The Histone Demethylase KDM4A Is Required to Sustain H3K9me3/H3K27me3 Epigenetic States and Oncogenesis in MLL-AF9 Acute Myeloid Leukemia

Acute Myeloid Leukemia (AML) is an aggressive heterogeneous hematological malignancy. Persistence of leukemia stem cells (LSC) drive AML leukemogenesis, responsible for drug resistance and disease relapse following conventional chemotherapy. Growing evidence demonstrates that epigenetic-based therapies pose a unique and rational avenue for eradication of LSCs, enabling long-term remission and cure. In this study, we identified the histone lysine specific demethylase 4A, KDM4A, as an essential regulator of AML oncogenic potential. KDM4A inhibition by shRNA knock-down (KD) or KDM4A inhibitors, selectively promoted myeloid differentiation of AML cells, resulting in significant AML cell death whilst sparing normal human CD34⁺ hemopoietic stem and progenitorcells (HSPC). Despite substantial evidence demonstrating that KDM4A is amplified and overexpressed in various malignancies including breast, lung, ovarian, prostate cancer and leukemia, there is little information to date as to its defined role in AML leukemogenesis nor whether it represents a viable therapeutic target.

To investigate the molecular mechanisms underlying the leukemia-selective dependence on KDM4A, we performed ChIP-seq and RNA-seq to profile the epigenomic and transcriptional consequence of KDM4A KD in MLL-AF9-driven human AML THP-1 cells, followed by validation in primary AML patient blasts and normal human CD34⁺ HSPCs. KDM4A KD leads to a significant global enrichment of its substrate, H3K9me3 and surprisingly H3K27me3 (substrate of the Polycomb Repressive Complex 2, PRC2), and within the promoter regions of a number of KDM4A bound genomic loci. Both H3K9me3 and H3K27me3 are the repressive histone modifications, which correlate with the transcriptional down-regulation of PRC2 target genes upon KDM4A KD including Nuclear Factor of Activated T Cells 2 (NFATC2) and RNA Polymerase Associated Factor 1 (PAF1). KD of either NFATC2 or PAF1 leads to AML cell apoptosis, while over-expression of NFACT2 in THP-1 cells partially overcomes KDM4A inhibition. Together we have established NFATC2 and PAF1 as two key KDM4A direct downstream targets, both of which have well-established roles in AML leukemogenesis, promoting AML cell survival. Interestingly inhibition of de novo H3K27me3 using the pharmacological inhibitor of histone methyltransferase EZH2 EPZ6438, resulting a reduction of H3K27me3, reduces accumulation of H3K9me3 and partially rescues the detrimental phenotype of KDM4A KD in THP-1 cells, suggesting that KDM4A epigenetic regulation in MLL-AF9 AML is PRC2 activity dependent. Taken together, our data have uncovered a new insight of cross-talk between two repressive epigenetic modifications of H3K9me3 and H3K27me3 regulated by KDM4A and required for MLL-AF9 cell survival.

To further address the clinical relevance of KDM4A and its downstream targets in AML, we employed meta-analysis on publicly available patient datasets. Our analysis of 461 AML patient samples, shows that KDM4A^highpatients are enriched amongst poorly differentiated subtype of AML (p=0.01505; Fisher's exact test) and rare amongst more differentiated AML (p=0.07153; Fisher's exact test). High expression of NFATC2relates to worse clinical outcome in a large cohort of AML patients. Furthermore, using Lasso regression algorithm, we relate KDM4A KD induced global transcriptional changes to patient survival in a large AML dataset, yielding an optimal 21-gene signature (KDM4A-21). The KDM4A-21 score can be calculated for each patient as the weighted sum of expressions of the 21 genes. The score^high patients in a number of large de novo AML cohorts confer significantly poor overall survival compared with score^low population (p<0.05; Mantel-Cox log-rank test). We are further evaluating these genes as potential cellular biomarkers of KDM4A inhibition for future preclinical trials.

To our knowledge, these data for the first time delineate an essential and selective role for KDM4A in AML oncogenesis. Our results reveal a novel tractable regulatory network of KDM4A-PAF1/NFATC2 that is sustained by KDM4A mediated H3K9me3/H3K27me3 epigenetic crosstalk and is indispensable in MLL-AF9 AML cells. Our study provides strong evidence to establish that KDM4A and its downstream targets could represent propitious novel therapeutic targets, and the associated transcriptional network may be used to guide personalised epigenetic therapies in AML.