Systematic genetic, biochemical, and lipidomic characterization of GSD III fibroblast models reveals mis-annotation and defines a conserved disease lipid signature.

Glycogen storage disease type III (GSD III) is an autosomal recessive disorder caused by pathogenic variants in AGL, which encodes the glycogen debranching enzyme (GDE). Loss of GDE function impairs glycogen degradation, leading to accumulation of abnormal glycogen in liver and skeletal muscle. Clinical manifestations arise in early childhood and include hypoglycemia, hepatomegaly, and progressive myopathy. Despite significant morbidity, no disease-modifying therapies exist, due in part to a lack of rigorously validated cellular models. Patient-derived fibroblasts provide an accessible and scalable system that preserves each patient's endogenous AGL genotype and broader genetic background, enabling investigation of disease heterogeneity, genotype-phenotype relationships, and potential genetic modifiers. We performed whole-genome sequencing and phenotypic characterization of nine fibroblast lines originating from patients who were diagnosed with GSD III. Eight lines harbored biallelic pathogenic or likely pathogenic AGL variants, while one was reclassified based on absence of AGL variants and the presence of a pathogenic variant in an alternative metabolic gene. Validated GSD III fibroblasts exhibited consistent disease features, including absence of GDE protein, impaired glycogen utilization, and distinct lipidomic signatures. We generated an isogenic immortalized AGL-deficient fibroblast model which recapitulated these phenotypes. Together, these patient-derived and isogenic models provide a robust, genomically defined platform for studying the pathogenesis and phenotypic heterogenity of GSD III. Refined characterization and accessibility of these models will be useful to evaluate candidate GSD III therapeutics.