The CAR is a fusion protein consisting of an extracellular antigen-binding domain, transmembrane domain and intracellular signalling domain. The extracellular antigen-binding domain is usually a single-chain variable fragment (scFv), which may recognise specific antigens on tumor cells. Intracellular signalling domains, including CD28, 4-1BB (CD137) and OX40, are often designed to promote the activation and cytotoxicity of T cells. CARs may improve the specificity of autologous T cells for lymphoid malignancies, which has demonstrated remarkable clinical efficacy (15-17). Although autologous CAR-modified T cells exhibit antitumor activity, its manufacture is complex and expensive. In a number of patients, treatment with CAR-T cells has been associated with substantial toxic effects, including cytokine release syndrome and neurotoxicity, which involve treatment in specialised care units (18-20). There is increasing evidence on PD-1 expression in small subsets of NK cells in patients with cancer. Compared with T cells, NK cells express lower PD-1 levels and promote dendritic cell migration; therefore, CAR-NKs cells are more effective in killing solid tumor CAR-T cells. Compared with CAR-T immunotherapy, CAR-NK cells have numerous advantages. To begin with, allogeneic NK cells may be safely used as effective cells without causing graft-versus-host disease (GvHD) because they mediate non-MHC-restricted killing of tumor cells (21). Additionally, ex vivo expanded primary human NK cells produce a different spectrum of cytokines to T cells, including IFN-γ, IL-3 and granulocyte-macrophage colony-stimulating factor, and CAR-NK cells are associated with a lower risk of CRS (22,23).

NK cells from autologous or allogeneic peripheral blood mononuclear cells are usually used for adoptive transfer of NK cells. Cord blood (CB) is a readily available allogeneic NK cell source and an ‘off-the-shelf’ frozen product, which provides an attractive and allogeneic NK cell for immunotherapy. Liu et al (24) demonstrated the effective cytotoxicity of CAR-CB-NK cells fixed with retroviral vectors containing CAR-CD19, IL-15 and inducible caspase-9 suicide gene to the CD19 positive cell line and primary leukaemia cells in vitro. It may significantly prolong survival time in a xenogeneic Raji lymphoma mouse model in vivo. Induced pluripotent stem cell (iPSC)-induced NK cells possess the advantages of NK-92 cell lines and primary NK cells. Researchers may obtain several NK cells with the same properties, expressing NKp44, NKp46, NKG2D and other activated receptors, which have promising prospects for development and application. Li et al (25) used human iPSCs to produce NK cells with novel CARs, which targeted cancer cells in an antigen-specific manner and improved the survival rate of xenotransplantation models of ovarian cancer. Apart from donor-derived NK cells, the NK cell line is also an important source for CAR-NK cell therapy. NK-92 cells lack almost all inhibitory KIRs and leukocyte function-related antigen-1 (LFA-1) and present higher cytotoxicity than primary NK cells (26). Preclinical and clinical studies have demonstrated the efficacy and safety of NK-92 cell line infusion in patients with cancer (27). A recent study demonstrated that KHYG-1 is a promising NK cell line for CAR-NK treatment (28).

Preclinical studies of CAR-NK cells demonstrated good efficacy in patients with colorectal cancer (29), ovarian cancer (24), glioblastoma (28,30), breast cancer (31), neuroblastoma (32) and other solid tumors.

NKG2D ligands are upregulated in the majority of cancer cells. Chang et al (33) constructed a CAR containing NKG2D transmembrane domain, DAP10 costimulatory domain and CD3ζ signal domain. This design enhanced the expression of NKG2D and amplified the downstream activation signal. NKG2D-DAP10-CD3ζ-NK cells presented significant cytotoxicity in prostate cancer and rhabdomyosarcoma cells. Furthermore, in the osteosarcoma mouse model, NK cells expressing NKG2D-DAP10-CD3ζ exhibited strong antitumor activity.

