Immune-Inhibitory Gene Expression is Positively Correlated with Overall Immune Activity and Predicts Increased Survival Probability of Cervical and Head and Neck Cancer Patients

Abstract

Background: Limited immunotherapy options are approved for the treatment of cervical cancer and only 10-25% of patients respond effectively to checkpoint inhibition monotherapy. To aid the development of novel therapeutic immune targets, we aimed to explore survival-associated immune biomarkers and co-expressed immune networks in cervical cancer. Methods: Using The Cancer Genome Atlas (TCGA) Cervical Squamous Cell Carcinoma (CESC) data (n = 304), we performed weighted gene co-expression network analysis (WGCNA), and determined which co-expressed immune-related genes and networks are associated with survival probability in CESC patients under conventional therapy. A "Pan-Immune Score" and "Immune Suppression Score" was generated based on expression of survival-associated co-expressed immune networks and immune suppressive genes, which were subsequently tested for association with survival probablity using the TCGA Head Neck Squamous Cell Carcinoma (HNSCC) data (n = 528), representing a second SCC cancer type. Results: In CESC, WGCNA identified a co-expression module enriched in immune response related genes, including 462 genes where high expression was associated with increased survival probability, and enriched for genes associated with T cell receptor, cytokine and chemokine signaling. However, a high level of expression of 43 of the genes in this module was associated with decreased survival probability but were not enriched in particular pathways. Separately, we identified 20 genes associated with immune suppression including inhibitory immune checkpoint and regulatory T cell-related genes, where high expression was associated with increased survival probability. Expression of these 20 immune suppressive genes (represented as "Immune Suppression Score") was highly correlated with expression of overall survival-associated immune genes (represented as "Pan-Immune Score"). However, high expression of seven immune suppression genes, including TWEAK-R, CD73, IL1 family and TGFb family genes, was significantly associated with decreased survival probability. Both scores also significantly associated with survival probability in HNSCC, and correlated with the previously established "Immunophenoscore." Conclusion: CESC and HNSCC tumors expressing genes predictive of T cell infiltrates (hot tumors) have a better prognosis, despite simultaneous expression of many immune inhibitory genes, than tumors lacking expression of genes associated with T cell infiltrates (cold tumors) whether or not these tumor express immune inhibitory genes.

Keywords: cervical cancer; head and neck cancer; immune checkpoints; immune inhibition; inflamed tumours; prognosis.

FIGURE 1

Idenditifaction of a co-expression Immune Module which correlates with CESC survival probability. (A) Flow chart describing analysis pipeline and summary of results. (B) Weighted gene co-expression network analysis (WGCNA) was used to identify survival-associated co-expressed immune networks in the TCGA CESC data. Clustering of module eigengenes with a threshold of 0.6 resulted in 21 modules. (C) Correlation was performed between eigengene expressions of each module with each other to represent association of modules with each other. (D) Correlation analysis of modules eigengene expression with clinical features. (E) Correlation analysis comparing the palevioletred1 Immune Module to clinical feature days_to_death.

FIGURE 2

Protein-protein interaction (PPI) network identifies hub network genes promoting increased CESC survival probability. Co-expressed survival-associated immune genes of the WGCNA palevioletred1 Immune Module were analyzed for PPI. (A) Shown is the network of 462 highly expressed genes in patients of increased survival probability (excluding non-connected genes). (B) Hub genes were identified as genes with highest connectivity in a particular KEGG pathway.

FIGURE 3

CESC FIGO stage is associated with tumour-promoting proinflammatory cytokines. (A) Clinical features were tested for association with survival. For BMI, four groups were defined based on diagnostic BMI values (underweight <18.5, normal 18.5–24.9, overweight 25–30, obese >30). For smoking, the lower and upper quartile of the clinical feature pack_years_smoked were compared. (B) Genes of the WGCNA palevioletred1 Immune Module (n = 505) which significantly associated with 5-years survival probability with cervical cancer were analyzed for correlation with FIGO stage using Spearman correlation analysis. Genes were ordered from largest to smallest correlation coefficient r (left panel). p-values were–log10 transformed and plotted as heatmap (right panel). Significant genes at the top are positively correlated with FIGO stage and significant genes at the bottom are negatively correlated with FIGO stage.

FIGURE 4

High expression of multiple immune suppression genes is associated with increased 5-years survival probability with CESC. A custom built list of 50 immune suppressive genes including immune checkpoint inhibitor genes, Treg related genes and soluble mediators was analyzed for significant association with 5-years survival probability. (A) Number of genes where high or low expression was associated with increased 5-years survival probability. (B) p-values of significant immune suppression genes were −log10 transformed. Genes for which low expression was associated with increased survival probability were further transformed to a negative value.

FIGURE 5

“Pan-Immune Score” and “Immune Suppression Score” predicting 5-years survival probability with CESC and HNSCC are highly correlated. (A, D) Subjects were assigned a “Pan-Immune Score” according to expression of survival-associated immune genes of the WGCNA palevioletred1 Immune Module, and 5-years survival probability of the upper and lower quartiles of the “Pan-Immune Score” representing high and low scores was compared in the TCGA-CESC (A) and TCGA-HNSCC (D) data. (B, E) Subjects were assigned an “Immune Suppression Score” according to expression of survival-associated immune suppression genes described in Figure 1B, and 5-years survival probability of the upper and lower quartiles of the “Immune Suppression Score” representing high and low scores was compared in the TCGA-CESC (B) and TCGA-HNSCC (E) data. (C, F) Correlation analysis of “Pan-Immune Score” and “Immune Suppression Score” in the TCGA-CESC (C) and TCGA-HNSCC (F) data.

FIGURE 6

“Pan-Immune Score” and “Immune Suppression Score” are positively correlated with “Immunophenoscore.” Correlation analysis of “Pan-Immune Score” and “Immune Suppression Score” with “Immunophenoscore” IPS4 in TCGA-CESC (A) and TCGA-HNSCC (B).