Pyridoxine-Dependent Epilepsy

1. Disease Information

Pyridoxine-dependent epilepsy (PDE), in its classic form, is an autosomal recessive neurometabolic epileptic encephalopathy caused by biallelic loss-of-function variants in ALDH7A1, the gene encoding α-aminoadipic semialdehyde dehydrogenase (α-AASA dehydrogenase), better known as antiquitin (ATQ). The defining clinical signature is early-onset, drug-refractory seizures that stop — clinically and electrographically — in response to large daily doses of pyridoxine (vitamin B6), and that recur if pyridoxine is withdrawn.

The current NCBI GeneReviews framing: "Pyridoxine-dependent epilepsy – ALDH7A1 (PDE-ALDH7A1) is characterized by seizures not well controlled with anti-seizure medication that are responsive clinically and electrographically to large daily supplements of pyridoxine (vitamin B6)" (GeneReviews, NBK1486).

Key identifiers:

Synonyms / alternative names: antiquitin deficiency; α-AASA dehydrogenase deficiency; pyridoxine-dependent seizures (PDS, older term); vitamin B6-dependent epilepsy (ALDH7A1 type); EPEO4. Note that "folinic acid-responsive seizures" (FARS) is now known to be the same disorder — allelic to PDE, caused by ALDH7A1 variants (Gallagher et al., 2009 [VERIFY PMID: 19128019]).

Data provenance: the knowledge here is drawn almost entirely from aggregated disease-level resources (OMIM, Orphanet, GeneReviews) and published cohort/case-series literature — not EHR-level individual patient records. The largest structured cohorts are the international PDE registry work behind Mills et al. 2010 and the consensus recommendations of the international PDE consortium.

Sources: GeneReviews NBK1486, OMIM #266100, MedlinePlus Genetics.

2. Etiology

Primary cause — genetic: biallelic (homozygous or compound heterozygous) pathogenic variants in ALDH7A1 (chromosome 5q23.2). This is a monogenic inborn error of lysine catabolism; there is no meaningful infectious or acquired etiology for classic PDE.

Causal chain in one breath: loss of antiquitin → block in the saccharopine/lysine-degradation pathway → build-up of Δ¹-piperideine-6-carboxylate (P6C) → P6C chemically inactivates pyridoxal 5′-phosphate (PLP, the active B6 cofactor) → functional B6 deficiency in the brain → seizures.

Genetic risk factors: - Causal variants: >165 published pathogenic ALDH7A1 variants (see §4). The single most common is c.1279G>C (p.Glu427Gln, historically "E399Q"), present in ~30% of European patient alleles. - Consanguinity raises risk (as for any AR disorder) and is over-represented in some reported cohorts. - Founder effects: a Dutch founder haplotype underlies the E399Q allele in several apparently unrelated Dutch families (Bennett/Salomons et al., 2007, "An intriguing 'silent' mutation and a founder effect in antiquitin (ALDH7A1)", PMID:17721876).

Environmental risk / protective / gene-environment factors: essentially none in the conventional sense — penetrance is complete and disease expression does not depend on exposures. The one true gene-environment interaction is therapeutic: dietary lysine load worsens metabolite accumulation (more substrate → more toxic product), while dietary lysine restriction and L-arginine supplementation reduce it (arginine competes with lysine for transport into brain and mitochondria). So "diet" behaves as a modifiable environmental lever on a fixed genetic defect rather than a cause. No protective genetic alleles are described.

Sources: Mills 2006, Nat Med, Bennett/Salomons 2007, Coughlin 2015.

3. Phenotypes

PDE is more than epilepsy — it's an encephalopathy with a developmental footprint. The phenotype spans a remarkable range, captured perfectly by Mills et al. 2010:

"...from ventriculomegaly detected on foetal ultrasound, through abnormal foetal movements and a multisystem neonatal disorder, to the onset of seizures and autistic features after the first year of life." [VERBATIM-VERIFIED, PMID:20554659]

Core phenotypes with suggested HPO terms:

Severity/progression: seizures are severe and drug-refractory until B6 is given, then dramatically responsive. Neurodevelopmental outcome, however, is frequently impaired independent of seizure control — the striking, clinically important dissociation. Coughlin 2015 states it plainly:

"75% of individuals with PDE have significant developmental delay and intellectual disability" [VERBATIM-VERIFIED, PMID:26026794]

Late-onset/atypical presentations tend to have more favorable cognition, attributed partly to the absence of neonatal seizure-induced injury.

