RNA Editing

Why RNA editing matters

• It breaks the one-gene-one-transcript assumption. The central dogma implies the mRNA faithfully copies the gene, but editing means the sequence that reaches the ribosome can differ from the DNA at defined positions. A single genomic locus can therefore yield edited and unedited protein isoforms in different tissues, at different developmental stages, or under different conditions.
• It is essential for a working nervous system. ADAR2 editing of the GluA2 AMPA-receptor Q/R site is nearly 100% efficient in healthy neurons and switches a glutamine to an arginine that blocks calcium flux through the channel pore. Mice lacking ADAR2 die of seizures within about three weeks; replacing their Gria2 gene with a pre-edited version rescues them entirely — proof this single edit is the enzyme's essential job.
• It marks self RNA and prevents autoinflammation. ADAR1 editing of endogenous double-stranded RNA (largely inverted Alu repeats) tags it as self. Without that mark, the cytosolic sensor MDA5 mistakes the cell's own duplex RNA for a virus and triggers a chronic type-I interferon response — the basis of Aicardi–Goutières syndrome.
• It builds two lipid-transport proteins from one gene. APOBEC1-mediated C-to-U editing of apolipoprotein B mRNA in the intestine produces ApoB48, the scaffold of chylomicrons that carry dietary fat, while the unedited liver transcript makes full-length ApoB100 for LDL. One editing switch, two arms of lipid metabolism.
• It is a lifetime-scale adaptation in cephalopods. Squid and octopus recode tens of thousands of mRNA sites, enriched in neural genes, and can retune ion channels in response to water temperature — physiological acclimation encoded in RNA rather than DNA.
• It confounds genome sequencing. Because inosine is read as guanosine by reverse transcriptase, A-to-I edits appear as A-to-G differences in cDNA. Distinguishing a genuine DNA variant from an RNA edit requires comparing DNA and RNA from the same sample — a routine headache in variant calling.
• It is a drug-delivery target in its own right. Programmable editing platforms (guide-RNA-directed recruitment of endogenous ADAR, and engineered dCas13–ADAR fusions such as REPAIR) aim to correct disease-causing G-to-A mutations at the RNA level, a reversible alternative to permanent DNA editing.

Common misconceptions

• "RNA editing changes the gene." It does not. The DNA is untouched; only the transcript is chemically altered. The same gene keeps producing unedited RNA, and the change is not heritable through the germline via the edit itself.
• "Editing and splicing are the same thing." Splicing removes whole introns/exons and never alters a retained base; editing transmutes individual bases in place. They can act on the same transcript, and an edit can even create the splice site that the spliceosome then uses — but the mechanisms are distinct.
• "Inosine is a typo the cell fixes." Inosine is the intended product. It is a real, stable modified nucleoside that the translational and reverse-transcription machinery deliberately interprets as guanosine; it is not an error awaiting repair.
• "Editing is rare and exotic." A-to-I editing occurs at millions of sites in the human transcriptome — although the overwhelming majority are in non-coding Alu repeats and do not recode protein. Recoding edits are the rare, functionally dramatic minority.
• "ADAR needs a template like DNA polymerase." ADAR needs double-stranded RNA structure, not a base-pairing template that dictates the product. It flips out a target adenosine and deaminates it; specificity comes from the local RNA fold and, in cephalopod-style guided systems, from complementary guide strands — not from copying a template.
• "C-to-U and A-to-I are done by the same enzyme." They are separate families: ADARs (adenosine deaminases acting on RNA) do A-to-I on double-stranded RNA; APOBEC/AID cytidine deaminases do C-to-U, and APOBEC1 needs a protein cofactor (A1CF or RBM47) plus a downstream mooring sequence to hit the right cytidine.

How RNA editing works

The two mammalian workhorses are both hydrolytic deaminases, but they attack different bases and rely on different targeting logic. A-to-I editing is carried out by the ADAR family. An ADAR enzyme requires a double-stranded RNA substrate, which in cells usually forms when an exon base-pairs with a nearby, inverted intronic repeat — very often an Alu element in humans. The enzyme's double-stranded RNA-binding domains clamp the duplex, and a base-flipping mechanism swings the target adenosine out of the helix into the catalytic pocket. There ADAR hydrolytically removes the amino group at the C6 position of adenine, yielding inosine. Because inosine pairs like guanosine, everything downstream — the ribosome during translation, the spliceosome, and reverse transcriptase during sequencing — reads the edited position as G. Humans have three genes: ADAR1 (interferon-inducible; its p150 isoform edits cytoplasmic duplex RNA to mark self), ADAR2/ADARB1 (the principal recoding enzyme, dominant in brain), and catalytically dead ADAR3.

