Intravenous and intracranial GD2-CAR T cells for H3K27M+ diffuse midline gliomas

1Chimeric antigen receptors (CARs) couple an antigen-binding domain to T cell signalling domains to redirect T lymphocytes to cancer cells expressing a target of interest. Autologous CAR T cells have mediated impressive results in refractory B and plasma cell malignancies–, but have not demonstrated high rates of sustained antitumour effects in solid cancers or brain tumours–, with the exception of promising responses in a recent trial of GD2-CAR T cell therapy for neuroblastoma. Differential rates of activity between liquid cancers and solid/brain tumours may relate to a dearth of safe targets with high homogenous expression, inadequate T cell trafficking and/or T cell dysfunction induced by the tumour microenvironment.

2H3K27M-mutated diffuse midline gliomas (DMGs) chiefly occur in children and young adults and originate in midline structures of the nervous system. Patients with pontine DMG (also called diffuse intrinsic pontine glioma, DIPG) have a median overall survival of 11 months from diagnosis, and 5 year overall survival below 1% (refs. –). Patients with DMGs outside of the brainstem, including the spinal cord, have a median overall survival of 13 months. DIPG is the most common cause of death due to brain cancer in children, with palliative radiotherapy the standard of care. Cytotoxic chemotherapy has not improved outcomes to date. Although targeted therapies and immuno-oncology strategies have begun to show early promise–, outcomes remain dismal.

3We discovered high, uniform expression of GD2, a disialoganglioside, on H3K27M+ DMG cells and demonstrated that intravenous (IV) administration of chimeric antigen receptor-modified T cells targeting GD2 (GD2-CART) eradicated established DMGs in patient-derived orthotopic xenograft mouse models. We and others also demonstrated increased potency and decreased systemic inflammation following intracerebroventricular (ICV) administration of CAR T cells compared with IV administration in preclinical brain tumour models–. These data provided rationale for this first-in-human/first-in-child phase 1 clinical trial (NCT04196413, a phase 1 clinical trial of autologous GD2-CART for DIPG and spinal diffuse midline glioma (sDMG)). We previously reported antitumour activity and correlative findings from the first three patients treated on NCT04196413 arm A at dose level 1 (DL1), and a fourth who enroled on trial but received therapy through a single-patient-compassionate investigational new drug (IND) application. Here we report final clinical results following enrolment completion of arm A, which demonstrate the tolerability of 1 × 106 GD2-CART cells per kg administered intravenously (DL1) followed by sequential ICV infusions, tumour regressions and, in some cases, sustained antitumour effects in patients with DIPG and sDMG.

4Eligible patients were 2–30 years of age, with biopsy-confirmed H3K27M-mutated DIPG or sDMG, who had completed standard frontline radiotherapy at least 4 weeks before enrolment, were not receiving corticosteroid therapy and had acceptable performance status (Lansky or Karnofsky score 60 or above). Although patients typically exhibited bulky disease in the brainstem or spinal cord, those with bulky disease involving the thalamus or cerebellum were not eligible due to increased risk of toxicity of GD2-CART observed with these tumour locations in murine models. Patients with clinically significant dysphagia (as an indicator of significant medullary dysfunction) were also ineligible. Detailed eligibility and exclusion criteria are given in Supplementary Methods 1. The clinical trial protocol was approved by the Stanford Institutional Review Board and registered with ClinicalTrials.gov (NCT04196413). Informed patient or parent consent and child assent were obtained.

