Abstract
Background: Acute myeloid leukemia (AML) is an aggressive hematologic malignancy with a high relapse rate and still limited therapeutic options. Natural killer (NK) cell-based immunotherapy has the potential to improve outcomes for patients with AML. Summary: Recent preclinical studies and early-stage clinical trials aim to enhance the intrinsic anti-leukemic properties of NK cells by selectively targeting AML cells with chimeric antigen receptors (CARs). Furthermore, NK and CAR-NK cells can be combined with other therapeutic modalities or engineered further to overcome the immunosuppressive microenvironment, and treatment resistance of AML blasts and leukemia-initiating cells (LIC). Key Messages: In this review, we summarize preclinical studies with cytokine-stimulated or genetically engineered NK cells derived from different cell sources for the treatment of AML and their translation into early-phase clinical trials. We also provide an overview of promising recent developments toward innovative NK cell-based therapies that may be implemented in the near future.
Introduction
Acute myeloid leukemia (AML) is one of the most common and aggressive types of leukemia in adults and is marked by a poor prognosis [1]. AML is characterized by the expansion of immature cells of the myeloid lineage and substantial heterogeneity and intrinsic variability of the leukemic cells, which contribute to elevated relapse rates and treatment resistance [2]. Currently, only allogeneic hematopoietic stem cell transplantation (HSCT) is able to provide a cure for AML [3]. However, relapse post-allogeneic HSCT is still observed, and the majority of elderly patients are ineligible for HSCT. Consequently, there is an urgent need for alternative therapeutic strategies for patients unsuitable for intensive treatment regimens and patients with relapsed or refractory (r/r) disease [4].
Immunotherapy has emerged as a promising approach for the treatment of different cancer types, aiming to empower the body’s immune system to specifically recognize and attack tumor cells [5]. Natural killer (NK) cells make up 10–15% of peripheral blood lymphocytes. As innate cytotoxic cells, they are capable of killing tumor cells without prior sensitization, with their activity regulated through a diverse array of germline-encoded activating and inhibitory receptors [6, 7]. Both patient-derived autologous NK cells and allogeneic NK cells from healthy donors have been safely used for cancer therapy. In the case of allogeneic NK cells, human leukocyte antigen (HLA) mismatch between donor and recipient does not induce graft-versus-host disease (GvHD) or cytotoxicity-related side effects, which are frequently encountered in T-cell-based therapy [7, 8]. In fact, a lack of recognition of HLA by inhibitory killer cell immunoglobulin-like receptors (KIRs) or the inhibitory C-type lectin receptors NKG2A/B facilitates the lysis of malignant cells that downregulate HLA in an autologous setting (missing-self hypothesis) and contributes to the graft-versus-leukemia effect of allogeneic NK cells [9].
To enhance their cytotoxic potential and increase selectivity for malignant cells, NK cells have been equipped with chimeric antigen receptors (CARs) targeting tumor-associated or lineage-specific surface antigens on cancer cells [10]. Despite the progress made in recent years, the application of CAR-NK cell-based immunotherapies targeting AML is still at an early stage, and the number of ongoing clinical trials investigating the efficacy and safety of these therapies is limited when compared to respective CAR-T cell studies.
Several FDA/EMA-approved CAR-T cell products are already in clinical practice for the treatment of B-cell malignancies, and significant efforts are being made to extend this success to other hematological malignancies including AML [11]. Nevertheless, CAR-T-cell therapy is generally associated with safety concerns such as the possible development of cytokine release syndrome (CRS) or immune effector cell-associated neurotoxicity syndrome (ICANS), which together with the current restriction to patient-derived autologous cells still limits their broad applicability [12]. CAR-NK cells show a more favorable safety profile and their intrinsic natural cytotoxicity is expected to lower the risk of tumor escape due to the treatment-induced selection of antigen-loss variants [9]. Hence, NK cells constitute a valid and promising alternative to T cells for the development of CAR-engineered effector cells (see Table 1 for a comparison of T and NK cell characteristics).
. | T cells . | NK cells . |
---|---|---|
Target recognition and cytotoxicity induction | - TCR mediated | - Balance of activating/inhibitory signals |
- MHC-restricted | - Non-MHC restricted | |
- Missing or altered self-recognition | ||
- Natural cytotoxicity receptors, NKG2D, DNAM-1, CD16 | ||
Main killing mechanism | - Perforin/granzymes | - Perforin/granzymes |
- Fas/TRAIL | - Fas/TRAIL | |
Available source for clinical use | - Autologous | - Autologous |
- Allogeneic (Risk of side effects) | - Allogeneic | |
- Off-the-shelf (donor-derived; cell lines; iPSC-derived) | ||
Side effects of CAR therapy | - CRS | - Low grade CRS |
- ICANS | - Low grade ICANS |
. | T cells . | NK cells . |
---|---|---|
Target recognition and cytotoxicity induction | - TCR mediated | - Balance of activating/inhibitory signals |
- MHC-restricted | - Non-MHC restricted | |
- Missing or altered self-recognition | ||
- Natural cytotoxicity receptors, NKG2D, DNAM-1, CD16 | ||
Main killing mechanism | - Perforin/granzymes | - Perforin/granzymes |
- Fas/TRAIL | - Fas/TRAIL | |
Available source for clinical use | - Autologous | - Autologous |
- Allogeneic (Risk of side effects) | - Allogeneic | |
- Off-the-shelf (donor-derived; cell lines; iPSC-derived) | ||
Side effects of CAR therapy | - CRS | - Low grade CRS |
- ICANS | - Low grade ICANS |
Many preclinical studies explore new strategies to combat AML with refined NK cell-based products that may be translated to clinical trials in the near future [14]. Among other aspects, NK cell therapies face the challenge of finding the optimal NK cell source to achieve effective cytotoxicity and persistence in the patient. In addition to peripheral blood from healthy donors, umbilical cord blood, hematopoietic stem and progenitor cells and induced pluripotent stem cells (iPSC) generated from somatic cells have been used to derive NK and CAR-NK cell products [15, 16].
Cord blood and HPSC have the advantage of being readily available from cord blood banks and registered bone marrow donors [16]. Indeed, the most promising CAR-NK cell trial so far has been performed with umbilical cord blood (UCB)-derived feeder-cell expanded CD19-CAR-NK cells employed for the treatment of B-cell malignancies [17, 18]. Similarly, CAR-NK cells targeting CD70 have successfully undergone preclinical evaluation, and a respective clinical trial has been launched (NCT05092451) (K. Rezvani, personal communication). While not yet engineered with a CAR, UCB-derived CD34+ HPSC NK cell products (oNKord®) have been evaluated in patients with AML, showing a favorable safety profile with disease-free survival until the end of the observation period in 4 of 10 patients [19]. Additional studies are currently recruiting patients to investigate UBC-derived NK cells for the treatment of AML (NCT04632316, NCT05333705, NCT05933070). Regarding iPSC-derived NK cells, a first-in-human phase I study recently started recruiting to evaluate the safety and efficacy of an iPSC-based CAR-NK cell product that targets CLL-1 in AML patients (NCT06027853).
In addition, in some approaches, the NK cell line NK-92 is employed as an alternative source of allogeneic NK cells. These continuously expanding cells are characterized by a high intrinsic cytotoxic potential due to the expression of most of the activating NK-cell receptors but a lack of the majority of inhibitory KIRs [20]. Research with NK-92 cells has contributed to the refinement of cellular therapies that target AML, with several preclinical studies aiming to improve respective CAR-engineered derivatives. Notably, the first clinical application of CAR-engineered NK-92 cells was based on a cell product targeting CD33-positive AML that demonstrated the safety and feasibility of infusion of high doses of the irradiated effector cells (NCT02944162) [21].
