Intra-tumoral T cells in pediatric brain tumors display clonal expansion and effector properties

Abstract

Brain tumors in children are a devastating disease in a high proportion of patients. Owing to inconsistent results in clinical trials in unstratified patients, the role of immunotherapy remains unclear. We performed an in-depth survey of the single-cell transcriptomes and clonal relationship of intra-tumoral T cells from children with brain tumors. Our results demonstrate that a large fraction of T cells in the tumor tissue are clonally expanded with the potential to recognize tumor antigens. Such clonally expanded T cells display enrichment of transcripts linked to effector function, tissue residency, immune checkpoints and signatures of neoantigen-specific T cells and immunotherapy response. We identify neoantigens in pediatric brain tumors and show that neoantigen-specific T cell gene signatures are linked to better survival outcomes. Notably, among the patients in our cohort, we observe substantial heterogeneity in the degree of clonal expansion and magnitude of T cell response. Our findings suggest that characterization of intra-tumoral T cell responses may enable selection of patients for immunotherapy, an approach that requires prospective validation in clinical trials.

Extended Data Fig. 1:. Clinical parameters do not correlate with T cell clonal expansion.

Correlation of CD4⁺ (left panel) or CD8⁺ (right panel) T cell clonal expansion with clinical and pathological characteristics of PBT patients (CD4⁺, n = 26; CD8⁺, n = 32). Patients with <50 CD8⁺ or CD4⁺ T cells with TCR data were excluded. Error bars represent mean ± s.e.m.

Extended Data Fig. 2:. Mutanome analysis of pediatric brain tumors.

a, Schema of mutanome analysis and neoantigen profiling in pediatric brain tumors. b, Number of genomic tumor-specific variants in pediatric brain tumors (n = 9) (top); table shows top 10 high-confidence tumor-specific variants detected in two index patients (MBL, medulloblastoma; HGG, high-grade glioma) (bottom). c, Kaplan-Meier survival curve based on T cell gene signature in pediatric high-grade glioma from the Pediatric Brain Tumor Atlas (PBTA); n = 36 per signature group; n.s. denotes P =0.078 by multivariate Cox regression; HR, hazard ratio; CI, confidence interval.

Extended Data Fig. 3:. CD8+ T cell subsets within pediatric brain tumors.

a, Violin plots show the per-cell distribution of unique genes, unique molecular identifiers (UMI), and percentage of UMI mapped to mitochondrial genome in 26,332 single CD8⁺ T cells across clusters in PBT (n = 38). Box plots extend from the 25th to 75th percentile and the center line represents the median. Whiskers are bounded by 25^th percentile - 1.5*interquartile range or 75th percentile + 1.5*interquartile range. b, UMAP shows Seurat clustering of 26,332 CD8⁺ T cell transcriptomes in PBT (n = 38). c, Heatmap shows top 50 differentially-expressed genes using MAST across CD8⁺ T cell clusters. d, GSEA plot shows enrichment of the indicated gene signatures in the indicated CD8⁺ T cell clusters. e, Gene set enrichment analysis (GSEA) plot showing enrichment of the ICB response signature in clonally-expanded versus non-expanded T cells from PBT patients (CD8⁺, n =38; CD4⁺, n = 35). In d and e, FDR-adjusted P value (q) and normalized enrichment score (NES) determined using fgsea package on R. f, Pie charts show TRAV, TRAJ and TRBV gene usage by T cells in MAIT cell clusters (below, key).

Extended Data Fig. 4:. Composition and phenotype of tumor-infiltrating CD8⁺ T cells.

a, Subset composition of tumor-infiltrating CD8⁺ T cells (stacked to 100%) across PBT patients (n = 38); numbers above bars represent total number of CD8⁺ T cells per patient and when <50, numbers are highlighted in red. b, Proportion of CD8⁺ T cell subsets among total CD8⁺ T cells in newly diagnosed (n = 34) versus recurrent (n = 4) tumors (above, key); all comparisons are non-significant by nonparametric two-tailed Mann-Whitney test. Error bars represent mean ± s.e.m. c, Analysis of canonical pathways from the Ingenuity Pathway Analysis database (horizontal axis; bars in plot) for which clonally-expanded CD8⁺ T cells from PBT show enrichment, presented as the frequency of differentially-expressed genes encoding components of each pathway that are upregulated or downregulated (key) in clonally-expanded CD8⁺ T cells relative to their expression in non-expanded cells (left vertical axis), and adjusted P values (right vertical axis; line; Fisher’s exact test); numbers above bars indicate total genes in each pathway. d, Crater plot displays genes differentially-expressed between clonally-expanded (clone size > 1) versus non-expanded (clone size = 1) CD8⁺ T cells from PBT (n = 38) (X axis) or pre-ICB tumors from ICB responders (n = 6) (Y axis). Top right quadrant displays genes upregulated in clonally-expanded CD8⁺ T cells that are shared between PBT and tumors from ICB responders. Size of dots represents significance (Benjamini-Hochberg FDR-corrected P value < 0.05 and log2 fold change > 0.35 or < −0.35 using MAST) and color of dots represents mean expression of displayed genes.

