Acute myeloid leukemia and NK cells: two warriors confront each other

ABSTRACT

Acute myeloid leukemia (AML) is a heterogeneous disease whose therapies currently show elevated toxicity and a high rate of relapse. Recently, the burgeoning of new anti-tumor therapeutic strategies aimed at enhancing the immune response has pushed natural killer cells (NKs) into the spotlight. These cells are powerful warriors that can bring about the lysis of tumor cells through their cytotoxic ability. However, tumor cells have developed strategies to evade recognition mediated by NKs. Here, we review the mechanisms triggered by AML cells and discuss the emerging immunotherapeutic strategies that potentiate the anti-tumor functions of NKs.

KEYWORDS:

• Acute myeloid leukemia
• natural killer cells
• NKG2D
• CAR-T
• immunotherapy
• immune checkpoints

1. Introduction

Acute myeloid leukemia (AML) is an aggressive malignancy of the bone marrow characterized by the uncontrolled proliferation of undifferentiated myeloid cells. It originates from a hematopoietic stem cell that acquires genetic and/or epigenetic mutations and is able to self-renew to maintain the disease.¹

Studies based on new sequencing technologies have made it possible to determine that the majority of AML cells originate from a founder clone with a driver mutation. A study of more than 1500 AML patients found that in the majority of them, the driver mutation was located in genes involved in epigenetic mechanisms (DNMT3A, ASXL1, IDH1/2, TET2).² Additionally, the founder clone can acquire subsequent mutations that generate subclones, each of which has different cellular morphologies and functionalities.³ These mutations are located in signaling pathways (FLT3, KIT, K-RAS), myeloid transcription factors (CEBPA, RUNX1) or in tumor suppressor genes (TP53, WT1, PHF6).⁴ The mutational pattern of each patient may be indicative of the prognosis of their disease. In this way, it has been reported that AML patients in complete remission (CR) but with persistent driver mutations have a higher incidence of relapse and poorer survival.⁵

AML has usually been classified by the French–American–British (FAB) group into different subtypes (M0-M7) depending on the type of cell that proliferates and its state of maturation. However, this classification has no prognostic value and the World Health Organization (WHO) established a new classification in 2008 based on cytogenetic analysis that is periodically revised.⁶

Therapeutic options for treating AML are still very limited and have changed very little in the last 30 years. The induction phase, known as “7 + 3”, based on an intensive chemotherapy regimen using standard doses of cytarabine for 7 days followed by anthracycline for 3 days. This therapy achieves CR from the disease in 60–80% of patients (age < 60 years) and in 40–60% of elderly patients (age ≥ 60 years).⁷ After CR, it is necessary to continue with an appropriate consolidation phase to eradicate the minimal residual disease and to avoid possible relapses. This treatment is based mainly on a high-dose of cytarabine followed by autologous or allogenic hematopoietic cell transplantation (HCT). Patients unsuitable for intensive therapy, especially elderly patients, receive treatments with a low-dose of cytarabine or therapy with hypomethylating agents such as decitabine (Dacogen®, DAC). Although these therapeutic strategies have improved survival rates of younger AML patients in recent years, they are still unsatisfactory in adverse-risk and most older patients.⁸ As a consequence, current research is directed towards identifying new therapeutic strategies that improve patients’survival and quality of life. In recent years, the greatest successes achieved in AML have been those that aim to enhance the immune response to increase the recognition and elimination of AML cells.

2. Impaired recognition of AML by natural killer cells

One of the first studies to demonstrate the role of natural killer (NK) cells in the immunosurveillance of cancer revealed that individuals who have NK cells with a low cytotoxicity capacity had a higher probability of developing some type of cancer during an 11-year follow-up.⁹ NK cells can recognize tumor cells independently of MHC, whereby inhibitory and activating receptors of NK cells bind to its ligands, which are expressed on the surface of tumor cells. Some of the first evidence that NK cells can kill AML cells was the absence of relapse of the disease after HCT transplantation containing alloreactive NKs.¹⁰ However, during cancer development, NK and AML cells can develop strategies to avoid recognition and allow the tumor to propagate.

