The main results of studies of these sites and their ligands obtained in recent years will be considered below.

3.1. The Orthosteric Binding Site for Agonists and Competitive Antagonists

As explained above, the orthosteric site is located inside the “clamshell” of the ligand-binding domain and provides the binding of full and partial agonists as well as competitive antagonists of the receptor. Although extensive research on its ligands has already been performed in previous years, it still attracts attention as a target for the development of potential drugs and/or pharmacological tools.

It should be noted that AMPA receptor agonists can enhance desensitization and often cause a number of side effects, including seizures and neurotoxicity [55]. In this regard, the development of competitive receptor antagonists is more promising for clinical use.The development of the quinoxaline-2,3-dione scaffold, a classic one for competitive AMPA receptor antagonists, has made it possible to obtain a large series of analogs with different selectivity profiles for ionotropic glutamate receptors, including a selective AMPA receptor antagonist 5 [56]. A detailed analysis of the binding of quinoxaline-2,3-diones showed that these compounds (for example, dinitro-derivative DNQX 6) are easily deprotonated at pH close to physiological and interact with the receptor specifically in the anionic form [57]. By modifying this scaffold, reversible photoswitchable antagonists such as ShuBQX-3 (7) were developed (only the trans-form has antagonistic activity; it is formed upon irradiation by light with a wavelength of 600 nm and turns into an inactive cis-form at 400–500 nm, making it possible to accurately and non-invasively control the receptor operation in space and time) [58]. Based on the combination of quinoxaline-2,3-dione and kynurenic acid scaffolds, a series of hybrid quinazolinediones was constructed, including the potential antiepileptic drug selurampanel (8) [59].Recently, new chemotypes of AMPA receptor agonists and competitive antagonists were being actively developed. For example, conformationally restricted bicyclic analogs of glutamic acid CIP-AS (9) and LM-12b (10) [60] act as partial or full agonists of the AMPA and kainate receptors, with a strong preference for the kainate GluK3 subunits. Their selectivity profiles and binding modes were analyzed using X-ray data. Based on the scaffold of aryl- and hetaryl-substituted phenylalanines, a series of competitive AMPA receptor antagonists such as compound 11 was obtained, their possible binding modes were analyzed using molecular docking, and their anticonvulsant and antioxidant properties were confirmed [61,62,63].By modifying the structure of the natural compound (S)-willardiine 12, which is a partial agonist of the AMPA receptor, a series of bicyclic derivatives of pyrimidinedione 13 was created [64,65]. It has been demonstrated that the replacement of the heterocyclic fragment leads to a significant (up to 500-fold) change in affinity. Molecular modeling and analysis of structural data obtained by X-ray crystallography have shown that this effect is caused by a significant difference in interactions with the protein and, especially, the binding site water molecules.In the development of a series of glutamate and aspartate analogs 14–16, the 4-hydroxy-1,2,3-triazole fragment was used as a bioisostere for the distal carboxylic group [66]. Compounds 14a–c and 15a were shown to be selective AMPA receptor agonists; moreover, compound 14b exhibited selectivity to certain types of AMPA receptor subunits. 3.2. The Positive Allosteric Modulator (PAM) Binding SitePositive allosteric modulators (PAMs) of the AMPA receptor (ampakines) bind at an allosteric site located at the interface between the subunits of the dimeric ligand-binding domain [7]. It is believed that their potentiating effect on the receptor is caused by the stabilization of its open form and/or the decrease in desensitization. Such modulators attract much attention due to their ability to improve learning and memory formation processes as well as their neuroprotective effect, which make them promising candidates for the development of drugs for the treatment of cognitive disorders (including the early stages of the Alzheimer’s disease), depression, and a number of other CNS pathologies [7,8,13] as well as the drug-induced respiratory depression [67]. The key advantage of the positive allosteric modulators of AMPA receptors is that, unlike agonists, they only potentiate the receptor in the presence of an endogenous ligand and therefore lead to fewer side effects [55,68]. However, in spite of the encouraging preclinical results, the majority of PAMs to date have not progressed beyond early clinical development [69]. Some of the problems in this field involve the insufficiently relevant in vitro and in vivo models of neurological and neuropsychiatric disorders, the pharmacodynamics and pharmacokinetics issues, and limited understanding of the processes underlying the activity and specificity of modulator action. For instance, in spite of significant progress in the development of the cell models [70] and animal models [71,72,73,74] for neurodegenerative diseases such as Alzheimer’s disease, amyotrophic lateral sclerosis, Parkinson’s disease, Huntington’s disease, as well as schizophrenia, no one of them is yet able to fully capture all the critical aspects and factors involved in these complex pathologies—especially as those factors are not yet fully understood. Nevertheless, the AMPA receptors remain an attractive target that has several unique advantages over other drug targets, and it is believed that the problems can be solved. Thus, identifying novel PAM chemotypes has become a priority for many pharmaceutical companies, and AMPAR PAMs offer many promising treatment avenues for neurological and neuropsychiatric disorders [69].From the structural perspective, reflecting the overall symmetry of the dimeric AMPAR ligand-binding domain, the positive allosteric modulator binding site has a symmetrical structure, including the central and two side pockets, and molecules of quite different sizes and chemical structures—from small nootropic compounds (for example, aniracetam 17) to rather large and complex ones—can bind in its various regions (Figure 5) [7]. Perhaps this is one of the reasons that hinder the development of general principles of PAM molecular design, despite the structure–activity relationships found in individual series of compounds, the vast available experimental data on the structure of the modulator–receptor complexes (recently analyzed in deep detail [75]), and the success of molecular modeling using pharmacophore analysis, 2D and 3D QSAR studies, molecular docking, and molecular dynamics [2,76,77,78,79].

