Human clinical datasets
The intratumoural multi-region sampling cohort for various cancer types utilizes RNA-seq data from the following studies: this paper, for multi-region sampling of GBM and LGG; ref. 24, for multi-region sampling of hepatocellular carcinoma; ref. 23, for multi-region sampling of hepatocellular carcinoma, STAD, renal cell carcinoma and COAD; ref. 22, for multi-region sampling of prostate cancer; and ref. 29, for multi-region sampling of MESO.
Analysis of NJ expression in multi-region samples was conducted immediately with our NJ prediction pipeline if the FASTQ file was available. If RNA-seq data were available only in BAM format, the sequencing file was first converted into FASTQ format utilizing the Picard software (version 2.7.7a). NJ prediction is detailed in the ‘Characterization of public NJs’ section of the Methods.
Data download
Bulk RNA-seq data for GBM (n = 167), LGG (n = 516), LUAD (n = 517), LUSC (n = 501), MESO (n = 516), LIHC (n = 371), STAD (n = 415), KIRC (n = 533), KIRP (n = 290), KICH (n = 66), COAD (n = 458) and PRAD (n = 497) samples were downloaded from TCGA in FASTQ format. Download of intratumoural multi-region sampling sequencing data is detailed in the previous section. Similarly, bulk RNA-seq data for 9,166 normal tissue samples in FASTQ format were downloaded from the GTEx repository. Bulk RNA-seq data for 66 patient-derived GBM cell lines were received from the Mayo Clinic Brain Tumor Patient-Derived Xenograft National Resource43. Proteomics data for 100 GBM samples were downloaded from the Clinical Proteomic Tumor Analysis Consortium44.
RNA-seq alignment
All downloaded RNA-seq datasets were individually aligned using a STAR aligner-based processing pipeline. Using the STAR software (version 2.7.7a), we constructed a genome index containing non-annotated junctions through the initial alignment pass of the input data. The complete set of command line parameters was as follows: –runThreadN 1 \ –outFilterMultimapScoreRange 1 \ –outFilterMultimapNmax 20 \ –outFilterMismatchNmax 10 \ –alignIntronMax 500000 \ –alignMatesGapMax 1000000 \ –sjdbScore 2 \ –alignSJDBoverhangMin 1 \ –genomeLoad NoSharedMemory \ –limitBAMsortRAM 80000000000 \ –readFilesCommand gunzip -c \ –outFilterMatchNminOverLread 0.33 \ –outFilterScoreMinOverLread 0.33 \ –sjdbOverhang 100 \ –outSAMstrandField intronMotif \ –outSAMattributes NH HI NM MD AS XS \ –limitSjdbInsertNsj 2000000 \ –outSAMunmapped None \ –outSAMtype BAM SortedByCoordinate \ –outSAMheaderHD @HD VN1.4 \ –twopassMode Basic \ –outSAMmultNmax 1 \ and aligned using the GRCH37 STAR index file.
TCGA sample selection and gene expression quantification
TCGA tumour samples with an absolute tumour purity greater than 0.60 were retained for downstream in silico analysis18,19. We selected non-mitochondrial, protein-coding transcripts defined by the Ensembl Homo sapiens GRCH37.87 gene annotation gene transfer format (GTF) file and utilized this curated list to select and retain protein-coding transcript isoforms in the TCGA RNA-seq data. Transcript-level expression data (log2[RSEM transcripts per million + 0.001]) for all TCGA samples were downloaded from the University of California, Santa Cruz Xena Toil pipeline and transformed into standard TPM values. Protein-coding transcript isoforms with a median TPM ≥ 10 were retained for downstream analysis. In the case of glioma TCGA cases, subsequent expression data in TPM were subset into six disease-type categories: all cases (n = 429), GBM cases (n = 115), LGG cases (n = 314), IDHwt cases (n = 166), IDHmut-A cases (n = 140) and IDHmut-O cases (n = 123). Protein-coding transcript isoforms with a median TPM ≥ 10 in at least one of the six disease types were retained for further analysis.
Characterization of public NJs
For public cancer-specific splicing event counting, we designed a custom R script that detected and quantified non-annotated, cancer-specific splicing events found across each corresponding patient cohort. From the output files derived from STAR aligner in the previous step, alternative splicing events were quantified in detected junction counts in the corresponding sj.out.tab file. We removed splicing events detected in the GRCh37.87 GTF sj.out.tab (GENCODE v33) file to define non-annotated splicing junctions. Non-annotated splicing junctions that overlap non-mitochondrial, protein-coding genes identified in the previous step were retained for continued analytical processing. We removed all splicing junctions with fewer than 10 of their target spliced reads (count) or fewer than 20 total spliced reads (depth) over the whole cohort. Similarly to previous studies14, we computed spliced frequency as the sum of the total number of target spliced reads divided by the collective sum of spliced reads from the target and canonical junctions. Splicing junctions with a read frequency greater than 1% were retained for downstream analyses. We defined public splicing junctions as ones that were putatively expressed with the aforementioned criteria of total read count, read depth and read frequency across at least 10% of the studied patient cohort and retained those for further analysis. To characterize cancer-specific splicing events, otherwise known as NJs, we removed all junctions that were putatively expressed with the same parameters in more than 1% of GTEx normal samples.
Detection of cancer-specific intron retention events
Intronic splicing events were detected and characterized using IRFinder v1.2.3. RNA-seq data from TCGA (GBM and LGG) and GTEx (central nervous system) aligned to GRCh37 (hg19) were imported into the software for the detection of intron retention events. General linear model-based analysis was used for differential intron retention assessment. The intron retention ratio is calculated as (intronic reads)/sum(intronic reads, normal spliced reads). Significant intron retention changes are defined as: no less than 10% in both directions; and adjusted P values less than 0.05. An intron retention event’s PSR in TCGA or GTEx is defined as the number of cases that fulfil these criteria divided by the total number of cases in the cohort. Putative cancer-specific intron retention NJs are characterized as intron retention events with a TCGA PSR ≥ 0.10 and a GTEx PSR < 0.01.
Transcriptomic validation of expressed NJs
Detection of expressed NJs in patient-derived GBM and LGG cell lines
RNA-seq data for cell lines derived from xenografts from patients with GBM were downloaded from the Mayo Clinic Brain Tumor Patient-Derived Xenograft National Resource. Patient-derived LGG cell lines were generated from surgically resected specimens in the Neurological Surgery Brain Tumor Center at the University of California, San Francisco (UCSF)41. RNA-seq data from GBM and LGG cell lines were aligned and processed as described above. Public NJs with splice junction CPM of >0 are considered detectable in cell line-derived RNA-seq data.
