The Silent Interval: Hidden Brain Remodeling Precedes the Return of Seizures

One of the strangest features of severe childhood epilepsy is that it does not always move in one direction. In West syndrome, the infantile spasms that convulse a newborn’s first months often fade away, and for a while the child appears better. Then, sometimes years later, a different and often more intractable epilepsy emerges, most commonly Lennox-Gastaut syndrome. Clinicians have long known that this deceptive calm exists. What they have not known is what the brain is doing during it. A new review in Frontiers in Synaptic Neuroscience argues that this seizure-free interval, far from being a period of rest, is likely one of the most biologically active phases of the entire disease, and that treating it as a blank space may be costing patients their best therapeutic opportunity.

The review, authored by Ruiying Ma, U Suk Kim, Yousun Chung and Kihoon Han of Korea University College of Medicine and collaborators, draws together recent time-resolved studies of a mouse model carrying the CYFIP2 Arg87Cys mutation, a recurrent genetic variant found in children with West syndrome. The animals recapitulate the human trajectory with unusual fidelity: seizure-like events in the neonatal period, a prolonged seizure-free juvenile interval, spontaneous recurrent seizures beginning around postnatal week 14, and premature death. By applying transcriptomic, ultrastructural, proteomic and lipidomic analyses across four postnatal stages, the researchers have produced the first multi-omic portrait of seizure evolution in a West syndrome model, and the picture it paints is one of continuous, quietly destructive renovation.

The clinical backdrop makes the case compelling. More than 900 epilepsy genes have now been catalogued, the majority linked to developmental and epileptic encephalopathies, the most severe class of epilepsies. In West syndrome, up to 60 percent of children who survive infantile spasms go on to develop other seizure types, and roughly one quarter to one third eventually meet criteria for Lennox-Gastaut syndrome. Critically, in a recent multicenter cohort, two thirds of children who evolved to that syndrome passed through an interval of complete seizure freedom first. Similar gaps appear elsewhere in the genetic epilepsy landscape: 38 percent of adults with KCNQ2 encephalopathy have experienced seizure-free periods of at least two years followed by recurrence, and the so-called honeymoon period is a recognized stage of CDKL5 deficiency disorder.

The therapeutic implication, as the review notes, has been stated explicitly in the Lennox-Gastaut literature: if the neural network is being reorganized during the gap, then the gap is precisely when intervention is most likely to alter the disease trajectory. The obstacle is that the child looks improved, brain tissue is inaccessible, and no tool exists to determine whether a patient is progressing toward a worse syndrome or has genuinely stabilized. Even the electroencephalogram offers only partial reassurance, because slow spike-and-wave activity, the electrical signature of Lennox-Gastaut syndrome, can appear before or during the clinically quiet interval, suggesting the brain never actually stopped misfiring.

This is where the mouse model becomes invaluable, because only in animals can the hidden interval be dissected cell by cell. One of the most unexpected structural findings concerns excitatory synapses in the hippocampus, the brain’s memory hub and a common seat of seizure genesis. At the time spontaneous seizures emerge and again two weeks later, the number of excitatory synapses in region CA1 falls, while inhibitory synapse numbers are preserved. On its face, fewer excitatory synapses should mean a calmer network, which sits awkwardly with the onset of recurrent seizures. The electron microscope resolved the paradox, at least as a hypothesis: dendritic spines, the tiny protrusions that normally receive excitatory inputs, were markedly reduced, and the remaining excitatory synapses had relocated onto the dendritic shaft itself, with multiple presynaptic boutons frequently converging on individual shafts.

That relocation matters because spine architecture is not merely decorative. The narrow neck of a spine electrically and biochemically isolates each synapse from its neighbors, compartmentalizing signals. Shaft synapses bypass that barrier entirely. The authors propose, carefully and as a hypothesis awaiting functional testing, that fewer synapses moved to the shaft could allow more effective integration of convergent excitatory inputs, producing stronger synchronization of neuronal activity even as the synapse count drops. The structural remodeling was accompanied by asymmetric molecular shifts, with postsynaptic proteins such as Shank3, Homer1 and GluA2 decreased while presynaptic markers like Synaptotagmin-1 and Synapsin 1 increased, and by altered splicing of the neurexin genes Nrxn2 and Nrxn3, whose isoform choices are known to instruct synapse specification. Axonal hypermyelination completed the picture, echoing other epilepsy models in which maladaptive myelin is thought to speed seizure propagation through the brain.

