A vaccine targeting mutant IDH1 in newly diagnosed glioma

Mutated isocitrate dehydrogenase 1 (IDH1) defines a molecularly distinct subtype of diffuse glioma1–3. The most common IDH1 mutation in gliomas affects codon 132 and encodes IDH1(R132H), which harbours a shared clonal neoepitope that is presented on major histocompatibility complex (MHC) class II4,5. An IDH1(R132H)-specific peptide vaccine (IDH1-vac) induces specific therapeutic T helper cell responses that are effective against IDH1(R132H)+ tumours in syngeneic MHC-humanized mice4,6–8. Here we describe a multicentre, single-arm, open-label, first-in-humans phase I trial that we carried out in 33 patients with newly diagnosed World Health Organization grade 3 and 4 IDH1(R132H)+ astrocytomas (Neurooncology Working Group of the German Cancer Society trial 16 (NOA16), ClinicalTrials.gov identifier NCT02454634). The trial met its primary safety endpoint, with vaccine-related adverse events restricted to grade 1. Vaccine-induced immune responses were observed in 93.3% of patients across multiple MHC alleles. Three-year progression-free and death-free rates were 0.63 and 0.84, respectively. Patients with immune responses showed a two-year progression-free rate of 0.82. Two patients without an immune response showed tumour progression within two years of first diagnosis. A mutation-specificity score that incorporates the duration and level of vaccine-induced IDH1(R132H)-specific T cell responses was associated with intratumoral presentation of the IDH1(R132H) neoantigen in pre-treatment tumour tissue. There was a high frequency of pseudoprogression, which indicates intratumoral inflammatory reactions. Pseudoprogression was associated with increased vaccine-induced peripheral T cell responses. Combined single-cell RNA and T cell receptor sequencing showed that tumour-infiltrating CD40LG+ and CXCL13+ T helper cell clusters in a patient with pseudoprogression were dominated by a single IDH1(R132H)-reactive T cell receptor.


Article
Two patients (ID19, ID21) were enrolled but could not be evaluated for immunogenicity testing and were therefore excluded from the immunogenicity analysis. Thirty out of the 32 patients in the SDS (93.8%) and 28

IDH1-vac is safe and immunogenic
The SDS comprised 249 vaccines administered to 32 patients. Twenty-nine out of 32 patients in the SDS (90.6%) and 27 out of 30 patients in the IDS (90.0%) received all 8 vaccinations; one patient received 7, one received 6, and one received 4 vaccinations. The duration of treatment in the SDS ranged from 44 to 162 days (median, 155 days) and the duration of observation ranged from 153 to 484 days (median 376 days). Quality control demonstrated that all vaccines contained 300 ± 30 μg peptide, and were sterile and free from endotoxins. No regime-limiting toxicity (RLT) was observed. The overall serum cytokine profile was indicative of an adverse cytokine release in response to IDH1-vac (Extended Data Fig. 2). Twenty-nine of the 32 patients (90.6%) had treatment-related adverse events, none of which was severe. One patient (3.1%) had treatment-related serious adverse events, and one patient (3.1%) temporarily discontinued the study drug due to treatment-related adverse events (Supplementary Tables 4, 5). Twenty-one (65.6%; 95% CI 46.81-81.43%) and 15 (46.9%; 95% CI 29.09-65.26%) of the adverse events classified as possibly related to IDH1-vac were local administration site conditions (injection site induration or erythema, respectively). Of the 30 patients in the IDS, 28 (93.3%; 95% CI 77.93-99.18%) displayed IDH1-vac-induced immune responses (Fig. 2a, b). IDH1-vac-induced T cell immune responses were observed in 26 of 30 patients and B cell immune responses in 28 of 30 patients across multiple human leukocyte antigen (HLA) alleles; these responses did not correlate with in vitro HLA affinities of the IDH1(R132H) peptide (Extended Data Fig. 3, Supplementary Table 6). Two patients (6.7%) developed neither T cell nor B cell immune responses (Extended Data Fig. 3). To incorporate the duration and level of IDH1-vac-induced T cell immune responses specifically to IDH1(R132H), we established an explorative mutation-specificity score (MSS; Supplementary Table 7, Extended Data Figs. 4,5). Flow cytometric effector sub-phenotyping of peripheral IDH1-vac-induced T cells from available patient samples with high MSSs showed predominant tumour necrosis factor (TNF), interferon-γ (IFNγ), and interleukin-17 (IL-17) cytokine production by T helper (T H ) cells upon in vitro re-stimulation with IDH1(R132H), which indicates the involvement of T H 1 and T H 17 subtypes of T H cells (Fig. 2c, Extended Data Fig. 4). Neither IL-10 production by regulatory T cells nor TNF or IFNγ production by cytotoxic T cells was observed (Fig. 2c). Moreover, the MSS was associated with intratumoral IDH1(R132H) antigen presentation in pre-treatment tumour tissue, as assessed by an in situ MHCII-IDH1(R132H) proximity ligation assay (PLA) 5 Table 8).

