaction networks (PINs). These core scaffolding proteins can be seen as three principal layers classified by protein family, with DLG proteins being at the top, SHANKs along the bottom, and DLGAPs connecting the two layers. Early studies utilizing yeast two hybrid enabled the identification of direct protein-protein interactions (PPIs) within the multiple layers of scaffolding proteins. More recently, mass-spectrometry has allowed the characterization of whole interactomes within the PSD. This expansion of knowledge has further solidified the centrality of core scaffolding family members within synaptic PINs and provided context for their role in neuronal development and synaptic function. Here, we discuss the scaffolding machinery of the PSD, their essential functions in the organization of synaptic PINs, along with their relationship to neuronal processes found to be impaired in complex brain disorders.

1. Introduction

The postsynaptic density (PSD) is a protein-rich specialization located at the postsynaptic membrane of excitatory synapses. Originally identified as a morphologically distinct region through electron microscopy [1, 2], a number of studies utilizing mass spectrometry have went on to characterize the protein composition of the PSD, identifying more than 2000 proteins [3–7]. This was soon followed by more detailed analyses aimed at identifying protein-protein interactions (PPIs) between individual components of the PSD [8–13]. Although challenging, these efforts have identified several PSD PPIs involving specialized signaling machinery capable of assembling large macromolecular protein complexes. These complexes work in concert with one another to interpret biochemical and electrical information received by neurotransmitter receptors along the postsynaptic membrane and induce subsequent molecular changes within the receiving neuron.

A major structural component of PSD protein complexes are scaffolding proteins that are both abundant and contain multiple protein domains specialized in PPIs (Figure 1A). This multi-domain architecture provides a platform for simultaneous interactions with multiple binding partners, can facilitate multimerization, and is a major contributing factor to the formation of macromolecular signaling complexes [14]. PPIs involving scaffolding proteins enable the presentation of their binding partners at the correct place and time. The most abundant and most highly researched scaffolding protein families of the PSD can be categorized into three main layers: a top layer represented by the discs large (DLG1–5) family of proteins which interacts with glutamate and other neurotransmitter receptors at the PSD membrane; a bottom layer composed of the SH3 and multiple Ankyrin repeat domains (SHANK1–3) protein family, which connects the scaffolding machinery to the cytoskeleton; and a middle layer composed of the DLG-associated proteins (DLGAP1–4) which connects the top and bottom layers of synaptic scaffolds (Figure 1B) [11, 13, 15]. These layers provide the core backbone of PSD protein interaction networks (PINs) through their ability to bind receptors, enzymes of diverse function, cytoskeletal proteins, and other scaffolding protein families. Each layer contains multiple corresponding family members that are essential to the organization of the PSD and can be dynamically regulated throughout development and in response to synaptic stimuli [11, 13].

Figure 1 –

Protein domain architecture and organization of PSD scaffolding proteins. A) Figure shows protein domain composition of select PSD scaffolding proteins described in this review. Dlg4, Dlgap1, and Shank3 are representative family members of the top, middle, and bottom layers of scaffolding protein families in the PSD, respectively. B) Illustration shows the interactions between the scaffolding proteins of the PSD. DLGs, DLGAPs and SHANKs, form the three main layers while the additional scaffolds are concentrated within the middle layer.

Scaffolding proteins are the fundamental organizers of the PSD signaling machinery and PINs. Due to this, they regulate numerous cellular processes essential for synaptic function. Damaging mutations impacting the function or expression of scaffolding proteins can have detrimental effects to the flow of information within the PSD. Consequently, PSD scaffolding proteins have been associated with multiple complex brain disorders such as autism spectrum disorder (ASD), schizophrenia (SCZ), intellectual disability (ID), and developmental delay (DD) [16–24]. Because of their centrality to PSD PINs, it has been a common practice for human genetics studies to analyze associations between disease and components of the PSD or particular protein interaction datasets. This underscores the importance of characterizing PPIs involving scaffolding proteins to understand their roles in PSD signaling mechanisms. However, there are several important factors to consider when interpreting the function of PSD proteins through their respective PPIs and their role in complex brain disorders. Here, we will focus on the organization of the core signaling machinery of the PSD and not on the description of functions assigned to individual proteins. We will first describe essential concepts and technical problems when determining PPIs of the PSD scaffolds. This includes the critical importance of spatio-temporal context as PSD proteins may be differentially expressed throughout development and localized within different neuronal subcellular compartments. We will then briefly discuss and present examples of the main components of the PSD scaffolding machinery, focusing on their protein domain architecture, PPIs, and how this relates to their function and disease.

