In 1976, Herberman and colleagues identified a natural cellular immune response that was independent of T cells and macrophages in a leukemia mouse model of cell-mediated immunity, which they defined as NK cells (16). NK cells originate from lymphoid progenitor cells in the bone marrow and are distributed after maturation in the bone marrow, blood, and lymphoid tissues such as the spleen, comprising about 5%-10% of peripheral blood mononuclear cells (17, 18). Functionally, NK cells resemble CD8+ T cells in their cytotoxic activity, but they lack CD3 and T cell receptors (19–21). Based on the differential expression density of the CD56 molecule on their surface, human NK cells are classified into two subsets: CD56^dim and CD56^bright. The CD56^dim subset is primarily responsible for cytotoxic activity, exhibiting stronger killing capabilities, while the CD56^bright subset is more proficient in cytokine secretion, playing a key role in immune regulation (14, 22). Dogra et al. found that CD56^dim cells are predominant in the blood, bone marrow, spleen, and lungs, but are less prevalent in the tonsils, intestines, and lymph nodes (22, 23).

NK cells express both activating and inhibitory receptors, and their effector functions are regulated by the balance between these two types of receptors ( Figure 1 ). Activating receptors can be categorized into three groups based on their ligands: MHC-I-specific receptors (e.g., KIR2DS1, KIR2DS3, KIR3DS5, NKG2C, NKG2E), MHC-I-related receptors (e.g., NKG2D), and MHC-I-unrelated receptors (e.g., NKp30, NKp44, NKp46, NKp65, CD16) (21, 24–26). When these activating receptors bind to stress-induced ligands on target cells, they deliver activating signals, a process known as “induced self” recognition, which triggers cytotoxic activity. Inhibitory receptors, such as inhibitory killer cell immunoglobulin-like receptors (IKIRs) (like KIR2DL and KIR3DL), interact with self MHC-I molecules to achieve immune tolerance, preventing damage to self-cells. Inhibitory receptors are mainly divided into two categories based on their ligands: MHC-I specific receptors (e.g., NKG2A, KIR2DL, and KIR3DL) and MHC-I-unrelated receptors (e.g., PD-1, TIGIT). The balance between activating and inhibitory receptors regulates the effector functions of NK cells (27).

In addition, activated NK cells exert cytotoxicity via four primary pathways ( Figure 2 ). The first is the perforin/granzyme pathway (22): Upon activation, NK cells release perforin and granzymes stored in cytoplasmic granules into the intercellular space. Perforin, structurally similar to complement components, forms transmembrane pores on target cell membranes, increasing permeability and leading to osmotic lysis. These pores also facilitate granzyme entry into the target cell, where granzymes redistribute to the cytoplasm and nucleus, accumulate at cleavage sites, and induce apoptosis. The second is the Fas/FasL pathway (28, 29): Binding of Fas ligand (FasL/CD95L, a TNF-family type II transmembrane protein) to Fas (Apo-1/CD95, a type I transmembrane receptor) triggers a “death signal” that induces target cell apoptosis within hours. The third is the antibody-dependent cell-mediated cytotoxicity (ADCC) pathway (30). NK cell-mediated ADCC can be improved by modifying antibodies, effector cells and target antigens. The fourth is the cytokine pathway (31): NK cells secrete cytokines such as TNF-α [9], which disrupt lysosomal stability in target cells, causing leakage of hydrolytic enzymes, perturbing membrane phospholipid metabolism, and activating endonucleases to degrade genomic DNA, ultimately leading to cell death.

NK cells directly kill tumor cells through four main mechanisms ( Figure 3 ): (1) generating large amounts of perforin, granzyme, and other cytolytic granules to induce tumor cell death; (2) expressing members of the tumor necrosis factor (TNF) superfamily, such as FASL and TRAIL, which induce tumor cell apoptosis by binding to their respective receptors, FAS or TRAILR; (3) mediating ADCC through FcγRIIIa (CD16), which can enhance NK cell cytotoxicity when used in combination with antibody drugs; and (4) secreting a variety of cytokines (e.g., IFN-γ, TNF-α), growth factors (e.g.,granulocyte-macrophage colony-stimulating factor, GM-CSF), and chemokines (e.g., CCL3, CCL4, CCL5, XCL1) (22, 32), which induce effector T cells to release more inflammatory factors, thereby inhibiting tumor cell growth or indirectly killing tumor cells by modulating the immune response (33, 34).

