Bulk and single-cell RNA sequencing analysis with 101 machine learning combinations reveal neutrophil extracellular trap involvement in hepatic ischemia-reperfusion injury and early allograft dysfunction

Hepatic ischaemia–reperfusion injury (HIRI) is a major clinical concern during the perioperative period and is closely associated with early allograft dysfunction (EAD), acute rejection (AR) and long-term graft survival. Neutrophil extracellular traps (NETs) are extracellular structures formed by the release of decondensed chromatin and granular proteins following neutrophil stimulation. There is growing evidence that NETs are involved in the progression of various liver transplantation complications, including ischaemia–reperfusion injury (IRI). This study aimed to comprehensively analyse the expression patterns of NET-related genes (NRGs) in HIRI, identify HIRI subtypes with distinct characteristics, and develop a reliable EAD prediction model. Microarray, bulk RNA-seq, and single-cell sequencing datasets were obtained from the GEO database. Initially, differentially expressed NRGs (DE-NRGs) were identified using differential gene expression analyses. We then utilised a non-negative matrix factorisation (NMF) algorithm to classify HIRI samples. Subsequently, we employed machine learning algorithms to screen the hub NRGs related to EAD and developed an EAD prediction model based on these hub NRGs. Concurrently, we assessed the expression patterns of hub NRGs at the single-cell level using the HIRI. Additionally, we validated C5AR1 expression and its effect on HIRI and NETs formation in a rat orthotopic liver transplantation (OLT) model. In this study, we identified 11 DE-NRGs in the HIRI context. Based on these 11 DE-NRGs, HIRI samples were classified into two distinct clusters. Cluster1 exhibited a low expression of DE-NRGs, minimal neutrophil infiltration, mild inflammation, and a low incidence of EAD. Conversely, Cluster2 displayed the opposite phenotype, with an activated inflammatory subtype and a higher incidence of EAD. Furthermore, an EAD prediction model was developed using the four hub NRGs associated with EAD. Based on risk scores, HIRI samples were classified into high- and low-risk groups. The OLT model confirmed substantial upregulation of C5AR1 expression in the liver tissue, accompanied by increased formation of NETs. Treatment with a C5AR1 antagonist improved liver function, reduced tissue inflammation, and decreased NETs formation. This study distinguished two apparent HIRI subtypes, established a predictive model for EAD, and validated the effect of C5AR1 on HIRI. These findings provide novel perspectives for the development of advanced clinical strategies to enhance the outcomes of liver transplant recipients.