Metagenomic analysis reveals severity-dependent microbial succession and correlation with host inflammatory response in oral and maxillofacial space infections

Background Oral and maxillofacial space infections (OMSI) vary widely in clinical severity, yet the relationships between microbial community patterns in the abscess niche and host inflammatory responses remain incompletely characterized. Methods We conducted a retrospective, cross-sectional, severity-stratified study of 197 patients diagnosed with OMSI between January 2020 and November 2023. Patients were stratified into mild (n=90), moderate (n=41), and severe (n=66) groups based on established clinical criteria. We performed mNGS on abscess pus samples to characterize the microbial community composition and assessed associations between these features and systemic inflammatory markers. Results Although α-diversity did not differ significantly among severity groups, β-diversity analysis revealed distinct microbial communities. Pairwise analyses indicated a threshold-like community shift, characterized by a significant divergence between mild and severe infections, while the moderate group exhibited an intermediate composition that overlapped with both. Severe infections were characterized by an enrichment of Prevotella. Furthermore, analysis of predominant taxa (>30% abundance) revealed considerable microbial heterogeneity, challenging a simple monoinfection model. Notably, a machine learning-identified microbial profile comprising Streptococcus , Corynebacterium , and Pseudomonas was significantly correlated with elevated systemic inflammatory markers. Conclusion This study characterizes associations between abscess-site microbial communities and host inflammatory profiles across OMSI severity strata. Given the cross-sectional design and the lack of an external validation cohort, the present findings should be interpreted as exploratory and non-causal. Future multicenter prospective studies including independent validation cohorts are warranted to test reproducibility and to evaluate whether any candidate features possess generalizable predictive value.