Induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells: Engineering innovations, translational hurdles and clinical prospects in immune therapy

Abstract

Induced pluripotent stem cell-derived chimeric antigen receptor natural killer (iPSC-CAR-NK) cells are an evolving off-the-shelf cellular immunotherapy platform with potential for scalable manufacturing, product standardization, and multiplex engineering. This review summarizes recent progress in clinical-grade manufacturing, quality-attribute definition, and natural killer cell-adapted engineering strategies, including chimeric antigen receptor design, gene editing, tumor microenvironment adaptation, single-cell and multi-omics-guided optimization, and synthetic biology-based control. In addition, this review discusses recent progress in the application of iPSC-CAR-NK cells to hematologic malignancies and solid tumors and cautiously examines the early translational signal suggested by a single compassionate-use report in systemic sclerosis. To move beyond a descriptive listing of engineering strategies, we propose a multilevel analytical framework encompassing product definition, effector execution, tissue delivery, host interaction, and translational implementation. This framework is used to evaluate the functional roles, interrelationships, and limiting factors of engineering modules during clinical translation. Current evidence suggests manageable safety profiles and preliminary antitumor or immunomodulatory activity in some settings, but major barriers remain, including limited persistence, insufficient solid-tumor infiltration, host immune clearance, antigen escape, multiplex-engineering risks, and immature quality-control and regulatory standards.

Key Words: Induced pluripotent stem cells; Chimeric antigen receptor natural killer cells; Off-the-shelf immunotherapy; Cell engineering; Clinical translation; Tumor microenvironment; Genome editing; Synthetic biology

Core Tip: Induced pluripotent stem cell-derived chimeric antigen receptor (CAR) natural killer cells provide a candidate standardized, cell-bank-driven platform for off-the-shelf immunotherapy. Their translational value should not be judged solely by CAR expression, enhanced cytotoxicity, or accumulated engineering modules. This review proposes a multilevel framework integrating product definition, effector execution, tissue delivery, host interaction, and translational implementation, and emphasizes dominant bottlenecks, evidence levels, engineering trade-offs, and the gap between preclinical functional enhancement and established clinical benefit.

INTRODUCTION

Current landscape and major challenges in cancer immunotherapy

Cancer remains one of the leading causes of death worldwide and continues to impose a substantial public health burden. According to data from the International Agency for Research on Cancer, approximately 20.0 million new cancer cases and 9.7 million cancer-related deaths occurred worldwide in 2022[1]. Immunotherapy has become an important treatment modality for cancer. Among immunotherapeutic approaches, immune cell therapies, particularly chimeric antigen receptor T-cell (CAR-T) therapy, have achieved substantial clinical progress in certain hematologic malignancies[2]. In recent years, the application of CAR-T cell therapy has gradually expanded beyond oncology. Small-scale clinical studies have reported signals of clinical remission in autoimmune diseases such as refractory systemic lupus erythematosus, with profound B-cell depletion considered a potential underlying mechanism[3]. However, autologous CAR-T cell therapy still has several limitations. It typically requires individualized manufacturing using the patient’s own T cells as the starting material, with production timelines that may extend to several weeks and relatively high treatment costs[4]. In addition, CAR-T cell therapy may induce severe toxicities, including cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome, thereby requiring rigorous clinical monitoring and management[5]. In solid tumors, CAR-T cell therapy also faces multiple barriers, including an immunosuppressive tumor microenvironment (TME), antigen heterogeneity, insufficient tumor infiltration, and limited in vivo expansion and persistence. As a result, its overall efficacy has generally not matched that observed in certain hematologic malignancies[6].

CAR-natural killer (NK) cell therapy refers to a therapeutic strategy in which CAR engineering technology is applied to NK cells. Unlike CAR-T cells, CAR-NK cells can not only mediate CAR-dependent antigen-specific killing but also retain NK-cell natural cytotoxicity, endogenous activating and inhibitory receptor-mediated regulation, and antibody-dependent cellular cytotoxicity (ADCC) when cluster of differentiation 16 (CD16) is expressed. Previous studies and reviews have suggested that, compared with CAR-T cells, CAR-NK cell therapy may be associated with lower risks of graft-vs-host disease (GvHD), CRS, and neurotoxicity[7]. However, this does not mean that these risks are negligible. CAR-NK cell therapy still requires systematic clinical monitoring and long-term safety validation. Table 1 briefly compares the major characteristics of CAR-T cell therapy and CAR-NK cell therapy. Different cellular sources used for CAR-NK cell therapy each have specific advantages and limitations. Peripheral blood- and cord blood-derived NK cells have already been explored in clinical applications; however, large-scale expansion, donor variability, and batch-to-batch consistency remain important limitations[8]. The NK-92 cell line is relatively easy to expand and is suitable for early-stage process development and mechanistic validation. However, NK-92 cells usually require irradiation before infusion, which limits their in vivo proliferation and sustained activity. In addition, because NK-92 cells generally lack endogenous CD16, their intrinsic ADCC capacity is limited, and this functional limitation often requires compensation through genetic engineering[9,10]. In contrast, induced pluripotent stem cell (iPSC)-derived CAR-NK cells theoretically offer platform-level advantages, including clonal cell banking, scalable production, batch-to-batch consistency, and multiplex engineering. However, these advantages still require cautious evaluation in the competitive context of other off-the-shelf cellular immunotherapy platforms. Accordingly, a subsequent section compares peripheral blood-derived CAR-NK cells, cord blood-derived CAR-NK cells, NK-92 cell line-derived CAR-NK cells, iPSC-derived CAR-NK (iPSC-CAR-NK) cells, allogeneic CAR-T cells, and γδ T-cell and CAR-γδ T-cell therapies.

Table 1 Comparison of chimeric antigen receptor T lymphocytes and chimeric antigen receptor natural killer cells.

Characteristic	CAR-T cells	CAR-NK cells
Source	Usually derived from patients’ autologous T cells, but they may also be derived from T cells from healthy donors	Can be derived from peripheral blood, cord blood, iPSCs, the NK-92 cell line, bone marrow, placenta, and tumor-infiltrating NK cells
Cell type	T cells	NK cells
Mechanism of action	CAR-mediated antigen recognition activates T-cell cytotoxic effector function, thereby killing target cells	CAR-NK cells mediate CAR-directed killing while retaining the endogenous cytotoxic activity of NK cells; when CD16 is expressed, they can also mediate ADCC
Target specificity	Can be engineered to target various tumor-associated surface antigens	Can be engineered to target various tumor-associated surface antigens and may supplement antigen recognition through endogenous NK-cell receptors
Product manufacturing and application model	Autologous CAR-T cells usually require individualized manufacturing; allogeneic CAR-T cells require additional immune-safety engineering	Theoretically more amenable to standardized allogeneic manufacturing and “off-the-shelf” production, and may reduce variability caused by interindividual differences
Safety	May be associated with CRS, ICANS, GvHD, and other treatment-related toxicities	Existing studies suggest that the risks of CRS and neurotoxicity may be lower; however, systematic clinical monitoring and long-term safety validation are still required
Cost	The manufacturing process is complex, and individualized production usually results in relatively high costs	More compatible with scalable manufacturing models and may theoretically reduce costs in allogeneic application settings; however, this still requires validation in terms of GMP manufacturing data, quality-control costs, and real-world evidence
Immune escape and in vivo limitations	Susceptible to antigen loss, T-cell exhaustion, and the immunosuppressive tumor microenvironment	Endogenous NK-cell recognition and ADCC may partially complement CAR-mediated recognition; these cells do not carry the risk of αβ TCR-mediated GvHD, but may still face host immune clearance, insufficient in vivo persistence, limited tumor infiltration, and suppression by the tumor microenvironment

CAR: Chimeric antigen receptor; NK: Natural killer; iPSCs: Induced pluripotent stem cells; ADCC: Antibody-dependent cellular cytotoxicity; CRS: Cytokine release syndrome; ICANS: Immune effector cell-associated neurotoxicity syndrome; GvHD: Graft-vs-host disease; GMP: Good Manufacturing Practice; TCR: T cell receptor.

Current research trends suggest that patient stratification, engineering optimization, and combination therapeutic strategies are becoming important directions in the field of cellular immunotherapy. Liquid biopsy can provide information on biomarkers such as circulating tumor DNA and tumor-associated extracellular vesicles, which may help support patient stratification and treatment response monitoring[11]. Artificial intelligence (AI)-assisted modeling has also been used to explore CAR design, manufacturing, and the prediction of efficacy and toxicity in CAR-related therapies; however, its translational value depends on high-quality data and reproducible validation workflows[12]. With regard to combination therapy, combining CAR-NK cell therapy with strategies such as radiotherapy, chemotherapy, or oncolytic viruses is mainly intended to modulate the immunosuppressive TME and enhance NK-cell function[13]. Against this background, iPSC-CAR-NK cells may be regarded as a candidate platform worthy of further investigation in the field of off-the-shelf cell therapy. However, their potential advantages still require further validation in terms of clinical evidence, long-term safety assessment, manufacturing consistency, and real-world application data.

Technical basis and translational potential of iPSC-derived CAR-NK cells

iPSCs have long-term self-renewal capacity and multilineage differentiation potential, and are relatively amenable to genetic engineering at the cell-banking stage[14,15]. Genetic modifications, including CAR introduction, are usually performed at the undifferentiated iPSC stage, which may theoretically facilitate the generation of NK cells with more stable and consistent engineered features after subsequent differentiation. The modified iPSCs can then be further differentiated into NK cells using either feeder-dependent or feeder-free protocols[16-18]. From a manufacturing perspective, iPSC-derived NK cells have the potential to serve as a more homogeneous off-the-shelf cell product, thereby reducing variability caused by donor heterogeneity[18]. However, whether this platform can substantially reduce costs still depends on the stability of the master cell bank, the feasibility of large-scale manufacturing, the success rates of differentiation and expansion, quality control (QC) costs, supply-chain maturity, and real-world data.

The development of iPSC-derived CAR-NK (iPSC-CAR-NK) cell technology has followed a staged evolutionary trajectory from foundational reprogramming technologies and NK-cell differentiation systems to NK-adapted CAR structural design and exploratory clinical translation (Figure 1). Milestone studies published in 2006 and 2007 established iPSC technology in murine and human cells, respectively[19,20]. In 2013, a study established a feasible strategy for deriving NK cells from human embryonic stem cells and iPSCs, and further suggested the potential of this approach for clinical-grade scalable manufacturing[17]. In 2018, another study introduced a CAR structure better adapted to the signaling characteristics of NK cells, and signals of enhanced antitumor activity were observed in an ovarian cancer model, thereby advancing the concept of NK-optimized CAR design[21]. With respect to industrial development and clinical translation, publicly available information indicates that iPSC-NK products such as FT500 and FT516 have entered clinical evaluation. FT596 has been described as an iPSC-CAR-NK cell product whose design integrates a CD19 CAR, an ADAM17 cleavage-resistant high-affinity CD16 variant (hnCD16), and an interleukin-15 (IL-15) signaling-enhancement module, highlighting the development potential of the iPSC platform for multi-module engineering. Beyond oncologic indications, the exploration of iPSC-CAR-NK cells in immune-mediated diseases such as systemic sclerosis remains at a very early stage. For systemic sclerosis, a single-case compassionate-use report published in Cell described the first-in-human (FIH) application of iPSC-derived CD19/B-cell maturation antigen (BCMA) CAR-NK cells in a patient with severe diffuse cutaneous systemic sclerosis, in which signals of B-cell depletion and clinical improvement were observed[22]. However, this result should still be interpreted as an early feasibility signal in the context of a single case. It should not be equated with disease-level validation of efficacy in systemic sclerosis, nor does it support broad extrapolation to other autoimmune diseases. For B-cell lymphoma, a phase I FIH trial published in The Lancet reported that FT596 was generally well tolerated, with no apparent neurotoxicity signal observed. Relatively deep clinical responses were observed in some patients and were maintained during the limited follow-up period[23].

Figure 1 Developmental timeline of induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells. The figure summarizes representative milestones in the development of induced pluripotent stem cell-derived chimeric antigen receptor natural killer (iPSC-CAR-NK) cell technology. Key foundational milestones include the generation of iPSCs from mouse somatic cells in 2006 and from human somatic cells in 2007, followed by the establishment of protocols for generating mature NK cells from human embryonic stem cells and iPSCs in 2013. In 2018, the unmodified iPSC-derived NK cell product FT500 received investigational new drug (IND) clearance, and an NK cell-specific CAR structure was designed to better match NK-cell signaling biology. In 2019, FT516 and FT596 received IND clearance. Subsequent advances included gene-editing strategies for universal cell products, including beta-2 microglobulin knockout and human leukocyte antigen E knock-in, followed by current efforts involving solid tumor trials, immune-mediated disease applications, and next-generation designs such as logic-gated, regulatable, and anti-exhaustion strategies. iPSCs: Induced pluripotent stem cells; NK: Natural killer; hESCs: Human embryonic stem cells; FDA: Food and Drug Administration; IND: Investigational new drug; CAR: Chimeric antigen receptor; B2M: Beta-2 microglobulin; HLA-E: Human leukocyte antigen E.

This review focuses on iPSC-CAR-NK cells and systematically summarizes their developmental trajectory from standardized manufacturing and engineering optimization to clinical translation. Rather than simply describing different engineering strategies, this review places greater emphasis on understanding this platform from the perspective of conceptual integration: iPSC-CAR-NK cells are not merely another source of conventional CAR-NK cells, but a cell-bank-based, off-the-shelf cell therapy platform amenable to modular engineering. The translational value of this platform depends on the balance among multiple levels, including product definition, effector execution, tissue delivery, host interaction, and translational implementation. On this basis, this review further discusses three emerging directions in the field. First, single-cell sequencing, multi-omics, and AI-assisted analyses can be used to identify functional cell subpopulations, differentiation states, TME-related constraints, and potential response biomarkers, thereby providing a basis for engineering decision-making and patient stratification. Second, synthetic biology-based programmable regulation may help improve the spatiotemporal control of cell activation and therapeutic safety through logic-gated CARs, self-regulatory feedback circuits, and safety switches. Third, the application of iPSC-CAR-NK cells in indications beyond cancer remains at an early exploratory stage. Accordingly, this review cautiously discusses only the early feasibility signal suggested by a single-case compassionate-use report in systemic sclerosis and avoids extrapolating it as evidence of broad applicability to autoimmune diseases. By integrating preclinical and early clinical evidence across hematologic malignancies, solid tumors, and immune-mediated diseases, this review aims to provide a conceptual integration framework for understanding the current progress, key barriers, and future directions of iPSC-CAR-NK cells as an off-the-shelf cellular immunotherapy platform.

Literature search strategy and scope of the review

iPSC-CAR-NK cells are still rapidly evolving, and related research spans multiple levels, including basic mechanisms, cell preparation, engineering strategies, preclinical validation, early clinical translation, QC, and regulatory considerations for translation. Different studies vary in evidence type, maturity, and extrapolability; therefore, it is necessary to interpret the existing literature in a stratified manner. This article is a narrative review rather than a systematic review or meta-analysis; therefore, no quantitative pooled analysis was performed. Instead, this review provides a comprehensive analysis and conceptual integration of iPSC-CAR-NK cells with a focus on manufacturing, engineering optimization, preclinical studies, early clinical translation, safety, QC, regulatory considerations, commercial feasibility, and future directions.

