ABSTRACT: Background Acute necrotizing encephalopathy (ANE) is characterized by fulminant neuroinflammation and highly heterogeneous responses to immunotherapy in children. The molecular drivers of this heterogeneity remain unclear, and early biomarkers that can reliably predict therapeutic response and clinical outcome are still lacking. Methods Twenty-five pediatric ANE patients were enrolled and stratified into good-prognosis (GP, n=10) and poor-prognosis (PP, n=15) groups based on clinical outcomes. Longitudinal cerebrospinal fluid (CSF) samplings were collected at baseline and early after treatment, and profiled using high-resolution 4D-DIA proteomics. Differential expression, pathway enrichment, machine-learning modeling (LASSO, random forest, BART), trajectory analysis, and external validation were performed to identify prognostic signature molecules. Results At baseline, the two prognostic groups showed pronounced immune and metabolic differences. A total of 649 differentially expressed proteins (DEPs) indicated early disruption of synaptic membrane integrity, dysregulated extracellular-matrix remodeling, and altered amino acid metabolism. Despite standardized immunotherapy, the PP group exhibited persistent complement hyperactivation (elevated C6, C8A, and CFB) and sustained IL-1/MAPK inflammatory signaling. In contrast, the GP group demonstrated restoration of immune homeostasis, activation of metabolic recovery pathways, and enhancement of tissue-repair processes. Machine-learning analyses identified IGFBP1 as the strongest baseline predictor (AUC = 0.838). Following early treatment, a biomarker panel consisting of IL1RAP, CD81, PTPRT, CHID1, CLASP2, BTD, MAPK15, and PIN4 provided superior predictive performance (combined AUC = 0.991). Longitudinal trajectory analysis further segregated patients into two distinct molecular evolution patterns, termed “functional recovery” and “persistent injury”, which closely aligned with their clinical outcomes. Conclusions This study provides the first longitudinal CSF proteomic landscape of pediatric ANE during immunotherapy and identifies four central biological processes driving disease progression: abnormal complement activation, intense inflammatory responses, disturbances in energy and amino acid metabolism, and impaired neural tissue repair. IGFBP1 and IL1RAP have emerged as promising biomarkers for prognosis and precision intervention.
Chaonan Fan 
