Chemical Shift Prediction Of RNAs And The Application Toward Characterizing RNAs Excited States

Chemical shift is the most valuable observable in NMR spectroscopy.It’s easy to measure and can be measured with great accuracy.In some senarios,chemical shifts are the only available NMR data.Chemical shift is extremely sensitive to even subtle change in molecular conformations and can be used to extract rich structural and dynamic information.In consequence,chemical shift prediction is a fundamental question in NMR study.An accurate predictor can greatly facilitate and validate chemical shift assignment that is the prerequisite to almost all biomolecular NMR studies.More importantly,it will make chemical shift a highly valueable restraint for structure calculation of biomolecules.Methods for chemical shift prediction have been well developed for proteins over the past four decades while such methodology development for RNAs nevertheless lags markedly behind that for proteins.There are currently no reliable predictors for imino group of RNAs that is a widely used probe of NMR,and the prediction for CH chemical shifts in non-canonical motif is also poor.This is mainly due to the lack of sufficient chemical shift data as well as an effective prediction model.To address these issues,we prepared more than 90 RNA samples containing canonical or non-canonical motifs,collected both NH and CH spectra for each sample,and obtained the chemical shift data.Combined with data in biological magnetic resonance bank（BMRB）,we developed several tools for RNA chemical shift prediction and achieved high accuracy.For imino chemical shift prediction,we found that the base pair triplet（BP-triplet）motif dictates imino chemical shifts in the central base pair.We collected ¹⁵N and ¹H^N chemical shift data for RNA base pairs with different sequence context and compiled them as a lookup table that links each different BP-triplet to experimental chemical shifts of its central base pair.This lookup table can be used to predict imino chemical shift of RNAs to unprecedented accuracy.We also obtained an accurate chemical shift predictor for CH group in canonical motif using the same strategy.Strikingly,we found that the chemical shift variations of the same central base pair in varied BP-triplets can be well interpreted by ring-current contributions caused by aromatic rings of the two flanking base pairs.This finding established that the semi-empirical model could predict NH and CH chemical shifts of more complicated motifs in RNAs.Based on this,we subsequently calculated chemical shifts of several more complicated motifs and found they correlated very well with experimental data.Along a separate line of strategy,we used machine learning method to predict the chemical shifts of more complicated RNA motifs,and also achieved better accuracy than previous studies.Finally,we combined the imino chemical shift prediction with NMR relaxation dispersion（RD）experiments targeting both ¹⁵N and ¹H^N of imino group,and developed a new tool to rapidly and accurately characterize secondary structure of RNA excited state（ES）.By combining ¹⁵N RD measurement and our predictor for imno chemical shifts,we applied a recently developed ¹H^N CEST experiment to ES of P5abc that has been well characterized previously,and successfully verified its transient state including an earlier speculated non-native G·G base pair.Furthermore,we applied this new method to NRS23,an intronic RNA element of Rous sarcoma virus RNA,and characterized its transient secondary structure for the first time.