Detection of expressed NJs in multi-region cases
In our cohort of spatially mapped glioma cases, approximately ten or more maximally distanced anatomical biopsies were collected from each patient, allowing for intratumoural assessment of genetic heterogeneity through bulk RNA-seq and whole-exome sequencing. Multi-region sequencing data of various other cancer types vary in the number of sampled regions per tumour and are detailed in the corresponding references (Extended Data Fig. 2). RNA-seq data collected from each multi-region sample were processed and aligned as described above. We searched for putative NJs previously characterized from TCGA in each multi-region sampling dataset. Public NJs with CPM > 0 were considered detectable. Public NJs with putative expression (≥10 spliced reads) in two or more mapped samples in the same case are considered spatial-conserved NJs. NJs detected in all multi-region samples in the same tumour are considered tumour-wide NJs.
Proteomic validation of expressed NJ-derived peptides
From the putative NJs detected in the above pipeline, we generated a database of all plausible polypeptides derived from all NJs. NJ-encoding transcripts were generated by mapping the junction coordinates to an hg19 human genome assembly in the Ensembl annotation database (AH13964, EnsDb.Hsapiens.v75). Prediction of NJ-derived amino acid sequences was subsequently performed, and appropriately translated sequences (methionine starting residue, removal of sequences following first stop codon) were retained for downstream n-base-polypeptide iteration. To detect NJ-derived polypeptides in GBM cases, we analysed RAW files of GBM and LGG MS data housed in the Clinical Proteomic Tumor Analysis Consortium (n = 99), ref. 45 (n = 99), ref. 53 (n = 92) and ref. 54 (n = 84). MaxQuant (v1.6.17.0) was used to identify tryptic sequences from the corresponding MS datasets. Predicted NJ-derived peptides, decoy sequences and a human reference proteome (UniProt Proteome ID: UP000005640) were input as a FASTA file into MaxQuant, and tryptic sequences derived from the input file were matched against the publicly available MS databases. Cancer-specific peptides spanning NJ-derived protein sequences were considered MS-confirmed. The relative detection levels of the NJ-derived peptides and normal-tissue-derived peptides were evaluated by their log2[peak intensities]. Aside from the default settings, the following commands and parameters were modified and used for MS analysis in MaxQuant: Digestion mode = Trypsin/P; Max missed = 3; Minimum peptide length = 5; Minimum peptide length for unspecific search = 5.
Peptide processing and HLA binding and presentation predictions
Cancer-specific transcripts with associated NJs were translated in silico into their corresponding amino acid sequences. A library of all possible peptides of 8 to 11 amino acids in length was then generated, and cancer-specific sequences were selected by removing those detectable in normal-tissue peptide isoforms in a reference human proteome dataset (UniProt Proteome ID: UP000005640). All cancer-specific peptides with their upstream and downstream flanking sequences (maximum flanking length of 30 amino acids) were independently analysed and ranked by MHCflurry 2.0 and HLAthena MSiC. HLA-I binding affinity was assessed against HLA-A*01:01, HLA-A*02:01, HLA-A*03:01, HLA-A*11:01 and HLA-A*24:02 in both cases. In the HLAthena evaluation of antigen binding and presentation to the corresponding HLA haplotypes, peptides were assigned to alleles by rank with a threshold of 0.1. Contexts of up to 30 flanking amino acids on both amino and carboxy termini were utilized with aggregation by peptide and no log-transformed expression. Baseline MHCflurry 2.0 models with both peptide–HLA binding affinity predictor and antigen-processing predictor were used. Overall, peptide–HLA presentation scores were characterized by mhcflurry_presentation_score and MSiC_HLA scores in MHCflurry 2.0 and HLAthena, respectively. To select for high-binders, we curated lists of peptide–HLA complexes in the top 10 percentile of scores from both prediction algorithms.
Cell culture
Culture of cells derived from xenografts from patients with GBM
GBM, GBM34, GBM43, GBM108, GBM115, GBM118, GBM102, GBM137, GBM148, GBM164 and GBM195, were obtained from the Mayo Clinic Brain Tumor PDX national resource. Xenograft lines were cultured according to recommended conditions in previous literature55 and passaged a maximum of 20 times before restoration to earlier passages. Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin and streptomycin. Cell culture plates were treated overnight at 4 °C with DPBS (with calcium and magnesium) and 10% laminin (Gibco catalogue number 23017015) before use.
Primary patient-derived GBM and LGG cell culture
Primary patient-derived IDHwt GBM (SF7996), IDHmut-A (SF10602) and IDHmut-O (SF10417) cell lines were previously internally generated from dissociated glioma biopsies and cultured as previously described41. Cells were cultured in serum-free, glioma neural stem cell medium, which comprises Neurocult NS-A (STEMCELL Technologies catalogue no. 05751) supplemented with N-2 supplement (Invitrogen catalogue no. 17502048), B-27 supplement minus vitamin A (Invitrogen catalogue no. 12587010), 1% penicillin and streptomycin, 1% glutamine and 1% sodium pyruvate. Before immediate use in culture, glioma neural stem medium was supplemented with 20 ng ml−1 EGF (Peprotech catalogue no. AF-100-15), bFGF (Peprotech catalogue no. AF-100-18B) and PDGF-AA (Peprotech catalogue no. AF-100-13A). As for cell lines derived from xenografts from patients with GBM, cell culture plates were incubated overnight at 4 °C with DPBS (with calcium and magnesium) and 10% laminin (Gibco catalogue no. 23017015) before use.
Jurkat76 cell culture
Jurkat76 cells were used as the TCR α- and β-negative human T cell derivative that allowed for non-competing introduction of exogenous TCRs. CD8+ Jurkat76 cells were cultured in RPMI supplemented with 10% fetal bovine serum and 1% penicillin and streptomycin.
T2 cell culture
T2 cells were used in the study to monitor immune cell response to the exogenous antigen of interest in a non-competitive environment. T2 cells are deficient in a peptide transporter involved in antigen processing (TAP), and as such, induction of these cells with exogenously administered peptides allows for their association and presentation by HLA molecules, HLA-A*02:01 in particular. We cultured T2 cells in IMDM medium supplemented with 20% FBS.
