UCSF turned immune cells into cancer killers inside the body, clearing leukemia in nearly all mice

Eighteen of 20 mice with an aggressive B-cell leukemia reached a complete response after a single injection that built cancer-fighting T cells inside their own bodies, with no cells ever removed for engineering. The mice carried human immune systems drawn from four donors, and the work came from the laboratory of Justin Eyquem at the University of California, San Francisco. Nature published it online on March 18, 2026.

Eyquem, an associate professor of medicine at UCSF, is senior author of the paper, titled “In vivo site-specific engineering to reprogram T cells” and printed in Nature volume 652; William Nyberg and Pierre-Louis Bernard, both postdoctoral fellows, are co-first authors. Collaborators came from the Gladstone Institutes, Duke University and the Innovative Genomics Institute, and Jennifer Doudna appears on the author list.

Two particles aimed at CD3

The treatment uses two delivery vehicles rather than one. According to the Nature paper, the first is an enveloped delivery vehicle carrying Cas9 protein and guide RNA, studded with an antibody fragment that recognizes CD3, a protein found on T cells. The second is an engineered adeno-associated virus, called AAV-hT7, which carries the DNA for the chimeric antigen receptor and was evolved on human T cells to resist neutralizing antibodies.

Together they drop the receptor gene into the T-cell-specific TRAC locus, a predetermined spot in the genome.

The UCSF news account frames the site-specific insertion as the point: because the DNA lands in a chosen location, it avoids the random insertion that can rarely cause secondary cancers, and an on-switch there is active only in T cells. The team reports it as the first time a large DNA segment has been inserted at a precise site in human T cells without taking the cells out of the body.

Eighteen of 20 mice, and a lentiviral comparison

In the leukemia model, the paper states, 18 out of 20 mice across four donors achieved complete responses after a single treatment. In a separate comparison, all six mice that received the two-vector treatment achieved complete tumor control, and they outperformed mice given conventional lentiviral CAR T cells. UCSF’s account describes all detectable cancer disappearing in nearly all mice within two weeks.

The engineered cells reached as much as 19.7% of splenic T cells in the paper, and UCSF reports that in some organs they made up as much as 40% of immune cells. The approach also worked in mouse models of multiple myeloma and of a solid sarcoma, a tumor type that has resisted CAR T therapy.

Eyquem told UCSF that the cells generated in the body look better than the ones made in the lab, a possibility he attributes in part to culture dishes eroding the stem-like, self-renewing qualities of T cells. Only the combination of anti-CD3 vehicles with AAV-hT7 produced complete B-cell aplasia, the on-target effect that signals the receptor cells are working.

Weeks of manufacturing and a $400,000 price

Conventional CAR T therapy collects a patient’s blood, separates the T cells, engineers them and expands them to hundreds of millions before infusion; the National Cancer Institute puts the span from blood draw to reinfusion at roughly three to five weeks. UCSF cites a typical cost of $400,000 to $500,000 per treatment, and patients usually need chemotherapy beforehand to make room in the bone marrow, which older or frailer patients may not tolerate.

Seven such products are approved for blood cancers, and all seven (Abecma, Aucatzyl, Breyanzi, Carvykti, Kymriah, Tecartus and Yescarta) appear on the Food and Drug Administration’s list of approved cellular and gene therapy products. The NCI notes their common side effects include cytokine release syndrome, which can cause high fever and dangerous drops in blood pressure, and neurologic problems such as confusion and impaired speech, both managed with drugs such as tocilizumab and steroids. Whether an in-body route would change that profile is not addressed by the mouse data, since the humanized animals cannot report confusion or fever the way a patient can.

Eyquem frames the stakes as access. “It’s become a global access issue,” he said, since many patients who would benefit “either can’t afford them or can’t get them fast enough.” A syndicated version of the UCSF story reappeared in October 2026, but the experiments are the March results. Eyquem and colleagues founded Azalea Therapeutics to carry the platform toward the clinic, and quality control after injection is not possible the way it is for a manufactured batch, so the system had to be tuned in advance to avoid editing the wrong cells.

The approach has not been tested in people. The central result is 20 leukemia mice across four donors, plus the six-mouse comparison, and UCSF states the approach still needs scale-up and clinical trials to establish safety and efficacy in people.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview