Epigenetic Adaptation Promotes Cellular Resilience in Senescence-like AML Cells

Despite initial response to treatment, most patients with acute myeloid leukemia (AML) will relapse and die of therapy-resistant disease. There is a significant need to understand the mechanisms that drive relapse of AML. Since selection of drug resistance mutations are rarely observed in relapsed AML, relapse in AML cannot be explained by genetic selection alone. Recent findings from our research group identified that residual AML cells surviving chemotherapy adopt a transient senescence-like phenotype (SLP), regardless of their genetic background and leukemia stem cell status. Upon recovery, we observed that post-senescent cancer cells gave rise to relapsed AMLs that are more tolerant to chemotherapy, consistent with progressive stress adaptation. However, the mechanisms enabling this adaptation and cellular resilience of AML cells remain unknown.

To investigate these processes in AML, we leveraged our recently developed ex vivo AML recurrence model involving primary patient-derived AML cases treated with cytarabine (Ara-C), the main component of induction chemotherapy. AML cells recovering from Ara-C-induced senescence exhibited significant changes in DNA 5-methylcytosine patterns. Analyzing the DNA methylome via enhanced reduced representation bisulfite sequencing revealed that SLP cells undergo DNA methylation changes. Subsequently, we isolated single leukemia cells from the residual fraction of chemotherapy-surviving SLP cells post-Ara-C treatment, as well as from untreated control cells. These single cells were expanded to generate monoclonal populations, effectively reducing genetic heterogeneity. Integrative analysis of the transcriptome and DNA methylome identified that independent AML clones acquired gene-specific DNA promoter hypermethylation following senescence recovery, with GFI1 emerging as one of the top candidate genes undergoing epigenetic silencing. To further investigate the functional role of GFI1 in SLP cells, we epigenetically silenced GFI1 using CRISPRi approaches and observed that silencing of GFI1 facilitated cell survival of SLP cells following chemotherapy treatment. Consistent with this finding, survival analysis of public databases including TCGA demonstrated that low mRNA expression levels of GFI1 in AML is linked to poor survival outcome in AML patients. Moreover, low expression of GFI1 was associated with promoter hypermethylation of GFI1.

Since DNA promoter methylation is established by DNA methyltransferases (DNMTs), we next tested the impact of DNMT inhibitors (DNMTis), such as 5-Azacitidine (5Aza) on SLP cells. Exposure to a low non-cytotoxic dose of 5Aza (≤100 nM) impaired SLP cells to initiate AML recurrence. To identify which specific DNMT facilitates the recovery of SLP cells, we conducted knockdown experiments in patient-derived AML cells targeting all three catalytically active DNMTs: DNMT1, DNMT3A and DNMT3B. Unlike DNMT3A/3B, knockdown of DNMT1 significantly impaired the recovery of SLP cells. To assess the therapeutic potential of targeting DNMT1, we treated SLP cells with a novel, potent specific DNMT1 inhibitor (DNMT1i; GSK3685032). While AML cells treated with either DNMT1i or Ara-C alone were able to recover from treatment, combination of Ara-C and DNMT1i abrogated recovery and colony-forming potential of SLP cells ex vivo. In addition, treatment of patient-derived xenograft mouse models with DNMT1i impaired leukemia recurrence from Ara-C-induced SLP cells in vivo.

Overall, our research demonstrates the importance of DNA methylation-mediated transcriptional changes for cellular resilience of AML cells. Targeting DNMT1 could disrupt the cellular resilience and epigenetic adaptation, providing a potential therapeutic strategy to reduce the risk of relapse arising from senescence-/diapause-like leukemia cells.