2. Determining PPIs Within the PSD

Protein complexes can be seen as functional units in which interacting proteins work together to carry out common cellular processes [25, 26]. Therefore, determining the composition of protein complexes at the PSD can provide insights into the multiple functions of a target protein. This is especially important for scaffolding proteins as they participate in the simultaneous regulation of multiple signaling pathways and cellular processes involved in neuronal development and synaptic transmission. However, PPIs are dynamic and biological variables such developmental time period, cellular location, or technical variables such as the experimental technique used for PPI characterization can all have a significant impact on the protein interactions identified.

A number of methods have been used for the identification of PPIs, each with their own set of advantages and disadvantages. The identification of direct, binary interactions between two proteins has been historically addressed through yeast two-hybrid systems [27]. Although they provide evidence of direct protein interactions, both the cellular environment and stoichiometric expression levels between the interacting proteins do not represent endogenous conditions. Thus, interactions identified through this method require further in-vivo validation. Apart from yeast two-hybrid, Immunoprecipitation (IP) of a protein of interest can be followed by immunoblotting (IP-IB) or mass-spectrometry (IP-MS) for either low- or high-throughput PPI identification, respectively. PPIs identified using IP-MS may include secondary and tertiary interactions in addition to direct interactions. These datasets are usually described as protein complexes, but they can be better described as a collection of direct and indirect protein interactions comprising protein interactomes.

Recent large-scale studies have characterized tens of thousands of PPIs through either yeast two-hybrid [28] or IP-MS following protein expression in cultured mammalian cell lines (e.g. HEK 293T or HCT116 cells) [29, 30]. While these studies have greatly expanded the knowledge of protein interactomes, the identification of PPIs involving synaptic proteins has proven to be more difficult. These studies are helpful in identifying interactions between two expressed components within the particular assay conditions as opposed to whether they are interactors within endogenous conditions. Therefore, they may contain large numbers of false positive and negative PPIs. Furthermore, comparisons of PPIs involving synaptic proteins determined under endogenous conditions versus PPIs for the same protein in systems such as HEK 293T cells have shown minimal PPI overlap between the two systems [8–11].

In addition to the experimental method used for PPI identification, characterization of PPIs within a distinct spatio-temporal context is essential. This is because proteins may have unique interacting partners dependent on both the developmental time period and the spatial location in which the protein resides. Unfortunately, variations of spatio-temporal context are still lacking for PSD PPIs as a large majority of studies focus solely on interactions within PSD fractions of adult rodent models. This has been the standard approach because the role of PSD proteins has been highly studied in the context of learning, memory and synaptic plasticity within mature synapses. However, the protein complex composition of the core-scaffolding proteins have been shown to change throughout neuronal development, with differential binding partners at embryonic stages compared to adulthood [11]. In addition, the spatial location of scaffolding proteins influences their function on many different levels, including anatomical brain region, cell type, subcellular location, and tissue expression. In line with this, the use of biochemical fractionation has been able to show differential PPIs for a number of proteins within the PSD and alternative neuronal locations such as the dendrite or soma [11, 12].

Several efforts have been made to characterize the protein interactions that provide the architectural framework for PSD function. Early studies using methods such as yeast two-hybrid assays were able to delineate individual PPIs involving the core PSD scaffolding proteins and were the first indication of these proteins forming “layers” within the PSD [14]. More recently, characterization of endogenous protein complexes via mass-spectrometry has both verified early yeast two-hybrid studies and greatly expanded the number of known intracellular binding partners of PSD scaffolding proteins by hundreds of interactors [8–13]. These large interactomes likely reflect the involvement of PSD scaffolding proteins in multiple protein complexes. At the same time, they provide context for the numerous cellular processes regulated by core PSD scaffolds and the phenotypes that arise in mutant mouse models.

3. Core Scaffolding Components of the PSD

3.1. The DLG family (Dlg4)

The DLG family forms the top layer of scaffolding proteins through their ability to stabilize glutamate receptors at the postsynaptic membrane and simultaneously bind several signaling molecules. DLGs belong to the MAGUK superfamily of proteins and have four family members commonly found at the PSD, including Dlg1 (SAP97), Dlg2 (PSD93/Chapsyn-110), Dlg3 (SAP102), and Dlg4 (PSD95/SAP90) [31]. A fifth family member, Dlg5, has also been observed to function in early neural development and synaptogenesis, but its function at the PSD has not been extensively studied [32, 33]. A conserved set of protein domains specialized in PPIs defines the protein domain architecture of the DLG family. These include three PDZ domains, an SH3 domain, and a GK domain towards the C-terminus (Figure 1A) [31, 34]. Through the use of yeast two-hybrid screening, initial PPI studies were able to show that each protein domain is involved in specific PPIs. For example, while the first two PDZ domains of Dlg4 have been shown to interact with C-terminal PDZ ligands of Shaker-type potassium channels and the NMDA receptor subunits, GluR2A and GluR2B [35–37], the third PDZ domain has been shown to specifically interact with neuroligins [38]. In contrast, Syngap1 and TARPs may interact with all three PDZ domains of Dlg4 [39–41]. The GK domain of the DLG family does not have the ability to bind or phosphorylate GMP [42] but instead acts as a protein interaction module with proteins such as Dlgap1 and Map1a [39, 43–45]. The SH3 domain has been shown to form both intra- and intermolecular interactions with the GK domain to regulate protein-protein interactions and MAGUK oligomerization [46, 47]. Furthermore, the SH3-GK intramolecular interaction has been proposed to undergo conformational changes upon ligand binding to the third PDZ domain which allows for alternate PPIs of the GK domain [48, 49].

