Postsynaptic Proteins at Excitatory Synapses in the Brain—Relationship with Depressive Disorders

Depressive disorders (DDs) are an increasingly common health problem that affects all age groups. DDs pathogenesis is multifactorial. However, it was proven that stress is one of the most important environmental factors contributing to the development of these conditions. In recent years, there has been growing interest in the role of the glutamatergic system in the context of pharmacotherapy of DDs. Thus, it has become increasingly important to explore the functioning of excitatory synapses in pathogenesis and pharmacological treatment of psychiatric disorders (including DDs). This knowledge may lead to the description of new mechanisms of depression and indicate new potential targets for the pharmacotherapy of illness. An excitatory synapse is a highly complex and very dynamic structure, containing a vast number of proteins. This review aimed to discuss in detail the role of the key postsynaptic proteins (e.g., NMDAR, AMPAR, mGluR5, PSD-95, Homer, NOS etc.) in the excitatory synapse and to systematize the knowledge about changes that occur in the clinical course of depression and after antidepressant treatment. In addition, a discussion on the potential use of ligands and/or modulators of postsynaptic proteins at the excitatory synapse has been presented.


Introduction
Depressive disorders (DDs) are widespread mental illnesses worldwide and pose a significant economic and psychosocial problem. The World Health Organization (WHO) estimates that depression affects about 3.8% of the human population (with a prevalence of 5% among adults), and it increases with age (in people over 60, its frequency is 5.7%) [1,2]. It has also been observed that the lifetime risk of developing DDs is twice as high in women as in men [3]. Moreover, mental illnesses (including DDs) are important health problem among adolescents and represent one of the leading causes of disease and disability in this age group [4][5][6]. Over 50 years ago, monoamine theory, assuming that symptoms of depression were related with deficiency or imbalances of monoamines systems, i.e., serotonin, norepinephrine, dopamine, was described. To this day, drugs able to modify (increase) the brain concentration of these neurotransmitters represent the most used type of pharmacotherapy [7,8]. Unfortunately, this therapy has many disadvantages, e.g., it takes weeks or months to achieve a therapeutic effect, low remission rate, and a high risk of relapse after responding to treatment [3,9]. An additional problem is the prevalence of treatment-resistant depression (TRD), characterized as the failure to achieve an inadequate response to at least two standard antidepressants. This condition is common in clinical practice. Up to 40% of major depressive disorder (MDD) patients suffer from TRD [10][11][12][13]. These factors can discourage patients from taking antidepressants regularly or even lead to their discontinuation, significantly reducing the chance of remission and triggering several 2 of 42 complications including suicide [14]. A better understanding of the molecular mechanisms underlying DDs, including MDD and bipolar disorder (BD), seems to be necessary to obtain new, more effective drugs (both antidepressant and anxiolytic profile) [3,[15][16][17].
Many environmental factors contribute to the development of DDs, among which stress is particularly important [18][19][20][21]. Numerous studies on depressed patients and animal models based on stress-related procedures have shown that specific brain areas (i.e., prefrontal cortex, hippocampus, amygdala, insula) have an altered volume [22][23][24][25][26][27][28]. It's well documented that stress factors impair the expression of neurotrophins and cause an increase in the level of the pro-inflammatory cytokines, which may result in atrophy, depleted neurogenesis, and consequently changes in neuroplasticity [21,[29][30][31][32][33][34]. Moreover, it was shown that acute stress causes an increase in extracellular glutamate (Glu) levels in the hippocampus and medial prefrontal cortex (mPFC), which may lead to excitotoxicity [35][36][37]. These results also suggest that imbalance between excitatory and inhibitory neurotransmission can be a potential substrate of DDs. The truth of the above evidence is confirmed by the ever-growing interest in the role of the glutamatergic system over the decades [15,[38][39][40]. A thorough understanding of the mechanisms responsible for the Glu metabolism and its influence on postsynaptic proteins at the excitatory synapse is a new research direction for achieving effective pharmacotherapy of DDs [13,15,41,42], as evidenced by the growing number of clinical studies documenting the rapid and robust antidepressant effect of ketamine (N-methyl-D-aspartate receptors (NMDAR) antagonist with additional effects on α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPAR), hyperpolarization-activated cyclic nucleotide-gated (HCN) channels, L-type voltage-dependent calcium channel (L-VDCC), opioid receptors, and monoaminergic receptors) [43][44][45].