5Primary objectives were the determination of the feasibility of manufacturing, assessment of safety and tolerability and identification of a maximally tolerated and/or recommended Phase 2 dose of IV GD2-CART following lymphodepleting chemotherapy in this population. Secondary objectives (detailed in Supplementary Methods 2, section 1) included preliminary assessment of benefit as measured by radiographic response on magnetic resonance imaging (MRI) and by clinical improvement in neurological dysfunction. IV doses were escalated using a 3 + 3 design, based on the occurrence of dose-limiting toxicities (DLTs) possibly, probably or most likely attributed to IV GD2-CAR T and occurring within 28 days following infusion. DLTs comprised any grade 5 toxicity, grade 4 cytokine release syndrome (CRS), grade 4 neurotoxicity lasting at least 96 h, new grade 3 neurotoxicity lasting at least 28 days, grade 4 neutropenia or thrombocytopenia lasting more than 28 days and grade 3 or greater non-haematologic toxicity, with the exceptions of those detailed in Supplementary Methods 2, section 5.4.5. Because sDMGs are rare, the protocol allowed safety in the DIPG cohort to inform dose escalation for patients with sDMG but, given the risk for neurotoxicity related to the location of DIPG, safety in sDMG patients did not inform dose escalation for patients with DIPG.

6In December 2020, a protocol amendment added a new secondary objective to evaluate safety and assess clinical benefit to patients treated with IV GD2-CAR T cells followed by ICV administration of GD2-CAR T cells (Supplementary Methods 2). Patients were eligible to receive second or subsequent IV or ICV infusions if they showed complete response, partial response, minor response or stable disease radiographically on MRI, or had evidence for clinical benefit from preinfusion baseline according to the protocol-specified clinical neurological examination, and if at least 28 days had elapsed since the initial GD2-CART infusion or 21 days from subsequent infusions, circulating CAR T cell levels were below 5% and toxicity had resolved to below grade 2. Lymphodepletion was not administered before ICV infusions. All subsequent infusions were delivered intracerebroventricularly.

7Following enrolment on the trial, no additional, non-protocol-directed chemotherapies, molecularly targeted therapies or immunotherapies were allowed. Local therapies such as intratumoural cyst drainage or re-irradiation were allowed; if re-irradiation was administered, subsequent infusions resumed after 4 weeks from the end of re-irradiation.

8GD2-CART were manufactured in the automated CliniMACS Prodigy with IL-7 (Miltenyi Biotec, 12.5 ng ml−1) and IL-15 (Miltenyi Biotec, 12.5 ng ml−1) in the presence of dasatinib (Extended Data Fig. 1a and Supplementary Methods 3), and cryopreserved at day 7 of culture. The GD2-CAR was encoded by a retroviral vector encoding an iCasp9 domain (Bellicum Pharmaceuticals, Inc.), and GD2.4-1BB.CD3z CAR separated by a P2A ribosomal skip sequence (Extended Data Fig. 1b).

9Patients received standard supportive care for seizure prophylaxis, immunosuppression associated with lymphodepleting chemotherapy, CRS and immune effector cell acute neurotoxicity syndrome (ICANS), as detailed in Supplementary Methods 4. Neurotoxicity that was distinct from ICANS and attributable to local tumour inflammation was designated tumour inflammation-associated neurotoxicity (TIAN), and graded using NCI Common Terminology Criteria for Adverse Events v.5.0. For mitigation of risks associated with TIAN, we implemented a toxicity-monitoring and toxicity-management algorithm that (1) placed Ommaya catheters or similar devices for intracranial pressure monitoring and potential treatment of hydrocephalus in all patients; (2) incorporated sequential neurological examinations and scheduled and symptom-prompted intracerebral pressure monitoring; (3) prescribed measures to lower elevated intracranial pressure (positioning, hypertonic saline (3%), cerebrospinal fluid (CSF) removal) in patients with documented elevated intracranial pressure; and (4) administered anakinra and corticosteroids to patients with significant neurological symptoms (detailed in Supplementary Methods 4).

