CAR-NK Engineering to Overcome TME Barriers.

1. Introduction
Chimeric antigen receptor (CAR)–natural killer (NK) cells carry lower risks of cytokine release syndrome, neurotoxicity, and graft-versus-host disease than CAR-T cells and are compatible with allogeneic, off-the-shelf production. However, efficacy in solid tumors remains constrained by the immunosuppressive and physically restrictive tumor microenvironment (TME) [1]. This review discusses the main barriers that limit the effectiveness of CAR-NK cells in the solid TME, including antigen heterogeneity, immune evasion, metabolic suppression, trogocytosis-induced fratricide, and inadequate cell persistence We also outline novel engineering and combinatorial strategies designed to reduce these hurdles and advance CAR-NK therapy towards successful clinical application in solid tumor settings (summarized in Table 1 and Table 2 and Figure 1).

2. NK Cells in the TME
The TME represents a complex, highly immunosuppressive environment that leads to significant challenges to the function and therapeutic efficacy of NK cells. The hostile TME comprises various immunosuppressive cells, including regulatory T cells, cancer-associated fibroblasts (CAFs), myeloid-derived suppressor cells (MDSCs), and tumor-associated macrophages (TAMs). Together, these populations limit immune cell infiltration into tumors and dampen their cytotoxic activity through inhibitory cytokines, nutrient depletion, and extracellular matrix (ECM) remodeling [84,85,86,87]. Furthermore, the TME contains numerous inhibitory factors, such as immunosuppressive cytokines, including interleukin (IL)-10 [87], transforming growth factor (TGF)-β [53], metabolic inhibitors, like prostaglandin E2 (PGE2) [88], adenosine [89], and enzymatic suppressors, such as indoleamine 2,3-dioxygenase (IDO), that converts tryptophan into immunosuppressive L-kynurenine metabolites [90,91]. All of these collectively contribute to the immunosuppressive TME, which is further characterized by hypoxia, nutrient depletion, and lactate-induced acidosis. These factors significantly impair NK cell function through several mechanisms, such as downregulation of activation receptors, including NKG2D (natural killer group 2, member D), and the natural cytotoxicity receptors (NKp30, NKp44, and NKp46). They also upregulate inhibitory checkpoints such as CD94/NKG2A, programmed cell death protein 1 (PD-1), T cell immunoglobulin and mucin domain containing-3 (TIM-3), T cell immunoreceptor with Ig and ITIM domains (TIGIT), cytotoxic T lymphocyte-associated protein-4, lymphocyte activation gene 3 (LAG-3), and interleukin 1 receptor 8 [92,93]. In addition, they suppress the production of granzyme B, perforin, and interferon (IFN)-γ [94,95]. They also alter cellular metabolism, inhibiting glycolysis and oxidative phosphorylation [96]. Furthermore, via chemokine expression modulation, the TME promotes the recruitment of less cytotoxic CD56bright NK cells over the more potent CD56dim NK cells, and the stromal cells create a physical barrier that further prevents the infiltration and trafficking of NK cells towards the tumor targets [97]. Therefore, extensive research has been carried out to understand these immunosuppressive mechanisms in order to engineer CAR-NK cells to better equip them to tackle TME-mediated inhibition and enhance immunotherapeutic potential against solid tumors.

