Trawling for drivers of T cell exhaustion
Although checkpoint blockade has markedly changed the landscape of cancer therapeutics, not all tumors are responsive to immunotherapy. Here, Kim et al. studied T cell exhaustion in two distinct types of colorectal cancers (CRCs), microsatellite instability–high (MSI) CRCs that are responsive to PD-1 blockade and microsatellite stable (MSS) CRCs that are resistant to PD-1 centric therapies. Although tumor-infiltrating T cells in both types of CRCs were exhausted, they report T cell exhaustion in MSS but not MSI CRCs to be driven by VEGF-A. Using mouse models, they demonstrate that combined blockade of both PD-1 and VEGF-A makes MSS CRCs sensitive to PD-1 blockade. Their study highlights how in-depth characterization of checkpoint-resistant tumors can lead to identification of pathways to sensitize them to immunotherapy.
Abstract
Although immune checkpoint blockade therapies have demonstrated clinical efficacy in cancer treatment, harnessing this strategy is largely encumbered by resistance in multiple cancer settings. Here, we show that tumor-infiltrating T cells are severely exhausted in the microsatellite stable (MSS) colorectal cancer (CRC), a representative example of PD-1 blockade–resistant tumors. In MSS CRC, we found wound healing signature to be up-regulated and that T cell exhaustion is driven by vascular endothelial growth factor-A (VEGF-A). We report that VEGF-A induces the expression of transcription factor TOX in T cells to drive exhaustion-specific transcription program in T cells. Using a combination of in vitro, ex vivo, and in vivo mouse studies, we demonstrate that combined blockade of PD-1 and VEGF-A restores the antitumor functions of T cells, resulting in better control of MSS CRC tumors.
INTRODUCTION
Tumor antigen–specific T cells within the tumor microenvironment are exhausted, and their antitumor functions are substantially impaired (1, 2). Chronic T cell receptor (TCR) stimulation by persistent antigen exposure results in T cell exhaustion during chronic viral infections and cancer (3, 4). Exhausted T cells progressively lose cardinal features of effector T cells, such as effector cytokine production, cytotoxicity against target cells, and proliferative capacity (5, 6). Further, they express multiple immune checkpoint receptors—such as programmed cell death protein–1 (PD-1), cytotoxic T lymphocyte–associated protein–4 (CTLA-4), T cell immunoglobulin and mucin–domain containing–3 (TIM-3), lymphocyte activation gene–3 (LAG-3), and T cell immunoreceptor with Ig and ITIM domain (TIGIT) (7, 8)—and exhibit transcriptional profiles associated with T cell exhaustion (9–12).
Currently, anti–PD-1, anti–programmed cell death ligand protein–1 (PD-L1), and anti–CTLA-4 antibodies have been developed to reinvigorate exhausted T cells in patients with cancer by interfering with the binding of immune checkpoint receptors and their ligands (13). Treatment with immune checkpoint inhibitors has become a part of standard-of-care treatment in various types of cancer (14–17). For example, PD-1 blockade showed durable clinical benefit in mismatch repair–deficient or microsatellite instability–high (MSI) colorectal cancer (CRC) (18–20). Nevertheless, patients with mismatch repair–proficient or microsatellite stable (MSS) CRC, which make up the majority of patients with CRC (about 80 to 90% CRC), do not respond to PD-1 blockade alone (response rate has been reported to be 0%) (14, 18, 21–23), limiting the clinical applicability of checkpoint blockade therapy and necessitating the development of alternative strategies to benefit this group of patients. Elucidating the mechanisms’ underlying resistance to checkpoint blockade in MSS CRC should remain an important goal for improving patient outcomes (24, 25).
Marked response to anti–PD-1 therapy in MSI CRC can be explained by high levels of tumor mutation burden observed in MSI CRC (18–20). Although MSS CRC has lower levels of tumor mutation burden than MSI CRC, tumor mutation burden alone cannot explain the lack of response to anti–PD-1 therapy in MSS CRC (26, 27). The objective PD-1 blockade response rates of other types of cancer with similar levels of tumor mutation burden as observed in MSS CRC, including Merkel cell carcinoma, cervical cancer, hepatocellular carcinoma, endometrial cancer, and breast cancer, range from 5.7 to 61.7% (27). Therefore, we hypothesized that factors other than tumor mutation burden are responsible for a lack of response to immune checkpoint blockade in MSS CRC.
Here, we examined the key mechanisms underlying T cell exhaustion and resistance to checkpoint blockade in patients with MSS CRC. We observed severe exhaustion of tumor-infiltrating T cells in MSS CRC and a coincident increase in vascular endothelial growth factor-A (VEGF-A), which induced exhaustion-related transcriptional programming in human T cells in a TOX-dependent manner. Inhibition of either VEGF-A or TOX reinvigorated T cells, and we demonstrated that combined blockade of PD-1 and VEGF-A effectively restores the antitumor functions of T cells in MSS CRC.
RESULTS
Tumor-infiltrating CD8+ T cells from MSS CRC are more exhausted than those from MSI CRC
First, we compared the size of T cell infiltrates into tumors between MSS and MSI CRCs by counting the number of CD3+ or CD8+ cells in semiquantitative immunohistochemical analyses. MSS CRC tissues had significantly fewer CD3+ and CD8+ cells at both the invasive margin and the tumor center than MSI CRC (Fig. 1, A to D). Flow cytometry analysis also showed that the relative numbers of CD3+ or CD8+ T cells compared with tumor cells were higher in MSI CRC compared with MSS CRC (fig. S1, A and B). Next, we examined the exhaustion status of T cells in patients with CRC by flow cytometry. The percentage of PD-1high, TIM-3+, LAG-3+, or TIGIT+ cells among tumor-infiltrating CD8+ T cells was significantly higher in tumors than in peripheral blood or adjacent normal mucosa (fig. S2), indicating that T cell exhaustion is limited to the tumor microenvironment. When the expression of immune checkpoint inhibitory receptors on tumor-infiltrating CD8+ T cells was evaluated, the percentage of PD-1high, TIM-3+, LAG-3+, or TIGIT+ cells among tumor-infiltrating CD8+ T cells was significantly higher in MSS CRC than in MSI CRC (Fig. 1E). However, the percentage of PD-1high, TIM-3+, LAG-3+, or TIGIT+ cells among peripheral blood CD8+ T cells or adjacent normal mucosa CD8+ T cells was not different between patients with MSS and MSI CRC (fig. S3, A and B). We also tried to substantiate these findings in a tumor-specific context using tumor-associated cancer-testis antigen NY-ESO-1, which was significantly up-regulated in tumor tissues compared with normal adjacent tissues (fig. S4). To this end, we examined the percentage of immune checkpoint receptor–expressing cells among NY-ESO-1157-165 human leukocyte antigen (HLA)–A2 multimer+ CD8+ T cells in HLA-A2(+) patients. Although the frequency of NY-ESO-1157-165–specific tumor-infiltrating CD8+ T cells was not different between the two groups (Fig. 1F), the percentage of PD-1high, TIM-3+, LAG-3+, or TIGIT+ cells among NY-ESO-1157-165–specific tumor-infiltrating CD8+ T cells was significantly higher in MSS CRC than in MSI CRC (Fig. 1G). We also examined whether the function of tumor-infiltrating CD8+ T cells is severely impaired in MSS CRC by analyzing anti-CD3–induced proliferation and cytokine production. The tumor-infiltrating CD8+ T cells from MSI CRC exhibited a high proliferative capacity, whereas those from MSS CRC exhibited diminished proliferation (Fig. 1H). Moreover, tumor-infiltrating CD8+ T cells from MSS CRC produced interferon-γ (IFN-γ) and tumor necrosis factor (TNF) to a lesser extent than those from MSI CRC (Fig. 1I and fig. S5). Together, these results indicate that MSS CRC tumors encompass fewer tumor-infiltrating CD8+ T cells with more severe exhaustion than MSI CRC tumors.
