Developmental and Epileptic Encephalopathy with Spike-Wave Activation in Sleep

1. Disease Information

What it is

DEE-SWAS is one of those syndromes that's really a state the brain slides into rather than a single disease. A kid — usually somewhere between 2 and 12, peaking around 4–5 — starts having seizures, and then, a year or two later, starts losing ground: language, attention, behavior, sometimes motor skills. The tell is on the overnight EEG: during non-REM sleep, spike-and-wave discharges bloom until they occupy most of the tracing. Wakefulness looks comparatively tame. Sleep is where the damage happens.

The ILAE's 2022 nosology (Specchio et al., PMID:35503717) folded a whole graveyard of older names into this single umbrella. The formal definition: "a spectrum of conditions with varied degree of cognitive, language, behavioral, and motor regression associated with marked spike-wave activation in sleep" [reported — epilepsydiagnosis.org syndrome page].

The split between the two halves is purely developmental history: - EE-SWAS — development was normal before the regression. The epileptic activity is doing all the harm. - DEE-SWAS — there was pre-existing developmental impairment, and then regression on top. Two encephalopathies stacked.

Think of it like a fever in a healthy adult versus a fever in someone already immunosuppressed — same insult, very different baseline, very different ending.

Landau-Kleffner syndrome (LKS) was retained as a clinically distinct EE-SWAS subtype: the variant where the regression is essentially all language — acquired auditory verbal agnosia in a previously normal child, ages ~3–9.

Identifiers

Synonyms (all from the MONDO synonym block, verified)

CSWS · CSWSS syndrome · DEE-SWAS · EE-SWAS · EESWAS · ESES · electrical status epilepticus of sleep · electrographic status epilepticus in sleep · continuous spike-wave in sleep · continuous spikes and waves during sleep · continuous spikes and waves during slow-wave sleep · continuous slow spike and wave of sleep · epileptic encephalopathy with continuous spike-and-wave during slow sleep (EE-CSWS) · epileptic encephalopathy with spike-and-wave activation in sleep · ESES with language regression · epileptic aphasia · Landau-Kleffner syndrome / LKS / LK syndrome. Also retired but still in the literature: atypical benign partial epilepsy (ABPE), pseudo-Lennox syndrome, Penelope syndrome.

Data provenance

Everything below is disease-level aggregated — case series, tertiary-center cohorts, systematic reviews, one small RCT. There is no EHR-derived phenotype library for this syndrome, and the OMOP/ICD coding is so coarse (8A62.Y "other specified") that EHR case-finding would be near-useless without EEG-report NLP. Worth flagging as a KNOWLEDGE_GAP if you're curating definitions.

2. Etiology

The big picture: heterogeneous as hell, and half of it stays unsolved

The best-powered modern etiology study is Viswanathan et al., Ann Neurol 2024 (PMID:39096015) — 91-patient Core cohort, all meeting ILAE D/EE-SWAS criteria.

"We identified the etiology in 42/91 (46%) patients in our Core cohort, including 29/44 (66%) with DEE-SWAS and 13/47 (28%) with EE-SWAS. A genetic etiology was identified in 31/91 (34%)." [verbatim-verified from references_cache/PMID_39096015.md]

"D/EE-SWAS genes were highly co-expressed in brain, highlighting the importance of channelopathies and transcriptional regulators. Structural etiologies were found in 12/91 (13%) individuals." [verbatim-verified]

That 66% vs 28% gap is the single most curation-relevant number in the whole literature: DEE-SWAS (pre-existing impairment) is more than twice as likely to have a findable cause as EE-SWAS. Makes intuitive sense — an already-abnormal brain usually got that way for a reason you can find.

Breakdown from the same cohort: - Genetic: 31/91 (34%) — 23 single-gene variants, 6 CNVs, 1 chromosomal abnormality [reported] - Structural: 12/91 (13%) — polymicrogyria ×5, thalamic lesions ×5, post-hemorrhagic hydrocephalus ×2 [reported] - Unsolved: 49/91 (54%)

Older tertiary-center series put the structural fraction much higher (~45–59%, with perinatal vascular lesions 21–78% and cortical malformations ~25%) [reported, PMC3929187] — the discrepancy is almost certainly ascertainment: an epilepsy-genetics research program enriches for undiagnosed kids, a general pediatric neurology clinic enriches for kids with obvious perinatal brain injury.

Causal factor classes

(a) Structural — early thalamic injury is the standout.

This one is mechanistically load-bearing, not just a bucket. From Sánchez Fernández / Leal et al., Epilepsy Behav 2018 (PMID:29133062):

"Early neonatal thalamic lesions account for about 14% of continuous spike-wave of sleep (CSWS) syndrome, representing the most common etiology in this epileptic encephalopathy in children, and promise useful insights into the pathophysiology of the disease." [verbatim-verified from cache]

Other structural causes: polymicrogyria (especially unilateral perisylvian), periventricular leukomalacia and other perinatal vascular insults, post-hemorrhagic hydrocephalus, porencephaly, hemimegalencephaly, cortical dysplasia, and — importantly — shunted hydrocephalus.

(b) Genetic. See §4 for the gene-by-gene detail. Headline: GRIN2A is the single most frequent gene; the two functional classes that dominate are ion channels/receptors and transcriptional regulators.

(c) Iatrogenic / drug-provoked — an under-appreciated and modifiable cause.

Sodium-channel-blocking and GABAergic ASMs can precipitate SWAS in a child with self-limited focal epilepsy who would otherwise have coasted to remission. Carbamazepine, oxcarbazepine, phenytoin and phenobarbital are all implicated (e.g. PMID:26415787, oxcarbazepine-induced ESES in idiopathic childhood focal epilepsy). Practically: "Carbamazepine is relatively contraindicated in ESES and should be discontinued" [reported, StatPearls NBK553167]. This deserves its own pathophysiology node — it's one of the few genuinely preventable routes into the syndrome.

(d) Unknown / presumed developmental. Over half. Age-dependency (onset window 2–12, remission around puberty) strongly implies the causal factor isn't the lesion or variant alone but its interaction with a developmental window — the same lesion in an adult brain doesn't do this.

Risk factors

Genetic risk: - Pathogenic/likely pathogenic variants in the genes in §4 (causal, not merely susceptibility, in solved cases) - Being a male carrier of an X-linked CNKSR2 variant (hemizygous males affected; most carrier mothers neurologically unremarkable [reported]) - Underlying self-limited epilepsy with centrotemporal spikes (SeLECTS) — sits on the same epilepsy-aphasia spectrum as DEE-SWAS; GRIN2A detection rate climbs from ~4.9% in BECTS/SeLECTS to ~17.6% in CSWS [reported, Lemke et al. Nat Genet 2013] - Incomplete penetrance and intrafamilial variability documented even for the same GRIN2A variant [reported, OMIM #245570]

Environmental / acquired risk: - Neonatal thalamic hemorrhage or infarction (often associated with neonatal sinovenous thrombosis) - Perinatal hypoxic-ischemic injury, prematurity, periventricular leukomalacia - Intraventricular hemorrhage → post-hemorrhagic hydrocephalus → shunt - Exposure to carbamazepine/oxcarbazepine/phenytoin/phenobarbital in a child with focal childhood epilepsy - Age 2–12 is itself the dominant risk factor — this is a developmental-window disease - Sex: mild male excess (~60:40) [reported, PMC3929187]; the Ann Neurol cohort was 53% male [reported]. Not a strong signal except in X-linked CNKSR2 families.

Protective factors: Honestly, nothing established. No protective allele, no dietary or lifestyle factor with evidence. The nearest thing to a protective factor is early recognition and early spike-suppressing treatment (see §12) plus avoiding the aggravating ASMs. Worth curating explicitly as absent rather than leaving the section blank.

