The development of CAR T cells for patients with CNS malignancies

Abstract

Chimeric antigen receptor (CAR) T cells have become standard-of-care therapies for patients with certain relapsed and/or refractory haematological malignancies over the past decade. However, this approach remains largely ineffective in patients with solid tumours, in part owing to limited CAR T cell persistence, the immunosuppressive tumour microenvironment of many solid tumours and limited trafficking of CAR T cells into tumours. Central nervous system (CNS) tumours, many of which are associated with a poor prognosis and require new treatment approaches, present additional challenges such as the presence of the blood–brain barrier as well as concerns over treatment-related neurotoxicities. Despite these difficulties, clinical trials involving both adult and paediatric patients with primary CNS tumours have provided signals of efficacy. In this Review, we discuss completed, ongoing and anticipated trials testing CAR T cells in patients with CNS tumours. We also highlight the most promising preclinical developments that might lead to novel clinical approaches in this area.

Introduction

Primary tumours developing in the central nervous system (CNS), such as glioblastoma, present unique challenges in terms of both clinical management and therapeutic development. Primary CNS tumours comprise approximately 1% of tumours diagnosed annually in the USA¹. Annual deaths from CNS tumours are equivalent to 76% of the annual diagnoses, highlighting the poor survival outcomes of many patients with these tumours compared to those with most other cancer types, including lung, breast and prostate cancers¹. Furthermore, the median overall survival (OS) durations of patients with CNS tumours have not improved substantially over the past decades. This scenario is in contrast with many extracranial tumour types, such as breast cancer and lung cancer, both of which have seen steady improvements in median survival outcomes over this period of time². These sustained poor survival outcomes showcase the unmet need for new treatment paradigms for CNS tumours. Metastatic disease to the CNS is associated with a similarly poor prognosis, with a median OS duration of 13–27 months depending on the primary cancer, the availability of effective systemic agents with CNS activity, and the extent and status of extracranial disease^3,4.

Chimeric antigen receptor (CAR) T cells are a form of adoptive cell therapy that redirects T cells to target tumour cells⁵. Autologous strategies involve harvesting of the patient’s T cells via apheresis, followed by engineering to express at least one type of CAR targeting one or more known tumour-associated antigens (TAAs), followed by in vitro expansion, and infusion of the modified cells back into the same patient. Currently, seven different CAR T cell therapies have been approved by the FDA, all for patients with B cell malignancies or multiple myeloma⁶. However, generating clinically significant responses in patients with advanced-stage solid tumours has thus far proved much more challenging⁷. These challenges include limited trafficking to the tumour and penetrance into the tumour mass, probably reflecting an immunosuppressive tumour microenvironment (TME), as well as tumour heterogeneity leading to loss of target antigen expression. However, several trials testing CAR T cells in patients with primary CNS malignancies, including those with glioblastoma and diffuse intrinsic pontine glioma (DIPG), have provided promising signals of efficacy. Despite generally limited progress, these positive results have been accompanied by encouraging findings with CAR T cells in patients with other solid tumours, such as neuroblastoma^8,9 and gastric cancer¹⁰. In this Review, we describe the available data from clinical trials testing CAR T cells in patients with CNS malignancies (Fig. 1) and provide an overview of future directions for this field (Fig. 2).

Fig. 1 |. Current and future CAR designs and targets.

a–h, Chimeric antigen receptor (CAR) constructs and their cognate target antigens tested in completed and/or planned clinical trials involving patients with central nervous system malignancies, including EGFR (panel a), EGFRvIII (panel b), IL-13α2 (panel c), GD2 (panel d), ephrinA2 (panel e), HER2 (panel f), CD19 (panel g) and B7-H3 (panel h). Axi-cel, axicabtagene ciloleucel; IL-13 receptor subunit α2, IL-13α2; Tisa-cel, tisagenlecleucel; TM, transmembrane.

Fig. 2 |. Future CAR T cell therapies for patients with CNS tumours.

a, Future chimeric antigen receptor (CAR) T cell designs include the use of multiple CAR constructs, such as IL-13 receptor subunit α2 (IL-13Rα2)–EGFR-targeted CAR T cells, which are capable of recognizing different targets and might address antigen heterogeneity and potentially mitigate antigen-negative recurrence. b, Combination therapies and/or CAR T cells capable of secreting additional molecules and/or targeted agents, such as T cell-engaging antibody molecule targeting EGFR (TEAM-E) CAR T cells, might have similar beneficial effects. c,d, More potent CAR T cells, such as ‘armoured’ designs capable of secreting specific cytokines, such as IL-18 (c), or those equipped with a dominant-negative TGFβ receptor (d), might address the immunosuppressive tumour microenvironment. e, Cellular engineering approaches, such as the synNotch system, allow combination targeting, addressing both on-target–off-tumour concerns and antigen heterogeneity. TCR, T cell receptor.

CAR T cells in adults with glioma

As the most common malignant primary brain tumour in adults¹¹, glioma has thus far been the main focus of attempts to develop CAR T cells as treatments for CNS tumours. Specifically, clinical studies of cellular therapy for such cancers have mostly involved patients with glioblastoma, a WHO grade 4 diffuse glioma defined by the absence of mutations in the isocitrate dehydrogenase gene (IDH) and either (1) microvascular proliferation or necrosis, or (2) at least one of the following molecular alterations: TERT promoter mutations, EGFR amplifications or co-occurring gain of chromosome 7 and loss chromosome 10 (+7/−10)¹². Despite an aggressive standard-of-care first-line therapy comprising maximal safe surgical resection, adjuvant radiotherapy and temozolomide chemotherapy¹³, glioblastoma remains incurable and all patients will inevitably have disease relapse. No therapeutic intervention for recurrent disease has ever been shown to improve OS in randomized phase III trials¹⁴ and median OS remains dismal at 12–15 months¹¹.

In the highly challenging setting of recurrent glioblastoma, several phase I or first-in-human trials have provided signals of therapeutic effects of CAR T cell therapy^15–17. These trials and the majority of other available clinical studies in glioblastoma have focused on approaches targeting EGFR or IL-13 receptor subunit α2 (IL-13Rα2), although safety and signals of biological activity have also been demonstrated for CAR T cells targeting other antigens.

IL-13Rα2-targeted CAR T cells

IL-13Rα2 is a cancer–testis antigen expressed in several human cancers, including in ~75% of glioblastomas, that is not expressed in non-malignant tissues apart from adult testes^18,19. IL-13 signalling via this receptor has been shown to mediate glioblastoma cell migration and invasion, and is associated with a mesenchymal gene expression signature and an inferior prognosis^20,21. Evidence of the safety and feasibility of IL-13Rα2-targeted CAR T cells was initially provided by a report describing the first three patients with recurrent glioblastoma who received the first-generation version of this CAR T cell product via repeated intracranial administration, with grade ≥3 adverse events limited to headache and one neurological event at higher dose levels²². Given the local method of administration, and lack of data suggesting that systemic lymphodepletion would augment efficacy in this setting, lymphodepleting chemotherapy was not administered. In a subsequent case report, these investigators described a complete response in a patient with glioblastoma with leptomeningeal disease who received a second-generation version of the product, equipped with a 4-1BB co-stimulatory domain²³.

