Antisense-mediated splice intervention to treat human disease: the odyssey continues

Introduction

The integrity of all gene expression is reliant on RNA, and the property of nucleic acids that permits complementary base pairing has long been an attractive option to alter gene expression by modulating RNA structure, function or abundance to treat human disease. Advances in the field have been predicated upon development of synthetic nucleic acid analogues (for review, see 1) that resist nuclease degradation and can be modified to evoke specific mechanisms. An antisense (the reverse complement) nucleic acid sequence (oligonucleotide) to a region of the target molecule can be designed and evaluated in a routine manner to identify compounds with the required specificity and affinity to alter the abundance, structure or functionality of the resulting transcript (Figure 1). Such outcomes have far-reaching potential to alter the course of human diseases, and at this time, the indications are mostly rare inherited conditions with otherwise limited treatment options. The field has attracted some debate regarding the high treatment costs, clinical trial and regulatory evaluation outcomes and approvals, some uncertainty regarding mechanisms of action¹ and emerging reports of off-target effects^2,3. Although only a small number of antisense drugs are in the clinic⁴ or undergoing translation at this time, ongoing efforts to improve oligonucleotide drug delivery and efficacy and to better understand antisense mechanisms will drive further advances in the field.

Figure 1. Currently approved antisense oligonucleotide drugs, indicating chemistry and mechanisms of action.

Antisense strategies that induce transcript degradation (left panel) include RNaseH degradation of the target transcript, activated by annealing of a complementary oligodeoxynucleotide (A) and transcript degradation by small interfering RNA (siRNA) (B). Transcript modification by RNA analogues (right panel) can be achieved by targeting splice motifs to exclude an exon to alter the reading frame or remove exons carrying mutations (C) or strengthen exon selection, otherwise compromised by base changes (D). 2′-O-MOE, 2′-O-(2-methoxyethyl); DMD, dystrophin; ISS-N1, intron splice silencer N1; PMO, phosphorodiamidate morpholino oligomer; PO, phosphodiester; PS, phosphorothioate; RISC, RNA-induced silencing complex; SMN2, survival motor neuron 2.

Natural nucleic acids are biologically labile and their application in drug development has required chemical modifications of the bases and nucleic acid backbone to increase resistance to endogenous nucleases and influence activity. The phosphorothioate backbone is the most widely applied chemical modification (Figure 2), and although the substitution of sulphur for a non-bridging oxygen confers nuclease stability, changes to the ribose moiety can confer RNA-like characteristics and hydrophobicity to the oligonucleotide (for example, 2′-O-methyl, 2′-O-methoxyethyl and 2′-O-fluoro). Phosphorothioate oligonucleotides carry a strong negative charge, allowing binding to various proteins and complex formation with cationic liposomes for efficient transfection in vitro. Other nucleobase and backbone modifications can confer specific characteristics and mechanisms of action, and include peptide nucleic acids and phosphorodiamidate morpholino oligomers (PMOs) (Figure 2) that carry a neutral charge but show inefficient cellular uptake (for review, see 5).

Figure 2. Natural and chemically modified nucleotides.

Upper panel: DNA, RNA and phosphorothioate (sulphur substituted for a non-bridging oxygen)-modified nucleotide (PS DNA). Lower panel: 2′-O-methyl phosphorothioate (2′-OMePS), 2′-O-methoxyethyl phosphorothioate (2′-O-MOE-PS) and phosphorodiamidate morpholino oligomer (PMO).

The more widely exploited antisense strategies use DNA analogues that activate RNase H1 degradation of the target transcript (Figure 1), and thereby downregulate gene expression, or double-stranded RNAs to induce gene silencing, comprehensively reviewed by Shen and Corey⁶. The concept of altering splice site selection to modify disease-causing transcripts was validated in vitro over 25 years ago⁷ and requires the design and application of an RNA analogue that targets motifs participating in exon selection and retention in the mature mRNA. Antisense targeting of splice sites or motifs necessary for exon selection can influence endogenous alternative splicing, block aberrant splicing, exclude a cryptic exon or an exon carrying a disease-causing mutation, or restore the open reading frame around a frame-shifting deletion, whereas targeting splicing silencers can enhance selection of an exon, otherwise compromised by nucleotide changes (for reviews, see 8,9).