COS-7 cell culture
We opted to use COS-7 (ATCC catalogue no. CRL-1651) cell lines as our respective primate and human artificial APC models52. These cell lines do not express HLA molecules, which allows for the introduction of the HLA allele of interest. COS-7 cells were cultured in DMEM medium supplemented with 10% FBS and 1% penicillin and streptomycin.
THP-1 cell culture
THP-1 cells (ATCC catalogue no. TIB-202) were used to investigate immune reactivity against neoantigen presentation by dendritic cells (DCs). THP-1 cells were cultured in RPMI-1640 supplemented with 10% FBS. All cell lines have been tested for forms of mycoplasma contamination. Cell lines were obtained from trusted sources and have not been authenticated.
siRNA-mediated knockdowns of splicing-related genes
Cells were seeded in 2 ml of antibiotic-free medium in a 6-well plate at the following densities: GBM115, 45,000 cells per well; SF10417, 100,000 cells per well; and SF10602, 100,000 cells per well. At 24 h post-seeding, cells were transfected by adding 400 μl reaction containing serum-free medium, 2.0 μl DharmaFECT 1 reagent (Horizon, no. T-2001-02), and their respective siRNA pools (four-siRNA equimolar mix) at a final concentration of 30 nM. At 24 h post-transfection, the medium was changed to complete medium. At 72 h post-transfection, RNAs were isolated and purified using the Zymo Quick-RNA microprep kit (Zymo Research, no. R1058).
CRISPRi
sgRNAs were designed using the Broad CRISPick webportal56,57. Top-ranked sgRNAs were ordered from IDT: top strands were appended with ‘CACCG’ on their 5′ end, bottom strands were appended with ‘AAAC’ on their 5′ end and ‘C’ on their 3′ end. The oligonucleotide names and sequences ordered from IDT are: sgSF3A3_CRISPRi_1_TopStrand, 5′-CACCGGAATTGAGAAGCCGCGACTA-3′; sgSF3A3_CRISPRi_1_BottomStrand, 5′-AAACTAGTCGCGGCTTCTCAATTCC-3′; sgSF3A3_CRISPRi_2_TopStrand, 5′-CACCGAAGCCGCGACTAAGGGAAGA-3′; sgSF3A3_CRISPRi_2_BottomStrand, 5′-AAACTCTTCCCTTAGTCGCGGCTTC-3′; sgSF3A3_CRISPRi_3_TopStrand, 5′-CACCGAGGGAAGATGGAGACAATAC-3′; sgSF3A3_CRISPRi_3_BottomStrand, 5′-AAACGTATTGTCTCCATCTTCCCTC-3′; sgSF3A3_CRISPRi_4_TopStrand, 5′-CACCGATTCAGACCACCAACACGGC-3′; sgSF3A3_CRISPRi_4_BottomStrand, 5′-AAACGCCGTGTTGGTGGTCTGAATC-3′; sgCELF2_CRISPRi_1_TopStrand, 5′-CACCGTCCCCTCCGAAATCCAGCGC-3′; sgCELF2_CRISPRi_1_BottomStrand, 5′-AAACGCGCTGGATTTCGGAGGGGAC-3′; sgCELF2_CRISPRi_2_TopStrand, 5′-CACCGGCCCCGGCGCTGGATTTCGG-3′; sgCELF2_CRISPRi_2_BottomStrand, 5′-AAACCCGAAATCCAGCGCCGGGGCC-3′; sgSNRPD2_CRISPRi_1_TopStrand, 5′-CACCGAGCGTAGTGACCATCATGTG-3′; sgSNRPD2_CRISPRi_1_BottomStrand, 5′-AAACCACATGATGGTCACTACGCTC-3′; sgSNRPD2_CRISPRi_2_TopStrand, 5′-CACCGCCTAGCCCGGCCTCACATGA-3′; sgSNRPD2_CRISPRi_2_BottomStrand, 5′-AAACTCATGTGAGGCCGGGCTAGGC-3′; sgROSA26_CRISPRi_TopStrand, 5′-CACCGACAGCAAGTTGTCTAACCCG-3′; sgROSA26_CRISPRi_BottomStrand, 5′-AAACCGGGTTAGACAACTTGCTGTC-3′; sgAAVS1_CRISPRi_TopStrand, 5′-CACCGGGGCCACTAGGGACAGGAT-3′; sgAAVS1_CRISPRi_BottomStrand, 5′-AAACATCCTGTCCCTAGTGGCCCC-3′.
sgROSA26 (ref. 58) and sgAAVS1 (ref. 59) were from previous literature. Top and bottom strands of each sgRNA were then annealed and ligated into the CRISPRi vector pLV hU6-sgRNA hUbC-dCas9-KRAB-T2a-Puro (Addgene plasmid no. 71236)60. Lentivirus was produced as described in the section of the Methods entitled Lentiviral transduction. For transduction, SF10417 was plated at 20,000 cells per well in 24-well plates, and GBM115 cells were plated at 60,000 cells per well in 6-well plates. At 24 h post-seeding, cells were transduced by addition of virus in complete medium supplemented with 4 μg ml−1 Polybrene. At 24 h post-transduction, medium was replaced with complete medium with 1 μg ml−1 puromycin, and cells were selected for 72 h and then allowed to recover in complete medium. Each sgRNA was assessed by three separate transductions.
Quantitative PCR with reverse transcription
A 1,000 ng quantity of DNAse-treated RNA was converted to cDNA using the iScript cDNA synthesis kit (BioRad, no. 1708891). This cDNA was then diluted 1:3 using ultrapure, nuclease-free water, and 2 μl was used per quantitative PCR (qPCR) reaction. qPCR with reverse transcription was performed using the Applied Biosystems POWER SYBR Green Master Mix (Applied Biosystems, no. 4367659). All samples were run in biological triplicates, with technical triplicates for each biological triplicate using the Quantstudio 5 (Thermo Scientific), and all gene expression data were normalized to the housekeeping gene GUSB. The cycling protocol was as follows: 2 min at 50 °C, 10 min at 95 °C, followed by 40 cycles at 95 °C for 15 s, and 60 °C for 60 s. Dissociation curves were plotted to confirm specific product amplification. Primer sequences corresponding to each gene for the mRNA expression analysis were designed using NCBI Primer.
