Introduction

Immune cells serve as the pillar of strength in antitumor and antiviral processes [1]. Through their rapid recognition and lysing of nascent transformed cells, immune cells can prevent tumorigenesis in the initial stage [2]. However, once malignant cells proliferate and metastasize uncontrollably, they change and depress the immunological responses mediated by host immune cells [3]. By infusing functionally active effector cells into immunocompromised patients, a process known as adoptive cell therapy (ACT), we can reconstruct host immunity and provide a promising strategy for disease treatment [4, 5]. Adoptive transfer of autologous immune cells that have been activated and amplified ex vivo has shown encouraging efficacy in patients with certain hematological cancers. However, the therapeutic efficiency in other tumors is far from satisfactory [6, 7]. With the advancement of gene engineering technology, cytotoxic T cells have been equipped with CARs, which endow T cells with superior and more precise killing capacity. In recent years, CAR-T cells have achieved numerous breakthroughs in cancer treatment, especially in hematologic malignancy treatment [8,9,10,11,12,13,14]. A multitude of CAR-T investigations regarding cancer treatment have progressed into the clinical trial stage, with a high rate of complete remission (CR) being exhibited, and some CAR-T cells have even developed into commercial products [15]. To date, six CAR-T products for treating hematological tumors have been approved by the US Food and Drug Administration (FDA), including Kymriah (Novartis), Yescarta (Gilead), Tecartus (Gilead), Breyanzi (Bristol Myers Squibb), Abecma (Bristol Myers Squibb and Bluebird Bio), and Carvykti (Legend and Janssen). CD19 (four products) and B-cell maturation antigen (BCMA) (two products) are the two primary antigens targeted by CAR-T cells to treat relapsed/refractory (R/R) B-cell-derived leukemia, lymphoma, and multiple myeloma [9, 13, 16,17,18,19,20,21]. Despite these promising outcomes of CAR-T cells in the treatment of hematological tumors, their limited efficacy in the treatment of solid tumors necessitates the exploration of novel strategies to help CAR-T cells break the barriers in solid neoplasm. CAR-T immunotherapy requires apheresis and time-consuming expansion of autologous immune cells from patients. For some patients with aggressively progressing cancer, costly and complicated procedures may result in delayed therapy. In addition, heavily pretreated cancer patients are unable to provide sufficient normal T cells, creating an additional barrier to CAR-T-cell development. Therefore, a surge of interest has recently focused on seeking other candidate immune cells to be engineered with CARs [22].

NK cells, a subset of innate lymphoid cells (ILCs) with diversified killing mechanisms, have recently become a focal point in the application of immunotherapy. The function of NK cells is regulated by a sophisticated array of activating and inhibitory receptors that can distinguish between healthy cells and transformed cells. The integrated signals from the engagement of these receptors and ligands can determine whether NK cells initiate killing activities against aberrant cells or maintain their tolerance of healthy cells [23, 24]. In contrast to T cells, NK cells recognize cancer cells in a human leukocyte antigen (HLA)-unrestricted manner, resulting in the lowest possibility of graft versus host disease (GVHD) development [25]. Furthermore, NK cells rarely induce severe toxicities such as cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS) in vivo [26]. Owing to these favorable attributes, NK cells engineered with CARs can overcome many hurdles that prevent CAR-T therapy from further application. The successful adoptive transfer of allogeneic NK cells into patients further identifies NK cells as a promising platform for CAR engineering and as “off-the-shelf” products for wide application [22]. The strategies of CAR design and transduction used in CAR-T therapy are applicable in NK engineering with encouraging outcomes. To date, CAR-NK cells have shown impressive efficacy in the treatment of hematological tumors and have been widely studied in the treatment of solid tumors, with numerous breakthroughs, such as in the treatment of glioblastoma, breast cancer, and ovarian cancer [27,28,29]. In this review, we present the favorable profile of NK cells as a potential platform for CAR-based engineering and then summarize the outcomes and strategies of CAR-NK cell therapy in up-to-date preclinical and clinical investigations. Finally, we evaluate the challenges remaining in CAR-NK cell therapy and describe existing strategies that can assist us in devising future prospective solutions.