Dlg4 is the most highly studied DLG family member and has been shown to have hundreds of binding partners through IP-MS studies utilizing either anti-Dlg4 antibodies [11, 13, 50] or endogenous Dlg4 TAP-TAG affinity purification [8]. TAP-TAG affinity purification is performed in two individual steps which allowed for the identification of proteins with less affinity in the initial purification step and those with increased affinity towards Dlg4 in the final step, termed the core Dlg4 complex [8]. Studies that have characterized Dlg4 interactors through IP-MS have identified a largely overlapping set of Dlg4 interacting proteins including NMDA and AMPA glutamate receptors, potassium channels, multiple DLG and DLGAP family members, cytoskeletal proteins, cell adhesion proteins, and signaling molecules. This illustrates the diverse set of cellular processes scaffolding proteins facilitate at the PSD including cytoskeletal dynamics and regulation of synaptic activity. As protein domains can also indicate protein function, protein domain profiling of Dlg4 complexes has shown an abundance of domains involved in PPIs, along with a specific enrichment in GKAP and PBPe domains (Figure 2) [11]. With GKAP and PBPe domains being specific to DLGAP family members and ionotropic glutamate receptors, respectively, this reflects the role of Dlg4 in facilitating the connection between glutamate receptors and the internal signaling machinery of the PSD.

Figure 2 –

Interactomes of the PSD Core scaffolding proteins and protein domain enrichment. (Left) Figure shows protein-protein interactions (PPIs) involving representative proteins of the three main layers of the postsynaptic density in adult mouse cortex. (Right) Protein domains found to be enriched in the interactomes of each corresponding core-scaffolding protein. All data is derived from [11].

In line with DLG proteins interacting with glutamate receptors, DLGs have been shown to be important for both AMPA receptor targeting to the synapse and NMDA receptor stabilization [34, 51, 52]. In addition, Dlg2 and Dlg4 have recently been shown to act cooperatively with NMDA receptors to form large supercomplexes involved in synaptic transmission. Interestingly, this large complex depends on the presence of Dlg2, Dlg4 and GluN2B as shown by its absence in mouse models lacking either Dlg2, Dlg4, or the C-terminal domain of Glun2B [10]. While Dlg2 and Dlg4 may act cooperatively in forming associated protein complexes, mutations in these proteins have been shown to have opposite effects on synaptic transmission. Here, Dlg4 knockout mice have been found to display increased LTP and impaired LTD while Dlg2 knockout mice exhibit impaired LTP and normal LTD [53]. Like Dlg4, mice lacking Dlg3 have been shown to have increased LTP, while conditional knockout of Dlg1 shows no effect on LTP [54, 55]. In addition to changes in electrophysiology, a lack of Dlg3 has been shown to increase the association of Dlg4 with NMDA receptor complexes indicating altered protein complex composition [54]. Through development, western blot analysis has shown Dlg3 to be more abundant than Dlg4 during early postnatal stages and then the opposite expression pattern during adulthood [56]. While mRNA expression profiling and mass spectrometry have been able to confirm their differential expression patterns throughout development, these studies have also shown both Dlg3 and Dlg4 to be similarly expressed and present in equal amounts in protein complexes during embryonic developmental stages [11, 57]. This, along with compensatory expression of alternate Dlg family members observed in Dlg4 mutant mice, illustrates the highly interconnected nature of the DLG family of proteins [54, 58, 59].

3.2. The DLGAP Family (Dlgap1)

The DLGAP family of proteins form the middle-layer of the PSD, connecting the DLG family of scaffolding proteins to the SHANK family. DLGAPs are the least studied family of PSD core scaffolds. Yeast two-hybrid studies originally identified DLGAPs as directly binding to the GK domain of the DLG family through multiple 14 amino acid repeats at the N-terminal region of DLGAP proteins [39, 43]. However, after these initial reports, the nature of DLG/DLGAP interactions have not been further investigated. DLGAPs were also characterized by their ability to bind to the PDZ domain of SHANKs through the PDZ ligand motif of Dlgap1 [60], effectively connecting the top and bottom layers of the PSD. DLGAPs have been shown to interact directly with the motor protein, dynein light chain (DLC), in association with Dlg4 and Myosin-V. This has been proposed to serve as a complex to regulate transport of DLGAP and DLG proteins [61]. Compared to the DLG family, DLGAPs are not rich in conserved protein domains. Instead, they contain a single conserved sequence of 160 amino acids in the C-terminal region termed the GKAP homology domain of unknown function 1 [62] (GH1 or GKAP domain) along with protein interaction motifs responsible for binding DLGs and SHANKs (Figure 1A).

