The Fanconi anemia pathway restrains MLL-rearranged leukemogenesis through suppressing non-homologous end joining-mediated genomic instability

Abstract

Mixed lineage leukemia (MLL) rearrangements drive approximately 10% of acute leukemias, including acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL). Individuals with Fanconi anemia (FA), a hereditary DNA damage repair (DDR) disorder, are at markedly increased risk of AML, yet the interplay between FA deficiency and MLL-rearranged (MLL-r) leukemia remains unclear. Using murine models and human MLL-r leukemia cells, we show that the FA pathway suppresses leukemogenesis by limiting error-prone non-homologous end joining (NHEJ)-mediated genomic instability. Loss of FA components accelerates leukemia development, promotes leukemic stem cell expansion, and increases DNA damage through hyperactive NHEJ. Pharmacological NHEJ inhibition selectively impairs the survival of FA-deficient MLL-r leukemia cells, revealing a synthetic lethal vulnerability. Consistent with these findings, reduced FA pathway gene expression in patient samples correlates with poor prognosis and increased sensitivity to NHEJ inhibition. These findings identify NHEJ blockade as a promising targeted therapeutic strategy for high-risk MLL-r leukemia.

Mutations in the Fanconi anemia (FA) DNA repair genes predispose to leukemia and bone marrow failure. Here, the authors discover that the FA pathway controls MLL-rearranged leukemia progression via blocking nonhomologous end joining-mediated genomic instability.

Introduction

Mixed lineage leukemia (MLL) is a histone methyltransferase that deposits H3K4me marks and plays a key role during early development. Genetic rearrangements involving MLL gene give rise to MLL-rearranged (MLL-r) leukemia, accounting for ~10% of all acute leukemia^1–3, generally associated with poor prognosis⁴. More than 80 different fusion partner genes have been identified, although the majority of cases involve fusions with one of about six common partners, with MLL-AF4 (MA4; KMT2A-AFF1) and MLL-AF9 (MA9; KMT2A-MLLT3) accounting for most cases^2,3. MLL fusion proteins aberrantly activate transcriptional programs to promote leukemic stem cell (LSC) self-renewal and block differentiation, thereby driving disease progression⁵. These oncogenic drivers impose substantial replicative and transcriptional stress on hematopoietic stem and progenitor cells (HSPCs), contributing to malignant transformation⁶.

Oncogenic stress plays a crucial role in driving genomic instability and tumorigenesis. Activated oncogenes disrupt normal cellular processes, including DNA replication, leading to replication stress and generating DNA double-strand breaks (DSBs)⁷, repair mechanisms of which involve two primary pathways: error-free homologous recombination (HR) and error-prone non-homologous end joining (NHEJ)^8,9. Failure to repair damage results in cellular senescence or apoptosis, while improper repair may introduce mutations^10–12. Indeed, dysregulation of DNA damage response (DDR) pathways is associated with various human disorders¹³. DDR has been proposed as an evolutionary safeguard balancing blood regeneration with leukemia suppression. Inadequate repair of DNA damage in HSCs is closely linked to bone marrow (BM) failure and leukemogenesis¹⁴.

One of the most well-characterized examples of DDR deficiency syndromes is Fanconi anemia (FA), an inherited disorder characterized by hypersensitivity to DNA interstrand crosslinks (ICLs) and a strong predisposition to cancer^15,16. To date, 23 FA genes have been identified, including the breast cancer susceptibility 1 and 2 (BRCA1 and BRCA2)¹⁷. The FA pathway facilitates the repair of ICLs primarily through FANCD2/FANCI monoubiquitination and interaction with HR repair machinery, thereby ensuring replication fork stability and genomic integrity^18,19. Clinically, FA commonly progresses from BM failure to a pre-leukemic myelodysplastic syndrome (MDS) stage and eventually evolves to AML^20,21. Beyond its canonical role, dysfunction of FA pathway is known to increase NHEJ activity, which can be detrimental due to the high error rates. Our previous studies have demonstrated that Poly (ADP-ribose) polymerase 1 (PARP1), a regulator of NHEJ through its catalytic activity^22,23, coordinates with the FA pathway to prevent excessive NHEJ in stressed HSPCs²⁴. Despite the significance of both MLL-r and DDR deficiency in leukemogenesis, the interplay between these two critical pathways remains to be elucidated.

PARP inhibitors and NHEJ inhibitors are both under investigation in cancer research based on the principle of synthetic lethality (SL), where a single genetic defect is manageable, but co-occurrence of multiple genetic events results in cell death^25,26. Leveraging SL has emerged as a promising therapeutic strategy, particularly for targeting cancers driven by specific genetic mutations. In the context of DDR, SL is often exploited in cancer therapy by targeting genes involved in HR and NHEJ. The most well-established example of this approach is the use of PARP inhibitors in breast and ovarian cancer harboring mutations in BReast CAncer gene (BRCA), which impair the HR repair pathway²⁷. However, the clinical efficacy of SL-based therapies is limited in HR-proficient cancers, which represent the majority of tumors.

In this study, we dissect the critical interplay between FA pathway deficiency and MLL-r-driven leukemogenesis, uncovering that FA proteins are essential gatekeepers against NHEJ-mediated genomic instability. We demonstrate that loss of FA function creates an SL dependency on the NHEJ pathway, exposing an exploitable therapeutic vulnerability in FA-deficient MLL-r leukemia.

Results

FA gene haploinsufficiency accelerates MLL-AF9-driven leukemogenesis in mice

Patients with FA exhibit a markedly elevated risk of developing AML^20,21,28. Interestingly, MLL-r fusions, which account for ~10% acute leukemia, are rarely seen in FA AML^29–31. We screened a panel of FA cell lines, including FA patient-derived AML & lymphoblast cell lines (Supplementary Table 1) and primary BM samples from FA patients for the presence of MA9 fusion, most frequently associated with AML (Supplementary Fig. 1a). MA9 fusion was not detected in any of FA cell lines or FA AML BM samples (Supplementary Fig. 1b), suggesting that MA9 is excluded from the leukemogenic landscape of FA. Therefore, we crossed heterozygous FA mice (namely Fanca^+/− and Fancd2^+/−, both of which are key FA components) with the well-established MA9 transgenic leukemia mouse model, which reliably develops AML within 6–7 months of age. Notably, complete loss of Fanca or Fancd2 in MA9 background resulted in embryonic lethality. However, viable Fanca^+/−; MA9 and Fancd2^+/−; MA9 progenies were obtained at reduced frequencies (Supplementary Fig. 1c, d).

Phenotypic analysis of these Fancd2^+/−; MA9 and Fanca^+/−; MA9 mice revealed that FA haploinsufficiency exacerbates leukemia development, evidenced by early onset of splenomegaly (Fig. 1a, Supplementary Fig. 2a), paler femurs indicative of hematopoietic disruption (Fig. 1b, Supplementary Fig. 2b), elevated white blood cell (WBC) count and thrombocytopenia (Fig. 1c, Supplementary Fig. 2c). Flow cytometry and cytomorphological analysis demonstrated increased leukemia blasts both in peripheral blood (PB) and BM (Fig. 1d, Supplementary Fig. 2d), accompanied by pathological alterations in BM and splenic architecture (Fig. 1e, Supplementary Fig. 2e). Moreover, flow cytometry analysis showed significantly higher frequencies of leukemia stem cells (LSCs; defined as Lineage⁻ Sca1⁻ c-Kit⁺ CD34⁺ CD16/32⁺; Supplementary Fig. 2f) in Fancd2^+/−; MA9 (Fig. 1f, g) and Fanca^+/−; MA9 mice (Supplementary Fig. 2g, h) compared to their WT; MA9 counterparts. Importantly, Fanca^+/−; MA9 and Fancd2^+/−; MA9 mice died profoundly earlier compared to their WT; MA9 littermates (Fig. 1h, Supplementary Fig. 2i). These findings suggest a cooperative role between defective DDR pathway and MA9 fusion in accelerating leukemia progression.

Fig. 1. Fancd2 haploinsufficiency in MA9 drives clonal expansion in vitro and promotes aggressive leukemia development in vivo.

a Representative spleen images (Left; representative of n = 5, 6, and 6 mice/group) and quantifications of spleen weight (Right) from WT (n = 5 mice), MA9 transgenic (n = 6 mice), and Fancd2^+/−; MA9 mice (n = 6 mice). b Representative tibia and femur images from WT, MA9 transgenic, and Fancd2^+/−; MA9 mice (representative of n = 5, 6, and 6 mice/group). Yellow arrows indicate regions of pale, leukemic-infiltrated BM, characteristic of diffuse leukemic replacement. c WBC and PLT count of WT (n = 5 mice), MA9 (n = 6 mice), and Fancd2^+/−; MA9 mice (n = 6 mice). d Morphological analysis of peripheral blood (PB) smears (Upper, scale bar, 25 μm) and BM cytospin preparations (Lower, scale bar, 10 μm) from WT (n = 5 mice), MA9 (n = 6 mice), and Fancd2^+/−; MA9 mice (n = 6 mice). PB smears and BM cytospins were stained with Giemsa. Red arrows indicate leukemic blasts, while red asterisks denote myeloid cells at different maturation stages. e H&E staining of BM and spleen sections from WT, MA9, and Fancd2^+/−; MA9 mice (scale bar, 150 μm; representative of n = 4 mice/group). f Flow cytometric analysis of LSCs (defined as Lineage⁻ Sca1⁻ c-Kit⁺ CD34⁺ CD16/32⁺) in the BM (Upper) and spleen (Lower) of WT, MA9, and Fancd2^+/−; MA9 mice (representative of n = 4 mice/group). g Quantification of LSC frequencies in BM and spleen from WT, MA9, and Fancd2^+/−; MA9 mice (n = 4 mice /group). h Kaplan–Meier survival curves comparing overall survival among WT (n = 5 mice), MA9 (n = 10 mice), and Fancd2^+/−; MA9 mice (n = 7 mice). Mice used for in vivo analyses were 16–30 weeks of age. All quantitative data are represented as mean ± SEM from three independent experiments. Data were analyzed using one-way ANOVA with Tukey’s multiple comparisons. p values in Source Data. *p < 0.05, **p < 0.01, ***p < 0.001). Survival differences were assessed by the log-rank (Mantel–Cox) test.

Forced expression of MA9 in FA-deficient HSPCs promotes clonal expansion

To elucidate the mechanisms, we transduced Lineage⁻ cells (enriched for HSPCs) from three FA knockout mouse models (Fanca^−/−, Fancc^−/−, and Fancd2^−/−), with retroviral vector expressing MA9 (MSCV-IRES-eGFP-MA9³², Fig. 2a, Supplementary Fig. 3a–c). We found that although WT + MA9 cells grew much faster than FA + MA9 cells during the first two weeks of ex vivo culture, there was a significant expansion of FA + MA9 cells after 2 weeks culture (Fig. 2b, c). The expansion was associated with reduced apoptosis (Fig. 2d, Supplementary Fig. 3d) and enhanced cell proliferation (Fig. 2e, Supplementary Fig. 3e). Consistently, the size of the colonies generated by FA + MA9 cells were profoundly bigger than those from WT + MA9 cells (Fig. 2f). FA + MA9 cells also gave rise to significantly higher numbers of colonies in CFU assays compared to WT + MA9 cells (Fig. 2g). Additionally, this proliferative and clonogenic advantage persisted through secondary and tertiary CFU plating (Supplementary Fig. 3f), indicating augmented self-renewal and progenitor activity.

