A prevailing hypothesis about the origins of life is the RNA World hypothesis, which posits that an ancestral state of modern life relied on RNA molecules that had developed the capacity to self-replicate and catalyze key biochemical reactions, paving the way for the evolution of increasingly more complex biochemistry and biology.
Yet, these days RNA is considered primarily a temporary information carrier. It does not store the information long term – this role is performed by DNA. It does not play active functions such as catalyzing reactions or self-replicating – this niche is occupied by proteins. Of course, there are notable exceptions: ribosomal RNAs, for example, comprise the catalytic center of the ribosome. Many non-coding RNAs (miRNAs, snRNAs, snoRNAs, and others) guide critical sequence and structure-specific regulation—though almost always requiring proteins to see those functions out. However, one of the major classes of RNAs, messenger RNAs, is by and large considered only an information “shuttle” from DNA to protein-making ribosomes. The prevailing thought is that although mRNAs are heavily regulated, they themselves are not the regulators; they do not intrinsically respond to changes. The RNA World hypothesis suggests that this was not always the case; in the early stages of life, RNA molecules would have needed to regulate their functions in response to environmental changes and catalytic needs. Does it mean mRNA lost the role of active, intrinsic regulator of gene expression and gave it away to proteins and other classes of RNA?
Well, not quite. In early 2000s, several groups discovered that mRNA can also be an active regulator of gene expression(Nahvi et al. 2002; Winkler et al. 2002; Mironov et al. 2002; Vitreschak et al. 2003). The RNA elements they identified were termed riboswitches; they are structural RNA elements that are poised to change in conformation upon direct binding of a small-molecule ligand. The change of 3D folding following the ligand binding opens up or closes down the regulatory sites that are important for ribosome binding, transcription termination, and other key processes. Importantly, these folding changes do not require any additional RNA or protein cofactors; they are a way for mRNAs to regulate their own expression. Subsequent studies demonstrated that riboswitches are abundant in bacterial transcriptomes, with known riboswitch families controlling many pathways in vitamin and amino acid biosynthesis in any given bacteria. Several riboswitches were later discovered in fungi and plants based on similarity to bacterial counterparts—although their evolutionary origins are less clear. This led to the idea that the mRNAs did not entirely lose this “RNA World” original function, and could still be the active regulators of gene expression, at least in bacteria.
However, the search for active roles of mRNA in animal transcriptomes has been less successful. No ligand-binding riboswitches were ever identified; what was found were ribozymes (Salehi-Ashtiani et al. 2006) and a couple of examples of protein-binding RNA switches (Ray et al. 2009; Liu et al. 2015). Such scarcity raised a number of unanswered questions. Is it possible that the mRNAs do not play active regulatory role in animals or, instead, that the precision of the search methods used is simply not enough to detect them? Is it possible that their role shifted from major regulation towards fine-tuning?
One possibility is that mRNAs play some of the same roles in animals as they do in bacteria, but that riboswitch-like mechanisms are much harder to identify. While bacterial genomes are small, their transcript architecture relatively simple and uniform, and many species have been sequenced, animal genomes are larger, with more diverse transcripts, and much more sparse sequence representation across evolutionary space. The major toolkit for identifying active mRNA elements—comparing closely related genomes to each other in search for evolutionarily conserved motifs—does not work as well on animal genomes because of their size and scarcity. Yet, what has changed since the early 2000s, when riboswitches were first discovered, is availability of extremely powerful high-throughput methods and publicly available data, which enabled us to answer some of these questions.
The specific question our paper addresses is: if we systematically search for RNA switches (elements that adopt two mutually exclusive conformations, each potentially leading to different gene-regulatory outcomes) in human protein-coding transcriptome, how many of them do we identify and what do they look like? Systematically searching for RNA switches in a metazoan transcriptome is something that has not been successfully done before; luckily, the recent development of advanced methods makes it possible today. First, there are enough families of riboswitches studied, that it is possible to imagine how the novel families could look like. We used machine learning techniques to capture general features shared by many riboswitch families, and searched for mRNAs that exhibit these features in human transcriptome. Second, thanks to the development of in cell RNA structure probing methods (Zubradt et al. 2017), we could now test whether mRNAs do exhibit riboswitch-like dynamic behavior in their native environment. Third, the massively parallel reporter assays (MPRAs) (Oikonomou et al. 2014) enabled us to study the functional outcomes of RNA conformation shifts in fine detail. These three methods combined allow for high-confidence identification of RNA switches.
Even having searched for RNA switches that work under homeostatic conditions, and not accounting for those active only under metabolic stresses, we have identified >200 high-confidence RNA switches. As very little is known about the RNA switches in humans, we have studied one of them—an RNA switch in the 3′UTR of the RORC gene that encodes the master regulator of Th17 cells—in detail. A combination of in vivo structure probing and cryogenic electron microscopy allowed us to confirm that it forms two distinct structural conformations. Using genome-wide CRISPRi screening, we found that one conformation binds strongly to nonsense mediated decay (NMD) machinery, leading to rapid mRNA degradation, while the other conformation does not invoke a strong NMD-mediated RNA decay. Moreover, the relative ratio of the two conformations shifts depending on the cellular conditions, and thus controls the level of RORC-encoded mRNA using a switch fully encoded in that same mRNA.
Our results show that the mRNA structure switches do play an active—and potentially pervasive—role in gene regulation not only in bacteria, but also in metazoa. This switchable nature of RNA for function and regulation is indeed present in animals, but it required new methods to discover. The switching behavior is used by many mRNAs (>200 identified in our study alone), and I expect that more will be found in the near future. We speculate that these RNA switches could act as sensors that transmit cellular signals into changes in gene expression by affecting mRNA stability or translation. The signal molecules sensed by RNA switches could vary all the way from metabolites to protein regulators. This could be especially important in cells where the local control of protein concentration is important, such as in neuronal processes. Evidence is arising that specific RNA regulatory mechanisms might be especially important in particular organisms and in particular tissues: for example, RNA editing is much more abundant specifically in neural tissue of cephalopods, and is linked to the extreme plasticity of their proteome, which might be connected to more advanced regulatory functions that in the closely related organisms (Voss and Rosenthal 2023). It is possible, therefore, that RNA switching behavior could be more important in particular tissues in certain organisms than in others. It is critical to understand how such mechanisms are wired into the cellular gene regulatory network; while their effect sizes might be less noticeable than those of transcription regulators, its understanding is crucial for building future understanding of gene expression.
Given the advances in direct targeting of RNA by small molecules(Warner et al. 2018), this work might lead to identifying the therapeutic targets that could be manipulated at the RNA level, which might be particularly useful for the many families of “undruggable” targets. Additionally, it is estimated that thousands of rare bacterial riboswitches are yet to be identified(Kavita and Breaker 2023), and we hope that our work could facilitate this development. Our findings underscore the potential of RNA switches as key regulatory elements, paving the way for novel therapeutic approaches and a deeper understanding of gene regulation in both bacterial and metazoan systems.