TET2 and DNMT3A Mutations Exert Divergent Effects on DNA Repair and Sensitivity of Leukemia Cells to PARP Inhibitors

Abstract

Somatic variants in TET2 and DNMT3A are founding mutations in hematological malignancies that affect the epigenetic regulation of DNA methylation. Mutations in both genes often co-occur with activating mutations in oncogenic tyrosine kinases such as FLT3^ITD, BCR-ABL1, JAK2^V617F, and MPL^W515L, or with mutations affecting related signaling pathways such as NRAS^G12D and CALR^del52. Here we show that TET2 and DNMT3A mutations exert divergent roles in regulating DNA repair activities in leukemia cells expressing these oncogenes. Malignant TET2-deficient cells displayed downregulation of BRCA1 and LIG4, resulting in reduced activity of BRCA1/2-mediated homologous recombination (HR) and DNA-PK -mediated non-homologous end-joining (D-NHEJ), respectively. TET2-deficient cells relied on PARP1-mediated alternative NHEJ (Alt-NHEJ) for protection from the toxic effects of spontaneous and drug-induced DNA double-strand breaks. Conversely, DNMT3A-deficient cells favored HR/D-NHEJ owing to downregulation of PARP1 and reduction of Alt-NHEJ. Consequently, malignant TET2-deficient cells were sensitive to PARP inhibitor (PARPi) treatment in vitro and in vivo, whereas DNMT3A-deficient cells were resistant. Disruption of TET2 dioxygenase activity or TET2 - Wilms tumor 1 (WT1) binding ability were responsible for DNA repair defects and sensitivity to PARPi associated with TET2 deficiency. Moreover, mutation or deletion of WT1 mimicked the effect of TET2 mutation on DSB repair activity and sensitivity to PARPi. Collectively, these findings reveal that TET2 and WT1 mutations may serve as biomarkers of synthetic lethality triggered by PARPi, which should be explored therapeutically.

Significance

TET2 and DNMT3A mutations affect distinct DNA repair mechanisms and govern the differential sensitivities of oncogenic tyrosine kinase-positive malignant hematopoietic cells to PARP inhibitors.

Introduction

Oncogenic tyrosine kinases (OTKs) are found in many types of neoplasms including hematopoietic malignancies (1). For example, mutations in FLT3 and MPL cell membrane tyrosine kinases (e.g., FLT3^ITD/TKD and MPL^W515L), in ABL1 nuclear/cytoplasmic tyrosine kinase (e.g., BCR-ABL1) and in JAK2 cytoplasmic tyrosine kinase (e.g., JAK2^V617F) are associated with numerous hematological malignancies, such as acute myeloid leukemia (AML), acute lymphoblastic leukemia (ALL), chronic myeloid leukemia (CML), and myeloproliferative neoplasms (MPNs). Standard treatment for these malignancies includes chemotherapy and/or targeted approaches with selective tyrosine kinase inhibitors (TKi). With the exception of CML in chronic phase, complete remissions are uncommon in TKi-treated patients, and after an initial response, diseases often progress to a more aggressive stage. There is an urgent need for new therapies that could improve the therapeutic outcome of patients with hematological malignancies driven by OTKs.

Previous reports (2-7) demonstrate that OTK-positive cells accumulate high levels of spontaneous and drug-induced DNA double-strand breaks (DSBs) in comparison to normal cells, but survive because of enhanced/altered DNA repair activities (8). Therefore, survival of OTK–positive cells depends on efficient repair of DSBs, the most lethal DNA lesions. DSBs are usually repaired by BRCA1/2-dependent homologous recombination (HR; key proteins: BRCA1, BRCA2, PALB2, RAD54, RAD51), DNA-PK –mediated non-homologous end-joining (D-NHEJ; key proteins: DNA-PKcs, Ku70, Ku80, LIG4), and PARP1–dependent alternative NHEJ (Alt-NHEJ; key proteins: PARP1, LIG3), which often serves as a backup DSB repair pathway (9).

The clinical success of the PARP inhibitor (PARPi) olaparib in BRCA1/2-mutated breast and ovarian cancers has established proof-of-concept for precision medicine in HR-deficient neoplasms by synthetic lethality (10). While OTK-positive cancer cells survive in spite of increased DSB repair activity, TKi-treated cells specifically display HR and/or D-NHEJ deficiencies that sensitize malignant hematopoietic cells to synthetic lethality triggered by PARPi (3,6,11). Intriguingly, reports from our lab (3,6) and others (12-14) suggest that the types of mutations co-occurring with OTKs strongly influence DNA repair in these cells and the sensitivity to PARPi.

Somatic mutations in TET2 (ten-eleven translocation 2) and DNMT3A (DNA (cytosine-5)-methyltransferase 3A) are common in age-related clonal hematopoiesis (CH) and are frequently detected in hematological malignancies (15). Most variants in both genes are point mutations that lead to loss of function. Although the proteins antagonistically regulate the epigenetic mark of 5-methylcytosine (5-mC), where DNMT3A catalyzes addition of 5-mC (16) while TET2 oxidizes 5-mC as a first step in DNA demethylation (17), mutations in both genes appear in a similar spectrum of human hematopoietic malignancies and occur with common co-operating mutations including OTKs. Moreover, while the mutations exert divergent effect on primitive hematopoietic progenitor cells (18), they lead to similar disease phenotypes, suggesting the roles of these mutations in hematopoietic malignancies may relate to mechanisms outside of DNA methylation.

The potential role of TET2 and DNMT3A mutations in DNA damage response (DDR) of hematopoietic malignancies was highlighted by the observation that loss of TET2 in AML reduced their sensitivity to AraC by altering the dynamics of transition between differentiated and stem-like states (19), and that AML with DNMT3A mutations displayed impaired sensitivity to anthracycline treatment due to their inability to sense and repair the drug-induced DNA torsional stress (20).

Here we report that mutations in TET2 and DNMT3A exert opposite effects on the sensitivity of OTK-transformed hematopoietic cells to PARPi; with TET2 mutations sensitizing malignant cells to PARPi, whilst DNMT3A mutations were associated with resistance. Moreover, in cells carrying DNMT3A mutations, the inactivation of TET2 overcomes PARPi resistance. The sensitivity of TET2-mutant cells to PARPi was dependent on the disruption of TET2 interaction with the Wilms tumor 1 (WT1) protein. Our observations suggest that certain combinations of co-occurring mutations with OTKs influence distinct DNA repair mechanisms and govern the differential sensitivities of OTK-positive leukemia cells to PARPi.

Materials and Methods

Primary human AML samples

AML samples (see Supplementary Table S1) were from the ECOG-ACRIN E1900 clinical trial (21), from Leukemia Tissue Bank, Princess Margaret Hospital, Ontario Cancer Institute, Toronto, Canada, from Mayo Clinic, Rochester, MN, USA, University of Bologna, Bologna, Italy and from the Department of Internal Medicine I, Division of Hematology & Hemostaseology, Medical University of Vienna, Austria. The investigators obtained informed written consent from the subjects.

