RNA State of Mind

Perspectives from the ‘Targeting RNA Using Small Molecules’ meeting @NYASciences #rSMs 

Our team at Arrakis Therapeutics recently had the pleasure to co-sponsor the ‘Targeting RNA Using Small Molecules’ symposium at the New York Academy of Sciences. While several conference sessions over the years have focused on our favorite topic, including the (in)famous Gordon Conference that catalyzed the formation of Arrakis (‘Discovering a New World of RNA-Targeted Medicines’), to our knowledge this is the first meeting solely focused on small-molecule targeting of RNA.

Several of our team traveled to the NYAS meeting. In today’s Dark Matter Blog I will highlight several of the key themes from the event and use them as a framework to discuss the tremendous opportunities and unique challenges that Arrakis, and the broader research community, face as we strive to make RNA-targeted drug discovery a reality.

“The key to pursuing excellence is to embrace an organic, long-term learning process, and not to live in a shell of static, safe mediocrity. Usually, growth comes at the expense of previous comfort or safety.” 

–Josh Waitzkin (chess prodigy, author, Duchenne Muscular Dystrophy advocate, New Yorker)
The Art of Learning

The Opening 

The introductory remarks were delivered by Samuel Hasson (Principal Investigator, Pfizer Neuroscience). Sam presented a compelling argument that targeting RNA with small molecules has the potential to open up a new landscape for drug discovery by enabling targeting of classically ‘undruggable’ targets supported by human genetics. He reasoned that leveraging small molecules as a modality should lead to the broader application of RNA targeting across multiple disease states and tissue types based on superior drug-like properties of small molecules compared to more complex biologics approaches including antisense oligonucleotides (ASOs), siRNA therapeutics, mRNA delivery, and CRISPR technology (e.g. oral availability, brain penetrance).

The high-level goal of the meeting was clearly laid out – let’s open a dialogue to enhance our shared understanding of the target landscape, and the related enabling technologies, to put RNA drug discovery on par with traditional protein-centric approaches.

This was all sounding quite familiar (Escaping the Gravitational Pull of “Druggability’). We were definitely in the right place.

The Middle Game

Identifying targetable RNA structure/function relationships in disease-relevant RNAs

Identifying targets of interest for RNA drug discovery is not particularly challenging. There is a wealth of high-value genes with causal links to human disease that have proven refractory to small molecule drug discovery and represent low-hanging fruit (see Drugging the Undruggable: Transcription Factors). In contrast, it is often less clear how to identify the most promising structures within RNA targets that are both ligandable with small molecules and predicted to modulate underlying biological function. Kevin Weeks (UNC Chapel Hill) presented a talk focused on this key concept.

Kevin argued that RNA structures with complex 2D/3D folds are likely to represent the best opportunities for success. He highlighted two chemical probing technologies developed in his lab that can identify RNA regions with these features. The first is selective 2’-hydroxyl acylation analyzed by primer extension and mutational profiling (SHAPE-MaP; Nature Methods 11, 959-965 (2014)). SHAPE-MaP serves as a ‘chemical flexibility meter’ that encodes 2D RNA structural information as mutation frequencies that can be read out at scale via next-generation sequencing (NGS). Low SHAPE reactivity nucleotides are less flexible and are likely to be paired, while nucleotides with high SHAPE reactivity are more flexible and are likely to be unpaired.  Importantly, SHAPE-MaP technology was able to identify stable structural motifs that overlapped with a high percentage of known functional elements in well-studied viral RNA targets such as HIV and HCV.

RNA also harbors higher-order 3D structures formed by ‘through space’ interactions that allow for sampling of multiple conformations in a dynamic ensemble. The second method highlighted in Kevin’s talk addresses these features by identifying RNA interaction groups by mutational profiling (RING-MaP). The technique combines information from multiple sites of dimethyl sulfate (DMS) chemical modification using NGS, which are subsequently analyzed to reveal correlated and clustered interactions (Proc Natl Acad Sci USA 111, 13858-63 (2014)). In unpublished work, the Weeks lab applied RING-MaP to identify structures of interest in Dengue virus. Using an orthogonal mutagenesis strategy several of these structures were shown to significantly inhibit viral replication, suggesting they represent bona fide functional elements.

Chemical probing methods for RNA, such as SHAPE-MaP and RING-MaP, represent a major advance in our ability to identify functional RNA structural elements in vitro and in cells. The scalability of these methods based on their adaptation to NGS-based readouts suggests they are ready for primetime and are likely to have a major impact going forward.