3. Targeting Antigen Heterogeneity in Solid Tumors and Overcoming Immune Escape
Unlike most hematological malignancies, solid tumors are highly heterogeneous in their antigen expression [98], where they often evade immune surveillance by downregulating their targeted antigens [99]. To reduce these challenges, CAR-NK cells can be designed to either increase the binding affinity of the CAR to the low-expression antigens or co-target multiple tumor-associated antigens (TAAs) (Figure 1A). Various techniques, including directed evolution [2], computational modeling [3,4], and phage display [5,6], have been utilized in CAR-T cell studies to design single-chain variable fragments (scFvs) highly specific to these low-density TAAs, resulting in enhanced efficacy while minimizing off-target toxicities. These platforms can also be leveraged for CAR-NK construct optimization. On the other hand, multi-targeting strategies using two independent “dual” CAR-NK cells have been shown to reduce tumor resistance. Dual CAR-NK cells targeting programmed death-ligand 1 (PD-L1) and human epidermal growth factor receptor (HER2) have shown enhanced cytotoxicity across diverse cancer cell lines, including breast, ovarian, pancreatic, and gastric cancer cells, when compared to mono-specific variants [7]. Similarly, dual CAR-NK cells against CD44-CD24 and CD44–mesothelin illustrated superior killing against triple-negative breast cancer cells, reducing tumor survival to 4–10% compared to 18–25% with single-target CARs [8]. NK cells expressing dual CARs targeting GD2 and NKG2D ligands showed improved recognition of glioblastoma (GBM) cells, achieving a ~2-fold reduction in tumor volume and weight compared to single CAR-treated mice [9]. Similarly, dual-targeting epidermal growth factor receptor (EGFR) and EGFRvIII CAR-NK cells improved the killing of GBM compared to mono-specific CARs [10] (Table 1).
Additionally, multi-specific engagers can be administered along with CAR-NK therapy to address antigenic heterogeneity in solid tumors (Figure 1A). Bispecific killer engager (BiKE) molecules are engineered to simultaneously bind tumor antigens and activate NK receptors such as CD16, NKp30, NKp46, and NKG2D. Building on the BiKEs concept, trispecific killer cell engagers (TriKEs) incorporate an additional functional domain, such as the cytokine IL-15, which further promotes activation, proliferation, and persistence, improving their efficacy in the TME [11]. Various studies have shown the improved cytolytic activity of NKG2D CAR-NK cells when incorporated with BiKEs, achieving over 60% lysis in HER2 breast carcinoma cells [12] and increasing glioblastoma cell killing from 11–13% to 32–43% in EGFR/HER2-targeted models [13]. Current CAR-NK clinical trials show activity against broad/stress-induced targets (for example, PD-L1 and NKG2D) and include multi-antigen recognition strategies (Table 1 and Table 2). In addition to antigen recognition, the CAR construct influences the quality of NK cell function. Incorporating NK cell-specific signaling and adaptor domains, such as 2B4, DNAM-1, DAP10, and DAP12, in combination with NKG2D, enhances activation signaling pathways, promotes degranulation, increases IFN-γ secretion, and improves overall cytotoxicity (Table 1) [14,15].

4. Modulating Trogocytosis to Enhance CAR-NK Function
Trogocytosis refers to the transfer of surface proteins, membrane fragments, or cytoplasmic components from another cell via direct cell-to-cell contact without cell fusion or death [100]. Trogocytosis-mediated immune escape has been documented in CAR-NK therapy [17] and occurs in two major ways. Firstly, it involves the transfer of cell surface receptors from cancer targets to the CAR-expressing effector cells, which leads to the downregulation of target antigens on the tumor cells. Secondly, it involves the transfer of CAR molecules to the tumor targets, resulting in antigen masking and depletion of CAR molecules on the engineered immune cells [101]. Furthermore, trogocytosis mediates fratricide, as CAR-bearing cells may target one another due to acquiring the target antigen, thereby reducing cell viability and anti-tumor efficacy.
Trogocytosis of tumor antigens on CAR-bearing cells can upregulate inhibitory molecules, such as LAG-3 and TIM-3, resulting in CAR cell exhaustion and dysfunction [16]. Trogocytosis in CAR-NK cells also causes metabolic dysregulation, leading to a significant suppression of glycolysis [17]. Addressing these challenges, Li and colleagues [17] designed a dual CAR (AI-CAR) with an activating CAR (aCAR) recognizing tumor antigens (ROR1 or CD19) and an inhibitory CAR (iCAR) targeting an NK-specific antigen (CS1) (Table 1; Figure 1B). When NK cells encounter their trogocytosed antigens, the inhibitory CAR sends out a “don’t-eat-me” signal, preventing fratricide, while the activating CAR ensures the effective recognition and elimination of tumor targets. Additionally, studies have shown that pretreatment of CAR-NK cells with immunological synapse inhibitors, such as Latrunculin A [18] and Cytochalasin D [19], prevents trogocytosis (Table 1; Figure 1B). While metabolic impairment following trogocytosis has been documented, strategies aimed at metabolic rescue have not yet been specifically investigated in detail in CAR-NK cells. However, studies in NK cells more broadly have examined metabolic dysregulation and potential modulatory pathways, which are discussed in more detail later in this review.

