CAR-NK cell therapy in colorectal cancer: A systematic review of early clinical signals and translational strategies.

79 Background: Colorectal cancer (CRC) demonstrates limited responsiveness to current immune checkpoint inhibitors and breakthrough in chimeric antigen receptor–T cell (CAR-T) strategies face barriers in solid tumors. CAR–natural killer (CAR-NK) therapy offers potential advantages in safety and allogeneic scalability, yet the development of CRC-directed CAR-NK therapies remains unclear. We conducted a systematic review to define clinical signals and characterize translational engineering strategies of CAR-NK in CRC. Methods: A systematic search of PubMed, Scopus, ScienceDirect, Cochrane, and Wiley Online Library was conducted from inception through February 2026 (PROSPERO CRD420261324246). We included clinical studies administering CAR-NK therapy to patients with CRC, preclinical investigations evaluating CAR-NK platforms in CRC models, and registry-based trials permitting CRC enrollment were analyzed separately. Risk of bias was assessed using ROBINS-I and the SYRCLE risk of bias tool. Results: The available evidence was limited and early-stage, with only a small number of CRC clinical studies (<15 patients) complemented by a larger body of preclinical investigations and pan–solid tumor trials permitting CRC enrollment. Clinically, CAR-NK administration, including intraperitoneal mRNA-electroporated NKG2D CAR-NK and allogeneic NKG2D CAR-NK expressing membrane-bound IL-15 ± PD-1 blockade, was feasible and well tolerated, with no reported grade ≥3 cytokine release syndrome or neurotoxicity. Locoregional tumor control and immune activation were observed. However, durable systemic responses were infrequent. Combination therapy demonstrated enhanced immune activation compared with CAR-NK monotherapy. Registry analysis identified only one CRC-dedicated CAR-NK trial. Most active studies enroll CRC within basket designs rather than disease-specific cohorts. The termination of a MICA/B-directed CAR-NK program further highlights the limited and evolving clinical development landscape in CRC. Preclinical models demonstrated consistent antigen-dependent cytotoxicity across multiple CRC targets. Convergent engineering strategies addressed trafficking, microenvironmental resistance, antigen density, and persistence, with combination approaches outperforming monotherapy in xenografts. Conclusions: CRC-specific CAR-NK evidence remains promising, with signals of safety and localized activity but limited systemic efficacy. In contrast, preclinical innovation demonstrates strategic convergence toward multi-layer engineering and rational combination design. Dedicated CRC-focused trials are urgently needed to translate these advances into clinically meaningful outcomes.