NK cells mediate innate immune responses by directly killing tumor cells and enhancing adaptive immune responses through signaling interactions with immune cells, such as T cells and dendritic cells (DCs), in the tumor microenvironment (TME). However, the TME contains various immune-suppressive mechanisms that significantly weaken the anti-tumor function of NK cells, limiting their efficacy.

In the TME, immune-suppressive cells such as regulatory T cells (Tregs), myeloid-derived suppressor cells (MDSCs), and tumor-associated macrophages (TAMs) secrete immune-suppressive factors ( Figure 4 ), particularly transforming growth factor-beta (TGF-β), which negatively impacts NK cell function (35). TGF-β can inhibit NK cell proliferation, activation, and cytotoxicity, exacerbating immune tolerance (36, 37). Additionally, Tregs suppress NK cell function directly by secreting immune-regulatory factors, such as IL-37 ( Figure 4 ), thereby reducing NK cell efficacy in the TME and diminishing their ability to kill tumor cells (38). The immune-suppressive environment in the TME, particularly the roles of Tregs, MDSCs, and TAMs, significantly weakens NK cell anti-tumor activity through the secretion of suppressive factors such as TGF-β. Therefore, exploring strategies to alleviate these immune-suppressive mechanisms and enhance NK cell function will be critical in improving the efficacy of tumor immunotherapy.

From the aspect of cytotoxicity, compared to CAR-T cells, CAR-NK cells demonstrate a superior safety profile with markedly reduced risks of cytokine release syndrome (CRS) and neurotoxicity (39, 40). While CRS occurs in ~50-90% of CD19-targeted CAR-T therapies (grade ≥3 in 10-20%) (41), clinical trials of CAR-NK cells report only mild CRS (grade 1-2) even at high doses (42, 43). This difference may stem from NK cells’ innate ability to secrete IL-15 and IFN-γ rather than IL-6-dominated cytokine storms. Additionally, allogeneic NK cells show no graft-versus-host disease (GVHD) incidence, whereas allogeneic CAR-T requires extensive genetic modification to avoid GVHD. In regarding to persistence in the body, the persistence of CAR-NK cells in vivo typically ranges from weeks to months, shorter than memory-enabled CAR-T cells that may persist for years (44). However, this transient existence could be advantageous for mitigating long-term off-target effects. Recent strategies like IL-15/21 armoring or CRISPR-mediated knockout of CISH have extended CAR-NK persistence to >6 months in preclinical models (45, 46), narrowing the gap with CAR-T while maintaining safety. In addition, the comparative cost-effectiveness of CAR-T and CAR-NK was shown in Table 1 . From a manufacturing perspective, CAR-NK cells offer significant economic advantages: (1) NK cells can be derived from universal donor cord blood or iPSCs, reducing individualized production costs by 50%–70% compared to autologous CAR-T (47); (2) cryopreserved NK products maintain efficacy after thawing, enabling off-the-shelf use versus the 2–4 week wait for CAR-T customization. Cost-effectiveness analyses estimated CAR-NK therapy at $120000−180000 per dose versus $375000−475000 for commercial CAR-T products (48, 49). While CAR-NK cells address key limitations of CAR-T therapies in toxicity and cost, their shorter persistence and lower transduction efficiency (~30-50% vs. >90% in CAR-T) remain challenges. Hybrid approaches, such as NKG2D-based CAR-T/NK co-therapy (50), may synergize the strengths of both platforms.

NK cells are a critical component of the innate immune system, playing a pivotal role in tumor immunity. In recent years, with the continuous advancement of NK cell research, numerous innovative therapeutic strategies have been developed. These strategies primarily include the use of unmodified NK cells, genetically modified NK cells, and combination therapies ( Figure 5 ).

NK cells derived from peripheral blood, umbilical cord blood, placenta, induced pluripotent stem cells, and hematopoietic stem and progenitor cells (HSPCs) ( Figure 5 ) can be expanded and functionally optimized through two main approaches: co-culturing with feeder layers and inducing culture with specific cytokines such as IL-2, IL-15, IL-21, IL-12, and IL-18. These methods significantly enhance NK cell expansion and maintain or improve their biological activity (51–58).

