AML arises from sequential somatic mutations in a primitive multipotential hematopoietic cell, which is a leukemia subtype with the highest incidence rate, poor outcomes (78, 79). The primary risk factors for AML include obesity, radiation exposure, prolonged exposure to high concentrations of benzene, and chronic tobacco smoke inhalation (80). A small but increasing proportion of AML cases (7% ~ 8%) develop after a patient with lymphoma, or an autoimmune disorder undergoing intensive chemotherapy, especially with alkylating agents platinum derivatives, or topoisomerase II inhibitors (81). According to French-American-British (FAB) classification (82), due to the origin and maturity of the leukemia cells, AML is generally divided into eight subtypes (M0-M7) (82).

LSCs are the initiation cells of AML, which have the ability of self-renewal and multidirectional differentiation to maintain the growth and recurrence of AML (83). LSCs are highly dependent on oxidative phosphorylation and glycolysis contributing to the aggressiveness of AML and are more drug-resistant (83–85). The onset of AML is closely related to gene mutations, such as NPM1, FLT3, IDH1/2, TP53 (86–88), as well as chromosomal rearrangements such as t(8;21) and inv16 (89, 90). AML development and progression are associated with dysregulated immune responses and induction of an immunosuppressive TME (91). Furthermore, AML blasts can hide from immune recognition by promoting T cell exhaustion and expansion of T regulatory cells (Tregs) (91). AML blasts have been found to increase the number of myeloid suppressor cells (MDSCs), polarize macrophages toward a pro-tumoral phenotype, and hamper NK cell effector functions (91). Moreover, AML cells can dysregulate the innate immune response by releasing cytokines and soluble factors or through direct contact with innate immune cells (92). In short, conventional dendritic cells (cDCs) are diminished in the AML BM compared to healthy donors, which may contribute to the lack of CD8⁺ T cells in the TME (93, 94) and increase the proportion of Tregs, MDSCs (95), which promotes the proliferation and metastasis of AML cells. BM stromal cells (BMSCs), specifically, CD73⁺CD105⁺CD271⁺ BMSCs subgroup in AML TME can promote the survival and proliferation of AML cells by secreting growth factors, thereby reducing treatment efficacy (96, 97).

Despite significant advances in conventional therapeutic approaches, including chemotherapy (e.g., cytarabine, anthracycline antibiotic doxorubicin) (98), targeted therapy (e.g., gilteritinib targeting FLT3 mutations) (99), radiation therapy (100), and HSCs transplantation (HSCT) have acquired great success in AML treatment (101), the overall survival rate of AML patients remain suboptimal, with 5-year survival rates ranging from 30% to 40% according to different studies (102–104). Moreover, while T cell-based immunotherapeutic approaches have shown promise in other hematologic malignancies, their efficacy in adult AML has been disappointingly limited, with pediatric trials only in the initial phases (105, 106).

Taken together, the treatment of AML is a long-term and complex process, and the current treatment plan has a certain effect on the alleviation of AML, but is far from achieving the goal of preventing recurrence or even complete cure. Therefore, the development of novel AML treatment methods have important clinical significance.

DNA hypomethylation drugs, currently FDA-approved for AML treatment, demonstrate interesting mechanistic effects: treatment with azacytidine and decitabine for 48 hours could decrease shedding of MICA, MICB, and ULBP2, consequently restoring NK cell function (107). Glycogen Synthase Kinase 3 Beta (GSK3β) expression is elevated in AML-NK cells and GSK3β pharmacological inhibition promotes conjugator formation by up-regulating LFA expression on NK cells and inducing ICAM-1 expression on AML cells, thereby enhancing the cytotoxic activity of AML-NK cells. This process only requires a short ex vivo exposure (16 hours) to 30 μM GSK3β inhibitors (SB415286, LY-2090314, Tideglusib) (108). Overactivation of NK cells by targeting GSK3β may be a novel strategy for the treatment of AML.