Establishing meaningful RNA drug discovery flowcharts 

There is a tremendous knowledge base regarding best practices for building early drug discovery pipelines for traditional protein target classes. If you are a kinase, GPCR or ion channel, pharma and biotech outfits have a pretty solid roadmap for how to identify and develop a lead molecule for you. The same cannot yet be said for RNA targets. How do we get there? This is a big topic with many considerations, but several of the speakers at NYAS touched on themes relevant to the discussion.

Nathan Baird (University of the Sciences) emphasized the importance of carefully considering assay conditions when studying RNA-small molecule interactions. During his postdoctoral training with Adrian Ferrè-D’Amarè (NHLBI/NIH), Nathan studied the idiosyncratic nature of riboswitch binding to cognate ligands, demonstrating that each is unique in its structural response to metabolite (RNA 16, 598-609 (2010)). Some riboswitches require the presence of their cognate ligand to adopt a native structure, while others required increased Mg²⁺ concentrations to collapse into the native fold prior to compound binding. The data imply that studying RNA-small molecule interactions under a single assay condition is likely to be inconclusive, if not utterly misleading.

In more recent work, Nathan addressed this issue by developing a FRET-based method to interrogate RNA-ligand interactions using a multi-dimensional analysis across a range of Mg²⁺, ligand and temperature conditions to yield a conformation/stability response surface for the cyclic diguanylate (c-di-GMP) riboswitch (Nat Commun 6: 8898 (2015)). The experiments rapidly identified the optimal parameters for cognate c-di-GMP ligand binding, and also showed that the non-cognate ligand kanamycin B stabilizes a non-native riboswitch confirmation that inhibits subsequent c-di-GMP binding. This is an interesting demonstration that allosteric control of folding may be a feasible approach for functional targeting of riboswitches and other RNA structures.

High-throughput screening (HTS) is a mainstay in the modern protein-focused drug discovery tool kit. There is a strong need to develop similarly predictive HTS assays for various classes of RNA targets and biology. Amanda Garner (University of Michigan) is paving the way toward assay formats that emphasize functional outcomes. The Garner Lab is focused on chemical probing of translational control and microRNAs as targets for therapy. They developed a catalytic enzyme-linked click chemistry assay (cat-ELCCA), a chemical ELISA, to monitor pre-miRNA stem-loop processing by Dicer (Bioconjugate Chem 26, 19-23 (2015)). Amanda’s team recently completed a robust pilot screen of ~50K small molecules and ~33K natural product extracts to monitor pre-miR-21 processing using the method (SLAS Discov doi: 10.1177/2472555217717944 [Epub ahead of print]). We look forward to learning more about the feasibility of targeting this important class of non-coding RNAs from Amanda’s team and others going forward.

Chemical biology is playing an increasingly prominent role in the pharma and biotech industry. Common applications of chemical biology in drug discovery include target ID following phenotypic screening campaigns and the exploration of new mechanisms of action for difficult-to-drug and novel target classes. At Arrakis, we firmly believe that chemical biology can be successfully applied to and have a major impact on exploration of RNA-small molecule interactions. This was one of many topics covered in a talk by Matt Disney (Scripps Florida). Matt discussed several interesting chemical biology strategies employed by his lab. Chemical Cross-Linking and Isolation by Pull-Down (Chem-CLIP) is an approach that adds functional moieties to small molecules including nucleic acid reactive groups that allow conjugation to cellular RNA targets and biotin handles that allow for subsequent isolation of conjugated products. The method was used to successfully demonstrate target engagement of small molecules for several microRNAs (miR-96, miR-210 and miR-18a) and RNA repeat expansions (FXTAS r(CGG) expansion, DM1 r(CUG) expansion) (Proc Natl Acad Sci USA 113, 5898-903 (2016), ACS Chem Biol 11, 2456-65 (2016), J Am Chem Soc 139, 3446-3455 (2017), ACS Cent Sci 3, 205-216 (2017), Nat Chem Biol 13, 188-193 (2017)).

