The Chromatin Accessibility Landscape and Regulatory Network of CD34 Positive Cells in Acute Myeloid Leukemia

Objective: Acute myeloid leukemia (AML) is an aggressive disease with diverse subtypes that exhibit distinct prognosis. Although genetically and molecularly well characterized, the underlying heterogeneity among these subtypes remains unexplained. The distinct regulatory patterns specific to each subtype have yet to be elucidated. Here, we constructed chromatin accessibility landscape and regulatory network on the CD34 positive cells from various molecular subgroups to define subtype-specific regulatory network in AML.

Methods: We collected bone marrow samples from AML patients (N=53). Samples were subjected to selection for mononuclear cells and purified using CD34 magnetic antibodies. Then, we performed ATAC sequencing and RNA sequencing on CD34 positive cells. Furthermore, we incorporated targeted capture sequencing data using an in-house gene panel including 267 AML related genes and the corresponding clinical data of these patients to define the molecular subtype-specific regulatory network of AML.

Results: Unsupervised clustering analysis of ATAC-seq on CD34 positive samples derived from 53 AML patients showed 4,8062 differential accessible peaks covering different subgroups and identified 5 distinct clusters according genetic data, including RUNX1::RUNX1T1, CBFβ::MYH11, NPM1 mutant, CEBPA double mutation (dmCEBPA) and SET::CAN-like class. Non-negative matrix factorization followed by consensus clustering of RNA-seq data uncovered similar transcriptomic clusters with 396 feature genes in purified AML blasts. These results indicated that epigenomic and transcriptomic features reflected genomic alterations in AML patients and further elucidated the underlying regulatory mechanisms. Based on this, we constructed the regulatory-expression networks by differential accessible peaks and feature genes in each subgroup. To examine the relation between genetic alteration and the regulatory network in AML, we analyzed transcription factor binding sites in a genome-wide manner and revealed distinct motifs and footprinting characteristics in each subgroup. Interestingly, the RUNX1::RUNX1T1 cluster exhibited more active epigenetic features and the HSC-related transcription factors like RUNX families were enriched. Besides, transcriptional profiling revealed that RUNX1::RUNX1T1 cluster exhibited higher expression of HSC signatures and lymphoid-related markers. Considering the impact of these transcription factors on cellular differentiation hierarchies, we then applied deconvolution analysis of bulk ATAC-seq and RNA-seq in purified AML samples to explore the composition of cell subpopulations in each patient. The RUNX1::RUNX1T1 and dmCEBPA cluster showed enrichment for HSPC and GMP signatures, while CBFβ::MYH11 and NPM1 mutant clusters exhibited enrichment for GMP and monocyte signatures, indicating differentiation blockade at different stages.

Conclusions: Overall, our study elucidated the aberrant gene regulatory network underlying AML, identified extensive networks of altered regulatory elements, and illuminated the interplay between the genetic, transcriptomic, and epigenetic landscape of the disease. This provided a comprehensive understanding of the heterogeneity among various AML subtypes, offering novel insights into potential therapeutic strategies.

Disclosures: Jianxiang Wang, Advisor of Abbvie.