Diagnosis-led treatment

Gene editing — CRISPR, base and prime editing

CRISPR-Cas9, base editing, prime editing, RNA editing, epigenome editing (CRISPRa)

The only approach that corrects the fault in the DNA itself rather than working around it. In children's neurology it has not yet treated a patient outside single-case protocols — but 2025 and 2026 changed what is realistic, and the first neurological programmes are now being built.

Where it stands, condition by condition

The same therapy can be well supported for one problem and completely untested for another. This is the single most common place families are misled.

What it is

Four generations of tool, increasingly precise, often all called «CRISPR» in news coverage:

Nuclease editing (classic CRISPR-Cas9). A guide molecule finds the target sequence and an enzyme cuts both strands of DNA. The cell repairs the break imperfectly, which reliably switches a gene off but is a blunt way to correct one. Good for removing something; poor for fixing a single letter.

Base editing. No double-strand cut. A chemical converter is carried to the exact spot and rewrites one DNA letter into another — the mechanism behind the first personalised treatment given to an infant. Most disease-causing variants in child neurology are single-letter changes, which is why this generation matters so much here.

Prime editing. More flexible again: it writes a short stretch of new sequence, so it can in principle fix insertions and deletions as well as single letters. This is the tool behind the mouse rescue of alternating hemiplegia.

RNA and epigenome editing. The DNA is left untouched. Either the message is edited as it is read, or the gene is switched up or down without changing its sequence. Both are reversible, which in a developing child is a serious advantage rather than a compromise.

The bottleneck is not the editor — it is delivery. Getting the machinery into brain cells is the whole problem. Lipid nanoparticles work well for the liver, which is why urea cycle disorders are first. For the brain, the current hope is an engineered AAV that crosses from the bloodstream using the transferrin receptor, avoiding neurosurgery.

How it is meant to work

Unlike an ASO, an edit is made once and stays. Unlike gene replacement, nothing foreign remains in the cell — the child's own corrected gene does the work, under its own natural control.

That last point is more important than it sounds. A delivered gene is expressed at whatever level the vector dictates. A corrected gene is read exactly as the body intends, in the right cells, at the right times. For a protein like the sodium channel in Dravet, where both too little and too much cause problems, that difference may be decisive.

Neurons do not divide, which cuts both ways: an edit made in a neuron should last a lifetime, and an unwanted edit in a neuron is equally permanent.

What has actually been tested

• Outside neurology the platform is proven. A CRISPR therapy for sickle cell disease and beta-thalassaemia has been licensed since late 2023 and is now approved down to age two.
• Inside paediatrics, the landmark is a single infant. In February 2025 a baby with a severe urea cycle disorder received a base editor designed for his own variant, built in about six months. He received two to three infusions, had no serious side effects, needed less medication and went home after a long admission; a year later he was reported to be developing on track. The team that treated him have been notably careful about the framing — their own public message has been that it was not a cure.
• The honest state in child neurology is preclinical. Base editing has corrected a Dravet variant in live mice and reduced seizures; prime editing has rescued alternating hemiplegia in mice. Mouse brain results have a long history of not transferring to children, and no child with either condition has been treated.
• What changed in 2026 is the surrounding machinery rather than the science: a regulator publishing draft guidance on how individualised editing treatments might be assessed, and a funded consortium building a platform aimed at paediatric epilepsies with a reusable regulatory route attached. Those two things are what turn a one-off heroic case into something repeatable.

What we still do not know

• Whether the delivery problem for the brain is solved. Everything in child neurology depends on this and it is not yet answered in humans.
• What proportion of cells must be corrected for a child to benefit. In the liver, roughly one in ten appears to be enough; in the brain, for most conditions, nobody knows.
• Off-target edits and their consequences over decades. The tools are far more precise than they were, but a permanent change in a child's neurons has a very long time horizon in which to reveal something.
• Whether editing after symptoms are established helps at all, or whether this becomes — like so much on this site — a treatment that must arrive before the damage.
• How a treatment built for one child is funded, overseen and repeated for the next child. This is now the rate-limiting step, not the biology.

Risks and unwanted effects

• Permanence. An edit cannot be undone. In an adult with a fatal illness that calculus is one thing; in a child who may live seventy more years it is another, and it is the central ethical argument in this field.
• Off-target editing elsewhere in the genome, and unintended larger rearrangements at the target site itself.
• Immune responses to the editing proteins, which are bacterial in origin, and to whatever vector delivers them.
• The delivery vehicle carries its own risks — AAV liver injury, or the effects of a lipid nanoparticle infusion.
• A risk specific to this moment: expectation. One widely reported infant has created an impression that bespoke editing is available. It is not. Outside a small number of research protocols there is no route to it, and a clinic offering «CRISPR treatment» for a child's neurological condition today is not offering what the headlines described.
• Germline editing — changes that would pass to future generations — is a different matter entirely and is prohibited. Everything described on this page edits the cells of one person only.

Questions to ask before you agree

1. Does my child have a confirmed single variant that an editor could in principle correct, and has anyone actually assessed that?
2. Is there a registered trial or platform programme for this gene — and if not, is there a registry or natural-history study we should join now?
3. For anything offered outside a trial: who oversees it, who manufactured the editor, and what is the long-term follow-up plan?
4. What is being edited — DNA, RNA, or the switch that controls the gene — and is it reversible?
5. What would success look like and when would we know, given that an edit cannot be withdrawn if the answer is no?

More in this section