Lipid nanoparticles reprogram NK cells inside the body
MIT researchers have built an in vivo cell-engineering strategy that could change how cancer immunotherapy and some autoimmune diseases are treated. Instead of removing immune cells from a patient, modifying them in a laboratory, and returning them days or weeks later, the new method sends genetic instructions directly to natural killer cells inside the body. The delivery vehicle is a lipid nanoparticle loaded with messenger RNA, and the payload instructs those cells to express a chimeric antigen receptor, or CAR.
The design matters because NK cells sit at an unusual intersection of immune surveillance and therapeutic flexibility. They can be redirected against malignant cells, but they are not the same as the T cells they may be asked to eliminate. In diseases driven by abnormal T cells, that distinction avoids a self-targeting problem that can complicate CAR T-cell strategies. The platform was tested in mice and humanized mice, where it eliminated tumors and, in a separate autoimmune model, reversed liver injury after removing autoreactive T cells.
One sentence captures the practical shift: a patient could someday receive nanoparticles rather than a bespoke cellular product.
Why NK cells offer a cleaner target than CAR T cells for some diseases
The appeal of CAR T-cell therapy is already well established in blood cancers, and the article notes that the U.S. Food and Drug Administration has approved seven genetically modified T-cell therapies over the past decade. These treatments work by equipping T cells with a receptor that recognizes a specific antigen on diseased cells. Once that receptor binds its target, the engineered cell activates and destroys the marked cell.
But the same logic becomes harder to apply when the disease itself involves T cells. A CAR T-cell product designed to hunt T cells can end up in a biologic tug-of-war, because the modified cells and their intended targets belong to the same immune lineage. NK cells avoid that conflict. They can be programmed to kill malignant or autoreactive T cells without creating the same risk of the engineered product attacking itself.
Daniel Anderson, senior author of the study, framed the manufacturing challenge directly: CAR NK-cell production is difficult, expensive, and labor-intensive. The in vivo approach aims to skip the extraction-and-expansion pipeline altogether. Instead, the nanoparticle carries mRNA that encodes the CAR construct, and the cell briefly uses that message to make the receptor after the particle reaches its destination.
Antibody-guided nanoparticles delivered genetic instructions with high efficiency
To improve delivery, the researchers coated the nanoparticle surface with antibody fragments that recognize NK-cell receptors. Those surface ligands help the particles home in on the intended immune cells rather than dispersing the genetic cargo indiscriminately. Once inside the NK cell, the mRNA directs production of the CAR protein, and the cell acquires the ability to identify and destroy cells carrying the matching antigen.
The same framework can be adapted to different targets. The article describes CARs directed against CD19, a B-cell marker, and CD3, which is present on T cells. That flexibility is central to the therapeutic concept, because a common delivery vehicle can, in principle, be paired with different receptor designs depending on whether the goal is to attack cancerous B cells or pathological T cells.
In mouse experiments, the nanoparticle approach modified approximately 80 percent of the targeted NK cells after injection. For a transient delivery platform, that level of uptake is biologically meaningful. It suggests the particles can reach circulating immune cells efficiently enough to produce a measurable therapeutic effect without requiring ex vivo manipulation.
Tumor clearance in melanoma and humanized lymphoma models
The anti-tumor data were striking. In mice with aggressive melanoma, three of five animals treated with nanoparticles designed to redirect NK cells against the cancer achieved complete remission. In a separate humanized mouse model of T-cell lymphoma, five injections of the nanoparticles eliminated the tumors entirely. Humanized mice carry transplanted human blood cells, which makes the model closer to the immune context that would be relevant for human translation than a standard murine system alone.
These results were compared with conventional CAR NK cells created outside the body and then infused back into animals. The in vivo strategy produced similar cancer-killing activity. That parity is important because it suggests the therapeutic biology can be preserved even when the laborious manufacturing step is removed. If the findings hold in humans, the treatment pathway could become shorter, cheaper, and easier to scale.
The tumor data also fit the underlying mechanism. CAR NK cells do not simply poison nearby tissue; they recognize a selected antigen and then direct immune attack against cells displaying that marker. In aggressive hematologic disease and melanoma models, selective recognition appears to have been enough to drive remission or complete tumor clearance in the treated animals.
Autoimmune hepatitis improved after one injection
The same platform was then used against autoimmune hepatitis, where abnormal T cells damage hepatocytes. Here the nanoparticle payload instructed NK cells to recognize the harmful T cells rather than a tumor antigen. The treatment eliminated the autoreactive cells in mice, and the liver tissue subsequently healed.
A single injection produced the improvement.
That result matters because autoimmune disease often forces clinicians to choose between suppressing immunity broadly and tolerating ongoing tissue injury. A targeted cellular response offers a different logic: remove the immune subset doing the damage while leaving the rest of host defense more intact. The liver findings suggest that the approach may not only interrupt active injury but also allow damaged tissue to recover once the attack stops.
A platform with broad disease potential, but still far from human use
The investigators say the same strategy could eventually be adapted for multiple sclerosis and rheumatoid arthritis, both of which can involve harmful T-cell responses. The biological premise is straightforward: if the nanoparticle can deliver a CAR message to NK cells safely and reproducibly, then the platform could be retuned for different targets without rebuilding the entire manufacturing pipeline each time.
Still, the evidence remains preclinical. Every efficacy result in this article comes from mice or humanized mice, not people. Safety, biodistribution, dose selection, durability of response, immune toxicity, and off-target effects all remain unresolved in humans. The article also makes clear that no CAR NK-cell therapy has received FDA approval yet.
The work was published in Nature Biotechnology, with support from Sanofi, the Marble Center for Cancer Nanomedicine, and a Koch Institute Support Grant from the National Cancer Institute. The next decisive question is not whether the platform can work in animals; it already does. The question is whether the same nanoparticle logic can be translated into a human immune system without losing precision, potency, or safety.