10Enrolment began in June 2020 and the data cut-off was 1 December 2023. A consort diagram is shown in Extended Data Fig. 2. Characteristics of the 13 enroled patients are shown in Table 1. Median age was 15 years (range 4–30 years), and seven patients were female; ten patients had DIPG and three had sDMG. The H3K27M mutation was demonstrated in biopsies from all tumours by either immunohistochemistry or DNA sequencing. Median time from diagnosis to enrolment was 5.0 months (range 3.9–11.6 months). Eight patients had disease progression or pseudoprogression on MRI imaging at enrolment, and five did not have documented progression at enrolment. Two patients (nos. 002 and 011) were removed from the study before treatment due to rapid tumour progression and a decline in performance status rendering them ineligible for protocol-directed therapy. Patient no. 002 was treated on a single-patient-compassionate IND, with outcomes for this patient reported previously.

11For the 13 enroled patients, 20 GD2-CART products were successfully manufactured (Extended Data Fig. 1), including seven CAR T cell products remanufactured to support multiple repeated ICV infusions. The product for patient no. 011 was not characterized, because this patient developed rapid tumour progression and died before CAR T cell infusion. Mean manufacturing duration was 7 days. The median time from enrolment to IV GD2-CART infusion was 22.9 days (range 15–38 days). Product functional assessment demonstrated GD2-specific reactivity (Extended Data Fig. 3a,b). Products demonstrated a mean fold T cell expansion of 21.97 (range 12.16–28.06), mean percentage viability of 93.51 (range 86.8–96.9), mean transduction efficiency of 57.27 (range 21.63–83.80) (Supplementary Table 1) and mean vector copy number per CAR+ cell of 3.17 (range 0.84–7.20) (Extended Data Fig. 3c,d). Detailed GD2-CART phenotypic composition of 19 products showed a predominance of central memory T cells (Extended Data Fig. 1c–f).

12Of 11 patients who received one IV GD2-CART infusion on study, 9 experienced clinical benefit, imaging benefit on MRI or both and received additional ICV infusions, as noted in Table 1 (median of four infusions, range 1–17), administered over 1.0–29.4 months at the time of data cut-off. Incidence and grade of CRS, ICANS and TIAN following each IV and ICV infusion are shown in Fig. 1a, Table 2 and Supplementary Tables 2 and 3. Following IV GD2-CART, all patients experienced CRS, with one of three patients on DL1 experiencing grade 2 CRS, six of eight patients on DL2 experiencing grade 2 or higher CRS and three patients on DL2 experiencing DLT attributed to grade 4 CRS (Table 2). CRS was similar in nature to that observed in other CAR T cell trials. The three instances of grade 4 CRS were characterized by hypotension requiring multiple pressors (one patient), pulmonary oedema and respiratory failure requiring either bilevel positive airway pressure/continuous positive airway pressure (one patient) or intubation (one patient). CRS was managed according to established guidelines using tocilizumab, anakinra, corticosteroids (dexamethasone and methylprednisolone), fluid resuscitation and supportive care. Following IV GD2-CART, we observed ICANS in one of three patients at DL1 (grade 2) and in four of eight patients at DL2 (n = 1 grade 3, n = 3 grade 1). Based on these results, we identified DL2 (3 × 106 GD2-CART per kg) as exceeding the maximally tolerated dose for IV administration in patients with DIPG/DMG.

13Among 62 ICV infusions, no DLTs occurred. Forty-one ICV infusions (66%) were associated with no CRS. Among 19 ICV infusions associated with CRS, most were low grade (n = 16 grade 1; n = 4 infusions in two patients grade 2; n = 1 grade 3 in the context of urosepsis in a patient with sDMG) (Table 2). No ICANS was observed following ICV infusions. One patient (no. 005) experienced a grade 3 sensory neuropathy (Supplementary Table 4) in a stocking-glove distribution following the first infusion, for which vitamin B6 toxicity was considered a contributory factor; this resolved following vitamin B6 supplement cessation. Of note, we did not observe any cases of transient painful peripheral neuropathy commonly associated with anti-GD2 antibody therapy. As previously reported, one patient (no. 003) experienced a grade 5 intratumoural haemorrhage (Supplementary Table 4) in a region of known intratumoural vascular anomaly. Intratumoural haemorrhages occur relatively frequently in DIPG; risk increases with time from diagnosis, and symptomatic intratumoural haemorrhages occur in around 20% of patients with DIPG by 12 months from diagnosis.