5. Addressing Tumor Stroma and Extracellular Matrix (ECM) Resistance
The TME consists of tumor cells, immunosuppressive immune cells, CAFs, complex stromal components, including ECM, basement membranes, and endothelial networks [85,86]. While the initial migration of immune cells towards the tumor targets is mediated via cytokine and chemokine gradients, their infiltration is substantially barricaded by the rigidity of the ECM [102]. CAFs, constituting 80% of the total tumor mass in certain malignancies [103], are the primary contributors of ECM deposition and TME remodeling. These cells secrete various mediators, including IL-6, PGE2, IDO, TGF-β, and matrix metalloproteinases (MMPs), all of which significantly impair NK cell cytotoxicity [20]. Administration of Nintedanib, an anti-fibrotic agent, significantly enhances CAR-NK tumor infiltration and cytotoxicity in pancreatic cancer models by suppressing CAF activation, increasing cancer cell killing from ~32% to ~42% (p < 0.0001) [21]. Engineering of CAR-NK cells targeting fibroblast activation protein (FAP), a serine protease overexpressed by CAFs (Figure 1C), has been shown to inhibit tumor progression in non-small cell lung cancer models [104] and cervical cancer models, where anti-FAP CAR-NK-92 cells demonstrated an 1.5-fold increase in CD107a expression and reduced CAF spheroid fluorescence by 50–83% within 34–72 h compared to controls (p < 0.05) [105]. Additionally, designing CAR-NK cells to secrete ECM-degrading enzymes such as heparanase [22], hyaluronidase [23,24], MMP [25], or collagenase [26] may overcome the stromal barriers and enhance CAR-NK penetration in the TME (Figure 1C). Several clinical trials are testing regional delivery of CAR-NK cells via the hepatic artery or intraperitoneally, which aligns with strategies to bypass stromal barriers (Table 2).

6. Enhancing CAR-NK Cell Infiltration and Homing to Tumors
Infiltration of NK cells is mediated via chemokine gradients generated by tumor, stromal, and immune cells within the TME; however, their infiltration remains suboptimal due to the chemokine system dysregulation [27]. Therefore, designing CAR-NK cells expressing tumor-matched chemokine receptors may enhance CAR-NK infiltration (Figure 1C). Studies reported that anti-mesothelin CAR-NK cells with CXCR2 showed enhanced pancreatic tumor infiltration [28], while NKG2D CAR-NK cells co-expressing CXCR1 showed improved migration and infiltration in xenograft models, resulting in a 20% increase in median survival compared to NKG2D CAR-NK cells lacking CXCR1 [29]. Similarly, EGFRvIII-targeting CAR-NK cells with CXCR4 demonstrated superior GBM infiltration and cytotoxicity in vivo, increasing intratumoral accumulation from 10–15 cells/field to around 35 cells/field [30]. Other chemokine receptors, such as CCR5 [31], CCR7 [32,106], and CCR4 [19], have been successfully engineered to enhance NK cell migration (Table 1).
Beyond genetic modification, chemokine receptor profiles can be drastically influenced during the ex vivo expansion of NK cells with specific cytokines. IL-2 and/or IL-15-mediated NK expansion upregulates CXCR3 expression [33,107]. While IL-2 alone upregulates CCR2 expression [34], combining it with glucocorticoids and IL-15 upregulates CXCR3 levels [108]. Finally, IL-18 in culture upregulates CCR7 expression [35]. Again, trogocytosis can also alter chemokine receptor expression; co-culturing of NK cells with CCR7+ K562 cells can lead to the transfer of CCR7 on NK cells, thereby facilitating their migration towards CCL19 and CCL21 ligands [109] (Table 1).
Trafficking of CAR-NK cells can be further enhanced through combination therapies and sophisticated delivery methodologies designed to optimize the TME for lymphocyte recruitment [110]. Approaches to modify the TME include using immunomodulatory agents, such as oncolytic viruses [36] or radiotherapy [25], which can boost tumor inflammation, stimulate immune responses, and improve CAR-NK cell efficacy. Finally, local injection of CAR-NK cells to the tumor sites is a strategic delivery route that has consistently demonstrated to be both safe and effective in preclinical and clinical investigations [37,38] (Table 1 and Table 2). Current clinical studies adopting local delivery/injection routes to enhance trafficking and exposure are listed in Table 2.

