Thermodynamic Characterization of RNA Duplexes Containing Naturally Occurring 1 × 2 Nucleotide Internal Loops^†

Thermodynamic data for RNA 1 × 2 nucleotide internal loops are lacking. Thermodynamic data that are available for 1 × 2 loops, however, are for loops that rarely occur in nature. In order to identify the most frequently occurring 1 × 2 nucleotide internal loops, a database of 955 RNA secondary structures was compiled and searched. Twenty-four RNA duplexes containing the most common 1 × 2 nucleotide loops were optically melted, and the thermodynamic parameters ΔH°, ΔS°, ΔG°₃₇, and T_M for each duplex were determined. This data set more than doubles the number of 1 × 2 nucleotide loops previously studied. A table of experimental free energy contributions for frequently occurring 1 × 2 nucleotide loops (as opposed to a predictive model) is likely to result in better prediction of RNA secondary structure from sequence. In order to improve free energy calculations for duplexes containing 1 × 2 nucleotide loops that do not have experimental free energy contributions, the data collected here were combined with data from 21 previously studied 1 × 2 loops. Using linear regression, the entire dataset was used to derive nearest neighbor parameters that can be used to predict the thermodynamics of previously unmeasured 1 × 2 nucleotide loops. The ΔG°_37,loop and ΔH°_loop nearest neighbor parameters derived here were compared to values that were published previously for 1 × 2 nucleotide loops but were derived from either a significantly smaller dataset of 1 × 2 nucleotide loops or from internal loops of various sizes [Lu, Z. J., Turner, D. H., and Mathews, D. H. (2006) Nucleic Acids Res. 34, 4912−4924]. Most of these values were found to be within experimental error, suggesting that previous approximations and assumptions associated with the derivation of those nearest neighbor parameters were valid. ΔS°_loop nearest neighbor parameters are also reported for 1 × 2 nucleotide loops. Both the experimental thermodynamics and the nearest neighbor parameters reported here can be used to improve secondary structure prediction from sequence.