2.1. Repertoire of NK cell receptors and their ligands

Conventional NK cells are key lymphocytes of the innate immune system, currently considered a prototype of the innate lymphoid cell (ILC) family. They exert anti-tumor effects mediated by their cytotoxic and cytokine-secreting capacity independently of MHC-based antigen recognition. Human NK cells comprise two main phenotypes based on the CD56 and CD16 markers. The CD56^dim CD16⁺ NK subset has great cytotoxic potential and is predominant in blood, whereas the CD56^high CD16⁻ subset produces abundant cytokines and is more abundant in lymph nodes. The functionality of NK cells is mediated and controlled by the balance between the expression of activating and inhibitory receptors.¹¹ The main activating receptors are natural cytotoxic receptors (NCR; NKp30, NKp46, NKp44), NKG2D and DNAM-1, whereas the killer-immunoglobulin-like receptor (KIR) family and NKG2A are inhibitory receptors. During NK cell maturation, the inhibitory receptors (mainly KIRs) interact with self-molecules (self-MHC class I) to avoid the recognition of self-cells in a process known as “licensing” or “education” of NK cells.¹² The unbalanced expression of these receptors and their ligands hinders the recognition of tumor cells and allows tumor progression.

In a high percentage of AML patients, the expression levels of some activating receptors (NKG2D, NKp30, NKp46, DNAM-1) are null or very low (Figure 1a).¹³ The inverse correlation between the DNAM-1 expression levels and one of its ligands (CD112) in AML patients suggests that the expression of activating receptors depends on their ligands being present on the surface of AML cells.¹⁴ Moreover, the status of AML patients also affects the repertory of NK cell receptors. Fauriat et al.¹⁵ reported that the NKp30 and NKp46 expression is reduced in AML patients at diagnosis but is partially recovered after CR and decreases again in an episode of disease relapse. It is also clearly established that high levels of expression of NKp30 and NKp46 are associated with higher rates of overall, relapse-free and progression-free survival compared with patients with reduced expression of their receptors.^16,17

Figure 1. Critical role of tumor microenvironment in NK-AML recognition. AML cells can create an immunosuppressive microenvironment that avoids recognition mediated by NK cells, triggering tumor immune escape. This tumor microenvironment results in decreased activity of NK cells by several mechanisms: a) Modulation of the NK receptor repertoire. NK functions are exhaustively regulated by the balance between activating and inhibitory receptors. However, during AML development the repertoire of NK cells is modified, reducing the level of expression of activating receptors (NKG2D, DNAM-1, NKp30) and increasing that of inhibitory receptors (KIRs and NKG2A). Additionally, ligands for NKG2D-activating receptor (NKG2DL) can be regulated by DNA methylation (red circles), reducing its level of expression on the cell surface of AML cells, or they can be released into the tumor environment in a soluble form that promotes the internalization and degradation of NKG2D; b) Defective NK maturation. NK cells of some AML patients feature defects in their maturation. AML patients show three different NK maturation profiles: hypomaturation (CD56^bright/dim KIRs⁻ CD57⁻), intermediate (CD56^dim KIR^−/+ CD57^−/+) and hypermaturation (CD56^dim KIRs⁺ CD57⁺). Elevated levels of miR-29b in AML patients reduce the cytotoxic ability of hypermature NK cells, indicating a poor level of recognition of tumor cells, and damage NK maturation, thereby increasing the number of hypomature NK cells; and c) Lysis inhibition by immune checkpoints. PD-1, TIM3 and TIGIT expressed on the cell surface of NK cells, recognize their ligands (PD-L1, Gal-9, or CD112/CD155 respectively), which are expressed on the cell surface of AML cells. As a consequence, activating pathways involved NK cell regulation (PI3K, ERK, PKCƟ) are inhibited, promoting NK cell anergy.

Expression of the ligands of the activating receptors is also modulated in AML (Figure 1a). Several research groups have shown that NKG2D ligands (NKG2DL; MICA/MICB and ULBPs1–3) are absent or expressed at very low levels on the cell surface of myeloid blasts.^18,19 Expression of these ligands depends on the cellular subtype showing little or no expression in myeloblastic cells but a high level of expression in monoblastic cells, suggesting that NKG2DL are acquired during late stages of maturation.²⁰ An important strategy developed by AML cells to avoid immune recognition is the release of these ligands in a soluble form from the cell surface through cleavage by metalloproteases or into exosomes. Two studies have shown that these ligands are frequently detected in the serum of AML patients, promoting the downregulation of NKG2D and thereby the loss of cytotoxicity capacity of NKs.^21,22 Therefore, AML patients with low levels of soluble MICA, MICB and ULBP1 are more likely to achieve CR and live more than one year.²¹ In contrast, little is known about the expression of other activating ligands. DNAM-1 ligands (CD112, CD155) are expressed in a high percentage (around 85%) of AML patients.¹⁴ However, B7H6, a ligand for NKp30, is expressed in a low number of AML blasts but is released into the serum of some patients (23.5%).¹⁶ The chronic exposure to these ligands might induce the downregulation of their receptors, leading to the evasion of NK cell immunosurveillance and facilitating tumor progression. Further studies are needed to confirm this hypothesis and to determine whether high levels of these activating receptors could improve the immune recognition of AML blasts.