Focusing on promising preclinical work and early-phase clinical trials, in this review, we provide an overview of the current state of advanced NK cell-based therapies against AML, covering unmodified and CAR-engineered NK cells derived from various sources, their activation using cytokine stimulation or different killer cells engagers, and respective combination approaches.
NK Cells for the Treatment of AML
From Autologous to Allogeneic NK Cell Therapy
NK cell-based therapies for AML have already been tested in clinical trials for 20 years. The emergence of refined treatment strategies and the development of CAR-engineered NK cell products have allowed to now introduce more sophisticated therapeutic approaches into the clinic. As of February 2024, there are 61 ongoing clinical trials listed that utilize NK cells (NK-92, primary and iPSC-derived) for the treatment of AML (see Fig. 1, www.ClinicalTrials.gov).
In early clinical studies, ex vivo expanded autologous NK cells were administered to cancer patients, expecting antitumor activity without the risk of GvHD and no need for immunosuppression. However, only a limited therapeutic effect was observed, possibly due to recognition of self-MHC antigens suppressing NK cell cytotoxicity via their inhibitory receptor repertoire [22, 23]. Furthermore, NK cells from cancer patients can display altered gene expression profiles associated with a dysfunctional phenotype, resulting in reduced cytotoxic potential and impaired ex vivo expansion capacity. Additionally, heavily pretreated patients often exhibit NK cell lymphopenia, complicating the isolation of NK cells in clinically relevant numbers [16]. This has led to the investigation of adoptive cell therapies with NK cell preparations from allogeneic donors.
In 2002, it was reported that KIR ligand mismatch in the context of haploidentical HSCT plays a crucial role in the NK cell-mediated graft-versus-leukemia effect, as well as in the prevention of graft rejection and GvHD [24]. Three years later, it was reported that haploidentical, related-donor NK cells (2 × 107 cells/kg) were able to expand in vivo after intensive lymphodepleting chemotherapy, and induced remission in 5 out of 19 AML patients. However, the observed anti-leukemic effect required the administration of interleukin (IL)-2, which caused severe side effects [22]. Later studies confirmed the anti-leukemic activity of NK cells in AML patients, and correlated clinical outcomes with the number of transferred alloreactive NK cells, without reporting severe IL-2-related toxicities [25, 26]. Thereby, no GvHD or other toxicities such as CRS and ICANS were observed (for reviews see [16, 23]).
Several studies were conducted to determine the most effective dosage of NK cells with respect to frequency and quantity of cell administration. Up to four injections of 5 × 106 to 2 × 108 NK cells/kg body weight were tested [27‒29]. GvHD higher than grade II was only observed when the remaining T cells contaminating the NK cell products exceeded 0.5 × 105 cells/kg body weight [30]. Although improved progression-free and overall survival were reported in some cases, the question of optimal dosage and time of administration remains.
In a recent phase I/II clinical study for the treatment of myeloid malignancies, high doses of up to 1 × 108 cells/kg of ex vivo stimulated NK cells of one healthy donor were administered 3 times to patients before or early after haploidentical HSCT. NK cells were expanded with the addition of replication-deficient K562 feeder cells that expressed membrane-bound IL-21 and 4-1BBL (FC21), resulting in a hyperfunctional NK cell phenotype. Treatment led to a low 2-year relapse rate of only 4%, which was most likely caused by the persistent and highly cytotoxic NK cells. Patients did not receive IL-2, eliminating the risk of IL-2-related toxicities. However, 1 male patient developed severe GvHD [31].
Different sources of NK cells and stimulation protocols can be used for the generation of clinical NK cell products [31], which in part were also employed in the studies mentioned above. The following sections will give an overview on the sources available for NK cell isolation and current NK cell manufacturing procedures.
Cytokines for the Improvement of NK Cell Therapy
Ex vivo NK cell expansion is limited, and it can be a hurdle to acquire clinically relevant cell numbers. Therefore, cytokines are utilized in ex vivo NK cell expansion, either individually or in combination. These factors influence NK cell development, expansion, and cytotoxicity, and can help overcome the suppressive tumor microenvironment (TME) in the recipient. Generally, IL-2, IL-12, IL-15, IL-18, or IL-21 as well as type I interferons (IFNs) are employed for ex vivo cultivation and expansion or can be administered in vivo to influence NK cell proliferation, activation, and persistence in the patients [32].
IL-2 and IL-15 are the most frequently used cytokines for NK cell stimulation. Their receptors have partly overlapping functions: Both promote the proliferation of NK cells and T cells. They share the IL-2/IL-15R β and the common γ chain receptor subunits, and signal via the JAK/STAT pathway. IL-2 additionally contributes to the elimination of self-reactive T cells and the stimulation of Treg cells. IL-15 maintains the CD8+ memory T-cell response [33] and plays a major role in cancer cell clearance by promoting antitumor properties of NK cells by upregulation of natural cytotoxic receptors (NCRs), NKG2D and DNAM-1. IL-21 is mainly produced by T helper cells. It influences the maturation of NK cells from the bone marrow and promotes NK cell proliferation and cytotoxicity [32].
The crucial role of IL-15 in NK cell homeostasis has been demonstrated in several genetic mouse models displaying massively reduced numbers and impaired functions of NK cells in the absence of IL-15. In mice, it was shown that IL-15 needs to bind to the IL-15 receptor α-chain (IL-15Rα) on neighboring cells to induce NK cell differentiation, proliferation, and survival via trans-presentation [34]. Based on this, IL-15 superagonists like N-803 (Anktiva™; previously termed ALT-803) have been developed. As an Fc-fusion complex consisting of an IL-15 high-affinity mutant and IL-15Rα, it mimics the trans-presentation of IL-15 by accessory cells, promoting NK cell stimulation [35]. A first-in-human study investigated the effect of N-803 in AML patients who had relapsed after allogeneic HSCT. The patients received 4 doses of up to 10 μg/kg N-803 by intravenous or subcutaneous application. The response rate was 19%, and 1 patient remained in CR for 7 months. It was found that the administration of N-803 led to the activation and expansion of CD8+ memory T cells and NK cells, particularly the CD56bright subset, which was accompanied by an increase in granzyme B production. Additionally, N-803 treatment induced elevated expression of activating NKp30 and NKG2D receptors on NK cells. No dose-limiting toxicities, severe forms of GvHD, CRS or capillary leak syndrome were observed [36]. In another study with systemic infusions of N-803 accelerated rejection of allogeneic NK cells by activated endogenous T cells was found, questioning its suitability in the context of adoptive NK cell therapies [37]. Nevertheless, N-803 could be efficacious for other applications, and further studies are planned, potentially in combination with checkpoint inhibitors or immune cell engagers [36].
Even though IL-15 is the most prominent cytokine for NK cell activation, the first NK cell-based therapy protocols primarily used IL-2 to stimulate NK cells ex vivo or in vivo [29, 36, 37]. Ex vivo IL-2 stimulation leads to a CD56bright CD16bright/dim NK cell phenotype with high surface expression of all NCRs and NKG2D, loss of homing receptor CD62L and increased cytokine secretion, as well as elevated expression of other activating receptors [30, 38, 39]. As IL-2 not only stimulates NK cells but also regulatory T cells, which in turn inhibit NK cell reactivity by producing TGF-β, the combination of a TGF-β antagonist with IL-2 and allogeneic NK cell therapy is being clinically investigated (NCT05400122) [40].
