Human NK92 MI cells and human acute myeloid leukemia (AML) cell line HL60 were obtained from Shanghai Institute of Biochemistry and Cell Biology, Chinese Academy of Sciences (Shanghai, China); human AML cell lines KG-1α and Kasumi-1 were purchased from ATCC (American Type Culture Collection) (Manassas, VA). NK92 MI were cultured in α-MEM (Alpha Minimum Essential Medium), complemented with 12.5% fetal bovine serum (FBS) (Gibco, Australia) and 12.5% horse serum (Gibco, Australia). HL60, Kasumi-1 and KG1α were cultured in IMDM (Iscove’s Modified Dulbecco’s Medium) complemented with 15% FBS (Gibco, Australia).

Primary AML cells were obtained from the bone marrow (BM) samples in clinical AML inpatients in the department of hemopathy in the First Affiliated Hospital of Gannan Medical University. The research protocol had been reviewed and approved by the Research Ethics Committee of the First Affiliated Hospital of Gannan Medical University, with Ethics No. LLSC-2023-353. Briefly, mononuclear cells were separated from BM by density-gradient centrifugation with Ficoll-Hypaque (Cytiva, Uppsala, Sweden), and cultured in IMDM (HyClone, Logan, Utah, USA) supplemented with 15% FBS (Gibco, Australia), 4 mM L-Glutamine (Gibco, Brazil), 100 ng/mL human IL-3 (Proteintech, China), 10 ng/mL human IL-6 (Proteintech, China), 50 ng/mL human M-CSF (Proteintech, China), TPO (Proteintech, China), 50 ng/mL human FLT3-ligand (Proteintech, China), 50 ng/mL human SCF (Proteintech, China) and 1% penicillin and streptomycin.

Human peripheral blood mononuclear cells (PBMCs) were firstly separated by density-gradient centrifugation with Ficoll-Hypaque (Cytiva, Uppsala, Sweden) from healthy donors, and then primary NK cells were isolated from PBMC by human NK Cell Isolation Kit (Cat#: 130-092-657, Miltenyi Biotec) via untouched negative MACS Separation as instructions. The purity of NK cells was analyzed by flow cytometry (see details in the supplementary information). Primary NK cells were cultured in RPMI-1640 Medium, complemented with 10% FBS (Gibco, Australia), 200 IU/mL rhIL-2 (Beyotime, China), 10 ng/mL rhIL-15 (Beyotime, China) and 1% penicillin and streptomycin.

The LSCs were isolated from KG-1α cells and primary AML cells by CD34⁺CD38^– Cell Isolation Kit (Cat#: 130-114-822, Miltenyi Biotec) via MACS Separation as instructions. Briefly, KG-1α cells or primary AML cells were collected in cold medium by centrifuge at 300×g for 10 minutes, following by the steps sequentially: First magnetic labeling (incubation with CD34 MultiSort MicroBeads)-First magnetic separation-Removal of MultiSort MicroBeads and second magnetic labeling (incubation with CD34 MultiSort MicroBeads and CD38 MicroBeads)-Second magnetic separation with LS Columns. After these steps, the target LSCs expressed CD34⁺CD38^– were isolated, and analyzed by flow cytometry to identify the expression level of CD34 and CD38 (see details in the supplementary information).