(A to D) The numbers of tumor-infiltrating CD3+ (A and B) and CD8+ (C and D) T cells were analyzed by immunohistochemical staining at the invasive margin (A and C) and tumor center (B and D) in MSS (n = 125) and MSI (n = 26) CRC. Cell density was quantified by averaging the number of stained cells in randomly selected high-power fields (number of positively stained cells per square millimeter). Representative images are presented on the left side. Scale bars, 100 μm. (E) Single cells were isolated from MSS (n = 110) and MSI (n = 14) CRC, and the percentages of PD-1high, TIM-3+, LAG-3+, and TIGIT+ cells among tumor-infiltrating CD8+ T cells were analyzed by flow cytometry. Representative flow cytometry plots are presented at the top. SSC, side-scattered light. (F) The percentage of NY-ESO-1157-165–specific cells among tumor-infiltrating CD8+ T cells from HLA-A2(+) patients with MSS (n = 45) and MSI (n = 5) CRC was analyzed by flow cytometry. (G) The percentages of PD-1high, TIM-3+, LAG-3+, and TIGIT+ cells among NY-ESO-1157-165–specific, tumor-infiltrating CD8+ T cells from HLA-A2(+) patients with MSS (n = 45) and MSI (n = 5) CRC were analyzed by flow cytometry. (H) Single cells from MSS (n = 10) and MSI (n = 3) CRC were stained with cell trace violet (CTV) and stimulated with anti-CD3 antibodies, and the proliferation of CD8+ T cells was analyzed by flow cytometry. Representative histograms are presented on the left side. (I) Single cells from MSS (n = 17) and MSI (n = 9) CRC were stimulated with anti-CD3 antibodies and intracellular cytokine staining performed for IFN-γ and TNF. Representative plots are presented on the left side. Bars represent mean ± SEM; NS, not significant; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.
VEGF-A is up-regulated in MSS CRC compared with MSI CRC
We hypothesized that the tumor microenvironment of MSS CRC is more immunosuppressive than that of MSI CRC, leading to the depletion and severe exhaustion of tumor-infiltrating CD8+ T cells. To validate this hypothesis, we categorized a The Cancer Genome Atlas (TCGA) CRC cohort (28) based on immune gene signature modules (29). MSS CRC was composed mostly of a wound healing subtype with elevated expression of angiogenic genes, whereas an IFN-γ dominant subtype was enriched in MSI CRC (Fig. 2A). Among various cancers based on TCGA database, MSS CRC had the highest frequency of a wound healing subtype with angiogenic features (Fig. 2B). In line with these findings, the VEGFA mRNA level was significantly up-regulated in MSS CRC compared with that in MSI CRC (Fig. 2C and fig. S6). Analysis of our own cohort confirmed the elevated expression of VEGF-A in MSS CRC tissues compared with MSI CRC tissues (Fig. 2D) when VEGF-A expression was quantified on the basis of the previous study (30). We examined the main source of VEGF-A by performing intracellular staining for VEGF-A (Fig. 2E). Among VEGF-A–producing cells, about 82% were tumor cells (CD45−EpCAM+ cells), and about 7% were tumor-infiltrating T cells (CD45+CD3+ T cells), indicating that VEGF-A is mainly produced by tumor cells. In plasma, VEGF-A also significantly increased in patients with MSS CRC compared with patients with MSI CRC (Fig. 2F). The level of plasma VEGF-A significantly correlated with the level of VEGF-A in tissue homogenates (fig. S7). Furthermore, the plasma concentration of VEGF-A demonstrated a positive correlation with stages of MSS CRC, but no clear association was found in MSI CRC (Fig. 2G). These data indicate that VEGF-A is highly expressed in MSS CRC, particularly in higher stages.
(A) The proportions of four different immune subtypes (wound healing, IFN-γ dominant, inflammatory, and lymphocyte-depleted) were analyzed in MSS and MSI CRC from TCGA CRC cohort. (B) The percentages of wound healing subtypes in TCGA pan-cancer cohort according to cancer type. Abbreviations for the cancer types were based on the suggestions of the National Cancer Institute Genomic Data Commons (http://gdc.cancer.gov) except for MSS CRC and MSI CRC. (C) The expression of VEGFA mRNA was analyzed in MSS and MSI CRC from TCGA colorectal cancer cohort. The log2 (RPKM + 1) value of VEGFA mRNA is presented. RPKM, reads per kilobase per million. (D) The expression of VEGF-A was analyzed by immunohistochemical staining of tumor tissues from MSS (n = 72) and MSI (n = 17) CRC. Representative images are presented on the left side. Scale bars, 100 μm. (E) Single cells from CRC tissues (n = 17) were cultured with brefeldin A and monensin for 12 hours and then intracellular staining was performed for VEGF-A. The proportions of tumor cells (CD45−EpCAM+ cells) and tumor-infiltrating T cells (CD45+CD3+ T cells) among VEGF-A–producing cells were analyzed by flow cytometry. (F and G) Plasma concentrations of VEGF-A in patients with MSS (n = 80) and MSI (n = 8) CRC measured by enzyme-linked immunosorbent assay (F) and correlated with stages (G). Bars represent mean ± SEM; *P < 0.05; ***P < 0.001; ****P < 0.0001.
VEGF-A up-regulates immune checkpoint inhibitory receptors in CD8+ T cells and impairs their function
Because VEGF-A was overexpressed in patients with MSS CRC, we examined the effect of VEGF-A on the phenotype and functions of T cells. First, we verified the expression of VEGF-A receptors in human CD8+ T cells. Ex vivo analysis of human peripheral blood CD8+ T cells from normal donors by both reverse transcription polymerase chain reaction (RT-PCR) and flow cytometry revealed that they expressed VEGFR2, but not VEGFR1, upon anti-CD3 stimulation (Fig. 3, A and B). Tumor-infiltrating CD8+ T cells from MSS CRC expressed VEGFR2 even without ex vivo stimulation (Fig. 3C). In addition, VEGFR2 expression was observed solely in tumor antigen–specific (NY-ESO-1157-165–specific) tumor-infiltrating CD8+ T cells, whereas bystander (cytomegalovirus pp65495-503–specific) tumor-infiltrating CD8+ T cells did not express VEGFR2 at all (Fig. 3D), indicating that, among the tumor-infiltrating CD8+ T cells, tumor antigen–specific CD8+ T cells are able to respond to VEGF-A.
(A and B) CD8+ T cells were obtained from normal donors and stimulated with anti-CD3 antibodies for 60 hours (A) or 84 hours (B). The expression of VEGFR1 and VEGFR2 mRNA was analyzed by RT-PCR (A). β-Actin was used as an internal control. The expression of VEGFR1 and VEGFR2 in CD8+ T cells was analyzed by flow cytometry (B). Human umbilical vein endothelial cells (HUVECs) were used as a positive control. (C) The expression of VEGFR2 in CD8+ T cells from peripheral blood and tumors from MSS CRC (n = 12) was analyzed by flow cytometry. Representative histograms are presented on the left side. (D) The expression of VEGFR2 in tumor antigen–specific (NY-ESO-1157-165–specific) and bystander (CMV pp65495-503–specific) tumor-infiltrating CD8+ T cells (n = 10) was analyzed by flow cytometry. Representative histograms are presented on the left side. (E to G) PBMCs from normal donors were stimulated with anti-CD3 antibodies and VEGF-A for 84 hours. The percentages of PD-1+, TIM-3+, LAG-3+, and TIGIT+ cells among CD8+ T cells were analyzed by flow cytometry (E; n = 8). Proliferation of CD8+ T cells was analyzed by cell trace violet dilution (F; n = 5). Intracellular cytokine staining was performed for IFN-γ and TNF after 6 hours of stimulation with anti-CD3 antibodies (G; n = 6). Representative histograms or plots are presented on the left side. (H) The correlation between plasma VEGF-A concentration and the percentages of PD-1high, TIM-3+, LAG-3+, and TIGIT+ cells among tumor-infiltrating CD8+ T cells was analyzed in patients with MSS CRC (n = 49). Bars represent mean ± SEM; **P < 0.01; ***P < 0.001; ****P < 0.0001.
