Membrane-associated guanylate kinases (MAGUKs) comprise one of the largest groups of scaffolding proteins at the postsynaptic density. MAGUKs have a critical role in the physiology and development of several tissues of metazoan organisms [22]. The family of MAGUKs includes a guanylate kinase-like domain, which contains members such as PSD-93, PSD-95, as well as the synapse-associated proteins SAP-97 and SAP-102. In humans, the SAP-97 and SAP-102 proteins are known to be encoded by genes, “discs large homolog” DLG1 and DLG3. MAGUKs contribute to maintaining excitatory and inhibitory synaptic inputs in the brain by regulating the formation of excitatory glutamatergic synapses. Therefore, MAGUKs are involved in processes of synaptic plasticity by regulating the localization and compartmentalization of glutamate receptors [23]. In the PSD architecture of excitatory synapses, MAGUKs—mainly PSD-95, participate in the organization of ionotropic glutamate receptors and associated signaling proteins, which adjust the synaptic strength. Diverse adaptations of MAGUK proteins in glutamatergic synapses, such as altering interactions between MAGUKs and ionotropic glutamate receptor subunits, often occur in neurodevelopmental disorders [24]. For instance, a mutation in the gene for the glutamate receptor subunit GRIN2B has been found in a patient with autism [25]. These authors explain that impaired GluN2B subunit binding to MAGUK proteins is associated with decreased dendritic spine density. Furthermore, it has recently been widely discussed that developmental clustering and coupling of N-methyl-${}_{\text{D}}$-Aspartate (NMDA) receptor subunits could be affected by MAGUK family proteins [9]. One recent study has suggested that a deficiency in a member of MAGUK family of proteins PSD-95 affects GABAergic inhibitory synapses. The authors of that study explain their results by a PSD-95 deficiency that enhances the inhibitory synapse activity by enhancing transport mechanisms of inhibitory cell adhesion molecules [26]. The relationship between MAGUKs and excitatory/inhibitory balance in the brain is therefore quite complex and depends upon the interactions and trafficking of cytoskeletal components from excitatory to inhibitory synapses.

Another large protein family at the postsynaptic density includes SHANK proteins. They are characterized as multidomain scaffolding proteins, which are associated with neurotransmitter receptors, signaling molecules and actin filaments. In conjunction with their binding partners, SHANK and guanylate kinase-associated proteins (GKAP) contribute to the organization of glutamate receptors in the PSD [27]. Based on their interactions with proteins, they recruit molecular machineries for G-protein-mediated signaling, leading to the control of calcium homeostasis in dendritic processes of neurons. They also participate in morphological changes which lead to the maturation of dendritic spines and synapse remodeling [28]. The proteins of the SHANK family are encoded by the three related SHANK genes: 1, 2 and 3. SHANK family proteins are mostly developed in excitatory glutamatergic synapses in mammalian brains [29, 30]. Although they share high, structural similarity and prevalently localize to the postsynaptic terminal, SHANK proteins are differentially-expressed during brain development, with SHANK2 and SHANK3 finalizing before SHANK1. SHANK2 and SHANK3 are therefore considered critical for several aspects of neuroplasticity during development [31]. Studies have demonstrated that SHANK1 protein levels increase from birth through early development, followed by a slight decrease in adulthood in the cerebellum and cortex [32]. In fact, SHANK1–3 protein components represent critical determinants of excitatory synaptic function. Therefore, whenever balanced expression of the different SHANK proteins is disturbed or diverse alterations in Shank isoforms occur, these events result in various dysfunctions of synaptic transmission (as described below in section 3.5).