Among the DLGAP family, Dlgap1 is the only member whose endogenous protein interactions have been identified through IP-MS [11, 13]. Proteins involved in Dlgap1-associated complexes are largely overlapping with those of Dlg4 including glutamate receptors, calcium channels, and signaling molecules. However, they do have distinct differences indicating the clustering of specific proteins via DLGAPs (Figure 2). Profiling of protein domain architecture within Dlgap1 complexes shows an enrichment of proteins containing domains involved in PPIs such as PDZ and SH3 domains (Figure 2) [11]. This highlights the centrality of Dlgap1 to PSD PINs as it interacts with scaffolding and adaptor proteins in both the upper and lower layers of the PSD. In line with yeast two-hybrid studies delineating the role of DLGAPs in connecting the two other major scaffolding layers of the PSD, functional studies have shown a disconnect between these layers in Dlgap1 knockout mice [15]. Here, mass spectrometry of immunoisolated DLG family members using a pan-MAGUK antibody was able to show decreased association with SHANK family members. Furthermore, immunoprecipitation of Shank3 followed by immunoblotting against Dlg4 was able to show a 60% decrease in association. The remaining fraction of Shank3 associated with Dlg4 is presumed to be connected through the other DLGAP family members [15]. This highlights the essential role of DLGAPs in maintaining the structure of the PSD.

The DLGAP family members have been shown to have both overlapping and distinct spatio-temporal expression throughout development. For example, while there are minor differences between the expression of the family members in the hippocampus and cortex, Dlgap3 has increased expression in the striatum relative to the other family members [63, 64]. With respect to development, Dlgap2 has the lowest expression in early post-natal development followed by relatively similar expression levels between the family members into adulthood [65]. In addition, different laboratories have reported a variety of phenotypes in individual mouse models lacking Dlgap 1–4. These phenotypes include behavioral abnormalities, deficits in synaptic transmission, and dysregulation of synaptic proteins [15, 64, 66, 67]. For example, a lack of Dlgap2 has been reported to decrease overall PSD thickness and length along with a decrease in the quantity of GluR1 and Camk2a PSD proteins. In contrast, a lack of Dlgap4 resulted in an overall increase in PSD size and no observed effects in the total protein levels of GluR1 and Camk2a [66, 67]. Although only investigated in the Dlgap4 mutant mouse model, elimination of Dlgap4 did not result in compensatory changes in the protein levels of other DLGAP family members [66]. The observed phenotypes reinforce the relevance of the DLGAP family of scaffolding proteins on the PSD structure and composition.

3.3. The SHANK Family (Shank3)

The SHANK family of proteins (Shank1–3) form the bottom layer of the PSD core scaffolding machinery through the direct interaction with the DLGAP family [60]. Like the DLG family, SHANKs contain a variety of protein domains and motifs specialized in PPIs. From the N-to C-terminal, these include multiple ankyrin repeat domains, an SH3 domain, A PDZ domain, a proline rich region, and a SAM domain (Figure 1A). Each member of the SHANK family has isoforms that lack particular domains and therefore influence their function. For example, the Shank3b isoform lacks the SAM domain and proline rich region which results in differential subcellular localization [68], whereas particular isoforms of Shank1 and Shank2 lack the N-terminal ankyrin repeats and SH3 domain [69, 70]. Profiling SHANK PPIs via yeast two-hybrid has been able to highlight their role in regulating cytoskeletal dynamics and glutamate receptor activity. Here, the ankyrin repeats were shown to directly interact with Sharpin and Sptan1 [71, 72] while the proline rich region has been shown to interact with the Homer family of adaptor proteins and cortactin [60, 73]. Sptan1 and Cortactin are both regulators of the actin cytoskeleton, while Homer proteins couple SHANKs to metabotropic glutamate receptors (mGluRs) [60, 72, 74]. The association of SHANKs to mGluRs, combined with their coupling to NMDA receptors through the DLGAP and DLG families, enable crosstalk between different classes of glutamate receptors [73], although it is not known if these protein interactions are part of one or multiple protein complexes. In addition, the SAM domain of SHANKs has been shown to facilitate SHANK multimerization [60] and regulate the localization of Shank2 and Shank3 [75]. These processes have been found to be regulated by Zn²⁺ binding to Shank2 and Shank3, while Shank1 does not appear to bind Zn²⁺ [76, 77]. In line with this, the interactome of Shank3 has been shown to be modulated by Zn²⁺ concentration [78].

