Small molecule targeting of RNA structures in neurological disorders

Abstract

Aberrant RNA structure and function operate in neurological disease progression and severity. As RNA contributes to disease pathology in a complex fashion, that is, via various mechanisms, it has become an attractive therapeutic target for small molecules and oligonucleotides. In this review, we discuss the identification of RNA structures that cause or contribute to neurological diseases as well as recent progress toward the development of small molecules that target them, including small molecule modulators of pre-mRNA splicing and RNA repeat expansions that cause microsatellite disorders such as Huntington’s disease and amyotrophic lateral sclerosis. The use of oligonucleotide-based modalities is also discussed. There are key differences between small molecule and oligonucleotide targeting of RNA. The former targets RNA structure, while the latter prefers unstructured regions. Thus, some targets will be preferentially targeted by oligonucleotides and others by small molecules.

Introduction

Since the dissemination of the first draft of the human genome¹ and subsequent sequencing efforts, the underlying genetic causes of many neurological disorders have been elucidated. In many cases, the discovery of RNA mutations has allowed for further insight into disease pathobiology and hence potential targets for therapeutic intervention.

RNA can contribute to disease by a variety of mechanisms depending on the location of the mutation, deletion, or insertion (i.e., in an intron or exon) or the function of the RNA that is mutated (i.e., coding or noncoding) (Fig. 1). For example, in frontotemporal dementia with parkinsonism-17 (FTDP-17) and spinal muscular atrophy (SMA) a mutation itself alters pre-mRNA splicing patterns, leading to changes in mature mRNAs and its encoded protein as well as protein expression levels.^2-4 In other neurological disorders, an insertion in a coding sequence can lead to the translation of a toxic protein, such as poly(Q) in Huntington’s disease (HD).⁵ Insertions in noncoding regions can activate a gain-of-function in which the RNA sequesters regulatory proteins, leading to dysregulation of alternative pre-mRNA splicing, as observed in fragile X–associated tremor ataxia syndrome (FXTAS)⁶ and amyotrophic lateral sclerosis (ALS).⁷ In all three cases above, the insertions are repeat expansions, a class of RNAs that cause microsatellite disorders that also undergo repeat-associated non-ATG (RAN) translation, an aberrant mechanism that is initiated without a canonical start codon and produces toxic polypeptides.^8-12 Notably, in each of these cases, the mutation or repeat expansion alters the structure of the RNA, thus leading to dysfunction. The influence of RNA structure on function and disease has been previously reviewed.¹³

Figure 1.

Mutant RNAs can cause or contribute to neurological diseases via multiple mechanisms. RNA repeat expansions fold into hairpin structures that form high affinity binding sites for RNA-binding proteins, thus causing disease via an RNA gain-of-function mechanism. This gain-of-function results in RNA processing defects and the formation of RNA-protein nuclear foci. Mutations in exon-intron junctions can alter binding of the spliceosome to the RNA, leading to aberrant pre-mRNA splicing.

Since RNA can cause or contribute to disease via various mechanisms, it has become an attractive therapeutic target for many neurological disorders. Targeting RNA could be advantageous as it lies upstream of protein targets, and modulation of the disease-causing RNA would result in direct improvement of disease phenotype. Indeed, many disease-causing RNAs have been targeted with antisense oligonucleotides (ASOs) that recognize sequence via Watson–Crick base pairing.¹⁴ ASOs have been invaluable tools to study disease biology, have entered clinical trials, and have been approved as therapeutics. For example, Nusinersen, which has been approved to treat SMA,¹⁵ is a life changing therapy that has demonstrated the great potential for ASOs to provide lead medicines. Small molecules that target RNA are an important alternative to oligonucleotide-based approaches as they can be more easily optimized using medicinal chemistry because of the large number of chemical structures that can be synthesized and tested and have potential for more favorable pharmacokinetic properties. Furthermore, small molecules can be designed to recognize a specific structure to improve selectivity and combat off-target effects. As ASOs prefer to bind unstructured regions,¹⁶ small molecules target a complementary set of RNAs, those that contain highly structured regions. Thus, while ASOs can be designed based on the sequence of an RNA, when designing small molecules, it is important to consider its structure.

RNA three-dimensional folds are composites of base paired regions and noncanonically paired regions, such as hairpins, internal loops, and bulges, as well as more complex structural motifs such as pseudoknots and quadruplexes (Fig. 2). Importantly, the structure of a given RNA can be modeled by using a dataset of thermodynamic parameters for base paired and noncanonically paired regions,^17,18 affording an ensemble of structures via free energy minimization.^19,20 Although this method can accurately predict structures of short RNA sequences, the structures of larger RNAs can be deduced by the addition of experimental restraints²¹ generated from chemical modification reagents^22-24 or selective 2′-hydroxyl acylation analyzed by primer extension (SHAPE)^25,26 in vitro or in cellulis. Indeed, free energy minimization and experimentally generated folding restraints have been used to determine structures of various RNAs and identify structural patterns within the human transcriptome.²⁷

Figure 2.

