1. Chemistry Modification

Chemical modification represents one of the most effective approaches to enhance Oligonucleotide drug Delivery. Modification of the nucleic acid backbone, the ribose sugar moiety and the nucleobase itself has been extensively employed in order to improve the drug-like properties of oligonucleotide drugs and thereby enhance delivery. Specifically, modification is utilized to improve oligonucleotide pharmacokinetics, pharmacodynamics and biodistribution. Specific patterns of modification are also required for the functionality of certain therapeutic modalities (for example, gapmers). The importance of chemistry is exemplified by the observation that extensive chemical modification of second-generation gapmer ASOs is sufficient to enable delivery to a wide variety of tissues, without the need for an additional delivery agent. Furthermore, of the ten approved oligonucleotide therapies approved to date, eight are lacking an additional delivery vehicle (that is ‘naked’) and so are dependent on chemical modification alone to facilitate their tissue delivery. This is also true of gapmer ASOs currently in development by Ionis Pharmaceuticals, including drugs for the treatment of amyotrophic lateral sclerosis, Alzheimer disease, centronuclear myopathy and, most notably, Huntington disease.

2. Backbone modification

The incorporation of phosphorothioate (PS) linkages (pictured above), in which one of the non-bridging oxygen atoms of the inter-nucleotide phosphate group is replaced with sulfur, is widely used in therapeutic oligonucleotides. There are many other kinds of backbone modification (for example, boranophosphate), although these have been less commonly used. PS backbone modifications are readily tolerated in ASO designs and do not disrupt RNase H activity. By contrast, siRNAs that contain PS modifications at every linkage are less active than the equivalent phosphodiester (PO) siRNA, and, as such, PS-containing siRNAs are typically modified at the termini only, if at all. Sulfated molecules, such as oligonucleotides containing PS linkages or thiol tails, are also taken up by scavenger receptors (such as the stabilins STAB1 and STAB2), which mediate their internalization into tissues such as the liver. The incorporation of PS linkages has the dual effect of conferring nuclease resistance and promoting binding to proteins in both plasma and within cells. Oligonucleotide interactions with plasma proteins such as albumin106 have the effect of improving drug pharmacokinetics by increasing the circulation time (and therefore reducing renal clearance). However, binding of a PS-containing gapmer ASO to plasma α2-macroglobulin (A2M) was found to be non-productive. PS modification of oligonucleotides also increases interactions with intracellular proteins (for example, nucleolin) that are believed to pro- mote their accumulation in the nucleus, the target site of action for splice-switching oligonucleotides.

Notably, resistance to cellular nucleases results in pro- longed tissue retention and long-lasting drug effects. In cases where this is undesirable, in the case of toxicity due to prolonged gene silencing for example, the incorporation of one or more PO linkages can be used to ‘tune’ the durability of the oligonucleotide by reducing its nuclease stability.

A disadvantage of PS backbone modifications is that they have the effect of reducing the binding affinity of an oligonucleotide for its target, a limitation that can be compensated for by incorporating additional types of modification.

3. Stereochemistry

The introduction of an additional sulfur atom in a PS linkage results in the generation of a chiral centre at each modified phosphorous atom, with the two possible stereoisomeric forms (designated S_p and R_p, respectively) (As shown above). As such, a fully PS backbone 20mer oligonucleotide is in fact a racemic mixture of the 2¹⁹ possible permutations (that is, more than half a million different molecules). The physicochemical properties of each stereocentre are distinct in terms of hydrophobicity/ionic character, nuclease resistance, target affinity and RNase H activity. In particular, a 3’-SpSpRp-5’ ‘stereochemical code’ contained within the ‘gap’ region of gapmer ASOs was found to be particularly active. Wave Life Sciences has developed a scalable method of synthesizing oligonucleotides with defined stereochemistry at each PS linkage, and is advancing oligonucleotide drugs with defined stereochemistry for various indications. However, they recently discontinued development of suvodirsen, a stereopure ASO designed to treat Duchenne muscular dystrophy (DMD) via skipping of dystrophin exon 51, owing to lack of efficacy in a phase I clinical trial114. Parallel clinical programmers with stereopure oligonucleotides targeting Huntington disease and C9ORF72 amyotrophic lateral sclerosis/frontotemporal dementia are ongoing.

It is interesting to speculate that the racemic mixtures of oligonucleotide drugs currently approved or in development contain many stereoisomers that exhibit low activity, thereby reducing the overall potency of the bulk mixture, and a small number of hyper-functional molecules. Identification of the most active stereoisomers would provide a major step forwards in oligonucleotide drug development, allowing for lower doses with more efficacious compounds. Notably, it has been suggested that a stereo-random mixture of S_p and R_p centres is required to balance stability and silencing activity. Nucleotide stereochemistry has also been exploited for aptamer development.

4. Nucleobase modification

Strategies to modify nucleobase chemistry are also being investigated. For example, pyrimidine methylation (5-methylcytidine and 5-methyluridine/ribothymidine) (As shown above) has the effect of increasing the oligonucleotide melting temperature by ~0.5 °C per substitution, and has been commonly incorporated into ASOs (such as those under development by Ionis Pharmaceuticals).