When we treated human CD8+ T cells from normal donors with recombinant VEGF-A protein in addition to anti-CD3, we found that VEGF-A increased the percentage of PD-1+, TIM-3+, LAG-3+, or TIGIT+ cells among total CD8+ T cells in a dose-dependent manner during antigen recognition (Fig. 3E and fig. S8), whereas VEGF-A did not up-regulate the expression of immune checkpoint receptors without anti-CD3 stimulation (fig. S9). Moreover, VEGF-A decreased anti-CD3–induced proliferation (Fig. 3F) and the production of IFN-γ and TNF (Fig. 3G). VEGF-A expression and T cell infiltration exhibited a negative correlation in both the invasive margin and tumor center in MSS CRC (fig. S10). Furthermore, the level of plasma VEGF-A significantly correlated with the percentage of PD-1high, TIM-3+, LAG-3+, or TIGIT+ cells among tumor-infiltrating CD8+ T cells in patients with MSS CRC (Fig. 3H). Together, these results indicate that VEGF-A may be one of the main causes of severe T cell exhaustion in patients with MSS CRC, as represented by the increased expression of immune checkpoint inhibitory receptors and diminished proliferation and cytokine production.
VEGF-A induces a TOX-mediated transcriptional program for T cell exhaustion
The transcriptional profiles of human CD8+ T cells from normal donors were analyzed by RNA sequencing (RNA-seq) analysis after anti-CD3 stimulation with or without VEGF-A treatment. Hierarchical clustering of whole transcriptomes revealed distinct expression profiles of CD8+ T cells after VEGF-A treatment (Fig. 4A). Among the 1502 differentially expressed genes (DEGs), VEGF-A up-regulated the expression of 820 genes, including various immune checkpoint and proapoptotic molecules, whereas it down-regulated the expression of 682 genes, including Kruppel-like factor (KLF) family members and prosurvival transcription factors (Fig. 4B). Gene set enrichment analysis (GSEA) revealed that the gene sets associated with exhausted CD8+ T cells, which were originally found in mice infected with chronic lymphocytic choriomeningitis virus (LCMV) (10), were enriched in VEGF-A–treated CD8+ T cells (Fig. 4C).
(A to D) CD8+ T cells from normal donors (n = 4) were stimulated with anti-CD3 antibodies for 60 hours in the absence or presence of VEGF-A. Total RNA was isolated and RNA-seq analysis was performed. Hierarchical clustering of the transcriptome with or without VEGF-A treatment (A). Heatmap of genes expressed differentially by CD8+ T cells with or without VEGF-A treatment (B). Gene expression is presented as row-wise z scores of normalized read counts. GSEA of gene sets up-regulated (top) or down-regulated (bottom) in exhausted CD8+ T cells from the chronic LCMV infection model was performed using the transcriptome of VEGF-A–treated versus untreated CD8+ T cells (C). NES, normalized enrichment score. Volcano plot depicting differential expression of 1045 transcription factors in VEGF-A–treated CD8+ T cells compared with untreated CD8+ T cells (D). (E and F) PBMCs from normal donors were stimulated with anti-CD3 antibodies for 84 hours in the absence or presence of VEGF-A. The expression of NFATc1 (E) and TOX (F) was analyzed by flow cytometry (n = 4). Representative histograms are presented on the left side. (G) Cyclosporin A (CsA) was added to the culture when PBMCs were stimulated in the presence of VEGF-A and the expression of TOX was analyzed by flow cytometry (n = 6). Representative histograms are presented on the left side. MFI, mean fluorescence intensity. Bars represent mean ± SEM; ****P < 0.0001.
Next, we focused on the expression of transcription factors to identify the main driver responsible for the VEGF-A–induced transcriptional program in T cell exhaustion. A volcano plot with transcription factor genes showed that NFATC1 and TOX were the most up-regulated genes after VEGF-A treatment in human CD8+ T cells (Fig. 4D and table S2). The up-regulation of NFATc1 and TOX proteins in VEGF-A–treated human CD8+ T cells was validated by flow cytometry (Fig. 4, E and F). We confirmed that NFATc1 mediated the VEGF-A–induced up-regulation of not only PD-1 but also other immune checkpoint inhibitory receptors, such as TIM-3, LAG-3, and TIGIT, in human CD8+ T cells using the NFATc1 inhibitor cyclosporine A (fig. S11A). However, NFATc1 inhibition also negated anti-CD3 stimulation–mediated T cell proliferation and cytokine production (fig. S11, B and C), indicating that NFATc1 is not an exhaustion-specific regulator. NFATc1 inhibition significantly abrogated the VEGF-A–induced expression of TOX (Fig. 4G), prompting us to investigate whether TOX expression can modulate T cell exhaustion during antigen recognition in the presence of VEGF-A. Thus, we investigated the role of TOX in VEGF-A–induced T cell exhaustion by knocking down the expression of TOX using small interfering RNA (siRNA). TOX-specific siRNA abolished the VEGF-A–mediated up-regulation of PD-1, TIM-3, LAG-3, and TIGIT (Fig. 5A) and restored the function of VEGF-A–treated CD8+ T cells in terms of anti-CD3–induced proliferation and the productions of IFN-γ and TNF (Fig. 5, B and C). GSEA showed that the changes in the gene sets associated with exhausted CD8+ T cells from the chronic LCMV infection model were reversed by knocking down TOX with siRNA (Fig. 5D). Moreover, the transcription profile of VEGF-A–treated, TOX siRNA–transfected CD8+ T cells significantly overlapped with that of VEGF-A–naïve CD8+ T cells (Fig. 5E).
(A to C) PBMCs from normal donors (n = 6) were treated with anti-CD3 antibodies and VEGF-A for 84 hours. Twenty-four hours after starting the treatment, PBMCs were transfected with TOX siRNA or control siRNA. The expression of TOX, PD-1, TIM-3, LAG-3, and TIGIT among CD8+ T cells was analyzed by flow cytometry (A). Representative histograms are presented at the top. Proliferation of CD8+ T cells was analyzed by cell trace violet dilution (B). Intracellular cytokine staining was performed for IFN-γ and TNF after 6 hours of stimulation with anti-CD3 antibodies (C). Representative histograms or plots are presented on the left side. (D and E) CD8+ T cells from normal donors (n = 4) were stimulated with anti-CD3 antibodies and VEGF-A for 60 hours. Twenty-four hours after stimulation, CD8+ T cells were transfected with TOX siRNA or control siRNA. Total RNA was isolated and RNA-seq analysis was performed. GSEA of gene sets up-regulated (top) or down-regulated (bottom) in exhausted CD8+ T cells from the chronic LCMV infection model was performed using the transcriptome of VEGF-A–treated, control siRNA–transfected versus VEGF-A–treated, TOX siRNA–transfected CD8+ T cells (D). Hierarchical clustering of VEGF-A–untreated CD8+ T cells, VEGF-A–treated CD8+ T cells, control siRNA–transfected CD8+ T cells, and TOX siRNA–transfected CD8+ T cells (E). Bars represent mean ± SEM; ****P < 0.0001.