The three SHANK family members have both overlapping and unique patterns of expression dependent on spatial location and developmental time period. It has been described that Shank1 and Shank2 have increased expression during early postnatal development [79]. However, all three family members have also been identified in embryonic day 14 mouse cortex through mass spectrometry. While Shank1 and Shank2 interactomes have not been characterized at this developmental stage, it has been shown that Shank3 associates to a variety of cytoskeletal proteins, synaptic adaptors, and scaffolds such as Homer 1–3, Baiap2 and Lrrc7 in embryonic day 14 mouse cortex [11]. In common with interactions found at the adult PSD, the DLGAP family members have been reported to associate with DLGs at early embryonic developmental stages. In contrast, the interaction between SHANK and DLGAP family members has not been observed during this same time period, illustrating how scaffolding proteins can have differential binding partners dependent on specific developmental stages and how the layers of PPIs within the PSD build on one another throughout time [11]. With respect to regional expression, SHANK proteins do not appear to have major differential expression between the family members in the cortex and hippocampus. However, similar to Dlgap3, Shank3 has increased expression in the striatum relative to Shank1 and Shank2 [64, 79, 80].

All three SHANK family members have been implicating in contributing to complex brain disorders along with regulating processes such as spine morphology and synaptic transmission [16, 81, 82]. Several mouse models have been generated that knockout particular isoforms or the proteins completely. As with the DLG family of proteins, deletion of SHANKs can result in compensatory mechanisms. For example, Shank3 protein expression is increased in Shank2 knockout mice, and in mice lacking Shank3, there is a significant upregulation of Shank2 and trend of increased Shank1 protein [83, 84]. However, as with other components of the PSD, how this compensatory differential protein expression of PSD scaffolds might affect the stoichiometry of postsynaptic protein complexes is unknown. Changes in total protein levels of NMDA receptors (Shank1, 2, and 3), AMPA receptors (Shank2 and 3), DLGAPs (Shank1 and 3), and Homer proteins (Shank1 and 3) are common within the several different SHANK mouse models, showing alterations in the protein binding partners of the SHANK family [82, 85].

Shank3 is the most widely studied member of the SHANK family due to its strong association to complex brain disorders and Phelan-McDermid syndrome. Multiple studies have identified endogenous Shank3 interactomes [11, 13, 78, 86, 87]. These studies have confirmed the presence of multiple members of the DLGAP and HOMER families associated with SHANKs at the PSD. Furthermore, the protein domain architecture shows an increased presence of proteins with GKAP, PDZ, PH, and S_TKc domains at the PSD (Figure 2). While GKAP domains are indicative of multiple DLGAP family members, the enrichment in the S_TKc domain shows that Shank3 is associated with an abundance of kinases likely involved in the propagation of cellular information relayed from the upper layers of the postsynaptic signaling machinery.

Haploinsufficiency of SHANK3 has been shown to be causative in Phelan-McDermid syndrome, which is characterized by global developmental delay, autistic behavior, hypotonia, and either absent or delayed speech [88–90]. In addition, mutations in SHANK3 have been shown to be associated to ASD, ID, and SCZ [16, 91–94]. Numerous Shank3 mouse models have been generated to examine the effects of disruptions within specific isoforms, loss of the whole protein, and specific mutations found in patients affected by complex developmental brain disorders [82]. The comparison of mutant mouse models containing a SCZ-associated mutation and an ASD-associated mutation has been able to show how different mutations can result in different effects on the translated protein product of Shank3 and phenotypic characteristics related to complex brain diseases. It has been reported that the mouse model containing the ASD-associated mutation resulted in a near complete loss of Shank3 accompanied by defects in synaptic transmission within the striatum and impairments in social behavior during early developmental stages. In contrast, the SCZ-associated mutation resulted in expression of a truncated Shank3 protein, defects in synaptic transmission within the prefrontal cortex, and social dominance behavior during adulthood [84]. Thus, Shank3 mutants exhibited numerous phenotypes related to those observed in patients with SHANK3 mutations and different complex brain disorders. Recently, CRISPR/Cas9 genome engineering has been used to generate cynomolgus macaques with indels that disrupt expression of SHANK3 [95]. This allowed the study of SHANK3-related phenotypes in brain structure, cognitive function, and social behavior in a model organism closer to humans. As seen in Shank3 mutant mice, impaired social interactions and other behavioral abnormalities were also observed. In addition, functional magnetic resonance imaging was able to show large scale changes in brain connectivity with global hypoconnectivity and local hyperconnectivity in areas such as the somatosensory cortex [95]. This represents the first successful study of exploring the functional consequences of mutations in a core scaffolding protein of the PSD in a cohort of non-human primates.