Inhibitors for PARP1/2 (Talazoparib Tosylate, Cat#: HY-108413) and ATM (CGK733, Cat#: HY-15520) were purchased from MCE. Antibodies against PARP1 (Cat #: 66520-1-Ig, Proteintech), cGAS (Cat#: 79978, Cell signaling technology), STING (Ca#: 13647, Cell signaling technology), phospho-STING (Ser366) (Cat#:72650, Cell signaling technology), phospho-TBK1 (Ser172) (Cat#: 5483, Cell signaling technology), TBK1 (Cat#: 3504, Cell signaling technology), IRF3(Cat#: 4302, Cell signaling technology), phospho-IRF3 (Ser396) (Cat#: 4947, Cell signaling technology), ATM (Cat#: 27156-1-AP, Proteintech), phospho-ATM (Ser1981) (Cat#: 5883, Cell signaling technology), phospho-γH2AX (Ser139) (Cat#: 83307-2-RR, Proteintech), H2AX (Cat#: 39689, Proteintech), GAPDH (Cat#: 60004-1-Ig, Proteintech), α-Tubulin (Cat#: 14555-1-AP, Proteintech) were used as primary antibodies in the western blotting assay; Goat Anti-Mouse IgG antibody, HRP conjugated (Cat#: 2900264, Millipore), and Goat Anti-Rabbit IgG antibody, HRP conjugated (Cat#: 3256751, Millipore) were used as second antibodies. Primary antibodies against human NKG2D ligands (MICA/B, ULBP1, ULBP2/5/6, and ULBP3) using in flow cytometry were purchased from R&D System, with Cat. No. as follows: FAB13001A, FAB1380A, FAB1298A, and FAB1517A, respectively; mouse anti-NKG2D antibody (Cat#: MAB139) used in functional blockade for NKG2D receptor and isotype control Goat anti-mouse IgG2A-APC (Cat#: IC003A) were obtained from R&D System. Mouse IgG (Cat#: SA00001-1) was purchased from Proteintech.

LSCs treated with PARPi were lysed on ice with RIPA buffer supplemented with Protease Inhibitors, phosphatase inhibitors (Cat#: 4693116001, Roche) and Phenylmethanesulfonyl fluoride (Cat#: HY-B0496, MEK). After a quantification for protein concentration using BCA assay kit (Cat#: 23227, Thermo Fisher), the lysates were subjected to the SDS-PAGE electrophoresis. The WB analysis was performed as previously described (20). The blots were observed using a Chemiluminescence Imaging System (SCG-W3000, Servicebio)

Restriction endonuclease and T4 DNA ligase were obtained from New England Biolabs (Ipswich, MA). The pLKO.1 vector was used to construct vectors expressed shRNAs targeting human PARP1, IRF3 and STING, with sequences synthesized by Genecreate (Wuhan, China) as follows: PARP1-1#: CTTCGTTAGAATGTCTGCCTT, PARP1-2#: CTCTCAAATCGCTTTTACA; STING-1#: GTCCAGGACTTGACATCTTAA, STING-2#: CATGGTCATATTACATCGGAT; IRF3-1#: GATCTGATTACCTTCACGGAA, IRF3-2#: GCAGGAGGATTTCGGAATCTT. The shRNA-carrying lentiviruses were produced in HEK293T cells by co-transfecting with packaging plasmids pCMV-dR8.91 and pMD2.G. PARP1, STING or IRF3 stable knockdown LSCs were established by lentiviruses infection, puromycin screening, and limited dilution.

Promoter fragments of human ULBP1 (601 bp), ULBP3 (501 bp), MICA (647 bp) and MICB (611 bp) were amplified by PCR from human genomic DNA using the following primers: ULBP1 forward, 5’-TGCGGTACCCCATTCACTCCCACAACTG-3’, ULBP1 reverse, CTGAAGCTTAGGAGGGGCCTGTGTGCAC; ULBP3: forward, 5’-CGCGGTACCAAAACATGCCTGAGTCTTG-3’, ULBP3 reverse, 5’-CGTAGATCTCCCCAAGCAGGACAGGAG-3’; MICA forward, 5’-TGCGGTACCACTCACCCGGATCCGAATC-3’, MICA reverse, 5’-CTAAAGCTTGGCCCAGCGCGGTCGCTTAC-3’; MICB forward, 5’-TGCGGTACCTCAGATTCCAGATCGCGCTC-3’, MICB reverse, 5’-CTGAAGCTTGCGCGGTCGCTCACAAAAT-3’. The PCR products were subcloned into the KpnI and HindIII or BglII sites of pGL3-Basic vector (Promega). The integrity of constructs was confirmed via DNA sequencing by Genecreate (Wuhan, China).

The ULBP1-, ULBP3-, MICA-, MICB-promoter reporter plasmid or pGL3-Basic plasmid and the internal control plasmid pRL-TK (Promega), along with pCMV-flag-IRF3/5D plasmid (a gift plasmid) (21) or pCMV-flag empty vector were transfected into HEK293T cells using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. After 24 hours, luciferase assays were performed with a dual-specific luciferase assay kit (Cat #: E1910, Promega) as described in our previous studies (22).