The recovery of NK cell functions after allo-HCT has been associated with protection against AML relapse (109) and NK cells have been identified as crucial players in the eradication of AML (110). Furthermore, donor NK cells, along with T cells, play a role in the graft-versus-leukemia (GVL) effect following HSCT for AML (111). In a phase I clinical trial (NCT01898793) ( Table 1 ), adoptive transfer cytokine induced memory like NK (CIML-NK) cells (dose range: 0.5 ~ 10×10⁷ cells/kg) proliferated and expanded in patients with AML and observed in 5 of 9 evaluable patients, including 4 complete responses (112). Another recent phase II clinical study (NCT02782546) ( Table 1 ) involved 15 AML patients who received consolidation therapy, followed by haploidentical HSCT and infusion of CIML-NK cells (dose level: 0.5 ~ 10×10⁶ cells/kg) and 13 patients (87%) achieved composite complete response after 28 days, the median event-free survival for all patients was 3.2 months, and 29% of the participants remained alive after 1 year (113). Another clinical trial (NCT03068819) conducted in 2021 ( Table 1 ), Jeffrey J Bednarski used donor derived CIML-NK cell to treat 8 pediatric and young adult AML patients with HSCT recurrence found that 4 patients achieved complete response CIML-NK cell (dose level: 4 ~ 6×10⁶ cells/kg) infusion. Interestingly, a patient showed sustained remission during a 2-year follow-up after CIML-NK cell infusion without any subsequent treatment (114).

CAR-NK cell therapy offers a promising therapeutic approach for treating AML. Primary CD33-targeting CAR-NK cells strongly reduce the burden of leukemia and prevents BM transplantation of leukemia cells without significant side effects (115, 116) ( Figure 2b ). For example, AML clearance in OCI-AML2-engrafted NSG-SGM3 mice was enhanced by injecting a total of three doses of 1 × 10⁷ of CD33 CAR-NK cells (115). The clinical trial, NCT05008575, currently underway at the Hematology Department of Chongqing Xinqiao Hospital in China targets leukemia cells expressing CD33, employing CAR-NK cells in combination with chemotherapy drugs ( Table 1 ). Among the 10 evaluated patients, only 1 developed grade II CRS and no higher grade CRS occurred. Regarding anti-leukemia efficacy, 60% (6/10) of patients achieved complete remission 28 days after CAR-NK cell infusion (117, 118). In a preclinical study, CD123 CAR-NK cells (5-day OS: 100%) also showed lower acute toxicity than CD123 CAR-T cells (5-day OS: 0%) in a mouse model of transplanted artificial blood cells, while their anti-leukemia efficacy was comparable in a mouse model of AML (119).

Other studies have investigated novel therapeutic strategies to kill AML cells by combination of NK cells, and trispecific killer engager (TriKE) molecules targeting CLEC12A, the AML mice treated with CLEC12A TriKE had significantly less tumor burden compared to tumor alone or tumor with NK cells (120) ( Figure 2b ). Killer immunoglobulin-like receptor-human leukocyte antigen (KIR-HLA) mismatched NK cells (median, 29 × 10⁶/kg NK cells) infusions shown a significant expansion of KIR-mismatched NK cells (median, 5,800/mL of blood on day 14) could decrease relapse rates without increasing mortality in children with AML (121). CD276 (B7-H3) is highly expressed in leukemia cells of AML patients, FC optimized CD276 mAb specifically binds to CD276 on primary AML cells, promoting activation and enhancing anti-AML effects of NK cells (122).

On the whole, preclinical studies using NK cells to treat AML have shown encouraging results. However, challenges remain in ensuring long-term sustainability and mitigating potential off-target toxicity. Future clinical trials will determine the true potential of NK cells to revolutionize the AML treatment landscape.