In 2024, the final results of the phase I trial testing this second-generation product in 58 response-evaluable patients with recurrent high-grade glioma (72% with glioblastoma) were reported. Treatment was well tolerated with no dose-limiting toxicities (DLTs), and grade ≥3 treatment-related toxicities in only two patients (grade 3 encephalopathy and grade 3 ataxia)¹⁶. Stable disease or better was observed in 50% of patients (29/58), with two partial responses, one complete response, and a second complete response after off-protocol administration of additional cycles of the CAR T cells. Despite these encouraging signals in a small subset, median OS among all patients was 7.7 months¹⁶, which is similar to the historical median OS durations of patients with recurrent glioblastoma²⁴. Importantly, the authors found a positive association between pretreatment intratumoural CD3⁺ T cell levels and OS, suggesting that an immunologically ‘hot’ TME featuring high levels of T cell infiltration might have a role in responsiveness to CAR T cell therapy¹⁶. Locoregionally delivered IL-13Rα2-targeted CAR T cells continue to be evaluated in patients with recurrent glioblastoma in clinical trials, both in combination with nivolumab and ipilimumab (NCT04003649) and in patients with leptomeningeal glioblastoma, ependymoma or medulloblastoma (NCT04661384).

EGFR-targeted CAR T cells

Over half of all glioblastomas harbour EGFR amplifications^25,26. Approximately 50% of these EGFR-amplified tumours also express EGFRvIII²⁷, a truncated form of EGFR arising from deletion of EGFR exons 2–7, resulting in a constitutively activated receptor with a tumour-specific extracellular epitope, and up to a third of EGFR-amplified tumours will also have missense mutations affecting the extracellular domain^26,28. Given this high prevalence of EGFR alterations and their roles in tumour growth, invasion, therapy resistance and immune evasion²⁹, EGFR and EGFRvIII have long been recognized as potential targets for CAR T cell therapy. Early evidence supporting this approach was provided by a first-in-human phase I trial including ten patients with recurrent glioblastoma who received a single dose of peripheral blood-infused EGFRvIII-directed CAR T cells produced using a lentivirally delivered second-generation CAR construct with a 4-1BB co-stimulatory domain³⁰. This approach was deemed safe and well-tolerated, albeit none of the patients had evidence of tumour regression at 1 month after infusion (one patient had stable disease lasting >18 months). Notably, seven of the ten patients in this study required surgical intervention after CAR T cell treatment, enabling tissue-specific analysis of CAR T cell trafficking and effects on the TME. Findings from these investigations³⁰ suggested that the CAR T cells had effectively trafficked and expanded in situ within active regions of the tumours, with nearly all patients having specific loss or decreased expression of EGFRvIII in resection specimens obtained after CAR T cell infusion. This analysis also revealed substantial infiltration of regulatory T (T_reg) cells as well as consistent upregulation of immune checkpoints and soluble immunosuppressive molecules, including indoleamine 2,3-dioxygenase 1, PD-L1, TGFβ and IL-10. These observations suggest that CAR T cells targeting EGFRvIII⁺ cancer cells induce a compensatory immunosuppressive response in the TME.

Similar results were observed in a phase I trial testing EGFRvIII-targeted CAR T cells utilizing a third-generation construct incorporating an additional 4-1BB co-stimulatory domain³¹. Unlike the initial first-in-human trial testing EGFRvIII-directed CAR T cells³⁰, patients in this trial received lymphodepleting chemotherapy with fludarabine and cyclophosphamide prior to CAR T cell administration. Among 18 patients treated, median progression-free survival (PFS) was 1.3 months and median OS was 6.9 months, albeit with one patient surviving for nearly 5 years. Notably, one patient died of acute hypoxic respiratory failure beginning approximately 1 h after receiving a 6 × 10¹⁰ dose of CAR T cells. The underlying mechanism of this toxicity is uncertain, although given the extremely rapid onset, this effect is unlikely to have been related to on-target (EGFR), off-tumour toxicity, and was more likely to have been related to an acute inflammatory reaction following massive cytokine release or congestion of the pulmonary vasculature with activated T cells.

On the basis of previous observations that EGFRvIII-targeted CAR T cells induce substantial upregulation of PD-L1 in the TME³⁰ and that EGFRvIII expression is often lost at disease recurrence³², the combination of EGFRvIII-specific CAR T cells plus the anti-PD-1 antibody pembrolizumab was tested in seven patients with newly diagnosed glioblastoma³³. Following maximal safe surgical resection and a 3-week course of radiotherapy (total dose 40 Gy) without temozolomide, patients received three cycles of combination therapy with peripherally administered CAR T cells (2 × 10⁸ cells) plus pembrolizumab followed by a fourth dose of pembrolizumab monotherapy. The combination was safe with no DLTs observed. However, the median PFS and OS durations of 5.2 and 11.8 months, respectively, with no notable exceptional responses suggest a lack of clinical activity. Comparisons of the TME of tumour material resected before versus after treatment demonstrated substantial evolution of the immune TME, with a greater proportion of tumour-associated macrophages and exhausted, regulatory and interferon (IFN)-stimulated T cells observed at relapse. Notably, a positive correlation was detected between the extent of IFN signalling in T cells at the time of disease progression and OS, consistent with the previous observation that a hot TME might promote responsiveness to IL-13Rα2-targeted CAR T cells¹⁶.

Other targets

Beyond IL-13Rα2 and EGFR, CAR T cells targeting various other antigens have previously been, or are currently being, evaluated in clinical trials involving patients with CNS tumours. A phase I trial involving 17 patients with recurrent glioblastoma tested peripherally infused CAR-modified virus-specific T cells targeting HER2 (ref. 34), a receptor tyrosine kinase expressed in up to 80% of patients, depending on the detection method³⁵. No lymphodepleting chemotherapy was administered and repeat CAR T cell dosing was permitted. Infusions were well tolerated, and no DLTs were observed. Of 16 evaluable patients, one had a partial response lasting >9 months, and seven had stable disease lasting between 8 weeks and 29 months (with three of these remaining progression-free for >2 years). HER2 continues to be explored as a CAR T cell target in an ongoing phase I trial involving patients with HER2-positive primary or metastatic CNS tumours (NCT02442297). Although no on-target, off-tumour toxicities were reported³⁴, wide-spread low-level expression of HER2 by non-malignant tissues remains a concern that must be accounted for in future studies.