This review is primarily based on journal articles indexed in PubMed/NLM and confirmed by DOI verification, while also including published conference abstracts and DOI-confirmed literature related to background contextualization, comparison with competing platforms, manufacturing, QC, regulatory considerations for translation, and methodology. The literature search and source verification were conducted up to June 2026. Search terms were based primarily on English terminology and supplemented by relevant Chinese keywords for cross-checking. The main terms included “induced pluripotent stem cell-derived natural killer cells”, “iPSC-derived CAR-NK cells”, “CAR-iNK”, “CAR-NK”, “iPSC-NK”, “FT596”, “FT516”, “FT576”, “FT538”, “synthetic biology”, “synthetic Notch”, “synNotch”, “interleukin-15”, “IL-15”, “tumor microenvironment”, “solid tumors”, “hematologic malignancies”, and “systemic sclerosis”. Corresponding Chinese terms were used for supplementary cross-checking where appropriate, and topic-specific combined searches and reference tracking were performed. Clinical trial registration information, corporate public disclosures, or other publicly available materials that were not listed as independent references were used only when necessary to verify the research background or translational progress. They were not treated as core evidence sources for independent analysis, nor were they considered equivalent to published clinical evidence.

Literature inclusion and selection were mainly based on titles, abstracts, full-text content, study type, level of evidence, and thematic relevance. This review primarily included the following categories of literature and evidence sources: Studies related to the generation, differentiation, expansion, cryopreservation, and QC of iPSC-derived NK cells or iPSC-CAR-NK cells; studies involving CAR structural optimization, gene editing, cytokine support, adaptation to the TME, synthetic biology-based regulation, safety-switch design, and single-cell- or multi-omics-guided optimization; preclinical studies, clinical studies, and early translational materials on iPSC-derived NK cells or iPSC-CAR-NK cells in hematologic malignancies, solid tumors, or immune-mediated diseases; studies or reviews related to the manufacturing processes, potency evaluation, safety, regulatory considerations, and translational barriers of cell therapy products; and literature related to CAR-T cells, allogeneic CAR-T cells, NK-92-derived CAR-NK cells, γδ T cells and CAR-γδ T cells, and other off-the-shelf or potentially off-the-shelf immune cell therapies that could inform background contextualization or comparisons with competing platforms. Materials that had a weak association with the iPSC-derived NK/CAR-NK platform, substantial redundancy, insufficient methodological or results-related information, or limited relevance to the topic of this review beyond general background information were not discussed as core evidence.

During evidence integration, this review prioritized peer-reviewed original studies, published clinical studies, and preclinical studies with relatively well-defined mechanisms. Published conference abstracts were discussed only as early translational signals or indications of research progress, and were not equated with established evidence of clinical efficacy. In this review, human studies, clinical trials, and single-case compassionate-use reports were classified as clinical evidence or materials related to early clinical translation; in vitro experiments, organoid studies, and animal model studies were classified as preclinical evidence; CAR structural design, gene editing, cytokine support, synthetic biology-based regulation, adaptation to the TME, and single-cell analyses or multi-omics analyses were classified as engineering and mechanistic evidence; Good Manufacturing Practice (GMP) manufacturing, master cell banks, critical quality attributes, potency assays, release criteria, comparability studies, manufacturing costs, and regulatory considerations were classified as evidence related to manufacturing, QC, and regulation; and literature on CAR-T cells, allogeneic CAR-T cells, NK-92-derived CAR-NK cells, γδ T cells and CAR-γδ T cells, and other related platforms was classified as background evidence and competing-platform comparison evidence. Different levels of evidence were distinguished wherever possible in the text to reduce confusion among clinical evidence, preclinical evidence, and theoretical engineering strategies. The literature sources, evidence stratification, and logic of conceptual integration in this review are shown in Figure 2.

Figure 2 Literature sources, evidence stratification, and conceptual integration framework. This figure summarizes the literature sources, evidence-type stratification, and process of conceptual integration in this review. This review is primarily based on journal articles indexed in PubMed/NLM and confirmed by DOI verification, while also including published conference abstracts and DOI-confirmed literature related to background contextualization, comparison with competing platforms, manufacturing, quality control, regulatory considerations for translation, and methodology. As this review is a narrative review rather than a systematic review or meta-analysis, no quantitative pooled analysis was performed. Instead, the evidence was stratified into clinical and early translational evidence, preclinical evidence, engineering and mechanistic evidence, evidence related to manufacturing, quality control, and regulation, and background evidence and competing-platform comparison evidence, thereby supporting the conceptual integration framework of this review. CMC: Chemistry, Manufacturing and Controls; QC: Quality control.

Conceptual framework: A multilevel analytical model for the clinical translation of iPSC-CAR-NK cells

To avoid interpreting the development of iPSC-CAR-NK cells as a simple linear accumulation of engineering modules, this review adopts a multilevel analytical framework to organize the subsequent discussion. This framework is shown in Figure 3 and divides the clinical translation process of iPSC-CAR-NK cells into five interrelated layers: The product definition layer, effector execution layer, tissue delivery layer, host interaction layer, and translational implementation layer. These five layers correspond to key issues including manufacturing consistency, effector function, lesion-site delivery, in vivo persistence, and clinical implementation, respectively. These layers are not independent components; rather, together they determine whether iPSC-CAR-NK cells can be translated from engineering designs into cell products that are evaluable, manufacturable, and clinically applicable. Accordingly, the subsequent sections on manufacturing workflows, engineering optimization, disease applications, barriers to clinical translation, and future directions are organized around this framework.

Figure 3 Multilevel conceptual framework for the clinical translation of induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells. This figure summarizes the multilevel analytical model for the clinical translation of induced pluripotent stem cell-derived chimeric antigen receptor natural killer (iPSC-CAR-NK) cells proposed in this review. This framework is not a linear manufacturing workflow, but rather a conceptual integration model for understanding different limiting layers and their interrelationships during the transition of iPSC-CAR-NK cells from engineering design to clinical application. The framework includes the product definition layer, effector execution layer, tissue delivery layer, host interaction layer, and translational implementation layer, which correspond to product quality attributes and manufacturing consistency, target-cell clearance function, lesion-site delivery capacity, in vivo host interactions, and conditions for translational implementation, respectively. This framework emphasizes that the clinical translation of iPSC-CAR-NK cells should not be understood as a simple accumulation of engineering modules, but should instead be pursued through a verifiable balance among functional benefits, cellular fitness, safety, manufacturing consistency, and clinical implementability. iPSCs: Induced pluripotent stem cells; NK: Natural killer; CAR: Chimeric antigen receptor; ADCC: Antibody-dependent cellular cytotoxicity; B2M: Beta-2 microglobulin; HLA: Human leukocyte antigen; CMC: Chemistry, Manufacturing and Controls; QC: Quality control; FDA: Food and Drug Administration; EMA: European Medicines Agency.

First, the product definition layer focuses on whether iPSC-CAR-NK cells can be manufactured in a stable and reproducible manner as cell products with clearly defined quality attributes. This represents an important basis for distinguishing the iPSC platform from conventional primary CAR-NK cells. This layer involves the selection of starting iPSC clones, establishment of the master cell bank, stability of genetic edits, consistency of directed differentiation, preservation of function after cryopreservation and thawing, batch-to-batch consistency, and critical quality attributes that are amenable to regulatory evaluation. Only when these issues are adequately controlled can the standardization, scalability, and comparability assessment of iPSC-CAR-NK cells have a practical basis. At the conceptual level, the differences between iPSC-CAR-NK cells and conventional CAR-T or conventional CAR-NK platforms lie not only in the type or source of effector cells, but also in the logic of product generation, the timing of engineering, and the structure of translational risks. Conventional CAR-T cell therapy usually starts with patient- or donor-derived mature T cells, and its effector function mainly depends on CAR-mediated T-cell activation and adaptive immune effector functions. In contrast, conventional CAR-NK cell therapies commonly start with mature or near-mature NK cells derived from peripheral blood, cord blood, or NK cell lines; these cells retain effector axes involving NK-cell natural cytotoxicity, endogenous activating and inhibitory receptors, and ADCC. In contrast, the iPSC-CAR-NK platform starts with iPSCs that can be expanded clonally, banked for long-term use, and engineered at multiple loci at an early stage, followed by directed differentiation to generate NK effector cells. Therefore, iPSC-CAR-NK cells should be understood more as a cell-bank-driven modular product platform than merely as another source of conventional CAR-NK cells.

Second, the effector execution layer focuses on whether the cells can effectively recognize and eliminate target cells. This layer encompasses functional components and readouts, including the CAR antigen-recognition domain, hinge and transmembrane regions, NK-adapted intracellular signaling domains, the CD16-mediated ADCC axis, cytokine-support modules, degranulation, cytokine release, and serial killing capacity. This layer is directly related to the immediate effector function of iPSC-CAR-NK cells in specific target-antigen contexts and disease models. However, effector execution capacity alone cannot determine final clinical efficacy. If in vivo delivery is insufficient, local microenvironmental suppression is pronounced, or host immune clearance occurs too rapidly, enhanced in vitro cytotoxicity may not necessarily translate into durable in vivo benefit.

Third, the tissue delivery layer focuses on whether the infused cells can reach lesion sites and enter effective sites of action. Compared with solid tumors, physical and stromal barriers are generally less prominent in hematologic malignancies. In solid tumors, however, abnormal vasculature, stromal barriers, chemokine mismatch, hypoxia, and metabolic stress may all limit CAR-NK cell homing, extravasation, and infiltration into the tumor parenchyma. Therefore, chemokine receptor engineering, adaptation to stromal barriers, and local microenvironment-responsive strategies should be regarded as optimizations at the tissue delivery layer rather than simply as strategies for enhancing cytotoxicity.

Fourth, the host interaction layer focuses on whether the infused cells can maintain sufficient exposure in patients and avoid premature clearance. This layer involves allogeneic immune clearance mediated by host T cells and NK cells, human leukocyte antigen (HLA)-related engineering, beta-2 microglobulin (B2M) or CIITA editing, HLA-E/HLA-G expression, immune checkpoint regulation, and potential risks of immune escape. Related strategies may influence the in vivo persistence of engineered cells, the need for repeated dosing, and long-term safety. For example, reducing HLA expression may decrease recognition by host T cells, whereas B2M deficiency may also trigger a “missing-self” response by host NK cells. Expression of molecules such as HLA-E or HLA-G may reduce the risk of NK cell-mediated clearance, but long-term immune escape and safety issues still require further evaluation.

Fifth, the translational implementation layer focuses on whether the product can enter clinical development in a manner that is manufacturable, amenable to regulatory evaluation, and affordable. This layer includes the Chemistry, Manufacturing and Controls (CMC)/QC system, potency assays, release criteria, comparability assessment for process changes, long-term safety monitoring, manufacturing costs, cold-chain logistics, and regulatory acceptability. Even if an engineering module shows functional advantages in vitro or in animal models, its translational value still needs to be reassessed in relation to manufacturability and clinical implementability if it also increases manufacturing complexity, genetic-stability risks, or regulatory uncertainty.

Based on this framework, the core question for iPSC-CAR-NK cells should not be framed simply as “how to stack more functional modules”, but rather should be understood as how to balance manufacturing consistency, effector function, tissue delivery, persistence within the host, safety, and regulatory feasibility. Enhancing cytotoxicity does not necessarily improve therapeutic efficacy in solid tumors, particularly when tissue delivery remains insufficient and local TME-mediated suppression persists. Reducing immunogenicity may prolong in vivo persistence, but it may also introduce issues related to “missing-self” recognition, immune escape, or long-term safety. Cytokine-armoring strategies may enhance cellular fitness and in vivo persistence, but they may also increase cytokine-related toxicity and regulatory complexity. Therefore, the subsequent sections will use this multilevel framework to evaluate the trade-offs among functional benefits, cellular fitness, safety, manufacturing consistency, and clinical translatability associated with different strategies.

Within this framework, this review further distinguishes core biological bottlenecks, key translational limitations, and specific engineering interventions. Core biological bottlenecks mainly refer to biological issues that limit effective in vivo exposure, lesion-site delivery, target-cell coverage, persistence within the host, and safety of iPSC-CAR-NK cells. Key translational limitations include clinical development-related issues such as manufacturing consistency, QC, genetic stability, potency evaluation, regulatory pathways, cost, and accessibility. Specific engineering interventions are technical approaches developed around these bottlenecks and limitations, including CAR structural optimization, gene editing, chemokine-receptor and homing engineering, TME adaptation, cytokine support, logic-gated control, and safety switches. These three categories are interrelated but are not interchangeable. When evaluating the translational value of a given engineering module, it is necessary to consider not only whether it enhances in vitro function, but also whether it addresses the dominant limiting factors in the specific disease context and whether it can establish a verifiable balance among functional benefits, cellular fitness, safety, manufacturing consistency, and clinical implementability.

MANUFACTURING, PRODUCT ATTRIBUTES, AND PLATFORM POSITIONING OF IPSC-CAR-NK CELLS

From the perspective of the multilevel analytical framework, this section mainly corresponds to the product definition layer and addresses whether iPSC-CAR-NK cells can be manufactured in a stable and reproducible manner as cell products with clearly defined quality attributes. Therefore, this section not only discusses their manufacturing workflow and potential advantages over conventional CAR-NK cells, but also emphasizes that these advantages have practical translational significance only when clonal stability, differentiation consistency, preservation of function after cryopreservation and thawing, batch-to-batch consistency, and the QC system have been validated.

Manufacturing workflow and key technical steps

Clinical-grade iPSCs are usually generated from somatic cells through reprogramming. Peripheral blood cells and skin fibroblasts are both commonly used starting materials; in clinical translation settings, the selection of starting materials should also take into account collection feasibility, genetic background, reprogramming efficiency, and subsequent GMP cell-banking requirements[24-26]. To reduce the risks of insertional mutagenesis and genetic instability, non-integrating reprogramming strategies are usually prioritized in clinical translation settings, such as modified synthetic mRNA- or Sendai virus vector-based methods[25,27]. After candidate iPSC clones are obtained, they should be expanded, screened, and banked under current GMP conditions. This process usually uses chemically defined and feeder-free culture systems, with xenogeneic components minimized where possible, to support subsequent master cell bank establishment and comparability assessment of quality attributes[24,26,28]. Candidate clones also need to undergo pluripotency characterization, karyotype analysis or genomic-stability assessment, and biosafety testing, including testing for sterility, mycoplasma, and viral safety[24,26,29].

The differentiation of iPSCs into NK cells usually does not occur as a single-step process, but proceeds through sequential stages, including hematopoietic lineage induction, NK-lineage specification, and maturation and expansion. Most protocols first induce iPSCs to generate CD34+ hematopoietic progenitor cells or related hematopoietic precursors. Subsequently, with the support of cytokines such as IL-3, IL-7, IL-15, and stem cell factor, hematopoietic precursors further differentiate toward NK-cell precursors and mature NK cells. During the maturation stage, IL-15 often serves as a key factor supporting NK-cell activation, expansion, and functional maturation[30-33]. Different differentiation protocols, culture systems, and state of the starting pluripotent stem cells may affect the phenotypic maturity, CD16 expression, KIR repertoire, and cytotoxic function of the final iPSC-NK cells. Therefore, differentiation consistency itself should be regarded as an important quality attribute at the product definition layer[33,34]. iPSC-NK cells generated through differentiation usually express NK-cell-associated markers such as CD56 and NK group 2 member D (NKG2D). Some protocols can also generate cell populations with CD16 expression and ADCC potential. These cells may exhibit perforin/granzyme-mediated cytotoxicity and ADCC[30-32]. After CAR engineering, iPSC-derived NK cells can further acquire CAR-mediated antigen-specific killing capacity[35-37]. For example, in a study of MSLN.CAR-IL-15-engineered iPSC-NK cells, these cells showed signals of antitumor activity in solid tumor models and under TME-related stress conditions. However, such findings still require further validation across different targets and tumor types, and in more clinically relevant models[35].

