Abstract
Acute myeloid leukemia (AML) is a heterogeneous disease with functionally diverse cells. While primitive leukemia cells are thought to be responsible for clonal expansion, other cell types may play roles in immune evasion and paracrine signaling. To analyze the complex AML ecosystem, we developed a technology for high throughput single-cell RNA-sequencing (scRNA-seq) combined with single-cell genotyping to capture mutations in cancer driver genes. We used this technology to parse normal and malignant hematopoietic systems.
We profiled 38,410 cells from bone marrow (BM) aspirates from five healthy donors and 16 AML patients that span different WHO subtypes and cytogenetic abnormalities. Within the normal donors, we identified 15 diverse hematopoietic cell types demarcated by established markers such as CD34 (HSC/Progenitors), CD14 (monocytes) and CD3 (T-cells), confirming expected differentiation trajectories. To systematically distinguish between malignant and normal cell types within tumors, we developed a machine learning classifier that integrated scRNA-seq and single-cell genotyping data. Malignant cells were classified into six types: HSC-like, progenitor-like, granulocyte macrophage progenitor (GMP)-like, promonocyte-like, monocyte-like and dendritic-like cells. Each cell type was represented by at least 1,000 cells and identified in at least ten patients. To assess the significance of these six malignant cell types, we estimated their abundance in an independent cohort of 179 AMLs that were analyzed by bulk RNA-seq (TCGA). We found that the cell type composition of a tumor closely correlates to its underlying genetic lesions. For example, RUNX1-RUNX1T1 translocations are associated with GMP-like cells and TP53 mutations with undifferentiated cells (P < 0.001). NPM1+FLT3-ITD mutated tumors are enriched for more primitive cells compared to NPM1+FLT3-TKD mutants, which may relate to the worse outcomes of patients with FLT3-ITD mutations. The correspondence between genetic lesions and tumor cell type composition can guide strategies for genotype-specific therapies that target appropriate cellular states.
Further investigation of primitive cells showed that gene expression programs associated with stemness (e.g. EGR1, MSI2) are mutually exclusive with myeloid priming (e.g. MPO, ELANE) in primitive cells of healthy donors. In contrast, these programs are often co-expressed within the same individual AML cells. When we applied our single cell-derived gene signatures to the TCGA dataset, stratification of these bulk expression profiles showed that patients with HSC-like progenitors had significantly poorer outcomes than patients with GMP-like progenitors (P < 0.0001). Aberrant co-expression of stemness and myeloid programs may underlie simultaneous self-renewal and proliferation, and expression of myeloid priming factors may provide a therapeutic window to target primitive AML cells while sparing normal HSCs.
Examination of T-cells in our single-cell dataset showed that AML patients have fewer CD8+ cytotoxic T-lymphocytes within the CD3+ T-cell compartment compared to healthy controls, which was validated by immunohistochemistry on BM core biopsies (69% in healthy controls vs. 54% in AML, P < 0.05). We observed increased CD25+FOXP3+ T-regulatory cells in AML patients (1.2% in healthy controls vs. 3.6% in AML, P < 0.001), indicating an immunosuppressive tumor environment. To investigate mechanisms of immunosuppression, we used a T-cell activation bioassay that reports Nuclear Factor of Activated T-cells (NFAT). We compared the immunosuppressive function of different AML cell types, and found that CD14+ monocyte-like cells most effectively inhibit T-cell activation (P < 0.0001). The malignant status of these differentiated AML cells was confirmed by genotyping, and they express multiple factors associated with immunosuppression and T-cell engagement, including TIM-3 (HAVCR2), HVEM (TNFRSF14), CD155 (PVR) and HLA-DR. These results suggest that AMLs can differentiate into monocyte-like cells that suppress T-cell activation.
In conclusion, we use novel technologies to parse heterogeneous cell states within the AML ecosystem. Our findings nominate strategies for precision therapies targeting AML progenitors or immunosuppressive functions of their differentiated progeny.
Disclosures
Pozdnyakova: Promedior, Inc.: Consultancy. Lane:N-of-one: Consultancy; Stemline Therapeutics: Research Funding.
Comprehensive mapping of immune perturbations associated with severe COVID-19