Extended Data Fig. 1: TGF-β limits the proliferation and accumulation of CD8+ TEFF TILs.
a, Expression levels of key marker genes defining the six CD8+ T cell clusters identified by 5′ scRNA-seq. b, Average expression of published gene signatures for T cell phenotypes (left) and general biological processes (right) across CD8+ clusters (see Supplementary Table 1). c, Cycling CD8+ TILs were defined by a cycling score > 0.2. Non-cycling cells were used to construct a reference map into which cycling cells were projected. d, Violin plots showing expression of Pdcd1, Cxcr3, Ly6a and Crtam across CD8+ clusters. Markers were selected based on differential expression, surface localization, and availability of validated antibodies. e, Flow cytometry gating strategy for CD8+ TIL subpopulations based on PD-1, CXCR3, SCA-1 and CRTAM. Naïve cells were PD-1-negative; T progenitor (TPR) were PD-1+ with high CXCR3 expression. PD-1hi cells were subdivided by SCA-1 and CRTAM expression into CRTAM+ T tissue resident (TTR), CRTAM−/SCA-1+ T effectors (TEFF), and CRTAM−/SCA-1− T exhausted (TEX) cells. f, CD8+ TIL composition in metastases following 3 days of treatment, analyzed by flow cytometry. Mice per group: vehicle, n = 26; galunisertib, n = 15; αPD-L1, n = 29; dual, n = 17. Mean ± S.D. g, TCR clonal size composition of scRNA-seq CD8+ clusters. Cells were categorized as single (=1), small (1 < x ≤ 5), medium (5 < x ≤ 20), large (20 < x ≤ 100), hyper expanded (100 < x ≤ 250) based on TCR sharing. h, Cell fate trajectories of expanded CD8+ TILs (>1 TCR copy) were inferred with Monocle 3, starting from naïve-like cells. i, Pseudotime trajectory values per CD8+ cluster. Box plots show median, IQR, and 1.5× IQR whiskers; outliers plotted individually. All statistical tests were two-sided and corrected for multiple comparisons where appropriate. Statistical analysis was performed using a generalized linear mixed-effects model (beta family) with Dunnett’s multiple comparisons (f).