Equally striking was the behavior of the glia, the brain’s support cells, which engaged in a discernible sequence rather than all at once. The number of Olig2-positive oligodendrocyte-lineage cells rose first, at postnatal week 7, during the seizure-free interval, and persisted thereafter. Microglia, the brain’s immune cells, followed at week 14, showing simplified, ameboid morphology, elevated CD68 phagocytic signal and increased proinflammatory cytokine transcripts. Reactive astrocytes appeared at weeks 14 and 28. Gene set enrichment analysis of the bulk transcriptome mirrored the same ordering. The sequence is broadly consistent with work in acquired epilepsy, where microglial activation has been shown to be required for subsequent reactive astrogliosis and epileptogenesis, but the mouse data place all three glial lineages on a single temporal framework within one disease model for the first time. The authors are careful to note that the widely spaced sampling points mean the ordering is suggested rather than established, and because CYFIP2 is expressed predominantly in neurons, the glial changes are almost certainly secondary responses to altered neuronal activity rather than direct effects of the variant.

The latest and arguably most ominous changes were metabolic. By week 28, electron microscopy revealed enlarged lipid droplets within astrocyte lysosomes, together with crystalline material never observed in wild-type animals, and astrocytic mitochondria that were smaller and lacked distinct cristae. Multiple independent datasets converged: single-nucleus transcriptomics identified an astrocyte subtype enriched for oxidative phosphorylation and fatty acid metabolism genes, the proteins APOD and PTGDS were prominently upregulated, and lipidomic profiling showed increased triacylglycerol, diacylglycerol and cholesteryl ester, the major lipid classes stored in lipid droplets. The findings fit an emerging framework in which hyperactive neurons offload excess fatty acids to astrocytes for oxidation, while lipid-laden reactive astrocytes in temporal lobe epilepsy show impaired glutamate uptake and drive disease progression. Whether the same feed-forward loop operates here remains unproven, since the lipid changes appeared only after weeks of seizures and could simply mark cumulative seizure burden rather than cause it.

Indeed, the review is as notable for what it admits it does not know as for what it reports. The deepest analyses, including electron microscopy and single-nucleus transcriptomics, were performed only at week 28, after seizures were established, while the critical transition between weeks 7 and 14 went unsampled. Higher-resolution assays were applied at fewer time points, a tradeoff the authors describe candidly: analytical depth was purchased at the expense of temporal coverage, leaving the clinically silent interval, the stage that matters most, the least well characterized. The only molecular data spanning the pre-onset period came from the bulk transcriptome, in which differentially expressed genes numbered just 20 and 41 at weeks 1 and 7 before exploding to 3,755 at week 14, hinting that molecular change simmered near the threshold of detection until seizures emerged. Electrophysiology was similarly sparse, with video-EEG recordings at a single time point in only three animals, meaning the interval is seizure-free by behavioral criteria alone and silent interictal activity cannot be excluded. A useful internal control, Cyfip2 haploinsufficient mice showing only 19 differentially expressed genes and no seizures, argues the knock-in phenotype reflects a toxic gain of function rather than simple gene dosage.

The downstream implications reach toward the clinic. The temporal data suggest that different interventions belong to different stages: allele-selective suppression of the mutant transcript, for instance with antisense oligonucleotides, would likely work best early, before glial engagement; anti-inflammatory or glia-targeted approaches would fit an intermediate stage; and metabolic or lipid-restoring strategies, such as liver-X receptor agonism, might apply later. None of this helps patients, however, unless the stage can be identified prospectively, and no validated staging biomarkers exist for these disorders. Candidate markers are accumulating, from the EEG detection of slow spike-and-wave activity before relapse, to plasma GFAP and neurofilament light chain, to the PTGDS and APOD proteins that emerged from the mouse work. The review’s central message is that the seizure-free interval should be reframed not as remission but as a remodeling phase, one that dense, onset-aligned sampling of the model, and parallel biomarker development in patients, could convert from a clinically silent enigma into a well-defined window for stage-specific therapy.

Subject of Research: Time-resolved multi-omic phenotyping of seizure evolution in the CYFIP2 Arg87Cys mouse model of West syndrome

Article Title: Time-resolved phenotyping of seizure evolution in developmental and epileptic encephalopathy: insights and gaps from the CYFIP2 Arg87Cys mouse model

Article References: Ma, R., Kim, U. S., Chung, Y., & Han, K. (2026). Time-resolved phenotyping of seizure evolution in developmental and epileptic encephalopathy: insights and gaps from the CYFIP2 Arg87Cys mouse model. Frontiers in Synaptic Neuroscience, 18, Article 1963161. https://doi.org/10.3389/fnsyn.2026.1963161

Image Credits: AI Generated

DOI: 10.3389/fnsyn.2026.1963161

Keywords: West syndrome, CYFIP2, developmental and epileptic encephalopathy, seizure-free interval, Lennox-Gastaut syndrome, synaptic remodeling, gliosis, astrocytes, lipid droplets, microglia, oligodendrocytes, biomarkers