Genetically modified mice

The Institutional Animal Care and Use Committee at Washington University approved all animal procedures. Dnmt3a^fl/fl (22), Tet2^fl/fl (23) and Parp1^−/− (11) mice were crossed to Flt3^ITD (24) and Vav-Cre (25) strains as described previously (18) to generate mice used for experiments. The following primers were employed for genotyping: Vav-Cre- AGATGCCAGGACATCAGGAACCT and ATCAGCCACACCAGACACAGAGATC; Tet2^fl- AAGAATTGCTACAGGCCTGC and TTCTTTAGCCCTTGCTGAGC; Flt3^ITD - common TCTGGTTCCATCCATCTTCC, wild-type AGGAAGTCGATGTTGGCACT and mutant CTTCGTATAATGTATGCTATACG; Dnmt3a^fl- ATCACATTACCTTTGTCCTCCCAGATCCAG and AGGCTGTCTGCATCGGACAGTGAGTGGTG, and Parp1 - common CATGTTCGATGGGAAAGTCCC, wild-type CCAGCGCAGCTCAGAGAAGCCA, and mutant AGGTGAGATGACAGGAGATC. Flt3^m/m;Wt1^+/− (Flt3^ITD;Wt1^+/−) mice were described previously (26). To verify efficiency of deletion of Dnmt3a floxed alleles, the following primers were used: Dnmt3a_loxForward CTGTGGCATCTCAGGGTGATGAGCA, Dnmt3a_1loxReverse TGAGTGGTGAGGCCCAGCTTATCGA, Dnmt3a_2loxReverse AAGCCTCAGGCCCTCTAGGCAAGAT using Phusion polymerase as previously described (27).

Ectopic expression constructs

Murine Tet2 catalytic domain (mTet2-CD) and enzymatically inactive mTet2-CD^CM mutant (H1295Y+D1297A) and human hTET2-CD were obtained from Dr. Yue Xiong, University of North Carolina at Chapel Hill (28). hTET2-CD^A1505T mutant was generated using QuikChange II XL site-directed mutagenesis kit from (Agilent Technologies, #200521) according to manufacturer’s instructions. hTET2-CD and hTET2-CD^A1505T were cloned into the custom-built lentiviral vector pLVX-CMV-flag-IRES-mRFP1 (VectorBuilder Inc.). mTet2-CD and mTet2-CD^CM were re-cloned into MIGR1 retroviral plasmid.

Viral infections

HEK 293T/17 cells (ATCC® CRL-11268, validated by ATCC) were used no longer than 4 weeks after thawing and were not tested for Mycoplasma. Retroviruses were prepared by co-transfecting HEK 293T/17 cells in a 10-cm plate with 10 μg of packaging pCL-ECO plasmid (Addgene #12371) and 10 μg MIGR1 based vectors by using Lipofectamine 2000 Transfection Reagent (Invitrogen) according to manufacturer’s protocol. Lentiviruses were generated by co-transfection of HEK 293T/17 cells with 10 μg of pLVX-CMV-hTet2(CD)-flag-IRES-mRFP1 based vectors, and 10 μg packaging pCMV delta R8.2 plasmid (Addgene #12263) and envelope pVSV-G plasmid (Clontech, #PT3343-5). Viruses were harvested 40h and 64h after the transfection and filtered through a 0.45-μm PES filter (Millipore). For infection, 5 × 10⁵ cells were suspended in 1 ml of virus-containing medium with 6 μg/ml polybrene (Sigma). GFP+ and RFP+ cells were sorted 48 hrs after the initial infection.

CRISPR/Cas9 targeting

Briefly, guide RNAs (gRNAs) targeting murine Tet2 and Dnmt3a were generated using CRISPR Design (Supplementary Table S2) and cloned into the lentiCRISPRv2 vector. The lentiCRISPRv2 vector with a gRNA insert, the packaging plasmid psPAX2, and the envelope plasmid VSVG were mixed together and packed in HEK293T/17 cells using Fugene 6. Lentiviruses were harvested at 48 and 72 h respectively. FLT3^ITD-positive 32Dcl3 cells (described before (3) and authenticated by Western blot detecting constitutively active FLT3^ITD kinase) were used for up to 4 weeks after thawing and were not tested for Mycoplasma. Cells were infected with freshly collected CRISPR-Cas9-gRNA lentivirus supplemented with 8 μg/ml polybrene (Sigma-Aldrich) for 48h. Infected cells were selected in media with puromycin (1 μg/ml) for up to 14 days. Single clones were harvested from methylcellulose and successful mutation of infected clones was confirmed with T7E1 assay. To identify the mutations, the PCR products were sequenced by Sanger sequencing. The primers used for Sanger sequencing are listed in Supplementary Table 2. Downregulation of TET2 and/or DNMT3A proteins was detected by Western blot.

Western blot

Nuclear and total cell lysates were obtained as described before (29) and analyzed by SDS-PAGE using primary antibodies against: TET2 (Cell Signaling #45010), DNMT3A (R&D Systems, MAB63151 and Cell Signaling #2160), ATM (Santa Cruz Biotech #sc-135663), 53BP1 (Abcam #ab-21083), CtIP (ThermoFisher Scientific #PA5-20963), SLFN11 (Abcam #ab121731), BRCA1 (EMD Millipore #OP92-100UG), BRCA2 (Abcam #ab75335), PALB2 (ThermoFisher Scientific #PA5-20796), RAD51 (Santa Cruz Biotech #sc-8349), RAD52 (Santa Cruz Biotech #sc-365341), RAD54 (Santa Cruz Biotech #sc-374598), DNA-PKcs (Bethyl #A300-518A), DNA ligase 4 (Santa Cruz Biotech #sc-271299), Ku80 (ThermoFisher Scientific #MA5-15873), Ku70 (Santa Cruz Biotech #sc-17789), PARP1 (Santa Cruz Biotech #sc-74470), PARP2 (Abgent #AP12265a), PARP3 (Abgent #AP6296b), DNA ligase 3 (Santa Cruz Biotech #sc-135883), Flag (Cell Signaling #2368), lamin B (Abcam #ab-16048-100), and β-actin (Santa Cruz Biotech #sc-47778) and the following secondary antibodies conjugated to HRP: goat anti-rabbit (EMD Millipore #12-348) and goat anti-mouse (EMD Millipore #AP181P).

5hmC dot-blot

Genomic DNA was isolated using DNeasy Blood Tissue Kit (Qiagen #69506), run in the Bio-Dot SF Microfiltration Apparatus (Bio-Rad #170-6542), cross-linked to Zeta-Probe Blotting Membrane (Bio-Rad # 1620153) and blotted with anti-5hmC primary antibody (Active Motif # 39791) followed by secondary HRP-conjugated goat anti-rabbit HRP-conjugated antibody (EMD Millipore #12-348).

