Type: Oral
Session: 803. Emerging Tools, Techniques, and Artificial Intelligence in Hematology: New Approaches to Predicting Patient Outcomes in Hematologic Malignancies
Hematology Disease Topics & Pathways:
Research, Acute Myeloid Malignancies, AML, MDS, Adult, Translational Research, MPN, Clinical Research, Chronic Myeloid Malignancies, Diseases, Real-world evidence, Adverse Events, Myeloid Malignancies, Biological Processes, Microbiome, Technology and Procedures, Multi-systemic interactions, Study Population, Human, Omics technologies
Methods. The study population consisted of 551 prospective patients with MN (including 198 AML, 196 MPN and 157 MDS) enrolled at Humanitas Cancer Center (Milan, Italy). Patients were evaluated at diagnosis and at multiple time points throughout the disease natural history. DNA extraction was performed on fecal material and subjected to next-generation shotgun metagenomics sequencing using the Illumina NovaSeq platform. Raw sequencing reads were mapped against the Integrated Gene Catalogue 2 of gut bacteria using METEOR whereafter MetaOMineR was used to generate abundances of metagenomic species pan-genomes. Differentially abundant taxa were calculated using Wilcoxon ranks-sum tests with FDR-adjustment. LC-MS was used to identify hydrophilic and hydrophobic metabolites on both blood and fecal samples. Metabolites annotation and quantitation were analyzed using Compound Discoverer 3.3.
Results. Microbiome richness and diversity were significantly decreased in AML compared to MDS, MPN and healthy controls, indicating overall community dysbiosis. Principal coordinate analysis showed that MDS, MPN and AML subtypes have distinct microbiome compositions (P<0.01). We identified several common and unique differentially abundant species associated with MN subtypes as well as with specific disease-related genomic profiles (including RNA splicing, NPM1, TP53, JAK/STAT mutations and 5q deletion). Pathobionts are enriched in MN while bacteria commonly associated with the gut microbiome of healthy individuals such as SCFAs producers, are depleted in MDS, MPN and AML: co-occurrence network analyses identified Eggerthella lenta, Clostridium innocuum and Actinomyces oris as common key MN driver species. Importantly, distinct gut microbiome signatures correlated with specific metabolomics profiles in plasma.
Next, we performed 16S sequencing of mononuclear marrow cells from a subset of MN patients (n=80) that showed the presence of several pathogenic bacteria genera that are enriched in the gut, such as Actinomyces and Streptococcus, thus suggesting the microbiome may have a direct ‘intratumoral’ role in shaping the microenvironment of MN.
In MDS/AML patients treated with chemotherapy (3+7 regimen) and hypomethylating (HMA) agents +/- venetoclax (VEN), we identified specific microbiome and metabolomics signatures associated with primary refractoriness/high-risk of disease relapse. By random forest analysis, microbiome, fecal and blood metabolomics significantly improved individual prediction of disease relapse vs. clinical-genomic features alone (AUCROC 0.94 vs 0.71; Model accuracy 0.86 vs 0.69, respectively). Feature selection allowed the identification of key species and fecal and blood metabolites that can be used as markers to improve individual prediction of disease relapse.
Microbiome diversity and richness were significantly reduced in MN patients with infections, with a further depletion observed in those with multi-drug resistant infections. Additionally, we found that in MDS/AML patients undergoing treatment, both the microbiome composition and metabolomics profile changed significantly. This change is consistent with gut mucosal damage in patients treated with chemotherapy. In contrast, HMA therapy +/- VEN was more protective against gut dysbiosis, aligning with a reduced risk of infectious complications vs. 3+7 regimen.
Conclusion. Integrative analysis of the gut microbiome and metabolomics provides crucial information to dissect the clinical heterogeneity and outcomes of patients with MN. Bacterial species, along with fecal and blood metabolites, can be utilized as biomarkers to enhance the prediction of relapse and the risk of infections associated with specific treatments, thus improving personalized patient management.
Disclosures: Harrison: Sobi: Consultancy; AOP: Consultancy, Honoraria, Speakers Bureau; BMS: Consultancy, Honoraria, Speakers Bureau; MSD: Consultancy, Honoraria, Speakers Bureau; Geron: Consultancy; IMAGO: Consultancy, Honoraria, Speakers Bureau; Janssen: Consultancy; Keros: Consultancy, Honoraria, Speakers Bureau; Galecto: Consultancy; MPN voice: Other: Leadership role; Incyte: Consultancy, Honoraria, Other: Teaching and Speaking; Research: PI, Speakers Bureau; AbbVie: Consultancy, Honoraria, Other: Teaching and speaking; Research: PI, Speakers Bureau; CTI: Ended employment in the past 24 months; MorphoSys/Constellation: Consultancy, Honoraria, Other: Research: PI, Research Funding, Speakers Bureau; GSK: Consultancy, Honoraria, Other: Teaching and speaking; Research: PI, Research Funding, Speakers Bureau; Novartis: Consultancy, Honoraria, Other: Teaching and speaking; Research: PI, Research Funding, Speakers Bureau. Della Porta: Bristol Myers Squibb: Consultancy.