Leukemic stem cell characterization in myelodysplastic syndromes and secondary acute myeloid leukemia using high dimensional mass cytometry
Approximately 35% of Myelodysplastic Syndromes (MDS) patients is at risk of transformation to acute myeloid leukemia (AML). Currently, these patients are not accurately identified; even 10-15% of the very low risk-group patients eventually progresses to AML, whereas approximately 35% of the very high risk-group patients do not. These patients do not receive optimal therapy, which affects patient outcome and quality of lives.
Characteristics of MDS hematopoietic stem and progenitor cells (HSPC) may relate to sub-types and risk-groups. Here we apply mass cytometry for comprehensive assessment of HSPC characteristics, ultimately enabling identification of MDS patients at risk of progression to AML and providing insights in MDS patho-biology.
Bone marrow samples from healthy donors (n=10), low risk (LR, n=10) and high risk (HR, n=10) MDS patients and 5 paired samples of MDS and secondary AML were collected. Mononuclear cell fractions of samples were enriched for HSPC by cell sorting (CD34+/CD117+/CD133+ cells), life-death stained and fixed. These fractions were barcoded and stained with 32 metal conjugated antibodies to assess the expression of HSPC and leukemia-associated immunophenotype (LAIP) markers by mass cytometry (CyTOF) and subjected to next generation sequencing for detection of recurrent gene mutations. For pre-processing and data analysis, unsupervised computational analyses were performed.
We used FlowSOM to group cells with similar immunophenotypes in nodes based on the expression of the 32 markers that were organized using a minimal spanning tree algorithm. Cells from healthy donor’s bone marrows were mapped to a grid of 64 nodes on which we then projected cells from MDS/sAML samples. A metaclustering step based on HSPC markers only, clustered nodes in 15 major populations with cells of a related phenotype. Unbiased marker enrichment modelling showed that these largely resembled known HSPC subsets. LR MDS showed a 1.5 fold changed abundance in common myeloid progenitor cells, whereas cells of HR MDS were present with a significantly increased abundance ( p < .05) in a branch that resembles granulocyte macrophage progenitors. Cells from AML samples were abundant in branches with a CD45dim phenotype, and significantly ( p < .05) enriched in nodes with a CD34+CD38- phenotype. Data from nodes with significant differential abundance (p< .05) over groups well separated healthy bone marrow and LR MDS from HR MDS or AML in cluster analyses. Paired analysis identified cells with a similar aberrant phenotype in individual HR MDS and matched AML samples. Samples with hallmarks of oligoclonality in masscytometry data showed concordant variant allele frequencies for recurrent gene mutations.
High-dimensional mass cytometry and computational data analyses enables characterization of HSPC subsets in MDS and identification of leukemia stem cells based on their LAIP.