In addition, abasic nucleotides (that is, nucleotides lacking a nucleobase) have been used to abrogate miRNA-like silencing while maintaining on-target slicer activity and for allele-specific silencing of mutant HTT and ATXN3 transcripts

5. Terminal modification

Phosphorylation of the 5’ terminus of the siRNA guide strand is essential for activity, as this group makes an important contact in the MID domain of AGO2. Removal of this terminal phosphate group by cellular phosphatases therefore has the effect of reducing siRNA potency. The addition of a 5’-(E)-vinyl-phosphonate modification acts as a phosphate mimic that is not a phosphatase substrate. This modification also protects against exonuclease degradation and enhanced silencing in vivo. Similarly, terminal inverted abasic ribonucleotides have been used to block exonuclease activity. The conjugation of delivery-promoting moieties to oligonucleotide termini is discussed below.

6. Ribose sugar modification

Oligonucleotides are frequently modified at the 2’ position of the ribose sugar. Combinations of DNA (2’-deoxy) and RNA bases are critical to the activity of gapmer ASOs (that is, for generating RNase H substrate heteroduplexes), and are utilized on the 3’ termini of some siRNA designs in order to confer nuclease resistance.

Similarly, 2’-O-methyl (2’-OMe), 2’-O-methoxyethyl (2’-MOE) and 2’-Fluoro (2’-F) (As shown above) are among the most commonly used 2’ substituents. These modifications increase oligonucleotide nuclease resistance by replacing the nucleophilic 2’-hydroxyl group of unmodified RNA, leading to improved stability in plasma, increased tissue half-lives and, consequently, prolonged drug effects. Furthermore, these modifications also enhance the binding affinity of the oligonucleotide for complementary RNA by promoting a 3’-endo pucker conformation (RNA-like) of the ribose. These 2’-ribose modifications are not compatible with RNase H activity, meaning they are typically used for steric block oligonucleotides, or for the flanking sequences in gapmer ASOs. Although 2’ substitutions that enhance binding affinity are not improvements in delivery per se, they can compensate for limited drug bioavailability as the fraction of the injected dose that reaches its intended target is more active.

For steric block and gapmer ASOs, the oligonucleotide simply needs to bind to its cognate target (and support RNase H cleavage with a DNA gap in the case of the latter). For siRNAs, the situation is more com- plex as the oligonucleotides must maintain the capacity for loading into an AGO2 protein and support slicer cleavage. However, extensive chemical modification has been reported using 2’-OMe and 2’-F chemistries, and active siRNAs have been generated in which every single 2’-hydroxyl is modified, suggesting that the RNAi machinery is remarkably tolerant of chemical modification. Furthermore, full chemical modification has been shown to be highly important for the activity of siRNA bioconjugates (discussed below). Conversely, the introduction of 2’-ribose modifications at certain specific positions can ablate RISC loading and silencing activity. This phenomenon has been exploited in order to inactivate the passenger strand of the siRNA duplex, and thereby minimize its potential to mediate off-target silencing effects.

2’-MOE modifications are not typically incorporated into siRNA designs (the one exception being single-stranded siRNAs). Alnylam Pharmaceuticals has developed two patterns of siRNA chemical modification that form the basis for many of the products in their development pipeline. The first is standard template chemistry, which consist of an alternating pattern of 2’-F and 2’-OMe modifications at all ribose positions. This design was shown to increase siRNA potency by more than 500-fold relative to the unmodified PO siRNA in some cases. The now-discontinued drug revusiran, targeting TTR for transthyretin amyloidosis, is an example of a standard template chemistry siRNA design. The second approach is termed enhanced stability chemistry, in which siRNAs contain a greater proportion of 2’-OMe than standard template chemistry siRNAs and also incorporate PS linkages at the two inter- nucleotide bridges at the 3’ terminus of the guide strand and the 5’ termini of both strands. The approved Alnylam drug givosiran (FIG. 1h) is an example of an enhanced stability chemistry siRNA design.

2’-OMe modifications can also abrogate the immune responses that can be induced by ASOs, siRNAs and CRISPR guide RNAs (BOX 2). These oligonucleotide drugs have the potential to stimulate immune reactions in both sequence and chemistry-dependent ways via cellular pattern recognition receptors located in the cytoplasm or endosome. Specifically, the Toll-like receptors can induce the interferon response: TLR3 recognizes double-stranded RNA motifs; TLR7 and TLR8 recognize single-stranded RNA; and TLR9 recognizes unmethylated CpG dinucleotides. Similarly, the RIG-I and PKR systems also recognize double-stranded RNA in the cytoplasm. Some of these immunogenic effects of siRNAs can be ablated by the inclusion of 2’-OMe modifications at key positions. Conversely, incorporation of 5’-triphosphate-modified oligonucleotides, or conjugation with CpG motif-containing TLR9 agonists, and other chemical modifications have been used to generate therapeutic immunostimulatory oligonucleotides.

