Nanoparticles aimed at NK cells turn them into CAR NK cells inside mice

The 80% is a ceiling. The abstract says lead formulations reach up to ~80% NK cell transfection in mice ^[3], so even in the best case roughly one NK cell in five was left unprogrammed ^[11]. The same sentence reports "minimal off-target effects", also in mice ^[3].

The B-cell result is a delivery test. CD19-targeted CAR NK cells completely depleted circulating B cells in healthy mice ^[5]. That shows the programmed cells found and killed their target. Healthy animals have no disease to improve, so the experiment cannot show a therapeutic effect.

The autoimmune result is the one the authors lean on. CD3-targeted CAR NK cells eliminated nearly all autoreactive T cells in a hepatitis model ^[6], and the abstract calls this "a potential advantage over CAR T therapies for T cell-driven disease" ^[12]. The readout is a cell count. The thing this doesn't tell you is whether the liver disease improved, how many animals were treated, or how a CAR T arm would have done.

"Off-the-shelf" is our phrase, not the authors'. The abstract's label is "a modular immunotherapy system" ^[13]. We think the phrase is fair as design logic: CAR mRNA is delivered systemically and the CAR NK cells form in the body ^[4], so no patient cell product has to be made. The paper's references on CAR T manufacturing and cost ^[9] point at the problem that design addresses.

The solid-tumour and humanized results are separate experiments. The NKG2A effect is reported in solid tumors, through relief of Qa-1b-mediated immunosuppression ^[7]. The regression is reported in two humanized hematologic malignancy models, after the authors optimized the nanoparticles for human CAR NK cells ^[8].

In our view, the abstract supports programmable CAR NK cells in mice and activity against two humanized blood-cancer models. The autoimmune edge over CAR T stays a hypothesis that the authors themselves label potential ^[12]. That is a verdict on the abstract alone.