🎗Integrative Transcriptomics and Machine Learning Identify Inflammatory and Profibrotic signalling as a Central Driver of Immune-Mediated Fibrosis in Chronic Liver Allograft Rejection.

Introduction: Chronic liver allograft rejection (CR) causes progressive graft fibrosis and late graft loss after transplantation. Reliable molecular predictors of early graft injury due to CR are lacking, and the immunopathogenic mechanisms linking alloimmune inflammation to fibrogenesis remain poorly understood. This study aimed to identify predictive biomarkers of CR using transcriptomic profiling and machine-learning approaches and to evaluate the role of the IL-33/ST2 signalling pathway as a potential driver of graft fibrogenesis.
Methods: Human liver biopsies from 29 transplant recipients were analyzed using RNA sequencing (RNA-seq) [No Rejection (NR), n=4; CR, n=4], immunohistochemistry (IHC) [NR, n=8; CR, n=10], and gene-expression profiling [NR, n=12; CR, n=6; controls, n=9]. Control samples were obtained from histologically normal donor liver tissue. Differential expression analysis (DESeq2) identified altered genes, followed by feature selection using machine-learning algorithms (LASSO, Random Forest, and SVM-RFE) to identify candidate biomarkers. Functional enrichment analyses (Gene Ontology, PanglaoDB, DisGeNET, KEGG) and protein–protein interaction networks (STRING/Cytoscape, MCC algorithm) identified key pathways and hub genes. Immune cell infiltration was inferred by transcriptomic deconvolution using CIBERSORTx. Histological and molecular validation included IHC, Sirius Red staining for fibrosis, immunofluorescence to assess ST2-positive cells, and RT-qPCR for inflammatory, fibrogenic, and IL-33/ST2 pathway genes. Biopsies were also cultured ex vivo with or without recombinant human IL-33. Data were analyzed using Kruskal–Wallis and Spearman tests (p<0.05). The study was approved by the HUFF IRB (DDI 2419).
Results: RNA-seq identified 947 upregulated and 647 downregulated genes in CR vs NR grafts. Enrichment analysis demonstrated activation of cytokine–receptor interactions, IL-17 signalling, and neutrophil chemotaxis pathways. Machine-learning models identified four predictive genes (PLK2, ACACB, RAPGEF5, NUP50-DT). PPI analysis revealed a central inflammatory–fibrogenic network involving IL6, IL1B, TGFB1, CCL2, and MMP9. Transcriptomic deconvolution indicated immune remodeling characterized by macrophage and B-cell infiltration.
CR biopsies showed periportal infiltration by CD3⁺ and CD4⁺ T cells and CD20⁺ B cells, with increased collagen deposition. Fibrogenic genes TGF-β and ACTA-1 were upregulated in CR; α-SMA correlated with METAVIR fibrosis score (p<0.05). Hepatic IL-33 and ST2 expression were significantly increased in CR and correlated with biochemical markers of liver injury (AST, ALT, ALP, TBIL) and fibrosis (p<0.05). Ex vivo IL-33 stimulation induced fibrogenic changes in liver biopsies.
Conclusions: CR appears driven by inflammatory and profibrotic signalling involving the IL-33/ST2 pathway. This integrative transcriptomic and machine-learning approach highlights potential biomarkers and therapeutic targets for early detection and intervention.