S125 PATHWAY AND GENE SET OVERDISPERSION ANALYSIS OF SINGLE CELL TRANSCRIPTOMIC DATA FROM PATIENTS WITH CHRONIC MYELOID LEUKAEMIA HIGHLIGHTS TRANSCRIPTIONAL HETEROGENEITY IN CD34+CD38‐ CELLS

Background: Chronic myeloid leukaemia (CML) is a malignancy arising from the haematopoietic stem cell (HSC). The 9;22 chromosomal translocation is the hallmark of the disease, giving rise to an aberrant fusion gene, BCR‐ABL, that initiates and maintains the cancer through production of a constitutively active tyrosine kinase. Treatment with imatinib mesylate, a tyrosine kinase inhibitor (TKI) that selectively inhibits BCR‐ABL and induces apoptosis of leukaemic cells, successfully induces remissions in the majority of patients. Despite this success, the disease frequently relapses on withdrawal of TKI, owing to the persistence of a reversibly quiescent, apoptosis‐resistant population of leukaemic stem cells (LSCs). Identification of deregulated pathway signalling distinguishing this subpopulation of LSCs is an important step to characterising and targeting this subset to achieve complete eradication of the disease. Aims: The primary aim of the project was to apply PAGODA (Pathway and Gene Set Overdispersion Analysis) to the publically available CML single cell RNA‐sequencing (scRNA‐seq) data to identify clinically relevant populations of LSCs and enrichment of pathways that could be attributed to LSC persistence. Methods: scRNA‐seq data from 1749 haematopoietic stem cells (HSCs) isolated from the bone marrow of five healthy donors, 18 CML patients at diagnosis and 20 CML patients on TKI was analysed. PAGODA was applied to these data to reveal transcriptional heterogeneity at the pathway level in a single cell transcriptional dataset. PAGODA condenses individual pathways and annotated gene sets into “aspects” of heterogeneity, to better delineate transcriptionally‐distinct cell subpopulations, and also allows for the discovery of de novogene sets with correlated expression within and between cells. Hierarchical clustering was applied to these aspects of heterogeneity and heatmaps of the first principle component of each aspect constructed to visualise pathway level transcriptional variability and transcriptionally distinct subsets in this refined HSC population. Results: Hierarchical clustering of cells from normal bone marrow and CML at diagnosis revealed the transcriptional profile distinguishing normal HSCs, BCR‐ABL+ leukaemic HSCs and BCR‐ABL‐ leukaemic HSCs. This also defined subsets of cells with a divergent transcriptional profile, including reduced expression of G2/M phase mitotic pathways, which may be consistent with a quiescent phenotype. Adjusting for cell cycle eliminated the distinction between BCR‐ABL+ and BCR‐ABL‐ cells whilst preserving the dichotomy between cells from normal and leukaemic marrow, suggesting an important role for extrinsic cell signaling in regulating transcription in this HSC population. Inclusion of TKI‐treated cells in the PAGODA analysis demonstrated persistence of this proposed quiescent subset following treatment. Summary/Conclusion: Aggregating pathway level information into “aspects” of heterogeneity using PAGODA, and applying hierarchical clustering has given new insights into the differing transcriptional profiles of leukaemic HSCs, and revealed a transcriptionally distinct subsets of cells that may represent a reversibly quiescent, apoptosis‐resistant cell population consistent with the disease‐maintaining leukaemic stem cell. This corroborates previous evidence and provides further support for an important role of extrinsic cell signaling in modifying the transcriptional state of HSCs.