CAR introduction and other genetic modifications can be performed either at the iPSC stage or at the post-differentiation NK-cell stage. For the iPSC platform, engineering at the iPSC stage facilitates subsequent clonal screening, genetic-stability assessment, and master cell bank establishment. In contrast, gene transfer at the post-differentiation NK-cell stage is closer to the conventional CAR-NK manufacturing workflow, but is usually constrained by NK-cell transduction efficiency and the limited expansion window[36,38-40]. Lentiviral vectors remain one of the commonly used gene-transfer methods for CAR-NK engineering, and approaches such as spinoculation can be used to improve transduction efficiency. In recent years, lentiviral packaging and delivery systems for clinical-grade CAR-NK manufacturing have also continued to be optimized[36,39,41]. In addition, electroporation and other non-viral delivery methods have been used in CAR-NK engineering studies to reduce manufacturing complexity and safety concerns associated with viral vectors[38]. After engineering, phenotypic indicators such as CAR expression, CD56, CD16, and NKG2D, together with functional readouts such as in vitro cytotoxicity, degranulation, and cytokine release, can be used to screen and evaluate CAR-positive cell populations with NK-cell functional characteristics[39]. The manufacturing workflow of iPSC-CAR-NK cells is shown in Figure 4.

Figure 4 Manufacturing workflow of induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells. The figure outlines the manufacturing workflow of induced pluripotent stem cell-derived chimeric antigen receptor natural killer (iPSC-CAR-NK) cells. Somatic cells, including adult human dermal fibroblasts and peripheral blood mononuclear cells, are first isolated and reprogrammed into iPSCs using synthetic mRNA or Sendai viral vectors. Candidate iPSCs are then cultured and expanded under current Good Manufacturing Practice-compliant, feeder-free, chemically defined culture conditions, followed by characterization and quality-control testing, including pluripotency validation, karyotype and genomic-stability analysis, sterility testing, mycoplasma testing, and viral detection. NK cell differentiation is initiated by CHIR99021 and bone morphogenetic protein 4 or by co-culture with OP9 stromal cells to induce cluster of differentiation 34-positive (CD34+) hematopoietic progenitor cells. These cells are further differentiated into NK precursor cells under stimulation with interleukin-3 (IL-3), IL-7, IL-15, and stem cell factor, followed by IL-15-mediated activation and expansion into mature NK cells with high expression of CD56, CD16, and natural killer group 2 member D. CAR introduction is performed using third-generation lentiviral vectors, with centrifugation-assisted infection to enhance transduction efficiency, or by non-viral electroporation. Finally, CAR-positive NK-cell populations are sorted and enriched by flow cytometry using fluorochrome-conjugated anti-CAR antibody staining combined with CD56 and CD16 staining. iPSCs: Induced pluripotent stem cells; NK: Natural killer; CAR: Chimeric antigen receptor; miRNA: MicroRNA; CD: Cluster of differentiation; IL: Interleukin; SCF: Stem cell factor; NKG2D: Natural killer group 2 member D.

Core features and potential advantages compared with conventional CAR-NK cells

A major feature of iPSC-CAR-NK cells is their distinct product-generation logic, which differs from that of conventional primary CAR-NK cells. Peripheral blood- or cord blood-derived CAR-NK cells are usually generated from donor-derived mature or near-mature NK cells, and their product quality is readily affected by donor variability, starting-cell composition, expansion efficiency, and functional status. In contrast, iPSC-CAR-NK cells can be generated from a screened single-clone-derived iPSC cell bank and differentiated into NK effector cells under standardized process conditions, which may theoretically facilitate improved product homogeneity, batch-to-batch consistency, and process comparability[18,24]. However, this advantage should not be assumed to be inherent. It still needs to be demonstrated through data on clonal stability, differentiation consistency, post-thaw functional preservation, batch-to-batch potency evaluation, and long-term QC data. From the perspective of production organization, once a fully characterized iPSC master cell bank has been established, the iPSC-CAR-NK platform may theoretically support large-scale production, long-term storage, and on-demand supply of off-the-shelf cell products, thereby reducing the uncertainty associated with repeated donor collection and potentially shortening patient waiting times[24,40,42]. However, no simple conclusion can yet be drawn as to whether this platform can substantially reduce the overall cost of treatment. Actual costs depend not only on production scale, but are also affected by factors such as the success rates of differentiation and expansion, GMP facility requirements, release testing requirements, cold-chain logistics, batch rejection rates, and real-world use scenarios[24,43]. Therefore, scalable manufacturing and cost control should be regarded as potential translational advantages of the iPSC-CAR-NK platform, rather than outcomes that have already been fully validated in clinical settings. The iPSC platform also provides a relatively clear product-development pathway for multiplex engineering. Compared with single-round engineering of mature NK cells, integrating CAR introduction, CD16 enhancement, cytokine-support modules, safety switches, or immunoregulatory modules at the iPSC stage facilitates clonal screening, genetic-stability assessment, and subsequent master cell bank establishment[14,15,18]. Recent original studies have also suggested that iPSC-CAR-NK cells can incorporate CARs, hnCD16, IL-15-related support modules, and other immunomodulatory designs, thereby forming multi-module engineered products[37]. However, such designs do not imply that adding more modules is necessarily better. The expression levels, signaling strength, cellular fitness, in vivo persistence, and safety of different constructs still need to be compared within the same model systems. Therefore, the engineering advantage of iPSC-CAR-NK cells should not be understood as simply allowing more modules to be added, but rather as the suitability of this platform for module combination, screening, and risk assessment within a well-controlled product platform. From the perspective of disease applicability, the iPSC-CAR-NK platform has begun to expand from cancer therapy toward immune-mediated diseases, although this direction remains at a very early stage. At present, direct evidence for the use of iPSC-CAR-NK cells in autoimmune diseases mainly comes from a single-case compassionate-use report in systemic sclerosis. In this case, signals of B-cell depletion and clinical improvement were observed after treatment with iPSC-derived CD19/BCMA CAR-NK cells. However, this result can only suggest the early feasibility of this strategy in this individual patient; it should not be equated with disease-level validation of efficacy in systemic sclerosis, nor does it support broad extrapolation to other autoimmune diseases[22]. Regarding hematologic malignancies, a phase I FIH trial published in The Lancet reported that FT596, an iPSC-derived CD19 CAR-NK cell product, was generally well tolerated in patients with relapsed or refractory B-cell lymphoma, with signals of clinical response observed in some patients[23]. Related commentaries have suggested that this study provides important early evidence for the clinical development of iPSC-CAR-NK cell products, but further validation in larger cohorts and with longer follow-up is still required[44].

Comparison with other off-the-shelf cellular immunotherapy platforms

Although iPSC-CAR-NK cells have theoretical advantages in standardized manufacturing, batch-to-batch consistency, and multiplex engineering, these advantages do not imply that they have already demonstrated overall clinical superiority over conventional CAR-T cells, conventional CAR-NK cells, or other off-the-shelf cellular immunotherapy platforms[43,45]. Compared with conventional CAR-T cells, the core differences between iPSC-CAR-NK cells and conventional CAR-T cells lie not merely in the replacement of T cells with NK cells, but also in changes in both immune effector logic and product-generation logic. In terms of effector mechanisms, iPSC-CAR-NK cells do not rely on endogenous αβTCR-major histocompatibility complex (MHC)-restricted recognition and retain effector axes involving NK-cell natural cytotoxicity, endogenous activating and inhibitory receptor-mediated regulation, and CD16-mediated ADCC[46]. In terms of product generation, this platform represents a shift from a manufacturing model based on patient-specific or donor-derived mature effector cells to a standardized and expandable product-development model that is based on a single-clone-derived iPSC cell bank and amenable to modular engineering[15,43,47].

Compared with conventional CAR-NK cells, iPSC-CAR-NK cells are not fundamentally distinct at the level of NK-cell effector mechanisms, but instead show platform-level differences in starting-cell source, the timing of engineering, cell-banking strategy, and manufacturability[45,47]. Peripheral blood-derived CAR-NK cells, cord blood-derived CAR-NK cells, and NK-92 cell line-derived CAR-NK cells all use mature or near-mature NK cells as the primary starting material for engineering. In contrast, iPSC-CAR-NK engineering can be performed upstream at the iPSC stage, and relatively consistent cell products can be generated through clonal screening, master cell bank establishment, and standardized differentiation workflows. Therefore, iPSC-CAR-NK cells should be understood more as a “cell-bank-driven modular product platform” rather than merely another source of conventional CAR-NK cells.

Peripheral blood-derived CAR-NK cells, cord blood-derived CAR-NK cells, NK-92 cell line-derived CAR-NK cells, allogeneic CAR-T cells, and γδ T-cell/CAR-γδ T-cell therapies each have their own technical basis, clinical development experience, and indication-specific positioning[8-10,48]. Therefore, the value of iPSC-CAR-NK cells should be understood more as a scalable platform option amenable to standardization and modular engineering rather than as a simple replacement for other platforms. At the same time, this platform-based nature also introduces challenges distinct from those associated with primary effector-cell platforms, including the genetic stability of iPSC clones, the risk of residual undifferentiated cells, the safety of multiplex editing, differentiation consistency, long-term QC, and regulatory complexity. Accordingly, evaluation of the iPSC-CAR-NK platform should take into account not only its potential for platform-based manufacturing and modular engineering but also unresolved issues related to in vivo persistence, host immune clearance, manufacturing complexity, QC, and regulatory feasibility.

Compared with peripheral blood- and cord blood-derived CAR-NK cells, iPSC-CAR-NK cells may theoretically be more conducive to achieving standardized manufacturing and batch-to-batch consistency through a master cell bank, thereby reducing the impact of donor variability on product phenotype and function[8,18,45,48-50]. Compared with NK-92 cell line-derived CAR-NK cells, iPSC-CAR-NK cells can avoid some limitations associated with tumor cell line origin and pre-infusion irradiation, while being more suitable for integrating multiplex engineering modules, such as CAR introduction, cytokine support, immune-evasion modulation, and safety switches, at an early cellular stage[9,10,48-50]. Compared with allogeneic CAR-T cells, iPSC-CAR-NK cells do not rely on αβTCR-mediated target-cell recognition and may theoretically reduce the risk of T cell receptor-mediated GvHD. However, as allogeneic cell products, they may still be subject to host immune clearance and face challenges such as insufficient in vivo persistence[51]. Compared with γδ T-cell and CAR-γδ T-cell therapies, iPSC-CAR-NK cells show platform-level potential in clonal cell banking, standardized differentiation, and multiplex engineering. In contrast, γδ T cells are characterized by MHC-unrestricted recognition, tissue homing, and innate-like immune responses, although their subset heterogeneity, expansion stability, and engineering consistency still require further optimization[52].

However, iPSC-CAR-NK cells still face several challenges, including complex differentiation workflows, relatively high manufacturing costs, risks related to genetic stability and residual undifferentiated cells, insufficient long-term in vivo persistence, and limited clinical comparative evidence[18,48-50]. Therefore, the advantages of iPSC-CAR-NK cells in scalable manufacturing, batch-to-batch consistency, and cost reduction should still be regarded as potential advantages rather than conclusions that have already been fully validated in clinical settings. Table 2 compares iPSC-CAR-NK cells with major competing off-the-shelf cellular immunotherapy platforms.

Table 2 Comparison of induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells with major competing off-the-shelf cellular immunotherapy platforms.

Platform	Main source/product form	Key features/potential advantages	Major limitations	Key differences compared with iPSC-derived CAR-NK cells	Maturity of clinical translation
Peripheral blood-derived CAR-NK cells	Peripheral blood NK cells from healthy donors or patients, followed by ex vivo expansion and CAR engineering	Relatively strong natural cytotoxic activity; mature cell source; relatively preserved primary NK-cell effector functions; an existing clinical research basis	Limited frequency of NK cells in the starting material; expansion efficiency and functional status are affected by donor variability; donor-to-donor differences may lead to insufficient batch-to-batch consistency; large-scale standardized manufacturing remains challenging	Cell function is closer to that of primary NK cells, but standardization, scalability, and batch-to-batch consistency are generally weaker than those of the iPSC-derived platform	Early clinical exploration has been reported
Cord blood-derived CAR-NK cells	Cord blood NK cells or cord blood-derived NK precursor cells, followed by expansion and CAR engineering	Relatively low immunogenicity; suitable for allogeneic application; relatively standardized sample source; potential for off-the-shelf development	Limited cell numbers from a single cord blood unit; ex vivo expansion and maturation still require optimization; inter-sample variability may affect manufacturing consistency; limited in vivo persistence	Existing clinical development basis, but difficult to achieve long-term, stable, large-scale production comparable to that enabled by a single-clone-derived iPSC	Early clinical exploration
NK-92 cell line-derived CAR-NK cells	NK-92 cell line engineered with a CAR	Easy to expand; relatively stable manufacturing workflow; suitable for early process development and mechanistic validation	Cell line origin raises safety concerns; pre-infusion irradiation is usually required, limiting in vivo proliferation and persistence; intrinsic lack of CD16 results in insufficient ADCC capacity	Standardized expansion is relatively straightforward, but in vivo persistence and clinical applicability are limited by pre-infusion irradiation and cell line-related properties	Preclinical research and early clinical exploration
iPSC-derived CAR-NK cells	Single-clone-derived iPSCs subjected to CAR introduction and multiplex engineering, followed by differentiation into NK cells	Master cell banks can be established; theoretically favorable for scalable and standardized manufacturing with improved batch-to-batch consistency; amenable to multiplex gene editing; suitable for modular engineering	Complex differentiation workflow; high manufacturing cost and quality-control requirements; genetic stability, residual undifferentiated cells, and risks associated with multiplex editing must be controlled; long-term in vivo persistence and real-world cost advantages still require validation	Compared with primary NK and NK-92 platforms, this platform has development potential in platform-based manufacturing and multiplex engineering, but its comparative clinical advantages still require further validation	Early-stage clinical validation
Allogeneic CAR-T cells	Healthy donor-derived T cells engineered with a CAR, often requiring TCR- and HLA-related gene editing	Strong antigen-specific cytotoxic capacity of T cells; relatively mature CAR-T development experience; potential for off-the-shelf development	Need to address TCR-mediated GvHD, host anti-graft responses, immune rejection, and gene-editing safety; CRS/ICANS may still occur; multiplex editing increases regulatory and quality-control complexity	Compared with iPSC-derived CAR-NK cells, CAR-T cell therapy has more extensive clinical development experience, but allogeneic T cell-related immune safety concerns and editing requirements are greater	Multiple early clinical studies are ongoing
γδ T-cell and CAR-γδ T-cell therapies	Peripheral blood- or tissue-derived γδ T cells, followed by ex vivo expansion or CAR engineering	MHC-unrestricted recognition; both innate-like and adaptive immune features; theoretically lower risk of GvHD; ability to recognize stress-associated antigens	Marked subset heterogeneity of γδ T cells; expansion stability and engineering efficiency require optimization; functional differences among subsets are substantial; consistency of clinical efficacy still requires validation	Compared with iPSC-derived CAR-NK cells, γδ T cells have T cell-like cytotoxicity and tissue homing- or tissue residency-related features, but product standardization and batch-to-batch consistency remain challenging	Preclinical to early clinical exploration

The “key features/potential advantages” listed in this table include theoretical platform advantages and technical features reported in early studies; they should not be equated with clinical superiority demonstrated by prospective randomized clinical trials. Some platforms should be understood as off-the-shelf or potentially off-the-shelf cellular immunotherapy strategies. The information in this table was synthesized from studies and reviews related to peripheral blood- and cord blood-derived chimeric antigen receptor (CAR)-natural killer (NK) cells, NK-92 cell line-derived CAR-NK cells, induced pluripotent stem cell (iPSC)-derived CAR-NK cells, allogeneic CAR-T cells, and γδ T-cell and CAR-γδ T-cell therapies[8-10,18,22-24,40,42,43,45,47,48,50-52]. The potential advantages of iPSC-derived CAR-NK cells in scalable manufacturing, batch-to-batch consistency, and cost control still require further validation through Good Manufacturing Practice manufacturing data, clinical studies, and real-world evidence. iPSCs: Induced pluripotent stem cells; CAR: Chimeric antigen receptor; NK: Natural killer; ADCC: Antibody-dependent cellular cytotoxicity; CRS: Cytokine release syndrome; ICANS: Immune effector cell-associated neurotoxicity syndrome; TCR: T cell receptor; HLA: Human leukocyte antigen; GvHD: Graft-vs-host disease.