Altered expression of the inhibitory receptors has also been described. CD158b (KIR2DL2/L3) and NKG2A inhibitory receptors are upregulated in AML patients with respect to healthy donors, irrespective of cellular subtype, impairing the cytotoxic ability and cytokine production (IFN-γ, TNF-α) of NK cells.^13,23

2.2. Defects in the maturation status of NK cells

NK cell maturation involves various stages of differentiation that are controlled by cytokines and transcription factors. An NK cell precursor originates from a common lymphoid progenitor, which is transformed into an immature NK cell that matures and acquires a migratory capacity and a repertoire of receptors that trigger its cytotoxic capability.²⁴ An accurate maturation process is necessary for NK cells to be able to recognize and eliminate cancer cells. However, the maturation status of those cells can be modulated during AML development. The analysis of the maturation profile of autologous NK from a murine model of AML, using specific murine maturation markers, showed that the immature phenotype (CD27⁺CD11b⁻) is upregulated, while the intermediate (CD27⁺CD11b⁺) is nearly absent, suggesting that the tumor microenvironment might modulate the maturation profile.²⁵

In humans, NK cell maturation is identified by the differential expression of the CD56, CD16, NKG2A, KIRs and CD57 markers.²⁶ CD56^bright CD16^low NK cells correspond to the most immature subset (Figure 1b). In CD56^dim NK cells, different maturation stages are characterized by the loss of NKG2A and the sequential acquisition of CD57 and KIR receptors. In AML patients, three different groups of patients were defined according to their NK maturation profile: hypomaturation (CD56^bright/dim KIRs⁻ CD57⁻), intermediate (CD56^dim KIR^−/+ CD57^−/+) and hypermaturation (CD56^dim KIRs⁺ CD57⁺) (Figure 1b).²⁷ Patients with the hypomaturation profile (around 9%) showed a poor 3-year overall survival and relapse-free survival, suggesting an altered maturation in some patients with relevant clinical outcomes. Defects in the NK cell maturation from leukemic mice and AML patients are associated with elevated levels of miR-29b in those cells.²⁵ This microRNA inhibits the expression of the T-bet and EOMES transcription factors, which are essential for controlling the final stages of NK cell differentiation. This has a dual effect: reduced T-bet expression leads to low levels of perforin in mature NK cells, decreasing their cytolytic ability, and the damage to the development of intermediate NK cells that require expression of both transcription factors.

NK cell maturation also enables the generation of memory-like NK cells. The first time that it was observed that NKs could acquire immunological memory was in a murine model deficient in T and B cells, in which prior contact with a hapten produced hypersensibility.²⁸ Subsequent studies showed that pre-activated NK cells with cytokines (IL-12, IL-15 and IL-18) induce the differentiation of memory-like NK cells, characterized by enhanced IFN-γ production and cytotoxic ability against human AML blasts.^29,30 The potent antileukemia effect was dependent on the increased expression of activating and cytokine receptors (NKG2D, NKp30, NKp44, CD25) and cytotoxic molecules, but regardless of the KIR-ligand interactions.

2.3. Expression of molecular immune checkpoint inhibitors

Immune checkpoint molecules are proteins that help maintain an adequate immune response, playing a key role in self-tolerance and the avoidance of indiscriminate lysis of self-cells. In recent years, targeting immune checkpoints with specific inhibitory antibodies has revolutionized the treatment of cancer by restoring the functionality of exhausted T cells against tumor cells. However, NK cell activity can also be regulated by immune checkpoints such as PD-1, TIM-3 or TIGIT.³¹

PD-1 is expressed in mature (CD56^dim CD16⁺) NK cells under sustained stimulation by MHC class I-deficient tumor cells or infected cells. PD-1⁺ NK cells have reduced proliferative and cytolytic abilities, and the impaired cytokine production typical of an exhausted phenotype. In fact, blockade of the PD-1/PDL-1 interaction with specific antibodies restores the functions of NK cells, promoting an anti-tumor response to several cancers, such as myeloma and digestive tumors.^32,33 Although the role of PD-1 in NK cell function in AML is poorly understood, there is clear evidence that PD-L1 is expressed in some AML patients, where it impairs immune recognition (Figure 1c).³⁴ Accordingly, several clinical trials evaluating PD-1/PD-L1 blockade are under way (Table 1). Recently, high levels of DNA methylation of PD-L1 gene have been associated with reduced PD-L1 expression in some AML cells, and a consequently lower risk of relapse and prolonged overall survival in those patients.³⁵ Methylation of PD-L1 could be an effective biomarker for stratifying patients who could benefit from PD-1 checkpoint inhibition.