7. Modulating Hypoxia and Immunosuppressive Factors in the TME
Tumor hypoxia critically impairs NK cell function, where the upregulation of hypoxia-inducible factors (HIFs), comprising oxygen-sensitive alpha subunits (HIF-1α, HIF-2α, HIF-3α) and constitutive beta subunits (HIF-1β), triggers detrimental cellular changes [39]. HIF-1α downregulates activating ligands, like MICA and MICB, on tumor cells through MMP-mediated shedding [111] and suppresses the expression levels of critical NK activation markers, including NKG2D, NKp30, NKp44, and NKp46 [40], resulting in reduced NK cell cytotoxicity. It further suppresses the production of the degranulation marker CD107a and the secretion of cytotoxic molecules, including granzyme B and perforin, by inhibiting the phosphorylation of extracellular signal-regulated kinases and signal transducer and activator of transcription 3 [41]. Therefore, blocking HIF-1α signaling in NK cells has been shown to improve their anti-tumor efficacy [44]. Furthermore, therapeutic agents targeting HIFs, such as those that inhibit their synthesis, dimerization formation, or trigger degradation, are currently being evaluated [112,113] with the potential to be integrated with CAR-NK therapy (Figure 1E). As shown in CAR-T cell studies, engineering CAR-NK cells with a hypoxia-response element [114,115] or oxygen-sensitive subdomains derived from HIF-1α [116,117] only ensures specificity in hypoxic conditions and reduces off-target effects in normoxic environments (Table 1; Figure 1E).
Hypoxic conditions force NK cells to switch to glycolysis from oxidative phosphorylation, which limits their energy availability and impairs their cytotoxicity. The shift towards glycolysis leads to the accumulation of lactate, which suppresses IFN-γ production, downregulates the expression of NKp46, CD107a, and granzyme B, and triggers mitochondrial damage and NK cell apoptosis [42]. Currently, glycolytic inhibitors and lactate dehydrogenase blockers that may counteract these metabolic challenges are being investigated (Figure 1E) [43]. Another way to improve the metabolic dysregulation is simultaneously integrating metabolic regulators, including adenosine monophosphate-activated protein kinase (AMPK) and peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC-1α), which enhances mitochondrial function and oxidative phosphorylation capacity [118]. While strategies aimed at metabolic rescue have been primarily investigated in NK cells, there remains a research gap in exploring these approaches within CAR-NK cells. Hypoxia also impairs NK cell functionality by modulating the immunosuppressive cells, such as regulatory T cells, M2 TAMs, and MDSCs, within the TME. CAR-NK cells have been engineered to directly target these immunosuppressive populations, particularly MDSCs [119] and TAMs [120], to enhance their anti-tumor activity (Table 1).
Hypoxia-driven expression of CD39 and CD73 by tumor and regulatory T cells elevates adenosine levels, which further reduces NK cell activity via the adenosine A2A receptor [45]. Blocking of CD73 with antagonist antibodies has been shown to improve CAR-NK efficacy (Figure 1E) [46]. Chambers and colleagues [47] also showed that engineering CAR-NK cells targeting CD73 not only exhibited improved infiltration and cytotoxicity against CD73+ tumor cells but also significantly reduced adenosine production (~0 µM), lowering it below the baseline levels observed in cancer cells cultured alone. Furthermore, deletion of the A2A receptor via genetic editing has shown enhanced CAR-NK mediated killing in preclinical models [48,121,122] (Table 1; Figure 1E).
The secretion of TGF-β from the immunosuppressive cells within the TME downregulates the expression of NK activation receptors NKG2D, DNAM-1, and NKp30 and reduces IFN-γ production [49]. Multiple strategies have been explored to block the immunosuppressive effects of TGF-β, including the silencing of TGFβ-induced miR-27a-5p in vivo [50], the deletion of the TGF-β receptor (TGFβR)-2 [51], and the administration of the TGF-β receptor kinase inhibitor [52], all of which led to enhanced NK cell functionality (Figure 1E). Furthermore, by engineering CAR-NK cells with high affinity, dominant-negative TGF-β receptors have been shown to resist the suppressive effects of TGF-β while maintaining their therapeutic potency [53,123]. TGF-β also reduces NK cell activity via the production of reactive oxygen species (ROS) [54]. Addressing this issue, Liu and colleagues [55] designed a CAR-NK cell targeting HER1 to co-express catalase that successfully neutralized the accumulated ROS hydrogen peroxide in the TME and enhanced CAR-NK cell functionality (Table 1; Figure 1E).
Hypoxia-induced loss of NK cytotoxicity can also be rescued by treating CAR-NK cells with cytokines, like IL-2, IL-15, IL-18, and IL-21, which improve survival, proliferation, and cytotoxic activity [44,56,124,125,126,127]. Additionally, BiKEs and TriKEs have been designed to further boost CAR-NK cell activity by simultaneously targeting tumor antigens and engaging CD16 on NK cells, with TriKEs additionally delivering IL-15 to support NK cell persistence and proliferation [56,128]. These approaches largely benefit CAR-NK anti-tumor efficacy, as the activation of BiKEs and TriKEs depends on CD16, and hypoxia has been shown not to impact CD16-mediated antibody-dependent cellular cytotoxicity [125]. Furthermore, research is being conducted on novel NK cell engagers that can activate receptors, like NKp46, NKp30, and NKG2D, which are significantly suppressed under hypoxic conditions [129,130] (Table 1).