An overview of NK cell biological properties

Development and classification

NK cells are a subgroup of innate lymphoid cells (ILCs) and are identified as the first line of defense against virally infected and/or transformed cells [30]. Derived from CD34+ hematopoietic progenitor cells in bone marrow, NK cells develop in a continuous process in bone marrow as well as in some secondary lymphoid organs (SLOs), such as the spleen, tonsils, thymus, and liver [31, 32]. However, it is unclear whether NK cells differentiate in a linear or nonlinear manner [33]. The developmental stages of NK cells differ significantly among different anatomical locations. Immature NK cells are predominantly distributed in lymph nodes and intestines and have tissue-adaptation signatures, whereas terminally differentiated NK cells mainly populate the blood, bone marrow, spleen, and lungs and have improved effector function [34]. According to the expression levels of CD56 and CD16, NK cells are divided into two major subgroups: CD56brightCD16− and CD56dimCD16+ NK cells [35]. CD56bright NK cells are immature populations and are mainly distributed in SLOs. They were previously thought to be involved in immunomodulation, but recently, they have been identified with robust cytokine-releasing potential after priming with proinflammatory cytokines such as interleukin-15 (IL-15). CD56bright NK cells are more similar to helper cells, secreting abundant cytokines such as interferon-γ (IFN-γ), tumor necrosis factor-β (TNF-β) and granulocyte–macrophage colony-stimulating factor (GM-CSF) [36, 37]. CD56dim NK cells represent the final stage of NK cell maturation and constitute approximately 90% of circulating NK cells. The increased expression of CD16a (FcγRIIIa) and cytotoxic molecules in CD56dim NK cells allows them to mediate serial killing activities toward malignant cells, for example, via antibody-dependent cellular cytotoxicity (ADCC) and death receptor-mediated apoptosis [38, 39]. Recently, high-resolution sequencing technologies further revealed increased heterogeneity of NK cells in different organs, indicating that more NK subpopulations can be further defined beyond the simple delineation of CD56brightCD16− and CD56dimCD16+ NK cells [34, 40, 41].

Activation and cytotoxicity

NK cells are critical for immune surveillance and antitumor responses in vivo. These biological functions are regulated by integrated signals from the stochastically expressed activating and inhibitory receptors on NK cells [42]. The activation and inhibition mechanisms of NK cells are depicted in Fig. 1. Ubiquitously expressed major histocompatibility complex class I (MHC-I) molecules (also known as HLA class I) on healthy cells can bind the inhibitory killer cell immunoglobulin-like receptors (KIRs) or NKG2A of NK cells, which can deliver predominant inhibitory signaling to maintain NK “self-tolerance” [24, 43]. However, tumor cells often downregulate their MHC-I molecule expression to evade the attack of CD8+ cytotoxic T cells, as the target recognition of CD8+ T cells relies on antigen presentation by MHC-I [44, 45]. Thus, the signaling balance of NK cells is broken, and they are inclined toward an activation state (also known as the “missing-self mechanism”). Other transformed or stressed cells expressing excessive activating ligands, such as NKG2D ligands, can directly stimulate NK cell activation through receptor‒ligand engagements called immune synapses [46]. The formulation of immune synapses can initiate firm adhesion and enable the focused delivery of lytic granules such as perforin and granzymes onto the target cells, inducing the apoptosis of target cells [45, 47]. Significantly, a single degranulation can be sufficient to lyse a target cell [48]. Following early lytic granule-mediated killing activities, delayed cell apoptosis responses can be initiated by target cells engaging with death ligands expressed on NK cells such as Fas ligand (Fasl) and TNF-related apoptosis-inducing ligand (TRAIL), conferring the serial killing ability of NK cells [49]. Additionally, CD16 is a potent activating receptor that allows NK cells to engage with antibody-opsonized target cells through ADCC. This crosslinking interaction can subsequently induce NK cells to release the cytotoxic substances mentioned above [50]. In addition to triggering their powerful killing ability, NK cells can secrete an array of cytokines and chemokines to stimulate broader cellular immune responses. For example, the IFN-γ and TNF released by activated NK cells can synergistically mediate the death of target cells [51]. IFN-γ can not only directly activate macrophages but also indirectly promote CD8+ T-cell-mediated immune responses by elevating MHC-II molecule expression on antigen-presenting cells [52]. The NK cell-dendritic cell axis also plays a critical role in tumor immunity. CCL5, XCL1, XCL2, and FLT3L secreted by NK cells are the major chemokines that recruit conventional type 1 dendritic cells (cDC1s). cDC1s can present tumor-associated antigens (TAAs) from apoptotic tumor cells to CD4+ and CD8+ T cells, thus inducing potent T-cell-mediated immune responses [52,53,54].

