Fig. 6: Prognostic value of spatial proximity between CD8⁺ T cells and SLC2A1⁺ TAMs.

a,b, Forest plots of SLC2A1⁺ TAM signature influence on PFS in the SYS (a) (n = 109) and TUSM (b) (n = 18) immunotherapy cohorts. Error bars indicate the 95% confidence interval (CI). The hazard ratio for SLC2A1⁺ TAM was 1.914 (95% CI, 1.042–3.518) in the SYS cohort and 2.151 (95% CI, 0.693–6.671) in the TUSM cohort. c,d, Patients were stratified into three subgroups based on combined infiltration levels of SLC2A1⁺ and SLC2A1⁻ TAMs: SLC2A1⁺ TAM^Low SLC2A1⁻ TAM^High, SLC2A1⁺ TAM^High SLC2A1⁻ TAM^Low and otherwise. Box plots comparing activated CD8⁺ T cell (left) or interferon gamma production (right) across groups in the SYS (c) (n = 109) and TUSM (d) (n = 18) immunotherapy cohorts. e,f, Kaplan–Meier survival curves of PFS stratified according to the above groups in the SYS (e) (n = 109) and TUSM (f) (n = 18) cohorts. g, Clinical nomogram integrating key variables to predict 1-, 2- and 3-year survival probabilities. Points from individual predictors are summed to estimate total survival risk. h, ROC curves evaluating 3-year survival prediction in the discovery (n = 76) and validation cohorts (n = 51) derived from TJ cohort 2. i, Area under the ROC curve analysis for the discovery (n = 76) and validation cohorts (n = 51). j,k, Kaplan–Meier curves of OS in the discovery (j) (n = 76) and validation cohorts (k) (n = 51). l, Schematics highlighting the major findings of this study. Statistical analysis was performed using the two-sided Cox proportional regression P values (a and b), two-tailed unpaired Student’s t-test (c and d), and log-rank (Mantel–Cox) test (e, f, j and k). For the box plots, the centre line denotes the median, the box encompasses the first and third quartiles and whiskers denote 1.5 times the interquartile range. The schematic in l was created using BioRender.com.

Source data