The classical chemotypes of the positive allosteric modulators of AMPA receptors are benzamides, benzothiadiazines, biarylalkylsulfonamides, and trifluoromethylpyrazoles.

Among the first AMPAR positive allosteric modulators were benzamides based on aniracetam 17. The most promising molecules of this chemotype were CX516 (18), CX614 (21), CX691 (22), and CX717 (23) [80,81] (the only difference between the two latter structures is the replacement of the piperidine fragment in CX691 by morpholine). Later, the more distant racetam analogs were obtained such as S47445 (CX1632, tulrampator) 24 [82] and 25 [83]. The analysis of the AMPAR LBD complexes with aniracetam, CX516 (18), Me-CX516 (19), and CX614 (21) [80] has shown that all these molecules bind in the central subpocket in a very similar way, with a key role played by the hydrogen bonds with the network of water molecules that mediate the interaction of a ligand with amino acid residues in the binding site (Figure 6A).To date, the most significant therapeutic effects of CX516 have been shown in the animal models of intellectual disability (at 5 mg/kg dose for 5 days) [84] and hyperactivity commonly observed in schizophrenia and autism spectrum disorder (at 10–40 mg/kg) [85]. In recent studies, CX546 has shown reasonably good results in the autism models and was found to effectively mediate neurogenesis and dendritogenesis [86,87]. However, its low oral bioavailability limits the clinical development of this agent. In its turn, during preclinical studies, CX614 has shown itself as a promising brain-derived neurotrophic factor (BDNF) inductor [88] and enhanced the effects of antidepressants such as imipramine and reboxetine.CX691 was studied for its procognitive effects in an animal model of Alzheimer’s disease, and increased hippocampal BDNF expression and improved spatial learning and memory were found [89]. For CX717, it has been established that it acts as a PAM and has a strong antidepressant effect [90]; in addition, CX717 is considered as a potential agent for the treatment of attention deficit hyperactivity disorder (ADHD).In recent preclinical studies, it was found that S47445 (CX1632, tulrampator) significantly enhances synaptic plasticity and increases neurotrophin levels both in the hippocampus and the prefrontal cortex of aged mice [82,91]. Interestingly, the antidepressant and anxiolytic effects were also detected for this ligand in three animal models. These behavioral effects were accompanied by increased levels of hippocampal neurogenesis and BDNF [92,93]. S47445 has also shown procognitive effects in animal models [94,95,96]. Based on these results, Phase I clinical studies have been started for S47445 as a potential agent for the treatment of Alzheimer’s disease and dementia-associated depression (NCT02626572, NCT02805439). However, the results of the first double-blind placebo-controlled clinical study in patients with mild to moderate Alzheimer’s disease and depressive symptoms suggested that S47445, although well tolerated, did not show significant improvement in cognitive functions [97].For compound 25, extensive in vitro and in vivo preclinical studies have been conducted, confirming that it acts as a neuroprotective agent and can significantly reduce neurological deficits and restore cognitive functions after ischemic brain injury [83].Although the derivatives and analogs of benzothiadiazine dioxide, starting with cyclothiazide (CTZ 26), represent one of the most thoroughly researched classes of the AMPAR positive allosteric modulators, their studies continue to attract considerable interest. During the search for new chemotypes of modulators and the investigation of their structure–activity relationships, a number of highly potent structures were found, for example, 27 [98], 28a and its unsaturated metabolite with comparable activity 28b [99]. In some cases, the “unexpected” structure–activity relationships discovered during the study could be explained using thermodynamic analysis, experimental data on the binding modes, and molecular modeling [100,101]. Unlike most of the studied benzothiadiazine dioxides, which admit independent binding of two modulator molecules in different regions of the binding site at the LBD dimer interface, the larger molecules of the phenoxy derivative 29 [102] and the specially designed dimeric compounds 30 [76] occupy several regions of the site by binding of one molecule (Figure 6B) and exhibit activity in the nanomolar range.Recently, an attempt was made to develop new positive allosteric modulators by switching from the benzothiadiazine dioxide scaffold to the thiochromane dioxide scaffold using a classical bioisosteric replacement [103]. However, the resulting series of compounds 31 exhibits lower activity toward the AMPA receptor compared to the original benzothiadiazine dioxides. The first biarylalkylsulfonamide modulator, compound 32, was found by high-throughput screening. Based on its structure, a number of analogs were created and detailed SAR studies were carried out. One of the most promising compounds exhibiting high activity and selectivity was LY-404187 (33, racemate) and its optically pure (R)-isomer LY-451646. In the earlier publications, its therapeutic effect was explained primarily by its antidepressant action clearly observed in the forced swim and the tail suspension tests [104,105]. Further studies were aimed at developing dimeric structures (LY-451395 34 and compounds 35) wherein, according to the X-ray diffraction data, sulfonamide groups bind to the symmetrical side pockets and the biaryl fragment occupies the central pocket [81] (Figure 7A). Interestingly, the replacement of the biaryl fragment by alkyl chains of different lengths does not lead to a decrease in activity, which makes it possible to vary the structure of the linker over a wide range. Compound LY-451395 (mibampator 34) was under phase II clinical trials for the treatment of Alzheimer’s disease, but the trials were terminated [106]. An