SHANK full-length proteins contain ankyrin repeat domains, a SH3 domain, a single PDZ domain, a Proline-rich region (PRD) and a Sterile alpha motif domain (SAM). The PDZ SHANK domain interacts with the “SAPAP” (Synapse-associated protein (SAP) 90/Postsynaptic density (PSD)-95-associated protein) family. The PRD binds to HOMER and Cortactin and this interaction is important for cytoskeletal regulation, as well as for modification of synaptic transmission [33]. SHANK3 is highly enriched in corticostriatal glutamatergic synapses [34], which are a part of neural circuits that are thought to be dysfunctional in ASD. High Shank1 gene expression levels occur in cortical and subcortical brain regions, including the amygdala and hippocampus, while Shank2 gene expression is enriched in cortical areas and the thalamus [34]. In the corticostriatal projections, activity from different areas of the cerebral cortex converges into the striatum and the output is then returned to the cortical layers through thalamic pathways, serving as an aid in motor decisions and behaviour. Therefore, different synaptic errors in different parts of the circuit will ultimately affect cortical activity—this could also be a convergence mechanism for recurrent behaviour with distinct genetic causes particularly in some disorders [35]. Functional deficits of a single copy of SHANK3 are one of the major causes of ASD, Phelan-McDermid syndrome (PMS) and mental disability [6].

The HOMER protein homolog represents an important postsynaptic density scaffolding element, which is involved in various processes of synaptic plasticity, including changes in intracellular calcium. To date, three gene families have been characterized—HOMER 1, 2 and 3 with both long and short-form variants. Long forms of HOMER proteins are a part of the PSD and are consecutively expressed, unlike the short forms, which are known as cytosolic molecules with low expression under basal conditions [36]. Alternative splicing of HOMER into long and short forms has been investigated in the context of the postsynaptic polymeric network structure. For instance, stoichiometric structures connecting HOMER, SHANK and MAGUK proteins have previously been suggested [37]. Indeed, one recent study found that the glutamatergic synapse protein interaction network includes interactions between HOMER and SHANK proteins [38]. The balance between long and short forms of HOMER contributes to normal synapse functioning; therefore, it is assumed that alterations in HOMER expression, folding, and binding to other scaffolding proteins could be associated with the pathological processes found in autism. Multiple controversies are associated with the functional consequences of mutations in HOMER proteins, particularly in the context of autism spectrum disorders and are currently being discussed [39]. HOMER proteins are responsible for the links between metabotropic glutamate receptors and PSD proteins [40], so disruption of the group-I metabotropic glutamate (mGlu) receptor mediated signaling can contribute to the etiology of autism [41].

The calmodulin-dependent protein kinase II-alpha (CaMKII$\alpha$) is a well-defined and abundant protein in the PSD and participates in the control of signal modality in synaptic connections. Activity of CaMKII$\alpha$ has previously been demonstrated as an important factor in the phosphorylation of synaptic Ras GTPase-activating proteins (synGAP —expressed in the postsynaptic density), distribution of synGAP from postsynaptic sites and for the activation of postsynaptic Ras-proximate-1 or Ras-related protein 1 (Rap1). SynGAP represents a neuron-specific Ras and Rap GTPase-activating protein (GAP) found in large amounts in PSD fraction [42]. Additionally, CaMKII$\alpha$ plays an essential role in the regulation of memory and learning processes, as well as an activity-dependent role in spine density limitation during postnatal development [42, 43]. One recent study found that CaMKII$\alpha$ is recruited to the scaffolding protein SHANK3 sub-compartment of the PSD [44]. Moreover, the binding of SHANK3 to CaMKII$\alpha$ potentially modulates dendritic spine morphology [45]. This complicated concept could be useful for explanation of the role of CaMKII$\alpha$ in the abnormal neuronal morphology observed in autism [1, 46]. In fact, mutation in CaMKII$\alpha$ contributes to alterations in dendritic morphology as well as synaptic transmission and thus consequently leads to ASD-related behavioural phenotypes [8].

