Background:An expanding body of evidence indicates that targeting cell metabolism is a promising therapeutic strategy in acute myeloid leukemia (AML). Pyrimidine biosynthesis, oxidative phosphorylation, amino acid and arginine metabolism are among the most relevant actionable pathways (Sykes et al. Cell 2016; Molina et al. Nat Med 2018; Jones et al. Cancer Cell 2018; Mussai et al. Blood 2015). However, few studies identified specific metabolic vulnerabilites (Fenouille et al. Nat Med 2017; Gallipoli et al. Blood 2018) due to the lack of a comprehensive characterization of AML cell metabolism.Aims:The study aimed to stratify AML patients based on their metabolic landscape and integrate their genomic and metabolic profiles.Methods:Genomic data of AML patients were obtained by whole exome sequencing (n = 165) or targeted NGS (n = 18). Of them, 119 AML were evaluated for serum metabolites by nuclear magnetic resonance (NMR), along with 145 healthy donors. Intracellular metabolites of AML bone marrow cells (35 CD34+ and 15 CD33+CD14− samples) and controls (21 cord blood CD34+ and 21 peripheral blood CD33+ samples) were analyzed by mass spectrometry (Metabolon).Results:Unsupervised hierarchical clustering of AML based on the intracellular metabolic landscape clearly defined 3 clusters (C, Fig.A). C1 was characterized by high metabolite concentration, C2 by low metabolite levels and C3 by a more complex metabolic profile. By integrating genomic data into the metabolic classification we obtained a significant clusterization (p = 0.029). C1 was enriched for NPM1‐mutated (mut) AML (50%), C2 for cases with altered chromatin/splicing genes or inv(3) (37% and 13%, respectively), C3 for TP53‐mut/aneuploid AML (35%). Moreover, C3 included all core binding factor AML. By training a Gaussian Process classifier on NMR data, we obtained a significant separation of C1 and C3 also at serum level (Fig. B,C). In particular, NPM1‐mut AML displayed abundance of tricarboxylic cycle byproducts, while glutamine and creatinine were increased in NPM1‐wildtype (wt) cases (Fig.D,E). At intracellular level, NPM1‐mut AML were characterized by increased intermediates of purine and pyrimidine metabolism, in line with an active biosynthetic activity. Moreover, unsupervised NMR data analysis identified a clear separation (0:7 accuracy) between TP53‐wt and mut/deleted AML, the latter showing reduced threonine, alanine, glucose, lactate and glutamine serum concentration (Fig.F).At genomic level, 88% of patients carried at least one mutation in a metabolism‐related gene including enzymes (encoded by nuclear or mitchondrial DNA) and metabolic regulators, with lipid metabolism, cellular respiration (IDH1/2 mutations and oxidative phosphorylation with altered UQCRH, FASTKD3, ND and COX genes), metabolism of carbohydrates (e.g. glycosyltransferase GXYLT1, GDP‐mannose biosynthetic process involving PMM2), glucose (e.g. glycolysis‐related genes HK2, HK3, PKLR, BPGM) and nucleotides (e.g. AK2, AK5: nucleotide phosphate group transfer, NME7: synthesis of nucleoside triphosphates) being the pathways most frequently targeted by mutations.Summary/Conclusion:AML is characterized by a number of functional and genomic‐driven metabolic alterations. The integration of genomic and metabolic profiles provided a novel refined AML classification and suggested subtype‐specific metabolic targets, including nucleotide metabolic pathways and bioenergetics in NPM1‐mut and TP53‐mut/del AML, respectively.imageSupported by: EHA research fellowship award, AIRC, FP7‐NGS‐PTL, Fondazione del Monte.