At this time, two mRNA splice-modulating drugs are reported to have delivered therapeutic benefit to a subset of patients with Duchenne muscular dystrophy (DMD)¹⁰ and to patients with spinal muscular atrophy (SMA)¹¹. Eteplirsen (Exondys51), Sarepta Therapeutics, Cambridge, MA, USA) is an antisense PMO that targets splice enhancer motifs in the DMD pre-mRNA to exclude exon 51 and restore the dystrophin mRNA reading frame, disrupted by most deletions beginning at exon 52 or ending at exon 50. Eteplirsen received accelerated approval from the US Food and Drug Administration (FDA) in September 2016 but is not yet approved in Europe. The drug nusinersen (Spinraza ^®, Biogen, Cambridge, MA, USA) for the treatment of SMA is a 2′-O-methoxyethyl antisense oligonucleotide (AO) on a phosphorothioate backbone. Nusinersen targets a splice silencer (ISS-N1) in survival motor neuron 2 (SMN2) intron 7 and promotes exon 7 selection and retention during pre-mRNA splicing¹² and was approved by the FDA in December 2016. The roughly 25-year history of antisense drug development parallels that of an increasing interest in therapeutics for rare diseases. Of patients with any of the more than 6,000 rare diseases described to date, only a few groups of patients benefit from an effective treatment. It is perhaps fitting that the first splice intervention antisense therapeutics to reach the clinic address two of the more common and debilitating of the life-limiting inherited rare diseases of childhood. This review focuses on the development and current status of AO drugs that alter exon selection and exert a therapeutic effect by restoring or altering protein structure and function.

RNA splice-modulating drugs in the clinic

The field of RNA splicing therapeutics was launched by the demonstration of AO-mediated suppression of a disease-causing cryptic splice site in the β-globin transcript in a cell-free system⁷. This was followed by parallel programs to develop RNA therapeutics to alter exon selection during processing of the dystrophin (DMD) pre-mRNA and generate internally truncated but functional dystrophin isoforms that reflect variants found in the less severe allelic disorder, Becker muscular dystrophy (for review, see 13). Over 60% of the DMD-causing mutations are deletions of one or more exons, predominantly between exons 43 and 53 (major deletion hotspot) or exons 3 and 7 (minor deletion hotspot), that disrupt the translational reading frame and yield a prematurely truncated, non-functional protein. Becker muscular dystrophy is less common and is the consequence of mutations that yield a protein of reduced function or abundance or both. Becker muscular dystrophy mutations are generally in-frame deletions and provide “templates” for partially functional dystrophin variants that could ameliorate the severe dystrophic muscle phenotype characteristic of DMD. While the dystrophin exon-skipping programs aimed to develop antisense oligomer sequences that prevent exon selection during dystrophin pre-mRNA splicing (exon skipping) to generate internally truncated dystrophin molecules, two different RNA-like analogue chemistries were used for translational development.

Splice-modifying therapies under development

Pre-clinical and clinical studies on novel splice-modulating experimental drugs are gaining momentum. (For review of splice-modifying oligonucleotides under development, see 9.) Current strategies include dystrophin exon-skipping programs, exons 45 casimersen (SRP-4045) and 53 golodirsen (SRP-4053) by Sarepta Therapeutics (https://www.sarepta.com/pipeline/exon-skipping-duchenne); an unrelated dystrophin exon 53 program^29,30; exon-skipping strategies to treat dysferlinopathies (for example, limb girdle muscular dystrophy type 2B) by excluding exons 37 and 38 in a Dysf mutant mouse model; targeting cryptic exons activated by deep intronic mutations causing choroideremia³¹, Leber congenital amaurosis³² and USH2A-associated retinal degeneration³³; correction of mis-splicing of harmonium in USH1C and exon inclusion to address a common splice variant causing adult-onset Pompe disease³⁴. Two different exon-skipping strategies to treat spinocerebellar ataxia type 3 were recently reported. The pathogenic expanded polyglutamine repeat can be removed by skipping exon 10³⁵; however, in an innovative strategy, exons 8 and 9 were skipped in order to prevent proteolytic cleavage and generation of toxic protein fragments³⁶.

