Tumor immunogenicity dictates reliance on TCF1 in CD8+ T cells for response to immunotherapy

Abstract

Stem-like CD8⁺ T cells are regulated by T cell factor 1 (TCF1) and are considered requisite for immune checkpoint blockade (ICB) response. However, recent findings indicate that reliance on TCF1⁺CD8⁺ T cells for ICB efficacy may differ across tumor contexts. We find that TCF1 is essential for optimal priming of tumor antigen-specific CD8⁺ T cells and ICB response in poorly immunogenic tumors that accumulate TOX⁺ dysfunctional T cells, but is dispensable for T cell priming and therapy response in highly immunogenic tumors that efficiently expand transitory effectors. Importantly, improving T cell priming by vaccination or by enhancing antigen presentation on tumors rescues the defective responses of TCF1-deficient CD8⁺ T cells upon ICB in poorly immunogenic tumors. Our study highlights TCF1's role during the early stages of anti-tumor CD8⁺ T cell responses with important implications for guiding optimal therapeutic interventions in cancers with low TCF1⁺CD8⁺ T cells and low-neo-antigen expression.

Keywords: T cell dysfunction; T cell fitness; T cell priming; TCF1 stem-like T cells; checkpoint blockade; tumor immunology; tumor-draining lymph node; vaccine.

Figure 1.. TCF1 is dispensable for ICB efficacy in highly but not poorly immunogenic tumors.

(A) Representative plots and frequency of the indicated populations in the thymus of WT or TCF1 cKO mice (n=4, one experiment). (B) Frequency of TCF1⁺ cells in the indicated populations in WT or TCF1 cKO mice from (A). Student’s t test, ****p<0.0001 (C) Representative histograms and frequency of MHC-I-SIINFEKL expression in B16-OVA or MC38-OVA tumors 14 days post implant (n=4, one experiment). (D) Experimental scheme. ICB: anti-PD-L1 + anti-TIM-3; ISO: IgG2a + IgG2b (E) Tumor area over time in WT and TCF1 cKO mice treated as in (D) (n=4–7 per group. B16-OVA: shown one out of 10 experiments; MC38-OVA: shown one out of 7 experiments). Two-way ANOVA and Sidak’s multiple comparisons test, *p<0.05, **p<0.01, ***p < 0.001, ****p<0.0001. (F) Gating strategy and marker expression in the indicated subsets of CD8⁺ TILs in MC38-OVA and B16-OVA tumors. (G-H) Frequency (top panel) of the TIL subsets defined in (F) and of TCF1-expressing cells (bottom panel) within total (G, n=11, two experiments combined,) or OVA-dextramer⁺ CD8⁺ TILs (H, n=5–15, two experiments combined) in MC38-OVA or B16-OVA tumors, 14 days post tumor implant into WT mice. Student’s t test, **p<0.01, ***p<0.001, ****p<0.0001. All values are reported as mean±SEM. See also Figure S1.

Figure 2.. Defective proliferation of TCF1 cKO T cells in the TDLN of poorly immunogenic tumors.

(A) Experimental scheme. ICB: anti-PD-L1 + anti-TIM-3; ISO: IgG2a + IgG2b. (B-D) Representative plots and frequency of proliferating WT or TCF1 cKO OTI T cells in the TDLN of mice implanted with (B) B16-OVA (n=6–7, two experiments combined out of 5 experiments), (C) MC38-OVA (n=3–5, one out of 3 experiments shown), or (D) a 1:5 mixture of MC38-OVA + MC38 tumor cells (n=4–5, one experiment) and treated as shown in (A). (E) Experimental scheme. ICB: anti-PD-L1 + anti-TIM-3; ISO: IgG2a + IgG2b. (F) Chimerism of WT and TCF1 cKO OTI T cells before and 3 days after adoptive transfer. (G) Frequency of proliferating WT or TCF1 cKO OTI T cells in the TDLN of mice treated as shown in (E) (n=6, one experiment). In Figure 2B–D and 2G, one-way ANOVA is used to calculate the statistics related to the total frequency of proliferating cells and within each cell division (all comparisons are done versus WT+ICB group). *p<0.05, **p<0.01. All values are reported as mean±SEM. See also Figure S2.

