NF-κB and Poly (ADP-ribose) Polymerase 1 Form a Positive Feedback Loop that Regulates DNA Repair in Acute Myeloid Leukemia Cells

This article is featured in Highlights of This Issue, p. 667

Acute myeloid leukemia (AML) is a highly aggressive hematologic malignancy characterized by the overproduction of immature white blood cells (1). Despite important advances in our understanding of the molecular basis of AML, survival outcomes of patients with AML have not improved over the past 20 years (2), indicating the need to develop novel therapies that are more effective and less toxic.

DNA damage is the common mechanism induced by radiotherapy and chemotherapy in the clinical treatment of cancer (3). Although therapy-induced DNA damage is widespread, AML cells utilize endogenous DNA repair mechanisms to enable their survival, recurrence, and resistance (4). In addition, enhancement of DNA repair pathways in chemotherapeutic-resistant AML cells is often considered as the major molecular change that alters its sensitivity to DNA-damaging agents. Given the key roles of DNA repair pathways in chemoresistance, it has been proposed to inhibit DNA repair as a rational sensitization method to improve genotoxicity therapy.

Constitutive NF-κB pathway activation has been found in different types of AML (5). DNA-damaging agents activate NF-κB initiating in the nucleus instead of via membrane-bound receptors (6–8). NF-κB activation induced by DNA damage is necessary for secondary resistance in AML cells (9, 10). Because efficient repair of DNA damage is required for the survival of AML cells (11, 12), we investigated whether NF-κB pathway activation also regulates DNA repair to promote cell survival after chemotherapy.

PARP1 is a nuclear protein that is mainly known for its ability to facilitate DNA repair by catalyzing the poly ADP-ribosylation (PARylation) of itself and other repair-associated proteins (13, 14). PARP1 hyperactivation in DNA repair is critical for the resistance to genotoxic agents, which has been confirmed not only in cancer cell lines and xenografts, but also in several clinical studies (15–17). In addition, PARP1 inhibitors are a new type of anticancer drugs that selectively kills cancer cells with homologous recombination (HR) repair defects (18, 19). However, deficient HR and the rapid emergence of resistance have largely limited the clinic usage of PARP1 inhibition (20–22). Thus, there is an urgent need to develop novel therapeutic approaches to extend the efficacy of PARP1 inhibitor–based therapies.

Because both NF-κB and PARP1 activities are indispensable to the establishment of resistance to genotoxic agents, we sought to explore the regulatory link between these two DNA repair components. Previous work identified NF-κB as a component of the DNA damage signaling pathway initiated by PARP1 and ATM that leads to IKKγ phosphorylation, inhibitor κBα (IκBα) degradation, and ultimately nuclear translocation of the NF-κB subunit RELA (23–25). Conversely, here we showed that RELA knockdown causes the downregulation of PARP1 expression. Our results demonstrate that NF-κB, as a critical transcription factor for the DNA damage response, can regulate PARP1 transcription and thus form a feedback loop. On the basis of this mechanism, we showed that combining a PARP inhibitor with an NF-κB inhibitor resulted in robust synergy in in vitro and in vivo AML models by decreasing DNA repair efficacy and increasing DNA damage accumulation and apoptosis induction. These findings suggest that dual inhibition of NF-κB and PARP1 may be an effective approach to increase the efficacy of DNA damage agents such as daunorubicin by blocking DNA repair jointly in AML therapy.

The AML cell lines KG1α and Kasumi-1 were obtained from ATCC, where they were characterized by DNA fingerprinting, Mycoplasma detection, and cell vitality detection. These cell lines were immediately expanded and frozen. They were cultured in complete Iscove's modified Dulbecco's medium (IMDM; Gibco) and RPMI1640 medium (Gibco), respectively, each medium was supplemented with 10% FBS (Gibco) and 1% penicillin–streptomycin. The cells were maintained in a humidified incubator at 37°C with 5% CO₂.

