The appeal of natural killer cells as a cellular therapy platform has grown steadily in recent years, particularly as the limitations of T cell-based chimeric antigen receptor therapies have become clearer. CAR-T cells have produced remarkable results in blood cancers such as leukemia and lymphoma, but their translation to solid tumors, and to brain tumors in particular, has been hampered by manufacturing delays, the risk of potentially fatal immune overactivation such as cytokine release syndrome and neurotoxicity, and the immunosuppressive environment that solid tumors create. Natural killer cells offer a distinct profile. They possess potent antigen-independent intrinsic cytotoxicity, meaning they can recognize and destroy stressed or malignant cells without needing a specific engineered target. They also carry self-regulating inhibitory mechanisms that reduce the likelihood of runaway immune activation, and they pose a low risk of graft-versus-host disease, which makes them candidates for off-the-shelf, donor-derived products that could be manufactured in advance and stored until needed.

In the new study, the researchers tackled one of the central practical challenges of natural killer cell therapy: obtaining enough highly functional cells to mount an effective attack. They expanded peripheral blood natural killer cells from healthy donors using genetically engineered K562 feeder cells that express membrane-bound interleukin-21, a powerful growth and activation signal for natural killer cells. This expansion system, designated K562-mb-IL-21, produced large numbers of highly activated and metabolically robust natural killer cells. Metabolic fitness matters enormously in cellular therapy, because cells that are exhausted or metabolically compromised after manufacturing tend to lose their killing capacity and their ability to persist in a patient’s body. The expanded cells generated in this study retained strong cytotoxic function against both established glioblastoma cell lines and, critically, tumor cells derived directly from patients, a more demanding and clinically relevant test than cell lines alone.

The breadth of activity observed in the expanded natural killer cells extended beyond glioblastoma. The researchers found that these cells also exhibited cytotoxicity against medulloblastoma, another aggressive brain tumor that primarily affects children and represents the most common malignant pediatric brain cancer. Testing against both a medulloblastoma cell line and a patient-derived medulloblastoma sample suggested that the expanded natural killer cells have broad targeting potential across different tumor types of the central nervous system. This is significant because patient-derived samples preserve much of the heterogeneity and biology of the original tumor, including the stem-like cell populations thought to drive recurrence and treatment resistance, whereas long-established cell lines can drift substantially from the clinical disease they are meant to model.

To sharpen the tumor-directed killing power of these cells, the team engineered them to express a chimeric antigen receptor targeting CD70, a cell surface molecule that has attracted growing interest as an immunotherapy target. CD70 is minimally expressed in healthy tissues but is overexpressed in recurrent glioblastoma, the form of the disease that ultimately kills most patients and for which no curative therapy exists. A receptor that recognizes CD70 should therefore direct the engineered cells preferentially toward tumor tissue while sparing normal brain and other organs. The researchers used CRISPR/Cas9 gene editing together with adeno-associated virus-based gene delivery to generate stable anti-CD70 CAR natural killer cells, a combination that allows precise insertion of the receptor construct and durable expression of the engineered receptor on the cell surface.

However, the engineering effort ran into a problem that is specific to natural killer cells and illustrates the subtle biology involved in designing such therapies. During the expansion process, natural killer cells naturally gain CD70 expression on their own surfaces. When the engineered cells carry an anti-CD70 receptor, this endogenous CD70 becomes a target: the CAR natural killer cells begin to kill one another, a phenomenon known as fratricide. This self-destruction prevents the large-scale expansion of a pure anti-CD70 CAR natural killer cell product, because the very cells being manufactured eliminate each other before a sufficient therapeutic dose can be produced. Fratricide has been a recognized obstacle in CD70-directed therapies, and overcoming it is essential for any clinical product targeting this antigen with natural killer cells.