NK cells in AML patients mainly exhibit impaired killing ability and overall functional impairment of exhausted phenotype ( Figure 2a ) that indicates poor prognosis and high recurrence (123). On the one hand, the content of resident CD3^-CD56⁺ NK cells in AML was decreased compared to healthy donors (124) ( Figure 2a ). NK cells from AML patients typically express more NCRs/NKG2D/DNAM-1, down-regulated NKG2A/iKIRs (113) and the ability of NK cells secreting IFN-γ is significantly reduced which may limit their ability to recognize and clear AML cells and patients with NK cell spectrum defects increased recurrence risk (p = 0.03) without regard for their cytogenetic classification (125). Turk et al. found that down-regulation of renin-angiotensin system (Ras) genes and neurotransmitter genes were involved in NK cell dysfunction in AML (126). For example, ATP6AP2 may modulate NK cell responses through regulation of pH homeostasis, autophagic flux, and NLRP3 inflammasome activation (127, 128). And, down-regulation of arginine may also impair cytokine secretion capacity of NK cells (129). Transcriptomic analysis of BM NK cells from AML patients reveals stress-induced inhibition of NK cell effector function. While CD160 expression is down-regulated in these NK cells, patients with CD160^high NK cells demonstrate significantly improved survival rates (130).

On the other hand, reduction in the number and function, high expression of T cell immunoglobulin and ITIM domain (TIGIT) of NK cells has been implicated in myelodysplastic syndrome (MDS), a condition that can progress to AML (131, 132). Through intercellular contact or secrete IL-10 or TGF-β by Tregs in AML TME contributed to NK cell exhaustion, which accelerates the occurrence of AML (22) ( Figure 2a ). And, AML development has been linked to impairing the BM homing capacity of infused NK cells (133). Unconventional CD56^-CD16⁺ NK cells which decreased expression of NKG2A/NKp30/NKp46 in AML has been observed, indicating adverse clinical outcome (134, 135) ( Figure 2a ), and high expression of NKp46 contribute to the prognosis of AML patients after allo-HSCT (135). Similarly, the results were also found in animal experiments, in the MLL-AF9-induced mouse AML model, more mNKs were detected, but their mature state may not be sufficient to fully exert their anti-AML effects (136).

In summary, NK cell exhaustion promotes the occurrence, progression, and recurrence of AML. Increasing the number of NK cells or activating their function can contribute to the treatment and prognosis of AML. Therefore, NK cell-based treatment strategies for AML show theoretical promise and warrant further investigation.

Significant challenges persist in maximizing the therapeutic potential of NK cells in AML treatment, particularly as AML cells have developed sophisticated mechanisms to evade NK cell-mediated immunosurveillance (137). Claudia Lengerke et al. demonstrated that NKG2D ligands are generally expressed on most of AML cells but not on LSCs, potentially explaining their resistance to NK cell-mediated cytotoxicity and their role in therapeutic resistance (138). In AML, the function of NK cells can be inhibited by various factors, including the presence of multiple immunosuppressive factors in the TME, such as TGF-β and IL-10 ( Figure 2a ), which inhibit the activity and proliferation of NK cells (139). In AML, Tregs and BMSCs typically expand and secrete inhibitory cytokines to reduce the anti-AML activity of NK cells (140, 141). AML cells may express inhibitory ligands such as PD-L1 ( Figure 2b ), which can bind to receptors on the surface of NK cells and inhibit their activity (142, 143). Some AML cells can induce NK cell death by activating the apoptotic pathway (144, 145) and overexpressing inhibitory immune molecules LILRB4 ( Figure 2b ), directly inhibiting the activity of NK cells (137, 146). AML cells can alter TME, for example, tumor associated fibroblasts may suppress NK cell function by secreting inhibitory factors (147, 148). One study suggested that combination of hypomethylating agents and NK cell infusion could be a promising strategy to overcome AML immune escape (137). Although there has been progress in understanding the immune escape mechanism of NK cells in AML, there are still some unknown areas.

Future research directions in AML should prioritize: i) elucidating the molecular mechanisms underlying AML-mediated NK cell dysfunction; ii) characterizing the dynamic interactions between NK cells and the leukemic microenvironment; iii) developing patient-stratified NK cell-based immunotherapeutic approaches to optimize clinical outcomes while minimizing adverse effects.