Gene–environment interaction: The clearest one is pharmacogenetic-ish rather than classical GxE: a child with a GRIN2A variant and a SeLECTS phenotype who gets started on carbamazepine may be tipped into full SWAS. Also: a genetic background (e.g. channelopathy) plus a structural thalamic hit appears additive in some series. No formal GxE study exists — flag as a gap.

3. Phenotypes

The two-act structure

Act I (age ~2–7): seizures appear, often nocturnal, often unimpressive. Up to 80% of children present with seizures as the first symptom [reported], and about 80% have only one seizure type at onset [reported]. Roughly 20% present the other way round — cognitive/behavioral change first, seizures later or never prominent [reported].

Act II (~1–2 years later): SWAS establishes on the sleep EEG, seizure frequency often jumps (up to 70% have multiple daily seizures once ESES appears [reported]), and the regression begins. This is the encephalopathy proper.

Phenotype table with HPO suggestions

All HP IDs below verified against sqlite:obo:hp via OAK.

⚠️ Frequency-band caution (per docs/frequency-evidence-guidelines.md): most of the percentages above are single-cohort tertiary-center figures, not pooled. I would only assign a frequency: enum to Developmental regression (definitional, obligate) and the EEG features. For the rest, omit the band rather than manufacture support.

The regression is not one thing

Worth splitting into nodes if you're curating carefully. Per the ILAE description, "All cognitive domains are affected including language and communication, temporo-spatial orientation, attention and social interaction" [reported]. The domain hit tracks the anatomy of the spike focus: - Perisylvian/temporal focus → LKS phenotype (auditory verbal agnosia, aphasia) - Frontal focus → CSWS phenotype (dysexecutive/frontal syndrome, behavioral disinhibition, global cognitive drop)

From Issa NP, Pediatr Neurol 2014 (PMID:25160535):

"Several pediatric seizure disorders have common electrophysiological features during slow-wave sleep that produce different syndromes based on which part of the developing brain is involved." [verbatim-verified from cache]

That sentence is a good anchor for a "topography determines phenotype" pathophysiology node.

Quality of life

No EQ-5D/PROMIS/SF-36 data specific to D/EE-SWAS that I could find — a genuine gap. Qualitatively: the burden is dominated by (1) permanent language/cognitive deficit rather than seizures, since seizures usually remit; (2) behavioral dysregulation and ADHD, which drive school placement and family stress; (3) in LKS, the profound communication loss — a child who could speak in sentences and now cannot understand speech at all. Caregiver burden is high across the active phase (typically 2–5+ years). Flag as KNOWLEDGE_GAP: no validated disease-specific QoL instrument.

4. Genetic / Molecular Information

Causal genes

The flagship: GRIN2A (HGNC:4585; OMIM 138253; 16p13.2; GluN2A subunit of the NMDA receptor).

• Lemke et al., Nat Genet 2013: "Heterozygous mutations in GRIN2A were detected in 27 of 359 affected individuals from independent cohorts with IFE (7.5%), with mutation detection rates ranging from 4.9% in individuals with BECTS to 17.6% in individuals with CSWS." [reported]
• In the Ann Neurol 2024 cohort, GRIN2A was the most common single gene (~23% of genetic cases) [reported]
• GRIN2A accounts for ~9–20% of epilepsy-aphasia syndromes overall [reported]

Functional consequence predicts phenotype — this is the therapeutically actionable bit. Strehlow et al., Brain 2019 (PMID:30544257):

"Null variants and mis_ATD+LBD_ of GRIN2A share the same clinical spectrum (milder phenotypes), but also result in similar electrophysiological consequences (loss-of-function) opposing those of mis_TMD+Linker_ (severe phenotypes; predominantly gain-of-function)." [reported]

"Individuals with developmental and epileptic encephalopathy due to misTMD+Linker are prone to having an underlying gain of NMDAR function and represent promising candidates for treatment with NMDAR blockers, such as memantine." [reported]

So: null / ATD+LBD missense → loss of function → milder, epilepsy-aphasia-spectrum end; TMD+Linker missense → gain of function → severe DEE end. Domain-level annotation, not just "pathogenic," is what determines whether memantine (block) or a positive allosteric modulator is the rational move.

The other established genes. From the Ann Neurol 2024 cohort (previously known): CNKSR2, SCN2A, ARID1B, CUL4B, GRIN2B, KCNH5, MECP2, SCN1A [reported].

Ten novel D/EE-SWAS genes from the same study, verbatim from the abstract:

"We identified 10 novel D/EE-SWAS genes with a range of functions: ATP1A2, CACNA1A, FOXP1, GRIN1, KCNMA1, KCNQ3, PPFIA3, PUF60, SETD1B, and ZBTB18, and 2 novel copy number variants, 17p11.2 duplication and 5q22 deletion." [verbatim-verified from cache]

From the systematic review of genetic ESES etiologies (PMID:29976148) — 16 studies, 151 cases, 11 monogenic genes: GRIN2A (34 cases), SCN2A (6), KCNA2 (5), KCNB1 (5), KCNQ2 (2), CNKSR2 (2), SLC6A1 (2), SLC9A6/NHE6 (1), ATN1/DRPLA (1), SRPX2/neuroserpin (1), OPA3 (1) [reported]. Key conclusion: "The most common underlying pathway was channelopathy" (56 cases) [reported].

From the Seizure 2023 systematic review (PMID:37352690), 172 cases: variants in GRIN2A, ZEB2, CNKSR2, and 17q21.31 deletions; conclusion that "presentations occurring before age five warrant genetic investigation" [reported].

From the Turkish cohort (PMID:38388889), 24 patients, 7 solved (29%): novel variants in SLC12A5, DLG4, SLC9A6; also SCN8A and Smith-Magenis syndrome [reported].

CNKSR2 (Xp22.12; connector enhancer of KSR-2) deserves its own node — it's the X-linked epilepsy-aphasia gene:

"The disease is characterized by intellectual disability, attention deficit-hyperactivity and abrupt lifelong language loss following a brief early-childhood epilepsy with continuous spike-waves in sleep." [reported]

~50% de novo; carrier mothers usually unaffected [reported]. Predominance of loss-of-function variants (PMC8281706).

Consolidated gene table

⚠️ HGNC IDs above are from memory and are NOT OAK-verified. Run just validate-terms before committing any of them — the repo uses lowercase hgnc:.

Copy number variants and chromosomal abnormalities

Recurrent CNVs from the systematic review (89 CNVs total, 9 recurrent) [reported]: - 15q11.2–13.1 duplication — 15 cases (also relevant to your existing 15q11q13 microduplication entry — likely a comorbidity/grouping link) - 3q29 duplication — 11 cases - Xp22.12 deletion (removing CNKSR2) — 6 cases - 16p13 deletion (removing GRIN2A) — 4 cases - 17q21.31 deletion (Koolen-de Vries) [reported, Seizure 2023] - 17p11.2 duplication (Potocki-Lupski) — novel, Ann Neurol 2024 - 5q22 deletion — novel, Ann Neurol 2024 - 17p11.2 deletion (Smith-Magenis) [reported, Turkish cohort]

Allele frequency / somatic vs germline

All reported variants are germline (constitutional). Somatic mosaicism has not been established as a mechanism in D/EE-SWAS — though it's plausible in cases with focal cortical dysplasia. Pathogenic variants are absent or vanishingly rare in gnomAD; GRIN2A is strongly constrained (missense- and LoF-intolerant). ⚠️ Specific gnomAD constraint scores not retrieved — look them up if you want to cite pLI/o/e values.

Modifier genes

None validated. The obvious candidate class — genes affecting sleep spindle generation and thalamocortical rhythm — is theorized but not demonstrated. The intrafamilial variability of the same GRIN2A variant strongly implies modifiers exist. Gap.