RNA secondary structure can be predicted from the sequence. Primary structure determines secondary structure, which consists of helices (gray), hairpins (green), internal loops (purple), multibranch loops (red), bulges (light blue), and dangling ends (orange). Interactions among secondary structural elements, such as the pseudoknot shown in dark blue, lead to tertiary structure. The sequence and structures correspond to the tetrahydrofolate (THF) riboswitch.^95,96 Atomic coordinates for the 3D structure were obtained from the RCSB Protein Data Bank (PDB ID 4LVV).

The ability to model RNA structure accurately from sequence has allowed for the development of a platform that can identify small molecules privileged for binding to certain RNA three-dimensional folds, a library-versus-library screening approach called two-dimensional combinatorial screening (2DCS).^28-30 The 2DCS method uses agarose-coated microarrays to display a library of small molecules, whether site-specifically immobilized or absorbed onto the surface. The small molecule microarray is then incubated with an RNA library containing unique three-dimensional folds generated by a randomized region in discrete patterns, for example, an internal loop or a single-nucleotide bulge.³¹ Incubation with competitor oligonucleotides that mimic the constant regions of the library restrict small molecule binding to the unique three-dimensional folds. Following incubation, bound RNAs are excised, amplified, and sequenced to identify RNA motifs privileged for interacting with the small molecule. RNA motif-small molecule interactions are scored using a statistical method dubbed high-throughput structure activity relationships through sequencing (HiT-StARTS), which computes the statistical significance for each selected RNA and its relative binding affinity.^32,33 Provided that there is sufficient fold coverage in the sequencing data that define bound RNA motifs (at least sixfold),^32,33 we anticipate that HiT-StARTS could define the nucleic acids, whether DNA or RNA, that bind other modalities including peptides, carbohydrates, and oligonucleotides. These RNA motif–small molecule interactions and their scoring functions comprise a database used by a lead identification strategy named Inforna (Fig. 3).^34,35 Inforna has provided lead small molecules targeting structural motifs in various disease-causing RNAs including a mutated microtubule-associated protein tau (MAPT) pre-mRNA that causes FTDP-17³⁶ and RNA repeat expansions,¹³ which are discussed below. Although structural information can assist with the design of small molecules targeting RNA, small molecules have been identified in the absence of structural information using screening approaches.^37,38 These screening approaches rapidly identified bioactive compounds; however, their mechanisms of action and RNA structures that they bind had to be determined at a later time.^37-39 In comparison, targeted screens such as 2DCS can rapidly identify small molecules that directly interact with RNA structures. Both approaches have successfully identified small molecules that target RNA structure in neurological disorders, which are discussed herein and summarized in Table 1.

Figure 3.

The lead identification strategy, Inforna, provides small molecules against disease-causing RNAs. Lead small molecules are identified by querying the structural motifs found in a disease-causing RNA against an annotated database of experimentally identified RNA motif–small molecule interactions.

Table 1.

RNA-targeting therapeutics discussed in this review

Type ofRNA	Disease	RNA target	Therapeutic	Stage of development	Molecular weight (g/mol)	Dosing	Ref.
Splice site	FTDP-17	MAPT exon 10-intron 10 junction	ASO	Preclinical (mouse)	~5500a	Osmotic pump in mice	50
Splice site	FTDP-17	MAPTexon 10-intron 10 junction		Preclinical (cell culture)	270–370		36
Splice site	SMA	SMN2 exon 7 5′ splice site	Nusinersen (ASO)	Approved	7501	Intrathecal injection (biweekly)	15,56,57,59
Splice site	SMA	SMN2 exon 7 5′ splice site		Clinical trials	377	Oral (weekly)	37
Splice site	SMA	TSL1 at SMN2 exon 7 3′ splice site		Clinical trials	389	Oral (daily)	38,60
Splice site	SMA	SMN2 TSL2 at exon 7-intron 7 junction		Preclinical (cell culture)	370		61
RNA repeat	HD	r(CAG)exp		Preclinical (cell culture)	312		81
RNA repeat	HD	HD mRNA	IONIS-HTTRX/ RG6024 (ASO)	Clinical trials	NDb	Intrathecal injection (monthly)	86
RNA repeat	HD	Mutant HD Allele	WVE-120101 and WVE-120102 (ASO)	Clinical trials	NDb	Intrathecal injection	87,88
RNA repeat	FXTAS	r(CGG)exp		Preclinical (cell culture)	1324		76,80
RNA repeat	FXTAS	r(CGG)exp		Preclinical (cell culture)	1887		77,79
RNA repeat	FTD/ALS	r(G4C2)exp		Preclinical (cell culture)	374		78
RNA repeat	FTD/ALS	C9orf72	IONIS-C9RX/ BIIB067 (ASO)	Clinical trials	NDb	Intrathecal injection	91
RNA repeat	SCA10	r(AUUCU)exp		Preclinical (cell culture)	1180		75

^aMolecular weight is an approximation based on the sequence of the ASO and does not take into account modifications.