CML is a multipotential hematopoietic stem cell disease (149). The hematopoietic cells contain a reciprocal translocation between chromosomes 9 and 22 in more than 90% of patients with classic morphologic findings, which leads to an overtly foreshortened long arm of one of the pair of chromosome 22, referred to as the Philadelphia chromosome (Ph) (150). The most iconic change in CML is Ph, caused by the reciprocal translocation t(9;22) (q34;q11.2), resulting in the formation of BCR-ABL fusion gene which was a clinical diagnostic marker (151). The chronic myelogenous leukemias (CMLs) include BCR rearrangement-positive CML, atypical CML, chronic myelomonocytic leukemia, juvenile myelomonocytic leukemia, chronic neutrophilic leukemia, chronic eosinophilic leukemia, and chronic basophilic leukemia (152).

Similar to AML, LSCs are the main cause of CML occurrence and recurrence. CML is caused by activation of BCR-ABL in HSCs and converting them into LSCs defined as CD34⁺CD38^- lead to expansion of myeloid progenitors (153). For example, in a phase 2 pilot study of 46 CML patients, patients who did not achieve major molecular response (MMR) at 18 months of treatment with imatinib or dasatinib had Ph⁺ cells (> 75%) in the CD34⁺CD38^- fraction (154). Tyrosine kinase inhibitors (TKIs) induce up-regulation of N-cadherin in LSCs and adhesion to MSCs leads to activation in typical Wnt signaling, protects LSCs from apoptosis and promotes relapse (155). LSCs may also avoid eradication by modulating host immune surveillance in TME, and cytotoxic T lymphocytes (CTLs) fail to induce an appropriate immune response against CML cells through CTLs exhaustion due to the interaction of the PD-1 receptor expressed on CTLs with the inhibitory ligand PD-L1 expressed on CML cells (156). Meanwhile, the number and activity of CD4⁺ helper T cells and CD8⁺ cytotoxic T cells are decreased in CML patients, manifested by up-regulation of surface markers such as PD-1 and CTLA-4, ultimately leading to a weakened immune response to CML cells (157). The expression of Lin^-CD11b⁺CD33⁺ MDSCs was increased at diagnosis (38.6 ± 6.5%) compared with MMR (11.8 ± 2.5%, p = 0.0004) and CD4⁺CD25^highCD127^-Foxp3⁺ Tregs was higher at diagnosis (2.3 ± 0.2%) compared with MMR (1.8 ± 0.2%, p = 0.02) (158) indicating the treatment of CML may require a combination of TKI and immunotherapy.

TKIs significantly improve patient outcomes by specifically inhibiting the activity of BCR-ABL tyrosine kinase such as asciminib, imatinib, nilotinib, dasatinib, etc (159). The advancement of TKIs has substantially prolonged the survival time of most CML patients, with 5-year survival rates reaching approximately 65% to 70% (160). Patients taking imatinib showed that about 86% to 93% were still alive 5 years (161). However, long-term TKI therapy may result in drug resistance in approximately 15 ~ 17% of cases, primarily due to T315I mutations and LSCs activation, presenting an urgent clinical challenge (162). HSCT, interferon therapy (Pegylated interferon), hormone treatment, radiation therapy were also used to treat CML patients, but recurrence of CML still occur because the immune system of patient is not durably reactivated. In order to solve the above problems, CD19 (163), CD26 (164) or CD38 (165) CAR-T was used to overcome TKIs and chemotherapy resistance which achieved satisfactory results. However, it is worth noting that there was some extratumoral cytotoxicity towards activated lymphocytes (164).

As described above, for the clinical treatment of CML and the reversal of chemotherapy resistance, it is a feasible strategy to combination of chemotherapy novel immunotherapies with high efficiency and low toxicity.

