Therapeutic oligonucleotides as genetic medicines: The current state and future perspective

Summary

Antisense therapeutic oligonucleotides are entering a transformative era in clinical development, with a total of 21 oligonucleotides approved and a robust pipeline poised for the coming years. These drugs have emerged as a cornerstone of genetic medicine, providing a programmable platform to modulate gene expression and treat otherwise untreatable disorders. This review covers the different mechanisms of action of these drugs and highlights the key advances in oligonucleotide chemistry and delivery that have supported the clinical successes witnessed in the last decade. It also provides a historical account and overview of the growing number of approvals and key phase 3 studies. Finally, it offers a perspective on the future of the field, initially focused only on rare diseases and now expanding to common disorders.

Graphical abstract

Godinho et al. provide an overview of nucleic acid therapies, a new class of programmable drugs that enable disease-modifying treatments by acting at the gene transcript level. The authors summarize key advances in chemistry and delivery and offer a perspective on the advances expected in the near future.

Introduction

Therapeutic antisense oligonucleotides are poised to become the third pillar of human medicine, alongside small molecules and biologics. These drugs usually consist of short pieces of chemically modified nucleic acids that target gene transcripts via Watson-Crick base pairing,¹ allowing the programmable and precise modulation of disease-related genes at their root cause. Oligonucleotides can also act in a “non-antisense manner,” where the tertiary structure of the oligonucleotide specifically binds and modulates proteins directly, as with aptamers. This review focuses exclusively on the chemistry and clinical advancements of oligonucleotides that act in the antisense manner, with emphasis on single-stranded antisense oligonucleotides (ASOs) and RNA interference (RNAi)-based drugs such as small interfering RNAs (siRNAs).

Oligonucleotide-based gene modulation generated early enthusiasm for its potential to reach targets considered undruggable by small molecules,² prompting a rapid, and premature, progression to clinical trials. The first-generation molecules showed poor stability, limited efficacy, and safety flags, eroding confidence in the field for years. Since then, heavier chemical modification has improved efficacy, durability, and safety while lowering effective doses,² and increasingly sophisticated ligand-based delivery now enables targeted uptake by hepatocytes in the liver.¹

Therapeutic oligonucleotides have now moved decisively past the proof-of-concept stage. Below, we review their mechanisms of action, the chemistry and delivery advances behind the past decade’s clinical successes, the expanding roster of approvals and ongoing phase 3 studies, and the field’s trajectory from rare into common disease.

Key features and mechanisms of action of therapeutic oligonucleotides

Most therapeutic oligonucleotides share a common molecular anatomy: a phosphate backbone composed of phosphodiester linkages, a sugar ribose, and a nitrous base. However, different antisense classes have distinct features and modes of action, which determine how they can be chemically modified (Figure 1) to improve stability and confer other drug-like properties.¹ Here, we summarize the key differences among modalities and explore their therapeutic applications.

Figure 1.

Mechanisms of oligonucleotide therapeutics

(A) Schematic illustrating the intrinsic flow of genetic information from DNA to RNA, to protein (not drawn to scale). Each step is heavily regulated, providing opportunities for therapeutic intervention.

(B) ASOs with gapmer (red) work via RNase H to mediate mRNA (blue) degradation.

(C) Steric blocker ASOs (green) prevent binding of proteins (e.g., splicing factors), to modulate pre-mRNA processing (e.g., to correct missplicing variants).

(D) siRNA duplexes load into the RISC complex where the passenger strand is cleaved (black), binds to, and blocks the translation of complementary mRNA (blue), decreasing protein expression.

(E) saRNA duplexes load into an RITA complex, which includes AGO2, and guides the RITA complex to a complementary promoter region, resulting in chromatin remodeling, recruitment of transcriptional machinery, and transcriptional activation.

(F) MicroRNAs (purple) are short RNAs that both bind to complementary mRNA sequences, inhibiting translation machinery, and load into RISC to function in an siRNA-like fashion resulting in RISC-mediated mRNA degradation.

(G) Anti-miRs (antagomiRs) (black) are short oligonucleotides that target the seed region of microRNAs, blocking microRNA-mediated regulation of RNA expression.

Asterisk denotes complementary binding. ASO, antisense oligonucleotide; mRNA, messenger RNA; siRNA, small interfering RNA; saRNA. small activating RNA; RISC, RNA-induced silencing complex; miRNA, microRNA; anti-miR, anti-microRNA; RITA, RNA-induced transcriptional activation complex. Figure created with BioRender.

ASOs are single-stranded 15- to 30-nucleotide (nt)-long molecules with a molecular weight of ∼6–8 kDa. These drugs may act by targeting RNA transcripts for degradation through recruitment of a highly conserved enzyme, RNase H (Figure 1B), or by sterically interfering with proteins involved in pre-messenger RNA (pre-mRNA) processing events, such as splicing (Figure 1C).

RNase H-dependent ASOs are an effective therapeutic approach for treating toxic gain-of-function diseasesby reducing harmful RNA in the nucleus and/or cytoplasm, consequently affecting target protein expression. RNase H-dependent ASOs require DNA or DNA-like residues to enable the formation of a DNA-RNA duplex recognized by the enzyme. Thereafter, the endogenous enzyme produces a cluster of cleavages on the RNA target. Although the first ASOs tested consisted of molecules that were all DNA (or modified DNA) residues throughout, it was soon understood that a central DNA gap was sufficient, but critical, to elicit gene silencing activity. All RNase H-dependent ASOs in the market today consist of gapmers with modified DNA at their core and flanked by RNA-like modifications that enhance binding to the target and increase ASO resistance to nucleases.⁴

ASOs can also function as steric blockers that bind to pre-mRNA and physically obstruct access of spliceosome components to induce the inclusion or exclusion of an exon in the mRNA (Figure 1C). This mechanism of action can be leveraged to restore production of a protein or to switch between protein isoforms.⁵ These steric blocker ASOs are more forgiving for chemical modifications, since they do not rely on enzymes for activity. In fact, some do not even have a phosphate backbone.

Like RNase H-dependent ASOs, siRNAs and microRNA (miRNA) mimics are being developed to treat gain-of-function diseases. These drugs leverage the endogenous RNAi pathway to downregulate gene expression post-transcriptionally (Figures 1D and 1F). Canonical siRNAs are ∼21 nt (nucleotides) long (∼15 kDa) and consist of fully complementary guide and passenger duplexes with a 19-nt paired region and 2-nt overhangs. miRNA mimics may contain hairpin loops and mismatches and require additional processing before becoming active.¹ Based on thermodynamic stability, the guide strand is loaded into an RNA-induced silencing complex (RISC), while the passenger strand is discarded. RISC contains the Argonaute 2 (Ago2) protein, which surveys the cytoplasmic transcriptome and induces a single cut on complementary mRNA (between positions 10 and 11 of the guide). siRNAs have high target specificity, whereas miRNAs regulate multiple targets via the 3′ untranslated region (3′ UTR). Because RISC acts catalytically rather than through the 1:1 binding of ASOs, a single RNAi trigger can direct the degradation of multiple mRNA copies, yielding greater potency and durability.¹

Although outside of this review’s scope, oligonucleotides are also being evaluated for therapeutic gene upregulation (Figures 1E and 1G). Double-stranded small activating RNAs (saRNAs) can form an RNA-induced transcriptional activation (RITA) complex that can increase the expression of target transcripts by base pairing with promoter regions (Figure 1E).⁶ Once an saRNA binds to a promoter sequence, a variety of enzymes are recruited to modify the chromatin structure and shift it to an active state. Therapeutic gene upregulation can also be achieved through antagomirs (Figure 1G), which directly interfere with endogenous post-transcriptional gene regulation. These single-stranded oligonucleotides bind to, and inhibit the action of, endogenous miRNAs, resulting in the de-repression of the target genes regulated by that miRNA.⁷

From chemical modifications to clinical success

Over years of research and clinical development, it became apparent that chemical modifications were crucial to give therapeutic oligonucleotide drug-like properties. Unmodified DNA and RNA are short-lived in circulation and prone to nuclease degradation and renal clearance,^8,9 and their negatively charged backbone limits membrane interaction, cellular uptake, and tissue distribution. A diverse toolbox of modifications has since been developed to enhance the stability, potency, durability, distribution, and safety (Figure 2).

