Single-Cell Analysis Reveals Molecular Mechanisms of NRAS-Mediated Leukemia Stem Cell Self-Renewal in a Murine Model of AML

Acute myeloid leukemia (AML) is a lethal malignancy because patients who initially respond to chemotherapy eventually relapse with treatment-resistant disease. Leukemia stem cells (LSCs) reestablish the disease by self-renewal: the ability of a stem cell to reproduce itself and give rise to progeny. LSC self-renewal is therefore critical to relapse. Most anticancer therapies are designed to inhibit proliferation. Yet, the mechanisms that direct hematopoietic stem cell (HSC) proliferation are distinct from the mechanisms that allow HSCs to self-renew (Li et al. Nature 2013). Consequently, targeting proliferation may explain the failure of traditional chemotherapy to eradicate this disease. To study leukemia self-renewal, we use a manipulatable, transgenic mouse model of AML with an Mll-AF9 fusion and a tetracycline repressible, activated NRAS (NRAS^G12V, Kim et al. Blood 2009). Doxycycline abolishes NRAS^G12V expression leading to leukemia remission. We demonstrated that expression of NRAS^G12V is required for self-renewal in this AML model and that NRAS^G12V-mediated signaling is distinct among leukemic subsets (Sachs et al. Blood 2014). We hypothesize that NRAS-activated pathways required for LSC self-renewal are limited to a subpopulation of cells with the LSC immunophenotype. Defining the mechanisms of self-renewal has been a challenge because cancer cells are highly heterogeneous and because disengaging proliferation from self-renewal can be difficult experimentally. To overcome these obstacles, we use single-cell technologies (single-cell, whole transcriptome, RNA sequencing and mass cytometry, CyTOF) to define the signaling and transcriptional profiles of individual cells. We performed single-cell RNA sequencing on unsorted leukemia cells and on a sorted, LSC-enriched population. The single-cell transcriptional profile of LSCs was distinct from the bulk population (Fig. 1A). The 100 most differentially expressed genes between these groups are involved in hematopoietic cell fate and differentiation, confirming the biological validity of this technique. Next, we sought to identify an NRAS^G12V-mediated self-renewing subpopulation among the LSCs. Unsupervised, two-dimensional, hierarchical clustering of LSC single-cell data identified three discrete subpopulations among the LSCs, each expressing a unique gene expression profile (Fig. 1B). Comparing the single-cell transcriptional profiles of NRAS^G12V-expressing LSCs to those of LSCs treated with doxycycline to extinguish NRAS^G12V ("RAS-On" and "RAS-Off" LSCs) revealed that two of the three LSC-expression profiles seen in RAS-on cells (Groups 1 and 3, Fig. 1B) are lost when NRAS^G12V is withdrawn (Fig. 1C). These data suggest that these two profiles (Groups 1 and 3) are NRAS^G12V-dependent, consistent with an earlier report that activated NRAS exerts bimodal effects on HSCs (Li et al., Nature 2013). Gene set enrichment analysis of these profiles, modified for single-cell data, revealed that Group 1 preferentially expresses genes associated with leukemia self-renewal. On the basis of this gene expression data, we identfied cell surface markers (CD36 and CD69) that delineate the two NRAS^G12V-responsive LSC-subpopulations (Groups 1 and 3). We sorted LSCs based on CD36 and CD69 status and found that CD36^-CD69⁺ LSCs (consistent with Group 1 gene expression) harbor nearly all of the colony-forming capacity of the LSCs, forming an average of 13 colonies versus 0.33 colonies for CD36+CD69- LSCs (Group 3) and versus 0.11 colonies for non-LSCs (per 10,000 cells plated, p < 0.00001 for each comparison). We have previously shown that colony-forming capacity is an accurate surrogate for in vivo leukemia reconstituting ability and self-renewal in our model (Sachs, et al. Blood 2014). These experiments characterize the NRAS^G12V-mediated self-renewal transcriptional signature and suggest that single-cell RNA sequencing data may be an effective tool for delineating the self-renewing subpopluation among immunophenotypically-defined LSCs. Using mass cytometry to query the activation status of signaling pathways simulteneously with multiple immunophenotypic markers, we show that Ki67^Low LSCs (the putative self-renewing LSCs) preferentially express increased levels of b-catenin and Myc. These data implicating AML self-renewal pathways can provide precise molecular targets for treating this deadly disease.