Especially after CD3/CD19 depletion, in vitro NK cell expansion with low-dose IL-15 stimulation induced higher NK cell cytotoxicity and expansion rates than IL-2/IL-15 combination protocols. An IL-21 booster shortly before harvest further increased expansion and cytotoxicity by 8–10% in preclinical evaluations [41, 42]. Furthermore, in a clinical trial, IL-15 treatment of AML patients following adoptive transfer of haploidentical NK cell infusions showed higher in vivo expansion and remission rates compared to historical IL-2 controls. Importantly, IL-15-induced side effects such as CRS and neurotoxicity were increased upon subcutaneous instead of intravenous application and were associated with elevated IL-6 levels [43]. The synergistic use of different NK cell stimulating cytokines is evaluated in a recent phase II clinical trial, treating patients with high-risk myeloid malignancies with ex vivo IL-15 and IL-21 stimulated NK cells after HSCT. Compared to patients who have only received HSCT, infusion of NK cells reduced disease progression, induced higher blood counts of NK and T cells, and increased memory-like NK cells [44].
In summary, stimulation with different cytokines initially leads to activation of NK cells, a boost in expansion rates and an increase in cytotoxicity through enhanced expression of activating receptors on the cell surface. However, it must be considered that continuous stimulation may also induce NK cell exhaustion, leading to a decrease in viability, function, persistence, and tumor control [45, 46]. Therefore, it is crucial to find the optimal dosing, administration route and combination of cytokines to enhance the anti-leukemic activity of NK cells against AML, without inducing NK cell exhaustion or adverse effects.
Cytokine-Induced Memory-Like NK Cells
Refinement of NK cell therapy is not only achieved by employing distinct cytokines but also critically depends on the phenotypic characteristics of the NK cell product itself. To increase NK cell in vivo persistence and ensure long-term clearance of leukemia-initiating cells (LICs), cytokine-induced memory-like (CIML)-NK cells are of particular interest [47], and will be discussed in this paragraph.
In recent years, it was found that not only cells of the adaptive immune system remember a previous activation but also NK cells can develop a so-called memory state when preactivated by viruses, haptens, or specific cytokine combinations. In this context, IFN-y has been reported as an immunostimulating cytokine relevant for the antiviral and antitumoral NK cell function [48]. This memory phenotype can be maintained over several cell divisions, even months after initial activation [49].
Data from a first phase I clinical trial suggest that CIML-NK cells might be a promising approach to further improve allogeneic cell therapy. Nine patients with active r/r AML that were preconditioned with fludarabine and cyclophosphamide, received HLA-haploidentical, allogeneic CIML-NK cells generated by in vitro cytokine stimulation (IL-12, IL-15, IL-18). The CIML-NK cells were administered at three dose levels ranging from 0.5 to 10 × 106 cells/kg, supported with low dose IL-2 following their application. Increased NK cell expansion, proliferation, and IFN-γ production were demonstrated and led to a measurable anti-leukemic effect. CIML-NK cells were proliferating for 3 weeks, and patients showed a complete remission (CR), CR with an incomplete count recovery rate of 45%, and an overall response rate of 55% [50, 51].
In a subsequent phase I clinical trial, 8 pediatric or young adult patients with post-HSCT-relapsed AML were treated with combination therapy of donor lymphocyte infusion and CIML-NK cells after salvage chemotherapy, without administration of exogenous cytokines. CIML-NK cells were derived from the original HSCT donor and expanded after infusion. Additionally, their memory-like phenotypes as well as anti-leukemic functions were maintained for more than 3 months. 28 days after CIML-cell infusion, the overall response rate was 75%, and 50% of the patients showed CR. One patient of the heavily pretreated cohort remained in CR for more than 2 years without additional therapies [52].
In both clinical trials, the administration of CIML-NK preparations was assessed as safe, with no occurrence of dose-limiting toxicities, CRS, or ICANS, and no increased incidence of GvHD [50, 52]. Several other phase I/II clinical trials testing CIML-NK cells in AML patients are ongoing (e.g., NCT04354025, NCT03068819, NCT02782546). These studies will provide further insights into the therapeutic potential of CIML-NK cells to limit the relapse of AML after HSCT.
Taken together, the efforts summarized above indicate that allogeneic NK cell therapy can indeed have a positive impact on the clinical situation of AML patients, which may be further improved by suitable adjustment of NK cell expansion and stimulation protocols. Nevertheless, treatment outcomes are still not satisfactory considering the remaining high numbers of relapses. Hence, different alternative treatment concepts are currently investigated in preclinical studies [53], which will be discussed in the following section.
NK Cell Engagers and Combinatorial Approaches to Enhance NK Cell Therapy
Many immunotherapeutics, including monoclonal antibodies, specific for tumor-associated antigens or immune checkpoint inhibitors, aim to increase the inherent cytotoxic efficiency of immune cells against cancer. Another therapeutic approach with the same goal is based on so-called “bispecific engagers,” which simultaneously bind to an activating receptor on cytotoxic immune cells and a specific tumor-associated antigen. This leads to an increase in tumor cell killing by steering the two types of cells into close proximity, channeling the inherent antitumor activity of the effector cells [54]. Up to now, most of such immunomodulatory interventions have focused on cytotoxic T cells. For example, the bispecific T-cell engager blinatumomab, which binds to CD3 on T cells and CD19 on B-ALL cells, has already been approved by the FDA and EMA. Nevertheless, increasing T-cell activity often results in toxic side effects related to immune overstimulation like CRS and ICANS, and in antigen escape [55].
In contrast, NK cell engagers may have a lower risk of side effects, due to the favorable cytokine and receptor repertoire of the effector cells. For example, the bispecific NK cell engager AFM13 binds to CD16a on NK cells and CD30 on lymphoma/leukemia cells [56]. AFM13 combined with cytokine pre-stimulated NK cells increased the killing of malignant cells in vitro and in vivo [57]. However, monotherapy with AFM13 only induced a 16.7% response rate in a recent phase II trial [58]. In order to increase killing efficiency, a trispecific NK cell engager was developed that in addition to CD16 also binds the activating NK cell receptor NKp46. This engager showed efficient clearance of LICs in vivo when targeting the common AML antigen CD123 [59]. Another trispecific NK cell engager, containing binding domains specific for CD16 and the AML antigen CD33 connected by an IL-15 sequence, has been clinically evaluated (NCT03214666), without dose-limiting toxicities, but enhanced cytotoxicity and in vivo NK-cell expansion [60]. An important challenge concerning NK cell engagers is the downregulation or shedding of CD16, triggered by prolonged contact with antibody-opsonized cells. This can be counteracted by inhibiting “a disintegrin and metalloprotease 17” (ADAM17), which is involved in the proteolytic cleavage of effector molecules like CD16 [61], or by overexpressing a non-cleavable CD16 version in NK cells for adoptive transfer [62].
Non-cleavable CD16 mutants are also used in approaches based on increasing cytotoxic NK cell functions like antibody-dependent cellular cytotoxicity. As such, a phase I clinical trial evaluating iPSC-derived NK cells genetically engineered to express a high-affinity non-cleavable CD16a (FT516) against AML reported an encouraging safety profile and efficacy [63]. Another phase I trial was conducted with the improved NK cell product FT538, which additionally expresses an IL-15 superagonist to increase persistence and carries a deletion of CD38 to avoid fratricide when combined with CD38-specific antibodies. No dose-limiting toxicities were observed in this study [64].
NK cell efficacy can also be enhanced by reducing inhibitory and increasing activating signals mediated by NK cell receptors via binding of their respective ligands on malignant cells. In this context, NKG2D and its ligands (NKG2DL) play a crucial role. NKG2D is one of the main activating receptors on NK cells, and the absence of its ligands on LICs leads to immune escape and relapse. However, inhibition of poly-ADP-ribose polymerase 1 (PARP) can result in the upregulation of NKG2DL on LICs, leading to their NK cell-mediated elimination and prevention of leukemia relapse and progression [65]. Accordingly, the PARP inhibitor talazoparib is being evaluated in a phase I/II trial in combination with allogeneic NK cells (NCT05319249). NKG2DL expression can also be upregulated by histone deacetylase inhibitors (HDACis) like valproic acid [66, 67]. Valproic acid treatment has already been shown to increase the cytotoxic efficiency of CAR-T cells in in vivo models of AML [68], and HDACis are under clinical evaluation for the treatment of AML [69].