Fig. 2. Forced expression of MA9 in FA-deficient HSPCs promotes clonal expansion in vitro and accelerates leukemogenesis in vivo.

a Schematic of experimental design (Created in BioRender. Xu, J. (2026) https://BioRender.com/7ftl19u). WBMCs were isolated from 6–8-week-old WT mice and three FA-deficient mouse models (Fanca^−/−, Fancc^−/−, and Fancd2^−/−), then subjected to lineage depletion to enrich for HSPCs. WT or FA-deficient HSPCs were transduced with a retrovirus expressing MA9-GFP. Following transduction, cells were cultured in HSPC medium for growth curve analysis and downstream functional assays, including CFU and BMT experiments. b Proliferation curves of WT + MA9 and FA-deficient + MA9 HSPCs over 30 days post-transduction (n = 3 independent experiments/group). c Doubling time at day 20 post-transduction (n = 9 independent experiments/group). d Flow cytometric analysis of apoptosis at day 20 post-transduction (n = 3 independent experiments/group). e Flow cytometric analysis of Ki67 expression at day 20 post-transduction (n = 3 independent experiments/group). f Representative GFP⁺ colony images from the CFU assay for each group (representative of n = 6 independent experiments/group). Scale bar, 100 μm. g Representative colony images (left) and colony number quantification (right) from the 1st plating (n = 6 independent experiments/group). h Representative flow cytometry plots showing engraftment (CD45.2⁺) and circulating leukemic progenitor-like cells (CD45.2⁺c-Kit⁺) in recipient mice (representative of n = 3 mice/group). i Quantification of total engraftment in recipient mice at 1- and 2-months post-transplantation (n = 3 mice/group). j Quantification of circulating leukemic progenitor-like cells in recipient mice at 1- and 2-months post-transplantation (n = 3 mice/group). k Representative Wright-Giemsa-stained peripheral blood (PB, upper; scale bar, 50 μm) and bone marrow (BM, lower; scale bar, 50 μm) smears from recipient mice (representative of n = 3 mice/group). l Kaplan–Meier survival analysis of the recipient mice transplanted with WT + MA9 (n = 8 mice), Fanca^−/− + MA9 (n = 6 mice), Fancc^−/− + MA9 (n = 6 mice), or Fancd2^−/− + MA9 (n = 6 mice) HSPCs, monitored for 90 days. All quantitative data are represented as mean ± SEM. Data were analyzed by one-way ANOVA with Tukey’s multiple comparisons; Log-rank (Mantel–Cox) Test for survival curve. p values in Source Data. *p < 0.05, **p < 0.01, ***p < 0.001.

In addition, mice transplanted with MA9 LSCs deficient for Fanca, Fancc or Fancd2 developed aggressive leukemia with significantly shorter latency compared to those transplanted with WT + MA9 LSCs, evidenced by higher levels of donor-derived chimera (Fig. 2h, i) and increased frequencies of donor-derived circulating leukemic progenitor-like cells (CD45.2⁺ c-Kit⁺) in BM at 1-month and 2-month post-transplantation (Fig. 2j). FA + MA9 cells transplanted recipients also exhibited elevated numbers of leukemia blasts both in PB and BM (Fig. 2k), accompanied by markedly increased WBC counts and reduced platelet numbers (Supplementary Fig. 3g, h). Moreover, FA + MA9 recipient cohorts exhibited significantly compromised survival compared to WT + MA9 cells transplanted group (Fig. 2l), highlighting leukemogenic potential of MA9 in the context of FA deficiency.

DNA damage accumulation and genomic instability of FA-deficient MA9-transduced HSPCs

The phosphorylation status of Tyr142 in H2AX plays a critical role in the regulation of DDR. Failure to dephosphorylate H2AX-Tyr142 (H2AX-Y142) has been associated with DDR impairment and apoptosis^33–35. Compared to their WT + MA9 counterparts, FA + MA9 cells exhibited significantly compromised DDR activation, as evidenced by persistent phosphorylation of H2AX-Y142 and markedly reduced phosphorylation of Chk2-T68, p53-S15 and ATM-S1981 detected by Flow cytometry (Fig. 3a), all of which are established signatures of DDR^36–38. Following treatment with DNA interstrand crosslinker mitomycin C (MMC), WT + MA9 cells showed dephosphorylation of H2AX-Y142 accompanied by increased phosphorylation of Chk2, p53 and ATM indicative of an active DNA damage repair response. In contrast, FA + MA9 cells retained elevated H2AX-Y142 phosphorylation and exhibited persistently low levels of Chk2-T68, p53-S15 and ATM-S1981, consistent with defective DDR signaling (Fig. 3a). We further confirmed DNA damage accumulation using comet assay³⁹, which revealed remarkably longer tail moments in FA + MA9 cells in comparison to WT + MA9 cells (Fig. 3b, c). In addition, FA + MA9 cells demonstrated increased chromosomal instability in Chromosomal breakage assays (Fig. 3d).

Fig. 3. DNA damage accumulation and genomic instability of FA-deficient MA9-transduced HSPCs.

a Representative flow cytometry histograms and quantification of mean fluorescence intensity (MFI) for H2AX-Y142, Chk2-T68, p53-S15 and ATM-S1981 in WT + MA9, Fanca^−/− + MA9, Fancc^−/− + MA9, and Fancd2^−/− + MA9 cells treated with vehicle or MMC (n = 3 independent experiments/group). b Representative comet assay images for DNA damage detection (representative of n = 3 independent experiments). Scale bar, 50 μm. c Quantification of comet tail moment (tail length × %DNA in tail). For each of n = 3 independent experiments, the mean tail moment was calculated from 100 cells/group; data represent mean ± SEM of the 3 independent experiment means). d Representative metaphase spread images showing chromosomal aberrations (breaks, dicentric chromosomes, fragments; red arrows) (representative of n = 3 independent experiments). Scale bar, 10 μm. Representative immunofluorescence images (e) and quantification (f) of γH2AX and RAD51/53BP1 co-localization in FA-deficient MA9-transduced HSPCs. Co-localized foci were defined as spatially overlapping γH2AX and RAD51 or 53BP1 puncta within the same nucleus. For each of n = 3 independent experiments, the mean percentage of co-localized foci was calculated from 25 cells/condition; data represent mean ± SEM of the 3 independent experiment means. Scale bar, 5 μm. g Western blot of NHEJ factors (KU70, KU80), DNA damage marker (γH2AX), and repair proteins (RAD51, 53BP1) in WT and FA-deficient MA9-transduced HSPCs. The samples derive from the same experiment but different gels for KU70/KU80, another for γH2AX, another for RAD51, and another for 53BP1 were processed in parallel. Band intensities quantified via ImageJ, normalized to β-actin, with WT + MA9 set to 1 per antibody; quantification reflects one representative experiment, blots representative of n = 3 independent experiments. All quantitative data are represented as mean ± SEM from three independent experiments. Statistical comparisons were performed using two-way ANOVA with Tukey’s multiple comparisons (a) and one-way ANOVA with Tukey’s multiple comparisons (c, f). p values in Source Data. *p < 0.05; **p < 0.01; ***p < 0.001.

DSBs can be predominantly repaired by two major pathways: error-free HR and error-prone NHEJ^8,9. To assess pathway preference for DSB repair in MA9-transduced cells, we performed immunofluorescence staining for RAD51 and 53BP1 foci, surrogates for HR and NHEJ, respectively²⁴. We found that MA9-HSPCs deficient for Fanca, Fancc or Fancd2 preferably utilized NHEJ for repair, while WT + MA9 cells exhibited significantly higher HR activity (Fig. 3e, f). This shift in DDR pathway usage was not attributed to RAD51 and 53BP1 protein level changes (Fig. 3g). FA-deficient MA9 cells exhibited markedly elevated levels of phosphorylated H2AX at Ser 139 (known as γH2AX, an established DSB marker⁴⁰) relative to WT + MA9 cells, accompanied by moderately increased levels of the core NHEJ components KU70 and KU80 (Fig. 3g). These findings indicate accumulation of DNA damage and a compensatory activation of error-prone NHEJ in the absence of functional FA pathway signaling.

Inhibition of NHEJ suppresses clonal expansion of FA-deficient MA9 HSPCs in vitro and in vivo

We then treated WT + MA9, Fanca^−/− + MA9, Fancc^−/− + MA9 and Fancd2^−/− + MA9 cells with NU7026, a selective inhibitor of DNA-PKcs, a critical component of the NHEJ pathway^24,41. Compared to WT + MA9, NU7026 treatment significantly limited the expansion of FA + MA9 circulating leukemic progenitor-like cells (Fig. 4a, Supplementary Fig. 3i), accompanied by increased apoptosis (Fig. 4b, c) and reduced cell proliferation (Fig. 4d, e). Additionally, DNA-PKcs blockage significantly reduced the leukemia burden in the recipients transplanted with FA + MA9 cells, shown by decreased donor-derived chimerism and lower frequencies of leukemic cells (Fig. 4f–h). NU7026 treated FA + MA9 cohort also exhibited delayed the onset of splenomegaly (Supplementary Fig. 3j), reduced WBC counts (Supplementary Fig. 3k), partial restoration of platelet levels (Supplementary Fig. 3l) and a decrease in leukemic blasts in the PB (Supplementary Fig. 3m) and BM (Supplementary Fig. 3n). Importantly, NU7026 treatment markedly improved overall survival of mice transplanted with FA + MA9 cells (Fig. 4i). Together, these results demonstrate that FA-deficient MA9-transformed HSPCs are highly dependent on NHEJ for survival and expansion, and that targeting the NHEJ pathway not only reduces leukemic burden but also selectively impairs their clonal outgrowth.