Quality-of-life impact: driven mainly by the intellectual/developmental disability rather than by seizures once controlled — implies lifelong caregiver support, special education, and communication/motor limitations. Formal EQ-5D/PROMIS data specific to PDE are sparse; QoL is inferred from developmental-outcome cohorts.

Sources: Mills 2010, Brain, GeneReviews NBK1486, Coughlin 2023 review, PMC12360241.

4. Genetic / Molecular Information

Causal gene: ALDH7A1 (aldehyde dehydrogenase 7 family member A1), OMIM *107323, chromosome 5q23.2. Encodes antiquitin, an NAD⁺-dependent dehydrogenase.

Enzyme function (verbatim):

"the nicotinamide adenine dinucleotide-dependent dehydrogenation of l-alpha-aminoadipic semialdehyde/L-Delta1-piperideine 6-carboxylate." [VERBATIM-VERIFIED, PMID:20554659]

Variant landscape: - >165 pathogenic variants published across the gene (missense, nonsense, frameshift, splice-site, and larger deletions). - Most common variant: c.1279G>C (p.Glu427Gln; legacy nomenclature p.Glu399Gln / "E399Q") — ClinVar RCV000019610; dbSNP rs121912707. Reported in ~30% of European PDE alleles. (Note the two coordinate systems: the transcript-based p.Glu427Gln and the historical mature-protein-based E399Q refer to the same variant — a common source of chart confusion.) - Founder effect: the E399Q allele carries a Dutch founder haplotype (PMID:17721876). - Variant classification follows ACMG/AMP; most recurrent alleles are Pathogenic/Likely Pathogenic in ClinVar. - Functional consequence: loss of function (abolished α-AASA/P6C dehydrogenase activity). Not gain-of-function, not dominant-negative — carriers are asymptomatic. - Origin: germline (constitutional). Somatic variation is not relevant. - Allele frequency: individual pathogenic alleles are rare in gnomAD; carrier-frequency modeling gives a disease incidence estimate of ~1:64,352 live births (see §9).

Modifier genes: none robustly established. Residual antiquitin activity of specific missense alleles correlates loosely with age of onset/severity, so genotype itself is the main modifier of expressivity.

Epigenetics / chromosomal abnormalities: no disease-specific methylation signature or recurrent large chromosomal rearrangement is characteristic; PDE is a classic single-gene point-mutation/small-indel disorder. Rare whole-gene or multi-exon deletions occur and can be missed by sequencing alone (argues for deletion/duplication analysis when only one variant is found).

Suggested annotations: gene → hgnc:877 (ALDH7A1, verify); GO molecular function GO:0004029 (aldehyde dehydrogenase (NAD+) activity) and GO:0047718 / lysine-catabolism-specific activity; GO:0030170 pyridoxal phosphate binding (for the downstream affected enzymes).

Sources: Mills 2006, Coughlin 2019 genotypic spectrum, PMC6345606, ClinVar RCV000019610, SNPedia rs121912707.

5. Environmental Information

Classic PDE is not driven by environmental, lifestyle, or infectious factors — it's a pure inborn error. The relevant "environmental" dimensions are entirely dietary/therapeutic and secondary:

• Dietary lysine intake modulates substrate flux into the blocked pathway (higher lysine → more α-AASA/P6C).
• Catabolic stress (fasting, febrile illness) can precipitate breakthrough seizures in some patients, an important management caveat.
• No toxin, radiation, occupational exposure, or pathogen is implicated. Infectious workup matters only because the neonatal presentation mimics sepsis/meningitis and delays diagnosis.

Sources: Coughlin 2015, GeneReviews NBK1486.

6. Mechanism / Pathophysiology

This is the heart of the entry — the causal chain from broken enzyme to seizure. Here's the cascade, upstream → downstream:

Step 1 — Enzyme block (upstream trigger). Antiquitin normally converts α-aminoadipic semialdehyde (α-AASA) to α-aminoadipate in the saccharopine pathway of lysine degradation (the brain's main route for breaking down lysine). Loss of antiquitin stalls this step. Cellular compartment: the reaction and its collapse center on cytosolic/mitochondrial lysine catabolism.

