Mark Leick, MD

 suggest germline genetic variants may help explain why CAR T cell therapy outcomes vary so widely between patients.

Investigators performed whole genome sequencing on patients with aggressive lymphoma treated with axicabtagene ciloleucel in the ZUMA-1 and ZUMA-7 clinical trials, in partnership with Kite Pharma.

Analysis focused initially on genes associated with hemophagocytic lymphohistiocytosis (HLH), a rare hyperinflammatory syndrome with biochemical similarities to cytokine release syndrome (CRS), the most common toxicity of CAR T cell therapy.

Investigators found that variants in STXBP2 were enriched among patients with high-grade toxicity in the ZUMA-1 cohort, a finding supported by in vitro experiments showing that STXBP2-deficient CAR T cells had impaired degranulation and increased inflammatory cytokine production.

Extending the analysis genome-wide, variants in ADAMTSL3 were linked to protection from toxicity, while variants in PTPN22 were strongly associated with greater CAR T cell expansion, across both ZUMA-1 and ZUMA-7.

HCPLive spoke with Mark B. Leick, MD, a medical oncologist and physician scientist at Massachusetts General Hospital who led the study, about its design and what the findings could mean for CAR T cell manufacturing and patient management going forward.

Q&A: Germline Variants May Predict CAR T Cell Toxicity, With Mark Leick, MD

HCPLive: What motivated you and your colleagues to investigate whether inherited germline genetics could influence outcomes after CAR T cell therapy?

Leick: CAR T cells are different from every other form of therapy that came before. It's not like a Bayer aspirin, where every dose is chemically identical. CAR T cells are manufactured from a patient's own T cells, so they carry all of that patient's ancestral genetic polymorphisms and mutations. Because of this, how CAR T cells behave varies from patient to patient in ways we can't totally understand.

Going into this study, we had some idea of which patients were at high risk of toxicity. Patients with a large tumor burden, or higher baseline levels of inflammation, tend to be more likely to have serious side effects. But beyond that, we didn't have a good understanding of why some patients go through catastrophic toxicities and others don't. It's a little like COVID-19, where some patients end up on a ventilator with severe immune inflammation and others sail through with minimal side effects.

Our hypothesis was founded on a rare disease called hemophagocytic lymphohistiocytosis (HLH), which classically affects pediatric patients, usually kids under age 5. In that disease, patients are born with defects in how their lymphocytes, T cells, and NK cells kill virally infected cells. When those lymphocytes can't kill effectively, they become continually activated and trigger an inflammatory cascade that can become life-threatening.

In the CAR T cell field, we've long recognized that the biochemical features of HLH and CRS after CAR T cell therapy are very similar, and the toxicity mechanism is thought to be similar as well: in HLH it's an inactivated T cell trying to kill a virally infected target, and in CAR T cells it's a CAR T cell trying to kill a tumor target, but the end result is the same, cytokines are produced that activate the myeloid system and drive high fevers, low blood pressure, and other symptoms.

Since HLH has a genetic basis, and we now know less severe forms of these mutations can present later in life, we hypothesized that some of those less severe variants might be present in patients who develop severe toxicity after CAR T cell therapy.

HCPLive: Can you walk our audience through what you eventually found?

Leick: We partnered with Kite Pharma, the largest producer of CAR T cell products, which gave us access to DNA samples, CAR T cell level measurements, clinical data, and serum biochemical data from patients across the ZUMA-1 and ZUMA-7 clinical trials of axicabtagene ciloleucel in patients with aggressive lymphoma.

We performed whole genome sequencing, but because these early CAR T cell trials only included around 100 patients, we needed a narrow hypothesis rather than an unrestricted genome-wide search. So we focused on fewer than 20 genes associated with HLH.

We defined a combined toxicity endpoint that captured patients with high-grade CRS, high-grade neurologic toxicity, or those who received tocilizumab, the antidote for CRS, since those patients likely would have had higher-grade toxicity without it. Among that pool, we found 6 patients with variants in the gene STXBP2, which is involved in lymphocyte degranulation, compared with none in the control cohort without high-grade toxicity.

We followed up with wet lab studies, comparing CAR T cells from patients with the variant to CAR T cells from controls, and found that cells with STXBP2 variants became more inflamed when they encountered target cells. We also introduced the CAR T cell gene and the specific variants into healthy donor T cells and saw the same result in vitro.

We then sought to replicate this finding in the ZUMA-7 cohort, an independent group of patients treated with the same CAR T cell product, but earlier in their line of therapy and with less inflammation overall. We didn't see the same STXBP2 finding there, which we think reflects that these patients had less severe disease and fewer prior treatments.

With both cohorts available, we could also look beyond toxicity. We found that variants in PTPN22 were strongly correlated with CAR T cell proliferation across both ZUMA-1 and ZUMA-7. PTPN22 helps regulate normal T cell biology and serves as a brake on T cell receptor activation, so there's a plausible mechanism: lose the brake, and T cells become more activated and proliferate more.

In fact, 2 patients had the exact same mutation, and that led them to have the highest CAR T cell expansion of all the 100 patients in ZUMA-1, suggesting that there is some sort of gain of function advantage to having that mutation in the context of CAR T.

HCPLive: Is there anything else you'd like to highlight, including any limitations?

Leick: I'd add a couple of limitations. Germline genetics analysis requires comparing genetic variation within populations of comparable ancestry, and because clinical trial accrual in the United States skews toward patients of European ancestry, our study was limited mostly to that population. We only had a couple of non-European patients, not enough to support any meaningful analysis. We'd love to study this in more diverse populations.

Our study was also focused on one product, axicabtagene ciloleucel, so these findings may not apply to other CAR T cell products or disease types beyond lymphoma. We're working in the lab now to understand whether this applies more broadly.

As for clinical implications, this doesn't mean patients with these variants got severely ill. No one died in either trial, and even knowing a patient had these variants, we likely wouldn't manage them differently since we're already fairly good at treating these toxicities. Where I think this becomes more relevant is in choosing donor T cells for allogeneic CAR T cell products, since a single donor's cells can be used to manufacture products for dozens or hundreds of patients, and in using natural human genetic variation as an alternative to CRISPR-based screens to identify genes, like PTPN22, that could be engineered into future CAR T cell products to improve efficacy.

Editor’s Note: Leick reports relevant disclosures with BioNTech, Cabaletta Bio, and Adaptimmune.

Leick MB, Sun B, Birocchi F, et al. Genomic correlates of clinical CAR T cell activity. 

. 2026;11(121). doi:10.1126/sciimmunol.aef4134

Neelapu SS, Locke FL, Bartlett NL, et al. Axicabtagene ciloleucel CAR T-cell therapy in refractory large B-cell lymphoma. N Engl J Med. 2017;377(26):2531-2544.

Articles

Whole-genome sequencing in aggressive lymphoma treated with axicabtagene ciloleucel implicates germline variation as a contributor to interpatient heterogeneity in CAR T efficacy and toxicity. An HLH-informed candidate approach associated STXBP2 variants with severe CRS/neurologic toxicity in ZUMA-1, supported by in vitro evidence of impaired degranulation and heightened cytokine release. Genome-wide analyses identified ADAMTSL3 variants linked to reduced toxicity and PTPN22 variants correlated with greater CAR T expansion across ZUMA-1 and ZUMA-7. Generalizability is constrained by ancestry and product specificity.

New research links germline genetic variants to CAR T cell toxicity risk and expansion in lymphoma patients treated with axi-cel.