^bND indicates that the sequence/molecular weight was not disclosed.

Collectively, the above provides a framework for important considerations for the discovery of therapeutics targeting RNA. From the perspective of the RNA target, one should consider where the functional, disease-causing element lies, whether in a highly structured or less structurally defined region. The former is better suited for a small molecule-targeting approach, while the latter is better suited for oligonucleotide-based modalities. From the standpoint of the potential therapeutic modality, blood–brain barrier (BBB) penetrance is required for the treatment of neurological diseases. BBB penetrance is a function of molecular weight and lipophilicity, among others. Indeed, the likelihood that a small molecule is BBB penetrant can be predicted by the Central Nervous System Multiparameter Optimization (CNS-MPO) score.⁴⁰ CNS-MPO is based on various physiochemical parameters including lipophilicity (clogP), calculated distribution coefficient at pH 7.4 (clogD), molecular weight, topological polar surface area, number of hydrogen bond donors, and most basic center (pK_a).⁴⁰ Because the number of modifications allowed to maintain bioactivity is limited for oligonucleotides, it is more likely that favorable properties that afford BBB penetrance can be engineered into small molecules. Although oligonucleotide therapeutics can be delivered directly to the central nervous system via intrathecal injection, as is the case with Nusinersen,¹⁵ repeated intrathecal injections can be painful, and oral delivery of therapeutics, which can be achieved with small molecules, may be favored. Furthermore, small molecules may also be favored from a metabolic standpoint, as ASOs are prone to instability due to nuclease degradation.⁴¹

Oligonucleotides are advantageous from a design perspective, as one simply uses Watson–Crick base pairing rules to target the RNA of interest, although the structure of the target region must be considered (unstructured being ideal). More recently, a platform that designs small molecules based on the RNA’s sequence and a fundamental understanding of RNA-small molecule interactions has been developed.³⁴ The hope is that this scalable platform can be generalized and streamline the drug discovery process for small molecules targeting RNA. Design of oligonucleotides as well as small molecules is certainly less time-consuming and cost-intensive than high-throughput screening, whether a targeted approach for a particular RNA or a phenotypic screen. Developing selective small molecules that target RNA has long been challenging. However, recent studies have shown that small molecules can be more than or at least as selective as their oligonucleotide counterpart,⁴² long considered a benchmark by which to compare small molecules. Furthermore, small molecule therapeutics can have fewer side effects than ASOs, as ASOs can cause nonspecific stimulation of the immune system,⁴³ and multiple clinical trials of ASO therapeutics have been halted due to reports of severe thrombocytopeonia.⁴¹ Despite their advantages and disadvantages, both ASOs and small molecules offer significant promise in treating incurable neurological disorders.

Neurological RNA targets and (potential) therapeutic modalities

Modulating tau pre-mRNA splicing

One class of neurodegenerative diseases in which RNA has been exploited as a therapeutic target is tauopathies, which are characterized by deposits of aggregated tau proteins in the brain.⁴⁴ There are currently no known cures for the tauopathies.⁴⁴ One tauopathy, FTDP-17, presents with cognitive impairment, behavioral changes, and motor symptoms⁴⁵ and is caused by a mutation in the gene MAPT (encodes the microtubule-stabilizing tau protein) located on chromosome 17.^44,46 Alternative splicing of the MAPT pre-mRNA produces six protein isoforms in the human brain.^44,46 The exclusion or inclusion of exon 10 results in the formation of protein with three or four microtubule binding repeat domains, respectively.^44,46 In normal human brains, the ratio of three repeat (3R) to four repeat (4R) tau is approximately 1.⁴⁴ In FTDP-17, increased exon 10 inclusion leads to excess 4R tau and the manifestations of disease.^44,46 Although the exact mechanism by which alterations in the 3R-to-4R ratio lead to disease has not been fully elucidated, it is known that 4R tau binds microtubules with greater affinity than 3R tau, owing to its additional microtubule-binding domain, and is more prone to aggregation.^44,46,47 Various mutations at the exon 10-intron 10 junction, such as G(+3)A, C(+14)U, and C(+16)U, lead to increased exon 10 inclusion, each weakening the stability of a hairpin structure that regulates alternative splicing (Fig. 4) 2,3,48,49 In particular, the C(+14)U mutation converts a GC pair to a thermodynamically less stable GU pair and causes the disinhibition-dementia-parkinsonism-atrophy (DDPAC) disorder.^2,3 This destabilization results in increased availability of the 5′ splice site to U1 small nucleolar ribonucleoprotein (snRNP).^2,3,49 This misregulated splicing event may be corrected by inhibiting U1 snRNP binding to the pre-mRNA with ASOs⁵⁰ or small molecules³⁶ to reduce the frequency that exon 10 is included and hence expression of 4R tau.