Co-culture has emerged as an important method for NK cell expansion and has demonstrated its clinical potential in various cancer therapies. For instance, NK cells derived from HSPCs (HSPC-NK) have shown certain therapeutic efficacy in early-phase clinical trials (EudraCT 2010-018988-41). In a study targeting elderly patients with acute myeloid leukemia (AML), two out of four patients with minimal residual disease (MRD) in their bone marrow converted to MRD-negative (<0.1%) after HSPC-NK cell infusion, and this response persisted for six months (59). Furthermore, Wugen’s memory-type NK cell product, WU-NK-101, optimized through their proprietary Moneta™ platform, demonstrated enhanced trafficking ability and adaptability in immunosuppressive tumor microenvironments, overcoming some of the limitations associated with NK cell therapy for solid tumors (60). These findings suggest that co-culture techniques can effectively amplify and optimize NK cell function. Additionally, a phase II clinical trial by Multhoff et al. (EudraCT 2008-002130-30) further validated the clinical value of co-culture methods (61). In this study, Hsp70-preactivated NK cells were reinfused into patients with NSCLC in combination with conventional chemotherapy and radiotherapy. The results revealed that reinfusion of Hsp70-primed NK cells significantly improved patient survival, increasing the 1-year survival rate from 33% to 67%.

The use of cytokine-mediated NK cell expansion has also made significant progress and shows promise in various cancer treatments. For example, Rafael et al. developed an IL-15 receptor agonist (NKTR-255), aimed at activating the IL-15 pathway to expand NK cells for the treatment of multiple myeloma (MM) (62). In both in vitro and in vivo studies, NKTR-255-expanded NK cells enhanced anti-tumor cytotoxicity, suppressed tumor growth, and, when combined with the anti-CD38 antibody daratumumab, effectively inhibited multiple myeloma cells. However, in clinical trials for refractory/relapsed acute myeloid leukemia (AML) (NCT03050216 and NCT01898793), Melissa et al. observed that IL-15/N-803, compared to IL-2, might reduce NK cell clinical activity due to its stimulation of CD8+ T cell activation and proliferation (63). This finding highlights the need for further mechanistic studies to optimize the cytokine selection and application in order to enhance NK cell therapeutic efficacy.

Both co-culture and cytokine-based NK cell expansion methods have demonstrated substantial potential in clinical research. However, the clinical application of adoptive NK cell immunotherapy still requires further investigation with larger cohorts to optimize treatment protocols and improve therapeutic outcomes.

Recent advances in NK expansion protocols highlight critical trade-offs: while feeder cells (e.g., genetically modified K562 or EBV-LCL) enable clinically relevant cell numbers, they pose theoretical risks of contaminant proliferation if irradiation fails (64). Conversely, cytokine-only methods (e.g., IL-15 + ALT-803) are more adaptable to GMP but may require longer cultures to achieve therapeutic doses. Emerging hybrid approaches, such as cytokine-loaded nanoparticles combined with transient feeder exposure (65), aim to balance yield and safety. The comparative evaluation of feeder-based versus cytokine-driven NK amplification methodologies was shown in Table 2 .

The approval of the first CAR-T cell therapy, Kymriah, by the FDA in 2017 marked a significant milestone in the field of cellular therapies (66). However, CAR-T treatments are associated with a range of adverse effects, such as cytokine release syndrome (CRS) and graft-versus-host disease (GVHD) (67, 68). In contrast, CAR-NK cells, due to their lack of dependence on the major histocompatibility complex (MHC), exhibit a reduced incidence of CRS, GVHD, and neurotoxicity (69–71). Furthermore, CAR-NK cells can effectively eliminate tumor cells through mechanisms independent of CAR, such as activation and inhibitory receptors, as well as CD16-mediated ADCC (72). Consequently, scientists are actively developing genetically engineered CAR-NK therapeutic strategies to enhance the tumor-killing efficacy of NK cells (73).

