Oral–Gut microbiome coalescence and ecosystem fragility promote drug-resistant colonization and infection in relapsed/refractory leukemia

Abstract The host microbiome is not a passive bystander but an active determinant in acute leukemia (AL) treatment outcomes. Our dual-cohort microbiome analysis reveals its critical role: Cohort 1 (Longitudinal time-series, n=18): From initial diagnosis through chemotherapy and hematopoietic stem cell transplantation, we performed serial sampling of oral and gut microbiota across multiple treatment stages. Patients with relapsed/refractory AL (R/R AL, n=4) exhibited pronounced temporal microbiome fluctuations and a dysbiotic shift. The initial community, dominated by Enterococcus and Lactobacillus, transitioned to include Klebsiella, Lautropia, Bacteroides fragilis, and Veillonella. Cohort 2 (Cross-sectional, n=23): This cohort was collected contemporaneously with Cohort 1 and consisted of additional patients in their relapsed phase, enabling cross-sectional comparison during the R/R stage. Compared to patients in complete remission (CR), R/R AL patients demonstrated significantly lower gut microbiome alpha diversity and higher abundances of Enterococcus, Klebsiella, and Stenotrophomonas. LEfSe analysis identified increased facultative anaerobes (e.g., Lactobacillaceae family, Lacticaseibacillus) in R/R patients, while CR patients were enriched in symbiotic and short-chain fatty acid (SCFA)-producing bacteria (e.g., Prevotellaceae, Bacteroidia). Functional profiling indicated heightened microbial demands for lipopolysaccharide (LPS) biosynthesis and the tricarboxylic acid (TCA) cycle VIII pathway in non-remission states. Oral–Gut Microbiome Coalescence and Colonization Resistance: R/R patients showed significantly greater oral–gut microbiome coalescence (similarity) than CR patients (65.2% vs. 35.7% of paired samples, p = 0.081). Crucially, 3 out of 15 coalesced R/R samples developed carbapenem-resistant organism (CRO) colonization during follow-up, whereas none of the 5 coalesced CR samples did, suggesting impaired colonization resistance in the R/R state. Network analysis revealed a negative correlation between Lactobacillaceae and Prevotellaceae in R/R patients, contrasting with a positive correlation in CR patients. In vitro validation further confirmed this functional impairment: Microbiota derived from R/R AL patients exhibited significantly reduced resistance to CRO colonization in an anaerobic in vitro competition assay, as indicated by higher CRO colony counts compared to CR-derived microbiota (44.00 vs. 11.20, p = 0.0234). This directly demonstrates the compromised colonization resistance capacity in the dysbiotic microbial communities of R/R patients. Clinical Validation and Causal Pathway: To validate these findings, we analyzed our center's database (2017–2023; n = 1821 AL patients with Gram-negative bacteremia [GNB BSI]; 19.9% R/R AL). R/R AL was an independent risk factor for 30-day mortality (OR = 3.35, 95% CI: 1.85–6.05, p < 0.001) and CRO infections (OR = 1.57, 95% CI: 1.03–2.41, p = 0.038). Other risk factors included CRO colonization, hypoproteinemia, and prior carbapenem use. Causal mediation analysis demonstrated that CRO colonization partially mediated the association between R/R AL and CRO infections (22.8% of the effect, p = 0.040). Conclusion: The microbiome actively influences AL treatment beyond immune modulation, crucially maintaining colonization resistance. R/R AL patients exhibit a vulnerable microbiome phenotype characterized by loss of diversity, dysbiosis, oral–gut coalescence, and depletion of protective taxa. This ecological fragility leads to reduced resistance to drug-resistant pathogens and predisposes patients to CRO colonization and infection, contributing to poorer clinical outcomes. These findings underscore the need for integrated microbiome surveillance and intervention strategies in high-risk leukemia populations.