Abstract
BACKGROUND: Excessive immunogenic cell death (ICD) drives early mortality in sepsis, but its dominant subtype and clinical relevance remain poorly defined. METHODS: We systematically screened 3,421 genes involved in 15 forms of cell death across multiple sepsis transcriptomic datasets. Machine learning and COX regression identified two ICD-associated genes (PDZD8 and ADRB2) and were used to build a Cell Death Index (CDI). CDI performance was verified in terms of discrimination, risk stratification and clinical relevance. Gene set scoring was used to assess the association of CDI with necroptosis. Single-cell RNA sequencing was performed to investigate PDZD8 and ADRB2 expression and function. Treatment with the necroptosis inhibitor Necrostatin-1 (Nec-1) was evaluated in a cecal ligation and puncture (CLP) mouse model. RESULTS: The CDI robustly predicted sepsis prognosis across multiple datasets (AUC: 0.628-0.70). Multivariate analysis identified CDI as an independent risk factor for 28-day mortality (OR = 1.63-2.91, p < 0.001). CDI levels were significantly elevated in sepsis patients compared to pneumonia and upper respiratory infection patients. Meanwhile, serum high-mobility group box 1 (HMGB1) was markedly elevated in sepsis cases, whereas caspase-8 (CASP8) remained unchanged, supporting an active necroptotic process. Single-cell RNA sequencing revealed that PDZD8 was highly expressed in neutrophils and associated with oxidative stress and necroptosis, while ADRB2 was enriched in NK cells and linked to suppressed inflammatory signaling. Treatment with Nec-1 in the CLP model significantly improved survival (from 0% to 90%), reduced inflammatory cytokine release, and alleviated organ damage. CONCLUSION: This study identifies necroptosis as the dominant ICD subtype driving immune imbalance in sepsis and establishes CDI as a clinically relevant prognostic biomarker, thereby providing mechanistic insight into sepsis pathogenesis and lay the groundwork for precision therapeutic interventions targeting necroptosis.