Inhibitors/compounds

PARPi olaparib and talazoparib, FLT3i quizartinib, JAK1/2i ruxolitinib, doxorubicin and hydroxyurea were from Selleckchem. 2-hydroxyglutarate (2-HG) was purchased from CaymanChemical [(2R)-Octyl-α-hydroxyglutarate-d17] and Toronto Research Chemicals [(2S)-2-Hydroxyglutaric Acid Octyl Ester Sodium Salt and (2R)-2-Hydroxyglutaric Acid Octyl Ester Sodium Salt, mixed 1:1]. Vitamin C (L-ascorbic acid = L-AA) was from Sigma. All compounds were dissolved, aliquoted and stored following manufacturer’s instructions.

In vitro sensitivity assays

Lin⁻CD34⁺ human cells and Lin-cKit+ murine cells were obtained from mononuclear fractions by magnetic sorting using the EasySep Lin negative selection followed by CD34 or cKit positive selection cocktail as described before (11). Human AML primary cells were incubated in StemSpan®SFEM medium (Stem Cell Technologies, Vancouver, Canada) supplemented with the cocktail of growth factors (100 ng/ml SCF, 20 ng/ml IL-3, 100 ng/ml FLT-3 ligand, 20 ng/ml G-CSF, 20 ng/ml IL-6) as described before (11). Murine normal and malignant hematopoietic cells were maintained in StemSpan®SFEM medium supplemented with: (normal CML-like and AML-like: 100 ng/ml SCF, and 20 ng/ml IL-3; MPN-like: 100 ng/ml SCF; 10 ng/ml FLT3 ligand; 20 ng/ml IL-3, IL-6, G-CSF and GM-CSF; 12 units/ml EPO; 2.5 ng/ml TPO] as described before (3,6,11). Compounds were added for 3 days following by trypan blue exclusion viability test and/or plating in Methocult (Stemcell Technologies) when indicated; colonies were scored after 5-7 days.

DNA fibers

Cells were left untreated or treated with 2.5 μM olaparib for 24 hours. DNA fibers were examined as described before (30). Briefly cells were treated with IdU (250 μM) for 50 minutes and washed three times. After lysis DNA spreads were obtained by tilting slides to 30 degrees and air dry for 3 hours. DNA was denatured with HCl for 2.5 hours and after blocking cells were stained with primary anti-BrdU antibody (BD #347580) for 2.5 hours. Secondary Alexa Fluor 594 rabbit anti-mouse IgG (H+L) antibody (Invitrogen #11062) was incubated for 1 hour. Images were taken with a Nikon NIU Upright Fluorescent Microscope and DNA fibers were measured using Image J.

DNA damage/repair

DSBs were detected by γ-H2AX immunofluorescence and neutral comet assay as described before (5) with modifications. Briefly, comet assays were performed using the Oxiselect Comet Assay Kit (Cell Biolabs #STA-355) according to the manufacturer’s instructions. Images were acquired by an inverted Olympus IX70 fluorescence microscope using a FITC filter, and the percentage of tail DNA of individual cells was calculated using the OpenComet plugin of ImageJ. HR, D-NHEJ and Alt-NHEJ were measured using DR-GFP (HR), EJ2-GFP (D-NHEJ) and EJ5-GFP (Alt-NHEJ) reporter cassettes as described before (11).

Bone marrow transplantations

Sub-lethally (600 cGy) irradiated CD45.1+ congenic mice were transplanted i.v. with 2.5 x 10⁶ CD45.2+ bone marrow cells from Flt3^ITD, Flt3^ITD;Parp1^−/−, Flt3^ITD;Tet2^−/−, Flt3^ITD;Tet2^−/−;Parp1^−/−, Flt3^ITD;Dnmt3a^−/− and Flt3^ITD;Dnmt3a^−/−;Parp1^−/− mice. Eight weeks later engraftment of CD45.2+ and CD45.2+Lin-Sca1+cKit+ cells was examined by flow cytometry (anti-CD45.2 - #60695, anti-mouse lineage antibody cocktail - #558074, anti-mouse Ly6A/E - #556162, anti-mouse CD117 – 560557; all from BD Pharmingen) and 2.5 x 10⁶ bone marrow cells were transplanted to the sub-lethally irradiated secondary CD45.1+ recipients. Donor engraftment in secondary recipients was analyzed 8-weeks post-transplant as above.

In vivo treatments

Sub-lethally (600 cGy) irradiated CD45.1+ congenic mice were transplanted i.v. with 2 x 10⁶ CD45.2+ bone marrow cells from Flt3^ITD;Tet2^−/−, Flt3^ITD;Dnmt3a^−/− and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− mice. Three weeks later engraftment of CD45.2+ cells was confirmed in peripheral blood of the individual CD45.1+ recipient mice. Engrafted mice were randomized and treated i.v. for 14 consecutive days with vehicle, 1 mg/kg quizartinib and/or 0.33mg/kg talazoparib. CD45.2+ and CD45.2+Lin-Sca1+cKit+ cells were quantified by immunofluorescence in peripheral blood, spleen, and bone marrow 3 days after the end of treatment.

Institutional approvals

The patient samples were collected after obtaining informed consent in concordance with Declaration of Helsinki and were approved by the institutional review boards. Animal experiments were approved by the IACUC of Temple University and Washington University School of Medicine.

Quantification and statistical analysis

All experiments were conducted at least in three independent replications. The mean and standard deviation (SD) and statistical analysis (p-value) was executed by Student t-test in SigmaPlot (Systat software, CA, USA). Data are considered statistically significant when p-value is less than 0.05.

RESULTS

TET2 and DNMT3A mutations exert opposite effect on the sensitivity of malignant cells to PARPi

To begin to assess how co-operating mutations influence the sensitivity of OTK-positive leukemia cells to agents that target DDR mechanisms, clonogenic assays were performed with Lin-CD34+ cells from AML patients with FLT3 ^ITD + NPM1 mutation (FLT3 ^ITD;NPM1^mut) and FLT3 ^ITD;NPM1^mut AML samples carrying TET2 mutation (TET2^mut), DNMT3A mutation (DNMT3A^mut) or both mutations as indicated in Supplementary Table S1. Results showed that FLT3^ITD;NPM1^mut;TET2^mut AML cells were sensitive to PARPi olaparib, whereas FLT3^ITD;NPM1^mut;DNMT3A^mut cells were resistant (Figure 1A, left panel). Remarkably, TET2^mut reversed the resistance phenotype of DNMT3A^mut AML cells, as FLT3^ITD;NPM1^mut;TET2^mut;DNMT3A^mut and FLT3^ITD;NPM1^mut;TET2^mut cells displayed similar sensitivity to olaparib. By contrast, TET2^mut and DNMT3A^mut did not affect the sensitivity of AML cells to the FLT3 kinase inhibitor (FLT3i) quizartinib (Figure 1A, middle panel). Additionally, while FLT3^ITD;NPM1^mut;DNMT3A^mut AML cells expectedly displayed resistance to doxorubicin (20), FLT3(ITD);NPM1^mut;TET2^mut cells displayed similar sensitivity to FLT3^ITD;NPM1^mut cells (Figure 1A, right panel). Moreover, TET2^mut sensitized FLT3^ITD;NPM1^mut and FLT3^ITD;NPM1^mut;DNMT3A^mut cells to the combination of olaparib + quizartinib +/− doxorubicin (Figure 1B).

