Therapeutics & platforms
ADAR RNA editing therapeutics
Hydrolytic adenosine deamination by ADAR, directed by an antisense oligonucleotide: reversible correction without a DNA break, bounded by A-to-I chemistry and by hepatic delivery.
Genome editing fixes the cause of a disease permanently — and gets things wrong permanently too. A double-strand break invokes repair whose outcome is only partly controlled, and whatever change results is inherited by every daughter cell. RNA editing is a different proposition: it alters the transcript rather than the gene, acts temporarily, and requires no break at all.
The enzyme is already in the cell
ADAR — adenosine deaminase acting on RNA — is a constitutively expressed human family, ADAR1 and ADAR2. Their natural role is to mark self double-stranded RNA so that innate immune sensors do not read it as viral.
The chemistry is a single reaction: hydrolytic deamination at position C6 of adenosine converts it to inosine. What makes this useful is how the cell’s reading machinery interprets the product. Ribosomes, the spliceosome and reverse transcriptase all read inosine as guanosine, so in biological terms an A has become a G.
ADAR’s substrate is double-stranded RNA, and that supplies the therapeutic trick: what is delivered is not an enzyme but a short antisense oligonucleotide complementary to the target mRNA. Hybridisation creates a local duplex that endogenous ADAR recognises and processes. The edited position is specified by a deliberate mismatch — a cytidine placed opposite the target adenosine, so that the A·C pair flips the base into the conformation the enzyme prefers. No foreign protein is introduced, so there is no immune response against an editor and no vector packaging limit to respect.
What bounds the approach
One chemistry. Only A-to-I, read as A-to-G. That corrects pathogenic G>A mutations — a real but limited class. The canonical example is the PiZ allele of SERPINA1 in alpha-1 antitrypsin deficiency, where the E342K substitution arose exactly as a G>A. Beyond direct correction, editing is also used indirectly: removing a stop codon, altering a splice site, or knocking out a start codon or a post-translational modification site.
Sequence context. ADAR has strong neighbour preferences: a 5’ U or A is favourable, a 5’ G distinctly unfavourable. Some clinically interesting sites sit in poor context, and that has to be compensated in duplex design rather than by choosing a different target.
Bystander editing. Adjacent adenosines within the same duplex are edited on their own account and must be protected by local chemical modification or by adjusting duplex length. A separate hazard is transcriptome-wide editing where the oligonucleotide is incidentally complementary to another mRNA; approaches that overexpress ADAR amplify this risk, which is why recruiting the enzyme at its endogenous level is generally preferred.
The oligonucleotide’s own chemistry. This is the platform’s central tension. Nuclease resistance and tissue residence come from a phosphorothioate backbone and 2’ ribose modification. But ADAR has to engage the duplex physically, and a fully 2’-O-methylated strand is no longer a substrate. Design is therefore a compromise between stability and recognition, resolved by placing modifications in a mosaic along the strand rather than uniformly.
Delivery. Conjugation to triantennary GalNAc addresses the oligonucleotide to the asialoglycoprotein receptor on hepatocytes — efficient and well established. Outside the liver there is no reliable equivalent, and that restricts the indication list more sharply than the editing biochemistry does.
Finally, the effect is transient by construction: transcripts turn over, the edited pool dilutes, and treatment must be repeated. That is simultaneously the platform’s chief drawback and its chief safety guarantee.