The protein level of NKG2D ligands expressed on cell membrane surface was detected by flow cytometry using FACS Calibur flow cytometer (BD Biosciences) as previously described (23). Data was analyzed by FlowJo_v10.9.0 analysis software.

Total RNA was extracted from LSCs using the Total RNA Extraction Reagent (Cat#: BSC51M1, BioFlux). The samples were subjected to reverse transcription using First Strand cDNA Synthesis Kit (Roche). SYBR Master Mix (Cat#: BSB25L1B, BioFlux) was used in a two-step PCR program (94 °C for 15 s and 60 °C for 60 s) for quantitative PCR. GAPDH was applied as internal standardizer to normalize the gene transcriptional level. Primers targeted human NKG2D ligands were listed as previously described (20).

The cytotoxic activity of NK92 MI or primary NK cells was examined by Cytotoxicity Assay Kit as previously described (20). Briefly, NK92 MI cells pretreated with NKG2D neutralizing antibody or control IgG were co-incubated with target cells at the indicated effector (E, NK92 MI cells)/target (T, AML or LSCs) ratio in CO₂ incubator for 4 hours. Then lactate dehydrogenase (LDH) released from the targeted cells killed by NK92 MI were detected followed the kit instruction (Cat#: G1780, Promega).

After the treatment of PARPi, LSCs were placed on glass slides and fixed with 4% paraformaldehyde for 20 min. Cells were permeabilized with 0.1% Triton in PBS for 30 min at room temperature, and incubated overnight in blocking solution containing antibody against γH2AX (Cat#: 83307-2-RR, Proteintech). Fluorescein (FITC)-labeled Goat Anti-Mouse IgG (Cat#: SA00003-2, Proteintech) was employed as a secondary antibody for observation. The nucleus was dyed with DAPI, and evaluated via a Zeiss fluorescence microscope. Images were processed via Adobe Photoshop.

Microarray data GSE127200 came from the Gene Expression Omnibus (GEO) was chosen to analyze correlation between PARP1 and NKG2D ligands. The effects of NKG2D ligands on the survival of AML patients were analyzed through the online survival analysis database Kaplan Meier plot from TCGA (http://kmplot.com/analysis/index.php?p=background); the effect of PARP1 on AML patients survival was conducted in UALCAN (http://ualcan.path.uab.edu/index.html) according to the instruction.

Data was shown as mean ± standard deviation. Two-tailed Student’s t-test was performed for statistical analysis between two groups; one-way ANOVA analysis was used for statistical analysis among groups more than three.

NKG2D ligands are widely expressed on tumor cell surface in solid and hematologic malignancies, which are closely related to its immunogenicity responding to NK cells. However, during the progression of cancers, the expression level of NKG2D ligands is downregulated or absent on cell surface, leading to an immune evasion from the surveillance of NK cells. AML is one of the most severe hematologic tumors with high incidence to be relapsed. It is predicated that the absence of NKG2D ligands may be an immune escape strategy for AML to break the immune balance to promote the development and progression of tumor. To find the expression level of various subtypes of NKG2D ligands on AML patients, primary AML cells were isolated from the BM and checked by flow cytometry. As data shown in Figures 1A, B, MICA and MICB were nearly undetectable on AML cell surface from all patients, with ULBP1, ULBP2 and ULBP3 also low expressed. To further evaluate the correlation between NKG2D ligands expression level and AML patients’ prognosis, the survival analysis plotters from AML patients with low and high NKG2D ligands were drawn based on the dada in public online database (http://kmplot.com). It was shown that although the survival time was not significantly different between ULBP1 low expression and high expression AML patients (P = 0.49) (Figure 1C), patients with high level of ULBP2, ULBP2, MICA and MICB gained a longer survival time than the low-expressed patients (Figures 1D–G), indicating that NKG2D ligands could be good predictors for prognosis in AML patients.

NK cells-mediated immunotherapy is a promising treatment against leukemia. The relapse of AML is traced back to the presence of LSCs in the bone marrow tumor microenvironment (24). LSCs is a unique population of therapy-resistant leukemia cells with long-term self-renewal capacity (25). Therefore, the targeted therapy against LSCs is considered to be an effective treatment to cure leukemia.