Figure 1. TET2^mut and DNMT3A^mut pre-determine the sensitivity of OTK-positive hematological malignancies to PARPi combined with TKi and/or cytotoxic drug.

(A, B) Clonogenic activity of Lin-CD34+ cells from individual AML patients carrying FLT3^ITD;NPM1^mut (FN, n=4 in panel A and 3 in panel B), FLT3^ITD;NPM1^mut;TET2^mut (FNT, n=3 in panel A and B), FLT3^ITD;NPM1^mut;DNMT3A^mut (FND, n=4 in panel A and 3 in panel B) and FLT3^ITD;NPM1^mut;TET2^mut;DNMT3A^mut (FNTD, n=3 in panel A and B) treated for 72 h with (A) the indicated concentrations of olaparib, quizartinib or doxorubicin, and (B) quizartinib (Q, 10 nM), olaparib (O, 1.25 μM), doxorubicin (D, 2.5 nM), and combinations. (C, D) Clonogenic activity of Lin-Kit+ BMC from Flt3^ITD (F), Flt3^ITD;Tet2^−/− (FT), Flt3^ITD;Dnmt3a^−/− (FD) and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− (FTD) mice (n=3/genotype) incubated for 72 h with (C) various concentrations of olaparib, and (D) 1 μM olaparib (O), 100 nM quizartinib (Q), 10 nM doxorubicin (D), and combinations. (E) 50:50 mixtures of RFP+ FD : GFP+ FT cells and RFP+ FD : GFP+ FTD cells were incubated with the indicated concentrations of olaparib for 10 days followed by flow cytometry analysis. Results show mean fold change ± SD of the percentage of the indicated cells in suspension when compared to untreated counterparts from triplicate experiments. Panels illustrate indicated cell mixtures after the treatment with 10 μM olaparib or diluent (Control). Pictures were acquired with Evos FL Auto 2 fluorescent microscope (Invitrogen), magnification 10X. (F) Clonogenic activity of BCR-ABL1 (BA), BCR-ABL1;Tet2^−/− (BAT), BCR-ABL1;Dnmt3a^−/− (BAD) and normal (N) murine Lin-Kit+ BMC incubated for 72 h with various concentrations of olaparib. (G, H) Clonogenic activity of JAK2^V617F (J), CALR^del52 (C) and MPL^W515L (M) -positive Tet2^−/− (T) and Dnmt3a^−/− (D) murine Lin-Kit+ BMC incubated with (G) the indicated concentrations of olaparib, and (H) 100 nM ruxolitinib (R), 2.5 μM olaparib (O) and/or 10 μM hydroxyurea (H). (A-D, F-H) Results show mean % ± SD of colonies from all tested samples compared to untreated cells; *p<0.05 in comparison to F and/or FD/JD.

NPM1^mut often accompany FLT3^ITD alone or together with TET2^mut and/or DNMT3A^mut (31). However, NPM1^mut did not alter the sensitivity of leukemia cells to olaparib (Supplementary Figure S1A,B).

Mouse models were generated to validate the unique roles of TET2^mut and DNMT3A^mut in regulating sensitivity of OTK–positive cells to PARPi. To model the somatic loss-of-function mutations found in human blood diseases, mice with conditional inactivation of Dnmt3a and Tet2 in the hematopoietic system were generated. Mice with loxP-flanked exons within Dnmt3a or Tet2 were crossed with Vav-Cre recombinase strain, which is active in all hematopoietic cells from E14.5 of development (henceforth referred to as Dnmt3a^−/− and Tet2^−/− respectively). The resultant mice were further crossed to mice harboring Flt3^ITD mutation. In concordance with human Lin-CD34+ AML cells, Lin-cKit+ bone marrow cells (BMC) from Flt3^ITD;Tet2^−/− and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− mice were highly sensitive to olaparib (Figure 1C) and to the combination of olaparib + quizartinib +/− doxorubicin (Figure 1D) when compared Lin-cKit+ BMC from Flt3^ITD and Flt3^ITD;Dnmt3a^−/− mice. The opposing effects of Tet2 and Dnmt3a mutations on sensitivity to PARPi was further supported by olaparib-induced selective elimination of Flt3^ITD;Tet2^−/− cells and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− cells from 50:50 mixtures with Flt3^ITD;Dnmt3a^−/− cells (Figure 1E).

The PARPi sensitivity phenotype of Tet2^−/− cells was not restricted to co-occurrence with Flt3^ITD as murine Tet2^−/− Lin-cKit+ BM cells transformed with BCR-ABL1, JAK2^V617F, MPL^W515L and CALR^del52 were also more sensitive to olaparib when compared to Dnmt3a^−/− counterparts (Figure 1F, G). In addition, JAK2^V617F;Tet2^−/− cells responded better to the combination of olaparib and ruxolitinib (JAK1/2 inhibitor) and/or hydroxyurea than JAK2^V617F;Dnmt3a^−/− cells (Figure 1H). Moreover, as OTKs signals through NRAS, Tet2^−/− Lin-cKit+ cells expressing the Nras^G12D activating mutation (32) were also highly sensitive to olaparib, compared to Nras^G12D;Dnmt3a^−/− counterparts (Supplementary Figure S1C).

To provide additional evidence that inactivation of TET2 and DNMT3A affect the sensitivity of OTK-positive cells to PARPi, expression of these proteins was inhibited by CRISPR/Cas9 approach in FLT3^ITD-positive 32Dcl3 murine hematopoietic clones (Figure 2A). Targeted deletion of Tet2 or Tet2 and Dnmt3a genes was associated with enhanced sensitivity to olaparib compared to control, whereas deletion of Dnmt3a caused olaparib resistance (Figure 2A). These effects correlated with the magnitude of downregulation of TET2 and DNMT3A in FLT3^ITD-positive 32Dcl3 clones (Figure 2B).

Figure 2. Inhibition of TET2 enzymatic activity enhances the sensitivity of FLT3^ITD-positive murine AML-like cells to PARPi.

