Exploring the RNA recognition motif: a computational analysis of RRM-RNA recognition and the interplay between multiple RRM domains

The RNA recognition motif (RRM) is the most prevalent RNA-binding protein domain in eukaryotes and is involved in most of RNA metabolism processes. Their RNA binding preferences have been investigated for almost two decades, but their RNA recognition code has remained elusive. In addition, RRMs are often accompanied by other RNA-binding domains to achieve higher specificities and affinities, but their interplay is not very well understood either. On such arrangements the inter-domain linker plays a key role, but its highly dynamic nature and predominant conformational ambiguity makes it challenging to study. The goal of this thesis is to help fill these gaps by providing novel insights on the RRM-RNA recognition preferences, explore the relative orientations of tandem RRMs, and help elucidate the role of the inter-connecting linker and proteins with a conformational ambiguous behaviour in general.
We gathered all the available structural information on experimental RRMs and RRM-RNA complexes, and based on a carefully curated sequence alignment, we developed RRMScorer, a computational scoring method to estimate canonical RRM-RNA binding. Our tool was reliable across several experimental sets and test-cases, and proved particularly useful on identifying single mutations to modify the RRM’s specificity. Then we focused on tandem RRMs, with two adjacent RRM domains, and defined a robust method to investigate the relative domain orientation in multidomain proteins using inter-domain vectors referenced to a stable secondary structure element. The orientations captured by the vectors correlate with intra-protein interactions, and by extending our analysis to AlphaFold2 (AF2) predicted structures, with particular attention to the inter-domain predicted aligned error (PAE), we demonstrate that inter-domain RRM orientations seem to be constrained.
According to our analysis, and in agreement with current literature, the linker region plays an important role in regulating many aspects of the tandem RRM’s interplay and sometimes participates directly on RNA binding. Connecting linkers have demonstrated to be not only dynamic, but also conformationally ambiguous in a number of cases, folding upon RNA binding for example. We studied two protein datasets that have shown conformationally ambiguous behavior experimentally, undergoing disorder-to-order transitions or switching folds. Our analysis highlighted the challenge that supposes studying such regions, still not well-captured by AF2, and requiring a probabilistic approach rather than the classic ordered/disordered vision.
By connecting these findings, this thesis helps to enhance our understanding of RRM-containing proteins and their dynamic behaviour, shedding light on a field with potential applications in both therapeutics and biotechnological settings.