Figure 2.

Schematic of oligonucleotide structure and modifications

(A) Example ASO (top) and siRNA (bottom) chemical modification pattern with popout highlighting the location of modifications including backbone (orange) and ribose modifications, including those commonly used in siRNAs (blue) and those commonly used in ASOs (light blue).

(B) Key for schematic in (A).

(C) Phosphorothioate backbone and extended nucleic acid (exNA) modifications.

(D) Ribose modifications commonly used in siRNAs including 2′-flouro (F) and 2′-O-methyl (OMe).

(E) Ribose modifications commonly used in ASOs including 2′-O-methoxyethyl (MOE), locked nucleic acids (LNAs), and constrained ethyl bridged nucleic acids (cEt).

(F) 5′-Phosphate and 5′-vinylphosphonate modifications (purple).

Figure created with ChemDraw.

Therapeutic oligonucleotides can be modified at different points within their structure: the phosphate backbone, ribose, or nitrogenous base. Alternatively, other moieties can be appended to the 3′- or 5′-end of the molecule to enhance activity or distribution. The selection of which chemical modifications to use and the overall chemical modification strategy depends on the oligonucleotide modality, target tissue, and disease indication. Below, we will review the modifications that brought the greatest clinical significance for each modality separately. Advances in delivery and trafficking will be addressed in a different section.

Modifying single-stranded oligomers: Steric blockers, RNase H-dependent ASOs, and gene expression enhancers

Single-stranded oligomers (gapmers and splice modulators) are more prone to degradation than double-stranded siRNAs and require extensive modifications for the stability, target affinity, and displacement of secondary structures or binding partners. While splice-modulating oligonucleotides can be fully modified, gapmers are constrained by the need for an unmodified DNA “gap” that forms the DNA:RNA heteroduplex required for RNase H recruitment. Below, we cover the modifications most commonly used in each ASO type.

Phosphorothioate (PS) backbone modifications (Figure 2C) are arguably one of the most widely applied and enduring oligonucleotide modifications.¹⁰ A PS modification is tolerated by RNase H, making it applicable to both classes of ASOs. The PS modification substitutes one of the non-bridging oxygens for a sulfur atom, improving stability against nucleases. Importantly, PS modification also plays a key role in protein binding, partially reducing immediate renal clearance, and mediating nonspecific tissue uptake. Unfortunately, the non-specificity of PS modifications reduces binding to the target RNA, requiring other modifications to compensate for this effect.

The use of PS linkages in the backbone gives rise to a mixture of diastereomers with the phosphorus acting as a chiral center. Work from the Stec lab has shown that nucleases interact with PS-modified ASOs in a diasterioisomeric-selective fashion,^11,12,13 with Rp diastereomers having higher binding affinity than Sp, but decreased nuclease resistance. Interestingly, ASOs that have been uniformly modified with all Rp or all Sp diastereomers are less effective than stereo-random PS oligomers.^11,12,13 As such, specific patterns of Rp/Sp modifications have been found to enhance the potency of gapmer ASOs¹⁴ and/or improve single-nucleotide polymorphism (SNP) discrimination.¹⁵ Systematic identification and selection of hyperfunctional stereoisomers is expected to allow for lower clinical doses by reducing the number of less-active isomers present. The utility of stereo-defined chemistries is being studied in the clinic by Wave Life Sciences and others for the treatment of Huntington’s disease (HD) and Duchenne muscular dystrophy (DMD).

Several ribose modifications have been used to improve the nuclease stability of ASOs further (Figure 2E), since PS modifications alone are not sufficient. In addition to improving stability, some of these modifications have also proven useful in enhancing potency by increasing binding to the RNA target. These modifications are generally introduced at the 2′ position of the sugar, and for gapmers, they are only used in flanking regions, as most are not tolerated in the central gap region of the RNase H-dependent ASOs. One of the first and most commonly used 2′-ribose modifications in ASOs is the 2′-methoxyethyl (2′-MOE), which is a derivative of the naturally occurring 2′-O-methyl (2′-Ome). 2′-MOE modifications enhance binding to the RNA target by increasing the melting temperature (∼1.7°C per modified nucleotide). 2′-MOE modifications also reduce immunogenicity of these oligomers.

Additional notable modifications are the locked nucleic acid (LNA) and its methylated variant constrained ethyl (cET), which link the 2′ oxygen to the 4′ carbon on the ribose ring. These modifications constrain the flexibility of the nucleotide, substantially increasing the Tm (∼4°C–8°C per modified nucleotide). These modifications are particularly useful in short oligomers (∼12–15 nt),¹⁶ since they enable strong binding to the target RNA, significantly enhancing potency.^17,18 However, LNA and cET are not tolerated by RNase H and, therefore, are only used in flanking regions for RNase H-dependent oligonucleotides (not the gap region).

Excessive LNA and cET modification use have been associated with liver toxicity, partly due to off-target effects on transcripts sharing sequence similarity with the target.^19,20 Several lead candidates containing a combination of LNA and cETs with other modifications (miravirsen and cobomarsen) have progressed to late-stage preclinical and clinical development. Other constrained nucleic acids, such as tricyclo DNA (tcDNA), consist of bicyclic sugar rings bonded to a cyclopropane ring forming a rigid three-ring structure that binds tightly to RNA targets (increase in Tm by 1°C per unit).^21,22 This chemical modification has recently advanced to a clinical trial as part of a steric blocker ASO for exon skipping strategy for the treatment of DMD (SQY51, SQY Therapeutics).

Fluoroarabinonucleic acid (2′ F-ANA) is a modification that confers a DNA-like structure to the single-stranded ASOs; it is compatible with RNase H, removing the requirement for a gap, and in combination with DNA, enables potent target downregulation.^23,24 Additional experimental chemistries include 3′-fluorohexitol nucleic acid (FHNA), 2-thiothymidine, and others that improve binding and selectivity against the RNA target and allow for discrimination based on SNPs.²⁵

Phosphorodiamidate morpholino oligomers (PMOs) consist of a nucleic acid analog that contains morpholine rings connected to a base and that uses phosphorodiamite linkages instead of phosphorodiester.²⁶ PMOs do not recruit RNase H and work exclusively via steric binding. PMOs are highly resistant to nucleases and have a neutral backbone, conferring longer plasma half-lives than negatively charged ASOs.²⁷ However, due to the lack of a PS backbone, bioavailability is poor, and upon systemic treatment, PMOs are rapidly cleared by the kidney. Currently approved steric blocker PMOs for exon skipping in eligible patients with DMD include eteplirsen, golodirsen, and casimersen from Sarepta and viltolarsen from NS pharma.

Modifying double-stranded oligomers: siRNAs and miRNA mimics

Lessons from chemically modifying ASOs accelerated RNAi drug development, though years of dedicated research were still needed to translate these modifications to double-stranded oligomers.

