Off-the-shelf CRISPR CAR-T tackles brain tumours

“In summary, this study of five subjects with recurrent high-grade glioma demonstrated the feasibility of treating these subjects with engineered allogeneic CAR-T cells via intrathecal delivery to tumor sites”Xuetao Li et al.

Patients needed a Karnofsky Performance Score of ≥40; all five carried adverse molecular features (IDH wild-type and unmethylated MGMT promoter). The mean age was 52 years, and four patients had baseline KPS below 70, indicating substantial functional impairment. The study ran from August 2020 to July 2022, with COVID-19 restrictions curtailing planned enrollment from twelve to five participants.

The universal CAR-T product, MT026, was engineered by disrupting the TRAC locus to silence the endogenous T-cell receptor and editing HLA class I α-chains (HLA-A/B/C) to damp host-versus-graft reactions – an alternative to β2-microglobulin knockout that preserves some class I expression and may reduce natural-killer (NK) cell vulnerability. In manufactured lots, 45.3% of cells expressed the CAR, with 99.52% TRAC loss and 95.23% HLA-I reduction by flow cytometry, whilst activation markers CD69 and CD137 were retained upon target engagement.

Preclinical assays supporting the clinical trial outline the trade-offs of partial HLA class I loss. In 24-hour co-cultures, allogeneic NK cells lysed only 20–30% of MT026 cells, compared with approximately 60% lysis of HLA-null K562 controls and minimal killing of wild-type T cells. Cytotoxicity against U251 glioma targets matched conventional CAR-T across effector-to-target ratios, and MT026 progressively cleared patient-derived organoids. In orthotopic xenografts, local MT026 dosing suppressed tumour signal and modestly extended survival compared with untreated controls, indicating that edited cells can remain potent despite immune-evasion edits.

Because multiplex editing raises safety concerns, the group profiled off-target effects using four orthogonal methods – AID-seq, GUIDE-seq, iGUIDE-seq, and PEM-seq. Across dual-edited samples, they detected few off-target sites and only one coding-region hit (MUC4) at a clone fraction of less than 0.1%; on-target translocations were low frequency and below the pre-specified concern thresholds.

»Collectively, these multilayered assessments demonstrate: 1) sgRNA-specific off-target profiles with limited coding region impact, 2) low structural variation risk at therapeutic thresholds, and 3) precise on-target editing, supporting the safety and clinical translatability of this dual-edited UCAR-T product,« the authors state in their Nature Communications paper.

In the trial, MT026 was administered monthly by lumbar puncture at 1.0–3.0×10^7 cells without lymphodepletion, with patients receiving four to nine infusions over 2.6–7.9 months. Two neuroradiologists, blinded to the outcome, read serial MRIs according to the Immunotherapy Response Assessment in Neuro-Oncology (iRANO) criteria. The primary endpoint was occurrence and severity of adverse events; secondary endpoints included overall survival, progression-free survival, objective response rate, duration of response and pharmacokinetic parameters.

“No grade ≥3 AEs, SAEs, immune effector cell–associated neurotoxicity syndrome (ICANS), GvHD, or infections were reported during the study”Xuetao Li et al.

Pharmacokinetics and cytokines suggested a localised response. CAR DNA in cerebrospinal fluid rose one day post-dose, peaked at two to four days, and remained detectable between cycles; no CAR DNA appeared in blood. IL-6 and other cytokines increased in CSF, with CSF IL-6 concentrations 79- to 647-fold higher than serum, consistent with activity confined to tumour-proximal compartments rather than systemic inflammation.

The primary endpoint was met: tolerability was favourable across repeated intrathecal injections. Study-drug-related events were grade 1–2 and transient – fever (4/5), hypoxia (2/5), vomiting (2/5) and headache (1/5) – with no high-grade toxicities, infections or alloreactivity syndromes. In the authors' words:

»No grade ≥3 AEs, SAEs, immune effector cell–associated neurotoxicity syndrome (ICANS), GvHD, or infections were reported during the study,« the authors emphasise.

Secondary efficacy endpoints, whilst preliminary given the small cohort, were notable for this setting: all five patients had radiographic shrinkage; one achieved a complete response and three partial responses, giving an objective response rate of 80% (95% CI 37.6–96.4). Mean duration of response was 3.4 months, and mean progression-free survival was 4.7 months; calculated from recurrence, mean overall survival was 13.1 months (maximum 33.2 months), though cross-study comparisons are fraught. Antigen loss emerged as a likely resistance route in one case, and two patients later received other investigational CAR-T products, complicating attribution.

»In summary, this study of five subjects with recurrent high-grade glioma demonstrated the feasibility of treating these subjects with engineered allogeneic CAR-T cells via intrathecal delivery to tumor sites. The absence of therapy-related serious adverse events and the potential clinical benefit warrant further investigation,« the authors conclude.

Taken together, the phase I data show that a dual-edited, off-the-shelf CAR-T can be delivered by lumbar puncture, persist in CSF between cycles, and induce tumour shrinkage without systemic toxicity – in a handful of patients, with caveats that warrant a larger, controlled trial. The immunological hurdles ahead are clear: NK-cell-mediated rejection and target antigen escape. The authors sketch potential refinements – such as augmenting HLA-E or CD47 and knocking out CIITA – to further tune the graft's visibility.

The study was led by Zhong Wang, Yuzhang Wu and Yulun Huang from the Hospital of Soochow University and the Army Medical University, Chongqing, China. It was published in Nature Communications on 6 January 2026.