Mechanisms of Azacitidine Chemotherapy Resistance in AML and MDS and New Therapy Options

BACKGROUND:TET2 (Ten-eleven-translocation-2) mutations found in 10-30% of MDS and AML patients are actionable for Azacitidine (AZA) treatment (PMID: 21828143). Although response rate of AZA in older AML patients is ~47%, rate of relapse is high due to evolution of disease clones and development of new mutations during treatment cycle. Currently, there is no standard of care for AZA-resistant MDS and AML, and treating these individuals is challenging. Thus, there is an urgent need to identify mechanisms for resistance to AZA for TET2-mutant cancers.

AIM: The objective of the study was to predict response to AZA in OCI-M2 disease model and its AZA-resistant (AZA-R) sub-clones using computational biology modeling (CBM) to determine mechanisms of resistance to AZA and identify new therapy options for AZA chemoresistance.

METHODS: Parental OCI-M2 cells, representing an AML transformed from MDS, were treated with 8 mM AZA added to media every 2 days (IC50=2µM) for a period of 6 weeks to generate OCI-M2 AZA resistant subclones. AZA-R subclones together with parental AZA-sensitive cells were profiled using cytogenetics and NGS (TruSight Myeloid Sequencing, Illumina). Genomic data were analyzed via CBM system (Cellworks Group), which generates disease-specific protein network maps and enables digital drug simulations. Impact of digital drug treatments were quantified by measuring predicted effects on AML cell growth score, which is a composite of cell proliferation, viability and apoptosis.

RESULTS: CBM correctly predicted the response of parental OCI-M2 AML cells to be sensitive to AZA and all its AZA-resistant subclones to be resistant to AZA treatment. Genomic analysis using CBM showed that OCI-M2 AML model harboured TET2mutation (Q1084P), KAT6A amplification and DNMT3B amplification, which are putative determinants of AZA sensitivity as the AZA-resistant subclones lacked cells harbouring these mutations (TET2, KAT6Aamp and DNMT3Bamp). Instead, AZA treatment selected clones with frameshift or nonsense mutations in the following genes: ASXL1, EZH2, CUX1, ATRX, BCOR, PTPN11, FLT3 along with preserved TET2 in all the clones. Mutations in ASXL1 and EZH2 were previously shown to reduce the importance of methylation in driving disease, thus potentially leading to subclone proliferation regardless of AZA. Also 4 of 6 AZA-resistant subclones had DNMT3A frameshift mutations which may enhance AZA resistance.

CBM identified new drug combinations based on overlapping as well as unique mutations in AZA-resistant subclones. These combinations include Ruxolitinib + Venetoclax, Nelfinavir + Ruxolitinib or Venetoclax, Sorafenib or Dasatinib + Venetoclax. Some of these drugs were shown to effectively substitute AZA on the parental cell lines (Venetoclax IC=0.9µM, Ruxolitinib IC50=1.7µM, Sorafenib IC50=5.3µM). Whether AZA enhanced sensitivity to these drugs in AZA-R clones is currently under investigation.

Conclusions: CBM-based genomic analysis identified secondary mutations in ASXL1, EZH2, CUX1, ATRX, BCOR, PTPN11 and FLT3 that associated with AZA chemoresistance. Digital drug screening identified new drug combinations including Venetoclax for validation testing in AZA-resistant MDS and AML.