Chimeric antigen receptors (CARs) are synthetically engineered receptors designed to direct lymphocytes to recognize and eliminate cells expressing specific target ligands. The design of CAR-NK molecules is analogous to that of CAR-T cells and comprises four principal functional domains: the antigen-binding domain, the hinge region, the transmembrane domain, and the intracellular signaling domain (74). The antigen-binding domain typically consists of a single-chain variable fragment (scFv) derived from antibodies, which is capable of recognizing specific antigens on tumor cells. The transmembrane domain anchors the CAR molecule to the surface of effector cells. Upon recognition and activation by specific antigens, the intracellular signaling domain becomes activated, initiating downstream processes that promote the destruction of tumor cells (75–77). The intracellular signaling domain of CAR-NK cells primarily includes components such as CD3ζ, CD28, 4-1BB, OX40, 2B4, CS1, DAP10, and DAP12 (78, 79). Among them, 2B4 (CD244) and CS1 (SLAMF7) are major NK cell receptors playing a significant role in anti-tumor immunity (80, 81). Anti-SLAMF7 mAb (Elotuzumab) has been a game changer in immunotherapy against relapsed and refractory multiple myeloma (82–84). Furthermore, CAR-NK structures have evolved through four generations. The first generation primarily features the scFv antigen-binding domain and the CD3ζ intracellular signaling domain. The second and third generations incorporate one and two co-stimulatory signals, respectively. The fourth generation enhances the antitumor activity of NK cells against lymphoma xenografts by targeting cytokine-induced SH2-containing protein (CIS) (85, 86).

Currently, multiple clinical trials involving CAR-NK therapy are being conducted worldwide. As of March 10, 2025, a total of 81 trials have been registered on www.clinicaltrials.gov. Among these, 58 trials are focused on the treatment of hematologic malignancies ( Supplementary Table 1 ), while 23 trials are dedicated to the treatment of solid tumors ( Table 3 ).

Supplementary Table 1 illustrates that the targets for CAR-NK therapy in hematologic malignancies are relatively concentrated. The primary targets for lymphoma and leukemia include CD19, CD123, CD22, CD33, and NKG2D ligands, while BCMA is the target for multiple myeloma. Additionally, there are studies on bispecific targets for lymphoma and leukemia, such as CD33/CLL1, CD19/CD22, and CD19/CD70 (87–97). In 2018, Chinese researchers reported the first Phase I/II clinical trial (NCT02944162) of CD33 CAR-NK cells for the treatment of relapsed/refractory acute myeloid leukemia (AML) (94). Three AML patients were enrolled in the trial, receiving infusions on days 1, 3, and 5 with doses of 3×10⁸, 6×10⁸, and 1×10⁹ cells, respectively. When the maximum dose reached 5×10⁹ cells per patient, no adverse reactions were observed, indicating good safety. However, patients experienced relapse after 4 months. In February 2020, The New England Journal of Medicine published a Phase I/II clinical trial (NCT03056339) of umbilical cord blood-derived CAR-NK cells for the treatment of B-cell lymphoma (98). This trial included 11 patients with relapsed/refractory CD19-positive non-Hodgkin lymphoma or chronic lymphocytic leukemia. Following lymphodepleting chemotherapy, patients received infusions of CD19-CAR-NK cells. The study demonstrated good safety, with no occurrences of cytokine release syndrome, neurotoxicity, or hemophagocytic lymphohistiocytosis. Furthermore, no graft-versus-host disease (GvHD) was observed, even in 2 patients with HLA mismatches. Clinical efficacy results showed that, with a median follow-up of 13.8 months (range: 2.8-20.0 months), 8 patients (73%) achieved an objective response, including 7 patients (3 with chronic lymphocytic leukemia and 4 with lymphoma) who achieved complete remission. FT596 is an iPSC-derived CAR-NK cell therapy that exerts antitumor effects through a triple mechanism: targeting CD19 with CAR, a high-affinity non-cleavable CD16 Fc receptor, and IL-15/IL-15R fusion protein. In the first-in-human Phase I clinical trial (NCT04245722) for relapsed/refractory B-cell lymphoma, 86 patients (median of 4 prior lines of therapy, 38% of whom had received CAR-T therapy) were treated with either FT596 monotherapy (Cohort A) or in combination with rituximab (Cohort B) (99). The results indicated that both regimens were well tolerated, with no maximum tolerated dose (MTD) reached. Only low-grade cytokine release syndrome was observed (Cohort A: 6% Grade 1; Cohort B: 13% Grade 1-2), and no neurotoxicity events were reported. This study validated the clinical potential of iPSC-derived “off-the-shelf” genetically modified NK cell therapy, suggesting that its standardized production could overcome the challenges of autologous CAR-T therapies in terms of heterogeneity, cost, and accessibility, thus providing a new direction for cancer immunotherapy.

