AUTHOR=Yu Haiyue , Guo Yongjing , Li Jialu , Fu Rong , Zhang Yunfeng , Guo Wanxu 

TITLE=Disruption of the gut bile acid-microbiota axis precedes severe bronchopulmonary dysplasia in preterm infants

JOURNAL=Frontiers in Microbiology

VOLUME=Volume 16 - 2025

YEAR=2025

URL=https://www.frontiersin.org/journals/microbiology/articles/10.3389/fmicb.2025.1705965

DOI=10.3389/fmicb.2025.1705965

ISSN=1664-302X

ABSTRACT=BackgroundBronchopulmonary dysplasia (BPD) remains a major cause of morbidity in preterm infants, yet current diagnostic criteria are delayed and underlying mechanisms are incompletely defined. Evidence suggests that intestinal dysbiosis may influence pulmonary outcomes via the gut–lung axis, but the metabolic mediators of this interaction remain unclear.MethodsWe conducted a prospective cohort study of 50 preterm infants (≤ 32 weeks gestation), stratified by BPD severity at 36 weeks. Stool samples collected on postnatal day 7 underwent 16S rRNA sequencing and targeted bile acid metabolomics. Differential features were identified via multivariate statistics and LEfSe. Spearman correlation analysis explored bile acid–microbiota interactions. An interpretable machine learning model (XGBoost) incorporating bile acid and microbial features was developed and validated using five-fold cross-validation and an independent test set.ResultsInfants with severe BPD showed significantly reduced levels of 16 bile acids—including primary, secondary, and sulfated species—compared to non-BPD controls. Gut microbiome β-diversity differed significantly among groups, with enrichment of opportunistic Proteobacteria (e.g., Brevundimonas) in severe BPD. Negative correlations were observed between depleted bile acids and enriched bacterial genera. The XGBoost model predicted BPD severity with 80% accuracy (AUC = 0.91), leveraging key features such as chenodeoxycholic acid (CDCA), hyocholic acid (HCA), and Brevundimonas.ConclusionsPreterm infants who develop severe BPD exhibit early disruption of the bile acid–microbiota axis, characterized by reduced bile acid levels and enrichment of opportunistic taxa. Integrating these features within interpretable machine-learning models enables accurate early risk stratification and provides mechanistic insights beyond traditional inflammation-based frameworks. Validation in larger, multicenter cohorts is warranted to refine biomarker panels and explore targeted interventions that modulate bile acid signaling or microbial ecology to prevent or attenuate BPD.