Figure 4.

Single mutations in an RNA structure at the MAPTexon 10-intron junction cause tauopathies. (A) The sequence and secondary structure of the exon 10–intron 10 junction of MAPT pre-mRNA, where blue nucleotides correspond to nucleotides found in exon 10. Mutations that destabilize the splicing regulatory element (SRE) hairpin at the exon 10–intron 10 junction are indicated by arrows pointing to the base that is mutated. The splice site is labeled with a red arrowhead. Nucleotides shown in red represent binding sites of the ASO splicing modulator.⁵⁰ (B) Destabilization of this hairpin enhances exon 10 inclusion via increased accessibility and binding to U1 snRNA. (C) Structures of small molecules that bind mutant MAPTpre-mRNAs and direct its alternative splicing toward exon 10 exclusion. MTX (left) binds to the tau SRE⁵³ and small molecules identified by Inforna³⁶ (middle and right) that bind to the tau SRE and reduce MAPT exon 10 inclusion in cells.

Indeed, a 2′-O-methoxyethyl phosphorothioate (2′O-MOE-PS) ASO reduced 4R tau levels in a human tau (htau) mouse model.⁵⁰ The ASO targets an intron splicing silencer (ISS) and an adjacent intron splicing modulator within the splicing regulatory element (SRE) hairpin, inhibiting selection of the 5′ splice site.^49,50 Varani and coworkers reported a nuclear magnetic resonance (NMR) structure of a complex between the tau SRE and neomycin, which thermodynamically stabilized the hairpin; however, the compound lacked therapeutic selectivity.⁵¹ Donahue et al. developed a high-throughput binding assay using a fluorescent probe to screen for ligands that bound to the tau SRE, identifying the anticancer drug mitoxantrone (MTX, Fig. 4C) from a library of ~110,000 compounds.⁵² They later reported an NMR structure of the compound in complex with the tau SRE, in which MTX stacks between the GC pairs that close the A bulge and displaces the A bulge from the helix.⁵³

Using Inforna, small molecules were identified that bound to the DDPAC mutant tau SRE.³⁶ Three compounds were identified that bind to the A bulge motif with affinities ranging from 10 to 50 micromolar.³⁶ Two of these small molecules reduced the ratio of 4R (exon 10 inclusion) to 3R (exon 10 exclusion) mRNA and protein in a cellular model at micromolar concentrations (Fig. 4C).³⁶ The most potent small molecule thermodynamically stabilized the DDPAC tau SRE, suggesting that small molecule stabilization of DDPAC tau can affect exon 10 inclusion and the production of 4R tau in cells.³⁶ Altogether, this study demonstrated a rational method for identifying a bioactive compound that could serve as leads for FTDP-17 therapeutic intervention.

Modulating SMN pre-mRNA splicing

Another neurological disorder which involves aberrant pre-mRNA splicing is SMA, a neurodegenerative disease characterized by loss of motor neurons in the anterior horn of the spinal cord, leading to muscle atrophy and weakness.⁵⁴ Humans carry two nearly identical forms of the survival motor neuron (SMN) gene, SMN1 and SMN2, which are located on chromosome 5q13, and differ by a single nucleotide in the coding region.^4,54 While SMN1 encodes full-length SMN protein,^4,54 a translationally silent C-to-U transition at position 6 of exon 7 in SMN2 increases exclusion of exon 7 in 90% of transcripts^4,54 and produces a shortened SMN protein isoform that is functional but rapidly degraded.⁵⁴ SMA is caused by homozygous loss of SMN1, and SMN2 produces insufficient levels of functional protein to compensate for this loss.⁵⁴ Proposed mechanisms of the effect of the C6T mutation on SMN2splicing include weakening of an SF2/ASF-dependent splicing enhancer, creation of a heterogeneous nuclear ribonucleoprotein (hnRNP) A1-dependent splicing silencer, and strengthening of an inhibitory stem-loop structure at the 3′ splice site of exon 7.^4,55 Thus, a therapeutic strategy is to regulate splicing of SMN2 exon 7, producing a full-length protein that substitutes for SMN1 (Fig. 5A).

Figure 5.