Fig. 1
figure 1

The mechanism of NK activation and self-tolerance. A In healthy conditions, self-HLA class I molecules of healthy cells bind the inhibitory receptors of NK cells such as KIRs and NKG2A/CD94. Dominant inhibitory signaling suppressed the cytolytic ability of NK cells to make autologous healthy cells “licensed”. B The majority of tumor cells downregulate or lost their MHC-I molecule expression to escape from the immune cells attacking. This results in decreasing tumor ligands combining with inhibitory receptors of NK cells, thus NK cells are activated to secret perforin and granzyme to lyse tumor cells. C Overexpressed activating ligands on stressed cells engage with NK cell receptors, leading to superior activating signaling surpassing inhibitory signaling. As a result, NK cells transform into activation state and initiate cell lysing. D Antibody-dependent cell-mediated cytotoxicity, ADCC. The tumor-specific Fc fragment binds CD16 (FcγRIII) of NK cells, resulting in ADCC development. In addition to ADCC, other killing mechanisms of NK cells include death-receptor-mediated and perforin/granzyme-mediated killing activities

Full size image

The strength of NK cells as immunotherapy candidates

Accessibility to abundant cell sources

NK cells can be obtained from autologous and allogenic sources. Initially, autologous NK cells were the major alternative in adoptive cellular therapy owing to their safety [55, 56]. The evidence suggests that autologous NK cells are not sufficient to exert robust antitumor responses, in part due to the NK inhibitory effects mediated by self MHC-I molecules and functional impairments caused by prior heavy treatment [57, 58]. These findings encourage transitioning the focus on autologous NK cells to allogenic NK cell sources, the use of which can avoid cumbersome collection processes and satisfy clinical doses [22, 59]. These NK cell sources include peripheral blood (PB), umbilical cord blood (UCB), NK cell lines, and stem cell-derived NK cells [42]. Each source of NK cells has its own set of strengths and limitations, as summarized in Fig. 2.

Fig. 2
figure 2

The research progress, advantages, and limitations of various NK cell sources. NK cells can be obtained from 5 different sources: PB, UCB, iPSC, hESC, and NK cell lines. Most cell sources have remarkable tumor-eliminating ability and provide clinically meaningful benefit, having transitioned into in-human studies of different stages. Each source of NK cells has its own set of strengths and limitations

Full size image

PB-derived NK cells, obtained through donor lymphocyte apheresis, represent a conventional option in CAR-NK investigations of cancer treatment. PB-NK cells are mainly a mature population characterized by CD56dimCD16bright cells, without obvious individual variations [60]. Additionally, PB-NK cells show relatively abundant expression of activating receptors such as NKG2D, NKp44, and NKp46, which significantly foster NK cell destruction potential against malignant cells [29]. However, the low proportion of NK cells in PB (approximately 10–15%) largely hinders the cell collection and ex vivo expansion process [34, 61]. NK cells isolated from PB are in various maturation stages and thus are characterized by heterogeneous receptor expression profiles, from maturing to fully mature phenotype variation [62, 63]. Thus, the standardization and stability of cell products are hard to guarantee.

UCB-NK cells are also a valuable and well-studied source, constituting up to 30% of UCB lymphocytes [64]. UCB-NK cells are easy to collect and can be frozen in a cell bank; thus, the incumbrances associated with the apheresis of healthy donors and time-consuming amplification can be avoided [65]. There are fewer contaminating T cells among UCB-NK cells. Furthermore, cell sorting techniques such as immunomagnetic cell separation can assist in attaining high-purity NK cells, minimizing the risk of GVHD as much as possible [66,67,68]. Additionally, UCB cells offer abundant cell sources, where hematopoietic stem cells and progenitor cells can be acquired and then differentiate into therapeutic NK cells with favorable phenotypes [69, 254]. Low-density lipid receptor (LDL-R), the main receptor of VSV-G, has a low level of expression on NK cells [255, 256]. This may be an explanation for the low transduction efficiency of VSV-G pseudotyped lentiviruses (VSV-G-LVs) into NK cells. Statins are widely prescribed medications for CLL patients, but they were recently identified with the function of upregulating LDL-R expression on immune cells, including NK cell lines and primary NK cells [256, 257]. Theoretically, transduction efficiency can be enhanced through statin administration. However, not all statins are suitable for boosting VSV-G-LV transduction efficiency, as most statins can negatively affect cell viability. In a comparative study, rosuvastatin was found to be the most potent substance to augment transduction, and its suppressive effect can be reversed in the presence of GGPP [256]. In addition to regulating LDL-R expression on target cells, employing other glycoproteins to pseudotype the viral vectors is also a feasible strategy to improve virus-mediated transduction efficiency [118]. The virtually unanimous conclusion is that the expression level of lentivirus receptors on target cells has a positive correlation with the transduction efficiency of the virus. BaEV, MV-, and RD114-pseudotyped viruses have been tested in NK cell transduction, and BaEV showed the best performance in large part due to the high expression of its receptors on NK cells [171, 255, 258].