alternative approach to the optimization of biarylalkylsulfonamides involves the creation of their conformationally restricted analogs (for example, compounds 36, 37, and BIIB-104 38). Among them, the compounds with an indane fragment are rather interesting; for one of those (37), detailed pharmacological studies have been carried out [107]. Compound BIIB-104 (PF-04958242, pesampator 38) is currently in development to target schizophrenia-associated cognitive impairment [3,108]. To date, however, no clinical trials have been conducted with these compounds.Trifluoromethylpyrazoles are the newest of the classical chemotypes of the AMPA receptor PAMs. Their binding in the central subpocket of the receptor is based primarily on hydrophobic interactions, with the trifluoromethyl group playing a key role (Figure 7B). The main development strategy was a structure-based design starting from compounds found by high-throughput screening. Based on the structural data, modifications were carried out that made it possible to obtain compounds 39–42 with more promising pharmacokinetic characteristics [11,109,110,111,112]. Moreover, a series of hybrid compounds containing an indane sulfonamide fragment (for example, 43 and 44) and, in some cases, replacing pyrazole moiety with furan, was also developed [11,113]. Currently, new chemotypes of the positive allosteric modulators of AMPA receptors are being actively developed. In this respect, great attention is paid to the reduction in side effects, since an expansion of the therapeutic window is required for the clinical use of the compounds. The risk of seizures and other side effects was found to be associated with agonistic effect at high concentrations and a bell-shaped concentration–response relationship [55,68].Recently, the compound HBT1 (45), a trifluoromethylpyrazole modulator, has been found to exhibit lower agonistic effects than other modulators and a normal (sigmoid) rather than a bell-shaped concentration–response relationship [114]. In addition, HBT1 exhibits specific receptor binding, including interaction with the S518 residue. Based on these structural data, promising derivatives of dihydropyridothiadiazine-2,2-dioxide, TAK-137 (46) and TAK-653 (47), were developed. Their agonistic effect is low, presumably due to the interaction with Ser743 in GluA1 [55,115,116,117]. It is important to note that the HBT1 compound in preclinical studies showed pronounced neurotrophic properties and increased BDNF levels in primary neurons [114]. However, at the moment, the exact mechanism of action of these agents remains unknown. More extensive preclinical studies have been conducted for TAK-137. It has been shown that this compound can be considered as a drug for the treatment of cognitive impairment in schizophrenia and as a potential new antidepressant [115,118]. In several animal models of schizophrenia, TAK-137 has shown significant improvements in social interaction, working memory, and other cognitive functions [119]. Another recent preclinical study compared the antidepressant properties of TAK-137 and ketamine. It was found that in rats treated with TAK-137 for three days, the same improvements were observed as in the treatment with ketamine, but no psychotomimetic side effects were found [116]. Finally, in preliminary Phase 1 studies with both healthy volunteers and ADHD patients, TAK-137 was found to be safe and well tolerated.A number of interesting structures of AMPA receptor modulators, such as a series of diacyl bispidine derivatives 48, have been recently developed [120,121,122]. For some of them (48a, 48b, 48c, and 48e), the potentiation of the AMPA receptor currents (e.g., for 48a in the concentration range of 0.01 to 10 nM, with the maximum increase by 110% achieved at 1 nM) and antiamnesic properties in vivo were found [122]; their specific binding to neurons was also studied [123]. According to the updated molecular docking data for these compounds, it is suggested that two modulator molecules bind symmetrically at the interface of the LBD dimer, similar to the binding mode of cyclothiazide (Figure 8) [122].Recently, a new promising series of bis-isoxazoles was obtained and compound 49 demonstrated very high positive modulator activity. The potentiation of the AMPA receptor currents was observed in a wide concentration range (10−12−10−6 M) and had a bell-shaped concentration dependence with maximum potentiation (up to 72%) at 10−11 M [124]. The molecular docking and molecular dynamics studies confirmed that the compound could indeed act as a positive AMPA receptor modulator binding in the validated PAM binding site (Figure 9).In addition, positive modulatory activity has recently been revealed for a number of well-known compounds that were not previously considered as AMPA receptor ligands. Among the nootropic compounds, we should mention the endogenous peptide cycloprolylglycine 50 and its analogs, for which a potentiating effect on the AMPA receptor (as well as interactions with a number of other targets) has been confirmed [125,126]. Ketamine 51 has been used for several decades as an intravenous anesthetic, and until recently it was considered only as an NMDA receptor blocker. However, in 2000, ketamine was found to exhibit antidepressant activity at subanesthetic concentrations, and in 2019 it was approved by the FDA as a rapid-action antidepressant with several advantages over the previously used drugs. The detailed mechanism of the antidepressant effect of ketamine, its metabolites, and analogs is not fully understood yet, but it has been shown that an important role belongs to the (direct or indirect) potentiation of the AMPA receptors, release or induction of BDNF, and the increase in molecular neuroplasticity [127,128,129,130,131]. Antidepressant effects have also been found for a number of other classes of AMPAR positive allosteric modulators [90,115,132]. 3.3. Binding Site for Negative Allosteric Modulators (Non-competitive Antagonists, Non-competitive Inhibitors)