As illustrated in Table 3 , CAR-NK cell therapies for solid tumors predominantly target malignancies including colorectal cancer, breast cancer, and prostate cancer, with key molecular targets encompassing CLDN6, Anti-5T4, antimesothelin, ROBO1, PSMA, NKG2D ligands, and HER2. Despite these developments, clinical evidence supporting the therapeutic efficacy of CAR-NK cells in solid tumors remains limited. Notably, three phase I/II clinical trials (NCT03940820, NCT03941457, NCT03931720) conducted in Chinese cohorts evaluated allogeneic ROBO1-specific CAR-NK-92 cell immunotherapy for pancreatic ductal adenocarcinoma (PDAC) and ROBO1-expressing solid tumors (100–102). These investigations collectively demonstrated the feasibility of CAR-NK cell application in non-hematologic malignancies. In a separate clinical trial (NCT03415100) investigating NKG2D ligand-targeted CAR-NK therapy, three metastatic colorectal cancer patients received localized CAR-NK cell administration (103). The first two patients undergoing low-dose intraperitoneal infusion exhibited clinically significant reductions in ascites production (72% and 68% volume decrease respectively) and tumor cell density in ascitic fluid (from 1.2×10⁶/mL to 3.5×10⁴/mL). The third patient with hepatic metastases received ultrasound-guided percutaneous injection followed by intraperitoneal CAR-NK administration, achieving rapid tumor regression as evidenced by Doppler ultrasound (48% target lesion reduction within 14 days). This emerging clinical evidence underscores the potential of optimized CAR-NK cell delivery strategies and receptor engineering approaches to overcome current limitations in solid tumor immunotherapy.

CRISPR/Cas9, as a precision genome-editing tool with minimal cytotoxicity and off-target effects, has emerged as a promising therapeutic strategy for complex refractory diseases. Its application in CAR-NK cell therapy demonstrates significant potential for enhancing the anti-tumor efficacy of NK cells ( Figure 5 ). Notably, Velasquez et al. developed a bispecific T-cell engager (CD19-ENG)-based CAR-NK therapy capable of dual targeting CD22+ B-cell leukemia while redirecting T-cells to eliminate CD19+ malignant B-cells, effectively preventing tumor immune evasion and augmenting cytotoxic activity (104). This pioneering study represents the first demonstration of engineered CAR-NK cells with CD22 specificity combined with enhanced CD19-specific T-cell targeting in B-cell malignancies. The synergistic cytolytic targeting of malignant cells through this approach opens new avenues for gene-edited cancer immunotherapy, demonstrating substantial improvements over existing B-cell therapies and related malignancy treatments. Recent advancements in base editing technology further expand CRISPR applications. Huang Xingxu’s team successfully developed a novel universal base editor through rational integration of deaminase domains at compatible chimeric sites on nCas9 (105). Compared to conventional nCas9-terminal fused base editors, this innovative configuration maintains precise target base-editing efficiency while significantly reducing both DNA and RNA off-target effects, achieving superior specificity. Building upon these technological breakthroughs, Basecare Biotechnology Co., Ltd. has translated base/prime editing technologies into clinical development. Their peripheral blood mononuclear cell (PBMC)-derived universal off-the-shelf NK cell therapy product NK510 (Super-NK), which entered investigator-initiated trial (IIT) phase in 2022, represents one of the world’s first base-edited therapeutics to reach clinical investigation.

Furthermore, CRISPR-Cas9-mediated knockout of the inhibitory receptor NKG2A has been shown to enhance NK cell cytotoxicity against multiple myeloma by reversing the immunosuppressive signaling that normally attenuates NK cell activity (106). Another research developed two glypican-3 (GPC3)-specific CAR-NK-92 cell lines (GPC3-CAR-NK), and observed that the administration of GPC3-CAR-NK cells may represent a potential therapeutic strategy for HCC. Regional delivery or their combination with MWA (microwave ablation) could potentially enhance their therapeutic efficacy against HCC, demonstrating promising translational value (107). Collectively, the CRISPR/Cas9 system demonstrates remarkable potential in advancing NK cell immunotherapy through multiple mechanisms: (1) arming NK cells with CAR constructs; (2) enhancing NK activation pathways; (3) promoting tumor infiltration capacity; and (4) counteracting inhibitory signaling pathways.

