Assessment of intratumor genetic heterogeneity by next-generation sequencing (NGS) is confounded by tumor purity and zygosity of mutations. Furthermore, bulk NGS cannot visualize the cell-level co-occurrence or exclusivity among multiple mutations, causing the inaccurate inference of subclonal architecture. Here, we performed single cell DNA sequencing (scDNA seq) in 82 bone marrow samples from 70 patients with acute myeloid leukemia (AML) using a novel microfluidics-based platform covering 40 amplicons in 19 AML genes (Tapestri, Mission Bio). All samples were concurrently sequenced by the bulk NGS using 295 gene exome capture sequencing. Selected cases were also analyzed by SNP array (Illuminia Omni 2.5 array) to obtain allele-specific copy number data. In total, 319,406 cells (median 3,755 cells/sample) were genotyped with median allele drop-out rate of 8.7% (population frequency inferred from commonly heterozygous SNP loci). Each amplicon was covered at a median 24x/cell. The scDNA seq detected 230 of 238 (97%) bulk NGS-confirmed mutations. RUNX1 mutations were frequently detected as homozygous mutations and concurrent SNP array analysis detected copy number neutral loss-of-heterozygosity of RUNX1 in these cases. Additionally, in cases with homozygous FLT3-ITD mutations, SNP array detected allele-specific copy number gain of mutant loci, which likely resulted in homozygous calls. scDNA seq data unambiguously resolved the single-cell level co-occurrence of driver mutations in AML such as DNMT3A/FLT3-ITD/NPM1, SRSF2/IDH2, and ASXL1/RUNX1, confirming the cooperative function of these driver mutations. On the other hand, the data also revealed mutually exclusive relationships at cellular level between IDH1/IDH2, FLT3-ITD/TKD, NRAS/KRAS, NRAS/PTPN11, and SRSF2/EZH2 mutations, which supports the functional redundancy of these genes in leukemogenesis. Inference of phylogenetic trees using SCITE algorithm (Jahn et al. Genome Biology 2016) uncovered distinct patterns of clonal evolution in AML. The majority of the analyzed cases had a linear evolution pattern where the founder mutations linearly acquired sub-clonal mutations in a step-wise manner. We also detected convergent evolution in some cases where functionally similar driver mutations were acquired in parallel. DNMT3A, IDH1, IDH2 and U2AF1 mutations were frequently detected as trunk mutations, whereas FLT3, NRAS, and NPM1 mutations were usually detected as branch mutations. Analysis of longitudinal samples from 11 patients revealed the remodeling of clonal architecture at relapse. Patients who had 2 or more major subclones at baseline had significantly worse overall survival than those with one subclone (2-year OS 13 vs. 70 months, p = 0.0493). For the first time our data provides a comprehensive landscape of driver mutations and detailed clonal evolution history in AML at the single-cell level using high-throughput scDNA seq genotype information.

Citation Format: Kiyomi Morita, Feng Wang, Katharina Jahn, Yuanqing Yan, Robert Durruthy-Durruthy, Anup Parikh, Jairo Matthews, Latasha Little, Curtis Gumbs, Jianhua Zhang, Xingzhi Song, Erika Thompson, Keyur Patel, Carlos Bueso-Ramos, Courtney DiNardo, Farhad Ravandi, Elias Jabbour, Michael Andreeff, Jorge Cortes, Marina Konopleva, Guillermo Garcia-Manero, Hagop Kantarjian, Dennis J. Eastburn, P Andrew Futreal, Niko Beerenwinkel, Koichi Takahashi. The single-cell atlas of driver mutations in acute myeloid leukemia [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2725.