Many of the studied negative allosteric modulators (non-competitive antagonists, non-competitive inhibitors) of the AMPA receptor bind in the linker region between the ligand-binding and transmembrane domains, preventing the displacement of the M3 helix and hindering the ion channel opening. Although for a number of antagonists with a confirmed non-competitive nature of action there is still no information on the binding site or the action mechanism, they will also be considered in this section.

A preeminent example of negative allosteric modulators (noncompetitive antagonists) of the AMPA receptor is perampanel 52 (PMP), approved by the FDA in 2012 as an antiepileptic drug [133]. Since then, research and clinical experience indicate that it is an effective antiepileptic agent with a broad spectrum and a novel mechanism of action [134,135], and may also be useful in alleviating functional and cognitive impairment after stroke [136,137]. In addition, its antitumor activity against brain tumors [138,139] and the possible effectiveness of negative allosteric modulators with a similar mechanism of action against tumors of other organs [140] have been demonstrated.Classical chemotypes of negative allosteric modulators of the AMPA receptor are 2,3-benzodiazepines and quinozalin-4-one derivatives. The binding mode and the mechanism of action of such modulators, using CP465022 (53), GYKI53655 (54), and perampanel as an example, were analyzed in detail by means of the X-ray diffraction analysis, site-directed mutagenesis, molecular and quantum mechanics, molecular docking, and molecular dynamics [141,142,143,144]. All these modulators bind in the interface region of the transmembrane domain (Figure 10). The large size and flexibility of the binding site “pocket”, which is able to adapt to different structures and orientations of the ligands, leads to the binding of modulators with significantly different structures in the same site. Modulator molecules play the role of “wedges” between the transmembrane segments, preventing the M3 helix from shifting when the channel opens. The binding occurs due to multiple weak interactions including hydrophobic contacts, π-stacking, and hydrogen bonds. At the same time, the interaction with certain amino acid residues depends on the state of the receptor, and the significance of some interactions remains debatable.Despite previous extensive studies of the 2,3-benzodiazepine scaffold, it continues to attract some interest in the search for new negative AMPA receptor modulators. Particularly, in recent years, its isoxazoline derivatives such as 55 [145] and m-chlorophenyl analog 56 (the o-chloro derivative is inactive) [146] have been obtained.In addition, new chemotypes of negative allosteric modulators of the AMPA receptor are currently being actively developed. The most interesting among them are the derivatives of phthalazine-1,4-dione (e.g., 57) [147], pyridothiazinone (e.g., 58) [148], benzodioxole (e.g., 59) [149], and 5-chloro-2-oxo-3H-benzoxazole (e.g., 60) [150]. A non-competitive inhibitory effect and influence on the kinetics of desensitization and deactivation for various subtypes of the AMPA receptor were also found for a number of curcumin derivatives 61 and 62 [151,152,153]. Several promising non-competitive AMPAR antagonists 63 were obtained in the