Due to the risk of viral insertional mutagenesis and quality variability in large-scale viral production [259, 260], nonviral transduction methods have gained more attention in recent years.

Electroporation is the most commonly used nonviral transfection method. It is considered a safer transduction approach because it induces short-term gene expression [261, 262]. Electroporating DNA into cells showed a limited transduction rate, but superior performance was presented in electroporating mRNA, which can reach up to a 95% transduction rate with minimal damage to cell viability [149, 263, 264]. The transfection efficiency is significantly enhanced in electroporating NK-92 cells but with limited improvement in transducing CB-NK and PB-NK cells [263, 265,266,267]. Several preclinical results have demonstrated the efficacy of electroporation-based CAR-NK cells in the treatment of both solid and hematologic tumors [76, 272,273,274,275]. SB and PB systems have several advantages over virus-mediated transduction approaches: (1) more random gene integration; (2) a large capacity for foreign genes; and (3) cost-effective production of the basic components [276, 277]. These attributes make the SB and PB systems attractive tools for CAR-based therapy. In recent years, transposon systems have been mainly applied to generate CAR-T cells in preclinical and clinical settings [278,279,280,281,282] but are in the minority of systems used in the CAR-NK engineering field. NK-92 cell lines were easier to engineer with CARs by transposon-based methods, and the resulting products showed effective antitumor responses [138, 283]. Recently, investigators have successfully engineered PB-NK cells with NKG2D CAR and the IL-15 gene using the PB system [162]. There are also some drawbacks to transposon-mediated methods, such as uncontrolled transposition events and transgene remobilization in target cells [284, 285]. Additionally, the transfer of transposon components (such as transposage and gene vectors) needs to be promoted by a virus or electroporation, which can lead to the negative effects mentioned above [272, 286].

CRISPR/Cas9 is a potent genetic modification technique that has been widely applied in cellular immunotherapy [287]. The CRISPR/Cas9 system generally consists of two components: a single guide RNA (sgRNA) and Cas nuclease protein [288]. The precise and highly efficient gene-editing process is initiated through the recognition of specific gene loci by the sgRNA, followed by interaction with Cas9. CRISPR/Cas9 has been utilized in the CAR-T therapy field to address multifaceted issues such as generating allogeneic CAR-T cells and overcoming CAR-T cell exhaustion and the negative factors of the TME [289,290,291]. Similarly, CRISPR/Cas9 was initially adopted to disrupt or insert functionally relevant genes to improve CAR-NK cell performance. Researchers have successfully knocked out CD38 to prevent NK cell fratricide [164] and conducted triple editing (disruption of ADAM17 and PDCD1, knock-in of CD16), achieving high manipulation efficiency and enhanced function [292]. More recently, CRISPR/Cas9 has also been utilized to realize highly efficient and locus-specific CAR transduction into immune cells. Directing CD19 CAR to the T-cell receptor α constant (TRAC) locus by CRISPR/Cas9 resulted in consistent CAR expression in human peripheral blood T cells as well as improved effector responses [293]. One group combined CRISPR/Cas9 with an adeno-associated virus (AVV)-mediated gene-delivery approach to insert anti-CD33 CAR to a safe-harbor locus of primary NK cells, acquiring a mean expression of 68% CAR-positive NK cells and enhanced anti-AML activity [294].