Fig. 4. Inhibition of NHEJ suppresses clonal expansion of FA-deficient MA9 HSPCs in vitro and in vivo.

a Proliferation curves of WT and FA-deficient (Fanca^−/−, Fancc^−/−, Fancd2^−/−) MA9 HSPCs. Cells were monitored for expansion over 30 days post-transduction. On day 15, a subset of each group was treated with NU7026 (n = 3 independent experiments /group). Flow cytometric analysis (b) and quantification (c) of apoptosis was performed on day 20 using Annexin V/DAPI staining, comparing untreated and NU7026-treated conditions (n = 3 independent experiments/group). d, e Ki67 staining was used to assess cell cycle status and proliferative index on day 20 post-transduction. Representative flow plots (d) and quantification (e) of Ki67⁺ populations are shown (n = 3 independent experiments/group). f Diagram of experimental design assessing FA deficiency and NHEJ inhibition on MA9-driven leukemogenesis in vivo (Created in BioRender. Xu, J. (2026) https://BioRender.com/k5r1nm4). MA9-transduced HSPCs from WT or FA-deficient (Fanca^−/−, Fancd2^−/−) mice (2-week ex vivo cultured) were transplanted into lethally irradiated BoyJ recipients via tail vein injection (Day 0). After 2 weeks of engraftment, mice were randomized into vehicle or treatment groups; treatment group received NU7026 (20 mg/kg, i.p.) once daily for 4 consecutive days, followed by longitudinal monitoring for leukemia progression. g, h Flow cytometry of PB from recipient mice. Engraftment was assessed as %CD45.2⁺ donor cells; circulating leukemic progenitor-like cells were identified as CD45.2⁺c-Kit⁺. Representative plots (g) and quantification (h) of CD45.2⁺ total donor cells and CD45.2⁺c-Kit⁺ cells (n = 3 independent mice/group). i Kaplan–Meier survival analysis of recipient mice receiving WT (vehicle n = 4 mice, NU7026 n = 4 mice), Fanca^−/− + MA9 (vehicle n = 3 mice, NU7026 n = 3 mice), or Fancd2^−/− + MA9 (vehicle n = 4 mice, NU7026 n = 3 mice) HSPC transplant after NU7026 treatment, monitored for 90 days post-transplant. All quantitative data are represented as mean ± SEM. HSPCs for MA9 transduction were harvested from 6 to 8-week-old WT or FA-deficient mice. Data were analyzed by two-way ANOVA with Tukey’s multiple comparisons; Log-rank (Mantel–Cox) Test for survival curve. p values in Source Data. *p < 0.05, **p < 0.01, ***p < 0.001.

Targeting NHEJ-dependent vulnerabilities in FA-knockdown human MLL-r leukemia cells

To directly assess DNA repair pathway choice, HR and NHEJ reporter assays⁴² were performed in MLL-r and non-MLL-r cell lines. FANCD2 knockdown significantly reduced HR and increased NHEJ efficiency in MLL-r THP-1, and MV4-11, but not in non-MLL-r HL-60 cells (Fig. 5a, Supplementary Fig. 4a), confirming that FA pathway loss selectively shifts DSB repair toward error-prone NHEJ in MLL-r leukemia cells. We next evaluated the effect of NHEJ inhibition on these cells. FANCD2-knockdown THP-1 and MV4-11 cells became significantly more sensitive to NU7026 treatment, as evidenced by a marked reduction in IC₅₀ values (Fig. 5b). In contrast, NHEJ inhibitors have only a modest effect on HL-60 cells, which lack MLL-r. Consistently, FANCD2-knockdown THP-1 and MV4-11 cells showed significantly reduced proliferation upon NHEJ inhibition (Fig. 5c), associated with increased apoptosis (Fig. 5d) and a decline in the proportion of Ki67-positive cells (Fig. 5e). Consistent results were obtained with NU1025, an alternative NHEJ inhibitor that targets PARP1 (Supplementary Fig. 4c). In addition, knockdown of another FA component, FANCA, in THP-1, MV4-11 and HL-60 cells yielded comparable outcomes (Supplementary Figs. 4b, 4d, 5a–d). Notably, increased sensitivity to NHEJ inhibitor was not observed in FANCD2- or FANCA-knockdown OCI-AML3 or U937 cells, two MLL-r negative human monocytic AML lines known to share some transcriptional features with MLL-r leukemia (Supplementary Fig. 4e–h)^43,44.

Fig. 5. FANCD2 knockdown sensitizes MLL-r human leukemia cells to NHEJ inhibition.

a HR and NHEJ repair efficiencies in MLL-r (THP-1, MV4-11) and non-MLL-r (HL-60) cell lines transfected with siCtrl or siFANCD2, as measured by DR-GFP and EJ5-GFP reporter assays, respectively. GFP⁺ cells were quantified within the DsRed⁺ gate by flow cytometry (n = 3 independent experiments/ group). b Dose-response curve illustrating the effect of NU7026 treatment on the viability of MLL-r (THP-1 and MV4-11) or non-MLL-r AML (HL-60) cells with and without FANCD2 knockdown (siFANCD2). IC50 values were calculated for each group based on dose-response data (n = 3 independent experiments/ group). c Cell growth curve of MLL-r or non-MLL-r leukemia cells with and without FANCD2 knockdown treated with NU7026 (n = 3 independent experiments/group). d Quantification of apoptosis. Annexin V/DAPI staining in THP-1, MV4-11, and HL-60 cells with or without FANCD2 knockdown following NHEJ inhibitor treatment (n = 3 independent experiments/group). e Quantification of cell cycle analysis. Flow Cytometry was performed using Ki67/DAPI staining that was conducted to evaluate proliferation status (n = 3 independent experiments/group). f IC50 assays for RS4;11 and MOLT-4 cells (n = 3 independent experiments/group). g Cell growth curve of RS4;11 and MOLT-4 cells with or without FANCD2 knockdown and NU7026 treatment. Cells were cultured for 6 days, and viable cell numbers were counted daily using trypan blue exclusion (n = 3 independent experiments/group). h Quantification of apoptosis analysis. Flow cytometry was performed by Annexin V/DAPI staining 48 h after NU7026 treatment (n = 3 independent experiments/ group). i Quantification of cell cycle analysis. Ki67 and DAPI staining 48 h post-treatment. FANCD2 knockdown combined with NU7026 treatment led to a reduction in Ki67⁺ proliferating population (n = 3 independent experiments/group). j HR and NHEJ repair efficiencies in MLL-r (RS4;11) and non-MLL-r (MOLT-4) cells (n = 3 independent experiments/group). All quantitative data are represented as mean ± SEM. Data were analyzed by two-way ANOVA with Tukey’s multiple comparisons. p values in Source Data. **p < 0.01, ***p < 0.001.

MLL-r occurs in both AML and ALL¹. Knockdown of FANCD2 in MLL-AF4⁺ RS4;11 ALL cells (Supplementary Fig. 4a) also significantly impaired cell proliferation in response to NU7026 treatment, whereas no such effect was observed in non-MLL-r ALL cell line, MOLT-4 (Fig. 5f, g, Supplementary Fig. 5e, f). Additionally, FANCD2- and FANCA-knockdown in RS4;11 cells led to a significant increase in Annexin V⁺ apoptotic cells upon NHEJ inhibition (Fig. 5h, Supplementary Fig. 4b, Supplementary Fig. 5g), along with a marked decrease in Ki67-positivity (Fig. 5i, Supplementary Fig. 5h), reduced HR and increased NHEJ efficiency (Fig. 5j). These were not seen in non-MLL-r MOLT-4 cells. Together, these findings indicate that NHEJ inhibitors selectively impair the survival and proliferation of both AML and ALL cells harboring MLL-r when FA pathway components are reduced.

NHEJ inhibition reduces leukemia burden in primary FA^Low MLL-r leukemia cell transplant mice

We also conducted a comprehensive analysis using the AML dataset from the web-based Kaplan–Meier Plotter database. Kaplan–Meier analysis revealed that lower FA gene expression correlated with significantly worse overall survival in patients with monocytic AML, a subtype strongly associated with MLL rearrangements (Fig. 6a)^45,46. In contrast, higher expression of genes involved in the NHEJ pathway, such as XRCC6 (KU70), XRCC5 (KU80), PRKDC (DNA-PKcs), and TP53BP1 (53BP1), also correlated with poor prognosis in this AML subset (Supplementary Fig. 6a), supporting a pathogenic role for NHEJ dysregulation in FA-deficient MLL-r leukemia.

Fig. 6. FANCD2^Low primary human MA9-AML patient cells are sensitive to NHEJ inhibition.

a Lower FA gene expression correlates with poor AML survival. Using the AML dataset from the Kaplan–Meier Plotter database, overall survival was analyzed relative to mRNA expression of FANCA (n = 1608 patients), FANCC (n = 255 patients), FANCD2 (n = 127 patients), ERCC4 (n = 255 patients), and BRCA2 (n = 255 patients). Patients were stratified into high/low expression groups by median value. b Schematic showing sensitivity of FANCD2^Low and FANCD2^High MA9-AML patient cells to NHEJ inhibition in vivo (Created in BioRender. Xu, J. (2026) https://BioRender.com/rtlv9do). Primary MA9-AML samples, stratified by qPCR-measured FANCD2 expression, were transplanted into sublethally irradiated 8–12-week-old NSGS mice via BMT. Two weeks post-transplant, mice received NU7026 or vehicle, and leukemia progression was monitored to evaluate therapeutic impact relative to FANCD2 status. c Representative flow plots showing human CD45⁺ (hCD45) versus mouse CD45⁺ (mCD45) cell chimerism in the BM (representative of n = 3 mice/group). d Quantification of human CD45⁺ cell engraftment in the recipient mice (n = 3 mice/group). e Representative flow cytometry plots showing donor-derived human CD34⁺CD38⁻ leukemia stem/progenitor cells within the human CD45⁺ compartment. CD34⁺CD38⁻ gates were defined using matched-fluorochrome IgG isotype controls (representative of n = 3 mice/group). f Quantification of human CD34⁺CD38⁻ leukemia stem/progenitor cells within the human CD45⁺ compartment (n = 3 mice/group). g Representative flow plots showing donor-derived human CD33⁺CD19⁻ myeloid blasts in the human CD45⁺ population (representative of n = 3 mice/group). h Quantification of human CD33⁺CD19⁻ myeloid blasts in the human CD45⁺ population (n = 3 mice/ group). i Kaplan–Meier survival analysis of FANCD2^Low or FANCD2^High MA9-AML patient cells in response to NHEJ inhibition. Kaplan–Meier survival curves showing the overall survival of NSGS recipient mice transplanted with FANCD2^Low (vehicle n = 4 mice, NU7026 n = 3 mice) or FANCD2^High (vehicle n = 3 mice, NU7026 n = 3 mice) MA9-AML patient-derived CD34⁺ cells and treated with NU7026 or vehicle, monitored for up to 90 days post-transplant. Data are mean ± SEM. Statistical significance: two-way ANOVA with Tukey’s multiple comparisons (d, f, h) and the log-rank (Mantel–Cox) test (i). p values in Source Data. *p < 0.05, ***p < 0.001.