The IKKβ inhibitor BMS-345541 (BMS) and the PARP1 inhibitor olaparib were obtained from Meilunbio Co., Ltd. BMS and olaparib were separately dissolved in dimethyl sulfoxide to achieve a concentration of 20 mmol/L, and then each solution was serially diluted to specific concentrations. Daunorubicin was purchased from Zhejiang Hisun Pharmaceutical Co., Ltd.

HR repair was measured in AML/DR-GFP cells, according to the previous publications as shown in Fig. 2A (26, 27). Specifically, 6 × 10⁵ AML cells were cotransfected with 2 μg of pDR-GFP (Addgene, catalog no 46085) and 2 μg of pCBASceI (Addgene, catalog no26477) plasmid DNA using Nucleofector solution and the appropriate Nucleofector program. After 48 hours of transfection, the cells were incubated with daunorubicin for 2 hours, washed, and incubated with vehicle, BMS, or olaparib for 12 hours. The cells were then collected and stained with 7-AAD (ClonTech) for 15 minutes at room temperature. The percentages of 7-AAD–negative and GFP-positive cells were analyzed by flow cytometry.

The nonhomologous end-joining (NHEJ) assay was carried out using the EJ5-GFP reporter assay, according to previous publications (Fig. 2B; refs. 27, 28). AML cells were treated as described in the HR repair assays, and the percentage of GFP-positive cells in AML cells was analyzed after cotransfection of EJ5-GFP (Addgene, catalog no 44026) and I-SceI to measure NHEJ activity.

Cells were incubated with 10 μg/mL anti-phospho-Histone H2AX (γH2AX; Ser139; BD Biosciences, catalog no 562253) conjugated to Alexa Fluor 647 for 20 minutes at room temperature and then analyzed using a flow cytometer (BD FACSCanto II). Data were acquired from 10,000 gated events.

The confocal microscopy assay was performed as described previously (29). Briefly, after washing with ice-cold PBS, cells were fixed, permeabilized, and incubated with γH2AX antibodies (Thermo Fisher Scientific, catalog no LF-PA0025) and fluorescence-labeled secondary antibodies. Immunostained cells were examined using a Leica Laser-Scanning Microscope (TCS SP8) with a 63X/1.4 objective. Image quantification was performed using ImageJ software (NIH, Bethesda, MD).

Cells were processed for neutral comet assay using a Comet Assay Kit from Trevigen (catalog no ADI-900-166), according the manufacturer's protocol. Approximately 50 nucleus images per slide were captured and processed by a Zeiss Axio Observer Z1 Microscope (Carl Zeiss). The tail moments from the cells were measured by CASP software.

Total RNA was extracted using BIOzol Reagent (BioFlux, catalog no BSC52M1), and cDNA was synthesized with a PrimeScript RT Reagent Kit (Takara Bio, catalog no RR047A). qPCR was performed in triplicate on a LightCycler 96 Real-Time PCR System (Roche, G10120-100G) with SYBR Premix Ex Taq (Takara Bio, catalog no RR420A). Relative expression was normalized to that of GAPDH by the 2^−ΔΔC_t method. The primers used in this study are shown in Supplementary Table S1.

Western blotting was performed according to standard protocols using a Chemiluminescence Detection System from Clinx Science Instruments. The following primary antibodies were used: anti-RELA (Proteintech, catalog no 10745-1-AP), anti-PAR (Calbiochem, catalog no AM80), and anti-PARP1 (Santa Cruz Biotechnology, catalog no sc-8007), as well as goat anti-rabbit (Proteintech, catalog no SA100001-2) or anti-mouse secondary antibodies (Proteintech, catalog no SA00001-1). Expression levels were normalized to those of the GAPDH mouse mAB (Proteintech, catalog no 60004-1-Ig); these antibodies were used as internal controls. Quantification was performed with ImageJ software.