Table 1. Ongoing clinical trials using NK cells to enhance the innate immune response in AML.

Therapy	NK strategy	Identifier	Phase	AML population (n)	Intervention	Estimated completion date
Allogeneic NK cells	Allogeneic donor-derived NK cells	NCT02477787	2	Refractory after allo-HSCT (90)	Biological: NK infusion, days + 13 (1–2 x 10^8 cells/kg) and + 20 (5 x 10^8 cells/kg). Drug: antihistamine.	June 2019
		NCT02229266	2	High-risk (56)	Biological: NK infusion. Drug: immunosuppression ^a, IL-2.	September 2020
		NCT02763475	2	Pediatric with cytological remission (35)	Biological: NK infusion, days 0 (5 x 10^7 cells/kg) and + 7 (5 x 10^8 cells/kg). Drugs: immunosuppression ^a, IL-2 (1 x 10^6 IU/m^2) 3 times/week, 2 weeks.	February 2022
		NCT03300492	1/2	Adult after allo-HSCT (10)	Biological: NK infusion (0.01–1 × 10^9 cells/kg) on days + 10, + 15, + 20.	December 2021
	CIML	NCT01898793	1/2	Refractory/relapsed (133)	Biological: CIML infusion (0.5 x 10^6 to maximum tolerable cells/kg). Drugs: immunosuppression ^a, IL-2 every other day for 2 weeks (total of 7 doses), ALT-803 (IL-15 agonist).	December 2020
		NCT02782546	2	High-risk after allo-HSCT (60)	Biological: CIML infusion, day + 7. Drugs: immunosuppression ^b, ALT-803 (10 mcg/kg) at 4h and on days + 12, + 17, + 22 post-transplant.	February 2022
		NCT03068819	1	Pediatric (24)	Biological: CIML infusion (0.5–10 x 10^6 cell/kg). Next day, infusion of donor lymphocyte (1–5 x 10^6 cell/kg). Drugs: chemotherapy ^c (2–4 weeks) prior infusion.	November 2024
	FATE-NK100	NCT03081780	1	Refractory/relapsed (20)	Biological: Infusion of maximum tolerable FATE-NK100 cells. Drugs: immunosuppression ^a, 6 IL-2 doses (6 x 10^6 IU/kg)	January 2021
	Allogeneic CD3^−CD19^− NK cells	NCT03152526	2	No CR after standard induction therapies (41)	Biological: CD3^−CD19^− NK infusion	June 2022
		NCT03050216	2	Refractory/relapsed (24)	Biological: CD3^−CD19^− NK infusion. Drug: immunosuppression ^a, ALT-803.	September 2020
Immune checkpoints inhibition	Blockage of PD-1/PD-L1 engagement	NCT02532231	2	In remission but high-risk for relapse (30)	Drug: Nivolumab (anti-PD-1) (3mg/kg), days + 1, + 15 of each 28-day treatment cycle (6 cycles).	October 2020
		NCT02775903	2	Older (272)	Drug: Durvalumab (anti-PD-L1) (1500 mg), day + 1 every 4 weeks, azacitidine (75 mg/m^2) for 7 days every 4 weeks.	April 2019
		NCT02771197	2	High-risk after allo-HSCT (20)	Drug: Prembrolizumab (anti-PD-1) at day + 1 every 3 weeks (8 total doses), melphalan, cyclophosphamide.	June 2021
		NCT02397720	2	Refractory/relapsed and new diagnosed (182)	Drug: Azacitidine (75mg/m^2), days 1–7 every 28 day cycle plus nivolumab (anti-PD-1) (3 mg/kg) and/or ipilimumab (1 mg/kg)	April 2020
CAR	NKG2D CAR-T	NCT02203825	1	Adult and older (12)	Biological: Single infusion of NKG2D CAR-T (0.1–3.0 x 10^Citation7cells/kg)	Completed without results
	Anti-CD33 CAR-NK92	NCT02944162	1/2	Refractory/relapsed (10)	Biological: Anti-CD33 CAR-NK92 infusion	September 2018
	Anti-CD7 CAR-pNK	NCT02742727	1/2	Refractory/relapsed (10)	Biological: Anti-CD7 CAR-pNK infusion	March 2018

CAR: chimeric antigen receptor; CIML: cytokine-induced memory-like cells; CR: complete remission; HCT: hematopoietic stem cell transplantation; ^acyclophosphamide and fluradabine;

^bcyclophosphamide, fluradabine, tacrolimus and mycophenolate mofetil; and ^c fluradabine, ara-C, and G-CSF (FLAG chemotherapy regimen).