We also performed H3 lysine 27 acetylation (H3K27ac) chromatin immunoprecipitation followed by sequencing (ChIP-seq), which is a useful tool for determining active regulatory elements, including enhancers and promoters (31, 32). We examined the effect of TOX on regulatory sequences by comparing H3K27ac ChIP-seq data between control siRNA– and TOX siRNA–transfected CD8+ T cells after anti-CD3 and VEGF-A treatment. A total of 7327 peaks were down-regulated in TOX siRNA–transfected cells, indicating that these peaks up-regulated in a TOX-specific manner (fig. S12A). Genome-wide analysis revealed that these 7327 differential H3K27ac peaks are highly enriched at intronic and intergenic regions (fig. S12B), suggesting a regulatory function of TOX on distal cis-regulatory elements. Next, we analyzed H3K27ac peaks around genes for immune checkpoint receptors, including PDCD1, HAVCR2, LAG3, and TIGIT. Down-regulation of peaks in TOX siRNA–transfected CD8+ T cells revealed several potential regulatory elements around immune checkpoint receptor genes (fig. S12, C to F), indicating that TOX epigenetically regulates genes for immune checkpoint receptors by controlling distal cis-regulatory elements.
Together, these data indicate that TOX mediates VEGF-A–induced T cell exhaustion and that TOX is responsible for not only the up-regulation of immune checkpoint inhibitory receptors but also the exhaustion-specific transcriptional program and functional impairment in VEGF-A–treated T cells. In addition, the epigenetic analysis revealed the link between TOX and the regulation of immune checkpoint receptor genes.
T cell exhaustion–specific program in MSS CRC is induced by VEGF-A in a TOX-dependent manner
Investigating whether TOX is responsible for the exhaustion status of tumor-infiltrating CD8+ T cells in patients with CRC, we found that the expression of immune checkpoint inhibitory receptors significantly associates with TOX expression (Fig. 6A). In addition, t-distributed stochastic neighbor embedding (t-SNE) analysis revealed that the expressions of TOX, PD-1, TIM-3, LAG-3, and TIGIT tightly overlapped to each other among tumor-infiltrating CD8+ T cells (Fig. 6B). Moreover, TOX expression was higher in tumor antigen–specific (NY-ESO-1157-165–specific) tumor-infiltrating CD8+ T cells compared with bystander (cytomegalovirus pp65495-503–specific) tumor-infiltrating CD8+ T cells (Fig. 6C). Next, we asked whether tumor-infiltrating CD8+ T cells from CRC have different expression profiles according to MSI status. To this end, we purified tumor-infiltrating CD8+ T cells from MSS and MSI CRC and performed RNA-seq analysis. Increased TOX expression was observed specifically in tumor-infiltrating CD8+ T cells from MSS CRC. Moreover, tumor-infiltrating CD8+ T cells from MSS CRC exhibited up-regulation genes for inhibitory receptors, including PDCD1, HAVCR2, LAG3, and TIGIT, and exhaustion-related transcription factors, including Eomes, BATF, and MAF (Fig. 6D). GSEA also revealed a highly significant overlap between the transcription profiles of tumor-infiltrating CD8+ T cells from MSS CRC and those observed in exhausted CD8+ T cells from the chronic LCMV infection model (fig. S13A) or VEGF-A–treated CD8+ T cells (fig. S13B). The transcription profile of tumor-infiltrating CD8+ T cells from MSI CRC resembled that of TOX siRNA–transfected CD8+ T cells (fig. S13C). We also compared TOX expression between MSS and MSI CRC at the protein level. Although we found no obvious difference in TOX expression in CD8+ T cells from peripheral blood or normal mucosa according to MSI status (fig. S14, A and B), TOX was highly up-regulated in tumor-infiltrating CD8+ T cells from MSS CRC compared with MSI CRC (Fig. 6E). We also investigated whether knocking down TOX can reverse the exhaustion of tumor-infiltrating CD8+ T cells from MSS CRC. TOX-silenced tumor-infiltrating CD8+ T cells from MSS CRC expressed reduced levels of inhibitory receptors and demonstrated enhanced cytokine production (Fig. 6, F and G). Collectively, our data suggest that tumor-infiltrating CD8+ T cells in MSS CRC undergo a VEGF-A–induced, TOX-mediated transcriptional program, resulting in severe exhaustion and impaired effector functions.
(A) The correlation between the expression of TOX and immune checkpoint inhibitory receptors in tumor-infiltrating CD8+ T cells from patients with CRC (n = 124) was analyzed by flow cytometry. Data are presented as fold change in the mean fluorescence intensity of TOX in PD-1neg, TIM-3−, LAG-3−, or TIGIT− tumor-infiltrating CD8+ T cells. (B) t-SNE analysis of the expression of TOX, PD-1, TIM-3, LAG-3, and TIGIT in tumor-infiltrating CD8+ T cells. (C) The expression of TOX in tumor antigen–specific (NY-ESO-1157-165 specific) and bystander (CMV pp65495–503 specific) tumor-infiltrating CD8+ T cells (n = 10) was analyzed by flow cytometry. (D) Tumor-infiltrating CD8+ T cells from MSS (n = 16, blue) and MSI (n = 8, red) CRC were isolated, total RNA was isolated, and RNA-seq analysis was performed. CRC cases are shown according to TOX expression level, and the CRC case with the highest TOX expression was placed on the left side. The gene expression of select genes is presented as row-wise z scores of normalized read counts. (E) The expression of TOX in tumor-infiltrating CD8+ T cells from MSS (n = 110) and MSI (n = 14) CRC was analyzed by flow cytometry. (F and G) TILs from MSS CRC were transfected with TOX siRNA or control siRNA. The expression of TOX, PD-1, TIM-3, LAG-3, and TIGIT in tumor-infiltrating CD8+ T cells was analyzed by flow cytometry (F; n = 6). Representative histograms are presented at the top. Intracellular cytokine staining was performed for IFN-γ and TNF after 6 hours of stimulation with anti-CD3 antibodies (G; n = 3). Representative plots are presented on the left side. Bars represent mean ± SEM; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.
Combined blockade of PD-1 and VEGF-A restores antitumor functions of T cells in MSS CRC
Last, we assessed whether blockade of the VEGF-A/VEGFR2 interaction enhances anti–PD-1–induced T cell reinvigoration. For this purpose, we established a tumor antigen–specific CD8+ T cell line that recognizes an HLA-A2–restricted NY-ESO-1 peptide (NY-ESO-1157-165, SLLMWITQV) with 99.1% purity (fig. S15A). Anti-CD3 stimulation up-regulated the expression of PD-1 and VEGFR2 in NY-ESO-1157-165–specific CD8+ T cells (fig. S15B). The NY-ESO-1157-165–specific PD-1+VEGFR2+ CD8+ T cells were then cocultured with HLA-A2+PD-L1+ Caco-2 MSS CRC cells pulsed with NY-ESO-1157-165 peptide. The cytotoxic activity of NY-ESO-1157-165–specific PD-1+VEGFR2+CD8+ T cells against Caco-2 cells was increased by either anti–PD-1 or anti-VEGFR2 blocking antibodies, which was further enhanced by combined treatment with both antibodies (Fig. 7A). Moreover, the production of IFN-γ and TNF was increased by anti–PD-1 or anti-VEGFR2 blocking antibodies and further enhanced by combined treatment (Fig. 7B).
(A and B) NY-ESO-1157-165–specific PD-1+VEGFR2+ CD8+ T cells were cocultured with HLA-A2+PD-L1+ Caco-2 MSS CRC cells pulsed with NY-ESO-1157-165 peptide for 6 hours. T cell–mediated cytotoxicity was evaluated by TO-PRO-3 staining of PKH26-labeled Caco-2 cells (A; n = 4). Intracellular cytokine staining was performed for IFN-γ and TNF (B). Representative plots are presented. (C and D) Single cells from MSS CRC were stimulated with anti-CD3 antibodies in the absence or presence of anti–PD-1 and/or anti-VEGFR2 for 84 hours (C; n = 22) or 36 hours (D; n = 20). Proliferation of CD8+ T cells was analyzed by cell trace violet dilution (C). Intracellular cytokine staining was performed for IFN-γ and TNF after addition of brefeldin A and monensin 24 hours after stimulation (D). Data are presented as fold change relative to isotype controls. Representative histograms or plots are presented on the left side. Bars represent mean ± SEM; ***P < 0.001; ****P < 0.0001.