All transfections were carried out using Lipo2000 (Invitrogen), according to the manufacturer's instructions. A total of 100 ng of PARP1 promoter construct, 50 ng of the expression plasmid pcDNA3.1-RELA, and 20 ng of the pRL-TK Renilla luciferase vector (GenePharma) were used for each transfection in 96-well plates. Firefly luciferase and Renilla luciferase assays were performed using a Dual-Luciferase Reporter Assay System (Promega, catalog no E1910). Activity was normalized according to the Firefly/Renilla ratio.

Experiments were performed using a ChIP Assay Kit according to the manufacturer's protocol (Millipore, catalog no 17-10086). Immunoprecipitation was conducted using an anti-NF-κB P65 antibody-ChIP Grade (Abcam, catalog no 19870) or normal rabbit IgG, followed by precipitation using protein A/G coupled to magna beads. After decrosslinking and protease digestion, DNA fragments were recovered, purified, and used as templates for PCR (TaKaRa, catalog no RR071A); the PARP1 promoter containing RELA-binding sites was amplified; and the PCR products were separated in 2% agarose gels and visualized with SYBR green. The primers used in this study are shown in Supplementary Table S2.

DNA pull-down assay was performed as described previously (30). Briefly, 5 μg of biotinylated double-stranded DNA probes were incubated with 500 μg of whole-cell extracts of KG1α in the presence of streptavidin-conjugated agarose beads and rocked at a gentle speed for 2 hours at room temperature; the mixture was isolated by centrifugation, and the proteins in the complex were dissolved and analyzed by immunoblotting with a RELA antibody. The DNA probes used in this study are shown in Supplementary Table S3.

AML cells were transiently transfected with a pNFκB-MetLuc2 (Clontech, catalog no 631743) construct containing an NF-κB promoter element upstream of the secreted Metridia luciferase (MetLuc) gene, according to the AmaxaTM 4D-NucleofectorTM protocol. Twenty-four hours after transfection, the cells were treated with daunorubicin for 2 hours, washed, and then incubated in medium containing BMS or olaparib alone or in combination for an additional 12 hours. The cell supernatant was harvested for secreted Metridia luciferase assays.

Apoptosis was analyzed with FITC Annexin V/PI Apoptosis Detection Kit I (KGA107, KeyGen Biotech). Collected cells were incubated with 100 μL of Annexin-V/PI staining solution for 15 minutes in the dark, according to the manufacturer's instructions, and analyzed by flow cytometry.

The viability of the treated cells was evaluated by MTT assay. Briefly, AML cells were seeded in 96-well plates at 8 × 10³ per well. MTT solution (20 μL) was added to each well, and then the cells were incubated at 37°C for 4 hours. The absorbance was measured at 570 nm on a microplate reader (Bio-Rad Laboratories, Inc.). Each sample was analyzed in triplicate.

Animal experiments were carried out at the Experimental Animal Center of Fujian Medical University (Fuzhou, China). The experiments were conducted in accordance with institutional guidelines and with prior approval from the Institutional Animal Care and Use Committee. A total of 40 male Balb/c athymic nude mice aged 5 weeks were purchased from the Slack Laboratory Animals Co., Ltd. (Shanghai, China). KG1α cells (5 × 10⁶) in medium containing 1:1 Matrigel (BD Biosciences#356234) were injected subcutaneously into the right flank. Two weeks after injection, the tumor-forming mice were randomly divided into five groups. Tumor volume was measured and calculated using the following formula: [(width)² × (height)]/2. Survival times were observed and analyzed using the Kaplan–Meier method and log-rank test (GraphPad Prism software).

Paraformaldehyde-fixed and paraffin-embedded tumor sections were stained with hematoxylin (Servicebio, catalog no G1004) and Ki67 (1:500 dilution, Abcam, catalog no AB15580) using routine methods. Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) staining was assessed using a TUNEL Kit (Roche, catalog no 11684817910) according to the manufacturer's instructions. Images were recorded with a digital microscope camera (Servicebio, catalog no ML-1600). Image quantification was performed using ImageJ software.

Values are presented as the mean ± SD of at least three independent experiments. Student t test was used to compare two groups of independent samples. One-way ANOVA analysis was used to evaluate the statistical significance of multiple groups. P < 0.05 were considered significant (*); no significant difference (ns).

