Graphical Abstract Highlights d Technology for high-throughput single-cell RNA sequencing and genotyping d Variable cell-type composition of AML correlates to genetics and outcome d Primitive AML cells aberrantly co-express stemness and myeloid priming genes d Differentiated AML cells express immunomodulatory factors and suppress T cells In BriefA combination of transcriptomics and mutational analyses in single cells from acute myeloid leukemia patients reveals the existence of distinct functional subsets and their associated drivers. SUMMARYAcute myeloid leukemia (AML) is a heterogeneous disease that resides within a complex microenvironment, complicating efforts to understand how different cell types contribute to disease progression. We combined single-cell RNA sequencing and genotyping to profile 38,410 cells from 40 bone marrow aspirates, including 16 AML patients and five healthy donors. We then applied a machine learning classifier to distinguish a spectrum of malignant cell types whose abundances varied between patients and between subclones in the same tumor. Cell type compositions correlated with prototypic genetic lesions, including an association of FLT3-ITD with abundant progenitor-like cells. Primitive AML cells exhibited dysregulated transcriptional programs with co-expression of stemness and myeloid priming genes and had prognostic significance. Differentiated monocyte-like AML cells expressed diverse immunomodulatory genes and suppressed T cell activity in vitro. In conclusion, we provide single-cell technologies and an atlas of AML cell states, regulators, and markers with implications for precision medicine and immune therapies.

Reconstructing lineage relationships in complex tissues can reveal mechanisms underlying development and disease. Recent methods combine single-cell transcriptomics with mitochondrial DNA variant detection to establish lineage relationships in primary human cells, but are not scalable to interrogate complex tissues. To overcome this limitation, here we develop a technology for highconfidence detection of mitochondrial mutations from high-throughput single-cell RNA-sequencing. We use the new method to identify skewed immune cell expansions in primary human clonal hematopoiesis. Main textSingle-cell RNA-sequencing (scRNA-seq) enables the unbiased assessment of cell states in health and disease 1,2 . Combined acquisition of cell state and genetic information can provide additional insight, such as targeted enrichment of cancer driver mutations from single-cell transcriptomes 3,4 . Separately, combining scRNA-seq with genetic cell barcodes is a powerful method to reveal clonal relationships and evolutionary dynamics of cells within organisms 5,6 . However, this has largely been limited to experimental model systems that can be genetically manipulated to insert cell barcodes. To infer clonal dynamics in primary human cells, recent methods have detected and utilized mitochondrial DNA (mtDNA) mutations as naturally occurring genetic cell barcodes [7][8][9] . The combination of scRNA-seq with mtDNA mutation detection can inform clonal relationships with high confidence, but is currently restricted to expensive, low-throughput, full-length transcript sequencing technologies like SmartSeq2 7,10 . To enable the reconstruction of clonal relationships in complex human tissues, we developed a method that captures genetic variants from high-throughput scRNA-seq platforms: MAESTER, or Mitochondrial Alteration Enrichment from Singlecell Transcriptomes to Establish Relatedness (Figure 1A). MAESTER is compatible with the most common high-throughput scRNA-seq platforms, including 10x Genomics 3' protocols, Seq-Well S 3 , and Drop-seq (Supplemental Figures 1-3) 11,12 . An intermediate step in each of these platforms yields full length cDNA transcripts, from which we enrich all 15 mitochondrial transcripts using pools of primers, while maintaining cell-identifying barcodes (Figure 1B, Supplemental Figure 4). Standard next-generation sequencing with 250 bp reads is then used to obtain the sequence of the amplified mitochondrial transcripts (Figure 1A). We developed a computational toolkit to call mtDNA variants from MAESTER data, the Mitochondrial Alteration Enrichment and Genome Analysis Toolkit (maegatk, Supplemental Figure 5, Methods). Building on previous tools that we developed 8 for mtDNA variant detection from single-cell ATAC or SmartSeq2, maegatk specifically handles technical biases implicit in high-throughput transcriptomic libraries. Critically, maegatk leverages unique molecular identifiers (UMIs) to collapse multiple sequencing reads of the same starting transcript, creating a consensus call for every