Synthetic siRNAs and miRNA mimics leverage the RNAi pathway to knock down the expression of defined RNAs. To harness this pathway, the guide strand needs to load into an enzymatic complex that subsequently surveys the transcriptome for the target. Thus, there are limitations on the type and location of chemical modifications that can be introduced in guide strands. Conversely, the passenger strand offers more flexibility for modification. In comparison to single-stranded ASOs, RNAi drugs require a lower relative number of modifications to significantly improve activity in vitro and in vivo.^28,29,30 In fact, in the early days of RNAi technology, minimal number of modifications were employed. Instead, there was heavy reliance on lipid/polymer formulations for delivery and protection against nucleases. The advent of conjugate-mediated delivery required more extensive modifications, and thus, the field has spent considerable time and effort optimizing chemical modification patterns to enable safe, potent, and durable activity. Today, the majority of products in the market and ongoing trials employ fully modified siRNAs, where every nucleoside is modified.

In siRNAs, PS modifications have been applied mostly at the terminal ends of the guide and passenger strands to improve stability against nucleases (Figure 2A). These modifications are generally not well tolerated in the seed region (2–8 nt from the 5′-end) of the guide strand,²⁹ as they substantially impact siRNA gene silencing activity. However, in some instances, PS interlinkages have been used in greater numbers in asymmetric siRNA constructs, specifically at the 3′-end of the guide strand to confer greater protein-binding ability.^9,31,32 Contiguous PS moieties have been shown to promote nonspecific protein binding, which can help delay clearance and improve tissue uptake and distribution. As with ASOs, most PS-modified siRNAs consist of a stereo-random isomer mixture, i.e., Rp and Sp at each position. Systematic evaluations of isomer effects in the context of typical 23(as)-22(s) siRNA duplexes have demonstrated that only a small number of isomers contribute to potency and durability,^33,34 suggesting opportunities for refinement.

Novel backbone modifications, such as extended nucleic acid (exNA) that consists of a methylene insertion between the 5′-C and 5′-OH of a nucleoside, significantly enhance resistance to nucleases (Figure 2C). The exNA modification, in combination with PS modifications, has shown to improve nuclease resistance up to 1,000-fold when compared to natural phosphodiester and 32-fold compared to stand-alone PS modification.³⁵ exNA modifications have rapidly progressed through the early stages of preclinical testing with encouraging safety and potency data for a variety of gene targets.^35,36 In fact, exNA modification has recently entered clinical trials in the context of a divalent-siRNA construct targeting PRNP in symptomatic patients with prion disease.³⁶

The most widely applied ribose modifications on siRNAs and miRNA mimics are 2′-OMe and 2′-F, which confer increased stability and potency (Figure 2D). However, careful consideration of modification positioning and pattern within the construct is required. The 2′-OMe modification is bulkier and hence, is less tolerated in certain positions of the guide strand (seed region), whereas 2′F resembles the natural hydroxyl in RNA, allowing for greater flexibility in guide strand design.³⁷ Although both modifications favor an A-form helical structure, they also introduce distortions to the duplex, with 2′-F allowing more stacking and 2′-OMe enabling less stacking of the bases.³⁸ Fully modifying guide strands with 2′-OMe renders them inactive, and fully modifying them with 2′-F results in poor gene-silencing performance.³⁷ Alternating 2′-OMe and 2′-F within the siRNA seems to provide a balanced combination that enables RISC loading and potent gene silencing.³⁰

Based on extensive research and mechanistic studies,^39,40 2′-OMe/2′-F patterns have been continuously refined to meet specific structural requirements for improved function. Optimized compositions allow for greater flexibility at the 5′-end of the guide strand, which is favorable for RISC loading; higher binding affinity within the seed region to facilitate target recognition and base pairing; as well as a region of lower affinity at the 3′-end of the guide strand to enhance cleavage rate.⁴⁰ Furthermore, introducing 2′-OMe modifications at specific positions allows for reduction of immunogenic effects observed with natural RNA^41,42 and also prevents RISC loading of the passenger strand.^43,44 This strategy is used in most advanced designs to mitigate potential off-target effects that may arise from unwanted passenger strand loading into Ago2.³⁹ Similarly, LNA, a constricted sugar modification described above, has also been used in siRNAs to increase binding and the thermodynamic stability of the duplex at the 5′-end of the passenger strand, blocking RISC loading.⁴⁵ LNA-modified siRNAs are RNAi competent and confer significant resistance against nucleases.⁴⁵ As with ASOs, a widespread adoption of this modification in the clinic has not materialized, most likely due to the liver toxicities observed with LNA-ASOs.

Finally, two other ribose modifications worth highlighting in this section are the unlocked nucleic acid (UNA) and glycol nucleic acid (GNA). UNAs consist of an RNA mimic where the ribose ring has been intentionally cleaved between the 2′ and 3′ carbons. As opposed to LNAs, UNAs confer greater flexibility to the guide strand and destabilize the duplex. Through careful consideration of positioning within the duplex, these modifications have been used to improve specificity against a given target.⁴⁶ Likewise, GNA, an acyclic glycol modification, is similarly used to improve potency and target discrimination, while reducing off-target liabilities by seed-pairing destabilization.^47,48,49,50 Both UNAs and GNAs have been the cornerstone of the development of allele-selective siRNAs targeting SNPs, where allele specific knockdown, with preservation of the wild-type allele, is required. Other modifications, such as internucleotide (E)-vinylphosphonate (E-VP), where the bridging oxygen is substituted by carbon, have also shown encouraging results for allele discrimination in models of HD,^51,52 but have yet to be evaluated in the context of other disease models.

To effectively load into the RISC, siRNA guide strands must be phosphorylated for recognition by the MID domain of Ago2.^53,54,55 However, fully modified siRNAs are not easily phosphorylated in vivo, unless a hydroxyl is present at the 5′-end of the guide.⁵⁶ Thus, to overcome this limitation and facilitate RISC loading of fully modified siRNAs, a 5′-phosphate is commonly added to the siRNA construct synthetically (Figure 2F). Although this improves gene silencing in vitro, research has shown that these chemically attached 5′-phosphates can be rapidly cleaved by phosphatases, limiting their in vivo efficacy.^57,58 Substitution with chemically stable phosphate mimics, such as 5'-(E)-vinylphosphonate, results in greater metabolic stability against exonucleases and enhanced delivery and activity in vivo.^57,59 Although it is unclear how many of the current clinical-stage drugs contain this modification, several companies (including Alnylam Pharmaceuticals, IONIS Pharmaceuticals, and Atalanta Therapeutics) have disclosed the use of this modification in their late preclinical efforts.

Delivery: The key to unlock the full potential of oligonucleotide-based therapies

Delivery remains one of the key challenges to unlock the therapeutic potential of oligonucleotides beyond the liver. Oligonucleotide drugs are relatively large molecules (∼6–8 kDa for single-stranded ASOs and ∼15 kDa for double-stranded siRNAs) that are negatively charged and do not readily cross cellular membranes. There are several biological obstacles that these drugs need to overcome to become effective medicines: survive nuclease degradation, bypass renal and reticulo-endothelial system clearance, cross the endothelium into the target tissue (notably restricted in specialized barriers like the brain), cross the cellular membrane, and escape the lysosomal pathway and arrive to the intracellular site of action in the cytoplasm or nucleus. The complexity of these challenges is dependent on the route of administration and the target organ and/or cell type. To enhance delivery, several strategies have been evaluated (Figure 3), from careful selection of route of administration to the use of formulation or conjugate-mediated approaches.

Figure 3.

Schematic of select delivery methods for oligonucleotide therapeutics

(A) N-acetylgalactosamine (GalNAc)-conjugated siRNA.

(B) Lipid nanoparticle (LNP) with siRNA cargo.