Another prominent target is GD2, a disialoganglioside with high levels of expression on cancer cells of neuroectodermal origin with minimal expression in non-malignant tissues³⁶. GD2-targeted CAR T cells have been evaluated both in patients with diffuse midline glioma (DMG)³⁷ and in those with glioblastoma. In a phase I trial, eight patients with GD2-positive glioblastoma received intravenously administered GD2-specific CAR T cells with or without additional intracavitary administration³⁸. This study included both adult and paediatric patients with recurrent glioblastoma, although paediatric patients are no longer classified as having glioblastoma according to 2021 WHO criteria. and would now be classified as having paediatric-type diffuse high-grade gliomas¹². Treatment was well tolerated with no severe adverse events observed. Although the authors reported a ‘partial response’ in four of the eight patients, these results must be interpreted with caution, as no objective radiographic response criteria, such as the Response Assessment in Neuro-Oncology (RANO) criteria, were used for tumour and/or response assessments, and some of the ‘responding’ patients did not have measurable enhancing tumours at the time of CAR T cell infusion. Nonetheless, the investigators were able to demonstrate a reduction in target antigen expression following treatment, along with concomitant increases in both CD8⁺ T cells and macrophages in a tumour specimen obtained from one patient who underwent repeat resection at 6 weeks after treatment. Other CAR T cell targets under investigation in clinical trials involving patients with glioblastoma include B7-H3 (NCT05366179, NCT05474378)³⁹, CD70 (NCT05353530)⁴⁰ and matrix metalloproteinase 2 via a chlorotoxin targeting domain (NCT04214392)⁴¹. Clinical trial data have not yet been published for these targets in adults with CNS tumours, although data on a B7-H3-directed CAR T cell approach presented at the 2025 ASCO Annual Meeting suggest that intracerebroventricular delivery of these cells is feasible and was safe in nine patients with recurrent glioblastoma⁴². A median OS of 14.6 months was reported at this cut-off.

Multi-antigen approaches

In light of the limited efficacy of monovalent targeting approaches in glioma thus far, as well as the established intratumoural heterogeneity of glioblastoma^43,44, researchers have focused on expanding the repertoire of target antigens recognized by CAR T cells. Phase I data are available from 18 patients with recurrent glioblastoma who received a bivalent CAR T cell product simultaneously targeting IL-13Rα2 and EGFR epitope 806 (EGFR806)^15,45. The latter is a cryptic, conformational epitope that is predominantly accessible following dysregulation of EGFR activation owing to EGFR amplification and/or the presence of mutations in the extracellular domain^46,47, and is estimated to be expressed in approximately 50–60% of glioblastomas⁴⁸. Patients underwent surgical debulking and placement of an Ommaya reservoir, followed by a single intracerebroventricular injection of 5 × 10⁶ (n = 6), 1 × 10⁷ (n = 6) or 2.5 × 10⁷ (n = 6) doses of EGFR and IL-13Rα2-targeted CAR T cells. No lymphodepleting chemotherapy was used. This approach consistently resulted in grade 2–3 neurotoxicities with elements of both immune effector-associated neurotoxicity syndrome (ICANS) and tumour-inflammation associated neurotoxicity (TIAN)⁴⁹, although these events were manageable and reversible in all patients. All patients also developed grade 1–2 cytokine-release syndrome (CRS). The authors demonstrated robust CAR T cell expansion in the CSF, with peak levels reaching orders of magnitude higher than those previously observed with peripherally administered EGFRvIII-targeted CAR T cells and that increased correspondingly with each increase in dose level³⁰. CAR T cell expansion in the CSF was accompanied by rapid increases in inflammatory cytokine levels in the CSF, consistent with CAR T cell activation and cytotoxic activity. Eight of 13 patients (62%) with measurable disease at the time of CAR T cell infusion had some degree of tumour shrinkage following treatment, with one patient having a partial response defined according to RANO criteria. Median PFS was 1.9 months and median OS was not reached at the data cut-off (median follow-up duration 8.1 months). Notably, one patient who entered the study with leptomeningeal disease and received a single infusion of 1 × 10⁷ cells had stable disease for >16 months that was ongoing at the time of the report, with CAR T cells remaining detectable in both CSF and peripheral blood samples beyond 1 year after infusion. Altogether, these results provide signals of efficacy, although the limited PFS durations of the majority of patients underscore the need for more durable disease control. This phase I trial is ongoing and is now utilizing repeat dosing with two planned CAR T cell infusions separated by 14 days.

An alternative dual-targeting approach has been described in a brief report of data from the first three patients with recurrent glioblastoma receiving CARV3–TEAM-E T cells, featuring an EGFRvIII-targeting second-generation CAR capable of secreting a T cell-engaging antibody molecule (TEAM) targeting wild-type EGFR¹⁷. This approach enables the recruitment of non-transduced bystander T cells to target wild-type EGFR⁵⁰, which is often more consistently expressed by glioblastoma cells than EGFRvIII. Patients received a single intracerebroventricular infusion of 10 × 10⁶ CAR T cells, which resulted in acute fever in all patients, albeit without grade ≥3 adverse events or DLTs. Similar to the experience with EGFR–IL-13Rα2-targeted CAR T cells, all patients had rapid radiographic tumour regression on MRI, occurring within days of cell infusion. However, responses were transient in two of the three patients, suggesting that resistance develops rapidly in a subset of patients. At the time of study publication, the third patient continued to have an ongoing response of >150 days after cell infusion. Similar to EGFR–IL-13Rα2-targeted CAR T cells, the exact mechanisms of this apparently rapid-onset resistance to CARV3–TEAM-E T cells remains unknown. Potential explanations that require further evaluation include tumour heterogeneity leading to antigen escape, compensatory TME-related factors including upregulation of immune checkpoints and immunosuppressive myeloid cell populations, recruitment of T_reg cells, the development of anti-CAR antibodies and/or limited trafficking into certain regions of tumours despite the intracerebroventricular route of administration. This trial continues to enrol patients and has adjusted the treatment plan to include lymphodepleting chemotherapy (NCT05660369). Data presented at the 2025 ASCO Annual Meeting suggest that the administration of lymphodepleting chemotherapy prior to CARV3–TEAM-E T cells is safe and was well tolerated in seven patients⁵¹.

Lastly, a phase I trial (NCT06186401) is testing synNotch CAR T cells in patients with either newly diagnosed or recurrent glioblastoma. This product utilizes EGFRvIII-targeted CAR T cells engineered to express a synthetic Notch (synNotch) receptor capable of undergoing induced transmembrane cleavage on engagement of the cognate antigen, thereby releasing the intracellular transcriptional domain to enter the T cell nucleus and promote the expression of target genes⁵². This construct is able to induce the local expression of CARs targeting the TAAs EphA2 and IL-13Rα2 in mouse models, with evidence of improved CAR T cell persistence and activity compared with CAR T cells constitutively expressing CARs targeting EGFRvIII or EphA2 (ref. 53). This study includes the use of lymphodepleting chemotherapy followed by a single intravenous infusion of the CAR T cell product. No results had been reported at the time of writing this Review.

CAR T cells in adults with CNS lymphoma

Lymphomas involving the CNS can be either primary or secondary. Primary CNS lymphoma (PCNSL) is a diffuse large B cell lymphoma involving the CNS without systemic disease, whereas secondary CNS lymphoma (SCNSL) refers to lymphoma that has spread to the CNS concurrently with systemic disease or following CNS relapse of an initially non-CNS lymphoma during or after systemic therapy. In both types of CNS lymphoma, clinical data suggesting efficacy of CAR T cell therapy are mainly limited to small retrospective reports describing patients who received CD19-targeted therapies^54–61, reflecting the fact that these agents (specifically tisagenlecleucel, axicabtagene ciloleucel and lisocabtagene maraleucel) are approved for use in patients with treatment-refractory systemic B cell non-Hodgkin lymphoma (NHL)⁶². In terms of prospective evidence, data from a pilot study testing axicabtagene ciloleucel in 18 patients with PCNSL (n = 13), SCNSL (n = 4) or vitroretinal lymphoma (n = 1) were presented at the 2024 ASCO Annual Meeting⁶³. The toxicity profile mirrored that of CD19-targeted CAR T cells in patients with systemic B cell NHL, with no new safety signals including no apparent additional risk of ICANS. The ORR was 94% with a 67% complete response rate and a median duration of response of 13.4 months, indicating encouraging efficacy. Similar results were observed in a phase I/II trial testing tisagenlecleucel in 12 patients with recurrent and/or refractory PCNSL⁶⁴, with a 58% objective response rate and complete responses in 50% of patients. Notably, four of the responses lasted beyond 6 months despite patients having heavily pretreated disease. Details of trials testing CAR T cells are provided in Table 1.