To functionally validate these findings in vivo, we transplanted primary MLL-r AML samples harboring t(9;11) or t(11;19) translocations into humanized NSGS mice⁴⁷, a well-established model for human leukemia xenografts due to their superior support for human hematopoiesis (Fig. 6b, Supplementary Fig. 6b, c). We found that FANCD2^Low MLL-r cells harboring t (9;11) translocation established significantly higher levels of human engraftment in NSGS mice compared to FANCD2^High cells. NU7026 treatment significantly reduced overall human leukemic burden in FANCD2^Low transplanted mice (Fig. 6c–h). Consequently, NU7026 treatment significantly prolonged survival of FANCD2^Low xenograft recipients (Fig. 6i), whereas only marginal effects were observed in FANCD2^High transplant cohorts. These findings were further confirmed in a second cohort of FANCD2^Low MLL-r primary AML cells carrying the t (11;19) translocation, which showed similar sensitivity to NHEJ inhibition in vivo (Supplementary Fig. 6d–j). Collectively, these results demonstrate that MLL-r AML cells with low FA gene expression are highly dependent on the NHEJ repair pathway and are preferentially targeted by NHEJ inhibition.

To further substantiate our findings, we conducted an in-depth analysis of multiple publicly available transcriptomic datasets and identified a significant association between the t (9;11) MLL-rearrangement and reduced expression of key FA pathway genes, including BRCA1, FANCA, FANCC, and FANCD2 compared to non-MLL-r controls (Supplementary Fig. 7a). These data suggest that a distinct subset of MLL-r AML, particularly those with monocytic differentiation, may have compromised FA pathway activity. To assess the functional consequences of reduced FA gene expression, we leveraged the CTRPv2 (Cancer Therapeutics Response Portal version 2), which integrates gene expression with drug response predictions⁴⁸. This analysis revealed a strong inverse correlation between FA gene expression and predicted sensitivity to the most accepted PARP inhibitor, Olaparib in MLL-r monocytic AML samples, in two large and independent cohorts: TARGET AML (Supplementary Fig. 7b) and BEAT AML (Supplementary Fig. 7c). Together, these findings strongly support the notion that MLL-r leukemias with low activity in the FA pathway rely more on error-prone NHEJ for repair, making them particularly vulnerable to drugs that block NHEJ and PARP pathways.

NHEJ inhibitor synergizes with cytotoxic and epigenetic therapies in FA^Low MLL-r leukemia

Next, we treated WT + MA9, Fanca^−/− + MA9, and Fancd2^−/− + MA9 cells with NU7026 in combination with either DNA-damaging agent cytarabine (AraC) or the Menin inhibitor Ziftomenib (KO-539), both of which are currently used in clinical for MLL-r leukemia^49,50. FA-deficient MA9 cells (Fanca^−/− and Fancd2^−/−) exhibited significantly enhanced sensitivity to both NU7026 + AraC or NU7026 + KO-539 combinations, with marked reductions in cell viability compared to monotherapy. In contrast, WT + MA9 cells demonstrated only modest responses, indicating that FA pathway deficiency confers a selective vulnerability to dual targeting of NHEJ and cytotoxic/epigenetic stress (Fig. 7a). Furthermore, inhibition of NHEJ synergized with either AraC or the Menin inhibitor KO-539 to significantly reduce FA-deficient leukemia-initiating cell (LIC) frequency detected by in vitro leukemic long-term culture-initiating cell (L-LTC-IC) assay (Fig. 7b)^51,52. Consistently, NHEJ inhibitor also exhibited strong synergistic responses with either AraC or KO-539 in FANCD2 knockdown human MLL-r cells (THP-1, MV4-11 and RS4;11), but not non-MLL-r cell lines, such as HL-60 and MOLT-4 cells (Fig. 7c). We then transplanted FA-deficient MA9-transduced HSPC (Fanca^−/− + MA9, Fancd2^−/− + MA9) into lethally irradiated BoyJ recipients. Combo therapy significantly reduced leukemic burden in mice compared to vehicle or single-agent controls (Fig. 7d, e) and extended overall survival in recipient mice (Fig. 7f), indicating therapeutic benefit beyond cytoreduction.

Fig. 7. NHEJ inhibitor exhibits synergistic effect with AraC and Menin inhibitor in killing MLL-r leukemia.

a WT + MA9, Fanca^−/− + MA9, and Fancd2^−/− + MA9 HSPCs were treated with vehicle, NU7026 (10 µM), KO-539 (500 nM), NU7026 + KO-539, AraC (1 µM), or NU7026 + AraC. Cell viability was assessed on Day 4 using CellTiter-Glo luminescent assay (n = 5 independent experiments/group). b L-LTC-IC frequencies of MA9-transduced HSPCs under the same treatment conditions. NU7026 combined with AraC or KO-539 synergistically reduced LSC frequency versus either agent alone (n = 3 independent experiments/group). c Human leukemia cell lines (THP-1, MV4-11, RS4;11, HL-60, and MOLT-4), with or without FANCD2 knockdown, were treated under identical drug conditions (n = 8 independent experiments/group for THP-1, MV4-11, and RS4;11; n = 5 independent experiments/group for HL-60 and MOLT-4). d, e NU7026 synergizes with AraC and KO-539 against FA + MA9 mouse cells in vivo. BoyJ mice transplanted with MA9-transduced WT, Fanca^−/−, or Fancd2^−/− HSPCs were treated 2 weeks post-transplant with vehicle, NU7026 (20 mg/kg), KO-539 (30 mg/kg), NU7026 + KO-539, AraC (50 mg/kg), or NU7026+AraC. At 2 months, BM engraftment (CD45.2⁺, d) and leukemic blast-like cells (CD45.2⁺c-Kit⁺, e) were quantified by flow cytometry (n = 4 mice/group). f Survival analysis of mice from (d, e), monitored for 90 days post-transplant (n = 4, 4, 3, 3, 3, 3 mice/treatment group). g–i Human CD34⁺ cells from MA9-AML patients were transplanted into sublethally irradiated NSGS mice; engraftment confirmed by hCD45. Two weeks post-transplant, mice received vehicle, NU7026, KO-539, NU7026 + KO-539, AraC, or NU7026 + AraC. Engraftment (g CD45), malignant HSPCs (h CD45⁺CD34⁺CD38⁻), and myeloid cells (i CD45⁺CD33⁺CD19⁻) quantified at 2 months (n = 3 mice/group). j Survival analysis of NSGS mice transplanted with FANCD2^High or FANCD2^Low primary MA9-AML cells under the same treatment regimens (n = 4, 4, 3, 3, 3, 3 mice/group), monitored for 90 days. Data are mean ± SEM. One-way ANOVA with Tukey’s multiple comparisons; log-rank (Mantel–Cox) test for survival. p values in Source Data. *p < 0.05, **p < 0.01, ***p < 0.001.

To evaluate translational relevance, we transplanted FANCD2^High and FANCD2^Low MLL-r AML primary patient samples into humanized NSGS mice followed by either NU7026 + AraC or NU7026 + KO-539 injection. In alignment with murine data, FANCD2^Low patient-derived leukemia exhibited markedly greater reductions in human engraftment (Fig. 7g), CD34⁺ -enriched HSPC fractions (Fig. 7h), malignant myeloid cells (Fig. 7i), and significantly improved overall survival (Fig. 7j) in response to combo therapy. Similar therapeutic effects were observed in NSGS mice transplanted with a primary MLL-r ALL patient sample (Supplementary Fig. 8a–e). These results demonstrate that the synergistic effect between NHEJ inhibition and genotoxic or epigenetic therapy is specific to MLL-r leukemias, particularly in the context of FA pathway deficiency.

NHEJ inhibitor combined with AraC or Menin inhibitor does not affect long-term normal hematopoiesis

We then investigated the impact of these regimens on normal human hematopoiesis by transplanting CD34⁺ human cord blood cells into sublethally irradiated NSGS mice⁵¹, followed by different treatments (Supplementary Fig. 9a, b). Although we observed a transient reduction in the total number of human CD45⁺ cells 1-month post-treatment across treatment groups that included AraC or KO-539, with or without NHEJ inhibition (Supplementary Fig. 9c), no significant difference in total human engraftment was found at 3 months post-treatment for all treatment groups compared to vehicle controls. Furthermore, while AraC + NU7026 and KO-539 + NU7026 treatment caused a decrease in human HSPCs (hCD45⁺CD34⁺CD38⁻) and more differentiated HPCs (hCD45⁺CD34⁺CD38⁺) at 1-month post treatment, comparable numbers of HSPCs and HPCs were observed at 3-month post-treatment in the transplanted recipients of all groups (Supplementary Fig. 9c, d). These results demonstrate that therapeutic regimens combining NHEJ inhibition with either AraC or Menin inhibition do not adversely affect the long-term function of normal human HSCs.

The FA pathway restricts aberrant NHEJ factor recruitment via the N-terminus of MLL in MLL-r leukemia

Leukemogenic MLL fusion proteins uniformly retain the N-terminal portion of MLL (MLL-N), known to mediate interactions with various DDR components^53–56, while lacking the C-terminal SET domain responsible for histone methyltransferase activity. To elucidate the mechanism by which the FA pathway modulates DNA repair pathway choice in MLL-r leukemia, we examined whether FA gene expression levels affect the interaction between MLL-N and core NHEJ machinery, specifically KU70, a key DNA end-binding protein. We observed low baseline interaction between KU70 and MLL-N in WT THP-1 and MV4-11 cells in proximity ligation assay (PLA). However, FANCD2 knockdown in these cells resulted in a substantial increase in KU70-MLL-N PLA signals (Fig. 8a, b), suggesting enhanced recruitment of NHEJ factors in the absence of FANCD2. This effect was not observed in non-MLL-r HL-60 cells (Supplementary Fig. 10a, b). Similarly, FANCD2 knockdown also led to a moderate increase in KU70-MLL-N association in MLL-r ALL cell line RS4;11 (Fig. 8a, b) but not in non-MLL-r ALL cell line MOLT-4 (Supplementary Fig. 10c, d). These findings support a model in which the FA pathway suppresses aberrant NHEJ engagement at sites associated with MLL fusion proteins.