Other emerging transduction strategies also include lipid nanoparticle (LNP)- and charge-altering releasable transporter (CART)-based transduction. LNPs and CARTs serve as protective carriers of nucleic acids, infusing into the cells without degradation by the nucleases. Once entering the cell cytosol, these substances can transform into a positively charged state to allow the release of internal mRNA and then proceed with protein expression. These strategies have been demonstrated to be effective in anti-CD19 CAR transduction into NK cells [295,296,297]. Altogether, these strategies have vast potential as genetic engineering tools in cellular immunotherapy but are still in their infancy of development. More investigations are required to test the safety performance and persistence of these cell products.

CAR-NK cell expansion and persistence

Large amounts of NK cells are required for clinical therapy to achieve sufficient responses. However, the weak in vitro expansion of NK cells significantly hinders CAR-NK cell production and broad application. Autologous NK cells from patients account for a smaller proportion of cells in PB, causing additional difficulty for NK cell expansion [298]. A common expansion method relies on a series of cytokines for stimulation, such as IL-2, IL-12, IL-15, IL-18, and IL-21 [299]. A specific cytokine cocktail can tune NK cells to a particular phenotype. For example, the combination of IL-12, IL-15, and IL-18 can facilitate the generation of memory-like NK cells, which exhibit optimal in vivo persistence and antitumor activity [103, 104, 187, 300]. Nevertheless, the proliferation of cytokine-induced NK cells is associated with limited fold changes. Furthermore, solely depending on cytokines, NK cells easily become cytokine-susceptible and cytokine-addicted, which may raise a major concern for in vivo persistence and vitality in the absence of abundant cytokines [301].

Feeder cells serve as large-scale culture systems that combine cytokine stimulation and receptor-mediated activation [299]. K562 cells are representative feeder cells. Other cells, such as the Epstein‒Barr virus-transformed lymphoblastoid cell line (EBV‑LCL), 721.221, and PBMCs, are also exploited as feeder cells [205, 302, 303]. They are generally engineered to express membrane-bound cytokines (IL-2, IL-15, and IL-21) and/or ligands of NK activating receptors (4-1BBL, OX40L, and HLA-E), which can synergistically promote persistent expansion and antitumor activity [303,304,305,306,307]. K562 cells expressing IL-21 and 4-1BBL have been tested clinically and are considered safe in patients [303]. Compared to the sole cytokine culture system, feeder cells can significantly extend the number of fold changes and alleviate the dysfunction and apoptosis of NK cells induced by cytokine deficiency post-infusion. However, the majority of feeder cells are derived from cancer cell lines and thus must be lethally irradiated prior to infusion. Many concerns are arising about whether surviving feeder cells and other unknown factors could pose potential risks in the context of a complex body environment.

To circumvent the administration of feeder cells to support activation and proliferation, several groups are endeavoring to manipulate CAR plasmids incorporating cytokine transgenes to facilitate expansion and persistence [248,249,250]. The expression form of cytokine gene cassettes can be either membrane-bound or constitutively autocrine. A team from MD Anderson Cancer Center managed to engineer CAR-CB-NK cells to express IL-15 in a constitutively autocrine manner, demonstrating enhanced proliferation and in vivo persistence. There were no signs indicating elevation of systematic IL-15 or other toxicities [26, 128]. In another study, ectopic expression of IL-15 significantly prolonged the persistence of NKG2D CAR- NK cells both in vitro and in vivo. Additionally, the effector function of CAR-PB-NK cells was also significantly facilitated in AML mouse models [162].

The trafficking and infiltration capabilities of CAR-NK cells

It is easier for infused CAR-NK cells to come into contact with hematological cancer cells in circulating peripheral blood; however, in their trafficking to solid tumor sites, multifaceted obstacles are encountered. The trafficking ability and infiltration amounts of NK cells have prognostic value for improved clinical outcomes [309,310,311].

To surmount the anatomical barriers in the treatment of solid tumors, orthotopic injections such as intraperitoneal injections, anterior prostatic lobe injections, and other ultrasound-guided injections have demonstrated effective tumor elimination in CAR-NK cell therapy without tissue damage [139, 192, 312]. In a phase I clinical trial in 9 patients with recurrent HER2-positive GB, NK-92/5.28.z cells targeting Her2 were injected into the wall of the resection cavity during relapse surgery. The disease progression of 5 patients was suppressed, lasting for 7 to 37 weeks, and no signs of dose-limiting toxicities were observed, demonstrating the feasibility and safety of intracranial injection of HER2-targeted CAR-NK cells [313].

