Fig. 3: Distinct microenvironment of different endothelial subtypes.

a Schematic plot illustrating cell type co-localization analysis of ST data. A square represents a bin50 spot that covers a 25 µm area in the ST slide. In the bin50 model, the strength of co-localization between different cell types is explored in a single bin50 spot, while in the bin150 model, co-localization between cell types is explored between a bin50 spot and its eight neighboring bin50 spots. b Heatmap displaying the scaled NES between different cells in the bin50 model across thirty slides. c, d The NES between target cells and EC subtypes in the bin50 model across ST slides from PT (left panel) and from LN (right panel). The target cells include tumor cells (c) and T/NK cells (d). The NES between target cells and vECs, aECs, cECs, and ECs are compared by the Wilcoxon rank sum test (paired samples, two-tailed); ***p < 0.001. e, f The ES between vECs, aECs, cECs, all ECs and various target cells in the bin50 model across ST slides from PT (e) and from LN (f). The ES values between vECs, aECs, cECs, all ECs and target cells in the ST slides are calculated and each dot represents the median value of an ES set. The four ES sets are compared using a Wilcoxon rank sum test (paired samples, two-tailed). The ES set that is significantly higher than the other three ES sets is highlighted; *p < 0.05; ***p < 0.001. g, h Spatial in situ projection illustrating the distribution of different cells in ST slides of paired PT and LN from NCC-P18 and NCC-P12 patients. i Distribution of vECs, aECs and cECs on the representative ST slides of samples containing tertiary lymphoid structures (TLS), with TLS area labeled in black. j Proportion of vECs, aECs and cECs within the TLS areas across seven samples that contained TLS. k Proportion of vECs, aECs, and cECs with the closest Euclidean distance to TLS across all samples containing TLS (n = 7). P-values are determined by the two-tailed Wilcoxon rank sum test.