IDH1-specific T cell responses and pseudoprogression
In the SDS, pseudoprogression (PsPD) occurred in 12 of 32 patients (37.5%) compared to 10 of 60 (16.7%) in a molecularly matched control cohort (Supplementary Table 3). Contrast-enhancing PsPD diagnosed by brain imaging is indicative of intratumoral inflammatory reactions (Fig. 3c). There was no apparent association with age, extent of resection, SOC treatment, or WHO grade. The longer median observation period (7.3 years) in the matched cohort caused a bias towards the detection of more PsPD. In NOA16, PsPD was associated with the onset of peripheral IDH1-vac-induced immune responses (Extended Data Fig. 7) and was restricted to patients with transient or sustained T cell immune responses; we did not detect PsPD in non-responder patients (Fig. 3d). ID01  ID02 ID03  ID04 ID05  ID06  ID07  ID08   ID09  ID10 ID11  ID12  ID13 ID14  ID15  ID17   ID18  ID19  ID20 ID21 ID22  ID23 ID24  ID25   ID26  ID27  ID28 ID29 ID30  ID31  ID32  Patients with PsPD had higher maximal levels of peripheral IDH1-vac induced T cell immune responses than patients who had progressive disease (Fig. 3e). Retrospective assessment of prognostic molecular markers in pre-treatment astrocytoma tissues (Supplementary Table 9, Extended Data Fig. 8) enabled further subgrouping of 24 out of 32 (75.0%) patients in the SDS. PsPD was not associated with any of the assessed tumour-intrinsic molecular markers, such as copy number variation load (CNV-L), methylation class, CDKN2A or CDKN2B deletion status, frequencies of peripheral immune cell subsets, or alterations in top peripheral T cell clonotypes (Supplementary Table 9, Extended Data Figs. 7,9,10). During followup, four out of ten patients (40%) with methylation class high grade glioma experienced progressive disease. Of these, patients with an MSS that stayed below median had a 2-year progression-free rate of 0.4 (95% CI 0.052-0.753) compared to 0.8 (95% CI 0.204-0.969) for patients with an MSS that reached above median, despite an equal distribution of unfavourable molecular markers (Fig. 3f, g). Seven out of 12 (58.3%) of the patients with PsPD, including patient ID08, still have stable disease with a median followup time of 53.1 months (95% CI 45.8-58.2 months).