Table 1 |.

Trials testing CAR T cells in adults with CNS malignancies

Trial	Treatment approach	Outcomes	Adverse events	Refs.
EGFRvIII
NCT02209376 (n = 10)	Patients with newly diagnosed or recurrent EGFRvIII-positive glioblastoma received a single i.v. infusion of EGFRvIII-targeted CAR T cells without LD chemotherapy	1 patient had durable SD (>18 months); all patients had evidence of CAR T cell expansion in peripheral blood; 5 of 7 patients had evidence of CAR T cell migration to areas of active disease	Grade ≥3 events in 3 of 10 patients, including seizure, headache and haemorrhage	30
NCT01454596 (n = 18)	Patients with recurrent EGFRvIII-positive glioblastoma received LD chemotherapy followed by a single i.v. infusion of EGFRvIII-targeted CAR T cells at various doses (6.3 × 106 to 2.6 × 1010)	mPFS 1.3 months, mOS 6.9 months; 3 patients survived for >1 year, 1 of whom remained alive at 59 months with no further treatment	All patients had grade ≥3 haematological events; 8 of 18 had grade ≥3 bacteraemia; grade ≥3 hypoxia and hypertension each in 2 patients	31
NCT03726515 (n = 7)	Patients with newly diagnosed EGFRvIII-positive glioblastoma received short-course RT (40 Gy × 15 fractions) followed by repeat 2-weekly i.v. infusions of EGFRvIII-targeted CAR T cells plus pembrolizumab for a maximum of three cycles followed by a fourth dose of pembrolizumab	mPFS 5.2 months, mOS 11.8 months	Grade 3–4 infection, cerebral oedema, encephalopathy, muscle weakness and seizure each observed in 1 patient	33
NCT05660369a (n = 7)	Patients with recurrent EGFRvIII-positive or EGFR-amplified recurrent glioblastoma received LD chemotherapy followed by ICV infusions of up to six doses of TCE-secreting TEAM-E CAR T cells	ORR 0%, SD in 5 patients, 1 patient survived for 12 months after first infusion, another was alive at >12 months	Grade 3–4 febrile neutropenia and decreased neutrophil count	17,51
IL-13Rα2
NCT02208362 (n = 58)	Patients with recurrent IL-13Rα2-positive glioblastoma received IL-13Rα2-targeted CAR T cells across five different treatment arms. Arm 1: intratumoural administration following biopsy (n = 2); arm 2: intratumoural administration following surgical resection (n = 17); arm 3: ICV administration (n = 10); arms 4 (n = 8) and 5 (n = 21): ICT and ICV with TCM-derived and Tmem-derived products, respectively	Arms 1–4: mOS 6.1 months. Arm 5: mOS 10.2 months; CR in 1 patient, PR in 2 patients; a second CR occurred following administration of additional CAR T cell cycles off-protocol	Possibly treatment-related grade ≥3 events in 35% of patients	16,22,23
NCT01082926 (n = 6)	Patients with recurrent IL-13Rα2-positive glioblastoma received local infusions of allogeneic GRm13Z40-2 CAR T cells plus IL-2 over a 2-week period	Transient tumour regression in 4 of 6 patients	No patients had dose-limiting or grade ≥3 treatment-related toxicities; 1 patient had stroke possibly related to the treatment and possibly related to dehydration	128
HER2
NCT01109095 (n = 16)	Patients with recurrent HER2-positive glioblastoma received on or more i.v. infusions of HER2-targeted CAR T cells without prior LD chemotherapy	mPFS 3.5 months, mOS 11.1 months, PR in 1 patient	Grade 3–4 events included lymphopenia, headache (each in 2 patients), neutropenia, fatigue, weakness, cerebral oedema, hydrocephalus and hyponatraemia (each in 1 patient)	34
GD2
NCT03170141 (n = 8)	Adult and paediatric patients with progressive GD2-positive glioblastoma received LD chemotherapy followed by a single i.v. infusion of 4SCAR T cells	mOS 10 months, PR in 4 patients	Grade 3–4 events (≤4 weeks of infusion) included headache (2 patients), lymphopenia and myalgia (each in 1 patient)	38
B7-H3
NCT05474378a (n = 11)	Adults with recurrent B7-H3 overexpressing glioblastoma received escalating intratumoural and intraventricular doses of B7-H3-targeted CAR T cells via Ommaya reservoirs	mOS 14.6 months, 4 patients remained on follow-up ≥22 months after treatment	Grade 3 hypertension in 1 patient	42
EPHA2
NCT03423992 (n = 3)	Patients with EphA2-positive recurrent glioblastoma received LD chemotherapy followed by a single i.v. infusion of EphA2-targeted CAR T cells	Survival durations in the range 86–181 days; SD in 1 patient	Grade 2 CRS in 2 patients	133
IL-13Rα2 and EGFR
NCT05168423a (n = 18)	Patients with EGFR-amplified recurrent glioblastoma received intrathecally administered EGFR–IL-13Rα2 co-targeted CAR T cells without LD chemotherapy	ORR 8% (PR in 1 patient), SD in 8 patients; mPFS 1.9 months, mOS not reached	Grade 3 neurotoxicity in 56% of patients, grade 3–4 lymphocyte count reductions in 33%, grade 3 hypoxia in 12%	15,45
CD19
NCT02445248 (n = 12)	Patients with PCNSL received LD chemotherapy followed by a single i.v. infusion of tisagenlecleucel	ORR 58%, CR in 6 patients, PR in 1 patient; 7 of 12 patients were alive at latest cut-off	Grade 3 ICANS in 1 patient	64
NCT04608487a (n = 18)	Patients with RR PCNSL or SCNSL received LD chemotherapy followed by a single dose of i.v. axicabtagene ciloleucel	ORR 94%, mPFS 14.3 months, mOS 26.4 months	Grade 3 ICANS in 5 of 18 patients, grade 3 seizure in 1 patient	63

^aStudy is ongoing.

CAR, chimeric antigen receptor; CNS, central nervous system; CR, complete response; CRS, cytokine-release syndrome; ICANS, immune effector cell-associated neurotoxicity syndrome; ICT, intracavitary; ICV, intracerebroventricular; i.v., intravenous; LD, lymphodepleting; mOS, median overall survival; mPFS, median progression-free survival; ORR, objective response rate; PCNSL, primary CNS lymphoma; PR, partial response; RR, relapsed and/or refractory; RT, radiotherapy; SCNSL, secondary CNS lymphoma; SD, stable disease; TCE, T cell-engager; T_CM, central memory T cell; T_mem, memory T cell; TEAM, T cell-engaging antibody molecule.

