Intra-Tumor Heterogeneity in Core-Binding Factor Acute Myeloid Leukemia

Intra-tumor heterogeneity (ITH) describes the coexistence of multiple genetically distinct subclones within the tumor of a patient resulting from somatic evolution, clonal diversification, and selection. It is a main causal driver of therapy resistance and treatment failure in the clinic by already containing subclones that are resistant to therapy or by subclones acquiring resistance to therapy. Core-binding factor (CBF) acute myeloid leukemia (AML) is cytogenetically defined by a translocation of chromosome 8 and 21 [t(8;21)(q22;q22)] or an inversion/translocation of chromosome 16 [inv(16)(p13.1q22) or t(16;16)(p13.1q22)] resulting in RUNX1::RUNX1T1 and CBFB::MYH11 fusion genes, respectively. Despite being classified as AML with favorable prognosis up to 45% of patients relapse and, therefore, understanding the history of somatic events of this heterogeneous AML subtype can contribute to improve clinical outcome.

To unravel ITH and clonal evolution for both small-scale (single nucleotide variants, insertions and deletions) and large-scale (somatic copy-number alterations (SCNAs), gene fusions) genetic alterations in CBF AML, we analyzed longitudinal samples from 9 patients with t(8;21) or inv(16) AML by whole exome, error-corrected targeted and nanopore long-read bulk sequencing (24 samples). Single-cell DNA sequencing (scDNAseq) of longitudinal samples was performed via the MissionBio Tapestri® platform using custom targeted sequencing panels including information on patient-specific somatic variants, SCNAs and CBF fusions. Samples were collected at the time of diagnosis (D), complete remission (CR), and relapse (Rel). We used a novel approach for reconstructing the order of somatic events in our patients that allowed us to detect SCNAs in phylogenetic trees inferred from somatic variants and CBF fusions.

In total, we identified 399 variants via bulk sequencing. In the diagnosis samples (n=9) 230 variants were detected and 169 variants in the relapse samples (n=8). Additionally, 7 SCNAs in 5 patients and patient-specific gene fusion breakpoints for all patients were identified using bulk and Nanopore sequencing. We were able to detect patient specific CBF fusions in single cells for 8/9 patients and for those 8 patients we detected between 61% and 96% (mean: 86%) of the variants identified via bulk sequencing. Additionally, we identified 10 variants by scDNAseq in the regions covered by the custom amplicon panels, that have not been detected in bulk sequencing, including FLT3-ITD.

In total, we inferred tumor phylogenies for 8 patients with detectable gene fusion at D and inferred clonal evolution under chemotherapy pressure for 3 patients. We were able to detect SCNAs by applying our method in small tumor subclones that were not detectable with bulk data nor using existing single-cell tools. By reconstructing the order of somatic evolutionary events from scDNAseq, we can demonstrate that the CBF gene fusions were the first events to occur in disease development in our patients. For patients with samples at D and Rel (n=3) we observed a persisting founding clone throughout the course of the disease. We identified remaining tumor cells in available CR samples (n=6) ranging from 4 to 34 cells (0.16%-1.54%) that harbor at least one variant or the CBF fusion. Among the 148 cells with any detectable variant/fusion, only 5 cells would have been detected when only investigating the fusion. Thus, simultaneous tracking of multiple genotypes showed a clear superiority compared to current strategies, which focus on fusion detection for MRD purposes. By assigning those cells to tumor clones inferred from samples at D and Rel, we can demonstrate that the AML clones were not completely eradicated during treatment.

In summary, our study presents a detailed picture of ITH and clonal evolution including somatic variants, gene fusions and SCNAs at single-cell resolution of 9 CBF AML patients. We reconstructed the mutational order of somatic events within one tumor sample including acquisition of CBF gene fusions and SCNAs, which adds a new layer into clonal evolution of CBF AML. Moreover, we were able to detect and assign remaining tumor cells at CR to inferred tumor clones of each patient. This highlights the importance of improving the understanding of tumor development and, in more detail, the necessity of identifying early events during tumorigenesis in CBF AML patients.