Epigenetics

Indirect but suggestive: SETD1B (H3K4 methyltransferase), ARID1B (BAF chromatin remodeling), and MECP2 (methyl-CpG binding) all appear as causal genes, i.e. chromatin/transcriptional regulation is one of the two major functional clusters. Per the Ann Neurol brain co-expression analysis, the D/EE-SWAS genes partition into Cluster 1 (ion channels: GRIN2A, GRIN2B, KCNH5, KCNQ3, CACNA1A, SCN1A, SCN2A) and Cluster 2 (transcriptional regulators: FOXP1, PUF60, MECP2, ARID1B, ZBTB18), both co-expressed above chance [reported]. No DNA-methylation episignature has been published for D/EE-SWAS as a syndrome (though episignatures exist for some individual causal genes, e.g. ARID1B/Coffin-Siris). Gap.

5. Environmental Information

Thin section, and that's the honest answer.

• Toxins / pollution / occupational / radiation: no established role. Nothing in CTD.
• Lifestyle: no established role. Sleep deprivation may worsen seizures generically but is not a syndrome-specific factor.
• Infectious agents: no causal pathogen. Post-encephalitic and post-meningitic acquired brain injury can be a structural substrate in individual cases, but this is generic acquired-lesion territory, not a specific infectious etiology. Notably, LKS was historically suspected to be inflammatory/autoimmune (hence steroid responsiveness), but no pathogen or autoantibody has been confirmed.
• The one real "environmental" exposure is pharmacological — the carbamazepine/oxcarbazepine/phenytoin/phenobarbital aggravation described in §2. I'd model this as an environmental/iatrogenic trigger node rather than leaving §5 empty.
• Perinatal events (hypoxia-ischemia, IVH, sinovenous thrombosis) are the most important non-genetic contributors, acting via structural injury — especially thalamic.

6. Mechanism / Pathophysiology

This is the interesting part. There are three distinct, partly complementary mechanistic models, and I'd curate them as competing/complementary mechanistic_hypotheses rather than blending them.

Model A — Thalamocortical disconnection and the "augmenting response" (CANONICAL for structural cases)

The thalamus is the metronome for non-REM sleep rhythms. Lesion it early and unilaterally, and the cortex on that side loses its normal pacing input — and, critically, gains an abnormal form of frequency-dependent synaptic potentiation.

From PMID:29133062 (nine patients with unilateral neonatal thalamic lesions):

"Thalamic volume loss ranged from 19% to 94%, predominantly on medial and dorsal nuclei and sparing the ventral thalamus. Lesions produced white matter loss and ventricle enlargement on the same hemisphere, which in four patients was associated with selective loss of thalamic-cortical fibers." [verbatim-verified from cache]

"Impact on EEG rhythms was mild, with a volume-loss-related decrease in alpha power and preservation of sleep spindles. The sleep continuous spiking was lateralized to the hemisphere with the lesion." [verbatim-verified]

"Unilateral selective thalamic-cortical disconnection is a common feature in our patients and is associated with both a focal pattern of CSWS and a pathological type of frequency-dependent excitability (peak: 10-20Hz). We propose that this excitability represents an abnormal synaptic plasticity previously described as the augmenting response. This synaptic plasticity has been described as absent in the corticocortical interactions in healthy experimental animals, emerging after ablation of the thalamus and producing a frequency-dependent potentiation with a peak at 10-20Hz. Because this response is potentiated by sleep states of reduced brainstem activation and by appropriate stimulating rhythms, such as sleep spindles, the simultaneous occurrence of these two factors in nonrapid-eye-movement sleep is proposed as an explanation for CSWS in our patients." [verbatim-verified — this is the money quote for the whole mechanism section]

The causal chain, node by node:

Early thalamic lesion (medial/dorsal nuclei)
  → selective loss of thalamocortical fibers (unilateral disconnection)
    → emergence of pathological corticocortical "augmenting response"
       (frequency-dependent potentiation peaking 10–20 Hz — absent in healthy cortex)
      → sleep spindles (10–16 Hz) + reduced brainstem arousal tone in NREM
 act as the ideal driving stimulus
→ runaway spike-wave activation confined to NREM sleep
  → [feeds Model B]

That's elegant: the spindle, a normal sleep rhythm, becomes the trigger pulse for a pathological potentiation the healthy brain doesn't have. Like a heart with a re-entrant circuit — the sinus beat isn't the problem, the abnormal pathway is; the normal rhythm just keeps lighting the fuse.

• UBERON:0001903 thalamic reticular nucleus (verified)
• UBERON:0001897 dorsal plus ventral thalamus (verified)
• UBERON:0000956 cerebral cortex, UBERON:0001950 neocortex (verified)
• GO:0021794 thalamus development (verified)
• CL:0000617 GABAergic neuron (TRN neurons), CL:0000598 pyramidal neuron, CL:0000679 glutamatergic neuron (all verified)

Model B — Disruption of sleep-dependent synaptic homeostasis (CANONICAL for the encephalopathy)

Model A explains why the spikes happen in sleep. Model B explains why the spikes make the child worse.

Normal picture: you potentiate synapses all day (learning), and slow-wave sleep runs a global downscaling program that renormalizes synaptic weight, preserving signal-to-noise and consolidating what matters. It's a nightly pruning shift — like the lymphatic system clearing the interstitium overnight, except for synaptic weight instead of fluid.

In SWAS, that shift doesn't happen. Per the literature summarized around Bölsterli et al. (impaired slow-wave downscaling in ESES):

"The profound spike activation in sleep found in CSWS disrupts synaptic homeostasis—the balanced synaptic potentiation during daytime and synaptic downscaling in sleep—leading to an inefficient cerebral network." [reported]

"Alterations in synaptic strength are shown through changes in sleep slow-wave activity (SWA), but notably during CSWS there are no sleep SWA changes, which occur again after CSWS remission" [reported]

That last observation is the strongest causal evidence available in humans: the overnight slope of slow-wave activity — the electrophysiological fingerprint of downscaling — flattens during the active phase and returns when SWAS remits. State-dependent, reversible, and time-locked to the clinical course.

And from PMID:25160535 (Issa 2014):

"Over the last 20 years, a variety of basic science findings suggest how spike-wave activity during sleep can cause the observed clinical outcomes." [verbatim-verified]

"The role of slow-wave sleep in normal cortical plasticity during developmental critical periods, how disruption of slow-wave sleep by electrographic seizures could affect cortical maps and development, and the organization and functional connectivity of the thalamic structures that when damaged are thought to produce these seizure disorders are reviewed." [verbatim-verified]

Chain:

Near-continuous NREM spike-wave
  → failure of sleep-dependent synaptic downscaling (flat overnight SWA slope)
    → saturated, non-selective synaptic weights → poor signal-to-noise
      → failed overnight memory consolidation + corrupted cortical map refinement
→ domain-specific regression matching the spike topography
  → (spikes remit at puberty) partial recovery, but the critical-period
     window for that cortical map has closed → residual permanent deficit

The critical-period framing is what explains the syndrome's cruellest feature: seizures stop, EEG normalizes, and the child still doesn't fully get the language back. The scaffolding came down before the building was finished.

• GO:0048167 regulation of synaptic plasticity (verified)
• GO:0060291 long-term synaptic potentiation, GO:0060292 long-term synaptic depression (verified)
• GO:0030431 sleep (verified)
• GO:0050803 regulation of synapse structure or activity (verified)
• GO:0050890 cognition, GO:0007613 memory (verified)

Model C — Molecular substrate: NMDAR/channel dysfunction and E/I imbalance

For the genetic cases, the proximate lesion is at the synapse.