NF-κB is a heterodimeric complex of REL family proteins (p50, p52, RELA, c-REL, and RELB) that is maintained in an inactive state in the cytoplasm through its interaction with IκB proteins (31). When stimulated, NF-κB is detached from IκBs, and binds to the promoter of its target genes in the nucleus to regulate their transcription (32, 33).

To assess the role of NF-κB in DNA repair in AML cells, we constructed RELA-knockdown KG1α and Kasumi-1 cells using lentiviral shRNA (lentivirus target sequences are shown in Supplementary Table S4) and then treated cells with daunorubicin, which is a DNA intercalating agent that causes DNA strand breaks and a subsequent DNA damage repair response. γH2AX was used as an indicator of DNA double-strand breaks (DSB). The ratio of γH2AX increased greatly after daunorubicin treatment for 2 hours in AML cells but decreased over time after removal of the drug, while RELA knockdown caused more γH2AX accumulation (Fig. 1A–E; Supplementary Fig. S1A–S1C), indicating that inhibiting NF-κB may impair DNA repair capacity in AML cells. To verify the effect of RELA in DNA repair, we further constructed RELA-overexpressing AML cells. The results suggested that RELA overexpression decreased DNA damage and promoted DNA repair, which were exactly opposite the effects of RELA knockdown (Fig. 1F–J; Supplementary Fig. S1D–S1F).

To gain more insight into the mechanisms by which RELA regulates DSB repair, we measured HR and NHEJ repair ability via DR-GFP and EJ5-GFP reporter assays, respectively (Fig. 2A and B). The ratio of GFP-positive cells in RELA-knockdown AML cells cotransfected with the DR-GFP/I-SceI or EJ5-GFP/I-SceI plasmids was significantly decreased (Fig. 2C–D; Supplementary Fig. S2A–S2B). Then, we examined the level of PARylation after DNA damage. The results showed that PARylation was also obviously reduced in RELA-knockdown AML cells (Fig. 2E; Supplementary Fig. S2C). Although there have been some reports that PARP1 regulates NF-κB activity through PAR activation during DNA damage (24, 25), our results demonstrated beyond our expectations that NF-κB can conversely regulate PAR activation.

Accumulating evidence suggests that PARP1 is involved in several DNA repair pathways, including HR, NHEJ, and BER (34); moreover, because RELA is a transcription factor, we focused on the precise role of RELA in regulating PARP1 transcription. RT-PCR and Western blotting results showed that PARP1 mRNA and protein levels were significantly downregulated in RELA-knockdown AML cells, but upregulated in RELA-overexpressing AML cells (Fig. 2F–H). Furthermore, we confirmed these findings using a putative luciferase construct containing the first 2 kb of the PARP1 5′-flanking region in the pGL3-Basic reporter vector. The results showed that transfection of RELA dramatically enhanced the activation of the PARP1 reporter, indicating that PARP1 may be one of the target genes regulated by RELA, which has not been identified previously (Fig. 2I).

To elucidate the underlying molecular basis of this pathway, we identified RELA-binding motifs in the PARP1 promoter using the bioinformatic prediction tool Jaspar (http://jaspar.genereg.net/). We found two potential RELA-binding sites, located at −35/−44 bp and −105/−114 bp, which were designated sites I and II, respectively (Fig. 2J). To investigate the roles of these two putative RELA-response elements in the RELA-mediated transactivation of the PARP1 promoter, we deleted these two sites from the PARP1-2 kb promoter reporter plasmid and then cotransfected the plasmid with the RELA expression plasmid into 293T cells. We found that deleting either site I or site II significantly attenuated transcriptional activity (Fig. 2K). A ChIP assay was performed to further investigate whether RELA regulates PARP1 gene expression through directly binding to the PARP1 promoter. The analysis of PCR products showed that RELA was able to bind to the promoter region of PARP1, covering two RELA-binding sites (Fig. 2L). Moreover, we assessed the potential physiologic recruitment of RELA to the PARP1 promoter using DNA pull-down assays. As shown in Fig. 2M, the endogenous RELA protein in AML cells was efficiently recruited to binding sites I and II. Together, these data suggest that RELA regulates the DNA repair pathway at least partly through modulating the transcriptional activity of the PARP1 gene by binding to its promoter.

