Chimeric antigen receptor (CAR) natural killer (NK)-cell therapy: leveraging the power of innate immunity

Summary

Chimeric antigen receptor (CAR) T cells are a rapidly emerging form of cancer treatment, and have resulted in remarkable responses in refractory lymphoid malignancies. However, their widespread clinical use is limited by toxicity related to cytokine release syndrome and neurotoxicity, the logistic complexity of their manufacturing, cost and time-to-treatment for autologous CAR-T cells, and the risk of graft-versus-host disease (GvHD) associated with allogeneic CAR-T cells. Natural killer (NK) cells have emerged as a promising source of cells for CAR-based therapies due to their ready availability and safety profile. NK cells are part of the innate immune system, providing the first line of defence against pathogens and cancer cells. They produce cytokines and mediate cytotoxicity without the need for prior sensitisation and have the ability to interact with, and activate other immune cells. NK cells for immunotherapy can be generated from multiple sources, such as expanded autologous or allogeneic peripheral blood, umbilical cord blood, haematopoietic stem cells, induced pluripotent stem cells, as well as cell lines. Genetic engineering of NK cells to express a CAR has shown impressive preclinical results and is currently being explored in multiple clinical trials. In the present review, we discuss both the preclinical and clinical trial progress made in the field of CAR NK-cell therapy, and the strategies to overcome the challenges encountered.

The past decade has witnessed the accelerated emergence of immunotherapy treatments that harness the power of the immune system to treat cancer and what many in the oncology community refer to as the ‘fifth pillar’ of cancer therapy, along with surgery, chemotherapy, radiotherapy and targeted therapy.¹ Two major immunotherapies offer new hope for potential curative responses to patients with terminal stages of cancer: immune checkpoint inhibitors² and adoptive cell therapy with chimeric antigen receptor (CAR)-T cells.³ CARs have been used extensively to redirect the specificity of autologous T cells against lymphoid malignancies with striking positive clinical activity,^4–8 with the first two such treatments, tisagenlecleucel (Kymriah, Novartis) and axicabtagene ciloleucel (Yescarta, Gilead), gaining United States Food and Drug Administration (FDA) approval in 2017. However, CAR-T cells have several drawbacks, including high manufacturing costs and serious toxicities such as cytokine release syndrome (CRS) and neurotoxicity. These challenges call for the investigation of novel, universal, efficient and safe cell therapy products. Natural killer (NK) cells do not carry the risk of graft-versus-host disease (GvHD) and therefore offer the potential for an off-the-shelf cellular product that could be readily available for immediate clinical use. NK cells for immunotherapy may be generated from different sources, both autologous and allogeneic, including peripheral blood (PB), cord blood (CB), haematopoietic stem cells (HSCs), induced pluripotent stem cells (iPSCs) and NK cell lines.^9–13 In preclinical studies, CAR NK-cell therapy has shown impressive efficacy and this approach is currently being investigated in multiple clinical trials. In the present review, we discuss advances made in the field of CAR NK-cell therapy, the innovative gene-editing technologies used to improve their activity, the sources of NK cells for clinical use, and the current limitations and perspectives on the application of this approach in the clinic.

NK cell biology

NK cells are essential players of the innate immune system and constitute the first line of defence against cancer cells.¹⁴ NK cells derive from CD34⁺ common lymphoid progenitor cells and differentiate into immature and mature NK cells in the bone marrow (BM).¹⁵ They are then distributed into lymphoid and non-lymphoid peripheral organs and tissues^15–18 including the PB, spleen, lung, liver and uterus.¹⁹

Classically, human NK cells are CD56⁺CD3⁻.²⁰ The activating receptor natural cytotoxicity triggering receptor 1 (NCR1 or NKp46) is also expressed on almost all human NK cells.^20,21 The level of CD56 expression distinguishes two major subsets of NK cells, CD56^dim and CD56^bright.²⁰ CD56^dim NK cells constitute ~90% of the PB population of NK cells. They are fully mature and predominantly function as cytotoxic cells, although they can also contribute to early cytokine production. In contrast, CD56^bright cells are less mature and comprise about 5–15% of the total PB NK-cell population, although they form the predominant population of NK cells in the lymph node.¹⁶ The primary function of CD56^bright NK cells is to produce cytokines.^22–24 The CD56^bright NK cells can also be primed to acquire potent cytotoxic effector function after exposure to cytokines.²⁵ Therefore, the function of NK cells is multifaceted. They possess potent cytotoxic properties while at the same time functioning as cytokine-producing cells.²⁶ NK cells infiltrate into tissues and kill pathogen-infected, malignant or stressed cells.^27,28 NK-cell activation is mediated through different mechanisms, including an interplay between the signals from activating [e.g. NK group 2 member D (NKG2D), NKp46, NKp33] and inhibitory receptors (e.g. NKG2A) (Table I), through direct CD16A signalling, which triggers antibody-dependent cell-mediated cytotoxicity (ADCC), and via various cytokines such as type I interferon (INF), interleukin (IL)-2, IL-12, IL-15 and IL-18.²⁹ NK cells can also receive signals through toll-like receptors (TLRs) expressed on their surface, which recognie pathogen-associated molecular patterns (PAMPs) expressed by injured cells.³⁰ NK cells kill their targets by releasing lytic granules such as perforin and granzymes, as well as by induction of death signals through the expression of death receptors [Fas ligand (FasL)/Fas, tumour necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL)/TRAIL receptors (TRAIL-R)].^31,32 Cytotoxic NK cells predominantly target cells that have downregulated type I major histocompatibility complex (MHC-I),³³ normally expressed by healthy cells of the body (Fig 1). Downregulation of MHC-I is a common mechanism by which cancer cells and virus-infected cells evade recognition by cytotoxic T lymphocytes (CTLs) through their T-cell receptors. Hence, NK cells may provide an important mechanism to target cancer cells that lose MHC-I expression (‘missing self’) in response to T-cell immune surveillance and thus, provide an additional layer of immune defence.^33–35

Table I.

Activating and inhibitory natural killer cell receptors.