8. Improving NK Cell Persistence and Survival
The inherent short lifespan and high turnover rates in vivo constrain NK cell persistence in solid tumors compared to other lymphocytes [131]. Their sustained functionality is dependent on the maintenance of the homeostatic balance between activating and inhibitory signaling, as overstimulation often leads to NK cell exhaustion [132]. This can be reduced by blocking the inhibitory receptors such as KIRs, PD-1, TIGIT, NKG2A, and TIM-3 (Figure 1D) [84,133]. Extensive research is being carried out investigating various checkpoint blockade therapies targeting NK cell inhibitory receptors (Section 9).
Another strategy to boost CAR-NK cell proliferation, persistence, and survival is by supplementing various cytokines such as IL-2, IL-12, IL-15, and IL-18. Armoring CAR-NK cells with IL-15 in their construct enhances their in vivo persistence and improves their metabolic fitness, leading to enhanced anti-tumor activity (Figure 1D) [57,126]. Another strategy to enhance their persistence is by targeting the negative regulators of cytokine signaling, such as the cytokine-inducible SH2-containing protein (CIS). CIS is a negative regulator of both IL-2 and IL-15 signaling, and knocking out of the cytokine-inducible SH2-containing protein gene (CISH) using clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated nuclease 9 (Cas9) in induced pluripotent stem cell-derived CAR-NK cells have been shown to significantly improve metabolic fitness through enhanced mitochondrial activity and mTOR signaling, resulting in overall improvement of NK performance and persistence (Figure 1D) [58]. Additionally, co-expressing anti-apoptotic genes, such as BCL-2 or BCL-XL, may also improve CAR-NK cell survival (Figure 1D) [59]. Another approach to ensure enhanced metabolic adaptation of CAR-NK cells within the TME is to increase the essential amino acid uptake. Increased expression of specific amino acid transport molecules, particularly SLC1A5 and the SLC3A2/SLC7A5 transporter complex, facilitates growth, proliferation, and persistence of CAR-NK cells (Table 1) [60].

9. Incorporating Immune Checkpoint Inhibitors

9.1. PD-1
Immune checkpoint inhibitors play a crucial role in regulating persistent immune responses. In the TME, the tumor cells exploit immune checkpoint pathways to suppress NK cell function, predominantly through the PD-1/PD-L1 axis (Figure 1F). PD-1 is abundantly expressed on activated NK cells, and upon ligation with the PD-L1 receptor overexpressed in tumor cells, leads to immunosuppression, thereby impairing NK cell functionality [134]. To counteract this, CAR-NK cells have been engineered with chimeric costimulatory converting receptors that convert the PD-1 inhibitory signals into activation. Combining the extracellular domain of PD-1 with NKG2D signaling components, along with the 4-1BB co-stimulatory domain, has demonstrated robust in vitro cytotoxicity in various cancer cell models [61] and induced significant gasdermin E-dependent pyroptosis in lung cancer models [62]. A dual CAR incorporating PD-1 and NKG2D recognition domains with DAP10 signaling showed an approximate 27% and 50% increase in TNF-α and TNF-related apoptosis-inducing ligand (TRAIL) production, respectively, enhanced cytotoxicity, and triggered apoptosis in gastric cancer cells [63]. Another innovative engineering of CAR-NK-92, combining IL-15Rα-sushi/IL-15 complexes with PD-1 signal inverters, demonstrated superior persistence and cytotoxicity against pancreatic cancer cells [64]. Additionally, using CRISPR/Cas9 to KO inhibitory genes, such as a disintegrin and metalloproteinase-17 (ADAM-17) and PD-1, enhances NK cell cytotoxicity and cytokine production [65] (Table 1).

