Mechanistic drivers of PD-L1/CTLA-4 checkpoint inhibitor–associated immune toxicity and systemic organ injury

Background Immune checkpoint inhibitors targeting PD-1/PD-L1 and CTLA-4 have substantially improved outcomes in multiple malignancies, but they may also trigger dysregulated immune activation accompanied by systemic toxic effects and multi-organ injury. Although immune-related adverse events are increasingly recognized in oncology practice, the clinical architectures through which immune toxicity progresses across organ systems, and the biological pathways that may support earlier risk assessment and mechanism-informed intervention, remain insufficiently defined. Methods We performed a translational multicohort study integrating clinical phenotyping, biomarker modeling, and mechanistic validation. In a development cohort derived from a large United States electronic health record database, cross-organ co-occurrence networks were used to characterize reproducible patterns of organ involvement, and data-driven phenotypes were derived using non-negative matrix factorization and hierarchical clustering, followed by landmark outcome analyses. A biomarker-informed risk framework integrating immune-toxicity features with conventional severity indices was subsequently evaluated in an independent United States external validation cohort. To assess biological concordance across clinical and experimental levels, parallel analyses were conducted in biospecimens from a prospectively enrolled single-center Chinese patient cohort and in preclinical models, with emphasis on pyroptosis-related signaling, myeloid inflammatory activation, endothelial–coagulation coupling, and mitochondrial injury. Results Immune-toxicity-associated systemic injury demonstrated stable cross-system architectures that segregated into clinically interpretable phenotypes with graded mortality risk, organ-support requirements, and recovery trajectories. Compared with conventional severity assessment alone, incorporation of immune-toxicity biomarker axes improved risk discrimination and supported more refined patient stratification across cohorts. Mechanistic analyses further identified prominent activation of pyroptosis-related pathways together with mitochondrial stress, while myeloid-dominated inflammation and endothelial activation were associated with propagation of systemic injury across organs. These findings linked phenotypic heterogeneity at the clinical level with biologically plausible injury programs observed in patient biospecimens and experimental systems. Conclusions This study establishes a translational framework connecting clinical phenotypes, biomarker-based risk stratification, and biologically supported injury pathways in immune checkpoint inhibitor-associated systemic toxicity. The findings may support earlier recognition and structured clinical assessment of patients with suspected immune-toxicity-associated organ injury and provide a rationale for future mechanism-guided intervention studies. Further multicenter, diagnostically controlled, and ethnically diverse investigations are needed to determine the generalizability of these findings and their utility in distinguishing immune-toxicity-associated critical illness from other causes of systemic inflammation and organ failure.