Article
Conclusions NOA16 met its primary endpoints by demonstrating the safety and immunogenicity of IDH1-vac in patients with newly diagnosed WHO grade 3 and 4 IDH1(R132H) + astrocytomas without further positive prognostic factors. Immunogenicity, irrespective of HLA type, and the high rate of PsPD warrant further clinical investigation of IDH1-vac. Patients who did not mount an IDH1-vac induced immune response showed reduced efficacy of the vaccine and disease progression within two years (Extended Data Fig. 3) compared to patients who did mount an immune response (Fig. 3b). IDH1-vac was immunogenic across multiple HLA alleles, supporting the concept of promiscuity of presentation on MHCII 4 ID02  ID05  ID30  ID14  ID29  ID13  ID18  ID07  ID08  ID26  ID24  ID09  ID17  ID10  ID15  ID32  ID04  ID31  ID20  ID01  ID23  ID28  ID33  ID03  ID25  ID22  ID11  ID27  ID12  ID06   0 ID06  ID02  ID04  ID01  ID13  ID07  ID14  ID20  ID27  ID10  ID15  ID23  ID29  ID18  ID33  ID17  ID08  ID22  ID19  ID32  ID26  ID11  ID31  ID25  ID28  ID24  ID12  ID21  ID05  ID03  ID09   to a molecularly matched cohort and previous reports (3 out of 60 patients, 5.0%) 11 may indicate an intratumoral immune reaction that results in disruption of the blood-brain barrier and contrast enhancement. The Response Assessment in Neuro-Oncology (RANO) criteria applied in this trial consider PsPD to be secondary to radiotherapy or combined radiochemotherapy with TMZ, particularly four weeks after completion of radiotherapy 12 , and most trials mandate follow-up imaging to demonstrate true progression in cases of suspected PsPD, according to RANO 13,14 . We excluded patients with suspected PsPD from the NOA16 trial, thereby enriching for patients with IDH1-vac-induced PsPD, but we acknowledge that late PsPD may occur as a result of radiotherapy 15 . Also, late PsPD six months after initiation of immunotherapy may occur, as acknowledged in the immunotherapy RANO (iRANO) criteria 13 , which were not defined at the time of initiation of this trial. Notably, the rate of PsPD in NOA16 did not differ when analysed according to iRANO criteria. However, there are limitations to definitive proof of PsPD, even with positron emission tomography (PET) imaging or histologic analysis of re-resection, as no firm criteria exist 16 .
NOA16 is based on strong preclinical data 4,6 and the decision to integrate IDH1-vac into the primary treatment of newly diagnosed patients provided a sufficient therapeutic window and allowed us to exploit potential positive immune interactions between SOC and vaccination. While this strategy has been chosen in other trials that have targeted shared 17 or personalized neoepitopes 18,19 , NOA16 targeted a shared clonal neoepitope to minimize the risk of immune evasion by clonal selection or spontaneous neoantigen loss 17 . Clonality of neoepitopes is a key determinant of efficacy for immune checkpoint inhibitors across many cancer entities 20 . Gliomas are particularly prone to the development of subclonal mutational events that contribute to resistance to immune checkpoint inhibitors 21 . Targeting a shared clonal driver mutation in newly diagnosed patients overcomes these limitations 6 and may provide a basis for future trials that target MHCII-restricted clonal shared and personalized neoepitopes in cancer immunotherapy.

Online content
Any methods, additional references, Nature Research 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 https://doi.org/10.1038/s41586-021-03363-z. a c -10 γδ T cell 1 2 3 4 5 6 7 8 9 10 11 12 13 14  evaluable patients had to be enrolled. Owing to an expected dropout rate of 20% (due to progression or other reasons), 39 patients had to be recruited. All patients received the trial-related intervention; the trial was not randomized and investigators were not blinded concerning trial related intervention during experiments and outcome assessment.

IDH1 vaccination
IDH1-vac consisted of 300 μg of an IDH1(R132H) 20-mer peptide (p123-142) manufactured by the GMP facility of the University of Tübingen, Germany and emulsified in Montanide (ISA50) as described earlier 23 by the GMP core facility at the University Hospital Heidelberg, Germany, a maximum of one day in advance. It was administered subcutaneously in combination with topical imiquimod (5%, Aldara). Quality controls for content, sterility and absence of endotoxin were performed for each emulsion at Labour LS s.e. & Co. KG, Germany.

