AUTHORS
Mengbao Du, Limengmeng Wang2, Shijun Xiong3, Yue Liang4, Lina Liu5, Jun Liu6, Yin Zhao7, Peihan Yu8, Shanshan Pei, PhD9, He Huang, MD, PhD10
AFFILIATIONS
1 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
2 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
3 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
4 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
5 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
6 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
7 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
8 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
9 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
10 Bone Marrow Transplantation Center of The First Affiliated Hospital & Liangzhu Laboratory, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China
ABSTRACT
Despite the clinical success of mIDH1/2 inhibitors for IDH1/2-mutated acute myeloid leukemia (AML), durable remissions remain elusive for most patients. Our current understanding of mIDH1/2 inhibitor resistance mechanisms largely centers on steady-state genetic mutations and signaling pathways. Given that the treatment course lasts for several months, we hypothesize that biological pathways directly but unintentionally induced by mIDH1/2 inhibitors can drive adaptive resistance by counteracting their own pro-differentiation effects. However, these mechanisms remain poorly understood and lack systematic characterization.
To address this challenge, we performed a genome-wide CRISPR-Cas9 knockout screen in a self-engineered IDH1-mutated TF-1 cell line, using 7-day ivosidenib (IVO) treatment as a continuous therapeutic pressure and the differentiation marker fetal hemoglobin as a functional readout (Commonly used by Losman et al., Science, 2013; Lyu et al., Cancer Discov, 2023, etc.). Our screen identified 1,277 candidate genes responsible for blocking IVO-induced differentiation. Integration of screen hits with transcriptomic signatures of IVO-resistant cells identified 53 overlapping genes. The top five include PIK3R3, PDCD1LG2, GGACT, ATM, and SMAD1. Among them, ATM upregulation was verified in multiple mIDH1/2 inhibitor-resistant settings, including longitudinal clinical relapse samples from published cohorts (Sirenko et al., Cell Stem Cell, 2025), in vitro selection-derived resistant cell lines, and in vivo therapy-resistant cell line-derived xenograft (CDX) models. In addition, both pharmacological inhibition (KU-55933/KU-60019) and genetic knockdown of ATM significantly sensitized AML cells to mIDH1/2 inhibitors, whereas CRISPRa-mediated upregulation of ATM expression (7.25-fold) drove robust resistance. These data identify ATM as a potentially actionable novel target driving adaptive resistance to mIDH1/2 inhibitors.
Mechanistically, we first performed comprehensive CUT&Tag and Methylated DNA Immunoprecipitation assays and revealed that the reduction of D-2-HG levels during mIDH1/2 inhibitor treatment was accompanied by the epigenetic derepression of the ATM promoter and subsequent transcriptional activation. Surprisingly, through comet assays and γH2AX quantification, we ruled out canonical DNA damage repair (DDR) as the driver of ATM-mediated resistance. Consequently, we pursued an unbiased proteomic approach to identify novel, non-DDR effectors of ATM. The ATM interactome was first mapped via Co-IP-MS, and the ATM-regulated proteome was identified by quantitative proteomic profiling. Integrating these two datasets revealed 58 proteins directly bound and upregulated by ATM, with the epigenetic player DOT1L ranking among the top hits. Subsequent quantitative phosphoproteomics mapped ATM-induced phosphorylation of DOT1L precisely to the S786 and T790 residues. Functionally, in ATM-overexpressing cells, DOT1L was stabilized, and the downstream HOXA9/MEIS1-mediated stemness program was activated. Ectopic expression of the phospho-dead DOT1L mutant (S786A/T790A) reduced DOT1L protein stability and impaired its ability to block differentiation.
To evaluate the translational potential, we treated IVO-resistant IDH1-mutated TF-1 cells with the ATM inhibitor KU-60019 plus IVO. This combinatorial approach yielded a robust synergistic anti-leukemic effect both in vitro and in CDX models. Crucially, we extended these findings to primary IDH1/2-mutated AML samples. In vitro, pharmacological inhibition or genetic knockdown of ATM drastically enhanced the terminal myeloid differentiation response to mIDH1/2 inhibitors. In patient-derived xenograft (PDX) models, in vivo treatment with combination therapy profoundly reduced bone marrow leukemic burden (hCD45+: Vehicle 98.87%, Enasidenib 77.8%, KU-60019 95.63%, Combination 21.7%; p<0.0001) and induced terminal myeloid differentiation compared to the control and monotherapy groups, without causing overt hematological toxicity.
In summary, we found a non-canonical ATM–DOT1L signaling axis directly activated by mIDH1/2 inhibitor treatment, serving as a novel mechanism underlying adaptive resistance. This study opens a new avenue to understand resistance mechanisms elicited by drug-induced resistance pathways. As a proof of concept, targeting ATM abolishes this resistance cascade and restores differentiation, offering a promising combination strategy for IDH1/2-mutated AML.
DISCLOSURES
Mengbao Du: No relevant conflicts of interest to declare.