The Rime of the Ancient Medicinal Chemist: Structure, Structure Everywhere

Our mission at Arrakis is to find small-molecule drugs that bind RNA, or RNA-targeted Small Molecules (rSMs). One of the articles of faith that informs and motivates this mission is that non-ribosomal RNA folds into structures that present small-molecule-compatible pockets and that those structures mediate important biological events. And, just as importantly, that binding of the small-molecule ligand will modulate those biological events to good therapeutic effect. These linked hypotheses—the reductionist drug discovery paradigm—have proved true for so many protein targets that they are the foundation for all modern drug discovery. But that path for drugging RNA remains largely untrodden and the landscape unfamiliar.

Recent successes suggest that small-molecule intervention into RNA dynamics and function is quite achievable: Novartis’s discovery of branaplam, which binds to and modulates the splicing of SMN2 pre-mRNA and is now in clinical trials for the treatment of spinal muscular atrophy (SMA); a mechanistically related agent from the Roche/PTC Therapeutics collaboration that is also in clinical trials of SMA; and Merck’s identification of ribocil, which binds to the FMN riboswitch and suppresses the expression of proteins from genes downstream of the riboswitch. These successes are perhaps not so surprising if you consider the emerging picture of RNA: that RNA folds, those folded structures are conserved, alterations in those structures impact function, and that RNA-binding proteins interact with those structures conferring significant—sometimes decisive—impacts on RNA function at those structures. That is all well and good, but are such functional RNA structures rare or ubiquitous?

Until recently, omic-level data have provided rather inconsistent perspectives on the prevalence of structured, functional RNA, especially in messenger RNA. But in March of this year, authors at UNC and NIBR published a paper in Cell describing a survey of RNA structure in E. coli. This paper provides the clearest and most comprehensive perspective to date, revealing a wide variety of RNA structures and their associated mechanisms of post-transcriptional regulation. The conclusions of this paper are strongly validating for the overall enterprise of identifying rSMs. In particular, the conclusions imply that if you find binders to regulatory RNA structures, their prospects for impacting translation are quite good. As the authors note, “…every single gene examined here is regulated in a meaningful way by RNA structure.”

The method at the center of this paper is SHAPE—selective 2’-hydroxyl acylation analyzed by primer extension—developed in the labs of Kevin Weeks at UNC. In SHAPE, an RNA is exposed to an acylating agent (in this case, 1-methyl-7-nitroisatoic anhydride or “1M7”) that forms esters at the 2’-OH positions of RNA. Those 2’-OHs that are either solvent accessible or locally activated are preferentially acylated, while those 2’-OHs that are stably hybridized or otherwise protected from the reagent exhibit lower levels of acylation. Upon denaturation and sequencing, the modified riboses induce either termination of the reverse transcriptase or misincorporation; the former is read as shorter DNAs and the latter as “mutations.” The relative levels of modification across the sequence enable inferences about the structure and refinement of structural models.

Predicting RNA structure with SHAPE software: Arrakis’s computational biologist Luis Barrera has created a set of software tools for working with SHAPE data. You can download this open-source package here.

One appealing aspect of SHAPE is that it allows rapid structure assessment using sequencing techniques, which helps account for its wide adoption in the RNA research community. Another is that SHAPE can be carried out either on isolated RNAs in solution or on RNA in intact cells—a nifty trick. But there is an art to this method and reliable conclusions benefit from meticulous execution and deep sequencing. In order to gain an appreciation of the ubiquity of functional RNA structures, this new paper examines the transcriptome of E. coli, yielding three especially important take-home messages:

1. There are other SHAPE-like methods with their fans and practitioners—dimethyl sulfate (DMS), hydroxyl radicals, other acylating agents, terbium cleavage, etc. Those methods and SHAPE have been used in various labs to survey RNA structures in cellular transcriptomes. The conclusions have been somewhat divergent: though they point to functional RNA structures, those structures appear rare and not robust. The thoroughness of the Weeks work reveals that RNA structure in E. coli is ubiquitous, especially in 5’UTRs. The implications of this observation for therapeutic intervention—should similar conclusions be reached for human cells—are profound. It means that most every gene of interest will yield potentially druggable RNA structures.
2. Second, the authors demonstrate a clear relationship between RNA structure and translational efficiency (TE), especially in the neighborhood of ribosome binding sequences. That is, the translational efficiency of a given mRNA is related to prevalent structures assumed by that mRNA and more specifically the magnitude of the activation free energy for the unfolding of those structures. If the therapeutic goal is the selective suppression of protein expression, this kind of insight is essential to develop ligands that modulate RNA function rather than merely bind to an RNA.
3. Finally, one of the most encouraging findings from the Weeks work is that RNA structures are not only prevalent but exhibit distinct architectures. This points to the prospect of selectivity for rSM intervention. One of the prevalent concerns about small molecules binding to RNA structures is whether selectivity is possible. One solution to this problem is to focus on binding modes that do not depend exclusively on canonical binding modes common to all nucleic acids—intercalation, groove binding, and electrostatic binding to phosphates. Rather, the focus should be on pockets created by the tertiary structure of a folded RNA. But is this adequate? Are those pockets sufficiently diverse? The present work points in the right way insofar as pocket diversity is likely correlated with overall structural diversity. Moreover, these observations improve the prospects for inter-target selectivity, a prerequisite for good pharmacology and acceptable toxicology.

There are two major risks facing those of us working to drug RNA—chemistry risk and biology risk. Chemistry risk derives from not just the question of whether drug-like small molecules can be found that bind into pockets in RNA. It also derives from the question of whether there are functionally significant RNA structures that present those potentially druggable pockets. And even if you stipulate the ready identification of RNA/ligand pairs, the biology risk is that ligand binding has no meaningful impact on RNA function and thus no therapeutic value. The Cell paper, insofar as its findings are generalizable to mammalian cells, reduces both chemistry and biology risk for the enterprise of drugging RNA.

At Arrakis, we share the perspective of Weeks et al. and we’re already working hard developing analogous concepts, methods, and data for mammalian cells. We are discovering exactly what you might predict: that functional RNA structures are ubiquitous and addressable using the drug discovery toolkit. This is truly the moment to drug RNA and we applaud this important contribution to the field.