The T-Box Brain Transcription Factor 1 (Tbr1) represents a gene that encodes a transcription factor with distinct prenatal and postnatal molecular actions. This gene has a high probability of alterations in ASD [50]. Neuron-specific Tbr1 adjusts the regional and laminar character of cerebral cortex regions in the developing brain [51]. Numerous studies have previously found that Tbr1 regulates the identity of neocortical regions in the regional and laminar pathways of the developing brain [51, 52, 53, 54]. The Calcium/Calmodulin Dependent Serine Protein Kinase (Cask) gene codes the MAGUK protein, wherein the N-terminal half consists of a calcium/calmodulin independent kinase (CaMK) domain and two LIN-2,7 (L27) domains, as well as the C-terminus half, which consists of the PDZ domain, Src homology 3 (SH3) and the guanylate kinase (GuK) domain [55]. Wang et al. [56] evidenced that overexpression of Tbr1 and Cask increased promoter activities of some of these genes, including the NMDA receptor 2b (Nr2b), glycine transporter and interleukin 7 receptor (IL7R). It was shown that expression of Nr2b was consistently downregulated in Tbr-1deficient mice. The TBR-1/CASK protein complex is involved in downstream control of the expression of the mentioned genes, thereby modulating brain function. The expression of Tbr1 is upregulated by neuronal activation together with the synaptic PDZ protein CASK and the nucleosome assembly protein CINAP (CASK-interacting nucleosome assembly protein), which forms a complex participating in the control of the glutamatergic receptor subunits [46]. Deriziotis et al. [57] and Sakai et al. [58] investigated direct interaction between TBR1 and Forkhead Box P2 (FOXP2), which is a transcription factor known for its role in brain development. TBR1 was identified as a binding partner for CASK since it requires a C-terminal region. This interaction results in an increase of transcriptional activity of Tbr1 by engagement of the nucleosome assembly protein CINAP (CASK interacting nucleosome assembly protein, also known as testis specific protein Y-encoded like 2, TSPYL2) into a promotor region containing the DNA binding motif for the T-box [59]. Numerous rodent studies have shown that the neurodevelopmental phenotypes associated with the mutation in PDZ protein CASK led to loss-of-function. Heterozygous mice display postnatal progressive microcephaly, hypoplasia of the optic nerves and cerebellum, growth retardation, as well as scoliosis [55, 60, 61]. Mutations in Tbr1 are accompanied by synaptic, neuronal, and behavioural dysfunctions associated with ASD. Tbr1${}^{\mathrm{+}\mathit{}\mathrm{/}K\mathit{}\mathrm{228}\mathit{}E}$ mice displayed increased self-grooming, altered social interaction and increased anxiety-like behaviours [49]. Tbr1-deficient mice develop axonal (presynaptic) defects and other deficiencies in neuron morphology, such as immature dendritic spines and reduced synaptic density [48]. Moreover, it has been found that Tbr1${}^{\mathrm{-}\mathit{}\mathrm{/}\mathrm{-}}$ mice suffer from several other postnatal abnormalities in lamination of the cerebral cortex. In heterozygous animals, the reduction of functional amygdala interconnections has been demonstrated along with autistic-like behaviours [62]. Huang et al. [63] demonstrated that Tbr1${}^{\mathrm{+}\mathit{}\mathrm{/}\mathrm{-}}$ mice manifested less-developed anterior commissures and also olfactory bulbs, a reduced population of interneurons and an atypical morphology of dendrites in olfactory mitral cells. The olfactory sense was not disturbed; however, Tbr1 haploinsufficiency affected olfactory discrimination. Characterization of cellular functions of Tbr1 with the use of mouse models is essential for understanding abnormalities in cortical lamination and in a broader context, it contributes to the deciphering of the complicated origin of neurodevelopmental disorders, including autism.