Glu acts through ionotropic and metabotropic receptors. However, NMDARs and AM-PARs seem to be the most important for synaptic plasticity [46][47][48][49]. Synaptic plasticity, that is, changes in the onset or magnitude of long-term potentiation (LTP) or long-term depression (LTD), can be regulated by changing the number, types, or properties of these receptors in the postsynaptic membrane. AMPARs and NMDARs trafficking underlie activityinduced changes in synaptic transmission, and therefore their abundance at synapses can significantly enhance or weaken it [50][51][52][53]. The excitatory synapse is a highly dynamic structure in which receptors constantly circulate between the synaptic membrane and the cytoplasm as well as between the extra-and synaptic matrix, while postsynaptic density (PSD) proteins as well as post-translational modifications of the polypeptide chains of the receptors' subunits play an important role in locating them and transmitting signals inside the cell [49,[53][54][55][56][57]. PSD proteins modulate the signaling cascade by linking synaptic transmission from presynaptic neurons and neurotransmitter systems, mainly through by NMDARs, AMPARs, and group I metabotropic Glu receptors (especially mGluR5) [58][59][60][61]. A thorough understanding of the mechanisms responsible for the Glu turnover and its influence on postsynaptic proteins at the excitatory synapse is a new target of direction effective pharmacotherapy of DDs [41]. There is much evidence of changes in postsynaptic proteins at the excitatory synapse in depressive disorders, both in human and animal tissues. On the other hand, there are many inconsistencies in these findings. Taking this into account, the main goal of this review was to synthesize knowledge about changes in both PSD protein levels and post-translational modifications in MDD and BD patients, as well as in animal models of depression, and to determine whether these changes are characteristic of selected areas of the brain. In addition, we also reviewed the changes in these proteins following the administration of antidepressants (or compounds with antidepressant-like activity), and the expression of the genes encoding them.
intracellular AMPARs to the PSD and phosphorylation of the GluA1 (S831) subunit, leading to enhanced synaptic transmission [75]. Additionally, CamKIIα, displaced from the cytoplasm to the PSD, also interacts with the GluN2B subunit, which is necessary for the induction of LTP [87].
In addition to rapid neurotransmission via ionotropic receptors, metabotropic receptors are also crucial at the excitatory synapse, among which metabotropic Glu receptor 5 (mGluR5) plays an important function in the development of DDs [88]. mGluR5 is encoded by GRM5 gene, belongs to the family of G-protein coupled receptors (GPCRs), and, together with mGluR1, is part of group 1 and mainly localized postsynaptically [89,90]. A characteristic feature of mGluRs is their occurrence in the form of homodimers and their structure because they contain a large extracellular domain at the N-terminus called a Venus flytrap (VFT), which possesses a ligand-binding site and, via a cystine-rich domain (CRD), binds to the 7-transmembrane domain (7TM). There is a C-terminal domain in the cytoplasm of mGluR, which allows the receptor to interact with other PSD proteins [91,92]. After Glu binds to mGluR5, a series of conformational changes occur, leading to activation of the phospholipase C (PLC) pathway and production of secondary messengers, e.g., inositol 1,4,5-triphosphate (IP 3 ) and diacylglycerol (DAG) consequently responsible for slow neurotransmission [93]. The effect of IP3 is a Ca 2+ influx, which affects calcium-dependent proteins such as CamKII [90,94]. Interestingly, mGluR5 also involves the NMDAR complex, as Jin et al. showed that the application of (RS)3,5-dihydroxyphenylglycine (3,5-DHPG, mGluR5 agonist) resulted in increased expression of membrane GluN1 and GluN2B subunits and reduced their levels intracellularly. This effect was abolished when 3-((2-methyl-1,3-thiazol-4-yl)ethynyl)pyridine hydrochloride (MTEP, mGluR5 selective antagonist) was used, indicating an essential role for mGluR5 as a molecule that determines the movement of NMDAR subunits to the surface [94].
To maintain the correct receptors composition, a balanced scaffold with many support proteins is necessary. One of them, especially widespread in forebrain, is postsynaptic density protein 95 (PSD-95), which is responsible for stabilizing and binding other PSD proteins [67]. PSD-95 is encoded by DLG-4 (discs large homolog 4) gene and belongs to the MAGUK (membrane-associated guanylate kinases) superfamily [65]. PSD-95 in its structure contains three PDZ (PSD-95/disc large/zonula occludens-1) domains followed sequentially by single SH3 (Src homology 3) and GK (guanylate kinase-like) domains [95]. PDZ domains allow PSD-95 to interact with many PSD proteins, including receptor (e.g., NMDAR, AMPAR, serotonin 5-HT 2 , dopamine D 2 ) subunits, thus exerting a vast influence on glutamatergic, serotonergic, and dopaminergic transmission [96]. Chen et al. showed that RNA interference knockdown of PSD-95 caused PSD deficit and decreased AMPAR (but not NMDAR) levels in hippocampal neurons, thus confirming the crucial role of PSD-95 in maintaining correct protein architecture [97]. It has been observed that overexpression of PSD-95 can promote the formation of multi-innervated spines with up to seven presynaptic connections. The interaction of the PDZ2-domain of PSD-95 with nNOS (neuronal nitric oxide synthase) plays a vital role in this process [96,98]. nNOS is one of the three isoforms of the NOS enzyme and is widely distributed in the CNS and has also been demonstrated in human neutrophils [99]. Nikonenko et al. showed that both small interfering RNA (siRNA)-mediated knockdown of nNOS and pharmacological blockade by administration of L-N G -nitroarginine methyl ester (L-NAME, NOS inhibitor) result in the inhibition of multi-innervated spines formation, thus demonstrating the role of PSD-95 and nNOS as essential factors for synapse building [98].