GRIN2A encodes GluN2A, the subunit that dominates NMDA receptors in cortex from late infancy onward — exactly the developmental window of this syndrome. GluN2A-containing NMDARs have fast deactivation kinetics; alter them and you alter the temporal integration window for coincidence detection, i.e. the machinery of plasticity itself. Both directions break things: - GoF (TMD/linker missense): prolonged current, excess Ca²⁺ influx, excitotoxic/hyperexcitable phenotype → severe DEE - LoF (null, ATD/LBD missense): reduced NMDAR signalling — including on GABAergic interneurons, so net disinhibition → milder epilepsy-aphasia spectrum

The other genes converge on the same theme from different angles: Nav/Kv/Cav channelopathies (intrinsic excitability), SLC6A1/GAT-1 and SLC12A5/KCC2 (GABAergic inhibitory tone and chloride gradient), DLG4/PSD-95, CNKSR2, PPFIA3 (postsynaptic scaffolding and active zone). Plus the transcriptional-regulator cluster acting further upstream on the whole developmental program.

• GO:0004972 NMDA glutamate receptor activity (verified)
• GO:0035249 synaptic transmission, glutamatergic (verified)
• GO:0060079 excitatory postsynaptic potential (verified)
• GO:0098978 glutamatergic synapse (verified)
• GO:0005248 voltage-gated sodium channel activity, GO:0005249 voltage-gated potassium channel activity (verified)
• GO:0006357 regulation of transcription by RNA polymerase II (verified — for Cluster 2)
• UniProt: GluN2A = Q12879; PDB structures of NMDAR GluN1/GluN2A heterotetramer available

Note for the KB: this maps cleanly onto your existing module epilepsy_excitation_inhibition_imbalance — #Excitation-Inhibition Imbalance is an obvious conforms_to target. Model B (sleep-dependent downscaling failure) is not covered by any existing module and might be worth one, since it also touches your glymphatic_dysfunction module's territory conceptually (both are "sleep does maintenance work; disease blocks the maintenance") without duplicating it — glymphatic is extracellular clearance, this is synaptic weight renormalization. Different plumbing, same night shift.

Network-level: functional imaging

FDG-PET and EEG-fMRI (De Tiège et al., Epilepsia 2009) show the syndrome is a network disease, not a focal one:

"Hypermetabolism in perisylvian regions bilaterally and hypometabolism in lateral and mesial prefrontal cortex, precuneus, posterior cingulate cortex and parahippocampal gyri characterized the acute phase of CSWS. Altered functional connectivity was found between hyper- and hypometabolic regions" [reported]

The hypometabolic set is essentially the default mode network (see "Default mode network hypometabolism in epileptic encephalopathies with CSWS," Epilepsy Res 2014). The mechanism proposed is remote inhibition — the hyperactive epileptic focus actively suppresses distant connected cortex. That explains how a perisylvian spike focus produces a frontal-executive clinical syndrome: the deficit is downstream of the focus, not at it.

• UBERON:0000451 prefrontal cortex, UBERON:0016525 frontal lobe, UBERON:0001871 temporal lobe (verified)

Not involved

• Metabolic changes: no primary metabolic derangement (excepting the rare OPA3/Costeff case). Regional cerebral glucose metabolism is altered (above), but that's a consequence.
• Immune system: steroid responsiveness has long tempted people toward a neuroinflammatory hypothesis, and microglial (CL:0000129) / astrocytic (CL:0000127) contributions are plausible, but no autoantibody, no CSF inflammatory signature, no confirmed immune mechanism. Steroids may work via non-immune routes (direct effects on neuronal excitability, BBB, or neurosteroid pathways). Curate this as an explicit KNOWLEDGE_GAP — it's a real open question and the biggest unexplained therapeutic observation in the syndrome.
• Tissue damage: no necrosis, no fibrosis, no gliotic signature attributable to the SWAS itself. The "damage" is functional/synaptic, which is precisely why partial recovery is possible.

Molecular profiling

Essentially absent for this syndrome specifically. The one systems-level result worth citing is the Ann Neurol brain-specific gene co-expression analysis showing the two functional clusters. No GEO dataset, no proteomics, no metabolomics, no single-cell or spatial data specific to D/EE-SWAS. Large gap, and a legitimate one to record.

7. Anatomical Structures Affected

Organ level - Primary: brain (UBERON:0000955), specifically cerebral cortex (UBERON:0000956) and thalamus (UBERON:0001897) (verified) - Body system: central nervous system only. No systemic organ involvement — this is a purely neurological syndrome unless the underlying genetic cause is syndromic (e.g. Mowat-Wilson, Smith-Magenis, Christianson, Costeff bring their own multi-organ features) - Secondary: none organ-wise; secondary consequences are behavioral/educational/psychosocial

Regional - Thalamus — medial and dorsal nuclei preferentially; ventral thalamus spared (verified from PMID:29133062). Thalamic reticular nucleus (UBERON:0001903) is the spindle generator and is the mechanistic linchpin - Perisylvian cortex — hypermetabolic in the acute phase; the LKS substrate (superior temporal / auditory association cortex) - Frontal cortex (UBERON:0016525) and prefrontal cortex (UBERON:0000451) — hypometabolic; the CSWS/dysexecutive substrate - Precuneus, posterior cingulate, parahippocampal gyrus — hypometabolic (DMN) - Centrotemporal / rolandic region — the SeLECTS-spectrum spike focus - Ipsilateral white matter and lateral ventricle — volume loss / enlargement in thalamic-lesion cases

Lateralization Genuinely variable and clinically informative: unilateral/focal SWAS in thalamic-lesion and other unilateral structural cases (spiking lateralizes to the lesioned hemisphere — verified in PMID:29133062); bilateral/diffuse SWAS in genetic and idiopathic cases. Bilateral secondary synchrony from a unilateral generator is common. Worth curating as a distinguishing feature.

Cell level - CL:0000598 pyramidal neuron (cortical, layer V — the augmenting-response substrate) - CL:0000679 glutamatergic neuron - CL:0000617 GABAergic neuron (TRN, cortical interneurons) - CL:0000099 interneuron - CL:0010012 cerebral cortex neuron - CL:0000127 astrocyte, CL:0000129 microglial cell — speculative, no direct evidence (all verified against sqlite:obo:cl)

Subcellular - GO:0045202 synapse; GO:0098978 glutamatergic synapse (verified) - Postsynaptic density (GO:0014069), dendritic spine (GO:0043197), plasma membrane, axon initial segment — ⚠️ these four IDs are from memory, verify with OAK - No mitochondrial, lysosomal, ER, or nuclear-envelope pathology (excepting rare OPA3)

8. Temporal Development

This syndrome has one of the tightest and most reproducible time courses in pediatric neurology, which makes it very curatable.

Onset - Seizure onset: 2–12 years, peak 4–5 years; medians in cohorts cluster at 3.3 yr (DEE-SWAS) / 4.4 yr (EE-SWAS) [reported, Ann Neurol 2024] and 4 years in the Turkish cohort [reported] - Regression onset: ~1–2 years after seizures, typically 5–6 years - LKS auditory verbal agnosia onset: 3–9 years - Pattern: insidious-to-subacute. Regression can be gradual over months or, disconcertingly, abrupt over weeks. Fluctuation (especially in LKS language) is characteristic and often misread as behavioral or psychiatric

Stages

Progression rate & course: stepwise/subacute during the active phase, then a spontaneous, age-dependent remission — one of the few epileptic encephalopathies that reliably self-terminates. But the neurodevelopmental damage does not fully reverse.