CAR T cells in paediatric CNS tumours

Paediatric CNS tumours are the leading cause of cancer-related death in children⁶⁵. Thus, substantial interest exists in developing effective cellular therapies for these tumours, with the goal of achieving the same level of success observed in paediatric leukaemias⁶⁶. The most common CNS tumours in young patients (0–14 years of age) include gliomas (low-grade and high-grade), medulloblastoma and other embryonal tumours⁶⁷. The landscape of paediatric CNS tumour histologies is relatively diverse compared with that of adult malignancies, albeit with fewer patients overall and, therefore, the majority of trials testing CAR T cells are open to a wider range of tumour types with selection based on expression of the target antigen. To date, all but one (NCT03638167)⁶⁸ CAR T cell trials in this setting remain ongoing, although several have early reports available, including from completed trial arms (Table 2).

Table 2 |.

Trials testing CAR T cells in paediatric patients with CNS malignancies

Trial	Treatment approach	Outcomes	Toxicities	Refs.
GD2
Arm A NCT04196413 (n = 13)	Patients with H3K27M-mutant pontine or spinal DMG who had completed radiotherapy received LD chemotherapy followed by i.v. infusion of GD2-targeted CAR T cells followed by repeat ICV infusions	mOS 20.6 months (17.6 months for patients with DIPG and 31.96 months for those with sDMG), CR in 1 patient ongoing at 30 months, PR in 3 patients	All patients had TIAN requiring intensive monitoring and care; all i.v. infusions resulted in CRS, declining to 33.9% with ICV infusion	37,81
NCT04099797 (n = 12)	Patients with recurrent GD2-expresing non-pontine H3K27-altered DMG or other CNS tumours received LD chemotherapy followed by i.v. infusions of GD2-targeted CAR T cells with (n = 8) or without (n = 3) co-expression of a constitutively active IL-7 receptor	ORR 16.7%, PR in two patients	Grade 1 CRS and TIAN in most patients; grade 4 CRS in 1 patient	82
B7-H3
BrainChild-03 Arm C NCT04185038 (n = 4)	Patients with recurrent DIPG received repeat locoregional infusions of B7-H3 CAR T cells without LD chemotherapy	SD was the best radiographic response, all 3 evaluable patients remained alive at 44, 45 and 52 months following initial infusion	No grade ≥3 events; all evaluable patients had headache, nausea/vomiting, and fever within 24 h of CAR T cell infusion	73,74
HER2
BrainChild-01 NCT03500991 (n = 3)	Patients with recurrent HER2-positive CNS tumours received LD chemotherapy followed by ICV or intratumoural repeat dose-escalated infusions of HER2-targeted CAR T cells	SD in 1 patient	Grade 3 events included headache (in 2 patients), back pain and fever (each in 1 patient)	84
IL-13RA2
NCT04510051 (n = 6)	Patients with recurrent IL-13Rα2-positive CNS cancers received LD chemotherapy followed by repeat ICV infusions of IL-13Rα2-targeted CAR T cells	Clinical and/or radiographic evidence of benefit in 3 of 5 patients	Grade 3–4 events included lymphocyte and white blood cell count reductions (in 3 patients), neutrophil count reductions, catheter-related infections, serum ALT increases and headache (each in 1 patient)	83
EGFR
NCT03638167 (n = 4)	Patients with recurrent EGFR-positive CNS tumours received ICV or intratumoural repeat dose-escalated infusions of EGFR806-targeted CAR T cells	Possible immune-related progressive disease/SD in 1 patient	No grade ≥3 events related to CAR T cells; all patients experienced grade 1–2 headache	68

All studies listed, except NCT03638167, were ongoing at the time of publication. ALT, alanine aminotransferase; CAR, chimeric antigen receptor; CNS, central nervous system; CR, complete response; CRS, cytokine-release syndrome; DIPG, diffuse intrinsic pontine glioma; DMG, diffuse midline glioma; ICV, intracerebroventricular; i.v., intravenous; LD, lymphodepleting; mOS, median overall survival; ORR, objective response rate; PR, partial response; SD, stable disease; sDMG, spinal diffuse midline glioma; TIAN, tumour inflammation-associated neurotoxicity.

B7-H3-targeted CAR T cells

B7-H3 or CD276 is an immunoregulatory protein identified as having high levels of expression in multiple paediatric tumour types, including embryonal tumours and medulloblastomas, high-grade gliomas and DMGs^39,69,70. Following initial demonstrations of safety in several trials testing antibodies targeting B7-H3 (refs. 71,72), CAR T cells targeting this protein were tested in a clinical trial. This phase I trial (BrainChild-03) included three cohorts: arm A for patients with localized recurrent and/or refractory CNS tumours, arm B for those with metastatic tumours, and arm C for those with DIPG (now classified by the WHO as DMG) who could be treated either upfront or at disease progression. Initial reports from arm C described the first three evaluable patients treated at dose level 1 (1 × 10⁷ CAR T cells), including the first patient with DMG ever treated with CAR T cells⁷³. Patients received repeat doses of CAR T cells administered to the lateral ventricle every 14 days, and this approach was well tolerated without DLTs.

The full report from BrainChild-03 arm C was published in January 2025, and describes 23 patients with DIPG who collectively received more than 250 doses of CAR T cells using an intrapatient dose escalation strategy⁷⁴. Dose regimen 4 (10 × 10⁷ CAR T cells) was established as the maximum tolerated dose. The most common adverse events included transient fever, headache, nausea and vomiting, and fatigue for up to 72 h after CAR T cell administration, with one DLT of intratumoural haemorrhage reported at dose level 1. Among 18 evaluable patients, CAR T cells were detectable in CSF samples from 13 patients, although radiographic responses were modest and included a partial response in one patient, stable disease in 14 and progressive disease in three; however, clinical improvement was evident with three long-term survivors. Although this trial was not powered to fully assess efficacy, median OS from the time of diagnosis was 19.8 months in this cohort, compared with a historical median OS among patients with DIPG of 11.2 months⁷⁵. Three patients enrolled in this study remained alive at the time of writing, all of whom received B7-H3-targeted CAR T cells at initial diagnosis, and two were continuing to receive therapy after >37 months and >31 months.

All patients in arm C of BrainChild-03 had radiographically defined DIPG, with 18 of 23 having a histopathologically or molecularly confirmed diagnosis of DMG. The longest ongoing survivor, >50 months from the time of diagnosis, elected to discontinue CAR T cell therapy after 1 year and has a histone wild-type high-grade glioma of the pons with TP53 and IDH mutations. This patient had a prolonged period of stable disease prior to eventual disease progression at around 44 months, which was probably also influenced by the molecular alterations associated with more benign disease biology. Of the two additional surviving patients who remained on therapy, one harboured the classic H3FRA^K27M mutation and the other had a non-diagnostic biopsy sample. This study also established the feasibility and safety of multiple intracerebroventricular infusions. Furthermore, the investigators created a protocol for CAR T cell preparation using a thaw-and-dilute strategy for DMSO removal, rather than thaw-and-wash, which resulted in improved T cell viability⁷⁶. This thaw-and-dilute protocol has now become the standard procedure across multiple trials testing locoregionally administered CAR T cells. One additional trial testing CAR T cells targeting B7-H3 is ongoing, with two arms including one for patients with brainstem DMGs or gliomas and the other for those with non-brainstem relapsed CNS tumours (NCT05835687).