(C) Antibody-siRNA conjugate depicted as full monoclonal antibody. Based on Roberts et al.³

Figure created with BioRender and Chemdraw. Not drawn to scale.

Route of administration

Selection of the adequate route of administration is key to minimizing clearance and maximizing exposure, while ensuring that critical concentrations are achieved at the target site of action.

Local delivery to the tissue of interest is often used as a strategy to bypass some of the biological hurdles described above, while limiting distribution to untargeted tissues and reducing the likelihood of unwanted side effects. Direct administrations to the site of action, or to a position in proximity, generally allows for the use of lower doses by reducing immediate dilution, metabolism, and clearance of the therapeutic oligonucleotide. Several approved therapeutic oligonucleotides have been approved for local administration in the eye (e.g., fomivirsen, administered intravitreally) and the central nervous system (CNS) (e.g., nusinersen and tofersen, administered intrathecally). Despite its usefulness for a subset of target organs, this strategy comes with its own set of limitations, including limiting volumes and doses, challenging dosing schedules, and a higher burden than intravenous or subcutaneous delivery.

Systemic delivery of therapeutic oligonucleotides is typically selected when the target tissue is not readily accessible and when leveraging the vasculature for distribution is feasible. Indeed, delivery of oligonucleotides to highly perfused organs that present discontinuous endothelium and fenestrated base membrane, such as the liver, is routinely used in clinical practice. Approximately one-third of clinical-stage therapeutic oligonucleotides are administered intravenously, and it is preferred in applications that rely on lipid nanoparticles for delivery (e.g., patisiran) and/or drugs that intend to target muscle tissues.⁶⁰ Intravenous dosing allows for rapid peak concentrations in the blood, with subsequent rapid clearance. Conversely, subcutaneous administration delays the initial clearance by allowing for sustained release from a subcutaneous oligonucleotide depot over time.⁶⁰ Subcutaneous administration is the most popular strategy in clinical practice, likely due to its low invasiveness compared to other routes of administration, but also because it removes the costly requirement of specialized personnel for administration. There are a significant number of oligonucleotide drugs that have been approved for subcutaneous administration (Table S1).

Formulation-based approaches

Oligonucleotide delivery systems can serve multiple purposes: they protect the oligonucleotide from degradation by encapsulating or binding, thus improving stability after administration, and they enhance delivery to target tissues and intracellular trafficking. Extensively studied synthetic carriers include lipid, polymer, inorganic, and peptide-based systems and their hybrids,^61,62 alongside biologically derived carriers such as exosomes, valued for biocompatibility and active targeting.^63,64 Ligand functionalization can add cell-type specificity, though at the cost of greater manufacturing and quality-control complexity.

Lipid-based systems are the most advanced in the clinic for oligonucleotide delivery (Figure 3B). These systems generally rely on electrostatic interactions between the negatively charged backbone of the oligonucleotide and the positively charged head of the lipid for condensation and complexation. Cationic lipids became a popular choice for their high efficiency in vitro; however, their use in a clinical setting was limited by in vivo toxicity.^61,65 More advanced lipid-based delivery systems use ionizable lipids, which are neutral in physiological pH and protonated in acidic conditions, such as the DLin-MC3-DMA (MC3) lipid developed by Tekmira/Arbutus (later licensed by Alnylam therapeutics). Patisiran is an siRNA therapy formulated in a lipid-based delivery system that was approved in 2018 for the treatment of transthyretin-mediated (TTR) amyloidosis.⁶⁶ Patisiran uses a proprietary mixture of MC3 lipid, a helper lipid (distearoylphosphatidylcholine, DSPC), cholesterol, and polyethylene glycol (PEG)-lipid (PEG-2000-C-DMG). This ionizable lipid nanoparticle is believed to further enhance delivery to hepatocytes by associating with apolipoprotein E (APOE), which facilitates delivery through low-density lipoprotein receptor (LDL-R).^67,68 In this formulation, the PEG-lipid plays an important role in stabilization of the nanoparticle, and in enhancing the circulating times after intravenous administration. Although significantly better tolerated than initial lipid formulations, the administration of patisiran often requires pre-treatment with steroids and antihistamine medications to reduce the likelihood of immune reactions.⁶⁹ Although there are other lipid nanoparticles been evaluated in humans for delivery of oligonucleotides, such as liposomal formulation SMARTICLES, patisiran is the only clinically established medicine.

Chemistry- and conjugation-based approaches

Most approved therapeutic oligonucleotides are administered without an excipient, historically relying on the PS backbone of single-stranded oligomers for nonspecific protein binding and uptake—mediated in the liver by scavenger receptors STAB1/STAB2.^70,71 PS also increases plasma protein binding (i.e., albumin⁷²), improving circulating times by delaying their glomerular filtration, and can determine their intracellular fate, for example, by enabling binding to nucleolin and leading to accumulation in the nucleus.⁷³

Examples of approved ASOs with PS backbones include nusinersen and tofersen, administered locally by intrathecal injection for activity in the CNS, and inotersen and volanesorsen, administered subcutaneously for gene modulation in the liver.

Although siRNAs traditionally rely less on PS modifications for delivery, some constructs use PS to enable broad delivery across the CNS, harnessing nonspecific protein binding. For example, the di-siRNA motif consists of two siRNAs connected to one another through a linker.^31,74 Each asymmetric monomer displays a fully PS-modified tail that is believed to enable tissue uptake through cooperative, nonspecific binding. The di-siRNA platform is currently being developed by Atalanta Therapeutics for HD,⁷⁵ congenital epilepsy⁷⁶ and pain disorders⁷⁷ and the aforementioned clinical trial for prion disease.³⁶ Interestingly, these novel multivalent oligonucleotide platforms (such as di-siRNAs, bis-siRNAs, and other hybrid ASO-siRNAs) provide the opportunity for dual-targeting approaches where two genetic targets are modulated simultaneously by a single molecular entity. These strategies bypass the need to co-formulate and/or administer two separate drugs and ensure both sequences reach the exact same cellular target. Currently, these bifunctional molecules are at the preclinical stage and are being investigated for multifactorial neurological disorders and cancer.

Conjugation of ligands to therapeutic oligonucleotides is used as a strategy to promote broad functional delivery, delay clearance, and enable targeted delivery to specific cell types or tissues.^8,78 The selection of sites on the oligonucleotide where to conjugate and the strategy of how to conjugate is crucial to maintain activity. On ASOs, ligands are generally conjugated at the 5′-end of the molecule, but there is some flexibility for conjugation at the 3′-end. On the other hand, ligand conjugation on siRNAs is preferentially carried out on the sense strand,⁷⁹ at the 3′-end, the 5′-end, and also, at specific 2′-O ribose positions. Although conjugation at the 5′-end of the guide strand can disrupt loading into RISC,⁷⁹ ligand conjugation can be designed to allow disassociation upon internalization by the target cell. This can be achieved by using labile linkers that disassemble in acidic conditions or other strategies. A chemically diverse number of bioconjugates have been appended to therapeutic oligonucleotides, including carbohydrates, lipids, peptides, sugars, aptamers, polymers, antibodies and small molecules. Below, we will discuss bioconjugates that have achieved significant clinical relevance.