Another flavor of RNA-centric chemical biology Matt presented was small molecule nucleic acid profiling by cleavage applied to RNA (Ribo-SNAP). In this method RNA-targeting small molecules are conjugated to the natural product bleomycin, which cleaves the targeted RNA. This is analogous to the PROTAC approach gaining traction in the protein world. This method was used to modestly, but significantly, reduce r(CUG) expansion RNA both in vitro and in DM1 patient cells in a selective manner. The Disney lab has also used in situ click chemistry to leverage the r(CUG) expansion in the templated synthesis of its own inhibitor. This involves adding azide and alkyne functional groups to r(CUG)-binding monomers to promote the self-assembly of high-affinity, multivalent chemical probes on their target. When the functional groups are swapped for FRET donor and acceptor pairs a similar strategy can be used to generate a chemical method for RNA imaging (Nat Chem Biol 13, 188-193 (2017)).

While these methods vary in their general applicability, each is a shining example of the innovation that chemical biology approaches can bring to bear on the RNA-targeted drug discovery process.

Exploring chemical space for RNA-targeting small molecules

Any conversation about targeting RNA with small molecules inevitably touches on whether we understand the most appropriate chemical space for the job. While the concept of privileged RNA scaffolds is attractive, some of the most prominent representative examples in the field suggest that productive small molecules are sitting in our screening decks already. John Howe, (Principal Scientist, Merck), presented the amazing ribocil story, an utterly unremarkable compound that does something remarkable:  it binds to and functionally impacts the bacterial FMN riboswitch. LMI070, the SMN2 splice modulator from Novartis that is currently in the clinic as a potential treatment for Spinal Muscle Atophy (SMA), is a similar story and is a compound that could easily be mistaken for a kinase inhibitor. How similar or different are RNA binders to the protein-friendly compounds in our libraries?

Amanda Hargrove (Duke University) helped shed some light in her presentation on a recently published assessment of key properties of RNA-targeted ligands (Angew Chem Int Ed Engl 56, 13498-13502 (2017)). In the study, the Hargrove lab performed a cheminformatics and shape-based analysis on 104 RNA-targeted ligands with biological activity to generate a RNA-targeted Bioactive ligaNd Database (R-BIND). The high-level takeaways are that bioactive RNA compounds are largely compliant to traditional medicinal chemistry rules, but are enriched for nuanced structural features such as having a rod-like shape. In principle, the insights from the analysis could facilitate the selection of RNA-focused small molecule libraries by sampling from our existing drug discovery compound screening sets.

At Arrakis we have performed similar analyzes and come to similar conclusions. That said, the limited size of the available small molecule training sets limits the conclusions we can draw. Due to the limited chemical space sampled, we run the risk of deluding ourselves into believing a self-fulfilling prophecy. There is a clear demand for more high-quality examples like ribocil and LMI070 as guideposts to show us the best path forward.

The Ending 

As we left the NYAS meeting, we were energized by the tremendous research and the growing scientific community dedicated to exploring the concept of targeting RNA with small molecules. We hope this is the first of many dedicated meetings on the subject and look forward to keeping in touch with the many friends and colleagues in attendance.

The Arrakis team also left feeling that, while it is early in the game, we are on the right track. We have assembled a great team of creative leading-edge RNA scientists and industry vets. The team is continuing to grow, allowing us to add talent with key skill sets to enrich our culture. Our TRYST platform is identifying RNA structures of interest in high-value ‘undruggable’ targets across several disease areas. We have multiple HTS campaigns underway to rapidly assess the ligandability of RNA as a class and explore a broad swath of available chemical space. We are assembling a comprehensive and integrated RNA flowchart, with multiple downstream biochemical and cell-based assays to support the hit-to-lead process. Our PEARL-seq chemical biology platform is progressing nicely and will provide key tools for assessing small molecule RNA target engagement and selectivity. We don’t expect it to be easy, but by capitalizing on our successes and learning from our failures, we are confident we can build an efficient drug discovery engine to unlock the full potential of RNA biology.

While many aspects of our work are proprietary, Arrakis is committed to engaging the scientific community to promote open science. As a first example, today we are glad to announce the public release of our first-generation SHAPEware software for analyzing SHAPE-MaP chemical probing data (www.shapewarerna.com). We believe that SHAPEware offers several advantages over existing tools including improved read mapping, the ability to generate more accurate measures of standard error using biological replicates, and a modular architecture scalable to the cloud to increase throughput. We encourage you to take it for a test drive. SHAPEware is freely available and the license is very permissive. All we ask is that you share your feedback and any modifications or improvements with the community. In the months and years to come we envision rolling out additional modules designed to analyze other streams of RNA chemical modification data. Happy MaP’ing!