7. Bridged nucleic acids

Bridged nucleic acids (BNAs) are types of nucleotides in which the pucker of the ribose sugar is constrained in the 3’-endo conformation via a bridge between the 2’ and 4’ carbon atoms. The most commonly used variations are locked nucleic acid (LNA), 2’,4’-constrained 2’-O-ethyl (constrained ethyl) BNA (cEt) and, to a lesser extent, 2’‐O,4’‐C‐ethylene‐bridged nucleic acid (ENA) (As shown above). BNAs enhance both nuclease stability and the affinity of the oligonucleotide for target RNA (typically by an increase of 3–8°C in melting temperature per modified nucleotide in the case of LNA). BNA modifications have therefore been incorporated into the flanking regions of gapmers to improve target binding. As such, cEt-flanking 3–10–3 gapmers are more efficacious than the MOE 5–10–5 equivalents. Conversely, BNA nucleotides are not compatible with RNase H-mediated cleavage and so are excluded from the DNA gap region. LNA modifications have also been utilized in steric block ASOs, such as miRNA inihibitors. For example, miravirsen and cobomarsen are both full PS ASO mixmers containing DNA and LNA modifications distributed throughout their sequences. Conversely, tiny LNAs are short (8mer) fully LNA-modified oligonucleotides designed to simultaneously inhibit multiple members of a miRNA family (as these may execute redundant physiological functions) through complementarity to the miRNA seed sequence that is common between family member.

8. Alternative chemistries

Whereas the majority of oligo-nucleotides are derived from RNA or DNA, other chemistries have been developed that differ substantially from these natural archetypes. PMO (phosphorodiamidate morpholino oligonucleotide) is a charge-neutral nucleic acid chemistry in which the five-membered ribose heterocycle is replaced by a six-membered morpho- line ring (As shown above). Sarepta Therapeutics is developing PMO-based steric block ASOs for exon skipping in the context of DMD. To date, two PMO drugs have been approved by the FDA, eteplirsen and golodirsen, which target exons 51 and 53 of the dystrophin mRNA, respectively. Sarepta is also developing further ASO products based on the same chemistry targeting different dystrophin exons. Additionally, Nippon Shinyaku Pharma recently published encouraging data on a rival exon 53-targeting PMO (viltolarsen, NS-065/NCNP-01) and received marketing authorization in Japan in March 2020.

Notably, PMO backbone linkages contain chiral centres, meaning that PMO drugs are necessarily racemic mixtures. In contrast with PS modifications described above, the effects of defined PMO stereochemistry have not been explored to date.

Another strategy that has been explored is the use of peptide nucleic acid (PNA), a nucleic acid mimic in which a pseudo peptide polymer backbone substitutes for the PO backbone of DNA/RNA (As shown above). As both PMOs and PNAs are uncharged nucleic acid molecules, they can be covalently conjugated to charged delivery-promoting moieties such as cell-penetrating peptides (CPPs) (discussed below). Conversely, a dis- advantage of these chemistries is that both PMO and PNA interact minimally with plasma proteins, meaning that they are rapidly cleared via urinary excretion.

The use of a constrained DNA analogue that increases the stability of RNA target–oligonucleotide duplexes by 2.4 °C per modification, known as tricyclo-DNA (tcDNA) (As shown above), is also being investigated. Interestingly, systemically administered tcDNA ASOs were shown to exhibit activity in the brain, suggesting that these molecules have the capacity to deliver oligonucleotides across the BBB. Given the non-natural structures of PMO, PNA and tcDNA, these chemistries are less suitable to RNase H and RNAi applications but have instead been used in steric block oligonucleotides, and for splice correction in particular. However, tcDNA has recently been incorporated into the flanking sequences of a gapmer designed to silence mutant HTT transcripts.

A final example is the development of a novel short interfering ribonucleic neutral (siRNN) chemistry. These siRNN molecules contain a modified phosphotriester structure that neutralizes the charge of the equivalent unmodified PO/PS linkages, and thereby promotes their uptake in recipient cells. siRNNs act as prodrugs, which are converted to classical siRNAs by thioesterases in the cytoplasm.

9. Bioconjugation

The delivery potential of ASOs and siRNAs can be enhanced through direct covalent conjugation of various moieties that promote intracellular uptake, target the drug to specific cells/tissues or reduce clearance from the circulation. These include lipids (for example, cholesterol that facilitates interactions with lipoprotein particles in the circulation), peptides (for cell targeting and/or cell penetration), aptamers, antibodies, and sugars (for example, N-acetylgalactosamine (GalNAc)). Bioconjugates constitute distinct, homogeneous, single-component molecular entities with precise stoichiometry, meaning that high-scale synthesis is relatively simple and their pharmacokinetic properties are well defined. Furthermore, bionconjugates are typically of small size (relative to nanoparticle approaches, discussed below), meaning that they generally exhibit favourable bio- distribution profiles (on account of being able to reach tissues beyond those with discontinuous or fenestrated endothelia).

For siRNAs there are four termini to which conjugates could potentially be attached. However, conjugation to the 5’ terminus of the guide strand is avoided as this terminal phosphate makes specific contacts with side-chain residues within the MID domain of AGO2 that are required for RNAi activity. Conjugation to the passenger strand is generally preferred so as not to encumber the on-target silencing activity of the guide strand and, conversely, to diminish the off-target gene silencing potential of the passenger strand. Conjugates can be designed such that they are disassembled following cellular entry. This can be achieved by using acid-labile linkers that are cleaved in the endosome, disulfide linkers that are reduced in the cytoplasm or Dicer substrate-type siRNA designs.

