A sizable slice of the drug development pie is an exercise in targeting proteins. Find an active site or pocket on a problematic protein, stuff in a small molecule to interfere with that protein’s function, and, if all goes well, treat the disease caused by that malfunctioning macromolecule.
But only about 15 percent of proteins have such a convenient pocket or active site, leading the rest to be considered “undruggable.” Several researchers and drug developers are realizing that the solution might be to target disease-linked proteins at the RNA level instead (See “Scientists Take Aim at Disease-Causing RNAs Using Small-Molecule Drugs,” The Scientist, April 2019).
Messenger RNAs are one obvious target, because they encode proteins and even interfere with cellular processes directly in some diseases. But other RNAs are in some drug developers’ sights: noncoding RNAs, such as microRNAs that regulate gene expression, could be worth altering as well. After all, only 1 to 2 percent of the human genome is translated into proteins, while about three-quarters is transcribed into RNAs (PNAS, 104:19428–33, 2007).
“With the ability to target RNA, you would dramatically expand the potential pool of drug targets,” says Rajeev Sivasankaran, head of rare diseases, neuroscience division, at the Novartis Institutes for BioMedical Research in Cambridge, Massachusetts.
One option under development is to use antisense oligonucleotides (ASOs), short sequences of nucleic acid that bind to target RNAs and block them or hasten their degradation. But these large molecules can’t easily reach the brain and spinal cord, and they can cause a drop in platelet counts, leading to bleeding problems.
Small molecules, in contrast, are tried-and-true drug candidates. RNA-targeted drugs are unusual, but not impossible: for example, certain antibiotics work on bacterial ribosomal RNAs. A small molecule binding to the functional site on an RNA could prevent interactions with the ribosome or other binding partners. Or a small molecule might promote cleavage and destruction of an RNA target. Scientists are now applying both high-throughput screening and rational design approaches to identify molecules that will influence RNA biology. Here, The Scientist surveys some of the key techniques making RNA targeting the next big thing in drug development.
Understanding RNA structure can help scientists find pockets on the nucleic acids where drugs might bind. Some RNA structures are just as complex as those of proteins, says Kevin Weeks, a chemist at the University of North Carolina at Chapel Hill and cofounder of Ribometrix, a company developing RNA-targeted drugs. Base pairing leads to secondary structure, with RNAs forming hairpins and helices. These can fold together into more-complex tertiary structures such as pseudoknots and multihelix junctions.
Weeks designed an easy-to-use technique called SHAPE (short for “selective 2′-hydroxyl acylation analyzed by primer extension”) to probe RNA structure (J Am Chem Soc, 127:4223–31, 2005; Nat Methods, 11:959–65, 2014). SHAPE identifies which parts of an RNA are loose and able to interact with a reagent, and which are bound up in a structure that blocks such interactions.
With the ability to target RNA, you would dramatically expand the potential pool of drug targets.—Rajeev Sivasankaran,
Novartis Institutes for BioMedical Research
The reagent, in this case, is one of a set of chemicals that can potentially attach itself to any of the four RNA nucleotides A, U, G, and C, tagging each with what’s called a 2′-O-adduct. The key trick: “It only reacts with the RNA at sort of wiggly or conformationally flexible nucleotides,” says Weeks. For example, in an RNA hairpin, the base-paired nucleotides along the pin’s length would be protected, but the loop at the end, where nucleotides are free, could be modified by the reagent.
The result is an RNA molecule in which certain bases are labeled with the adduct, and others are untagged. Then, Weeks and his team use reverse transcriptase to make a DNA copy of that RNA. As the enzyme travels along the RNA, it makes mistakes with the labeled bases, putting in mismatched nucleotides. When the scientists sequence the DNAs, those mistakes show up as mutations, and point to places where the original RNA was unbound.
Researchers can feed their sequences into software, freely available on Weeks’s website, that predicts the structure. It’s “pretty reliable, but not 100 percent perfect,” says Weeks. The technique works both in test tubes and inside cells.
Straight to the Screen
In recent high-throughput screens, Novartis stumbled onto small molecules that bind RNA and the spliceosome to alleviate splicing defects associated with spinal muscular atrophy (Nat Chem Biol, 11:511–17, 2015). Those researchers weren’t specifically looking for RNA binders, though. Amanda Garner, a chemical biologist at the University of Michigan in Ann Arbor, designed a screen that aims for RNAs from the start. Specifically, she’s after small molecules that bind microRNAs.
Her method is a variant on the ELISA (enzyme-linked immunosorbent assay), which uses antibodies to detect peptides or proteins. It’s called a catalytic enzyme-linked click-chemistry assay, or cat-ELCCA (Chem Commun, 52:8267–70, 2016). The assay detects small molecules that interfere with maturation of a hairpin-shaped pre-miRNA into a functional microRNA.
The first step is to link the pre-miRNA of interest to a molecular label to make it detectable later on. At the loop on the top of the hairpin, Garner attaches an 8-carbon ring called TCO (trans-cyclooctene). She immobilizes these tagged pre-miRNAs on the bottom of a 384-well plate. Then she adds different compounds from a chemical library, giving them a chance to bind the miRNAs. After that, she adds Dicer, the enzyme responsible for cutting pre-miRNAs to make them active.
Dicer cleaves any hairpins that remain unbound by small molecules, cutting off the TCO. The miRNAs attached to small molecules, however, are protected from Dicer’s blade. Finally, Garner adds horseradish peroxidase bound to tetrazine, which can bind to TCO. The activity of the peroxidase, only present on the uncleaved miRNAs, can be detected with a chemiluminescent assay and the glowing wells identified with a plate reader. Any well that glows is a potential hit.
