Skip to content

GD2-CAR-T for pediatric DIPG and diffuse midline glioma — Stanford Ph1 update

Practice-defining pediatric CAR-T cohort update: autologous GD2-targeted CAR-T (IV + intracerebroventricular) in H3K27M-mutant diffuse midline glioma and DIPG.

  • Speaker(s): Michelle Monje, MD, PhD (also Brain 26 co-chair and gave the neuroscience-of-glioma plenary thread); Crystal L. Mackall, MD (cell-therapy lead)
  • Affiliation: Stanford University (Department of Neurology & Neurological Sciences; Center for Cancer Cell Therapy); Howard Hughes Medical Institute
  • Anchor type: cell therapy / trial
  • Day / session: Day 2 (Tue Mar 24, 2026), Pediatric / DIPG session
  • Status: post-meeting; details captured from AACR Brain 26 program notes, prior Stanford Ph1 reports (Nature 2022 / Nature Medicine 2024 follow-ups), and Monje/Mackall public statements; expanded 2026 cohort numerics TBD pending abstract supplement

Headline finding

Updated, expanded cohort of autologous GD2-CAR-T (4-1BB costim) in H3K27M-mutant DIPG and diffuse midline glioma (DMG), extending the earlier Stanford report (~9/10 evaluable clinical-benefit signal) into a larger multi-site Ph1 with longer survival follow-up and biomarker correlates. This remains the strongest CAR-T efficacy signal in any pediatric CNS solid tumor, in a disease (H3K27M DMG) where nothing has been practice-changing since ONC201 / dordaviprone emerged. Specific updated cohort N, PFS, and OS from the Brain 26 talk TBD.

Mechanism / methodology

Autologous patient T cells transduced with a GD2-directed CAR (4-1BB / CD3ζ); GD2 is uniformly and highly expressed on H3K27M DMG. Initial cell dose is delivered intravenously after lymphodepletion, followed by intracerebroventricular re-dosing via Ommaya to address CNS pharmacokinetics. ICANS-class neuroinflammatory toxicity is managed proactively given tumor location in brainstem / thalamus. Biomarker workup includes CSF cytokine profiles, TCR-seq of expanded CAR-T product, and serial MRI volumetric response.

Mechanism / methodology deep dive

Underlying biology. H3K27M mutation in histone H3.3 / H3.1 globally reduces H3K27me3, locking DMG cells in an OPC-like / neural-precursor-like state with uniformly high GD2 disialoganglioside surface expression — a developmentally restricted antigen otherwise largely absent from postnatal CNS at high density. The H3K27M-driven OPC-like state also confers responsiveness to neural-circuit / activity-dependent inputs (the Monje thread linking neuron-glioma synapses to growth).

Therapeutic design. Autologous patient T cells lentivirally transduced with a GD2-CAR using the 14g2a-derived scFv, a CD8α hinge/transmembrane, 4-1BB costimulatory domain, and CD3ζ (the Mackall lab construct also used in pediatric neuroblastoma). Manufacturing is standard 9–14-day expansion. Lymphodepletion with fludarabine + cyclophosphamide precedes a single IV dose, followed by Ommaya-administered ICV re-doses at escalating cadence. Tocilizumab, anakinra, and intrathecal steroids are pre-positioned for tumor-inflammation-associated neurotoxicity (TIAN), a Monje-defined entity distinct from classical ICANS.

Resistance / failure modes. Documented: TIAN-driven brainstem swelling requiring aggressive steroid / anakinra and occasional EVD placement; partial / transient radiographic responses; on-treatment progression without clear antigen loss. Anticipated escape: cell-state drift away from OPC-like (toward AC-like / MES-like) reducing GD2 surface density; T-cell exhaustion in CNS TME.

Translational gaps. Multi-site replication of Stanford manufacturing and TIAN management capacity remains the key barrier; sequencing with ONC201/dordaviprone, RT, and re-irradiation TBD.

Where it fits in the corpus

  • AACR 2026: Monje/Mackall pediatric-CAR-T sessions echoed in the AACR Annual neuro-onc and pediatric tracks in San Diego.
  • SITC 2026: the cellular-therapies track frames GD2-DMG as the lead pediatric solid-tumor CAR-T signal.
  • Related Brain 26 entries: upenn-icv-bivalent-cart.md, city-of-hope-il13ra2-cart.md, neuroscience-of-gliomas-tam.md, parrinello-injury-programs.md.

Open questions

  • Can the Stanford signal be replicated across multi-institution sites with comparable manufacturing and ICV-management capacity?
  • Combinatorial space: where do ONC201 / dordaviprone, RT, and GD2-CAR-T sequence — and does cell-state plasticity (Suva framework) create predictable escape via antigen / lineage drift?