Low Exposure Extended Dosing Mimicking Clinical Exposures of the Oral Formulation of Azacitidine Results in a Sustained Hypomethylation and Targets Leukemic Stem Cells

Acute Myeloid Leukemia (AML) is characterized by uncontrolled proliferation of incompletely differentiated myeloid stem/ progenitor cells. Despite recent advances in therapy, high rates of clinical relapse, even in patients who achieve complete remission, remains a critically unmet need. One of the strategies to prolong remission post-induction in AML is to employ effective maintenance therapies. Oral Azacitidine (Oral-AZA; CC-486) is the first and only currently approved maintenance therapy in AML. However, the mechanism of action of Oral-AZA and whether it is differentiated from the Injectable-AZA used in AML induction therapy, remains unclear.

In this work, we explore the mechanism of Oral-AZA by in vitro modelling of the relative clinical exposures of Oral-AZA vs Injectable-AZA in AML cell line models. Following Vu et al (Nat. Commun. 2020), we used the Injectable-AZA concentration (1 μM aZA) as a single dose (HELD - High Exposure Limited Duration). A fractionated dose of 0.2 μM each day over 5 days (LEED - Low Exposure Extended Duration) was used to model Oral-AZA.

Azacitidine incorporates into RNA and DNA of AML cells leading to hypomethylation driven gene expression changes. We found that HELD but not LEED dosing produced an acute anti-proliferative effect in sensitive AML cell lines suggesting potential for a non-hypomethylation mediated/integrated stress response (ISR) driven mechanism. This ISR effect was rescued by co-treatment with ISRIB, an ISR inhibitor. With HELD, we observed robust ATF4 activation as early as 6 hours that was sustained up to 24 hours. In contrast, LEED induced modest and transient ATF4 activation. Thus, an Injectable-AZA-like regimen activated the ISR pathway more robustly than an Oral-AZA-like regimen. Interestingly, decitabine, a DNA-only incorporating cytidine analog did not activate ISR.

In AML cell lines, the target of Azacitidine, DNMT1 was depleted (about 90% compared to control) within 24 hours in both HELD and LEED regimens. However, LEED produced a more sustained DNMT1 loss, up to 7 days whereas in HELD, DNMT1 protein levels recovered 96 hours post-dosing. Given the relative level and duration of DNMT1 loss, we hypothesized that LEED would lead to a more durable hypomethylation. We performed whole genome bisulfite sequencing (WGBS) in 3 AML cell lines (OCI-AML2, MV-4-11 and SKM1) at 48- and 96-hours post-initiation of HELD and LEED dosing. Consistent with the DNMT1 depletion kinetics, at 48 hours we observed almost 75% global DNA hypomethylation with both HELD and LEED treatments. At 96 hours, HELD demonstrated a recovery effect and reverted to 50% hypomethylation. In contrast, LEED showed up to 85% hypomethylated sites, demonstrating the durability of the hypomethylation mediated by an Oral-AZA-like (LEED) regimen.

Next, we explored the effect of Oral-AZA-like dosing in the differentiation of leukemic stem cells (LSCs) towards a more mature phenotype. In an in vitro LSC model (OCI-AML-20) with flow cytometry analysis, LEED dosing resulted in a greater depletion (2-fold more) of LSCs (CD34+/38- or 38 low) and concomitant enrichment of cells with more differentiated phenotype (CD34+/38+) compared to HELD dosing. To further validate these observations at the transcriptomic level, we performed single cell RNAseq in OCI-AML-20 cells at different timepoints (3, 5 and 7 days). Data were analyzed using a classifier to identify leukemic myeloid cell lineages (Van Galen et al., Cell 2019). Compared to untreated cells, at day 7, both LEED and HELD dosing resulted in an increase of GMP and promonocytes and those differences were more pronounced with Oral-AZA-like (50% GMP and 20% promonocytic) than the Injectable-AZA-like regimen (28% GMP and 12.5% promonocytic).

Taken together, our in vitro modeling of the clinically relevant exposures of Oral-AZA vs Injectable-AZA by using HELD and LEED dosing strategies show a shift of the molecular mechanism between the clinical entities. Our work demonstrates that an Injectable-AZA-like regimen mediates cytotoxicity through an early ISR driven effect. Oral-AZA mechanism leads to a more sustained pharmacodynamic DNA hypomethylation effect that in turn could be linked to a differentiation inducing effect on the LSC population.