C-to-U editing in mammals is dominated by APOBEC1, a cytidine deaminase of the AID/APOBEC family. On its own APOBEC1 is not sequence-specific enough; it works as a holoenzyme with the cofactor A1CF (APOBEC1 complementation factor) or the alternative cofactor RBM47, which recognize an 11-nucleotide "mooring sequence" located 4–8 nucleotides downstream of the target cytidine. The canonical substrate is apolipoprotein B mRNA, where APOBEC1 deaminates cytidine 6666, changing the codon CAA (glutamine 2153) into UAA — a premature stop. In the liver the transcript is left unedited and the ribosome makes the 4536-residue ApoB100; in the intestine editing is switched on and translation stops early, producing ApoB48, roughly the N-terminal 48% of the protein and the backbone of chylomicrons.

The functional consequences fall into a few classes. Recoding changes an amino acid: the GluA2 Q/R site (CAG glutamine → CGG arginine) is the paradigm, and similar edits retune serotonin 5-HT_2C receptors, kainate receptors, and the Kv1.1 potassium channel. Splice-site editing creates or destroys the AG/GU dinucleotides the spliceosome reads, changing which exons appear — ADAR2 even edits its own pre-mRNA to create a new 3′ splice acceptor, an autoregulatory feedback loop. Stop-codon editing, as in apoB, truncates or read-throughs proteins. And editing within microRNAs and their target sites can redirect an entire regulatory network by changing which mRNAs a miRNA silences. Editing efficiency at any site ranges from a few percent to essentially 100%, and that dial is itself regulated by cell type, RNA structure, and enzyme abundance.

RNA editing vs alternative splicing

Feature	RNA editing	Alternative splicing
Unit of change	Single base	Whole exon / intron
Chemistry	Deamination (A→I, C→U)	Phosphodiester cut & ligation
Machinery	ADAR, APOBEC1 deaminases	Spliceosome (snRNPs)
Sequence altered in place?	Yes — base transmuted	No — retained bases unchanged
Can change one amino acid?	Yes (recoding)	Only by including/excluding exons
Can create a stop codon?	Yes (apoB CAA→UAA)	Indirectly, via frame shift
Interaction	Can create/abolish splice sites	Provides the introns that fold ADAR duplexes
Read as by sequencing	A→G (inosine) or C→U	Different exon junctions

A-to-I (ADAR) vs C-to-U (APOBEC) editing

Property	A-to-I editing	C-to-U editing
Enzyme family	ADAR (ADAR1, ADAR2, ADAR3)	APOBEC / AID (mainly APOBEC1)
Base change	Adenosine → inosine	Cytidine → uridine
Read by ribosome as	Guanosine (I ≈ G)	Uridine (U)
Substrate requirement	Double-stranded RNA fold	Single-stranded RNA + mooring sequence
Cofactor	None obligatory (dsRNA suffices)	A1CF or RBM47 required
Canonical target	GluA2 Q/R site; Alu repeats	Apolipoprotein B (C6666)
Number of sites (human)	Millions (mostly Alu)	Tens (highly restricted)
Disease link	Aicardi–Goutières (ADAR1 loss), ALS, cancer	Dysregulated APOBEC → mutagenesis

Famous experiments and history

• Benne and trypanosome kinetoplasts (1986). Rob Benne's group found that the mitochondrial coxII mRNA of Trypanosoma brucei contained four uridines with no counterpart in the DNA. They coined the term "RNA editing" for this guide-RNA-directed insertion of bases — the first demonstration that a transcript's sequence could deviate from its gene.
• Scott, Chan, and apolipoprotein B (1987). Two groups independently reported that intestinal apoB mRNA carried a C-to-U change absent from the gene, converting a glutamine codon into a stop and producing the shorter ApoB48. This was the first substitutional editing described in animals and remains the textbook C-to-U example.
• Bass and Weintraub's unwinding activity (1988). Studying Xenopus egg extracts, Brenda Bass and Harold Weintraub described an activity that appeared to "unwind" double-stranded RNA. It was soon recognized to be an adenosine deaminase converting A to I — the founding observation of ADAR-mediated editing, and the reason the enzyme was originally called a dsRNA unwindase.
• The GluA2 Q/R site and ADAR2 knockout (Seeburg lab, 1990s–2000). Peter Seeburg's group mapped the single edited adenosine that changes glutamine to arginine in the AMPA receptor pore and showed it controls calcium permeability. ADAR2-null mice die of seizures within weeks; knocking a pre-edited Gria2 allele into those mice rescues them, pinning the enzyme's essential role to this one site.
• Cephalopod recoding (Rosenthal, Eisenberg, and colleagues, 2015–2017). Transcriptome-wide surveys of squid and octopus revealed tens of thousands of recoding A-to-I sites concentrated in neural genes, some temperature-dependent, and showed that conserving ADAR's double-stranded substrates constrains genomic evolution — cephalopods trade DNA-level change for RNA-level flexibility.