9.2. PD-L1
Similarly, CAR-NK cells targeting PD-L1 have demonstrated enhanced cytotoxicity across multiple solid tumor models. High-affinity NK (haNK) cells engineered NK-92 cells expressing endoplasmic reticulum-retained IL-2, with PD-L1 CAR effectively recognized and eliminated various tumor targets, including triple-negative breast cancer and lung, urogenital, gastric, and head and neck squamous cell carcinoma [66,135]. Furthermore, these engineered CAR-NK cells significantly reduced immunosuppressive macrophages and MDSCs expressing high PD-L1 levels [66,135]. Integrating these PD-L1 haNK cells with anti-PD1 and N-803, an IL-15 super agonist, demonstrated superior tumor growth control in preclinical models [66]. Fourth-generation PD-L1 CAR-NK cells co-expressing IL-15 further enhanced the persistence and functionality of CAR-NK cells against various tumor cell lines [68] (Table 1).
In heterogeneous tumor models, PD-L1 CAR-NK cells successfully eliminated resistant populations that escaped T cell-mediated killing due to antigen presentation defects and PD-L1 upregulation [67]. A novel atezolizumab-based PD-L1 CAR-NK cell showed potent activity against tumors overexpressing PD-L1 and exhibited a self-amplifying effect by inducing PD-L1 expression on PD-L1 low tumor cells, thereby enhancing cytotoxicity [69]. CAR-NK cells derived from hematopoietic progenitor cells and human pluripotent stem cells targeting PD-L1 showed enhanced anti-tumor activity both in vitro and in vivo. In addition to lysing tumors expressing high levels of PD-L1, they also revived exhausted T cells, thereby promoting a stronger immune response [136,137]. Dual CAR-NK cells targeting PD-L1 and folate receptor alpha showed enhanced cytotoxicity and promoted the development of a memory-like phenotype in NK cells [138]. Similarly, another dual CAR-NK construct targeting PD-L1 and HER2 demonstrated superior therapeutic efficacy against multiple solid tumors, including breast, ovarian, pancreatic, and gastric cancer cell lines, while minimizing immune escape through antigen loss [7] (Table 1).
Lastly, combining PD-L1 CAR-NK cells with immune checkpoint inhibitors may further enhance anti-tumor activity, as reported in CAR-T studies involving engineered CARs to secrete anti-PD-1 scFvs [70], co-transduction with PD-1 short hairpin ribonucleic acid (RNA) [71], or adenine base editing to downregulate PD-1 expression [72]. All these approaches reduced exhaustion and improved the cytotoxic function of CAR-T cells, and similar strategies can be explored in CAR-NK therapies (Table 1).

9.3. Human Leukocyte Antigen-G (HLA-G)
HLA-G is a non-classical major histocompatibility complex (MHC) class I molecule that binds with immunoglobulin-like transcript 2 (ILT2) and killer cell immunoglobulin-like receptor (KIR) DL4 on NK cells. This triggers an immunosuppressive microenvironment by encouraging the expansion of immunosuppressive cells, like regulatory T cells and MDSCs, and by diminishing NK cell functionality, resulting in immune evasion of tumor cells [73]. HLA-G targeting CAR-NK cells with a DAP12 intracellular signaling domain demonstrated significant cytotoxicity both in vitro and in vivo against several solid tumor models, including breast, brain, pancreatic, and ovarian cancer cells (Figure 1F) [74]. Co-culture of the anti-HLA-G CAR-NK cells with tumor cells led to the activation of the Syk/Zap70 signal transduction cascade of NK cells, thereby reversing the HLA-G-mediated immunosuppressive signals [74] (Table 1).

9.4. B7-H3 (CD276)
The overexpression of the B7 homologous 3 protein (B7-H3) immune checkpoint molecule in several cancers has been shown to suppress NK-mediated cell lysis [139]. Anti-B7-H3 CAR-NK-92 cells demonstrated significantly higher cytotoxic potential and elevated cytotoxic markers, including perforin, granzyme B, and CD107a, against tumor targets both in vitro and in vivo (Figure 1F) [75]. Similarly, Grote and colleagues [76] developed a second-generation B7-H3 CAR-NK-92 cell line, which showed high specificity by lysing neuroblastoma cells expressing high B7-H3 levels while sparing B7-H3-negative cell populations. The same study investigated the resilience of the engineered CAR-NK cells against the TME challenges, including low pH, hypoxia, and diverse immunosuppressive factors. The B7-H3 CAR-NK cells exhibited remarkable therapeutic potential by overcoming the TME-mediated immunosuppression and successfully lysing melanoma cells [76] (Table 1).

