Systems biology analysis identifies TNFRSF9 as a functional marker of tumor-infiltrating regulatory T-cell enabling clinical outcome prediction in lung cancer

Abstract

Regulatory T cells (Tregs) are enriched in the tumor microenvironment and play key roles in immune evasion of cancer cells. Cell surface markers specific for tumor-infiltrating Tregs (TI-Tregs) can be effectively targeted to enhance antitumor immunity and used for stratification of immunotherapy outcomes. Here, we present a systems biology approach to identify functional cell surface markers for TI-Tregs. We selected differentially expressed genes for surface proteins of TI-Tregs and compared these with other CD4⁺ T cells using bulk RNA-sequencing data from murine lung cancer models. Thereafter, we filtered for human orthologues with conserved expression in TI-Tregs using single-cell transcriptome data from patients with non-small cell lung cancer (NSCLC). To evaluate the functional importance of expression-based markers of TI-Tregs, we utilized network-based measure of context-associated centrality in a Treg-specific coregulatory network. We identified TNFRSF9 (also known as 4-1BB or CD137), a previously reported target for enhancing antitumor immunity, among the final candidates for TI-Treg markers with high functional importance score. We found that the low TNFRSF9 expression level in Tregs was associated with enhanced overall survival rate and response to anti-PD-1 immunotherapy in patients with NSCLC, proposing that TNFRSF9 promotes immune suppressive activity of Tregs in tumor. Collectively, these results demonstrated that integrative transcriptome and network analysis can facilitate the discovery of functional markers of tumor-specific immune cells to develop novel therapeutic targets and biomarkers for boosting cancer immunotherapy.

Keywords: Context associated hub; Coregulatory network; Functional markers; Tumor microenvironment; Tumor-infiltrating regulatory T cell.

Graphical abstract

Fig. 1

Schematic overview of a systems biology approach to identify functional marker genes of tumor-infiltrating regulatory T cells (TI-Tregs).

Fig. 2

Expression-based marker genes of TI-Tregs in mouse and human. (a) Volcano plots that show differential expression analysis from comparisons of tumor-bearing mouse TI-Treg (TBM-TI-Treg) with five other CD4⁺ T cell types in different tissue contexts: conventional CD4⁺ T cell from tumor-bearing mouse (TBM-TI-Tconv), spleen-derived Treg from tumor-bearing mouse (TBM-SP-Treg) and those from normal mouse (NM-SP-Treg), spleen-derived conventional CD4⁺ T cell from tumor-bearing mouse (TBM-SP-Tconv) and those from normal mouse (NM-SP-Tconv). The final 25 candidate markers by intersection of the comparisons are marked in red. (b) Violin plots that show the expression of six human genes differentially expressed in tumor Treg (TTR), normal Treg (NTR), PBMC Treg (PTR), tumor Tconv (TTH). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Network-based prioritization of functional markers of tumor-infiltrating regulatory T cells (TI-Tregs). (a) Layout of coregulatory network for Tregs (Treg-specific network). Genes associated with rheumatoid arthritis (RA) are represented as red nodes. (b) Within-group connectivity for each group of genes associated with multiple sclerosis (MS), psoriasis, and RA. Within-group connectivity by Treg-specific network and read disease gene sets are represented as red stars. (c) Illustrative summary of context-associated centrality (CAC). (d) Genes sorted by CAC score (-log₁₀(q-value) of enrichment of network neighbors among context-associated signature genes) and six expression-based markers in human TI-Tregs are represented as red points. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4

TNFRSF9 expression level in tumor-infiltrating regulatory T cells (TI-Tregs) is predictive of clinical outcomes in anticancer treatment. (a-b) Kaplan-Meier analysis curves by high (top 30%) and low (bottom 30%) expression levels of FOXP3 (a), expression level of TNFRSF9 normalized by that of FOXP3 (b) in tumor samples from The Cancer Genome Atlas-lung adenocarcinoma (TCGA-LUAD) and TCGA-lung squamous cell carcinoma (LUSC). (c-d) Kaplan-Meier analysis curves by high (top 40%) and low (bottom 40%) expression levels of FOXP3 (a), expression level of TNFRSF9 normalized by that of FOXP3 (b) in tumor samples from Swedish cohort of lung cancer (GSE81089) (e-f) Waterfall plots showing expression changes from baseline (median) for FOXP3 (e) and TNFRSF9 normalized by FOXP3 (f) in tumor samples from patients with non-small cell lung cancer (NSCLC) treated by anti-PD-1 inhibitors (GSE135222).