A common theme in oligonucleotide bioconjugation approaches is the promotion of interactions between the conjugate and its corresponding cell surface receptor protein, leading to subsequent internalization by receptor-mediated endocytosis. The interaction of bio-conjugates with cell type-associated receptors thereby enables targeted delivery to specific tissues, or cell types within a tissue. To date, the physiological effects of saturating such receptor pathways have not been extensively studied.

10. Lipid conjugates

Covalent conjugation to lipid molecules has been used to enhance the delivery of siRNAs and antagomir ASOs. Cholesterol siRNAs (conjugated to the 3’terminus of the passenger strand) (As shown above) have been utilized for hepatic gene silencing (for example, Apolipoprotein B, Apob) and, more recently, to silence myostatin (Mstn) in murine skeletal muscle (a target organ in which it has historically been particularly challenging to achieve effective RNAi) after systemic delivery. Other lipid derivatives have also been exploited to enhance siRNA delivery. For example, siRNAs conjugated to α-tocopherol (vitamin E) were reported to induce potent silencing of Apob in the mouse liver. In this case, the lipid moiety was con- jugated to the 5! terminus of the passenger strand of a Dicer substrate siRNA 27/29mer duplex. Upon cellular entry, the siRNA is cleaved by Dicer so as to generate the mature 19’+ 2mer active RNAi trigger and to simultaneously cleave off the α-tocopherol. Similarly, siRNAs conjugated to long-chain (>C18) fatty acids via a trans-4-hydroxyprolinol linker attached to the 3’ end of the passenger strand were capable of inducing comparable levels of Apob silencing to cholesterol-conjugated siRNAs. The in vivo activity of lipid-conjugated siRNAs was demonstrated to be dependent on their capacity to bind to lipoprotein particles (for example, HDL and LDL) in the circulation and, thereby, hijack the endogenous system for lipid transport and uptake. Pre-assembly of cholesterol siRNAs with purified HDL particles resulted in enhanced gene silencing in the liver and jejunum relative to cholesterol siRNAs alone. Furthermore, lipoprotein particle pre-assembly was also shown to affect siRNA biodistribution, with LDL siRNA particles taken up almost exclusively in the liver, and HDL siRNA particles primarily taken up by the liver, and also the adrenal glands, ovary, kidney and small intestine. Accordingly, endocytosis of cholesterol siRNAs was shown to be mediated by scavenger receptor-type B1 (SCARB1, SR-B1) or LDL receptor (LDLR) for HDL and LDL particles, respectively. In vivo association of siRNAs with the different classes of lipoprotein is governed by their overall hydrophobicity, with the more hydrophobic conjugates preferentially binding to LDL and the less lipophilic conjugates preferentially binding to HDL.

11. GalNAc conjugates

GalNAc is a carbohydrate moiety that binds to the highly liver-expressed asialoglycoprotein receptor 1 (ASGR1, ASPGR) with high affinity (K_d = 2.5 nM) and facilitates the uptake of PO ASOs and siRNAs into hepatocytes by endocytosis. ASGR1 is very highly expressed in the liver, and is rap- idly recycled to the cell membrane, making it an ideal receptor for effective liver-targeted delivery. The interaction between GalNAc and ASGR1 is pH-sensitive, such that dissociation of the receptor and oligonucleotide conjugate occurs during acidification of the endosome. The GalNAc moiety is subsequently subject to enzymatic degradation that liberates the oligonucleotide. GalNAc-conjugated ASOs are preferentially delivered to hepatocytes in vivo, whereas unconjugated ASOs are primarily detected in non-parenchymal liver cells.

Typically, a triantennary GalNAc structure (As shown above) is used as the conjugated moiety, although there are other structural variants. GalNAc conjugation enhanced ASO potency by ~7-fold in mouse, specific to the liver, and by ~30-fold in human patients. As such, GalNAc conjugation is now one of the leading strategies for delivering experimental oligonucleotide drugs currently in development, given its high liver silencing potential, small size relative to nanoparticle complexes, defined chemical composition and low cost of synthesis. In particular, GalNAc conjugation features heavily in the drug development pipelines of several pharma companies, most notably Alnylam, who are developing drugs for the treatment of diseases such as haemophilia A and B and primary hyperoxaluria type 1. Furthermore, a GalNAc-conjugated siRNA, givosiran (developed by Alnylam), received FDA approval in November 2019. Givosiran is a GalNAc-conjugated, blunt-ended, enhanced stability chemistry siRNA duplex targeting 5’-aminolevulinate synthase 1 (ALAS1) for the treatment of acute hepatic porphyria. Similarly, inclisiran, a second GalNAc- conjugated siRNA containing 2’-F, 2’-OMe and PS modifications (developed by Alnylam/The Medicines Company and acquired by Novartis), is in late-stage clinical trials for the treatment of familial hypercholesterolaemia. Inclisiran targets PCSK9 (pro- protein convertase subtilisin/kexin type 9), which is a circulating factor that negatively regulates expression of LDLR and is primarily expressed in the liver. Hepatic PCSK9 knockdown therefore increases the availability of LDLR to remove LDL cholesterol from the circulation. Subcutaneous injection of inclisiran resulted in long-term downregulation of circulating PCSK9 and LDL cholesterol (~6 months) suggesting that an infrequent treatment regimen may be a sufficient lipid-lowering strategy.