(A, B) Trypan blue excluding cells relative to untreated control ± SD after 96 hrs of treatment with various concentrations of olaparib (3 experiments in triplicates). Western blot showing expression of DNMT3A and/or TET2 in FLT3^ITD-positive 32Dcl3 cells (F) and in clones with complete (A) or complete and partial (B) targeted deletion of TET2 (FT, Ft), DNMT3A (FD, Fd) and DNMT3A+TET2 (FTD) by CRISPR/Cas9. (C) Clonogenic activity of GFP+ Lin-cKit+ BMC from FLT3^ITD;Tet2^−/−;Evi1-GFP, FLT3^ITD;Tet2^+/−;Evi1-GFP and FLT3^ITD;Tet2^+/+;Evi1-GFP mice treated for 72 hrs with various concentrations of olaparib. Results show mean % ± SD of cells/colonies compared to untreated cells. (D) Clonogenic activity of murine Lin-cKit+ FLT3^ITD;Tet2^−/− BMC expressing GFP (control), and Flag-tagged TET2-CD and TET2-CD(CM) and treated with 1 μM olaparib for 72 hrs. Results represent % of colonies ± SD compared to untreated cells. *p<0.05 when compared to other groups. Western blot showing expression of the Flag-tagged proteins in total cells lysates. (E) Clonogenic activity of murine Lin-cKit+ FLT3^ITD;Tet2^−/− BMC treated with 1 μM olaparib (O) and/or 250 μM L-AA for 6 days. Results show mean % ± SD of colonies compared to untreated cells. Representative dot-blot detecting 5hmC is shown. (F) Clonogenic activity of murine Lin-cKit+ FLT3^ITD;Dnmt3a2^−/− BMC treated with 1 μM olaparib (O) and/or 500 μM 2-HG for 96 hrs. Results show mean % ± SD of colonies compared to untreated cells; *p<0.05 when compared to O. Representative dot-blot detecting 5hmC is shown. (G) Mean % of colonies ± SD from FLT3^TKD;NPM1^mut;DNMT3A^mut Lin-CD34+ AML cells treated for 96 hrs with 1.25 μM olaparib (O), 50 nM quizartinib (Q) and/or 250 mM 2-HG when compared to untreated cells; * and ** p<0.05 when compared to single and dual treatments, respectively.

Dose-dependent effects of TET2 on PARPi sensitivity may have clinical relevance because both heterozygous and homozygous TET2 mutations have been detected in hematopoietic malignancies (33), which affected 5hmC levels: Tet2+/+ >Tet2+/− >Tet2−/− (34). Clonogenic assays with mouse BMC cells confirmed a dose-dependent effect of Tet2 on sensitivity to olaparib: Flt3^ITD;Tet2^+/+ > Flt3^ITD;Tet2^+/− > Flt3^ITD;Tet2^−/− (Figure 2C).

To determine if the 5-methylcytosine hydroxylase activity of TET2 regulated sensitivity to PARPi, the TET2 catalytic domain (TET2-CD) or catalytically inactive mutant (TET2-CD^CM) (28) were expressed in Flt3^ITD;Tet2^−/− murine Lin-cKit+ BMC cells followed by treatment with olaparib. The results indicated that loss of TET2 enzymatic activity was responsible for sensitizing the cells to PARPi (Figure 2D). However, restoration of TET1/2/3-mediated global 5hmC by L-ascorbic acid (L-AA) in Flt3^ITD;Tet2−/− cells did not result in resistance to olaparib (Figure 2E) suggesting that regulation of the response to PARPi is TET2 dioxygenase -specific.

Incubation of the Flt3^ITD;Dnmt3a^−/− murine Lin-cKit+ BMC with 2-HG, a competitive inhibitor of α-ketoglutarate (α-KG)-dependent dioxygenases including TET1-3 (35), reduced global 5hmC and induced sensitivity of these cells to olaparib (Figure 2F). However, 2-HG did not affect the sensitivity of Flt3^ITD;Tet2^−/− and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− cells to olaparib (Supplementary Figure S2) strongly suggesting that 2-HG –mediated inhibition of TET2 enzymatic activity contributed to this effect. 2-HG treatment also enhanced the sensitivity of FLT3^TKD;NPM1^mut;DNMT3A^mut human primary Lin-CD34+ AML cells to olaparib +/− quizartinib (Figure 2G). Cumulatively, these data suggest loss of enzymatic activity of TET2 modulates the sensitivity of TET2^mut cells to PARPi.

Tet2 and Dnmt3a loss-of-functions determine the therapeutic effect of PARPi +/− TKi against Flt3^ITD-positive murine leukemias in vivo.

To determine the clinical impact of in vitro findings, combination therapies were applied to murine leukemia cells in vivo. Congenic mice (CD45.1+) were transplanted with leukemic BM cells from Flt3^ITD;Tet2^−/−, Flt3^ITD;Dnmt3a^−/− and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− mice (CD45.2+). Representative mice with AML-like disease were chosen as donors of leukemia BM cells from the cohorts described before (18) and also from these characterized in Supplementary Figure S3A-D. Following engraftment of donor cells, recipient mice were treated with vehicle, FLT3i quizartinib, PARPi talazoparib, or quizartinib + talazoparib (Figure 3A). Talazoparib was used here because it displays better pharmacokinetic parameters in mice than olaparib (36). Therapeutic effect was examined by quantification of CD45.2+ cells in peripheral blood, spleen and bone marrow, and CD45.2+ Lin-Sca1+cKit+ stem/progenitor cells in the bone marrow.

Figure 3. Divergent effect of the loss of Tet2 and Dnmt3a on sensitivity of Flt3^ITD-positive murine AML-like cells to talazoparib +/− quizartinib in mice.

(A) Experimental scheme. (B) CD45.1+ congenic mice (n=6-9/group) bearing CD45.2+ Flt3^ITD;Tet2^−/− (FT, green), Flt3^ITD;Dnmt3a^−/− (FD, orange), or Flt3^ITD;Tet2^−/−;Dnmt3a^−/− (FTD, purple) leukemia cells were treated with vehicle (−), 1 mg/kg quizartinib (Q), 0.33 mg/kg talazoparib (T) and Q+T. Results represent mean % ± SD of CD45.2+ and/or CD45.2+Lin-Sca1+cKit+ (LSK) cells in peripheral blood leukocytes (PBL), spleen cells (SPL) and bone marrow cells (BMC) relative to control (vehicle-treated mice). * and ** p<0.05 when compared to control and single treatment, respectively, using One Way ANOVA.

Untreated animals developed myeloid leukemia without maturation-like disease (≥25% Lin-CD45.2+ cells and ≤1% Mac1+, Ter119+, B220+ and CD3e+ cells in bone marrow) (37). Talazoparib alone and especially in combination with quizartinib exerted a strong inhibitory effect against Flt3^ITD;Tet2^−/− AML, and to a lesser degree against Flt3^ITD;Tet2^−/−;Dnmt3a^−/− cells (Figure 3B). Conversely, talazoparib was ineffective against Flt3^ITD;Dnmt3a^−/− cells, nor did it enhance the effect of quizartinib against this genotype.

Flt3^ITD-positive Tet2-mutant leukemia cells accumulate DSBs due to increased speed of replication fork progression and attenuated HR and D-NHEJ activities

Because OTK-positive cells show accumulation of DNA damage, DSBs were examined in mice from the different genotypes as a mechanism to explain the differential responses to PARPi. This was quantified via accumulation of toxic DSBs [by neutral comet assay and histone H2AX phosphorylation on serine 139 = γ-H2AX (38,39)] and cell survival after exposure to olaparib +/− quizartinib and doxorubicin. In general, olaparib +/− quizartinib -treated Flt3^ITD;Tet2^−/− and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− murine Lin-cKit+ cells accumulated significantly more DSBs than their Flt3^ITD;Dnmt3a^−/− counterparts (Figure 4A), which was associated with reduced survival rate of the former cells (Supplementary Figure S4A). Moreover, Flt3^ITD;Tet2^−/− cells treated with the combination of olaparib +/− quizartinib +/− doxorubicin also displayed higher levels of DSBs when compared to Flt3^ITD;Dnmt3a^−/− cells (Supplementary Figure S4B).