We tested whether the functions of tumor-infiltrating CD8+ T cells are enhanced by ex vivo blockade of PD-1 and VEGF-A in the culture of single-cell isolates from MSS CRC. We evaluated the proliferation and production of IFN-γ and TNF by tumor-infiltrating CD8+ T cells upon anti-CD3 stimulation. Both the proliferation and cytokine production of tumor-infiltrating CD8+ T cells were increased by anti–PD-1 or anti-VEGFR2 blocking antibodies and further enhanced by combined treatment (Fig. 7, C and D). Anti–VEGF-A neutralizing antibody also had similar effects on T cell functions in the culture of single-cell isolates from MSS CRC in terms of proliferation (fig. S16A) and effector cytokine production (fig. S16B) as anti-VEGFR2 blocking antibody. Anti-VEGFR2 (fig. S17A) and anti–VEGF-A (fig. S17B) each significantly decreased the expression of TOX and immune checkpoint receptors in tumor-infiltrating CD8+ T cells, whereas anti–PD-1 treatment increased the expression of TIM-3, LAG-3, and TIGIT.
We corroborated these findings in an in vivo model established with syngeneic MC38 MSS CRC cells expressing ovalbumin (MC38-OVA) (33), which have an up-regulated wound healing signature including VEGF-A expression (fig. S18, A to D). First, we generated T cell–specific VEGFR2 conditional knockout (cKO) mice by crossing VEGFR2-floxed (VEGFR2fl/fl) mice with LCK-Cre (LCKcre) mice to assess the role of VEGFR2 in T cells. MC38-OVA cells inoculated in VEGFR2 cKO mice grew more slowly than those in wild-type mice, resulting in improved overall survival (Fig. 8, A and B). T cell depletion accelerated tumor growth and shortened survival in both wild-type and VEGFR2 cKO mice (fig. S19, A to C). Specifically, we found no differences in tumor growth kinetics and mouse survival between wild-type and VEGFR2 cKO mice when T cells were depleted, indicating that better tumor control in VEGFR2 cKO mice depends on T cells.
(A and B) Wild-type (WT) or T cell–specific VEGFR2 cKO mice (n = 12 for each group) were inoculated with MC38-OVA MSS CRC cells, and the tumor growth kinetics was analyzed (A). Estimated tumor volume is presented. Overall survival was analyzed with a Kaplan-Meier survival curve (B). (C to G) WT or T cell–specific VEGFR2 cKO mice were inoculated with MC38-OVA MSS CRC cells, and TILs were harvested 14 days after tumor cell inoculation (n = 5 for each group). The percentage of OVA257-265-specific cells (C), expression of TOX (D), and expression of PD-1, TIM-3, LAG-3, and TIGIT (E) among tumor-infiltrating OVA257-265-specific CD8+ T cells were analyzed. Representative plots or histograms are presented on the left side or at the top. TILs were stimulated with OVA257-265 peptide for 84 hours (F) or 36 hours (G). Proliferation of OVA257-265-specific CD8+ T cells was analyzed by cell trace violet dilution (F). Intracellular cytokine staining was performed for IFN-γ and TNF after addition of brefeldin A and monensin 24 hours after stimulation (G). Representative histograms or plots are presented on the left side. (H to J) WT mice were inoculated with MC38-OVA MSS CRC cells. Treatment with anti–PD-1 and/or anti-VEGFR2 started 14 days after tumor cell inoculation (n = 12 for each group) and the tumor growth kinetics of each treatment group was analyzed (H). Estimated tumor volume is presented. The overall survival of each treatment group is presented in Kaplan-Meier survival curves (I). Tumor tissues were harvested 21 days after tumor cell inoculation, and the expression of TOX in tumor-infiltrating OVA257-265-specific CD8+ T cells was analyzed (n = 6 for each group). ∆MFI is defined as a difference of MFI of TOX and isotype control (MFI of TOX − MFI of isotype control). MFI, mean fluorescence intensity (J). Bars represent mean ± SEM; **P < 0.01; ***P < 0.001; ****P < 0.0001.
We isolated and analyzed tumor-infiltrating lymphocytes (TILs) 14 days after MC38-OVA inoculation and confirmed the absence of VEGFR2 expression in tumor-infiltrating CD8+ T cells from VEGFR2 cKO mice by flow cytometry (fig. S20). The relative frequency of OVA257–264 (SIINFEKL)–specific cells among tumor-infiltrating CD8+ T cells, which was detected using a specific major histocompatibility complex I (MHC-I) (H-2Kb) multimer, was significantly higher in VEGFR2 cKO mice than in wild-type mice (Fig. 8C). In addition, the expression of TOX (Fig. 8D) and immune checkpoint inhibitory receptors (Fig. 8E) in tumor-infiltrating OVA257-265-specific CD8+ T cells was significantly down-regulated in VEGFR2 cKO mice compared with wild-type mice. When TILs were stimulated with OVA257-264 (SIINFEKL) peptide, the percentages of proliferating cells and cytokine (IFN-γ and TNF)–producing cells among OVA257-264 (SIINFEKL)–specific CD8+ T cells were significantly higher in VEGFR2 cKO mice compared with wild-type mice (Fig. 8, F and G).
We then evaluated the therapeutic efficacy of the combined blockade of PD-1 and VEGFR2 in wild-type mice. Although monotherapy with anti–PD-1 or anti-VEGFR2 blocking antibodies inhibited tumor growth and prolonged survival in mice, combined treatment with anti–PD-1 and anti-VEGFR2 antibodies further enhanced the therapeutic efficacy in terms of tumor growth and survival (Fig. 8, H and I). When we examined the phenotypes of tumor-infiltrating OVA257-264 (SIINFEKL)–specific CD8+ T cells, anti-VEGFR2 treatment significantly decreased the expression of TOX (Fig. 8J), PD-1, TIM-3, LAG-3, and TIGIT (fig. S21), whereas anti–PD-1 treatment did not alter the level of TOX (Fig. 8J) and increased the expression of TIM-3, LAG-3, and TIGIT (fig. S21). Collectively, our data demonstrate that cotargeting the PD-1/PD-L1 and VEGF-A/VEGFR2 axes cooperatively contributed to the reinvigoration of exhausted TILs, resulting in enhanced control of MSS CRC tumors.
DISCUSSION
Cancer immunotherapies that revitalize tumor-specific T cells through blockade of the PD-1/PD-L1 axis have emerged as a standard of care in metastatic MSI CRC (18–20). However, patients with MSS CRC, which make up most of the patients with CRC, respond poorly to PD-1 blockade, necessitating alternative strategies to benefit this group of patients. Here, we explored the immune microenvironment of CRC and identified VEGF-A–induced TOX as a core regulator of T cell exhaustion in MSS CRC, demonstrating that the exhausted T cells are reinvigorated upon TOX inhibition. Moreover, we supported our findings in vitro, ex vivo, and in vivo, with inhibition of VEGF-A/VEGFR2 signaling alleviating T cell exhaustion, and combined anti–PD-1 and anti-VEGFR2 therapy improving the antitumor response. Collectively, our study provides mechanistic insights that can be exploited to overcome anti–PD-1 unresponsiveness in MSS CRC by blocking both the PD-1/PD-L1 and VEGF-A/VEGFR2 pathways.