9.5. CD47/Signal Regulatory Protein Alpha (SIRPα)
CD47 is an inhibitory ligand overexpressed in multiple cancer cells. It binds with SIRPα on innate immune cells, which triggers “don’t-eat-me” signaling, ultimately leading to tumor immune evasion. This interaction suppresses macrophage phagocytosis [77] and NK cell cytotoxicity against tumor cells [78]. Additionally, elevated CD47 messenger RNA levels correlate with poor patient survival outcomes across multiple solid malignancies [77,79]. Therefore, various approaches have been investigated to disrupt this CD47/SIRPα axis, including blocking antibodies, recombinant SIRPα proteins, and strategic CAR designs. Anti-CD47 antibodies combined with CAR-NK cells demonstrated robust anti-tumor activity in gastrointestinal cancer models [80]. While limited studies are available on anti-CD47 CAR-NK cells, CAR-T cells targeting CD47 show great efficacy against ovarian [81], pancreatic [140], lung [141], and osteosarcoma models (Figure 1F). [82]. CAR immune cells have also been engineered to secrete soluble CD47/SIRPα-blocking agents, including chimeric monoclonal CD47 antibodies [142], SIRPα-Fc fusion proteins [83], and high-affinity CD47 blockers, like CV1 [80,143] and SIRPγ-derived protein (SGRP) [144]. Disrupting the CD47-SIRPα axis enables macrophage phagocytosis, induces phenotype switch from M2 to M1 TAMs [77,83,145], reduces CAR cell therapy resistance, and improves NK cell anti-tumor activity [78] (Table 1). Clinically, CD47 targeting is constrained by on-target, off-tumor toxicity since CD47 is expressed on normal cells, posing safety risks. For example Magrolimab, a monoclonal antibody that blocks CD47, causes anemia in patients with myelodysplastic syndrome [146].

10. Clinical Translation and Ongoing Trials
To highlight the strategies reviewed above, we compiled a table of current CAR-NK clinical trials in solid tumors, summarizing CAR constructs, targets, and phases (Table 2). These trials show first-in-human studies targeting TROP2, PD-1, PD-L1, NKG2D, and others as indicated, with several testing intraperitoneal delivery, hepatic artery transfusion, or combinations with checkpoint blockades and cytokine agonists. A phase II trial of irradiated PD-L1 CAR-NK plus pembrolizumab and N-803 in gastroesophageal and head and neck cancers (NCT04847466, Table 2) reflects emphasis on using CAR-NK cells with immune modulation. Additionally, TROP2 CAR-NK studies, including IL-15 transduced NK cells, aim to enhance persistence, and NKG2D CAR-NK cells may mitigate antigen heterogeneity by recognizing multiple stress ligands (Table 2). In solid tumors, both CAR-T and CAR-NK therapies remain largely confined to early phase I/II studies with modest response rates compared with hematological malignancies. Direct comparisons between CAR-T and CAR-NK in the same indications are not yet available, and cross-trial comparisons are complicated by differences in targets, constructs, and trial design. At present, the rationale for prioritizing CAR-NK engineering in solid tumors rests mainly on their favorable safety profile, off-the-shelf potential, and encouraging preclinical data in TME-relevant models rather than on proven clinical superiority over CAR-T cells [147].