Duration: active phase typically 2–5 years. From the Ann Neurol cohort: DEE-SWAS median epilepsy duration 10.0 years vs EE-SWAS 5.2 years [reported], and:

"Although developmental regression patterns were similar in both syndromes, DEE-SWAS was associated with a longer duration of epilepsy and poorer intellectual outcome than EE-SWAS." [verbatim-verified from cache]

Remission patterns: both spontaneous (age-dependent, near-universal for the EEG pattern and seizures) and treatment-induced (steroids/benzodiazepines can abolish SWAS in weeks). Relapse after treatment-induced remission is common — the diazepam and steroid literature is full of it — which is why prolonged/pulsed courses are used.

Critical period — the whole therapeutic rationale: the vulnerable window is the intervention window. Because the deficit accrues from cumulative SWAS exposure during an active cortical-map-refinement period, duration of ESES is the main predictor of neurocognitive outcome [reported, PMC3929187]. Every month of unsuppressed SWAS is irreversible developmental opportunity cost. This is the argument for early aggressive treatment and for annual sleep EEG surveillance in at-risk children.

9. Inheritance and Population

Epidemiology - Prevalence among childhood epilepsies: "0.5% to 0.6% of all childhood epilepsy cases" at tertiary referral epilepsy centers [reported, PMC3929187]; other sources give a wider 0.2%–2% of epilepsies [reported] - Population prevalence: ⚠️ Not reliably published. Orphanet classes it as rare (< 1 in 2,000). Back-of-envelope from a childhood epilepsy prevalence of ~0.5–1% and a 0.5% share gives an order of magnitude around 2–5 per 100,000 children, but that's a derived estimate, not a cited figure — do not curate it as a sourced prevalence. Use prevalence_class: UNKNOWN or NOT_YET_DOCUMENTED with a notes field, or cite the ORPHA:725 epidemiology row directly via the structured Orphanet cache - Incidence: no published incidence figure found. Gap. - Sex ratio: ~60:40 male:female [reported, PMC3929187]; 53% male in the Ann Neurol cohort [reported]. ILAE states both sexes equally affected [reported]. Treat as "slight male predominance or none" — sources disagree - Age distribution: exclusively pediatric onset (2–12), remitting around puberty. Adults exist only as survivors with residual deficits

Inheritance (for the genetic subset — ~34% of cases) - Predominantly de novo autosomal dominant. Most single-gene cases are de novo heterozygous variants - X-linked for CNKSR2 (~50% de novo; carrier mothers usually unaffected [reported]), MECP2, CUL4B, SLC9A6 - Autosomal recessive rarely (OPA3/Costeff) - HPO inheritance terms: HP:0000006 (AD), HP:0001417 (X-linked), HP:0001423 (X-linked dominant), HP:0001419 (X-linked recessive), HP:0000007 (AR) — ⚠️ verify these IDs with OAK; I did not check them - Penetrance: incomplete, explicitly documented for GRIN2A — OMIM #245570 notes "incomplete penetrance and intrafamilial variability, even among family members who carry the same GRIN2A mutation" [reported]. Use penetrance: INCOMPLETE - Expressivity: highly variable — the same GRIN2A variant can produce anything from asymptomatic to severe DEE across one family. This is the single best-documented genotype-phenotype caveat in the syndrome - Anticipation: not a feature, except in the one ATN1/DRPLA repeat-expansion case - Germline mosaicism: not documented specifically; theoretically possible for any de novo dominant gene, and standard recurrence-risk counselling (~1%) applies - Founder effects: none reported - Consanguinity: no established role (the recessive fraction is tiny) - Carrier frequency: not applicable at syndrome level; not meaningfully estimable

Population demographics - Ethnic/geographic: no established variation. Cohorts published from Australia/NZ, UK, Netherlands, Italy, France, Germany, Spain, Denmark, Turkey, Serbia, Romania, Malaysia, Hong Kong, USA — the syndrome appears globally with no reported prevalence differences. Ascertainment is heavily skewed to high-income countries with routine overnight EEG access, which is itself worth noting: you cannot diagnose this without a sleep EEG, so under-diagnosis in low-resource settings is near-certain - Variant geography: no population-specific variants reported

10. Diagnostics

The single indispensable test

Overnight / sleep EEG. Nothing else diagnoses this. A routine awake EEG can be entirely normal or show only modest focal spikes; the syndrome hides in NREM sleep. If you take one thing from this section: a child with unexplained developmental regression needs a sleep EEG, not a waking one.

EEG features: - Bilateral (or, less often, unilateral) continuous or near-continuous slow spike-wave during NREM sleep - Frequency 1.5–3 Hz (often stated as 1–2 Hz) - Marked attenuation in REM sleep and wakefulness — the state-dependence is the diagnostic signature - SWI highest in the first sleep cycle, declining across the night - Localization typically frontotemporal or centrotemporal - HP:0011182 Interictal epileptiform activity; HP:0010841 Multifocal epileptiform discharges; HP:0002353 EEG abnormality (verified)

Spike-wave index (SWI) — and its controversy. SWI = (minutes containing spike-wave × 100) / total NREM minutes.

Thresholds are genuinely unsettled, and this matters for any computable phenotype: - Classic/strict: ≥85% of NREM ("typical ESES") - Commonly used pragmatic: ≥50% - SWI <85% sometimes labeled "atypical ESES" - Resolution often defined as SWI <50% - The ILAE 2022 criteria deliberately do NOT specify a minimum percentage — they require "marked activation" clinically judged, precisely because the thresholds were never validated against outcome

⚠️ This is a real curation trap. Do not write "SWI ≥85% is the ILAE criterion" — it isn't. Model it as an open methodological question (KNOWLEDGE_GAP) with the competing thresholds recorded. The systematic review noted ~67.6% of published genetic cases were diagnosed using a >50% threshold [reported] — i.e. the literature isn't even internally consistent about who has the disease.

Imaging

• Brain MRI is mandatory. Looking for: thalamic lesion (often small, requires deliberate attention to medial/dorsal nuclei — easy to miss), polymicrogyria, periventricular leukomalacia, porencephaly, hydrocephalus/shunt, cortical dysplasia. Volumetric analysis may be needed for subtle thalamic volume loss (range in the published series: 19%–94%)
• FDG-PET (research/selected): perisylvian hypermetabolism + prefrontal/precuneus/posterior cingulate hypometabolism in the active phase
• SPECT: focal hyperperfusion with remote hypoperfusion
• EEG-fMRI: research tool for mapping the generator and remote effects
• Polysomnography with spindle quantification: emerging — sleep spindle density is both a biomarker and, per the rTMS work, a treatment-response correlate

Genetic testing

Given a 34% genetic yield, this is not optional. Recommended approach:

1. Chromosomal microarray (CMA) first or in parallel — CNVs are ~19% of the genetic yield (6/31 in the Ann Neurol cohort), and several recurrent CNVs (15q11.2-13.1 dup, 3q29 dup, Xp22.12 del, 16p13 del, 17q21.31 del, 17p11.2 dup/del) are microarray-detectable
2. Exome or genome sequencing — best single-test yield; genome adds CNV/structural resolution and non-coding coverage. The Seizure 2023 review concluded that "presentations occurring before age five warrant genetic investigation" [reported]
3. Epilepsy gene panels — acceptable but will miss the syndromic and novel genes; must include GRIN2A, GRIN2B, GRIN1, CNKSR2, SCN1A, SCN2A, SCN8A, KCNQ2, KCNQ3, KCNA2, KCNB1, KCNH5, KCNMA1, CACNA1A, ATP1A2, SLC6A1, SLC12A5, SLC9A6, DLG4, MECP2, FOXP1, ZBTB18, SETD1B, ARID1B, PUF60, CUL4B, ZEB2, PPFIA3
4. Single-gene GRIN2A — reasonable only when the phenotype is classic epilepsy-aphasia/LKS and cost is limiting
5. Karyotype/FISH: low yield; reserve for suspected specific rearrangements
6. Mitochondrial DNA testing: not indicated
7. Repeat expansion testing: not indicated (except the DRPLA outlier if there's a suggestive family history)
8. Variant classification per ACMG/AMP; interpret in ClinVar/ClinGen context. For GRIN2A, push for protein-domain annotation (ATD/LBD vs TMD/linker) because it changes the therapeutic hypothesis

Omics diagnostics: RNA-seq, proteomics, metabolomics, methylation episignature — none established for this syndrome. Metabolic workup is generally low-yield unless the phenotype suggests a specific IEM.

