Investigating Mechanisms of Response or Resistance to Immune Checkpoint Inhibitors by Analyzing Cell-Cell Communications in Tumors Before and After Programmed Cell Death-1 (PD-1) Targeted Therapy: An Integrative Analysis Using Single-cell RNA and Bulk-RNA Sequencing Data

Figures & data

Figure 1. Study overview, analysis of the tumor microenvironment in patients before and after anti-PD-1 treatment with scRNA-seq, and marker ligands/receptors in specific cell types

(A) Study overview. (B, C, D & E) UMAP plots of cells from pretreatment responders, posttreatment responders, pretreatment nonresponders, and posttreatment nonresponders, with each cell color coded to indicate the associated cell type. (F, G, H & I) Proportions of each cell type in the four groups. The chi-square test was used to compare the cell composition among groups. (J) Tumor mutation burden of responders vs. nonresponders (pretreatment status). The Mann-Whitney U test was used to compared the tumor mutation burden of the two groups. (K) PD-L1 (CD274) expression level in specific cell types in pretreatment responders vs. nonresponders. The Mann-Whitney U test was used to compared the expression levels of the two groups. (L) Marker ligand/receptor pairs in specific cell types. The top 4 representative markers for each cell type are shown. Notes: *, p < .05; **, p < .01; ***, p < .001; ns, not significant.

Figure 2. Comparison of pretreatment responders and nonresponders

(A & B) Cell-cell interactions across all cell pairs in pretreatment responders and pretreatment nonresponders. (C) Comparison of cell-cell interactions between pretreatment responders and pretreatment nonresponders. We calculated the ratio of the total interaction scores taking into account all interaction pairs within each specific cell pair between pretreatment responders and pretreatment nonresponders. (D) Specific comparison of each ligand-receptor interaction in pretreatment responders with each ligand-receptor interaction in pretreatment nonresponders. Overlapping genes within the ligand or receptor genes and the 693 DEGs identified between pretreatment responders and pretreatment nonresponders in the Hugo et al. study are shown. Differentially expressed ligands/receptors identified in the Hugo et al. study are shown in bold and italic text. (E) Significantly different ligand-receptor pairs with adjusted p value <.05 between pretreatment responders and pretreatment nonresponders are shown. We called these genes the “Ligand-receptor Pairs Related to Response before Treatment”. The Mann-Whitney U test was used to compare the scores of the specific ligand-receptor pair between pretreatment responders and nonresponders, taking into account the scores of that ligand-receptor pair in all cell pairs. The adjusted p value was calculated by “FDR” method. Notes, in Fig. 2A & B, the number of ligand-receptor pairs is represented by circles: larger circles reflect more ligand-receptor pairs within the cell pair, and a deeper red color indicates a higher interaction intensity. In Fig. 2 C, D & E, a Ratio (pretreatment responders/nonresponders) >1 indicates that a higher interaction intensity existed in responders than in nonresponders (shown in red). A Ratio (pretreatment responders/nonresponders) <1 indicates that a lower interaction intensity exists in responders than in nonresponders (shown in blue).

Figure 3. Relative differences in changes from pretreatment to posttreatment status between responders and nonresponders

(A, B, C & D) Cell-cell interactions across all cell pairs in responders and nonresponders before and after anti-PD-1 treatment. A deeper red color indicates a higher intensity. (E & F) Changes in responders and nonresponders from pretreatment to posttreatment status; the posttreatment total interaction scores and numbers of ligand-receptor pairs in each cell pair were divided by the respective pretreatment values. (G) Relative differences (responders vs. nonresponders) in changes from pretreatment to posttreatment status. Notes, in Fig. 3A, B, C & D, larger circles reflect more ligand-receptor pairs within the cell pair, and a deeper red color indicates a higher interaction intensity. In Fig. 3 C & D, a Ratio (post/pre) >1 indicates a higher interaction intensity exists in posttreatment data than in pretreatment data (shown in red). A Ratio (post/pre) <1 indicates that a lower interaction intensity exists in posttreatment data than in pretreatment data (shown in blue). The size of the circles reflects the ratio of the number of ligand-receptor pairs (post/pre). In Fig. 3E, a “Relative Ratio” between responders and nonresponders >1 means that the interaction intensity was relatively increased in responders or relatively decreased in nonresponders during treatment (shown in red). A “relative ratio” between responders and nonresponders <1 means that the interaction intensity was relatively decreased in responders or relatively increased in nonresponders during treatment (shown in blue). The size of the circles reflects the relative changes in the number of ligand-receptor interactions within cell pairs during treatment between responders and nonresponders.