NK cell-based therapeutic strategies for CML have garnered increasing attention in recent years (166). To enhance NK cell expansion and cytotoxicity, modified dendritic cell-derived exosomes activated NK cells could improve anti-CML effects via NKG2D/NKG2D-L pathway (167). K562 cells were modified by expressing a membrane-bound form of IL-15 which could induce higher expansion of CD56⁺CD3^- NK cells from PB (168, 169) ( Figure 3b ). Modification effects of IL-15 are also used in MM (170), MDS (171), colon cancer (CRC) and pancreatic cancer (PDAC) (172). Qi Li also reported that IL-21(50 ng/mL) could increase the number of CD56⁺CD3⁺ NK cells among PBMCs (173) revealing the feasibility of IL-15 or 21-secreting CAR-NK in treating CML. At present, CAR-NK based CML therapy is in the animal experimental stage ( Figure 3b ), Jusuf Imeri found that CD25 CAR-NK92 cells can effectively treat NSG mice transplanted with K562-CD25 cells and significantly increased the survival rate as compared to the untreated and NK92 WT treated cohorts (p < 0.01) (174) indicating that it is feasible to treat CML with CAR-NK.

Multiple clinical trials have been conducted exploring NK cell-based treatments for CML. For example, the clinical trial (NCT00720785) conducted in 2021 ( Table 1b ) verified the anti-CML effects by combination of Bortezomib and NK cells infusion. In another phase I/II study (NCT03348033) assessing the safety and feasibility of autologous activated and expanded NK cells in CML, infusion of NK cells significantly reduced BCR-ABL gene expression compared to untreated controls ( Table 1 ). NCT02727803 and NCT02727803, both CML clinical trials, similarly demonstrated that infusion of the allogeneic NK cell line NK92 or primary CD56⁺CD3^- NK cells, respectively, significantly reduced BCR-ABL gene expression compared to controls ( Table 1 ). These results provided a clinical reference for NK cell-based CML therapy and further confirmed its feasibility.

NK cells can be also used to reverse TKIs resistance. For reversing TKIs resistance, NK cell-derived extracellular vesicles (EVs) exhibit stronger cytotoxicity against imatinib resistant cells than parental ones by reducing the CD34⁺/CD38^- sub-populations (166). In advanced EGFR-mutated non-small-cell lung cancer, Ye Fei et al. found that combination of CD8⁺CD56⁺ NKT cells with gefitinib can overcome EGFR-TKIs resistance, revealing the universality of NK cells in reversing TKIs resistance (175). NK92-CD16 cells preferally kill TKIs-resistant cells by targeting ICAM-1, especially combination with cetuximab (an EGFR-targeted mab), the effect of reversing resistance is further enhanced (176, 177). CML-specific Dendritic cell-derived exosomes (CML-RAE-1γ-Dex) rich in RAE-1γ activate both NK cells and T lymphocytes ( Figure 3b ), and inhibit the proliferation of imatinib-resistant CML cells with T315I mutations (167).

Future directions in CML should prioritize: i) optimization of NK cell expansion and activation protocols; ii) advancement of CAR-NK development and clinical validation; iii) longitudinal assessment of NK cell therapy safety and efficacy in CML patients through rigorous clinical trials.

Most NK cells in CML were active CD56^dim cluster in PB and interacted with leukemia cells through inhibitory LGALS9-TIM3 and PVR-TIGIT interactions (178). Other studies also found the expression of TIGIT was increased on CD56^dim NK cells in CML PB while CD57 was increased on CD56^dim NK cells in CML BM (179), indicating that reversing immune suppression of PB NK cells by blocking TIGIT while improving proliferation of BM NK cells via targeting CD57 may be more effective in anti-CML ( Figure 3a ). Similar to AML, the absolute number of mature CD56^dimCD16/57^bright NK cells and circulating NK cells in PB were significantly reduced in CML patients (180). For example, the percentage of CD56⁺ NK subset in total circulating NK pool was significantly reduced in 21 CML patients (2.5% ± 0.5%) compared with normal donors (5.7% ± 0.8%) (p < 0.001) (181). Meanwhile, NK cells were dysfunctional during CML progression from chronic phase to blast crisis because BCR-ABL decreased the natural cytotoxicity of NK cells and the acquisition of KIRs (182). BCR-ABL could also interference with NK cell differentiation and exogenous addition of BCR-ABL transduced autologous CD34⁺ cells could inhibit NK cell differentiation of normal umbilical cord blood CD34⁺ and CD38^- cells (183). Another study found that CML cells could effectively inhibit the cytotoxicity of baseline and IL-2-induced NK cells to K562 cells through reducing NADPH oxidase-mediated formation of ROS (184) revealing the complexity of interrelation between CML and NK cells.

