Prevotella stercorea links gut microbiome ecology to respiratory infection protection through host-context-dependent, richness-independent associations

Using a longitudinal cohort of 633 Gambian children (IHAT-GUT, NCT02941081), we investigated two ecologically distinct pathways potentially linking Prevotella stercorea abundance to infection risk: a community-mediated pathway operating through overall gut microbiome richness, and a species-specific pathway operating independently of it. We modelled P. stercorea abundance in relation to gut microbiome richness, adverse-event incidence (diarrhoea, acute respiratory infection [ARI], fever), inflammatory biomarkers (CRP, AGP; measured at enrolment [D1] only), and weight-for-age z-score (WAZ), using richness-mediation analysis, negative binomial and joint-species models, and WAZ-stratified interaction models, with Benjamini–Hochberg correction applied within each family of related tests. P. stercorea abundance positively predicted gut microbiome richness, consistent with a role in community-level colonisation resistance for enteric outcomes. Its association with reduced ARI persisted largely unchanged after richness adjustment, consistent with a richness-independent, species-autonomous pathway. WAZ was uncorrelated with microbiome richness within strata, consistent with WAZ serving as a proxy for host immune-metabolic reserve rather than a determinant of microbiome composition. In Low-WAZ children, P. stercorea abundance at Day 1 was associated with suppressed CRP; in High-WAZ children, elevated Day 1 inflammation predicted subsequent P. stercorea abundance at Day 85. ARI and fever protection was richness-independent and concentrated in Low-WAZ children (q≤ 0.006); P. copri did not retain an independent protective association when modelled jointly with P. stercorea (q = 0.381 vs. q = 0.016). These findings are consistent with a richness-independent, host-context-dependent association between P. stercorea abundance and respiratory infection risk. Given the observational design, causal and mechanistic interpretations remain hypothesis-generating; experimental validation and independent replication are needed before translational application can be considered.