It was reported that the NKG2D ligands are absent on the cell surface of LSCs (26). KG-1α is one type of stem cell-like, which is reported to be resistant to most chemotherapeutic drugs (27). We isolated LSCs from KG-1α by magnetic beads to obtain the CD34 positive expression and CD38 negative expression cell population (Supplementary Figure S1). Thereafter, we checked the expression level of subtypes ligands by flow cytometry. It was found that, all five subtypes of ligands including ULBL1~3, and MICA/B were not detected on the surface of LSCs, while the NKG2D ligands were expressed on AML cell lines including HL60 and Kasumi-1 (Figure 2A). Furthermore, we examined whether LSCs could arouse the cytotoxic activation of NK92 MI through the NKG2D receptor-ligands axis. As shown in Figures 2B, C, compared with HL60 cells, NK92 MI cells exhibited poorer cytotoxicity against KG-1α; especially when co-incubated with LSCs, NK92 MI cells nearly exhibited no cytotoxicity, indicating that LSCs lost their ability to sensitize NK cells’ activity. Altogether, these data demonstrated that the absence of NKG2D ligands assisted LSCs to escape from NK cells’ immune killing.

Multiple signaling pathways are implicated in the modulation of NKG2D ligands expression in various cancers. However, the regulatory mechanisms were not clearly clarified in LSCs. Some cancer treatments especially chemotherapy and targeted therapies were widely reported to induce NKG2D ligands expression in tumor cells. To explore the potential regulatory strategy in LCSs, a broad screening of small molecules that promoted NKG2D ligands expression in LSCs derived from KG-1α was performed by quantitative real-time PCR. Intriguingly, PARP1/2 inhibitor Talazoparib Tosylate was demonstrated to significantly induce the transcription of ULBP1, ULBP2 and ULBP3 (parts of results from the screening were shown in Figure 3A). To further confirm the effects of PARPi on NKG2D ligands, two other PARP1/2 inhibitors Olaparib and Niraparib were also employed in LSCs. It was shown that all these PARP inhibitors obviously promoted the transcription of NKG2D ligands in LSCs (Supplementary Figure S2). In primary LSCs (Supplementary Figure S3), the transcriptional levels of NKG2D ligands were also found to be upregulated (Figure 3B).

Meanwhile, the cytotoxic activity of primary NK cells from healthy donor (Supplementary Figure S4) was examined when co-incubated with LSCs pretreated with PARPi (Talazoparib Tosylate). As data shown in Figure 3C, under the treatment of PAPRi, LSCs significantly enhanced the cytotoxicity of primary NK cells; however, the blockade of NKG2D receptor by primary antibody against NKG2D in primary NK cells obviously reversed its cytotoxicity towards LSCs even with the treatment of PARPi, indicating that the improved cytotoxic activity for primary NK cells against LSCs was dependent on the increasing expression of NKG2D ligands in LSCs, as well as the binding of ligands to NKG2D receptor.

To understand the role of PARP1 on the regulation of NKG2D ligands in AML, we chose a microarray dataset GSE127200 to analyze the correlation between PARP1 and NKG2D ligands from AML patients. As data shown in Figure 3D, the transcription of ULBP1, ULBP2, ULBP3, MICA and MICB were negatively correlated with PRAP1 in NKG2D ligands-negative cells. We further measured the transcriptional mRNA level of PARP1 and NKG2D ligands in AML cells from our hospitalized patients using qPCR assay, and three groups: low, medium, and high were divided according to the transcriptional level of PARP1 (Supplementary Table S1). As data shown in Figure 3E, the transcription of ULBP1, ULBP2, ULBP3, MICA and MICB was decreased with the increasing of PARP1, suggesting that PARP1 was also negatively correlated with NKG2D ligands in our AML samples.