NK cell exhaustion could promote CML recrudescence ( Figures 3a, b ). Firstly, inhibition of ROS can restore NK cell numbers and enhance their cytotoxicity against CML (185). And, knockout of CXCR4 leads to NK cell depletion and TKIs resistance (186, 187). Secondly, both CD56⁺CD3^-NK cells and CD56⁺CD3⁺ NK-T cells suppress granulocyte-macrophage colony formation in BCR-ABL⁺ progenitors through JAK-2/STAT-5 pathway activation, while sparing normal CD34⁺ cells (188). Meanwhile, CML cells may release MICA into plasma, leading to NKG2D down-regulation on CD56⁺ NK cells and subsequent NK cell dysfunction (188). Thirdly, hyper-functional adaptive-like NK cells in CML MMR patients exhibited a 56-fold expansion of a normally rare subset (p < 0.01) which diminished following TKIs resistance (189). Degranulation of NK cells can be partially saved by inhibiting CIS or TNF-β to overcome NK cell suppression (190). Consistent with these findings, Amandine Decroos et al. found that high frequencies of perforin-expressing NK cells is associated with treatment-free remission (191), and these results suggest that targeting inflammatory signals can enhance NK cell-based CML immunotherapy.

Clinical evidence demonstrates that CML patients with favorable responses to imatinib exhibit elevated levels of CD3^-CD56⁺ NK cells (p = 0.0043), CD16⁺ NK cells (p = 0.0046) and CD57⁺ NK cells (p = 0.0208) (192). Moreover, NK cells from TKIs-treated CML patients show enhanced expression of NKp30/NKp46/NKp80 (193), suggesting that NK cell maturation status correlates with TKIs response. The clinical trial, NCT03239886, revealed that patients experiencing relapse after 6-month imatinib discontinuation demonstrated significantly lower NK cell proportions compared to non-relapsing patients, indicating NK cell levels as potential predictive biomarkers for molecular relapse risk ( Table 1 ). In the context of allo-HSCT, patients receiving HLA-matched but KIR3DL1-mismatched transplants showed reduced BCR-ABL transcription levels and enhanced NK cell activity (194). Notably, KIR3DL1⁺ NK cells exhibited rapid recovery (17.1% median at days 28-56, increasing to 41.7-86.0% by days 28-41), suggesting that KIR3DL1-HLA-B interactions may modulate anti-tumor immunity.

The above studies indicate that NK cells play an important role in the occurrence and recurrence of CML and can be used as an effective weapon in the treatment of CML. Next, we will further summarize the mechanism of NK cell immune escape in CML.

Immune escape represents a critical barrier to NK cell-based immunotherapy. The CML microenvironment is characterized by abundant MDSCs and Tregs ( Figure 3a ), which suppress NK cell activity and proliferation, facilitating immune evasion (195, 196). Changes in receptors on the surface of NK cells also mediate immune escape. Ya-Ching Hsieh et al. found that up-regulation of inhibitory receptors NKG2A on NK cells leads to loss of effective recognition of CML cells by NK cells (197), Meanwhile, the CML cells themselves can secrete a variety of cytokines such as IL-10 and TGF-β, which could inhibit the function and proliferation of NK cells (107). CML cells may also evade NK cell attacks by altering surface antigens, such as down-regulating the expression of NKG2DL or MHC, to inhibit the action of NK cells by secreting soluble MICA (sMICA) and reduce NK cell recognition (198). CML cells also inhibited NK cell activation signals by up-regulating TIM-3 or PD-1 that interact with receptors on the NK cell surface (178). Many CML cells exhibit a deficiency of the HLA-DR antigen, which is an important molecule for NK cells to recognize target cells (199). When HLA antigen expression decreases, the recognition ability of NK cells is limited (199). Future studies need to focus on how to restore the function of NK cells and prevent immune escape of NK cells and treatment failure.