The fragile X mental retardation 1 (Fmr1) gene is located on the X chromosome and the expansion of trinucleotide CGG leads to its silencing with a consequent absence of the encoded fragile X mental retardation protein (FMRP), which is a synaptic RNA binding protein that modulates the duration of the action potential, as well as synaptic plasticity. These consequences result in a disease known as Fragile X syndrome (FXS) [64]. It has been demonstrated that neuroplasticity disorders belong to the main findings in animal models of FXS and that the lack of balance in inhibitory and excitatory neuronal circuits is the basis of many clinical manifestations of this disorder [65]. Fmr1 gene knockout mice have shown many deformities in neuronal plasticity and learning. Recent studies have suggested that dendritic spines are reduced in juvenile Fmr1 knock-out mice; however, the differences are not noticeable in adult mice [66, 67]. Sitzman et al. [64] demonstrated that Fmr1 deficient mice display deficits in synaptic plasticity and neurite extension. Several studies have also shown metabolic changes, namely that Fmr1 knockout mice have an increased rate of glucose metabolism, exalted oxidative and metabolic stress and deformities in their nitric metabolism of oxygen [64, 68, 69]. Todd et al. [70] demonstrated that FMRP levels are rapidly up-regulated in primary cortical neurons in response to the activation of the type I metabotropic glutamate receptor (mGluR1 and mGlur5) by agonist S-3,5-dihydrophenylglycine. Further, that mRNA for PSD-95 is a scaffolding protein involved in synaptic plasticity, which contains a highly conserved canonical FMRP binding site within its 3’ UTR. Furthermore, PSD-95 translates rapidly in response to the activation of glutamate receptors, since mGluR5 changes in PSD-95 expression are lost in neurons derived from FMRP deficient mice [71]. They hypothesize that FMRP is required for mGluR1 and mGluR5-dependent translation of PSD-95 and provide insight into the pathophysiology of FXS. DeMarco et al. [72] have suggested that FMRP regulates PSD-95 translation via activation of miRNAs, thus working together to mediate the reversible inhibition of the PSD-95 expression in neurons. In addition to alterations in excitatory glutamatergic neurotransmission, recent studies have suggested that Fmr1 knockout mice also suffer from abnormalities in inhibitory GABAergic neurotransmission [73]. Sabanov et al. [73] found that a significantly-reduced amplitude of evoked inhibitory postsynaptic currents decreased expression levels of GABA receptor subunits. Overall, most of the above-mentioned studies support the integrated view of the FMRP role in synaptic plasticity and postsynaptic structural remodeling. It is also clear that mutations or modifications of FMRP play an important role in developmental delays and the etiology of ASD. With the exception of PSD-95, it remains to be elucidated how FMRP might target other postsynaptic density proteins and affect synaptic remodeling.

Contactin-associated protein-like 2 (CNTNAP2) is a transmembrane cell-adhesion molecule type I which occurs in the central and peripheral nervous system. It is highly expressed throughout the spinal cord and brain, especially in the frontal and temporal lobes, dorsal thalamus, striatum and specific cortical layers [74]. It is a member of the neuronexin family and has an essential role in neural development, belonging to one of the most susceptible genes for autism spectrum disorders [17]. Cntnap2 deficient mice are characterized by social deficits and repetitive seizures, and behaviours. Vogt et al. [75] assessed the CNTNAP2 function in mouse development and infection of cortical interneurons, utilizing a Cntnap2 null mouse. Lazaro et al. [76] found a significant decrease in excitatory and inhibitory synaptic inputs on pyramidal neurons of the prefrontal cortex of Cntnap2 knock-out mice. This means that reduced synaptic inputs can gain temporal coordination of neuronal firing in cortical collections. Scott et al. [77] report loss of CNTNAP2 in the rat causes autism-related alterations in social interactions, stereotypic behaviour and sensory processing. Cntnap2${}^{\mathrm{-}\mathit{}\mathrm{/}\mathrm{-}}$ rats lack sociability and social behaviour. Other findings include alterations in sensory processing, increased avoidance to sounds of moderate intensity, deficit of rapid audiovisual temporal recalibration and exaggerated acoustic startle responses. Gao et al. [78] characterized an interaction between CNTNAP2 and the Partitioning-defective 3 protein (PAR3). These authors demonstrated that the two proteins interact through PDZ-mediated binding, since CNTNAP2${}^{+}$/PAR3${}^{+}$ complexes are largely associated with clathrin-coated endocyte vesicles in heterologous cells. Overexpression of Par3, however, not a knockout lacking all PDZ domains, can cluster endogenous CNTNAP2 in primary neurons. Altogether, the aforementioned studies suggest a strong association of CNTNAP2 alterations in excitatory glutamatergic neurons with autism-related behavioural phenotypes. Although the mechanisms associated with how CNTNAP2 regulates glutamatergic synapses are not fully understood, it has been suggested that CNTNAP2 plays a role in the AMPA receptor subunit’s trafficking [79]. Moreover, Cntnap2 knockdown decreased the number of functional excitatory synapses and the consequences observed in altered social behaviour were compensated by enhancement of the AMPA receptor function [80]. Taken together, CNTNAP2 is clearly a part of the molecular architecture of the postsynaptic density, which may be pathologically modified in autism and other neuropsychiatric disorders.

