IFNγ signaling in cytotoxic T cells restricts anti-tumor responses by inhibiting the maintenance and diversity of intra-tumoral stem-like T cells

Abstract

IFNγ is an immune mediator with concomitant pro- and anti-tumor functions. Here, we provide evidence that IFNγ directly acts on intra-tumoral CD8 T cells to restrict anti-tumor responses. We report that expression of the IFNγ receptor β chain (IFNγR2) in CD8 T cells negatively correlates with clinical responsiveness to checkpoint blockade in metastatic melanoma patients, suggesting that the loss of sensitivity to IFNγ contributes to successful antitumor immunity. Indeed, specific deletion of IFNγR in CD8 T cells promotes tumor control in a mouse model of melanoma. Chronic IFNγ inhibits the maintenance, clonal diversity and proliferation of stem-like T cells. This leads to decreased generation of T cells with intermediate expression of exhaustion markers, previously associated with beneficial anti-tumor responses. This study provides evidence of a negative feedback loop whereby IFNγ depletes stem-like T cells to restrict anti-tumor immunity. Targeting this pathway might represent an alternative strategy to enhance T cell-based therapies.

Fig. 1. IFNγ receptor expression in circulating CD8 T cells from metastatic melanoma patients undergoing checkpoint therapy and correlation with progression.

a–d Analysis of IFNγR expression in scRNAseq data of CD8 T cells from 8 metastatic melanoma cancer patients treated with checkpoint blockade (pembrolizumab (n = 4), ipilimumab/nivolumab (n = 4)). Samples were taken before (Day 0, pre) and after (Day 21, post) treatment. a–b Frequency of cells displaying detectable expression of the receptor (left panels) and level of expression (right panels) of IFNγR1 (a) and IFNγR2 (b) in multiple T cell subsets. c–d Proportion of cells displaying detectable expression of the receptor (left panels) and level of expression (right panels) of IFNγR1 (c) and IFNγR2 (d) according to clonal size. For (c, right panel and d), statistics compare smallest clonal size to others. a–d Percent of cells is showed as Mean +/− SEM. Level of expression is showed as box-plot with lower and upper hinges of boxes represent 25th to 75th percentiles, the central line represents the median, and the whiskers extend to highest and lowest values no greater than 1.5× interquartile range. Pairwise group comparisons with two-sided Wilcoxon signed-rank test. e–h Bulk-RNA seq analysis of CD8 T cells from metastatic melanoma patients (n = 110) 21 days after starting checkpoint blockade therapy. e–f GOBP pathway analysis of genes correlated (blue) and inversely correlated (red) with IFNγR1 (e) and IFNγR2 (f) after checkpoint blockade. Overlap (n), number of significant genes from a pathway (hypergeometric test). g- Correlation between disease progression and IFNγR1 (left panels) or IFNγR2 (right panels) expression after first cycle of checkpoint blockade. N = 68 no progression; n = 44 h- Correlation between autoimmune event and IFNγR1 (left panels) or IFNγR2 (right panels) expression after first cycle of checkpoint blockade. N = 46 with autoimmunity; n = 44 (g–h) Lower and upper hinges of boxes represent 25th to 75th percentiles, the central line represents the median, and the whiskers extend to highest and lowest values no greater than 1.5× interquartile range. Two-sided Wilcoxon signed-rank test.

Fig. 2. IFNγ sensing by CD8 T cells restricts anti-tumor immunity in mice.