Receptor	Ligand(s)
Activating receptors
Activating KIR	HLA-A11,-Bw4, -C
NKG2D	MICA/B, ULBP1 to -6
DNAM-1 (CD226)	PVR (Necl5, CD155), nectin 2 (CD112)
CRTAM (CD355)	Necl-2
NKp80	AICL1
NKp65	KACL
NKp46 (CD335)	Viral HA/HN, CFP (properdin)
NKp44 (CD336)	Viral HA/HN, PCNA, MLL5, PDGF-DD
NKp30 (CD337)	B7-H6, BAT3, HCMV pp65
CD16	IgG (Fc portion)
2B4 (CD244)	CD48
NKG2C/E/H	HLA-E (for NKG2C)
CD96	CD155
Inhibitory receptors
Inhibitory KIR	MHC class I
NKR-P1A	LLT1
NKG2A/B	HLA-E
TIGIT	PVR (Necl5, CD155), nectin 2 (CD112)
LIR-1 (ILT-2/CD85j/LILRB1)	HLA (a3), HCMV UL18
IRp60 (CD300a)	PS, PE
CEACAM1 (CD66)	CEACAM1 (CD66), TIM-3 (HAVCR2)

KIR, killer-cell immunoglobulin-like receptor; HLA, human leucocyte antigen; MIC, major histocompatibility complex (MHC) class I chain-related protein; ULBP, UL16-binding proteins; DNAM-1, DNAX accessory molecule 1; PVR, poliovirus receptor; Necl, nectinlike molecules; CRTAM, class I-restricted T cell-associated molecule; AICL, activation-induced C-type lectin; KACL, keratinocyte-associated C-type lectin; HA, haemagglutinin; HN, haemagglutinin-neuraminidase; CFP, complement factor P; PCNA, proliferating cell nuclear antigen; MLL5, mixed lineage leukaemia 5; PDGF, plateletderived growth factor; IRp60, inhibitory receptor protein 60; BAT3, HLA-B-associated transcript 3; HCMV, human cytomegalovirus; IgG, immunoglobulin G; LLT1, lectin-like transcript 1; TIGIT, T cell immunoglobulin and immunoreceptor tyrosine-based activation motif (ITIM) domain; LIR-1, leucocyte immunoglobulin-like receptor 1; PS, phosphatidylserine; PE, phosphatidylethanolamine; CEACAM, carcinoembryonic antigen-related cell adhesion molecules; TIM-3, T-cell immunoglobulin domain and mucin domain 3; HAVCR2, hepatitis A virus cellular receptor 2.

Fig 1.

Regulation of natural killer (NK) cell response according to the ‘missing self’ and ‘altered self’ models. (A) The presence of major histocompatibility complex (MHC)-I, which are ligands for inhibitory NK-cell receptors, and the lack of stress-induced ligands on the surface of healthy cells, lead to a major inhibitory signal of NK cells. (B) The presence of stress-induced ligands for the activating NK-cell receptors and the downregulation of MHC-I by tumour cells lead to a major activating signal of NK cells.

Other than their known cytotoxicity, NK cells are a major source of cytokines and chemokines, such as type 1 cytokines, INF-γ, TNF-α, granulocyte-macrophage colony-stimulating factor (GM-CSF) and others. Due to their capacity to secrete INF-γ, NK cells are classified as prototypical members of group 1 innate lymphoid cells (ILCs).^21,36

Sources of NK cells for immunotherapy

Considering the critical role of NK cells in defence against tumour growth and metastasis, NK cell-based immunotherapies, including the adoptive transfer of NK cells, have been widely explored. NK cells for adoptive immunotherapy can be generated from a myriad of autologous or allogeneic sources, including PB,³⁷ BM-derived haematopoietic progenitor cells,³⁸ iPSCs,³⁹ human embryonic stem cells (hESCs),⁴⁰ umbilical CB,⁴¹ NK cell lines (such as NK-92)⁴² or memory-like NK cells,⁴³ each with their advantages and disadvantages. While after activation with cytokines such as IL-2, IL-12, IL-15, IL-18 and type-I IFNs, autologous NK cells are potentially highly active in vitro, their clinical efficacy appears to be limited.^44,45 This limited effect could be due to the inhibition of autologous NK cells by self-human leucocyte antigen (HLA) molecules expressed on tumour cells. Moreover, endogenous NK cells might not retain sufficient cytotoxicity against cancer cells due to the overall immunosuppressive tumour milieu.⁴⁶

Allogeneic NK cells from healthy adults can be obtained from PB through apheresis, or from the BM via harvesting. However, this requires a donor willing to undergo this procedure. NK cell lines, including NK-92, NKL, KHYG-1, YT and NKG offer an attractive and potentially unlimited source of cells for immunotherapy. Their robust in vitro anti-tumour activity is well-established;^42,47 however, these cell lines are derived from patients with NK lymphoma, raising potential safety concerns related to their engraftment following infusion.⁴⁸ To prevent engraftment, NK cell lines are irradiated prior to infusion. While the safety of irradiated NK cell lines has been reported in multiple Phase I clinical trials in patients with cancer,^49–51 the short duration of persistence of the irradiated cells may ultimately limit their clinical efficacy. Indeed, in a Phase I clinical trial with irradiated NK-92 cell lines in patients with relapsed/refractory acute myeloid leukaemia (AML), there was no significant increase in the number or cytotoxicity of NK cells in the circulation after NK-cell adoptive transfer. Instead, a significant drop in the cytotoxicity of PB NK cells was observed at 4 h after infusion, pointing to the potentially negative impact of irradiation on the function of the infused product.⁵⁰