Instead of enhancing activating signals, also decreasing inhibitory receptor expression on NK cells is effective. For example, CRISPR/Cas9-mediated deletion of the immune checkpoint NKG2A in NK cells increased cytotoxicity of the resulting effector cells against multiple myeloma [70], and can also be combined with the expression of AML-targeting CARs in NK cells (own unpublished data).
Adoptive cell therapy is most often preceded by a lymphodepleting chemotherapy regimen, which increases the persistence and efficacy of the subsequent treatment. While commonly a combination of fludarabine and cyclophosphamide is used [71], also the combination of cellular therapies with different chemotherapeutic regimens is currently under clinical investigation to enhance engraftment and overcome drug-induced resistance. Table 2 gives an overview of clinical trials that combine adoptive NK cell therapies with chemotherapy.
Identifier . | Phase . | Type of chemotherapy . | Timepoint of chemotherapy . | Type of NK cell therapy . | Country . |
---|---|---|---|---|---|
NCT05744440 | I | Azacitidine | Prior to NK cells | 2 × 1 × 107/kg or 5 × 107/kg allogeneic NK cells | China |
NCT04221971 | I | Cyclophosphamide/fludarabine | Combination | 3×1 × 107/kg haploidentical NK cells | China |
NCT05834244 | Ib | Azacitidine/venetoclax | Combination | Dose escalation/expansion allogeneic NK cells | USA |
NCT04347616 | I/II | Cyclophosphamide/fludarabine | Prior to NK cells | 1.0–3.0 × 109 allogeneic UCB-NK cells±IL-2 | The Netherlands |
NCT06152809 | I | Venetoclax | Prior to NK cells | Allogeneic CIML-NK cells±IL-2 | USA |
NCT05503134 | I/II | Cyclophosphamide/fludarabine | Prior to NK cells | 6 × 1 × 107/kg to 1 × 108/kg mbIL-21 expanded NK cells | USA |
Identifier . | Phase . | Type of chemotherapy . | Timepoint of chemotherapy . | Type of NK cell therapy . | Country . |
---|---|---|---|---|---|
NCT05744440 | I | Azacitidine | Prior to NK cells | 2 × 1 × 107/kg or 5 × 107/kg allogeneic NK cells | China |
NCT04221971 | I | Cyclophosphamide/fludarabine | Combination | 3×1 × 107/kg haploidentical NK cells | China |
NCT05834244 | Ib | Azacitidine/venetoclax | Combination | Dose escalation/expansion allogeneic NK cells | USA |
NCT04347616 | I/II | Cyclophosphamide/fludarabine | Prior to NK cells | 1.0–3.0 × 109 allogeneic UCB-NK cells±IL-2 | The Netherlands |
NCT06152809 | I | Venetoclax | Prior to NK cells | Allogeneic CIML-NK cells±IL-2 | USA |
NCT05503134 | I/II | Cyclophosphamide/fludarabine | Prior to NK cells | 6 × 1 × 107/kg to 1 × 108/kg mbIL-21 expanded NK cells | USA |
Several other NK cell-based combinatorial approaches are currently under preclinical and clinical investigation, e.g., monoclonal antibody therapy and immune checkpoint inhibitors, and were recently described in detail [53]. Additionally, targeting the immunosuppressive TME and epigenetic landscape produces promising results (for further reading, see [72]).
Enhancing the Specificity of NK Cells for AML by Expression of Chimeric Antigen Receptors
As discussed above, modifying the expression of endogenous receptors and ligands on NK cells and tumor cells can increase NK cell cytotoxicity against AML. Another promising approach to enhance the activity of NK cells is the expression of a CAR that can mediate selective recognition of malignant cells and targeted cytotoxicity.
CARs are synthetic activating receptors with distinct functional domains. The extracellular cell-targeting domain typically comprises a single-chain fragment variable (scFv) domain sourced from an antibody that recognizes a specific antigen on the surface of target cells. A spacer or hinge region separates the cell-targeting domain from the transmembrane domain (TMD) and provides flexibility for antigen recognition. Typically, hinge regions from CD8α, CD28, IgG4 or IgG1 are utilized in current CAR constructs. The TMD, frequently derived from CD8α or CD28, anchors the CAR within the cell membrane. It largely determines the degree of oligomerization of CAR molecules on the cell surface and therefore impacts receptor signaling [73]. The signal-transducing intracellular domain of a CAR usually includes the CD3ζ chain of the T-cell receptor complex. CD3ζ contains three immunoreceptor tyrosine-based activation motifs that initiate downstream signaling cascades and is expressed endogenously by NK cells, where it is involved in NCR and CD16 signaling [20]. Signaling is further potentiated in second- and third-generation CARs by costimulatory domains derived from CD28, 4-1BB (CD137), OX40 (CD134), or ICOS to augment effector cell activation, cytotoxicity, and proliferation [73]. CAR-T cells have been extensively studied and are used in the clinic for the treatment of B-ALL, B-cell lymphomas, and multiple myeloma. There are also many clinical studies investigating CAR-T cells for the treatment of AML. However, as of now none of these products have achieved regulatory approval due to unwanted toxicities and/or a lack of efficacy. As mentioned above, T cells can induce severe side effects that are typically not observed when using NK cells for adoptive therapy, which also applies to CAR-engineered cells [74, 75]. CAR-NK cells have already shown safety and clinical efficacy when targeting CD19 in r/r lymphoid malignancies in a recent phase I/II trial [18]. Although promising CAR designs that are optimized for NK cells have been generated [10], up to now clinically applied CAR-NK cells still employ CARs with “classical” CD28 and CD3ζ signaling domains typically used in CAR-T cell products [17].
The primary challenge in targeting AML blasts and LICs in AML with CARs is the lack of leukemia-specific target antigens, resulting in unwanted on-target/off-leukemia toxicity or fratricide, if the chosen target antigen is also expressed on NK cells themselves. In an effort to identify potential AML-specific CAR targets, one study utilized proteomic and transcriptomic analyses. Nevertheless, their algorithm did not identify promising single target candidates but suggested a combinatorial approach [76]. This leaves for now a limited number of target structures like CD123, CD33, NKG2DL, CLL-1, CD70, and CD7. Although not strictly selective for AML, they are frequently expressed on AML cells and are currently investigated in clinical trials with respective CAR-NK cells (see Table 3). As with cytokine-stimulated and ex vivo expanded but otherwise unmodified NK cells, different sources such as donor-derived primary cells or cell lines can be used for the generation of CAR-NK cells. These approaches are discussed in detail below, focusing first on established NK-92 cells that have been used as a basis for many of the initial studies on CAR-NK cells.