ENGINEERING OPTIMIZATION STRATEGIES FOR IPSC-CAR-NK CELLS

This section primarily maps onto the effector execution layer, tissue delivery layer, and host interaction layer of the multilevel framework. To avoid interpreting engineering strategies as isolated or indefinitely stackable functional modules, this review further classifies the engineering optimization of iPSC-CAR-NK cells into four interrelated mechanistic layers. First, the recognition and activation layer comprises the antigen-recognition domain, hinge and transmembrane regions, and NK-adapted intracellular signaling domains; this layer determines whether the cells can recognize target cells and initiate NK-cell effector programs. Second, the effector amplification and functional durability layer comprises ADCC enhancement, cytokine support, and optimization of serial killing capacity; this layer is intended to enhance effector activity and sustain functional output over time. Third, the in vivo exposure and tissue adaptation layer comprises in vivo persistence, chemotaxis and homing regulation, TME adaptation, and evasion of host immune clearance; this layer addresses whether the cells can reach lesion sites, survive locally, and exert sustained functions. Fourth, the control and safety layer comprises synNotch systems, logic-gated control, self-regulating feedback circuits, and safety switches; this layer is used to control the timing, spatial localization, intensity, and termination of cellular activation. It should be emphasized that single-cell sequencing, CRISPR screening, and computational modeling are more appropriately understood as data-driven tools for identifying limiting nodes and guiding engineering decisions, rather than as functional enhancement modules parallel to the effector modules described above. Therefore, this section does not simply list different engineering strategies side by side, but instead focuses on their functional positioning, applicable scenarios, and potential trade-offs within different mechanistic layers.

NK cell-specific optimization of CAR structure

CARs usually consist of three functional modules: The antigen-recognition domain, the hinge and transmembrane regions, and the intracellular signaling domain. These modules influence targeting specificity, the spatial distance and geometry of receptor engagement, and the strength of NK-cell activation, respectively (Figure 5)[53].

Figure 5 Schematic diagram of chimeric antigen receptor natural killer cell structural optimization. The figure illustrates key structural modules involved in the optimization of chimeric antigen receptor natural killer (CAR-NK) cells. The antigen-recognition domain can be designed using different binding modules, including a single-chain variable fragment, a binding protein targeting peptide-major histocompatibility complex, or a nanobody, also known as the variable domain of a heavy-chain-only antibody. These antigen-binding modules differ in molecular size, binding properties, hydrophobicity, and potential immunogenicity. The extracellular recognition module is connected to intracellular signaling components through linker, hinge, and transmembrane regions, such as a cluster of differentiation 8 alpha (CD8α) hinge domain and a natural killer group 2 member D transmembrane domain. NK cell-associated surface molecules, including cluster of differentiation 56 (CD56) and CD16, and NK-adapted signaling domains, including DNAX-activating protein 10, 2B4, also known as CD244, and CD3ζ, contribute to CAR-NK cell activation. After target recognition, activated CAR-NK cells mediate cytotoxic attack through effector molecules such as perforin and granzymes. scFv: Single-chain variable fragment; pMHC: Peptide-major histocompatibility complex; CD: Cluster of differentiation; DAP10: DNAX-activating protein 10; NKG2D: Natural killer group 2 member D.

The design of intracellular signaling domains is critical for CAR-NK cell activity. DNAX-activating protein 10 (DAP10), DAP12, and other signaling adaptor proteins are common endogenous components in NK cells. Mechanistic studies have shown that these molecules can assemble into distinct but functionally cooperative receptor complexes, thereby providing an immunological rationale for incorporating NK-cell-intrinsic signaling elements into CAR constructs[54]. Some studies have evaluated multiple complementary signaling combinations and reported that the DAP10/2B4/CD3ζ configuration showed enhanced functional activity in specific experimental systems[55]. However, this does not mean that CAR structures commonly used in CAR-T cell therapy are ineffective in the NK-cell context. A phase I study reported in 2025 evaluated CD19-BBζ CAR-NK cells, and the results provided early signals that this type of construct may be clinically feasible in the NK-cell context and have a manageable safety profile, with relatively durable remissions also observed in some patients[56].

The hinge and transmembrane regions can influence cell-surface CAR expression, epitope accessibility, and the geometry of the immunological synapse. Current evidence supports selecting hinge length according to the membrane-proximal or membrane-distal location of the target epitope, rather than assuming that either a long or short hinge is intrinsically optimal[53]. Among publicly reported iPSC-NK constructs, a design combination incorporating a CD8α hinge region, an NKG2D transmembrane domain, a 2B4 intracellular domain, and a CD3ζ signaling domain has been used, suggesting that these modules may provide useful reference points for engineering design on the iPSC platform[35].

The antigen-recognition domain is still most commonly based on single-chain variable fragments (scFvs), but its binding affinity requires careful optimization. Previous studies have shown that changes in both affinity and epitope selection can affect the activation pattern and functional performance of CAR-NK cells. Therefore, the design objective should be to balance cytotoxic potency, target specificity, and cellular fitness, rather than simply pursuing high affinity[57]. In addition to conventional scFvs, nanobodies (VHHs) have also been used in the design of antigen-recognition domains for CAR-NK cells. For example, anti-MICA VHH-based CAR-NK cells can recognize and kill MICA-positive tumor cells, suggesting that VHHs may serve as an alternative format for CAR-NK recognition modules[58]. In addition, computationally designed binding proteins targeting peptide-MHC (pMHC) have been incorporated into CAR frameworks and can induce peptide-specific immune cell activation[59]. However, CARs incorporating pMHC-binding proteins should currently be viewed primarily as an instructive design concept for alternative recognition domains. Their applicability, specificity, and safety in CAR-NK systems, particularly in iPSC-CAR-NK systems, still require further validation. Overall, VHHs and pMHC-binding proteins represent two alternative approaches beyond conventional scFv-based recognition domains, but their advantages still need to be quantitatively compared under conditions matched for the target antigen, effector-cell type, and experimental system.

Gene editing-mediated functional enhancement

Gene editing is not merely a tool for enhancing cytotoxicity; rather, it can modulate iPSC-CAR-NK cell functions across multiple mechanistic layers. Its effects can be broadly divided into three categories: First, reducing inhibitory-signaling thresholds; second, strengthening the effector-amplification axis; and third, improving in vivo exposure and host interaction. It should be noted that these three types of edits are not “functional enhancement” modules operating at the same mechanistic level. Rather, these categories act, respectively, on distinct limiting nodes related to activation thresholds, effector amplification, and in vivo exposure, and their combinatorial value depends on the dominant bottleneck in a given disease context. Studies have shown that NKG2A (KLRC1) knockout can enhance NK-cell activation and overcome HLA-E-mediated inhibition[60]. With respect to the effector-amplification axis, augmentation of ADCC can promote synergy among multiple effector mechanisms. Engineering iPSCs before their differentiation into iNK cells to express a high-affinity CD16a variant resistant to ADAM17-mediated cleavage (hnCD16) can improve antibody-mediated tumor-cell killing[61]. Publicly available abstract-level evidence further suggests that the iPSC platform can be used to combine CAR, hnCD16, and an IL-15/IL-15Rα fusion protein (IL-15RF), thereby simultaneously enhancing targeting capacity, ADCC, and cytokine-driven autocrine support[62]. Strategies aimed at improving in vivo exposure and host interaction mainly include reducing the risk of host immune clearance and optimizing tissue distribution. B2M editing can reduce HLA class I expression, thereby decreasing host T-cell-mediated rejection; however, B2M deficiency may also trigger a host NK-cell-mediated “missing-self” response. Studies have shown that CD54 and CD58 knockout can reduce the risk of host NK-cell-mediated clearance of such allogeneic cell products[63]. In terms of migration and homing, C-X-C receptor 4 (CXCR4) is a representative example. Studies have shown that low CXCR4 expression in iPSC-NK cells limits their bone marrow infiltration. Regulated expression of CXCR4 during iNK-cell maturation can improve bone marrow tropism and enhance therapeutic efficacy in acute myeloid leukemia (AML) models; these findings also suggest the feasibility of further integration with CAR engineering[64].

Engineering adaptation to the TME

The constraints imposed by the TME on iPSC-CAR-NK cells should not be reduced to a single “immunosuppressive environment”. Rather, they should be further dissected into interwoven mechanistic layers, including immunosuppressive cytokines, metabolic-suppression axes, inhibitory receptor-ligand signaling, tissue delivery barriers, and remodeling of the antigenic landscape. These mechanisms may differentially affect CAR-NK cell homing, local persistence, activation thresholds, degranulation, cytokine release, and serial killing capacity. Therefore, TME-adaptation strategies should be understood as engineering decisions at the in vivo exposure and tissue adaptation layer, rather than simply as effector-enhancement modules[65]. First, the transforming growth factor-β (TGF-β)/TGFBR2/SMAD signaling axis is currently one of the TME-associated suppressive pathways for which relatively direct evidence is available in iPSC-derived NK-cell or CAR-NK-cell systems. TGF-β can suppress NK-cell activation, cytotoxic-molecule expression, and effector-cytokine production, and may also affect functions related to activating receptors such as NKG2D and DNAM-1. In hepatocellular carcinoma (HCC) models, iPSC-NK cells that only express anti-GPC3 or anti-AFP CARs may still be susceptible to TGF-β-mediated suppression. In contrast, TGFBR2 knockout or expression of a dominant-negative TGF-β receptor can render iPSC-NK cells resistant to TGF-β-mediated suppression and enhance their anti-HCC activity[66]. These findings indicate that TGF-β pathway adaptation is not a general strategy for “enhancing cytotoxicity”, but rather a targeted engineering intervention against local TME-mediated suppressive signaling. Second, metabolic suppression in the solid-tumor TME may also limit CAR-NK cell function through specific pathways. Hypoxia can alter the metabolic states of tumor and immune cells through HIF-related pathways. CD39/CD73-mediated ATP catabolism can promote adenosine accumulation and suppress NK-cell activation through A2A receptor signaling. Lactate accumulation, an acidic microenvironment, nutrient competition, and the IDO-kynurenine-AhR pathway may also impair NK-cell metabolic fitness, degranulation, and cytokine secretion[67,68]. Therefore, strategies such as adenosine clearance, inhibition of A2A receptor signaling, and enhancing hypoxia resistance or metabolic adaptability can be mechanistically categorized as metabolic-adaptation strategies. However, whether these strategies can consistently enhance the in vivo efficacy of iPSC-CAR-NK cells in solid tumors still requires direct validation in CAR-NK-specific systems and more clinically relevant models. Third, the TME can also affect CAR-NK cell recognition and sustained effector function through inhibitory receptor-ligand axes and changes in the antigenic landscape. TME components, including tumor cells, myeloid-derived suppressor cells, regulatory T cells, tumor-associated macrophages, and cancer-associated fibroblasts, can raise NK-cell activation thresholds through inhibitory signals such as programmed cell death ligand 1/programmed cell death protein 1 (PD-L1/PD-1), HLA-E/NKG2A, and CD155/TIGIT[54,69]. Recent reviews on emerging immune targets and signaling pathways in tumor immunotherapy have also emphasized that immunotherapy resistance and heterogeneity in therapeutic efficacy are often associated with the combined effects of dysregulated immune surveillance, tumor immune escape, checkpoint signaling, innate immune regulation, and inflammation-related pathways. Therefore, TME-adaptation strategies for iPSC-CAR-NK cells should, whenever possible, be grounded in specific immune signaling networks and functional validation, rather than remaining at the level of a generalized description of “immunosuppression”[69]. In addition, antigen heterogeneity, target-antigen downregulation, and changes in stress-ligand expression may weaken a single CAR-recognition axis. A study of MSLN-CAR-IL-15 iNK cells further reported that single-cell transcriptomic analysis identified increased MHC expression and decreased MSLN expression in a TGF-β-associated tumor-cell population after treatment, suggesting that TME remodeling may be accompanied by changes in the antigenic landscape and immune-recognition state and may contribute to immune escape[35]. Based on these mechanisms, future TME-adaptation strategies should integrate single-cell sequencing, spatial omics, and functional validation to determine whether the dominant limitation arises from TGF-β-mediated suppression, the adenosine metabolic axis, hypoxia/metabolic stress, checkpoint signaling, insufficient chemotaxis and infiltration, or antigen escape. Corresponding strategies, such as TGF-β resistance, metabolic adaptation, chemotaxis and homing engineering, multi-target recognition, or logic-gated design, can then be selected.

Single-cell, multi-omics, and AI-assisted precision optimization: From data to engineering decisions

During the development of iPSC-CAR-NK cells, single-cell sequencing, multi-omics analysis, and AI-assisted modeling should not be understood merely as tools for describing cellular states or displaying cellular heterogeneity. Rather, they should be incorporated into a closed-loop optimization pathway of “data-insight-engineering decision-making-validation”. Reviews on cross-cancer multi-omics integration suggest that integrated analysis of multilayered data, including genomic, transcriptomic, epigenetic, and proteomic data, can help identify tumor driver genes, shared oncogenic pathways, and potentially actionable nodes, thereby providing a data-driven basis for target screening, patient stratification, and optimization of therapeutic strategies. Therefore, in the development of iPSC-CAR-NK cells, the value of multi-omics analysis should not be limited to describing the molecular features of tumors. Instead, it should be further used to generate testable engineering hypotheses, such as selecting more appropriate target-antigen combinations, identifying tumor-intrinsic pathways associated with immune escape, or prioritizing key limiting factors that need to be addressed through CAR structural optimization, chemotaxis and homing engineering, TME adaptation, or logic-gated modules[70]. Specifically, single-cell transcriptome sequencing, single-cell chromatin-accessibility analysis, CITE-seq, spatial omics, functional genomics screening, and clinical follow-up data can serve as multilevel data input. These data can further reveal key insights into differentiation maturity status, functional NK-cell subsets, inhibitory signaling pathways, patterns of antigen escape, TME-induced states of functional suppression, and post-infusion in vivo dynamics[71-73]. Functional genomics screening and AI-assisted modeling can further be used to identify candidate regulatory nodes and provide guidance for prioritizing engineering designs[74,75]. Subsequently, these insights can be translated into specific engineering decisions, such as optimizing the differentiation process from iPSCs to NK cells, selecting more appropriate CAR signaling domains, identifying candidate nodes that may require knockout, enhancement, or regulation, or incorporating modules for chemotaxis and homing, TME adaptation, and logic-gated control. Finally, these designs still need to be validated using readouts such as in vitro cytotoxicity, serial killing, degranulation, cytokine release, in vivo distribution, tumor infiltration, safety, cellular fitness, and batch-to-batch consistency, followed by iterative optimization based on the validation results. Therefore, the core value of single-cell, multi-omics, and AI-assisted strategies does not lie in generating more data per se, but in converting complex data into testable and actionable engineering design hypotheses that can be fed back into subsequent optimization. Along this closed-loop pathway, their applications can be organized into at least three major aspects. First, in the characterization of differentiation processes, different iPSC-to-NK differentiation routes may lead to different degrees of functional maturation. Studies comparing iNK cells generated using two differentiation strategies have identified differences in degranulation, granzyme B and interferon-γ secretion, and short-term cytotoxic activity, suggesting that the choice of differentiation protocol can directly influence the final functional phenotype[76]. Second, in the context of engineering optimization, single-cell data can be used to identify cell states that more closely resemble highly functional NK-cell phenotypes and to further infer regulatory nodes that may be amenable to engineering intervention. Examples consistent with this strategy include genome-wide CRISPR screens in primary NK cells, which have identified key negative regulators such as MED12, ARIH2, and CCNC, as well as studies in iPSC-derived CAR-iNK cells that used single-cell transcriptomic analysis to characterize state changes in both tumor cells and tumor-infiltrating iNK cells[35,74]. Third, in dynamic clinical monitoring, the CAR-T field has already established relatively well-developed approaches for post-infusion single-cell tracking, which can be used to identify expression signatures associated with activation, exhaustion, and immune escape. However, the application of this framework to CAR-NK cells still requires further validation through cohort-level evidence[77].