GD2-targeted CAR T cells

The disialoganglioside GD2 is an established drug target originally identified in melanoma and other neuroectodermal tumours that rose to prominence in paediatric oncology following the FDA approval of dinutuximab, an anti-GD2 antibody, for use in paediatric patients with high-risk neuroblastoma in 2015 (ref. 77). CAR T cells or other cell therapies targeting GD2 have been studied for more than two decades as treatments for solid tumours, but have only begun to show promising signs of clinical activity over the past few years^8,78–80. A breakthrough occurred when GD2 was found to be overexpressed on the surface of DMG cells, and could be effectively targeted using GD2-directed CAR T cells in preclinical models³⁶. These studies led to a phase I trial testing this approach in patients with pontine or spinal DMGs; those with thalamic tumours were excluded following CAR T cell-related deaths in thalamic xenograft mouse models. The treatment plan initially stipulated intravenous CAR T cell administration, although this was amended shortly after study opening to permit intracerebroventricular delivery of subsequent doses if patients derived clinical benefit, with repeat doses delivered 1–3 months following the initial intravenous dose. Early reports indicated clinical or radiographic improvements in three of the first four patients treated, with impressive albeit not durable radiographic improvements in two of these patients³⁷. Evaluations of single cells obtained from the CSF of patients during treatment have revealed marked differences in immune cell composition and phenotype between samples obtained after the intravenous dose and samples obtained after the intracerebroventricular doses. Myeloid populations in the CSF had an immune-activating phenotype during clinical responses following intracerebroventricular administration only; intravenous dosing resulted in the accumulation of immune-suppressive myeloid cells even in the setting of a clinical response³⁷. Increased T_reg cell levels were also observed in the CSF following intravenous delivery³⁷, highlighting several potential benefits of intracerebroventricular dosing, and prompting the study chairs to open additional trial arms that involved intracerebroventricular administration only.

In this initial report, the authors also characterized a specific toxicity that seems to be unique to CNS-directed CAR T cells, known as TIAN^37,49. Owing to localized inflammation at the tumour site following CAR T cell administration, both oedema (type 1) and electrophysiological disruption (type 2) resulted in neurotoxicities with a similar effect observed in preclinical models. Given the risks of herniation and death, patients enrolled in the trial testing GD2-targeted CAR T cells at Stanford had a CSF reservoir placed before starting therapy to enable expeditious removal of excess fluid, and required intensive supportive care around the time of CAR T cell administration^37,81. TIAN has mainly been observed and described in patients with midline tumours; however, prospective studies aim to incorporate TIAN grading⁴⁹ to gain a better understanding of the full spectrum that might occur in CNS-directed CAR T cell therapy.

The full report from the intravenous followed by intracerebroventricular dosing arm of the trial testing GD2-directed CAR T cells has now been published, with 11 patients receiving treatment⁸¹. Among nine patients with pontine DMG and two with spinal DMG, dose-limiting CRS occurred at intravenous dose level 2 (3 × 10⁶ CAR T cells per kilogram), indicating dose level 1 (1 × 10⁶ CAR T cells per kilogram) as the maximum tolerated dose. No DLTs occurred following intracerebroventricular dosing, with doses safely escalated up to 30 × 10⁶ CAR T cells. TIAN was observed in 100% of patients following intracerebroventricular administration, with a decreasing intensity of events on subsequent infusions. Radiographic response rates and clinical improvements were notable, with 4 of 11 patients having a best radiographic response of 50–100% volumetric reductions in tumour size. Clinical improvements in neurological symptoms were also noted, which exceeded radiographic responses in some patients. Most responses were unfortunately not durable; however, the outcomes in two outstanding responders are described in detail in the report, including a 17-year-old male with pontine DMG who had a complete response, and at the time of writing had not received repeat radiotherapy, and continued to receive CAR T cells via intracerebroventricular infusion 30 months from the first infusion. Median OS among patients with pontine DMG in this trial was 17.6 months, and in those with spinal DMG was 32.0 months; the authors note that their study population was highly selective and excluded patients with high-risk disease, and therefore, OS data must be interpreted with caution. Nonetheless, these data have led to substantial enthusiasm for the idea that cellular therapy might indeed hold the key to improving the outcomes in patients with DMG.

In parallel, a phase I trial tested GD2-targeted CAR T cells equipped with a constitutively active IL-7 receptor (IL-7R) in patients with DMG and other GD2-expressing CNS tumours⁸². CAR T cells were administered intravenously at doses escalating from 1 × 10⁷ cells/m² to 3 × 10⁷ cells/m², and repeated every 12–14 weeks in patients with stable disease or following radiographic evidence of clinical benefit. The cohort included four patients with thalamic DMG, three with pontine DMG, one with spinal DMG, two with medulloblastoma and one with atypical teratoid/rhabdoid tumour, and the first three patients treated received non-IL-7R-modified GD2-directed CAR T cells and subsequent patients received IL-7R-modified cells. Overall, this approach was well tolerated, with one patient having grade 4 CRS, with otherwise grade 0–1 CRS and TIAN observed at all CAR T cell doses when using a fractionated dosing regimen. Transient clinical improvements in neurological symptoms lasting up to 3 weeks, without radiographic evidence of a response, were observed in the three patients who received non-IL-7R-modified CAR T cells. In patients receiving IL-7-modified products, two patients had a partial response, five had stable disease and one had progressive disease, with clinical improvements in neurological deficits lasting up to 27 months. Of note, both patients with a partial response had DMG, one with thalamic and one with pontine disease. In addition, the one patient with progressive disease after receiving IL-7R-modified GD2-targeted CAR T cells was subsequently found to have GD2-negative DMG following cell infusion, highlighting the potential for acquired resistance with this therapy. Follow-up from cohorts receiving IL-7R-modified GD2 CAR T cells via intracerebroventricular administration is ongoing (NCT04099797).

GD2-directed CAR T cells are also being tested in a separate phase I trial sequentially enrolling patients with relapsed embryonal tumours including medulloblastoma and high-grade glioma including DMG, who will receive a single intravenous infusion of GD2-targeted CAR T cells equipped with an inducible caspase 9 suicide switch (NCT05298995). Elsewhere, another phase I trial is testing GD2-directed CAR T cells initially administered intravenously followed by intracerebroventricular doses (NCT05544526), similar to the original trial in this area⁷⁹.

Other targets and multi-antigen approaches

Ongoing studies are evaluating alternative CAR T cell targets across patients with paediatric CNS tumours, with results still pending. Similar to adults with glioblastoma, IL-13Rα2 is also expressed in patients with paediatric high-grade gliomas including ependymomas and subsets with other disease histologies. Data from a phase I trial testing IL-13Rα2-directed CAR T cells in six patients with paediatric IL-13Rα2-positive CNS tumours are available as a preprint⁸³. Patients received repeat intracerebroventricular doses after lymphodepletion, and this approach was well tolerated without CAR-specific DLTs. Three of five evaluable patients had transient reductions in tumour volume of 10–40%. Analysis of CSF samples obtained during this study also revealed clonal expansion of CAR-negative CD8⁺ T cells, suggesting a localized immune response.

