Integrated Analysis Reveals Immunogenic Cell Death in Sepsis-induced Cardiomyopathy

BackgroundSepsis-induced cardiomyopathy (SIC) poses a significant challenge in critical care, necessitating comprehensive understanding and innovative diagnostic approaches. This study explores the immune-related molecular intricacies underlying SIC, employing bioinformatics analyses and machine learning techniques. MethodsRNA-seq and scRNA-seq datasets (GSE79962 and GSE190856) were obtained from the Gene Expression Omnibus (GEO). After initial quality control and preprocessing, scRNA-seq data (GSE190856) were analyzed using the Seurat package, including cell clustering and annotation. The CellChat package was then used to analyze immune cell interactions. Unsupervised clustering of SIC patients was performed based on differentially expressed ICD-related genes (GSE79962). Immune cell infiltration and gene set variation analysis were conducted, and weighted gene co-expression network analysis identified co-expression modules. A predictive signature for SIC was constructed through machine learning methods. ResultsTrough analyzing the GSE190856 scRNA-seq dataset, the communication between macrophages/monocytes and lymphocytes was found to be enhanced in mouse myocardial tissue during the early onset of SIC. Meanwhile, the expression level of ICD-related genes was upregulated in the monocytes infiltrating to the heart. These results suggestted that ICD may play a crucial role in the pathogenesis of SIC, which had been verified by the upregulated expression of ICD-related genes in the hearts of SIC patients in the GSE79962 dataset. The SIC patients were classified to 2 clusters, with cluster 1 exhibited an upregulation of the renin-angiotensin system, while cluster 2 displayed heightened activity in the RIG-I-like receptor signaling pathway. After comparing four machine learning models, the support vector machine (SVM) model exhibited better discrimination for SIC patients. By correlating the expression levels of the five crucial genes contained in this model with the clinical features of SIC patients, we found that JARID2 was negatively related to the Left Ventricular Ejection Fractions, while TNIP2 was negatively related to the variety of inotropes and vasopressors used in the SIC patients. ConclusionThis research unveils the correlation between ICD and SIC, offering insights into immune activity in the hearts during sepsis. The constructed SVM model with selected genes provides a promising molecular strategy for SIC diagnosis.