To test whether PARP1 plays a critical role in NF-κB activation after DNA damage in AML cells, as reported in other cell types (23–25), we knocked down PAPR1 (lentivirus target sequences are shown in Supplementary Table S4) in KG1α and Kasumi-1 cells and then investigated the effect on NF-κB activity by examining the expression of the pNFκB–MetLuc2 reporter after DNA damage. Our results showed that PARP1 knockdown indeed reduced NF-κB activity during DNA repair (Fig. 3A and B), which was consistent with previous reports. Therefore, NF-κB and PARP1 interfere with each other, which indicate that these two components form a closed loop in DNA repair in AML cells. Next, we knocked down RELA in PARP1-knockdown AML cells and found that double-knockdown of these two targets caused more serious DNA damage accumulation than knockdown of either single target (Fig. 3C–E). These results indicated that dually targeting this positive feedback loop with simultaneous NF-κB and PARP1 inhibition would generate more pronounced therapeutic effects than single inhibition.

To verify the mechanism of positive feedback between NF-κB/RELA and PARP1, as well as the possibility of clinical translation, we used BMS and olaparib, which are small-molecule inhibitors of IKKβ and PARP1/2, respectively (35, 36). We found that NF-κB was activated by daunorubicin-induced DNA damage; BMS obviously inhibited NF-κB activation; and the inhibitory effect of BMS was potentiated in the presence of olaparib (Fig. 4A and B). We further investigated the interaction of these pathways in DNA damage repair. The results showed that the PARylation and HR pathways were activated by daunorubicin in KG1α cells, which might be one of mechanisms underlying the resistance of AML cells to daunorubicin (Fig. 4C–F). Olaparib clearly inhibited PARylation (Fig. 4C) but slightly increased daunorubicin-induced HR activation, which was significantly inhibited by BMS (Fig. 4D and E). In addition to HR, BMS also suppressed PARylation and NHEJ to some extent (Fig. 4C, D, and F). Similar to the results obtained in KG1α cells, daunorubicin activated the PARylation and HR repair pathways (Supplementary Fig. S3A and S3B), and BMS suppressed PARylation, HR, and NHEJ in Kasumi-1 cells (Supplementary Fig. S3A–S3C); however, in contrast to the results obtained in KG1α cells, olaparib decreased HR in Kasumi-1 cells (Supplementary Fig. S3B).

Next, we monitored the clearance of γH2AX in KG1α cells. More γH2AX remained when cells were treated with BMS or olaparib during the repair process, and most of the γH2AX remained following treatment with the combination of olaparib and BMS (Fig. 5A–D). Treatment with BMS and/or olaparib without daunorubicin pretreatment caused negligible DSBs in AML cells (data not shown). These findings were validated by the neutral comet assay based on the tail moment, which is an indicator of DNA damage (Fig. 5E and F). Treatment of Kasumi-1 cells produced similar results (Fig. 5G and H). Collectively, these observations indicated that pharmacologic inhibition of both NF-κB and PARP1 blocks DNA repair.

To determine whether double inhibition of NF-κB and PARP1 influences the tumor response to daunorubicin in vitro, cell apoptosis and proliferation were measured in AML cells. Because NF-κB plays an important antiapoptotic role, we speculated that BMS, which is an upstream inhibitor of NF-κB, would induce apoptosis and sensitize cells to chemotherapy. The results showed that both BMS and olaparib mono-treatment increased apoptosis induction by daunorubicin, and combination treatment further increased the apoptosis ratio at 36 hours after the induction of DSBs (Fig. 6A–D). We then determined whether BMS and olaparib increased the efficacy of daunorubicin in inhibiting proliferation. We treated cells with different concentrations of daunorubicin, either alone or in combination with BMS and olaparib, for 72 hours. Both BMS and olaparib mono-treatment increased the inhibition of proliferation by daunorubicin, and the combination treatment further increased the inhibition ratio (Fig. 6E and F). The cytotoxicity of the single drug in AML cells is shown in Supplementary Fig. S4.

