Regulatory T-cell Genes Drive Altered Immune Microenvironment in Adult Solid Cancers and Allow for Immune Contextual Patient Subtyping

Immune gene expression shows pattern differences between cancer types but shows no correlation with number of mutations. A, Gene expression heatmap of the 129 common immune genes across all solid cancer types of TCGA. Red indicates higher mean absolute expression of the patients of one tumor type. Thymoma has the highest mean expression of the correlated genes but shows a distinct pattern when compared with other cancers. Kidney and lung have high mean immune gene expression, while adrenocortical carcinoma and uveal melanoma have the lowest mean expression. B, Scatterplot depicting the correlation between number of mutations and mean immune gene expression. The graph shows the number of mutations of each tumor sample versus the mean of immune gene expression. Only a low correlation (R² = 0.015, straight line fit Pearson correlation) is observed, showing mutation number is not definitive in the expression of immune-correlated genes. Standard TCGA abbreviations for cancer types are shown in Fig. 1B, colors indicate different cancer types.

Normal tissue has higher immune gene expression in some cancers. A, Gene expression heatmap comparing the immune gene expression of matched normal and tumor tissue samples across solid cancers. Top heatmap shows the mean absolute gene expression of normal samples while the bottom shows the tumor samples. Below the heatmap a scaled heatmap showing for each cancer type whether the normal or the tumor tissue has higher mean immune gene expression (red means higher; blue means lower). Tumor types are grouped by having a higher mean immune gene expression (High) in normal tissue, lower (Low) or similar (Sim). Five of 16 cancer types have higher immune gene expression in normal tissue compared with tumor tissue, 3 have lower, and 8 are similar. B, Boxplot showing the difference in expression of normal and matched tumor tissue of the 11 seed genes. The median of each gene was taken for the matched samples across solid cancers. Pink, normal tissue; blue, tumor. FOXP3 showed the highest increase from normal to tumor tissue, while NCAM1 showed the highest decrease.

Treg gene signature expression is higher in most cancers compared with normal tissue, and FOXP3 shows more hypomethylation in tumor tissue. A, Gene expression heatmap of the Treg and CD8 signatures across cancer types, comparing matched normal tissue to tumor tissue. Below the individual absolute gene expression heatmap is the summary heatmap showing whether normal or tumor tissue had higher expression of both signatures (red means higher expression). Tumor types are ordered in the same groups as Fig. 3A. The CD8 signature was higher in the normal tissue of more than half of the cancer types (12 of 16). However, the Treg signature was higher in all tumor tissue compared with normal tissue with the exception of liver cancer. B, Boxplots showing that more tumor samples have FOXP3 hypomethylation compared with normal tissue. Many samples of the tumor tissue (blue) fall below the third quarter of samples and have a lower median of FOXP3 methylation compared with normal tissue (pink). FOXP3 methylation data for all matched samples of solid cancers in TCGA were included. All seven methylation sites assigned to the FOXP3 gene were included in the analysis. C, Boxplots showing the correlation between the Treg gene signature and the methylation of the 205 site of FOXP3. The methylation of samples of cg10858077_X_49121205 were binned according to their SD and the expression of the Treg signature for each bin was graphed. No correlation between methylation and Treg signature expression was found.

Colorectal cancer patients can be clustered into two groups based on Treg signature that show distinct clinical characteristics. A, Gene expression heatmap of the genes forming part of the Treg signature. Using the clustering method described in the methods, two clusters were found and named the high and low Treg cluster. Each gene is scaled individually from red (high gene expression) to blue. B, Bar charts showing proportions of samples within each CMS (left) and anatomical site (right) by Treg cluster. The majority of samples in the high Treg cluster are in CMS1 or CMS4, whereas the low Treg cluster has mainly CMS2 samples. Similarly, samples from the High Treg cluster are mainly right-sided tumors while samples from the low Treg cluster are mainly left-sided. NAs in CMS are samples that are not significantly assigned to a CMS. The P values shown indicate that distribution between Treg cluster and CMS or anatomic sites were independent as calculated by the χ² test of independence.