Numerous additional pharmaceutical companies-namely Dicerna Pharmaceuticals, Silence Therapeutics, Arbutus Biopharma and Arrowhead Pharmaceuticals — are also developing GalNAc-conjugated oligonucleotide products.

12. Antibody and aptamer conjugates

Although there is a plethora of technologies capable of delivering nucleic acids to hepatic cells, there is still a need for strategies that can target cell surface receptors specific to other tis- sues. Antibodies have been used as delivery vehicles for other kinds of drugs, although their utility for oligo- nucleotide delivery is still in the early stages of development. Specific interactions between an antibody and a cell surface receptor have the potential to enable delivery to tissues and/or cell subpopulations that are not accessible using other technologies. Various receptors have been successfully targeted for siRNA delivery (as shown above), including the HIV gp160 protein, HER2, CD7 (T cell marker), CD71 (transferrin receptor, highly expressed in cardiac and skeletal muscle) and TMEFF2. Similarly, ASOs have also been conjugated with antibodies against CD44 (a neural stem cell marker), EPHA2 and EGFR. In these cases, the ASO was delivered as a duplex with a DNA carrier strand to which the antibody was attached via click chemistry. Such a design allows the DNA passenger to be degraded after cellular entry, thereby releasing the ASO from the complex. Antibody–siRNA and antibody–ASO conjugates targeting tissues such as skeletal muscle are currently being developed by Avidity Biosciences and Dyne Therapeutics, respectively.

Similarly, the conjugation of therapeutic oligo- nucleotides to nucleic acid aptamers (BOX 1) has also been explored for enhancing delivery of siRNAs and ASOs to specific target cells (as shown above). Aptamers can be considered ‘chemical antibodies’ that bind to their respective target proteins with high affinity, but present numerous advantages over antibodies as they are simple and inexpensive to manufacture (that is, by chemical synthesis), are smaller in size and exhibit lower immunogenicity.

13. Peptide conjugates

Peptides are an attractive source of ligands that may confer tissue/cell-targeting, cell-penetrating (that is, CPPs) or endosomolytic properties onto therapeutic oligonucleotide conjugates. CPPs (also known as protein transduction domains) are short (typically Drosophila Antennapedia protein) and Transportan (a chimeric pep- tide consisting of part of the galanin neuropeptide fused to the wasp venom, mastoparan)) or are based on polymers of basic amino acids (that is, arginine and lysine). One of the most promising applications of CPPs is their direct chemical conjugation to charge-neutral ASO chemistries, such as PMO and PNA.

Several groups have pioneered the use of peptide–PMO (PPMO) conjugates (as shown above) for the treatment of various diseases, most notably for dystrophin splice switching in the context of DMD. Early PPMO dystrophin exon skipping studies demonstrated efficacy using (RXR)4-PMO and the ‘B’ peptide (with sequence (RXRRBR)2XB), where X and B are 6-aminohexanoic acid and β-alanine spacer residues, respectively. The spacer residues are important for the optimal positioning of the charged arginine side chains. This approach was further modified by generating a chimeric peptide consisting of B peptide fused with a muscle-targeting peptide (MSP). The resulting B–MSP–PMO conjugates demonstrated further dystrophin restoration efficacy in the mdx mouse model of DMD (although the relative arrangement of the peptide constituents was found to be important, with MSP–B–PMO exhibiting low activity). Exon skipping activity has also been reported when PMOs were conjugated to a different muscle-targeting peptide (M12) identified by phage display, although activity in the heart was minimal. Subsequently, several series of peptides known as ‘Pip’s (PMO/PNA internalization peptide) consisting of R, B and X amino acids with an internal core containing hydrophobic residues have been developed. Current-generation Pip–PMO conjugates (FIG. 4e) are much more potent than naked PMO in dystrophic animal models and, importantly, reach cardiac muscle (a tissue critical to the lethality of DMD) after systemic delivery. The PPMO hydrophobic core is required for cardiac delivery, but can itself be scrambled, inverted, or individual residues substituted with only minimal changes to efficacy. A major challenge for PPMO technology is toxicity, with evidence of renal damage in both rat (at very high doses) and cynomolgus monkey studies using arginine-rich CPP–PMOs. Notably, the arginine content of the CPP is correlated with both exon skipping activity and nephrotoxicity, and so current research efforts are directed towards the optimization of peptide chemistry to mitigate renal toxicity without compromising splice correction efficacy. Sarepta Therapeutics is developing SRP-5051, a PPMO designed to skip dystrophin exon 51. Additionally, PepGen Ltd is commercializing PPMO technology.

PPMO uptake is energy dependent and appears to involve distinct endocytic pathways in skeletal and cardiac muscle cells. It has been reported that treatment of chloroquine can enhance PPMO activity, suggesting that many conjugate molecules may not escape the endolyosomal compartment. PPMOs have also been shown to spontaneously form micelles of defined sizes and surface charge, meaning that they are more readily taken up by endocytosis, in part mediated by scavenger receptors.