Umbilical CB is another attractive source of NK cells with a number of advantages, such as off-the-shelf availability. In addition, CB has fewer T cells than PB and the cells have a naïve phenotype, thus reducing the risk of GvHD in the event that contaminating T cells are transferred to the patient in the infused NK-cell product.¹² NK cells constitute up to 30% of lymphocytes in CB,⁵² with a higher proportion of CD56^bright cells.^52–54 In addition, our group has shown that compared to PB NK cells, CB NK cells have higher expression of genes related to cell cycle and proliferation, suggesting that they may have a greater potential to proliferate and persist in vivo.⁵⁵ However, the small volume of blood in a CB unit has historically made the application of CB-derived NK cells for therapy challenging. To overcome these obstacles, different platforms to expand CB NK cells have been developed, including the use of engineered K562-based feeder that express membrane-bound IL-21 and 4-1BB ligand.⁴¹ Following ex vivo expansion, CB NK cells were shown to display a full array of activating and inhibitory receptors and killed tumour targets to the same extent as PB-derived mature NK cells.⁴¹ Another group used CB as a source of NK cells by isolating CD34⁺ human stem and progenitor cells (HSPCs) from umbilical CB, and expanding and differentiating them into NK cells in the presence of growth factors and cytokines such as stem cell factor (SCF), IL-7, IL-15 and IL-2. Using this strategy, they could generate clinically relevant and scalable doses of NK cells for therapeutic application.⁵⁶ Other groups have investigated the use of iPSC-derived NK cells from BM or umbilical CB-derived CD34⁺ stem cells cultured in the presence of cytokines and various growth factors.^38,57 Hermanson et al.⁵⁸ showed that iPSC-derived NK cells can be produced from a standardised cell population to manufacture a homogeneous NK-cell population at clinical scale. This NK-cell source is currently an active area of research by multiple groups,^59–62 including Fate Therapeutics, Inc. (San Diego, CA, USA).

Even though investigators continue to explore the use of NK cells, the jury is still out as to what is the best source for NK-cell immunotherapy. However, it is probable that as we increase our understanding of the rules governing NK-cell effector function, we will be able to develop more effective NK cells for the treatment of cancer.

Clinical experience with adoptive transfer of non-engineered NK cells

The infusion of NK cells derived from different sources has been explored in several settings. In an early trial of adoptive transfer with allogeneic NK cells, 43 patients with poor prognosis AML, metastatic melanoma or renal cell carcinoma (RCC) received lymphodepleting agents followed by infusion of NK cells from healthy haploidentical donors and daily IL-2 infusions for 14 days. Complete remissions were reported in five of 19 patients with poor-risk AML. Among patients with melanoma and RCC, six had stable disease; however, most ultimately progressed.³⁷

NK cell lines are another source of NK cells being investigated in the clinic, with the NK-92 cell line being the only FDA-approved NK cell line for clinical use. Several Phase I clinical trials investigated the use of NK-92 cells in advanced treatment-resistant malignancies. Clinical trials run by groups in Chicago and Frankfurt enrolled patients with RCC, lung cancer or other solid tumours.^49,51 While the approach was found to be safe, clinical responses were at best modest. Among 12 patients with melanoma and RCC treated by the group at Rush University, four patients had stable disease, two had a mixed response and six patients had progressive disease at the designated 4-week monitoring period.⁴⁹ After a 4-year follow-up period, 10 patients had died from progressive disease and one of post-haematopoietic stem cell transplant (HSCT) complications.⁴⁹ In the study from Germany, three of 15 treated patients with advanced lung cancer achieved a clinical response (one stable disease and two mixed response) 28 days after NK-92 cell infusion.⁵¹ Another clinical trial led by the group at Princess Margaret in Toronto investigated the safety and efficacy of escalating doses of irradiated NK-92 cells in 12 patients with refractory lymphoma or multiple myeloma who relapsed after autologous HSCT.⁶³ The treatment was well tolerated except for occasional minor infusion-related toxicities, including fevers, chills, nausea and fatigue. One patient with Hodgkin lymphoma and another with multiple myeloma achieved a complete response (CR), two patients had minor responses and one patient had some clinical improvement.⁶³ Of the patients who achieved a CR, one patient with myeloma received concomitant therapy with lenalidomide and dexamethasone during and after the NK-92 cell infusion.⁶³ While these results are encouraging, more research is needed to understand the clinical value of immunotherapy with NK-92 cells, especially in view of concerns related to the need for irradiation and its potential negative impact on the persistence and clinical efficacy of the cells following infusion.

Another approach uses in vitro differentiated NK cells from CB HSCs. In a Phase I clinical trial, 10 elderly patients with AML and myelodysplastic syndrome in morphological remission who were not eligible for HSCT received partially HLA-matched CB HSC-derived NK cells following fludarabine and cyclophosphamide conditioning.⁶⁴ Of the 10 treated patients, four remained disease-free at the time of reporting (60, 52, 22 and 16 months after NK-cell infusion). While this study established the safety of NK-cell transfer, definitive conclusions on the long-term efficacy of this approach cannot be drawn, as the patients were already in remission at the time of NK-cell therapy.

The use of directly isolated and ex vivo expanded NK cells from CB has also been clinically explored. In a Phase I trial, partially HLA-matched CB-derived NK cells were adoptively transferred to patients with multiple myeloma undergoing high-dose chemotherapy and autologous HSCT. In all, 10 of 12 treated patients achieved a very good partial response or better responses, including eight near CRs and six had minimal residual disease (MRD) negativity by flow cytometry. The infusions were well tolerated with no dose-limiting toxicities. Notably, the HLA-mismatched NK cells were detectable for up to 26 days after infusion.⁶⁵ While this study established the safety of combining allogeneic, partially mismatched CB NK cells with autologous HSCT, it is difficult to draw firm conclusions on the clinical efficacy of the CB NK cells in this setting due to the small number of patients, the combination with autologous HSCT and the subsequent administration of maintenance therapy.