Therefore, future data-driven strategies in the development of iPSC-CAR-NK cells should move beyond merely describing cellular heterogeneity and toward guiding engineering decisions. Single-cell and multi-omics data generated during differentiation can be used to define more stable differentiation windows and quality-assessment metrics. Omics data from tumor co-culture systems or in vivo models can be used to identify nodes associated with TME-mediated suppression, antigen escape, and functional attenuation. Functional genomics screening and AI-assisted modeling can further be used to prioritize candidate targets, predict risks associated with module combinations, and generate testable engineering design hypotheses. Only when these data can be translated into explicit design choices and reproducible functional validation results will their translational value in iPSC-CAR-NK platform development become more clearly defined.

Synthetic biology-mediated programmable regulation

Synthetic biology-based strategies are mainly used to enhance the controllability of iPSC-CAR-NK cell activation and to provide engineered cells with inducible safety-switch mechanisms. With regard to logic-gated regulation, synNotch is a representative strategy. In this system, recognition of a priming antigen induces the subsequent expression of a second receptor or effector module, thereby enabling combinatorial antigen recognition and spatially restricted activation[78]. Preclinical studies related to combinatorial gating have also been conducted in NK-cell models. For example, in a colorectal cancer model, a strategy designed to use HER2-triggered synNotch signaling to induce CEA-CAR expression showed potential for enhancing targeting specificity[79]. In addition, clinical trial registry data indicate that NCT06186401 is an ongoing phase I clinical trial of synNotch-CAR T cells. In this study, EGFRvIII-activated synNotch signaling is intended to induce the expression of a CAR targeting EphA2/IL-13Rα2. Although this study does not involve iPSC-CAR-NK cells, it nevertheless suggests that the clinical translation of this design paradigm is being explored. In terms of safety control, iCasp9 is one of the more extensively validated suicide switches. Following administration of a small-molecule dimerizer, iCasp9 can rapidly induce apoptosis of engineered cells and facilitate their clearance, thereby providing a risk-control approach for reducing cellular exposure in the event of severe toxicity or uncontrolled cellular activity[80]. Published human studies have also characterized the pharmacokinetic properties of the dimerizing agent AP1903 (rimiducid) and supported its safety profile[81]. However, in iPSC-CAR-NK cells, protective switch designs that use normal-tissue markers as input signals remain at an early proof-of-concept stage, and substantial further research is still needed to demonstrate their feasibility.

Self-regulating feedback circuits

A common strategy in synthetic biology is the “sense-and-respond” design, in which cells first detect specific input signals and then generate effector outputs in a conditional manner. In CAR-NK cells, the most relevant input signals are usually not cytokine concentrations, but rather tumor antigens or other tumor-localized signaling cues. The outputs typically include locally restricted cytokine expression or enhanced cytokine-related signaling programs[82]. In the CAR-T field, this strategy has been supported by original studies. A representative example is the use of a tumor-specific synNotch circuit to locally induce IL-2 expression within the TME. Studies in CAR-T cells suggest that, in specific models, this type of design can enhance cellular infiltration and expansion and may reduce the risk of toxicity associated with systemic cytokine exposure[83]. Other cytokine-engineering strategies in CAR-T cells include the use of receptor engineering to convert inhibitory cytokine signals, such as IL-4- or TGF-β-mediated signaling, into stimulatory outputs, as well as the incorporation of conditional cytokine-expression modules to achieve local immune modulation[84]. However, it should be emphasized that, in iPSC-CAR-NK cells, these strategies still largely remain transferable design blueprints rather than validated strategies in this system. Their practical value must ultimately be validated through direct experiments in iPSC-CAR-NK systems, with particular attention to three key questions: Whether these strategies can improve in vivo persistence, whether they may increase treatment-related toxicity, and whether they may generate new selective pressures that promote immune escape[82]. From an engineering perspective, two strategies appear to be more readily translatable. The first is cytokine armoring combined with conditional expression to restrict its effects as much as possible to the TME[84]. The second is the reprogramming of cytokine-receptor signaling to convert inhibitory cytokine signals into stimulatory signals[82]. However, the safety window of these strategies still needs to be clarified through further research.

Taken together, the engineering optimization of iPSC-CAR-NK cells should be understood as a form of multilevel systems regulation rather than as the simple accumulation of individual functional modules. CAR structural design primarily addresses recognition and activation; ADCC enhancement and cytokine support mainly operate at the level of effector amplification and functional durability; TME-adaptation strategies and the regulation of chemotaxis and homing are primarily directed toward tissue delivery and maintenance of local function; HLA-related engineering and immune-evasion strategies mainly influence host interaction and in vivo exposure; and synthetic biology circuits and safety switches are mainly used to improve activation selectivity and risk control. Different layers may complement one another, but their combination may also introduce redundancy, impaired cellular fitness, increased safety risks, or greater manufacturing complexity. Therefore, future engineering design should shift from “increasing the number of modules” to “layered combination guided by dominant bottlenecks”.

Building on the mechanistic layers described above, the relationships among engineering modules can be further classified into three types: Cross-layer synergy, intra-layer redundancy, and cross-layer conflict. First, cross-layer synergy usually occurs when different modules act on distinct limiting nodes. For example, CAR structural optimization primarily enhances antigen recognition and initial activation; hnCD16 can complement the ADCC effector axis when combined with tumor-targeting antibodies; IL-15-related modules may support functional durability; and chemotaxis and homing regulation or TME-adaptation strategies mainly improve tissue delivery and maintenance of local function. Relevant studies on multi-module engineering and receptor engineering provide preclinical support for this type of combinatorial strategy[65,85]. Early clinical studies of products such as FT596 suggest that multi-module iPSC-CAR-NK products have shown a degree of feasibility for clinical development, but they do not yet demonstrate the independent therapeutic contribution of each module or clearly define the synergistic relationships among modules[23]. Studies related to the solid-tumor TME also suggest that tissue delivery and local microenvironmental adaptation often need to be considered together with effector-enhancement strategies[65]. Second, intra-layer redundancy may occur when multiple modules are directed toward similar functional outputs. For example, strongly activating CAR signaling, sustained cytokine support, and immune-checkpoint knockout may all increase cellular activation. If the dominant limiting node has not been clearly defined, repeatedly enhancing the same effector axis may increase the risks of impaired cellular fitness, excessive activation, or greater manufacturing complexity[86]. Third, cross-layer conflicts may occur when an engineering design improves one limiting node while increasing risks at another layer. For example, HLA-related engineering may reduce host T-cell-mediated recognition, but HLA-I downregulation or loss may also enhance host NK-cell-mediated “missing-self” recognition. IL-15-related armoring strategies may improve in vivo persistence, but they may also increase cytokine-related toxicity and the complexity of regulatory evaluation. Incorporating safety switches may improve risk control, but it may also increase construct complexity and the difficulty of potency evaluation[87,88]. Therefore, the combinatorial engineering design of iPSC-CAR-NK cells should follow a dominant-bottleneck-guided design principle that explicitly weighs potential trade-offs, rather than using the number of engineered modules as the primary evaluation criterion. More engineered modules do not necessarily lead to stronger in vivo efficacy or greater translational value. An added module is more likely to have a sound rationale for inclusion in a combinatorial design only when it targets a key limiting node in a specific disease context and when its expected or demonstrable functional benefit outweighs potential costs, including impaired cellular fitness, increased safety risks, increased construct complexity, increased manufacturing complexity, increased difficulty in maintaining batch-to-batch consistency, and greater regulatory-evaluation complexity. Therefore, future multi-module engineering of iPSC-CAR-NK cells should shift from “functional stacking” toward systems optimization characterized by bottleneck matching, layered complementarity, clearly defined marginal benefit, and manageable risk.

THERAPEUTIC APPLICATIONS AND EVIDENCE LEVELS OF IPSC-CAR-NK CELLS IN DISEASE TREATMENT

It should be emphasized that the overall clinical translation of iPSC-CAR-NK cells remains at an early stage, and the level of evidence differs substantially across disease contexts. Early clinical evidence has emerged for some hematological malignancies, whereas most solid-tumor applications are still mainly supported by in vitro experiments, organoid models, or animal studies. Data on products developed for multiple myeloma (MM) remain limited to early study results or interim disclosures, and evidence in autoimmune diseases is currently largely limited to a single compassionate-use case report. Therefore, the findings discussed in this section should be interpreted in light of the evidence type, study design, sample size, and follow-up duration. Existing data are more appropriately viewed as signals of safety, feasibility, and preliminary activity, rather than as established evidence of clinical efficacy. The long-term in vivo persistence of iPSC-CAR-NK cells, durability of therapeutic effects, comparative effectiveness relative to existing treatment strategies, repeat-dosing regimens, and commercial feasibility still require further validation. Table 3 summarizes the evidence levels and translational maturity of iPSC-CAR-NK cells across different disease contexts.

Table 3 Evidence levels and translational maturity of induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells across different disease contexts.

Disease/application context	Representative product or strategy	Ref.	Current primary evidence type	Conclusions supported by current evidence	Major limitations	Translational maturity
Relapsed/refractory B-cell lymphoma	FT596; FT516 may serve as a reference for a non-CAR-engineered iPSC-derived NK-cell platform	[23,92]	Early phase I clinical studies	The FT596 study suggests early safety and preliminary antitumor activity signals of iPSC-derived CAR-NK cells in B-cell lymphoma. FT516 may serve as a clinical safety and feasibility reference for a non-CAR-engineered iPSC-NK platform	The sample size and follow-up duration are limited. Randomized controlled studies and direct comparisons with CAR-T cells, bispecific antibodies, or antibody-based combination therapies are lacking. FT516 is not a CAR-engineered product and therefore cannot serve as direct evidence for the efficacy of iPSC-CAR-NK cells	Early clinical stage
Multiple myeloma	FT576; multi-module iPSC-derived NK/CAR-NK strategies involving BCMA-CAR, hnCD16, IL-15-related support modules, CD38-related engineering, and other modifications	[50,91,114]	Early clinical data, interim disclosures, and related mechanistic studies	FT576-related data suggest that BCMA-targeted, multi-module engineered iPSC-derived NK/CAR-NK strategies have a degree of translational potential. Antigen-escape studies may help explain insufficient durability of response and relapse risk	Full peer-reviewed clinical publications and long-term follow-up remain limited. The duration of response, patterns of treatment failure, antigen escape involving BCMA, GPRC5D, and other targets, and comparative effectiveness against existing BCMA-targeted therapies remain unclear	Preliminary clinical evidence stage
AML	FT538; anti-TIM3 iPSC-CAR-NK cells; NKG2C-KE; a registered clinical study of CLL1/CD33-targeted iPSC-derived NK cells	[93-95]; ClinicalTrials.gov: NCT06367673	Preclinical studies, early clinical recruitment, or preliminary translational data	Current evidence supports the feasibility of target identification, engineering design, and functional validation for iPSC-derived NK/CAR-NK strategies in AML	Most evidence remains preclinical. AML targets are often shared with normal hematopoietic cells, creating a risk of on-target, off-tumor toxicity. In vivo safety, the therapeutic dose window, and the GMP scale-up pathway still require validation	Preclinical to early clinical exploratory stage
Solid tumors	CD276-, MSLN-, and GPC3-targeted iPSC-derived CAR-NK strategies, as well as iPSC-derived NK/CAR-NK strategies incorporating CCL19, CCR2B, IL-15, NKG2D, or other functional-enhancement modules	[21,35,98-100,106,107]	In vitro experiments, patient-derived organoid studies, animal models, and preclinical mechanistic studies	Current evidence suggests the preclinical feasibility, antigen-dependent killing, and some in vivo antitumor activity signals of iPSC-derived CAR-NK cells in solid tumors	Immunodeficient animal models have limited clinical predictive value. Stromal barriers, abnormal vasculature, immunosuppressive networks, metabolic stress, and antigen heterogeneity in human solid tumors have not been sufficiently modeled. Current evidence cannot be directly extrapolated to clinical efficacy	Mainly preclinical stage
Single-case compassionate-use exploration in systemic sclerosis; autoimmune-disease applications have not yet been established	CD19/BCMA dual-targeting iPSC-derived CAR-NK cells	[22,103]	Single compassionate-use case report and related commentary articles	The original case study suggests that B-cell depletion, clinical improvement, and manageable early safety signals were observed in this individual patient. This result should be regarded only as an early feasibility signal in the context of single-patient compassionate use	This is a single, uncontrolled case and cannot demonstrate reproducibility of efficacy. It should not be equated with disease-level proof of concept in systemic sclerosis, nor does it support broad extrapolation to other autoimmune diseases. Long-term safety, B-cell reconstitution, infection risk, host immune responses, and the respective contributions of CD19 and BCMA dual targeting still require validation in larger cohorts with longer follow-up	Single-case early translational signal

CAR: Chimeric antigen receptor; iPSCs: Induced pluripotent stem cells; NK: Natural killer; BCMA: B-cell maturation antigen; GPRC5D: G protein-coupled receptor class C group 5 member D; AML: Acute myeloid leukemia; GMP: Good Manufacturing Practice; IL: Interleukin; NKG2D: Natural killer group 2 member D.

Based on the multilevel framework described above, this review interprets evidence for disease applications by considering not only whether signals of in vitro cytotoxicity, organoid killing, or tumor control in animal models have been observed, but also whether these signals extend across key layers such as tissue delivery, in vivo persistence within the host, safety, and translational implementation. Therefore, the translational significance of the same engineering strategy in different disease contexts should be assessed separately according to the dominant limiting layer and the level of evidence. It should also be noted that a translational gap remains between engineering-based functional enhancement and clinical efficacy. Designs such as CAR structural optimization, hnCD16-mediated ADCC enhancement, IL-15-related cytokine support, multi-target recognition, immune evasion, and TME adaptation are usually first supported by functional signals observed in vitro, including those derived from cytotoxicity assays, organoid models, animal models, or early dose-finding studies. However, these signals cannot be directly equated with durable remission, disease control, or long-term benefit at the patient level. Particularly for multi-module engineered products, clinically observed therapeutic responses may be influenced by multiple factors, including target selection, cell dose, dosing frequency, lymphodepletion regimens, concomitant antibody therapy, host immune status, and disease burden. Therefore, it is difficult to determine the independent therapeutic contribution of a given engineering module based solely on early clinical results. On this basis, when interpreting application evidence across different disease contexts, this review further distinguishes among “engineering feasibility”, “preliminary activity signals”, and “established clinical efficacy”, and avoids interpreting preclinical functional enhancement or early clinical remission signals as definitive clinical benefit.