Extended Data Fig. 5:. T cell responses in pediatric brain tumors versus adult brain tumors.

a, UMAP (left) and violin (right) displays TCF7 expression across subsets in tumor-infiltrating CD8⁺ T cells from PBT (n = 38). Inset (above left) shows proportion of TCF7-expressing cells per subset. b, UMAP (left) and violin (right) displays KLRB1 expression across subsets in tumor-infiltrating CD8⁺ T cells from PBT (n = 38). Inset (above left) shows proportion of KLRB1-expressing cells per subset. c, Expression of CLEC2D transcripts in adult glioblastoma (GBM, n = 96), pediatric high-grade glioma (pHGG, n = 25) or pediatric low-grade glioma (pLGG, n = 93) from PedcBioPortal datasets. In a-c, box plots extend from the 25th to 75th percentile and the center line represents the median. Whiskers represent minimum and maximum values. d, Single-cell trajectory analysis showing relationship between cells in different CD8⁺ T cell subsets (line) in pediatric brain tumors, constructed using Monocle 3.

Extended Data Fig. 6:. Tumor-infiltrating PD-1⁺CD8⁺ T cells display clonal expansion and cytokine production.

a, Proportion of clonally-expanded CD8⁺ T cells in PDCD1-nonexpressing versus PDCD1-expressing CD8⁺ T cells (n = 32) (top); ****P = 1.6×10⁻⁵. Proportion of PDCD1-expressing CD8⁺ T cells in nonexpanded versus clonally-expanded CD8⁺ T cells in PBT (n = 32) (bottom); ****P = 1.8×10⁻⁵. P value determined by nonparametric two tailed Wilcoxon matched-pairs signed rank test in both analyses. Patients with <50 CD8⁺ T cells with TCR data were excluded. b, Representative flow-cytometric gating strategy for the assessment of CD103, cytotoxic molecule, and cytokine expression in total CD8⁺ T cells and in PD-1^negCD8⁺ T cells versus PD-1⁺CD8⁺ T cells from PBT (also related to Fig. 4 b,c). c, Proportion (left plot) of cells expressing IL2 in PDCD1-non-expressing versus PDCD1-expressing CD8⁺ T cells in PBT (n = 38). Flow-cytometric analysis (right) of the expression and proportion of IL-2⁺ cells in PD 1^negCD8⁺ T cells versus PD-1⁺CD8⁺ T cells from PBT patients (n = 7); n.s. denotes P = 0.16 by non-parametric two-tailed Wilcoxon matched-pairs signed rank test. d, Bar chart shows polyfunctionality based on the production of multiple cytokines in PD-1^negCD8⁺ T cells versus PD-1⁺CD8⁺ T cells from PBT (n = 7). e, GSEA plot shows enrichment of cell cycle gene signature in PDCD1-expressing versus PDCD1-non-expressing CD8⁺ T cells in PBT (n = 38). P value and NES as in Fig. 2a.

Extended Data Fig. 7:. Expression of transcripts encoding HLA molecules and features of LAG3-expressing CD8⁺ T cells in pediatric brain tumors.

Expression of (a) HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DPB1, and HLA-DQB1 transcripts and (b) antigen processing and presentation signature genes across six diagnoses in the PBTA; error bars represent mean ± s.e.m. CPP, choroid plexus papilloma (n = 16); CrPh, craniopharyngioma (n = 36); LGG, low-grade glioma (n = 302); HGG, high-grade glioma (n = 148); MBL, medulloblastoma (n = 119); AE, anaplastic ependymoma (n = 93). c, Volcano plot shows differentially-expressed genes between LAG3-non-expressing versus LAG3-expressing CD8+ T cells from PBT patients with low expression of PDCD1 in CD8+ T cells (n = 10) (Benjamini-Hochberg FDR-corrected P value < 0.05, log2 fold change > 0.35 or < −0.35 using MAST); dot size and color as in Fig. 6b.