In a recent study, Garner used cat-ELCCA to screen for compounds that block cleavage of the pre-miRNA for miR-21, a microRNA overexpressed in most cancers. Of about 83,000 small molecules and natural products, she reported 1,819 hits, 217 of which were confirmed after testing them in triplicate.
It’s not enough for a small molecule to bind the RNA of interest; it must also ignore all the other RNAs in a cell. To begin to address this specificity, Garner and colleagues checked their miR-21 binders for interactions with a control pre-microRNA, called pre-let7-d. Some of the natural extracts were selective for miR-21, and Garner has worked with collaborators to identify a pure compound that could be a candidate for an eventual cancer drug. (SLAS Discov, 23:47–54, 2018).
Not everyone has access to a large chemical library. And most small-molecule libraries are designed with protein targets in mind, notes Matt Disney, a chemical biologist at the Scripps Research Institute in Jupiter, Florida, and cofounder of Expansion Therapeutics, a drug development company targeting pathologically expanded RNA repeats. That is, the parameters best suited for targeting proteins might not apply to RNAs. For example, RNA binders might not require the same stringent molecular weight standards—typically less than 500 daltons—set for protein-targeting drugs.
Disney takes a rational design approach to targeting RNAs. His lab has designed a pipeline to identify and exploit promising interactions between RNAs and small molecules (J Am Chem Soc, doi: 10.1021/jacs.8b13419, 2019).
It starts with 2-dimensional combinatorial screening, or 2DCS (J Am Chem Soc, 130:11185–94, 2008). Disney’s idea was to pit two libraries—one of diverse RNA structures, the other of small molecules—against each other by affixing the small molecules in a microarray, and exposing them to a mixture of tens of thousands of RNA motifs. The RNAs are labeled with fluorescent or radioactive compounds, so each hit, defined as a successful binding between small molecule and RNA motif, creates a detectable spot on the microarray. The researchers excise those hits and use RNA sequencing to identify them.
Next comes data processing. Disney’s group developed a statistical procedure called HiT-StARTS, for High Throughput Structure–Activity Relationships Through Sequencing. Essentially, HiT-StARTS crunches the data to score the relative affinity between each RNA motif–small molecule pair (ACS Cent Sci, 3:205–16, 2017).
The researchers then feed that data and the RNA sequences into their database, Inforna, which is accessible on Disney’s website (Nat Chem Biol, 10:291–97, 2014). The database currently includes about 40,000 RNA–small molecule interactions, and is regularly updated with more. Inforna mines the database to pick out the best RNA–small molecule interactions.
Inforna has pointed to several potentially druggable RNAs. Researchers can also come to Inforna with a specific RNA in mind, and look for possible binding pockets and small molecules to try.
For example, the Scripps team used Inforna to identify a potential lead compound for myotonic dystrophy, an adult-onset muscle wasting disease characterized by muscle rigidity. It’s caused by an out-of-control expansion of repeats in the gene DMPK, resulting in an mRNA with hundreds or thousands of CUG’s in a row. This expanded RNA sticks to a splicing factor, interfering with the processing of other mRNAs.
The expanded RNA doubles back on itself, forming a series of links with the C’s and G’s base-paired and the U’s sticking out. Disney’s team used Inforna to design small molecules that bind to this abnormal chain, but not to normal-length DMPK transcripts, which don’t form this structure. Then, to one of those compounds, they attached a nucleic-acid cleaving antibiotic and antitumor agent, bleomycin. The combo, dubbed Cugamycin, finds and chops up the abnormal RNA (Nat Chem Biol, 13:188–93, 2017). When injected into a mouse model of myotonic dystrophy, Cugamycin normalized the majority of disease-linked RNA mis-splicing events and reduced muscle rigidity (PNAS, 116:7799–804, 2019).
In vivo validation
A lead compound that binds an RNA in test tubes may not work the same way in cells, so it’s important to validate small-molecule action in living material, and to check for off-target RNA binding. Disney has a starter toolbox for these questions, too.
One option is chemical cross-linking and isolation by pull down, or Chem-CLIP (Angew Chem Int Ed Engl, 52:1001–13, 2013). Researchers hook their RNA-binding small molecule to a crosslinker, such as the alkylating agent chlorambucil, and to a convenient pull-down agent, such as biotin. The crosslinker attaches the small molecule to the RNA target sequences. The scientists can then pull down the bound RNAs and identify them by sequencing.
For researchers who’d rather not fiddle with the chemistry Chem-CLIP requires, Disney developed a different option, called antisense oligonucleotide ligand binding site mapping, or ASO-Bind-Map, to check for small-molecule interactions with specific RNAs (Chem, 4:2384–404, 2018). It relies on the small molecule in question to stick to the target sequence and block ASO binding. If an ASO can’t bind, it won’t promote destruction of RNAs that match its sequence. In ASO-Bind-Map, researchers treat cells with the small molecule and a series of ASOs for the target RNA. They can then amplify and sequence that target RNA, looking for the regions that were protected from ASO-mediated destruction.
Even with these recent advances, the RNA-drugging toolbox is far from full. Disney says the pharmaceutical industry will need more assays like these to validate targets. And Weeks hopes to improve SHAPE’s ability to infer the tertiary shape of RNAs. Design tools like Weeks’s and Disney’s can work together with screens, notes Garner, if scientists use them to identify the right bits of RNAs—ones with likely-druggable pockets—to use as bait in their screens.
Programs like the spinal muscular atrophy drugs—now in clinical trials—and the other lead compounds identified by researchers suggest it’s a toolbox worth filling. “All this stuff says, yeah, you can drug RNA with small molecules,” says Disney. “And we really need to be doing it.”