The Mechanism of Therapy Resistance By Lineage Plasticity in AML and How to Overcome It

Acute myeloid leukemia (AML) is an aggressive blood cancer driven predominantly by mutations in epigenetic regulators. This characteristic allows AML to thrive in a hostile and rapidly changing environment where it needs to adapt to stressors both extrinsic (chemotherapy) and intrinsic (replicative and oxidative stress) to survive. Our study is the first to demonstrate the mechanism of how AML uses epigenetic reprogramming to adjust to these stressors via lineage plasticity. Venetoclax and Azacitidine (Ven/Aza) co-treatment is increasingly the front-line therapy for AML, functioning through BH3 inhibition and DNA hypomethylation. However, most patients will relapse with a highly aggressive refractory AML. Some studies suggest that monocytic character of the AML is responsible for relapse, but more recent evidence demonstrates that Ven/Aza treatment itself can induce a lineage switch to monocytic character. A potent selection mechanism facilitated by epigenetic dysregulation generating plastic clones is likely driving resistance, though the mechanism remains unclear and represents a significant barrier for AML treatment.

Using a combination of single cell transcriptomics with surface immunophenotype sequencing in multiple human xenograft and primary patient samples, we discovered that this plasticity is widespread and mediated by epigenetic reprogramming of the tumor in direct response to the therapy. Broadly, we found that AML responds to Ven/Aza by transitioning into a reversible state of cell cycle arrest with transcriptional character similar to hematopoietic stem cells. This transient state protects AML from Azacitidine through replication arrest and protects from Venetoclax by upregulation of anti-apoptotic transcriptional programs. These protections operate simultaneously with reprogramming events driven by a switch from MYC-driven transcriptional programs to a PU.1-driven program. Reprogramming is both facilitated by and dependent on repression of the Chromatin Assembly Factor 1 (CAF1) complex, a chromatin-bound master epigenetic regulator responsible for maintaining lineage fidelity through transcriptional regulation via its role as a nucleosome assembly factor. Repressing CAF1 facilitates reprogramming where the output is an altered cell population that is distinct from the initial tumor both transcriptionally and by ATACseq. AML can repeatedly return to this plastic state, regardless of the number of cycles of Ven/Aza, and introducing other cytotoxic compounds does not remove this plastic stem-like population but instead initiates the iterative process of reprogramming anew until a different resistant population emerges.

To find therapeutic vulnerabilities, we generated several PDX models of Ven/Aza resistance so that we could track the stages of acquired resistance in real time. When single cell data from these experiments were projected onto a reference map of normal hematopoiesis, we found that Ven/Aza resistance stems from a plastic population of cells that actively samples the niche by generating short-lived multi-lineage hybrid differentiated products. While the predominating output of this process was monocytic, we found that the induced plastic stem-like population was durably retained even after Ven/Aza withdrawal. To undermine this process, we took advantage of the plastic state to drive differentiation through an external stimulus. We expected this strategy to be particularly effective because CAF1 was repressed (CAF1 repression facilitates differentiation in every metazoan model) and, based on our single cell data, these plastic populations of AML cells express pro-differentiation surface receptors. In particular, Ven/Aza induced the expression of CSF3R (the G-CSF receptor) which when stimulated, drives neutrophil maturation through the JAK/STAT pathway. While G-CSF as a single agent or in sequence with Ven/Aza was unable to achieve maturation, when used concurrently with Ven/Aza (or recapitulated by CAF1 depletion using shRNA) we were able to drive neutrophil differentiation in vitro and fully resolve primary mouse and PDX tumors in vivo, even when using Ven/Aza resistant tumors as starting material. Together, these data demonstrate the possibility that Ven/Aza resistance can be completely subverted by taking advantage of the plastic state induced by the therapy to stimulate differentiation and resolve the disease.