PPMO technology has also been demonstrated to be effective for targeting CUG repeat-expanded transcripts in the context of myotonic dystrophy type I (DM1) (whereas naked PMO was completely ineffective), for splice correction to restore BTK expression for the treatment of X-linked agammaglobulinaemia and for delivery to the CNS in animal models of spinal muscular atroph. Similarly, brain delivery (to the cerebellum and Purkinje cells in particular) of an arginine-rich CPP–PMO conjugate after systemic delivery has also been demonstrated.

PPMOs are also promising antibacterial (as they are capable of traversing the bacterial cell wall) and antiviral agents. Intranasal administration of (RXR)4 peptide–PMO conjugates targeting an essential bacterial gene acpP in murine infection models was shown to be bactericidal and increased survival. Further, arginine-rich peptide–PMO conjugates were shown to exert protective effects in murine viral infection models of SARS-CoV and Ebola.

Conjugation of peptides to charged-backbone oligonucleotides has been explored to a much lesser extent, as charge–charge interactions between the constituents complicate synthesis and purification, and conjugates may have the potential to self-aggregate. Nevertheless, there are a few examples of such conjugates. It was recently demonstrated that conjugation of an ASO gapmer to a ligand for the glucagon-like peptide 1 receptor (GLP1R) conferred targeted gene silencing in pancreatic β-cells, the pancreas being a particularly challenging organ to deliver to. In this case, the targeting moiety was a 40-amino acid peptide consisting of a modified GLP1 sequence covalently conjugated to the ASO via the carboxy terminus. Peptide conjugation has also been explored for siRNA delivery. For example, the cyclic RGD peptide (recognized by αvβ3 integrin receptors) has successfully been used to deliver anti-VEGFR2 siRNA conjugates to mouse tumours. Similarly, the CPPs TAT (48–60) and penetratin have been utilized as siRNA conjugates for delivery to the lung via the intratracheal route. Although modest silencing of the target gene was observed, administration of the unconjugated peptides alone also exhibited a repressive effect. Furthermore, treatment with the penetratin–siRNA conjugates was associated with the release of pro-inflammatory markers TNF, IL-12 p40 and IFNα. These observations high- light that potential peptide-mediated non-specific effects on gene expression and innate immune activation must be carefully considered.

14. Nanocarriers

Advances in nanotechnology and material science offer advantages and potential solutions to the challenge of oligonucleotide drug delivery, in particular the requirements for crossing biological barriers and transmembrane intracellular delivery. The major advantages of nanoparticle delivery systems include bespoke optimization of nanoparticle biophysical (for example, size, shape and chemical/material composition) and biological (for example, ligand functionalization for targeting) properties, allowing for highly tailored delivery platforms. A wide range of nanocarriers for nucleic acid drug delivery are at various stages of development, including non-covalent complexation with cationic polymers (for example, polyethylenimine), dendrimers, CPPs (for example, MPG-8, PepFect6, RVG-9R, and Xentry-KALA) and inorganic methods (for example, calcium phosphate nanoparticles). Below, we focus on lipid-based formulations for oligonucleotide delivery and emerging novel approaches including endogenously derived exosomes, SNAs and self-assembling DNA nanostructures.

15. Lipoplexes and liposomes

Formulation with lipids is one of the most common approaches to enhancing nucleic acid delivery. Mixing polyanionic nucleic acid drugs with lipids leads to the condensing of nucleic acids into nanoparticles that have a more favourable surface charge, and are sufficiently large (~100 nm in diameter) to trigger uptake by endocytosis. Lipoplexes are the result of direct electrostatic interaction between polyanionic nucleic acid and the cationic lipid, and are typically a heterogeneous population of relatively unstable complexes. Lipoplex formulations need to be prepared shortly before use, and have been success- fully used for local delivery applications. By contrast, liposomes comprise a lipid bilayer, with the nucleic acid drug residing in the encapsulated aqueous space. Liposomes are more complex (typically consisting of cationic or fusogenic lipids (to promote endosomal escape) and cholesterol PEGylated lipid) and exhibit more consistent physical properties with greater stability than lipoplexes. For example, some lipid nano- particles (LNPs), also known as stable nucleic acid lipid particles (as shown above), are liposomes that contain ionizable lipid, phosphatidylcholine, cholesterol and PEG–lipid conjugates in defined ratios and have been successfully utilized in multiple instances. Landmark examples are the silencing of hepatitis B virus and APOB by siRNAs in preclinical animal studies and, more recently, the approval of patisiran, an siRNA that is delivered as an LNP formulation. Encapsulation of nucleic acid cargos provides a means of protection from nuclease digestion in the circulation and in the endosome. Additionally, ionizable LNPs also associate with APOE, which further facilitates liver uptake via LDLR-mediated endocytosis. Similarly, LNPs containing lipidoid or lipid-like materials have demonstrated robust siRNA-mediated silencing in rodents and non-human primates.

A disadvantage of LNPs is that their delivery is primarily limited to the liver and reticuloendothelial system as the sinusoidal capillary epithelium in this tissue pro- vides spaces large enough to allow the entry of these relatively large nanoparticles. However, local delivery of LNPs has been used to successfully deliver siRNAs to the CNS after intracerebroventricular injection. Conversely, the large size of nanoparticles is advantageous as it essentially precludes renal filtration and permits delivery of a higher payload.

