CRISPR Base Editing: Rewriting DNA One Letter at a Time

The problem with double-strand breaks

Standard CRISPR-Cas9 works by creating a double-strand break (DSB) at a target site. The cell must then repair the break — and it does so through one of two pathways:

• Non-homologous end joining (NHEJ) — fast but error-prone; introduces random insertions and deletions (indels) that disrupt the gene. Useful for knocking out genes, but unreliable for correcting a specific mutation.
• Homology-directed repair (HDR) — accurate if a repair template is provided, but only active in dividing cells during S/G2 phase, and inefficient even then.

For correcting a point mutation in a post-mitotic cell (a neuron, a cardiomyocyte) or in a stem cell that should remain undivided, both pathways are inadequate. Base editing was developed to sidestep the break entirely.

For the molecular details of DNA structure and how the double helix is replicated, see our DNA, genes and heredity guide.

How cytosine and adenine base editors work

Both types of base editor share the same architecture: a guide RNA targets the editor to a specific site, and a catalytically impaired Cas9 holds the DNA strands apart without cutting both of them. This creates a short single-stranded DNA bubble — the "editing window" — where a fused deaminase enzyme can act.

Together, CBEs and ABEs cover four of the twelve possible point mutation types. Prime editing (2019) extends this to all twelve plus small insertions and deletions, using a reverse transcriptase to write new sequence into the target site from a pegRNA template.

Clinical applications in 2026

Base editing's clinical translation has been rapid:

• Sickle cell disease and beta-thalassaemia — multiple approaches target the HBB gene point mutation or reactivate foetal haemoglobin (HbF) by disrupting BCL11A enhancers. Base-edited haematopoietic stem cells are delivered after conditioning. Early trials report near-complete elimination of vaso-occlusive crises.
• Hutchinson-Gilford progeria — this rapid-ageing syndrome is caused by a single C→T mutation in LMNA generating the toxic protein progerin. ABE-correcting the mutation in mice (2021, Nature) extended lifespan by ~25%. Human trials are in preparation.
• Familial hypercholesterolaemia — silencing PCSK9 via base editing in the liver achieves LDL reductions >60% in non-human primates with a single dose. Duration of effect appears permanent in treated animals.
• T-cell leukaemia — a multiplex base editing approach knocked out four genes simultaneously in donor T-cells to create off-the-shelf CAR-T cells, successfully treating T-ALL in a paediatric patient at Great Ormond Street Hospital (2022, reported in Nature Medicine).

See our stem cell therapy article for how base-edited cells are manufactured and transplanted, and our mRNA vaccines article for delivery mechanisms shared with base editing approaches.

Off-target risk and safety

The absence of a double-strand break significantly reduces the risk of large chromosomal rearrangements — a major safety advantage over standard CRISPR. However, base editors have their own off-target concerns:

• DNA off-target editing — deaminases can edit cytosines or adenines at non-target sites where the guide RNA partially matches, though fourth-generation editors have reduced this substantially through protein engineering.
• RNA off-target editing — some base editors (particularly early CBEs using APOBEC variants) also deaminate cytosines in cellular RNA at a high rate. Evolved deaminase variants with reduced RNA affinity have addressed this in newer editors.

The mutation-correction precision of base editing — operating at the level of individual nucleotides — represents a qualitative advance over earlier gene therapy methods that used viral vectors to insert full gene copies into random genomic positions.