Step 2 — Metabolite accumulation. α-AASA piles up and sits in a spontaneous chemical equilibrium with its cyclic Schiff-base form, Δ¹-piperideine-6-carboxylate (P6C). Pipecolic acid also rises (parallel lysine-degradation branch).

Step 3 — The chemical sabotage (the crux). P6C reacts with pyridoxal 5′-phosphate (PLP) — the active form of vitamin B6 — via a Knoevenagel condensation, forming an inactive adduct. Straight from the founding paper:

P6C "inactivates pyridoxal 5'-phosphate (PLP) by forming a Knoevenagel condensation product." [VERBATIM-VERIFIED, PMID:16491085]

This is a chemical trap, not an enzyme-cofactor competition — the P6C literally consumes and neutralizes PLP.

Step 4 — Functional B6 (PLP) deficiency. PLP is the cofactor for ~140 enzymes. The seizure-critical casualty is glutamic acid decarboxylase (GAD), the PLP-dependent enzyme that makes GABA (the brain's principal inhibitory neurotransmitter). PLP depletion → less GABA synthesis.

Step 5 — Excitation/inhibition imbalance → seizures (downstream clinical output). Falling GABAergic inhibition (with likely secondary glutamate/neurotransmitter dysregulation, since PLP also serves aromatic amino acid decarboxylase and others) produces neuronal hyperexcitability and hypersynchrony → refractory seizures. This is why the disease is a downstream conformer of the generic epilepsy_excitation_inhibition_imbalance module (#Excitation-Inhibition Imbalance is the natural conforms_to target).

Step 6 — Independent neurotoxicity (the reason B6 alone isn't enough). Accumulated α-AASA/P6C (and possibly the metabolite 6-oxo-pipecolic acid) are thought to be directly neurotoxic and neurodevelopmentally damaging, which explains the ~75% intellectual-disability rate despite seizure control — and the entire rationale for substrate-reduction (triple) therapy.

Cell types & structures: GABAergic neurons (CL:0000617 GABAergic neuron), broadly cortical/subcortical neurons; the corpus callosum (isthmus) is structurally hypoplastic.

Suggested GO / CHEBI terms: - GO biological process: GO:0019477 L-lysine catabolic process; GO:0009448 GABA metabolic process; GO:0042816 vitamin B6 metabolic process; GO:0006536 glutamate metabolic process. - CHEBI chemicals: pyridoxal 5′-phosphate CHEBI:18405; pyridoxine CHEBI:16709; L-lysine CHEBI:18019; GABA CHEBI:16865; pipecolic acid CHEBI:17964; α-aminoadipic acid CHEBI:37024 (verify all IDs with OAK).

Molecular profiling / omics: untargeted metabolomics has been the productive omics angle — recent work identified novel pyridoxine-independent diagnostic markers (6-hydroxy-2-aminocaproic acid [HACA] and a C₉H₁₁NO₄ isomer), plus 2-oxopropyl-P6C / 6-oxo-pipecolic acid as emerging biomarkers. No characteristic transcriptomic/proteomic/lipidomic disease signature is established beyond the lysine-pathway metabolite fingerprint. Functional genomics: the zebrafish CRISPR knockout (see §15) is the main perturbation model.

Sources: Mills 2006, Global metabolomics, PMC9784804, Biomarkers review, doi:10.3390/biom16040486.

7. Anatomical Structures Affected

• Organ / system level: the central nervous system is the primary and essentially sole target (UBERON:0000955 brain; nervous system UBERON:0001016). The neonatal multisystem picture (feeding, respiratory, metabolic disturbance) is a functional/metabolic spillover rather than fixed organ pathology.
• Regional/structural: corpus callosum (UBERON:0002336) — thin posterior segment (isthmus) is near-universal; ventricular system (ventriculomegaly), cisterna magna (mega cisterna magna), and scattered white-matter abnormalities, cortical dysplasia, and hydrocephalus in subsets.
• Tissue / cell level: neurons, especially GABAergic neurons (CL:0000617); the defect is biochemical/global rather than a focal lesion.
• Subcellular: lysine catabolism spans cytosol and mitochondrion (GO:0005739 mitochondrion; GO:0005829 cytosol); the PLP-dependent reactions affected are largely cytosolic.
• Lateralization: structural changes (callosal thinning, ventriculomegaly) are typically bilateral/midline; seizures may be focal or generalized.