HER2 expression can be detected across a spectrum of paediatric CNS tumour types including ependymoma, high-grade glioma and medulloblastoma. CAR T cells targeting HER2 in any HER2-positive relapsed and/or refractory CNS tumours are being tested in a single-centre phase I trial (NCT03500991) and in a multi-centre trial enrolling only patients with HER2-positive ependymoma (NCT04903080). An initial report from the former study indicates that three patients were able to receive repeat intracerebroventricular CAR T cell infusions without DLTs, with correlative data suggesting immune activation in the CSF⁸⁴. However, only one patient had radiographically stable disease, whereas the other two patients had progressive disease. EGFR806-targeted CAR T cells were also recently tested in paediatric patients with EGFR-expressing CNS tumours, although the trial was terminated early owing to poor accrual. Results from four patients showed tolerability and best response of stable disease (NCT03638167)⁶⁸.

Given that paediatric CNS tumours often express several potentially targetable antigens, CAR T cells directed at B7-H3, EGFR806, HER2 and IL-13Rα2 are being tested in a phase I trial that is currently recruiting patients (NCT05768880). The trial includes two arms, one for those with pontine DMG/DIPG, and the other for those with non-pontine DMG and other relapsed and/or refractory tumours; no specific antigen expression is required for enrolment. CAR T cells are created using a mix of viral vectors, so that each individual T cell might express from one to four CARs on the cell surface. With tumour heterogeneity and antigen-negative resistance observed in many patients receiving CAR T cells, this study is expected to provide an important evaluation of methods of mitigating antigen-negative resistance.

Upcoming or anticipated trials

Many trials testing CAR T cells in patients with paediatric CNS tumours are either recruiting patients or ongoing with results awaited, inspired by some of the early work that has provided glimmers of efficacy in patients with otherwise refractory tumours. Examples not previously mentioned include a trial testing B7-H3-targeted CAR T cells with a novel scFv⁸⁵. ‘Armoured’ GD2-targeted CAR T cells further engineered to secrete IL-18 are also being tested in patients with non-CNS solid tumours (EU CT 2022-501725-21-00), with plans to test this approach in patients with CNS tumours once safety has been established. Investigations of novel targets are also in various stages of development, including CAR T cells targeting CD99 (ref. 86) or GPC2 (ref. 87). Enthusiasm also exists regarding the use of gating technologies, such as synNotch systems, which enable conditional CAR expression in the presence of certain stimuli, to enhance safety^53,86. Additional armouring strategies as well as multi-antigen targeted CARs are also expected, as well as the use of small molecules to enhance antigen presentation, such as tazemetostat, which drives GD2 expression⁸⁸.

Future approaches for CAR T cell therapy in CNS tumours

Combination therapies

Despite the clinical and radiographic responses seen with monovalent and bivalent CAR T cell approaches in patients with CNS tumours, combination therapies are more likely to yield durable responses. For example, regimens integrating CAR T cells with immune checkpoint inhibitors, such as anti-PD-1/PD-L1 or anti-CTLA4 antibodies, have the potential to overcome the immunosuppressive TME observed in many CNS malignancies^33,89. Preclinical and clinical data suggest that approaches combining CAR T cells with agents such as pembrolizumab might sustain T cell activity, enhance infiltration and prolong persistence^90–92. However, we have previously shown that combining CAR T cells with pembrolizumab in patients with newly diagnosed glioblastoma results in reduced systemic CAR T cell engraftment and limited efficacy, potentially by limiting the size of a subpopulation of PD-1⁺ CAR T cells with an important role in the anti-glioblastoma response^33,93. The experience with combinations with radiotherapy or oncolytic viruses suggests the potential to sensitize tumours to CAR T cells by increasing antigen expression and disrupting the blood–brain barrier (BBB), thus facilitating CAR T cell entry into the CNS^94–96. Chemotherapeutic agents that modulate the immune microenvironment, such as temozolomide⁹⁷, could further potentiate these effects, making combination strategies a key focus for future studies. Lastly, targeted inhibitors of oncogenic signalling pathways (such as kinase inhibitors) might potentiate the antitumour activity of CAR T cells^98,99.

CAR modifications and cytokine armouring

Armoured CAR T cells, which are engineered to also express cytokines such as IL-12 or IL-15, have demonstrated improved activity within the immunosuppressive TME in various other solid tumours. For example, the capacity for localized IL-15 production increases the expansion and promotes the antitumour activity of GPC3-targeted CAR T cells in patients with advanced-stage liver tumours¹⁰⁰. CAR T cells engineered for local IL-12 production have also demonstrated improved antitumour activity in preclinical models^101–103. Other next-generation approaches that are being developed for the treatment of patients with glioblastoma include modulation of TGFβ signalling^104–106. Applying this approach to the glioblastoma TME would probably have pleiotropic effects on both immunosuppressive and pro-tumorigenic myeloid cell populations. Several strategies involving modulation of TGFβ signalling are at various phases of preclinical and clinical development. Our group has utilized a dominant-negative TGFβ receptor II (TGFβDN) that has previously been tested clinically in combination with a PSMA-targeted CAR T cell product in mouse models of prostate cancer¹⁰⁷ and demonstrated that this approach promotes the cytolytic activity of EGFR–IL-13Rα2-targeted CAR T cells as well as central memory phenotypes and long-term proliferative capacity in preclinical models of glioblastoma. We observed that TGFβDN-expressing trimodular CAR T cells have enhanced proliferative capacity at 2–4 weeks following administration in mouse models and are more cytolytic than their TGFβ receptor II-unmodified counterparts in repeated antigen challenge assays, while also adopting a more pronounced central memory profile¹⁰⁶. Additional strategies include modifying CAR T cells for TGFβ-dependent cytotoxic activity and IL-13Rα2–TGFβ co-targeted CAR T cells^104,105. Moreover, incorporating safety switches and/or suicide genes into CAR constructs might enable more controlled CAR T cell activation, and thus mitigate CNS toxicities^108–110. These advances aim to create robust CAR T cells capable of sustained activity and precision targeting of CNS tumours.

Addressing antigen heterogeneity

Heterogeneous antigen expression in CNS tumours remains a major challenge, most notably in patients with glioblastoma. Both de novo and recurrent glioblastomas harbour a complex landscape of EGFR alterations¹¹¹, in addition to alterations in other receptor tyrosine kinases and their related signalling pathways^112–117. Incorporating dual or tandem CAR designs capable of targeting multiple antigens simultaneously might mitigate antigen loss, a common mechanism of treatment resistance. Early data demonstrate preservation of amplified wild-type EGFR, even in tumours with selectively edited EGFRvIII loss, which necessitated the use of a more cross-reactive EGFR CAR with activity against several EGFR isoforms^30,118. Various other multi-target CAR T cell strategies enabling simultaneous targeting of multiple tumour-specific antigens or TAAs using either tandem or parallel CAR T cell designs are either being tested clinically or are in development¹⁵. For example, data from both preclinical models and from patients with recurrent glioblastoma suggest activity of CAR T cell–bispecific T cell engager combinations^{17,50,119,120}. Other research has aimed to integrate a broader range of TAAs, informed by single-cell RNA sequencing. Pooled mRNA CAR T cells are under evaluation preclinically as a method for countering target heterogeneity in patients with recurrent glioblastoma¹²¹. CAR T cells with synNotch receptors enable antigen-responsive CAR expression following detection of an initial tumour-specific priming antigen, thus enabling adaptation of the level of T cell activity to the heterogeneous and evolving microenvironments of both adult and paediatric glioblastomas^53,86. CAR T cells featuring such synNotch systems are being tested in a phase I trial involving patients with EGFRvIII-positive glioblastoma who are receiving E-SYNC CAR T cells (NCT06186401). In this design, engagement of EGFRvIII-positive glioblastoma cells is required for dual-targeting EphA2–IL-13Rα2 CAR expression. Such approaches might increase the likelihood of tumour-specific eradication and/or reduce the risk of off-tumour toxicities.

