Base editing

Base editing removes the step that the genome editing page identifies as the source of most trouble: it does not cut both strands, so the outcome is not handed to a repair pathway. Instead it performs chemistry directly on a base, in place, using an enzyme that has nothing to do with nucleases.

Why an R-loop makes the chemistry possible

The deaminases used are single-strand specific — they act on unpaired DNA and cannot reach a base inside an intact duplex. Cas9 supplies exactly that substrate as a side effect of how it binds. When the guide RNA pairs with the target strand, the other strand is displaced as unpaired single-stranded DNA, forming a structure called the R-loop. A deaminase tethered to a catalytically impaired Cas9 is therefore held against a short stretch of exposed, chemically accessible DNA at a programmable position.

A cytosine deaminase converts cytosine to uracil, which DNA polymerase reads as thymine, giving a C·G to T·A change. Cytosine base editors also carry a uracil glycosylase inhibitor, because the cell’s base-excision repair otherwise removes the uracil and reverses the edit before replication. Adenine base editors use a laboratory-evolved deoxyadenosine deaminase — no natural enzyme acting on DNA adenine was available — which yields inosine, read as guanine, giving A·T to G·C. Nicking the unedited strand biases repair toward using the edited strand as the template, which is what raises efficiency to the levels the field reports.

The window is the specification and the defect

The tether length and Cas9 geometry place the accessible single-stranded region at a roughly five-nucleotide window within the protospacer. Anything of the target base inside that window is a substrate. If the pathogenic cytosine has another cytosine two positions away, both are edited — bystander editing — and the guide cannot separate them. Narrowed-window variants and alternative PAM-recognising Cas proteins that reposition the window are the response, and the design problem is often finding a guide that puts one base in the window and no other.

The limits worth stating plainly

Only transitions are directly available: C to T and A to G, plus their complements. The transversions are not reachable by deamination chemistry and require other approaches — glycosylase-based editors for some, and prime editing for the general case.

Two off-target classes are distinct from a nuclease’s. Guide-independent deamination happens wherever transient single-stranded DNA occurs — at replication forks, in transcription bubbles — because the deaminase is catalytically active whether or not Cas9 has found its site, and it is not detected by assays that only examine sites resembling the guide. The cytosine deaminases in early editors also act on RNA, producing widespread transcriptome-wide edits; engineered variants with reduced RNA activity address this, and reporting it is now expected.

Delivery, immunogenicity and the on-target structural damage of double-strand breaks are covered elsewhere; base editing’s own risk profile is chemical.