A68-05 Multi-omic Integration Reveals Functional Signatures of Lung Microbiome and Microbe-host Interaction in Acute Respiratory Distress Syndrome in Pneumonia

Rationale The signature of the lung microbiota and its functional interaction with the host in pneumonia-induced acute respiratory distress syndrome (ARDS) remains poorly characterized, hindering the development of microbiota-targeting therapies. Methods Metagenomic analyses of bronchoalveolar lavage fluid (BALF) and sputum from a multicenter, prospective cohort of pneumonia patients with (n = 273) and without (n = 551) ARDS were performed to characterize microbiota dysbiosis. Concurrently, host responses were profiled through an integrated multi-omics approach on BALF samples, including proteomics (LC-MS/MS), transcriptomics, and metabolomics (untargeted LC-MS). To identify coordinated host response patterns, we applied Weighted Gene Co-expression Network Analysis (WGCNA) to the multi-omics datasets. Furthermore, integrative analyses were conducted to correlate key microbial features with the host multi-omics modules. Results Alpha diversity of microbiome declined in patients with ARDS in both BALF and sputum samples. The number of core microbes decreased in non-ARDS patients compared to ARDS patients and shifted to pathogenic microbes. The top three genus with increased abundance in BALF of ARDS patients were Elizabethkingia, Aspergillus, and Corynebacterium, while Prevotella, Peptostreptococcus, and Parvimonas were significantly decreased in abundance. We used five machine-learning models (random forest, XGBoost, SVM, LightGBM, and Lasso) to predict the ARDS-related microbiota phenotype using 44 different abundance genus and XGBoost achieved the best performance with AUC = 0.720. When comparing survival curve, patients with XGBoost model predicted ARDS-related microbiota phenotype had higher risk of death than those without (Log-rank test, p < 0.0001). After adjusting for age and sex, ARDS-related microbiota phenotype is an independent risk of in-hospital death (HR = 2.338 [95%CI: 1.005 - 1.037], p < 0.001), underscoring the significant association between this specific microbial shift and adverse outcomes in pneumonia patients. Integrative analysis of the lung microbiome and host multi-omics modules identified Prevotella spp. as the sole differentially abundant genus exhibiting a positive correlation with host endocytic pathways alongside a significant negative correlation with immune and inflammatory responses. This association pattern was inversely related to that of pathogenic taxa such as Klebsiella spp. Complementary metagenomic and metabolomic data revealed that this Prevotella spp. was associated with an active core carbon metabolism, notably enriched in pathways for 2-oxocarboxylic acid metabolism and the biosynthesis of amino acids including lysine, serine, and threonine. Conclusions This study defines an ARDS-specific lung microbiome signature associated with mortality in pneumonia patients. The unique association between Prevotella spp., its distinct carbon metabolism, and a tempered host immune response highlights a potential target for microbiota-directed interventions. This abstract is funded by: Shanghai Municipal Science and Technology Major Project (ZD2021CY001), Shanghai Municipal Science and Technology Major Project (ZD2021CY001),Science and Technology Commission of Shanghai Municipality (25ZR1401052)