In Vivo Production of Novel RNA Molecules and Computational Modeling of Their 3D Structures

Abstract ID 93311 Poster Board 150 Developing therapeutic RNAs and RNA-targeted small molecule drugs involves producing new RNA agents and understanding RNA structures. In this study, we aimed to employ our tRNA/pre-miRNA-based technology to produce two novel recombinant RNA (BioRNA) molecules and investigate their 3D structures through computational modeling. BioRNA expression plasmids were constructed through molecular cloning and verified via DNA sequencing. Following transformation and fermentation production, target BioRNAs were purified from total bacterial RNAs using an anion exchange FPLC method. High Purity (> 99%) was confirmed through urea-PAGE and HPLC analyses, with high yields (9.91 and 4.56 mg per 250 mL culture) and minimal endotoxin levels (≤ 0.13 EU/μg RNA). Further, we employed Rosetta FARFAR2 and RNAComposer to predict the 3D structures of BioRNAs and assess potential differences as indicated by the root mean square deviation (RMSD). Our results showed that RNAComposer excelled Rosetta in modeling the known malachite green aptamer structure (RMSD = 2.558Å) while the latter demonstrated refinement with the increase of structure models (RMSD = 9.702Å to 6.895Å from 600 to 2000 models). On the other hand, Rosetta (RMSD = 12.125Å) outperformed RNAComposer (RMSD = 22.944Å) in modeling the structure of human selenocysteine tRNA. Likewise, the 3D structures of novel BioRNAs predicted by Rosetta and RNACompoer were obviously different. As both programs recapitulated the empirical cloverleaf structure of the human leucine tRNA segment, the hsa-pre-miR-34a components within BioRNAs predicted by Rosetta were more stretched out while more curved by RNAComposer. Interestingly, the deviations between Rosetta and RNAComposer predictions increased with the size of RNAs (e.g., hsa-pre-miR-34a, RMSD=15.21Å; entire BioRNA, 21.231Å). These distinctions are likely attributed to the fact that RNAComposer uses larger fragments and allows efficient recognition of homologous elements or even the whole known structures, offering faster processing. By contrast, Rosetta relies on the Monte Carlo process, involving repeated random sampling to simulate near-native structures, which takes a rather much longer time. These predicted 3D structures of novel BioRNAs are amenable to experimental validation.