Tumor hypoxia is a negative prognostic factor that is implicated in oncogenic signal activation, immune escape, and resistance to treatment. Identifying the mechanistic role of hypoxia in immune escape and resistance to immune-checkpoint inhibitors may aid the identification of therapeutic targets. We and others have shown that V-domain Ig suppressor of T-cell activation (VISTA), a negative checkpoint regulator in the B7 family, is highly expressed in the tumor microenvironment in tumor models and primary human cancers. In this study, we show that VISTA and HIF1α activity are correlated in a cohort of colorectal cancer patients. High VISTA expression was associated with worse overall survival. We used the CT26 colon cancer model to investigate the regulation of VISTA by hypoxia. Compared with less hypoxic tumor regions or draining lymph nodes, regions of profound hypoxia in the tumor microenvironment were associated with increased VISTA expression on tumor-infiltrating myeloid-derived suppressor cells (MDSC). Using chromatin immunoprecipitation and genetic silencing, we show that hypoxia-inducible factor (HIF)-1α binding to a conserved hypoxia response element in the VISTA promoter upregulated VISTA on myeloid cells. Further, antibody targeting or genetic ablation of VISTA under hypoxia relieved MDSC-mediated T-cell suppression, revealing VISTA as a mediator of MDSC function. Collectively, these data suggest that targeting VISTA may mitigate the deleterious effects of hypoxia on antitumor immunity.

Acquired resistance to immune-checkpoint inhibitors is prohibitive to achieving maximum treatment response (1). Acquired resistance due to upregulation of alternate immune-checkpoint pathways contributes to treatment failure (2–5). Combination or sequential therapies designed to target nonredundant immune-checkpoint pathways may be able to overcome acquired resistance. In fact, multiple clinical trials have demonstrated positive results from such therapies (1), which compels continued investigation of immune-checkpoint regulation in the tumor microenvironment (TME).

V-domain Ig-containing Suppressor of T-cell Activation (VISTA, also known as VSIR, PD-1H, Dies1, DD1α, and Gi24) is a negative checkpoint regulator in the B7 family (6, 7). Anti-VISTA monotherapy enhances antitumor immunity and reduces tumor burden in murine colon carcinoma and melanoma (6, 8). VISTA is nonredundant with the PD-1 pathway and anti-VISTA synergizes with anti–PD-1 blockade to improve tumor remission (6–9). VISTA is abundant in murine and human carcinomas due to its expression on tumor-infiltrating immune cells, especially on myeloid cells, including myeloid-derived suppressor cells (MDSC; ref. 8). In contrast to PD-L1, VISTA is predominantly expressed on infiltrating hematopoietic cells (8). VISTA upregulation as a mechanism to resist checkpoint inhibition has been suggested by multiple studies (5, 10–12). In prostate cancer, which exhibits primary resistance to immune-checkpoint therapy, VISTA was upregulated in the TME following treatment with anti-CTLA4 (5). In metastatic melanoma, increased VISTA expression on infiltrating hematopoietic cells was associated with progression of disease in patients who initially responded to either anti–PD-1 monotherapy or the combination of anti–PD-1 and anti-CTLA4 (10). VISTA expression in the TME in melanoma is an independent prognostic factor associated with worse survival (12). Collectively, these studies suggest that VISTA mediates both primary and acquired resistance. We seek to understand how VISTA is regulated by TME factors.

Tumor hypoxia is an independent negative prognostic factor that promotes resistance to therapy through multiple complex mechanisms (13). In numerous studies, tumor hypoxia is associated with poor local control of disease, increased metastasis, and reduced overall survival (13). Mechanistically, hypoxia promotes immune escape through deleterious metabolic and genetic adaptations in tumor cells (reviewed in ref. 14) and by altering the immunologic signature of the TME. Hypoxia promotes tumor infiltration of suppressive regulatory T cells (Treg; refs. 14, 15) and MDSCs by upregulating the chemokine CCL28 (16) and numerous CXC family members (15), respectively. In addition, hypoxia blunts antitumor immunity through facilitating differentiation of MDSCs from precursor cells (15) and upregulation of suppressive mediators Arg1 and Inos (17). Hypoxia also increases expression of functional PD-L1 in MDSCs (18, 19).