Sources: GeneReviews NBK1486, Coughlin 2023, PMC12360241.

8. Temporal Development

• Onset: classically neonatal — seizures within the first hours to weeks of life, often with prolonged seizures and recurrent status epilepticus. A meaningful minority present late/atypical, with onset up to ~2–3 years (rarely into later childhood/adolescence). Fetal presentations (abnormal fetal movements, ventriculomegaly on prenatal ultrasound) are documented.
• Onset pattern: acute/dramatic seizure onset on a background of a chronic, lifelong metabolic defect.
• Course: chronic, lifelong — pyridoxine dependence is permanent; withdrawal reliably brings seizures back (a diagnostic feature historically, though rechallenge is now discouraged when genetic/biochemical confirmation is available). With treatment the course is stable with respect to seizures but the developmental disability is largely fixed/static rather than progressive-degenerative.
• Breakthrough seizures: can occur with intercurrent illness, fasting, or medication lapses; myoclonic seizures and status epilepticus are risk factors for breakthroughs; folinic acid is added when pyridoxine responsiveness is incomplete.
• Critical window (key actionable point): early treatment matters. Reported observation — a delay of up to ~4 days may not add harm, but delays >1 week associate with increased risk of learning difficulties and cerebral palsy; the substrate-reduction (triple) therapy also works best when started early.

Sources: GeneReviews NBK1486, Coughlin 2015, nationwide age-span study, ScienceDirect S0920121123000244.

9. Inheritance and Population

• Inheritance: autosomal recessive (HP:0000007). Bind inheritance_term to HP:0000007 Autosomal recessive inheritance.
• Penetrance: effectively complete in biallelic pathogenic-variant carriers; heterozygous carriers are unaffected.
• Expressivity: variable — from severe classic neonatal encephalopathy to milder late-onset, correlating loosely with residual enzyme activity.
• Genetic anticipation: not applicable (not a repeat-expansion disorder).
• Germline mosaicism: not a described recurrence mechanism.
• Founder effect / consanguinity: Dutch founder haplotype for E399Q (PMID:17721876); consanguinity elevates risk as for any AR condition.
• Carrier frequency: used to derive incidence estimates (below).

Epidemiology: - Historical clinical-diagnosis incidence estimates vary widely: ~1:20,000 (a single German center), 1:396,000 (Netherlands), 1:783,000 (UK). - Carrier-frequency-based modeling gives ~1:64,352 live births — likely a better population estimate, since clinical case-finding under-ascertains atypical/late-onset cases. - Orphanet classes it as a rare disease (prevalence <1/1,000,000 to a few per million depending on region).

Suggested Prevalence records (dismech structured format): - population: Germany (single-center) · measure_type: BIRTH_PREVALENCE · rate ~5 per 100,000 (1:20,000) · notes: highest regional estimate. - population: Worldwide (carrier-frequency model) · measure_type: BIRTH_PREVALENCE · rate ~1.55 per 100,000 (1:64,352) · prevalence_class: BAND_1_9_PER_100000. - population: United Kingdom · measure_type: BIRTH_PREVALENCE · rate ~0.13 per 100,000 (1:783,000).

Demographics: no strong sex bias (AR disorder; M:F ≈ 1:1). Reported across many ethnic groups worldwide; specific variants show regional clustering (E399Q in European/Dutch populations). Age distribution is dominated by neonatal/infant diagnosis, with a long tail of later-recognized atypical cases.

Sources: GeneReviews NBK1486, Coughlin 2019, PMC6345606, Bennett/Salomons 2007.

