Despite promising preclinical studies and early clinical data (Liu et al., NEJM, 2020), chimeric antigen receptor natural killer cells (CAR NK) are not yet on par with CAR T cells for the treatment of malignant diseases. Much effort has gone into developing NK-optimized CAR architectures (Li et al., Cell Stem Cell, 2018) and to increasing the persistence of CAR NK cells. We hypothesized that CAR NK cells could also benefit from the targeted integration of a CAR transgene, as compared with semi-random integration obtained using gammaretroviral vectors. Precise targeting has proven highly advantageous for TRAC CAR T cells owing to the optimal regulation and homogeneous expression conferred by the endogenous TCR alpha promoter (Eyquem et al., Nature, 2017). We first optimized an editing protocol for primary NK cells that uses electroporation of a CRISPR/Cas9 ribonucleoprotein followed by transduction with AAV6 bearing a homology-directed repair template. With this protocol we were able to knock out (KO) genes (including CD38, B2M, and different NK receptors), knock in transgenes (GFP-fusion protein, truncated EGFR, CAR), or both, at high efficiency. There was only moderate cell toxicity, and proliferation and cytotoxicity were conserved. Because of the high constitutive expression of B2M, we initially chose to focus on this locus and targeted a 1928z CAR transgene for regulation under the endogenous B2M promoter; our template design also rescues the B2M gene upon knock-in, enabling intermediate B2M expression (Figure 1). This strategy conferred high expression of the CAR and significantly more homogeneity as compared with gammaretroviral delivery (Figure 1). Furthermore, the rescued B2M expression protected B2M CAR NK cells from fratricide killing, whereas the B2M KO NK cells and the CAR NK cells with disrupted B2M expression were susceptible. B2M CAR NK cells also displayed higher in vitro cytotoxicity as compared with WT NK cells against two different CD19-expressing tumor cell lines (NALM6, SupB15). We tested B2M CAR NK cells in an NSG mouse xenograft model with an intraperitoneally injected human ovarian cancer cell line engineered to express CD19 (CD19 + SKOV3). Unexpectedly, compared to retroviral CAR-NK cells, B2M CAR NK cells conferred less tumor control (assessed by bioluminescence imaging) and less of a survival benefit. Therefore, we screened a panel of other loci for CAR targeting: five for NK activating receptors ( CD16A, CD56, DNAM1, NKp30, NKp46) and four for NK inhibitory receptors ( CD161, LILRB1, NKG2A, TIGIT). For activating receptors, the transgene was designed to use the endogenous promoter and to maintain expression of the targeted locus. For inhibitory receptors, the transgene was placed under the control of an EF1a exogenous promoter and designed to disrupt expression of the targeted locus. We compared these constructs and the initial B2M CAR NK design for CAR expression (by flow cytometry) and in vitro functionality (cytotoxicity assays) and observed wide variation in knock-in efficiency and CAR expression (Figure 2). For the top constructs based on knock-in efficiency, we conducted in vitro cytotoxicity assays against NALM6 and SupB15 cell lines. We found that when normalized for CAR percentage using the respective KO conditions, all tested knock-in loci performed better than B2M CAR NK cells, despite the lower CAR expression. Our current focus is on validating these top loci for the targeted integration of a 1928z CAR using in vivo models for comparison with B2M CAR NK and retroviral CAR NK. In summary, we have identified advantageous in vitro characteristics for the targeted integration of a CAR transgene in primary NK cells and screened 10 loci for CAR expression and in vitro cytotoxicity against tumor target cells. Confirmatory in vivo studies are ongoing and will be presented.