Figure 3.. TCF1 lowers the activation threshold and poises T cells for optimal responses to low TCR triggering.

(A) WT and TCF1 cKO T cells were activated with a high (1ug/ml each, n=3, shown one out of 3 experiments) or low (0.3ug/ml each, n=5–6, two experiments combined) dose of anti-CD3 + anti-CD28 for 48 hrs. Representative histograms and frequency of WT and TCF1 cKO OTI T cells in each cell division three days post activation. Student’s t test, *p<0.05, **p<0.01, ***p<0.001. (B) Frequency of PD-1 and CD28 positive cells in WT and TCF1 cKO OTI T cells activated as in (A). Two-way ANOVA and Sidak’s multiple comparison test, *p,0.05, **p<0.01, ***p<0.001, ****p<0.0001. n=3, one experiment. (C) Level (MFI) of phospho-AKT, phospho-LAT, phospho-SHP2, phospho-ERK and phospho-ZAP70 in WT and TCF1 cKO OTI T cells, 24 hours post in vitro activation with anti-CD3 + anti-CD28 (1ug/ml each). n=3, shown one out of 2 experiments. (D) Representative histograms and frequency of WT and TCF1 cKO OTI T cells in each cell division three days post co-culture with BM-DC pulsed with a high (17.78nM) or low (3pM) dose of SIINFEKL. The phenotype of WT and TCF1 cKO T cells at day three is shown. Student’s t test, *p<0.05, **p<0.01. n=3, one experiment. (E) Frequency of WT and TCF1 cKO OTI T cells in each cell division three days post co-culture with BM-DC pulsed with a high (17.78nM) dose of SIIVFEKL (V4). n=3, one experiment. (F) Phenotype and frequency of PD-1 positive cells in WT and TCF1 cKO T cells from (E). (G) Expression of mScarlet-SIIN or mScarlet-SIIQ in MC38 tumor cells engineered with the indicated lentiviral vectors. (H) Frequency of proliferating WT or TCF1 cKO OTI T cells in the TDLN three days post transfer in mice implanted with MC38-mScarlet-SIIN or MC38-mScarlet-SIIQ tumors. n=8–10 mice, one experiment. All values are reported as mean±SEM. See also Figures S3 and S4.

Figure 4.. Defective activation of TCF1 cKO OTI T cells in the TDLN of poorly but not highly immunogenic tumors.

(A) Combined Uniform Manifold Approximation and Projection (UMAP) of single-cell transcriptomes from WT (n=3–5) and TCF1 cKO OTI T cells (n=3–5) sorted from the TDLN of B16-OVA- or MC38-OVA-bearing mice treated as shown in Figure 2A. Left, cluster annotation. Right, single-cell density. (B) Dot plot showing expression of cluster defining genes. Circle size indicates the percentage of cells expressing the gene in each cluster. Color scale indicates average gene expression level. (C) Pseudotime trajectory of single-cells (starting node: naïve cell cluster). Left, pseudotime projected on the UMAP. Right, expression profiles of selected genes along the pseudotime. (D) Single-cell density in WT or TCF1 cKO OTI T cells sorted from the TDLN of MC38-OVA- or B16-OVA-bearing mice treated as in Figure 2A. (E) Frequency of single-cells in each cluster and condition analyzed.

Figure 5.. TCF1 cKO T cells contain a higher proportion of CXCR3⁺Ly6C⁺ naïve cells poised to become short-lived effectors.

