How splice-switching ASOs work

Antisense oligonucleotides are a small molecular class that does a lot of work. Nusinersen, the spinal muscular atrophy drug Biogen markets as Spinraza, is an antisense oligonucleotide. So is milasen, the drug Timothy Yu's lab made for one child in Colorado. The two molecules are roughly the same size, share the same backbone chemistry, and bind their RNA targets by the same mechanism. They differ in the 18-to-25 letters of nucleotide sequence that tell each one where in the genome to act. The molecule is the platform. The sequence is the address.

The shared chemistry is what makes the second sentence true. Once a researcher has a chemistry class that survives in cerebrospinal fluid for months, binds RNA tightly enough to displace splicing factors, and produces a known set of side effects at known doses, the rest of the work is finding the address. That is the central thing to understand about why the field has accelerated so quickly since nusinersen's 2016 approval.

The chemistry stack

A clinical-grade splice-switching antisense oligonucleotide is a short strand of synthetic DNA-like nucleic acid carrying three layers of chemical modification.

The first layer is the phosphorothioate backbone. In natural DNA, the phosphate groups that link nucleotides each have one non-bridging oxygen. In a phosphorothioate ASO, that oxygen is replaced with sulfur on every linkage. The substitution is small in atomic terms and large in pharmacological terms. Phosphorothioate ASOs survive nucleases that would shred natural DNA in minutes, circulating for hours to days; they bind serum proteins, which extends tissue residence time; and they are taken up by most cell types through receptor-mediated endocytosis without a delivery vehicle. The phosphorothioate backbone is the modification that turns an oligonucleotide from a research reagent into a drug.

The second layer is sugar modification at the 2' position of the ribose. Natural RNA has a hydroxyl at 2'. Natural DNA has hydrogen. Therapeutic ASOs typically carry a 2'-O-methoxyethyl group (2'-MOE), which Stanley Crooke's group at Ionis Pharmaceuticals settled on in the 1990s after evaluating dozens of alternatives. The 2'-MOE modification adds nuclease resistance beyond what phosphorothioate alone provides, raises the binding affinity of the oligonucleotide for its target RNA by several degrees of melting temperature per nucleotide, and reduces non-specific protein interactions. Other 2' modifications in clinical use include 2'-O-methyl, 2'-fluoro, and the constrained-ribose locked nucleic acid (LNA) class. A separate chemistry class, the morpholino oligomer (PMO), replaces the entire ribose-phosphate backbone with a charge-neutral synthetic equivalent.

The third layer is sequence design. The 18-to-25-letter sequence is what selects the target, and most of the engineering work in a new ASO program goes into picking the sequence that binds the right RNA region with the right affinity and specificity. Sequence design is computational at first and empirical at the end: candidate sequences are ranked by predicted binding energy and off-target risk, then synthesized and tested in cells.

Four mechanisms

Antisense oligonucleotides act through four distinct mechanisms, each tied to a chemistry choice. The Shen and Corey 2018 review in Nucleic Acids Research lays out the framework that the field still uses.

RNase H cleavage is the mechanism for "gapmer" designs that have a central stretch of unmodified DNA flanked by 2'-MOE wings. The DNA stretch in the middle, when bound to a target RNA, recruits the cellular enzyme RNase H, which cleaves the RNA strand of the duplex. The result is a knockdown of the target transcript. Mipomersen for familial hypercholesterolemia (FDA approval 2013) and inotersen for hereditary ATTR amyloidosis (2018) work this way.

Splice modulation, the mechanism most relevant to the n-of-1 cases, is what happens when an ASO binds pre-mRNA at a splicing-control site without recruiting RNase H. A fully 2'-modified ASO, with no DNA gap, cannot recruit RNase H. The ASO sits on the splicing factor's binding site and blocks it. The cell's splicing machinery, deprived of the signal, includes or excludes a different exon than it otherwise would. Nusinersen works this way: it binds an intronic splicing silencer in SMN2 pre-mRNA, suppresses the silencer, and shifts the cell toward including exon 7 and producing functional SMN protein. Eteplirsen for Duchenne muscular dystrophy (2016) works similarly through a morpholino that promotes skipping of exon 51 in dystrophin pre-mRNA. Milasen works similarly: it masks a cryptic splice site introduced by a transposable-element insertion in MFSD8, restoring normal splicing.

Translation block is direct steric interference with ribosome assembly. The ASO binds near the start codon or 5' untranslated region of an mRNA and prevents the small ribosomal subunit from initiating translation. This mechanism is less common in approved drugs.