N-Acetylgalactosamine

GalNAc (N-acetylgalactosamine) is by far the most well-investigated ligand conjugation strategy in the field, which has been applied to both single- and double-stranded oligonucleotides (Figure 3A).⁸⁰ GalNAc is a carbohydrate molecule that binds with high affinity (K_d = 2.5 nM) to the asialoglycoprotein receptor (ASGPR), which is highly expressed by hepatocytes.⁸¹ After binding the ASGPR, the oligonucleotide is endocytosed and acidic conditions facilitate GalNAc dissociation from the oligonucleotide.⁸² The ASGPR receptor is then rapidly recycled to the cell surface of the hepatocyte (∼15 min). Several GalNAc variants have reached late preclinical and clinical stages, yet the most widely implemented structure is the triantennary GalNAc, which consists of three GalNAc monosaccharides attached to a common scaffold that provides optimal binding to the ASGPR. Compared to non-conjugated PS-modified oligos that mostly deliver to non-parenchymal cells,^82,83 when administered subcutaneously, GalNAc conjugates improve delivery to the target cell (hepatocyte) up to 10-fold in rodents⁸² and 30-fold in humans,⁸⁴ thus enhancing ASO and siRNA potency. Several GalNAc-conjugated siRNAs and one ASO have received FDA approval (Table S1). The success of the GalNAc platform for hepatic indications has driven rapid industry adoption. Major companies, including Alnylam, IONIS Pharmaceuticals, Arrowhead Pharmaceuticals, Dicerna Pharmaceuticals (now part of Novo Nordisk), Novartis, and Roche, have established or licensed GalNAc-based programs to strengthen their drug pipelines.

Antibodies

Antibodies bind with high affinity to their cognate target and, therefore, have been evaluated for tissue- and cell-type-specific drug delivery.⁸⁵ Interest in using antibodies for extrahepatic delivery of therapeutic oligonucleotides has resurged, with a variety of antibody-oligonucleotide conjugate (AOC) architectures being explored, which incorporate either full monoclonal antibodies (mAbs) or smaller antibody fragments (Fabs and others) (Figure 3C). Full-length mAbs (∼150 kDa) provide high target affinity and long half-lives, whereas Fabs and others confer better tissue penetration without immune effector functions as they lack the Fc region.

Transferrin receptor 1 (TfR) antibodies are the most advanced technology in the clinic and have been used to enhance delivery of siRNAs and ASOs to muscle tissue.^86,87 TfR is critical in the transport and uptake of iron and is highly expressed in cardiac, smooth, and skeletal muscle cells,⁸⁸ as well as in the blood-brain barrier (BBB) endothelium. The recycling time of TfR is ∼17 min, making it particularly attractive for drug delivery. Avidity Biosciences (now a Novartis company) has developed a TfR mAb conjugated to an siRNA through a non-cleavable linker (AOC 1001/Del-desiran),⁸⁹ and Dyne therapeutics engineered a TfR Fab conjugated to a gapmer ASO (DYNE-101) through a cleavable linker. Both target DMPK mRNA for the treatment of myotonic dystrophy type 1 (DM1), and both have shown encouraging data in phase 1/2 trial, with AOC1001/Del-desiran recently progressing to a phase 3 trial. In parallel, phase 1/2 clinical trials are also ongoing in facioscapulohumeral muscular dystrophy (FSHD) and DMD by these companies with the same mAb and Fab.⁹⁰ When linking charged siRNA or RNase H ASOs to the antibody, usually it is not possible to link multiple ASOs to one mAb due to the electrostatic interference. Interestingly, for the DMD trial, a PMO ASO is conjugated to the mAb. As the PMO is uncharged, it was possible to link multiple PMOs to a single antibody, with a drug to antibody ratio of up to 9.7.⁸⁹

TfR antibodies are also evaluated for therapeutic delivery across the BBB for the treatment of neurological disorders. JCR Pharmaceuticals and Denali Therapeutics pioneered the first two approvals for TfR brain delivery vehicles for iduronate-2-sulfatase (pabinafusp alfa and tividenofusp alfa, respectively), a lysosomal enzyme deficient in mucopolysaccharidoses type II. These advances sparked renewed interest in TfR-mediated BBB transit for delivery of oligonucleotides to the brain. Many companies, including Denali,⁹¹ and also Roche, Vect-Horus, Manifold Bio, Takeda, and Janssen (now J&J), have explored the TfR-mediated ASO and siRNA delivery to the brain. Despite significant promise, none of these technologies have reached clinical stage yet.

Lipids

Lipid conjugates have been leveraged for broad functional delivery of therapeutic oligonucleotides to multiple organs, including the liver and muscle, after systemic administration.⁸ Preferential delivery likely occurs via binding to circulating lipoproteins, which facilitate uptake in target organs.^92,93,94 Binding to circulating lipid transport systems correlates with the overall hydrophobicity, with more hydrophobic siRNAs associating primarily with LDL and less hydrophobic associating primarily with high-density lipoprotein (HDL).⁹³ LDL-bound lipid-oligo conjugates preferentially distribute to the liver, whereas HDL-bound lipid-oligo conjugates distribute to the adrenal gland, ovary, and small intestine.⁹³

Systemically administered polysaturated and polyunsaturated fatty acids are currently being evaluated in the clinic for a diverse group of diseases, including pre-eclampsia, myotonic dystrophy, and DMD. Docosanoic acid (DCA)-siRNA conjugates targeting soluble FLT1 in the placenta (CBP-4888) for the treatment of pre-eclampsia by Comanche Biopharma^95,96; oleic acid-conjugated antagomir miR-23b (ATX-01) targeting the muscle for the treatment of DM1 and DM2 by Arthex Biopharma⁹⁷; palmitic acid-conjugated tcDNA ASO for exon 51 skipping (SQY51) in the muscle as a treatment for DMD by SQY therapeutics⁹⁸; and other companies, such as Biomarin, have focused on the development of lipid conjugates based on aliphatic lipids, such as C16, for the delivery of splice switching ASOs to the muscle for the treatment of DMD (BMN351).^99,100

Lipid conjugates have also been considered for local delivery of therapeutic oligonucleotides to the CNS and skin. Alnylam Pharmaceuticals developed a linear aliphatic lipid (C16) conjugated at the 2′-O position of the ribose of sense strand sixth nucleotide.¹⁰¹ This C16 conjugate enables broad distribution through the CNS and has shown promising results in late preclinical studies for modulation of gene expression in a variety of cell types.¹⁰¹ Thus, C16 conjugate has progressed to the clinic and is being investigated for familial Alzheimer disease (mivelsiran) targeting amyloid precursor protein (APP) and for HD targeting huntingtin (HTT). Also in the CNS space, DTx Pharma (acquired by Novartis) is developing long-chain fatty acid conjugates to deliver siRNAs to Schwann cells after intrathecal administration for the treatment of Charcot-Marie-Tooth disease type 1A (DTx-1252). Alys Pharmaceuticals (previously Aldena Therapeutics) is currently developing lipid-conjugated siRNAs, including DCA conjugates, for intradermal injections¹⁰² or for delivery to the skin with star-shaped particles (STAR).¹⁰³ STAR particles consist of small particles made of aluminum oxide or stainless steel with micron-scale projections that create pores in the skin, significantly increasing permeability to drugs. Alys Pharmaceuticals is developing treatments for a variety of skin conditions including alopecia areata, atopic dermatitis, and chronic hand eczema. In addition, some historical programs include cholesterol-conjugated siRNAs (RXI-109) for reduction of hypertrophic scars and keloids by targeting connective tissue growth factor (CTGF) from RXi Pharmaceuticals. This program was deprioritized based on unfavorable path to commercialization, rather than poor performance of the drug.

Peptides

Peptides have for long been used as ligands or adjuvants of therapeutic oligonucleotides to enable cell-type-specific delivery, to aid with cell penetration and/or release from the endosomal compartment. Nonetheless, the development of peptide conjugates has somewhat fallen behind lipids and antibodies, likely due to compatibility hurdles and/or toxic liabilities in the clinic.