TIM-3 is highly expressed on the surface of human NK cells and binds to its ligand Galectin-9 (Gal-9), which is expressed in AML blasts (Figure 1c). As a consequence of this interaction, TIM-3⁺ NK cells increase IFN-γ production, leading to increased expression of the indoleamine 2,3-dioxygenase 1 enzyme by AML cells, which, in turn, impairs NK cell degranulation activity and ensures leukemic immune escape.³⁶ Moreover, high levels of soluble Gal-9 were detected in the serum of AML patients, where Gal-9 is able to bind to TIM-3 expressed on the surface of leukemic stem cells that activates the NF-kB and β-catenin pathways involved in the self-renewal of these malignant stem cells.³⁷ Targeting this stimulatory loop could be an efficient target to avoid the development of human AML.

TIGIT is expressed in cytotoxic T cells and in NK cells, and binds to the same ligands as DNAM-1 (CD112 and CD155) (Figure 1c). In AML patients, a large number of dysfunctional CD8⁺ T cells with increased expression of TIGIT and PD-1 but low levels of DNAM-1 was reported, but TIGIT expression in the NK cells of these patients has not been characterized.³⁸ As previously described, AML patients show low levels of DNAM-1 expression, while its ligands are highly expressed.¹⁴ This could favor the binding of TIGIT with CD112 and CD155 ligands, promoting inhibitory signaling and, thereby, tumor immune escape. Recently, it has been shown that blockade of TIGIT prevents NK cell exhaustion and promotes NK cell-mediated anti-tumor effects in several mouse models of cancer (colon, breast, melanoma and fibrosarcoma), but the therapeutic potential in AML patients needs further study.³⁹

Given these findings, the development of new immunotherapies aimed at restoring the functionality of NK cells and the recognition of AML cells is a promising strategy, and one with important clinical implications.

3. Targeting AML cells by NK cell-based immunotherapy

The development of immunotherapy, an approach that aims to fight cancer using the body’s own immune system, has revolutionized the treatment of cancer. Numerous lines of evidence, summarized in this review, support the use of NK cell-based immunotherapy as a strategy for better recognition of AML cells. We also review the promising strategies that are currently proposed in the context of AML.

3.1. Adoptive transfer of NK cells

3.1.1. Autologous NK cells

Infusing NK cells to treat AML was first proposed by the American National Cancer Institute more than thirty years ago.⁴⁰ This therapy is based on the administration of autologous NKs (1–2 x 10⁷ cells/kg) previously activated with IL-2 or IFN-γ. Subsequently, IL-2 is administered in vivo to allow NK cells to proliferate in patients (Figure 2a). The advantage of this therapy over allo-HCT is that immune suppression and HLA-matching are not necessary. Nevertheless, it also has limitations due to the high toxicity arising from the sustained in vivo administration of IL-2, and to the absence of any clear clinical benefit.⁴¹ The lack of improvement of immune recognition may be due to the interaction between KIR receptors expressed in autologous NKs and HLA-I molecules in AML cells.⁴² This implies that modulation of the KIR/HLA-1 axis could enhance the clinical effect of autologous NKs transplant in AML.

Figure 2. Enhancing AML recognition by immunotherapy techniques. Several strategies based on the use of NK cells have been proposed to allow the recognition of AML cells, such as: a) NK infusion. Autologous NK cells or NK cells from a KIR-ligand mismatched donor, are expanded in vitro in the presence of IL-2, IFN-γ, and/or anti-CD3. AML patients are infused with these cells and treated with IL-2 to promote the expansion of NK cells. An alternative is the infusion of allogeneic CIML cells that are expanded in vitro in the presence of a cytokine cocktail (IL-12, IL-15 and IL-18). An advantage of this treatment is that there is no need to treat the patient with IL-2; b) Epigenetic treatments. Treatment with HDACi and DNMTi restores NKG2DL (MICA and ULBPs 1–3) expression on the cell surface of AML cells. The TIMP3 gene, which is methylated in some AML patients, is also expressed, leading to the inhibition of the main protease involved in the release of NKG2DL, ADAM17. As a consequence, NKG2DL (MICA/B and ULBP2) are not shed from the cell surface and are released in their soluble form (sMICA/B and sULBP2), maintaining the high expression levels on the AML cell surface. c) Immune checkpoint blockade. Specific antibodies against PD-1 (nivolumab, pembrolizumab) or its ligand PD-L1 (durvalumab) block the PD-1/PD-L1 interaction, avoiding the anergy of NK cells; and d) CAR technology. T cells or NK cells collected from the AML patient are transduced with CAR with specific genes (NKG2D, NKp30) or antibodies (α-CD33, α-CD7). Further, these cells are infused in AML patients and when CAR recognizes its antigen, expressed on the surface of AML cell, CAR-T or CAR-NKs are activated.

