Fitness landscape of a dynamic RNA structure

Abstract

RNA structures are dynamic. As a consequence, mutational effects can be hard to rationalize with reference to a single static native structure. We reasoned that deep mutational scanning experiments, which couple molecular function to fitness, should capture mutational effects across multiple conformational states simultaneously. Here, we provide a proof-of-principle that this is indeed the case, using the self-splicing group I intron from Tetrahymena thermophila as a model system. We comprehensively mutagenized two 4-bp segments of the intron that come together to form the P1 extension (P1ex) helix at the 5' splice site and, following cleavage at the 5' splice site, dissociate to allow formation of an alternative helix (P10) at the 3' splice site. Using an in vivo reporter system that couples splicing activity to fitness in E. coli, we demonstrate that fitness is driven jointly by constraints on P1ex and P10 formation and that patterns of epistasis can be used to infer the presence of intramolecular pleiotropy. Importantly, using a machine learning approach that allows quantification of mutational effects in a genotype-specific manner, we show that the fitness landscape can be deconvoluted to implicate P1ex or P10 as the effective genetic background in which molecular fitness is compromised or enhanced. Our results highlight deep mutational scanning as a tool to study transient but important conformational states, with the capacity to provide critical insights into the evolution and evolvability of RNAs as dynamic ensembles.Our findings also suggest that, in the future, deep mutational scanning approaches might help us to reverse-engineer dynamic interactions and critical non-native states from a single fitness landscape.

mutscan - a flexible R package for efficient end-to-end analysis of multiplexed assays of variant effect data

“…As for the variant counts, we see a strong correlation between the fitness scores from the different methods. This corroborates findings from (32), where fitness scores from DiMSum were found to be highly correlated with log-fold changes calculated using DESeq2 (33) on the same count matrix. As for the detected variants, the correlation between the fitness scores is much stronger for variants with only a single mutation, and decreases as the number of mutations increases (Additional file 2:Figure S11-S14), possibly due to the lower absolute counts observed for variants with more mutations.…”

Fast free-energy-based neutral set size estimates for the RNA genotype–phenotype map

“…Since the minimum-free-energy criterion is not always unique [22] and suboptimal structures are often close in free energy to the minimum-free-energy structure [21], several low-energy structures per sequence can be relevant functionally [23] and in evolutionary processes [24,25]. Therefore, realistic models of the RNA GP map should not restrict themselves to the mfe structure for each sequence, but more complex many-to-many models should be used, which include several low-energy structures for each sequence.…”

“…A given sequence can therefore fold into several structures. This is not just an artefact of the thermodynamic model: RNAs fold into several structures [23,24] and this can be relevant for their function [23] and observed in their evolution [25]. If low-energy structures matter in the function and evolution of an RNA molecule, then they should be included in the neutral set size and thus the low-energy set size would be a more appropriate way of quantifying how many sequences correspond to a given structure and how likely a given structure is predicted to evolve.…”

Higher-order epistasis and phenotypic prediction

“…Understanding the relationship between genotype and phenotype is difficult because the effects of a mutation often depend on which other mutations are already present in the sequence, a phenomenon known as epistasis [1–3]. Recent advances in high-throughput mutagenesis and phenotyping have for the first time provided a detailed view of these complex genetic interactions, by allowing phenotypic measurements for the effects of tens of thousands of combinations of mutations within individual proteins [4–18], RNAs [19–24], and regulatory or splicing elements [25–31]. Importantly, it has now become clear that the data from these experiments cannot be captured by considering simple pairwise interactions, but rather higher-order genetic interactions between three, four, or even all sites within a functional element are empirically common [2, 12, 32–44] and indeed often expected based on first-principles biophysical considerations [12, 23, 32, 35, 36, 41, 45, 46].…”