GRIN-related epilepsy in Chinese pediatric patients: a retrospective study of clinical, genetic, and mutant protein structural features in a tertiary center cohort.

BACKGROUND: N-methyl-D-aspartate receptors (NMDARs) play a pivotal role in neurodevelopment. While pathogenic variants in GRIN genes cause a broad spectrum of epileptic and developmental disorders, the precise molecular determinants that drive specific neurodevelopmental courses remain poorly understood. This study aims to utilize high-resolution structural modeling to identify associations between localized conformational alterations in NMDARs and distinct clinical phenotypes. METHODS: We retrospectively evaluated a cohort of 31 Chinese pediatric patients with GRIN-related epilepsy (GRIN1, n = 3; GRIN2A, n = 14; GRIN2B, n = 11; GRIN2D, n = 3) using detailed electroclinical and neurodevelopmental phenotyping. Concurrently, three-dimensional structural models for 23 missense variants were generated using AlphaFold2. Domain-specific root-mean-square deviation (RMSD) and Euclidean displacements of transmembrane helices were quantified. RESULTS: Patients in this cohort predominantly presented with early-onset epilepsy, multiple seizure types, and significant pharmaco-resistance, with 41.9% achieving seizure freedom after treatment. GRIN2B variants were primarily associated with developmental and epileptic encephalopathy (DEE) (63.6%, 7/11), whereas GRIN2A variants tended to be related with the epileptic encephalopathy (EE) course (71.4%, 10/14). Pathogenic variants were significantly clustered within the ligand-binding domain (LBD) and transmembrane domain (TMD). In silico structural analysis suggested that spatial displacement at the extracellular end of the M1 helix (M1 Top) might be associated with electrical status epilepticus during sleep (ESES) (median RMSD 0.54 Å vs. 0.37 Å, P = 0.038). Furthermore, a potential divergent structure-phenotype correlation was noted within the LBD: conformational alterations in the LBD-S1 region appeared more frequent in the developmental encephalopathy (DE) phenotype than in the EE group (median RMSD: 0.45 Å vs. 0.36 Å, adjusted P = 0.017), whereas LBD-S2 alterations were more prominent in the EE group than in the DEE group (median RMSD: 1.30 Å vs. 0.64 Å, adjusted P = 0.017). CONCLUSIONS: Localized conformational alterations within specific NMDAR micro-domains provide insights into pathogenic clinical phenotypes. However, the observed divergence suggests that geometric deviations alone may not account for clinical severity in a linear fashion. Our findings highlight the potential for in silico modeling to complement clinical phenotyping, but emphasize that precision medicine requires a multimodal framework combining structural predictions with functional validation.