Another interesting approach is the application of NK cells with memory-like properties. Although NK cells have traditionally been considered members of the innate immune system, several important studies have reported that NK cells can acquire memory-like properties after exposure to pathogens or activation by cytokines.^66,67 In 2009, the Yokoyama group first reported that murine splenic NK cells activated with IL-12, IL-15 and IL-18 in vitro exhibit an enhanced capacity to produce IFN-γ upon restimulation.⁶⁸ Preclinical studies confirmed that cytokine-preactivated NK cells maintained enhanced anti-tumour functions in vivo.⁶⁹ The memory-like responses induced by IL-12, IL-15 and IL-18 have also been described in human NK cells, with enhanced IFN-γ production upon restimulation with cytokines or target cells.⁴³ In the context of a first-in-human clinical trial, allogeneic cytokine-induced human memory-like NK cells proliferated and expanded in patients with AML (NCT01898793) with clinical responses observed in five of nine evaluable patients, including four complete remissions.¹³

In summary, clinical studies using non-modified NK cells have confirmed the safety of this approach; however, final conclusions regarding efficacy will require longer follow-ups and larger study populations. These challenges, discussed in detail in the subsequent section, have fuelled the development of novel strategies using genetic engineering to enhance the activity of NK cells for therapy.

Limitations of NK-cell immunotherapy

A major consideration for the successful clinical application of NK-cell immunotherapy is their lack of persistence after adoptive transfer in the absence of cytokine support.⁷⁰ While this could be helpful in reducing long-term adverse effects and toxicity, it may also significantly reduce their clinical efficacy. Indeed, a number of studies have shown that the in vivo persistence and proliferation of NK cells after adoptive transfer may be predictive of clinical response.^37,71 Thus, enhancing NK-cell persistence is the subject of active research by many groups. A number of strategies have been explored, including the administration of exogenous cytokines such as IL-2 or IL-15.⁷² This approach, while enhancing the in vivo expansion and proliferation of NK cells, is associated with systemic toxicity, including capillary leak syndrome following IL-2 infusion⁷³ and neutropenia after IL-15 infusion.⁷⁴ In addition, IL-2 induces the expansion of regulatory T cells (Tregs), which can in turn suppress the expansion, persistence and anti-tumour activity of NK cells.⁷⁵ To mitigate this effect, Bachanova et al.⁷⁶ administered the IL-2-diphteria fusion protein (IL2DT) prior to NK-cell infusion to deplete Tregs and reported superior expansion of the donor NK cells associated with improved complete remission and disease-free survival rates in patients with refractory AML. Another strategy that could help with the persistence of NK cells after adoptive infusion involves the administration of lymphodepleting agents such as cyclophosphamide and fludarabine prior to NK-cell transfer. Conditioning therapy with lymphodepleting agents improves the success of cell therapy through a number of mechanisms,⁷⁷ including eliminating the recipient lymphocytes to allow for the transferred cells to expand and proliferate,⁷⁸ removing homeostatic cytokine sinks and increasing the availability of cytokines such as IL-2 and IL-15 to support the proliferation of the infused cells,⁷⁹ eliminating immunosuppressive elements such as Tregs and myeloid-derived suppressor cells (MDSCs), inducing expression of co-stimulatory molecules and downregulating immunosuppressive molecules such as indoleamine 2,3-dioxygenase (IDO) in tumour cells.⁸⁰ However, even this approach does not fully protect against early loss of response,³⁷ which could be due to a myriad of factors, including host factors imposing a complex network of competition between the donor NK cells and the recipient tumour-related immunosuppression.⁸¹

The tumour microenvironment (TME) induces dysfunction and exhaustion in NK cells⁸² through a number of mechanisms, including: (i) suppressive immune and non-immune cells such as Tregs and MDSCs,⁸³ (ii) suppressive cytokines such as tumour growth factor-β (TGF-β),^84,85 (iii) overexpression of ligands for inhibitory NK-cell receptors by tumour cells, such as HLA-E (ligand for the inhibitory receptor NKG2A),⁸⁶ CD200 (ligand for the inhibitory receptor CD200R),⁸⁷ and galactin-9 (ligand for the inhibitory receptor Tim-3)⁸⁸ and (iv) downregulation of ligands for activating receptors such as CD48 (ligand for the activating receptor 2B4), by oncogenic proteins.⁸⁹

Genetic engineering of NK cells to improve their function for adoptive cell transfer

The past decade has witnessed tremendous progress in the field of genetic engineering and gene editing. These tools have been applied to optimise the therapeutic potential of NK cells, including improving their in vivo proliferation and persistence through expression of cytokines, increasing their cytotoxicity by skewing the receptor profile towards a predominantly activated phenotype, as well as targeting the immunosuppressive TME (Fig 2).

Fig 2.

Natural killer (NK) cell engineering strategies to make the next-generation chimeric antigen receptors (CARs). IL, interleukin; TME, tumour microenvironment; MDSCs, myeloid-derived suppressor cells; DN, dominant negative; TGFβr, tumour growth factor-β receptor.

Enhancing expansion and persistence

Strategies to enhance the in vivo persistence of NK cells include the genetic engineering of cells to autonomously secrete a cytokine such as IL-2 or IL-15.^71,90,91 Retroviral transduction of NK-92 cells with IL-2 enhanced their anti-tumour activity both in vitro and in vivo compared to NK-92 cells stimulated with exogenous IL-2.⁹¹ Similarly, retroviral transduction of NK cells with membrane-bound (mb)IL-15 improved NK-cell proliferation and persistence.⁹⁰ Our group has successfully engineered CB-derived NK cells with a retroviral vector to secrete IL-15 to enhance their in vivo proliferation and persistence and to express a CD19-directed CAR to improve their in vivo anti-tumour activity.⁷¹ Another interesting approach is the genetic engineering of memory-like NK cells to express a CAR, recently shown to result in potent in vivo activity in a mouse model of lymphoma.⁹²