CNS-specific targeting strategies

Locoregional CAR T cell administration strategies have demonstrated considerable promise in several trials. In particular, CSF engraftment of CAR T cells following intracerebroventricular administration has been found to be markedly superior to peripheral intravenous administration¹⁵ (Table 3). Combining intracerebroventricular with intratumoural or systemic delivery methods might further enhance antitumour efficacy¹⁶. Convection-enhanced delivery is another approach that ensures the precise localization of drugs within CNS tumours while bypassing the BBB. However, this technique has yet to be applied to the administration of CAR T cells to patients with CNS malignancies. Advances in CAR T cell engineering might further their potential for CNS targeting. For example, researchers have identified the brain-specific extracellular matrix protein brevican (BCAN), which is uniquely expressed in the CNS, as a particulated receptor that could enable more effective targeting of CAR T cells to brain tumours via synNotch or potentially other targeting mechanisms¹²². In this example, incorporating a synthetic receptor that recognizes brevican endowed CAR T cells with enhanced specificity and retention within the CNS microenvironment. This innovative strategy enables CAR T cells to be tailored to the unique challenges created by CNS tumours and provides a promising alternative to conventional approaches. The synNotch receptor (anti-brevican scFv) expressed on the surface of engineered T cells binds brevican in the extracellular matrix, leading to local activation of the dual-targeting EphA2–IL-13Rα2 CAR, thus allowing only local CAR T cell activation in the regions immediately surrounding the glioblastoma cells. This system allows for the activation of synNotch by antigens not involved in oncogenesis, or potentially even those not expressed by glioblastoma cells, and might enable a more diverse range of antigens to be targeted, given the restricted local release. Other studies have explored the potential of engineering CAR T cells to co-express chemokine receptors, such as CXCR4 and CCR2, to improve their trafficking to specific tumour sites, but might also allow homing to areas of non-tumour inflammation^123,124. While these strategies have shown promise in non-CNS settings (such as preclinical models of lung, pancreatic and haematological cancers), they have thus far not been investigated as methods for improving CAR T cell CNS tropism. Given the unique biology of T cell entry into the brain, additional efforts are likely to be required to optimize CAR T cell accumulation in CNS cancers^125–127. Taken together, these developments highlight ongoing research efforts to optimize CAR T cell localization and activity within the brain.

Table 3 |.

Routes of CAR T cell administration in patients with CNS malignancies

Route of administration	Advantages	Disadvantages	Refs.
Intravenous	Logistically simple Can be administered in an outpatient setting Permits serial monitoring of peripheral immune interactions	CAR T cells must cross the BBB Lack of interactions with local immune environment Higher risk of systemic toxicities than with local delivery	30,31,33,34,37,38,63,64,81,82,133
Intracerebroventricular	Locoregional administration Enables acquisition of repeat CSF samples	Requires surgical placement of an Ommaya reservoir Risk of infection	15–17,23,37,42,45,51,73,74,81,83,84
Intracavitary/intratumoural	Locoregional administration	Requires surgical placement of a catheter Parenchymal injection might limit uptake	16,22,23,42,73,84,128

BBB, blood–brain barrier; CAR, chimeric antigen receptor; CSF, cerebrospinal fluid.

Allogeneic CAR T cell approaches

Allogeneic, or ‘off-the-shelf’, CAR T cells provide a scalable solution to many of the limitations of autologous therapies, including manufacturing costs, turnaround times and quality of the source material. Gene-editing technologies such as CRISPR–Cas9 and TALENs enable the creation of universal CAR T cells by eliminating endogenous T cell receptors and MHCs to prevent graft-versus-host disease and graft rejection, respectively. One example that has already been tested in clinical trials is GRm13Z40-2, an allogeneic IL-13Rα2-targeted CAR T cell product that has been genetically modified for resistance to immunosuppression by glucocorticoids¹²⁸. These CAR T cells were generated using biallelic inactivation of gene encoding the glucocorticoid receptor (NR3C1) via zinc finger nuclease targeting in IL-13-zetakine⁺ CD8⁺ cytolytic T lymphocytes. This product was administered intracranially in combination with recombinant IL-2 in six patients with recurrent glioblastoma who were also receiving systemic dexamethasone, and the investigators demonstrated that this approach is well tolerated with transient reductions in tumour volume and/or evidence of tumour necrosis in four patients. Universal CAR T cells such as this can be manufactured in advance, enabling rapid clinical deployment. The standardized manufacturing processes for allogeneic CAR T cells also promise to reduce costs and improve patient access, paving the way for broader testing of this novel therapy in clinical trials. Immune cell types other than conventional αβ T cells might be more amenable to allogeneic off-the-shelf applications. Both natural killer (NK) cells and γδ T cells are being evaluated preclinically and in early phase trials involving patients with glioblastoma^129–131. All of these immune cell types or subtypes, including αβ T cells, NK and γδ T cells, can be obtained from donors without cancer, although the additional innate, MHC-independent recognition properties of the latter two might provide certain advantages in patients with solid tumours. These properties theoretically make cell therapy a more practical therapeutic option at an increased number of time points during the disease course of patients with glioblastoma. Evidence also exists that donor-derived conventional CAR T cell approaches are feasible and efficacious in patients with neuroblastoma¹³².

Conclusions

Advances in CAR T cell engineering, combination treatment strategies and CNS-specific targeting approaches have the potential to transform the landscape of cellular immunotherapy for CNS tumours. By addressing antigen heterogeneity, enhancing T cell fitness and leveraging allogeneic platforms, researchers are overcoming many ofthe existing barriers and beginning to evaluate the clinical potential of CAR T cells in this challenging domain. Continued interdisciplinary efforts will be crucial to translating these innovations into effective clinical applications in order to improve the outcomes in patients with CNS malignancies.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1038/s41571-025-01102-1.

Key points.

• Advances in chimeric antigen receptor (CAR) T cell engineering, combination treatment strategies and central nervous system (CNS)-specific targeting approaches have the potential to transform the landscape of cellular immunotherapy for CNS tumours.

• CNS tumours harbour several novel tumour-specific, and tumour-associated targets for CAR T cell approaches.

• The immunosuppressive effects of CNS solid tumours often limit the efficacy of CAR T cells.

• Continued interdisciplinary efforts will be crucial to translating preclinical innovations into effective clinical applications to improve the outcomes of patients with CNS malignancies.