Identifier . | Phase . | Indication . | Target . | NK source . | Country . |
---|---|---|---|---|---|
NCT05734898 | N.A. | r/r AML | NKG2DL | Unknown | China |
NCT05987696 | I | AML, MRD | CD33 + CLL-1 | iPSC | China |
NCT06027853 | I | r/r AML | CLL-1 | iPSC | China |
NCT05247957 | N.A. | r/r AML | NKG2DL | Cord Blood | China |
NCT05574608 | Early I | r/r AML | CD123 | Unknown | China |
NCT02944162 | I/II | r/r AML | CD33 | NK-92 | China |
NCT04623944 | I | r/r AML, r/r MDS | NKG2DL | Unknown | USA |
NCT06006403 | I/II | r/r AML, BPDCN | CD123 | Unknown | China |
NCT05215015 | Early I | AML | CD33 + CLL-1 | Unknown | China |
NCT06201247 | Early I | r/r AML | CD123 | Unknown | China |
NCT05092451 | I/II | AML, MDS, B-cell lymphoma | CD70 | Cord Blood | USA |
NCT05008575 | I | r/r AML | CD33 | Unknown | China |
NCT02742727 | I/II | r/r AML, Lymphoma | CD7 | Unknown | China |
Identifier . | Phase . | Indication . | Target . | NK source . | Country . |
---|---|---|---|---|---|
NCT05734898 | N.A. | r/r AML | NKG2DL | Unknown | China |
NCT05987696 | I | AML, MRD | CD33 + CLL-1 | iPSC | China |
NCT06027853 | I | r/r AML | CLL-1 | iPSC | China |
NCT05247957 | N.A. | r/r AML | NKG2DL | Cord Blood | China |
NCT05574608 | Early I | r/r AML | CD123 | Unknown | China |
NCT02944162 | I/II | r/r AML | CD33 | NK-92 | China |
NCT04623944 | I | r/r AML, r/r MDS | NKG2DL | Unknown | USA |
NCT06006403 | I/II | r/r AML, BPDCN | CD123 | Unknown | China |
NCT05215015 | Early I | AML | CD33 + CLL-1 | Unknown | China |
NCT06201247 | Early I | r/r AML | CD123 | Unknown | China |
NCT05092451 | I/II | AML, MDS, B-cell lymphoma | CD70 | Cord Blood | USA |
NCT05008575 | I | r/r AML | CD33 | Unknown | China |
NCT02742727 | I/II | r/r AML, Lymphoma | CD7 | Unknown | China |
AML, acute myeloid leukemia; r/r, relapsed/refractory; MRD, minimal residual disease; BPDCN, blastic plasmacytoid dendritic cell neoplasm; MDS, myelodysplastic syndrome.
Genetically Engineered NK-92 Cells for the Treatment of AML
Continuously expanding NK cell lines like NK-92 have not only been valuable tools to study NK cell biology, but due to their stable and homogeneous phenotype and the ease of genetic modification have also been instrumental to first establish the concept of genetically engineered NK cells as a standardized off-the-shelf therapeutic [20, 77, 78]. Initially derived from the peripheral blood of a patient with Non-Hodgkin lymphoma (NHL), NK-92 cells are large granular lymphocytes, which share many features with activated primary NK cells and display potent cytotoxic activity against a broad range of target cells [79]. Not only their growth but also their cytolytic functions are dependent on the pro-inflammatory cytokine IL-2. Phenotypically, NK-92 cells lack CD16 expression as well as most of the KIRs, similar to CD56bright NK cells. While most inhibitory NK cell receptors are absent, NK-92 cells express a broad variety of activating receptors such as NKG2D, 2B4, and the NCRs NKp30 and NKp46 [80]. Additionally, high amounts of cytolytic granules and expression of ligands like TRAIL and Fas-L enable the killing of malignant cells through the release of granzymes and perforin as well as in a death-receptor-dependent manner. A number of early-phase clinical trials with unmodified NK-92 cells demonstrated that repeated infusions even of high cell numbers were in general well tolerated without severe side effects [77]. This included one study with 7 AML patients, where stable or reduced blast counts were transiently observed in the bone marrow of 3 of the patients [81]. Due to their origin from an NHL patient, up to now NK-92 cells are irradiated as a safety measure before clinical application, which prevents further proliferation and permanent engraftment, but permits to retain the functionality of the cells for several days [82].
Early approaches for genetic engineering of NK-92 cells addressed their dependence on exogenous IL-2. One such strategy was based on particle-mediated transfer of human IL-2 cDNA, resulting in NK-92MI cells that secrete IL-2 into the culture medium and maintain growth and cytotoxicity in an autocrine fashion [83]. Recently, NK-92 cells were modified to express an endoplasmic reticulum (ER)-retained IL-2 derivative together with a high-affinity variant of CD16 (CD16F158V), in order to circumvent potential side effects of systemic IL-2 exposure and to provide endogenously CD16-negative NK-92 cells with the ability to mediate antibody-dependent cellular cytotoxicity [84, 85]. Combined with therapeutic IgG1 antibodies, these high-affinity NK-92 cells are investigated in early-phase clinical trials in different solid tumor indications [77]. With respect to AML, NK-92 cells carrying CD16F158V were applied in a preclinical study together with an anti-CD123 antibody, displaying selective cytotoxicity against primary AML cells and prolonging survival in an AML xenograft model [86]. To further enhance tumor specificity and efficacy, NK-92 cells were also engineered with CARs targeting different antigens expressed by solid tumors or hematological malignancies [20]. One such approach based on NK-92 cells expressing a HER2 (ErbB2)-specific second-generation CAR is currently being investigated in a phase I clinical trial in patients with recurrent glioblastoma (NCT03383978), with direct intracranial injection of the irradiated CAR-NK cells shown to be feasible and safe in the completed dose-escalation part of the trial [87].
With respect to hematological malignancies, CAR-NK-92 cells demonstrated enhanced in vitro and in vivo activity in preclinical models of B-cell leukemia and lymphoma [84, 88‒91], multiple myeloma [92, 93], T-cell malignancies [94‒96], and AML [97]. In a first-in-human clinical trial, NK-92 cells were lentivirally transduced to express a CD33-specific third-generation CAR with 4-1BB and CD28 costimulatory domains. Safety of infusions with doses of up to 5 × 109 irradiated cells was shown in 3 patients with r/r AML (NCT02944162). Although treatment was well tolerated, no clinical efficacy was demonstrated, which was mainly attributed to the limited in vivo persistence of the irradiated cells for 1 week [21].
Nevertheless, this trial prompted a series of preclinical studies aimed at increasing the efficacy of CAR-NK-92 cells against AML by improving their functionality and employing alternative target antigens (see Table 4). These include CD4 [98], CD123 [97], CD135 (fms-like tyrosine kinase 3 [FLT3]) [99], CD276 (B7-H3) [100], and mesothelin (MSLN) [101]. In a recently reported study, a CD123-CAR was co-expressed together with IL-15, which further enhanced efficacy against AML in a patient-derived xenograft (PDX) model [97]. There is also an ongoing phase I/II study with CD7-targeting NK-92 cells that includes patients with CD7-positive AML (NCT02742727). However, the current status of this trial is unknown.