Extended Data Fig. 8:. CD4⁺ T cell subsets within pediatric brain tumors.

a, Violin plots show the per-cell distribution of unique genes, unique molecular identifiers (UMI), and percentage of UMI mapped to mitochondrial genome in 14,994 single CD4⁺ T cells across clusters in PBT (n = 35). Box plots extend from the 25th to 75th percentile and the center line represents the median. Whiskers are bounded by 25^th percentile - 1.5*interquartile range or 75th percentile + 1.5*interquartile range. b, UMAP shows Seurat clustering of 14,994 CD4⁺ T cell transcriptomes in PBT (n = 35). c, Heatmap shows top 50 differentially-expressed genes using MAST across CD4⁺ T cell clusters. d, GSEA plot shows enrichment of the indicated gene signatures in the indicated CD4⁺ T cell clusters. e, GSEA plot shows enrichment of CD4-CTL gene signature in clonally-expanded versus non-expanded non-T_REG CD4⁺ T cells. In d and e, P value and NES determined as in Fig. 2a. f, Proportion of clonally-expanded CD4⁺ T cells in PDCD1-non-expressing versus PDCD1-expressing non-T_REG CD4⁺ T cells (n = 24); Patients with <50 CD4⁺ T cells with TCR data or patients with 0 PDCD1⁺CD4⁺ T cells were excluded; **P = 0.0018 by nonparametric two-tailed Wilcoxon matched-pairs signed rank test. g, Flow-cytometric analysis of IL-2 expression and proportion of IL-2⁺ cells in PD-1^negCD4⁺ T cells versus PD-1⁺CD4⁺ T cells from PBT patients (n = 10); n.s. denotes P = 0.37 by non-parametric two-tailed Wilcoxon matched-pairs signed rank test.

Fig. 1:. T cells within pediatric brain tumors show marked clonal expansion.

a, Study overview. b, T cell receptor (TCR) clone size distribution amongst tumor-infiltrating CD4⁺ and CD8⁺ T cells across pediatric brain tumor (PBT) patients (n = 38); grey bars indicate non-expanded cells with a clone size of 1; colored lines delineate expanded clonotypes with clone size >1. c, Proportion (left) of clonally-expanded CD4⁺ or CD8⁺ T cells and top five clones (right) among total CD4⁺ or CD8⁺ T cells from each patient (CD4⁺, n = 26; CD8⁺, n = 32) (dashed line marks 10%); each circle represents a patient (above, key). d, Normalized Inverse Simpson and Shannon-Wiener diversity index of CD4⁺ T cells (left panel) or CD8⁺ T cells (right panel) from PBT (CD4⁺, n = 26; CD8⁺, n = 32) versus NSCLC (n = 10) (above, key); **P = 0.0090, ****P = 4 × 10⁻⁶ and n.s. denotes P = 0.67 by nonparametric two-tailed Mann-Whitney test. In c and d, patients with <50 CD4⁺ or CD8⁺ T cells with TCR data were excluded; error bars represent mean ± s.e.m.

Fig. 2:. Tumor-infiltrating T cells display neoantigen-specific T cell gene signatures which are associated with improved survival.

a, Gene set enrichment analysis (GSEA) plot showing enrichment of neoantigen-specific CD8⁺ or CD4⁺ T cell gene signatures in clonally-expanded versus non-expanded T cells from PBT patients (CD8⁺, n = 38; CD4⁺, n = 35); FDR-adjusted P value (q) and normalized enrichment score (NES) determined using fgsea package on R. b, Kaplan-Meier survival curves showing significant prognostic separation according to CD8 (left) or CD4 (right) neoantigen-specific T cell gene signature in pediatric high-grade glioma from the Pediatric Brain Tumor Atlas (PBTA); n = 36 per signature group; ***P = 0.006, *P = 0.013, by multivariate Cox regression; HR, hazard ratio; CI, confidence interval. c, GLIPH2 analysis showing 1071 CD8⁺ (left) or 1104 CD4⁺ (right) specificity groups containing TCRβ sequences with shared motifs in tumor-infiltrating T cells from PBT patients (CD8⁺, n = 38; CD4⁺, n = 35). Each node represents a specificity group; node size represents the sum of TCRβ clone sizes within a specificity group; node color represents the number of unique TCRβ sequences; grey linkages represent the sharing of at least one TCRβ (Vβ-CDR3β-Jβ) sequence between specificity groups. d, Number of TCR specificity groups in tumor-infiltrating T cells from PBT patients (CD8⁺, n = 38; CD4⁺, n = 35) (dashed line marks 10 specificity groups); each circle represents a patient (above, key); error bars represent mean ± s.e.m.