Fig. 8. FA proteins prevent NHEJ factor recruitment-mediated by N terminal of MLL-r leukemia.

a Representative PLA images of KU70–MLL-N interactions in MLL-r AML (THP-1, MV4-11) and MLL-r ALL (RS4;11) cell lines transfected with siCtrl or siFANCD2 (representative of n = 3 independent experiments). Cells were stained with anti-KU70 and anti-MLL-N antibodies, followed by PLA to detect protein-protein interactions. Scale bar, 10 μm. b Quantification of PLA puncta per nucleus from (a), comparing siCtrl and siFANCD2 conditions (for each of n = 3 independent experiments, the mean number of PLA puncta/cell was calculated from 20 cells). c IP of KU70/KU80 binding to MLL-N under FANCD2-proficient vs. -deficient conditions. Lysates from THP-1, MV4-11, and RS4;11 cells (siCtrl or siFANCD2) were immunoprecipitated with anti-MLL-N (or IgG control), then immunoblotted for KU70 and KU80. Band intensities normalized to β-actin, with siCtrl/IgG (lane 1) set to 1 per cell line/antibody. The samples derive from the same experiment but different gels for KU70, KU80 and β-actin, and another for FANCD2 were processed in parallel. Blots representative of n = 3 independent experiments. d Structural overlay of AlphaFold3-predicted models of KU70 (purple), KU80 (red), and FANCD2 (yellow) onto KU70, KU80, and MLL-N1400 (green). FANCD2 occupies a region overlapping MLL-N1400’s interaction domain, indicating both cannot occupy this space simultaneously. A horizontal (f) and vertical (e) view of a model of KU70 (purple), KU80 (red), MLL-N1400 (green), with and without FANCD2 (yellow). FANCD2 forces MLL-N1400 from a more compact configuration to a more spread-out configuration as well at the orientation of its interaction domain. g PAE charts for FANCD2–MLL-N, KU70–KU80–MLL-N1400, and KU70–KU80–MLL-N1400–FANCD2 complexes. FANCD2 and KU70/80 each show moderate binding to MLL-N1400 individually, but this is reduced when both are present, as they instead preferentially bind each other. h Working model: FA pathway suppresses NHEJ in MLL-r leukemia by limiting KU70/KU80 recruitment to the MLL N-terminal region; absence of this axis in non-MLL-r lines suggests a subtype-specific mechanism protecting genome integrity in MLL-driven leukemogenesis (Created in BioRender. Xu, J. (2026) https://BioRender.com/rtlv9do). Data are mean ± SEM. Unpaired Student’s t test. p values in Source Data. *p  <  0.05; **p  <  0.01, ***p  <  0.001.

Additionally, FANCD2 knockdown resulted in a robust increase in co-precipitation of KU70/KU80 with MLL-N in MLL-r cells (THP-1, MV4-11 and RS4;11) (Fig. 8c), but not in MLL-r-negative HL-60 and MOLT-4 cells (Supplementary Fig. 10e, f), further emphasizing the specificity of this interaction to MLL-r contexts. In contrast, FANCD2 overexpression did not affect KU70/KU80 interaction with MLL-N (Supplementary Fig. 10g, h), confirming that FA deficiency is the primary driver of aberrant KU70/KU80 recruitment in MLL-r leukemia. Together, these results demonstrate that FANCD2 functions as a critical suppressor of error-prone NHEJ activity in MLL-r leukemia by limiting the recruitment of KU70/KU80 to the MLL N-terminal region. The absence of this regulatory axis in non-MLL-r cells suggests a subtype-specific mechanism by which the FA pathway safeguards genome integrity in MLL-driven leukemogenesis across both AML and ALL contexts.

Given that the N-terminal 1400 residues of MLL (MLL-N1400) are retained in virtually all MLL-r fusion proteins irrespective of the fusion partner⁵⁷, we modeled this shared region to capture interactions relevant across MLL-r leukemia subtypes. Comparative analysis using AlphaFold3 revealed a notable overlap between the predicted interaction interfaces of FANCD2-MLL-N1400 and KU70/KU80-MLL-N1400 complexes, suggesting potential competitive binding or steric interference (Fig. 8d). Superimposition of the KU70/KU80-MLL-N1400 model with the FANCD2-KU70/KU80-MLL-N model revealed a broader conformational spread of MLL-N1400 domain in the ternary complex (Fig. 8e, f). Notably, interactions inferred from the Predicted Aligned Error (PAE) maps between MLL-N and FANCD2, as well as MLL-N1400 and KU70/KU80 in the binary complexes, were absent in PAE analysis of the ternary complex (MLL-N1400, FANCD2, and KU70/KU80) (Fig. 8g). Instead, PAE maps of the ternary complex revealed putative interactions between FANCD2 and KU70/KU80, indicative of mutual sequestration or occlusion effects (Fig. 8g). Collectively, these findings support a model in which the FA pathway acts as a critical suppressor of error-prone NHEJ in MLL-r leukemia by limiting the accessibility of KU70/KU80 to the MLL N-terminal domain. This regulatory mechanism appears to be specific to MLL-r cells and is absent from non-rearranged counterparts, suggesting that the FA pathway enforces genomic stability through a context-dependent suppression of aberrant DNA repair engagement in MLL-driven leukemogenesis (Fig. 8h).

Discussion

Our study delineates a previously unappreciated tumor-suppressive role of the FA DNA repair pathway in the pathogenesis of MLL-r leukemia. Through comprehensive genetic and pharmacological analysis, we demonstrate that: (1) Genetic ablation of FA pathway components accelerates leukemic transformation in MLL-r models; (2) MA9 leukemic cells deficient in FA genes accumulate excessive DNA damage and genomic instability, driven by aberrant hyperactivation of NHEJ; (3) Targeted pharmacologic inhibition of NHEJ selectively eliminates MLL-r leukemic cells both in murine and human models; (4) NHEJ inhibitors synergize with AraC or Menin inhibitors, enhancing anti-leukemic efficacy both in vitro and in vivo; (5) Mechanistically, FA proteins constrain aberrant recruitment of NHEJ components to MLL-N, thereby attenuating NHEJ-mediated mutagenesis and impeding leukemic progression.

MLL-r occurs in ~10% of acute leukemia cases and is characterized by early onset, poor response to therapy, and dismal clinical outcomes^58–60. While biallelic loss of FA genes is embryonic lethal in MLL-r mice, even partially FA gene loss (haploinsufficiency) significantly exacerbates leukemogenesis, highlighting the indispensable role of FA pathway in maintaining genomic integrity under oncogenic stress. These findings align with the striking clinical rarity of MLL-r AML in FA patients^29,30, suggesting that a functional FA pathway actively suppresses malignant transformation driven by MLL fusions. We further demonstrate that ectopic expression of MA9 in FA-deficient HSPCs leads to robust in vitro clonal expansion and dramatically shortened leukemia latency in vivo. This synergistic effect is tightly associated with accumulation of DSBs and heightened chromosomal instability, establishing a direct connection between defective DNA repair and enhanced leukemic potential.

At molecular level, our findings reinforce and extend the emerging paradigm that the FA pathway not only facilitates HR but also actively suppresses aberrant NHEJ activity. Prior studies from our group and others have implicated PARP1, a modulator of DDR pathway choice^61,62 as a critical cofactor, that coordinates with FA proteins to restrain excessive NHEJ in stressed HSPCs^24,63. Here, we build on this framework, providing direct evidence that FA haploinsufficiency drives leukemogenesis through dysregulated and mutagenic NHEJ activity. Our data identify NHEJ not merely as a bystander but as a pivotal effector of leukemic proliferation in the context of FA-deficient MLL-r leukemia, positioning NHEJ as both a mechanistic driver and a promising therapeutic target in this disease setting.

Indeed, we show that pharmacological inhibition of NHEJ^22,23, selectively eliminates FA-deficient MLL-r leukemic cells across multiple species and experimental models. This selective vulnerability is further supported by a strong correlation between MLL-r leukemia and downregulated FA pathway genes, as well as predicted sensitivity to Olaparib. These findings substantially extend the therapeutic scope of PARP inhibitors beyond HR-deficient cancers, encompassing a wider spectrum of malignancies, particularly hematological cancers. Notably, combining NHEJ inhibitors with AraC or Menin inhibitors^49,50, results in synergistic cytotoxicity against leukemic cells, while minimizing off-target effects on normal hematopoiesis. These results uncover a distinct therapeutic window in which actively proliferating leukemic cells can be effectively targeted without compromising long-term hematopoiesis. Given the central role of LSCs in therapeutic resistance and disease relapse, our data provides compelling evidence that NHEJ-targeted strategies can selectively deplete proliferative leukemic populations while preserving the quiescent HSPC compartment, offering a rational and potentially durable approach to eradicating disease at its root.

Under physiological conditions, MLL functions as a bipartite complex comprising MLL-N and MLL-C. However, oncogenic MLL fusion proteins retain only MLL-N, which has been directly implicated in interactions with DDR regulators including ATM, ATR, and Menin^53–55. Through mechanistic interrogation, we demonstrate that FA pathway plays a critical role in restraining aberrant recruitment of NHEJ factors to MLL-N domain of fusion proteins. This discovery provides the molecular explanation for the heightened NHEJ reliance observed in FA-deficient contexts. Our findings further establish that FA proteins actively suppress mutagenic DNA repair at sites of oncogene-induced replication stress. Loss of FA function, therefore, permits unrestrained engagement of error-prone NHEJ machinery, promoting genomic instability and accelerating leukemogenic progression.

In summary, our findings identify a previously unappreciated role of the FA pathway in modulating NHEJ during MLL-r leukemogenesis. These results not only provide mechanistic insights into the pathophysiology of FA and MLL-r AML but also establish the therapeutic potential of targeting NHEJ to selectively eliminate MLL-r leukemic cells while preserving normal hematopoiesis.

Methods

Mice

Fanca^+/−; MA9 or Fancd2^+/−; MA9 mice were obtained by crossing Fanca^+/− or Fancd2^+/− mice^64–66 with MA9 mice (JAX; Stock #: 009079)³². All mouse lines were maintained on a C57BL/6 background.16–30-week-old MA9 mice were used for experiments. BoyJ mice aged 8–12 weeks were used as transplantation recipients, and humanized NSGS mice (JAX; Stock # 013062) served as recipients for patient-derived xenograft (PDX) models. For isolating cells for MA9 transduction, 6–8-week-old WT or FA mice were used. Both male and female mice were used in this study. All animal experiments were performed in accordance with a protocol (IS00024702) approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Pittsburgh. Mice were monitored regularly for signs of leukemic burden, including weight loss, decreased appetite, and hematologic abnormalities (anemia, leukocytosis), as assessed by peripheral blood counts. Mice were humanely euthanized upon reaching predefined endpoints, including body weight loss exceeding 20% of baseline, severe anemia (hemoglobin <6 g/dL), or other signs of distress (e.g., hunched posture, decreased mobility, or unresponsiveness). These limits were not exceeded in any experiment reported in this study.

Human cell lines and human samples

Human leukemia cell lines were purchased from the American Type Culture Collection (ATCC) or the German Collection of Microorganisms and Cell Cultures (DSMZ): MV4-11 (ATCC, Cat# CRL-9591), THP-1 (ATCC, Cat# TIB-202), HL-60 (ATCC, Cat# CCL-240), U937 (ATCC, Cat# CRL-1593.2), MOLT-4 (ATCC, Cat# CRL-1582), RS4;11 (ATCC, Cat# CRL-1873), and OCI-AML3 (DSMZ, Cat# ACC 582). Cells were cultured in RPMI-1640 supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin at 37 °C with 5% CO₂. Healthy human BM aspirates were purchased from Stem Express or STEMCELL Technologies as commercially sourced, de-identified specimens. Primary AML and ALL specimens (Supplementary Table 4) were obtained through Pittsburgh Biospecimen Core under a protocol approved by the University of Pittsburgh Institutional Review Board (IRB protocol #CR19030357-02). Samples were procured as residual post-surgical materials and de-identified prior to use, with corresponding clinical information, including disease type and stage, provided for research purposes. Informed consent for specimen collection and banking for research purposes was obtained from patients at the time of specimen procurement. No separate or additional informed consent was obtained specifically for the present study, as the samples were de-identified and provided through the established biorepository under an IRB-approved protocol. All human specimen collection and research procedures were conducted in accordance with the principles of the Declaration of Helsinki. The collection, handling, and evaluation of protected health information were performed in compliance with the Health Insurance Portability and Accountability Act (HIPAA).