The activity of individual elements included in the NMDAR-PSD-95-nNOS complex is modulated by various isoforms of the Homer 1 protein (Homer protein homolog 1) [100]. Homer 1 is widely distributed in the CNS and skeletal muscles and has nine isoforms, including Homer 1a-h and Ania-3 (Activity and neurotransmitter induced early gene 3) [101,102]. Homer-1 proteins consist of a Homer family specific EVH1 domain (enabled/vasodilatorstimulated phosphoprotein homology 1), a proline motif found in all Homer 1 proteins, and CC (coiled coil) domain, allowing the formation of homo-and heterooligomers. Short isoforms such as Homer 1a and Ania-3 lack a CC domain and function as negative modulators of Glu receptors [103]. Through the EVH1 domain, Homer 1 forms connections with other proteins, e.g., mGluRs (group 1), and can regulate their function. Wang et al. showed that Homer 1a is responsible for uncoupling between GluN2B, PSD-95, and nNOS, which was associated with the reduction of NMDAR-mediated transmission and thus had a neuroprotective effect. At the same time, Homer1 b/c facilitated mutual interactions among the NMDAR-PSD-95-nNOS complex [100]. Furthermore, an association between rs7713917 variant in the HOMER1 gene and an increased risk of suicide attempts and a worse response to antidepressant treatment with sleep deprivation and light therapy has been shown [104,105].
Homer 1 co-occurs with Shank (SH3 and multiple ankyrin repeat domain), forming a mesh-like structure that provides a scaffold for the other PSD proteins [106]. Shank3 (SH3 and multiple ankyrin repeat domains 3) is a member of the SHANK family, abundant at the excitatory synapse, and plays a vital role in the proper maturation and formation of dendritic spines [107]. Shank3 consists of a Shank/ProSAP N-terminal (SPN) domain followed by ankyrin repeats, src homology 3 (SH3) domain, PDZ domain, and sterile alpha motif (SAM) domain. PDZ domain binds to guanylate kinase-associated protein (GKAP). It thus indirectly allows Shank3 to interact with NMDARs and AMPARs [108], while the SAM domain is located at the C-terminal and binds zinc ions [109]. Duffney et al. showed that Shank-3 knockdown using si-RNA in cortical structures resulted in decreased NMDAR-mediated ionic current and decreased GluN1 subunit expression, significantly confirming the importance of Shank3 in correct excitatory synapse function [110]. Shank3 occurs in the Shank3a-f isoforms and is expressed in the nervous system, e.g., cortex, cerebellum, amygdala, hippocampus, spinal cord, striatum, dorsal root ganglia, but also in the heart, thymocytes, and spleen [107,111]. Interestingly, in a group of seven patients with Phelan-McDermid syndrome (22q13.3 deletion, which is associated with a deletion of the SHANK3 gene), four showed the presence of bipolar depression, which suggests a close relationship between Shank3 and this disorder [112].

Alterations in Postsynaptic Density Proteins in Depressive Disorders
Recent years, especially the last decade, have seen abundant research performed on postsynaptic proteins in the context of DDs. This review focuses only on the most important proteins, i.e., NMDAR, AMPAR, PSD-95, CamKII, Homer 1, Shank3, ZnT-1, and nNOS.
In the publications, we found clinical studies showing differences in PSD proteins expression in both depressed patients (Table 1) and preclinical studies, using animals (both mice and rats) showing depressive-like (mainly stress-induced) behaviors, as well as after treatment with drugs with antidepressant or antidepressant-like activity (Tables 2-5). The observed alterations were sometimes varied and ambiguous, which may also be related to the diversity of the brain regions (hippocampus, prefrontal cortex, amygdala and locus coeruleus) in which the analyses were carried out. In addition, we found studies showing changes in peripheral blood cells (Table 1). Due to the large number of studies with the use of animal models, we divided our review according to the type of tested proteins (NMDAR and AMPAR, Tables 2 and 3; other PSD proteins, Tables 4 and 5) as well as by the species of animals used in studies (mice , Tables 2 and 4; rats, Tables 3 and 5). In this article, we included both research on protein and mRNA level changes.   ↓NOS activity (L-Citrulline/L-Arginine ratio) in whole MDD group ↓NOS activity in drug-free MDD ↑NOS activity in whole MDD group at 3 months and 6 months ↑NOS activity in MDD responders group at 3 months and 6 months ↔NOS activity in MDD non-responders group at 3 months and 6 months Loeb et al. [123] SH3 and multiple ankyrin repeat domains 3

Shank3
No control MDD = 24 women BD = 32 women (All participants were treated with antidepressants and/or mood stabilizers of the 1st and 2nd generation)

Human Studies
Based on the review, it can be concluded that among the PSD proteins, NMDAR analyses in post-mortem brains of patients with DDs constitute the vast majority. Most authors indicate no changes in both the protein and mRNA levels of the GluN1 subunit [113,114,[116][117][118]120].