The commonly used injection method is intravenous (i.v.) injection. CAR-NK cells need to extravasate from the blood and migrate to the solid tumor bed. This homing process is regulated by the dynamic chemokine receptor-and-ligand interactions between NK cells and tumor cells [314, 315]. Thus, increasing the expression of chemokine receptors on NK cells is a major strategy initially applied in NK cell-based therapies. A growing body of studies has equipped NK cells with chemokine receptors (CCRs), such as chemokine receptor chemokine (C-X-C motif) receptor 2 (CXCR2) [316, 317], CXCR4 [318] and CXCR7 [319], to match their cognate ligands expressed on tumor cells, and improved chemoattraction in the antitumor response of NK cells was shown. Concomitant expression of the CXCR1 or CXCR4 transgene on CAR-NK cells has also demonstrated enhanced migration to CD19+ hematological tumors [320], glioblastoma [321] and ovarian tumors [322]. The release level of chemokines in the TME varies greatly following different kinetics of the tumors, indicating the need for artificial infusion of sufficient chemokines to attract NK cells [318]. The feasible approaches include direct local administration of stimulation factors [318] or delivery of fusion proteins loaded with chemokine ligands [323]. The latter can release chemokine ligands upon engagement with tumor cells, thus facilitating the connection of chemokine receptor/ligand axes. However, these chemokine interactions are versatile and vary in different milieus, possibly causing a totally reversed effect (either promoting or diminishing) on the trafficking ability of NK cells [324, 325]. Thus, more efforts should be made to investigate the comprehensive mechanisms and responses of the chemokine interplay in the intricate context of the human body.

As hinted above, the complex TME is another major barrier that hinders the homing and function of CAR-NK cells. The obstructive factors existing in the TME network can be generally divided into three parts: counterproductive cells, immunosuppressive soluble substances, and a harsh metabolic milieu [180]. In the normal environment of tissues, immune cells such as myeloid cells, regulatory T (Treg) cells, or regulatory B (Breg) cells can provide positive feedback to the proinflammatory cytokines secreted from NK cells. They similarly produce cytokines such as IL-12, IL-15, and IL-18 to promote NK cell growth, maturation, and functionality [53, 326]. However, the formulation and exacerbation of the TME in solid tumors render some immune cells, including Treg cells, myeloid-derived suppressor cells (MDSCs), and tumor-associated macrophages (TAMs), to have suppressive roles. Under TME pressure, NK cells can also be forced to transform to have suppressive phenotypes, which discount their activity and infiltration ability [325]. Traitorous immune cells release inhibitory cytokines such as transforming growth factor beta 1 (TGF-β), prostaglandin E2 (PGE2), and IL-10 or “absorb” cytokines that are favorable for NK cells (IL-2), both of which directly or indirectly impede NK cell response and survival [327,328,329,330]. Thus, circumvention of these negative modulators is critical for CAR-NK cell therapy efficacy in vivo [331]. Pharmacological interventions such as chemotherapies have been adopted to eliminate MDSCs and Treg cells [233, 332, 333]. Additionally, NK cells were manipulated with the NKG2D.ζ CAR construct to target NKG2DL-expressing MDSCs, and cytotoxicity toward MDSCs in the xenograft TME model was observed. NKG2D.ζ CAR-NK cells have been tested in clinical studies to show that they can effectively eliminate intertumoral MDSCs in neuroblastoma patients and facilitate the infiltration and efficacy of infused CAR-T cells [334].

