NMR and molecular dynamics demonstrate the RNA internal loop GAGU is dynamic and adjacent base pairs determine conformational preference

Abstract

The conformational variability of RNA duplexes with the internal loop 5GAGU/3UGAG was investigated by a combination of nuclear magnetic resonance spectroscopy (NMR) and all-atom molecular dynamics (MD) simulations. A previous study showed that the CG-flanked internal loop in the sequence 5GACGAGUGUCA/3ACUGUGAGCAG existed in a major conformation (conformation I) characterized by Uridines U7 and U7* bulging into solution, A5 and A5* stacking, and with G4-G6* and G6-G4* base pairs closing the loop on either end. It was also determined that a minor conformation existed with a set of closing G-U pairs with a bifurcated hydrogen bond and A-G non-canonical pairs with a single hydrogen bond (conformation II). A maximum hydrogen bonding structure with wobble G-U pairs and A-G pairs with two hydrogen bonds (conformation III) was not observed by NMR or predicted by molecular dynamics. In this work, alternative flanking base pairs A-U, U-A and G-C were studied by substituting the C-G base pair adjacent to the GAGU internal loop. NMR spectra demonstrated changes in conformational preference depending on the identity of the flanking pair. Clustering analyses of structures from MD simulations of C-G- and U-A-flanked duplexes showed transitions between minor conformations II and III with a greater fraction of the structures in conformation II, in spite of the simulations starting in conformation III. In addition, U-A-flanked simulations contained a substantial amount (25%) of structures in an intermediate state between conformations II and III. A-U- and G-C-flanked structures were all in a single cluster whose centroid structure was in conformation II. MD simulations showed a dominance of structure II over structure III, in agreement with NMR data for C-G closure, but in contrast with the NMR data for other closures. Simulations starting in conformation I did not transition to either of II and III, with all structures being in a single cluster for all flanking base pairs except for G-C-closure where 22% of the structures were in a state defined by a C3endo sugar pucker for G4 and G4*. Bases G4 and G4* were also in syn orientation around the glycosidic bond for all flanking pairs while G6 and G6* transitioned between syn and anti orientations for C-G- and A-U-flanked simulations, consistent with NMR observations.