On the basis of the effects observed in vitro, we next addressed whether double inhibition of NF-κB and PARP1 would increase the efficacy of daunorubicin in vivo. A KG1α xenograft mouse model was established via subcutaneous injection. Two weeks after tumor cell injection, the tumor-forming mice were randomized into five groups to receive treatment with vehicle, daunorubicin, BMS, and olaparib alone or in combination every other day until six doses were administered, as depicted in Fig. 7A. The tumor volumes were measured every day during treatment. We found that the KG1α xenograft tumors grew rapidly without drug treatment. Daunorubicin mono-treatment and daunorubicin and BMS or olaparib combination treatment decreased tumor growth slightly; in contrast, the combination of daunorubicin, olaparib, and BMS significantly decreased the tumor volume (Fig. 7A). Survival time analysis showed that daunorubicin mono-treatment had almost no effect on median survival times (34 vs. 29.5 days; P>0.05), while combination treatment with BMS (46 vs. 34 days) or olaparib (44 vs. 34 days) increased the efficacy of daunorubicin to some extent, and the simultaneous combination of all three drugs significantly extended mouse survival times compared with the combination of only two drugs (62 vs. 46/44 days; Fig. 7B). In addition, the treatment caused no obvious systemic toxicity, as assessed by weight loss (Fig. 7C). We performed TUNEL and Ki67 assays of xenograft tumor tissues to measure the apoptosis and proliferation of AML cells in the xenograft tumors; the results shown in Fig. 7D–F suggested that combination treatment with BMS or olaparib increased the efficacy of daunorubicin in proapoptosis and proliferation inhibition to some extent, while the simultaneous combination of all three drugs significantly increased the efficacy of the combination of two drugs, similar to the results obtained in cells. Comprehensive statistical analysis of these data illustrated that the combination of three drugs caused significant tumor regression. Together, double inhibition of NF-κB and PARP1 significantly increased the efficacy of daunorubicin in vivo.

In this study, we provide evidence that RELA can regulate DNA repair after treatment with a DNA-damaging agent, and how the underlying mechanism regulates this pathway was demonstrated by us. We found that RELA regulates DNA repair by binding to the promoter region of the PARP1 gene and affecting PARP1 gene transcription (Fig. 2). RELA knockdown or NF-κB inhibition reduced PARylation in daunorubicin-damaged AML cells. Conversely, PARP1 knockdown reduced NF-κB activity, indicating that NF-κB and PARP1 create a positive feedback loop in DNA repair. This new loop complements to the DNA damage and repair mechanism of NF-κB upon exposure to a DNA-damaging agent. What we have found in human AML cell lines in this study provided an insight into improving treatment for AML.