14We observed TIAN in 91% of patients following IV infusion and in 100% of patients following the first ICV infusion; TIAN grade typically diminished with subsequent infusions (Fig. 1a). Forty-four ICV infusions (71%) were associated with TIAN. Nine of nine (100%) patients with DIPG developed TIAN following IV GD2-CART (n = 2 grade 4, n = 3 grade 3 and n = 4 grade 2), and 67% of patients with DIPG developed TIAN following ICV GD2-CAR T cell infusions (n = 29 of 43 infusions with TIAN: grade 4, n = 1; grade 3, n = 3; grade 2, n = 9; grade 1, n = 16). TIAN reversed in all patients following treatment according to our toxicity management algorithm, and no patient experienced a DLT due to TIAN. In one patient with sDMG, high-grade communicating hydrocephalus was observed during peak tumour inflammation. The agent available for ablation of CAR T cells by the induction of caspase-9, AP1903, was not administered to any patients treated on arm A.

15Experience with the first three patients enroled suggested a hypothesized risk of progression approximately 2–3 months following IV infusion, including patient no. 003, who showed an impressive response to IV infusion, followed by progression at approximately day 70, with response to subsequent ICV infusion. Based on this experience, beginning with patient no. 004, the protocol was amended to allow patients with clinical or imaging benefit following IV infusion to receive multiple sequential ICV infusions. Patients were eligible to receive ICV infusions every 1–3 months, providing they had experienced stable disease or clinical or imaging benefit with the previous infusion.

16Several patients exhibited reduction in tumour size following GD2-CART, with the best volumetric change in tumour size for each patient shown in Fig. 1b. In this patient cohort, this ‘best response’ change in tumour volume data fits a Gaussian distribution by the Shapiro–Wilk test of normality, with no outliers (Fig. 1b and Extended Data Fig. 4a). Overall survival from time of diagnosis to trial enrolment, and from first GD2-CART administration to time off trial, death or data cut-off, is shown in Fig. 1c. Patient no. 010 (DIPG), who was enroled with evidence of progression or possible pseudoprogression by imaging following completion of upfront radiotherapy, demonstrated a continuing reduction in tumour volume to the point of complete response within months following the first GD2-CAR T cell infusion. This complete response was sustained at 30 months following enrolment, the time of data cut-off, and is ongoing (Figs. 1b and 2a,b). Patient no. 006 (sDMG), who was in clear clinical and radiographic tumour progression at the time of first treatment, demonstrated 91% reduction in tumour volume by 7 months following the first GD2-CAR T cell infusion (Figs. 1b and 3a,b). Changes in tumour volume over time for additional patients are shown in Extended Data Fig. 4b.

17Clinical improvements, as measured by changes in clinical improvement score (CIS), were also observed (Supplementary Table 5 and Extended Data Fig. 9). In general, clinical improvement coincided with tumour response on MRI (Figs. 2c and 3c). For example, before the first infusion, patient no. 006 (sDMG) had severe paraplegia, neuropathic pain and bowel and bladder dysfunction. At the time of her best overall response (over 90% tumour reduction), she experienced intact bowel and bladder continence, markedly improved pain, and improved lower extremity function enabling ambulation with a cane. Similarly, patient no. 010 (DIPG) had sensory and motor deficits requiring assistance with gait at enrolment (used a wheelchair for longer distances) and experienced a complete response by imaging accompanied by significant clinical improvements, including improved left-sided hearing, improved hemifacial, hemibody and taste sensation, improved motor coordination and improved gait to the point of independent ambulation. For some patients, however, the relationship between imaging findings on MRI and clinical improvement was not direct, with patients no. 004 and no. 007 demonstrating sustained clinical improvement without substantial changes in overall tumour volume by MRI, and patient no. 012 exhibiting early clinical improvement but overall increase in tumour volume (Fig. 1b, Extended Data Figs. 4b and 9 and Supplementary Table 5).