a–f Control (red) and CD8-IFNγRKO (grey) mice were engrafted with B16-OVA tumors. a- Mouse survival over time (n = 14). Data is from 3 independent experiments. Mantel-Cox test. b- Average tumor volume over time. Data is from 3 independent experiments (n = 14). Data shows Mean +/− SEM. Mixed-effects model analysis with Šidák’s multiple comparison test. c- Tumor volume of individual mice over time. d- Tumor weight between 12 and 15 days. Each point represents one mouse (n = 18 Control; n = 16 CD8-IFNγRKO), from 3 independent experiments. Data shows Mean +/− SEM. Unpaired t test. e–f Analysis of total number of TILs per tumor (e, n = 9) and the number of TILs according to the tumor volume (f, n = 6 Control and n = 5 CD8-IFNγRKO) after 12 to 15 days. Each point represents one individual mouse from 3 independent experiments. Data shows Mean +/− SEM. Unpaired t test. g- TILs were harvested and restimulated in vitro with 1ug OVA peptide. Quantification of IFNγ and TNFα producing cells. Every dot is a mouse (n = 4 Naïve control; n = 8 WT/ CD8-IFNγRKO) from 2 independent experiments. Data shows Mean +/− SEM. Two-way Anova with Šidák’s multiple comparison test. h- IFNγ quantification by Elisa in tumor supernatants from control and CD8-IFNγRKO mice after 12 to 15 days. Every dot is a mouse (n = 16 Control; n = 11 CD8-IFNγRKO) from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. i- TILs were restimulated in vitro with 1ug OVA peptide. Quantification of surface LAMP-1 staining. Each symbol represents an individual mouse (n = 8 Control; and n = 9 CD8-IFNγRKO) from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. j–k B16-OVA tumor sections from Control and CD8-IFNγRKO mice were stained with for CD8 (green). Tumors express mCherry (red) and nuclei were stained with Dapi (blue). The tumor core was defined as mCherry+ Dapi+. The stroma network and margin were defined as mCherry- Dapi+. j- Representative images of tumor sections from control (left panels) and CD8-IFNγRKO (right panels) mice. k- Proportion of CD8 T cells in the core of tumors from Control (red) and CD8-IFNγRKO (grey) mice. Every dot is a mouse (n = 6 Control; n = 5 CD8-IFNγRKO) from 3 independent experiments. Data shows Mean +/− SEM. Unpaired t test.

Fig. 3. Unbiased characterization of CD8 T cells infiltrating B16-OVA tumors.

Control (n = 3) and CD8-IFNγRKO (n = 3) mice were engrafted with B16-OVA tumors. CD45⁺CD3⁺CD8⁺Tomato⁺ cells were sorted from tumors and subjected to scRNA-seq analysis. a- Graph-based clustering (n = 15706 total) identified 5 clusters. b- The violin plots show the expression levels (y axes) of selected exhaustion markers (Pdcd1 (encoding PD-1), Ctla4, Lag3, Tigit, Havcr2 (encoding Tim-3), and Tox) in each of the identified clusters (x axes). c- UMAP visualization of CD8 Stem-like signature. d- UMAP visualization of the distribution of selected transcripts defining the Stem-like gene signature. Genes include Tcf7 (encoding Tcf1), Il7r, Ccr7, IFNγ, Gzmb (encoding granzyme B) and Pdcd1 (encoding PD-1). e- Pseudotime trajectory across the 5 TIL clusters. f- The violin plots show the expression score (y axes) of the exhaustion gene signature in each of the identified clusters (x axes). Clusters have been re-ordered to follow the exhaustion score and pseudo-time. Box plots indicate median (middle line), 25th, 75th percentile (box). (n = 15706 total). g- UMAP visualization of CD8 T cell terminal exhaustion signature. h- The violin plots show the expression score (y axes) of the TCR signaling gene signature in each of the identified clusters (x axes). Box plots indicate median (middle line), 25th, 75th percentile (box). Pairwise group comparisons with two-sided Wilcoxon signed-rank test. (n = 15706 total). i- The dot plot shows the relative expression and percent of cells expressing the transcription factor Tox and the receptor Entpd1 (encoding CD39) within the different clusters.

Fig. 4. IFNγ sensing by CD8 T cells in tumors alters their cell states and TCR clonality.