Targeting checkpoints

Immune checkpoint-based therapies, which target the regulatory pathways of immunocompetent cells, have resulted in major clinical advances and long-term remissions and possible cures in some patients with cancer.² Most of this progress has been achieved with T cells, but there is compelling evidence to support that targeting checkpoints in NK cells may be just as effective. NKG2A is an inhibitory receptor that recognises a non-classical MHC-I, HLA-E. Monalizumab, an anti-NKG2A monoclonal antibody, was investigated in combination with cetuximab in a Phase II trial in 26 patients with refractory squamous cell carcinoma of the head and neck.⁹³ The results were promising with eight patients achieving a partial response (31%), 14 patients (54%) having stable disease and only four patients (15%) showing disease progression.⁹³ Silencing NKG2A in NK cells using short hairpin RNA (shRNA) or short interfering RNA (siRNA) was shown to suppress inhibitory signalling and subsequently enhance the cytotoxicity of NK cells against HLA-E expressing cancer cells in vitro and in vivo.⁹⁴ More recently, Kamiya et al.⁹⁵ devised a way to generate NK cells lacking NKG2A using a retroviral vector encoding a single-chain variable fragment (scFv) derived from an anti-NKG2A antibody linked to endoplasmic reticulum-retention domains. The adoptively transfused NKG2A^null NK cells demonstrated an enhanced cytotoxicity against HLA-E-expressing tumours. TIGIT (T cell immunoglobulin and immunoreceptor tyrosine-based inhibitory motif domain) is another NK-cell checkpoint receptor that binds to CD155 [also called poliovirus receptor (PVR)] on the surface of tumour cells.⁹⁶ TIGIT blockade through genetic knockout or the use of monoclonal antibodies has been shown to unleash NK-cell anti-tumour activity, leading to better tumour control in various preclinical mouse models.⁹⁷ The suppressor of cytokine signalling (SOCS) family of proteins plays an important role in NK cell biology by attenuating cytokine signalling and effector function against cancer. One of its members, cytokine inducible SH2 containing protein (CIS), encoded by the CISH gene, is as an important checkpoint molecule in NK cells and is upregulated in response to IL-15.⁹⁸ CISH-knockout in iPSC-derived NK cells resulted in increased IL-15-mediated Janus kinase-signal transducers and activators of transcription (JAK-STAT) signalling and superior anti-tumour activity in an AML xenograft mouse model.⁹⁹ Our group has developed a novel approach that combines CAR19/IL-15 transduction and CISH deletion that resulted in enhanced anti-tumour activity in a lymphoma mouse model. Mechanistically, we showed that the NK-cell anti-tumour activity was associated with enhanced Akt/mTORC1 and c-MYC signalling and increased aerobic glycolysis, thus, promoting their metabolic fitness and endowing them with superior anti-tumour activity.¹⁰⁰

Targeting the TME

Targeting the TME involves strategies directed at its complex immunosuppressive network, including immune cells, stromal cells, cytokines, metabolites and regulatory pathways. TGF-β is an important immunosuppressive cytokine in the TME. Engineering NK cells to express a high-affinity non-signal transducing receptor derived from TGF-βR2 (dominant-negative receptor) was shown to neutralise the negative effects of TGF-β on NK cells and restore their cytotoxicity.¹⁰¹ Another strategy targets the microRNA (miRNA) miR-27a-5p, which is upregulated by TGF-β in NK cells.¹⁰² Nucleofection of an inhibitor of miR-27a-5p increased NK-cell effector function both in vitro and in vivo.¹⁰² Our group used Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-Cas9 technology to knockout the TGF-β receptor 2 gene (TGFβR2) in human NK cells, which rendered NK cells insensitive to TGF-β and maintained their effector activity against AML.¹⁰³ Adenosine is another immunosuppressive metabolite of the TME that is generated from ATP in response to hypoxia and stress by the ectonucleotidases CD73 and CD39.¹⁰⁴ Blocking the high-affinity A2A adenosine receptor on NK cells enhanced their proliferation and maturation as well as their anti-tumour activity in mouse models of melanoma, breast cancer and fibrosarcoma.^105,106 The TME has an abundance of other immunosuppressive factors such as IL-10, arginase-1, nitric oxide, prostaglandin E₂ (PGE2), IDO and reactive oxygen species that could be targeted by genetic engineering or with combination therapies, such as expressing anti-oxidants in adoptively transferred NK cells.

MDSCs are immune modulators that suppress anti-tumour immune responses¹⁰⁷ and pose a challenge to the success of immunotherapy.¹⁰⁸ Treatment with ipilimumab [anti-cytotoxic T-lymphocyte antigen 4 (CTLA4)] has been shown to decrease the frequency of MDSCs in patients with melanoma.¹⁰⁹ Therefore, approaches that aim to combine NK-cell immunotherapy with MDSC depletion may further improve responses. Tregs are also attractive therapeutic targets as they suppress NK-cell activity by different mechanisms, including by TGF-β production⁸³ and by limiting IL-2 availability to NK cells.¹¹⁰ Chemotherapies such as fluorouracil (5-FU), cyclophosphamide, doxorubicin and sunitinib have been shown to target MDSCs and Tregs,¹¹¹ and in combination with NK-cell immunotherapy, may result in superior outcomes. It is important to stress that these chemotherapeutic agents retain their broad and immunosuppressive effects, which call for the use of novel technologies to specifically and potently target and eliminate MDSCs and Tregs.

In brief, reprogramming NK cells to circumvent immune evasion mechanisms imposed by the TME is a promising approach to improve the efficacy of NK-cell therapy.

Redirecting the specificity of NK cells using CAR engineering

The advent of CAR technology allowed for major enhancements in the cytotoxic activity of immune effector cells for adoptive therapy. CARs are genetically engineered transmembrane receptors that have two main functions: (i) specific antigen recognition independent of MHC restriction and (ii) activation of the modified cells through signal transduction.¹¹² CAR T-cell therapies have shown remarkable responses in lymphoid leukaemias and lymphomas and in myeloma,^4,7,8 but their use is associated with certain limitations, including cost as well as serious toxicities such as CRS and neurotoxicity. NK cells are attractive candidates for CAR engineering because of their short lifespan, lasting nearly 2 weeks in vivo, and their different cytokine profile that may reduce the risk of toxicity after adoptive transfer.¹¹³ Other advantages of NK cells are related to their intrinsic capacity to recognise and target tumour cells through their own receptors,¹¹⁴ which allows them to kill cancer cells even when antigen escape mechanisms such as CAR target antigen downregulation evolve.¹¹⁵ Importantly, both preclinical and clinical data indicate that the risk of GvHD with allogeneic CAR-NK cells should be minimal.^37,116–118 These unique characteristics, including the potential for an off-the-shelf and readily available product that could be produced at lower cost, make NK cells attractive contenders for CAR engineering.