Treatment

For in vitro drug treatment, cells were seeded at a density of 2.5 × 10⁴ cells/mL and allowed to equilibrate overnight before drug treatment. NU7026 (10 µM; Cat# N1537, Sigma-Aldrich, USA)⁶⁷ was added for 24 h; NU1025 (500 µM; Cat# N7282, Sigma-Aldrich, USA) for 48 h; cytarabine (AraC; 1 µM; Cat# C1768, Sigma-Aldrich, USA)⁵¹ for 48 h; and KO-539 (500 nM; Cat# HY-125880, MedChem Express, USA) for 72 h⁶⁸. For combination treatments, drugs were added simultaneously at the indicated concentrations and durations. Stock solutions of NU7026, NU1025, and KO-539 were prepared in DMSO and diluted into culture medium to a final DMSO concentration of ≤0.1% (v/v); AraC was dissolved in sterile PBS immediately before use. Vehicle controls contained equivalent concentrations of the respective solvents. All treatments were performed in triplicate independent experiments, and cell viability, IC50, proliferation, and apoptosis were assessed at the indicated time points.

For in vivo drug treatment, two weeks after BM transplantation (BMT), recipient mice were randomly assigned to treatment groups. Drug treatments began two weeks after BMT. NU7026 (20 mg/kg, IP)⁴¹ was administered once daily from Day 1 to Day 4. KO-539 (30 mg/kg, PO)⁶⁸ was administered once daily from Day 1 to Day 7. Cytarabine (AraC; 50 mg/kg, IP)⁶⁹ was administered once daily from Day 1 to Day 7. For combination regimens, NU7026 was administered first (~09:00), followed by KO-539 (~13:00) and AraC (~17:00), maintaining a minimum 4-h interval between consecutive doses. On Days 5–7, when NU7026 was no longer administered, KO-539 and AraC were given at ~09:00 and ~13:00, respectively. NU7026 and KO-539 were each dissolved in DMSO and diluted in sterile saline to the indicated concentrations immediately before administration; AraC was dissolved in sterile PBS. All drug solutions were freshly prepared on the day of dosing, and injection volumes were adjusted to 10 mL/kg body weight. Dosing occurred during the light phase, and administration times were recorded for each animal.

Genotyping of Fanca^+/−; MA9 and Fancd2^+/−; MA9 mice

Genomic DNA was extracted from ~1–2 mm tail tip biopsies using a rapid alkaline lysis method as follows. Tail tips were placed in 100 µL of 50 mM NaOH solution in a 1.5 mL microcentrifuge tube and incubated at 95 °C for 30 min to lyse cells and denature proteins. Samples were then cooled to room temperature, and 10 µL of 1 M Tris-HCl (pH 8.0) was added to neutralize the alkaline solution. The mixture was briefly vortexed and centrifuged at 10,000 × g for 1 min to pellet debris. The supernatant containing genomic DNA was collected and used 2 µL of DNA as template for PCR genotyping using primers listed in Supplementary Table 2.

RT-PCR detection of MA9 fusion transcripts

Total RNA was extracted from FA cells lines or FA patients diagnosed with AML using the RNeasy Mini Kit (74104, Qiagen) according to the manufacturer’s instructions, including on-column DNase I treatment to remove genomic DNA contamination. RNA concentration and purity were determined using a NanoDrop spectrophotometer (ND-2000c, Thermo Fisher Scientific), and 500 ng of total RNA was reverse transcribed in a 20 µL reaction using the SuperScript IV First-Strand Synthesis System (18091050, Invitrogen) with random hexamer primers. To amplify a partial/full-length MA9 fusion transcript (~4500 bp), long-range PCR was performed using the LongAmp Taq DNA Polymerase kit (M0323S, New England Biolabs, USA). Primers were designed to flank the fusion junction with the forward primer located in MLL exon 7 (5′-CAGGAGAGAGCTGTTGGAAG-3′) and the reverse primer located in AF9 exon 6 (5′-TGTCTTGGGTCAGGTTCTTG-3′)⁷⁰. Each 50 µL PCR reaction contained 1× LongAmp buffer, 0.2 mM of each dNTP, 0.4 µM of each primer, 1 µL LongAmp Taq DNA Polymerase, and 1–2 µL of cDNA template. Thermal cycling was carried out on a Bio-Rad T100 Thermal Cycler with the following program: initial denaturation at 94 °C for 2 min; 35 cycles of 94 °C for 20 s, 60 °C for 30 s, and 65 °C for 5 min; followed by a final extension at 65 °C for 10 min. PCR products were analyzed on a 0.8% agarose gel stained with ethidium bromide (EtBr, 1610433, Bio-Rad) and visualized using a Bio-Rad Gel Doc system.

Histological and cytological staining

Mice were euthanized according to institutional animal care guidelines. Spleen was harvested intact and bisected to ensure adequate fixative penetration, and femurs were dissected to expose the BM. Tissues were immediately fixed in 10% neutral-buffered formalin (NBF) for 72 h at room temperature. For femurs or BM-containing tissues, decalcification was performed in 10% EDTA (pH 7.4) at 4 °C for 10 days with daily solution changes. Adequate decalcification was confirmed by needle penetration testing and/or absence of resistance during test sectioning. Decalcified tissues were thoroughly rinsed in running tap water for 2 h to remove residual EDTA before further processing. Fixed and decalcified tissues were processed through a graded ethanol series (70%, 80%, 95%, 100%), cleared in xylene, and embedded in paraffin. Tissue blocks were sectioned at 5 µm thickness using a rotary microtome. Sections were mounted on charged glass slides and dried overnight at 37 °C.

For Hematoxylin and Eosin (H&E) staining, slides were deparaffinized in xylene (3 × 5 min), rehydrated through a descending ethanol series (100%, 95%, 70%, 5 min each), and rinsed in distilled water. Slides were stained with Mayer’s hematoxylin for 5 min, rinsed under running tap water for 5 min, optionally differentiated with 0.3% acid alcohol, and nuclei were blued in Scott’s tap water substitute or ammonia water for 30 s to 1 min. Slides were then stained with 1% alcoholic eosin Y solution for 2 min, dehydrated through graded ethanol (70%, 95%, 100%, 5 min each), cleared in xylene (2 × 5 min), and coverslipped using a permanent mounting medium.

PB smears and BM cytospin preparations were air-dried and fixed in methanol for 3 min. Slides were then stained by immersion in freshly prepared Giemsa working solution (1:20 dilution in distilled water, pH 6.8) for 20 min at room temperature. After staining, slides were rinsed gently with buffered distilled water, air-dried, and coverslipped for microscopic examination using an EVOS™ M7000 Imaging System (Invitrogen).

Generation of MSCV-MA9-GFP retrovirus and transduction

HEK293T cells were seeded at 70–80% confluency in 10 cm dishes one day before transfection. The MSCV-MA9-GFP retroviral construct³² was co-transfected with the packaging plasmid pCL-Eco using Lipofectamine 3000 (Cat# L3000015, Thermo Fisher Scientific) according to the manufacturer’s instructions. After 16 h, the medium was replaced with DMEM containing 2% FBS. Viral supernatants were collected at 48 and 72 h, filtered through 0.45 µm polyethersulfone filters, and concentrated using Retro-X™ Concentrator (Cat# 631455, Takara Bio). Viral pellets were resuspended in 50 µL sterile IMEM plus 2% FBS and stored at −80 °C until use.

Isolation and transduction of HSPCs from Fanca^−/−, Fancc^−/− and Fancd2^−/− mice

Whole BM cells were isolated from 6–8-week-old Fanca^−/−, Fancc^−/− and Fancd2^−/− mice. Femurs and tibias were flushed with HBSS, and mononuclear cells were collected and lyased with ammonium-chloride-potassium (ACK) buffer for 30 min at RT. Lin⁻ cells were enriched using the EasySep™ Mouse Hematopoietic Progenitor Cell Isolation Kit (Cat#19856, Stem Cell Technology, Canada) and according to the manufacturer’s protocol.

For retroviral transduction, Lin⁻ cells were cultured in StemSpan SFEM medium (Cat# 09650, Stem Cell Technologies, Canada) supplemented with 50 ng/mL SCF, 50 ng/mL FLT3-L, 50 ng/mL TPO, and 10 ng/mL IL-3. Cells were seeded into Retronectin (20 µg/mL, Takara)-coated plates and transduced with MSCV-MA9-GFP retrovirus in the presence of 8 µg/mL polybrene. Spinoculation was performed at 1000 × g for 90 min at 32 °C, and transduction was repeated after 24 h to enhance efficiency. After 72 h, transduction efficiency was assessed by GFP expression using flow cytometry.

Serial replating leukemia colony formation assay

MA9-transduced cells were subjected to serial colony-forming assays to assess self-renewal capacity in a 35-mm tissue culture dish in 4 mL of semisolid medium containing 3 mL of MethoCult M3134 (Cat # 03134; Stem Cell Technologies, Vancouver, BC, Canada) and the following growth factors: 100 ng/ml SCF (Cat #: 250-03), 10 ng/ml IL-3 (Cat # 213-13), 100 ng/ml GM-CSF (Cat # 315-03) (all from Peprotech, Burlington, NC), and 4 units/mL erythropoietin (Cat # 587602; BioLegend, San Diego CA). For secondary and tertiary plating, cells were harvested from first plating, dissociated into single-cell suspensions, and replated at equal cell numbers for subsequent rounds. Colonies were evaluated under an inverted microscope at 5× magnification. A colony was defined as a spatially independent cluster containing ≥20 cells; clusters with fewer than 20 cells were excluded. Where indicated, colonies were further categorized as micro-colonies (20–49 cells) or mature colonies (≥50 cells), with total colony counts per well including both categories.

siRNA transfection

For siRNA-mediated gene silencing, leukemic cells were transfected by nucleofection using the Lonza 4D-Nucleofector™ system, which provides high-efficiency delivery in suspension hematopoietic cells. Briefly, 1 × 10⁶ cells were resuspended in 100 µL SE Nucleofector™ Solution (Lonza) and mixed with 100 nM FANCA or FANCD2 siRNA (Silencer® Select, Thermo Fisher Scientific; Assay ID listed in Supplementary Table 3). The cell-siRNA mixture was transferred to a nucleocuvette and electroporated using program CM-137 (recommended for myeloid leukemia lines). Immediately after nucleofection, cells were transferred into pre-warmed complete RPMI-1640 medium and allowed to recover at 37 °C. Knockdown efficiency was assessed by 72 h post-nucleofection by Western blot analysis.