Amplicon sequencing for validation of NJ expression
RNAs from respective cell lines were isolated and purified using the Zymo Quick-RNA microprep kit (Zymo Research, no. R1058). A 1,000 ng quantity of DNAse-treated RNA was converted to cDNA using the iScript cDNA synthesis kit (BioRad, no. 1708891). This cDNA was then diluted 1:3 using ultrapure, nuclease-free water, and 2 μl was used per PCR reaction. Sixteen reactions were carried out per amplicon per cell line using Q5 High-Fidelity 2× master mix (NEB, no. M0492L) with primers containing partial Illumina adaptors. Reaction mixtures were set up according to the manufacturer’s guidelines. These products were then purified by separation on a 1.0% agarose gel at 100 V (constant) for 1 h and were then purified using the Monarch DNA gel extraction kit (NEB, no. T1020L). Purified products were quantified with a qubit high-sensitivity dsDNA kit (Invitrogen, no. Q32851) and prepared and submitted according to Azenta (Genewiz) guidelines for amplicon sequencing.
IVS of healthy-donor PBMCs
HLA-A*02:01:01+ PBMCs were purchased from StemExpress in either fresh or cryopreserved format. Approximately 1 × 109 fresh PBMCs (StemExpress catalogue no. LE001F) were immediately proportioned into aliquots of 3 × 108 cells and cryopreserved in liquid nitrogen, with one aliquot actively used for downstream IVS. Cryopreserved PBMCs (StemExpress catalogue no. PBMNC300C) totalling approximately 3 × 108 cells per cryovial were used in one vial per IVS procedure. PBMCs were thawed with 1:1,000 Benzonase/RPMI (Sigma Aldrich catalogue no. E8263). The CD14+ population was isolated from the PBMCs using CD14+ Miltenyi microbeads (Miltenyi Biotec catalogue no. 130-050-201) as per the manufacturer’s instructions. The CD14− flowthrough was cryopreserved for 6 days before naive CD8+ T cell isolation. Isolated CD14+ cells were cultured in CellGenix GMP DC medium (CellGenix catalogue no. 20801-0500) supplemented with 1% human serum (Sigma Aldrich catalogue no. H6914), 1% penicillin and streptomycin, 1,000 U ml−1 recombinant human IL-4 (Peprotech catalogue no. 200-04) and GM-CSF (Peprotech catalogue no. 300-03) in non-treated 24-well plates at a seeding density of 5 × 105 cells per well. On day 3, recombinant human IL-4 and GM-CSF (1,000 U ml−1 each) were added to the DC culture. On day 5, the DC culture was matured with 250 ng ml−1 LPS (Sigma Aldrich catalogue no. L6529) in addition to supplementation of recombinant human IL-4 and GM-CSF (1,000 U ml−1 each). Naive CD8+ T cells were isolated from the thawed CD14− population on day 6 using the EasySep Human Naive CD8+ T Cell Isolation Kit (STEMCELL Technologies catalogue no. 19258) as per the manufacturer’s instructions. Isolated naive CD8+ T cells were cultured in X-Vivo 15 medium (Lonza catalogue no. 04-418Q) supplemented with 5% human serum, 1% penicillin and streptomycin and 10 ng ml−1 of recombinant human IL-7 (Peprotech catalogue no. 200-07) in 48-well plates at a seeding density of 5 × 105 cells per well. On day 8, adherent matured DCs were collected from the plate using cold PBS. The collected DCs (1 × 106 cells ml−1) were exogenously pulsed with 1 μM of the neoantigen peptide, influenza peptide or no peptide for 1 h at 37 °C. The peptide-pulsed or non-pulsed DCs were then co-cultured with naive CD8+ T cells at an optimal DC/T cell ratio of 1:4 in 48-well plates. The co-culture was maintained with X-Vivo 15 medium supplemented with 10 ng ml−1 of recombinant human IL-7, 10 ng ml−1 recombinant human IL-15 (Peprotech catalogue no. 200-15) and 60 ng ml−1 of recombinant human IL-21 (Peprotech catalogue no. 200-21) for 10 days with IL-7 and IL-15 restimulation every 2 days. Cells were reseeded into subsequent 24-well, 12-well and 6-well plates on the basis of confluency. This concluded the first cycle of IVS of the neoantigens and influenza peptides. On days 19 and 29, sensitized CD8+ T cells were reintroduced to a second and third round of stimulation with newly pulsed DCs, and the co-culture was maintained for 10 additional days until the end of the second and third cycle of IVS. Cytokine assays were performed at the end of the second and third cycles of IVS to determine whether a peptide-reactive T cell population has expanded.
Mutation-specific ELISA screen
Aliquots containing CD8+ T cells from individual parent IVS wells were collected and split equally into 96-well plate daughter wells containing 1 × 105 cells per well. Daughter wells in triplicate were stimulated with T2 cells pulsed with the neoantigen peptide of interest, control peptide, no peptide or no T2 cells at all for 16 h at an effector-to-target (E/T) ratio of 1:1. T2 cells were pulsed with 1 pM to 1 μM of the neoantigen peptide of interest, control peptide or no peptides for 1 h at 37 °C. Influenza-reactive T cells were co-cultured against influenza peptide-pulsed T2 cells as a positive control. Co-culture supernatant was collected and diluted for use in IFNγ (BD Biosciences catalogue no. 555142) and TNF (BD Biosciences catalogue no. 555212) ELISAs as per the manufacturer’s instructions. ELISA readouts were performed on the Epoch Microplate Spectrophotometer (BioTek Instruments) using the BioTek Gen5 Data Analysis software (version 1.11). Wells with significantly increased expression levels of IFNγ and TNF were selected for downstream single-cell immune profiling using single-cell RNA and V(D)J sequencing.