3.1.2. Allogeneic NK cells

Further studies were performed using allogeneic NKs from healthy donors that maintain their function and can be safely administered. As previously described, during NK cell development and to guarantee ‘self-tolerance’, KIR receptors bind with their ligands to ‘license’ the NKs and avoid the recognition and lysis of self-cells.⁴³ Studies were carried out in AML patients using alloreactive NKs pre-activated with IL-2 and with one or more KIR-ligands mismatched to avoid the recognition of self-HLA class I molecules (Figure 2a).^44,45 All patients receive immunosuppressive chemotherapy before NK cell infusion, and further exogenous administration of IL-2 in order to activate and expand circulating donor NK cells. One benefit of this therapy is the low incidence of graft versus host disease (GvHD) and the production of a strong graft versus leukemia (GvL) that is associated with better survival and a lower probability of relapse. In elderly patients, whose therapeutic options are very limited, consolidation therapy with these cells is feasible and promotes a better disease-free survival rate.⁴⁶ Moreover, allogeneic clones can persist for up to 12 months, allowing the elimination of residual blasts.⁴⁷ Several clinical trials are currently under way using haploidentical NKs as consolidation therapy in AML, or in combination with other therapies, to determine their anti-leukemic effect and any possible secondary effects (Table 1).

Ex vivo stimulation of NKs with a cytokine cocktail of IL-12, IL-15 and IL-18 is an established alternative to avoid the in vivo administration of IL-2 after NK cell infusion.⁴⁸ These cells, known as CIML (cytokine-induced memory-like), proliferate and produce high levels of IFN-γ during the first week, after which its production decreases. However, the in vitro re-stimulation of CIML cells, with the same activation cytokine cocktail or with a target cell (cell line or primary AML cells), restores the production of IFN-γ, suggesting that these cells have memory properties.⁴⁹ Moreover, these cells more strongly express activating receptors (NKG2D, DNAM-1, NKp30, NKp44, NKp46) relative to unstimulated NKs, implying a greater cytotoxicity to tumor cells and, thereby, a higher frequency of CR in elderly relapsed/refractory AML patients.³⁰ In fact, there have been several phase I and II trials based on relapsed/refractory AML patients using CIML cells, alone (0.5–10 x 10⁶ cells/kg) or in combination with other treatments (Table 1).

Alternative ex vivo expansion methods for NK cells are currently being evaluated in several clinical trials for the treatment of AML patients (Table 1). One of them is the infusion of allogeneic CD3⁻CD19⁻ NKs, a method that increases the proliferation and cytolytic capacity of infused NK cells in patients. Another is the use of an IL-15 antagonist (ALT-803) that increases the activation and cytotoxic capacity of NK cells with respect to IL-2. In addition, one of the most important advances is the enrichment and expansion of terminally differentiated adaptive NK cells (commercially known as FATE-NK100) and that are currently being evaluated in refractory or relapsed AML patients (Table 1). These cells are obtained from some specific NK cell subsets (CD57⁺ NKG2C⁺) that are expanded following CMV infection, creating a pool of adaptive NK cells with a lower threshold of activation.⁵⁰ To generate these cells, peripheral blood mononuclear cells from CMV-seropositive donors were depleted of CD3⁺ T and CD19⁺ B cells and cultured with IL-15 (a cytokine that drives NK cell proliferation) and a small molecule inhibitor of glycogen synthase kinase 3-beta (GSK3β) for 7–9 days.⁵¹ This facilitates the generation of a large number of mature CD57⁺ NK cells with increased IFN-γ production. In addition to AML, two clinical trials are currently ongoing, one (NTC03319459) in subjects with advanced solid tumors (gastric, colorectal, and head and neck squamous cell carcinoma), and the other (NTC03213964) in women with advanced ovarian, fallopian tube or primary peritoneal cancer.

3.2. Regulation of NK cell ligands by epigenetic mechanisms

To enhance the cytotoxic ability of NK cells it is essential to recognize the engagement between activating receptors and their ligands expressed on the cell surface of target cells. However, as we have remarked above, the expression of these ligands is reduced or absent in AML cells, so their expression must be enhanced in order to facilitate immune recognition mediated by NKs.