11. Manufacturing Challenges and Limitations of CAR NK Therapy
Despite showing fewer severe toxicities than CAR-T cells due to their shorter lifespan, limited in vivo expansion, and lower cytokine secretion, CAR NK cells still possess some limitations. Off-target binding and on-target, off-tumor toxicity remain a major concern, as few tumor-associated antigens are exclusively expressed on malignant cells. Therefore, it is crucial to develop CAR constructs that can strictly discriminate between malignant and healthy cells. Emerging computation and AI-based antigen discovery tools may help design scFvs with greater tumor specificity while reducing off-target cytotoxicity. Additionally, to improve their persistence, CAR NK cells are often armored with cytokines, such as IL-15, which may lead to systemic toxicity, immune dysregulation, or uncontrolled proliferation [148]. The most common side effects reported after CAR NK transfusion include fever and fatigue, arising from the increase in IL-6 and C-reactive protein [1]. To address these concerns, inducible or logic-gated CAR NK cells (“on-switch” AND/OR-gated CARs) have been designed to control unintended activation or include safety switches, such as inducible caspase-9, that allow selective depletion of CAR-NK cells in cases of severe toxicity [149].
Although CAR NK therapy has shown promising results in preclinical and early clinical studies, clinical-scale manufacturing of CAR NK cells remains a significant barrier to widespread adoption, largely due to the lack of standardization. Various allogeneic sources have been used to generate CAR NK cells, including peripheral blood, umbilical cord blood, NK cell lines (such as NK-92), and iPSCs, each with distinct advantages and limitations previously reviewed in detail [150]. The diverse sources of NK cells, while advantageous, introduce production challenges as different sources exhibit highly varying transduction and expansion efficacy, persistence, and cytotoxic activity, leading to varying performances of the product. Furthermore, unlike other cells, the transduction and transfection efficiency is much lower in NK cells [151], therefore making it much more difficult to generate high levels of genetically modified NK cells. In addition, the batch variability in titer and purity often seen in clinical-grade lentiviral and retroviral vectors ultimately impacts the gene transduction efficiency [152]. Moreover, gene transduction via viral vectors may lead to insertional mutagenesis, which poses the risk of genome instability and secondary malignancies [153]. iPSC-derived NK cells address many of these limitations by providing a renewable, genetically defined, and clonally stable source. iPSCs can be precisely genetically engineered at the pluripotent stage using CRISPR-Cas9 and then cloned to generate uniform, fully functional NK cells with stable CAR expression, therefore eliminating the need to virally transduce mature NK cells. Moreover, these cloned, genetically edited iPSCs allow the generation of CAR-engineered master cell banks that can be expanded, differentiated into defined effector NK cells, cryopreserved, and distributed as standardized “off-the-shelf” products, greatly reducing production time, cost, and variability while ensuring improved persistence and cytotoxicity [154]. The lack of a standardized CAR design, along with significant variation in genetic engineering methods, vector systems, expression strategies, and NK cell sources across different research institutions, further adds to the challenges for clinical application. Therefore, further research determining which genetic engineering approaches and combinations work best for producing effective CAR-NK therapies is required. While advanced genetic engineering strategies, such as quadruple gene-engineered CAR NK cells, have shown durable anti-tumor activity in preclinical studies [155], each genetic modification further adds to the manufacturing complexity. Finally, generating consistent product quality under Good Manufacturing Practice (GMP) conditions requires strict control of GMP-grade cytokines, serum-free media, and seed cell thawing protocols. Feeder-free expansion systems in bioreactors often fail to achieve yields comparable to feeder-based protocols while maintaining regulatory compliance [156]. Furthermore, cryopreservation impairs CAR NK cell activity [157]; therefore, research addressing this and post-thaw viability is crucial. For off-the-shelf use, it is essential to develop a standardized, scalable, and reproducible manufacturing pipeline ensuring the safety, efficacy, and accessibility of next-generation CAR-NK therapies.

12. Conclusions
Effective solid tumor control with CAR-NK therapy will require addressing TME barriers (summarized in Figure 1). This consists of broadening target recognition, preventing antigen loss, limiting fratricide, overcoming stromal barriers, hypoxia, and metabolic stress, enhancing homing, sustaining cell persistence, and blocking inhibitory checkpoints. Early clinical data support the feasibility of these strategies, including multi-antigen targeting, engineered cytokine support, and regional delivery of CAR-NK cells. The future of CAR NK therapy lies in addressing these limitations using innovative approaches and combination strategies. Artificial intelligence has been integrated into CAR therapy, where deep learning approaches enable multi-antigen targeting and antigen escape prevention, natural language processing facilitates clinical data extraction from reports and the literature, computer vision analyzes CAR cell morphology and phenotypes, reinforcement learning optimizes dosing schedules, and predictive algorithms assess therapeutic efficacy and toxicity profiles [158]. Although emerging artificial intelligence and machine learning tools may aid CAR-NK design in the future, their current use in solid tumors is still exploratory.
Prioritizing CAR-NK combinations should be driven by the main barrier in each tumor. In desmoplastic, stroma-rich cancers, pairing CAF/ECM targeting or anti-stroma approaches with multi-antigen targeting may best improve both access and specificity. In hypoxic TMEs, combinations that couple metabolic or mitochondrial support (for example, cytokine armoring or metabolic re-programming) with persistence tools and checkpoint modulation may provide impact, whereas in checkpoint or soluble factor-driven settings, multi-antigen targeting, checkpoint blockade, and cytokine support may provide better efficacy. Additionally, combination strategies such as administering immune checkpoint inhibitors, kinase inhibitors, protease inhibitors, chemotherapy, radiotherapy, STING agonists, oncolytic viruses, and photothermal therapy that can regulate or reshape the TME present opportunities to improve CAR NK homing, persistence, and cytotoxicity [159].