Loss of SMN1 causes spinal muscular atrophy (SMA). Fortuitously, humans have a related gene, SMN2, that could compensate for SMN1 loss. (A) A C-to-U substitution in SMN2 exon 7 leads to exon 7 exclusion and production of truncated SMN protein that has a shortened half-life as compared to SMN1. A therapeutic strategy for SMA therefore is to direct exon 7 splicing towards its inclusion with a small molecule, producing a full-length SMN2 protein, a functional equivalent for SMN1. (B) Indeed, small molecules have been discovered through phenotypic screening that increase exon 7 inclusion to produce full-length SMN2 protein. Their structures are shown here. (C) In vitro SHAPE-directed modeling of the RNA secondary structure of SMN2 exon 7. Blue-labeled residues, corresponding to nucleotides 68–73, represent the binding site for SMN-C2. PK4C9 interacts with blue-labeled residues corresponding to nucleotides 82, 99, and 100. Residues of the ISS-N1 region, which binds to Nusinersen, are labeled in red. The 5′ and 3′ splice sites are labeled with red arrowheads.

Nusinersen (SPINRAZA™) is an ASO drug recently approved for the treatment of SMA.¹⁵ Nusinersen targets intronic splicing silencer N1 (ISS-N1), a 15 nucleotide (nt.) RNA regulatory element located downstream of the 5′ splice site of exon 7.^56,57 The ASO blocks binding of hnRNP A1/A2 to two motifs in the region and disrupts the formation of inhibitory secondary structures that facilitate exon 7 skipping.^56,57 One of the structures, a terminal stem-loop structure (TSL2), sequesters the 5′ splice site of exon 7 and overlaps a cis- element (3′ cluster) near the 5′ splice site that regulates exon 7 splicing.^56-58 Two separate clinical trials of infants and children with SMA demonstrated improved motor function among patients who received Nusinersen compared to those who received a sham control.^15,59 The success of this ASO confirmed that altering splicing of SMN2 by targeting the RNA is a promising therapeutic strategy, and recently, small molecule drugs that alter the splicing of SMN2 pre-mRNA have also been explored.^37-39,60,61

Efforts to increase SMN protein using small molecules have relied on high-throughput screens to identify modulators of SMN production. Because these small molecules have been identified through phenotypic screening, their targets, and hence modes of action, had to be determined. Some of these small molecules act through inhibition of decapping enzyme scavenger (DcpS),⁶² an enzyme that functions in mRNA decay and modulates mRNA stability,^63-65 although the role of DcpS in SMA remains unclear.⁶⁶ Other small molecules directly alter SMN2 splicing (Fig. 5B)^37-39,60 or have mechanisms that have yet to be determined.^67,68

To identify small molecules that facilitate SMN2 exon 7 inclusion, Palacino etal. (Novartis) screened a small molecule compound library in a motor neuron cell line that expressed a SMN2 exon 7 minigene.³⁷ Although the hit rate was less than 1%, many of the hit compounds contained a pyridazine core that were subsequently optimized for potency.³⁷ Two compounds emerging from lead optimization, NVS-SM1 and NVS-SM2, increased SMN2 protein levels in SMA patient-derived fibroblasts and in the brain and spinal cord of an SMA mouse model.³⁷ NVS-SM1, which displayed efficacy at lower doses, was further evaluated in induced pluripotent stem cell (iPSC)-derived neurons, where it increased full-length SMN2 mRNA and SMN protein levels upon treatment.³⁷ In a long-term study with severe SMNΔ7 mice, NVS-SM1 increased body weight and survival as well as elevated SMN protein levels in the brain.³⁷ Mutational analysis, surface plasmon resonance, and NMR spectroscopy studies suggest that the NVS-SM1 acts by enhancing the association of U1 snRNP with the 5′ splice site of exon 7 in a sequence specific manner.³⁷ NVS-SM1 (branaplam) is currently undergoing clinical trials to treat SMA (NCT02268552). Altogether, this work demonstrates that small molecules can direct splicing events, including those that cause disease.³⁷

Another high-throughput screen for small molecules that promote SMN2 exon 7 inclusion was performed by Roche and PTC Therapeutics.³⁸ Further optimization of hit compounds led to SMN-C1, SMN-C2, and SMN-C3, which increased full-length SMN2 mRNA and SMN protein levels in SMA patient-derived fibroblasts and patient-derived iPSCs.³⁸ Analysis of splice junctions in transcripts from type 1 fibroblasts treated with SMN-C3 indicated that SMN-C3 selectively targeted exon 7 junctions.³⁸ Mutations to two splicing enhancers, TSL2 and ISS-N1, which affected exon 7 inclusion did not decrease SMN-C3 activity, indicating that the mutations did not affect RNA structures critical for compound activity.³⁸ In the more severe SMNΔ7 mouse model, SMN-C3 increased SMN protein levels in the brain and quadriceps muscle by up to 150% and 90%, respectively. SMN-C3, administered as multiple treatments, increased body weight, improved motor functions, and prolonged survival time.³⁸ A derivative of these splicing modulators, RG-7916 (risdiplam), is currently being investigated in three phase 2 clinical trials for SMA (NCT03032172, NCT02913482, and NCT02908685). Collectively, this study demonstrated that orally bioavailable compounds increased SMN2 exon 7 inclusion and SMN protein levels in patient-derived cells and SMA mice.³⁸