Endpoints
The primary endpoints were safety and immunogenicity. The safety endpoint was the RLT, which was defined as one of the following that was related to IDH1-vac administration: any injection site reaction of National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) version 4.0 grade 4; any injection site reaction of CTCAE grade 3 that persisted after two weeks; any other hypersensitivity, anaphylaxis or local allergic reaction of at least CTCAE grade 3; brain oedema (CTCAE grade 4); autoimmunity of CTCAE grade 3 or more; CTCAE grade 3 or more toxicity to organs other than the bone marrow, but excluding grade 3 nausea, grade 3 or 4 vomiting in patients who had not received optimal treatment with anti-emetics, grade 3 or 4 diarrhoea in patients who had not received optimal treatment with anti-diarrheals, and grade 3 fatigue; and death. Adverse events were counted as treatment-related if the relationship to treatment was 'certain', 'related', 'probable', 'possible', or not reported. For safety assessment, patients were medically reviewed at each visit.
To exclude unexpected IDH1-vac-induced immunological tolerance against IDH1(R132H), shortening of progression-free survival (PFS), defined as an observed decrease in the estimated 12-months PFS rate of at least 10% compared to the anticipated value of 70.7% derived from previous studies, was defined as a safety criterion for early trial termination. The safety analysis was based on all enrolled patients who received one or more administration(s) of IDH1-vac. The immunogenicity endpoint was defined as the presence of an IDH1(R132H)-specific T cell and/or antibody response at any time point during the trial. IDH1(R132H)-specific T cell and antibody responses were measured on PBMCs using IFNγ ELISpot and on serum using peptide-coated ELISA, respectively. For IFNγ ELISpot, a cut-off of 50 IFNγ spots after subtraction of negative control was defined as positive. For ELISA, the cut-off for positivity was defined as optical density related to negative control ≥5.

Disease assessment
Disease assessment, including overall response rate and diagnosis of PsPD, was performed using standardized three-monthly MRI according to the RANO criteria by central neuroradiology review 14 . In NOA16 and the molecularly matched control cohort, PsPD, which may mainly indicate an intratumoral inflammatory reaction 24 , was defined as an increase in the size of the tumour on T2-FLAIR MRI sequences and/or the novel appearance or enlargement of contrast-enhancing lesions followed by stabilization or regression on follow-up MRI up to three months after initiation of SOC and/or immunotherapy 14 .

Preparation of peptides for analyses
Lyophilized peptides were reconstituted in 100% DMSO and diluted to a final concentration of 10 mg ml −1 with aqua ad iniectabilia (Braun). The final DMSO concentration was 10%.

Isolation of serum
Serum tubes were kept standing upright at room temperature for 15 min before isolation. Serum tubes were centrifuged at 1,000g for 10 min at room temperature. Supernatant was aliquoted on ice and frozen at -80 °C.

Isolation of LILs
Lesion tissue was dissected into small pieces (2 × 2 mm) and transferred into 24-well tissue culture-treated plates at three pieces per well in 2 ml human tumour-invading lymphocyte (TIL) medium (RPMI1640 (Pan Biotec) with 10% human serum (Sigma Aldrich), 2 mM l-glutamine, 1.25 μg/ml amphotericin B (both Gibco), 1,000 U/ml IL-2 (Proleukin)) containing 30 ng/ml anti-human CD3 (clone OKT-3, eBioscience). Medium was exchanged every 2-3 days and tissue pieces removed on day 7. LILs that migrated out of the tumour into the medium were further expanded until day 14 and cryopreserved as above.

Generation of patient REP cells
To enable HLA-autologous testing for antigen-specific reactivity of LILs and TCR-transgenic cells, patient-autologous rapidly expanded PBMCs (REP cells), which express high levels of MHC molecules and can serve as antigen-presenting cells (APCs), were generated. PBMCs (1 × 10 5 ) were co-cultured in a high-density culture with 3 × 10 7 irradiated (40 Gy) feeder cells (PBMCs from non-autologous donors) in X-vivo15 medium supplemented with 2% human AB serum (Sigma-Aldrich) and 30 ng/ml OKT-3 antibody (Invitrogen) in T-25 flasks in 25 ml total volume. After 24 h, cells were supplemented with 300 IU/ml hIL-2. Medium was replaced every 5 days with hIL-2 supplementation and cells were split as needed. Cells were collected after 14 days of co-culture and cryopreserved.