10. Diagnostics

Biochemical (first-line, and the historical breakthrough): - Urinary/plasma/CSF α-AASA (α-aminoadipic semialdehyde) — the classic, robust diagnostic marker. Mills 2006 established that measurement of urinary α-AASA provides a simple diagnostic confirmation. - P6C (Δ¹-piperideine-6-carboxylate) — in equilibrium with α-AASA; sum of AASA+P6C tracked in therapy. - Pipecolic acid — elevated in plasma/CSF/urine, but less specific (rises in other conditions, e.g. peroxisomal disorders) and pyridoxine treatment lowers it, so it can normalize on treatment. - Emerging pyridoxine-independent markers: 6-hydroxy-2-aminocaproic acid (HACA), a C₉H₁₁NO₄ isomer, and 2-oxopropyl-P6C / 6-oxo-pipecolic acid — useful because they stay elevated even after treatment starts (helpful when a patient is already on B6).

Genetic confirmation (definitive): - ALDH7A1 sequencing (single-gene or via gene panel). Because rare exonic/whole-gene deletions are missed by sequencing, add deletion/duplication (CNV) analysis if only one variant is found. - WES/WGS increasingly first-line, especially for atypical/late presentations; rapid genome sequencing has diagnosed late-onset B6-dependent epilepsy. - Prenatal diagnosis feasible once familial variants are known (Mills 2006: gene analysis enables prenatal diagnosis).

Clinical / therapeutic test: - Pyridoxine trial: IV pyridoxine (with EEG and cardiorespiratory monitoring — apnea/hypotonia can follow the first dose) producing prompt clinical + electrographic seizure cessation. Historically a diagnostic pyridoxine-withdrawal rechallenge was used; now discouraged in favor of biochemical + genetic confirmation. - Folinic acid consideration when pyridoxine responsiveness is incomplete (given the FARS = PDE identity).

Imaging: MRI shows thin posterior corpus callosum (isthmus) (near-universal, demonstrable by geometric morphometry), plus ventriculomegaly, mega cisterna magna, white-matter changes, occasional cortical dysplasia/hydrocephalus. Imaging supports but does not confirm.

EEG: variable — burst-suppression, multifocal/generalized epileptiform discharges; the electrographic response to pyridoxine is itself informative.

Differential diagnosis — the other vitamin B6-dependent epilepsies (critical to distinguish):

The unifying frame: "Vitamin B6-dependent epilepsies are caused by mutations in at least five different genes involved in B6 metabolism... The ALDH7A1, PNPO, ALPL, ALDH4A1, and more recently PLPBP genes have been implicated" (PLPBP review, PMC7932866). NEC caution for curation: because these disorders share the "vitamin B6-dependent epilepsy" label, deep-research tools are prone to conflating ALDH7A1 with PNPO/PLPBP — verify that every cited paper is specifically about ALDH7A1 before quoting.

Screening: not yet in most standard newborn-screening panels, though α-AASA is being evaluated as a newborn-screening analyte; cascade/carrier testing for relatives once a familial variant is known.

Sources: Mills 2006, Metabolomics biomarkers, PMC9784804, PLPBP review, PMC7932866, GeneReviews NBK1486.

11. Outcome / Prognosis

• Survival / mortality: with prompt, sustained pyridoxine therapy, survival is generally good; the danger is undiagnosed/untreated disease, where refractory status epilepticus can be fatal. No robust disease-specific survival percentages, but early death is uncommon once treated.
• Morbidity — the defining prognostic reality: ~75% have significant developmental delay / intellectual disability despite good seizure control (Coughlin 2015 [VERBATIM-VERIFIED]). Motor impairment, language delay, and behavioral/autistic features are common.
• Prognostic factors: (1) time to diagnosis/treatment — delays >1 week worsen outcome; (2) phenotype/onset — late-onset patients tend to have better cognition; (3) genotype/residual activity; (4) use of adjunctive substrate-reduction therapy started early.
• Disease course: seizures become controllable and stable; the neurodevelopmental deficit is largely static (present from early on rather than degenerative). Breakthrough seizures occur with illness/fasting/nonadherence.
• Recovery potential: seizures — excellent with B6; cognition — limited once established, which is exactly why the field has pushed toward earlier and substrate-reducing treatment.

Sources: Coughlin 2015, Coughlin 2023 review, PMC12360241, GeneReviews NBK1486.