Figure 4. Tet2 deletion is associated with accelerated fork progression and impaired DSB repair resulting in accumulation of DSBs in PARPi +/−OTKi treated Flt3^ITD-positive cells.

Flt3^ITD (F), Flt3^ITD;Tet2^−/− (FT), Flt3^ITD;Dnmt3a^−/− (FD) and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− (FTD) murine Lin-cKit+ BMC were tested. (A) Cells were treated with 10 μM olaparib (O) +/− 100 nM quizartinib (Q) for 24 and 48 hrs. DSBs were assessed by γ-H2AX immunofluorescence (upper panels) and neutral comet assay (lower panels). Results represent mean % of control ± SD from triplicates. (B) DNA fibers from cells untreated or treated with 2.5 μM olaparib (O) for 24 hrs. Results represent forks speed; p value was calculated using Mann-Whitney rank sum test. (C) HR, D-NHEJ and Alt-NHEJ activities were measured using DR-GFP, EJ-GFP and EJ2-GFP reporter cassettes, respectively. Results represent mean % ± SD from triplicates. *p<0.05 in comparison to Flt3^ITD cells. (D) Western blot analysis of the nuclear lysates obtained from indicated cells; DDR = DNA damage response, HR = homologous recombination, D-NHEJ = DNA-PK –mediated non-homologous end-joining, and (Alt-NHEJ) = alternative NHEJ.

Olaparib-induced accumulation of DSBs might result from enhanced induction and/or inadequate repair of DSBs. PARPi-triggered DSBs could be generated by increased speed of replication fork progression (40). Replication fork kinetic analyses indicate that olaparib accelerated fork elongation in Flt3^ITD and Flt3^ITD;Tet2^−/− cells, but inhibited fork elongation in Flt3^ITD;Dnmt3a^−/− cells (Figure 4B). Olaparib did not further accelerate already high speed fork progression in Flt3^ITD;Tet2^−/−;Dnmt3a^−/− cells.

The impact of Tet2 and Dnmt3a loss-of-function on DSB repair was tested using specific reporter cassettes measuring HR, D-NHEJ and Alt-NHEJ repair activities as described before (11) (Figure 4C). Both HR and D-NHEJ responses were sharply attenuated in Tet2-mutant Flt3^ITD murine Lin-cKit+ BMC compared to Flt3^ITD and Flt3^ITD;Dnmt3a^−/− cells. Conversely, Alt-NHEJ activity was selectively diminished in Flt3^ITD;Dnmt3a^−/− cells.

To evaluate the specific alterations in DSB repair pathways induced by Tet2 and Dnmt3a loss-of-function, Western blot analyses were performed to examine expression of the key components of DDR, including HR, D-NHEJ and Alt-NHEJ. BRCA1 and LIG4 were downregulated in Tet2-mutant mouse AML cells (Figure 4D), in concordance with abrogation of BRCA1-dependent HR and LIG4-dependent D-NHEJ (Figure 4C). PARP1 was downregulated in Flt3^ITD;Dnmt3a^−/− cells (Figure 4D), which underlies their reduced Alt-NHEJ (Figure 5C) and cellular PARylation activity (Supplementary Figure S5).

Figure 5. PARP1 protects Flt3^ITD;Tet2^−/− cells from lethal effect of DSBs.

(A) Genotypes of the bone marrow donor mice. (B) Mean number of colonies ± SD from 10³ LSK bone marrow cells (3 mice/group). *p<0.05 when compared to FT and/or F counterparts. (C) Mean number of CD45.2+ and CD45.2+ Lin-Sca1+cKit+ (LSK) BMC ± SD in bone marrow of sub-lethally irradiated CD45.1+ congenic mice transplanted with 2.5 x 10⁶ bone marrow cells from the indicated mice (6-8 mice/group). *p<0.01 when compared to FT. (D) Mean % ± SD of colonies from triplicates of LSK cells treated with 50 nM quizartinib (Q) +/− 5 nM doxorubicin (D) when compared to untreated cells. *p<0.05 when compared to FT, and FP and FDP counterparts. (E, F) Indicated LSK cells were treated with 100 nM quizartinib (Q) +/− 10 nM doxorubicin (D) for 24 hrs. DSBs were assessed by neutral comet assay (E) and γ-H2AX immunofluorescence (F). Results represent mean % of control ± SD from triplicates.

Mutations in TET2 and DNMT3A are known to deregulate DNA methylation and mRNA expression of genes involved in pathogenesis of leukemia (41). We found that mRNA expression of DSB repair genes, including BRCA1, LIG4 and PARP1 was similar in FLT3^ITD, FLT3^ITD;TET2^mut, FLT3^ITD;DNMT3A^mut and FLT3^ITD;TET2^mut;DNMT3A^mut AMLs (Supplementary Figure S6A). Moreover, single-cell RNA-seq analysis of murine Lin-cKit+ Flt3^ITD, Flt3^ITD;Tet2^−/− and Flt3^ITD;Dnmt3a^−/− BMC did not reveal any significant differences in expression of DSB repair genes except upregulation of Parp1 in Flt3^ITD;Tet2^−/− cells (Supplementary Figure S6B,C; Supplementary Table S3), which however did not translate into upregulation of the protein expression (see Figure 4D). Altogether, we hypothesize that deregulation of BRCA1, LIG4 and PARP1 protein expression in OTK-positive TET2 and/or DNMT3A -mutated hematological malignancies may depend on post-transcriptional modifications.

PARP1 protects Flt3^ITD;Tet2^−/− but not Flt3^ITD;Dnmt3a^−/− leukemia initiating cells from lethal DSBs

The prior data suggest that PARP activity and reliance on the Alt-NHEJ is required to protect OTK-positive TET2^mut leukemia cells from the toxic effect of DSBs. Among the 18 members of the PARP family, only PARP1, PARP2 and PARP3 have been identified to catalyze PARylation during the DNA damage response and to be involved in DSB repair (42). PARP1 does not show a clear preference for any type of strand break, while PARP2 and PARP3 are preferentially activated by DNA strand breaks harboring a 5′ phosphate, suggesting that among PARP1-3, PARP1 plays a more universal role in DSB repair (43).

To examine the role of PARP1 in response to DSBs in hematological cells with Tet2 and Dnmt3a loss-of-function, clonogenic potential, engraftment capability and response to DNA damage was compared in Lin-Sca1+cKit+ (LSK) BMC from mice of the aforementioned genotypes lacking Parp1 (Figure 5A).

The genetic absence of Parp1 significantly reduced the clonogenic potential of LSK cells with Flt3^ITD and Tet2 mutations, exerted only a modest effect on LSK cells with Flt3^ITD alone, but did not affect the clonogenic potential of LSK cells with Flt3^ITD and Dnmt3a mutations (Figure 5B). To assess the impact of Parp1 deletion on LSCs serial bone marrow transplantations were performed. Secondary transplant engraftment capacity of Flt3^ITD;Tet2^−/−;Parp1^−/− BMC was reduced when compared to Flt3^ITD;Tet2^−/− cells (Figure 5C). Conversely, lack of Parp1 had no effect on engraftment capabilities of Flt3^ITD and Flt3^ITD;Dnmt3a^−/− BM. Therefore, PARP1 appears to plan important role in maintaining LSCs of Flt3^ITD;Tet2^−/− but not Flt3^ITD and Flt3^ITD;Dnmt3a^−/− genotype.