Interestingly, these observations apply to both MDD and BD patients [116]. Although slightly contrasting results (decrease in GluN1 mRNA level) were shown by Beneyto et al. [115], different gender ratios may be of importance. Similarly, according to Gray et al., GRIN1 gene expression was higher in women with MDD, while it did not change in men. Hence, we can conclude that an important factor influencing the NMDAR proteins level is, among other things, sex [68]. Analyses of different NMDAR subunits also confirm this. As shown by Gray et al., the expression of GRIN2A and GRIN2B was higher in the group of women with MDD, while it remained unchanged in men [68]. Other authors also indicate no changes in GRIN2A expression in the MDD group with predominantly males [113,115,116]. These observations seem to be consistent, despite the different areas of the brain studied [113] or the analytical methods used [68,113,116]. On the other hand, decreased expression of GRIN2B has been observed post-mortem in the locus coeruleus of the brain, which may indicate a multifactorial etiology of DDs and the participation of the noradrenergic-glutamatergic component [113]. GRIN2A and GRIN2B expression also appears to be relatively similar in MDD and BD patients. However, when we look at the levels of these proteins in different studies, the results are more varied and inconsistent with the expression of the genes that code for them. For example, two studies [74,117] showed a decrease in GluN2A level, while subsequent studies [118,119] showed its increase. The reason for these differences seems to stem from the dissimilarity of the studied structures (prefrontal cortex vs. lateral amygdala and hippocampus), which appeared to be of secondary importance in mRNA analysis. This hypothesis may not be entirely accurate when we analyze the changes in the GluN2B protein, which show a decrease in both the prefrontal cortex and the hippocampus [117,119] and no changes in the lateral amygdala [118]. The very cause of subjects' death in the case of protein analysis may be irrelevant, which seemed to be essential for the study of gene expression [68]. Observed discrepancies may be explained by the different functional morphology and physiology of NMDARs in other parts of the CNS. As already mentioned above, NMDARs form heterotetramers, most often composed of two GluN1 subunits that bind to two additional subunits: GluN2 (A-D), which binds Glu, or less commonly GluN3 (A-B) with high affinity for glycine [180]. NMDARs most often consist of GluN1 and GluN2 subunits, particularly GluN2A and GluN2B [181,182]. Importantly, NMDARs containing GluN2A subunits show three times faster degradation times and reduced Glu affinity compared to GluN2Bcontaining receptors [183]. Mellone et al. showed that ZnT-1 in hippocampal neurons binds to GluN2A (1049-1464) C-terminal (but not GluN2B) and modulates PSD-95 (postsynaptic density protein 95) activity and dendritic spike morphology [184]. Importantly, studies on MDD subjects showed an increase in the level of ZnT-1 protein, which may suggest significant changes in NMDAR signaling caused by this transporter [74]. Recent studies have shown that activation of NMDA GluN2A receptors exerts a survival-promoting effect, while NMDA GluN2B receptors, which are mainly segregated into extrasynaptic sites, show deleterious effects [185]. Two further subunits, GluN2C and GluN2D, which are known to be positively involved in synaptic transmission and working memory, are also essential [186]. In addition, it is known that NMDARs in the CNS mainly comprise triheteromeric receptors consisting of the GluN1/GluN2A/GluN2B subunits. Other triheteromeric NM-DARs, including GluN1/GluN2A/GluN2D and GluN1/GluN2B/GluN2D, have been observed in the human spinal cord as well as in the rat thalamus and midbrain. Much research to date has focused on diheteromeric NMDARs containing identical GluN2 or GluN3 subunits. However, the triheteromeric NMDAR containing a combination of GluN2 and/or GluN3 subunits has different channel gating kinetics and pharmacology from diheteromeric receptors [180]. With that in mind, extending protein research on NMDARs with additional subunits and studies showing their coexistence and variety in selected regions of the brain is essential.