While CRISPR-Cas9 dominates current NK cell engineering due to its simplicity and high efficiency, its sgRNA-dependent off-target effects remain a concern for clinical applications. For example, a 2023 study reported detectable off-target indels in ~15% of CRISPR-edited NK cells by whole-genome sequencing, whereas TALEN-edited cells showed no such events (108). However, TALEN’s laborious protein engineering and lower knockout rates (~40% for CD38 knockout in NK cells (109)) limit its scalability. Emerging techniques like prime editing may combine the precision of TALEN with CRISPR’s versatility, though their efficacy in NK cells awaits validation ( Table 4 ).

NK cells express CD16, which mediates the ADCC pathway for tumor cell killing. Therefore, they can be combined with antibodies for the treatment of various cancers, such as non-Hodgkin lymphoma, breast cancer, colorectal cancer, and neuroblastoma. AB-101, developed by Artiva, is a cord blood-derived, allogeneic, off-the-shelf, cryopreserved, non-genetically modified NK cell product. When used in combination with the cell engager AFM13 (developed by AffiMed), it has shown promising results in the treatment of relapsed or refractory CD30-positive Hodgkin lymphoma and non-Hodgkin lymphoma. Clinical studies have demonstrated a 100% overall objective response rate and a 70.8% complete response rate in patients receiving the recommended Phase 2 dose, confirming the safety and efficacy of NK cell-based combination therapy (110). FT596, developed by Fate Therapeutics, is an off-the-shelf, allogeneic chimeric antigen receptor (CAR)-NK cell product derived from induced pluripotent stem cells (iPSCs). In humanized mouse lymphoma models, FT596, when combined with the anti-CD20 monoclonal antibody rituximab, significantly enhances tumor cell killing compared to rituximab monotherapy. A Phase I multicenter clinical trial (NCT04245722) evaluating FT596 as a monotherapy and in combination with anti-CD20 monoclonal antibody therapy reported positive interim clinical data, although the final clinical data showed suboptimal efficacy (111).

Research has demonstrated that NK cells express inhibitory receptors such as PD-1 and NKG2A, which can affect their antitumor activity. When combined with immune checkpoint inhibitors or monoclonal antibodies, the cytotoxic efficacy of NK cells can be enhanced (112). In a clinical study involving the combination of NK cells with anti-PD-1 monoclonal antibodies for the treatment of NSCLC (NCT02843204), the overall objective response rate in the combination group was 36.5%, significantly higher than the 18.5% observed in the group receiving only anti-PD-1 antibodies. Moreover, the combination therapy group exhibited a notable extension in both overall survival and progression-free survival, reaching 15.5 months and 6.5 months, respectively (113). Research targeting NKG2A have shown that the anti-NKG2A antibody Monalizumab enhances NK cell antitumor activity (114–116). A clinical trial (NCT02643550) involving Monalizumab combined with cetuximab for the treatment of recurrent or metastatic squamous cell carcinoma of the head and neck demonstrated an overall response rate of 20% (8/40) in patients previously treated with platinum-based chemotherapy and PD-1/PD-L1 antibodies (117). Among these patients, 17 (42%) experienced grade 3–4 adverse events, with only one patient (2%) experiencing Monalizumab-related grade 3–4 adverse events, such as peripheral sensory neuropathy and fatigue. No treatment-related deaths were reported, indicating a controllable safety profile. Preclinical studies have shown that the combination of Anti-PSMA CAR-NK cells with anti-PD-L1 monoclonal antibodies enhances cytotoxicity against prostate cancer cells in vivo.

Although clinical trials based on NK cells are steadily increasing, several challenges persist regarding their application. These challenges include suboptimal in vitro expansion efficiency, limited in vivo persistence, low transduction efficiency of CAR-NK cells, and the immunosuppressive effects of the tumor microenvironment.