IFNγ ELISpot of PBMCs
ELISpot white-bottom multiscreen HTS plates (MSIPS4W10, Millipore) were coated with anti-human IFNγ (1-D1K, Mabtech) and blocked with X-Vivo-20 (Lonza) containing 2% human albumin (HA). PBMCs were thawed, rested overnight in X-Vivo medium and seeded at 4 × 10 5 cells per well and stimulated with 2 μg peptides per well in 100 μl volume. PBMCs were stimulated with IDH1(R132H) (p123-142), wild-type IDH1 (p123-142), or MOG (p35-55) at equal concentrations or with peptide diluent aqua ad iniectabilia (Braun) with 10% DMSO (vehicle) at equal volume as negative controls, or with 1 μg staphylococcal enterotoxin B (Sigma-Aldrich) per well and 0.05 μg CMV with 0.05 μg AdV per well (both in 100 μl volume) as positive controls. After 40 h, IFNγ-producing cells were detected with biotinylated anti-human IFNγ antibodies (7-B6-1), streptavidin-ALP (both Mabtech) and ALP colour development buffer (Bio-Rad) and quantified using an ImmunoSpot Analyzer (Cellular Technology Ltd). Quality control was performed and reviewed by a second person. For categorization of T cell responses, transient T cell responses were defined as a spot count above 50 followed by a spot count of less than 50 at EOS. Sustained T cell responses were defined as a spot count above 50 followed by a spot count of more than 50 at EOS.

Flow cytometry
Antibody amounts were titrated previously or used according to manufacturer's instructions, and scaled up according to cell numbers at time of seeding. In all experiments, corresponding FMO controls were used (Extended Data Fig. 4). As many events as possible were measured on a Lyric Flow Cytometer (BD Bioscience) using BD FACSuite sotware version 1.3.
Data analysis for all experiments was done using FlowJo software v.10.5.0.

Detection of cytokines in serum
Serum was analysed using multiplex bead technology (Bio-Plex Pro Human Cytokine 27-plex panel, order no. M500KCAFOY, Bio-Rad, Hercules, CA) according to the manufacturer´s instructions. Serum was diluted 1:2. Standard curves were generated by using the reference cytokine sample supplied in the kit and were used to calculate the cytokine concentrations in the samples. Acquisition and data analysis were performed by bio-plex Manager.

Proximity ligation assay
PLA was performed on baseline paraffin-embedded glioma tissues as described previously 5 . For image acquisition, a nonlinear adjustment (gamma changes) was used for visualization purposes.

TCRB deep sequencing
Genomic DNA was isolated from patient EDTA blood using the DNeasy Blood and Tissue Kit (Qiagen). TCR beta chain (TCRB) deep sequencing was performed to detect rearranged TCRβ gene sequences using hsTCRB Kit (Adaptive Biotechnologies) according to the manufacturer's protocol. The prepared library was sequenced on an Illumina MiSeq by the Genomics & Proteomics Core Facility, German Cancer Research Center (DKFZ). Data processing (demultiplexing, trimming, gene mapping) was done using the Adaptive Biotechnologies proprietary platform. Data were visualized using the Treemap Visualization package version 2.4.2 (https://cran.r-project.org/web/packages/treemap/index. html). TCRB sequencing data are available at https://clients.adaptivebiotech.com/pub/platten-2021-nature.

Next-generation HLA typing
Genomic DNA was isolated from patient EDTA blood using the QIAamp DNA Blood Mini Kit (Qiagen). Subsequently, peptide-binding domains were sequenced as described previously 25 .
850k methylation arrays 850k methylation arrays were performed as described previously 26 .
Panel sequencing DNA from FFPE tissue was extracted on the Promega Maxwell device (Promega) following the manufacturer's instructions. Extracted DNA was then sheared on a Covaris M220 (Covaris). DNA integrity and fragment size were determined on a Bioanalyzer 2100 (Agilent). Sequencing was performed on a NextSeq 500 instrument (Illumina) with an average coverage of 550-fold 27 .

