Structural Prediction of Coronavirus s2m Kissing Complexes and Extended Duplexes

Abstract

The three-dimensional (3D) atomistic-resolution structure and dynamics of RNA kissing complexes (KCs) and extended duplexes (EDs), homodimers formed through palindromic base pairing, are crucial for understanding viral replication and structure-informed therapeutic design. Polyacrylamide gel electrophoresis (PAGE) evidence suggests KC and ED dimer formation between stem-loop II motif (s2m) elements in SARS-CoV, SARS-CoV-2, and Delta SARS-CoV-2, which may regulate host immune response. However, the absence of 3D structural data on s2m dimers limits structural interpretation needed to explain differences in stability indicated by native PAGE and biophysical implications. In this work, we evaluate the VFold3D/LA-IsRNA pipeline for resolving 3D structures of s2m KCs and EDs by validating its accuracy with blind and referenced predictions against experimental HIV-1 DIS KC and ED structures. Engendering confidence in the approach for blind prediction of KC and ED structures, HIV-1 DIS predictions achieved an average RMSD of 3.28 Å relative to crystal structures, while local interactions, such as palindrome-flanking purine stack orientations in the terminal loops, were in closer agreement with reported solution-phase NMR (RMSD ∼ 2.5 Å), cryo-EM maps, and previous molecular dynamics (MD) simulations. We find that the predicted 3D dimer structures of s2m resulted in kinked or linear shapes of s2m KC complexes that provide an interpretation consistent with native PAGE migration differences, where KCs are more kinked (63° to 133°) than linear ED dimers (127° to 156°). Following MD refinement, the SARS-CoV s2m KC adopts stacking palindromic basepair triplets, whereas SARS-CoV-2 and Delta s2m only form canonical palindrome basepairs, explaining their relative dimer instability suggested by PAGE band intensity. Ultimately, our results support the use of the VFold3D/LA-IsRNA pipeline for KC and ED generation, yielding predictions consistent with experimental data and providing an atomistic foundation for data-driven design of antiviral therapies to disrupt the lifecycle or immune response of viruses.