Recent progress has been made in translating animal findings

Recent progress has been made in translating animal findings on memory performance and amnestic effects – that are largely determined by tonically active receptors in which α5 contributes to the benzodiazepine binding site. Based on the observation that deletion of this subunit led to improved spatial learning performance [56], it was hypothesized that amnestic benzodiazepine effects (also seen in sedative anesthetics) can be attributed to excessive activity of this receptor population, and that nootropic effects can be elicited by α5-selective negative allosteric modulators [57]. This notion was confirmed in animal studies using several experimental compounds 58, 59, 60. This line of research produced the so far furthest developed α5-selective negative modulator basmisanil (RG1662). The SC 560 receptor was tested in a clinical trial (https://clinicaltrials.gov/ct2/show/NCT02024789) in a Down syndrome cohort for its ability to alleviate cognitive disabilities. No benefit was noted, but the compound may improve cognition in healthy subjects. Cognitive deficits also occur as negative symptoms in schizophrenic patients, and benzodiazepines have a long history as candidates for this indication as well. A small-scale trial with bretazenil was terminated owing to excessive sedative effects despite promising efficacy [61]. α2 and α5 subunit-containing receptors were traditionally considered as the most promising targets to address schizophrenia symptoms 62, 63. At time of writing, basmisanil is being tested in a cohort of schizophrenic patients as an add-on for antipsychotic therapy to test whether negative symptoms can be reduced (https://clinicaltrials.gov/ct2/show/NCT02953639). Overall, subtype selective targeting has been developed with success in rodent models. Proof of concept in humans is also well established for some receptor species – such as α5 subunit-containing receptors as targets for nootropic or amnestic effects. Translation of animal findings to human remains challenging, and trial failures may result from substantial sequence differences in several receptor subunits that impact on drug effects, as well as from differences in the expression patterns of individual subunits in distinct anatomical and cellular environments [64]. Moreover, individual receptor subtype composition and arrangement may also not be identical in different species for all receptor subtypes.
Novel Ligands for the High-Affinity Benzodiazepine Binding Sites The development of benzodiazepines was traditionally based on large ligand series and experimental exploration of (quantitative) structure–activity relationships. There still is a paucity of selective compounds, and this nowadays could potentially be addressed by structure-guided ligand design. Box 4 reviews principles of structure-guided drug design. Computational models based on available homologous proteins have been in overall agreement regarding the relative positions of individual protein segments in the binding site [17]. Based on early mutational work, these segments were originally termed ‘loops’ A–G, with some being loops in the structural sense while others are short pieces of β strands 27, 31. The use of such homology models, in combination with a chemical biology approach to identify residues in contact with diazepam, has allowed relative positioning of diazepam into homology structures of an α1γ2β2α1β2 GABAA receptor. Virtual ligand screening into this structure has led to the discovery of novel ligands with high affinity for the canonical benzodiazepine binding site [38], thus validating the overall correctness of the structural models. Owing to the high conservation of loops A–G in individual subunits, such as α2 and α3, or γ2 and γ3, the overall architecture of the individual sites is similar in different receptor isoforms. Only a few amino acids in the binding site confer unique properties to the pocket to enable ligand selectivity (Figure 1C,D). Owing to methodological limitations, models of highly homologous proteins (such as α2 and α3) based on a more distant family member (such as β3) will not reliably predict the subtle differences between individual isoforms of the high-affinity benzodiazepine binding sites.