We next checked the protein expression level of PARP1 in AML patients. As data shown in Figures 3F, G, compared with PBMCs from the healthy donors, PARP1 was significantly increased in primary AML cells derived from BM in AML patients (Supplementary Figure S5). Also, AML patients with PARP1 low expression had a longer survival compared with those PARP1 high expression patients (Figure 3H), indicating that PARP1 was a poor prognostic factor for AML patients. Taken together, these data indicated that PARP1 displayed a negative correlation with NKG2D ligands in AML patients. The inhibition of PAPR1 in LSCs would increase NK cell-mediated cytotoxicity towards LSCs.

We further detected the regulatory roles of PARP1 on NKG2D ligands in LCSs. A PARP1stable knock-down cell line was established in LSCs derived from KG-1α cells, and the knock-down efficiency was checked by WB (Figure 4A; Supplementary Figure S6). The knock-down of PARP1 could obviously increase the expression level of ULBP1, ULBP2, and ULBP3 in LSCs (Figure 4B).

PARP1 is a pivotal regulator in the DNA repair process. The inhibition of PARP1 will result in the accumulation of damaged DNA, which further activates the DDR pathways (28). We next detected whether the induction of NKG2D ligands in LSCs was associated with the activation of DDR pathway caused by PAPR1 inhibition. The immunofluorescence images showed that the inhibition of PARP1 increased γH2AX foci accumulation (Figure 4C). In addition, WB analysis revealed that the inhibition of PARP1 elevated γH2AX expression, as well as the phosphorylated forms ATM (Figures 4D, E; Supplementary Figure S7). However, compared with the inhibition of PARP1 itself, the combined inhibition of ATM (CGK733) impeded the upregulation of NKG2D ligands expression induced by PARP1 inhibitor (Figure 4F). Hence, these data suggested that the knock-down or inhibition of PARP1 activated the intracellular DDR pathway, which further increased the expression of NKG2D ligands in LSCs.

The accumulation of damaged DNA in the cytoplasm will activate the cytosolic DNA sensors. We therefore assessed their activation and found that the cGAS expression level and phosphorylated STING were significantly increased upon the treatment of PARP1 inhibitor in LSCs; meanwhile, the downstream major signals including TBK1 and IRF3 were also found to be activated with phosphorylated forms increased (Figures 5A, B; Supplementary Figure S8). Besides, the activation of cGAS/STING/IRF3 signaling upon the treatment of PARPi was also checked in primary LSCs isolated from AML patients. As shown in Figure 5C, the treatment of PARPi significantly increased γH2AX expression, as well as the phosphorylated-STING, TBK1 and IRF3 in primary LSCs (Supplementary Figure S9), which were consistent with the results from LSCs cell line. Taken together, it was suggested that the inhibition of PARP1 in LSCs could stimulate the activation of cGAS-STING signaling pathway.

To elucidate whether the cGAS-STING signaling pathway was involved in the upregulation of NKG2D ligands in LSCs mediated by PAPR1 inhibition, we constructed various shRNA targeting STING and IRF3, respectively (Figures 5D, E; Supplementary Figures S10, S11). Results from flow cytometry indicated that, in wild type LSCs, the treatment of PARP1 inhibitor could upregulate NKG2D ligands expression on cell surface; while in STING or IRF3 knock-down cells, the upregulation effects were not observed upon the inhibition of PARP1 (Figures 5F, G). IRF3 is an important transcriptional regulatory factor that is involved in type I interferon (IFN-I)-mediated innate immune response via a direct binding to the promoter of targeting gene. To determine whether IRF3 also acted as a transcription factor to regulate the transcription of NKG2D ligand genes, we predicted the binding sequences for IRF3 in their promoter regions in the JASPAR database (Supplementary Figure S12). According to the predicted results, we constructed luciferase reporter genes system to perform the DLR assays to measure the ULBP1, ULBP3, MICA and MICB promoter activity stimulated by IRF3/D5 (an activated form of IRF3). As shown in Figure 5H, transfection of IRF3/5D plasmid could significantly stimulate the activation of ULBP1, ULBP3, MICA and MICB promoter, indicating that IRF3 is a potential transcriptional regulatory factor for the transcription of NKG2D ligands. All these data demonstrated that the inhibition of PARP1 could activate the cGAS/STING/IRF3 signaling pathway and subsequently promote the expression of NKG2D ligands in LSCs.