Aberrant angiogenesis is a key feature of various cancers and is regulated by diverse pro- and antiangiogenic factors (34). Among these factors, VEGF-A is the main driver of angiogenesis in the tumor microenvironment (35–37). In addition to its roles in angiogenesis, VEGF-A has also been implicated in the antitumor immune response; VEGF-A enhances the immunosuppressive activity of myeloid-derived suppressor cells and regulatory T cells (38, 39). Furthermore, VEGF-A has been shown to interfere with the infiltration of the tumor microenvironment by CD8+ T cells in murine tumor models (40, 41). However, the direct effects of VEGF-A on human T cells and the underlying mechanisms were not elucidated. In particular, the VEGF-A–induced transcriptional program that regulates fate of T cells was largely unknown. In the current study, we demonstrated that VEGF-A, which is abundant in the tumor microenvironment of MSS CRC, induces the expression of PD-1 and other immune checkpoint inhibitory receptors in human CD8+ T cells, consistent with previous studies (42–44). We also demonstrated that VEGF-A induces an exhaustion-related transcriptional program observed in CD8+ T cells during chronic LCMV infection (10). Furthermore, our functional studies also showed that VEGF-A diminished cytokine production and proliferation of CD8+ T cells during antigen recognition. Collectively, these findings indicate an immunosuppressive function of VEGF-A directly driving T cell exhaustion in the tumor microenvironment of MSS CRC.
Mechanistically, the VEGF-A/VEGFR2 interaction up-regulated DNA-binding factor TOX in T cells, which is responsible for the exhaustion-specific transcriptional program during antigen recognition. Several transcription factors were implicated in T cell exhaustion, including T-bet (45), Eomes (46), TCF-1 (47), Blimp-1 (48), FoxO1 (49), NFATc1 (50), BATF (51), and MAF (52). Recently, Palazon et al. (53) demonstrated that hypoxia-inducible factor–1α (HIF-1α) induces the expression of immune checkpoint coinhibitory receptors in tumor-infiltrating T cells. They also showed that T cells express VEGF-A by HIF-1α–driven transcriptional program. Others have suggested a potential association between TOX and T cell exhaustion (6, 12, 50), but the direct causal relationship, molecular mechanism, and biological implications of this association have not been elucidated. In the current study, we explored the role of TOX in the context of T cell exhaustion. siRNA-mediated knockdown of TOX reversed T cell exhaustion phenotypes induced by VEGF-A, suggesting that TOX is a driver responsible for VEGF-A–induced T cell exhaustion.
TOX was originally identified as a factor essential for the development of CD4+ T cells in the thymus (54), but TOX has also been implicated in the development of innate lymphoid cells (55). Recently, Page et al. (56) demonstrated that TOX is involved in microbe-induced autoimmunity against host tissues, highlighting the role of TOX in CD8+ T cells during priming. In the current study, we shed light on the role of VEGF-A in T cell exhaustion via TOX up-regulation. VEGF-A–mediated up-regulation of TOX is responsible for not only the up-regulation of multiple checkpoint inhibitory receptors but also exhaustion-related transcriptional changes in CD8+ T cells (14). Very recently, a critical role of TOX in driving T cell exhaustion has been reported in mouse models of chronic viral infection or tumors (57–61). Conflicting roles of TOX in T cell–mediated autoimmunity (56) and T cell exhaustion (57–61) can be explained by context-dependent action of TOX, particularly regarding T cell differentiation status and apoptosis/survival signals.
To enhance the therapeutic benefits of checkpoint blockade in patients with MSS CRC, we tested the antitumor efficacy of combined PD-1/PD-L1 and VEGF-A/VEGFR2 blockade. In the experiments using a tumor antigen–specific CD8+ T cell line, enhanced antitumor activity was achieved with combination treatment with anti–PD-1 and anti-VEGFR2 antibodies. We also substantiated this finding using ex vivo assays with tumor-infiltrating CD8+ T cells from MSS CRC and an in vivo syngenic mouse model. Recent clinical trials combining checkpoint blockade and VEGF-A/VEGFR2 blockade have shown promising clinical activity (62–65). Along with these encouraging results, our study justifies broadening clinical trials investigating the safety and efficacy of combined VEGF-A/VEGFR2 and PD-1/PD-L1 blockade for patients with recalcitrant solid tumors such as MSS CRC.
In summary, we propose a fundamental role of VEGF-A in shaping the MSS CRC immune microenvironment. To the best of our knowledge, this is the first study dissecting the direct cross talk between angiogenic factor VEGF-A and human T cells in the tumor microenvironment. Our findings have potential implications for combinatory immunotherapy and target agents in MSS CRC and other anti–PD-1–resistant cancers. In a broader context, our results reveal that VEGF-A promotes tumor progression not only via its established role in angiogenesis but also by directly affecting T cells in the tumor microenvironment. Last, our study provides a framework for rationally designed treatment strategies targeting the VEGF-A/VEGR2 and PD-1/PD-L1 pathways.
MATERIALS AND METHODS
Study design and patients
The goal of this study was to examine mechanisms of T cell exhaustion in CRC. The study population was a prospective cohort of 151 treatment-naïve patients with CRC who underwent surgical resection between May 2016 and October 2018 (table S3). This study was approved by the institutional review board, and all patients provided written informed consent.
Cell lines
Caco-2 cells were purchased from the Korean Cell Line Bank and grown in media according to American Type Culture Collection recommendations. MC38-OVA cells were grown in RPMI 1640 (Welgene) supplemented with 10% fetal bovine serum (Welgene) and 1% penicillin/streptomycin (Sigma). Peripheral blood mononuclear cells (PBMCs) obtained from HLA-A*0201(+) donors were used to generate NY-ESO-1–specific CD8+ T cells. CD8+ T cells were negatively isolated using a magnetic bead separation kit (Miltenyi Biotec). A phycoerythrin-conjugated HLA-A*0201 dextramer loaded with NY-ESO-1157-165 peptide (SLLMWITQV) was used to label NY-ESO-1–specific CD8+ T cells. Anti-phycoerythrin microbeads (Miltenyi Biotec) were then conjugated for positive selection by magnetic separation. Purified NY-ESO-1–specific CD8+ T cells were expanded in RPMI 1640 containing anti-CD3 antibody (50 ng/ml), IL-2 (200 IU/ml), IL-7 (10 ng/ml), and IL-15 (100 ng/ml) with irradiated autologous PBMCs. The purity of NY-ESO-1–specific CD8+ T cells was maintained above 95% throughout the experiments.
Animals
Specific pathogen-free C57BL/6J mice were purchased from the Jackson Laboratory. C57BL/6J mice carrying VEGFR2 loxP-flanked alleles were crossed with C57BL/6J LCK-Cre mice to achieve T cell–specific gene deletion.
Reagents
All information is organized in table S4.
Sample preparation
Lymphocytes from human peripheral blood were obtained by density gradient centrifugation with lymphocyte separation medium (Corning). Single cells from human normal adjacent mucosa and tumor tissue were obtained by mechanical dissociation with MACS C-tubes (Miltenyi Biotec) and enzymatic dissociation using the Tumor Dissociation Kit (Miltenyi Biotec) according to the manufacturer’s instructions. Tumor-infiltrating CD8+ T lymphocytes were purified using CD8 microbeads (Miltenyi Biotec) from single cell isolates according to the manufacturer’s instructions. Single cells were obtained from mouse tumor tissue by mechanical dissociation with MACS C-tubes (Miltenyi Biotec) and separated by Percoll-Paque (GE Healthcare) density gradient centrifugation.
MSI analysis
DNA was extracted from formalin-fixed, paraffin-embedded tumor tissue using the MiniBEST Universal Genomic DNA Extraction kit (TaKaRa) and amplified. A Bethesda microsatellite panel (BAT25, BAT26, MFD15, D2S123, and D5S346) was used to identify MSI status. MSI-H was defined as the presence of at least two MSI markers. The presence of a single instability marker and the absence of instability markers were grouped together as microsatellite stable (MSS) CRC.