Growth curve measurement of transduced or knockdown cells

Cells were seeded in triplicates at a density of 5 × 10⁴ cells per well in a 24-well plate and cultured at 37 °C, 5% CO₂, and their growth was monitored daily. Each day, cells were gently resuspended, and an aliquot (10 µL) was taken for counting. Viable cell numbers were determined using the trypan blue exclusion assay and counted using an automated cell counter (Countess, Thermo Fisher). The total viable cell number was recorded daily from Day 1 to Day 6. Cell growth was plotted as cell number (y-axis) vs. culture time (x-axis, days 1–6). Doubling time was calculated using the formula:

$$\mathrm{Doubling\; Time}=t\times \ln \left(2\right)/\ln \left(N_t/N_0\right),$$

where t is the time interval, N₀ is the initial cell number, and Nₜ is the final cell number.

Human CD34⁺ cell isolation and quantitative PCR (qPCR) analysis of FANCD2 expression

Human CD34⁺ hematopoietic progenitor cells from AML/ALL patients (Supplementary Table 4) were enriched by fluorescence-activated cell sorting (FACS) using a BD FACSAria II cell sorter (BD Biosciences). Cells were stained with anti-human CD34 antibody conjugated to APC following the manufacturer’s protocol. CD34⁺ cells were gated using a matched-fluorochrome IgG isotype control, which was included in each staining experiment to define the positive boundary and distinguish specifically from nonspecific antibody staining. Sorted CD34⁺ cells were collected into tubes for RNA extraction. Purity of sorted CD34⁺ cells routinely exceeded 95% as assessed by post-sort analysis. Total RNA was extracted from sorted CD34⁺ cells using the RNeasy Micro Kit (74004, QIAGEN). RNA concentration and purity were determined using a NanoDrop spectrophotometer. First-strand cDNA synthesis was performed using 100 ng of total RNA and the SuperScript IV First-Strand Synthesis System (Invitrogen) with random hexamer primers. qPCR for human FANCD2 expression was conducted on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) using PowerUp^TM SYBR^TM Green Master Mix (A25741, Thermo Fisher Scientific). Primer sequences were as follows: FANCD2 forward 5′- CAAAATCAGCTAGGTGTGGATCA-3′ and reverse 5′-CCAGGCCATTAACAAACTCTTCT-3′; GAPDH forward 5′-AGGTCGGTGTGAACGGATTTG-3′ and reverse 5′-GGGGTCGTTGATGGCAACA-3′ (housekeeping control). Reactions were set up in 20 µL volumes containing 1× SYBR Green Master Mix, 0.3 µM of each primer, and 10 ng cDNA. Thermal cycling conditions included initial activation at 95 °C for 2 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Gene expression levels were calculated using the 2^−ΔΔCt method with GAPDH as the internal reference. All qPCR reactions were performed in technical triplicates.

Bone marrow transplant (BMT)

For the murine model, 1 × 10⁶ unsorted MA9-transduced HSPCs (CD45.2⁺) after two weeks ex vivo culture was transplanted into lethally irradiated congenic Boy J recipient mice (CD45.1⁺). Leukemia engraftment (CD45.2⁺) and circulating leukemic progenitor-like cell (CD45.2⁺c-Kit⁺) frequencies were assessed at defined time points post-transplantation by flow cytometry.

For the patient-derived xenograft (PDX) model, primary CD34⁺ cells from either healthy donors or MLL-r AML/ALL patients were transplanted into sublethally irradiated NSGS mice. Mice subsequently received the indicated treatments^51,69. Human engraftment and malignant cell populations were analyzed at various time points post-treatment by flow cytometry⁴⁷.

Flow cytometry analysis

For apoptosis analysis, cells were stained with Annexin V-APC and DAPI, following manufacturer protocols. Early apoptotic cells were defined as Annexin V⁺DAPI⁻ and late apoptotic/necrotic cells as Annexin V⁺DAPI⁺. For cell cycle analysis, cells were fixed in ice-cold 70% ethanol overnight, followed by intracellular staining with anti-Ki67-APC-Cy7 and DAPI.

For DDR profiling, cells were fixed with 1.6% paraformaldehyde, permeabilized with 90% methanol, and stained with the following phospho-specific antibodies: H2AX-Y142, Chk2-pT68, p53-S15 and ATM-S1981, followed by a PE-conjugated secondary antibody (one marker per tube). Cells were analyzed using phospho-flow cytometry, and DDR marker-positive populations were quantified relative to isotype controls.

For LSC phenotyping, single-cell suspensions from spleen and BM were lineage-depleted using the Lineage Cell Depletion Kit. Cells were stained with the following antibodies: Sca-1-APC-CY7, c-Kit-PE-Cy7, CD34-APC, CD16/32-PE, LSCs were defined as Lin⁻ Sca1⁻ c-Kit⁺ CD34⁺ CD16/32⁺.

For quantification of retroviral transduction efficiency, cells were analyzed 72 h post-transduction by flow cytometry based on native GFP fluorescence, without additional antibody staining. Transduction efficiency was calculated as the percentage of GFP⁺ cells among live cells, using non-transduced cells as a negative control baseline.

For detecting mouse engraftment and donor-derived leukemic blast in the recipient mice, peripheral blood, spleen, and BM cells were stained with anti-mouse CD45.2-APC. Donor-derived circulating leukemic progenitor-like cells were co-stained with CD45.2-APC and c-Kit-PE and gated as CD45.2⁺c-Kit⁺ (Supplementary Table 5 for antibody details).

For detecting human cell engraftment and leukemic burden in NSGS recipient mice, peripheral blood and BM cells were stained with anti-human CD45-APC-Cy7 and anti-mouse CD45-FITC to distinguish donor (hCD45⁺) from host (mCD45⁺) populations. Within the hCD45⁺ compartment, malignant HSPCs were identified as CD34⁺CD38⁻, and myeloid blasts were identified as CD33⁺CD19⁻, using anti-CD34-APC, anti-CD38-PE/PE-Cy7, anti-CD33-PE-Cy7, and anti-CD19-APC antibodies (Supplementary Table 5 for antibody details).

Sequential gating strategies for donor-derived circulating leukemic progenitor-like cell phenotyping, apoptosis analysis, cell cycle analysis, DDR profiling, GFP⁺ transduction efficiency, mouse engraftment/leukemic progenitor identification, and human cell engraftment/leukemic stem-progenitor identification are shown in Supplementary Figs. 2f, 11a–e.

Immunoprecipitation and western blotting

Immunoprecipitation was performed from lysates of the indicated human leukemia cell lines (THP-1, MV4-11, RS4;11, and, where indicated, HL-60 and MOLT-4). Cells were lysed in a non-denaturing buffer supplemented with protease and phosphatase inhibitors. To minimize DNA- or chromatin-mediated bridging between proteins, clarified lysates were divided, and aliquots were treated with DNase I prior to antibody incubation. Pre-cleared lysates were incubated with an antibody against the MLL N-terminal region or isotype control IgG, followed by capture with protein A/G beads. Immunocomplexes were washed under conditions preserving native interactions, eluted, and analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting with antibodies against KU70, KU80, and FANCD2. Input and IgG controls were included in all experiments.

For standard Western blotting, cells were washed with ice-cold PBS and lysed in ice-cold buffer containing 50 mM Tris-HCl (pH 7.4), 0.1% NP-40, and 1 M NaCl, supplemented with protease and phosphatase inhibitors (10 μg/mL aprotinin, 25 μg/mL leupeptin, 10 μg/mL pepstatin A, 2 mM phenylmethylsulfonyl fluoride, 0.1 M NaP₂O₄, 25 mM NaF, and 2 mM sodium orthovanadate). Lysates were incubated for 30 min on ice and clarified by centrifugation. Protein samples were resolved by SDS-PAGE, transferred to PVDF membranes, and probed overnight at 4 °C with the indicated primary antibodies (see Supplementary Table 6 for antibody details). After incubation with appropriate horseradish peroxidase (HRP)-conjugated secondary antibodies, signals were detected using enhanced chemiluminescence (ECL; Thermo Scientific, Cat# 32106) and acquired on a ChemiDoc Go imaging system (Bio-Rad). Band intensities were quantified by densitometry using ImageJ v1.8.0 (NIH, Bethesda, MD). Uncropped scans of all blots used in this study are provided in the Source Data file.

Immunofluorescence staining and co-localization analysis

Cells were fixed with 4% paraformaldehyde for 10 min at room temperature, followed by permeabilization with 0.2% Triton X-100 for 5 min. After blocking with 5% BSA for 60 min, cells were incubated overnight at 4 °C with primary antibodies against γH2AX and either RAD51 or 53BP1. Following three washes with PBS, cells were incubated with appropriate fluorescent secondary antibodies for 1 h at room temperature in the dark. Nuclei were counterstained with DAPI, and samples were mounted for imaging. Fluorescence microscopy (EVOS M7000, Invitrogen) was performed to assess the co-localization of γH2AX with RAD51 or 53BP1.

For RAD51 and 53BP1 foci analyses, MA9 transduced WT or FA-deficient HSPCs were maintained under low-cytokine culture conditions (StemSpan SFEM supplemented with 10 ng/mL SCF, 10 ng/mL FLT3-L, 10 ng/mL TPO, and 5 ng/mL IL-3) under hypoxic conditions. HSCs were synchronized into S phase by cytokine deprivation prior to RAD51 analysis. For NHEJ analysis, cell-cycle synchronization was achieved using aphidicolin (2 μg/mL) for 24 h as previously described⁷¹. Co-localized foci were defined as spatially overlapping γH2AX and RAD51 or 53BP1 puncta within the same nucleus. The number of co-localized foci per cell was quantified manually from fluorescence images using ImageJ, and ≥25 cells per condition were analyzed across at least 3 independent experiments.

Comet assay (single-cell gel electrophoresis, neutral conditions)

WT + MA9, Fanca^−/− + MA9, Fancc^−/− + MA9, and Fancd2^−/− + MA9 transduced cells were collected and washed twice with ice-cold PBS (Ca²⁺/Mg²⁺ free). Approximately 1 × 10⁵ cells were resuspended in chilled PBS at a concentration of 1 × 10⁶ cells/mL. An aliquot (10 µL) was gently mixed with 90 µL of molten low-melting-point agarose (1% w/v, maintained at 37 °C) and immediately spread onto Comet slides pre-coated with 1% normal-melting-point agarose. Slides were placed at 4 °C in the dark for 10 min to allow agarose solidification. Slides were then immersed in neutral lysis buffer (2.5 M NaCl, 100 mM EDTA, 10 mM Tris-HCl, 1% sodium lauroyl sarcosinate, pH 10, supplemented with 1% Triton X-100 and 10% DMSO) at 4 °C for at least 1 h to lyse cells and remove proteins. Following lysis, slides were equilibrated in neutral electrophoresis buffer (90 mM Tris-base, 90 mM boric acid, 2 mM EDTA, pH 8.5) for 30 min at 4 °C. Electrophoresis was conducted in the same buffer at 25 V (~0.6 V/cm) for 30 min at 4 °C. After electrophoresis, slides were neutralized with three washes of 0.4 M Tris-HCl (pH 7.5), dehydrated in 70% ethanol for 5 min, and air-dried. DNA was stained with SYTO™ 60 Red Fluorescent Nucleic Acid Stain (S11342, Thermo Fisher Scientific, USA) and visualized using a fluorescence microscope equipped with appropriate filters. For each sample, at least 100 comets were analyzed using ImageJ software, and DNA damage was quantified by calculating the tail moment (product of tail length and the percentage of DNA in the tail)³⁶.

