Systematic Bias Introduced by Ficoll-Based Isolation in AML Sample Composition and Downstream Analyses

Acute myeloid leukemia (AML) is a heterogeneous malignancy whose characterization relies on immunophenotyping and molecular profiling. While red blood cell lysis is the recommended method for leukocyte isolation in clinical diagnostics, Ficoll-based density gradient centrifugation remains widely used in research and biobanking. Here, we systematically compare these two isolation approaches using paired primary AML samples. We show that Ficoll processing consistently enriches lymphocytes and AML blasts while depleting granulocytes. The increased T-cell content introduced by Ficoll impaired AML engraftment in immunodeficient NSG mice, as residual T cells in the graft triggered graft-versus-host disease, precluding reliable evaluation of leukemia-initiating activity. Although Ficoll had minimal impact on ex-vivo AML blast expansion or chemotherapy response, RNA sequencing identified 1,136 differentially expressed genes between isolation methods, with Ficoll samples exhibiting enrichment of proliferation-associated pathways and leukemic stem cell (LSC) transcriptional programs. Immunogenomic deconvolution further demonstrated that Ficoll artificially inflates estimated CD8⁺ T-cell and monocyte abundances while reducing inferred neutrophil frequencies. Mutation calling from RNA-seq data revealed substantial discrepancies between methods, including failure to detect a clinically relevant DNMT3A R882 mutation in a Ficoll-processed sample. These findings identify leukocyte isolation as a critical pre-analytical variable in AML research and support the systematic use of red blood cell lysis to preserve cellular diversity and avoid unpredictable biases introduced by Ficoll-based isolation.