Immunohistochemistry
Sections from surgically resected cancer were stained using the automated staining platform (Ventana Benchmark XT automated staining system, Ventana Medical Systems) according to the manufacturer’s instructions. Four-μm-thick sections from tissue blocks were deparaffinized and rehydrated with xylene and alcohol solutions. Antigen retrieval was performed using cell conditioning solution (CC1; Ventana Medical Systems). The tissue sections were subsequently incubated with primary antibodies. After a chromogenic visualization step using the ultraView Universal DAB Detection Kit (Ventana Medical Systems), slides were counterstained with hematoxylin and cover-slipped. Positive and negative controls were stained concurrently to validate the staining procedure. To measure CD3+ or CD8+ T lymphocyte density, all stained slides were scanned (magnification ×200) using a VENTANA iScan HT slide scanner (Ventana Medical Systems). The digital slides were analyzed by VENTANA Virtuoso Digital Pathology Image Analysis software version 5.3 (Ventana Medical Systems), and areas of insufficient quality were avoided to prevent scanning errors. The cell counter plug-in of the ImageJ platform (http://rsb.info.nih.gov/ij/index.html) was used to quantify the number of positively stained cells per square millimeter. TILs at the invasive margin were determined as the number of stained lymphocytes in the invasion front as counted in 10 fields (magnification, ×200), with the exception of tumor areas with crush artifacts or necrosis. TILs at the center of the tumor were determined as the number of stained lymphocytes in the tumor center in 5 to 10 fields (magnification, ×200). VEGF H-score was used to quantify VEGF-A expression (30). The vascular density was measured by counting CD31-positive microvessels in the areas showing high vascularization at the invasion front. The mean of three fields was determined (magnification, ×200). Samples were stored and supplied by Liver Cancer Specimen Bank/Gene Bank of Severance Hospital. The antibodies used for immunohistochemistry are detailed in table S4.
Flow cytometry
The LIVE/DEAD fixable dead cell stain kit (Invitrogen) was used to exclude dead cells. To detect NY-ESO-1– or cytomegalovirus (CMV) pp65–specific CD8+ T cells, cells were stained with phycoerythrin-conjugated HLA-A*0201–restricted dextramers specific to NY-ESO-1157-165 (SLLMWITQV) or CMV pp65495-503 (NLVPMVATV) for 20 min at room temperature. Surface protein was stained with fluorochrome-conjugated antibodies for 30 min at 4°C. For intracellular staining, cells were permeabilized and fixed with the FoxP3 staining buffer kit (Thermo Fisher Scientific) and further stained with fluorochrome-conjugated antibodies. For detection of human PD-1 after ex vivo treatment with anti–PD-1 blocking antibody (clone EH12.2H7, mouse IgG1 isotype), we first stained PD-1 using clone EH12.2H7 and further stained using fluorochrome-conjugated anti-mouse IgG1 antibody as the secondary antibody. All flow cytometric analyses were performed with an LSR II instrument (BD Biosciences), and data were analyzed using FlowJo software (Treestar). The antibodies used for flow cytometry are detailed in table S4.
TCGA data analysis
RNA-seq data from TCGA were obtained from Firebrose (Broad Institute). The log2 (RPKM+1) values were obtained and compared between patients with MSS and MSI CRC. The immune subtypes of each sample were categorized according to previous studies (29).
Proliferation and cytokine secretion assays
To assess cellular proliferation, we stained cells using CellTrace Violet cell proliferation kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, cells were incubated with Cell Trace Violet (5 μM) for 20 min at room temperature then washed with fetal bovine serum-supplemented phosphate buffered saline. Plate-bound anti-CD3 antibody (1 μg/ml; Miltenyi Biotec) or SIINFEKL peptide (10 μg/ml) was used to stimulate T cells. After 84 hours, the cells were harvested, and the dilution of Cell Trace Violet was evaluated by flow cytometry. A proliferation score was calculated by averaging the number of divisions in responding cells and presented as a percentage. To assess cytokine secretion we stimulated T cells with plate-bound anti-CD3 antibody (1 μg/ml; Miltenyi Biotec) or SIINFEKL peptide (10 μg/ml). After 24 hours, brefeldin A (GolgiPlug, BD Biosciences) and monensin (GolgiSTOP, BD Biosciences) were added to the culture. After 12 hours, the cells were harvested, and cytokine production was evaluated by flow cytometry. Anti-human PD-1 antibody (10 μg/ml, EH12.2H7), anti-human VEGFR2 antibody (10 μg/ml, 7D4-6), anti-human VEGF-A antibody (10 μg/ml, A15136B) or mouse IgG isotype control (10 μg/ml, MOPC-21) was used to block PD-1 or VEGFR2/VEGF-A.
Enzyme-linked immunosorbent assay
The Human VEGF-A Quantikine ELISA kit (R&D Systems) was used to measure the concentration of VEGF-A in plasma and tumor tissues from CRC patients according to the manufacturer’s instructions. To measure VEGF-A in tissue homogenates, we homogenized 150 mg of tumor tissues with Precellys Lysing Kit in 350 μl of phosphate-buffered saline with 1% protease inhibitor, followed by centrifugation for obtaining supernatants.
RT-PCR
Total RNA was isolated from cells using the RNeasy Mini Kit (Qiagen). Complementary DNA was synthesized from total RNA using the LeGene cDNA Synthesis Master Mix (LeGene Biosciences) according to the manufacturer’s instructions. Polymerase chain reactions was performed with specific primers for VEGFR1 (forward: 5′-TCCTTTGGATGAGCAGTGTG-3′, reverse: 5′-AGCCCCTCTTCCAAGTGATT-3′), VEGFR2 (forward: 5′-CCAGTCAGAGACCCACGTTT-3′, reverse: 5′-TCCAGAATCCTCTTCCATGC-3′), and β-actin (forward: 5′-AGAGCTACGAGCTGCCTGAC-3′, reverse: 5′-AGCACTGTGTTGGCGTACAG-3′) using the following cycle conditions: 30 s at 95°C, 30 s at 95°C, and 30 s at 72°C. Amplification was carried out for 30 cycles and the product exposed to gel electrophoresis with SYBR (Invitrogen). The presence of amplified DNA was visualized using ultraviolet light exposure. TaqMan Gene Expression Assays (Applied Biosystems) were used to determine mRNA levels of CTAG1B and β-actin in tissue homogenates according to the manufacturer’s instructions.
RNA-sequencing
RNA was extracted from cells using TRIZOL reagent (Thermo Fisher Scientific). The quality and quantity of RNA were assessed using an Agilent 2100 bioanalyzer with the RNA 6000 Nano Chip (Agilent Technologies) and ND-2000 spectrophotometer (Thermo Fisher Scientific), respectively. The QuantSeq 3′ mRNA-Seq Library Prep Kit (Lexogen) was used to construct a library according to the manufacturer’s instructions. Briefly, each set of total RNA was prepared to be 500 ng, and an oligo-dT primer containing an Illumina-compatible sequence at its 5′ end was hybridized to the RNA. Reverse transcription was then performed. After degradation of the RNA template, second strand synthesis was initiated by a random primer containing an Illumina-compatible linker sequence at its 5′ end. The double-stranded library was purified by using magnetic beads to remove all reactive components. The library was amplified to add the complete adapter sequences required for cluster generation. The finished library was purified from PCR components. High-throughput sequencing was performed by single-end 75 sequencing using NextSeq 500 (Illumina). The QuantSeq 3′ mRNA-Seq reads were aligned using Bowtie2. Bowtie2 indices were generated from the genome assembly sequence or representative transcript sequences to align to the genome and transcriptome. The alignment file was used to assemble transcripts and EdgeR to process the data. Differentially expressed genes were determined based on the counts from unique and multiple alignments using coverage in Bedtools. Differentially expressed genes were defined by a false discovery rate < 0.05 and an absolute fold change > 2. GSEA was used to test for enrichment of a specific gene set. A normalized enrichment score (NES) was specified for each GSEA.
siRNA transfection
Control or TOX-specific siRNA (Thermo Fisher Scientific) was transfected with NEON transfection electroporator (Life Technologies). Transfection was performed with 200 nM siRNA at 2000 V for 10 ms with 3 pulses. T cells were stimulated with plate-bound anti-CD3 antibody (1 μg/ml; Miltenyi Biotec) for 24 hours to enhance the transfection efficiency before transfection.
