Abstract
In RNA secondary structure prediction, nearest-neighbor parameters are used to determine the stability of a given structure. We derived the nearest-neighbor parameters for RNAs containing inosine-cytosine pairs. For parameter derivation, we developed a method that combines UV adsorption measurement experiments with free-energy calculations using molecular dynamics simulations. The method provides fast drop-in parameters for modified bases. Derived parameters were compared and found to be consistent with existing parameters for canonical RNAs. A duplex with an internal inosine-cytosine pair is 0.9 kcal/mol more unstable than the same duplex with an internal guanine-cytosine pair, and is as stable as the one with an internal adenine-uracil pair (only 0.1 kcal/mol more stable) on average.
Introduction
RNA secondary structure prediction is fundamental to understanding biological functions. The Nearest-Neighbor (NN) model is the standard computational framework for estimating the stability of RNA duplexes. However, this model relies on experimentally derived thermodynamic parameters for all possible adjacent base-pair combinations. While parameters for canonical bases (A, C, G, U) are well-established, post-transcriptional modifications introduce significant complexity.
A-to-I editing, where adenosine is deaminated to inosine, is a widespread modification. Inosine (I) pairs with cytosine (C) via two hydrogen bonds, unlike the three bonds in a guanine-cytosine (G-C) pair. Historically, the lack of specific parameters for I-C pairs forced researchers to approximate them as G-C pairs in folding algorithms, leading to consistent overestimation of stability in edited RNAs. This study addresses this gap by providing high-precision thermodynamic data for I-C pairs.
Methodology: A Hybrid Experimental-Computational Approach
To derive these parameters efficiently, the researchers developed a robust method combining physical experiments with high-throughput simulations:
- UV Adsorption Measurements: Eight RNA sequences containing I-C pairs were synthesized. By measuring UV absorbance as a function of temperature, the team determined the melting temperatures and thermodynamic properties (free energy, $\Delta G$) of these duplexes. This provided a high-accuracy experimental baseline.
- Molecular Dynamics (MD) and Free-Energy Perturbation (FEP): To cover the full range of 30 possible nearest-neighbor environments, the team used FEP simulations. This allowed them to calculate the relative free-energy change when a known base pair is mutated to an I-C pair in various stacking contexts.
- Linear Least-Squares Integration: A mathematical model integrated the sparse experimental data with the dense simulation results. This approach "anchors" the computational predictions to experimental reality, mitigating systematic biases inherent in MD force fields.
Key Insights: The Thermodynamics of Inosine
The study revealed critical differences between I-C pairs and their canonical counterparts, highlighting why treating Inosine as Guanine is problematic.
- The "Two-Bond" Reality: An internal I-C pair is significantly less stable than a G-C pair, with an average difference of 0.9 kcal/mol. Interestingly, I-C pairs are thermodynamically comparable to A-U pairs (which also have two hydrogen bonds), being only 0.1 kcal/mol more stable on average.
- Stacking vs. Pairing: Stability in RNA is driven not just by horizontal hydrogen bonds, but by vertical "stacking" interactions. The study found that tandem I-C pairs are significantly more stable than tandem I-U (wobble) pairs, despite both having two hydrogen bonds, due to the structural preference for Watson-Crick-like geometry in I-C stacks.
- Terminal Effects: A surprising finding was a "terminal bonus" for I-C pairs. At the ends of a duplex, I-C pairs appeared slightly more stable than terminal G-C pairs, possibly due to unique solvation or entropic effects at the strand termini.
Validation and Impact
The derived parameters were validated through cross-validation, yielding a Mean Absolute Difference (MAD) of 0.41 kcal/mol. This level of accuracy ensures that the parameters can be reliably integrated into existing RNA folding software. By providing these missing values, the study enables more accurate modeling of RNAs undergoing A-to-I editing, which is crucial for understanding gene regulation and transcriptomics.
Future Directions
While this work provides $\Delta G$ parameters at 37°C, further research is needed to decompose these values into enthalpy ($\Delta H$) and entropy ($\Delta S$) to allow for predictions across different temperatures. Additionally, extending this hybrid methodology to non-helical structures like loops, bulges, and junctions will complete the thermodynamic profile of inosine-modified RNA.
Summary updated