Pharmacophore/receptor models for three recombinant GABA(A)/BzR subtypes (alpha1beta3gamma2, alpha5beta3gamma2, and alpha6beta3gamma2) have been established via an SAR ligand-mapping approach. This study was based on the affinities of 151 BzR ligands at five distinct (alpha1-3,5,6beta3gamma2) recombinant GABA(A)/BzR receptor subtypes from at least nine different structural families. Examination of the included volumes of the alpha1-, alpha5-, and alpha6-containing subtypes indicated that region L(2) for the alpha5-containing subtype appeared to be larger in size than the analogous region of the other receptor subtypes. Region L(Di), in contrast, appeared to be larger in the alpha1 subtype than in the other two subtypes. Moreover, region L(3) in the alpha6 subtype is either very small or nonexistent in this diazepam-insensitive subtype (see Figure 16 for details) as compared to the other subtypes. Use of the pharmacophore/receptor models for these subtypes has resulted in the design of novel BzR ligands (see 27) selective for the alpha5beta3gamma2 receptor subtype. alpha5-Selective ligand 27 when injected directly into the hippocampus did enhance memory in one paradigm (Bailey et al., unpublished observations); however, systemic administration of either 9 or 27 into animals did not provide an observable enhancement. This result is in complete agreement with the observation of Liu (1996). It has been shown (Liu, 1996; Wisden et al., 1992) that in the central nervous system of the rat (as well as monkeys and pigeons) there are several native subtypes of the GABA(A) receptor which exhibit different functions, regional distributions, and neuronal locations. Although 27 binds more potently at alpha5beta3gamma2 receptor subtypes and is clearly an inverse agonist (Liu et al., 1996; Liu, 1996), it is possible that this ligand acts as an agonist at one or more subtypes. Liu (1996) clearly showed that a number of imidazobenzodiazepines were negative modulators at one subtype and agonists at another. Therefore, selectivity for a particular subtype at this point is not sufficient to rule out some physiological effect at other GABA(A)/BzR subtypes. The inability of 27 to potentiate memory when given systemically is again in support of this hypothesis, especially since alpha1beta2gamma2 subtypes are distributed throughout the brain (Wisden et al., 1992). A drug delivered systemically is far more likely to interact with all subtypes than one delivered to a specific brain region. This observation (systemic vs intrahippocampal) provides further support for the design of more subtype-specific ligands at the BzR to accurately define their pharmacology, one key to the design of new drugs with fewer side effects.

The synthesis and pharmacological properties of imidazobenzodiazepines with both high affinity and selectivity for alpha 5-containing GABAA receptors are described. Four of these compounds (5, 6, 8, and 9) inhibited [3H]flunitrazepam binding to recombinant alpha 5 beta 2 gamma 2 GABAA receptors with IC50 values between approximately 0.4 and 5 nM. These compounds were > or = 24-75-fold more selective for recombinant receptors containing alpha 5 subunits compared to other, "diazepam-sensitive" GABAA receptors containing either alpha 1, alpha 2, or alpha 3 subunits. Imidazobenzodiazepine 9 (used as the prototypical alpha 5 selective ligand) inhibited [3H]flunitrazepam binding to hippocampal membranes with high- and low-affinity components (IC50 0.6 +/- 0.2 and 85.6 +/- 13.1 nM, respectively), representing approximately 16% and approximately 84% of the receptor pool. Inhibition of [3H]flunitrazepam binding to cerebellar membranes with imidazobenzodiazepine 9 was best fitted to a single population of sites with an IC50 of 79.8 +/- 18.3 nM. These imidazobenzodiazepines behaved as GABA negative ligands in recombinant GABAA receptors expressed in Xenopus oocytes and were convulsant in mice after parenteral administration. The relative potencies of flumazenil and zolpidem in blocking convulsions induced by 9 and DMCM, respectively, indicated that occupation of alpha 5-containing GABAA receptors substantially contributed to the convulsant properties of acetylene analog 9. These 8-substituted imidazobenzodiazepines (5, 6, 8 and 9) should prove useful in examining the physiological roles of GABAA receptors bearing an alpha 5 subunit and may also lead to the development of novel, subtype selective therapeutic agents.

Affinities of a series of substituted imidazobenzodiazepines at recombinant alpha1beta3gamma2, alpha2beta3gamma2, alpha3beta3gamma2, alpha5beta3gamma2, and alpha6beta3gamma2 GABAA/benzodiazepine receptor subtypes are reported. Many of these ligands displayed high affinities (low-nanomolar to subnanomolar scale) at all five receptor subtypes. Furthermore, a number of imidazobenzodiazepines exhibited relatively good selectivity at the alpha5-containing receptor isoform. For example, ligand 27 (RY-023) demonstrated a 55-fold higher selectivity at alpha5beta3gamma2 isoforms in comparison to other receptor subtypes. The affinity ratio of alpha1 (the most prevalent subtype in the brain) to alpha5 of this series of ligands ranged from 60- to 75-fold for the most selective ligands. Studies of quantitative structure-activity relationships (QSAR) by means of comparative molecular field analysis (CoMFA) were carried out. As a result, examination of CoMFA models for all five receptor subtypes demonstrated their predictability for affinities of imidazobenzodiazepines at the five receptor subtypes. Regions of molecular fields which would favor or disfavor the binding affinity of a ligand at a specific receptor subtype were examined via CoMFA for alpha1, alpha2, alpha3, alpha5, and alpha6 subtypes. A CoMFA regression analysis was applied to predict the ratio of Ki alpha1/Ki alpha5, an index for the selectivity of a ligand at the alpha5 subtype. All of the CoMFA models offered good cross-validated correlations for the ligands in the test set as well as the ratios of Ki alpha1/Ki alpha5, which demonstrated their potential for prediction.