Though not widely exploited, modulation of endogenous alternative splicing to alter isoform ratios could have broad applicability to a number of acquired and inherited phenotypes. AO-mediated switching of Bcl-x splicing from the anti-apoptotic isoform to the pro-apoptotic isoform induced apoptosis of glioma cell lines³⁷ and of hepatic stellate cells in liver fibrosis studies³⁸. Endogenous aberrant splicing of GNB3 was corrected by AO-mediated splice intervention³⁹. The LMNA gene encodes two alternatively spliced products that are major components of the nuclear envelope. Mutations in the pre-lamin A transcript can result in aberrant splicing, causing Hutchison–Gilford progeria syndrome that can be addressed either by AO targeting of the cryptic splice site (for review, see 40) or by manipulating splicing of exon 11 to favour the lamin C isoform⁴¹.

Although development of novel molecular therapeutics is eagerly anticipated by patients and families and has instigated efforts to find new targets and intervention opportunities, current limitations to broader applicability need to be overcome, primarily inefficient delivery to target tissues. The charge-neutral PMOs have an excellent safety profile but are rapidly cleared from the circulation, showing poor uptake into cells. Pre-clinical development of exon-skipping PMOs in a mouse muscular dystrophy model showed moderate restoration of skeletal muscle dystrophin but minimal uptake in the myocardium⁴². Systemic delivery of these molecules can be greatly improved by conjugation to cell-penetrating peptides^43,44, the use of novel strategies, such as exploiting exosomes and targeting peptides⁴⁵, and gaining greater insights into the mechanisms of cell uptake⁴⁶. Successful translation of peptide-phosphorodiamidate morpholino drugs that are currently in development and clinical evaluation (for example, for DMD) may well begin a new era for splice-modifying therapies, particularly if tissue or cell targeting can be achieved, perhaps leading to reduced dosages.

Non-antisense and off-target effects of synthetic oligonucleotides

Oligonucleotide backbone and sugar modifications influence resistance to nucleases, mechanism of action, tissue distribution and binding to various proteins but can also cause hybridization-dependent (off-target hybridization of RNase H1-competent AOs resulting in hepatotoxicity⁴⁷) and independent effects. Phosphorothioate oligonucleotides administered in vivo bind to serum and cellular proteins^2,48 (for review, see 49) and show good tissue uptake and distribution, in the brain in particular⁵⁰. However, in a systematic literature review of original trial reports, by van Meer et al.⁵¹, all studies involving phosphorothioate oligonucleotides administered subcutaneously report injection site reactions, while intravenous delivery was associated with thrombocytopenia. Phosphorothioate oligonucleotide sequences were reported to be immunostimulatory⁵² and to bind to platelets, eliciting strong platelet-activating effects⁵³, while intracerebroventricular injection of 2′-O-methyl phosphorothioate AOs upregulated immune system-associated genes⁵⁰. More recently, antisense drugs include modifications to enhance affinity (for example, locked nucleic acids or constrained ethyl substituted gapmers) that are reported to cause hepatotoxicity due to off-target RNA hybridization and downregulation^47,54. In an elegant study, Dieckmann et al.⁵⁴ showed that hybridization-dependent toxicity of RNase H1-competent phosphorothioate oligonucleotides can be mitigated by limiting the affinity of the oligonucleotide and improving target specificity.

The propensity for phosphorothioate oligonucleotides to cause injection site reactions^15,17, flu-like symptoms, and hepatotoxicity was well known prior to the drisapersen exon-skipping studies and was reported in patients who received the approved drug mipomersen (RNase H1-competent gapmer) for the treatment of homozygous hypercholesteremia^55,56.