The cellular targets of the small molecule modifiers of SMN2 splicing discussed above were later elucidated in two different studies. Sivaramakrishnan et al. studied SMN-C5, a close analog of RG-7916 and determined that it binds two distinct sites in SMN2 pre-mRNA and stabilizes an unidentified ribonucleoprotein complex.⁶⁰ A different investigation into the mechanism of action used a derivative of SMN-C2 which was appended with a photocrosslinking probe and a biotin tag (SMN-C2-BD).³⁹ This probe was used in an RNA pull-down experiment, named Chemical Cross-Linking and Isolation by Pull Down (Chem-CLIP),⁶⁹ which showed that SMN-C3 bound to a purine-rich motif in exon 7, GAGGAAGA.³⁹ The putative binding site within this motif, AGGAAG, had the highest binding affinity for SMN-C2, a fluorescent analog of SMN-C3, compared to other sequences within exon 7 and adjacent intron 6 and 7.³⁹ In vitro and in cell SHAPE probing revealed that the AGGAAG motif is part of a helix within a stem-loop (TSL1) at the 3′ splice site (Fig. 5C).³⁹ Proteomic analysis using the photocrosslinking probe revealed that SMN-C2 binds to splicing activators FUBP1 and KHSRP and increases their binding affinity for exon 7.³⁹ Collectively, these two mechanistic studies suggest that these small molecule splicing modulators bind to the exon 7 AGGAAG motif and affect protein binding to modulate SMN2 splicing.

Although the small molecule splicing modulators discussed above were identified through phenotypic high-throughput screens, targeted screens have also been used to identify small molecule splicing modulators.⁶¹ Garcia-Lopez et al. reported a small molecule that targets the terminal step-loop 2 (TSL2) located at the exon7/intron 7 junction and increases exon 7 inclusion in SMN2.⁶¹ The group identified homocarbonyltopsentin (PK4C9) via a fluorescence displacement assay of 304 compounds for those that bound to and stabilized a tri-loop conformation of TSL2with improved accessibility of the 5′ splice site.⁶¹ Binding of PK4C9 was confirmed with 2-aminopurine labeled TSL2 RNAs.⁶¹ In a cellular model and patient-derived fibroblasts, PK4C9 increased SMN2 exon 7 inclusion and levels of SMN2 protein at 40 μM.⁶¹ The small molecule also increased SMN2 exon 7 inclusion in a transgenic Drosophila model.⁶¹ Mutations in SMN2 mini-genes that increased the strength of base pairing near the 5′ splice site decreased the effect of PK4C9 on SMN2 splicing, demonstrating that its activity is mediated by its interaction with TSL2.⁶¹ These data show that small molecules that bind to TSL2 and modulate SMN2 splicing can be identified through targeted screens.⁶¹ The identification of numerous small molecules that modulate SMN2 splicing shows promise for treating SMA and other diseases caused by aberrant alternative pre-mRNA splicing.

Small molecules targeting RNA repeat expansions

Another class of RNA structural elements that contribute to neurological disorders are RNA repeat expansions (or r(SEQ)^exp where “exp” denotes expansion), which include HD [r(CAG)^exp],⁷⁰ FXTAS [r(CGG)^exp],⁷¹ C9ORF72 ALS and frontal temporal dementia [c9ALS/FTD, r(G₄C₂)^exp],⁷² and spinocerebellar ataxia 10 [SCA10, r(AUUCU)^exp],⁷³ among others (Fig. 6A). How the expanded RNA repeats cause disease depends on both the sequence and its location. For example, in HD, r(CAG)^exp is located in exon 1 of huntingtin (HTT) mRNA.Since it is found in the coding region of the gene, the repeat is translated into a mutant HTT protein containing a toxic stretch of polyQ. Other RNA repeat expansions cause disease through translation of mutant proteins even though they are not found in coding regions. The r(G₄C₂)^exp in intron 1 of C9ORF72 and r(CGG)^exp in the 5′ untranslated region (UTR) of the fragile X mental retardation 1 (FMR1) gene undergo RAN translation.^9-12 Many expanded repeats also have gain-of-function mechanisms in which they bind and sequester regulatory proteins, such as muscleblind-like 1 (MBNL1) in the case of r(CAG)^exp74 and hnRNP K and other proteins in the case of r(CGG)^exp, r(G₄C₂)^exp, and r(AUUCU)^exp.^6,7 Sequestration of these proteins by the RNA repeat results in the formation of nuclear foci and pre-mRNA splicing defects as MBNL1 and hnRNP K are important regulators of pre-mRNA splicing. Because RNA is the central toxic agent in these neurological disorders, many efforts have been made to target them and inhibit disease pathobiology. Fortuitously, RNA repeats fold into hairpin structures containing repeating structural motifs that have allowed for the design of small molecules that selectively bind the three-dimensional fold of the disease-causing RNA and improve disease-associated defects.^75-81

Figure 6.