In colorectal cancer, a leading cause of cancer-related death in the United States, hypoxia plays a role in the epithelial-to-mesenchymal transition that underlies progression to metastatic disease (20). Hypoxia also promotes tumor progression through cooperation with other oncogenic pathways (21), directly facilitating neovascularization (13), supporting immunosuppressive tumor-associated immune infiltrates (18), and promoting radiation resistance (22, 23). In this study, we found that high expression of HIF1A, a surrogate for hypoxia, was associated with VISTA expression in a cohort of patients with colorectal adenocarcinoma from The Cancer Genome Atlas (TCGA) database. High VISTA expression was associated with shorter overall survival. This observation, together with the presence of hypoxia response element in the VISTA promoter, led us to identify HIF1α as a transcriptional activator of VISTA in MDSCs in the TME. Results from antibody blockade and genetic silencing identified VISTA as a mediator of MDSC suppression of T cells, thus implicating hypoxia-driven VISTA expression in immune escape in colon cancer.

Mice and tumor models

All animal experiments were approved by the Institutional Animal Care and Use Committee of Geisel School of Medicine at Dartmouth. Mice were maintained in a specific pathogen-free facility. Experimental groups were age, gender, and strain matched. Female BALB/c mice were purchased from Charles River Laboratories (8–10 weeks old). VISTA−/− (KO) BALB/c mice were bred in-house. CT26 colon carcinoma cell line was a gift from Janssen Biotech Inc. The cells were obtained from ATCC in 2015 and frozen aliquots made after passaging the cells 3 times. For each experiment, cells were grown from the frozen aliquots of the same batch for 3 to 5 days in standard culture conditions until ∼50% to 70% confluent. Cells were harvested and used in experiments the same day. Cells were not authenticated in the past year. Mycoplasma testing was performed by IDEXX BioAnalytics (Columbia, MO). To establish tumors, 1 × 105 CT26 cells were injected intradermally. Tumor size was tracked and mice with tumors 10 to 15 mm in diameter were used for experiments.

Subjects

Peripheral blood samples were obtained from healthy volunteers (25–60 years of age). The protocol was approved by the Institutional Review Board of Dartmouth College and conducted in accordance with the ethical principles of the Declaration of Helsinki and Good Clinical Practice as defined by the International Conference on Harmonization. All donors gave written informed consent. Peripheral blood mononuclear cells were prepared from Terumo BCT leukoreduction system chamber content (following platelet pheresis) obtained from the Dartmouth-Hitchcock Medical Center and enriched by density gradient centrifugation over Ficoll (GE Healthcare Life Sciences) using the manufacturer's protocol.

Reagents and antibodies

RPMI-1640 was obtained from Corning Technologies. Antibiotics were purchased from Sigma. FBS was purchased from HyClone. Dead cells were excluded using Invitrogen Fixable LIVE/DEAD in Near-IR, Yellow, or Violet. The following antibodies were used for flow cytometry or immunofluorescence staining: from BioLegend: anti-VISTA (clone MH5A), anti-CD45 (30-F11), anti-CD11b (M1/70), anti-CD4 (RM4-5), anti-CD8 (53-6.7), anti-Ly6C (HK1.4), anti-Ly6G (IA8), anti-Gr1 (RB6-8C5), anti-F4/80 (BM8), and Armenian Hamster IgG Isotype control (HTK888); from eBioscience: anti-CD11c (N418), anti-CD16/CD32 (clone 93), and anti-FoxP3 (FJK-16s); anti-Armenian Hamster IgG (Jackson ImmunoResearch), and anti-VISTA (clone 13F3, made in-house). Antibodies for flow cytometry staining of human PBMCs: Hu FcR Binding Inhibitor (eBioscience), anti-VISTA (GG8, made in-house), anti-CD14 (clone TÜK4, Miltenyi), and from BioLegend: anti-CD11b (M1/70), anti-CD33 (WM-53), anti-HLA-DR (L243), anti-CD3 (SK7), anti-CD19 (HIB19), and mouse IgG1 κ isotype control (MOPC-21). For blocking experiments, antibodies used were anti-VISTA (clone 13F3, made in-house) and Armenian Hamster IgG1 Isotype Control (clone PIP, Bio X Cell).