Excessive TGF-β is produced by immunosuppressive cells and tumor cells themselves in the TME. Pathologic levels of TGF-β are often correlated with serious disease progression and an impaired immune system [335,336,337,338]. Multiple strategies have been employed to disrupt TGF-β signaling to preserve NK cell therapy efficacy. Genetically engineering negative TGF-β receptors into NK cells [339,340,341] or knocking out TGF-β signaling-related downstream mediators on NK cells [342] successfully diminished or eliminated the blocking effects of TGF-β on NK cells. In vivo studies have identified enhanced NK cytolytic efficacy in GBM-engrafted mice after silencing TGF-β signaling. Additionally, molecular kinase inhibitors [341, 343, 344] and monoclonal antibodies [345] targeting TGF-β have exhibited equivalent antagonistic effects and improved the performance of infiltrating NK cells. Other TME-abundant suppressive soluble factors, including adenosine, PGE2, IL-10, and IL-37, are likewise potential targets to reverse the unfavorable TME [346, 347]. Adenosine is a metabolic byproduct in response to hypoxia in the TME, and its accumulation can paralyze NK cell functional activity [348]. The mainstream strategies to circumvent the negative effects of adenosine are targeting the hypoxic ectoenzyme CD73 [219, 349,350,351,352] and/or knocking out A2A receptors expressed on immune cells [352,353,354,355]. The results showed a plunging concentration of adenosine in the TME as well as the revival of NK cells. A triple-functional NK was manipulated with a locally released CD73 antibody fragment concomitant with dual-targeting (NKG2D and GD2) CAR expression to target GBM [219]. Enhanced cytolytic ability and persistence of intratumoral NK cells have been observed. Regional regulation of adenosine did not cause metabolic disorders in the whole body. Lever aging eminent preclinical performances, successional clinical trials using CD73 blocking antibodies have been carried out to overcome the dilemma of treating solid tumors (NCT04148937, NCT03454451, and NCT03616886).

Hypoxia, a typical hallmark of the TME, develops as a result of the malignant outgrowth of barely vascularized solid tumor tissues [356]. Restricted oxygen concentrations and deficient nutrients in solid tumor regions induce the downregulation of NK cell functional molecule expression (such as activation/inhibitory receptors, cytokines and death receptors) and substantially suppress their killing performance and migration ability [357, 358]. Nevertheless, relying on the hypoxic TME, tumor cells can escape immune cell monitoring and attack [359, 360]. Thus, mitigating intratumor hypoxia is a potential strategy to improve the treatment of solid tumors. Multiple pharmacological and physical strategies have been widely investigated in a series of preclinical and clinical studies to directly adjust the hypoxic state [361,362,363,364,365,366,367]. For example, hypoxia-activated prodrugs (HAPs) [368] or inhibitors targeting the hypoxia-inducible factor (HIF) protein family may be applied, or patients may even be physically exposed to a hyper oxygenated environment [369]. These strategies may provide directions to ameliorate and transform infiltrating-tumoral CAR-NK cell functional and metabolic exhaustion. We can also draw inspiration from a recent novel hypoxia-sensing CAR-T cell structure, which made full use of deleterious hypoxia as a switch to favor itself, launching a CAR-mediated killing procedure [370, 371]. This approach may establish a good pattern for the future design of CAR-NK cells to overcome hypoxia. Another unfavorable condition is the low pH in the TME. As a result of hypoxia, more anaerobic glycolysis reacts to support the activity of tumor cells; thus, increasingly accumulated lactic acid can damage the functionality of NK cells and promote suppressive immune cells [372]. The depletion of excessive metabolites is a direct strategy. However, it is noteworthy that preventing excessive immune metabolite modulation and maintaining a physiologic balance of the inner environment would be critically important for NK cell function.

Conclusion and future perspectives

Paradigm shifting CAR-T-cell therapy has pioneered the development of the CAR technique and yielded promising outcomes in treating hematological tumors. Benefiting from the technologies and valuable lessons learned from CAR-T cell therapy, CAR-NK cell therapy has advanced rapidly with continuous innovations. Preclinical and early clinical outcomes have demonstrated the vast potential of CAR-NK cells as “off-the-shelf” products for cancer treatment. To date, numerous strategies have been applied in CAR-NK cell therapies to address the challenges discussed above, and satisfactory outcomes have been observed in preclinical studies. However, some of these strategies are difficult to translate into clinically approved procedures, such as the systematic infusion of NK-cell-stimulating cytokines. In contrast, multiplexed CAR-NK cell design systems or combinatorial approaches based on radiotherapy and other FDA-approved drugs may hold great potential to overcome the barriers in the CAR-NK cell therapy field and provide clinical benefit. With the development of cutting-edge technologies such as single-cell RNA sequence analysis (scRNA-seq), we have access to elucidating key parameters associated with CAR-NK cell biological function and therapeutic efficacy, which may provide investigators and clinicians with critical insights into how to optimize the promise of NK cell-based cancer therapy. In the coming years, clinical translation-oriented research and in-depth clinical testing of CAR-NK cell therapies are urgently needed to determine their potential market authorization.