Cell-penetrating peptides (CPPs) are generally short, cationic, arginine-rich peptides that bind and transiently disrupt the cell membrane, thereby enabling cellular uptake. Sarepta Therapeutics and PepGen Ltd have developed peptide-PMO technologies for DMD with the latest generation of conjugates allowing for potent exon skipping. Nonetheless, these clinical developments have been stopped, due to dose-limiting nephrotoxicity, which is correlated to the arginine content.¹⁰⁴ Entrada Therapeutics has generated an alternative CPP that is circular, contains less arginine residues, and is suggested to play a role in cellular uptake and endosomal escape.¹⁰⁵ This conjugate is currently tested in phase 1/2 clinical trials for DMD (ENTR-601-44 and ENTR-601-45).

Endosomolytic peptides, such as melittin or melittin-like peptides (MLPs), have also been conjugated or co-administered with therapeutic oligonucleotides to enhance release from the endosomal compartment. Endosomyolytic peptides disrupt cellular membranes and form pore-like structures at high concentrations. When used in the context of GalNAc-conjugated siRNAs, MLPs have been shown to enable a 500-fold increase in gene silencing of apolipoprotein B (APOB) and hepatitis B virus (HBV)-related transcripts.^106,107 This strategy progressed into the clinic with acceptable tolerability profile in healthy volunteers, but long-term toxicities were observed in non-human primates in extended toxicology studies, halting clinical programs.^8,108 Although endosomolytic peptides substantially improve the potency of therapeutic oligonucleotides by assisting escape to the cytoplasm, toxicity liabilities reduced interest in clinical development.

Therapeutic oligonucleotides in the clinic: Approved drugs and the evolving landscape

At the time of writing, 21 antisense oligonucleotides have received regulatory approval from the FDA, European Medicines Agency (EMA), and/or Japan’s Pharmaceutical and Medical Devices Agency (PMDA) (Table S1); two were later withdrawn, for poor sales (fomivirsen) and hepatotoxicity risk (mipomersen). The clinical pipeline continues to expand, with a documented rise in phase 3 trials over the last decade suggesting the field is maturing.^109,110 Below, we trace the approval history for each antisense class and highlight key phase 3 studies.

RNase H-dependent and steric blocker ASOs

ASOs were the first oligonucleotide modality that reached the clinic in the early 90s, a little over 10 years from their discovery in the late 70s.¹¹¹ The first oligonucleotide drug approved was intravitreal-administered fomivirsen, in 1998 and 1999 by the FDA and EMA, respectively. Fomivirsen consists of an RNase H-activating ASO chemistry (fully PS DNA), developed for cytomegalovirus (CMV)-induced retinitis, previously the most common infectious cause of blindness in patients with acquired immune deficiency syndrome (AIDS).¹¹² With the availability of triple combination antiviral therapy for individuals with human immunodeficiency virus (HIV) infection, the incidence of individuals with AIDS and CMV-induced retinitis vastly decreased. Thus, the drug was withdrawn from the market in 2002 and 2006 in the European Union and the United States, respectively.

The next ASO approval—mipomersen, an RNase H-dependent antisense oligonucleotide targeting APOB for hypercholesterolemia—did not occur until 2013, 15 years after fomivirsen.¹¹³ Mipomersen was only approved by the FDA, based on clinical data showing significant reductions in blood cholesterol after subcutaneous administration of the drug in a placebo-controlled trial.¹¹⁴ Mipomersen was later voluntarily withdrawn from the market in 2019 due to high patient discontinuation related to tolerability issues. Nonetheless, this approval represents a tipping point for ASO approvals with the third drug, eteplirsen,⁹⁰ approved only 3 years later by the FDA and the fourth ASO, nusinersen, approved 3 months later.¹¹⁵ Ever since, at least 1 ASO therapy has been approved annually by regulatory agencies (Table S1). Here, we will provide context for some notable approved ASOs that illustrates the different chemical modifications, modalities, formulations, and routes of administration.

There are currently 5 splice-modulating ASOs approved: nusinersen for spinal muscular atrophy (SMA) and eteplirsen, golodirsen, casimersen, and viltolarsen all for patients with DMD with eligible mutations.⁹⁰

Nusinersen is a fully modified MOE PS ASO that increases the inclusion of exon 7 in SMN2 transcripts by binding to the intron 7 intronic splicing silencer. The result is an increased production of SMN protein and delay in disease progression. Nusinersen is injected intrathecally, with 4 loading doses, given on days 1, 15, 29, and 59, followed by maintenance doses every 4 months.^115,116 Clinical trials in patients with severe type 1 SMA showed a significant decrease in death or permanent ventilation for nusinersen-treated patients, compared to sham-treated patients.¹¹⁵ Furthermore, some treated patients achieved milestones that are by definition not met in patients with type 1 SMA, such as independent sitting and standing. In patients with milder type 2 SMA, nusinersen treatment resulted in significantly improved motor function compared to the sham-treated patients.¹¹⁶ Based on these data, nusinersen was approved for all types of SMA by the FDA, EMA, and PMDA.¹¹⁷

Splice modulation as a therapeutic strategy for DMD intends to allow the production of internally deleted, but partially functional, dystrophin proteins by inducing exon skipping at the pre-mRNA level.⁹⁰ Thus, ASOs are able to restore the reading frame, allowing patients with DMD to make a Becker muscular dystrophy (BMD)-like dystrophin. BMD is a milder disease, with a later onset of symptoms and a slower disease trajectory. The approach is mutation specific: depending on the size and location of deletions, different exons need to be targeted to restore the open reading frame. While animal models show that dystrophin restoration reduces muscle pathology and improves function, clinical trials have only confirmed low, single-digit dystrophin restoration with variability across patients and compounds.⁹⁰ Based on the dystrophin restoration data, which served as a surrogate endpoint, and the absence of severe side effects, the FDA granted accelerated approval to four ASOs: eteplirsen (exon 51), golodirsen (exon 53), casimersen (exon 45), and viltolarsen (exon 53, approved first by the PMDA). Confirmatory evidence of slowed-down disease progression is so far lacking: confirmatory studies did not meet the primary endpoint for golodirsen, casimersen, and viltolarsen and eteplirsen studies are still ongoing. All four use PMO chemistry, whose limited bioavailability requires frequent, high-dose weekly intravenous infusions (30–80 mg/kg). Improved skeletal muscle delivery using either chemical modification or conjugation is currently under investigation, with 8 ongoing clinical trials for DMD testing compounds targeting various exons (as described above).⁹⁰

Six additional ASOs have also been approved (Table S1). These consists of second- (and third-) generation gapmer ASOs with evolved chemistry designs that enhance targeting activity while reducing tolerability issues. Inotersen (2018) and volanesorsen (2019) exemplify this class: inotersen treats hereditary TTR (hTTR) amyloidosis, caused by toxic gain-of-function TTR mutations that produce aggregate-prone transthyretin and multi-organ disease,¹¹⁸ and volanesorsen targets APOC-III to lower triglycerides in familial chylomicronemia syndrome.¹¹⁹ Second-generation ASOs have also been used for treating CNS diseases through local intrathecal administrations. Tofersen was the first gapmer ASO approved for CNS applications (in 2023) targeting superoxide dismutase 1 (SOD1) in patients with amyotrophic lateral sclerosis (ALS).¹²⁰ Tofersen received approval based on improvements seen in key disease biomarkers, such as neurofilament chain light (NfL), in placebo-controlled trials. The primary clinical endpoint (ALS severity score) was not met in the clinical trial, but longer-term data from the open-label treatment stage, which showed functional improvements on ALS severity scores after longer treatment, were supportive of a treatment effect.¹²⁰