Fig. 3:. Clonally-expanded CD8⁺ T cells display cell states linked to anti-tumor immunity.

a, Uniform manifold approximation and projection (UMAP) displays single-cell transcriptomes and corresponding clone size of the CD8⁺ TCR clonotypes across PBT patients (n = 38); circle size in UMAP indicates degree of clonal expansion. Inset (right) shows the frequency of each cell subset among all clonally-expanded CD8⁺ T cells. b, Subset composition of clonally-expanded CD8⁺ T cells (stacked to 100%) across PBT patients (n = 32) (above, key); numbers above bars represent total number of expanded CD8⁺ T cells per patient. c, Proportion of GZMK^high or CD16⁺ effector subset among clonally-expanded intra-tumoral CD8⁺ T cells in low-grade (LGG, CPP, CrPh; n = 14) versus high-grade (MG, AE, HGG, MBL; n = 18) tumors; *P = 0.011, GZMK^high; *P = 0.034, CD16⁺ effector, by nonparametric two-tailed Mann-Whitney test. Error bars represent mean ± s.e.m. In b and c, patients with <50 CD8⁺ T cells with TCR data were excluded.

Fig. 4:. Clonally-expanded CD8⁺ T cells display effector properties.

a, Crater plot displays genes differentially-expressed between clonally-expanded (clone size > 1) versus non-expanded (clone size = 1) CD8⁺ T cells from PBT (n = 38) (X axis) or NSCLC tumors (n = 10) (Y axis). Top right quadrant displays genes upregulated in clonally-expanded CD8⁺ T cells that are shared between PBT and NSCLC. Size of dots represents significance (Benjamini-Hochberg FDR-corrected P-value < 0.05 and log2 fold change > 0.35 or < −0.35 using MAST) and color of dots represents mean expression of displayed genes. b, Flow-cytometric analysis of the expression and frequency of CD103⁺CD8⁺ T cells (top panel) or GZMK⁺CD8⁺ T cells (bottom panel) among tumor-infiltrating CD8⁺ T cells in PBT patients (n = 10). Error bars represent mean ± s.e.m. c, Bar chart shows production of multiple cytokines in GZMK^negCD8⁺ T cells versus GZMK⁺CD8⁺ T cells from PBT (n = 10). Gating strategy for b and c are shown in Extended Data Fig. 6b. d, Single-cell TCR sequence analysis showing TCR sharing between CD8⁺ T cells from the indicated subsets (rows in grid, connected by lines); colored vertical bars (top) indicate number of cells with TCR sharing between the indicated subsets (columns); colored horizontal bars (bottom) indicate number of CD8⁺ T cells per subset; color key as in Figure 3b.

Fig. 5:. PDCD1-expressing intra-tumoral CD8⁺ T cells are not dysfunctional.

a, UMAP displays PDCD1 expression in tumor-infiltrating CD8⁺ T cells from PBT (n = 38). Inset (above) shows proportion of PDCD1-expressing cells per subset. b, Flow-cytometric analysis of the expression and frequency of PD-1⁺CD8⁺ T cells among tumor-infiltrating CD8⁺ T cells in PBT patients (n = 10); error bars represent mean ± s.e.m. c, Volcano plot shows differentially-expressed genes between PDCD1-non-expressing versus PDCD1-expressing CD8⁺ T cells from PBT (n = 38) (Benjamini-Hochberg FDR-corrected P value < 0.05, log2 fold change > 0.35 or < −0.35 by MAST); dot size represents the difference in percentage of expressing cells (for respective genes) between PDCD1-non-expressing and PDCD1-expressing CD8⁺ T cells; dot color indicates mean expression of respective gene. d, Proportion of cells expressing the indicated differentially-expressed genes between PDCD1-non-expressing versus PDCD1-expressing CD8⁺ T cells in PBT (n = 38). GZMA, IFNG, TNF, ITGAE, and GZMK are differentially-expressed genes in PDCD1-expressing CD8⁺ T cells (adjusted P value < 0.05, log2 fold change > 0.35). e, Flow-cytometric analysis of the expression (left) and proportion (right) of expressing cells for each of the indicated effector molecules or T_RM marker in PD-1^negCD8⁺ T cells versus PD-1⁺CD8⁺ T cells from PBT patients (n = 7 for IFNγ, TNF; n = 10 for granzyme A, granzyme B, granzyme K, CD103); **P = 0.002 for granzyme A, granzyme B, granzyme K, CD103; * P = 0.031 for IFNγ, TNF using non-parametric two-tailed Wilcoxon matched-pairs signed rank test.