CAR structure.

CAR constructs used to engineer T cells have four essential components: a scFv that binds a specific antigen, an extracellular spacer domain, a transmembrane domain and an intracellular signalling domain derived from the CD3ζ immunoreceptor tyrosine-based activation motifs (ITAMs).¹¹⁹ Progress in the field of genetic engineering have driven the development of next-generation CARs with stronger target binding affinity and more effective anti-tumour activity. While first-generation CARs only carry one intracellular signalling region,¹²⁰ second-generation CARs contain one signalling and one co-stimulatory domain such as CD28 or 4-1BB (CD137),^121,122 and third-generation CARs contain a signalling domain and multiple co-stimulatory domains.^123,124 Fourth-generation CARs incorporate innovative approaches that aim to mitigate challenges governing the use of CARs, including poor proliferation, limited persistence, suboptimal cytokine secretion and target antigen loss.¹²⁵ Some of these innovative approaches include the use of gene knockout or knock-in,^126,127 overexpression of cytokines,¹²⁵ simultaneous expression of multiple CARs^128,129 or introduction of suicide genes¹³⁰ (Fig 2).

Role of signalling and costimulatory domains.

CD3ζ and FcR-γ domains generate the initial signals required for cytotoxicity in both T and NK cells.^131–133 However, these signals alone result in relatively weak cytokine production and low anti-tumour efficacy, as also reported with first-generation CARs that express a single intracellular signalling domain.^134,135 The addition of co-stimulatory domains to the primary CAR platform forms the hallmark of second- and third-generation CARs, leading to superior effector function^136–138 and persistence of cells following adoptive transfer.¹³⁹ Several co-stimulatory domains have been studied, including members of the immunoglobulin superfamily, e.g. CD28 and inducible T-cell co-stimulator (ICOS), members of the TNF receptor superfamily, including 4-1BB, CD27, OX40 and CD40, and others such as CD40L and TLRs.¹⁴⁰ Co-stimulatory domains that are important in NK-cell signalling such as the 2B4 co-stimulatory domain,⁵⁹ DAP10, which naturally associates with NKG2D,¹⁴¹ or DAP12, that is involved in signal transduction of the activating NK-cell receptors NKG2C and NKp44, have also been investigated. NK cells transduced with either anti-CD19-DAP10 or anti-CD19-CD3ζ could both trigger NK-cell cytotoxicity,¹⁴² but the best response was seen when the CAR construct contained both DAP10 and CD3ζ signalling domains.¹⁴³ Similarly, incorporation of DAP12 with an scFv against prostate stem cell antigen (PSCA) resulted in an improved cytotoxicity and higher INF-γ secretion in primary NK cells compared to anti-PSCA CAR-NK cells expressing CD3ζ only.¹⁴⁴ It is not clear why a DAP12-based CAR with only one ITAM might be at least as efficient or even superior in downstream signalling than a CD3ζ-based CAR with three ITAMs.¹⁴⁵ It is possible that DAP12-mediated ITAM phosphorylation may directly form a docking site for downstream signalling, while CD3ζ-mediated ITAM phosphorylation might influence other docking sites that could in turn modulate the effector mechanisms of the NK-cell response.

Preclinical data with CAR NK-cell therapy

Preclinical studies with CAR-NK cells initially investigated the safety and efficacy of anti-CD19 and anti-CD20 CAR-NK cells against B-cell malignancies, although results with first-generation CARs were modest.^142,146 The addition of a costimulatory domain to the CAR construct resulted in significant improvement in the efficacy of CAR-NK cells. Chu et al.¹⁴⁷ genetically modified PB NK cells from healthy donors to express anti-CD20.4-1BB.CD3ζ CAR using mRNA nucleofection. This approach resulted in 67% CAR expression and significant in vitro and in vivo activity against Burkitt lymphoma in preclinical models. Our group has developed an approach for the generation of ‘armoured’ anti-CD19 CAR-NK cells that aims to improve their persistence and cytotoxicity, while also ensuring safety.⁷¹ Using retroviral transduction, we genetically engineered CB-derived NK cells to express a vector that encodes the genes for (i) anti-CD19 CAR, (ii) IL-15 to improve NK-cell proliferation and persistence,¹⁴⁸ and (iii) inducible caspase-9 (iC9), a safety switch that can be pharmacologically activated to eliminate transduced cells in the event of toxicity.^149,150 We showed that iC9.CAR19.CD28-ζ-2A-IL-15 transduced NK cells exert significant anti-tumour activity in xenograft mouse models of Raji tumour⁷¹ and in patients with relapsed or refractory lymphoid malignancies.¹⁵¹

In T-lymphoid malignancies, CAR-NK cells may be advantageous over CAR-T cells, as the shared expression of targetable antigens¹⁵² on both malignant and normal T lymphocytes (e.g. CD5) will result in the fratricidal killing of the CAR-expressing T cells during manufacturing. Chen et al.¹⁵³ reported that third-generation anti-CD5 CAR-transduced NK-92 cells that incorporate both CD28 and 4-1BB co-stimulatory domains mediate strong in vitro cytotoxicity against primary patient-derived T-cell acute lymphoblastic leukaemia (T-ALL), T-cell lymphoma and Sézary cells and improve the survival of immunodeficient mice engrafted with Jurkat lymphoma. Similarly, Xu et al.¹⁵⁴ compared the anti-tumour activity of NK-92 cells transduced with a lentiviral vector expressing anti-CD5 CAR-CD3ζ with a 4-1BB or 2B4 co-stimulatory domain. While NK-92 cells transduced with either CAR construct-mediated effective cytotoxicity against CD5⁺ malignant cells, the anti-tumour activity of the 2B4-incorporating CAR-NK cells was superior both in vitro and in a T-ALL xenograft mouse model. Recently, Tang et al.¹⁵⁵ developed a novel CAR targeting CD7 using a nanobody. Nanobodies are antibody fragments consisting of a single monomeric variable domain derived from the heavy-chain of the camelidae antibody with a number of advantages for CAR construction, including small molecular weight, high specificity and weak immunogenicity.¹⁵⁶ They tested two anti-CD7 CARs, one carrying a monovalent and the other carrying a bivalent nanobody sequence (dCD7 CAR).¹⁵⁶ The CARs were then introduced into IL-2 expressing NK-92 cells (NK-92MI).^156,157