Control and CD8-IFNγRKO mice were engrafted with B16-OVA tumors. a–b and d–e CD45⁺CD3⁺CD8⁺Tomato⁺ cells were sorted from tumors and subjected to scRNA-seq analysis. a- Graph-based clustering (n = 15,706 total) of the assembled cell states according to the signatures identified in Fig. 4. b- The bar plots show the percentages of CD8 T cells from Control (red) and CD8-IFNγRKO (grey) mice that were found in each cell state. Permutation test. c- Tumors were harvested 10-12 days post-tumor engraftment and cell suspensions were stained for CD45, CD3, PD-1, and TCF1/7 and subjected to flow cytometry. Percentage of TCF1/7⁺ PD-1⁺ CD8 T cells in tumors from Control (red) and CD8-IFNγRKO (grey) mice Each point represents one individual mouse (n = 27 for Ctrl; n = 18 for CD8-IFNγRKO), from 4 independent experiments. Data shows Mean +/− SEM. Unpaired t test. d- Heatmap of control TILs shows the relative average expression of IFNγR1 and IFNγR2 in the different cell states in control mice. The average expression of IFNγR1 and IFNγR2 in CD8-IFNγRKO cells from the exhausted state was used as a control for negative IFNγR1 expression. e- The dot plot shows the relative average expression and percent of cells expressing transcripts implicated in IFNγ signaling and response within the different cell states. f–h CD45⁺CD3⁺CD8⁺Tomato⁺ cells were sorted from tumors and subjected to scRNA-seq and scTCR-seq analysis. f- The relative frequency of the top 20 clones and their distribution throughout the cell states was analyzed within Control and CD8-IFNγRKO mice. Each color corresponds to a different clone. g- Relative TCR clonal abundance by cell state and genotype. Clonotype abundance (as frequency of total repertoire per cell state per genotype) is defined as: Rare = 0 < X < = 0.01%; Small = 0.01 < X < = 0.05%; Medium = 0.005 < X < = 0.1%; Large 0.1 < X < = 0.25%; Hyperexpanded 0.25 < X < = 1%. h- Graph-based clustering of TILs (n = 7000 for each genotype) from control (left) and CD8-IFNγRKO (right) mice overlaid with the frequency of clonotypes. The black line corresponds to the pseudotime trajectory, as in Fig. 3e. Clonotype abundance is defined as in (g).

Fig. 5. IFNγ sensing by CD8 T cells restricts proliferation by targeting stem-like T cells.

a–c Analysis of transcriptomics from Fig. 3. a- Percentages of CD8 T cells from Control (red) and CD8-IFNγRKO (grey) mice in G1, G2/M, or G1 phase of the cell cycle (n = 15706). (b–c) Normalized Enrichment score (NES) of cell cycle-related Hallmark (top) and Reactome (bottom) pathways in Stem-like (b) or Exhausted (c) T cell states. Bars are colored by adjusted p-values, derived from the fgsea R package (d–f) CFSE-labelled total immune cells from tumor-bearing WT mice were restimulated ex vivo with anti-CD3 and anti-CD28 and treated with blocking IFNγ antibody, recombinant IFNγ, or media only (Mock). Data are from 3 independent experiments. Each dot is one well of immune cells pooled from multiple mice. d- Ratio between Tcf1/7+ and Tcf1/7^- TIL ratio over time (n = 7 for Day 0; n = 12 for Day 4.5 mock, n = 9 for Day 4.5 rIFNγ and Day 4.5 anti-IFNγ; n = 8 for Day 7 mock, n = 7 for Day 7 rIFNγ and Day 4.5 anti-IFNγ). Data shows Mean +/− SEM. Mixed-effects model analysis with Tukey’s multiple comparison test. e- Absolute number of Tcf1/7⁺ PD-1⁺ stem-like T cells over time normalized to day 0 (n = 9 for Day 4.5 mock, n = 7 for Day 4.5 rIFNγ and Day 4.5 anti-IFNγ; n = 8 for Day 7 mock, n = 7 for Day 7 conditions). Data shows Mean +/− SEM. Mixed-effects model analysis with Šidák’s multiple comparison test. f- Percentage of stem-like T cells that divided, assessed by CFSE dilution (n = 10 for Day 4.5 mock, n = 9 for Day 4.5 rIFNγ and Day 4.5 anti-IFNγ; n = 8 for Day 7 mock, n = 7 for Day 7 conditions). Data shows Mean +/− SEM. Mixed-effects model analysis with Šidák’s multiple comparison test. g–i Stem-like TILs (orange) and exhausted TILs (violet) from B16-OVA tumors bearing WT mice were labeled with CellTrace violet dye (CTV) and restimulated with anti-CD3 and anti-CD28. g- Representative examples of CTV dilution after 4 days. Control represents TILs at day 0. h- Percentage of dividing cells. Each point represents one individual sample (n = 8 for Stem-like; n = 10 for exhausted), from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. i- Percentage of cells that completely diluted CTV. Each point represents one individual sample (n = 7), from 3 independent experiments. Data shows Mean +/− SEM. Unpaired t test. j–k CTV-labeled exhausted TILs from B16-OVA tumors were restimulated as in (g–i) in the presence (grey) or absence (red) of anti-IFNγ. j- Representative examples of CTV dilution profiles. k- Percentage of exhausted T cells that divided after 4 days. Each point represents one individual sample (n = 7), from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. l–n In vitro-generated stem-like and exhausted IFNγRKO and WT OTI cells were ad-mixed and injected into B16-OVA tumor-bearing mice. l- Experimental set-up, created with BioRender.com. m- Expression of selected exhaustion markers and Tcf1/7 of in vitro differentiated WT and IFNγRKO OTI before transfer. Mock represents unstained OTI T cells. n- Ratio between IFNγRKO and WT OTI from stem-like (orange) or exhausted (violet) conditions in tumors 6 days after transfer. Each point represents one individual mouse (n = 20 for Stem-like; n = 12 for exhausted), from 3 independent experiments. Data shows Mean +/− SEM. Non-parametric Mann-Whitney test.