Single-cell immune profiling
Once an expanded neoantigen-reactive CD8+ T cell population from IVS was identified, single-cell RNA and V(D)J sequencing were performed using the 10x Genomics platform. Before sequencing, CD8+ T cells from the expanded neoantigen-reactive (ELISA screen-positive) wells were collected and co-cultured with T2 cells pulsed with 1 μM of the neoantigen peptide of interest, a control peptide or no peptides at an E/T ratio of 1:1. One co-culture replicate was performed for 3 h for single-cell RNA-seq analysis, and another was performed for 16 h for IFNγ and TNF ELISA confirmation. The final cell concentration was adjusted to approximately 1 × 104 cells per microlitre with an initial cell viability of at least 90% to maximize the likelihood of achieving the desired cell recovery target. Independent CD8+ T cell and non-pulsed T2 single cultures were sequenced alongside the co-culture conditions for differentiating cell types in the downstream single-cell sequencing analysis. The Chromium Next GEM Single Cell 5′ Reagent Kit v2 (Dual Index) (10x Genomics, catalogue no. CG000331) was used for preparation for single-cell sequencing analysis. Gel beads in emulsions (GEMs) were generated by combining the single-cell 5′ gel beads, partitioning oil and the master mix containing the cells onto the Chromium Next GEM Chip K. Cell lysis and barcoded reverse transcription of RNAs in all single cells were finished inside their corresponding GEM. Barcoded cDNA product was recovered through post-GEM-RT cleanup and PCR amplification. cDNA quality control and quantification were performed on the Fragment Analyzer System (Agilent Technologies). A 50 ng quantity of cDNA was used for the construction of the 5′ gene expression library, and each sample was indexed by a Chromium i7 Sample Index Kit. This process was performed on an Illumina NovaSeq 6000 sequencer at the UCSF Institute of Human Genetics (IHG) with a minimum of 20,000 read pairs per cell for the 5′ Gene Expression library. The enriched product was measured by the Fragment Analyzer System. A 50 ng quantity of enrichment TCR product was used for library construction. Single-cell V(D)J-enriched libraries were subsequently sequenced on the Illumina NovaSeq 6000 with a minimum of 5,000 read pairs per cell for the V(D)J library. Cell Ranger 7.0.0 (10x Genomics Cloud Analysis) was used to pre-process raw single-cell RNA-seq and identify V(D)J clonotypes. The annotation files vdj_GRCh38_alts_ensembl-3.1.0-3.1.0 and GRCh38-3.0.0 were used for demultiplexing cellular barcodes, performing read alignments and generating feature–barcode matrices. Only cells for which clonotype information was available were retained for downstream analysis. Single-cell gene expression and corresponding V(D)J sequences of candidate T cell clonotypes were analysed on the Loupe V(D)J browser. Single cells with detectable CD8A expression were specifically isolated and characterized as the CD8+ T cell population and subsequently grouped according to their TCR clonotypes. To identify T cell clonotypes associated with a neoantigen-specific response, we selected expanded TCR clonotypes with significantly increased levels of IFNG, TNF and GZMB expression in the T cell:neoantigen-pulsed T2 condition compared to the T cell:control-pulsed T2 and T cell:non-pulsed T2 conditions.
HLA typing
OptiType 1.3.1 was used for genotyping HLA alleles from available whole-exome sequencing data available for glioma cell lines with default parameters.
Plasmids and peptides
HLA-A*02:01 and NEJ-derived gene sequences were all synthesized and cloned into the pTwist Lenti SFFV Puro WPRE vector (Twist Biosciences). Constructs encoding full-length and truncated multi-base-polypeptide versions of the wild-type and mutant GNAS and RPL22 sequences were generated. TCR α and β was synthesized and cloned into the pTwist Lenti SFFV vector (Twist Biosciences). HPLC-grade NEJ-derived neoantigen peptides (>95%) were manufactured by TC Laboratories.
Lentiviral transduction
HEK293T cells were plated in 6-well culture plates at a density of 1 × 106 cells per well with 2 ml DMEM supplemented with 10% FBS without antibiotics. After approximately 18 to 24 h or at 90% confluency, HEK293T cells were transfected with the expression construct, see above, and lentiviral packaging plasmids, pMD2.G (Addgene, no. 12259) and psPAX2 (Addgene, catalogue no. 12260).
TCR α/β transduction
A 1.0 μg quantity of TCR α/β transfer plasmid, 0.75 μg psPAX2 and 0.25 μg pMD2.G were combined with 200 μl Opti-MEM (Thermo Fischer Scientific catalogue no. 31985062). A 6 μl volume of Xtremegene HP was added to this mixture, and complex formation was allowed to occur for 15 min at room temperature, at which point this reaction mixture was added to the corresponding HEK293T cells. Transfection medium was replaced with fresh DMEM after 24 h. Viral supernatant was collected after 48 h, and the functional virus titre was measured on 6-well plates seeded with Jurkat76/CD8 cells or PBMC-derived CD8+ T cells at 60–70% confluency. Viral transduction was performed with threefold serial dilutions of the virus stock supplemented with Polybrene at a final concentration of 4 μg ml−1. Medium was changed 24 h following viral transduction. Cells were assessed for transduction efficiency after 3–4 days by measuring surface expression of TCR α/β and CD3 by fluorescence-activated cell sorting (FACS) analysis. Cells demonstrating a high level of double-positive expression of TCR α/β and CD3 were flow-sorted and maintained for downstream co-culture and reactivity assays.
HLA and neoantigen transduction
Constructs expressing HLA-A*02:01 were linearized and restricted with BamHI and XhoI (New England Biolabs) and purified using the Zymoclean Gel DNA Recovery Kit (Zymo Research catalogue no. D4007). The HLA-A*0201 sequence was then ligated into a lentiviral construct downstream of an EF1A-core promoter and upstream of an IRES followed by a blasticidin resistance gene. A 1.0 μg quantity of either HLA-A*02:01 or neoantigen transfer plasmid, 0.75 μg psPAX2 and 0.25 μg pMD2.G were combined with 200 μl Opti-MEM (Thermo Fischer Scientific catalogue no. 31985062). A 6 μl volume of Xtremegene HP was added to this mixture, and complex formation was allowed to occur for 15 min at room temperature, at which point this reaction mixture was added to corresponding HEK293T cells. As stated above, neoantigen constructs encode either the full-length or truncated version of the NJ-derived peptide. The transfection medium was replaced with fresh DMEM medium after 24 h. HLA-A*02:01 lentiviral transduction and screening were performed first before neoantigen lentiviral transduction and screening for streamlined drug selection. Viral supernatant was collected after a subsequent 48 h, and the functional virus titre was measured on 6-well plates seeded with COS-7 cells at 60–70% confluency. Viral transduction was performed with threefold serial dilutions of the virus stock supplemented with 4 μg ml−1 Polybrene. Medium was changed 24 h following viral transduction and replaced with complete medium supplemented with blasticidin. Cells were assessed for transduction efficiency after 3–4 days by drug screening. HLA-A*02:01-transduced APCs were cultured in medium treated with 10 μg ml−1 blasticidin for approximately 7 days before assessing for cell viability across titres. Neoantigen-lentiviral transduction was subsequently performed, and APCs transduced with both HLA-A*02:01 and neoantigen-expressing constructs were then cultured in medium treated with 3 μg ml−1 puromycin for approximately 7 days. Cell viability was assessed afterwards across all titre conditions. Cells were assessed for transduction efficiency after 3–4 days by measuring surface expression of HLA-A2 FACS analysis.