18Median overall survival for patients treated on arm A was 20.6 months from diagnosis, with two patients with DIPG alive and followed subsequent to data cut-off (patient no. 007, length of follow-up 33 months; and patient no. 010, length of follow-up 30 months). For patients with DIPG, median overall survival was 17.6 months and for sDMG was 31.96 months (Extended Data Fig. 4c); comparison with historical controls was not possible in this study due to the highly selected nature of the trial cohort.

19Peripheral blood demonstrated GD2-CART expansion following IV infusion at levels similar to those seen in other active CAR T cell trials–, which persisted during ICV infusions (Figs. 2c and 3c and Extended Data Fig. 5a) but decreased over time. GD2-CART expansion was also evident in CSF following multiple repeated infusions (Extended Data Fig. 5b). Patient nos. 006 (Fig. 3c) and 009 (Extended Data Figs. 4a and 5a) experienced loss of detectable GD2-CART in the peripheral blood, as measured by PCR, temporally correlated with clinical and/or imaging progression. Increased cytokine/chemokine levels were present in peripheral blood following IV GD2-CART, whereas cytokine/chemokine levels were more marked in CSF following ICV GD2-CART (Extended Data Fig. 6), consistent with our previous report of DL1 patients. Sequential ICV GD2-CART infusions induced repeated, transient elevations in CSF cytokine/chemokine levels (Figs. 2d and 3d and Extended Data Fig. 6).

20Grade 2 and greater CRS correlated with increased plasma levels of the myeloid cell chemokine MCP1 (also called CCL2), and a trend towards increased levels of the T cell cytokine IL-2 following IV GD2-CART administration (Extended Data Fig. 7a). Grade 2 and greater TIAN correlated with increased CSF levels of IL-10 and MCP1, and a trend towards increased levels of TNF following ICV GD2-CART (Extended Data Fig. 7b). For examination of the correlates of response, we assigned patients to ‘responder’ and ‘non-responder’ groups based on their best tumour volume changes (Fig. 1b), with three patients defined as non-responders (overall increase in tumour size) and eight defined as responders (decrease in tumour size or overall stable tumour size and improvement in neurological function). Using this stratification, we found no differences in the CD4:CD8 ratio of the CAR T cell product between responders and non-responders (Extended Data Fig. 7c), but responders exhibited higher levels of IL-2 in plasma following IV GD2-CART administration (Extended Data Fig. 7d), potentially reflecting increased T cell activation in these patients. In individual patients, progression was temporally associated with a decrease in plasma IP-10 or CXCL10 (Extended Data Fig. 7e). CSF levels of immunosuppressive TGF-beta factors increased following GD2-CART administrations in some patients, and higher TGF-beta1 levels early following GD2-CART infusion trended with progression of disease (Extended Data Fig. 8a–c). Notably, TGF-beta levels in CSF were relatively low in patient no. 010, who had a complete response, throughout his course (Extended Data Fig. 8a,b).

21Cell-free tumour DNA (ctDNA) in CSF, as measured by ddPCR, demonstrated decreased H3K27M mutations per millilitre CSF in some patients associated with tumour regression, and increased levels at times of peak inflammation or when patients experienced clinical and/or imaging progression (Figs. 2c and 3c and Extended Data Fig. 5a). Taken together, these data demonstrate evidence for sustained expansion and persistence of GD2-CART in the blood of patients with DIPG and sDMG, associated with clinical and biological evidence of antitumour activity, and begin to identify potential contributors to, or correlates of, loss of durable response in patients.