Like NMDARs, AMPARs are abundantly located in the postsynaptic membrane and their effects are interdependent. During LTP, presynaptic Glu release activates AMPARs, and the triggered depolarization removes the Mg 2+ blockade of the NMDA channel and allows Ca 2+ influx. Strong activation of NMDARs triggers the Ca 2+ -calmodulin protein kinase II (CamKII) signaling cascade, which leads to LTP, brain-derived neurotrophic factor (BDNF) secretion and synaptic amplification [185]. Human post-mortem studies of AMPAR-building proteins show much greater variability compared to NMDAR analyses. The vast majority of these studies concern gene expression [68,113,120], while one study shows changes in protein levels [74]. Two papers show no change in the GRIA1 gene expression [68,113], and one of them shows its decrease [120]. In this situation, however, attention should be paid to the structural heterogeneity of the analyzed brain tissue. On the other hand, sexual differentiation is a likely cause of heterogeneous observations in the case of GRIA2 gene expression. Duric et al. [120] in the hippocampus and Chandley et al. [113] in the locus coeruleus showed no changes in the GRIA2 level, while Gray et al. [68] showed an increase in the expression of this gene in women and no changes in the group of men. Notably, most of the groups studied in the work of Chandley et al. and Duric et al. were men [113,120]. In addition, a decrease in GRIA3 expression and no change in GRIA4 level in MDD patients has been shown [120]. The observed AMPAR subunits changes have implications for the functionality of the entire receptor. It is indicated that AMPARs being a combination of GluA1 and GluA2 are essential for the plasticity of neurons and are rapidly recycled, therefore their number in the cell membrane reflects the balance between endo-and exocytosis processes. GluA1 subunits are delivered to the synapses in an activity-dependent manner, and GluA2/3 subunits are continuously provided to synapses independently of synaptic activity. Therefore, the trafficking process is an essential mechanism underlying synaptic plasticity since the recruitment of AMPAR to the postsynaptic membrane is positively correlated with LTP, and their endocytosis negatively correlated with LTP [185].
The most crucial PSD protein, necessary to maintain the molecular organization of postsynaptic density, anchors NMDARs and AMPARs, and mediates intracellular signaling is PSD-95 [187] (Figure 1). So, it is not surprising that it has been extensively studied in the context of DDs pathophysiology. Among the studies, almost all showed a reduced level of PSD-95 protein in both suicides and patients with MDD, including those who reported suicide [74,114,117,119]. Only one study indicates an increase in PSD-95 levels, but this may be due to a different brain structure (lateral amygdala) [118]. On the other hand, studies of DLG4 gene expression indicate its decrease, and this effect may vary depending on the diagnosis (MDD vs. BD) [121].
Peripheral blood is also a source of PSD protein mRNA. It has been observed that the level of SHANK3 in peripheral blood mononuclear cells (PBMCs) correlates with treatment response among women with MDD, indicating the potential employment of SHANK3 as a marker of antidepressant response [124]. Although some studies suggest a genetic predisposition to DDs, Somani et al. showed that neutrophils of drug-naïve MDD patients significantly have higher nNOS mRNA expression, but no such correlation was found among first-degree relatives of these patients [99]. In contrast, the locus coeruleus showed reduced nNOS protein immunoreactivity, which was absent in the cerebellum [122]. Different results were observed in the lateral amygdala as there were no significant differences in nNOS protein levels in patients with MDD and adjustment disorder with depressed mood [118]. In both nNOS analyses, post-mortem toxicology studies did not show antidepressant drugs use. This is a fundamental fact, because a case-control treatment study showed that the activity of NOS in the blood was initially lower in patients with depression than in healthy controls but increased in patients who responded to treatment. In this study, no changes in NOS activity were found in depressed patients who were not receiving drugs and were not responding to treatment. There is also no correlation between the class of antidepressants used and changes in NOS activity [123].

Animal Studies
In research on the pathophysiology of depression and amid the search for new, more effective antidepressants, animal models that initiate behavior similar to depression in humans are most often used. The models based on the use of various stressors are the best validated and most frequently applicated. Among them, we find: Chronic Unpredictable Stress (CUS), Chronic Unpredictable Mild Stress (CUMS), Chronic Restraint Stress (RST), Chronic Social Defeat Stress (CSDS), and Chronic Mild Stress (CMS) [188,189].
The stress response in different brain regions can vary dramatically. As is already known, one of the causes of depression is neurotransmission dysfunction in the brain. Available research suggests that the glutamatergic system is involved in the pathophysiology and treatment of depression [145]. NMDARs, especially GluN2A-and GluN2B-containing, are essential for regulating neuronal plasticity [128]. Molecular analysis in the hippocampus of female mice showed an increase in the expression of Grin2A and Grin2B genes after CUS and after administration of two hormonal compounds (estradiol, progesterone) [131]. Reports suggest estrogens' role in structural and functional synaptic plasticity and long-term potentiation (LTP) in the hippocampus [131]. In the same brain region, the levels of GluN2A, GluN2B proteins were elevated after CUMS [128]. Interestingly, the increased levels of genes encoding NMDAR proteins were demonstrated by Tamasi et al. also after venlafaxine treatment in rats [152]. On the other hand, administration of lurasidone and fluoxetine to mice and infenprodil to rats also lowered the levels of these proteins in the hippocampus, prefrontal cortex, and medial prefrontal cortex (p-GluN2B), respectively [127,145]. Moreover, the analysis of NMDA genes and proteins after 28 days of ketamine administration showed a significant decrease in the hippocampus [130]. In contrast, in the basolateral and inferior limbic prefrontal cortex, ketamine decreased the expression of only GluN2B protein [129]. Other authors also indicate that GluN2A protein levels in the hippocampus were significantly lower than GluN2B in FSL rats [146]. In turn, in a rat model of zinc deficiency, GluN2A and GluN2B protein levels were reduced, while fluoxetine treatment had no significant effect [73]. Given the above, it can be concluded that changes in NMDAR expression levels may be largely dependent on factors inducing behavioral changes and appear more critical for the antidepressant response.