Dose-dependent assessment of TCR reactivity against neoantigen
Specificity of neoantigen-reactive CD8+ T cells and TCR-transduced T cells was assessed by human IFNγ (BD Biosciences catalogue no. 555142), IL-2 (BD Biosciences catalogue no. 555190) and TNF (BD Biosciences catalogue no. 555212) ELISA. Assessment of TCR recognition against exogenously introduced neoantigen peptides presented by HLA molecules was conducted by co-culturing T cells with peptide-pulsed T2 cell conditions. T2 cells were pulsed with neoantigen peptide of interest at a concentration between 1 pM and 1 μM, decoy peptide or no peptides for 1 h at 37 °C. Influenza-reactive T cells were co-cultured against influenza peptide-pulsed T2 cells as a positive control. T cells and T2 cells were co-cultured in a 96-well round-bottom plate at a concentration of 1 × 105 of each cell type in 200 μl of medium for 16 h. Supernatant was collected and diluted for cytokine release assays per the manufacturer’s instructions. ELISA assay readouts were performed on an Epoch Microplate Spectrophotometer (input wavelength 450 nm and output wavelength 570 nm) using the BioTek Gen5 Data Analysis software. To characterize the dose-dependent activation of the TCRs in transduced triple-reporter Jurkat76/CD8 cells, we performed flow analysis to assess the level of expression of NFAT–GFP, NF-κB–CFP and AP-1–mCherry following 16 h of co-culture. Similarly, the reactivity of TCR-transduced PBMC-derived CD8+ T cells was evaluated by flow analysis following anti-CD107a (BioLegend, catalogue no. 328620) and anti-CD137 antibody (4-1BB; BioLegend catalogue no. 309804) staining.
In vitro transcription synthesis of mRNA
All constructs were subcloned into pcDNA3.1 (Invitrogen, 2520855) and linearized by XhoI restriction enzyme with the plasmid DNA template transcribed downstream from the bacteriophage T7 promoter sequence. For long (>0.5 kilobase (kb)) and short (<0.5 kb) transcripts, 1 μg and 0.5 μg of template were used, respectively. Reactions were assembled at room temperature using the mMESSAGE mMACHINE T7 Transcription Kit as per the manufacturer’s instructions (Invitrogen, 2582905) and incubated at 37 °C for 1 h for long transcripts and 16 h for short transcripts. Following DNase treatment, a poly(A) tailing reaction was performed for 1 h according to the HiScribe T7 ARCA manual (NEB, E2060S). Subsequently, the synthesized mRNA was purified by LiCl precipitation using 70% DEPC-based ethanol. Synthesized mRNA was heat-shocked (70 °C, 5 min) with the formaldehyde loading dye to verify quality through gel electrophoresis.
mRNA transfection of HLA-A*02:01, truncated neoantigen and full-length NEJ-encoding mRNA
Transfection of in vitro transcription-synthesized mRNA into COS-7 cells was performed with electroporation using the Neon Transfection System 100 μl Kit (Invitrogen, MPK10096) as per the manufacturer’s instructions. A total of 1 × 106 COS-7 cells were washed and resuspended with 100 μl of Neon Resuspension Buffer. A 5 μg quantity of HLA-A2 and 5 μg of candidate (either the truncated neoantigen sequence or the full-length NEJ sequence) mRNA were added into the cell solution. Electroporation was performed on the Neon NxT Electroporation System (Invitrogen, NEON1). Electroporation of COS-7 cells was performed with the following optimized conditions: pulse voltage of 1,200 V, width of 30 ms and 2 pulses. Transfected cells were immediately transferred into warm RPMI with no antibiotics. Aliquots of transfected cells were retained for validation of HLA-A2 expression by staining with HLA-A2 monoclonal antibody (BB7.2, Thermo Scientific, 17-9876-42) and subsequent flow cytometry analysis.
Evaluation of TCR specificity against endogenously processed and HLA-presented neoantigen
Characterization of neoantigens that are endogenously processed and presented by surface HLA was conducted by co-culturing HLA-A*02:01/neoantigen-transfected COS-7 cells with TCR-transduced T cells. Similarly, T cells and COS-7 cells were co-cultured in a 96-well flat-bottom plate at a concentration of 1 × 105 of each cell type in 200 μl of medium for 16 h. Supernatant was collected and diluted for cytokine release assays as per the manufacturer’s instructions, and cytokine release levels were assessed with the Epoch Microplate Spectrophotometer and BioTek Gen5 Data Analysis software. In all cytokine release assay experiments, maximum cellular cytokine release per well was determined by the addition of 0.2 μl Cell Activation Cocktail (without brefeldin A) (BioLegend catalogue no. 423302) per 100 μl cell solution. Evaluation of endogenously processed and presented neoantigens in glioma cell lines was performed by co-culturing TCR-transduced triple-reporter Jurkat76 cells with glioma cells at a 1:1 E/T ratio (1 × 105 per well in a 96-well plate). Flow analysis was performed to assess the level of expression of NFAT–GFP, NF-κB–CFP and AP-1–mCherry following 16 h of co-culture.