Applications in hematological malignancies

Hematological malignancies often have relatively well-defined target-antigen landscapes, such as CD19 and BCMA, and effector cells can more readily access tumor cells. In contrast, solid tumors are usually constrained by limited infiltration, immunosuppression, and antigen heterogeneity, which make in vivo responses to cell-based therapy more difficult to predict[89]. Therefore, for iPSC-CAR-NK cells, hematological malignancies represent an important early clinical setting for evaluating safety, the therapeutic dose window, the feasibility of repeat dosing, and strategies for enhancing in vivo persistence[90]. Overall, hematological malignancies provide a relatively more clinically translatable disease setting for iPSC-CAR-NK cells. However, current studies still mainly support signals of early safety, feasibility, and preliminary antitumor activity. For products incorporating a CAR module, hnCD16, IL-15-related support modules, or other multiplex engineering designs, longer follow-up, larger sample sizes, exposure-response analyses, analyses of treatment-failure patterns, and comparative studies with existing therapeutic strategies are still needed to further determine whether these engineering designs can be consistently translated into durable clinical benefit.

MM: In MM, target antigens are relatively well defined. Among them, BCMA is one of the most extensively developed and investigated therapeutic targets. In addition, antigens such as G protein-coupled receptor class C group 5 member D and CD38 have also attracted considerable attention[50]. FT576 is a representative multicomponent engineered product. According to publicly available information, it contains four major design elements: A BCMA-targeting CAR; a high-affinity, cleavage-resistant CD16 module designed to enhance ADCC when combined with monoclonal antibodies; a membrane-bound IL-15 receptor-fusion-related module intended to support in vivo persistence; and a CD38-related engineering modification designed to reduce the potential risk of anti-CD38 antibody-mediated clearance of the product itself when combined with anti-CD38 monoclonal antibodies[91]. Currently available efficacy and safety data for FT576 are derived in part from interim disclosures. In these disclosures, no CRS or neurotoxicity was reported. Among evaluable patients, confirmed responses were observed, including very good partial responses. Although these results suggest a degree of clinical activity, the follow-up duration remains limited, and formal peer-reviewed publications are still needed to further clarify the durability of response and patterns of treatment failure[91].

Overall, currently available FT576-related data suggest that this engineered iPSC-derived NK/CAR-NK platform strategy may have translational potential in MM. However, the available evidence is still mainly derived from early study results or interim disclosures, with limited follow-up duration. The durability of response, patterns of disease relapse, and comparative effectiveness relative to existing BCMA-targeted therapies have not yet been sufficiently clarified. Therefore, this strategy should currently be regarded as an exploratory therapeutic strategy with preliminary signals of clinical activity, rather than as an established clinical therapeutic platform.

Relapsed/refractory B-cell lymphoma: In relapsed/refractory B-cell lymphoma, iPSC-derived NK-cell platforms have accumulated some early clinical evidence. The phase I study of FT596 has been published in a peer-reviewed journal, and the reported results indicated an overall favorable tolerability profile. In the combination-treatment cohort, CRS occurred in approximately 13% of patients, and all cases were grade 1-2, with no neurotoxicity observed[23]. With respect to efficacy, the full published report reported an overall response rate (ORR) of approximately 54% and a complete response (CR) rate of approximately 37% for the combination regimen. In the follicular lymphoma subgroup, the CR rate was approximately 85%. The study also included patients who had previously received CAR-T cell therapy, and some patients who achieved a CR remained in remission at the data cutoff[23]. Another published clinical study evaluated FT516, an iPSC-derived NK-cell product without CAR engineering. In this phase I trial, 55 patients were treated, and the reported ORR was 58%. Only one case of grade 1 CRS was observed, and no neurotoxicity was reported[92].

Overall, studies such as the FT596 trial provide early clinical evidence for the application of iPSC-CAR-NK platforms in relapsed/refractory B-cell lymphoma, suggesting a relatively manageable safety profile and some antitumor activity. However, these studies remain at an early stage, with limited sample sizes and follow-up duration, and direct comparisons with existing CAR-T cell therapies, bispecific antibodies, or antibody-based combination therapies are lacking. Therefore, the current evidence is more appropriately interpreted as signals of early clinical feasibility and activity, rather than as an established therapeutic advantage. Studies of non-CAR-engineered iPSC-derived NK-cell products, such as FT516, may also provide supplementary reference information regarding the safety and clinical feasibility of iPSC-derived NK-cell platforms, but they should not be regarded as direct evidence for the efficacy of iPSC-CAR-NK cells.

AML: In the iPSC-NK field, the published report on FT538 describes it as an engineered iPSC-derived NK-cell product rather than a CD33-CAR therapy. Its major engineering features include a high-affinity, non-cleavable CD16, CD38 knockout, and an IL-15/IL-15 receptor fusion construct. The study mainly evaluated its in vitro cytotoxic activity against AML cells and its effects when combined with commonly used anti-AML therapeutic agents[93]. In the iPSC-CAR-NK field, one study constructed a third-generation anti-TIM3 CAR-NK construct and reported enhanced cytotoxicity against TIM3-positive AML cells, with less pronounced effects on TIM3-negative cells. This type of study more closely follows the standard research workflow of “target identification-CAR design-functional validation”[94]. Another related strategy involves the use of killer cell engagers. A study published in Molecular Therapy in 2021 reported NKG2C-KE (anti-NKG2C/IL-15/anti-CD33), which enhanced the targeted response of NKG2C+ NK cells against CD33+ myeloid leukemia cells and showed activity in both primary NK cells and iPSC-derived cells[95]. In AML, a FIH phase I clinical trial has also begun recruiting to evaluate an engineered iPSC-derived NK-cell therapy expressing CARs targeting CLL1 or CD33 (NCT06367673).

Although iPSC-CAR-NK cells have shown some exploratory value in AML models and early translational studies, several important challenges and limitations remain. First, target selection in AML is particularly complex because many candidate antigens are also expressed in normal hematopoietic cell populations, thereby increasing the risk of on-target/off-tumor toxicity. For example, although CD33 is broadly expressed on AML cells, it is also present on normal hematopoietic cells, which may lead to on-target/off-tumor toxic effects. Second, despite the strong engineering potential of iPSC-CAR-NK cells, limited in vivo persistence and host immune-mediated clearance remain major barriers. Host clearance may shorten the duration of therapeutic benefit, and overcoming this issue will likely require continued optimization of immunomodulatory strategies and further refinement of the therapeutic platform. In addition, before these products can advance to registrational clinical development, it remains necessary to further validate their in vivo safety, therapeutic dose window, and GMP-compliant scalable manufacturing processes.

Therefore, the application of iPSC-CAR-NK cells in AML should currently still be regarded as being at the preclinical to early clinical exploratory stage. The true translational value of this approach will depend on whether target selection, the safety window, in vivo persistence, and the risk of toxicity to normal hematopoiesis can be adequately addressed.

Exploratory applications in solid tumors

In solid tumors, potent in vitro cytotoxicity, organoid-killing results, or inhibition of tumor growth in xenograft models does not necessarily predict clinically meaningful therapeutic efficacy. Therefore, the solid-tumor setting most clearly illustrates the translational gap between engineering-based functional signals and clinical efficacy. Even if a given iPSC-CAR-NK construct shows antigen-dependent killing and tumor-inhibitory activity in vitro, in organoid models, or in immunodeficient animal models, it still needs to be further demonstrated that it can overcome multiple constraints in patients, including tissue-delivery barriers, suppression by the TME, antigen heterogeneity, and host immune-mediated clearance. Current solid-tumor studies of iPSC-CAR-NK cells are mostly based on in vitro co-culture systems, patient-derived organoids, subcutaneous or intraperitoneal xenograft models, and immunodeficient mouse models. These models can be used to validate target-antigen-dependent killing, preliminary in vivo biodistribution, and signals of antitumor activity, but their clinical predictive value for the complex microenvironment of human solid tumors remains limited. In particular, immunodeficient mice usually lack an intact human immune system and do not adequately recapitulate the immunosuppressive network in patients that is jointly shaped by T cells, myeloid-derived suppressor cells, tumor-associated macrophages, regulatory T cells, cancer-associated fibroblasts, and endothelial cells. Human solid tumors also impose multiple constraints, including dense extracellular matrix and stromal barriers, abnormal tumor vasculature and impaired vascular access, hypoxia and metabolic stress, chemokine mismatch, antigen-expression heterogeneity, and antigen downregulation. These factors may limit CAR-NK cell homing, extravasation, infiltration into the tumor parenchyma, local persistence, and sustained effector function[96]. Therefore, when interpreting results from solid-tumor models, it is necessary to further distinguish whether the major limiting factors arise from tissue-delivery barriers; local suppressive mechanisms such as TGF-β signaling, adenosine accumulation, or hypoxia; inhibitory receptor-ligand axes such as PD-L1/PD-1, HLA-E/NKG2A, or CD155/TIGIT; or target-antigen heterogeneity and antigen downregulation. The engineering decisions corresponding to these different mechanisms are not the same. Thus, in vitro killing and efficacy signals observed in immunodeficient mouse models in solid-tumor studies should be understood as signals of preclinical feasibility, rather than being directly extrapolated as evidence of clinical efficacy[89,97].

In esophageal squamous cell carcinoma (ESCC), one study generated iPSC-derived CD276-CAR-NK cells and performed multi-level validation. Immunohistochemistry showed that approximately half of the ESCC specimens were positive for membranous CD276 staining (54/105), with positivity observed in 38 of 74 stage III-IV cases. The investigators compared patient-derived tumor organoids with adjacent normal epithelial control organoids and observed a cytotoxicity pattern more consistent with antigen-dependent killing. In a BNDG mouse xenograft model, the CAR-NK group showed slower tumor growth, with no obvious loss of body weight[98]. Overall, these findings provide valuable preclinical proof-of-concept evidence for ESCC. However, the current evidence remains at the preclinical stage, and any conclusions regarding clinical efficacy still require support from studies with higher levels of evidence.

In ovarian cancer, a Chinese-language study reported a relatively high frequency of positive mesothelin (MSLN) staining in ovarian cancer tissues and subsequently generated iPSC-derived MSLN-CAR-NK cells. In vitro co-culture assays showed that these cells enhanced tumor-cell apoptosis and the release of effector molecules[99]. With respect to in vivo evidence in ovarian cancer, a study published in Cell Stem Cell used an A1847-luc intraperitoneal ovarian cancer model and reported that iPSC-CAR-NK cells carrying an NKG2D-2B4ζ construct showed more pronounced signals of tumor-growth inhibition and prolonged survival in this model[21]. Another study published in Cancer Science used GPC3-targeting CAR-NK/ILC cells derived from HLA-homozygous iPSCs and showed, in a KOC7c intraperitoneal dissemination model, that repeated intraperitoneal administration could suppress tumor burden and prolong survival[100]. However, these studies are limited by relatively small sample sizes, and most of the currently available evidence still comes from in vitro cell-based experiments or animal models. Therefore, the antitumor effects of these ovarian cancer-directed iPSC-CAR-NK/ILC-related strategies require further validation in more robust in vivo experimental systems. Future studies should continue to integrate in vitro and in vivo approaches to further evaluate their antitumor activity and underlying mechanisms. To improve the clinical relevance of preclinical evidence, subsequent studies should also systematically assess their in vivo pharmacodynamics, persistence, tissue distribution, and potential safety window in larger cohorts of patient-derived samples and more appropriate animal models.

Therefore, the translational significance of iPSC-CAR-NK cells in solid tumors should still be interpreted with caution. Findings from solid-tumor models such as ESCC and ovarian cancer provide valuable proof-of-concept evidence for target selection, CAR structural design, and TME-adaptation strategies. However, these results mainly demonstrate antigen-dependent killing and short-term tumor control under specific model conditions. Future studies may integrate orthotopic tumor models, early-passage patient-derived xenograft models, humanized immune-system mouse models, tumor organoid-immune cell co-culture systems, and technologies such as spatial transcriptomics or multiplex immunoimaging to improve the clinical relevance of evaluating CAR-NK cell infiltration, maintenance of effector function, antigen escape, and safety in complex solid-TMEs[101,102]. However, these models also have their own limitations, and their results should still be interpreted cautiously in light of the degree of immune reconstitution, extent of stromal preservation, vascular architecture, and species differences. Only when consistent evidence has been obtained across multiple clinically relevant models and early clinical studies have further validated safety, the therapeutic dose window, in vivo persistence, and preliminary efficacy can the translational value of iPSC-CAR-NK cells in solid tumors be more reliably assessed.

Single-case compassionate-use exploration in systemic sclerosis

A study published in Cell in 2025 reported the FIH use of iPSC-derived CD19/BCMA dual-targeting CAR-NK cells in a patient with severe diffuse cutaneous systemic sclerosis[22]. This engineered product integrates multiple modifications designed to reduce allogeneic immune-mediated clearance and improve in vivo functional performance, including knockout of B2M, CIITA, and CD16, knock-in of HLA-E and HLA-G, and introduction of IL-2RF and truncated epidermal growth factor receptor (tEGFR), with tEGFR serving as a potential marker for conditional cell elimination or cell tracking. After treatment, B-cell depletion, clinical improvement, and manageable early safety signals were observed[22]. A subsequent commentary article suggested that this study provided important insights for exploring off-the-shelf iPSC-CAR-NK cells in systemic sclerosis, but the evidence should still be interpreted only as an early feasibility signal in the context of single-patient compassionate use[103].

It should be particularly emphasized that this case cannot be equated with disease-level validation of efficacy in systemic sclerosis, nor can it be used to infer that iPSC-CAR-NK cells are applicable to other autoimmune diseases. The improvement observed in a single patient may have been influenced by multiple factors, including disease heterogeneity, previous treatment history, concomitant or supportive therapies, individual immune status, follow-up duration, and fluctuations in the natural disease course. Therefore, this case cannot demonstrate that the strategy has reproducible efficacy. In addition, the B-cell depletion and clinical improvement observed in this case do not allow the relative contributions of CD19 targeting, BCMA targeting, the natural effector functions of NK cells, engineered immune-evasion design, in vivo persistence of the infused cells, and the individual immune status of the patient to be directly distinguished. Therefore, these findings cannot be used to determine whether the dual-targeting engineering design itself has produced a reproducible disease-modifying effect. Future studies should further evaluate safety, the therapeutic dose window, the reproducibility and durability of clinical improvement, B-cell reconstitution, infection risk, host immune responses, and the relative contributions of CD19 targeting and BCMA targeting in larger patient cohorts, prospective clinical studies, and longer-term follow-up. Equally important, the in vivo persistence of NK cells, dynamic changes in immune reconstitution, and potential delayed safety risks should be continuously monitored. Overall, this study should be understood as a single-case early translational signal in systemic sclerosis that warrants further exploration, rather than as evidence supporting broad applicability in the field of autoimmune diseases.

