Figure 5.
Cross-cohort analysis shows differences in proportion and functional states of immune subphenotypes between DLBCL and Hodgkin lymphoma (HL). (A) After batch normalization, the PCA visually confirms the immune subphenotypes identified are well distributed across the 2 cohorts, indicating no cohort bias. Visual inspection confirmed using k-nearest neighbor quantitative batch effect test (kBET) (see supplemental Figure 12). (B) TME compartment proportions (%) relative to each cohort (DLBCL/HL/RLN) are denoted on the y-axis, and the x-axis denotes major phenotypes. Comparing differences with RLN, macrophages were increased 21.3% (95% CI, 12.4-30.2; BH q < 1e-03) in DLBCL compared with differences between HL and RLN. Comparing TREG relative differences between RLN, DLBCL showed a 9.1% (95% CI, -13.6 to -4.7; q = 3.2e-04) less enriched abundance compared with differences between HL and RLN. (C) Analyses of cell-state protein expression (Z-score) on each immune subset across the 2 cohorts show differences in functional states of immune subsets in DLBCL compared with HL. PD-L1 expression on macrophages is significantly lower in DLBCL compared with HL (logFC= -0.08, q < 1e-05), whereas PD-1 expression is higher on TREG (logFC = 0.23, q < 1e-04) in DLBCL compared with HL. TIM-3 expression is significantly higher in CD8 T cells comparing DLBCL to HL (logFC = 0.57, q < 1e-05), and significantly higher in CD4 T cells compared DLBCL with HL (logFC = 0.31, q < 1e-05). (D) For each experiment, global expression gating identified CCR4+, ICOS+, Ki67+, PD1+ separately for each cohort, and relative proportions among +TREG cells are depicted. The grids on the radar denote the cohort proportion (%) among TREG phenotypes positive for a corresponding phenotype. Significant differences comparing TREG case relative proportions (P < .05) are denoted with an asterisk. There were significant higher case relative proportions of CCR4+PD-1+ TREG (6.087%; 95% CI, 2.56-9.62; P < .001) in DLBCL compared with HL, and significantly higher PD-1+ TREG relative proportions (6.85%; 95% CI, 3.76-9.93; P < .001). Alternatively, there were significantly fewer Ki67+ T-reg (-28.34%; 95% CI, -33.42 to -23.26%; P < .001) comparing DLBCL with HL relative proportions (supplemental Figure 15). (E) Spatial interactions between PD1+CCR4+ TREG and T cells were computed using a radius of interaction of 15 microns (1000 permutations, P < .01), and the relative proportion within each cohort of significant interactions/repulsions (P < .01) between PD1+CCR4+TREG and PD1+/mid/− CD8 (left) and CD4 (right) T cells were compared across cohorts using an ANOVA. The y-axis denotes the relative proportion of significant interactions (P < .01), and significant differences (BH adjusted p values) of proportions comparing DLBCL with HL and RLN are depicted.

Cross-cohort analysis shows differences in proportion and functional states of immune subphenotypes between DLBCL and Hodgkin lymphoma (HL). (A) After batch normalization, the PCA visually confirms the immune subphenotypes identified are well distributed across the 2 cohorts, indicating no cohort bias. Visual inspection confirmed using k-nearest neighbor quantitative batch effect test (kBET) (see supplemental Figure 12). (B) TME compartment proportions (%) relative to each cohort (DLBCL/HL/RLN) are denoted on the y-axis, and the x-axis denotes major phenotypes. Comparing differences with RLN, macrophages were increased 21.3% (95% CI, 12.4-30.2; BH q < 1e-03) in DLBCL compared with differences between HL and RLN. Comparing TREG relative differences between RLN, DLBCL showed a 9.1% (95% CI, -13.6 to -4.7; q = 3.2e-04) less enriched abundance compared with differences between HL and RLN. (C) Analyses of cell-state protein expression (Z-score) on each immune subset across the 2 cohorts show differences in functional states of immune subsets in DLBCL compared with HL. PD-L1 expression on macrophages is significantly lower in DLBCL compared with HL (logFC= -0.08, q < 1e-05), whereas PD-1 expression is higher on TREG (logFC = 0.23, q < 1e-04) in DLBCL compared with HL. TIM-3 expression is significantly higher in CD8 T cells comparing DLBCL to HL (logFC = 0.57, q < 1e-05), and significantly higher in CD4 T cells compared DLBCL with HL (logFC = 0.31, q < 1e-05). (D) For each experiment, global expression gating identified CCR4+, ICOS+, Ki67+, PD1+ separately for each cohort, and relative proportions among +TREG cells are depicted. The grids on the radar denote the cohort proportion (%) among TREG phenotypes positive for a corresponding phenotype. Significant differences comparing TREG case relative proportions (P < .05) are denoted with an asterisk. There were significant higher case relative proportions of CCR4+PD-1+ TREG (6.087%; 95% CI, 2.56-9.62; P < .001) in DLBCL compared with HL, and significantly higher PD-1+ TREG relative proportions (6.85%; 95% CI, 3.76-9.93; P < .001). Alternatively, there were significantly fewer Ki67+ T-reg (-28.34%; 95% CI, -33.42 to -23.26%; P < .001) comparing DLBCL with HL relative proportions (supplemental Figure 15). (E) Spatial interactions between PD1+CCR4+ TREG and T cells were computed using a radius of interaction of 15 microns (1000 permutations, P < .01), and the relative proportion within each cohort of significant interactions/repulsions (P < .01) between PD1+CCR4+TREG and PD1+/mid/ CD8 (left) and CD4 (right) T cells were compared across cohorts using an ANOVA. The y-axis denotes the relative proportion of significant interactions (P < .01), and significant differences (BH adjusted p values) of proportions comparing DLBCL with HL and RLN are depicted.

Close Modal

or Create an Account

Close Modal
Close Modal