Abstract IA12: Single cell multi-omics to define normal and malignant differentiation topologies

Abstract

Abstract Cancer progression, relapse and resistance are the result of an evolutionary optimization process. Vast intra-tumoral diversity provides the critical substrate for cancer to evolve and adapt to the selective pressures provided by effective therapy. Our previous work has shown that genetically distinct subpopulations compete and mold the genetic makeup of the malignancy. Additionally, we have shown that epigenetic changes in cancer may be similar to the process of genetic diversification, in which stochastic trial and error leads to rare fitness enhancing events. These studies demonstrate the need to integrate genetic, epigenetic and transcriptional information in the study of cancer evolution, specifically at the single-cell resolution – the atomic unit of somatic evolution. To enable this work, we have developed a single-cell multi-omics toolkit, and apply it to chart the evolutionary history and developmental topographies of normal and malignant cells. We demonstrated that epimutations serve as a molecular clock, and that heritable epimutation information therefore allows to infer high-resolution lineages with single-cell data, directly in patient samples. Multi-omic single-cell Integration of methylome sequencing with whole transcriptome and genotyping capture validated tree topology inferred solely on the basis of epimutation information. To examine potential lineage biases during therapy, we profiled serial samples, and identified clades of cells that were preferentially affected by targeted therapy. The single-cell integration of genetic, epigenetic and transcriptional information thus charts the lineage history of cancer and its evolution with therapy. Second, charting the transcriptomes of clonally mutated cells is challenging in the absence of surface markers that distinguish cancer clones from one another, or from admixed non-neoplastic cells. To tackle this challenge, we developed Genotyping of Transcriptomes (GoT), a technology to integrate genotyping with high-throughput droplet-based single-cell RNA sequencing. With GoT we profiled thousands of CD34+ cells from patients myeloproliferative neoplasms to study how somatic mutations corrupt the process of human hematopoiesis. These data allow to superimpose the two differentiation trees; the native wildtype tree and the one corrupted by mutation. High-resolution mapping of malignant versus normal progenitors showed increased fitness with myeloid differentiation with CALR mutation. We identified the unfolded protein response as a predominant outcome of CALR mutations, with dependency on cell identity. Notably, stem cells and more differentiated progenitors show distinct transcriptional programs as a result of somatic mutation, suggesting differential sensitivity to therapeutic targeting. We further extended the GoT toolkit to genotype multiple targets and loci that are distant from transcript ends. Together, these findings reveal that the transcriptional output of somatic mutations in blood neoplasms is dependent on the native cell identity. Finally, The expanding application of single cell RNA-seq to primary human samples has shown that transcriptional cell state diversity is found ubiquitously across cancer, often independent from genetic clonal diversity. Thus, tumors are composed of admixtures of cells that differ in central phenotypes such as stemness and resistance to therapy. These observation prompts key questions in cancer biology: (i) How are cell states encoded epigenetically? (ii) What is the heritability of cell states? (iii) What is the ability of cells to toggle between cell states? While the exploration of these central aspects of cancer has only begun in model organisms using artificial constructs for lineage tracing, in primary patient samples these questions remains completely unexplored. To address this fundamental knowledge gap in cancer biology, and to define cell state heritability in human somatic evolution, we leveraged our unique ability for joint profiling of DNA methylation and RNA-seq within the same single cells, and performed a direct comparison of the epigenome of transcriptionally defined stem-like cells vs. differentiated malignant cells in glioblastoma. This led to the discovery of PRC2 (Polycomb Repressive Complex 2) as a key switch for differentiation in glioblastoma. Moreover, we integrated high-resolution lineage trees with transcriptional cell states to address critical question of cell state heritability and transitions. We developed an innovative analytic framework, drawing on principles in the field of quantitative ecology, to measure for the first time heritability and transition probability of transcriptional cell states directly in clinical samples. Our findings contrast IDH-mutant gliomas that follow a traditional stem cell hierarchy model with IDH-wildtype glioblastomas that are dominated by cellular plasticity, illuminating a longstanding debate in the field. Thus, we show that in some cancers cell state transitions follow a unidirectional hierarchy akin to normal development, while in others a more plastic bidirectional hierarchy emerges that allows for significant de-differentiation. Citation Format: Dan A. Landau. Single cell multi-omics to define normal and malignant differentiation topologies [abstract]. In: Proceedings of the AACR Virtual Special Conference on Tumor Heterogeneity: From Single Cells to Clinical Impact; 2020 Sep 17-18. Philadelphia (PA): AACR; Cancer Res 2020;80(21 Suppl):Abstract nr IA12.