Editing Out Seizures? Base Editing Shows Promise for SCN8A DEE

Commentary

Developmental and epileptic encephalopathies (DEEs) arise from the convergence of genetic mutations, seizure biology and neurodevelopmental disruption, creating complex challenges for effective treatment. In many cases, the causative variant is known, the affected ion channel or synaptic protein is well characterized, and yet treatment still relies largely on broad suppression of downstream excitability. This is especially clear for SCN8A developmental and epileptic encephalopathy (SCN8A DEE), in which gain-of-function variants in the voltage-gated sodium channel NaV1.6 produce severe seizures, developmental impairment, movement abnormalities, and increased risk of sudden unexpected death in epilepsy.^1‐3 Antiseizure medications that block sodium channels can be useful in some patients, but they do not correct the molecular defect, offer only partial symptomatic relief, and are frequently limited by tolerability issues and side effects. ⁴

In an attempt to target the recurrent R1872W variant (c.5614C > T), which accounts for a substantial minority of SCN8A DEE cases (10-15%), ⁵ the study by Reever et al asks the direct question: can the disease-causing nucleotide itself be repaired in vivo, and is that sufficient to prevent the major features of SCN8A DEE? ⁶ The R1872W mutation alters channel inactivation and increases the persistent sodium current, a pathological current strongly linked to neuronal hyperexcitability and epileptiform activity.^5‐7 From a therapeutic perspective, this variant represents an attractive case study: it is recurrent, mechanistically interpretable, and caused by a single base substitution that is, in principle, editable.

The authors used an adenine base editor (ABE) to target the complementary DNA strand and revert the mutant adenine base back to the wild-type guanine, achieving a T > C conversion on the leading strand. ⁶ Conventional CRISPR-Cas9 editing relies on double-strand DNA breaks, with unwanted repair outcomes that can include insertions, deletions and larger rearrangements. Base editing offers a more conservative approach: a catalytically impaired Cas9 is fused to a deaminase and guided to a defined locus, enabling single-base conversion without cutting both DNA strands.^8,9 This distinction makes base editing a more attractive therapeutic strategy.

The authors performed extensive in vitro screening to identify a guide and editor combination with satisfying on-target activity and minimal bystander editing, then packaged the selected SCN8A-ABE into a dual PhP.eB-adeno-associated viruses (AAV-PhP.eB) system for delivery. Treated mice showed improved survival, reduced seizure frequency and severity and, in many cases, absence of seizures. Electrophysiological recordings demonstrated reduced neuronal hyperexcitability and suppression of the aberrant persistent sodium current. Behavioral abnormalities, including diminished mobility and anxiety-like phenotypes, were also improved but not fully rescued. Perhaps the most interesting result is not that editing worked, but that incomplete editing worked. SCN8A-ABE treatment produced an approximately 32% conversion of the mutant transcript into wild-type transcript, and this level of correction was sufficient to achieve a significant therapeutic effect in vivo. ⁶ This result has two implications: first, complete correction of every affected neuron may not be necessary, at least for this mutation and this model; second, SCN8A DEE may have a “therapeutic threshold” whereby reducing NaV1.6 gain-of-function below a critical point is enough to allow circuits to regain normal functionality. This is encouraging from a translational standpoint, because achieving uniform editing throughout the human brain is unrealistic with current delivery systems.

However, the delivery strategy used in this study also highlights a major translational challenge. The SCN8A-ABE was delivered using AAV-PhP.eB, a capsid with exceptional central nervous system tropism in mice but one that does not efficiently translate across species and is not currently suitable for clinical use. The therapeutic rationale for SCN8A-DEE would likely require widespread brain transduction through systemic administration, yet no blood–brain barrier-crossing viral vector has so far demonstrated equivalent efficacy in human patients. Although substantial efforts are underway to develop next-generation neurotropic vectors, delivery remains one of the principal barriers to translating base-editing approaches from mouse models to the clinic.

From a therapeutic perspective, the timing of treatment could represent a potential limitation of the study. Reever et al administered the SCN8A-ABE at postnatal day 2, which is before the onset of seizures in the mouse model. This is a powerful way to test whether editing can prevent disease emergence, but it is not the usual clinical scenario. Most children will be treated only after seizures begin, after genetic diagnosis, and after some degree of developmental disruption has already occurred. If altered sodium channel function and epileptic activity have already reshaped cortical circuits, correction at a later timepoint may not result in the same level of rescue as shown in this study. Furthermore, treatment after the onset of epileptic symptoms may reduce seizures, but it is unlikely to address developmental or behavioral phenotypes. The incomplete rescue of behavioral phenotypes observed even after very early treatment may provide an additional clue. Voltage-gated sodium channels have important roles during prenatal brain development, influencing neuronal differentiation, migration and circuit assembly. ¹⁰ Some aspects of SCN8A DEE may therefore emerge before seizure onset and potentially before birth, raising the possibility that not all disease manifestations are reversible through postnatal correction of the causative variant. This has important implications for future clinical trials: while seizure control may remain an achievable endpoint, complete neurodevelopmental rescue may be substantially more challenging if critical pathogenic processes occur during fetal development. Such considerations inevitably raise difficult questions regarding the optimal therapeutic window for genetic intervention and how early treatment would need to be administered to fully modify disease trajectory.

As always, safety remains the biggest practical challenge in the field. The authors approach this thoughtfully by using editor modifications intended to reduce unwanted DNA editing and they report no significant off-target effects in predicted sites as assessed by transcriptome-wide RNA sequencing and whole-genome sequencing. Despite these efforts, the safety standard for permanent editing in a developing human brain will remain a major barrier to the clinic. Nonetheless, this study is an important proof of principle to show that base editing in DEEs can move treatment beyond excitability suppression toward direct repair of a pathogenic epilepsy variant. The next questions are less about whether editing can work in principle and more about therapeutic timing, delivery, durability and safety. For SCN8A-DEE, the key translational questions may be simple to state but difficult to answer: how much correction is required, in which neurons, how can it be delivered across the human brain, and at what developmental stage does intervention need to occur?