Epigenetic modifications that modulate the expression of activating ligands are known to arise during cancer development. Several studies have reported that treatment with inhibitors of histone deacetylases (HDACi; valproic acid, trichostatin A or all-trans-retinoic acid) increases the expression of NKG2DL (MICA/B, ULBPs1-3) and decreases the release of their soluble forms in cell lines and primary AML cells, thereby enhancing the lytic capacity of NKs (Figure 2b).^20,52,53 Histone acetylation is also important for regulating other activating ligands in AML, such as CD48 (2B4 ligand).⁵⁴

DNA methylation is another important epigenetic mechanism of gene regulation. Results from our group revealed that DNA methylation can contribute to the absence of NKG2DL expression observed during tumor development in AML patients.⁵⁵ We found significantly higher DNA methylation levels in the promoter region of MICA, ULBP1 and ULBP2 genes in AML patients compared with healthy donors, and this was correlated with the absence of transcription for these genes. Moreover, 59% of AML patients had at least one methylated NKG2DL, although the combination of ligands varied between patients. DNA methylation is a reversible mechanism, so treatment with DNA methyltransferase inhibitors (DNMTi) such as azacitidine (Vidaza®, AZA) and DAC, which are commonly used to treat elderly AML patients, restores the expression of NKG2DL on the surface of AML cells (Figure 2b). Our group recently reported that treatment with these drugs also reduces the release of soluble NKG2DL, maintaining their expression on the surface of AML cells by the hypomethylation of TIMP3, an inhibitor of protease ADAM17 (Figure 2b).⁵⁶ However, the role of DNA methylation in regulating other activating ligands is unknown.

All these studies suggest that the use of epigenetic treatments in combination with other conventional therapies could be an important method for enhancing the expression of activating ligands in AML blasts and thus the recognition mediated by allogeneic NKs.

3.3. Blockade of immune checkpoints

Given the role of immune checkpoints in inhibiting NK cell activation, the effect of blocking these molecules has been evaluated. Blockade of some immune checkpoints (LAG-3, TIM-3, PD-1) with specific antibodies increases the lytic capacity of CIML cells against AML (Figure 2c).⁵⁷ In an experimental mouse model of lung cancer, the specific blockade of LAG-3 and PD-1 increases the number of mature NKs and thereby reactivates the antitumoral properties (greater production of cytotoxic granules and IFNγ) of CIML cells, allowing the control of the metastasis.⁵⁸

Within immune checkpoints, PD-1 and its ligand (PD-L1) are the most thoroughly studied due to the excellent results obtained from T cells in several tumors. It has also been reported that PD-1 or PD-L1 blockade increases the degranulation and cytotoxic capacity of NKs, shrinking the tumor grown in xenograft models of digestive and myeloma cancers.^33,59 In AML, several clinical trials are currently underway to evaluate the response of this blockade (Table 1). To this end, patients with relapse/refractory or high-risk AML are treated with specific antibodies against PD-1 or PD-L1 alone or in combination with other therapies (AZA, chemotherapy, dendritic cell vaccine). Unpublished preliminary data from the NCT02397720 trial show that the combined use of AZA with anti-PD-1 (nivolumab) is well tolerated and produces durable responses in relapsed AML. Therefore, the results of current clinical trials are needed to corroborate the efficiency of this therapy, as has been observed in solid tumors.

Recent significant studies have shown TIGIT and IL-1R8 to be new checkpoint molecules or negative regulators of NK cell function in several solid tumors.^39,60 Blockade of these proteins reverses the exhaustion of tumor-infiltrating NK cells and reduces the threshold for NK cell activation, respectively. The study of these mechanisms in AML, in combination or not with other therapeutic approaches, could help enhance the NK-cell mediated anti-tumor response.

3.4. Chimeric antigen receptor-based immunotherapy

Chimeric antigen receptor (CAR) therapy is an emerging immunotherapy that combines specific antigen recognition with signaling capacity to recognize specific targets in cancer cells.⁶¹ These CARs can be expressed in immune cells as T lymphocytes (CAR-T therapy) by viral transduction or mRNA transfection, and frequently its specificity is based on scFv regions or TCR binding domains. Several CAR-T therapies directed against some of the most important targets in AML blasts (such as CD33, FLT3 and CD123) have been assayed in recent years, revealing potent activity against tumor cells.^62–64 An ideal CAR target should be expressed in a large number of AML blasts and with high-density but in as few normal tissues as possible to avoid undue toxicity. These targets are widely expressed in the majority of AML blasts but, unfortunately, also in some normal hematopoietic stem cells, myeloid cells and other tissues. CAR-T cell immunotherapy aimed at simultaneously targeting two AML antigens (CD33 and CD123) or other candidates identified by integrating proteomic and genomic datasets is a promising strategy for effectively eliminating AML cells and leukemic stem cells, while avoiding relapse and without incurring greater toxicity.^65,66