4. Expanded Families of Scaffolding Proteins

4.1. Baiap2 (IRSp53)

Baiap2 (also known as IRSp53) is an abundant PSD scaffolding protein implicated in regulating actin dynamics through its interactions with F-actin bundles and small GTPases [96]. From the N- to the C-terminus, all isoforms of Baiap2 contain an IMD (IRSp53 and MIM homology) domain, a CRIB-PR (Cdc42/Rac interactive binding-proline rich) region, and a SH3 domain (Figure 1A). At the C-terminus, alternate isoforms of Baiap2 may either contain a WH2 (WASP homology 2) domain or a PDZ ligand [96]. Similar to the DLGAP family of proteins, Baiap2 interacts with both DLGs and SHANKs, albeit through different domains. Here, Baiap2 has been shown to directly interact with the first two PDZ domains of the DLG family through the its C-terminal PDZ ligand and interact with a proline-rich region in SHANKs through its N-terminal SH3 domain [97–99]. Akin to the intramolecular interaction of the GK and SH3 domains of the DLG family, the CRIB-PR region has been shown to interact with the SH3 domain of Baiap2, effectively reducing its capacity to interact with other proteins. Binding of the small GTPase Cdc42 has been shown to reverse this inhibition by inducing a conformation change [100]. Baiap2 has been shown to interact with a host of cytoskeletal proteins through its SH3 domain in order to regulate actin dynamics including dynamin and Wasf2, a member of the WAVE complex [101, 102]. In addition, the IMD domain binds F-actin to regulate cytoskeleton dynamics and is involved in Baiap2 dimerization [103, 104]. Although Baiap2 has not been directly subjected to high-throughput protein interaction identification at the PSD, it has been identified in Dlg4, Dlgap1, and Shank3 interactomes at different developmental stages within mouse cortex [11]. In addition, co-fractionation of protein complexes in whole brain has identified Baiap2 in complex with Cdc42 [105]. These studies confirm previously identified interactions and provide further support for the central role of Baiap2 as a PSD scaffold.

Baiap2^−/− mice display a variety of abnormalities consistent with the protein-protein interactions of Baiap2. Apical Baiap2^−/− pyramidal neurons in the prelimbic area of the medial prefrontal cortex have reduced complexity and decreased number of dendritic spines [106]. However, hippocampal neurons in Baiap2^−/− mice showed no morphological abnormalities, illustrating a potential region-specific role of Baiap2 [107, 108]. In addition, Baiap2^−/− mice have enhanced LTP and decreased NMDA-dependent LTD. The decrease in LTD was shown to be due to a stabilization of F-actin and a subsequent decrease in NMDA receptor trafficking due to a lack of Baiap2 [106, 107]. Abnormalities in LTP and LTD, coupled with behavioral phenotypes illustrating deficits in learning and memory, shows the importance of Baiap2 in organizing PSD PINs.

4.2. Cnksr2 (MAGUIN)

Cnksr2 (also known as MAGUIN) is a multi-domain scaffolding protein located at the PSD directly implicated in psychiatric disorders. From the N- to C- terminus, Cnksr2 contains a SAM domain, a conserved region in CNK (CRIC) domain, a PDZ domain, a PH domain surrounded by poly-proline regions on either side, and a PDZ ligand at the N-terminus (Figure 1A). Through the PDZ ligand, Cnksr2 has been shown to directly interact with the DLG family along with other scaffolding molecules including Magi2 (S-SCAM) and Lrrc7 (Densin-180) [109, 110]. The PH domain has been attributed to synaptic targeting of Cnksr2 [111]. Unlike the SHANKs, the SAM domain has been proposed to not contribute to multimerization of Cnksr2, with a leucine rich region at the C-terminus instead being responsible for this [109]. Through immunoprecipitation, Cnksr2 has been shown to interact with Raf1 implicating it in Ras/MAPK signaling [112] and multiple proteins regulating the Rac/Cdc42 GTP signaling pathway, including Arhgap39 [11, 113]. The interaction between Arhgap39 and Cnksr2 has been shown to mediate spine maturation. Overexpression of Cnksr2 lacking a binding site for Arhgap39 failed to rescue an increased protrusion number but decreased mature synapse number observed in hippocampal neurons post knockdown of Cnksr2. In contrast, overexpression of wildtype Cnksr2 was able to rescue this phenotype, illustrating the importance of Cnksr2 in targeting Arhgap39 activity to the PSD [113]. High throughput characterization of endogenous Cnksr2 PPIs has been carried out in two separate studies [11, 113]. These studies were able to confirm scaffolding interactions identified through yeast two-hybrid such as Dlg4 and Lrrc7, but also expand this to including additional scaffolding proteins such as Baiap2, Dlgap1, and Shank3 [11, 113], showing the integral role Cnksr2 may have in PSD PINs involving the scaffolding proteins of PSD.

