Single Cell Transcriptome Analysis Reveals Changing Levels and Distributions of Stemness across Disease States in Pediatric AML

The occurrence of relapse after achieving remission is one of the main obstacles in curing paediatric AML. This phenomenon may be linked to the heterogeneity found within tumours, and how the clonal composition of tumours can change over time across disease states (Ding et al. 2012).

To begin to explore profile changes in clonal compositions across various stages of AML, single cell RNA sequencing was performed using the 10X 3' Chromium Single Cell platform on two matched primary, remission and relapse paediatric AML samples. Samples were extracted from the bone marrow or peripheral blood of patients enrolled in the Children's Oncology Group study AAML0531 (Gamis et al. 2015).

The gene expression profiles of individual cells were correlated to the LSC+ (leukemic stem cell) reference expression profile (Ng et al. 2016) to quantify the degree of similarity of each cell to a functionally defined stem cell population. Consensus clustering was performed to identify subgroups within disease states. The level of correlation to LSC+ for each cell was then mapped onto the subgroups to investigate patterns of distribution of "stemness" within cell populations.

A greater proportion of cells with a correlation value to LSC+ of greater than 0.3 was observed at primary and relapse stages compared to remission. At primary and relapse, subgroups showed substantial inter- and intra-cellular heterogeneity in their gene expression profiles, though the levels of correlation to LSC+ were similar across all subgroups. In one patient, cells within the relapse phase displayed an increase in CD34 expression (a primitive hematopoietic cell marker) compared to cells at the primary stage, consistent with the possibility of an increase in stem-like populations at relapse. This increase in CD34 was accompanied by decreases in the expression of genes encoding for major histocompatibility complex molecules, HLA-DRA and HLA-DPB1, at relapse versus primary.

Certain subpopulations at remission showed a distinct lack of cells with high correlation to LSC+; these subgroups had a much lower correlation to the average expression levels of cells at primary and relapse stages (r~ 0.6 versus r~0.93 for remission subgroups with higher stemness levels). The low correlation observed for subgroups at remission which lacked a stem-like profile was driven by high expression of HBA1, HBA2 and HBB (>6 versus <1 on log2(UMI) scale). These genes were previously shown to be highly expressed within myelopoietic and erythropoietic cells (Huisman et al. 1996), which is consistent with the types of cells that are expected to be found in the normal bone marrow (http://www.proteinatlas.org). Within the same patient with increased CD34 and decreased MHC II gene expression at relapse versus primary, the proportion of cells displaying erythropoietic gene expression profiles showed a distinct contraction from primary to relapse stages.

The results from this study demonstrated, at the single cell level, the dynamic changes in the degree of stemness and the expansion and contraction of subpopulations across disease states in pediatric AML. These observations are consistent with the evolving genetic and cellular landscapes from diagnosis to relapse as shown in previous studies (Farrar et al. 2016; Jan et al. 2012). As well, we were able to parse out cells at remission that had a more malignant-like gene expression profile from cells expected to be found within the milieu of normal blood and bone marrow. This study thus provides a first step towards understanding patterns of change in the composition of tumour cell populations over time.