Clinical criteria

ILAE 2022 (Specchio et al., PMID:35503717) — DEE-SWAS/EE-SWAS, in substance: - Regression or plateauing in development affecting one or more of cognition, language, behavior, motor function - Marked spike-wave activation in NREM sleep, temporally related to the regression - Onset 2–12 years (peak 4–5) - Seizures usually present but not mandatory - EE-SWAS if development was normal before; DEE-SWAS if impaired before - LKS as a distinct EE-SWAS subtype defined by acquired auditory verbal agnosia

⚠️ I could not retrieve the position paper's formal mandatory / alerts / exclusionary tables — Epilepsia and the ILAE site both blocked automated fetch. Get the actual table from the PDF before curating definitions for this entry. The cached references_cache/PMID_35503717.md is abstract-only.

Differential diagnosis

Screening

• No newborn or population screening exists or is warranted
• Targeted surveillance is the meaningful intervention: serial sleep EEG in children with (a) early thalamic lesion, (b) polymicrogyria, (c) shunted hydrocephalus, (d) SeLECTS with new cognitive/behavioral change, (e) known pathogenic variant in a D/EE-SWAS gene. There is no formal guideline endorsing an interval — I'd flag "annual, or on any cognitive change" as expert-practice, not evidence-based
• Cascade testing where a familial variant is identified; be explicit about incomplete penetrance in counselling
• Carrier screening: not applicable

11. Outcome / Prognosis

Mortality: Not a fatal syndrome. No excess mortality established, no survival statistics, no life-expectancy reduction attributable to D/EE-SWAS itself. SUDEP risk is presumably that of the underlying epilepsy generally, but has not been quantified for this syndrome. The RESCUE ESES trial explicitly recorded "No deaths were reported" [verbatim-verified]. This is a morbidity disease, not a mortality disease — curate disease-specific mortality: none established rather than leaving it blank.

Morbidity — the actual endpoint. Persistent intellectual disability, language impairment, ADHD, and learning disability in most patients. Educational placement and independent adult functioning are the outcomes that matter.

Concrete numbers from the Ann Neurol 2024 cohort [reported]:

Plus, verbatim: "Phenotypic analysis highlights valuable clinical differences between DEE-SWAS and EE-SWAS which inform clinical care and prognostic counseling." [verbatim-verified]

From the older CSWS literature: "Most patients continue to demonstrate some degree of impairment" [reported], and "Duration of ESES seems to be the main predictor of neurocognitive function" [reported].

Recovery potential: Partial. Seizures and the EEG pattern remit near-universally around puberty. Cognition recovers partially — the earlier and more completely SWAS is suppressed, the more is recovered. Full return to premorbid function is uncommon, and in LKS specifically, complete language recovery is the exception; many are left with lasting receptive language impairment into adulthood.

Prognostic factors (best supported → weakest): 1. Duration of SWAS — the strongest predictor 2. Etiology — from the 50-child Serbian cohort: SeLECTS-background patients had "shorter symptom duration and superior prognosis, whereas those with structural etiologies experienced prolonged manifestations and reduced treatment efficacy" [reported, PMID:41076959] 3. DEE-SWAS vs EE-SWAS — pre-existing impairment predicts worse outcome (49% vs 8% moderate-severe ID) 4. Age at SWAS onset — earlier onset, worse (more of the critical period consumed) 5. Spike-wave index magnitude — higher SWI associated with more severe developmental disturbance [reported], though weaker/less consistent than duration 6. Time to effective treatment — the modifiable one

Prognostic biomarkers: SWI and its trajectory; sleep spindle density (emerging — the rTMS study found spindle increase correlated with IQ improvement, p=0.035 [reported]); overnight slow-wave-activity slope (research). No molecular/fluid biomarker exists.

Complications: educational failure, behavioral/psychiatric comorbidity, social exclusion; steroid-related complications from prolonged treatment (weight gain, hypertension, immunosuppression, bone effects) — treatment toxicity is a genuine part of the disease burden here.

12. Treatment

The uncomfortable headline: treatment for this syndrome rests on a single small, prematurely terminated RCT plus a lot of retrospective case series. Every decision below is made in an evidence twilight.

The one randomized trial: RESCUE ESES (PMID:38081201, Lancet Neurol 2024)

Corticosteroids vs clobazam, 8 tertiary centres, 7 European countries, children 2–12 diagnosed within 6 months, steroid- and clobazam-naive.

"At the 6-month assessment, an improvement of 11·25 IQ points or greater was reported for five (25%) of 20 children assigned corticosteroids versus zero (0%) of 18 assigned clobazam (risk ratio [RR] 10·0, 95% CI 1·2-1310·4; p=0·025)." [verbatim-verified from cache]

"An improvement of 0·75 points or more in the cognitive sum score was recorded for one (5%) of 22 children assigned corticosteroids versus one (5%) of 21 children assigned clobazam (RR 1·0, 95% CI 0·1-11·7, p=0·97)." [verbatim-verified]

"The trial was terminated prematurely, and the target sample size was not met, so our findings must be interpreted with caution. Our data indicated an improvement in IQ outcomes with corticosteroids compared with clobazam treatment, but no difference was seen in cognitive sum score. Our findings strengthen those from previous uncontrolled studies that support the early use of corticosteroids for children with EE-SWAS." [verbatim-verified]

Read that carefully before curating: the two co-primary outcomes disagreed. IQ favored steroids with a confidence interval you could drive a bus through (1.2 to 1310.4); the cognitive sum score showed literally nothing. 45 children enrolled against a target of 130 over eight years. This is suggestive evidence for steroids, not established efficacy — and it should be curated with that caveat intact, not laundered into "steroids are proven effective."

Regimens used: prednisolone 1–2 mg/kg/day oral continuous, or methylprednisolone 20 mg/kg/day IV ×3 days every 4 weeks (pulse). Clobazam 0.5–1.2 mg/kg/day.

Safety [verbatim-verified]: "Adverse events occurred in ten (45%) of 22 children who received corticosteroids, most frequently weight gain, and in 11 (52%) of 21 children who received clobazam, most often fatigue and behavioural disturbances."

Real-world effectiveness (PMID:41076959, Seizure 2025, n=50)

"corticosteroids (80.9%), clobazam (55.8%), levetiracetam (54.1%), and sulthiame (52.9%) were the most effective treatments." [reported]

Pooled historical data

A pooled analysis of 575 treated ESES cases found improvement in cognition or EEG most often with surgery (90%), steroids (81%), or benzodiazepines (68%), with standard ASMs least effective (49%) [reported]. Note the surgery figure is drowning in selection bias — only carefully chosen structural cases get operated.

Treatment table with NCIT/CHEBI annotations

† NCIT IDs taken from the verified list in CLAUDE.md. ‡ CHEBI IDs verified via sqlite:obo:chebi. Everything marked ⚠️ needs an OAK lookup before curating.

Contraindicated / to avoid

Carbamazepine (CHEBI:3387 ‡), oxcarbazepine, phenytoin, phenobarbital — can induce or worsen SWAS. "Carbamazepine is relatively contraindicated in ESES and should be discontinued" [reported]. This belongs in the entry as an explicit negative treatment recommendation — it's the cheapest intervention in the whole syndrome.

