Modeling RNA-Ligand Interactions: The Rev-Binding Element RNA-Aminoglycoside Complex Fabrice Leclerc and Robert Cedergren* De ´ partement de Biochimie, Universite ´ de Montre ´ al, C.P. 6128, Succursale Centre-ville Montre ´ al, Que ´ bec H3C3J7, Canada Received June 6, 1997 X An approach to the modeling of ligand-RNA complexes has been developed by combining three- dimensional structure-activity relationship (3D-SAR) computations with a docking protocol. The ability of 3D-SAR to predict bound conformations of flexible ligands was first assessed by attempting to reconstruct the known, bound conformations of phenyloxazolines complexed with human rhinovirus 14 (HRV14) RNA. Subsequently, the same 3D-SAR analysis was applied to the identification of bound conformations of aminoglycosides which associate with the Rev- binding element (RBE) RNA. Bound conformations were identified by parsing ligand conformational data sets with pharmacophores determined by the 3D-SAR analysis. These “bioactive” structures were docked to the receptor RNA, and optimization of the complex was undertaken by extensive searching of ligand conformational space coupled with molecular dynamics computations. The similarity between the bound conformations of the ligand from the 3D-SAR analysis and those found in the docking protocol suggests that this methodology is valid for the prediction of bound ligand conformations and the modeling of the structure of the ligand-RNA complexes. Introduction Research over the past several years has amply demonstrated the importance of RNA-ligand complexes in cellular processes. Consequently, there is a growing interest in targeting RNA complexes for therapeutics. A case in point is the aminoglycoside inhibition of the association between the Rev protein and the Rev- binding element RNA (RBE) of HIV-1, which determines the fate of viral mRNA. 1,2 Although the conformation of the RBE bound to the Rev peptide has been predicted by modeling 3 and subsequently confirmed by two NMR studies, 4,5 this structure offers little indication as to how an aminoglycoside could bind to the RNA in a way to prevent the interaction with the Rev protein. Modeling and docking a ligand to a receptor is computationally complex in general because of the requirement to find mutually complementary sites in two conformationally flexible molecules. But in the case of RNA complexes, more difficulties arise, because many of the computational tools available for studying the structure of molecular complexes have been developed with proteins in mind. Since NMR and X-ray structures of RNA molecules are becoming more commonplace and a method of predicting bound conformations of RNA from low-resolution chemical data and in vitro selection from random libraries 3,6 has been developed in our laboratory, we have recently focused more on methods which determine the bioactive conformation of the ligand and the structure of the complex rather than the bound RNA conformation. Previously, we showed that structure-activity relationship (SAR) analysis of con- formationally rigid ligands can be used to obtain infor- mation on the three-dimensional structure of the com- plex between DNA and quinolones 7 and that a docking protocol based on electrostatic and van der Waals energies can be applied to modeling of the Rev peptide- RBE complex. 8 Here, we first assess the validity and reliability of the SAR method to identify bioactive conformations for the case of conformationally flexible inhibitors, the phenyl- oxazolines, whose bound structures with human rhi- novirus 14 (HRV14) have been determined by X-ray crystallography. 9 Then, the same approach is applied to the study of the RBE-binding aminoglycosides where the bioactive conformation is unknown. A docking protocol that incorporates the binding properties of the aminoglycosides inferred from the 3D-SAR study is then used to predict the binding conformation of the ami- noglycosides within the RNA binding site. The final model of the complex is supported by the high similarity between the bioactive conformations of the ligands produced by the SAR study and the docking protocol as well as the ability of the model to rationalize available experimental data from the complex. Results and Discussion We have devised a scheme based on a 3D-SAR analysis and a docking protocol to model the bound, bioactive conformation of aminoglycosides and the RBE RNA-aminoglycoside complex (Figure 1). The method involves the prediction of the biologically significant conformations by a 3D-SAR analysis using extensive conformational generation (Figure 1, steps 1-4). The bioactive conformers thus identified are used as starting conformations for a docking and modeling protocol. During docking extensive conformational sampling is used again which, together with molecular mechanics, produces a new series of bioactive conformations (steps 5 and 6). The reliability of models is then evaluated by comparison of the bioactive conformations identified by 3D-SAR and docking (step 7); this step provides a type * Corresponding author. Tel: (514) 343-6320. Fax: (514) 343-2177. E-mail: ceder@poste.umontreal.ca. X Abstract published in Advance ACS Abstracts, December 15, 1997. 175 J. Med. Chem. 1998, 41, 175-182 S0022-2623(97)00372-5 CCC: $15.00 © 1998 American Chemical Society Published on Web 01/15/1998