AMPARs for Glu, particularly those containing the GluA1 and GluA2 subunits, also contribute to normal plasticity of neurons, and various factors can modulate their in a multidirectional manner. For example, in rats, stress increased the GRIA1 expression in the paraventricular nucleus of the hypothalamus and the level of the GluA1 in the hypothalamus [161]. On the contrary, in stressed mice, the levels of AMPAR protein (and selected subunits) were significantly lower in the medial frontal cortex [135] and the prefrontal cortex [137]. The surface receptor crosslinking of BS 3 showed a decrease in GluA1 and GluA2 levels associated with stress in the hippocampus [139]. Emerging data suggest that some fast-acting drugs in DDs, such as scopolamine, increase Glu release and induce neurotrophic factors through AMPAR activation [137]. It was observed that the administration of fluoxetine [135] and scopolamine [137] reversed stress-induced changes. In contrast, chronic stress increased in GluA1 and p-s845-GluA1 levels in the basolateral amygdala, while fluoxetine therapy decreased their levels [136]. Chronic ketamine administration caused a decrease in AMPARs, both gene and protein levels in the whole hippocampus and CA1 region [130]. Moreover, treatment with infenprodil, fluoxetine, and S-ketamine, venlafaxine, and NaHS normalized GluA1 protein stress-induced changes in the medial prefrontal cortex and hippocampus, respectively [145,150,154,156,162]. In its turn, the application of ketamine and memantine, induced a significant increase in the pS845-GluA1 protein subunit in the hippocampus [160]. Another interesting line of research turned out to be lipopolysaccharide-induced inflammation that caused a substantial decrease in GluA1 protein in the hippocampus while the administration of ketamine and Ac-YVAD-CMK (the selective NLRP3 inflammasome inhibitor) reversed its adverse effects [134].
PSD proteins, such as PSD-95, CamKII, Homer 1, and Shank3 may also be involved in the development of depressive disorders. Analysis of the Dlg4 gene of stressed animals revealed a decrease in this gene in the prefrontal cortex, hippocampus, and hypothalamus in mice [167] and gene and protein increases in the hypothalamus in rats [161]. On the other hand, in the hippocampus, it showed a decrease in protein in stressed animals [166]. Interestingly, PSD-95 protein levels were significantly higher in the basolateral amygdala after stress [136]. Application of fluoxetine, asioaticoside in mice, and ketamine or sodium hydrosulfide-NaHS (CA1, CA3 region) in rats increased PSD-95 protein levels in the hippocampus [156,162,166]. In contrast, in the basolateral amygdala, fluoxetine increased protein levels [136]. Chronic administration of ketamine decreased PSD-95 protein levels in the hippocampus, while in combination with betaine it increased them [164]. In stressed rats, administration of lurasidone increased Dlg4 gene levels in the prefrontal cortex [175]. In contrast, in male mice, lurasidone and fluoxetine induced a decrease in PSD-95 after both drugs in the hippocampus and prefrontal cortex [127]. It was also demonstrated that infenprodil and YY-21 decreased protein expression after CUMS in the medial prefrontal cortex in rats [145,150]. In the RST model, imipramine administration increased PSD-95 expression in lateral nuclei and basal nuclei and decreased in medial prefrontal cortex [165]. The application of CX717 therapy increased protein levels in mPFC (2h after administration) [174]. Leem et al. also showed a decrease in p-CAMKII expression in basolateral amygdala (BA) and an increase in medial prefrontal cortex (mPFC) after drug administration [165]. CAMKII dysfunction is involved in many neurological disorders, including depression. Short-term manipulation of CAMKII can result in long-term effects on disease-related behavior. On this basis, it can cause structural changes which in turn contribute to the progression of the disease and its duration [170]. Sleep deprivation caused a decrease in the CAMKII protein subunit in the prefrontal cortex and hippocampus and administration of citalopram reversed its effects in mice [168,171]. In addition, venlafaxine treatment increased the levels of Camk2g and Camk2b genes in the frontal cortex in rats [152].
