CAR‐NK cell‐based therapies: translational and regulatory breakthroughs

In this article for the “Highlights of 2025” series, we discuss advances in chimeric antigen receptor (CAR)–engineered natural killer (NK) cell therapies. We highlight iPSC-derived manufacturing, expansion into autoimmune disease, discovery of intrinsic regulators, and rational antigen targeting, illustrating how discovery-driven biology and engineering are reshaping CAR-NK therapies toward broader clinical applicability. Figure made in ©BioRender—biorender.com. For over three decades, natural killer (NK) cell–based adoptive cell therapies have been safely explored in cancer and immune-mediated diseases. NK cell–based therapies have since progressed with the emergence of chimeric antigen receptor (CAR)-NK cells by driving advances in genetic engineering,1 cell manufacturing,2, 3 and NK cell signaling biology4 Despite these advantages, early clinical and preclinical experiences with CAR-NK cells had limited clinical impact due to challenges in scalable manufacturing, limited in vivo persistence, narrow clinical application, and largely empirical approaches to CAR design.5 As a result, CAR-NK therapies initially occupied a complementary rather than transformative role within the adoptive cell therapy landscape. Early translational development of CAR-NK therapies provided evidence of safety and on-target activity, establishing the feasibility of engineered approaches in humans. However, early approaches relied on donor-derived primary NK cells, which were limited by donor variability and expansion capacity,1 or on established NK cell lines derived from the parental NK-92, which, despite scalability, were constrained by regulatory concerns.6, 7 Furthermore, CAR-NK development remained focused on oncology and was frequently positioned as a safer alternative to CAR-T cells rather than a distinct therapeutic platform. This limited exploring applications in immune dysregulation and autoimmune diseases, despite the central role of NK cells in immune homeostasis. Early CAR-NK designs followed T-cell signaling rationale, most commonly CD28 costimulatory domains coupled with CD3ζ, leaving NK-tailored designs underexplored. Consequently, CAR-NK engineering relied on empirical receptor and cytokine design, with limited use of unbiased screening or checkpoint insight, constraining rational optimization and predictability of therapeutic behavior. By 2025, however, accumulating evidence suggested that the field had reached a critical inflection point. A focused search of ClinicalTrials.gov (March 2026), excluding NK-92 and CAR-T studies, identified 128 interventional CAR-NK clinical trials, with only six withdrawn (~5%), suggesting confidence in feasibility and manufacturing. Notably, 32 trials were initiated in 2025 alone, representing over one quarter of all registered CAR-NK studies, while two studies had reported results at the time of analysis. This places the field in an era of rapid clinical expansion in which new trials have accelerated ahead of outcome data. Rather than minor tuning, recent studies now address the core translational and mechanistic constraints that previously limited CAR-NK development, such as standardizing NK-cell production platforms, expanding CAR-NK therapy into nononcologic applications, identifying regulatory checkpoints, and rationally engineering signaling and cytokine use. These advances reframe CAR-NK cells from donor-dependent cellular products toward engineered immune therapies that can be designed, tuned, and predicted, providing the basis for the key breakthroughs highlighted below. In The Lancet, Ghobadi et al.8 reported a fundamental advance toward standardization and scalability with the first-in-human phase 1 trial of induced pluripotent stem cell–derived (iPSC) CD19-directed allogeneic CAR-NK cells in patients with B-cell lymphomas. This study implemented a CAR design tailored to NK cell biology, incorporating NKG2D–2B4–CD3ζ signaling, a high-affinity CD16 receptor to enable antibody-dependent cellular cytotoxicity (ADCC), and an IL-15/IL-15Rα fusion to support autonomous persistence. These features address donor variability, limited persistence, and activation arising from T-cell-derived signaling mechanisms. The platform enabled standardized, high-dose administration of CAR-NK cells, including repeated infusions at the 108–109 cell range per dose, without reaching a maximum tolerated dose, and persisted in the peripheral blood up to 15 days postinfusion. Consistent with previous reports, no neurotoxicity or GvHD were reported. The favorable safety profile and antitumor activity support combining standardized manufacturing with NK-tailored signaling and cytokine support, establishing a scalable platform for CAR-NK therapies. Expanding applications beyond oncology, another key advance reported in The Lancet by Gao et al.9 was the first-in-human application of peripheral/cord blood–derived anti-CD19 allogeneic CAR-NK cells in relapsed/refractory systemic lupus erythematosus. In this open-label case series, patients received three CAR-NK infusions per cycle, with dosing beginning at 0.75 × 109 cells and escalating to 4.5 × 109 cells per dose across different schedules. Representing a high scale, repeat-dose administration that was previously challenging to achieve with donor variable products. The treatment induced rapid and sustained peripheral B-cell depletion during follow-ups. The construct itself followed T-cell principles (4-1BB–CD3ζ) but was paired with membrane-bound IL-15 to support NK persistence. Consistent with other trials' safety reports, no dose-limiting toxicities or neurotoxicity were observed, and cytokine release syndrome occurred in 1/18 patients (grade 1). Together, these findings support CAR-NK cells as a feasible immune-resetting platform in severe autoimmunity and reinforce that CAR-NK therapy can be administered in a scalable manner with early predictable safety. A major advance in 2025 moved CAR-NK optimization from empirical design toward systematic discovery of intrinsic regulators. In Cancer Cell, Biederstädt et al.10 reported PreCiSE, the first genome-wide CRISPR screening platform optimized for primary human NK cells. The platform was designed to run screens under repeated tumor challenge and immunosuppressive stress, conditions that mimic the pressures CAR-NK cells face in hard-to-treat tumors. The screens uncovered critical checkpoints that constrain NK cell function, including MED12, ARIH2, and CCNC, that