study of arctigenin analogs, with molecular docking results suggesting a binding mode in the transmembrane domain similar to that of the known AMPA receptor non-competitive antagonists such as perampanel [154]. It is interesting to note that the RNA aptamers found by systematic evolutionary selection as well as their chemically modified analogs can also act as noncompetitive inhibitors of AMPA and kainate receptors with different selectivity profiles [155,156,157].Recently, studies have been published on the simultaneous binding of competitive antagonists and negative allosteric modulators to the AMPA receptor using ZK200775 (64) and GYKI53655 (54) [50] as an example. It was shown that the interaction of the ligand with the binding site is not affected by the presence of another ligand, and the respective conformational changes occur independently. 3.4. Chemotypes with Activity Cliffs

For some new scaffolds of the allosteric AMPA receptor modulators, the activity cliffs are observed, that is, significant changes in the magnitude and even nature of activity occur with minor changes in the structure of the compound. Possible reasons for this phenomenon are currently under investigation.

For instance, among bis(tetrahydroquinazoline) derivatives [158,159,160], compounds 65a, 65b, 65d, 65e, 65f, 65h, and 65i are positive modulators of the AMPA receptor in a broad range of concentrations (maximum potentiation by 70%, 55%, 66%, 53%, 77%, 51%, and 61% at 1 nM, respectively), while compounds 65c, 65g, and 66 are negative modulators (maximum current reduction by 30% at 0.1 nM; by 50% at 1 μM or 30% at 0.1 nM; EC50 = 14 µM, respectively). In the bis(amide) series, compound 67a is a positive modulator in a broad range of concentrations (maximum potentiation by 40% at 1 nM), while compounds 67b–67d are negative modulators (maximum current decrease at 1 nM by 20%, 50%, and 40%, respectively) [161,162]. Interestingly, in the case of tricyclic scaffold derivatives, an inversion of the modulatory activity can be achieved either by slight structural modifications of the substituent at the N atom or by the presence/absence of a carbonyl group in the bispidine skeleton. Thus, compound 68a is inactive, compounds 68b, 68e, and 68g are positive modulators of the AMPA receptor in a wide concentration range (maximum potentiation by 30% at 1 nM, 40% at 0.001 nM, and 65% at 1 nM, respectively), and compounds 68c, 68d, and 68f are negative modulators (maximum current decrease by 20% at 0.1 nM, 35% at 0.01 nM, and 35% at 1 nM, respectively) [163,164,165,166]. To date, compounds 68e and 68f have undergone a full cycle of preclinical trials, the results of which are being prepared for publication. Compound 69a is a highly potent positive AMPA receptor modulator (maximum potentiation of 62% at 1 nM), while compound 69b is a negative modulator (maximum current reduction of 39% at 0.1 μM) [167]. Based on the molecular docking and molecular dynamics results, compound 69a can indeed interact with the validated PAM binding site at the interface between the dimeric LBDs (Figure 11) while the binding stability of compound 69b is much lower and it should have substantially weaker PAM activity. On the other hand, it is suggested that compound 69b can be tentatively expected to act via the NAM binding site at the LBD–TMD interface.