Fig. 6. IFNγ signaling in intratumoral Stem-like CD8 T cells impairs their maintenance.

a–b; e Analysis of transcriptomics data from Fig. 3. a- NES of chosen Reactome pathways up- and down-regulated in CD8-IFNγRKO vs control Stem-like T cells. Bars are colored by adjusted p-values, derived from the fgsea R package. b- Volcano plot of Differentially Expressed Genes between control and CD8-IFNγRKO Stem-like T cells. Green dots: genes with log2 (fold-change) value >0.5 or <−05; Blue dots: genes with an adjusted p value < 0.05; Red dots: genes with log2 (fold-change) value >0.5 or <−05 and an adjusted p value < 0.05. Total variables = 195. Wilcoxon rank sum test. c- Control (red) and CD8-IFNγRKO (grey) mice were engrafted with B16-OVA tumors. Percentage of apoptotic cleaved Caspase 3⁺ stem-like or exhausted TILs. Each dot represents an independent mouse (n = 9 for Control, n = 6 for CD8-IFNγRKO), from 2 independent experiments. Data shows Mean +/− SEM. Mixed-effects model analysis with Šidák’s multiple comparison test. d- Naïve CD8 T cells were stimulated with anti-CD3, anti-CD28 and IL-2 and treated with anti-IFNγ when indicated. Percentage of cleaved-Caspase 3⁺ apoptotic CD8 T cells. Each dot represents a mouse (n = 8 for Mock and n = 10 for anti-IFNγ) from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. e- Stem-like signature scoring of Control (red) and CD8-IFNγRKO (grey) stem-like cells. Box plots indicate median (middle line), 25th, 75th percentile (box). Pairwise group comparisons with two-sided Wilcoxon signed-rank test (n = 15706 total). f–g In vitro-generated stem-like WT (red) and IFNγRKO (grey) OTI cells were injected at a 1:1 ratio in tumor-bearing mice as described in Fig. 5L. Percentage of PD-1⁺ Tcf7⁺ OTI in tumors (f) and in draining lymph nodes (g) after 6 days. Lines link cells from the same mouse. Each dot represents a mouse (n = 9) from 3 independent experiments. Data shows Mean +/− SEM. Paired t test. h- stem-like TILs were sorted and restimulated with anti-CD3 and anti-CD28 and treated with anti-IFNγ when indicated. Percentage of PD-1⁺ Tcf7⁺ cells after 4 days. Each dot is a mix of 5 mice (n = 5 for Control^; n = 3 for CD8-IFNγRKO) from 3 independent experiments. Data shows Mean +/− SEM. Paired t test. i- Correlation between the percentage of IFNγR1 expressing stem-like T cells and Stem-like signature scoring in transcriptomics data from human tumor patients. R = Pearson’s correlation coefficient; Fisher’s test. Each dot is a patient (left n = ; right n = 14).

Fig. 7. IFNγ signaling in TILs is not sufficient to alter their metabolism but reinforces exhaustion in vivo.

