Study Shows Potential for Low-Dose CAR T, With Implications for Increased Access

Chimeric antigen receptor (CAR) T-cell therapies are normally manufactured and infused at high doses to maximize a patient’s odds of achieving a durable remission. But manufacturing can fail to hit the target cell count, producing “out-of-specification” (OOS) products that are often withheld from patients or given only at a physician's discretion. Curiously, some patients who receive far lower CAR-T doses than intended still respond well to therapy. This has not been understood, until now.

A study, led by researchers at Charité and Heidelberg University Hospital and appearing in Nature Communications this week, set out to characterize exactly what characteristics distinguish these low-dose responders.

The researchers drew on the HD-CAR-1 phase I/II trial (NCT03676504), which tested third-generation anti-CD19 CAR T cells across a wide range of doses in patients with relapsed or refractory B-cell lymphomas. Among 28 patients, they identified 3 groups: high-dose responders (n=8), low-dose non-responders (n=12), and the group of interest, low-dose responders (n=6), the patients who achieved robust, durable clinical responses despite receiving CAR T-cell therapy doses far below the typical range. Notably, response in this trial was unrelated to sex, age, prior stem cell transplant, or disease type. The overall trial results were themselves notable: 40% of lymphoma patients achieved long-term responses, with an excellent safety profile; patients showed only low-grade cytokine release syndrome and no serious neurotoxicity.

Using 36-plex full-spectrum flow cytometry on the infused CAR-T products, the team found that low-dose responders had a cellular signature distinct from high-dose responders. Whereas high-dose responders' products were enriched in naïve and central-memory-like T cells, which is consistent with prior literature linking these "young" cell states to durable expansion, low-dose responders' products were instead dominated by effector and effector-memory-like CAR-T cells that showed no signs of exhaustion.

The authors say this makes biological sense: with few cells to work with, low-dose responders seem to depend on cells that can kill tumor targets immediately, rather than cells built for long-term persistence and expansion.

“For patients receiving low CAR-T cell doses, the number of functional effector-memory and effector-like CAR T cells mediating immediate and potent target cell killing represents the critical factor,” the authors write.

A key insight was that absolute cell counts, not relative percentages, mattered most. Low-dose responders and high-dose responders had comparable absolute numbers of effector-like CAR-T cells despite the low-dose group receiving a much smaller total infusion — suggesting a response requires crossing some minimum threshold of functional effector cells, regardless of overall dose. The team distilled this into a simple, practical biomarker: CAR-T cells lacking both CD27 and CD39 (CD27⁻CD39⁻), which identifies non-exhausted effector cells. This combination outperformed previously published biomarker panels at distinguishing responders from non-responders, and follow-up lab experiments confirmed that CD27⁻CD39⁻ cells are stable, potent killers with limited ability to proliferate further— a bona fide terminal effector state, while CD27⁺CD39⁺ cells were more proliferative and could give rise to CD27⁻CD39⁻ progeny.

Blood Before Manufacturing Matters

The study then asked what determines whether a low-dose product turns out functional or dysfunctional in the first place, and traced this back to the patient's blood before manufacturing even begins. Low-dose responders' pre-manufacturing blood was enriched in CD8 T cells with effector/effector-memory phenotypes and relatively low in myeloid cells. Low-dose non-responders, in contrast, had blood rich in monocytes and dendritic cells with a "regulatory" (M2-like) molecular profile: reduced antigen-presentation machinery, blunted inflammatory signaling, and elevated complement-regulatory proteins like CD55 and clusterin. Using single-cell proteogenomics, the authors showed these M2-like myeloid programs actively promote CAR-T dysfunction: when they deliberately exposed healthy donor cells to M2-polarizing cytokines (IL-4, IL-10, IL-13, TGF-β) before CAR-T manufacturing, the resulting CAR-T cells acquired the dysfunctional CD39⁺ phenotype and killed leukemia cells less effectively.

Of note, this effect was reversible—washing out the cytokines before manufacturing prevented the dysfunctional phenotype from developing. This suggests that manufacturing conditions, not fixed patient biology, drive this deficit.

“Consistently, CAR-T cells generated after cytokine washout displayed cytotoxic killing capacities comparable to the control,” the authors wrote. “As expected, cytotoxic activity was significantly associated with the abundance of CD39⁺ CAR-T cells.”

Finally, the CD27⁻CD39⁻ biomarker was validated in an independent 42-patient cohort that spanned various types of lymphoma—mantle cell lymphoma, diffuse large B-cell lymphoma, and follicular lymphoma—as well as multiple myeloma. CAR designs, and different commercial products that included lisocabtagene maraleucel (Breyanzi; Bristol Myers Squibb), axicabtagene ciloleucel (Yescarta; Kite/Gilead), tisagenlecleucel (Kymriah; Novartis) and ciltacabtagene autoleucel (Carvykti; Johnson & Johnson), where it again reliably distinguished responders from nonresponders.

Clinical and Access Implications

For payers and for patients, the findings are significant. Recent analyses in the United States suggest only about 20% of the patients eligible for CAR T-cell therapy have access, for reasons of cost, proximity to a treatment center, and availability of a caregiver. Community clinics are reluctant to take on the risk associated with CAR T. But these results could change that equation.

“For a considerable proportion of patients eligible for CAR-T cell therapy, the number of manufactured CAR-T cells may not meet current end product specifications, thus resulting in an OOS product,” the authors wrote. “As a result, these patients may not receive treatment, or CAR-T cells may only be administered at the discretion of the treating physician, leading to uncertain therapeutic outcomes.”

Although some reports show CAR T products can be generated in as little as 3 days, “these accelerated protocols often fail to produce sufficient cell numbers within this timeframe,” the authors continued. “This, in particular, can exacerbate the risk of OOS for already challenging patient samples and clinical situations. On the other hand, pharmaceutical enterprises manufacturing OOS products face the challenge of non reimbursement, which introduces major financial risks.”

However, if a simple 2-marker flow cytometry panel can flag which "failed," under-dosed CAR-T products still carry real therapeutic potential, it could let physicians salvage OOS products that would otherwise be discarded, expanding patient access to a very expensive therapy, reducing manufacturing waste and cost, and informing revised regulatory thresholds for what counts as an adequate CAR-T dose.

Reference

Yousefian S, Schubert ML, Minafra AR, et al. A distinct CAR-T cell phenotype mediates therapeutic response at limited doses. Nature Comm. 2026;17:7589. DOI: 10.1038/s41467-026-76068-4