Target . | Targeting moiety . | Other modifications . | Response . | Reference . |
---|---|---|---|---|
CD4 | 3rd generation CAR | - | Effective against AML cell lines and patient-derived AML samples in vitro | You et al. [96] (2019) |
CD4scFv-CD28-4-1BB-CD3ζ | Reduced leukemia burden and prolonged survival in a systemic MOLM-13 leukemia model | |||
CD7 | 3rd generation CAR | - | Phase I/II clinical trial (status unknown) | NCT02742727 |
CD7scFv-CD28-4-1BB-CD3ζ | ||||
CD33 | 3rd generation CAR | - | Phase I clinical trial | NCT02944162 Marin (2024) [18] |
Infusion of up to 5 billion cells into AML patients with no severe side effects | ||||
CD33scFv-CD28-4-1BB-CD3ζ | No obvious clinical efficacy | |||
CD123 | 3rd generation CAR | IL-15 | Effective against primary AML cells in vitro and in an in vivo AML PDX-model | Chen et al. [95] (2016) |
CD123scFv-CD28-4-1BB-CD3ζ | ||||
CD16176V + CD123 mAb (7G3) | Prolonged survival in primary human AML xenograft model | Boissel et al. [84] (2016) | ||
CD135 (FLT3) | 2nd generation CAR | - | In vitro efficacy against AML cell lines, prolonged survival of treated mice in a MOLM-13 AML xenograft model | Morgan et al. [97] (2021) |
CD135scFv-CD28-CD3ζ | ||||
CD276 (B7-H3) | 2nd generation CAR | TIGIT, CBLB, NKG2A knock-out | Increased cytotoxicity against U937 cells; no consistent activity enhancement against other AML cell lines | Salman et al. [98] (2019) |
CD276scFv-CD28-CD3ζ | ||||
MSLN | 2nd generation CAR | - | Specific targeting of MSLN+ AML cells in vitro | Mansour et al. [99] (2023) |
MSLNscFv-4-1BB-CD3ζ | Reduced tumor burden in a nomo-1 xenograft mouse model |
Target . | Targeting moiety . | Other modifications . | Response . | Reference . |
---|---|---|---|---|
CD4 | 3rd generation CAR | - | Effective against AML cell lines and patient-derived AML samples in vitro | You et al. [96] (2019) |
CD4scFv-CD28-4-1BB-CD3ζ | Reduced leukemia burden and prolonged survival in a systemic MOLM-13 leukemia model | |||
CD7 | 3rd generation CAR | - | Phase I/II clinical trial (status unknown) | NCT02742727 |
CD7scFv-CD28-4-1BB-CD3ζ | ||||
CD33 | 3rd generation CAR | - | Phase I clinical trial | NCT02944162 Marin (2024) [18] |
Infusion of up to 5 billion cells into AML patients with no severe side effects | ||||
CD33scFv-CD28-4-1BB-CD3ζ | No obvious clinical efficacy | |||
CD123 | 3rd generation CAR | IL-15 | Effective against primary AML cells in vitro and in an in vivo AML PDX-model | Chen et al. [95] (2016) |
CD123scFv-CD28-4-1BB-CD3ζ | ||||
CD16176V + CD123 mAb (7G3) | Prolonged survival in primary human AML xenograft model | Boissel et al. [84] (2016) | ||
CD135 (FLT3) | 2nd generation CAR | - | In vitro efficacy against AML cell lines, prolonged survival of treated mice in a MOLM-13 AML xenograft model | Morgan et al. [97] (2021) |
CD135scFv-CD28-CD3ζ | ||||
CD276 (B7-H3) | 2nd generation CAR | TIGIT, CBLB, NKG2A knock-out | Increased cytotoxicity against U937 cells; no consistent activity enhancement against other AML cell lines | Salman et al. [98] (2019) |
CD276scFv-CD28-CD3ζ | ||||
MSLN | 2nd generation CAR | - | Specific targeting of MSLN+ AML cells in vitro | Mansour et al. [99] (2023) |
MSLNscFv-4-1BB-CD3ζ | Reduced tumor burden in a nomo-1 xenograft mouse model |
AML, acute myeloid leukemia; FLT3, fms-like tyrosine kinase 3; mAb, monoclonal antibody; MSLN, mesothelin; PDX, patient-derived xenograft; scFv, single-chain fragment variable.
As already mentioned above, disrupting inhibitory checkpoints can enhance the anti-leukemic activity of NK cells. In NK-92 cells, it has been observed that antibody blockade of TIGIT increased killing of AML target cell lines MV-4-11, TF-1, and OCI-AML3. Enhanced lysis of AML cells by NK-92 was further augmented by combined targeting of TIGIT, CD39 and the adenosine A2A receptor (A2AR) [102]. Nevertheless, while CRISPR/Cas9-mediated knock-out of TIGIT, CBLB, and NKG2A increased cytotoxicity of CAR-NK-92 cells against the AML cell line U937, the effects on the activity against other AML cell lines were variable [103].
Genetically Engineered Primary NK Cells for the Treatment of AML
A prominent target used for CAR-NK approaches for the treatment of AML is CD123, which was already mentioned in the context of NK-92-CAR cells. CD123 is the α-subunit of the IL-3 receptor (IL-3R), which forms a heterodimer with the IL-3R β-subunit to activate JAK/STAT signaling during hematopoiesis [104]. CD123 is highly expressed by AML cells, but at a significantly lower level by hematopoietic stem cells (HSCs), which may provide a sufficient degree of selectivity. In a recent study, second-generation CD123-CAR-NK cells were generated from the peripheral blood of healthy donors and demonstrated anti-leukemic activity in in vitro and in vivo settings. As compared to CD123-CAR-T cells, CD123-CAR-NK cells showed reduced on-target/off-tumor toxicity in immunodeficient mice engrafted with human hematopoietic cells. While CD123-CAR-T cells induced toxicity against hCD34+ CD38-stem cells, leading to the death of mice, treatment with CD123-CAR-NK cells did not result in toxicity ensuring the survival of all mice until the end of the experiment [105]. Another study utilized peripheral blood NK cells to generate CD123-CAR-NK cells with composite 2B4 zeta or 4-1BB zeta signaling domains, both of which exhibited anti-AML activity in vitro. However, in vivo, only 2B4 zeta CD123-CAR-NK cells demonstrated improved antitumor activity when compared to non-transduced NK cells. To enhance persistence, CAR-NK cells were further modified to secrete IL-15 (sIL-15). In vivo, these cells showed anti-AML activity in one model but also induced IL-15-associated lethal systemic toxicities in a second model, highlighting the need for a tightly controlled or inducible cytokine secretion [106]. CD123-CAR-NK cells generated from peripheral blood mononuclear cells (PBMCs) were also used in a study investigating the fully automated separation of NK cells with the CliniMACS Prodigy®, and different expansion protocols with GMP-compliant media. The Prodigy manufacturing process revealed high effector cell viability and purity. However, manually performed experiments resulted in significantly higher expansion rates. The expanded NK cells were successfully modified by retroviral transduction to express a CD123-CAR and subsequently displayed potent antitumor activity against the AML cell line KG-1a and primary AML blasts [107]. Following further optimization, an improved expansion protocol was integrated into the Prodigy workflow to achieve GMP-grade manufacturing of CAR-NK cells. This process involved immunomagnetic separation that was followed by NK cell expansion over 14 days, resulting in high purity and viability [108]. Therefore, it is feasible to produce CAR-NK cells in clinically relevant numbers and a GMP-compliant manner. With respect to clinical application, currently, three trials in China are employing CD123-CAR-NK cells for the treatment of AML (see Table 3).
CD33 serves as a myeloid-specific transmembrane sialic acid-binding receptor that is highly expressed by almost all AML cells and leukemic stem cells, whereas healthy hematopoietic cells show lower CD33 levels [109]. Although CD33 represents a promising target for AML treatment, preclinical models have indicated hematopoietic toxicity associated with CD33-targeted therapies [110]. In one preclinical study, peripheral blood-derived CD33-CAR-NK cells were generated and demonstrated cytotoxic antitumor activity against CD33-positive OCI-AML2 and primary AML cells in vitro. The treatment of OCI-AML2 xenograft mouse models with CD33-CAR-NK cells significantly reduced leukemic burden and inhibited bone marrow engraftment of leukemic cells, with no obvious side effects [111]. In addition, automated generation of CD33-CAR-NK cells using the CliniMACS Prodigy device was demonstrated, with the functionality of the Prodigy produced CD33-CAR-NK cells comparable to that of CD33-CAR-NK cells produced manually at a small scale, showing the potential to generate high numbers of CAR-NK cells under GMP conditions for clinical application [112]. Another study targeting CD33 aimed to further increase NK cell cytotoxicity by the expression and secretion of an anti-CD16 antibody (B16) that binds to Fc receptors, thereby activating the NK cells. In vitro, the bifunctional CD33-CAR/B16-antibody expressing NK cells demonstrated an approximately four-fold increase in efficiency compared to CAR-NK cells carrying only the CD33-CAR. An in vivo THP-1 xenograft mouse model displayed clearance of the leukemic cells and increased survival of the mice [113].