Intrathecally administered CNS oligonucleotides such as tofersen and nusinersen pose a practical burden to the patient due to the need of frequent spinal taps, potential for post-lumbar-puncture syndrome, hydrocephalous, aseptic meningitis, and other inflammatory or cerebrospinal fluid (CSF) complications (elevated CSF protein and cell count), as described in post-marketing reports, and warrant close monitoring^121,122,123

Second-generation gapmer ASOs brought significant improvements over first-generation gapmer ASOs, but still require relatively high doses and close monitoring of side effects (i.e., thrombocytopenia). The newest generation of therapeutic oligonucleotides instead leverages cell-type-specific delivery to improve tissue targeting, effectively reducing the dose required for therapeutically meaningful effects and improving tolerability. As examples, (1) GalNAc-conjugated eplontersen (approved in 2023)¹²⁴ targeting hTTR requires a ∼45 mg dose administered monthly vs. ∼285 mg administered weekly for inotersen and (2) GalNAc-conjugated olezarsen (approved in 2024)¹²⁴ targeting APOC-III has a typical starting dose of ∼80 mg administered monthly vs. ∼285 mg administered weekly (for 3 months and then every 2 weeks for maintenance) for volanesorsen.

RNAi-based drugs

The elucidation of the RNAi pathway followed the initial discovery of ASOs by approximately two decades,¹²⁵ a breakthrough that led to the Nobel Prize being awarded to Andrew Fire and Craig Mello in 2006. Twenty years after the seminal publication in Nature and 12 years after the Nobel Prize was awarded, the first RNAi drug, patisiran was approved by the FDA. It was also the first genetic medicine approved for hTTR amyloidosis, preceding inotersen, and is a partially chemically modified siRNA formulated in a lipid nanoparticle for intravenous administration.⁶⁶ The phase 3 APOLLO trial and long-term open-label data showed that patisiran is able to improve neuropathy scores and patient quality of life over years, with side effects mostly limited to site infusion reactions.⁶⁶ Although patisiran requires the administration of premedication to reduce the likelihood of inflammatory responses, it does not require continuous monitoring of renal function, like inotersen. Currently, patisiran is the only approved siRNA drug delivered in a lipid formulation, with all other drugs consisting of fully modified siRNAs conjugated to GalNAc.

The first GalNAc-conjugated siRNA approved by the FDA was givosiran in 2019 (EMA 2020),¹²⁶ targeting the hepatic delta-aminolevulinic acid synthase 1 gene (ALAS1) for the treatment of acute hepatic porphyria. Givosiran reduced porphyria attack rates by 70% in the ENVISION phase 3 trial when compared to the placebo-controlled group. This represented a pivotal approval for the field of therapeutic oligonucleotides, providing the first clinical proof of concept for the emerging era of conjugate-mediated delivery. Note that this approval preceded the approval of the first GalNAc-conjugated gapmer ASO, eplontersen. Furthermore, GalNAc-conjugation facilitated the final shift from intravenous to subcutaneous delivery, which allows for patient self-administration and has become the preferred route for treating liver diseases. Since the first approval, a total of 6 additional GalNAc-siRNAs are now on the market, including lumasiran, inclisiran, vutrisiran, nedosiran, fitusiran, and plozasiran. Some of these drugs feature new chemistry improvements beyond ligand conjugation that significantly enhance durability and/or reduce off-targeting.

Lumasiran was approved in 2020 for the treatment of primary hyperoxaluria type 1 by targeting HAO1 transcripts to reduce glycolate oxidase and prevent kidney damage.¹²⁷ Patients treated with lumasiran showed dramatic reductions in urinary oxalate levels in the ILLUMINATE-A phase 3 trial. Inclisiran was approved in 2020 by the EMA, and later in 2021 by the FDA, for the treatment of familial hypercholesterolemia by targeting proprotein convertase subtilism-kexin type 9 (PCSK9). In the ORION phase 3 trials, inclisiran enabled reductions of LDL-C by 50%, and the effect was sustained for at least 18 months.^128,129 Vutrisiran targets hTTR and was approved in 2022 based on data from the HELIOS-A phase 3 trial that showed improvements in neuropathy scores and quality of life.¹³⁰ A second GalNAc-siRNA drug for primary hyperoxaluria targeting lactate dehydrogenase A (LDHA), nedosiran, received approval in 2023.¹³¹ In PHYOX2 phase 2/3 trial, 50% of patients taking nedosiran showed normal levels of oxalate in the urine. Although nedosiran has only received approval for PH1, the drug targets the final enzyme that gives raise to oxalate and, therefore, may be potentially effective for patients with variants in PH2/3. Fitusiran, targets antithrombin for the treatment of hemophilia and was approved in 2025.¹³² The ATLAS phase 3 trial showed a ∼72% reduction in annualized bleed rate in patients with hemophilia A/B when compared to patients receiving on-demand therapy. Finally, plozasiran, was approved in 2025 for the treatment of familial chyloicronemia syndrome (FCS) and other severe hypertriglyceridemias by targeting APOC-III.¹³³ The PALISADE phase 3 trial demonstrated that patients treated with plozasiran show significant reductions in triglycerides in the blood and a reduced risk of acute pancreatitis. These GalNAc-conjugated siRNAs exhibit remarkable potency and durability, with typical doses ranging from 25 to 50 mg and maintenance schedules of every 2–6 months (except for inclisiran, 284 mg every 6 months), following the completion of loading doses.

Beyond the 21 approved antisense oligonucleotides, the clinical landscape continues to expand with numerous clinical trials ongoing. Table S2 highlights key phase 3 studies that have been recently initiated or completed. The therapeutic potential of other closely related modalities—such as miRNA mimics and saRNAs—is currently being interrogated in phase 1 and 2 clinical trials to establish clinical proof of concept.

The future of therapeutic oligonucleotides as genetic medicines

Artificial intelligence: A force multiplier for more effective and safer therapeutic oligonucleotides

Artificial intelligence (AI) is rapidly reshaping many industries and is expected to change and accelerate the development of therapeutic oligonucleotides. In the coming years, AI will likely become a discovery tool that unifies target selection, sequence design, chemical modification, delivery optimization, and safety into a single predictive pipeline, while also improving manufacturing through optimized synthesis, impurity control, and batch consistency.

AI-driven platforms can mine genomics, transcriptomics, and splicing data to fast-track identification of novel genetic targets and the right oligonucleotide class for a given disease¹³⁴—already in use for rare and ultra-rare neuromuscular disorders. Through computational modeling of vast amounts of data, AI and machine learning algorithms will also support oligonucleotide drug design beyond the current in silico tools, to propose positional chemical modifications that improve potency, stability, durability, and safety. The analysis of large amounts of high-quality data generated from in vitro and in vivo screening campaigns will enable optimization of chemistry architectures based on target sequence. Similarly, AI tools are already playing an important role in assisting to tackle drug delivery challenges by streamlining the discovery of novel ligands for extrahepatic delivery and endosomal escape. However, and arguably, one of the highest value areas where AI-driven platforms can impact development of therapeutic oligonucleotides is off-target safety. In the future, the predictive power of AI will provide insights about transcriptome-wide effects before any empirical data are produced, allowing for prioritization of lead candidates with low safety risk profile for subsequent follow-up experiments.

However, it is important to highlight that AI will not be a replacement for empirical testing and will, in fact, be reliant on large and high-quality datasets to learn from. As such, it will be of crucial importance to establish clear processes to determine the suitability of datasets for machine learning. Researchers must collect and share quality data, not only when an oligonucleotide treatment is successful but also when it is not, or when it is not safe in both preclinical and clinical settings. The continued effort of generating and timely sharing empirical off-target and toxicology data will be of great importance in the coming years to enable machine learning and AI to reach its full potential.