RNA repeat expansions cause or contribute to >40 microsatellite disorders, which are neurological and neuromuscular in nature. (A) Secondary structures of various RNA repeat expansions. r(CAG)^exp, which causes HD, forms repeating 5′-CAG/3′-GAC loops; r(CGG)^exp, which causes FXTAS, forms repeating 5′-CGG/3′-GGC loops; r(G₄C₂)^exp, which causes c9ALS/FTD, contains repeating 1 × 1 nucleotide G/G internal loops; and r(AUUCU)^exp, which causes SCA10, forms repeating 5′-AUUCU/3′- UCUUA loops. (B) Small molecules targeting RNA repeat expansions can be modularly assembled into dimeric compounds which simultaneously bind adjacent internal loops, increasing potency and selectivity. Such an approach is illustrated for r(CGG)^exp but has demonstrated for other repeat expansions. (C) Structures of small molecules targeting RNA repeat expansions.

The r(CAG)^exp repeat causing HD is one RNA repeat that has been targeted with small molecules. The HD repeat was one of the first genetic diseases where the mutational cause was assigned to a specific chromosome.^5,70 Despite the root cause of HD being known for over two decades, there are no known cures. Targeting the HD RNA offers a promising therapeutic option because r(CAG)^exp operates in more than one pathomechanism. The r(CAG)^exp is found in exon 1 of HTT mRNA and thus is translated into a toxic polyQ-containing HTT protein, which causes neurodegeneration and other defects.⁸² The RNA repeat is directly implicated in the translation of this toxic protein; however, r(CAG)^exp contributes to HD disease progression through other mechanisms. For example, longer repeats of r(CAG)^exp sequester regulatory proteins such as MBNL1, leading to splicing defects in its natural substrates.⁷⁴ Furthermore, the mutant HTT mRNA itself can be aberrantly spliced producing a mini-HTT mRNA, which contains r(CAG)^exp-containing exon 1 and a portion of intron 1 containing a cryptic poly(A)-site.⁸³ This aberrant splicing is due to the sequestration of serine-arginine splicing factor 6, and the aberrantly spliced mini HTT mRNA has implications in disease progression.⁸³ Thus, targeting this RNA repeat with small molecules may improve disease phenotype by multiple mechanisms.

The r(CAG)^exp folds into a hairpin structure with a repeating array of 1 × 1 nucleotide A/A internal loops, providing binding pockets for small molecules. Indeed, small molecules that bind r(CAG)^exp have been identified through chemical similarity searching and virtual screening followed by binding measurements.⁸¹ These studies identified a small molecule that binds to r(CAG)^exp, displaces MBNL1, and improves pre-mRNA splicing defects that are regulated by MBNL1 in an HD-patient derived cell line.⁸¹ Although this compound was not studied for improving other modes of r(CAG)^exp-mediated toxicity, it demonstrated that small molecules can improve HD-associated defects. ASOs that target r(CAG)^exp, usually around 20 nucleotides in length,⁸⁴ are also being investigated, further demonstrating the utility of targeting the toxic RNA species in HD.^85-88

Another RNA repeat that contributes to neurological disease is r(CGG)^exp, which causes FXTAS and is found in the 5′ UTR of FMR1.⁷¹ This RNA repeat, which folds into a hairpin structure containing repeating 1 × 1 nucleotide G/G internal loops, causes toxicity via two mechanisms: (1) sequestration of regulatory proteins such as DiGeorge Syndrome critical region 8, Src-associated substrate during mitosis of 68 kDa (Sam68), and hnRNP, resulting in pre-mRNA splicing defects and the formation of nuclear foci (i.e., gain-of-function);⁶ and (2) producing toxic polymeric proteins via RAN translation.^9,10 Multiple small molecules have been designed that bind the repeating three-dimensional structure of r(CGG)^exp and improve disease phenotypes (Fig. 6A). One of these small molecules is a hydroxyellipticine derivative that was modularly assembled to form a potent, dimeric compound (contains two RNA-binding modules such that two 1 × 1 nucleotide G/G internal loops are bound simultaneously; Fig. 6B and C).^76,80 The dimeric compound preferentially binds to 1 × 1 nucleotide G/G internal loops (K_d = 50 nM) over base-paired RNAs, as saturable binding to fully base paired RNAs is not observed ⁷⁶This small molecule completely restored Sam68-dependent splicing defects in a model cell line and inhibited RAN translation of r(CGG)^exp without affecting canonical translation of the message.⁷⁶