Both CD7-CAR-NK-92MI and dCD7-CAR-NK-92MI cells exerted potent in vitro and in vivo anti-tumour activity against T-leukaemia cell lines and primary tumours, with a slightly stronger cytotoxicity observed for the bivalent CAR.¹⁵⁶

The application of CAR therapy in AML remains challenging. Although none of the ongoing clinical trials of CAR T-cell therapy in AML have yet yielded mature data, the clinical responses reported to date have been at best modest.¹⁵⁸ Major barriers to the success of CAR therapy in AML include shared expression of antigen on AML cells and normal HSCs (such as CD123), thus increasing the risk of marrow aplasia, and heterogeneous expression on progenitors or absence of target antigen on blasts (e.g. CD33), resulting in leukaemia escape.¹⁵⁹ AML cells are highly susceptible to NK cell-mediated killing as they express many of the ligands recognised by NK-cell activating receptors, such as MHC chain-related antigens (MICA/B) and the UL16-binding proteins (ULBPs) recognised by NKG2D, and CD155 (recognised by DNAM).¹⁶⁰ Thus, CAR-NK cells may overcome some of the challenges related to antigen escape and tumour heterogeneity through their innate ability to recognise and target AML cells. In addition, their short lifespan may limit the extent and length of cytopenias when targeting antigens that are also expressed on normal HSCs. Thus, efforts to develop CAR-NK cells against AML are underway. In a recent study, primary human NK cells transduced with a self-inactivating alpha-retroviral vector incorporating anti-CD123, and dual co-stimulation with CD28 and 4-1BB¹⁶¹ were reported to efficiently kill AML cell lines as well as primary AML blasts in vitro.¹⁶¹

Myeloma cells are also highly susceptible to NK cell-mediated killing as they express multiple ligands for NK receptors.¹⁶² Therefore, there is a great interest in developing CAR-NK cells targeting myeloma antigens such as CS1, CD138 and B-cell maturation antigen (BCMA). Chu et al.¹⁶³ reported that NK-92 and NKL cell lines transduced with a second-generation lentiviral vector encoding anti-CS1 CAR with CD3ζ signalling and a CD28 co-stimulatory domain exert potent in vitro and in vivo anti-tumour activity against CS1⁺ cell lines and primary patient-derived myeloma cells. Similarly, Jiang et al.¹⁶⁴ confirmed the anti-myeloma activity of NK-92MI cells lentivirally transduced with an anti-CD138 CAR-CD3ζ construct in a xenograft mouse model of CD138-positive myeloma cell line (RPMI8226).

Despite the progress made in treating haematological malignancies, a number of challenges limit the application of CAR-cell therapy for the treatment of solid tumours.¹⁶⁵ These include lack of specific targetable antigens,¹⁶⁶ limitations imposed by the physical and functional barriers that prevent effective trafficking and penetration of CAR cells to tumour sites,^167,168 and the hostile TME characterised by hypoxia, acidic pH, nutritional depletion and immunosuppression.^169,170

CAR-NK cells have also been used to target solid tumour antigens. Human epidermal growth factor receptor 2 (HER-2)-specific CAR-NK cells (HER-2, C6.5-scFv-Fc-CD3ζ-CD28) were shown to mediate effective in vitro and in vivo activity in a mouse model of ovarian cancer,¹⁷¹ and epidermal growth factor receptor (EGFR)-redirected CAR-NK cells could effectively target glioblastoma when administered intracranially in a xenograft mouse model.¹⁷² However, CAR-NK cells are also susceptible to the evasion tactics developed by tumours. Kailayangiri et al.¹⁷³ reported that NK cells expressing a second- or third-generation ganglioside antigen (G_D2) CAR construct (G_D2-t2B4ζ, G_D2-BBζ or G_D2-t2B4.BBζ) could kill several G_D2⁺ cancer cell lines in vitro, but failed to control the tumour in a xenograft mouse model. The suboptimal in vivo efficacy was related to the upregulation of HLA-G [the ligand for the inhibitory killer-cell immunoglobulin-like receptor (KIR), KIR2DL4] on Ewing sarcoma cells, resulting in NK-cell inhibition and immune escape.^173,174 To strengthen the antigen-specific NK-cell signalling, Li et al.⁵⁹ devised an anti-mesothelin CAR containing the transmembrane domain of the activating NK-cell receptor NKG2D, the 2B4 co-stimulatory domain, and the CD3z signalling domain and showed that iPSCs-derived NK cells expressing this CAR mediate strong anti-tumour effects in an ovarian cancer xenograft model.

A novel approach to overcome the multiple challenges presented by solid tumours is to combine CAR cells with oncolytic viruses.¹⁷⁵ Oncolytic viruses synergise with CAR cells through a number of mechanisms, including direct lysis of both CAR antigen positive and negative cancer cells, as well as the release of inflammatory cytokines that can change an immunologically ‘cold’ tumour into a ‘hot’ tumour (with influx of immune cells and cytokines). Chen et al.¹⁷⁶ reported that the addition of an oncolytic virus to EGFR-transduced NK-92 cells could further augment their anti-tumour response in a breast cancer brain metastasis mouse model.

In summary, although the preclinical data with CAR-NK cells in the setting of solid tumour hold promise, their successful translation to the clinic requires an appreciation of the potential blocking interactions between the tumour cells and NK cells, and the development of strategies to overcome them.

