Similarities and differences of bone marrow and peripheral blood samples from acute myeloid leukemia patients in terms of cellular heterogeneity and ex‐vivo drug sensitivity

Abstract

AbstractBackgroundBone marrow (BM) evaluation is the de facto standard for diagnosis, molecular analysis, risk stratification, and therapy response assessment in acute myeloid leukemia (AML), but in patients with a high number of circulating blast cells, the peripheral blood (PB) sample could provide similar information as BM. However, there is no large‐scale molecular study comparing the two specimens in terms of their gene expression profiles, cellular heterogeneities, and ex‐vivo drug sensitivity.MethodologyWe used (i) the BEAT‐AML cohort each with detailed molecular data; (ii) cell‐type deconvolution to estimate leukemic and immune cell proportions between specimen types; (iii) differential expression (DE) and drug‐cell type association analysis; and (iv) logistic regression models to assess the association between induction therapy response, cell‐type composition and first‐line drug treatment.ResultsResults: We identified 207 patients having BM and 116 patients having PB samples. There was a total of 1271 DE genes (false discovery rate < 0.05) between BM and PB; the top enriched pathways in terms of DE genes belong to the immune system pathways. Aggregated ex‐vivo drug response profiles from the two specimens were largely similar, as were the cellular components, except for the GMP‐like cell type (17% in BM vs. 5% in PB, p‐value = 2 × 10−7). Among the specimen‐specific results, the GMP‐like subtype was associated with multiple drug resistance in BM and the ProMono‐like subtype in PB. Several cell types were associated with the response to induction therapy, but the impact of specimen type on the interaction of cell type and cytarabine‐associated induction therapy was not statistically significant for most cell types.ResultsConclusions: Even though there are molecular and cellular differences between BM and PB samples, they show many similarities in ex‐vivo drug response profiles, indicating the clinical utility of the substantially less‐invasive PB samples.