Homer 1 also has its role in synaptic plasticity. Homer genes encode a family of proteins in the PSD, where they act as multimodal adaptors. Overexpression of Homer 1a protein may contribute to decreased density of postsynaptic proteins such as Shank and inhibit postsynaptic AMPAR and NMDAR currents. Moreover, it is also involved in Gluinduced changes in the distribution of pre-and postsynaptic proteins [96]. In stressed mice, a reduction in the Homer 1a gene expression in the prefrontal cortex and protein level in both the prefrontal cortex and hippocampus were noted [128,172]. Contrary observations were made in the prefrontal cortex of stressed rats in which an increase in the Homer 1 protein was shown [177]. Furthermore, the administration of imipramine, ketamine and fluoxetine increased Homer 1a, Homer 1b/c gene expression in the cortex of mice [172].
At the cellular level, zinc is one of the main enzymes involved in biochemical processes, and disturbance of homeostasis leads to physiological or pathological problems. Despite the high demand for zinc in cells, its levels must be kept low. Among others, the ZnT protein contributes to the maintenance of zinc balance. Rafalo-Ulinska et al. showed a decrease in this protein after zinc and imipramine supplementation in the prefrontal cortex, and an increase in the hippocampus. In contrast, supplementation with zinc alone contributed to a decreased ZnT-1 protein in the prefrontal cortex [173].
An important signaling molecule in CNS regulating anxiety behavior is nNOS. It has been shown that nNOS-derived free radical (NO) contributes to depression caused by chronic stress [190]. nNOS activity is also coupled to NMDAR activity via the membranebound postsynaptic density protein PSD-95. While PSD-95 expression is inhibited, calcium ion-activated NO production via NMDAR activation is blocked, and excitotoxicity is reduced [191]. In a rat model of CUS, memantine therapy reduced nNOS protein levels [179]. Interestingly, Yin et al. demonstrated the importance of gender on nNOS protein level.
In a mouse stress model, female hippocampal nNOS protein levels decreased while male increased [169]. In addition, it seems of interest to use escitalopram and Yueju-Ganmaidazao (YG), which reversed the levels of this protein in stressed animals [169].

Postsynaptic Density Proteins as Therapeutic Targets for DDs
Among the discussed post-synaptic proteins of the excitatory synapse, the most promising seem to be research on the possibilities of modulating glutamatergic transmission by influencing Glu receptors. Because numerous preclinical studies have shown the function of various ligands or modulators of glutamate receptors, in this paper, the authors will focus mainly on clinical trials, which give more information about the safety and effectiveness of new antidepressant therapy. Targeting pharmacotherapy of DDs to disrupt glutamatergic signaling began with using low sub-anesthetic doses of ketamine, a drug commonly used in anesthesiology [192]. The first clinical trial in patients with MDD and BD using a 40-min infusion of racemic ketamine intravenously at a subanesthetic dose of 0.5 mg/kg was conducted by Berman et al. In the ketamine group, a sustained 72 h change in scores expressed by the 25-item Hamilton Depression Rating Scale (HDRS/HAM-D) have been noticed. A the same time, in the control group, there were no significant differences [193]. Zarate et al. used the same protocol of ketamine treatment (dose, route of administration) in TRD patients. Results indicate that ketamine induced a rapid antidepressant effect (measured by the 21-item HAM-D scale and Beck Depression Inventory (BDI) that remained for seven days [194]. This finding contributed to the subsequent studies on ketamine with multiple intravenous administration in a larger group of patients. Among 97 patients with TRD, 205 ketamine intravenous infusions (0.5 mg/kg/40 min; performed 2006 and 2012) showed antidepressant activity (improvement in Montgomery-Asberg Depression Rating Scale (MADRS) scores) in 67% of TRD individuals [195]. Ketamine was also administered by intramuscular (i.m.) injection with nearly 100% bioavailability. At 0.5 mg/kg dose, there is an improved antidepressant response in MDD subjects, similar to effects observed in the group receiving electroconvulsive therapy (ECT). Moreover, oral ketamine administration at a 1 mg/kg dose (despite its relatively low bioavailability on range 20-25%) for three weeks showed pharmacological efficacy (measured by the 17-item HAM-D and Beck Scale for Suicidal Ideation (BSSI) scales) compared to ECT [196]. On the other hand, sublingual ketamine shows a slightly higher bioavailability (30%), but even more importantly, indicates an improvement in mood, sleep, and cognitive functions in 26 patients with MDD and BD. Transient light-headedness was reported side-effects of this type of treatment [197]. Moreover, the Food and Drug Administration (FDA) approved esketamine nasal spray in 2019 and it is now a promising new therapeutic option for treating TRD [198].
