Glucose deprivation and identification of TXNIP as an immunometabolic modulator of T cell activation in cancer

Abstract

Background: The ability of immune cells to rapidly respond to pathogens or malignant cells is tightly linked to metabolic pathways. In cancer, the tumor microenvironment (TME) represents a complex system with a strong metabolism stress, in part due to glucose shortage, which limits proper T cell activation, differentiation and functions preventing anti-tumor immunity.

Methods: In this study, we evaluated T cell immune reactivity in glucose-restricted mixed lymphocyte reaction (MLR), using a comprehensive profiling of soluble factors, multiparametric flow cytometry and single cell RNA sequencing (scRNA-seq).

Results: We determined that glucose restriction potentiates anti-PD-1 immune responses and identified thioredoxin-interacting protein (TXNIP), a negative regulator of glucose uptake, as a potential immunometabolic modulator of T cell activation. We confirmed TXNIP downregulation in tumor infiltrating T cells in cancer patients. We next investigated the implication of TXNIP in modulating immune effector functions in primary human T cells and showed that TXNIP depletion increased IFN-γ secretion and tumor cell killing.

Conclusions: TXNIP is at the interface between immunometabolism and T cell activation and could represent a potential target for immuno-oncology treatments.

Keywords: T cells activation; TXNIP; cancer immunotherapy; glucose deprivation; mixed lymphocyte reaction; single-cell RNA-sequencing; tumor microenvironment.

Figure 1

Impact of glucose deprivation on CD4⁺ T cell immunoreactivity. (A) Schematic representation of the MLR protocol for glucose deprivation study. (B) Representative flow cytometry UMAP dimensionality reduction representation of CD4⁺ T cells upon MLR stimulation colored by non-supervised clustering (Donor 1, n=1). (C) Box plots of CD4⁺ T cells proportions in Tregs (left panel), highly reactive (middle panel), or non-reactive (right panel) clusters, displaying group of numerical data through their 3^rd and 1^st quantiles (box), median (central band), minimum and maximum (whiskers) (n=4). HG, High Glucose (11 mM); LG, Low Glucose (1mM).

Figure 2

In vitro glucose deprivation upon anti-PD-1 stimulation impacts both CD4⁺ T cell immunoreactivity and immunometabolism. (A) Heatmap of the log2 fold change (FC) of each cytokine and chemokine concentration in anti-PD-1-treated MLR over untreated MLR (Medium) (n=4). (B) Heatmap of the mean fluorescent intensity (MFI) of CD25 and CTLA-4 per conditions and clusters upon MLR stimulation (n=4). (C) UMAP representation of in vitro scRNA-seq data colored by cell phenotype (n=1). (D) Barplot of cell proportions per condition colored by main cell phenotype (n=1). (E, F) Violin plot of glycolysis score per main cell phenotype (E), and condition (F) (n=1). Statistical analyses for E and F: ANOVA with Tukey’s multiple comparisons tests. *p-value < 0.05, **p-value < 0.01 and ***p-value < 0.001 between indicated groups. HG, High Glucose (11 mM); LG, Low Glucose (1mM).

Figure 3

TXNIP is down-regulated both upon CD4⁺ T cell activation and glucose deprivation in MLR. (A) Venn diagram of up-regulated (left panel) and down-regulated (right panel) genes in low glucose (LG) compared to high glucose (HG) for the indicated stimulation conditions (n=1). (B) Venn diagram of up-regulated (left panel) and down-regulated (right panel) genes in low glucose (LG) for the indicated stimulation conditions comparisons. For A and B, genes exhibiting an absolute log2 fold change value superior to 0.25 and a Bonferroni-adjusted p-value inferior to 0.05 were considered significantly up- or down-regulated (n=1). (C) UMAP representation of TXNIP gene expression (n=1). (D-F) Heatmap of TXNIP mean gene expression per main cell phenotype (D), MKI67-IL2RA co-expression status (E), and conditions (F) (n=1). (G) Representative immunoblot of TXNIP and actin expression under high or low glucose concentration. (H) Barplot of the corresponding TXNIP expression over actin quantification (n=5). HG, High Glucose (11 mM); LG, Low Glucose (1 mM).

Figure 4

In patients TXNIP is highly expressed in lymphocytes. Heatmap of TXNIP mean gene expression computed as log(TPM/10 + 1) across nearly 2 million single cells in total, per cell phenotype (as rows) and dataset (as columns) in TISCH database. All metadata available in TISCH for each dataset was included in annotation on top of the heatmap.

Figure 5

In patients TXNIP expression level decreases in tumor-infiltrating CD4+ T cells. (A) UMAP representation of the pan-cancer scRNA-seq CD4⁺ T cell atlas colored by cell phenotype. (B, C) Violin plot of glycolysis score within the pan-cancer scRNA-seq CD4⁺ T cell atlas per main cell phenotype (B), and sample type (C). Statistical analyses: ANOVA with Tukey’s multiple comparisons tests. *p-value < 0.05, **p-value < 0.01 and ***p-value < 0.001 between indicated groups. (D) UMAP representation of TXNIP RNA expression within the pan-cancer scRNA-seq CD4⁺ T cell atlas. (E) Heatmap of TXNIP mean gene expression per sample type within the pan-cancer scRNA-seq CD4⁺ T cell atlas.

Figure 6

TXNIP knockout in immune T cells. (A) Schematic representation of the in vitro assay protocol for TXNIP functional evaluation. (B) Representative immunoblot of TXNIP and actin expression upon TXNIP-targeting sgRNA in MLR donors: WT (scramble sgRNA), KO_1 (sgTXNIP1), KO_2a (sgTXNIP2a), KO_2b (sgTXNIP2b). (C) Barplot of the corresponding TXNIP expression over actin quantification (n=6). (D) Box plot of predicted percentage of frameshift for the indicated TXNIP-targeting sgRNA compared to WT determined by Sanger sequencing for MLR samples (n=6). (E) Representative immunoblot of TXNIP and actin expression upon TXNIP-targeting sgRNA in TCR donors: WT (scramble sgRNA), KO_2b (sgTXNIP2b). (F) Barplot of the corresponding TXNIP expression over actin quantification (n=4). (G) Box plot of predicted percentage of frameshift for KO_2b compared to WT determined by Sanger sequencing for TCR killing assay samples (n=4).

Figure 7

TXNIP knockout increases T cell immune responses. (A) Box plots of indicated soluble factor concentrations in untreated MLR (Medium), displaying group of numerical data through their 3^rd and 1^st quantiles (box), median (central band), minimum and maximum (whiskers) (n=6). Statistical analyses: Wilcoxon or T test, p-value is considered significantly relevant when p< 0.05 for the corresponding soluble factor. (B) Heatmap of the log2 concentration of the indicated cytokines and chemokines in untreated MLR (Medium) (n=6). (C) Box plots of the mean fluorescent intensity (MFI) or percentage (%) for the indicated markers in untreated MLR (Medium), displaying group of numerical data through their 3^rd and 1^st quantiles (box), median (central band), minimum and maximum (whiskers) (n=5). Statistical analyses: Wilcoxon or T test, p-value is considered significantly relevant when p<0.05 for the corresponding soluble factor. (D, E) Red Area Per Well of NY-ESO-1⁺ A375 tumor cells was plotted over time alone or after addition of NT (non-transduced), NY WT (TCR only) and NY KO (TCR with TXNIP KO) T cells for donor 1 (D1) (D) or donor 2 (D2) (E).