AI-powered discovery of a novel CB2R agonist as potential therapy for radiation-induced pulmonary fibrosis

Radiation-induced pulmonary fibrosis (RIPF) is an irreversible, often lethal late-onset lung injury induced by exposure of thorax to ionizing radiation, including thoracic radiotherapy and nuclear accidents, whose pathogenesis remains obscure and for which no approved targeted therapy exists. Here we identify the cannabinoid CB2 receptor (CB2R) as an ideal intervention node for RIPF, and apply artificial-intelligence drug design (AIDD) to discover YX2125, a nanomolar agonist ( K i = 3.8 nmol/L). In a murine RIPF model, oral administration of YX2125 (20 mg/kg) preserved lung architecture and restored exercise capacity. Bulk-RNA-seq coupled with qRT-PCR, Western blot and functional assays revealed that YX2125 abolishes the Ca 2+ –NFATc4–ATF3 feed-forward loop, suppresses the ferroptosis driver ACSL4, restores the xCT–GPX4 antioxidant axis and lipid peroxidation, and consequently blocks fibrogenic signaling (TGF- β , COL1A1, FN1). Protection of YX2125 was nullified by genetic CB2R deletion or the selective antagonist SR144528, confirming its on-target activity. YX2125 also reduced early neutrophil/macrophage infiltration and promoted M1-to-M2 polarization in a CB2R-dependent manner. Together, these data establish CB2R agonism as a tractable anti-RIPF strategy and position YX2125 as a first-in-class candidate. CB2R agonist YX2125 generated by AIDD, alleviates RIPF by suppressing ACSL4 and restoring GPX4 to inhibit ferroptosis, while promoting M2 macrophage polarization, serves as a promising first-in-class therapeutic candidate.