Clinical experience with CAR-NK cells

The past year has witnessed a rapid surge in the number of clinical trials using CAR-NK cells to target haematological malignancies and solid cancers. Our group is currently leading a first-in-human Phase I/II study to test the safety and efficacy of CAR.CD19-CD28-zeta-2A-iC9-IL-15-transduced HLA-mismatched CB NK cells in patients with relapsed/refractory CD19⁺ B-lymphoid malignancies (NCT03056339). Our preliminary data are highly encouraging with responses observed in eight of 11 patients (seven CRs) and with no evidence of neurotoxicity, CRS or GvHD.¹⁵¹

CAR-NK cells are also being explored in the setting of other cancers. A group reported the results of a Phase I study in three patients with relapsed or refractory AML treated with anti-CD33 CAR NK-92 cells transduced with a third-generation CAR incorporating both CD28 and 4-1BB.¹⁷⁷ The study established the safety of escalating doses of CAR-transduced NK-92 cell infusions; however, no durable responses were achieved. The authors questioned the impact of irradiation on the persistence and cytotoxicity of the infused product and concluded that future modifications of the construct, as well as a better trial design to select patients with high CD33-expressing leukaemia cells, may improve the efficacy of their approach.¹⁷⁷

Additional clinical studies with CAR NK cells targeting CD19 in B-cell malignancies (NCT02892695), CD7 in leukaemia and lymphoma (NCT02742727), BCMA in multiple myeloma (NCT03940833), CD33 in AML (NCT02944162), mucin 1 (MUC1) in colorectal cancer, gastric cancer, pancreatic cancer, breast cancer, non-small cell lung cancer and glioma (NCT02839954), prostate-specific membrane antigen (PSMA) in prostate cancer (NCT03692663), HER-2 in glioblastoma (NCT03383978), mesothelin in ovarian cancer (NCT03692637) are underway at multiple centres (Table II).

Table II.

Clinical trials involving chimeric antigen receptor natural killer (CAR NK)-cell therapy registered in clinicaltrials.gov.

Clinical trial identifier	Start date	Phase	NK-cell source	Construct	Disease	Target	Location	Recruitment status
Multiple myeloma
NCT03940833	May 2019	I/II	NK-92 cell line	N/A	Multiple myeloma	BCMA	China	Recruiting
B-lymphoid malignancies
NCT03579927	October 2019	I/II	CB	CAR.CD19-CD28-zeta-2A-iCasp9-IL-15	B-cell lymphoma	CD19	MDACC, USA	Not yet recruiting
NCT03692767	March 2019	Early Phase I	N/A	N/A	B-cell lymphoma	CD22	China	Not yet recruiting
NCT03690310	March 2019	Early Phase I	N/A	N/A	B-cell lymphoma	CD 19	China	Not yet recruiting
NCT03824964	February 2019	Early Phase I	N/A	N/A	B-cell lymphoma	CD19/CD22	China	Not yet recruiting
NCT03056339	June 2017	I/II	CB	CAR.CD 19-CD28-zeta-2A-iCasp9-IL-15	B-cell lymphoma	CD19	MDACC, USA	Recruiting
NCT02892695	September 2016	I/II	NK-92	CAR.CD 19-TCRζ-CD28–4-IBB	Leukaemia and	CD19	China	Recruiting
NCT01974479	September 2013	Phase I	Haploidentical	CAR.CD 19-BB-zeta	lymphoma BALL	CD19	Singapore	Suspended for an
NCT00995137	October 2009	I	donor Donor	CAR.CD 19-BB-zeta	BALL	CD19	St. Jude	interim review Completed
AML
NCT02944162	October 2016	I/II	NK-92 cell line	CAR.CD33-CD28-CD137-CD3ζ	AML	CD33	Children’s Research Hospital, USA China	Unknown
T-cell leukaemia and lymphoma/AML
NCT02742727	March 2016	I/II	NK-92 cell line	CAR.CD7-TCRζ-CD28-4-1BB	CD7 positive leukaemia	CD7	China	Unknown
Solid tumours
NCT03940820	May 2019	I/II	N/A	N/A	and lymphoma Solid tumours	ROBOl	China	Recruiting
NCT03692637	March 2019	Early Phase I	N/A	N/A	Epithelial ovarian cancer	Mesothelin	China	Not yet recruiting
NCT03692663	December 2018	Early Phase I	N/A	N/A	Prostate cancer	PSMA	China	Not yet recruiting
NCT03415100	January 2018	I	N/A	N/A	Solid tumours	NKG2D	China	Recruiting
NCT03383978	December 2017	I	NK-92	5.28.Z	Glioblastoma	HER-2	Germany	Recruiting
NCT02839954	July 2016	I/II	NK-92	N/A	Solid tumours	MUC1	China	Unknown

ALL, acute lymphoblastic leukaemia; AML, acute myeloid leukaemia; BCMA, B-cell maturation antigen; CB, cord blood; HER-2, human epidermal growth factor receptor 2; MDACC, MD Anderson Cancer Center; MUC1, mucin 1; N/A, not available; NK, natural killer; NKG2D, NK group 2 member D; PSMA, prostate-specific membrane antigen; ROBO1, Roundabout Guidance Receptor 1.

Future and perspective

NK cells are powerful innate immune cells with unique characteristics that make them ideal candidates for cancer immunotherapy. They are readily available, can be derived from multiple sources and expanded to clinically relevant numbers, and their safety profile has been shown in multiple clinical studies in the past 15 years. Moreover, their ability to recognise and kill tumour cells through their germ-line encoded receptors further contributes to the anti-tumour activity of the CAR and can potentially reduce the risk of relapse mediated through downregulation of the CAR target antigen. While the preclinical and early phase clinical results with CAR-NK cells are encouraging and support their further development, in comparison to CAR-T immunotherapy, the field of CAR-NK therapy is still in its infancy and a number of questions remain to be addressed. These include determination of the best source of NK cells for immunotherapy, the optimal vector system, the most biologically relevant signalling domain for CAR activation and the ideal ex vivo expansion strategy, to name a few.

Nonetheless, the promising preclinical data and the exciting early clinical results with CAR-NK cells support their incorporation into the current armamentarium of cell-based cancer therapeutics. It is likely that in the near future, we will witness an array of combinatorial strategies for the therapeutic applications of NK cells to improve their persistence and their trafficking, enhance their cytotoxicity while preserving their safety and increase their resistance to the immunosuppressive TME.