H3K27ac ChIP-seq analysis
We conducted ChIP-seq to profile genome-wide landscape of H3 lysine 27 acetylation (H3K27ac) in control siRNA- and TOX siRNA-transfected primary CD8+ T cells after anti-CD3 and VEGF-A treatment. The cells were cross-linked in resuspension buffer (phosphate-buffered saline, 1% fetal bovine serum, pH 7.4) with 1% formaldehyde for 10 min at 25°C. The cross-linking was quenched by adding 125 mM glycine and incubating in 25°C for 5 min with rotation and 15 min in ice. The cells were suspended in SDS lysis buffer (1% SDS, 50 mM tris-HCl, pH8.0, 10mM EDTA) with protease inhibitors. Mono- and di-nucleosome size chromatin was obtained through sonication. The sonicated chromatin was incubated with H3K27ac antibody (Active Motif) and protein beads (Thermo Fisher Scientific) for 4 hours in 4°C with rotation. The chromatin-antibody-bead complex was subjected to serial washing with varying salt concentrations optimized for the antibody used. The immunoprecipitated complex was treated with RNaseA (Qiagen) and reverse–cross-linked overnight at 68°C. The immunoprecipitated DNA was recovered using AMPure XP beads (Beckman Coulter), and ChIP-seq libraries were prepared using NEBNext Ultra II DNA library Prep Kit (NEB) following the manufacturer’s instruction. The ChIP-seq libraries were sequenced using Illumina Nextseq 550 platform. The sequenced DNA reads were mapped to human reference genome (hg38) using BWA-mem (ver. 0.7.17, “–M” option). Reads with low alignment quality (MAPQ < 10) were removed, and PCR duplicates were discarded using Picard. ChIP peaks, relative to input, were identified using MACS with a cutoff (P < 10–5) and narrow peak options. The differential peaks were identified using HOMER “mergePeaks” with “-d given” option.
Coculture experiment
Target cancer cells were labelled with PKH26 dye (Sigma-Aldrich) according to the manufacturer’s instructions to assess in vitro cytotoxicity. They were pulsed with 10 μg/ml NY-ESO-1157-165 peptide (SLLMWITQV) for 1 hour at 37°C in a 5% CO2 incubator. NY-ESO-1–specific CD8+ T cells were then added (effector:target ratio 4:1) in the presence of anti–PD-1 antibody, anti-VEGFR2 antibody, or IgG isotype control. Cells were harvested and stained with TO-PRO-3 dye (Thermo Fisher Scientific) to detect dead target cancer cells after 6 hours. For intracellular cytokine staining, brefeldin A (GolgiPlug, BD Biosciences) and monensin (GolgiSTOP, BD Biosciences) were added to the culture. VEGF-A–enriched culture medium derived from Caco-2 MSS CRC cells was added to the coculture as a conditioned medium.
In vivo experiments
Female mice between 6 and 8 weeks of age were used in all experiments. Syngeneic MC38-OVA cells (1.0 × 106 cells) were inoculated into the right flank via subcutaneous injection. For the analysis of TILs, mice were sacrificed 2 weeks after tumor inoculation. After verifying tumor growth, treatment with anti–PD-1 antibody (RMP1-14), anti-VEGFR2 antibody (DC101), or rat IgG isotype control (2A3) was started 2 weeks after tumor inoculation. For T cell depletion study, anti-CD3 antibody (145-2C11) was administered three times at 3-day intervals before tumor inoculation. Mice were euthanized if the tumor volume exceeded 3000 mm3. All in vivo experiments were approved by the KAIST Animal Care Committee.
Statistical analysis
GraphPad Prism (GraphPad Software) or R software (R Foundation for Statistical Computing) was used for statistical analysis. Paired or unpaired t tests were used to compare continuous variables. The χ2 test was used to compare qualitative variables. Pearson or Spearman correlation analyses were used to evaluate correlations between parameters. The one-tailed Fisher’s exact test was used to assess the significance of overlapping genes between different groups, and the log-rank test used to compare survival outcomes. Significance was set as P < 0.05.
SUPPLEMENTARY MATERIALS
immunology.sciencemag.org/cgi/content/full/4/41/eaay0555/DC1
Fig. S1. Relative numbers of tumor-infiltrating T cells to tumor cells in MSS and MSI CRC.
Fig. S2. Expression of immune checkpoint inhibitory receptors in CD8+ T cells from the peripheral blood, adjacent normal mucosa, and tumors of patients with CRC.
Fig. S3. Expression of immune checkpoint inhibitory receptors in CD8+ T cells from the peripheral blood and adjacent normal mucosa of patients with MSS and MSI CRC.
Fig. S4. Expression of CTAG1B in normal adjacent mucosa and tumor tissues.
Fig. S5. Production of IFN-γ and TNF in CD8+ TILs upon anti-CD3 and anti-CD28 stimulation.
Fig. S6. Expression of upstream regulators of wound healing signature genes in CRC.
Fig. S7. Correlation of VEGF-A levels between plasma and tissue homogenates.
Fig. S8. Representative histograms for the expression of immune checkpoint receptors on CD8+ T cells stimulated with anti-CD3 antibodies and VEGF-A.
Fig. S9. Expression of immune checkpoint receptors on CD8+ T cells treated with VEGF-A in the absence of anti-CD3 stimulation.
Fig. S10. Correlation between VEGF-A expression and T cell infiltration in MSS CRC.
Fig. S11. Effects of NFATc1 inhibition on CD8+ T cells.
Fig. S12. H3K27ac ChIP-seq analysis for control siRNA– or TOX siRNA–transfected CD8+ T cells after anti-CD3 and VEGF-A treatment.
Fig. S13. GSEA analysis of tumor-infiltrating CD8+ T cell transcriptomes.
Fig. S14. Expression of TOX in CD8+ T cells from the peripheral blood and adjacent normal mucosa of MSS and MSI CRC patients.
Fig. S15. Characteristics of NY-ESO-1157-165–specific CD8+ T cell lines.
Fig. S16. Effects of the blockade of PD-1 and VEGF-A on the function of tumor-infiltrating CD8+ T cells.
Fig. S17. Effects of the blockade of PD-1, VEGFR2, and VEGF-A on the phenotype of tumor-infiltrating CD8+ T cells.
Fig. S18. Expression of wound healing signature genes and VEGF-A in MC38-OVA tumor tissues.
Fig. S19. Effects of T cell depletion in vivo.
Fig. S20. Expression of VEGFR2 in tumor-infiltrating CD8+ T cells from wild-type and T cell–specific VEGFR2 conditional knockout mice.
Fig. S21. Effects of in vivo blockade of PD-1 and VEGFR2 on the phenotype of OVA257-265-specific, tumor-infiltrating CD8+ T cells.
Table S1. Raw data (Excel).
Table S2. List of transcription factors up-regulated by VEGF-A treatment in CD8+ T cells during antigen recognition [Log2(fold change) > 2 and adjusted P < 0.05; Excel].
Table S3. List of patients (Excel).
Table S4. Key resources (Excel).
This is an article distributed under the terms of the Science Journals Default License.
REFERENCES AND NOTES
- Copyright © 2019 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works