The MAGEL2 protein occurs at high levels in the hypothalamus and plays a crucial role in the regulation of subcellular processes that contribute to the recycling of membrane proteins from endosomes [81]. Protein truncation mutations in MAGEL2 result in Schaaf-Yang syndrome and MAGEL2 comes from a small array of genes deleted in Prader-Willi syndrome. Abnormal daytime sleepiness and nocturnal and/or early morning awakening occur in people with Schaaf-Yang syndrome and Prader-Willi syndrome, while mice bearing a deletion of the Magel2 gene suffer from disrupted circadian rhythms [82]. Although Prader-Willi syndrome is a multigene neurodevelopmental disorder with various physiological and behavioural deficits, the clinical characteristics include severe hypotonia, a newborn’s poor feeding due to reduced suckling reflexes, childhood-onset obesity, lower cognitive functioning and autistic symptoms [83]. Magel2-deficient mice manifest altered gene and protein expression of PSD-95 in early brain developmental stages [5]. Magel2-deficient mice have shown altered levels of oxytocin in the hypothalamus, as well as 50% neonatal mortality due to suckling problems [84]. Male mice who survive were shown to have deficits in the exploration of novel objects, social novelty and dimensional learning [84, 85].

Numerous Shank-deficient mouse models (with a deleted gene belonging to the Shank family) have been generated to provide clarification of the role of Shank proteins in vivo. By comparing individual Shank members, deletion of Shank1 proved to have milder phenotypic deficits than Shank2 and Shank3. Shank2 and Shank3 genes are more structurally similar compared to Shank1 [86]. Here, it needs to be emphasized that marked differences in alterations in behaviour and synaptic function of Shank deficient mice can be explained by remaining Shank isoforms [87, 88]. Moreover, endophenotypes within ASD patients with different types of mutations likely depend on isoform-specific disruption of SHANKs [2].

Shank1 mRNA and proteins are found almost exclusively in the central nervous system when compared to Shank2 and Shank3 in rodents, which are expressed in many tissues, such as the kidney, brain, heart, spleen and liver [34]. Overexpression of Shank1 alters spine morphology [89]. The first mouse Shank model was created by Hung et al. [90] via disruption of the Shank1 gene through deletion of exons 14 and 15, which leads to complete knockout of all Shank1 isoforms. Many studies have provided evidence of an ASD-related phenotype, which is defined by deficits of social communication, modifications in repetitive behaviour and deterioration in cognitive function. Behavioural phenotypes of Shank-mouse models depend on specific gene mutations and modifications, resulting in intellectual disability, schizophrenia and autism symptoms [91]. Shank1 deficient mice are characterized by lower psychomotor activation and higher levels of anxiety. Additionally, they have impaired contextual fear memory and improved acquisition [90, 92]. A study by Mao et al. [93] showed that Shank1 is highly present in Parvalbumin-expressing (PV+) fast-spiking inhibitory interneurons in the hippocampus. It was shown that a lack of Shank1 in these neurons leads to devaluation of afferent excitatory synaptic activity and inhibitory efferent synaptic activity to pyramidal neurons. The hippocampal CA1 pyramidal neurons in Shank1 deficient mice display a change in the excitatory/inhibitory balance, which can be considered a pathophysiologic feature of ASD. Dysregulation of excitatory/inhibitory balance in the hippocampus was displayed by lower expression of gephyrin (a scaffolding component of inhibitory synapses). Mossa et al. [86] investigated how simultaneous deletion of Shank1 and Shank3 affects brain development and behaviour of mice. They showed a low survival rate and a big devaluation in the activation of intracellular signaling pathways associated with Akt, S6, ERK1/2 and eEF2 during development, as well as serious behavioural disorders. It was suggested that Shank1 and Shank3 proteins are essential for developmental adjustment of Akt activation and other intracellular pathways that are critical for mammalian postnatal brain development and synaptic plasticity. It is important to note that a full description of the complex role of the PI3K/AKT/mTOR pathways in brain development is beyond the scope of this review. Nevertheless, the diversity of binding partners is being expanded, as a recent study has discovered new components of mammalian targets of rapamycin (mTOR) signaling pathways in neurodevelopment [94]. In this context, another scaffolding protein which is present in high concentrations in the PSD of neurons located in various brain areas is the Insulin receptor substrate p53 (IRSp53, also known as BAIAP2) [95]. IRSp53 directly interacts with PSD95, SHANK proteins and small GTPases like “Cell division control protein 42 homolog” — Cdc42 [96]. Activation of the Rac/Cdc42 cascade regulates actin filaments polymerization in neuronal dendritic spines [97]. Therefore, IRSp53 belongs to crucial elements linked to the Akt pathway that participates in recruiting cytoskeletal elements and synaptic scaffolds, including PSD-95 and SHANK proteins [96]. Studies have shown that IRSp53 is involved in various brain disorders, including ASD [98, 99], attention deficit hyperactivity disorder [100] and schizophrenia [101]. A deletion of Irsp53 leads to sex-dependent changes in behaviour, such as social deficit, hyperreactivity and excitatory/inhibitory imbalance [102]. A lack of IRSp53 activity leads to NMDA receptor hyperreactivity in the hippocampus [103] and prepulse inhibition in cortical excitatory neurons [102]. The suppression of the mGluR5 and NMDA receptors resulted in improvement of the social deficit in the Irsp53 mutant mice [104].