22Recent advances in fundamental understanding of DMG have begun to translate to clinical advances–,–. The present study establishes GD2-CAR T cell therapy as a promising modality for a historically lethal CNS cancer. The rate of clinical improvements and tumour regressions on MRI imaging, including a complete response by RANO 2.0 criteria sustained for over 30 months and throughout the duration of this report, is reason for cautious optimism, both for DIPG/sDMG therapy and, more broadly, for CAR T cell therapy of solid tumours. Therapeutic activity was clearly demonstrated in several patients, based on tumour regression by imaging and improvements in neurological symptoms. Whether GD2-CAR T cell therapy favourably impacts overall survival is difficult to discern, given that the eligibility criteria excluded patients with lower performance status, the need for steroid therapy, bulky thalamic or cerebellar tumour involvement and other factors. Moreover, multiple patients demonstrated MAPK pathway and ATRX or DAXX mutations, which can be associated with relatively longer survival and more localized disease. The effects of GD2-CAR T cell therapy on overall survival will need to be assessed in future Phase II trials.

23The toxicities observed here were generally anticipated, but with some notable observations. Cytokine release syndrome was dose limiting following IV infusion of GD2-CART for DIPG/DMG. By contrast, CRS did not commonly complicate ICV infusions. Given that the antitumour immune response occurs within the CNS in both contexts, this marked difference in systemic cytokine responses leading to CRS following IV administration is somewhat unexpected, and underscores the unique features of immune cell trafficking into and out of the CNS only recently coming to light,. Unexpectedly, ICANS was observed only following IV administration, but not with ICV infusions, despite the higher levels of cytokines/chemokines evident in CSF following ICV GD2-CART infusions. Whereas the rate of intratumoural haemorrhage observed in this cohort (1 of 11) is in line with the natural history of this disease, future study is needed to determine whether CAR T cell therapy alters the risk of intratumoural bleeding. Inflammation of neural structures can cause oedema with consequent mechanical complications such as obstruction of CSF flow, and immune signalling can also affect neuronal function. Varying grades of TIAN occurred commonly, as would be expected with inflammation of tumours located in structurally constrained and functionally eloquent neuroanatomical areas such as the brainstem and spinal cord. Concordantly, tumours located in the lower brainstem were associated with more severe TIAN. TIAN was transient and manageable with intensive support. In general, earlier infusions and larger tumours were associated with greater toxicity. The correlation of more severe TIAN with higher levels of cytokines/chemokines related to myeloid response, such as MCP1 or CCL2, raises the prospect that myeloid cells may contribute to neurological symptoms induced by CAR T cell therapy for CNS tumours, a hypothesis that requires testing in future studies. The observed decrease in TIAN severity with repeated infusions is notable and could be due to decreased tumour burden with repeated infusions, increased immunomodulatory or immunosuppressive mechanisms in the tumour microenvironment, or both.

24Given the promising results noted in arm A with repeated ICV administration of GD2-CAR T cells, but with dose-limiting CRS following IV administration at high dose levels, we have now initiated arms B and C to test ICV-only administration with and without lymphodepleting chemotherapy. Given the burden of monthly infusions on the patient, we have also included formalized assessments of quality of life and patient-reported outcomes in arms B and C. Ongoing and future work will define the optimal route of administration, the role of lymphodepleting chemotherapy and potential combination strategies for optimization of this promising therapy to achieve more complete and durable responses for children and adults with H3K27M-mutated diffuse midline gliomas.

25The Stanford Institutional Review Board approved this trial, which was registered with ClinicalTrials.gov (NCT04196413). Informed patient or parent consent and child assent were obtained. Included in consent is permission to share the (deidentified) results of the trial in scientific publications or presentations. Permission to publish images, photographs and videos was also obtained.