NF-κB has a variety of physiologic functions, including roles in immune and inflammatory responses, oncogenesis, and responses to genotoxic stress (37–39), as a result, it might cause severe adverse effects like immunodeficiency if it was inhibited nonselectively (40). Thus, selective inhibition of NF-κB is expected to effectively reduce resistance to genotoxic treatments with minimal toxicity. PARP1 inhibitors can selectively inhibit NF-κB activation by radiation and chemotherapeutics, and the inhibitory effect will be amplified by the positive feedback loop between NF-κB and PARP1 observed in this study. Therefore, when used in combination with PARP1 inhibitors, NF-κB inhibitors may resensitize treatment-refractory cancer cells to conventional chemotherapy. In addition, the dose of NF-κB inhibitors can be reduced in combination treatments, thereby minimizing their side effects on other physiologic processes. Moreover, the baseline DNA repair activity varies with tumor type. Therefore, it is vital to understand the intrinsic DNA repair activity of each tumor type prior to DNA-damaging agent treatment. For instance, the deficient HR pathway in BRCA1/2-deficient breast or ovarian cancer creates a distinct vulnerability to PARP inhibitions (41, 42). AML cells are more dependent than another tumor cells on NF-κB, and here, we clarified that NF-κB plays a critical role in DNA repair in AML cells (Figs. 1 and 2A–E). BMS is a potent and specific inhibitor of IKKβ that binds to an allosteric site on the IKKβ catalytic subunit and inhibits IκBα phosphorylation and degradation, thereby inhibiting NF-κB activation (35). The results from our current study revealed that BMS inhibits PARylation, HR, and NHEJ in daunorubicin-damaged AML cells (Fig. 4C–F) and decreases the repair of daunorubicin-induced DSBs (Fig. 5), causing cell apoptosis (Fig. 6).

Exploration of the DNA repair mechanisms and pathways is likely to provide favorable strategies for sensitizing tumor cells to drug therapy. Intracellular repair mechanisms are quite complex, one disordered repair pathway may cause compensatory activation of other repair pathways, leading to incomplete inhibition of DNA repair. Thus, only the simultaneous suppression of multiple pathways can achieve better therapeutic results. Our results showed that olaparib inhibits PARylation, but slightly activates the HR repair pathway in AML cells with daunorubicin damage (Fig. 4C–E). Possible reasons for these findings are as follows: olaparib impairs BER-mediated single-strand break (SSB) repair, and the accumulated SSBs are converted to DSBs during DNA replication (43); because DSBs are primarily repaired via the HR and/or NHEJ pathways to allow cell survival (44), the increased DSBs will activate HR, which may lead to olaparib resistance. Thus, the approach of combining DNA damage agents, a PARP1 inhibitor, and an NF-κB inhibitor may block DNA repair more robustly than any combination of two drugs, and as a result, drug resistance is less likely to occur. Our results demonstrated that the combination of three drugs simultaneously inhibits PARylation, HR, and NHEJ (Fig. 4C–F), effectively decreasing DNA repair, increasing DNA damage accumulation (Fig. 5), and inducing apoptosis, thereby resulting in reduced AML cells proliferation (Fig. 6), delayed disease progression, and increased survival in an AML xenograft tumor model (Fig. 7).

This study on DNA repair regulation by PARP1 and NF-κB in AML cells has significantly improved our understanding of the molecular signaling induced by genotoxic agents. This study has also provided a promising drug combination approach for selectively inhibiting NF-κB activation by chemotherapeutics and DNA damage in AML cells, and this combination may resensitize treatment-refractory AML cells or leukemia-initiating cells to conventional chemotherapy. Patients with AML may benefit from these cost-effective “old” chemotherapeutic drugs. Thus, systematically testing whether the triple combination approach exhibits synergistic therapeutic effects in the clinic may be worthwhile. We hope that this innovative approach will lead to more complete clinical remissions and improved patient outcomes.

No potential conflicts of interest were disclosed.

Conception and design: D. Li, J. Xu, Y. Chen, L. Wu

Development of methodology: D. Li, Y. Luo, X. Chen, Y. Zhuang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Li, Y. Luo, L. Zhang, T. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Li, Y. Luo, Y. Fan

Writing, review, and/or revision of the manuscript: D. Li, L. Wu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Li, L. Wu

Study supervision: Y. Chen, L. Wu

The authors gratefully acknowledge the support of this project by National Natural Science Foundation of China (81872898, 81273541, and 81572662), Joint Funds for the Innovation of Science and Technology, Fujian Province (2016Y9057), Joint Research Program of Health and Education of Fujian Province (WKJ2016-2-33), University Industry Cooperation Project of Fujian Province (2016Y4005), Natural Science Foundation of Fujian Province (2018J01842 and 2017J01823), and Startup Fund for Scientific Research, Fujian Medical University (2016QH014 and 2017XQ2019).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.