Figure 4. Relative differences in changes in specific ligand-receptor pairs between responders and nonresponders with “Relative Ratio” >2 or <0.5 (overlapping with DEGs from the Riaz et al. study)

(A) CD8 T cells and NK cells as the ligand-expressing cells, and other cells as the receptor-expressing cells. (B) CAFs and myofibroblasts as the ligand-expressing cells, and other cells as the receptor-expressing cells. (C) Macrophages, DCs and pDCs as the ligand-expressing cells, and other cells as the receptor-expressing cells. Notes, the left and middle columns show the change in the cell-cell interaction in responders and nonresponders from pretreatment to posttreatment status; if ratio (post/pre) >1, there was a higher interaction intensity in the posttreatment data than in the pretreatment data (shown in red). If Ratio (post/pre) <1, there was a lower interaction intensity in the posttreatment data than in the pretreatment data (shown in blue). The right column shows the relative differences in changes in responders and nonresponders from pretreatment to posttreatment status. If the “Relative Ratio” between responders and nonresponders was >1, the interaction intensity was relatively increased in responders or relatively decreased in nonresponders during treatment (shown in red). If the “Relative Ratio” between responders and nonresponders was <1, the interaction intensity relatively decreased in responders or relatively increased in nonresponders during treatment (shown in blue).

Figure 5. Relative differences in changes in specific ligand-receptor pairs between responders and nonresponders with “Relative Ratio” >2 or <0.5 (overlapping with DEGs identified in the Riaz et al. study)

(A) B cells, plasma cells, endothelial cells, proliferative T cells and Tregs as the ligand-expressing cells, and other cells as the receptor-expressing cells. (B) Tumor cells as the ligand-expressing cells, and other cells as the receptor-expressing cells. Notes, the left and middle columns show the change in the cell-cell interaction in responders and nonresponders from pretreatment to posttreatment status; if Ratio (post/pre) >1, there was a higher interaction intensity in the posttreatment data than in the pretreatment data (shown in red). If Ratio (post/pre) <1, there was a lower interaction intensity in the posttreatment data than in the pretreatment data (shown in blue). The right column shows the relative differences in changes in responders and nonresponders from pretreatment to posttreatment status. If the “Relative Ratio” between responders and nonresponders was >1, the interaction intensity relatively increased in responders or relatively decreased in nonresponders during treatment (shown in red). If the “Relative Ratio” between responders and nonresponders was <1, the interaction intensity relatively decreased in responders or relatively increased in nonresponders during treatment (shown in blue).

Figure 6. “Ligand-receptors Pairs Related to Response On Treatment” with significant relative differences in changes between responders and nonresponders in our study

(A) Changes from pretreatment to posttreatment status in responders. (B) Changes from pretreatment to posttreatment status in nonresponders. (C) Relative differences in the changes between responders and nonresponders. The Mann-Whitney U test was performed to compare the “Ratio (post/pre)” values between responders and nonresponders. The false discovery rate (FDR) method proposed by Yoav Benjamini and Yosef Hochberg was applied to calculate the adjusted p value. Ligand-receptor pairs with adjusted p value <.05 were called the “Ligand-receptor Pairs Related to Response On Treatment”. ***, p < .001; **, p < .01; *, p < .05. Notes, in Fig. 6A, B & C, a Ratio (post/pre) >1 indicates that a higher interaction intensity existed in the posttreatment data than in the pretreatment data (shown in red). A Ratio (post/pre) <1 indicates that a lower interaction intensity existed in the posttreatment data than in the pretreatment data (shown in blue).

Figure 7. Trajectory Analyses of All Cell Types and Tumor-Specific CD8 T cells Using Ligand/Receptor-Related Genes

(A) Trajectory analyses of all cell types in responders and nonresponders. (B) Distribution of CD39 (ENTPD1) and CD103 (ITGAE) coexpression in tumor-specific CD8 T cells. The figures shows that tumor-specific CD8 T cells were enriched in the exhausted CD8 T cells. (C) Trajectory analysis of the abovementioned exhausted (tumor-specific) CD8 T cells of responders using ligand receptor-related genes. The results showed that these cells developed along four branches. (D) Differentially expressed gene analysis among the four branches using ligand-receptor-related genes. The left two columns show highly expressed genes in branch 1, and the right column shows highly expressed genes in branch 2 or 3. (E) Coexpression analysis of the abovementioned exhausted (tumor-specific) CD8 T cells using ligand receptor-related genes. The results showed the expression correlations among genes. The left annotation column shows genes that were highly expressed in the corresponding cell branch. (F) GSVA of the four branches using GO terms. (G) Differentially expressed gene analysis comparing branch 1 cells between pretreatment status and posttreatment status.

Figure 8. Validation of the “Ligand-receptor Pairs Related to Response Before Treatment” and the “Ligand-receptor Pairs Related to Response On Treatment) with additional immunotherapy datasets

The receiver operating characteristic (ROC) curves of the predictive models in the validation sets and the confusion tables of the best model’s classification of the validation sets. (A & B) The “Ligand-receptor Pairs Related to Response Before Treatment” model was validated by validation set 1 (including 41 patients who received nivolumab or pembrolizumab, with samples obtained before treatment) and validation set 2 (including 14 patients who received nivolumab or pembrolizumab, with samples obtained before treatment). (C) The “Ligand-receptor Pairs Related to Response On Treatment” model was validated by the pretreatment and posttreatment data in validation set 1 (including 9 patients who received nivolumab or pembrolizumab, with samples obtained both before treatment and after treatment from the same patients).