U-78875 [imidazo[1,5-a]quinoxalin-4(5H)-one, 3-(5-cyclopropyl-1,2,4-oxadiazol-3-yl)-5-(1-methylethyl)] belongs to a series of imidazoquinoxaline derivatives, recently discovered ligands with high affinity for benzodiazepine receptors. In this study, we have examined the drug and its analogs for their modes of interaction with the receptors, with a particular emphasis on finding molecular determinants for their functional properties. Changes in the substituents on N5 and C6 of the heterocyclic ring produced no major effects on binding characteristics but yielded drugs of widely varying efficacy (antagonist to full agonist), measured as gamma-aminobutyric acid (GABA)-mediated 36Cl- uptake and t-butylbicyclophosphoro[35S]thionate binding in rat cerebrocortical membranes. The relative binding affinity and efficacy of the analogs measured in brain membranes were similar to those in cloned GABAA receptors of the alpha 1 beta 2 gamma 2 (type I) and alpha 3 beta 2 gamma 2 (type II) subtypes. The imidazoquinoxalines showed no marked subtype selectivity. Their Ki value against [3H]flunitrazepam binding for type I was only 2-3 times lower than that for type II, and their rank order for agonistic activity was the same in the two subtypes, measured as GABA-mediated Cl- currents in human kidney cells (A293) expressing the subtypes of GABAA receptors. According to computational modeling of the drugs using both molecular and quantum mechanics, the agonistic activity of the imidazoquinoxaline derivatives depends on the presence of a bulky alkyl substituent at N5 and the deformation of the substituted portion of the otherwise planar ring system induced by a bulky moiety at N5 or C6. With a fixed N5 substituent (isopropyl), the relative efficacy in the brain membranes, as well as in the cloned receptors, appeared to be dependent on the degree of the ring deformation. This out-of-plane portion of the imidazoquinoxalines can be assigned to the general region occupied by the 5-phenyl group of diazepam and other agonistic functional groups of several nonbenzodiazepine ligands. It seems that this region, apparently common to various agonistic ligands, interacts with an agonistic pocket in type I and type II subtypes of the benzodiazepine receptors in the brain. Our results also provide direct support for the view that the agonists and nonagonists share largely overlapping binding regions in the benzodiazepine receptor, which has been proposed earlier from in vivo efficacy measurements of other series of ligands.