26Clinical response was assessed using the CIS generated by protocol-directed neurological examinations performed by a neuro-oncologist at prescribed times following GD2-CAR T cell administration. The CIS represents a simple quantification of the neurological examination, which added or subtracted one point for each symptom/sign that improved or worsened, respectively, from the patient’s preinfusion baseline examination (Supplementary Methods 5). For patients who received corticosteroid therapy for toxicity management, CIS scoring was deferred until at least 7 days following corticosteroid discontinuation. Imaging responses were assessed by MRI scans of the brain and/or spinal cord with and without gadolinium. Because DMGs are diffusely infiltrative of CNS structures and difficult to measure in linear dimensions, volumetric segmentation of tumour corresponding to abnormal T2 signal was performed by a neuroradiologist to measure radiographic change in tumour volume, consistent with RANO 2.0 recommendations.

27Sample size estimation was based on clinical considerations and the Phase I 3 + 3 design. The targeted DLT rate was 30% or less. All patients who enroled into arm A and received one infusion of GD2-CAR T cells on trial were included in the analysis. Descriptive statistics were used to summarize baseline patient and disease characteristics, toxicity data and correlative and clinical outcomes. The Shapiro–Wilk test of normality was used to assess normal (Gaussian) distribution of tumour volumetric response data. Overall survival was measured from date of diagnosis to the date that event occurred, or censored at the time of data cut-off. Survival probability was estimated using the Kaplan–Meier method. The confidence interval of median survival time was constructed by the method of Brookmeyer–Crowley. All comparisons made in correlative outcomes were exploratory. Mann–Whitney U-test was used, with no adjustments made for multiple comparisons. Statistical analyses were conducted using Prism software.

28Peripheral blood and CSF samples were collected before and following IV and ICV infusions to measure cytokine/chemokine levels in blood and CSF, GD2-CART persistence by qPCR and CAR-FACS and cell-free tumour DNA in CSF, as detailed in the previous report of the first four patients and reiterated below.

29Patient blood samples were processed and mononuclear cells viably cryopreserved. DNA was extracted from whole blood (2 × 106–5 × 106 peripheral blood mononuclear cells (PBMCs)) using the QIAmp DNA blood Mini Kit (Qiagen, catalogue no. 51306) at baseline and at multiple time points following CAR administration. CAR presence was measured by qPCR using the primer and probe sequences provided below. For the standard curve, a custom Minigene plasmid (IDT) was designed containing a partial GD2.41BB.z sequence and a partial albumin sequence, which served as a control for normalization. The standard curve contained a tenfold serial dilution of plasmid ranging between 5 × 108 and 5 × 100 copies. Both plasmid and patient DNA from each time point were run in triplicate, with each reaction containing 5 µl of DNA (50 ng total), 200 nM forward and reverse albumin primers (or 300 nM forward and reverse GD2.41BB.z primers), 150 nM probe suspended in 10 µl of TaqMan Fast Universal PCR Master Mix (2X), no AmpErase UNG (Thermo Fisher Scientific) and either 24.5 µl (albumin) or 22.5 µl (GD2.41BB.z) of TE buffer (Invitrogen, catalogue no. AM9935). The Thermo Fisher Scientific QuantStudio 6 Pro Real-time qPCR Instrument was used for qPCR, with 20 µl per reaction. Quality metrics for all qPCR standard curve results were R2 > 0.95 and efficiency 70–110%.

30Albumin results from the plate were normalized to average albumin, then GD2-CAR copy number (copies per 50 ng of DNA), adjusted to albumin and modified to copies per 100 ng of DNA, was calculated by the following equation: copy number (copies per 100 ng of DNA) = 2 × (GD2-CAR copy number × (albumin copy number/average albumin)).