Recognition of the NKG2D ligands (NKG2DL) MHC class-I related molecules MICA and MICB, or the UL16-binding proteins ULBP1 to ULBP6 by their receptor NKG2D on NK cells can trigger effective cytotoxicity [114]. The expression of NKG2DL is frequently increased during and after malignant transformation and has been observed in various cancers, including AML [110, 115]. Hence, NKG2DL can serve as promising universal targets for multiple cancers, and preclinical studies are exploring CAR-NK cells recognizing NKG2DL for the treatment of multiple myeloma [116, 117]. In an approach to target AML, non-viral piggyBac transposon technology was employed to engineer peripheral blood-derived NK cells to express a CAR containing the extracellular domain of NKG2D for specific targeting of NKG2DL. The efficacy of such cells was demonstrated in vitro against AML cell lines and in vivo in a KG-1 AML xenograft model. In order to enhance in vivo persistence, the NKG2D-CAR-NK cells were further modified to co-express IL-15, resulting in improved in vivo tumor control and prolonged survival [118].
The C-type lectin-like molecule-1 (CLL-1, CLEC12A) is highly expressed on AML LICs and blast cells but not on healthy HSCs, making CLL-1 a promising target. One patient with secondary AML responded to treatment with CLL-1-CAR-T cells with morphological, immunophenotypic, and molecular CR for over 10 months [119]. Currently, three different CAR-NK clinical trials are ongoing that target CLL-1 alone or CLL-1 together with CD33 (see Table 3).
Also, CD70 is a promising target for AML as it is highly expressed on LICs and blasts, but absent on healthy HSCs [76, 120]. However, in response to activation, CD70 is upregulated on peripheral blood-derived NK cells, leading to significant fratricide upon introduction of a CD70-CAR [120]. Despite this complication, genetically engineered cord blood-derived CD70-CAR-NK cells which secrete IL-15 are currently evaluated in a phase I/II clinical trial (NCT05092451; Table 3).
CD7 is an alternative AML-associated marker targeted in a clinical trial. This molecule is mostly expressed in healthy T cells, NK cells, and their precursors and serves as a costimulatory receptor during maturation [110]. CD7 expression is found in approximately 30% of AML cases and appears to be correlated with aggressive disease progression and resistance to other therapies. Although preclinical data with CAR-T cells indicate efficacy in AML, CAR-NK cell-based strategies to target CD7 may cause extensive fratricide and depletion of respective CAR-NK cells [121]. A possible solution for this problem is genetic deletion of the respective antigen in the CAR cells [122] or in HSCs used for repopulation after treatment [123].
There are also other surface antigens that are frequently expressed on AML cells, which have not yet been targeted in clinical CAR-NK trials. One of these molecules is TIM3, which is present on most AML LICs at higher levels than on normal HSCs [124, 125]. Recently, TIM3-CAR iPSC clones were generated and differentiated into TIM3-CAR-NK-like cells, which showed enhanced cytotoxicity against TIM3-positive AML cells when compared to unmodified iPSC-derived NK-like cells [126].
Another surface receptor that is commonly expressed on AML blasts is FLT3, a receptor tyrosine kinase with mutations in approximately 30% of AML patients [127]. Although FLT3 inhibitors are available, they do not significantly improve clinical outcomes for patients with FLT3 mutations. FLT3 has thus been targeted in different CAR-T studies, despite its expression also on healthy HSCs and HPCs [110]. Indeed, one study reported CAR-T cell reactivity against human HSCs, indicating a possible need to include an off-switch in the CAR cells for the regulation of myelotoxicity [128]. In another recent study, FLT3 targeting CAR-NK cells that secrete soluble IL-15 were generated from cord blood NK cells. These FLT3-CAR_sIL-15 NK cells showed significantly prolonged survival in MOLM-13 and patient-derived xenograft AML models even after one freezing and thawing cycle [99].
CD244 (also known as SLAMF4 or 2B4) was identified as a potential AML target, as the knockdown of CD244 significantly hampered the proliferation of leukemic cells both in vitro and in vivo without compromising the repopulating capacity of HSCs [129]. However, since CD244 is normally also expressed on HSCs and HPCs (hematopoietic progenitor cells) [130], there are concerns about potential on-target/off-tumor toxicity when targeted by a CAR. Furthermore, CD244 is naturally expressed in NK cells at high levels, where it serves an activating function [131]. Accordingly, extensive fratricide will likely prohibit its use as a target for CAR-NK cells.
Recently, CD86 (B7-2) has been suggested as a new potential AML immunotherapy target based on pan-cancer analyses that used public databases [109]. Physiologically expressed on professional antigen-presenting cells, CD86 provides costimulatory signals via CD28 that are crucial for T-cell activation. However, due to stimulation of both CD28 and CTLA-4, CD86 also plays a role in down-modulating T-cell activity [132]. In the context of AML, CD86 was found to be expressed at high levels and this overexpression was associated with poor prognosis [133].
Another target antigen that recently attracted attention is the colony-stimulating factor 1 receptor (CSF1R), which is primarily expressed by monocytes, macrophages, and dendritic cells. Targeting CSF1R would thus address the tumor-supporting microenvironment [134]. In a study from 2019, small-molecule and small-interfering RNA (siRNA) screens revealed sensitivity for CSF1R inhibitors for 23% of AML samples. Although responses varied widely from highly sensitive to non-sensitive, CSF1R might be a promising target expressed on tumor-supportive cells that promote AML progression [135].
Typically, the expression of the different target antigens described above is very heterogenous in AML, possibly enabling some malignant cells to escape CAR-dependent immune recognition [136]. One strategy to circumvent this type of immune escape and subsequent relapses driven by antigen-negative cells is to increase the selective pressure on AML cells by targeting two antigens simultaneously. Such bispecific or dual-targeting CARs are studied extensively in preclinical and clinical settings. Dual antigen targeting can be realized by combining two NK cell products each carrying a different monospecific CAR, by equipping NK cells with two independent CARs that are expressed in the same cell, by employing recombinant CAR designs which include two antigen binding domains in a single CAR, or by using adapter CAR systems that are triggered by different soluble engager molecules [73, 137, 138]. As these cells recognize one or the other antigen, CAR effectors used for this strategy are frequently termed OR-gated CAR immune cells. There is one combinatorial clinical trial with CAR-NK cells targeting CD33 and CLL-1 (NCT05215015), but no data regarding the safety and efficacy of these cells in patients have been published yet. Also, other target combinations for AML have been suggested based on proteomic analyses, which include CD33 and ADGRE2, CLL-1 and CCR1, CD33 and CD70, or LILRB2 and CLL-1, but require further preclinical investigation [76].
Conclusions and Future Prospects
As outlined above, the safety and efficacy of CAR-NK cells in AML patients are already under investigation in different clinical trials. Nevertheless, these approaches still require further optimization to overcome general limitations of CAR strategies for the treatment of AML such as the lack of safe and truly AML-specific target antigens, and limitations of CAR-NK cells like insufficient in vivo persistence, relatively poor tumor infiltration and NK cell inhibition and dysfunction in the TME. Indeed, the AML niche in the bone marrow is hostile to immune cells and hampers the effects of immunotherapy [139]. Hence, different approaches are under development that aim at overcoming these limitations by next-level gene editing of NK cells (summarized in Fig. 2). On-target/off-tumor toxicity is a major concern in AML therapy as CAR target antigens are shared between blasts and healthy cells. In addition, heterogeneity of AML cells can lead to immune escape and relapse driven by antigen-negative leukemic cells. While dual-targeting approaches could limit this type of immune escape, on-target/off-tumor effects may potentiate when addressing two targets simultaneously. A promising strategy to avoid this complication and protect stem cells while selectively killing malignant cells is the use of NOT-gated CAR-NK cells. These cells combine activating and dominant inhibitory CARs, killing cells only when the inhibitory CAR is not triggered [137].