(A) UMAP of single-cell transcriptomes from naive WT and TCF1 cKO OTI T cells from Figure 4A. (B) Dot plot showing cluster defining genes. Circle size indicates the percentage of cells expressing the gene in each cluster. Color scale indicates average gene expression level. (C) Frequency of single-cells in each naïve cell cluster expressed as percentage of the overall naïve population in each condition analyzed. (D) Heatmap showing the inferred activity (red, active; blue, inactive) of 34 selected regulons (i.e. regulons relevant in stem-like CD8⁺ T cells) in the naïve cell clusters. Single cells are ordered in columns and grouped by cluster. Rows represent regulons named by the relative transcription factor. For each regulon, the number of the downstream target genes co-expressed with the indicated transcription factor is given. (E) Representative plots and frequency of the three naïve subsets defined using CXCR3 and Ly6C markers among naïve (CD62L⁺CD44⁻) T cells expressing either a OTI (n=8–9 mice, three experiments combined) or a native (n=6 mice, two experiments combined) TCR repertoire from tumor-free WT or TCF1 cKO mice. Student’s t test, ***p<0.001, ****p<0.0001. (F) Representative histograms and frequency of EOMES⁺ cells in the naïve subsets defined in (E) from tumor-free WT or TCF1 cKO mice (n=6 mice, two experiments combined). One-way ANOVA, *p<0.05, **p<0.01, ****p<0.0001 (G) Frequency of EOMES⁺ cells in the naïve subsets defined in (E) and present in adoptively transferred WT or TCF1 cKO OTI T cells harvested from the TDLN of MC38-OVA-bearing mice (n=8, shown one out of two experiments) three days post transfer. One-way ANOVA, ****p<0.0001. (H) Frequency of CXCR3⁺Ly6C⁺ naïve WT or TCF1 cKO OTI T cells harvested from the TDLN of MC38-OVA-bearing mice three days post adoptive transfer. Student’s t test, **p<0.01. (n=4, shown one out of two experiments). (I) Frequency of CXCR6⁺ cells in naive (CTV^highCD62L⁺CD44⁻ ) WT or TCF1 cKO T cells harvested from the TDLN of MC38-OVA-bearing mice (n=4, shown one out of two experiments) three days post transfer. Student’s t test, **p<0.01, ****p<0.0001. All values are reported as mean±SEM. See also Figure S5 and Table S1.

Figure 6.. Poorly versus highly immunogenic tumors instruct distinct CD8⁺ T cell fates.

(A) UMAP of single-cell transcriptomes from CD8⁺ TILs harvested from WT or TCF1 cKO mice (n=3–5) inoculated with B16-OVA or MC38-OVA tumors and treated with two doses of anti-PD-L1 + anti-TIM-3 (ICB) or control antibodies (ISO) at days 6 and 9 post tumor inoculation. Left, cluster annotation. Right, single-cell density. (B) Dot plot showing cluster defining genes. Circle size indicates the percentage of cells expressing the gene within each cluster. Color scale indicates average gene expression level. (C) Density of single-cell events in WT or TCF1 cKO CD8⁺ TILs from MC38-OVA- or B16-OVA-bearing mice treated as indicated. (D) Frequency of single-cells in each cluster from (A) expressed as percentage of the overall population in WT or TCF1 cKO CD8 TILs from B16-OVA or MC38-OVA tumors treated as indicated. (E) Projection of a transitory effector signature and of a signature comprising transcription factors enriched in dysfunctional cells from melanoma patients on the UMAP from (A). (F) Pseudotime trajectory of single cells (starting node: naïve/stem-like cell cluster). Cells are colored along the trajectory based on a blue-to-yellow color gradient. See also Figure S6 and Tables S2 and S3.

Figure 7.. TOX⁺ dysfunctional CD8⁺ TILs from B16-OVA-bearing TCF1 cKO mice have a destabilized phenotype resembling cells from patients that fail to respond to ICB.

