Single-Cell Mass Cytometry Characterizes Phenotypic and Functional Heterogeneity in Acute Myeloid Leukemia at Diagnosis, in Remission and Relapse

Patients with acute myeloid leukemia (AML) continue facing poor long-term survival due to high relapse rate. Persistence of dormant self-renewing leukemia stem/progenitor cells (LSPCs) has been implicated as driver of subsequent relapse, and stem cell gene signatures are associated with poor outcome (Shlush et al. Nature 2017; Ng et al. Nature 2016; Eppert et al Nat Med 2011). Identification of the unique phenotypes and functional proteins in LSCs surviving induction therapy may aid in understanding the mechanisms of chemoresistance and provide novel therapeutic targets in the residual leukemia clones.

In this study, we have developed and optimized a comprehensive single-cell mass cytometry (CyTOF) panel, including 36 markers to define LSPCs, with the goal to identify and characterize expression of multiple intracellular signaling pathways and anti-apoptotic proteins in residual AML cells. The validated CyTOF panel was applied in 21 samples collected from 7 AML patients at diagnosis, in remission and at relapse and 5 healthy donors. Data were analyzed using SPADE (Qiu et al, Nat Biotechnol 2011) or Cytofkit (Chen et al. PLoS Comput Biol 2016) tools.

We first generated SPADE trees for all diagnostic samples (n=7, Fig 1A), annotating 7 distinct cell populations based on the median expression of selected surface markers as shown in the heatmap (Fig 1B). From these annotations, populations A1, A2 and A6 were positive for CD34 expression, with A6 representing phenotypically the most primitive fraction CD34^hiCD38^low population (frequency range 0.04%-17.32%). Fractions A1 and A2 expressed more committed myeloid markers, positive for CD135 and CD33 progenitor markers and differentiation markers including CD15, CD11b, and CD7 (Fig 1B). Variability was observed in terms of cell composition, and non-stem fractions A3-A5 were abundant populations in all AML samples except AML2. We further studied the activation/expression of functional proteins in these populations and found that pro-survival BCL-2 protein was highly expressed in the primitive A6 population across AML samples (median intensity 9.0 ± 5.3 in A6 vs 3.5 ± 3.9 in other populations). Variable p-AKT activation was observed in both A6 (4.9 ± 4.2) and differentiated A3-A5 populations (4.5 ± 3.6).

We next examined how multi-parametric CyTOF analysis will aid in characterization of MRD populations by comparing samples from 7 patients collected at the time of diagnosis, in remission and at relapse. Using the Cytofkit bioconductor analysis and FlowSOM algorithm, we identified distinct patterns of relapse (Fig 1C). In AML 1-3, major populations were markedly reduced by induction chemotherapy, but residual cells re-grew and contributed to relapse. In AML 4-7, the major populations at diagnosis were eliminated by the therapy, but minor (or undetectable) populations at diagnosis progressed over treatment and represented the bulk of leukemia upon relapse. This finding is consistent with genomic studies that relapse may originate from either the founding clones or subclones that acquire additional mutations (Ding et al. Nature 2012). In AML#3, a major population (cluster 7, CD34⁻CD38⁺CD123⁺CD64⁺HLA-DR⁺CD99⁺, 75.6%) present at diagnosis was identified as persisting in remission at 2.5% by CyTOF analysis but not by conventional MRD flow cytometry and gave rise to the overt leukemia (78.1%) at relapse (Fig 1D). In this patient whose AML harbored mutation in negative MAPK regulator phosphatase PTPN11, persistent AML cells expressed BCL-2, MCL-1, and p-p38MAPK (Fig 1E), consistent with dominant activation of MAPK signaling and anti-apoptotic proteins. We found highly enriched BCL-2 expression and p38MAPK activation in relapse-driving clones in AML5 and AML7, and in diagnostic clone in AML6 (Fig 1F).

In summary, using CyTOF, SPADE and Cytofkit analysis tools, we characterized LSPC-specific intracellular signaling pathways in AML samples at diagnosis, in remission and at the time of relapse. Distinct populations were identified to contribute to relapse, indicating that use of additional targeted therapies such as BCL-2 inhibitors may be instrumental post remission to prevent relapse. In conclusion, analysis of the multi-parametric single cell CyTOF mass cytometry may aid in understanding clonal evolution during chemotherapy and identify potential therapeutic targets in individual patients.

View largeDownload slide

View largeDownload slide

Close modal