LNPs can be further functionalized with peptides, PEG or other ligands that confer cell-specific targeting (for example, GalNAc (hepatocytes), anisamide (lung tumours), strophanthidin (various tumours) and vitamin A (hepatic stellate cells)). Notably, an increase in the complexity of LNPs complicates manufacture and may increase their toxicity, which is a major concern that may limit their clinical utility. For example, LNP siRNA particles (such as patisiran) require pre- medication with steroids and antihistamines to abrogate unwanted immune reactions.

16. Exosomes

An area of nanotechnology that is gaining interest is based on the application of natural bio- logical nanoparticles known as exosomes (a class of extracellular vesicle). Exosomes are heterogeneous, lipid bilayer-encapsulated vesicles approximately 100 nm in diameter that are generated as a result of the inward budding of the multivesicular bodies. Exosomes are thought to be released into the extracellular space by all cells, where they facilitate intercellular communication via the transfer of their complex macromolecular cargoes (that is, nucleic acids, proteins and lipids). Exosomes present numerous favourable properties in terms of oligonucleotide drug delivery: exosomes are capable of traversing biological membranes, such as the BBB; the presence of the marker protein CD47 protects exosomes from phagocytosis, thereby increasing their circulation time relative to liposomes; exosomes are considered non-toxic and have been safely administered to patients with graft-versus-host disease; exosomes have the potential to be produced in an autologous manner; exosomes from some sources have been shown to have inherent pro-regeneration and anti-inflammatory properties that may augment the effects of therapeutic oligonucleotide delivery; and engineered exosomes can serve as a modular platform whereby combinations of therapies and/or targeting moieties can be deployed.

A major challenge for exosome therapeutics is the efficient loading of therapeutic oligonucleotide cargo. Vesicles can be loaded either endogenously (for example, by overexpression of the cargo in the producer cell line) or exogenously (for example, by electroporation, sonication, co-incubation with cholesterol-conjugated siRNAs and so forth). The loading of exosomes with splice-switching PMOs has been achieved via conjugation with the CP05 peptide, which binds to CD63 (a marker commonly found on exosomes) so as to decorate the exosomes with PMO cargo.

The pattern of exosome biodistribution can be favourably altered through the display of surface ligands, such as peptides like rabies virus glycoprotein (RVG) to enhance brain penetration and facilitate delivery to cells within the nervous system (as shown above) or GE11 that promotes binding to tumour cells by interacting with EGFR. Similarly, exosomes decorated with an RNA aptamer targeting PSMA (prostate-specific membrane antigen) were capable of delivering siRNAs to xenograft tumours and inducing tumour regression.

Methods for the manufacture of therapeutic exosomes at high scale, including clinical grade, have been reported. The use of mesenchymal stem cell lines (with the potential for immortalization) and culture in bioreactors enables large volumes of exosome-containing conditioned media to be generated. Subsequently, methods such as tangential flow filtration and size-exclusion liquid chromatography provide a scalable means of isolating therapeutic exosomes from these supernatants. Therapeutic applications of engineered exosome technology are presently at an advanced preclinical stage for two companies: Codiak Biosciences and Evox Therapeutics.

17. Spherical nucleic acids

An alternative nanoparticle- based delivery strategy is the SNA approach. SNA particles consist of a hydrophobic core nanoparticle (comprising gold, silica or various other materials) that is decorated with hydrophilic oligonucleotides (for example, ASOs, siRNAs and immunostimulatory oligo- nucleotides) that are densely packed onto the surface via thiol linkages (as shown above). In contrast to other nano- particle designs, SNA-attached oligonucleotides radiate outwards from the core structure. While exposed, the oligonucleotides are protected from nucleolytic degradation to some extent as a consequence of steric hindrance, high local salt concentration, and through interactions with corona proteins.

SNA particles carrying an siRNA targeting the anti-apoptotic factor Bcl2l12 were able to promote tumour apoptosis, reduce the tumour burden and extend survival in glioblastoma xenograft-bearing mice. Importantly, this study demonstrated that SNAs have the potential to cross the BBB in both tumour-bearing mice (with impaired BBB integrity) and also in wild- type mice, although the majority of SNA particles were deposited in the liver and kidneys. SNA particles have also been applied for topical delivery to skin keratinocytes in the context of diabetic wound healing (that is, GM3S-targeting siRNA) and psoriasis (that is, TNF-targeting ASO). SNA particles are currently being commercialized for oligonucleotide delivery applications by Exicure, Inc.