Precision medicine — the frontier

The rational-therapy story here is genuinely good, and it's the reason etiologic workup matters:

• GRIN2A gain-of-function (TMD/linker missense) → memantine (CHEBI:64312 ‡), an NMDAR open-channel blocker. Per Strehlow et al.: these individuals "represent promising candidates for treatment with NMDAR blockers, such as memantine" [reported]
• GRIN2A loss-of-function (null, ATD/LBD) → NMDAR positive allosteric modulators (investigational; L-serine has been tried for GRIN LoF)
• Sodium-channel GoF (SCN2A early-onset, SCN8A) → high-dose sodium channel blockers — note this directly conflicts with the general "avoid carbamazepine" rule, which is exactly why genotype matters
• Sodium-channel LoF (SCN1A, SCN2A late-onset) → avoid sodium channel blockers
• SLC6A1/GAT-1 LoF → GABAergic strategies under investigation

From the Ann Neurol conclusion: "Our etiological findings pave the way for the development of precision therapies." [verbatim-verified]

Emerging: rTMS via sleep-spindle restoration

Small (n=9) open-label study, PMID:40620003, and mechanistically the most interesting thing in the treatment literature because it tests Model B directly. Low-frequency (0.3–1 Hz) rTMS over the central facial motor area or a PET-identified hypometabolic zone, 10 workdays, 1,000–1,500 pulses/day: - Sleep spindle density: 55 → 91 (3 mo) → 147 (6 mo), p=0.002 [reported] - Median SWI: 81% → 68% → 57% (p=0.045, p=0.035) [reported] - Median IQ 72 → 83, and "improvement correlated significantly with sleep spindle increase (p = 0.035)" [reported] - "The mean probability of the sleep spindle coupling in the slow wave 'up' state increased from 28% to 55%" [reported]

n=9, open-label, no control — nowhere near practice-changing. But it's the first study to move a mechanistic variable (spindle–slow-wave coupling) and show cognition follow it. If Model B is right, this is what right looks like.

Treatment strategy in practice

1. Diagnose with sleep EEG; get MRI and genetics going in parallel
2. Stop any carbamazepine/oxcarbazepine/phenytoin/phenobarbital
3. First-line: corticosteroids (continuous or pulse) and/or high-dose benzodiazepine (clobazam/nocturnal diazepam) — early, because duration is the prognostic driver
4. Adjunct/alternative ASMs: levetiracetam, sulthiame, ethosuximide, valproate
5. Genotype-guided adjustment once results return
6. Ketogenic diet or IVIG in refractory cases (weak evidence)
7. Surgery if a resectable structural lesion is driving it
8. Throughout: speech therapy, special education, neuropsychological monitoring, serial sleep EEG to track SWI

Note the guideline vacuum. Per the European steroid survey (PMID:40301922), 60 centres in 18 countries, 11 different published steroid regimens, only 7 used as published: "Steroids are part of the first line therapy of (D)EE-SWAS across Europe, but heterogeneity in formulations, dosages, and regimens persists due to limited guideline availability." [reported]. And per the Seizure 2023 review: "Uniformity concerning the new definition of EE/DEE-SWAS, guidelines for management and more frequent genetic screening will be needed to guide best practices." [reported]

Clinical trials: RESCUE ESES = Dutch Trial Register NL43510.041.13 / ISRCTN42686094 (note: ISRCTN, not an NCT — your clinical_trials block wants NCT IDs, so this one may not fit the standard pattern). No large active NCT-registered interventional trial specific to D/EE-SWAS was identified; GRIN-disorder trials (memantine, L-serine) exist but enroll by gene, not by this syndrome.

13. Prevention

Short section, honestly assessed.

Primary prevention: No means of preventing the syndrome. The genetic cases are overwhelmingly de novo. The one partial lever is preventing the perinatal brain injuries (neonatal thalamic hemorrhage/infarction, IVH, HIE) that constitute the commonest structural etiology — i.e. general perinatal and neonatal care quality, not anything syndrome-specific. No vaccine, no immunization strategy applies.

The genuine preventive intervention is avoiding iatrogenic precipitation: don't put a child with SeLECTS on carbamazepine/oxcarbazepine. Small, cheap, real.

Secondary prevention (early detection) — this is where the value is: - Sleep EEG surveillance in the at-risk groups listed in §10 - Low threshold for sleep EEG in any child with unexplained developmental regression, language loss, or new behavioral deterioration - Rationale is explicit: because outcome tracks SWAS duration, shortening the delay to diagnosis is itself the therapy. Median diagnostic delay is not well quantified — gap.

Tertiary prevention (limiting damage in diagnosed patients): - Early aggressive SWAS suppression - Serial sleep EEG to confirm suppression and catch relapse - Serial neuropsychological assessment - Speech/language therapy and educational support to compensate for what can't be prevented - Monitor for steroid toxicity during prolonged courses (BP, weight, glucose — 98%/93%/64% of European centres respectively [reported])

Genetic screening / counselling: - No newborn or population screening - Post-diagnosis genetic counselling for solved cases; recurrence risk usually low (de novo) but not zero (germline mosaicism ~1%) - X-linked CNKSR2 families need proper carrier counselling — 50% recurrence for sons of carrier mothers - Prenatal testing / PGT technically available for known familial variants but complicated by incomplete penetrance and extreme variable expressivity — a GRIN2A variant carrier may be asymptomatic. This is a genuine counselling difficulty and worth recording - NCIT:C15240 Genetic Counseling

Public health / environmental interventions: not applicable.

14. Other Species / Natural Disease

Naturally occurring animal disease: none. There is no reported spontaneous DEE-SWAS analog in companion animals, livestock, or wildlife. I checked OMIA-adjacent literature and found nothing — this is unsurprising, because the diagnosis depends on (a) a sleep EEG and (b) documented loss of language, neither of which transfers to veterinary medicine. Curate as explicitly not applicable rather than leaving blank.

• Taxonomy: NCBITaxon:9606 Homo sapiens (only)
• Breed (VBO): not applicable
• Zoonotic potential / cross-species transmission: not applicable (non-infectious)

Comparative biology / evolutionary conservation: - The molecular substrate is deeply conserved: NMDA receptor subunits, voltage-gated Na⁺/K⁺/Ca²⁺ channels, and the thalamocortical circuit itself are conserved across mammals. Sleep spindles and slow-wave sleep are present in all mammals studied, and the synaptic homeostasis hypothesis was developed largely in rodents and Drosophila - Orthologs: Grin2a (mouse MGI, NCBI Gene 14811), Cnksr2 (mouse), Grin2a (rat, zebrafish grin2aa/grin2ab) - What does NOT conserve is the phenotype: the defining clinical feature is acquired aphasia / language regression, which has no animal correlate. This is a fundamental HUMAN_MODEL_MISMATCH and should be curated as one — models can reproduce the seizures, the sleep-EEG abnormality, and social/vocalization deficits, but never the syndrome's core clinical feature

15. Model Organisms

Mouse — the workhorse

Cnksr2 knockout mouse (PMID:34580165, J Neurosci 2021):

"Cnksr2 KO mice have increased seizures, impaired learning and memory, increased levels of anxiety, and loss of ultrasonic vocalizations (USV)." [reported]

Follow-up (eNeuro 2025, "The Epilepsy–Aphasia Syndrome Gene, Cnksr2, Plays a Critical Role in the Anterior Cingulate Cortex Mediating Vocal Communication") localized the USV deficit to excitatory neurons of the anterior cingulate cortex [reported].

This is the best available model, and the USV loss is the closest thing to a rodent "aphasia" readout that exists. But be careful how you phrase it: mouse ultrasonic vocalization is a social/affective signal, not language. It is analogous, not homologous, and that gap is exactly the HUMAN_MODEL_MISMATCH to record.