HLA immunoprecipitation and liquid chromatography with tandem MS
COS-7 cells were co-electroporated with 10 μg of each mRNA encoding the HLA-A*02:01 allele and the full-length coding sequence of the mutated GNAS or RPL22 using the Neon Transfection system (100-μl tip, setting: 1,050 V, 10 ms and 2 pulses). A total of 20 × 106 cells were electroporated per condition and plated in 6-well non-TC plates overnight. For the GMB115 cell line sample, approximately 100 × 106 cells were used. Cells were collected by incubating with 1 mM EDTA (Millipore Sigma) for 10 min at 37 °C. For the immunoprecipitation experiments, cells were lysed in 8 ml of 1% CHAPS (Millipore Sigma) for 1 h at 4 °C; the lysates were then spun down for 1 h at 20,000g and 4 °C, and supernatant was collected. For the affinity-column-based immunopurification of HLA-I ligands, 40 mg of cyanogen bromide-activated Sepharose 4B (MilliporeSigma) was activated with 1 mM hydrochloric acid (MilliporeSigma) for 30 min. Subsequently, 1 mg of W6/32 antibody (Bio X Cell) was coupled to Sepharose in the presence of binding buffer (150 mM sodium chloride, 50 mM sodium bicarbonate, pH 8.3; sodium chloride) for 2 h at room temperature. Sepharose was blocked for 1 h with glycine and washed three times with PBS. Supernatants of cell lysates were run through an affinity column using peristaltic pumps at 6 ml min−1 flow rate overnight at 4 °C. HLA complexes and binding peptides were eluted from the column five times using 1% TFA. Peptides and HLA-I complexes were separated using C18 columns (Sep-Pak C18 1 cc Vac Cartridge, 50 mg of sorbent per cartridge, 37–55-μm particle size, Waters). C18 columns were pre-conditioned with 80% ACN (Millipore Sigma) in 0.1% TFA and equilibrated with two washes of 0.1% TFA. Samples were loaded, washed twice with 0.1% TFA and eluted in 300 μl of 30%, 40% and 50% acetonitrile in 0.1% TFA. All three fractions were pooled, dried down using vacuum centrifugation and stored at −80 °C until further processing. HLA-I ligands were isolated by solid-phase extractions using in-house C18 mini-columns. Samples were analysed by high-resolution, high-accuracy liquid chromatography with tandem MS (Lumos Fusion, Thermo Fisher Scientific). COS-7 samples were run in DDA mode, and GMB115 samples were run in DIA mode. MS and tandem MS were operated at resolutions of 60,000 and 30,000, respectively. Only charge states 1, 2 and 3 were allowed. The isolation window was chosen as 1.6 Th, and collision energy was set at 30%. For tandem MS, maximum injection time was 100 ms with an automatic gain control of 50,000. MS data were processed using FragPipe. Protein false discovery rate was set at 1%. Oxidization of methionine, phosphorylation of serine, threonine and tyrosine, and N-terminal acetylation were set as variable modifications for all samples. Samples were searched against a database comprising UniProt Cercopithecus aethiops or UniProt human-reviewed proteins supplemented with the human HLA-A*02:01 allele sequence, mutRPL22 and mutGNAS, as well as common contaminants.
Characterization of CD8+ T cell-mediated anti-tumour reactivity
To determine whether TCR-transduced T cells were capable of mounting an anti-tumour response, TCR-transduced Jurkat76/CD8 or PBMC-derived CD8+ T cells were co-cultured with patient-derived GBM or LGG cell lines. CD8+ T cells were isolated from healthy-donor-derived PBMCs using the EasySep Human CD8+ T Cell Isolation Kit (STEMCELL Technologies, catalogue no. 17953). CD8+ T cells were then activated with Dynabeads Human T-Activator CD3/CD28 for T Cell Expansion and Activation (Thermo Scientific, catalogue no. 11161D) at a concentration of 25 μl per 1 × 106 cells. CD8+ T cells were cultured for 7 days with IL-7 (30 μl per 1 × 106 cells) supplemented every 2 days. CD8+ T cells were then lentivirus-transduced with neoantigen-specific TCRs with a hybridized mouse TCR constant region using the above transduction procedure. This additional step removes the likelihood of TCR α-chain and β-chain mispairing and allows us to evaluate TCR-transduction efficiency by staining with anti-mouse TCR constant region antibody (clone H57-597; BioLegend catalogue no. 109208). Flow sorting was performed to isolate highly transduced CD8+ T cells by selecting for cells stained strongly with anti-CD3 and anti-mouse TCR constant region antibody. Sorted transduced CD8+ T cells were expanded for 7 days before use in co-culture assays. Killing assays were performed using an xCELLigence RTCA S16 Real-Time Cell Analyzer. Tumour cells were cultured in medium pre-treated with 100 ng ml−1 IFNγ (Peprotech, catalogue no. 300-02) for 48 h and washed twice with PBS before seeding. A total of 1 × 104 tumour cells were plated per well in a 96-well E-plate (Agilent), and impedance was read for 16 h during incubation. TCR-transduced CD8+ T cells were introduced to each well at an E/T ratio of either 1:1 or 2:1, and tumour-specific killing was measured by changes in cell index over 24–48 h.
Identification of HLA-restricted CD8+ T cell-mediated reactivity against neoantigens
Evaluation of HLA-restricted T cell reactivity was performed by perturbing TCR and HLA–peptide interactions with the introduction of anti-HLA antibodies. In dose-dependent reactivity assays, T2 cells at a concentration of 1 × 105 tumour cells per well of a 96-well plate were washed twice with PBS and incubated for 30 min with blocking anti-HLA antibody (50 μg per well; clone W6/32, Bio X Cell, catalogue no. BE0079) or isotype control (50 μg per well; Bio X Cell, catalogue no. BE0085) at a total volume of 100 μl. Without any additional washes, T cells were added to achieve a final volume of 200 μl. In tumour-killing assays, tumour cells were added to each well of a 96-well E-plate in a total volume of 50 μl for initial seeding. Anti-HLA antibody or isotype control (50 μg per well) was added to each well 30 min before the addition of T cells to reach a total volume of 100 μl. T cells were added to each well to achieve a final volume of 200 μl, and impedance was measured for the following 24–48 h.