In clinical trials, other NMDAR antagonists were also used, e.g., memantine (an uncompetitive, low-affinity, and selective open-channel blocker of NMDAR). A growing body of evidence shows that memantine significantly reduced depressive symptom scores in MDD patients [199]. Moreover, limited evidence shows its effectiveness in bipolar patients as an add-on treatment [200]. Recently, the increasing number of substances targeted to regulation-specific NMDAR subunits show potential antidepressant efficacy [201]. One of them is traxoprodil (CP-101,606), an antagonist of the GluN2B subunit, the use of which in monotherapy and in combination with escitalopram, imipramine, or fluoxetine in male Swiss mice resulted in a reduction of immobility time in a forced swim test (FST) [202]. Moreover, in a group of patients using paroxetine (but not responding to selective serotonin reuptake inhibitors; SSRIs), intravenous infusion of traxoprodil (0.75 mg/kg/h for 1.5 h and then 0.15 mg/kg/h for 6.5 h) resulted in a significant reduction in MADRS scores and a higher response rate in 17-item HAM-D compared to the placebo group [203]. Furthermore, a small clinical trial involving five patients with TRD showed that oral intake 4-8 mg/day of another GluN2B antagonist, rislenemdaz (MK-0657), resulted in significant changes in 17-item HDRS and BDI scale scores compared to placebo [204]. In addition, other GluN2B subunit antagonists such as EVT-101 (also known as ENS-101) or MIJ821 are the subject of clinical trials [201,205].
Another strategy of experimental pharmacology of DDs is based on the use of partial NMDAR agonists. [192]. The example is D-cycloserine (DCS), a partial agonist at NMDAR-associates glycine. Heresco-Levy et al. showed that oral administration of DCS at a gradually increasing dose (up to max 1000 mg/day) resulted in significant clinical improvement as measured by the 21-item HAM-D and BDI scales in TRD patients when receiving higher doses [206]. The following partial NMDAR agonist is rapastinel (GLYX-13), which affects the glycine site of this receptor. In male Sprague-Dawley rats subjected to 14 days of adrenocorticotropic hormone (ACTH) injection, it was observed that the administration of rapastinel (10 mg/kg, i.p.) significantly reduced immobilization time in the FST but did not affect the distance traveled in the open field test (OFT) [207]. A clinical trial with intravenous rapastinel in patients with MDD showed that 5 mg/kg and 10 mg/kg doses of GLYX-13 had a significant clinical effect as measured by the 17-item HAM-D scale, which lasted from 2 h to 7 days after administration [208].
Potential antidepressant effectiveness was found for "ampakines", the compounds that potentiate AMPAR function. Intraperitoneal injection of 3 mg/kg/day of S47445 decreased olfactory bulbectomy-induced motor hyperactivity in male C57BL/6J mice. Still, oral intake of 400 mg of another AMPAR positive allosteric modulator (ORG 26576) by MDD patients did not significantly produce clinical antidepressant effects [209,210].
Due to the involvement of mGluRs in the pathogenesis of DDs, they are also targets for new pharmacological therapy. Many of them are focused on negative allosteric modulators of mGluR5. Among them, compounds, such as basimglurant (RG-7090, RO-4917523) and AZD2066, are in clinical trials, but their effects are inconclusive and further studies are required to determine their therapeutic benefits [192,201].

Conclusions
DDs represent a serious and growing health problem among both young and older adults, and if left untreated, they can lead to a suicide death. The appearance of a series of biochemical, electrophysiological, and behavioral studies in the 1990s indicated significant changes in the glutamatergic system in depressed people which, after antidepressant treatment or ECT, contributed to the formulation of the glutamatergic theory of depression and initiated intensive research in this field. It was essential to describe the changes taking place in the Glu synapse, both in depression and after effective antidepressant therapy. The first studies focused mainly on the NMDA and AMPA ionotropic receptors, while subsequent studies considered mGluRs. Despite the large variety of available studies, taking into account the research methodology, brain region, or heterogeneity of the studied groups, it can be concluded that changes in the GluN2A and GluN2B subunits in DDs were most often identified. In the case of the AMPARs, most studies show changes in the GluA1 subunit. Therefore, it seems that these should be the most important Glu therapeutic targets for potential new antidepressant therapies. This is confirmed by clinical studies of these receptors' new ligands/modulators (e.g., traxoprodil). Over the years, the view of excitatory synapse function has changed dramatically, such that the static structure has become highly dynamic, in which postsynaptic density proteins play an important role in both the localization and function of Glu receptors and can modify intracellular signaling. The importance of these proteins, in the development of DDs is also evidenced by numerous studies that indicate a reduced level of PSD-95, CamKII, and Homer 1 proteins. Hence, apart from the receptor mentioned above, these proteins should also become an exciting direction in the searching for more effective drugs to combat DDs.

Funding:
The study was supported by grants from the National Science Centre (contracts UMO-2016/21/B/NZ7/01623 to M. Sowa-Kućma). The funder had no role in the study design, data and literature collection and analysis, decision to publish, or preparation of the manuscript.
Institutional Review Board Statement: Not applicable.