Grin2a mouse models — heterozygous and null lines exist. The Grin2a⁺/⁻ mouse shows changes in prefrontal cortex, insular cortex, superficial cortical layers, and thalamic reticular nucleus [reported], which is intriguing given the TRN's role in spindle generation. Caveat: most Grin2a mouse work is framed around schizophrenia, not epilepsy-aphasia, so the literature is oriented elsewhere and should be read with that in mind.

Thalamic lesion models: the "augmenting response" — the pathological frequency-dependent potentiation at the heart of Model A — was originally characterized in cats and rodents after experimental thalamic ablation (see PMID:29133062's discussion). These aren't disease models per se, but they're the physiological foundation of the leading mechanistic hypothesis. Worth curating as MODEL_ORGANISM evidence for the mechanism node even though they long predate the syndrome's molecular era.

Other genetic models available: Scn1a, Scn2a, Scn8a, Kcnq2, Kcnb1, Kcna2, Slc6a1, Mecp2, Arid1b, Foxp1, Zeb2 mouse lines all exist (MGI, IMPC/KOMP), constitutive and conditional. None were built as DEE-SWAS models; all are DEE models more broadly.

In vitro

• Xenopus oocyte and HEK293 two-electrode/patch electrophysiology — the workhorse for GRIN2A variant functional classification (this is how Strehlow et al. established the LoF/GoF domain split). This is the assay that makes the precision-medicine claim actionable
• Patient iPSC-derived neurons and cortical organoids — feasible and being pursued for GRIN and channelopathy genes; nothing published specific to D/EE-SWAS
• Cellosaurus/ATCC: no disease-specific cell line

Model limitations — be explicit

1. No rodent model reproduces the defining EEG pattern (near-continuous NREM spike-wave with sleep-state dependence) together with regression. This is the single biggest gap
2. No language phenotype is possible — the clinical core of the syndrome is untestable in animals
3. The age-dependent spontaneous remission at puberty — the syndrome's most distinctive natural-history feature — has no established animal correlate
4. Mouse sleep architecture (polyphasic, much shorter cycles, different spindle characteristics) differs enough that translating slow-wave downscaling findings to a child's overnight EEG requires real caution
5. Most single-gene models were made for a different indication and are studied under different phenotyping batteries

Research applications the models do support

• Variant functional classification → therapeutic stratification (in vitro electrophysiology; strongest translational value)
• Circuit dissection of thalamocortical/TRN contributions
• Sleep-dependent synaptic homeostasis mechanisms
• Preclinical testing of NMDAR blockers/PAMs
• Vocalization/social-communication readouts as a partial language proxy (Cnksr2)

Resources

MGI (informatics.jax.org), IMPC, KOMP/EuMMCR, IMSR, RGD, ZFIN, Alliance of Genome Resources, Cellosaurus.

Curation notes for the dismech entry

A few things I'd flag before this goes into kb/disorders/:

1. disease_term: MONDO:0800501 is verified and correct. It's a relatively new-ish MONDO term — per the repo's known gotcha, seed both cache/enums/diseaseterm_*.csv and cache/mondo/terms.csv from the local sqlite adapter, in the worktree and the primary checkout, or CI will fail with an OLS timeout dressed up as "term not found."

2. Model this as two subtypes, not one blob: DEE-SWAS and EE-SWAS, with LKS as a sub-subtype or a distinguishing-feature block under EE-SWAS. The 66%/28% etiologic-yield split and the 49%/8% ID-outcome split are the strongest sub-entity evidence in the literature and they'd be destroyed by lumping.

3. Existing module conformance candidates: epilepsy_excitation_inhibition_imbalance#Excitation-Inhibition Imbalance is a clean fit. The sleep-dependent-downscaling mechanism (Model B) has no existing module and is a plausible future one — it recurs conceptually across DEEs.

4. Curate the mechanism as competing hypotheses, not one chain. Model A (thalamocortical disconnection / augmenting response, status: CANONICAL for structural cases) and Model B (synaptic homeostasis failure, CANONICAL for the encephalopathy) are complementary; the neuroinflammatory/steroid-responsiveness explanation is a genuine KNOWLEDGE_GAP.

5. Three explicit gaps worth discussions entries: (a) the unvalidated SWI threshold (85% vs 50% vs ILAE's deliberate silence); (b) why steroids work, given no confirmed immune mechanism; (c) no disease-specific QoL instrument and no population prevalence/incidence figure.

6. One HUMAN_MODEL_MISMATCH: the Cnksr2 USV phenotype as a language proxy. Real, useful, and not the same thing as aphasia.

7. Evidence discipline: every [reported] quote above needs just fetch-reference PMID:xxxxx and a manual substring check before it becomes a snippet:. The [verbatim-verified] ones came out of full cached text and should survive validate-references as-is — and remember the ≥5-word minimum and no square brackets.

Sources

• ILAE syndrome page: DEE-SWAS / EE-SWAS
• Specchio et al., ILAE classification of childhood-onset syndromes, Epilepsia 2022 (PMID:35503717)
• Viswanathan et al., Solving the Etiology of D/EE-SWAS, Ann Neurol 2024 (PMID:39096015)
• van Arnhem et al., RESCUE ESES RCT, Lancet Neurol 2024 (PMID:38081201)
• Leal et al., Anatomical and physiological basis of CSWS after early thalamic lesions, Epilepsy Behav 2018 (PMID:29133062)
• Issa NP, Neurobiology of CSWS and Landau-Kleffner syndromes, Pediatr Neurol 2014 (PMID:25160535)
• Kravljanac et al., D/EE-SWAS cohort of 50 children, Seizure 2025 (PMID:41076959)
• The genetic landscape of DEE-SWAS, Seizure 2023 (PMID:37352690)
• Expanding the clinical and genetic landscape of (D)EE-SWAS: Turkish cohort, Neurogenetics 2024 (PMID:38388889)
• Genetic etiologies of ESES: systematic review (PMID:29976148)
• Strehlow et al., GRIN2A-related disorders: genotype and functional consequence predict phenotype, Brain 2019 (PMID:30544257)
• Lemke et al., Mutations in GRIN2A cause idiopathic focal epilepsy with rolandic spikes, Nat Genet 2013
• CNKSR2-related neurodevelopmental and epilepsy disorder: cohort of 13 families
• Cnksr2 Loss in Mice Leads to Increased Neural Activity and Behavioral Phenotypes of Epilepsy-Aphasia Syndrome, J Neurosci 2021 (PMID:34580165)
• The Epilepsy–Aphasia Syndrome Gene, Cnksr2, Plays a Critical Role in the Anterior Cingulate Cortex, eNeuro 2025
• Singhal & Sullivan, Continuous Spike-Wave during Slow Wave Sleep and Related Conditions, ISRN Neurol 2014
• European experience of steroid therapy in (D)EE-SWAS (PMID:40301922)
• rTMS improves cognition in SeLECTS with ESES via increase of the sleep spindle (PMID:40620003)
• ESES induced by oxcarbazepine in idiopathic focal epilepsy in childhood (PMID:26415787)
• Electrical Status Epilepticus in Sleep — StatPearls
• De Tiège et al., Insights into psychomotor regression in CSWS from FDG-PET and EEG-fMRI, Epilepsia 2009
• Default mode network hypometabolism in epileptic encephalopathies with CSWS, Epilepsy Res 2014
• Successful treatment of a child with EE-SWAS and GRIN2A variant using sulthiame, Cureus 2023
• Landau-Kleffner Syndrome: Current Etiopathogenesis and Management
• OMIM #245570 — Epilepsy, focal, with speech disorder (FESD)
• Treatment of DEE-SWAS — Practical Neurology
• Orphanet ORPHA:725