A patient's DNA may explain who gets dangerously sick from a 'living' cancer drug

Some of a lymphoma patient's own immune cells are removed, genetically reprogrammed to hunt their cancer, and put back as a one-time infusion. The drug is Yescarta, a CAR-T cell therapy already approved for several blood cancers. The cancer often responds. A few days later, the same patient can spike a fever, see blood pressure collapse, and end up in intensive care.

That cascade is the most dangerous side effect of CAR-T: cytokine release syndrome, the runaway inflammation that can turn a successful cancer treatment into a medical emergency. For years, the field has treated it as a problem with the engineered cells. A new study in Science Immunology argues the cause sits somewhere else, in the patient's own DNA.

Researchers at Massachusetts General Hospital reanalyzed data from Kite Pharma's Zuma-1 and Zuma-7 trials of axi-cel in pretreated lymphoma patients. The lead investigator Leick. The question was simple. Because every approved CAR-T is manufactured from the patient's own T cells, could the patient's germline, the DNA they were born with, shape how those engineered cells behaved, including the toxicity profile?

To narrow the search, the team focused on genes tied to hemophagocytic lymphohistiocytosis (HLH), a rare childhood hyperinflammatory disease. HLH shares clinical features with cytokine release syndrome, so a gene that drives HLH might also push a CAR-T patient toward severe CRS. The top hit was STXBP2, a gene that helps immune cells release the chemical messengers they use to communicate. Six patients in Zuma-1 carried loss-of-function mutations in STXBP2, meaning their copies of the gene could not do that job. All six developed toxicity after Yescarta treatment.

That 6-of-6 result is the cleanest signal the team found, but it is also a narrow one. STXBP2 loss-of-function mutations are rare, and the six carriers were all in Zuma-1, which enrolled sicker patients than Zuma-7. The same gene-toxicity pattern did not show up in Zuma-7. Leick's explanation, reported to FierceBiotech, is that Zuma-1 patients had higher baseline inflammation going into treatment, which is why a hyperinflammatory germline variant was more likely to tip them into severe CRS. That asymmetry is itself a finding about who the risk applies to: the most vulnerable patients, on the sickest end of the spectrum, are the ones whose DNA seems to matter most.

Leick is careful with the framing. The STXBP2 signal is "hypothesis-generating," not a clinical test, and a single gene should not be treated as a screening tool yet. The team is also pursuing a second, separate direction: a gene tied to how well CAR-T cells can replicate once they are inside the patient. They describe it as an active area of further investigation rather than a settled result.

That second thread matters because it points to where the field can actually act on the finding. If a patient's STXBP2 status could be checked before the T cells are removed, doctors could warn the patient, prepare the ICU, or steer them toward a different therapy. And if next-generation CAR-T designers can engineer around the most vulnerable patients' biology, choosing a different co-stimulatory domain or building in a safety switch, the class of therapy that currently fits only a narrow, very sick population could reach more of them safely.

A 2022 ASH annual meeting abstract already reported deleterious germline STXBP2 variants in patients who developed toxicity after axi-cel, and a separate 2022 paper in Blood linked clonal hematopoiesis, another kind of acquired genetic change, to increased risk of severe neurotoxicity in large B-cell lymphoma patients treated with Yescarta. The new Science Immunology study is the first to put the germline question on a peer-reviewed footing in a Kite-sponsored pivotal trial dataset, with a named gene and a clean, if small, count.

The germline finding also comes as the FDA is taking a closer look at CAR-T safety more broadly, for reasons that do not overlap with germline genetics. The FDA has moved toward a classwide boxed warning on commercial CAR-T products over secondary T-cell malignancy risk, then partially reconsidered the language. That is a separate safety question from germline toxicity genetics, and the two should not be conflated. The regulatory attention is real, though, and it gives the field a reason to widen the safety conversation, from the engineered cells to the patient those cells came from.

Leick frames the open question plainly. Some CAR-T products are known to be more toxic than others, and tumor biology influences response, "but there's a lot of other features about toxicity that we don't entirely understand." Germline DNA is now one of those features, with a candidate gene attached.