Fig 6:. Heterogeneity in the expression of immunotherapy targets in CD8⁺ T cells

a, Proportion of cells expressing the indicated transcripts (one per column, left) or ICB response signature score (column, right) in clonally-expanded intra-tumoral CD8⁺ T cells in PBT (n = 32); patients with <50 CD8⁺ T cells with TCR data were excluded; TIM3 gene name, HAVCR2; 4–1BB gene name, TNFRSF9. b, Volcano plot shows differentially-expressed genes between LAG3-non-expressing versus LAG3-expressing CD8⁺ T cells from PBT (n = 38) (Benjamini-Hochberg FDR-corrected P value < 0.05, log2 fold change > 0.35 or < −0.35 using MAST); dot size represents the difference in percentage of expressing cells (for respective genes) between LAG3-non-expressing and LAG3-expressing CD8⁺ T cells; dot color indicates mean expression of respective gene. c, Proportion of cells expressing the indicated differentially-expressed genes (Benjamini-Hochberg FDR-corrected P value < 0.05, log2 fold change > 0.35) between LAG3-non-expressing versus LAG3-expressing CD8⁺ T cells in PBT (n = 38). d, Proportion of clonally-expanded CD8⁺ T cells in LAG3-non-expressing versus LAG3-expressing CD8⁺ T cells in PBT (n = 32); patients with <50 CD8⁺ T cells with TCR data were excluded; ****P = 1 × 10⁻⁶ by nonparametric two-tailed Wilcoxon matched-pairs signed rank test; in PDCD1-low patients (n=10), **P = 0.0039. In (c) and (d), orange lines denote PBT donors with low PDCD1 expression in CD8⁺ T cells (n = 10).

Fig. 7:. CD4-CTLs are clonally-expanded in pediatric brain tumors.

a, UMAP displays single-cell transcriptomes and corresponding clone size of the CD4⁺ TCR clonotypes across PBT patients (n = 35); circle size in UMAP indicates degree of clonal expansion Inset (top) shows the frequency of each cell subset among all clonally-expanded CD4⁺ T cells. b, Subset composition of tumor-infiltrating CD4⁺ T cells (stacked to 100%) across PBT patients (n = 35); numbers above bars represent total number of CD4⁺ T cells per patient and when <50, numbers are highlighted in red. c, UMAP shows CD4-CTL gene signature score (left) or PDCD1 expression (right) in tumor-infiltrating CD4⁺ T cells from PBT (n = 35). Inset (above, right) shows proportion of PDCD1-expressing cells per subset. d, Flow-cytometric analysis of the proportion of tumor-infiltrating CD4⁺ T cells expressing granzyme A, granzyme B or PD-1 in PBT (n = 12); error bars represent mean ± s.e.m.

Fig. 8:. PD-1⁺CD4⁺ T cells display increased expression of cytotoxic molecules and cytokines

a, Volcano plot shows differentially-expressed genes between PDCD1-non-expressing versus PDCD1-expressing non-T_REG CD4⁺ T cells from PBT (n = 35) (Benjamini-Hochberg FDR-corrected P value < 0.05, log2 fold change > 0.35 or < −0.35 using MAST); dot size represents the difference in percentage of expressing cells (for respective genes) between PDCD1-non-expressing and PDCD1-expressing non-T_REG CD4⁺ T cells; dot color indicates mean expression of respective gene. b, Proportion of cells expressing transcripts encoding the indicated effector molecules in PDCD1-non-expressing versus PDCD1-expressing non-T_REG CD4⁺ T cells in PBT (n = 35); GZMA and IFNG are differentially-expressed genes in PDCD1-expressing CD4⁺ T cells (Benjamini-Hochberg FDR-corrected P value < 0.05, log2 fold change > 0.35 using MAST). c, Flow-cytometric analysis of the expression and proportion of cells expressing the indicated effector molecules in PD-1^negCD4⁺ T cells versus PD-1⁺CD4⁺ T cells from PBT (n = 10 for IFNγ, TNF; n = 12 for granzyme A, granzyme B). ***P = 0.0005 for granzyme A, granzyme B using nonparametric two-tailed Wilcoxon matched-pairs signed rank test. d, Proportion of T_REG cells and e, CD4-CTL among tumor-infiltrating CD4⁺ T cells in pediatric brain tumors (n = 35) versus adult high-grade gliomas (n = 26); **P = 0.0013, ****P = 1.4 × 10⁻⁷ respectively by nonparametric two-tailed Mann-Whitney test; error bars represent mean ± s.e.m. f, Proportion of clonally-expanded intra-tumoral CD4⁺ T cells expressing the indicated transcripts (one per column) in PBT (n = 26) (one per row); patients with <50 CD4⁺ T cells with TCR data were excluded; TIM3 gene name, HAVCR2; 4–1BB gene name, TNFRSF9.