Single-cell RNA and TCR sequencing
Single-cell capturing and downstream library constructions of FACS-sorted cells were performed using Chromium Single Cell V(D)J Reagent kit v1 chemistry (10x Genomics; PN-1000006, PN-1000020, PN-1000005, PN-120262) according to the manufacturer's protocol. The constructed scVDJ library and scGEX libraries were sequenced on HiSeq2500 rapid and HiSeq4000 platforms (Illumina), respectively.
Single-cell RNA data were processed using cellranger pipeline (version 3.1.0) with GRCh38 genome assembly (version 3.0.0, 10x Genomics) with default setting. The filtered matrices were then analysed using Seurat 28 . Cells with fewer than 2,000 unique molecular identifiers, fewer than 900 genes, and/or more than 10% mitochondrial gene expression were excluded from the analysis. Genes detected in fewer than three cells were excluded. Gene expression was transformed and normalized using regularized negative binomial regression as implemented in sctransform 29 . VDJ genes were removed from the variable genes to prevent clustering of cells on the basis of TCR clones. Highly variable genes were selected using principal component analysis, and 40 principal components were selected on the basis of inflection point in the elbow plot. Cells were clustered using graph-based clustering with Louvain modularity of 0.45 and UMAPs were plotted for visualization. Differential gene expression analysis was performed using MAST 30 to determine the identity of each cluster and highly upregulated genes were used to label each cluster. Clusters with upregulated heat shock proteins and CD3 − cells were excluded and cells were re-normalized and re-clustered as described above. Single-cell VDJ data were processed similarly using cellranger pipeline. Barcodes of individual top TCRs were then mapped onto single-cell RNA data to determine the distribution of TCR clones in the clusters.
Single-cell sequencing data have been deposited in the NCBI Sequence Read Archive with the accession codes SRR12880623 and SRR12880624.

TCR cloning
Synthetic alpha and beta VDJ fragments of the variable region of the TCR compatible with BsaI-mediated Golden Gate Assembly cloning were obtained from Twist Biosciences. An S/MAR sequence-bearing expression vector (pSMARTer) that allows extrachromosomal replication of the vector in eukaryotic cells was used and designed to harbour mouse alpha and beta constant TCR regions and a p2a self-cleaving peptide linker to facilitate production of separate alpha and beta polypeptide chains of the TCR. The TCR variable fragments were inserted into the expression vector using a single-step Golden Gate reaction and transformed into NEB5-alpha-competent Escherichia coli (NEB). Colonies were screened for the transgene by antibiotic resistance, and an endotoxin-free plasmid was prepared using NucleoBond Extra Maxi EF kit (Macherey-Nagel) for transfection.

TCR-NFAT reporter assay
The cloned TCR expression vector and a nano-luciferase-based NFAT reporter vector (pDONR, with 4× NFAT-response elements) were delivered into Jurkat Δ76 cells (obtained from TRON gGmbH, authenticated using the Multiplexion STR profiling and compared to normal Jurkat cells, regularly tested for mycoplasma contamination and tested negative at all time points) using electroporation (Neon Transfection system, ThermoFisher Scientific). In brief, 2 × 10 6 cells were used per electroporation with Neon 100-μl tips (8 μg TCR expression vector with 5 μg NFAT reporter vector). Cells were prepared according to the manufacturer's protocol; electroporated with 1,325 V, 10 ms, 3 pulses; and transferred to antibiotic-free RPM1 1640 medium containing 10% FCS. Patient-autologous PBMCs or REP cells were used as APCs as indicated and thawed 24 h before co-culture in X-VIVO 15 medium (Lonza) containing 50 U/ml benzonase (Sigma-Aldrich), rested for 6-8 h before seeding into 96-well white-opaque tissue culture-treated plates (Falcon) at 1.5 × 10 5 cells per well, and loaded with peptides at a final concentration of 10 μg/ml in a total volume of 150 μl for 16 h. A pool of human IDH1(R132H) peptides (p122-136, p124-138, p126-140) was used. MOG (p35-55) at equal concentrations and PBS + 10% DMSO (vehicle) at equal volume were used as negative controls. Forty-eight hours after electroporation, Jurkat Δ76 cells were collected and co-cultured with peptide-loaded PBMCs for 6 h at a 1:1 ratio. Human T cell TransAct beads (Miltenyi) were used as positive control. Nano-luciferase induction, indicating TCR activation, was assayed using the Nano-Glo Luciferase assay system (Promega) according to the manufacturer's protocol and signal was detected on a PHERAstar FS plate reader (BMG Labtech).