In line with Cnksr2 being central to PSD PINs, Cnksr2 has also been implicated in contributing to multiple complex brain disorders. Cnksr2 is found within a single-gene, genome-wide significant locus, contributing to schizophrenia risk [20]. In addition, mutations resulting in a loss of expression in Cnksr2 have been shown to be causative for X-linked intellectual disability. Over 9 families have been identified with individuals affected by a deficiency in Cnksr2 [114–117]. Furthermore, large scale exome sequencing efforts to characterize genetic causes of developmental disorders have identified several additional damaging mutations in Cnksr2 [22]. With common variation contributing to schizophrenia and loss of function contributing to intellectual disability, this illustrates how different types of mutations in a single gene can result in differing outcomes related to complex brain disorders.

4.3. Sorbs2 (nArgBP2)

Sorbs2 is an abundant scaffolding protein at the PSD that has multiple isoforms, including nARGBP2, an isoform specifically expressed in neural tissues. Sorbs2 contains several conserved protein domains, including a sorbin homology (SoHo) domain at the C-terminus, a C2H2-type zinc finger, and three SH3 domains at the N-terminus (Figure 1A) [118]. Sorbs2 directly interacts with the DLGAP family of proteins through its third SH3 domain, exemplifying its association with the core scaffolding components of the PSD [119]. Although Sorbs2 complexes have not been subject to high-throughput identification, characterization of Dlgap1 complexes at the PSD have confirmed this association [11]. By using specific domains of Sorbs2 for immunoprecipitation, the SoHo domain was shown to interact with alpha2-spectrin and the SH3 domains were shown to interact with multiple proteins involved in regulation of cytoskeleton dynamics such as dynamin, synaptojanin, and the WAVE complex [120]. These interactions are identical to what is observed with the SH3 domain of Baiap2, indicating the possibility of a functional complex. The role for Sorbs2 in the regulation of cytoskeletal dynamics has been confirmed in excitatory neurons as the Sorbs2 knockout mouse model has decreased dendritic complexity and knockdown of Sorbs2 results in abnormal spine morphology [121, 122]. In addition, Sorbs2 knockout mice display defects in excitatory synapse transmission and behavioral tasks associated with learning and memory [122].

4.4. Lrrc7 (Densin-180)

Lrrc7 (also known as Densin-180) is central to the network of PSD scaffolding proteins as it has been shown to directly interact with both Cnksr2 and SHANKs [109, 123]. Lrrc7 is concentrated at the PSD and contains 16 leucine rich repeats (LRRs) and a PDZ domain (Figure 1A) [124]. Proteins containing this domain architecture have been collectively termed LAP (leucine rich repeat and PDZ domain) proteins [125]. While the LRR domain has been implicated in targeting Lrrc7 to the plasma membrane [126], the PDZ domain has been shown to directly interact with delta-catenin and Cnksr2 [109, 127]. Lrrc7 has also been shown to directly interact with alpha-actinin, a cytoskeletal protein that binds to F-actin to anchor structures within the PSD [128]. With both Cnksr2 and alpha-actinin binding to Dlg4, these interactions provide a link between Lrrc7 and Dlg4 [109, 129]. In addition, Lrrc7 interacts with key proteins involved in calcium signaling and synaptic plasticity including Camk2a and the L-type calcium channels, Cav1.2 and Cav1.3 [128, 130, 131]. Here, Lrrc7 has been shown to recruit Camk2a to calcium channels in order to facilitate the modulation of neuronal excitability [130, 131].

In line with Lrrc7 interacting with both cytoskeletal proteins and facilitating the regulation of synaptic plasticity, Lrrc7 mutant mice display reduced levels of both alpha-actinin and mGluR5 in PSD fractions [132]. As alpha-actinin directly interacts with mGluR5, this has been postulated to be contributing factor to the reduction in mGluR5 [133]. Mutant mice also displayed deficiencies in both NMDA and mGluR-dependent forms of LTD along with abnormalities in spine morphology. While a lack of Lrrc7 results in decreased dendritic complexity, overexpression of Lrrc7 in cultured hippocampal neurons has been shown to produce the opposite effect [123, 134]. In addition, mutant mouse models have been shown to exhibit abnormal behavioral traits such as impaired short-term memory tasks, aggressive behavior during juvenile developmental stages, and social dysfunction in adulthood [132, 134]. Positive allosteric modulation of mGluR5 has recently been shown to rescue both behavioral abnormalities and neuronal morphology, highlighting the critical link between metabotropic glutamate receptors and Lrrc7 [134].

5. PSD scaffolds in Complex Brain Disorders

Complex brain disorders such as ASD, SCZ, ID, and DD are polygenic neurodevelopmental disorders that are influenced through a variety of genetic and environmental risk factors [135]. Substantial progress has been made in identifying the genetic underpinnings of complex brain disorders with respect to single nucleotide polymorphisms (SNPs) identified through Genome Wide Association Studies (GWAS), inherited or de novo single nucleotide variants (SNVs), and copy number variations (CNVs) [16–24]. Together, these studies have identified thousands of genetic risk factors associated to complex brain disorders. Although these risk factors contribute through several different forms of genetic abnormalities and with varying levels of penetrance, a number of them have been found to converge on postsynaptic proteins involved in neuronal development and synaptic function.

