Single Cell Correlation Analysis: A Novel Method to Analyze Single Cell RNA Sequencing Data Identifies a Self-Renewing Subpopulation of Human Acute Myeloid Leukemia Stem Cells

Introduction Acute myeloid leukemia (AML) is an aggressive hematopoietic malignancy. Despite early response to standard chemotherapy, 60-70% of patients relapse with treatment resistant disease and die of relapsed disease. Leukemia stem cells (LSCs), the subpopulation of leukemia cells with self-renewal capacity, can recapitulate leukemia to cause relapse. Ongoing work to define therapeutic targets that can block self-renewal would prevent relapse and disease progression in AML. Prior work has established that LSCs express a distinct gene expression profile. We sought to define the gene expression profile of self-renewal at the single-cell level because a precise understanding of the transcriptional basis for self-renewal could provide more effective therapeutic targets. Our group previously demonstrated that activated NRAS (NRAS G12V) mediates self-renewal in the LSC-enriched subpopulation of AML in the Mll-AF9/NRAS G12V mouse model of AML (Sachs 2014). Single cell RNA-sequencing (scRNA-seq) of the LSC-enriched subpopulation identified three distinct transcriptional profiles. Two genes that encode cell surface markers, Cd36 and Cd69, delineate these distinct profiles. Using in vivo leukemia engraftment assays, we found that only the CD36 -CD69 + LSC subgroup repopulates leukemia while the CD36 +CD69 - LSC subgroup does not, demonstrating that self-renewal capacity is limited to the CD36 -CD69 + subset of the LSC-enriched compartment (Sachs 2020). In contrast, the CD36 +CD69 - LSCs were highly proliferative while the CD36 -CD69 + LSCs were poorly proliferative. These experiments show that self-renewal and rapid proliferation are mutually exclusive features of LSCs, as has previously been shown in normal hematopoietic stem cells (Li 2013). Methods and Results We investigated whether primary human AML stem cells express the murine-derived single-cell self-renewal and proliferation gene expression profiles at the single-cell level. Current computational methods to compare single-cell gene expression profiles use arbitrary similarity scores and do not assess the statistical significance of their findings. Therefore, we developed a new computational method called Single cell Correlation Analysis (SCA). SCA is designed to quantitate the expression of a predefined profile in single-cell RNA sequencing data. The quantification of the expression profiles is assessed on a per-cell basis in an unsupervised manner, meaning that SCA does not rely on any pre-sorting or