A55-23 Two Stable, Lung Microbiome Assemblages in Severe Pneumonia

Rationale While the lung microbiome demonstrates relevance to respiratory health, a pressing need to identify stable, reproducible signatures reflective of prognosis in severe pneumonia exists. Methods We prospectively assessed n = 16,319 individuals with severe pneumonia from five regions in China (i.e. Shanghai, Zhengzhou, Tianjin, Chengdu and Guangzhou). A ScanTrack algorithm, employing single-nucleotide polymorphism (SNP) signatures was developed to distinguish microbial colonization from environmental contaminants. The lung microbiome (by metagenomics) was related to clinical outcomes and pneumonia severity including assessment of host immune status. PERMANOVA was employed to elucidate clinical and demographic factors contributing to microbiome variation and cluster-specific co-occurrence networks constructed for age-stratified sub-groups to assess for potential microbial assemblages. An external prospective cohort from a separate region in China (Xi’an: n = 155; including 29 in-hospital deaths) and two retrospective cohorts of inpatients at Shanghai Ruijin Hospital (n = 577) and Jiangsu Provincial People’s Hospital (n = 216) with severe pneumonia were used to validate the microbial assemblages and their prognostic performance. Results Age-dependent changes to the lung microbiome were detected. We identified two stable, core microbial assemblages which were reproducible and conserved across all cohorts, a 76 species N-Core and 34 species P-Core signature. The N-Core assemblage correlates with reduced host inflammation and enhanced pathogen resistance, while the P-Core assemblage links to pro-inflammatory states, reduced microbial diversity and greater pathogen susceptibility. N-Core/P-Core features serve as robust biomarkers for clinical prognosis with log(N/P) values > 0 linked to improved survival and shorter hospitalization. Integrating N-Core/P-Core features alongside host immunity and age in machine-learning classifiers markedly enhanced diagnostic accuracy, achieving a mean diagnostic AUC of 0.80 across five centers and external validation datasets. The validation cohort demonstrated that individuals with a log(N/P) > 0 exhibit significantly prolonged survival and shorter duration of hospitalization. Patients with elevated CURB-65 or CUB-65 scores (indicating greater disease severity) demonstrate significantly lower log(N/P) values. Conclusions Our findings describe two core microbial architectures modulating host and pathogen dynamics in severe pneumonia. These data suggest that the lung microbiome has predictive and prognostic in severe pneumonia. This abstract is funded by: Noncommunicable Chronic Diseases-National Science and Technology Major Project(2023ZD0506200)