In order to comprehensively characterize the molecular features of chromosomal structural changes and their phenotypic outcomes at the individual cell level in Acute Myeloid Leukemia with complex karyotype (CK-AML), we integrated two single-cell multi-omics methods to study genetic and non-genetic heterogeneity in parallel. We have established an algorithm that allows computational assembly and linkage of resulting different datasets. We have implemented our method for the analysis of patients at diagnosis, treatment and relapse stages. We have revealed unprecedented genetic heterogeneity between individual CK-AML patient cells. We could monitor the growth patterns of tumor subclones, their transcriptome and cell-surface proteome over time and space to identify therapeutic vulnerabilities. We uncovered genetic and non-genetic factors that promote disease resistance and relapse capabilities of leukemic stem cells. Based on these datasets, we evaluated the drug-response profiles of leukemic stem cells.
In Reporting Period 1, we have successfully implemented the scNOVA-CITE technology in first set of CK-AML patients and identified a combination of genetic and non-genetic mechanisms at single-cell resolution (Research Line 1) that mediate tumor evolution, progression and identified novel therapy resistance mechanisms (Research Line 2). Major parts of the results in Reporting Period 1 have recently been published in Nature Genetics (Leppä AM*, Grimes K*, Jeong H*,….Korbel JO#, Trumpp A#. Single-cell analysis reveals dynamic clonal evolution and targetable phenotypes in complex karyotype AML. Nature Genetics, 2024 Dec;56(12): 2790-2803. doi: 10.1038/s41588-024-01999-x *Equal Contribution, #Correspondence). In the following, we will provide the progress since last report.
We have acquired a large cohort of longitudinal CK-AML samples from 19 patients. This cohort includes 39 samples from the time of diagnosis, remission and relapse or refractory disease. Using our scNOVA-CITE framework, we are analyzing this larger cohort to further characterize the genetic and non-genetic mechanisms of therapy resistance used by CK-AML cells in response to chemotherapy. We have so far analyzed 1,072 single-cell genomes for SVs from 31 samples and observed a high degree of inter- and intra-patient heterogeneity with SVs ranging from 3-24 per cell. Comparing longitudinal samples, we found that at diagnosis, the number of subclones varies widely (1-33 subclones per sample), indicating high clonal heterogeneity. In contrast, at relapse/refractory disease, the range is narrower (5-10 subclones per sample). Monoclonal samples tend to diversify, whereas polyclonal ones reduce heterogeneity, suggesting a clonal “sweet spot” for leukemia survival. We hypothesize that high degree of clonal diversity may lead to instability while too few clones may limit the adaptability upon therapy. Moreover, we identified that TP53 mutated CK-AMLs frequently exhibit branched polyclonal growth, whereas CK-AMLs with recurrent chromosomal aberrations (e.g. KMT2a-rearrangements) tend to show either monoclonal or linear growth patterns.
In addition, we have performed CITE-seq analysis of 267,703 cells in total from 39 samples including the ones above. We projected CK-AML samples onto the healthy reference dataset to assess cellular differentiation status of leukemic cells. Overall, differentiation was highly heterogeneous, with most samples appearing primitive and enriched for Hematopoietic Stem-like (HSC-like) cells. However, projections spanned all compartments, including erythroid, megakaryocytic, MEP, LMPP, GMP, promonocytic, monocytic, and dendritic-like cells, with some even resembling MPAL (mixed phenotype acute leukemia). Notably, individual samples often contained multiple cell types. Comparing diagnosis to relapse/refractory disease, we consistently observed a reduction in promonocytic/monocytic cells and a relative increase in erythroid and megakaryocytic populations, suggesting a shift in differentiation over time. Integrating CITE-Seq and Strand-Seq data further revealed subclone-specific lineage biases.
We have developed a computational framework called translocatoR+ that uses Strand-seq data to characterize highly complex balanced and unbalanced translocations at the single-cell level. Using translocatoR+ we were able to reconstruct complex events including up to six chromosomes as well as chromoplexy events at the single-cell level. We also identified sub-clonal translocations that emerged upon relapse. We have validated translocatoR+ using orthogonal methods (optical genome mapping and in silico mixture of cells with known translocations at different ratios. Currently, we are identifying the fusion genes created by these translocations and use scNOVA to study functional consequences.
