A Single Letter in the Genetic Code Was Causing Seizures. Scientists Just Corrected It

Inside every neuron in the human brain, sodium channels open and close in milliseconds, letting ions flood through in precisely timed pulses that generate thought, movement, sensation.

The Nav1.6 channel, encoded by a gene called SCN8A, is especially important to this process: concentrated at the neuron’s firing trigger, it helps set the threshold between silence and action. In some children, one nucleotide in SCN8A has mutated (one letter changed among roughly three billion), and the channel can no longer close properly. Sodium keeps flowing. Neurons keep firing. The result is uncontrollable seizures, developmental delay, and, not infrequently, sudden death.

Now researchers at the University of Virginia have done something that would have seemed implausible not long ago: they corrected that single letter, in living brain cells, in mice, and the seizures largely stopped.

The condition is called SCN8A developmental and epileptic encephalopathy, or SCN8A DEE. It affects roughly one in 56,000 births, though clinicians reckon the true number is higher because the diagnosis is still widely missed. The specific mutation the UVA team targeted, known as R1872W, accounts for around 10 to 15 percent of all SCN8A DEE cases. What it does is remove an arginine residue from the channel’s inactivation gate, causing it to linger open longer than it should. A pathological current called INaP builds up. Cortical neurons become hyperexcitable in a way that antiseizure medications often cannot contain, and the risk of sudden unexpected death is high.

Standard treatments suppress the downstream chaos; they do not address its origin. That gap is precisely what drew Manoj Patel and his team toward gene editing.

Base editing is a refinement of CRISPR technology, though calling it merely that undersells how different it is in practice. Traditional CRISPR cuts both strands of the DNA helix and relies on the cell’s own repair machinery to patch the break, a process that can introduce errors. Base editing skips the cutting entirely. A chemically modified Cas9 protein, unable to cleave, is fused to a deaminase enzyme and guided by a short RNA sequence to a specific nucleotide. At that address, the enzyme converts adenine to inosine, which the cell reads as guanine. One letter swapped. No breaks, no collateral damage.

The UVA team spent considerable effort finding a base editor that could reach the R1872W locus at all, because there was no standard recognition site nearby for the Cas9 to dock. They settled on a variant called ABE8e-NRCH, further modified with a V106W mutation to tighten specificity, then split the full construct across two separate viral delivery vehicles because it was too large for one. The resulting tool, SCN8A-ABE, was injected into the cerebrospinal fluid of two-day-old mice carrying the mutation.

“Historically, treatments addressed only the downstream effects of genetic mutations; today, we can correct the mutations themselves, targeting the root cause of disease,” said Patel, of UVA’s Department of Anesthesiology and Brain Institute. “Base editing opens the door to the treatment of numerous genetic diseases, not only those associated with epilepsy, and has the potential to significantly improve patients’ quality of life.”

The results, published in the Journal of Clinical Investigation, were striking. Spontaneous seizures were completely eliminated in seven of eleven treated mice. In three more, seizure frequency dropped sharply and severity was reduced to mild, localised twitching rather than the full convulsive events that typically killed untreated animals within weeks. Survival curves tell the broader story: untreated mice with the R1872W variant almost universally died before postnatal day 65; most treated mice were alive past day 150, at which point the researchers began euthanising them for sequencing, so the true survival benefit was probably larger still. The team found a threshold effect: mice in which the editor corrected at least 17 percent of mutant DNA copies showed complete seizure cessation; those below about 13 percent correction still died prematurely. At the RNA transcript level, editing efficiency reached approximately 32 percent, reflecting the fact that the viral capsid preferentially targets neurons, where SCN8A is expressed most strongly.

Off-target effects were scrutinised across 290 potential sites by whole-genome sequencing. No biologically meaningful edits were found anywhere other than the intended target.

Beyond seizure counts, treated mice moved more freely and spent less time hugging the walls of an open-field arena, a standard measure of anxiety-like behaviour. Electrophysiology confirmed the mechanism directly: the aberrant persistent sodium current was suppressed in treated animals’ cortical neurons, bringing their firing properties back toward those of healthy controls. “This shows that the devastating impact of the mutation is not permanent — and can be reversed,” said Caeley Reever, who led the project as first author. “We were able to effectively ‘cure’ mice carrying this specific gene mutation — a mutation that is known to cause epilepsy in some children.”

Significant questions remain before anything like this could reach a child with SCN8A DEE. The mice were treated at two days old, before seizure onset; in human patients, diagnosis typically comes after seizures have already begun, at four to six months. Whether a therapeutic window exists for post-onset treatment is still unknown. AAV delivery to the human brain carries its own regulatory and safety hurdles. And the R1872W variant, while relatively common within SCN8A DEE, is just one of hundreds of pathogenic mutations in the gene; each, in principle, would need its own tailored editor. “Our goal is to assess this gene therapy in children with this specific SCN8A variant,” Patel said.

The underlying logic, though, extends well beyond this one condition. SCN8A DEE is caused by a point mutation; so are thousands of other inherited diseases. Each is, at bottom, a letter in the wrong place. The UVA work is a proof of concept that a sufficiently precise molecular pencil can erase and rewrite that letter in living brain tissue without disturbing the surrounding text, something that was, until recently, largely aspiration rather than experiment. What remains now is narrowing the distance between a mouse at postnatal day two and a child at six months, between proof of principle and proof of treatment.

Source: Journal of Clinical Investigation, doi:10.1172/JCI196402

Could this approach work for other types of epilepsy, not just SCN8A?

The researchers designed their tool for one specific mutation, but the technology is not inherently limited to SCN8A. Dozens of genetic epilepsy syndromes, including Dravet syndrome, are caused by single-nucleotide variants in related channel genes, and base editors can in principle be retargeted to any of them. The main constraint is that each variant requires its own custom guide RNA and editing construct, so development would need to be run separately for each target, which is not trivial at scale.

Why didn’t the treatment work equally well in all of the treated mice?

The team identified a clear threshold: animals that achieved at least 17 percent correction of mutant DNA showed complete seizure elimination, while those below roughly 13 percent still died from seizure-related causes. The variation most likely reflects differences in how efficiently the viral vectors reached enough neurons in each individual animal. Improving delivery consistency, through higher doses or next-generation capsids, is one of the stated priorities for follow-up work.

How is base editing different from antisense oligonucleotide therapy?

Antisense oligonucleotides (ASOs) bind to RNA and alter how a gene’s message is processed, reducing the production of a pathological protein; they do not change the underlying DNA and need to be re-administered repeatedly as the molecules degrade. Base editing makes a permanent change to the DNA itself, meaning a single treatment could, in principle, correct the mutation for the lifetime of the edited cell. The UVA team’s 32 percent correction of mutant transcripts exceeded the roughly 25 percent reduction that ASO studies have shown is sufficient for phenotypic rescue in the same disease model.

What is the risk of off-target edits damaging other genes?

It was a central concern in the study, and the team went to considerable lengths to address it. They used a modified base editor with a V106W mutation specifically to reduce off-target activity, then checked for unintended edits across 290 candidate sites using whole-genome sequencing and whole-transcriptome RNA analysis. No biologically meaningful edits were found outside the intended target, though the researchers acknowledge that larger-scale clinical studies would need to replicate this finding before the approach could be considered safe in human patients.