A common theme amongst studies identifying genetic risk factors for complex brain disorders is statistically testing the enrichment of risk factors in particular gene sets that may be relevant to the disease. Common lists include targets of the fragile X mental retardation 1 (FMR1) protein, protein complexes (such as NMDA or ARC), and the proteins found to be at the PSD [18, 20, 136]. However, these enrichment analyses test a small number of PSD interactomes, creating a limited view of synaptic PINs within the PSD. Moreover, there is a large degree of overlap between the proteins found to harbor damaging mutations within different neurodevelopmental disorders [11, 16–24]. This is not surprising since the core scaffold machinery of the PSD is a highly connected component within PSD PPI networks. In accordance with this, interactomes of Dlg4, Dlgap1, and Shank3 have all been shown to be highly enriched in proteins found to harbor de novo coding mutations in patients affected by ASD, ID, and DD, with several risk factors shared between the three complexes (Figure 3A). In addition, numerous risk factors found to interact with all three core scaffolding proteins were also implicated in contributing to multiple disorders through different mutations (Figure 3B) [11]. This emphasizes both the degree of overlap within individual risk factors contributing to multiple complex brain disorders and the centrality of the core scaffolds in PSD PINs affected by them. Supporting their essential role in synaptic function, highly connected proteins in the PSD PINs that are risk factors for complex brain disorders have also been shown to be differentially regulated by phosphorylation upon the induction of LTP. Here, gene-set enrichment analysis has been able to show that while the PSD as a whole is enriched in risk factors for complex brain disorders, the depletion of the phospho-regulated binding partners of Dlg4, Dlgap1, and Shank3 completely eliminates this enrichment [13]. This shows that the core-scaffolding components not only serve as a hub for PPIs within the PSD but also for the regulation of multiple processes involved in synaptic function and complex brain disorders.

Figure 3 –

Core scaffolding PSD interactomes and complex brain disorders. (A) Distribution of risk factors found to harbor nonsynonymous mutations in patients affected by autism spectrum disorder (ASD), intellectual disability (ID), developmental delay (DD), schizophrenia (SCZ), or multiple disorders within the interactomes of Dlg4, Dlgap1, and Shank3. Multiple risk factors are common interactors between the proteins representing the three main layers of the PSD. (B) Heatmap showing proteins found to interact with all three core scaffolding proteins (Dlg4, Dlgap1, and Shank3). Columns indicate if the protein has been implicated in contributing to complex brain disorders in the Online Mendelian Inheritance in Man (OMIM) database and through de novo nonsynonymous mutations in ASD, ID, DD, and SCZ. The abbreviations in the OMIM column are as follows: ASD – susceptibility to autism spectrum disorder, CNH – congenital hypomyelinating neuropathy, DD – related to developmental disorders, ID – Related to intellectual disability, E – related to epilepsy, MC – microcephaly. All data for PPIs and nonsynonymous mutations are derived from [11].

6. Conclusion

Scaffolding proteins within the PSD are essential for the organization of synaptic PINs and play key roles in regulating synaptic function. The core-scaffolding components provide the architectural framework for signaling mechanisms within the PSD and share several characteristics such as possessing an abundance of protein domains specialized in PPIs. These domains allow for the occurrence of multiple PPIs simultaneously throughout the length of the protein and dependent on the scaffold, can also allow for protein multimerization. Multimerization can result in multiple copies of a scaffold within a complex and macromolecular complex assembly. In line with this, characterization of endogenous PPIs involving the core scaffolding proteins have identified hundreds of interactors. These interactomes have a core of highly connected proteins within the PSD PIN that contain the “PSD-risk” for complex brain disease. While significant progress has been made in determining the PPIs of PSD scaffolding proteins, there is still a need for increased quantitative analyses to determine stoichiometric ratios between binding partners and increased coverage of multiple developmental time points to determine the differential association of PPIs throughout time. With the advent of cell-type neuronal differentiation techniques in induced pluripotent stem cells, this may provide a platform to determine PPIs of PSD proteins during early development and model patient-specific mutations in complex brain disorders at a cell-type specific resolution. In this regard, it will be important to develop a correspondent bioinformatic pipeline and database to provide a clear annotation of cell-type specific protein interactions. In addition, the determination of interactomes following the induction of synaptic activity will allow the characterization of PSD PINs during these dynamic processes. With the scaffolding components of the PSD organizing multiple cellular processes and signaling pathways essential for synaptic function, increased characterization of their interactomes under multiple cellular and functional contexts will further elucidate their relationship with complex brain disorders.