31A high-dimensional, immunophenotyping, flow cytometry panel was designed for immune profiling of CAR T cells in real time. PBMCs were isolated from fresh whole blood by gradient centrifugation on ficoll (Ficoll paque Plus, GE Healthcare, Sigma-Aldrich). Between 2 million and 5 million PBMCs were stained with fixable Live/Dead aqua (Invitrogen) amine-reactive viability stain. Cells were preincubated with Fc block (trustain, BioLegend) for 5 min, then stained at room temperature with fluorochrome-conjugated mAb in a 15-colour, 17-parameter staining combination as previously described. Antibody–fluorochrome conjugates used were: anti-CD3-FITC (BioLegend, 0.5 µg of antibody per 1 million cells in 100 µl); anti-CD8-PerCP Cy5.5 (BD Biosciences, 0.5 µg of antibody per 1 million cells in 100 µl); anti-CD45-BV785 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anto-CD4-BV711 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anti-CD95-BV650 (BioLegend, 0.5 µg of antibody per 1 million cells in 100 µl); CD39-BV605 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anti-CD57-BV421 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anti-CCR7-BUV805 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anti-CD45RA-Alx700 (BioLegend, 0.5 µg of antibody per 1 million cells in 100 µl); anti-GD2-CAR-DyLight650 (custom, clone 1A7, 1 µg of antibody per 1 million cells in 100 µl); anti-CD14-PE-Cy7 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anti-CD11b-APC-Cy7 (BioLegend, 1 µg of antibody per 1 million cells in 100 µl); anti-CD33-PE-Dazzle (BioLegend, 5 µl of antibody per 1 million cells in 100 µl); and anti-GD2-PE (BioLegend, 5 µl of antibody per 1 million cells in 100 µl). For determination of the optimal concentration of antibody for staining in the CAR-FACS panel, each fluorochrome-conjugated antibody was titrated to determine the saturating amount of antibody needed to stain a test of 1 million cells. Dose–response curves for each antibody informed the saturating amount of fluorochrome-conjugated antibody required for staining 1 million cells in 100 ml of staining volume.

32CAR-tranduced T cells were used as positive control included in daily staining experiments. Immunostained and fixed cells were acquired on an LSR (BD BioSciences) five-laser (blue, 488 nm; violet, 405 nm; ultraviolet, 355 nm; red, 640 nm; green, 532 nm) analyser. A minimum of 106 cells were acquired, unless restricted by the number of cells isolated from 8 ml of whole blood or when acquiring isolated cells from CSF. The assay limit of detection for cells was calculated as 1 in 104 of total acquired PBMCs. Representative gating is shown in Extended Data Fig. 10.

33Patient blood and CSF samples were collected at predetermined and trigger time points throughout treatment. Samples were spun at 250g for 6 min. Supernatant was frozen at −80 °C until batched for assessment. Cytokine assessment was performed with the Immunoassay Team-Human Immune Monitoring Center at Stanford University. Panels include Luminex–EMD Millipore HIMC H80 (panel 1 is Milliplex HCYTA-60K-PX48; panel 2 is Milliplex HCP2MAG-62K-PX23; panel 3 includes Milliplex HSP1MAG-63K-06 and HADCYMAG-61K-03 (resistin, leptin and hepatocyte growth factor) to generate a nine-plex) and TGF-b (TGFBMAG-64K-03). Kits were purchased from EMD Millipore and used according to the manufacturer’s recommendations, with modifications described. The assay set-up followed the recommended protocol. Briefly, samples were diluted threefold for panels 1 and 2 and tenfold for panel 3. Diluted sample (25 µl) was mixed with antibody-linked magnetic beads in a 96-well plate and incubated overnight at 4 °C with shaking. Both cold and room-temperature incubation steps were performed on an orbital shaker at 500–600 rpm. Plates were washed twice with wash buffer in a Biotek ELx405 washer (BioTek Instruments). Following incubation for 1 h at room temperature with biotinylated detection antibody, streptavidin-PE was added for 30 min with shaking. Plates were washed as described above, and PBS added to wells for reading in the Luminex FlexMap3D Instrument with a lower bound of 50 beads per sample per cytokine. Custom Assay Chex control beads (Radix Biosolutions) were purchased and added to all wells. Wells with a bead count under 50 were flagged, and data with a bead count under 20 were excluded. Data are presented as picograms per millilitre, based on standard curves or heat maps of fold change from the baseline time point. All samples were run in technical duplicate.

34Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

35Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41586-024-08171-9.