12. Treatment

Foundation — pyridoxine (vitamin B6) supplementation, lifelong. - Restores the depleted PLP pool by mass action, rescuing GAD/GABA synthesis. Immediate seizure control is the hallmark. - Dosing individualized; caution at first IV dose (apnea/hypotonia risk → monitor). Excess pyridoxine risks sensory neuropathy, so dose is balanced. - CHEBI:16709 pyridoxine. Suggested MAXO: dietary/vitamin supplementation (MAXO:0000088 dietary intervention as the closest broad term) + pharmacotherapy (NCIT:C15986) with therapeutic_agent pyridoxine (CHEBI:16709). Verify best MAXO term with OAK.

Substrate-reduction "triple therapy" (pyridoxine + dietary lysine restriction + L-arginine supplementation) — targets the neurotoxic-metabolite arm that B6 alone doesn't fix:

triple therapy "further reduced toxic metabolites, and in some subjects appeared to improve neurodevelopmental outcome" and "early diagnosis and treatment with this new triple therapy may ameliorate the cognitive impairment in PDE." [VERBATIM-VERIFIED, PMID:26026794] - Lysine restriction = less substrate feeding the blocked pathway (MAXO:0000088 dietary intervention; CHEBI:18019 L-lysine). - L-arginine competes with lysine for the brain/mitochondrial transporter, lowering intracerebral lysine (CHEBI:16467 L-arginine). - Best results when started early.

Adjunct — folinic acid for incomplete pyridoxine responsiveness or breakthrough seizures (folinic-acid-responsive seizures are the same ALDH7A1 disorder; CHEBI:63606 folinic acid).

Anti-seizure medications: generally insufficient alone (that refractoriness is diagnostic), but sometimes used adjunctively during stabilization.

Pharmacogenomics: not a major factor — treatment is genotype-agnostic vitamin/dietary therapy rather than metabolized small-molecule drugs.

Advanced / experimental therapeutics: no approved gene therapy, cell therapy, or RNA therapy for PDE as of this review; substrate-reduction optimization and earlier diagnosis (newborn screening) are the active translational frontiers. Check ClinicalTrials.gov for current lysine-restriction / arginine and biomarker studies (no landmark NCT to cite as established standard here).

Supportive/rehabilitative: developmental services, PT/OT/speech therapy, special education for the ID component.

Sources: Coughlin 2015, Effect of lysine restriction + arginine, PMID:27324284, Consensus recommendations, ScienceDirect S1096719211001661, GeneReviews NBK1486.

13. Prevention

• Primary prevention: not preventable at the individual level (genetic). Population approach = carrier screening / genetic counseling for at-risk families, prenatal diagnosis, and preimplantation genetic testing where familial variants are known.
• Secondary prevention (early detection): the biggest opportunity — earlier diagnosis (biomarker + genetic) to start treatment before neurodevelopmental damage accrues. Newborn-screening evaluation of α-AASA is an active area precisely because early treatment improves outcomes.
• Tertiary prevention (limiting complications in diagnosed patients): sustained pyridoxine adherence, triple therapy to reduce neurotoxic metabolites, sick-day management to prevent breakthrough seizures during fasting/illness, and developmental support.
• Counseling: AR recurrence risk 25% per pregnancy for carrier couples → genetic counseling is standard.

Sources: GeneReviews NBK1486, Coughlin 2015.

14. Other Species / Natural Disease

• Taxonomy: human disease (NCBITaxon:9606). ALDH7A1 is highly evolutionarily conserved — hence the name antiquitin ("ancient" gene).
• Orthologs: conserved orthologs in mouse (Aldh7a1), zebrafish (aldh7a1), and beyond; the deep conservation of the lysine-degradation/aldehyde-dehydrogenase function is what makes cross-species modeling informative.
• Naturally occurring animal disease: no well-established spontaneous companion-animal or wildlife equivalent of PDE is described (check OMIA for any veterinary antiquitin phenotype; none prominent).
• Comparative biology: conservation of the mechanism (lysine catabolism → α-AASA/P6C → PLP inactivation) is what allows the zebrafish model to faithfully reproduce the human biochemistry.

Sources: Zebrafish model, PMC5714462, OMIM *107323.

