One lipid to predict them all Free

In this issue of Blood, O’Brien et al¹ demonstrates that distinct gene mutations in acute myeloid leukemia (AML) result in unique plasma metabolite and lipid signatures. A total of 231 diagnostic AML plasma samples isolated prior intensive chemotherapy treatments were investigated using unbiased metabolic and lipidomic analysis. Although metabolites were associated with the mutational landscape of AML cells, lipids were highly associated with refractory status. Among all lipids, sphingomyelin (d44:1) [SM(d44:1)] was sufficient to predict overall survival.

Lipid metabolism differs between healthy hematopoietic stem and progenitor cells (HSPCs), leukemic stem/initiating cells (LSCs/LICs), and AML blasts, and plays a critical role in cancer progression.^2,3 Patients with AML show distinct lipid profiles, both in cells and plasma, compared to healthy individuals, with specific lipid species correlating with cytogenetic and prognostic subtypes.^4,5 

Mutations in AML can profoundly reprogram lipid metabolism and create specific metabolic dependencies. For instance, in IDH1-mutated but not IDH2-mutated AML, the lipid synthesis enzyme acetyl coenzyme A carboxylase 1 (ACC1) has been identified as a synthetic lethal target.⁶FLT3 mutations drive fatty acid biosynthesis and regulate polyunsaturated fatty acids (PUFAs) via C/EBPα, making leukemic cells susceptible to lipid peroxidation when FLT3 or C/EBPα is inhibited. Inhibition of stearoyl-CoA desaturase (SCD) increases oxidative lipid stress and activates ferroptosis, highlighting that the role of lipid intermediates depends on the genetic context.⁷ Additionally, mannose metabolism via mannose-6-phosphate isomerase (MPI) has been implicated in chemoresistance; loss of MPI promotes lipid accumulation and sensitizes FLT3-mutated AML cells to ferroptosis and FLT3 inhibition.⁸ Altogether, these findings illustrate how driver mutations can rewire AML’s lipidome, revealing metabolic vulnerabilities and biomarker value that may be leveraged therapeutically.^9,10 

Here, O’Brien et al, identified by mass spectrometry 177 metabolites and 1988 lipids across all samples. Five clinical features showed significant association to specific metabolites (eg, gender, age, blast count) and 11 clinical features showed significant association to specific lipids (eg, LDH, LCV, WBC blast count). Leveraging their comprehensive unbiased data, machine learning models were trained to identify refractory AML and generate a model predictive of patients’ survival. The circulating lipid signature outperformed the metabolic signature (AUC = 0.94 vs AUC = 0.56) and holds promise as prospective biomarker for patients’ outcomes. The gradient booster classifier revealed that SM(d44:1) has the greatest lipid-marker to predict survival across AML in their discovery and validation cohorts, with high SM(d44:1) being significantly associated with survival.

The association between plasma lipids and patient genotypes is intriguing and advocates for deeper investigation. One possibility is that plasma lipid changes mirror the lipid distribution within leukemic cells; alternatively, they may originate from extracellular vesicles, reflect cellular breakdown from leukemic cells or the surrounding microenvironment, or even represent a systemic response of the microenvironment to leukemia. These mechanistic possibilities raise further questions: does the lipid signature correlate with established classification systems such as French-American-British (FAB), or with morphological and immunophenotypic features of the disease? Clarifying these relationships will contribute to understanding the biological basis of the observed lipid alterations. Notably, these findings point toward the relevance of sphingomyelin biology in AML and its potential as a source of circulating biomarkers and treatment-response stratifiers, reinforcing the need to identify the upstream drivers of these lipidomic patterns. Although SM(d44:1) emerged as a top predictive lipid for primary refractory disease, its specific functional role and regulatory mechanisms remain unclear. Furthermore, although the lipid signature demonstrated strong predictive power for short-term outcomes, its relevance for longer-term end points like relapse or survival remains to be validated through longitudinal studies. Hence, to support continued exploration, the authors have developed an interactive platform (https://plasmalipids.shinyapps.io/AML_circulating_metabolites), which serves as a valuable resource for the research community to further investigate and validate these lipidomic signatures.

Conflict-of-interest disclosure: The author declares no competing financial interests.