Rationalizing the effects of RNA modifications on protein interactions

Abstract

RNA modifications play a crucial role in regulating gene expression by altering RNA structure and modulating interactions with RNA-binding proteins (RBPs). In this study, we explore the impact of specific RNA chemical modifications-N⁶-methyladenosine (m⁶A), A-to-I editing, and pseudouridine (Ψ)-on RNA secondary structure and protein-RNA interactions. Utilizing genome-wide data, including RNA secondary structure predictions and protein-RNA interaction datasets, we classify proteins into distinct categories based on their binding behaviors: modification specific and structure independent, or modification unspecific and structure dependent. For instance, m⁶A readers such as YTHDF2 exhibit modification-specific and structure-independent binding, consistently recognizing m⁶A regardless of structural changes. Conversely, proteins such as U2AF2 display modification-unspecific and structure-dependent behavior, altering their binding preferences in response to structural changes induced by different modifications. A-to-I editing, which causes significant structural changes, typically reduces protein interactions, while Ψ enhances RNA structural stability, albeit with variable effects on protein binding. To predict these interactions, we developed the catRAPID 2.2 RNA modifications algorithm, which computes the effects of RNA modifications on protein-RNA binding propensities. This algorithm enables the prediction and analysis of RNA modifications' impact on protein interactions, offering new insights into RNA biology and engineering.

Keywords: A-to-I; MT: RNA and epigenetic editing Special Issue; RNA modifications; RNA secondary structure; RNA-binding proteins; interactions predictions; m6A; machine-learning; protein-RNA interactions; pseudouridine.

Graphical abstract

Figure 1

Impact of RNA modifications on secondary structure stability Distribution of RNA fragments based on changes in secondary structure upon modification with m⁶A, A-to-I editing, and pseudouridine (Ψ). (A) The fraction of RNA fragments that retain or alter their secondary structure after modification is depicted. (B) The differences in RNA free energy (ΔΔG) between modified and unmodified fragments are shown, with positive ΔΔG values indicating decreased stability and negative values indicating increased stability.

Figure 2

Protein interactions with modified RNAs as identified through CLIP data The total number of protein interaction peaks with RNA fragments containing specific chemical modifications—m⁶A, A-to-I editing, and pseudouridine (Ψ)—is presented. Bars represent the interaction frequency for each modification, emphasizing the differential binding preferences of proteins.

Figure 3

Protein binding specificity relative to RNA modifications and structural changes Protein interactions with m⁶A, A-to-I, and Ψ modifications are displayed, focusing on both structurally stable and unstable RNA fragments. (A) The bar plot illustrates the percentage of protein interactions specific to each modification, normalized by the number of structurally stable sequences. (B) The bar plot illustrates the percentage of protein interactions specific to each modification, normalized by the number of structurally unstable sequences. Proteins without any RNA binder are highlighted in gray.

Figure 4

Protein classification according to binding specificity for each modification (CLIP) (A) Bar plot showing the modifications preferentially bound (the maximum percentage of binding sites among the three modifications) by the CLIP proteins. (B) Modification-specific and structure-independent proteins or proteins binding in >90% of the cases with fragments with a specific modification without any preference for the structural stability of the sequence. The absolute number of bindings with m⁶A structurally stable fragments for each protein is reported in the figure.

Figure 5

Binding preferences of m⁶A readers in response to RNA structural changes Changes in single-stranded content of RNA fragments are shown, comparing the RNA structure before and after m⁶A modification. The y axis quantifies the percentage change in single-stranded regions, with positive values indicating increased linearization and negative values indicating a shift toward more structured conformations.

Figure 6

Influence of RNA modifications on protein-RNA interaction predictions The impact of m⁶A and A-to-I modifications on the binding propensity of proteins to RNA is depicted through Z scores (called “Target recognition ability”), calculated using the catRAPID 2.2 RNA modifications algorithm. (A) Z scores represent the effect of m⁶A on interaction propensity, with higher scores indicating a stronger modification-induced change in binding. Bars are color coded to reflect enrichment levels, which indicate how frequently positive CLIP interactions score higher than randomly selected negative sequences. (B) A similar analysis is shown for A-to-I modifications, demonstrating a variable impact on protein-RNA binding. The Z score and the enrichment significantly correlate with each other for both m⁶A (r = 0.56, p < 1.18e-05) and A-to-I (r = 0.73, p < 2.61e-05).