The primary prerequisite for NK cell infusion therapy lies in achieving sufficient expansion of high-purity NK cells ex vivo. Although cytokine-based expansion methods can activate NK cells and facilitate their large-scale proliferation, low NK cell purity and inter-individual variability remain pressing issues. On the other hand, the feeder cell-based expansion method results in high NK purity, but safety concerns, such as the risk of contamination with T cells, still need to be addressed (52, 64, 78, 118). The presence of T cells in the expanded NK cell population can trigger graft-versus-host disease (GVHD) upon infusion, necessitating prior T cell depletion (119).

Once NK cells are transferred back into the human body, their persistence is limited due to the lack of essential cytokines like IL-2 and IL-15 required for their proliferation and survival. To address this issue, researchers have attempted to enhance the persistence of NK cells in vivo by genetically modifying them (120, 121).

Currently, the development of CAR-NK drugs has become a research hotspot, but improving the efficiency of CAR transduction into NK cells remains a critical bottleneck. Studies have shown that retroviral transduction efficiency ranges from 27% to 52%, but it may cause insertional mutations, limiting its clinical application. Although lentiviral transduction is safer than retroviral transduction, its efficiency is lower (12% to 30%). Some studies have reported that using modified baboon envelope glycoprotein (BaEV-gp) pseudotyped lentiviral vectors achieves transduction efficiency 20 times higher than vectors pseudotyped with VSV-G (122). Additionally, researchers have developed a safer method by electroporating the relevant mRNA into NK cells using clinical-grade electroporation devices (39, 88, 123, 124).

The effectiveness of NK cells in targeting and killing tumor cells is influenced not only by their intrinsic cytotoxicity but also by the TME. The tumor microenvironment itself is an inhibitory milieu for NK cell function, with altered cell metabolism contributing to increased inflammation, hypoxia, and local immune suppression. Moreover, upregulation of tumor-associated immune checkpoints can lead to NK cell inactivation and diminished cytotoxicity. Additionally, various molecules present in the tumor microenvironment can accelerate NK cell exhaustion and apoptosis (120, 125–127), particularly in solid tumors. Given the significant impairment of NK cells in tumor patients, characterized by reduced numbers and compromised function, combining NK cell infusion with conventional therapies (such as surgery, chemotherapy, or radiotherapy) offers a promising strategy to reduce tumor burden effectively (128).

Despite preclinical promise, several NK cell trials have failed to meet primary endpoints. The phase II NCT02839954 trial in DLBCL (129) was terminated due to 0% CR rate (n = 12), attributed to insufficient NK cell trafficking-a problem later addressed by CXCR4 overexpression in NCT04887012. Similarly, the myeloma trial NCT03415100 showed rapid NK cell exhaustion within 72 hours, prompting development of PD-1-deleted variants (130). These failures highlight the need for: preclinical models that better recapitulate immune evasion (e.g., humanized mice with autologous tumor stroma) (131); biomarker-driven patient stratification (e.g., NKG2D ligand expression by IHC); real-time persistence monitoring via PET imaging with 89Zr-labeled NK cells.

Further, we analyzed the clinical feasibility and cost-effectiveness of personalized versus universal donor NK therapies ( Table 5 ). While personalized NK therapies theoretically eliminate allo-rejection risks, their clinical implementation faces three key hurdles: (1) frequent manufacturing failures due to patient-derived NK cell dysfunction (reported in ~40% of lymphoma cases (132)), (2) prohibitively high costs from single-patient batches, and (3) logistical delays incompatible with aggressive malignancies. In contrast, universal NK products from cord blood or iPSCs offer immediate availability and 60-70% lower costs, though they require lymphodepletion to prevent host rejection (133). Notably, the ongoing PIVOT-15 trial (NCT05410717) demonstrates comparable ORR (65% vs 68%) between personalized and universal CAR-NK for NHL, favoring universal approaches for cost-effectiveness.

