Chromatin Accessibility Analysis Reveals Epigenetic Evolution Is a Common Mechanism of Relapse in Acute Myeloid Leukemia

Introduction: Acute myeloid leukemia (AML) is associated with a poor prognosis and high rates of relapse despite aggressive treatments including high dose chemotherapy. To understand the clonal dynamics and genetic evolution of relapsed AML, we analyzed a cohort of 142 previously published genotyped and paired diagnosis-relapse AML samples. 40% of queried cases exhibited no major changes in somatic mutations upon relapse, and genetically stable clonal structure correlated with increased relapse probability. Thus, we hypothesized epigenetic reprogramming plays a role in these cases and AML relapse in general. Here, we examine the epigenetic landscape of relapsed AML and characterize the cis and trans regulatory elements that correlate with AML relapse in the presence or absence of genetic evolution.

Methods: We identified 27 viable cryopreserved paired diagnosis and relapse samples from the same patient treated at Stanford with high-dose chemotherapy. Leukemic blasts and (when possible) leukemia stem cell (LSC) enriched populations were FACS purified and prepared for genotyping with a myeloid malignancy targeted sequencing panel as well as ATAC-seq for chromatin accessibility profiling. Single-cell ATAC-seq was further performed on select samples to investigate regulatory reprogramming in cell subpopulations. We then performed integrative analysis to uncover the interplay between genetic lesions, epigenetic regulatory programs, and gene accessibility in relapsed AML.

Results: Genotyping analysis of these AML specimens revealed that 40% of samples exhibited no changes in AML-related genetic alterations upon relapse (hereby referred to as "stable" samples). Chromatin accessibility analysis revealed these stable samples had a distinct epigenetic signature, modulating similar gene accessibility programs and sharing enhancer loci that become accessible across all stable relapse samples. These sequences included genes involved in chromatin organization and compaction, as well as those involved in transcriptional control of hematopoietic differentiation and myeloid cell maturation. We also observed several regulatory signatures present at relapse specific for AML subtypes, including NPM1/FLT3 double mutant AML and those with mutations in transcription factors such as CEBPA or RUNX1.

We then performed single-cell ATAC-seq on genetically stable samples to further characterize the sub-clonal epigenetic dynamics between the diagnosis and relapse cells. Several samples exhibited regulatory heterogeneity in multiple cell subpopulations that changed significantly at relapse. One subpopulation of interest was characterized by increased GATA and RUNX family transcription factor motif accessibility at relapse, indicating a shift toward a less differentiated progenitor cell phenotype. In addition, we identified a subpopulation of cells at diagnosis that were epigenetically similar to the major epigenetic states present at relapse, indicating that selection for specific epigenetic subclones may occur in AML patients during therapy in the absence of additional genetic lesions.

Finally, given the critical role of LSCs in AML pathogenesis and their possible role as a reservoir for AML relapse, we analyzed LSC-enriched subpopulations in a subset of our cohort. ATAC-seq analysis indicated these cells shared several LSC-specific epigenetic features between samples, are distinguished from leukemia blasts by distinct regulatory programs, and undergo epigenetic remodeling between initial diagnosis and relapse. Gene accessibility analysis also revealed a shared LSC gene expression signature that also shifted at relapse. These data indicate a specific, distinct epigenomic signature for LSC enriched cell populations, and that epigenetic evolution at relapse occurs intracellularly, rather than reflecting heterogeneity in cellular subpopulations upon AML relapse.

Conclusion: This study reveals that epigenetic remodeling in the absence of genetic evolution is a mechanism through which AML relapse occurs. We show that chromatin reorganization of genes and regulatory sequences occurs in these AML cells, leading to a permissive cell state that might be resistant to conventional treatment. Ongoing work includes dissecting the subclonal structure of these AML cells and identifying the relationship between gene regulatory networks that contribute to relapse.