Recurrent glioblastoma (rGBM) remains a major unmet clinical problem with limited effective therapies. This phase 1 study investigated locoregional administration of autologous chimeric antigen receptor T cells targeting B7‑H3 (TX103) delivered directly into the intracranial compartment. The trial aimed to define safety, tolerability, pharmacokinetics and preliminary efficacy of intracranial CAR‑T infusion in patients with B7‑H3‑positive rGBM.
This was an open‑label, 3 + 3 dose‑escalation phase 1 trial enrolling adults aged 18–75 years whose recurrent glioblastoma expressed B7‑H3 in at least 30% of tumor cells. Primary endpoints were safety, maximum tolerated dose (MTD) and recommended phase 2 dose (RP2D). Secondary endpoints included survival, pharmacokinetics and immunological response. The trial is registered on ClinicalTrials.gov under NCT05241392.
TX103 is an autologous CAR‑T product engineered to recognize B7‑H3. In this study the product was delivered intracranially, allowing locoregional exposure of the tumor and cerebrospinal compartment to the infused cells. The intracranial route enabled assessment of on‑target effects in the central nervous system while monitoring peripheral circulation for systemic exposure.
Three dose levels were evaluated per infusion: DL1 = 2 × 10^7 cells, DL2 = 6 × 10^7 cells and DL3 = 1.5 × 10^8 cells. Fifteen patients received a total of 72 intracranial infusions; 13 patients underwent repeated infusions. Based on safety and tolerability across dose cohorts, the recommended phase 2 dose (RP2D) was designated as DL2 (6 × 10^7 cells per infusion).
TX103 intracranial infusion was overall well tolerated. No dose‑limiting toxicities (DLTs) were observed and no maximum tolerated dose (MTD) was reached. Treatment‑related adverse events (TRAEs) reported included predominantly low‑grade events: cytokine release syndrome in 86.7% of patients, sinus tachycardia in 53.3%, vomiting in 53.3%, hypertension in 53.3% and elevated intracranial pressure in 46.7%.
Three grade 3 TRAEs considered serious occurred: elevated intracranial pressure, epilepsy and depressed consciousness; two of these events occurred in patients treated at the highest dose level (DL3). The study reported no cumulative toxicity after repeated intracranial infusions.
Among patients with measurable disease (14 patients reported), disease control (stable disease or better) was achieved in 8 patients. The dataset included one complete response that was sustained through the most recent follow‑up reported in the publication.
Survival outcomes included a 12‑month overall survival (OS) rate of 66.7% and a median OS of 19.1 months from the time of first infusion; the 95% confidence interval was reported as 8.93 months to not reached. These efficacy signals were presented as preliminary outcomes in this phase 1 cohort.
Pharmacokinetic and immunological assessments showed a marked increase in CAR transgene copy numbers and cytokine concentrations in cerebrospinal fluid (CSF) after intracranial TX103 infusion. Peripheral (systemic) activity was minimal, consistent with a locoregional effect. The CSF cytokine rise and CAR detection indicate engagement of the infused cells within the central nervous system compartment.
The study additionally noted absence of cumulative toxicity with repeated intracranial dosing, supporting feasibility of multiple administrations in this setting.
Intracranial infusion of autologous B7‑H3‑targeting CAR‑T cells (TX103) demonstrated an acceptable safety profile in this phase 1 3 + 3 dose‑escalation trial, with no DLTs or identified MTD. Early signals of anti‑tumor activity included disease control in more than half of patients with measurable disease and one sustained complete response. CSF pharmacodynamics showed on‑target CAR presence and cytokine induction with minimal peripheral exposure.
Based on safety and preliminary efficacy, the authors support advancing TX103 to a phase 2 evaluation at the RP2D (6 × 10^7 cells per infusion). These findings add to a growing body of locoregional CAR‑T studies in central nervous system tumors and provide rationale for further controlled evaluation.
The clinical study protocol is provided in the Supplementary Information of the article. Raw patient‑level clinical data are not publicly available due to privacy and consent restrictions. De‑identified clinical data, processed pharmacological data and other data underlying reported results are provided in the Source Data files associated with the publication.