Mass Cytometry of Acute Myeloid Leukemia Captures Early Therapy Response in Rare Cell Subsets

Introduction: The plasticity and stemness of acute myeloid leukemia (AML) cells are thought to be driving factors in disease aggressiveness and poor patient survival. These factors also contribute to the challenge of designing analytical cytometry panels to study AML over time during treatment. A high content single cell approach was designed to pinpoint rare populations of AML cells present prior to treatment that emerge and dominate following therapy resistance or disease relapse. Twenty-seven diagnostic and differentiation markers were measured on AML cells in order to track rare cell subsets pre- and post- treatment, measure kinetics of early therapy response, and identify any novel leukemic cell populations.

Methods: AML patients undergoing induction chemotherapy were identified and consented for this study according to a local Institutional Review Board-approved protocol. Samples of peripheral blood and bone marrow were collected and cryopreserved after mononuclear cell separation. Thawed samples were stained with a mass cytometry panel to quantify 27 AML biomarkers on every cell. From each patient, up to 12 samples of blood and marrow were obtained before, during, and after induction chemotherapy (n = 46 total samples; 14 marrow and 32 blood from 5 individual donors). Stained cells were analyzed by mass cytometry (CyTOF). Dimensionality reduction and visualization was performed using viSNE. Leukemic cell subsets and differentiated cells from healthy donors were characterized in terms of abundance in each sample, 27-marker expression phenotype, heterogeneity of marker expression within the subset, and differentiation status in each sample. A stem/progenitor index was created to quantify the phenotypic distance of cells in a population from the CD34⁺stem/progenitor cells.

Results: In unsupervised viSNE analysis of 27-marker cellular phenotype, leukemic blasts formed phenotypically distinct groups of cells, were CD45^lo, and had an expression profile that closely matched the diagnostic fluorescent flow cytometry immunophenotype. The 27-marker panel grouped cells into 11 major populations: CD34⁺ hematopoietic stem/progenitor cells (HSPCs), 5 differentiated non-malignant populations (myeloid, CD4⁺ T, CD8⁺ T, B, and NK cells), and 3 major populations of leukemia cells. Remission was apparent on viSNE when no more than 5% of leukemic blasts remained. The remaining cells were instead part of the non-malignant populations. In poor outcome cases, at-diagnosis bone marrow leukemia cells were initially close in 27-dimenional phenotype to the non-malignant stem/progenitor cells (i.e. higher stem/progenitor index). In a case of primary refractory disease, six therapy resistant cell (TRC) populations became dominant within the leukemic blast area. These same TRC populations were present but initially rare and constituted only 0.6% to 2% of total pre-treatment AML blasts at diagnosis. TRC populations displayed increased per-cell expression of markers associated with stemness and leukemia initiating ability, including CD34, CD38, and CXCR4/CD184 (respective increases of 0.7, 0.9, and 0.6 fold on the log-like asinh₁₅ scale). In contrast, expression of CD34, CD38, and major cell type identity markers (e.g. HLA-DR, CD4, CD19) did not significantly change over time on non-AML cells (all <0.2 fold).

Conclusion: These results demonstrate the ability of high-dimensional mass cytometry to reveal, characterize, and compare rare and highly plastic cell populations over time in primary human tissue samples. Use of 27 markers meant that cells that dramatically changed expression of a few markers could still be matched to phenotypically similar cells using the other 20+ markers. Such phenotypic similarities were captured well by viSNE computational analysis. This approach offers greatly expanded longitudinal monitoring of AML while 1) effectively capturing differentiation along with cellular abundance, 2) identifying biomarkers of therapy resistance for cell sorting and targeting, and 3) enabling the single cell analysis of signaling networks in concert with critical markers of cell identity.