Knockout mice for the Shank2 gene also display ASD-like behaviour, including reduced social interaction, reduced social communication through ultrasound vocalizations, repeated jumping and also showed a significant decrease in function of the NMDA (N-methyl-${}_{\text{D}}$-aspartate) glutamate receptor (NMDAR) [105]. Conditional Shank2 deficient mice with deletion in excitatory and inhibitory neurons were used in a study by Kim et al. [106]. Significant changes in synaptic transmission in the hippocampus and striatum were demonstrated, which are associated with different behavioural abnormalities in the social, recurrent, locomotor and anxiety domains. Lee et al. [107] found that Shank2 is expressed in GABAergic neurons, including parvalbumin (PV)-positive neurons, which is a neuronal cell type involved in brain excitation and rhythms, as well as brain dysfunctions, including ASD. A novel Shank2 transcriptional variant in a Shank2 deficient mouse model of ASD has shown that two lines of Shank2 knockout mice generated by deleting different exons (exon 6-7 or exon 7) showed distinct cellular phenotypes. Lee et al. [7] also found expression of a novel exon (exon 4${}^{\prime }$ or e4’) between the existing exons 4 and 5 in the Shank2 model of e6-7 Shank2 knockout mice. The results suggested an example of genetic indemnity, leading to phenotypic heterogeneity among ASD patients with the same gene mutation. The causative contribution of rare point SHANK2 gene mutations in schizophrenia has also been suggested [108]. Furthermore, new transcriptional variants can work as modifier genes, thereby contributing to differences between two SHANK2 lines of mutation. Bey et al. [109] investigated determined preliminary behaviour in Shank3-knockout mice in striatal inhibitory neurons and self-grooming lesions were noticed in excitatory neurons of the forebrain. In contrast, social, communicative and instrumental learning behaviours were not significantly affected in these mice. Unique patterns of changes in biochemical and electrophysiological findings in the relevant brain regions have been discovered, which reflects the complex nature of Shank3 transcriptional regulation. Decreased HOMER1b/c and hyper-excitability of the membrane were observed with striatal loss of SHANK3. In comparison, the Shank3 deletion in hippocampal neurons led to an increase in NMDA receptor mediated excitatory current [109]. In individuals with autism spectrum disorders, hyper-reactivity with sensory input is a common feature. In a study by Chen et al. [110], both male and female Shank3-deficient mice were more sensitive to relatively weak tactile stimulation. Spontaneous and stimulus-induced calcium jets in pyramidal neurons were detected in population-based calcium imaging of the population; however, activity in interneurons was reduced. Preferential deletion of Shank3 in inhibitory interneurons indicated pyramidal neuron hyperactivity. This study suggests that cortical GABAergic interneuron dysfunction plays a key role in sensory hyper-reactivity in the Shank3mouse model of ASD and identifies a potential cellular target for the exploration of therapeutic interventions.