C046 | PREDICTIVE VALUE OF MICROBIOME AND METABOLOMIC BIOMARKERS IN MYELODYSPLASTIC SYNDROMES AND ACUTE MYELOID LEUKEMIA

Acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) are heterogeneous hematologic malignancies characterized by genetic mutations, immune dysregulation, and microenvironmental alterations. Although the genetic landscape has been extensively explored, the contribution of the gut microbiome to disease evolution and treatment outcomes remains largely undefined. This study aimed to investigate the predictive value of gut microbiome in relation to infection risk, relapse, and survival in MDS and AML patients. A total of 101 AML and 105 MDS patients were enrolled at Humanitas Research Hospital (Milan, Italy). Blood, fecal, and bone marrow samples were collected at diagnosis and during treatment with hypomethylating agents (HMA) ± venetoclax (VEN) or intensive chemotherapy (IC). Microbiome analysis was performed through shotgun metagenomic sequencing, while metabolomic profiles were obtained from blood and fecal samples using LC-MS. Distinct gut microbiome signatures were associated with disease subtype, mutational status, infection risk, and treatment response. High-risk MDS exhibited depletion of commensal bacteria and enrichment of pathogenic species such as Olsenella sp., while secondary AML showed loss of beneficial taxa, including F. prausnitzii and Blautia spp. Microbiome composition differed according to genomic alterations, with NPM1- and CBF-mutated AML displaying distinct profiles compared to TP53-mutated cases. Patients experiencing infections, especially with multidrug-resistant organisms, had reduced microbial diversity and increased E. faecium abundance. IC treatment resulted in greater microbiome disruption and was associated with enrichment of antibiotic resistance genes (resistotypes). Unsupervised clustering identified four resistotypes, with resistotype R3 enriched in AML patients exposed to chemotherapy. Integrating microbiome and resistome features with clinical and genomic data in random forest models improved prediction of overall survival (AUCROC 0.75 vs 0.72) and relapse (AUCROC 0.91 vs 0.78). Key microbial predictors included the enrichment of Sellimonas intestinalis and depletion of Ruminococcus spp., correlating with fecal vitamin A derivatives and amino acids. Depletion of Bacteroides dorei, associated with anti-cancer immunity, emerged as a strong predictor of relapse. These findings support the integration of gut microbiome profiling into personalized risk stratification strategies in MDS and AML.