Occupancy without cleavage describes ASOs that bind microRNAs or non-coding RNAs and prevent them from doing their normal job. Anti-miR therapeutics fall in this category.

The point of the four-mechanism map is that the chemistry choices and the sequence choices together determine which mechanism the molecule will use. A program targeting an aberrant splice site needs a fully 2'-modified ASO without a DNA gap. A program targeting a toxic transcript for knockdown needs a gapmer. The choices are made early.

The Krainer and Bennett collaboration

Adrian Krainer at Cold Spring Harbor Laboratory and Frank Bennett at Ionis Pharmaceuticals published a series of papers between 2008 and 2014 establishing the mechanism, the chemistry, and the clinical pathway for nusinersen. Krainer's group identified the splice-modulation mechanism in SMN2 and screened candidate sequences. Bennett's group at Ionis manufactured the candidates with the 2'-MOE phosphorothioate chemistry and ran the toxicology and clinical trials. The 2010 paper by Hua, Sahashi, Krainer, Bennett, and others in Genes and Development, demonstrating systemic correction of SMA in mice with the lead candidate, was the proof of concept that produced the drug.

What made nusinersen the platform precedent was not just that it worked. It worked at intrathecal doses below 12 mg, in children, with a safety profile that the 18-month placebo-controlled phase 3 trial of 121 infants confirmed clearly enough that the trial was stopped early for benefit. The FDA approved the drug in December 2016. The same chemistry, the same intrathecal route, and the same dosing schedule could now be applied to other splice-switching targets without re-establishing the safety profile of the route or the chemistry class. Milasen was dosed intrathecally less than two years later.

What makes a mutation ASO-amenable

Not every ultra-rare mutation has an antisense oligonucleotide solution. The mutation has to produce its disease through a mechanism the chemistry can address. The clearest cases are splice-site disrupting mutations and intronic insertions that introduce cryptic splice sites: an ASO can mask the cryptic site and restore normal splicing. Mutations that reduce expression of a needed protein can sometimes be addressed by ASOs that suppress an antisense RNA or a degradation signal, raising expression of the functional copy. Mutations producing a toxic gain-of-function protein can be addressed by gapmer ASOs that knock down the toxic transcript.

Mutations that produce a misfolded protein with no mechanistic toehold for an ASO are harder. Missense mutations in a coding region, where the protein is translated but defective, are usually outside the ASO playbook unless they happen to disrupt splicing.

The other constraint is tissue access. Intrathecal delivery reaches the central nervous system, which is why most of the n-of-1 ASO cases have been pediatric neurological conditions. Intravitreal injection reaches the retina, the route used for ASO programs in inherited retinal degeneration. Subcutaneous and intravenous routes reach the liver well, which is why ATTR amyloidosis and familial hypercholesterolemia were the first non-neurological ASO indications. Skeletal muscle, heart, and kidneys are less accessible. Conjugating an ASO to a targeting ligand, such as the GalNAc moiety that directs molecules to hepatocytes, is the active engineering frontier for non-CNS, non-eye indications.

Why the platform compounds

The reason the splice-switching ASO has become the workhorse molecule for n-of-1 rare-disease programs is that its design parameters generalize. A sponsor-investigator designing an ASO for a new mutation does not have to redo the toxicology of the chemistry class. The FDA's nonclinical guidance for individualized ASOs, issued April 2021, allows a single three-month rodent study if the chemistry and route are sufficiently precedented. The CMC guidance, issued December 2021, allows a simplified manufacturing package for known chemistries. The clinical guidance, issued July 2022, describes how to dose-escalate and monitor for class-known adverse events such as intrathecal-route hydrocephalus and CSF pleocytosis.

The next-generation chemistry frontier is already visible. Salanersen, an ASO Bennett's group at Ionis described in a 2025 preprint with a 2'-O-(N-methylacetamido) ribose modification, was three to four times more potent than nusinersen against the same SMN2 target in human SMN2 transgenic mice. The same kind of incremental chemistry improvement is likely to produce drugs at lower doses with fewer route-class adverse events over the next decade.

The molecule is the platform. The sequence is the address. The infrastructure that lets the next ASO program design its sequence faster, manufacture it at smaller batch size, run shorter toxicology, and dose its first individual on a faster schedule is what the FDA guidance series is trying to formalize, and it is what the cases keep proving works.