Li et al (25) identified a CAR containing NKG2D transmembrane domain, 2B4 co-stimulatory domain and CD3ζ signal domain, which mediate strong antigen-specific NK cell signalling. NK cells (NK-CAR-iPSC-NK cells) extracted from human iPSCs expressing this type of CAR had a typical NK cell phenotype and stronger antitumor activity than NK cells extracted from T-CAR-expressing iPSC (T-CAR-iPSC-NK cells) and non-CAR-expressing cells. NK-CAR-iPSC-NK cells significantly inhibited tumor growth and prolonged overall survival (OS) in an ovarian cancer xenotransplantation model. This is the first time that human iPSCs have been engineered at the stem cell level to produce iPSC-derived NK cells that express CARs. Further research on this technology may provide standardised and targeted ‘off-the-shelf’ lymphocytes for anticancer immunotherapy.

Xiao et al (29) constructed a CAR by fusing the extracellular domain of natural killer cell receptor NKG2D with DAP12. Although electroporation may result in lower toxicities, the expression of NKG2D RNA-CAR significantly enhanced the cytotoxicity of NK cells on several solid tumor cell lines in vitro and had a significant therapeutic effect on mice with solid tumors. Local injection of CAR-NK cells exhibited good clinical efficacy in three patients with metastatic colorectal cancer. The results demonstrated the potential of NK cells modified with RNA-CAR in the treatment of metastatic colorectal cancer. In previous studies, NK cells expressing NKG2D CARs were usually constructed with viral vectors. When applied clinically, the DNA CAR expression mediated by integrating viral vectors cannot be easily inhibited when severe toxicity is associated with cytokine storm or on-target and off-target tumors. Using short-lived CAR-expressing cells modified with this non-integrating method, the duration and potency of CAR effects may be controlled by different dosing and infusion schemes, which may improve the safety of this CAR-NK regimen. Electroporation may result in lower toxicities, but different dosing and infusion schemes may result in immune rejection (29).

Murakami et al (28) constructed a lentivirus vector composed of the transmembrane domain of EGFR variant III (EGFRvIII), CD3ζ signal domain, CD137 (4-1BB) and CD28 costimulatory domain. These lentivirus vectors were introduced into the human NK cell line KHYG-1. It was found that CAR-KHYG-1 cells had strong antigen specificity and cytotoxicity to U87 cells stably expressing EGFRvIII in vitro. As an ‘off-the-shelf’ treatment method, continuous expansion of CAR-KHYG-1 cells may have a clinical effect in cancer patients. EvCAR-KHYG-1 may be an effective treatment for glioblastoma. Müller et al (30) reported that NK cells that overexpressed CXCR4 EGFRvIII-specific CAR promoted immunotherapy of glioblastoma that secretes XCL12/SDF-1α. The engineered NK cells expressing chemokine receptor and tumor-specific CAR are a promising tool for improving adoptive immunotherapy for tumors.

A preclinical study (31) demonstrated that NK cells co-expressing CAR and IL-15 continued to proliferate even in the absence of exogenous cytokines and exhibited highly selective cytotoxicity against breast cancer cells expressing EpCAM, which were resistant to the cytotoxicity of unmodified NK cells.

Esser et al (32) constructed a clone derivative of the NK-92 cell line, which stably expressed GD2-specific CAR. The GD2-CAR-NK92 cells demonstrated strong cytotoxicity against neuroblastoma cell lines, which depended on the specific recognition of target antigens. The GD2-specific NK cells exhibited cytotoxicity against melanoma cells, breast cancer cells and glioblastoma cells that express GD2, indicating the potential clinical application value of GD2-CAR-NK92 cells.

To date, 10 clinical trials utilising CAR-expressing NK cells for the treatment of solid tumors (NCT02839954, NCT03692637, NCT03692663, NCT03415100, NCT03940820, NCT03940833, NCT03941457, NCT03931720, NCT03656705 and NCT03383978) have been initiated (Table I).

The first study (clinical trial no. NCT02839954, https://www.clinicaltrials.gov/ct2/show/NCT02839954) was initiated by researchers from the First People's Hospital of Hefei to investigate the safety and efficacy of CAR-pNK cell immunotherapy for MUC1-positive recurrent or refractory solid tumors. The enrolled patients received CAR-pNK cell immunotherapy with specific CAR targeting the MUC1 antigen. The main inclusion criteria were as follows: MUC1-positive tumors that lack available treatment methods and have a poor prognosis. Currently available treatments were included. Eligible MUC1-positive tumors in the clinical study included brain malignant glioma, colorectal cancer, gastric cancer, hepatocellular carcinoma, non-small cell lung cancer, pancreatic cancer and triple-negative basal breast cancer. MUC1 expression in tumor tissues was confirmed by immunohistochemistry. CD27 and 4-1BB (CD137) are included as signal domains of CAR in the present study. All patients were followed up until August 2017. A total of 10 patients with non-small cell lung cancer, pancreatic cancer, liver cancer, colon cancer and ovarian cancer were enrolled. Serious side effects, including cytokine release syndrome or myelosuppression, were not observed in these 10 patients, indicating that the treatment was safe and tolerable. However, there is no further report about this research including efficacy data.