a–d; g–i Analysis of transcriptomics data from Fig. 3. a–b NES of chosen KEGG (a) and Hallmark (b) pathways differentially regulated in CD8-IFNγRKO vs control Exhausted TILs. Bars are colored by adjusted p-values, derived from the fgsea R package. c–d Violin plots show Glycolysis scoring (c) and Oxphos scoring (d) of Control (red) and CD8-IFNγRKO (grey) TILs within each cell state. Box plots indicate median (middle line), 25th, 75th percentile (box). Pairwise group comparisons with two-sided Wilcoxon signed-rank test. (n = 15706 total). e–f Metabolic analysis of PD-1⁺ Lag-3⁺ TILs from Control and CD8-IFNγRKO mice engrafted with B16-OVA. Each dot represents a mouse (Control n = 17 and CD8- IFNγRKO n = 13) from 3 independent experiments. Data shows Mean +/− SEM. Unpaired t test. e- Ratio between MitoTracker Deep red (MDR) and MitoTracker green (MDG). f- BEC index (ratio between ATPB and GADPH expression). g- Heatmap shows the relative average expression of selected exhaustion markers in control and CD8-IFNγRKO TILs over the different cell states. h- Volcano plot representing Differentially Expressed Genes between control and CD8-IFNγRKO TILs in Exhausted T state. Blue dots: genes with an adjusted p value<0.05; Red dots: genes with log2 (fold-change) value >0.5 or < −05 and an adjusted p value<0.05. Total variables = 543. Wilcoxon rank sum test. i- Enrichment Score of PD-1 (upper panel) and CTLA-4 (lower panel) between CD8-IFNγRKO and control TILs in the Exhausted state. j–m Phenotypic analysis of PD-1⁺ Lag-3⁺ TILs from Control and CD8-IFNγRKO mice engrafted with B16-OVA. Each dot represents a mouse from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. j- Representative flow cytometry plots with gate highlighting a population with intermediate expression (mid) of PD-1, Tim-3, Lag-3, and Tigit. k–m Percentage of TILs with intermediate expression of PD-1 and Tim-3 (k; Control n = 16 and CD8-IFNγRKO n = 12), Lag-3 (l; Control n = 16 and CD8-IFNγRKO n = 12) and Tigit (m; Control n = 15 and CD8-IFNγRKO n = 13). Each dot represents a mouse from 3 independent experiments. Data shows Mean +/− SEM. Unpaired t test.

Fig. 8. IFNγR ablation in CD8 T cells improves tumor control in a T cell transfer model.

a–c WT mice bearing B16-OVA were transferred with Control (red), IFNγRKO (grey) OTI T cells or not transferred (green). a- Mouse survival over time (n = 7 control, n = 17 WT and n = 18 IFNγRKO). Data is from 3 independent experiments. Mantel-Cox test. b- Tumor volume of individual mice over time (n = 15). Black square highlight tumors with a volume of at least 300 mm³. Data is from 3 independent experiments. c- Average tumor weight 7 days post-OTI transfer. Each dot is a mouse (n = 9 WT and n = 10 IFNγRKO) from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. d- WT mice bearing B16-OVA were either co-transferred with Control and IFNγRKO OTI T cells. Ratio between the number of IFNγRKO and WT OTI cells in Lymph nodes and Tumors after 5 days. Each dot is a mouse (n = 4) from 2 independent experiments. Data shows Mean +/− SEM. One sample t and Wilcoxon test. e- WT and IFNγRKO OTI were stimulated in vitro with BMDCs loaded with N4, T4, or V4 peptides. Percentage of CD69⁺ OTI cells after 24 h. Each dot is an independent sample (n = 5 WT-naive, n = 6 WT-T4/V4, IFNγRKO-naïve/T4/V4; n = 18 WT-N4 IFNγRKO-N4) from 2 experiments. Data shows Mean +/− SEM. Mixed-effects model analysis with Šidák’s multiple comparison test. f–g WT mice bearing B16-OVA tumors were transferred with Control (red) and IFNγRKO (grey) OTI T cells. Intra-tumoral OTI cells were restimulated in vitro with 1ug OVA peptide. f- Representative plot showing IFNγ and TNFα production. g- Quantification of IFNγ and TNFα production by OTI WT (red), OTI IFNγRKO (grey), and non-stimulated OTI control (black). Each dot is an independent sample (n = 9 WT, n = 6 IFNγRKO) from 3 independent experiments. Data shows Mean +/− SEM. Mixed-effects model analysis with Šidák’s multiple comparison test. h–j WT mice bearing B16-OVA tumors were transferred with Control (red) and IFNγRKO (grey) OTI T cells. h–i In vivo cytotoxic assay was performed after 12 days. h- Representative histogram of CFSE labelled target cells after 24 h. i- Quantification of cell lysis. Each dot is an independent sample (n = 6 WT, n = 8 IFNγRKO) from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test. j- Lamp-1 expression following restimulation. Each dot is an independent sample (n = 8 WT, n = 8 IFNγRKO) from 2 independent experiments. Data shows Mean +/− SEM. Unpaired t test.