Following this strategy, CAR-NK cells were generated that target CD33 and FLT3 through an activating CAR, enabling lysis of CD33+ and/or FLT3+ AML cells. Simultaneously, the CAR-NK cells expressed an inhibitory CAR that recognizes endomucin expressed by healthy HSCs and early hematopoietic progenitor cells and can overrule signaling of the activating CAR. Conceptually, stem cells are protected from off-tumor toxicity by this approach. In addition, soluble IL-15 (sIL-15) released by the CAR-NK cells is supposed to promote NK cell expansion, persistence, and tumor killing. These engineered NK cells were shown to be far superior to unmodified NK cells and efficiently cleared AML blasts both in vitro and in patient-derived xenograft AML mouse models [140‒142]. Hence, these and similar logic-gated CAR-NK cells represent an attractive strategy for the development of advanced CAR-NK cells for future AML therapies [137, 138], likely allowing to also target antigens that cannot be safely addressed with traditional CAR designs. Likewise, intracellular antigens that are presented via HLA may be targeted by NK cells engineered to express a cognate T-cell receptor or CAR recognizing the respective peptide/MHC complex. Similar to bispecific T-cell engagers and CAR-T cells [143, 144], several approaches with CAR-NK cells are under development targeting intracellular antigens such as ERP57 [145], PRAME [145], and NPM1c [145].
IL-15 is commonly used not only to expand NK cells in vitro but also to promote cellular survival and proliferation in vivo. Due to the superior performance of CD19-directed CAR-NK cells which overexpress soluble IL-15 in mouse models [146], this approach has been translated to a CAR-NK cell clinical trial for the treatment of B-cell malignancies. Thereby, long-term persistence of the CAR-NK cells was observed in the patients, without compromising on safety. While in responders CAR-NK cell DNA was detected for at least 1 year, no neurotoxicity or CRS higher than grade 1 was observed [17, 18]. Apart from sIL-15, also a membrane-bound derivative of IL-15 (mbIL-15) has been overexpressed in NK cells. Anchoring of this molecule at the NK cell surface was achieved by connecting the IL-15 sequence to the CD8α hinge and transmembrane domains. Notably, mbIL-15 showed superior effects on survival and proliferation when compared to sIL-15 [147]. These studies demonstrate that the genetic engineering of NK cells to express cytokines like IL-15 can have major beneficial effects on NK cell functions. Similar approaches to equip CAR-NK cells with cytokines other than IL-15 or engineered designer cytokines with novel activities, potentially combined with additional genetic modifications, hold promise to further boost antitumor efficacy, and recruit endogenous bystander immune cells for a concerted antitumor attack.
This is supported by a study that not only showed the positive effects of sIL-15 on NK cell survival and function but also demonstrated enhanced activity of bystander T cells, possibly potentiating the antitumor immune response [99]. While so far not reported for IL-15-engineered CAR-NK cells, continuous IL-15 signaling has been described to exhaust unmodified NK cells, concomitant with detrimental metabolic changes [45]. To overcome this limitation and potentiate the effects of IL-15, genetic deletion of CISH, a key negative regulator of IL-15 signaling in NK cells, has been performed, successfully restoring metabolic programs and preventing cellular exhaustion [148‒150]. In this context, it is important to note that a deficit in metabolic fitness of CAR-NK cells resulted in the loss of tumor control in long-term xenograft mouse models [151], which can also be heavily influenced by the presence of cytokines and metabolites in the TME [152, 153]. To circumvent immunosuppressive mechanisms, in addition to CISH deletion, also the adenosine-generating CD73 molecule was successfully targeted [154]. Likewise, different approaches to (genetically) enhance metabolic signaling by modulating SREBP, mTOR, or cMyc activity are under investigation [152, 153].
Other strategies to armor NK cells against the hostile and immunosuppressive TME encompass the ablation of distinct immune checkpoints, making NK cells resistant to NKG2A-mediated [70, 155, 156] or TIGIT-mediated inhibition [157]. Likewise, TGF-β-mediated immune suppression can be overcome by overexpressing a dominant-negative receptor [158]. As NK cells express a variety of inhibitory receptors and checkpoint molecules which balance their activity, many more of these molecules are of potential interest for genetic editing to increase the NK cells’ antitumor activity [159, 160].
In addition, attempts have been made to overcome the susceptibility of donor NK cells to therapeutic antibody treatment. For example, by ablating CD38, NK cells can be redirected against CD38-expressing tumor cells with the antibody daratumumab without inducing fratricide. The efficacy of this approach was further enhanced by overexpressing a high-affinity CD16 receptor in the NK cells [161].
To enhance NK cell trafficking to tumor sites, distinct chemokine receptors, whether already produced endogenously or not, can be overexpressed in NK cells. CXCR4 overexpression was shown to enhance migration to the bone marrow [6, 162]. Overexpression of CCR7 redirected gene-edited NK cells toward CCL19-expressing lymphoma cells and improved tumor clearance in xenograft models [163]. Likewise, overexpression of CXCR2, CCR2, or CCR4 was shown to successfully guide NK cells to tumor sites [164, 165], and CCR5 overexpression was effectively utilized in combination with CCL5-armed oncolytic viruses to augment their antitumor efficacy [166].
To avoid the costs and complexity of manufacturing virally transduced CAR-NK cell products in a GMP-compliant manner, non-viral methods have successfully been applied for the genetic modification of NK cells [167]. This also addresses safety concerns and regulatory constraints associated with the use of viral vectors. Likewise, sophisticated methods for gene editing such as designer nucleases and CRISPR/Cas systems have been developed and are being adapted and improved to efficiently and safely modify NK cells [168].
While not all of the very promising developments summarized above have already been evaluated for the treatment of AML, they hold enormous potential to enhance functionality, specificity, safety, and long-term efficacy of CAR-engineered NK cells and redirect them more efficiently against AML. Certainly, advancing CAR designs, unlocking new target molecules, increasing persistence, proliferation, and tumor infiltration of NK cells, and armoring them against immunosuppression will all be decisive for the development of highly effective next-generation CAR-NK cell therapies for AML.
Acknowledgments
We thank the authors of the literature cited in this review for their pioneering work, and the patients who contributed to the clinical trials summarized in this study. We apologize to all colleagues in the field whose work we could not include due to space restrictions. Figures were generated with BioRender.
Conflict of Interest Statement
E.U. is an advisory board member of Phialogics and CRIION and has sponsored research projects with Gilead and BMS. The other authors declare no competing financial interest related to the presented work. D.S., U.K., W.S.W., and E.U. are named as inventors on patent applications and patents related to optimize CAR designs. U.K. is a consultant and received speaker fees from AstraZeneca, Affimed, Glycostem, GammaDelta, Zelluna, Miltenyi Biotec, Novartis Pharma GmbH, and Bristol-Myers Squibb GmbH & Co. KGaA.
Funding Sources
The authors of this study were supported by the Deutsche Kresbhilfe (German Cancer Aid), Deutsche Forschungsgemeinschaft (DFG, German Research Foundation, Project-ID 318346496), Collaborative Research Center CRC/IRTG 1292/TP12, the Gilead Research Award, the Foundation “Hilfe für krebskranke Kinder Frankfurt e. V.” in frame of the C3OMBAT-AML consortium, the Foundation “Menschen für Kinder,” the LOEWE Center Frankfurt Cancer Institute, and the German Cancer Consortium (DKTK).
Author Contributions
All authors contributed to individual parts of this review article. F.G. and J.S. reviewed the literature and wrote the initial manuscript draft. E.U. and W.S.W. supervised the work.