(A) Tox expression on the UMAP from Figure 6A. (B) TOX expression (frequency, mean±SEM) in PD-1⁺TIM-3⁺ CD8⁺ TILs in MC38-OVA- or B16-OVA-bearing WT or TCF1 cKO mice (n=4–6, one experiment) treated with anti-PD-L1 + anti-TIM-3 (ICB) or control (ISO) antibodies at day 6 and 9 post tumor injection. Analysis is done on day12. Student’s t test, *p<0.05. (C) Heatmap showing differentially expressed genes between WT and TCF1 cKO cells within the proliferating TOX⁺PD-1⁺TIM-3⁺CD8⁺ TIL (cluster 8) from Figure 6A in B16-OVA tumors treated with ICB at day 6 and 9 post tumor inoculation. (D) CASPASE-3 expression (frequency, mean±SEM) in PD-1⁺CD8⁺ TILs from B16-OVA-bearing WT or TCF1 cKO mice, three days after the second dose of ISO/ICB (n=6–13 mice, two experiments combined). ANOVA, ***p<0.001 (comparison is done vs WT ICB group). (E) Expression of the indicated signature on proliferating TOX⁺TIM-3⁺PD-1⁺CD8⁺ TILs (cluster 8) from figure 6A in B16-OVA tumors in WT or TCF1 cKO mice treated with two doses of ICB at day 6 and 9 post tumor inoculation. Mann-Whitney test, ****p<0.0001. (F) Violin plot of Jun expression in the naïve/stem-like T cell cluster (cluster 1) from Figure 6A from WT or TCF1 cKO TILs in B16-OVA- and MC38-OVA-bearing mice treated with ICB at day 6 and 9 post tumor inoculation. Mann-Whitney test, ****p<0.0001. Standard violin plot (G) Expression of the TCF1 cKO signature in CD8⁺ TILs from post-treatment melanoma patients treated as indicated and stratified as responder (R) or non-responder (NR). Dots represent individual cells. (H) Left: True positive (y axis) and true negative (x axis) rates when predicting the clinical response based on expression of the TCF1 cKO signature in pre-treatment CD8⁺ TILs from melanoma patients subsequently treated with anti-PD-1. Right: expression of the TCF1 cKO signature in CD8⁺ TILs from pre-treatment melanoma patients stratified as responders (R, n=4) and non-responders (NR, n=8). Dots represent individual cells. For box plot: Bottom, middle and upper lines denote lower quartile (Q1), median and upper quartile (Q3), respectively. Whiskers denote minimum (Q1 − 1.5 × interquartile range) and maximum (Q3 + 1.5 × interquartile range). See also Figure S6 and Tables S3–S5.

Figure 8.. Enhancing antigen presentation on melanoma cells or T cell priming by therapeutic vaccination can rescue defective ICB responses in TCF1 cKO mice.

(A) Representative histograms of MHC-I (H-2K^b/H-2D^b) expression on B16-F10 or B16-K1 cells. (B) Frequency (top panel) of the TIL subsets defined in Figure 1F and of TCF1 positive cells (bottom panel) in these CD8⁺ TIL subsets present in B16-F10 or B16-K1 tumors. Analysis is done at comparable tumor dimensions (B16-F10: day 13; B16-K1 day 17 post tumor implant into WT mice). Student’s t test, *p<0.05, ****p<0.0001, n=6, one experiment. (C) Experimental scheme. (D) Tumor area over time in WT and TCF1 cKO mice treated as shown in (C) (n = 5–6 mice per group; B16: one experiment. B16K1: shown one out of two experiments). Two-way ANOVA, Sidak’s multiple comparisons test, *p<0.05, **p<0.01, ***p < 0.001, ****p<0.0001. (E) Experimental scheme. (F) Tumor area over time in TCF1 cKO mice treated as shown in (E) (n=5–6, one experiment). Two-way ANOVA, Sidak’s multiple comparisons test, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. (G-H) Frequency of total (G) or OVA-specific (H) CD8⁺ TILs present in tumors from mice in (F). One-way ANOVA. **p<0.01; ****p<0.0001 (I) Frequency of stem-like cells (population 1 as in Figure 1F: PD-1⁺Slamf6⁺CX3CR1⁻) and effector cells (population 2 as in Figure 1F: PD-1⁺Slamf6^low/negCX3CR1⁺) within total CD8⁺ TILs in tumors from mice in (F). ICB: anti-PD-L1 + anti-TIM-3; ISO: IgG2a + IgG2b. All values are reported as mean ± SEM. See also Figure S7.