In vitro HLA affinity analyses
Peptides were synthesized by Genscript and dissolved in DMSO followed by dilution in assay buffer. The final DMSO concentration was 10%. Peptides did not contain cysteines so no reducing agent was added. As positive controls, peptides CLIP (PVSKMRMATPLLMQA), KLAT (HA306-318, YKYVKQNTLKLAT) or PADRE (AKFVAAWTLKAAA) were used. Peptides were titrated in assay buffer (10,000, 1,000, 100, 10, 1, 0.1, 0.01, and 0.001 nM) and recombinant MHC II of different alleles and paralogues was added. After at least 24 h of refolding, solutions were transferred to optiplates AlphaScreen acceptors and donor beads were added. Raw data was imported into Microsoft Excel and deconvoluted. For some peptides the highest concentrations led to a reduction in signal (hooking effect). These datapoints were deleted. Data were imported into GraphPad Prism software version 9.0.0 and analysed by sigmoid curve fitting. All experiments were done in duplicate with good correlation.

Statistics
For statistical analyses of primary endpoints, two patient analysis populations were defined. The safety population included all enrolled patients who had at least received one dose of IDH1-vac. This was the analysis dataset for evaluating patient characteristics, study administration, efficacy (overall response rate, i.e. stable disease), and safety endpoints (safety dataset, SDS). The immunogenicity population (immunogenicity dataset, IDS) included all patients who could be evaluated for immunogenicity assessment. A patient was defined as evaluable if they had completed the study up to and including V07, had received at least four vaccinations through V07 and had all intended blood samples collected for immune monitoring through V07; or had received at least 6 of 8 vaccinations, and baseline plus at least two further blood samples had been collected for immune monitoring through V12. Non-evaluable patients were replaced for assessment of immunogenicity, except for patients who left the study early owing to RLT. For the primary endpoints (RLT and immune response), summary tables, percentages and exact 95% CIs according to Clopper-Pearson were generated.
All secondary variables were analysed using explorative and mainly descriptive methods using GraphPad Prism software version 9.0.0. For PLA, Pearson correlation coefficient was calculated. For contingency analyses, Fisher's exact test was performed. For multiple comparisons, a Kruskal-Wallis test (KWT) by ranks was performed and multiplicity adjusted P values (Dunn's test) are presented. All statistical tests were two-tailed to a significance level of 5%. For detailed description of exploratory analyses, see Supplementary Table 8. For analysis of selected secondary variables, a molecular dataset was defined. The molecular dataset included all patients whose astrocytomas could retrospectively be defined molecularly according to copy number variation load (CNV-L), methylation class, and CDKN2A/B status.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
Single-cell RNA-seq data that are associated with Fig. 4 and Extended Data Figs. 11, 12 have been deposited in the NCBI Sequence Read Archive with the accession codes SRR12880623 and SRR12880624. TCRB sequencing data that are associated with Extended Data Fig. 10 are available at https://clients.adaptivebiotech.com/pub/platten-2021-nature. Fig. 3 | Relationship between MHC alleles and T cell response. a, Venn diagram of T cell non-responders and B cell non-responders in the IDS. b, c, Allele prevalence of MHC class I supertype families (b), and MHC class II DRB1* alleles with a total prevalence of three or more, and paralogues (c). Grey, numbers of alleles present or absent (for paralogues) in patients with T cell responses to IDH1-vac (T cell response); black, numbers of alleles present or absent (for paralogues) in patients without T cell responses to IDH1-vac (T cell non-response). n (total alleles) = 64 for 32 patients in the SDS. d, IDH1