Survival outcome for patients with acute myeloid leukemia (AML) remains poor overall. A notable exception is acute promyelocytic leukemia (APL), where differentiation therapy with all-trans retinoic acid (ATRA) has transformed outcome. Attempts to replicate this success in non-APL AML have been unsuccessful so far. One emerging target is LSD1, an epigenetic regulator that maintains an undifferentiated state in AML. While LSD1 inhibitors (LSD1i), alone or in combination with ATRA, show promise in preclinical models, their clinical efficacy remains limited, underscoring the need for more effective strategies. Importantly, ATRA's efficacy enhances significantly when combined with arsenic trioxide (ATO), which elevates reactive oxygen species (ROS), a group of reactive oxygen-containing molecules including superoxide anions that collectively contribute to oxidative stress.

To examine the role of ROS during LSD1 inhibition, we treated primary patient-derived AML samples and cell lines with LSD1i (GSK-LSD1) and monitored ROS dynamics using selective intracellular biosensors and ROS type-specific fluorescent probes. LSD1i induced rapid accumulation of superoxide within 24 hours in a non-mitochondrial pathway and preceded the onset of AML differentiation. Notably, scavenging ROS with N-acetylcysteine (NAC) significantly impairs differentiation, indicating that ROS plays a critical role in mediating the pro-differentiation effects of LSD1i. To validate that LSD1 regulates ROS, we developed a murine leukemogenic model by transducing hematopoietic stem and progenitor cells (HSPCs) from Lsd1fl/flconditional knockout mice with the MLL-AF9 oncogene and tamoxifen-inducible Cre recombinase. Tamoxifen-induced deletion of Lsd1 led to pronounced ROS accumulation, whereas deletion of Ezh2, encoding another epigenetic regulator, did not affect ROS levels, demonstrating that ROS accumulation is specific to LSD1 loss.

Analyzing public AML datasets revealed a strong negative correlation between LSD1 expression and ROS-modulating/inflammatory gene signatures (p<2.3e-55). We confirmed that LSD1 inhibition upregulated ROS and inflammatory genes. To determine the mechanism of LSD1i-induced ROS accumulation, we performed RNA-seq and ChIP-seq in patient-derived AML cells. We identified that LSD1 directly binds to and represses upstream enhancer activity regulating CYBB, the gene encoding NADPH oxidase 2 (NOX2), a key enzyme involved in superoxide ROS generation. Using a selective NOX2 inhibitor (GSK2795039) abolished LSD1i-induced ROS accumulation.

Surprisingly, despite sustained superoxide elevation after LSD1 inhibition, oxidative stress levels and DNA damage rose only transiently, before returning to levels comparable to untreated AML. To investigate this adaptive response, we performed RNA-seq on patient-derived AML samples after oxidative stress had subsided. All cases displayed significant negative enrichment of ATR target gene signatures. We validated that LSD1i treatment activated ATR in a ROS-dependent manner. In turn, ATR activation downregulated key enzymes involved in generating highly reactive ROS (hROS), including myeloperoxidase (MPO), thereby attenuating oxidative stress. Inhibition of ATR prevented MPO suppression, restored hROS levels, and abrogated redox homeostatic adaptation, resulting in sustained oxidative stress.

Combined inhibition of LSD1 and ATR caused ROS overaccumulation, pronounced AML differentiation with morphological changes, and AML cell death. In PDX models, the combination therapy significantly improved survival compared to the monotherapies. The enhanced anti-leukemic effects, including increased differentiation and cell death were reversed by ROS scavenging and inhibition of NOX2 or MPO.

Overall, our findings reveal LSD1 as a critical epigenetic gatekeeper that constrains oxidative stress in AML. Inhibition of LSD1 triggers rapid superoxide accumulation, functioning as a pro-oxidant agent. However, we further uncovered a novel resistance mechanism wherein AML cells dynamically adapt to LSD1i-induced oxidative stress. Rather than decreasing overall ROS levels after LSD1i treatment, AML cells selectively suppress specific hROS (MPO-derived HOCl), thereby rebalancing the ROS composition to avert excessive oxidative stress and maintain survival. Co-inhibition of ATR disrupts this redox compensation and potentiates LSD1i-mediated differentiation therapy by forcing lethal oxidative stress escalation.

This content is only available as a PDF.
Sign in via your Institution