when knocked-out, boosted CAR-NK antitumor activity across models, supporting clinical exploration of gene-edited cells. Complementing these findings, in Nature, Rafei et al.11 applied a systematic discovery approach by interrogating scRNA-seq profiles of anti-CD19-IL-15 CAR-NK cells before and after infusion, which highlighted the cAMP responsive element modulator (CREM) among the top upregulated transcription factors in vivo. CREM knockout increased CAR-NK effector function in vitro and in vivo, including after tumor rechallenge. Beyond any single hit, these studies establish complementary discovery pipelines, such as CRISPR screening and in vivo scRNA-seq, which provide ways to uncover additional checkpoints that can be edited, combined, and engineered to tune CAR-NK therapies toward more predictable therapeutic behavior. If CRISPR screens and in vivo profiling identify the molecular “dials” that govern CAR-NK fitness, construct and biologic engineering determine how these dials are wired into therapeutic function. In Molecular Therapy, Wong et al.12 developed a high-affinity anti-ROR1 antibody (A9) that improved NK cell–mediated ADCC against triple-negative breast cancer cells compared with the benchmark anti-ROR1 antibody UC-961. Using computational and structural analyses, the study addressed a key limitation of prior ROR1 targeting: potential off-target toxicity driven by homology with other receptors. The authors then bypassed this specificity constraint by engineering antibodies against an alternative extracellular region of ROR1, generating a new high-affinity anti-ROR1 scFv that served as a platform for multiple therapeutic formats, including a CAR, a monoclonal antibody, and an antibody–drug conjugate. Importantly, the study showed that antibodies can be engineered to enhance NK-cell effector function, highlighting that next-generation antibodies can be designed beyond tumor-antigen binding alone. These approaches also guide construct-level engineering. For example, recent work in NK-92 cells illustrates how swapping intracellular domains and applying pharmacologic control with dasatinib tunes CAR signaling strength and function,13 supporting combination-ready CAR-NK products. Together with the advances above, this work reinforces a central theme of 2025: CAR-NK therapies are becoming increasingly scalable and predictable through standardized manufacturing and rational engineering. These articles highlight several technologies that now enable CAR-NK cells' transition from cell products to engineered therapeutic platforms. High-throughput genetic engineering tools, including CRISPR screening and targeted editing, allow investigators to identify and validate intrinsic regulators that govern NK cell function in the tumor microenvironment. Single-cell multi-omics profiling before and after infusion maps how CAR-NK cells reprogram in vivo, revealing transcriptional checkpoints that can be targeted. Advances in automated, closed manufacturing and renewable sources, such as iPSCs, are making CAR-NK therapies more scalable and consistent. Finally, advances in antibody engineering and CAR domain design allow for therapies that not only target tumor antigens but also actively recruit and tune cytotoxicity, expanding combination-ready NK cell-based immunotherapy strategies. Looking forward, the field is entering a new phase in which discovery and delivery converge, grounded in a clearer understanding of what is required to make CAR-NK therapies effective, scalable, and accessible. Integrated computational–experimental platforms will become increasingly important to predict CAR-NK cytotoxicity across tumor contexts, guiding design decisions and reducing reliance on empirical testing alone. Recent multiscale in silico models demonstrate this potential that can be refined and expanded over the next few years.14 These approaches could enable in silico screening of CAR architectures and signaling configurations before experimental validation. Emerging 2025 work demonstrates that engineered lentiviral platforms can deliver CARs in vivo to NK cells, generating functional CAR-NK cells and pointing toward next-generation in vivo delivery strategies that bypass ex vivo manufacturing.15 Regulatory checkpoints uncovered in 2025, such as CREM, are likely to be among the first edits tested in clinical trials, and additional targets like ARIH2 or NR4A may be co-opted to enhance durability. Unbiased CAR-NK discovery pipelines will drive more rational CAR designs, expanding beyond cytokine support to include metabolic and epigenetic programs that sustain long-term activity. Finally, nononcology indications are expected to move beyond autoimmunity, into areas such as fibrosis, inflammatory bowel disease, and viral infections. Together, the 2025 breakthroughs indicate that CAR-NK development will increasingly combine disease-specific biology with computationally guided engineering to deliver cell-specific designs, accelerating the shift toward immune therapies with more predictable behavior. Funding for the Guimaraes's Laboratory was partially provided by the Medical Research Future Fund, Australia (with the support of the Queensland Children's Hospital Foundation, Microba Life Sciences, Richie's Rainbow Foundation, Translational Research Institute and The University of Queensland, Australia) under award number 2019485; and funding from the Cooper Rice-Brading Foundation, Australia, The Tie Dye Project, Bricks & Smiles, The Kids Cancer Project, Australia, Tour de Cure, the PA Research Foundation, the National Breast Cancer Foundation (award number 2023/IIRS0063). A.F.B.B. was supported by São Paulo Research Foundation (FAPESP) process number 2025/01430-6. The content is solely the responsibility of the authors and does not necessarily represent the official views of the organizations and funding agencies. The authors thank Emma Rose Mei Wong for assistance with English language editing and grammar revision. Alison Felipe Bordini Biggi: Conceptualization; writing – original draft; writing – review and editing; visualization. Fernando Souza-Fonseca-Guimaraes: Conceptualization; funding acquisition; writing – review and editing; visualization; supervision. F.S.F.G is a Board Member of Cure Cancer Australia Foundation and member of the Scientific Advisory Committee of ANZSA. Sanofi TSH, Microba Life Sciences, and Cartherics sponsor research in the laboratory of F.S.F.G. The authors have no commercial, proprietary, or financial interest in this study.