AUTHORS
Ruyan Rahnama, MD, PhD, Khadija Kone, N/A2, Samika Arun3, Bonell Patino-Escobar, MD4, Amrik S Kang, PhD5, Allison Arnold6, Arun Wiita, MD, PhD7
AFFILIATIONS
1 University of California, San Francisco, San Francisco, California, United States
2 University of California, San Francisco, San Francisco, California, United States
3 University of California, San Francisco, San Francisco, California, United States
4 University of California, San Francisco, San Francisco, California, United States
5 University of California, San Francisco, San Francisco, California, United States
6 University of California, San Francisco, San Francisco, California, United States
7 University of California, San Francisco, San Francisco, California, United States
ABSTRACT
Introduction: Acute myeloid leukemia (AML) remains a major cause of cancer-related mortality in children and adults. Chimeric antigen receptor (CAR)–engineered natural killer (NK) cells are a promising off-the-shelf immunotherapy platform; however, limited persistence, tumor antigen heterogeneity, and on-target, off-tumor toxicity remain major barriers to clinical translation. CD70 is a novel and specific AML target antigen. Our group previously showed that natural ligand (CD27) based anti-CD70 CAR-T cells outperform scFv-based designs (Kasap et al, Blood, 2025) and applied machine learning strategies to generate a CD27-variant (N88A) CAR-T with enhanced anti-AML activity (Kang et al, ASGCT Annual Meeting, 2025). Here we continue to validate the superior activity of the N88A CAR in the context of NK cells and leverage NK-specific biology to inform engineering modifications that further boost performance.
Methods: The CD27-variant (N88A) was sub-cloned into a NK-specific (2B4.CD3ζ) CAR backbone with membrane-bound IL15 (mbIL15) and delivered into primary NK using a γ-retrovirus vector. NK cells were activated and expanded in the presence of K562.mbIL15.41BBL feeder cells. CD70 upregulation in feeder-expanded NK cells necessitated CRISPR/Cas9-mediated CD70 knockout (KO) to avoid fratricide. CD70 KO and CAR transduction efficiencies were measured by flow cytometry. Short-term (18-hour) cytotoxicity at multiple effector-to-target ratios was measured by residual target cell luminescence following CAR-NK coculture. Long-term cytotoxicity was evaluated using a serial target stimulation assay in which AML target and CAR-NK cells were plated 1:1 on Day 0 and challenged with additional target every 48-72 hours. IFN-γ secretion by CAR-NKs after target cell coculture was used as an additional measure of NK activation. In vivo evaluation of CAR-NK cells was performed in NSG mice, with tumor burden measured by bioluminescence imaging and CAR-NK expansion by flow cytometry.
Results: We tested three parallel workflows to optimize the editing sequence of CD70 KO and CAR transduction in primary NK cells after demonstrating CD70 upregulation with feeder-based activation. Upfront CRISPR/Cas9-mediated knockout of CD70 on Day 0 of NK cell isolation from PBMC followed by four days of recovery prior to CAR transduction allowed for the most robust NK cell expansion and editing efficiency. CD70 KO and CAR transduction efficiencies were consistently high with >60% CAR surface expression across all tested constructs. Notably, even under conditions of suboptimal CD70 knockout (<90%), high CAR surface expression was sufficient to eliminate residual CD70⁺ NK cells, with selective enrichment of CD70⁻ CAR⁺ NK cells. All CD27-based CARs demonstrated antigen-specific activity in vitro against CD70+ AML target cell lines (MOLM-14, NOMO-1, THP-1, MV411, MOLM-13) and loss of CAR-mediated activity in CD70- controls (MOLM14.CD70KO, THP-1.CD70KO). In a head-to-head comparison, the NK-specific co-stimulatory backbone 2B4.CD3ζ outperformed the conventional T-cell derived 41BB.CD3ζ backbone, driving greater NK expansion and cytotoxicity. In vivo, the N88A.2B4.CD3ζ CAR demonstrated superior peripheral expansion and antitumor activity against NOMO-1 when supported with twice-weekly systemic IL-2. To obviate the need for exogenous cytokine support, we armored our N88A.2B4.CD3ζ CAR to express mbIL-15 for local autocrine signaling. This engineering enhancement yielded remarkable improvements in CAR-NK expansion, short-term, and long-term cytotoxicity and suppressed tumor progression in an AML xenograft model without systemic IL-2 support.
Conclusion: We report NK-specific engineering enhancements that improve upon the anti-AML activity of N88A CARs previously described in CAR-T cells. These findings demonstrate the value of tailoring CAR optimization strategies to the specific effector cell platform. Specifically, NK-informed modifications to the CAR design improved both in vitro and in vivo antitumor activity, highlighting the potential to advance natural ligand-based CAR- NK therapies to the clinic. Future studies will include transitioning from a retrovirus-based editing system to a CRISPR/Cas9, AAV-mediated knock-in approach for targeted integration of our lead CAR into NK inhibitory loci (ex. KLRC1, TIGIT). This innovative engineering strategy and downstream mechanistic analyses will help us to further optimize our CD70-targeting CAR-NK cell designs.
DISCLOSURES
Ruyan Rahnama, MD, PhD: No relevant conflicts of interest to declare.