Immune monitoring of patients with cancer expressing mutGNAS-NEJ
PBMCs obtained from HLA-A*02:01+ patients with cancers expressing GNAS NEJs were tested for the presence of mutGNAS-specific CD8+ T cells by FACS using dual-colour HLA-A*02:01 dextramers loaded with the MS-identified mutGNAS peptide. Patient CD8+ T cells were stimulated in vitro with NEJ-expressing HLA-A*02:01-matched monocyte-derived DCs (moDCs) for 2 weeks before the FACS staining. To generate moDCs, HLA-A*02:01 healthy-donor PBMCs were plated in tissue culture flasks at 1 × 106 cells per square centimetre in complete medium without cytokines for 2 h at 37 °C to separate the adherent (monocyte-containing) and non-adherent (T cell-containing) fractions. The adherent fraction was washed with PBS, and fresh human A/B serum-containing medium supplemented with recombinant human IL-4 and GM-CSF (400 IU ml−1) was provided every 3 days. On day 6, moDCs were matured with LPS (Invitrogen) and IFNγ (Miltenyi Biotec) for 24 h before transfection. moDCs were electroporated with 100 μg ml−1 of mRNA encoding full-length mutGNAS using the Neon Transfection system (10-μl tip, setting: 1,325 V, 10 ms and 3 pulses). A bulk population of patient CD8+ cells were enriched from PBMCs by negative selection (STEMCELL Technologies) and co-cultured with mutGNAS-NEJ-expressing HLA-A*02:01-matched moDCs at a 2:1 ratio in non-tissue-treated 24-well plates (FALCON) in the presence of 300 IU ml−1 IL-2 and 50 ng ml−1 of IL-7, IL-15 and IL-21. Cytokines were replenished every 3 days. As a control, similarly isolated and co-cultured HLA-A*02:01-matched CD8+ cells from a healthy donor were used. For dextramer labelling, HLA-A*02:01 multimers bound to mutGNAS and conjugated to PE or APC were purchased from Immudex. As a specificity control, HLA-A*02:01 multimers bound to the nine-amino acid polypeptide from P53(R175H) (HMTEVVRHC) were used. Cells were labelled with dual-fluorophore-conjugated dextramers for 15 min at room temperature, followed by surface antibodies against CD3–BV785, CD4–BV421 and CD8–BV650 (BioLegend) for an additional 15 min at 4 °C. Cells were washed twice, stained with the viability dye 7-AAD (BioLegend) and acquired on a BD Fortessa X20 flow cytometer.
FACS analysis and antibodies
TCR-transduced cell lines were stained with anti-human TCR α/β (clone IP26, BioLegend catalogue no. 306717) and anti-human CD3 antibody (clone HIT3a, BioLegend catalogue no. 300307) to assess the surface-level expression of the transduced TCR. CD8+ T cells were stained with anti-CD107a (BioLegend, catalogue no. 328620) and anti-CD137 antibody (4-1BB; BioLegend catalogue no. 309804) to assess CD8+ T cell degranulation and TCR activation, respectively. The viability of cells was assessed with the Zombie Green Fixable Viability Kit (BioLegend, catalogue no. 423111) APCs and patient-derived glioma cell lines were stained with HLA-A2 monoclonal antibody (clone BB7.2, Thermo Fisher Scientific catalogue no. 17-9876-42). Approximately 1 × 106 cells per 100 μl FACS buffer (PBS supplemented with 1% BSA (Sigma Aldrich catalogue no. L6529)) were incubated with one test volume of antibody for 20 min as indicated by the manufacturer. Stained cells were washed once with FACS buffer before resuspension to a concentration of 4 × 105 cells per 100 μl FACS buffer. Cells were then analysed with the Attune NxT flow cytometer (Thermo Fischer Scientific). Unless otherwise stated, the concentration of antibody used was the one recommended by the manufacturer.
Gene set enrichment analysis
Differential gene expression of TCGA, GTEx and UCSF GBM and LGG RNA-seq was performed and quantified using DESeq2 (ref. 61). Only genes with an absolute fold change of >1.5 and a Benjamini–Hochberg-adjusted P value < 0.05 called by DESeq2 were considered to be differentially expressed62. Pre-ranked gene set enrichment analysis (GSEA) was carried out by ranking genes with the product of their fold-change sign and the −log10[adjusted P value].
Disease subtype-specific differential gene analysis
GSEA comparison was performed between IDH mutation subtypes (IDHwt and IDHmut) as well as glioma disease subtypes (IDHwt, IDHmut-A and IDHmut-O). Splicing-related gene sets were selected on the basis of keyword search, and gene sets with an adjusted P value of <0.05 when comparing two groups are considered differentially enriched. Unbiased hierarchical clustering of differentially enriched gene sets allows the characterization of subgroup-specific upregulated genes.
NJ-load-specific differential gene analysis
TCGA LGG and GBM samples were ranked according to the total putative NJs expressed per sample. High (NJHI) and low (NJLO) NJ load samples in each disease subtype were characterized as the upper and lower 0.10 percentile of ranked samples, respectively. GSEA was carried out between the NJHI and NJLO samples of each disease subgroup. Gene sets with a unidirectional fold-change and adjusted P value of <0.05 were considered to be enriched gene sets associated with NJ load. Splicing-related gene sets were selected on the basis of keyword searches. Leading-edge genes shared across all disease subgroups in the same gene set are defined as enriched genes associated with NJ load.
NJ and splicing-related gene correlation analysis
Selection of IDHmut upregulated genes was determined by splicing-related genes expressed with a significant (P < 0.05) log2[fold increase] of 1.5 in IDHmut cases when compared to their wild-type counterpart. Selection of splicing genes affected by oligodendroglioma-specific loss of chromosomes 1p and 19q was determined by chromosome 1p and 19q splicing-related genes expressed with a significant (P < 0.05) log2[fold decrease] of 1.5 in IDHmut-O cases compared to both IDHmut-A and IDHwt cases. Splicing-related genes that were selected for in vitro validation were chosen on the basis of previously reported confirmation of aberrant splicing based on their dysregulated expression40,42. To determine correlation factors between each of the identified public NJs with each splicing gene of interest, we performed a Pearson correlation analysis against each NJ and splicing-related gene pair. NJs with the highest positive correlation score against the select IDHmut upregulated genes (CELF2 and ELAVL4) averaged across all three glioma subtypes were tested in downstream qPCR assays. Similarly, NJs with the most negative correlation score against select chromosome 1p or 19q splicing-related genes downregulated in IDHmut-O cases (SNRPD2 and SF3A3) averaged across all three glioma subtypes were also tested in downstream qPCR assays.
AlphaFold2 structure predictions
AlphaFold v2.3.2 and its reference databases were installed. AlphaFold was run in multimer mode with default options and the highest rank resulting pdb file was visualized using Pymol. The image was exported with the settings ray 5000,5000 and png image,dpi=2400.
Quantification and statistical analysis
All statistical analysis was performed in R statistical software (v.4.3.3) or GraphPad Prism (v.9.2.0). Data shown in column graphs represent mean ± standard error of the mean (s.e.m.) or mean ± standard deviation (s.d.), as indicated in the figure legends. Individual data points are plotted. Details of statistical testing can be found in the figure legends. Significance values: *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001; NS, not significant. Statistical information for individual figures is provided in Supplementary Table 3.
Materials availability
We have cloned TCR cDNAs that have anti-tumour properties. We have filed an invention disclosure (UCSF-743PRV) and will share these with academic investigators as per the material transfer agreement.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