18. DNA nanostructures

DNA nanostructures, of which there are many varieties, have also been utilized for oligonucleotide delivery. These structures include DNA origami, whereby long DNA molecules are held in defined structures using short DNA ‘staples’ that enable a wide variety of complex shapes to be formed, including polygonal nanostructures such as DNA cages. DNA nanostructures typically self-assemble owing to base pairing complementarity of their constituent parts, and can be designed with precise geometries such that their physical properties (for example, size, flexibility and shape) can be fine-tuned in order to maximize their delivery potential. DNA nanostructures used for nucleic acid delivery applications will typically be modular structures that incorporate nucleic acid drugs (and targeting ligands such as aptamers) within the design of the structure itself. For example, DNA nanostructures have been designed that incorporate ASOs (as shown above), siRNAs and immunostimulatory oligonucleotides displayed on the structure surface. A highly interesting property of DNA nanostructures is that they have been reported to not accumulate in the liver, and can be engineered to be small (~20 nm), meaning extrahepatic delivery is possible. However, further reducing the size of DNA nanostructures without additional functionalization will likely result in enhanced renal filtration, and overall lower bioavailability.

19. Stimuli-responsive nano-technology

Stimuli-sensitive, activatable drug delivery nanotechnologies are emerging as oligonucleotide delivery solutions. Activatable CPP conjugates consist of an oligonucleotide covalently attached to a peptide that is folded into a hair- pin structure. One half of the hairpin is arginine rich and positively charged, whereas the second half of the peptide is negatively charged and acts as a neutralizing inhibitory domain. The loop of the hairpin contains an enzymatic cleavage motif, allowing for activation of the conjugate when it reaches the desired site of action. Activation by matrix metalloproteinase 2 (MMP2) has been utilized to enable targeted delivery of siRNA to hepatoma cells in vitro and in xenograft tumours in vivo. Similarly, dynamic polyconjugates consist of oligonucleotides conjugated to a scaffold that is linked to multiple delivery-assisting moieties. For example, an siRNA-dynamic polyconjugate consisting of a PBAVE scaffold polymer linked to PEG and NAG (N-acetyl glucosamine, for liver targeting) moieties by acid-labile linkers induced potent gene silencing in mouse hepatocytes after intravenous administration. The acidic environment of the endosome induces cleavage of the PEG and NAG groups, leading to exposure of the PBAVE tertiary amines, buffering of luminal pH and consequent endosomal escape.

Even more complex ‘smart’ delivery vehicles are possible with DNA nanotechnology. For example, DNA origami has been used to generate a box that is opened and incorporates two structure-shifting aptamer ‘locks’. When both aptamers interact with their target proteins, the lock opens and the DNA box changes conformation to release its contents. This technology was utilized to deliver gold nanoparticles and antibody fragments but could potentially be modified for nucleic acid delivery. Such logic-gated delivery vehicles present numerous advantages as therapeutic payloads could be concentrated at the desired sites of action, leading to higher efficacy and reduced off-target effects. Technologies that can confer stimuli-responsiveness (for example, to pH, temperature, redox state, enzymatic activity, magnetic fields and light) in nanoparticle drug delivery systems have been reviewed elsewhere.

Challenges and considerations

The establishment of therapeutic platforms capable of delivering oligonucleotide drugs to specific organs or tissues will likely involve defined patterns of chemical modification, combined with conjugation/complexation strategies that confer predictable pharmacokinetic and pharmacodynamic properties, and well-understood mechanisms of action. In this manner, oligonucleotide therapeutics have the potential to extend the range of possible pharmaceutical targets and provide a means by which new drugs can be rapidly developed to meet unmet and emerging clinical needs, without the need to ‘start from scratch’. Such a situation is exemplified by the case of milasen, an ASO drug designed as a personalized drug for a single patient suffering from Batten disease. Milasen was designed using the same full PS–2’-MOE design as nusinersen, an FDA-approved chemistry that distributes favourably throughout the CNS after intrathecal administration, resulting in a period of less than 1 year between identification of the patient’s mutation and the first administration of the ASO drug. Whether such a model of drug development is scalable to more patients remains to be seen, although the n-Lorem Foundation aims to use bespoke ASO drugs to treat patients with ultra-rare diseases, where conventional clinical trials are impossible. However, a key issue will be how the safety of different drug sequences using the same chemistry/delivery platform will be viewed by regulatory authorities.

Although LNPs and GalNAc conjugates offer excel- lent hepatic delivery in both preclinical and clinical studies, systemic delivery beyond the liver will require further investigation, innovation and development. In many cases, approved oligonucleotide drugs will likely be extremely, and possibly prohibitively, expensive. For example, nusinersen currently costs $750,000 for the first year and $375,000 in subsequent years. Similarly, the cost of eteplirsen is $300,000 per annum. The cost– benefit ratio for highly effective, life-changing medications such as nusinersen is likely to be favourable. By contrast, reimbursing the cost of eteplirsen, which has demonstrated very limited efficacy, will be much more difficult to justify. Importantly, drug costs should also be weighed against the cost of care for patients left untreated. As oligonucleotide therapeutics are further combined with novel delivery modalities, such advances may compound the cost of materials. However, the improved efficacy and/or better targeted delivery afforded by such delivery technologies may mean that lower doses of drug can be administered, which may conversely reduce costs. Another key consideration is safety. While the immune-stimulating properties of standard nucleic acid modification chemistries are relatively well understood, the potential immunogenic- ity of delivery agent components or ligand conjugates may present additional challenges to safe and effective oligonucleotide drug delivery.

Given the large number of nucleic acid chemistries, delivery technologies and therapeutic modalities, direct head-to-head comparisons are unlikely to be possible in many c