The solution the researchers adopted was elegant in its directness: they knocked out the CD70 gene itself in the natural killer cells using CRISPR/Cas9. With the endogenous CD70 removed from the cell surface, the anti-CD70 receptor had nothing to recognize on neighboring engineered cells, and fratricide was eliminated. This genetic edit enabled the large-scale expansion of anti-CD70 CAR natural killer cells that could be grown to therapeutic numbers without destroying themselves. Importantly, the CD70 knockout did not appear to compromise the fundamental function of the cells. The resulting fratricide-resistant CAR natural killer cells remained functional and demonstrated potent activity against glioblastoma models both in vitro, in laboratory culture, and in vivo, in animal models bearing glioblastoma.

The in vivo component of the study carries particular weight, because laboratory dish experiments cannot capture the challenges that cellular therapies face inside a living organism, including trafficking to the tumor site, survival in the hostile tumor microenvironment, and persistence over time. The demonstration that the engineered cells retain anti-tumor activity in living models of glioblastoma, including models based on patient-derived tumor cells, provides stronger evidence that the approach could translate toward clinical testing. Patient-derived xenograft and organoid approaches have become increasingly important in neuro-oncology precisely because glioblastoma is so heterogeneous between patients, and a therapy that works across patient-derived samples has a better chance of helping a broad patient population than one validated only against a single cell line.

The study also highlights the value of combining innate and adaptive strategies in a single therapeutic cell. Because the expanded natural killer cells retain their intrinsic, antigen-independent cytotoxicity even after CAR engineering, they can potentially attack tumor cells through two parallel mechanisms: the engineered receptor directed at CD70, and their natural activating receptors responding to stress ligands on malignant cells. This dual capability could matter in glioblastoma, where antigen escape, the loss or downregulation of a targeted molecule, is a common route by which tumors evade single-target therapies. A cell that can fall back on natural recognition when CD70 expression varies may be harder for the tumor to outmaneuver. The self-regulating inhibitory mechanisms of natural killer cells further suggest a safety profile that could avoid the severe toxicities associated with T cell therapies, an especially important consideration for a disease treated in the confined and unforgiving environment of the brain.

The work, led by Misaal Mehboob and colleagues in the laboratory of Ali A. Ashkar at the McMaster Immunology Research Centre and the Centre for Discovery in Cancer Research, with contributions from Sheila K. Singh’s group and collaborators including Dean A. Lee and Meisam Naeimi Kararoudi, was supported by funding from the Canadian Institutes of Health Research. The research involved human sample collection under approval from the Hamilton Integrated Research Ethics Board, with peripheral blood obtained from consenting healthy volunteers, and all animal studies conducted under approved institutional protocols. As with all preclinical research, substantial steps remain between these laboratory findings and any approved therapy for patients, including optimization of manufacturing, formal safety testing, and clinical trials. Nevertheless, by solving the fratricide problem that has constrained CD70-directed natural killer cell therapy and showing potent activity against patient-derived glioblastoma both in culture and in living models, the study delivers a concrete technical advance in the effort to bring cellular immunotherapy to a disease that has seen tragically little progress in decades. For patients facing recurrent glioblastoma, where treatment options are nearly exhausted, each such advance carries real significance.

Subject of Research: CAR-engineered natural killer cell immunotherapy for glioblastoma

Article Title: Expanded NK and fratricide-resistant CD70 CAR-NK cells are potent against patient-derived glioblastoma cells in vitro and in vivo

Article References: Mehboob, M., Rätsep, M., Moinuddin, A., Portillo, A., Vahedi, F., Sookhaklari, M., Troy, E. C., Sezgin, Y., Shaikh, V., Maich, W., Venugopal, C., Kararoudi, M. N., Lee, D. A., Singh, S. K., & Ashkar, A. A. (2026). Expanded NK and fratricide-resistant CD70 CAR-NK cells are potent against patient-derived glioblastoma cells in vitro and in vivo. Cancer Immunology, Immunotherapy. https://doi.org/10.1007/s00262-026-04536-3

Image Credits: AI Generated

DOI: 10.1007/s00262-026-04536-3

Keywords: glioblastoma, natural killer cells, CAR-NK, CD70, CRISPR/Cas9, immunotherapy, medulloblastoma, brain cancer, cell therapy, interleukin-21, fratricide, patient-derived models