The design of another dimeric small molecule with bis-benzimadzole RNA-binding modules was also informed by Inforna.⁷⁹ This small molecule improves Sam68-dependent splicing defects at micromolar concentrations.⁷⁹ Appending the bioactive compound with a cross-linking module to covalently modify the RNA and a biotin purification module (1) afforded a more potent compound that improved Sam68-dependent splicing defects and inhibited RAN translation of r(CGG)^exp;⁷⁷ and (2) enabled the isolation of cross-linked cellular targets via Chem-CLIP, confirming that r(CGG)^exp is indeed the target of the small molecule.⁷⁷ Interestingly, a 2′-O-methyl (2′OMe) modified ASO with a phosphorothioate backbone that targeted r(CGG)^exp, 2′-OMePS-(CCG)₁₂, inhibited both RAN and canonical translation, suggesting that small molecules may offer a more selective approach to modulating FXTAS biology.⁷⁹ Collectively, these studies show that small molecules can improve multiple aspects of disease by binding to r(CGG)^exp and can be lead optimized to more potently and selectively target this disease-causing RNA.

The RNA repeat that causes the most common genetic form of FTD and ALS (r(G₄C₂)^exp) shares a structure related to that of r(CGG)^exp, namely the GG internal loops.⁷⁸ The r(G₄C₂)^exp in intron 1 of the gene C90RF72 causes toxicity through sequestration of regulatory proteins resulting in pre-mRNA splicing defects,⁷ nuclear trafficking defects,^89,90 and producing toxic “c9RAN proteins” that form neuronal inclusions throughout the central nervous system.^11,12 Since r(G₄C₂)^exp and r(CGG)^exp contain structurally similar 1 × 1 G/G internal loops, the hydroxyellipticine compound that improved r(CGG)^exp-related disease was used to identify structurally related compounds that were then tested for improving ALS-associated defects (Fig. 6A).⁷⁸ This study identified three small molecules that inhibited RAN translation and decreased the number of r(G₄C₂)^exp-containing nuclear foci.⁷⁸ Importantly, one small molecule decreased the production of c9RAN proteins and nuclear foci in induced neurons (iNeurons) derived from c9ALS/FTD patients.⁷⁸ Small molecule inhibition of ALS-related defects in patient-derived cells represents an important avenue to develop chemical probes and lead therapeutics for ALS/FTD. ASOs that target C9orf72 have been developed and are also being investigated in the clinic.⁹¹

The RNA repeat that causes spinocerebellar ataxia 10 (SCA10) has also been targeted with small molecules (Fig. 6).⁷⁵ SCA10 is caused by r(AUUCU)^exp in intron 9 of the ataxin 10 (ATX 10) mRNA and forms nuclear foci with proteins such as hnRNP K.^73,92 This RNA repeat also results in translocation of protein kinase C-δ (PKCδ) in mitochondria, resulting in dysfunction and activation of caspase 3 and induction of apoptosis.⁹² This disease-causing RNA folds into a secondary structure with repeating 3×3 pyrimidine-rich UCU/UCU internal loops flanked by AU pairs.⁹³ To modulate SCA10-related disease, small molecules that bind to the flanking AU base pairs were identified via chemical similarity searching and high-throughput screening.⁷⁵ This small molecule was used to synthesize a multivalent compound that binds multiple adjacent AU pairs.⁷⁵ In SCA10 patient-derived fibroblasts, this rationally designed small molecule reduced the number of nuclear foci at nanomolar concentrations.⁷⁵ The small molecule also reduced the amount of PKC-δ that is translocated to the mitochondria and decreased caspase-3 activity.⁷⁵ Small molecules that bind to RNA base pairs therefore could offer therapeutic benefit for SCA10 and other diseases, provided that the RNA-binding modules are placed at the exact distance that separates the pairs to be targeted to garner selectivity. A variety of small molecules that bind selectively to RNA base pairs have recently been identified and can allow for the design of chemical probes to target these disease-causing RNAs.⁹⁴

Conclusions

It is clear that RNA structure and function contribute to the progression of neurological diseases. The small molecules discussed herein represent lead compounds that can improve phenotypic changes observed in RNA-mediated disorders. The small molecule splicing modulators of SMN2 show significant promise for treating SMA, providing proof-of-principle that other small molecules could be discovered or designed that modulate aberrant alternative pre-mRNA splicing associated with diseases, including tauopathies. Furthermore, the small molecules that modulate RNA repeat expansions can be further optimized as lead therapeutics. As the roles of RNA in disease continue to expand, the development of novel RNA-targeting small molecules will become increasingly important to provide novel chemical probes of function or lead therapeutics for neurological disorders. There is much to learn about how to develop therapies that target disease-causing RNAs. As various compounds progress into preclinical and clinical trials, the emerging information set will hopefully accelerate advancements to treat the most serious of currently untreatable maladies.