Automated real-time imaging of intestinal barrier integrity and molecular profiling for early outcome prediction in inflammatory bowel disease: endo-histo-barrier-omics study
Marietta Iacucci, Snehali Majumder, Irene Zammarchi, Giovanni Santacroce, Ivan Capobianco, Cecilia Lina Pugliano, Ujwala Chaudari, Pablo Meseguer Esbri, Brian Hayes, Rory Crotty, María R. Aburto, Maria Rocio del Amor, Bisi Bode Kolawole, Julia Eckenberger, Asma Amamou, Valery Naranjo, Enrico Grisan, Subrata Ghosh, Endo-Histo-Barrier-Omics Group, Miguel Puga‐Tejada, Louise Burke, Jane McCarthy, J. Alejandro Morales
Abstract
BACKGROUND: Barrier healing is an emerging therapeutic target in inflammatory bowel disease (IBD), though its assessment remains challenging. We evaluated automated advanced imaging for real-time barrier assessment, its correlation with epithelial/vascular barrier markers, and ability to predict adverse outcomes. METHODS: IBD patients and healthy controls undergoing endoscopic assessment were prospectively enrolled. The intestinal barrier was evaluated using ultra-high-magnification endocytoscopy and probe-based confocal laser endomicroscopy. Targeted biopsies were obtained from inflamed and non-inflamed segments. Epithelial and vascular barriers were assessed through automated multiplex immunofluorescence for Claudin-2, ZO-1, E-cadherin, PV-1, and CD31. Gene expression profiling was performed in epithelial and lamina propria compartments. Artificial intelligence (AI)-based analysis was employed for automated evaluation of barrier features captured by advanced imaging. RESULTS: In total, 103 patients were included (38 ulcerative colitis [UC], 54 Crohn's disease [CD], 11 healthy controls). Advanced imaging revealed barrier healing in 21% (8/38) of UC and 30% (16/54) of CD patients. In UC, Claudin-2 moderately correlated with abnormal crypt architecture (ρ = 0.49), goblet cell depletion (ρ = 0.5), and overall endocytoscopy activity (ρ = 0.49). In CD, PV-1 moderately correlated with altered blood flow (ρ = 0.41) and vessel architecture (ρ = 0.40). An integrated assessment of advanced imaging, combined with Claudin-2 and PV-1 expression, effectively predicted adverse outcomes in UC and CD, respectively. AI tools accurately classified epithelial and vascular barrier features captured by advanced imaging. Finally, gene expression confirmed upregulation of Claudin-2 and PV-1 in IBD. CONCLUSION: Automated advanced imaging enables real-time barrier assessment in IBD and correlates with markers of epithelial and vascular barrier impairment. AI integration can enhance standardization toward broader clinical applicability.