MGMT-silencing as a novel biomarker in AML: Rational application of a dual alkylator-DNA crosslinker for targeted therapy

Abstract Acute Myeloid Leukemia (AML) remains a therapeutically challenging malignancy with poor outcomes. Standard chemotherapy regimens or hypomethylating agents with BCL2 inhibitors frequently fail upfront and relapses are common. Addition of targeted agents (IDH1/2 or FLT3 inhibitors) improves outcomes but do not cure leukemia emphasizing the need for novel biomarkers and therapies. Epigenetic silencing of O6-methylguanine-DNA methyltransferase (MGMT) occurs as a common DNA damage response (DDR) defect, present in approximately 11% of tumors according to TCGA data, with low-grade glioma and AML among the lowest MGMT-expressing cancers (Knijnenburg., et al. Cell Reports, 2018). MGMT repairs O6-guanine alkylation damage and its loss sensitize cells to alkylating agents. While MGMT is a predictive biomarker for temozolomide (TMZ) in glioma, its role in AML remains underexplored, despite evidence of MGMT silencing in subsets of AML. We conducted a multi-omics analysis of 34 de novo and 15 relapsed/refractory AML samples from the Yale Hematology Tissue biobank, alongside MGMT methylation and expression analysis in 170 TCGA AML cases. Interestingly, 25-30% of AMLs showed high MGMT promoter methylation and low transcript/protein levels. These alterations did not correlate with any specific cytogenetic abnormalities or mutations, and thus, could serve as an independent exploitable vulnerability. Importantly, CD34+ cells from healthy individuals retained MGMT protein expression, highlighting MGMT silencing as a leukemia-specific event. To explore therapeutic strategies for MGMT-silenced AML, we screened 10 alkylating agents generating diverse array of DNA lesions, including TMZ and our novel compound N3-(2-fluoroethyl) imidazotetrazine (KL50), which produces O6-(2-fluoroethyl)guanine lesions that undergo time-dependent conversion to DNA inter-strand crosslinks. The screen utilized isogenic MOLM13, U937 and K562 AML cell lines engineered to exogenously express MGMT, and THP1 and Kasumi1 lines engineered with knock down of endogenous MGMT Interestingly, TMZ and KL50 emerged as the hits imparting the maximum differential activity in MGMT non-expressing compared to expressing isogenic cells. In parallel, PRISM screen results across 22 AML cell lines showed a strong inverse correlation between KL50 sensitivity and MGMT expression, with moderate correlation for TMZ. In vivo, KL50 showed superior efficacy and >6 months tumor-free survival at ~5-fold lower doses than TMZ in multiple MGMT-silenced, AML cell-derived xenografts (CDXs) including p53-null CDXs. To evaluate efficacy in a more clinically relevant setting drugs were tested in MGMT-silenced and MGMT-proficient primary AML patient-derived xenografts (PDXs) engrafted into cytokine-humanized MISh/mTRG (CSF1h/h, IL3/CSF2h/h, SIRPAh/m, THPOh/h, Rag2-/-, Il2rg-/-) mice. Both KL50 and TMZ induced complete remission of MGMT-silenced PDX, eliminating human CD45+ and CD34+ cells in peripheral blood with no relapse for over a month (study ongoing). In contrast, MGMT-proficient PDXs did not respond to either TMZ or KL50. Based on clinical evidence from glioma showing that mismatch repair (MMR) loss drives TMZ resistance, we investigated its resistance mechanisms in multiple MGMT-silenced, MMR-proficient AML CDX models. Notably, TMZ resistance developed rapidly (3 weeks to 2 months), whereas KL50 sustained durable responses. The whole exome sequencing revealed loss-of-function mutations in MMR genes, often involving C>T or G>A transitions, in multiple resistant models. Clonal re-expression of MGMT post TMZ was observed rarely and was associated with MGMT promoter methylation changes. In isogenic CDX models with MMR (MSH6) knockdown, TMZ resistance was confirmed, consistent with TMZ's need for functional MMR to elicit apoptotic signaling post O6-guanine alkylation. Conversely, MSH6-deficient tumors retained sensitivity to KL50, in accordance with the ability of inter-strand DNA crosslinking to induce apoptosis independent of MMR. We are now screening the MGMT-deficient AML samples for MMR protein expression. Identification of MMR deficiency as a key resistance mechanism to TMZ in these patients could further explain its poor efficacy observed in previous AML trials. This comprehensive study explains the limited potential of TMZ in MGMT-silenced AML and provides the rationale to initiate clinical trials with a novel time-dependent crosslinker (KL50) in a biomarker driven fashion in AML.