Chromosomal breakage analysis

Chromosome breakage analysis was performed on mouse MA9 transduced HSPCs as previously described⁷². Colcemid mid (0.1 µg/mL; Gibco, Grand Island, NY, USA) was added during the final 2 h to arrest cells in metaphase. Cells were harvested by trypsinization or direct centrifugation, treated with 0.075 M KCl hypotonic solution for 20 min at 37 °C, and fixed in methanol: acetic acid (3:1) with at least three changes. Metaphase spreads were prepared by dropping fixed cells onto chilled, humidified glass microscope slides. Slides were air-dried and stained with 5% Giemsa (Sigma-Aldrich, USA) in phosphate buffer (pH 6.8) for 5–10 min.

Cell viability assay

Mouse MA9 transduced HSPCs or human leukemia cell line with FANCD2/FANCA knockdown were cultured and subjected to various drug treatments as follows: NU7026 (10 µM) was added for 24 h; NU1025 (500 µM) for 48 h; cytarabine (AraC; 1 µM) for 48 h; and KO-539 (500 nM) for 72 h. Cells were treated with each agent alone or in combination as indicated. Following treatment, cell viability was measured using the CellTiter-Glo Luminescent Cell Viability Assay Kit (Promega, Madison, WI), according to the manufacturer’s instructions. Luminescence was detected using a Synergy H1 Hybrid Multi-Mode Reader (BioTek, Winooski, VT).

In situ PLA

Human leukemia cell lines (THP-1, MV4-11, RS4;11, HL-60, and MOLT-4), including cells with stable FANCD2 knockdown and corresponding non-targeting control cells, were maintained under standard culture conditions. Cells were harvested and washed twice with PBS and fixed with cold methanol (−20 °C) for 5 min to preserve cellular and nuclear morphology. Following fixation, cells were permeabilized with 0.5% Triton X-100 in PBS for 5 min at room temperature to allow antibody access to intracellular targets. The PLA was performed using the Duolink® In Situ Red Starter Kit Mouse/Rabbit (Cat# DUO92101, Sigma-Aldrich) according to the manufacturer’s instructions⁷³. After permeabilization, cells were blocked in 3% BSA in PBS for 1 h at room temperature. Primary antibody incubation was carried out overnight at 4 °C using a combination of mouse anti-MLL-r N-terminal antibody (specific to the MLL-r fusion N-terminus, 05-764, Sigma-Aldrich, IL USA) and rabbit anti-Ku70 (4588S, Cell Signaling Technology, USA) antibody, each diluted 1:500 in PBS. After thorough PBS washes, PLA probes (anti-mouse PLUS and anti-rabbit MINUS) were applied and incubated at 37 °C for 1 h. This was followed by ligation and rolling circle amplification as per the Duolink protocol. Nuclei were counterstained using Duolink® In Situ Mounting Medium with DAPI. Fluorescent PLA signals, indicative of close proximity between MLL-r N-terminal and Ku70 proteins, were visualized using a Nikon AXR confocal microscope. Comparative analysis was performed between control and FANCD2-deficient cells to assess the impact of FANCD2 loss on MLL-r and Ku70 interaction dynamics.

L-LTC-IC assay

L-LTC-IC frequency was determined by in vitro limiting dilution assay as previously described^51,52. Briefly, MS-5 feeder cells were maintained in IMDM supplemented with 10% heat-inactivated FBS and 2% penicillin/streptomycin at 37 °C in 5% CO₂ and subcultured at 80% confluency. WT + MA9, Fanca^−/− + MA9, and Fancd2^−/− + MA9 transduced HSPCs were plated at limiting dilutions onto freshly prepared irradiated MS-5 feeder layers and maintained under normoxic conditions for 4 weeks with weekly half-medium changes performed without disrupting the feeder layer. Wells were scored for leukemic colony formation, and L-LTC-IC frequency was calculated using ELDA online tool (http://bioinf.wehi.edu.au/software/elda/index.html)⁷⁴.

DNA repair reporter assay

To assess DNA repair activity, human leukemic cells were transfected with linearized DR-GFP (HR) and EJ5-GFP (NHEJ) constructs^42,75 using a Lonza 4D-Nucleofector, followed by G418 (10131035, Thermo Fisher Scientific) selection. Leukemia cells carrying chromosomally integrated reporters were then subjected to siRNA-mediated gene knockdown. After 48 h, cells were nucleofected with a plasmid encoding I-SceI endonuclease to induce DSBs. To exclude the potential transfection difference, 0.1 μg of plasmid encoding DsRed (pDsRed2-N1) was used as a control. 72 h post-nucleofection, the ratio between GFP⁺ and DsRed⁺ cells was determined by Flow cytometry⁴².

Analysis of FA pathway gene expression and predicted olaparib sensitivity in primary AML samples

Data acquisition and processing

Gene expression data from the BEAT AML cohort were downloaded from the Vizome platform (https://vizome.org), while data for the TARGET AML cohort were obtained from the National Cancer Institute’s TARGET Data Matrix (https://ocg.cancer.gov/programs/target). Both datasets include RNA-sequencing profiles of primary AML patient samples, and expression values were used as provided by each respective platform. Gene expression and Olaparib AUC data for AML cell lines used to train the elastic net model were obtained from the Cancer Therapeutics Response Portal v2 (CTRPv2). Prior to downstream analysis, gene expression values across datasets were log2-transformed [log2(x + 1)] to approximate normality and ensure comparability of scale.

FA pathway signature and expression stratification

Expression levels of core FA pathway genes (BRCA1, FANCA, FANCC, and FANCD2) were compared across three groups stratified by t(9;11) cytogenetic status: healthy controls/normal BM (HL), non-t(9;11) AML samples, and t(9;11)-positive AML samples. Pairwise comparisons of expression levels between groups (HL vs. non-t(9;11), non-t(9;11) vs. t(9;11), and HL vs. t(9;11)) were performed using two-sided Wilcoxon rank-sum tests for each gene independently. Multiple testing correction was applied using the Benjamini-Hochberg procedure where applicable. All data processing and analyses were conducted using R (version 4.2.2). A significance threshold of p < 0.05 was applied for all statistical tests.

Prediction of olaparib sensitivity

A machine learning approach was used to estimate Olaparib sensitivity from gene expression signatures. Specifically, we trained an elastic net regression model using gene expression and Olaparib AUC data from AML cell lines available in the CTRPv2 dataset, implemented via the glmnet R package. The trained model was then applied to the BEAT AML and TARGET AML expression data to generate predicted Olaparib AUC values for each sample. Lower predicted AUC values were interpreted as higher predicted sensitivity to Olaparib. For visualization and group-wise comparison, predicted AUC values were dichotomized into “Sensitive” and “Resistant” categories based on the median predicted AUC value within each cohort. Expression levels of individual FA genes (BRCA1, FANCA, FANCC, and FANCD2) were each independently dichotomized into “High” and “Low” groups based on median expression within each cohort. The association between FA gene expression groups and predicted drug response categories was assessed using the chi-square test of independence. Multiple testing correction was applied using the Benjamini-Hochberg procedure where applicable. All data processing and analyses were conducted using R (version 4.2.2). A significance threshold of p < 0.05 was applied for all statistical tests.

Structural modeling using AlphaFold3

Protein structural models of FANCD2, KU70 (XRCC6), KU80 (XRCC5), and the N-terminal region of the MLL fusion protein⁵⁷ were generated using AlphaFold3 via the AlphaFold Server (https://alphafoldserver.com). Residues 1–1400 correspond to the N-terminal region of MLL (MLL-N1400) that is retained across virtually all MLL-r fusion proteins regardless of the fusion partner (e.g., AF4, AF9), as chromosomal breakpoints cluster within this region; this fragment includes the AT-hook and CXXC domains but excludes the C-terminal SET domain, which is lost in all MLL fusions. Protein sequences were retrieved from UniProt (FANCD2: Q9BXW9; XRCC6/KU70: P12956; XRCC5/KU80: P13010; KMT2A/MLL: Q03164). Binary complexes (KU70/KU80–MLL-N1400 and FANCD2–MLL-N1400) and the ternary complex (KU70/KU80–MLL-N1400–FANCD2) were modeled by submitting the corresponding protein sequences as multimeric inputs. For each complex, five independent predictions were generated, and the model with the highest predicted confidence (based on ipTM/pTM scores) was selected for downstream analysis. PAE matrices were extracted from the AlphaFold3 output to assess the confidence of relative inter-chain positioning; lower PAE values between residue pairs indicate higher confidence in their predicted relative orientation. Structural superimposition and visualization of predicted models, including alignment of binary and ternary complexes, were performed using UCSF Chimera X (version 1.7).

Statistics and reproducibility

No statistical method was used to predetermine sample size; sample sizes were chosen based on standard practice in the field and prior experience with similar experimental systems. No data was excluded from the analysis. Recipient mice in in vivo drug treatment experiments were randomly assigned to treatment groups; no randomization was applied to other experiments. The investigators were not blinded to allocation during experiments and outcome assessment. All statistical analyses were performed using GraphPad Prism. All tests were two-sided. Comparisons between two groups were performed using an unpaired Student’s t test. For comparisons involving more than two groups, one-way ANOVA with Tukey’s multiple comparisons test or two-way ANOVA with Tukey’s multiple comparisons test was used, as appropriate. Survival data were analyzed using Kaplan–Meier plots and the log-rank (Mantel–Cox) test. Results are presented as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001 were considered statistically significant. Statistical analyses were conducted using GraphPad Prism version 9.0 and R version 4.2.2.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.X. performed the research, analyzed the data and wrote the paper. L.S., E.V.W., A.Z.Z, R.W. and Z.G. performed some of the research, assisted with data analysis; S.I. provided human samples and contributed to some data interpretation; W.D. designed the research, analyzed the data, and wrote the paper.

Peer review

Peer review information

Nature Communications thanks the anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding statement

This work was supported by NIH/National Heart, Lung, and Blood Institute grants R01HL151390 and R56HL169348 (to W.D.), a NIH/National Cancer Institute grant R01CA285400 (to W.D.), and a NIH/National Institute on Aging grant RF1AG093995 (to W.D.). W.D. is also supported by a Blood Cancer United (formerly the Leukemia and Lymphoma Society) Scholar Award.

Data availability

All the other data are available within the article and its Supplementary Information. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77335-0.