Flt3^ITD;Tet2^−/−;Parp1^−/−, but not Flt3^ITD;Parp1^−/− and Flt3^ITD;Dnmt3a^−/−;Parp1^−/− LSK cells were more sensitive to quizartinib + doxorubicin than their Parp1^+/+ counterparts (Figure 5D). These effects were associated with an enhanced accumulation of spontaneous and drug-induced DSBs assessed by neutral comet assay (Figure 5E) and γ-H2AX (Figure 5F). In conclusion, PARP1 plays an important role in maintaining OTK-positive leukemia cells with Tet2, but not Dnmt3a, mutations.

Disruption of TET2 – WT1 axis sensitizes leukemia cells to PARPi.

TET2 interacts with multiple proteins to exert its regulatory role. TET2 mutations detected in AML patients disrupt its interaction with Wilms tumor 1 (WT1) protein (28) and WT1 mutations result in loss of TET2 function in AML (28). In keeping with this observation, WT1 and TET2 mutations are in general mutually exclusive in AML and suggest that a single mutation in TET2 or WT1 is sufficient to block the activity of the TET2/WT1 complex (44).

To test the role of WT1 mutations (WT1^mut) in sensitivity to PARPi, hTET2-CD wild-type and its enzymatically active but WT1 binding-deficient hTET2-CD^A1505T mutant (28) were expressed in FLT3^ITD;NPM1^mut;TET2^mut Lin-CD34+ primary AML cells. Remarkably, while hTET2-CD reduced the sensitivity to olaparib (as also seen in murine cells, Figure 2D), cells expressing the TET2-CD^A1505T mutant remained sensitive to olaparib (Figure 6A). If WT1 binding to TET2 was important for this phenotype, we hypothesized that WT1^mut AML cells may also be sensitive to PARPi. In concordance, FLT3^ITD;WT1^mut Lin-CD34+ AML primary cells were highly sensitive to olaparib as compared to FLT3^ITD;DNMT3A^mut cells (Figure 6B); the former cells were also sensitive to olaparib +/ quizartinib +/− doxorubicin (Figure 6C).

Figure 6. WT1^mut mimics TET2^mut to increase the sensitivity of FLT3^ITD cells to PARPi.

(A) Clonogenic activity of human FLT3^ITD;NPM1^mut;TET2^mut Lin-CD34+ AML cells expressing mRFP1 (control), mRFP1 + Flag-tagged TET2-CD and mRFP1 + TET2-CD^A1505T mutant treated with 1 μM olaparib. Results from 3 individual AML samples represent % of colonies ± SD in comparison to untreated cells; *p<0.05 when compared to TET2-CD. Western blot showing expression of the Flag-tagged TET2-CD and TET2-CD^A1505T proteins in total cells lysates. (B, C) Clonogenic activity of Lin-CD34+ cells from individual AML patients carrying FLT3^ITD;DNMT3A^mut (FD, n=5) and FLT3^ITD;WT1^mut (FW, n=4) treated with (B) the indicated concentrations of olaparib and (C) quizartinib (Q, 10 nM), olaparib (O, 1.25 μM), doxorubicin (D, 2.5 nM), and combinations. Results show mean % ± SD of colonies compared to untreated cells; *p<0.05 and **p<0.05 in comparison to single- and dual- drug treatment, respectively. (D) Clonogenic activity of murine Flt3^ITD;Dnmt3a^−/− (FD), Flt3^ITD;Tet2^−/− (FT) and Flt3^ITD;Wt1^+/− (FW) Lin-cKit+ BMC treated with the indicated concentrations of olaparib. Results show mean % ± SD of colonies compared to untreated cells. (E) Neutral comet assay from Lin-cKit+ cells from individual AML patients (3 patients/group) carrying FLT3^ITD;DNMT3A^mut (FD), FLT3^ITD;NPM1^mut;TET2^mut (FNT), and FLT3^ITD;WT1^mut (FW) treated with 10 μM olaparib. Results represent mean % tail DNA ± SD; *p<0.05 in comparison to FD cells. (F) HR, D-NHEJ and Alt-NHEJ activities were measured in Lin-CD34+ AML cells using specific reporter cassettes. Results represent mean % ± SD from 3 patients/group; *p<0.05 in comparison to corresponding FD cells. (G) Western analysis of the indicated proteins in the nuclear lysates of Lin-CD34+ AML FD, FW and FNT cells from individual patients #1-6.

To validate the role of WT1 in response of FLT3^ITD-positive AML cells to PARPi, we studied BM from Flt3^ITD;Wt1^+/− haploinsufficient mice, which developed a lethal fully penetrant AML (26). Olaparib sensitivity of murine Flt3^ITD;Wt1^+/− leukemic cells mimicked that of Flt3^ITD;Tet2^−/− cells (Figure 6D).

Similar to FLT3^ITD;NPM1^mut;TET2^mut cells, FLT3^ITD;WT1^mut Lin-CD34+ AML cells accumulated a high number of olaparib-induced DSBs as detected by neutral comet assay when compared to FLT3^ITD;DNMT3A^mut cells (Figure 6E). FLT3^ITD;WT1^mut Lin-CD34+ AML cells also displayed HR, D-NHEJ and Alt-NHEJ repair activities similar to FLT3^ITD;NPM1^mut;TET2^mut cells (Figure 6F). Analogous to mice, BRCA1 and LIG4 were selectively downregulated in Lin-CD34+ FLT3^ITD;WT1^mut and FLT3^ITD;NPM1^mut;TET2^mut AML cells compared to FLT3^ITD;DNMT3A^mut counterparts (Figure 6G). Cumulatively, these data show that WT1^mut mimics the impact of TET2^mut on DSB repair and sensitivity to PARPi in FLT3^ITD –positive AML.

Discussion

PARPi combination therapy with TKi has the potential to become a novel clinical approach against cohorts of hematological malignancies and solid tumors driven by OTKs (45). Since OTKs are usually accompanied by additional mutations, it is paramount to determine if these mutations alter the sensitivity to PARPi (3,6,13,14). Mutations in genes encoding epigenetics-modifying enzymes TET2 and DNMT3A are common in myeloid malignancies such as AML, CML and MPNs (46-48).