Although the hepatotoxicity may be attributable in part to off-target hybridization, the exact cause of the injection site reactions remains unknown. Shen et al.⁵⁷ examined cells transfected with phosphorothioate oligonucleotides with various 2′ modifications. They found that these compounds bind a large number of intracellular proteins and can replace the architectural long non-coding RNA NEAT1 in subnuclear bodies called paraspeckles. The repertoire of proteins bound by phosphorothioate AOs and the effects of protein depletion or altered cellular localization are influenced by any 2′ modification⁵⁸. The most widely used modification, 2′-O-methoxyethyl, was found to have an acceptable tolerability profile and consistent behaviour between sequences³, whereas 2′-fluoro-modified phosphorothioate oligonucleotides disturb nuclear biology, impair cell proliferation and cause loss of cellular proteins⁵⁹.

We explored the interaction of 2′-O-methyl phosphorothioate AOs with nuclear proteins and report backbone-specific effects of modified oligonucleotides on subnuclear organelles, altered distribution of nuclear proteins, the appearance of novel large structured nuclear inclusions, modification of RNA processing and marked disturbance of the transcriptome in cultured cells. We speculate that some side effects and adverse events reported after clinical evaluation of phosphorothioate nucleic acid drugs may be mediated, at least in part, by non-specific interaction of nuclear components with the phosphorothioate backbone⁶⁰. The sequence-independent effects of phosphorothioate compounds indeed may enhance or confound apparent antisense effects⁶¹ as a consequence of extensive disturbance of nuclear proteins and organelles, particularly those critical for RNA processing and splicing. Strategies to reduce cellular protein binding by phosphorothioate AOs and mitigate toxic effects, thereby improving the therapeutic profile of these molecules by backbone⁶² and 2′-O-modifications⁶³ are important steps in guiding AO design. Other considerations that may warrant comprehensive investigation are whether phosphorothioate nucleotide degradation products are incorporated into cellular nucleic acids, a question that has been raised⁶⁴, and whether the immune activation by intra-cerebroventricular injection of a 2′-O-methyl phosphorothioate AO reported in mice⁵⁰ should be of concern in humans who receive similar compounds.

Molecular medicines in the future

In an era of promising gene replacement and gene correction strategies, exploiting novel recombinant viruses that can be selected and tailored for tropism to specific tissues, the rationale for pursuing transcript-modifying therapies might be questioned. Many of these emerging molecular and gene therapeutics target rare and life-limiting diseases caused by multiple gene variants that may show variable progression and phenotypes within one patient group and, in the event that more than one treatment modality is investigated, compete for the same restricted patient population in recruiting clinical trial participants.

However, consider the situation where different treatment modalities could be used in tandem. For example, an antisense strategy might be the only option following vector gene delivery if the virus results in seroconversion and precludes re-delivery. PMOs designed to alter exon selection are specific to the target transcript and, to the best of our knowledge, have not been reported to cause off-target effects on splicing and have no effect when delivered to cells lacking the target transcript. In the case of cryptic exon or aberrant splice site usage, antisense-mediated strategies can generate a normal product, unlikely to be immunogenic, that is expressed under endogenous, tissue/cell/developmental regulation. Gain-of-function mutations may be particularly amenable to splice interventions, whereas for loss-of-function mutations, splice variants can be designed to encode products that may exhibit reduced function but could ameliorate otherwise severe disease.

We conclude by acknowledging the contribution of the patients and the combined efforts of chemists, researchers and clinicians and those in the biopharmaceutical industry who championed these drugs and the insightful regulators who have yielded two splice interventions, now available to patients. Although there is general agreement that better drug uptake and greater efficacy are desirable, these drugs, by all accounts, are altering the course of previously untreatable disease. There is potential for numerous therapeutics in the “splice modulating” class. The lessons learnt over the course of translating eteplirsen and nusinersen and from the failure of drispersen will inform the development of emerging and future molecular medicines.

Abbreviations

AO, antisense oligonucleotide; DMD, Duchenne muscular dystrophy; FDA, US Food and Drug Administration; PMO, phosphorodiamidate morpholino oligomer; SMA, spinal muscular atrophy; SMN, survival motor neuron