The expansion of primary NK cells in vitro is the primary challenge of CAR-NK cell therapy. Since the number of NK cells from a single donor is far from enough for the treatment, the expansion and activation of NK cells are important (34). Interleukin may be used to expand and activate NK cells. The process of amplifying NK cells by adding IL-2 or IL-15 alone usually requires 2-3 weeks (35). The combination of IL-2 and IL-21 is better than IL-2 alone in promoting NK cells to curb tumor cells (36,37). Irradiated K562 cells may be used as donor cells during the expansion of primary NK cells; however, they must be completely removed prior to infusion as K562 is a cancer cell line (38).

The CAR structure is important in the construction of CAR-NK cells. An extracellular binding domain is the core component of CAR, which determines the specificity of effective cells. Presently, there is no specific antigen that is exclusively expressed on tumor cells; instead, tumor cells express several tumor-related antigens. Therefore, CAR structures with several antigen receptor complexes may be designed to recognise tumor-related antigen combinations to improve the specificity and versatility of NK cells. Activated or inhibited receptors may be used in the construction of CAR. NKG2D is an important activated receptor in NK cells. Inhibition of MICA and MICB may promote the antitumor immunity of NK cells mediated by NKG2D and CD16 Fc receptors (39,40). CD244 (2B4) is a signal domain of the lymphocyte-activating molecule-related receptor, which regulates the effective stimulation and co-stimulation signals in NK cells, and may promote signal transduction in NK cells that are relocated to tumor cells (41).

Non-engineered NK cells usually disappear 2 weeks after infusion, which limits the clinical application of NK cells (42,43). Liu et al (44) reported that infused CAR-NK cells lasted for at least 12 months at a low level in vivo. The long-term persistence of HLA-mismatched CAR-NK cells may be mediated by the loose environment created by lymphocyte depletion therapy and ectopic expression of IL-15 by CAR-NK cells, which enhanced the expansion and persistence of transformed NK cells. However, these persistent cells were insufficient to prevent recurrence, which may limit the application of CAR-NK cells in tumors. IL-15 may be used to enhance the persistence of CAR-NK cells in the future.

CAR can be transferred into NK cells through viral and nonviral vectors. Viral vectors are widely used in CAR-NK cell therapy as they can be stably integrated into the genome. Retroviral vectors have high transfection efficiency but have certain disadvantages, including high risk and cost, limited viral vector capacity, insertion mutation, carcinogenesis and induction of immune response (45,46). Lentivirus vectors have a moderate number of insertion mutations, although the lentivirus transduction efficiency in peripheral blood-derived NK cells is relatively low (8-16%) (47). The transfection efficiency of viral vectors is extremely high, but there are certain obvious defects; therefore, it is imperative to investigate and optimise nonviral vectors. The Sleeping Beauty (SB) transposon vector provides a safe, effective and economic method to integrate genetic information through a gene transfer vector, which may overcome the shortcomings of viral vectors. However, the applicability of the SB transposon system in transferring CAR into primary NK cells remains to be verified (48-50). It is considered a safe and economic method to transfect CAR-NK cells with mRNA, which demonstrated significant cytotoxicity in xenotransplantation cancer models (33,51).

NK cells are extremely sensitive to freezing and thawing when activated by cytokines, and the cytotoxicity of NK cells is significantly decreased following thawing (52,53). The sensitivity of NK cells to cryopreservation makes it difficult to store and transport NK cells, which is the limitation of CAR-NK cell therapy. Adequate cell density is also important in the transportation process. High NK cell levels will lead to cell inactivity. As a result of the temperature sensitivity of NK cells, they need to be transported at body temperature to maintain the cytotoxicity.