We show here that loss-of-function mutations of Tet2 and/or Dnmt3a in murine hematopoietic cells play divergent roles in regulating PARPi sensitivity in hematological malignancies expressing OTKs such as FLT3^ITD, JAK2^V617F, BCR-ABL1, MPL^W515L and in those harboring activating mutations in OTKs-mediated signaling pathways, e.g., CALR^del52 and NRAS^G12D; TET2^mut caused PARPi sensitivity; conversely, DNMT3A^mut induced PARPi resistance. In addition, genetic and biochemical inactivation of TET2 in PARPi-resistant Flt3^ITD;Dnmt3a^−/− cells restored their sensitivity to olaparib. These results from multiple in vitro experimental models translated in to the anti-leukemia activity of PARPi +/− TKi in mice bearing Flt3^ITD;Tet2^−/− and Flt3^ITD;Tet2^−/−;Dnmt3a^−/− leukemias, but not against these with Flt3^ITD;Dnmt3a^−/− AML.

Our observation that TET2-deficiency enhances the sensitivity of OTK-positive leukemia cells to PARPi is supported by recent report demonstrating that CRISPR/Cas9-targeted deletion of TET2 in AML cells lines OCI-AML2 and THP1 was associated with increased sensitivity to olaparib and talazoparib (49). Since each of these cell lines harbor >200 somatic mutations and chromosomal translocations/aberrations (50), specific genetic contribution of TET2 mutations to the sensitivity to PARPi could not be identified. Our findings pinpoint new role of TET2 mutations, divergent from that of DNMT3A mutations, in regulating the DSB repair and sensitivity to PARPi in OTK-induced leukemias. Although we cannot exclude the possibility that other undetected genetic aberrations may affect sensitivity to PARP inhibitors, results from murine genetic models of leukemias mimicking these in primary human malignancies support our conclusion about the specific roles of TET2 and DNMT3A mutations.

While most TET2 mutations result in loss of the protein expression, some are missense mutations inactivating its enzymatic activity while preserving the protein expression (51). Inability of the enzymatically-inactive TET2-CD^CM mutant to restore olaparib resistance in Flt3^ITD;Tet2^−/− cells suggested that OTK-positive AML cells carrying TET2 mutations encoding for enzymatically-inactive protein should be sensitive to PARPi +/− standard treatment (TKi and/or doxorubicin).

Increased sensitivity of OTK-positive TET2^mut leukemia cells to PARPi was associated with accumulation of lethal DSBs due to accelerated fork progression and inhibition of HR and D-NHEJ activities. Downregulation of BRCA1 and LIG4, the key proteins in HR and D-NHEJ, respectively, is in concordance with impaired HR and D-NHEJ in these cells. Altogether, we show that PARPi triggers robust synthetic lethality in OTK-positive TET2^mut malignant hematopoietic cells displaying HR and D-NHEJ deficiencies (11).

RNA-seq performed in naïve and ruxolitinib-treated JAK2^V617F;Tet2^−/− and JAK2^V617F;Tet^+/+ Lin⁻cKit⁺ cells (6) identified numerous pathways affected by Tet2 knockout (Supplementary Figure S7A-D). Among the unique regulators were transcription factors, such as KLF4, NOTCH1, CEBPβ, PML, p53 and MYC, which can modulate DDR to regulate sensitivity of OTK-positive TET2^mut cells to PARPi independently of HR and D-NHEJ (Supplementary Figure S7E,F).

Non-transformed murine Tet2^−/− hematopoietic stem/progenitor cells display a signature of inefficient/aberrant DNA repair (abundant accumulation of transitions and transversions) and were more sensitive to olaparib than their Tet2^+/+ counterparts (49,52). However, Tet2^−/− cells did not accumulate DSBs and their DNA repair defects are most likely due to deregulation of translesion synthesis, mismatch repair and/or glycosylases involved in base excision repair, and may not involve HR and D-NHEJ (53,54). Therefore, mechanisms responsible for PARPi-mediated synthetic lethality in OTK-positive TET2^mut leukemia cells are different from those increasing cytotoxicity of PARPi in non-transformed Tet2−/− hematopoietic cells.

TET2 and WT1 interact to stimulate transactivation of target genes (28). We have shown that WT1 mutations recapitulate the alterations associated with TET2 mutations in DSB metabolism (enhanced accumulation of DSBs, downregulation of BRCA1 and LIG4 proteins, and inhibition of HR and D-NHEJ) and in sensitivity to PARPi of the primary OTK-positive leukemia cells. The impact of the TET2-WT1 axis on DSB repair is also supported by other reports that downregulation of TET2 and WT1 was associated with reduced mRNA expression of BRCA1 in U2OS osteosarcoma cell line and LIG4 in HL60 AML cell line, respectively (13,28).

Collectively, the accumulation of toxic DSBs in FLT3^ITD;Tet2^−/− and FLT3^ITD;Tet2^−/−;Dnmt3a^−/− cells when compared to FLT3^ITD and FLT3^ITD;Dnmt3a−/− cells appears to result from an increased speed of fork progression and/or impaired HR/D-NHEJ repair activities.

Experiments involving knockout mice pinpointed PARP1, which is one of the key elements of Alt-NHEJ, as the predominant factor protecting HR/D-NHEJ –deficient Flt3^ITD;Tet2^−/−, but not HR/D-NHEJ –proficient Flt3^ITD;Dnmt3a^−/− leukemia initiating cells from toxic DSBs. However, in addition to DNA repair, PARP1 has been broadly implicated in regulation of epigenetic events, thus we cannot exclude the possibility that the detrimental effect of Tet2 and Parp1 double knockout in Flt3^ITD-positive cells depends also on abnormal histone modification, chromatin state and/or transcriptional regulation (55).

It has been reported that leukemia cells harboring IDH1/2 mutants, which generate high levels of 2-hydroxyglutarate (2-HG), a competitive inhibitor of α-ketoglutarate –dependent dioxygenases (e.g., TET1-3 and KDM4A/B), or cells treated with 2-HG were sensitive to PARPi (14,56). 2-HG –mediated PARPi-triggered synthetic lethality was due to downregulation of DNA damage sensor ATM and inhibition of RAD51 in HR, but these effects depended on inhibition of histone demethylases KDM4A/B, but not of TET2.

In conclusion, mutations in TET2 and DNMT3A exert opposite effects on the sensitivity to PARPi in malignant hematopoietic cells displaying activation of OTK-mediated signaling (Supplementary Figure S8A). Since mutations in TET2 are found in cohorts of OTK-positive AML, MPN and CML (Supplementary Figure S8B) we postulate that PARPi may have a broad clinical application. In addition to mutations, numerous leukemias may display functional deficiency of TET2 due to reduced expression levels of non-mutated gene (Supplementary Figure S8C-F), further increasing the spectrum of patients, which might benefit from such an approach. Moreover, OTKs (e.g., BCR-ABL1) can promote cytoplasmic compartmentalization of TET2 thus leading to its loss-of-function (57).

We recently showed that TKi-treated leukemia cells displaying deficiencies in HR and D-NHEJ were less sensitive to PARPi in the in vitro bone marrow microenvironment (BMM) mimicking conditions and in vivo in immunodeficient mice (58). However, the protective effect of BMM could be at least partially abrogated by longer treatment regimen with higher doses of PARPi and/or by anti-tumor immune response elicited by PARPi in immunocompetent mice [this work and (59)]. In conclusion, our report warrants clinical trial testing PARPi +/− TKi for OTK-positive TET2^mut or WT1^mut hematological malignancies.