CHALLENGES AND RESPONSE STRATEGIES IN CLINICAL TRANSLATION

Based on the multilevel framework described in the preceding sections, this section provides a layered discussion of the barriers to the clinical translation of iPSC-CAR-NK cells. To avoid simply listing heterogeneous challenges in parallel, this review further distinguishes among core biological bottlenecks, key translational constraints, and specific engineering-based modulatory approaches. Core biological bottlenecks mainly refer to biological limitations that directly affect the ability of iPSC-CAR-NK cells to exert effective therapeutic activity in vivo. These include insufficient in vivo persistence and limited duration of effective exposure, restricted tissue delivery and infiltration in solid tumors, host immune-mediated clearance, target-antigen heterogeneity/immune escape, and TME-mediated functional suppression. Key translational constraints mainly refer to issues that affect whether a product can be manufactured consistently, evaluated for quality, reviewed by regulatory authorities, and implemented clinically. These include differentiation consistency, genetic stability, the risk of residual undifferentiated iPSCs, safety risks associated with multiplex engineering, potency assays, release criteria, CMC/QC requirements, and cost manageability.

In contrast, detailed optimization of CAR structure, selection of a specific cytokine module, introduction of specific chemokine receptors, knockout of a single immune checkpoint, or design of an individual safety switch should be more appropriately understood as specific engineering-based modulatory approaches directed toward the bottlenecks and constraints described above. These strategies do not necessarily determine the success or failure of clinical translation by themselves. Their value depends on whether they can address the dominant limiting layer in a specific disease context and achieve a relative balance among functional benefit, cellular fitness, safety, manufacturing consistency, and clinical implementability. Therefore, this section does not treat all challenges as equally important. Instead, according to their hierarchical position and translational impact, it attempts to distinguish the dominant bottlenecks that mainly limit clinical translation from engineering-based modulatory elements that can be further optimized.

To present the layered framework outlined above more intuitively, Table 4 summarizes the major issues that should be prioritized during the clinical translation of iPSC-CAR-NK cells using the structure of “bottleneck layer-core limiting factor-engineering/mitigation strategy-key validation metrics-key considerations for translational interpretation”. This table is intended to illustrate that the translational value of different engineering strategies cannot be judged solely on the basis of enhanced in vitro function. Instead, such value should be evaluated comprehensively by considering the bottleneck layer being targeted, the corresponding validation metrics, manufacturing feasibility, and clinical implementability.

Table 4 Bottleneck layers, engineering and mitigation strategies, and key validation metrics in the clinical translation of induced pluripotent stem cell-derived chimeric antigen receptor natural killer cells.

Bottleneck layer	Core bottlenecks/key translational constraints	Representative engineering/mitigation strategies	Key validation metrics	Key considerations for translational interpretation
Product definition layer	Insufficient differentiation consistency, batch-to-batch variability, risk of genetic instability, and risk of residual undifferentiated iPSCs	Clonal screening, establishment of master cell banks, standardized differentiation workflows, control of residual iPSCs, and optimization of cryopreservation and thawing processes	NK-cell purity, CAR positivity rate, cell viability, differentiation yield, post-thaw function, karyotype, CNV, genomic stability, residual pluripotency markers, and sterility, mycoplasma, and endotoxin testing	Determines whether the product can be manufactured consistently, assessed for quality, and compared across batches
Effector execution layer	Insufficient target-cell recognition, inadequate cytotoxic function, target-antigen heterogeneity or antigen downregulation, and insufficient serial killing capacity	NK-adapted CAR structural optimization, optimization of the antigen-recognition domain and affinity, hnCD16-mediated ADCC enhancement, multi-target design, logic gating, and cytokine support	Antigen-dependent killing, cytotoxicity at different E:T ratios, CD107a degranulation, granzyme B and perforin expression, cytokine release, serial killing capacity, antigen-negative escape, and killing of non-target cells	Enhanced in vitro effector function alone cannot be equated with in vivo efficacy or clinical benefit
Tissue delivery layer, including TME adaptation	Insufficient homing to solid tumors, extravasation, and infiltration into the tumor parenchyma; local TME-mediated suppression; hypoxia/metabolic stress; and remodeling of the antigenic landscape	Chemokine receptor engineering, adaptation to stromal barriers, TGF-β resistance, adenosine/A2A-axis adaptation, hypoxia-resistant or metabolically adaptive designs, and local TME-responsive modules	Chemotactic migration ability, tumor-tissue infiltration ratio, tumor-to-peripheral-blood cell ratio, spatial distribution, local persistence duration, markers related to TGF-β, adenosine, and hypoxia, and in vivo tumor-growth inhibitory activity	Requires determination of whether the main limitation arises from tissue delivery, local suppressive pathways, or antigen escape
Host interaction layer	Allogeneic immune-mediated clearance, insufficient in vivo persistence, accelerated clearance after repeat dosing, immunogenicity, and risk of “missing-self” recognition	HLA-related engineering, B2M/CIITA editing, HLA-E/HLA-G expression, evasion of host NK-cell-mediated clearance, and in vivo persistence-enhancing modules	In vivo expansion and persistence, cellular exposure levels in peripheral blood and tissues, host T/NK-cell clearance responses, anti-product immune responses, changes in cellular exposure after repeat dosing, and long-term safety	Determines the effective exposure duration and feasibility of repeat dosing; immune-evasion benefits need to be balanced against safety risks
Translational implementation layer	Safety risks associated with multiplex engineering, inadequate potency assays, non-uniform release criteria, insufficient comparability after process changes, and uncertainties in regulatory pathways and costs	Potency assay systems, release criteria, comparability studies, safety switches, long-term safety monitoring, manufacturing cost control, and optimization of cold-chain and supply-chain logistics	Batch production success rate, release pass rate, potency consistency, vector copy number, gene-editing off-target risk, replication-competent virus detection, manufacturing cycle, cost per dose, and long-term AE/SAE monitoring	Determines whether an engineering strategy can become a cell product that is amenable to regulatory evaluation, manufacturable, and clinically deployable

iPSCs: Induced pluripotent stem cells; NK: Natural killer; CAR: Chimeric antigen receptor; CNV: Copy number variation; ADCC: Antibody-dependent cellular cytotoxicity; TME: Tumor microenvironment; TGF-β: Transforming growth factor-β; B2M: Beta-2 microglobulin; HLA: Human leukocyte antigen; AE: Adverse event; SAE: Serious adverse event.

Limited in vivo persistence

The capacity of iPSC-CAR-NK cells for in vivo expansion and persistence may be intrinsically limited, which may affect the duration of effective cellular exposure and potential therapeutic activity. Therefore, in some early clinical studies, clinical regimens have used repeat dosing or combination with other treatment modalities to attempt to prolong the duration of effective in vivo exposure of engineered cells[23]. NK-cell homeostasis and functional maintenance are closely linked to IL-15 signaling[104]. Some iPSC-CAR-NK products have incorporated an IL-15/IL-15 receptor fusion module into the engineered construct to support cell survival and functional maintenance[23]. In the allogeneic treatment setting, host immune responses may further limit the in vivo persistence of infused cells[63]. A study of iPSC-derived CD70 CAR-NK cells reported that part of the product design was intended to mitigate allogeneic immune responses and that the product could suppress or eliminate allogeneic T cells[37]. As mechanistic evidence, human iPSC-derived NK cells with CISH knockout have been reported to exhibit improved metabolic fitness and in vivo persistence, together with enhanced antitumor activity[105].

Insufficient infiltration in solid tumors

In solid tumors, the major limitations of CAR-NK cells often lie not only in tumor recognition, but also in whether they can traffic to tumor tissue, persist locally within the tumor, and maintain effector function. Therefore, “insufficient infiltration” in solid tumors should not be understood merely as an insufficient number of cells entering tumor tissue. It should also involve assessment of the maintenance of local function after tumor entry, including whether TGF-β signaling, adenosine accumulation, hypoxia, checkpoint ligand-mediated inhibition, and myeloid-cell-associated suppressive signals limit CAR-NK cell degranulation, cytokine secretion, and serial killing capacity. In some solid-tumor models of iPSC-CAR-NK cells, studies commonly assess tumor infiltration using histological methods or flow cytometry and use these observations to explain differences in in vivo antitumor activity[98,106]. Taking CD276-targeting iPSC-CAR-NK cells as an example, mouse tissue-section analysis provided evidence that exogenously infused NK cells were able to infiltrate tumor tissue[98]. A potential feature of the iPSC platform is that multiple functional modules can be introduced simultaneously at the progenitor-cell stage. For example, one study reported that MSLN-CAR-IL-15 iPSC-derived NK cells showed enhanced in vivo antitumor activity signals in specific models, and further characterized cellular states in the TME through single-cell transcriptomic analysis[35]. To enhance homing and tumor infiltration, some engineered iPSC-NK cell strategies have introduced CCR2B expression and CCL19 secretion modules to improve cell trafficking and promote immune-cell recruitment[107]. In addition, in HCC models involving iPSC-derived NK cells, overcoming TGF-β-mediated suppression may be one factor contributing to enhanced antitumor activity[66].

Host-mediated immune clearance and rejection

As an allogeneic cell product, iPSC-CAR-NK cells may be subject to host immune-mediated clearance in recipients, thereby shortening the duration of effective in vivo exposure. Previous studies have shown that genetic modification of HLA class I expression can reduce the recognition of NK cells by HLA-mismatched T cells and the cytotoxicity mediated by these T cells[108]. However, excessive downregulation of HLA class I expression may also trigger “missing-self” recognition by host NK cells, thereby leading to immune clearance. Therefore, such strategies usually need to be combined with complementary strategies that inhibit host NK-cell recognition or provide protective signals[109]. Another issue that requires attention is that, even in autologous applications, iPSC-derived products may still be recognized and killed by NK cells because of imbalances in NK-cell receptor-ligand expression patterns. Therefore, it is recommended that the NK-cell receptor-ligand profile be incorporated into routine product characterization[110]. Engineered immune-evasion strategies also provide a direction worthy of further exploration. For example, expression of HLA-G, PD-L1, and PD-L2 in iPSCs may reduce innate and adaptive immune-mediated attack. In addition, immune-checkpoint-related immune-evasion designs may also be used to reduce the risk that HLA-deficient iPSC-derived products are recognized and cleared by innate immune cells[111].

Antigen heterogeneity and immune escape

Heterogeneity in tumor-antigen density and the presence of antigen-negative subclones may limit the coverage of single-target CARs and increase the risk of tumor immune escape[112]. To address this issue, one study developed 3MICA/B CAR-iNK cells targeting the membrane-proximal α3 domain of MICA/B, with the aim of broadening target recognition and reducing incomplete tumor-cell clearance caused by antigen heterogeneity[113]. In addition to multi-scFv or dual-CAR designs, the iPSC platform can also be used to construct multimodal recognition architectures, in which the CAR-recognition axis is combined in parallel with the hnCD16-mediated ADCC axis. In MM models, one representative strategy integrates a BCMA-CAR recognition axis, an hnCD16-mediated ADCC axis, an IL-15-related support module, and CD38-related engineering modification, and can be combined with anti-CD38 monoclonal antibodies. This design thereby forms a multi-antigen/multimodal recognition strategy that combines CAR targeting with antibody-mediated recognition[114]. From the perspective of early clinical translation, FT576 is a related multicomponent engineered product. Publicly disclosed interim phase I data have evaluated FT576 as monotherapy and in combination with daratumumab for relapsed/refractory MM. However, this evidence should still be interpreted as an early clinical signal rather than as an established conclusion regarding clinical benefit[91]. In the clinical-study setting, the FT596 study also included cohorts receiving combination treatment with rituximab, thereby allowing ADCC to potentially function as a second recognition pathway in addition to the CAR-mediated recognition axis[23].

Safety and reliability issues associated with multiplex engineering

As iPSC-CAR-NK cell products increasingly move toward multiplex genome editing, risk assessment should not focus only on off-target effects at individual edited loci. Instead, it should systematically evaluate DNA damage responses, structural genomic alterations, clonal selection bias, and long-term safety issues arising from the cumulative effects of multi-site editing. Nuclease-mediated DNA double-strand breaks, including those induced by CRISPR-Cas9, can trigger p53-associated DNA damage responses and may reduce the efficiency of precise editing. During cell banking and clonal screening, this selection pressure may lead to the relative enrichment of cells with impaired p53 function or other abnormal clones with growth advantages[115,116].

Therefore, for iPSC clones that have undergone multiple rounds of editing and long-term expansion, assessing only the editing efficiency at the intended target loci is insufficient. It is also necessary to evaluate the integrity of the p53 pathway, cellular proliferative advantages, changes in clonal composition, and the risk of enrichment for tumor-associated mutations[116-118]. In addition to conventional off-target mutations, large deletions, inversions, duplications, chromosomal translocations, and complex chromosomal rearrangements are also major risks that require particular attention in multiplex genome editing. Existing studies have shown that CRISPR-Cas9-induced DNA double-strand breaks can cause large deletions and complex rearrangements near the target sites[119]. In certain cellular contexts, chromosome bridges, micronucleus formation, and chromothripsis-like alterations may also contribute to more extensive structural abnormalities[120].

Therefore, for iPSC-CAR-NK products that simultaneously involve site-specific CAR integration, immune-evasion editing, cytokine-support modules, safety switches, and other functional-enhancement modules, the different edited loci may interact in unpredictable ways. These interactions may include aberrant joining between different DNA break sites, chromosomal translocations, copy-number alterations, and local or genome-wide structural variants. Such alterations may not be fully captured by conventional karyotyping. Therefore, depending on product-specific risk and the stage of development, a tiered assessment may integrate methods such as karyotyping, copy-number variation analysis, targeted deep sequencing, long-read sequencing, and whole-genome sequencing[117].

Multiplex editing may also amplify pre-existing clonal selection bias and epigenetic drift associated with iPSC culture. During reprogramming, clonal isolation, long-term expansion, gene editing, and differentiation, iPSCs may progressively accumulate genetic and epigenetic alterations, including acquired chromosomal abnormalities, copy-number changes, alterations in DNA methylation status, changes in chromatin accessibility, and drift in differentiation propensity[121]. These alterations may affect the maturation state, effector function, in vivo persistence, immunogenicity, and batch-to-batch consistency of the final NK-cell product. This issue may be particularly important in a development model based on a “universal cell bank plus disease-specific modular engineering” strategy. If different engineered versions are derived from different clones, different passage levels, different rounds of editing, or different differentiation batches, phenotypic and functional differences may arise across batches. Therefore, during manufacturing, the starting clone, passage range, editing sequence, and differentiation conditions should be fixed as much as possible. Genetic stability, epigenetic stability, clonal composition, key functional readouts, and batch-to-batch consistency should also be incorporated into a continuous quality-monitoring system[122].

Potential long-term tumorigenic risks also need to be evaluated more cautiously. The tumorigenic risk of iPSC-derived products may be associated not only with teratoma formation caused by residual undifferentiated iPSCs, but also with pro-proliferative mutations, chromosomal abnormalities, epigenetic abnormalities acquired during gene editing or long-term culture, and safety-switch failure[121,123]. For multiplex-engineered iPSC-CAR-NK cells, theoretical safety advantages cannot substitute for systematic safety evaluation. In addition to strict control of residual undifferentiated iPSCs at the release stage, in vitro transformation-risk assessment, long-term expansion stability testing, in vivo tumorigenicity evaluation, integration-site and copy-number analyses, long-term follow-up, and functional validation of safety switches when necessary should be integrated to reduce potential tumorigenic risks[122]. Overall, although multiplex genome editing provides iPSC-CAR-NK cells with stronger modular engineering capacity, it also substantially raises the technical thresholds for genetic stability, epigenetic stability, manufacturing consistency, and long-term safety[124,125].