15. Model Organisms

Zebrafish (the flagship model): aldh7a1⁻/⁻ knockout generated with CRISPR-Cas9 — the first genetic PDE animal model. It recapitulates the human disease remarkably well: - Seizure behavior: spontaneous rapid locomotion and circling swim, earliest ~8 dpf; EEG shows large-amplitude spike discharges vs wild type. - Pharmacology matches humans: "the seizures show an almost immediate sensitivity to pyridoxine and pyridoxal 5′-phosphate, with a resulting extension of the life span" (paraphrased from the model paper). - Biochemistry matches humans: "Impaired lysine degradation with accumulation of PDE biomarkers, B6 deficiency, and low γ-aminobutyric acid levels were observed in the aldh7a1−/− larvae" — directly ties the metabolite build-up to the low-GABA/seizure mechanism. - Evidence source for these = MODEL_ORGANISM (in vivo animal).

Two independent zebrafish reports: Pena et al., Genetics 2017 (doi:10.1534/genetics.117.300137) and Zabinyakov et al., PLOS One 2017 (PMID:29053735).

Mouse: no widely characterized Aldh7a1 knockout mouse recapitulating the full PDE phenotype was found in this search (a notable model gap — worth flagging as a KNOWLEDGE_GAP or HUMAN_MODEL_MISMATCH candidate in the entry).

In vitro / cellular / computational: recombinant antiquitin biochemistry and structural/computational analysis of missense variants (e.g., ScienceDirect S000927972400139X) support loss-of-function interpretation; iPSC/organoid PDE models are not yet established.

Applications: the zebrafish model is used to study seizure mechanism, screen B6-vitamer and substrate-reduction responses, and validate biomarkers.

Sources: Zebrafish, Genetics 2017, Zabinyakov PLOS One, PMID:29053735, Zebrafish model PMC5714462.

Curation cheat-sheet (for the dismech entry)

• Module conformance: epilepsy_excitation_inhibition_imbalance#Excitation-Inhibition Imbalance — PDE is a clean downstream conformer (PLP depletion → GAD failure → low GABA → E/I imbalance → seizures). Consider also flagging the lysine-catabolism/metabolic-block chain (it's an intoxication-flavored inborn error, though the acute-decompensation module fits less cleanly than for the classic organic acidemias).
• Highest-confidence PMIDs with verified verbatim quotes: 16491085 (Mills 2006, mechanism/Knoevenagel), 20554659 (Mills 2010, phenotype spectrum + enzyme reaction), 26026794 (Coughlin 2015, 75% ID + triple therapy).
• Leads to fetch-and-verify before quoting: 17721876 (founder), 29053735 (zebrafish), 27324284 (lysine/arginine), 19128019 (Gallagher, FARS=PDE — [VERIFY PMID]), plus the biomarker/metabolomics papers.
• Watch for NEC: confirm every source is ALDH7A1-specific, not PNPO/PLPBP/ALPL — they all wear the "vitamin B6-dependent epilepsy" jacket.

Sources (primary set): - Mills PB et al., Nat Med 2006 — PMID:16491085 - Mills PB et al., Brain 2010 — PMID:20554659 - Coughlin CR et al., Mol Genet Metab 2015 — PMID:26026794 - GeneReviews: Pyridoxine-Dependent Epilepsy – ALDH7A1, NBK1486 - OMIM #266100 (EPEO4) · OMIM *107323 (ALDH7A1) - Coughlin genotypic spectrum 2019, PMC6345606 - Bennett/Salomons founder effect 2007 — PMID:17721876 - Global metabolomics biomarkers, PMC9784804 - PLPBP/B6-epilepsy differential review, PMC7932866 - Zabinyakov zebrafish model — PMID:29053735 · Zebrafish PMC5714462 - Coughlin current-perspectives review 2023, PMC12360241 - MedlinePlus Genetics: PDE

That's the whole map, Harry. The one-sentence version if you're skimming: a broken lysine-cleanup enzyme (antiquitin/ALDH7A1) lets a metabolite build up that chemically strangles vitamin B6, the brain runs out of the cofactor it needs to make its "calm down" signal, and you get B6-rescuable seizures — but the metabolite is also quietly toxic, which is why ~75% of kids still have intellectual disability even when the seizures are under control, and why the field moved to triple therapy. Just remember every PMID here is a lead until just fetch-reference confirms the snippet — I verified three of them verbatim, the rest need the usual cache-and-check pass before they land in the entry.