Fig. 2: Associations of T cells with irAEs.

a–d The uniform manifold approximation and projection (UMAP) plots of 248,557 high-quality T cells (a) and CD8⁺ T cell subcluster (b), CD4⁺ T cell subcluster (c) and NK T cell subcluster (d). Each dot represents individual cell and colored by their cell subclusters. e Stacked bar plots showing the distribution of the T cell subclusters in different tissue type, treatment, irAE status and disease relapse. f Proportional analysis of T cell subclusters comparing patients with and without irAEs at different time points. In all boxplots, the center line represents the median, the bounds of the box indicate the 25th and 75th percentiles (interquartile range, IQR), and the whiskers extend to 1.5 × IQR. Each dot represents an individual patient. P values were calculated using the two-sided Dirichlet-multinomial regression model and FDR values were calculated by the Benjamini–Hochberg method. Sample sizes: pre-radio and on-radio tumor (irAEs absent: n = 6; irAEs present: n = 12); pre-radio PBMC (irAEs absent: n = 5; irAEs present: n = 4); pre-imm and on-imm PBMC (irAE absent: n = 3; irAE present: n = 8). All sample sizes reported are biological replicates, with the patient serving as the independent unit of study. Tn naive T cell, Tm memory T cell, NK natural killer, Tem effector memory T cells, Th T helper cell, Tex exhausted T cell, Temra terminally differentiated effector memory or effector T cells, ICI immune checkpoint inhibitor, irAEs immune-related adverse events, Radio chemoradiotherapy, UK, unknown, NoICI not treated by ICI. Source data are provided as a Source Data file.