From rare diseases to personalized therapies, to N-of-many: Unique challenges and approaches

The clinical development of therapeutic oligonucleotides first concentrated on rare genetic disorders, where defined targets and high unmet need enabled rapid translation. While 80% of over 6,000 rare diseases known have a genetic origin, only 5% have disease-modifying therapies that target the root of the disease.¹³⁵ Their programmable nature makes them a perfect tool for developing individualized genetic medicines. Notably, the distribution and safety profile depend more on the chemical modification than the sequence of the oligonucleotides. Thus, the same chemistry, once optimized for tissue distribution, activity, and safety, can be reprogrammed to target different disease-causing transcripts, with minor impact on the pharmacokinetics and safety profile.

Milasen, a splice-modulating ASO for the treatment of a single patient with CLN7 Batten disease (Mila Makovec) is a model example. Intrathecal delivery of milasen resulted in a reduction of the duration and frequency of seizures.¹³⁶ The work was a collaborative effort of Tim Yu (Boston Children’s Hospital) and Mila’s Miracle Foundation, as well as many others involved in the preclinical and clinical development. Milasen provided a crucial proof of concept in the development of individualized oligonucleotide therapies, which ignited the field to support the development of now over ∼40 different ASOs to treat over 80 patients.^137,138 In fact, through extensive cooperation across academia and non-profit organizations, such as n-Lorem and the Dutch Center for RNA Therapeutics, there are currently more than twice the number of individualized ASOs used to treat patients than there are approved ASOs! While there are as yet no individualized siRNA-based therapies, application of lessons learned from ASOs stand to accelerate development.

The development of individualized oligonucleotide therapies poses unique challenges, including the need for accelerated development and complex regulatory processes. In general, individualized therapies are being developed for patients with progressive diseases that might not see clinical benefit if timelines to first dose are too long. Thus, step-by-step frameworks are being designed to help guide and accelerate the process from diagnosis to variant identification and interpretation, to determining amenability to ASO treatment. Additional barriers include complex infrastructure for clinical implementation, monitoring, and data capture, which is beyond the scope of this review. Despite the challenges, developing individualized therapeutic oligonucleotides offers an opportunity to obtain key pharmacokinetic, pharmacodynamic, and safety data for many oligonucleotide chemistries, albeit in a small number of patients. The transparent sharing of experiences, both positive and negative, is key for advancing customized oligonucleotide therapies. This exchange streamlines nonclinical and clinical development protocols and supports reverse translation efforts, particularly in cases where unexpected human toxicity emerges. For further information on frameworks and guidelines, we refer the interested reader to a new review from the N = 1 Collaborative¹³⁷ and a publication from the International Rare Disease Research Consortium (IRDIRC) N = 1 therapy taskforce.¹³⁹

To date, most individualized ASOs have been developed in the United States, where the FDA has been an active partner in repurposing existing regulatory pathways. Currently, policy requires separate investigational new drug (IND) applications for each treatment. Promisingly, a new guidance for considerations for the use of a plausible mechanism framework was published in 2026 (https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-plausible-mechanism-framework-develop-individualized-therapies-target-specific). The framework, developed for base editing and oligonucleotide therapeutics, outlines a new system where multiple individualized treatments targeting the same gene or pathway would be part of the same IND family, utilizing one master document as a template for initial treatment. In a sequential process, additional treatments would cross-reference to the “parent IND” and leverage information from the first treatment to allow for accelerated manufacturing and abridged or waived safety studies. Similarly, in the European Union, the European Commission and EMA are preparing a new directive for platform technology development (https://www.ema.europa.eu/en/events/webinar-use-platform-approaches-non-clinical-clinical-domains).¹⁴⁰

Expanding to tackle common disorders: The present and the prospects

Therapeutic oligonucleotides are advancing into an exciting new era, with several molecules being explored for the treatment of high-prevalence disorders, including hypercholesterolemia. Inclisiran, a GalNAc-conjugated siRNA targeting PCSK9, was first approved in 2021 for adults with heterozygous familial hypercholesterolemia or atherosclerotic cardiovascular disease (ASCVD).¹²⁸ In 2023, the FDA expanded the indication to include adults with high LDL cholesterol and other co-morbidities, and in 2025, for use as a monotherapy in adults with primary hypercholesterolemia.¹²⁸ This is a major landmark in the field and broadens the application of therapeutic oligonucleotides beyond their traditional use for treating rare diseases.

Moving forward, the field is likely to see the approval of more oligonucleotide treatments for common disorders, particularly those targeting cardiometabolic indications. Currently, there are several siRNAs and ASOs in development for hypercholesterolemia, targeting PCSK9 and ANGPTL3, and hypertriglyceridemia, targeting APOC3 and ANGPTL3. These can also be potentially used in combination for the treatment of common mixed dyslipidemias. As an example, a phase 3 study (NCT06813911) was recently initiated to study Lp(a) lowering using pelacarsen in inclisiran-treated patients. Furthermore, Arrowhead Pharmaceuticals has initiated the first dual-targeting RNAi therapeutic trial for the treatment of mixed hyperlipidemia targeting PCSK9 and APOC3. ARO-DIMER-PA combines two siRNAs in a single molecule for modulation of two different genetic targets. The future and success of therapeutic oligonucleotide treatments for high-prevalence disorders hinges on identifying and validating relevant therapeutic targets and solving targeted delivery challenges. With each tissue where delivery is “unlocked,” oligonucleotide treatments become an option for additional diseases.

Final considerations

With 21 approved drugs and more likely on the way, therapeutic oligonucleotides now offer disease-modifying treatment for many otherwise incurable disorders. This success followed years of chemistry evolution and deepening RNA biology insight, pivoting on the discovery of the GalNAc delivery platform, which accelerated drug development for liver disorders and now fuels a broader effort to unlock extrahepatic delivery.

As the field matures and applications broaden from rare to common disorders, oligonucleotides are firmly establishing themselves as the third pillar of human medicine.

Declaration of interests

B.M.D.C.G. owns stock at Atalanta Therapeutics and is a named inventor on several patents and patent applications regarding RNA interference technologies and therapeutic targets. B.M.D.C.G. serves as a Director on the Board of the Oligonucleotide Therapeutics Society. C.M.F. is a named inventor on several patents and patent applications regarding RNA interference technologies and therapeutic targets. C.M.F. serves on the Scientific Advisory Board of the Oligonucleotide Therapeutics Society. A.A.-R. discloses being employed by LUMC, which has patents on exon skipping technology, some of which have been licensed to BioMarin and subsequently sublicensed to Sarepta. As co-inventor of some of these patents, A.A.-R. was entitled to a share of royalties. A.A.-R. further discloses being an ad hoc consultant for PTC Therapeutics, Sarepta Therapeutics, Regenxbio, Dyne Therapeutics, Lilly, BioMarin Pharmaceuticals Inc., Eisai, Entrada, Takeda, Splicesense, Galapagos, Sapreme, Italfarmaco, and Astra Zeneca. A.A.-R. also reports being a member of the scientific advisory boards of Hybridize Therapeutics (past), Silence Therapeutics, Sarepta Therapeutics, Sapreme, and Mitorx. Remuneration for consulting and advising activities is paid to LUMC. In the past 5 years, LUMC also received speaker honoraria from Alnylam Netherlands, Italfarmaco, and Pfizer and funding for contract research from Sapreme, Eisai, BioMarin, Galapagos, and Synaffix. Project funding is received from Sarepta Therapeutics and Entrada via unrestricted grants. Annemieke serves as a Director on the Board of the Oligonucleotide Therapeutics Society.