Among potential safety-control strategies, iCasp9 is one of the inducible suicide switches that has been extensively investigated in CAR-T cells and other engineered immune-cell therapies, and it has been systematically discussed as a representative synthetic genetic switch for improving the controllability and safety of cell therapy[126]. Targeted insertion of iCasp9 into safe-harbor loci such as AAVS1, with biallelic integration when conditions permit, may reduce the risk of the emergence of escape clones caused by loss of heterozygosity or promoter methylation[127,128]. As a small-molecule trigger, AP1903 is supported by safety and pharmacokinetic data from studies in healthy volunteers[81]. In addition, iCasp9 has been clinically validated as a safety switch in adoptive cell therapy, supporting its use as a mechanism for rapidly reducing cellular exposure when necessary[80].

Manufacturing processes and QC

The batch-to-batch consistency of iPSC-CAR-NK cell products depends to a large extent on a complete and coherent evidence chain spanning upstream cell banking and downstream differentiation workflows. Targeted integration of the CAR construct, followed by clonal screening, helps control issues related to integration site, copy number, and random insertion at an early stage. For example, a study of EpCAM-CAR iPSCs used Southern blot analysis to verify that no random integration events had occurred[129].

In terms of manufacturing scale-up, iPSC-CAR-NK products need to undergo expansion, cryopreservation, and thawing within a scalable, verifiable, and traceable manufacturing system, and pre- and post-cryopreservation quality attributes should be incorporated into the stability evaluation framework[130-132]. It should be noted that conventional two-dimensional systems or short-term in vitro potency readouts may be insufficient to reflect solid-tumor-relevant functions. Existing studies suggest that cryopreservation can impair NK-cell migratory capacity and cytotoxic activity in three-dimensional collagen matrices. Therefore, stability evaluation should distinguish between functional readouts obtained in two-dimensional and three-dimensional systems[133].

For product release potency assessment, a multiparameter assay panel should be established based on the mechanism of action. This panel may include cell viability and phenotypic analysis, target-antigen-dependent cytotoxicity, and functional indicators such as degranulation or cytokine secretion. Among these, flow cytometry-based cytotoxicity assays can serve as one commonly used analytical tool[134,135]. Residual undifferentiated iPSCs constitute a platform-specific safety concern and can be addressed through a dual-protection strategy that integrates high-sensitivity detection with process-directed removal. Representative approaches include marker-based assays, label-free/high-throughput separation, and rBC2 LCN-based magnetic bead depletion of pluripotent cells[123,136,137].

In addition to manufacturing processes and QC themselves, the clinical translation of iPSC-CAR-NK cells also needs to meet broader regulatory science and regulatory review requirements. Under the regulatory frameworks of the United States Food and Drug Administration (FDA) and the European Medicines Agency (EMA), the development of genome-edited or genetically modified cell therapy products usually requires an evidence chain covering investigational new drug applications in the FDA context, clinical trial applications or authorizations in the EMA context, CMC information, quality, nonclinical, and clinical evidence, potency assays, comparability studies, and long-term safety monitoring. For cell therapy products involving genome-editing steps, quality and safety assessments generally need to cover the genome-editing strategy, editing efficiency, off-target risks, genomic integrity, product identity, purity, potency, and manufacturing-process consistency[138,139].

For multiplex-engineered iPSC-CAR-NK cell products, regulatory assessment should not be confined to the final cellular phenotype. It should also encompass edited genomic loci, transgene copy number, residual vectors or genome-editing tools, structural chromosomal abnormalities, clonal drift, residual undifferentiated iPSCs, post-thaw functional recovery, and, when necessary, long-term safety monitoring[138,139]. Potency assays should also be developed and validated in alignment with the product’s mechanism of action and critical quality attributes. Rather than relying on a single phenotypic indicator, these assays should incorporate functional readouts, including target-antigen-dependent cytotoxicity, degranulation, cytokine release, ADCC activity, and maintenance of effector function after repeated stimulation or serial killing.

At the same time, harmonization of release specifications remains an important challenge in this field. Across different studies and products, fully harmonized standards are still lacking for CAR positivity, NK-cell purity, cell viability, limits for residual iPSCs, potency assay methods, post-thaw functional recovery, and stability indicators. If changes occur in the manufacturing process, genome-editing strategy, cell bank, critical raw and ancillary materials, manufacturing site, or cryopreservation and thawing procedures, comparability studies are also required to demonstrate that the pre- and post-change products have acceptable consistency in identity, purity, potency, safety, and stability[140]. Therefore, although multiplex genome editing enhances the potential for platform-based development and functional expansion of iPSC-CAR-NK cells, it also substantially increases the complexity of regulatory review. In the future, more standardized frameworks for potency evaluation, release specifications, and process-change comparability are urgently needed.

In addition to the CMC/QC and regulatory science issues described above, cost structure and commercial feasibility are also key issues that must be addressed during the clinical translation of iPSC-CAR-NK cell products. In theory, the iPSC platform may amortize upfront development costs through the establishment of master cell banks, batch-scale differentiation, standardized cryopreservation, and off-the-shelf supply, while reducing batch-to-batch variability caused by donor-to-donor differences. However, this potential economic advantage should not be simply equated with a cost advantage that has already been demonstrated by clinical studies or real-world data. The final cost of goods for iPSC-CAR-NK cell products depends on multiple steps, including the establishment of GMP-grade iPSC cell banks, multiplex genome editing, directed differentiation and expansion, product cryopreservation and thawing, release testing, long-term storage, cold-chain logistics, and regulatory compliance requirements. Therefore, in the absence of systematic cost accounting, GMP manufacturing data, and real-world pharmacoeconomic evidence, the cost-reduction advantage of this platform should still be regarded as a potential advantage rather than an established conclusion regarding commercial feasibility[141,142].

First, the establishment of GMP-grade iPSC master cell banks and working cell banks requires substantial upfront investment. This process includes not only donor screening, starting-cell collection, reprogramming, clonal screening, and expansion, but also pluripotency characterization, karyotype analysis, genomic stability testing, sterility testing, mycoplasma testing, adventitious agent testing, viral safety assessment, and long-term traceability management. Although an established master cell bank can support the production of multiple manufacturing batches and the development of different engineered versions, the costs of its construction, validation, and maintenance are substantial. Cost amortization through scaled manufacturing may be achievable only when the production volume is sufficiently large, the batch success rate is stable, the process scale-up pathway is clearly defined, and the target-indication size is relatively well defined[24,143].

Second, multiplex genome editing and complex engineering designs may further increase R&D, QC, and regulatory-compliance costs. iPSC-CAR-NK cell products often incorporate CAR-introduction strategies, cytokine-support modules, immune-evasion strategies, safety switches, ADCC-enhancement modules, and other functional modifications. Each additional engineering module may introduce further requirements for vector construction, genome-editing reagents, clonal screening, off-target analysis, integration-site assessment, copy-number testing, detection of residual editing tools or vectors, genomic stability testing, potency assays, and functional validation. For products involving multi-site editing, it is also necessary to assess structural chromosomal abnormalities, clonal selection bias, genetic drift during long-term expansion, and potential interactions among different editing modules. If the manufacturing process, editing strategy, or critical raw materials change, comparability studies are also required to demonstrate acceptable consistency between the pre- and post-change products in terms of identity, purity, potency, safety, and stability. Therefore, although multiplex engineering may help enhance function and safety control, it also increases process complexity, the risk of batch failure, the burden of release testing, and the complexity of regulatory review[130,144].

Finally, long-term storage and logistics systems also affect real-world costs. Off-the-shelf cell products usually rely on cryopreservation and low-temperature transportation. Their commercial application requires stable liquid-nitrogen or ultra-low-temperature storage systems, verifiable cold-chain transportation procedures, standardized thawing operations, pre-infusion quality verification, and complete chain-of-identity and chain-of-custody tracking systems. Although cryopreservation helps establish off-the-shelf inventory, the processes of cryopreservation, thawing, and transportation may affect cell viability, migratory capacity, and cytotoxic function. Therefore, post-cryopreservation potency, stability, and post-thaw functional recovery should be incorporated into the QC system. Overall, the commercial feasibility of iPSC-CAR-NK cell products ultimately depends on multiple factors, including single-batch yield, cost per dose, batch failure rate, release testing cycle, storage and transportation costs, target-indication size, dosing frequency, reimbursement systems, and regulatory acceptability. At the current stage, claims regarding the cost-reduction and commercialization advantages of this platform should be made with caution[131,145].

CONCLUSION

Overall, the clinical translation of iPSC-CAR-NK cells should not be understood as a simple compilation of manufacturing workflows, engineering modules, and early application evidence. Instead, it should be conceptually integrated within a multilevel framework encompassing “product definition-effector execution-tissue delivery-host interaction-translational implementation”. This review emphasizes that the evaluation of this platform should not be limited to whether it can express a CAR or enhance NK-cell cytotoxic function. Rather, greater attention should be paid to the logic of product generation, the dominant bottlenecks, the hierarchical relationships among engineering modules, and the boundaries of clinical translation.

From the perspective of platform positioning, the fundamental distinction between iPSC-CAR-NK cells and conventional CAR-T or conventional CAR-NK therapies lies not only in differences in effector-cell type or cell source, but also in the fact that this platform combines the natural effector mechanisms of NK cells with an iPSC-cell-bank-driven logic of product development. Based on clonal iPSC cell banks, this platform moves product definition, modular engineering, standardized differentiation, and batch-consistency control upstream to an early stage of cell-product development. This feature gives it potential value as an off-the-shelf cell therapy platform. However, whether this potential can be translated into actual clinical advantages still depends on further validation across multiple layers, including in vivo persistence, tissue delivery, host immune-mediated clearance, safety, CMC/QC, and regulatory feasibility[23,43].

Based on this framework, iPSC-CAR-NK cells are gradually moving beyond a stage focused solely on technical feasibility and entering a phase in which product development, mechanistic optimization, and early clinical evaluation are being advanced in parallel. The value of this platform is mainly reflected in the establishment of master cell banks, standardized manufacturing, batch-to-batch consistency, multi-site engineering, and the potential for off-the-shelf supply. However, these advantages should still be regarded as platform potential requiring continued validation, rather than as therapeutic superiority that has been sufficiently demonstrated in clinical settings. Whether this platform value can ultimately be realized will depend on whether a verifiable balance can be achieved among manufacturing consistency, effector function, tissue delivery, in vivo persistence, safety, and the feasibility of clinical implementation.

At the current stage, the factors limiting the clinical translation of iPSC-CAR-NK cells are not equally important. It is necessary to distinguish among core biological bottlenecks, key translational constraints, and specific engineering-based modulatory approaches. Insufficient in vivo persistence, restricted delivery and infiltration in solid tumors, host immune-mediated clearance, antigen heterogeneity and immune escape, and TME-mediated functional suppression are better understood as core biological bottlenecks that more directly determine whether effective in vivo therapeutic activity can be generated[146]. In contrast, differentiation consistency, genetic stability, the risk of residual undifferentiated iPSCs, the safety of multiplex engineering, potency assays, release specifications, CMC/QC, and cost control more directly determine whether a product can be manufactured consistently, evaluated by regulatory authorities, and deployed clinically. By comparison, detailed optimization of CAR structure, selection of individual cytokine modules, introduction of specific chemokine receptors, knockout of individual immune checkpoints, or safety-switch design should be understood as engineering-based modulatory approaches directed toward the bottlenecks described above. Their value depends on whether they address the dominant limiting layer in a specific disease context.

Therefore, combinatorial engineering of iPSC-CAR-NK cells should not be based on the default assumption that more engineering modules are necessarily better. Strategies such as CAR structural optimization, hnCD16-mediated ADCC enhancement, IL-15-related support modules, chemokine/homing engineering, TME adaptation, immune evasion, and safety switches may generate synergistic effects across different layers, but they may also lead to functional redundancy or cross-layer conflicts. Only when an added module targets a major limiting step in a specific disease context, and when its functional benefit is expected to outweigh potential costs such as reduced cellular fitness, increased safety risks, greater construct complexity, increased manufacturing difficulty, and regulatory uncertainty, can that module have clear translational value. Future multi-module engineering should shift from “functional stacking” toward systematic optimization based on bottleneck matching, cross-layer complementarity, clear marginal benefit, and controllable risk[85,114].

Single-cell sequencing, multi-omics integration, CRISPR screening, AI-assisted modeling, and synthetic biology strategies provide important tools for the systematic optimization described above. Their value does not lie in generating more data or displaying more technical modules, but rather in translating differentiation states, functional NK-cell subsets, inhibitory signaling pathways, antigen-escape patterns, TME-induced functional attenuation, and dynamic in vivo changes after infusion into testable engineering hypotheses. Accordingly, TME-adaptation strategies should not remain at the level of a general description of an “immunosuppressive microenvironment”. Instead, they should further define specific cytokines, metabolic axes, receptor-ligand signals, tissue-delivery barriers, and changes in the antigenic landscape. Synthetic biology designs, including logic gating, self-regulatory feedback circuits, and safety switches, may help control the timing, spatial localization, intensity, and capacity for termination of cellular activation. However, these strategies still require systematic validation through in vitro function, biodistribution, long-term safety, cellular fitness, batch-to-batch consistency, and clinical endpoints, rather than being listed only as conceptual enhancement modules[70,74,75,82].

From the perspective of clinical translation, the currently available evidence mainly supports safety and feasibility, along with preliminary activity signals, for iPSC-CAR-NK cells, but remains insufficient to demonstrate that their clinical advantages have been established. Hematological malignancies currently appear to represent an application setting relatively closer to clinical translation; however, the durability of response, exposure-response relationships, optimal dosing regimens, and attribution of individual module contributions still require further validation. In solid tumors, combined constraints such as insufficient tissue delivery, TME-mediated suppression, antigen heterogeneity, and host immune-mediated clearance are more prominent. The single compassionate-use report in systemic sclerosis should be interpreted only as an early feasibility signal. It cannot be equated with disease-level efficacy validation, nor does it support broad extrapolation to other autoimmune diseases. Therefore, future evaluation of this platform should avoid directly equating engineering complexity, enhanced in vitro function, preclinical tumor-growth inhibition, or early clinical remission signals with definitive clinical efficacy advantages[22,23,103].

From the perspective of translational implementation, the maturation of the iPSC-CAR-NK platform also depends on factors such as GMP manufacturing scale, batch success rate, potency consistency, release specifications, process-change comparability, cold-chain logistics, long-term safety monitoring, real-world costs, and regulatory acceptability. Platform-based manufacturing and off-the-shelf supply do not inherently equate to low cost or high accessibility. Only when a verifiable closed loop is established among manufacturing, QC, clinical benefit, and commercial implementation can the advantages of scale-up truly be substantiated.

In summary, iPSC-CAR-NK cells represent an important direction in the development of off-the-shelf cellular immunotherapy, but this field remains at an early stage of clinical translation. Future studies should be based on clear bottleneck stratification and validation metrics, and should use more rigorous clinical endpoints, long-term follow-up, exposure-response analyses, failure-mode analyses, and comparative studies to determine whether different engineering strategies can truly translate into stable, reproducible, and clinically meaningful patient benefit. Only when a verifiable balance is achieved among manufacturing consistency, effector function, tissue delivery, in vivo persistence, host immune interactions, safety, regulatory feasibility, and commercial implementation can iPSC-CAR-NK cells progress from a promising engineering platform to a mature and broadly deployable clinical treatment modality.