Inherited genetics linked to CAR-T toxicity and expansion

CAR-T therapies carry meaningful safety risks, and toxicity is known to vary with factors including disease characteristics, patient factors and the therapy itself. A new study suggests another variable deserves attention: the patient's inherited genome.

In a study published in Science Immunology, investigators analyzed patients from Kite Pharma's ZUMA-1 and ZUMA-7 trials of axicabtagene ciloleucel (Yescarta). In the ZUMA-1 pivotal cohort, grade ≥3 cytokine release syndrome (CRS) and neurologic events were reported in 11% and 31% of patients, respectively. The team found evidence that inherited genetic variation can influence both CAR-T cell expansion and clinical toxicity. Because autologous CAR-T therapies are manufactured from a patient's own T cells, those engineered cells retain the patient's inherited genetic variants. This raises a mechanistic question about how much inherited variation shapes treatment response and safety.

Loss-of-function variants in STXBP2 linked to toxicity

The team started from a targeted hypothesis. Hemophagocytic lymphohistiocytosis (HLH) is a rare pediatric hyperinflammatory disease with clinical features resembling severe CRS. The researchers screened genes associated with HLH and found that STXBP2, which regulates cytotoxic granule release, stood out. In the ZUMA-1 trial, six patients had loss-of-function STXBP2 mutations, and all six experienced toxicity.

This signal did not replicate in ZUMA-7, which enrolled patients with less advanced disease. The researchers hypothesize that the difference may reflect baseline inflammation; ZUMA-1 patients were generally more severely ill and more inflamed, potentially amplifying the effect of STXBP2 variants.

Mechanistically, the study team reported that donor T cells engineered to lack STXBP2 or to express these silencing variants triggered inflammation. This provides a plausible biological explanation for why these patients developed CRS-like toxicity.

Genome-wide findings

Beyond STXBP2, the genome-wide analysis identified additional signals that were consistent across the two trials. Variants in ADAMTSL3 were associated with protection from treatment-related toxicity, while variants in PTPN22 were strongly associated with CAR-T cell expansion after infusion.

Carl June, M.D., a pioneer of CAR-T therapy at the University of Pennsylvania (not involved in the study), commented that the research "convincingly demonstrates that the underlying germline genetics of the patient act as a major, intrinsic determinant of both CAR-T cell expansion and clinical toxicity."

He described the PTPN22/expansion link as "particularly striking" and suggested that intentional PTPN22 knockdowns could be incorporated into manufacturing to enhance expansion and reduce the required cell dose. This points toward a future where human genetic variation could inform both risk stratification and cell therapy design.

The diversity gap limits generalizability

A key limitation of this study is that the analysis was restricted to patients of European ancestry. The original Kite trials enrolled too few participants from other ancestry groups to support a meaningful analysis. Dr. Mark Leick, an oncologist at Massachusetts General Hospital who led the study, acknowledged this directly: "We just did not have the numbers of non-Europeans to conduct any serious type of analysis."

This means we don't yet know whether the STXBP2, ADAMTSL3 and PTPN22 associations generalize across ancestry groups. If cell therapy sponsors eventually want to use germline genetics to stratify patients, they will need trials that are representative enough to validate these findings more broadly.

What this could mean for future cell therapy trials

The study raises the possibility that germline genetics could eventually become another variable in cell therapy risk stratification. Before that becomes clinically actionable, the associations will need validation in larger and more diverse populations, and prospective studies will need to establish whether genotype adds meaningful predictive value beyond known clinical risk factors.

This has practical implications:

• Genetics as a prospective trial variable. Future studies could incorporate germline sequencing to test whether candidate variants prospectively predict toxicity or expansion.
• Genotype-informed product development. PTPN22 provides an intriguing example of how human genetic variation might inform CAR-T engineering; this remains a research opportunity rather than an established manufacturing strategy.
• Diverse recruitment as a scientific requirement. If genetic associations are going to influence clinical decisions, they need to be established across the populations who will receive these therapies.

Operationally, this would add another layer to an already complex patient journey. Cell and gene therapy programs increasingly need to connect recruitment, biomarker or genetic testing, clinical eligibility and referral without creating additional burden for patients or sites. If germline risk markers eventually become clinically actionable in CAR-T, sponsors will need to think about how that information is generated and used much earlier in the trial pathway.

Implications for the field

The immediate implication isn't that every CAR-T candidate should now undergo germline screening. The evidence isn't there yet. What's important is that patient genetics may represent a previously underappreciated source of variation in how engineered cell therapies behave once they enter the body.

For developers, that creates a new set of questions. Can germline variants improve prospective models of toxicity? Do their effects change with disease stage or inflammatory state? Are the same associations present across ancestries and across CAR-T products? And could naturally occurring human variation eventually inform how the therapies themselves are engineered?

Answering those questions will require larger, more diverse and genotype-rich clinical datasets. As cell therapies become increasingly precise, understanding the biology of the patient receiving them may become just as important as understanding the cells being engineered.