CAR technology may also be useful for enhancing the recognition and lysis mediated by NK cells (Figure 2d). Two strategies have been proposed for this: CAR-T encoding NK cell-activating receptors, and CAR-NK. In the first strategy, the use of a CAR-T with a full-length human NKG2D (NKG2D CAR-T) to target cancer cells expressing NKG2D ligands has been reported.⁶⁷ In murine models of several tumors (glioblastoma, osteosarcoma) this NKG2D CAR-T can infiltrate the tumor, downregulating its activity and increasing survival.^68,69 This therapy protects against future relapses because surviving animals can respond to a new infiltration of tumor cells. However, one disadvantage of this CAR-T model is its high toxicity as a consequence of the increased production of cytokines/chemokines (IFN-γ, IL-6, MCP-1) by activated T cells.⁷⁰ This could be avoided by employing the alternative strategy of using memory T cells, which are known to be less toxic after CAR-T infusion in murine models of glioblastoma and osteosarcoma.^68,69 Although the antitumoral effect of NKG2D CAR-T is unknown in AML models, the Dana-Farber Cancer Institute is conducting a phase I clinical trial (NCT02203825) in which patients with AML, myelodysplastic syndrome and multiple myeloma are infused with a single dose (0.1-3 x 10⁷ cell/kg) of NKG2D CAR-T (Table 1). NKp30 is another activating receptor whose use in CAR-T models has been proposed.⁷¹ This NKp30 CAR-T increases the production of IFN-γ and the cytotoxic capacity towards malignant cells that express its ligand, B7H6. NKp30 CAR-T improves survival in lymphoma murine models and prevents the development of the tumor after a subsequent challenge of malignant cells, suggesting that an immunological memory is generated. CAR-T technology can also be used to target the ligands of the activating receptors and to enhance the lysis of tumor cells. In this way, a B7H6 CAR-T therapy enhances the release of IFN-γ by transfected T cells against B7H6-expressing cell lines, and prolongs survival in ovarian and lymphoma murine models.^72,73

The second strategy for enhancing the recognition between NK and tumor cells is the use of CAR-modified NK cells (Figure 2d). This therapy is attracting attention because of its advantages over T lymphocytes: (1) sensitization is not necessary; (2) there is no MHC restriction; (3) antigen-specific receptors are not required; and (4) there are fewer side effects than with GvHD.⁷⁴ There are currently two phase I and II clinical trials underway using CD7- or CD33-CAR NK-92 cells in AML patients with relapsed or refractory disease (Table 1). The first findings, which have recently been published, show that high doses of CD33-CAR NK cells have no significant adverse effects on the three analyzed AML patients, showing the safety of the treatment, although further studies are needed to establish the clinical efficacy.⁷⁵

4. Concluding remarks

NK cells are key effectors in cancer immunosurveillance that promote the recognition and lysis of tumor cells. However, the evolution and frequent relapses in AML patients suggest that leukemia cells can escape this recognition. As mentioned above, several mechanisms that impair NK cell functions in AML have been described in recent years that have prompted great efforts aimed at developing new therapeutic strategies that enhance the recognition and lysis of AML cells by NK cells. Currently available immunotherapeutic treatments have yielded promising results in vitro and in mouse models of AML. Nevertheless, it is important to perform in-depth studies to determine the toxicity of these immunotherapies in AML patients. The ability to inhibit AML targets specifically using immune checkpoint blockade or CAR-T/CAR-NK therapies opens up new possibilities for NK cell-based immunotherapy. A considerable number of clinical trials are currently underway using both therapies, alone or in combination with well-established AML treatments. The results of these studies will clarify the clinical value of both immunotherapies as treatments for AML.

Abbreviations

AML	=	acute myeloid leukemia
CAR	=	chimeric antigen receptor
CIML	=	cytokine-induced memory-like
CR	=	complete remission
FAB	=	French-American-British classification
GvHD	=	graft versus host disease
GvL	=	graft versus leukemia
HSCT	=	hematopoietic stem cell transplant
NK	=	natural killer
WHO	=	World Health Organization.

Conflicts of interest

The authors declare no conflicts of interest.

Additional information

Funding

This work was supported by grants from Plan Nacional de I+D+I 2013–2016 ISCIII (Institute of Health Carlos III; PI16/01318 and PI17/01244), FEDER Funding Program from the European Union; and Red Española de Investigación Renal (REDinREN, grant number RD16/0009/0020).