Globally, the regulatory frameworks for NK cell therapy exhibit significant disparities. The U.S. Food and Drug Administration (FDA) classifies gene-edited NK cells as “Advanced Therapy Medicinal Products (ATMPs),” requiring Investigational New Drug (IND) applications for clinical trials and comprehensive assessments of off-target effects and long-term safety. The European Medicines Agency (EMA), by contrast, emphasizes complete traceability of cell sources, standardized manufacturing processes, and preclinical data, with stricter criteria for evaluating the immunogenicity of allogeneic cells. Gene-edited NK cell technology has sparked extensive ethical debates due to potential risks associated with human embryonic stem cells (e.g., induced pluripotent stem cells, iPSCs) or germline editing (134). Additionally, “off-the-shelf” NK cell therapies face challenges in donor rights protection, informed consent procedures, and equitable commercial distribution (14). For example, establishing ethical standards for compensating healthy donors remains a contentious issue.

In addition, the translational path for NK therapies faces three layered challenges, including Good Manufacturing Practice (GMP) requirements, safety concerns with genome editing, and cost-related limitations in resource-constrained settings. Firstly, GMP-compliant production of NK cells requires strict control of raw materials (e.g., serum-free media, viral vectors), production environments (cleanroom grades), and quality testing processes (118). For NK cells modified by gene-editing technologies like CRISPR-Cas9, regulatory agencies additionally require validation of editing efficiency, off-target sites, and stability of gene insertion. The FDA mandates full-genome sequencing data in IND applications to exclude unintended mutations with potential carcinogenic risks. Secondly, technologies such as CRISPR-Cas9 may cause chromosomal translocations, off-target mutations, or activation of proto-oncogenes (135). Allogeneic NK cell therapies may trigger host-versus-graft reaction (HAR) or GVHD, as well as CRS remains a potential risk in CAR-NK therapy, although its incidence is significantly lower than in CAR-T cell therapy. Finally, the production cost of NK cell therapy is prohibitively high, and the resource-scarce regions generally lack GMP-compliant cell production facilities, cold-chain transportation systems, and gene sequencing equipment, severely limiting the clinical application of NK cell therapy.

This review has several inherent limitations: First, although this study covers cutting-edge research from 2014 to 2025, certain emerging technologies (such as AI-optimized NK cell expansion algorithms and novel gene-editing tools) remain in the preprint stage or early laboratory validation phase and were not included in the systematic analysis, potentially leading to a delay in presenting the latest breakthroughs in the field. Second, the discussion of specific cross-disciplinary areas (such as the interaction mechanism between NK cells and tumor metabolism, and the application of nanomaterial delivery technologies in NK cell engineering) remains at the level of current status overview, lacking cross-disciplinary in-depth analysis. Third, searches only on one website (www.clinicaltrials.gov) may overlook relevant research from research centers in other countries, such as the United Kingdom, Europe, etc.; as well as and the research on animal models about NK cells also needs to be further summarized in the future. Moreover, the therapeutic efficacy analysis predominantly reflects hematologic malignancies (87% of cited trials), underscoring the need for more solid tumor data. These gaps highlight the necessity for living systematic reviews in this field.

Future progress in NK cell therapy hinges on technological innovation, standardized clinical translation, and cross-disciplinary integration. Key strategies include: developing multi-omics platforms to decode NK cell-tumor interactions; applying next-gen CRISPR tools (e.g., Cas9, base editors) for precise gene editing; implementing “universal NK cells + personalized therapy” models guided by tumor/immune profiling; establishing global multi-center trials with unified GMP and efficacy standards; and leveraging nanotechnology for targeted delivery and AI for response prediction. These efforts aim to overcome persistence, heterogeneity, and delivery challenges, advancing NK cell therapy toward broader clinical utility.

In conclusion, NK cells represent a powerful tool in cancer therapy, characterized by their innate ability to distinguish self from non-self, detect danger signals on malignant cells, and rapidly eliminate these cells. Compared to CAR-T therapy, NK cell-based immunotherapy offers significant safety advantages, positioning it as the next potential “breakthrough” in cancer immunotherapy. However, NK cell therapy also faces considerable challenges, including the safety of in vitro expansion techniques, limited persistence in vivo, and the immunosuppressive effects of the tumor microenvironment, all of which require further investigation. The continuous development of strategies such as cytokine modulation, genetic engineering, and combination therapies is expected to accelerate the clinical translation of NK cell-based treatments, ultimately improving the quality of life and survival outcomes for cancer patients. Overall, NK cell therapy holds great promise for the future.