Résumé : Arteriovenous malformations (AVMs) are high-flow vascular anomalies arising from defects in vascular remodeling. This thesis focuses on the germline mutation–associated AVM syndromes: Capillary Malformation–Arteriovenous Malformation (CM-AVM), caused by RASA1/EPHB4, Through comparative clinical analysis and experimental modeling, we investigated how genotype, vascular architecture, and mechanosensory signaling interact to drive AVM development and localization.Clinically, we compared CM-AVM with another typical germline mutation AVM: HHT (Hereditary Hemorrhagic Telangiectasia) caused by ENG/ACVRL1/SMAD4/GDF2 gene mutation, and we show distinct lesion patterns and organ involvement. HHT-related AVMs often affect the visceral circulation, while CM-AVMs are more frequent in neonates and involve vein of Galen, pial AVF, musculoskeletal and deep dermal vessels. We identified that capillary curvature correlates with AVM localization: low-curvature vascular beds are more susceptible in CM-AVM, whereas high-curvature networks are more affected in HHT.In zebrafish models deficient in rasa1a or ephb4a, we recapitulated human-like vascular malformations, including persistent fistulas in the dorsal longitudinal vein (DLV). These defects result from failure of flow-mediated vessel fusion, a distinct angiogenic program. Mechanistically, RASA1 deficiency leads to an imbalance in the PI3K/MAPK pathway and impaired calcium signaling, thereby disrupting endothelial communication under flow. Pharmacological intervention with Selumetinib and YS-49 partially restored vessel fusion and signaling balance.This work advances our understanding of AVM pathogenesis by connecting molecular defects to vascular structure and hemodynamic context, and offers translational insights for mechanism-based therapies targeting AVMs.