Detection of hypoxia in vivo

Pimonidazole hydrochloride (pimo), which forms stable covalent adducts with thiol groups generated in hypoxic cells, was used to detect hypoxia. Pimonidazole adducts were detected by monoclonal antibody (mAb; antipimonidazole FITC, hypoxyprobe) for subsequent flow cytometry and immunofluorescence. For in vivo detection, mice were injected intraperitoneally (i.p.) with pimo (60 mg/kg) 90 minutes prior to tissue harvest. For flow, single-cell suspensions were prepared by mechanical dissociation and Tris-buffered ammonium chloride (ACT) red blood cell lysis (spleen and lymph nodes), then incubated with hypoxyprobe-1 mAb-FITC after surface staining and permeabilization.

MDSC-mediated T-cell suppression assay

Spleens were isolated from tumor-bearing mice. MDSCs were enriched using the Miltenyi MDSC Isolation Kit according to the manufacturer's instructions. Enriched MDSCs were stained with anti-CD11b, anti-Ly6G, and anti-Ly6C to confirm equal purity among groups before coculture with T cells. Naïve T cells were isolated from naïve age- and sex-matched mice using EasySep Naïve T-cell Kit (Stemcell Technologies) according to the manufacturer's instructions. Enriched T cells were labeled with CellTrace Violet (Thermo Fisher) and stimulated by plate-bound anti-CD3 (2C11) and soluble anti-CD28 (PV1), at 5 μg/mL (precoated in 100 μL PBS) and 1 μg/mL, respectively. In anti-VISTA (13F3) antagonism (or control antibody) experiments, MDSCs were incubated with respective antibody for 30 minutes on ice prior to coculturing with T cells at a 1:8 ratio.

Induction of hypoxia in vitro and in vivo

PBMCs and MDSC suppression assays were cultured in a hypoxia chamber (Stemcell Technologies) with a hypoxic gas mixture (1% O2, 5% CO2, N2 balance) or control normoxic conditions (standard culture, 21% O2, 5% CO2). For chemical induction of hypoxia in vivo, naïve mice were injected i.p. with 60 mg/kg of CoCl2 in PBS. After 6 hours, mice were sacrificed, and peripheral lymph nodes were harvested and prepared by mechanical dissociation for immediate staining.

In vitro shRNA knockdown

Validated lentiviral short hairpin RNA (shRNA), and packaging vectors were obtained from Sigma (MISSION TRC-Hs 1.5). The targeted sequences for Hif-1α were 5′-GTGATGAAAGAATTACCGAAT-3′ and 5′-TGCTCTTTGTGGTTGGATCTA-3′; for Hif-2α 5′-CGACCTGAAGATTGAAGTGAT-3′ and 5′-GCGCAAATGTACCCAATGATA-3′; and p53 5′-CGGCGCACAGAGGAAGAGAAT-3′ and 5′-GTCCAGATGAAGCTCCCAGAA-3′, respectively. The SHC-202 vector containing a nontargeting shRNA was used as a negative control. Viral particles were generated and concentrated according to The RNAi Consortium protocols (Broad Institute). For infections, freshly isolated PBMCs were plated (1 × 106 cells/well) in 12-well plates and spin-infected for 30 minutes at 2,300 rpm with concentrated virus (∼3 MOI) and 8 μg/mL polybrene (Sigma-Aldrich). After overnight incubation, cells were washed and seeded in fresh medium and returned to normal culture medium.

Flow cytometry and analysis

Stained cells were acquired using Miltenyi MACSQuant 8-color cytometer. Data were analyzed by FlowJo 9.8.3 software (Tree Star). Quadrant gates were set using fluorescence minus one (FMO) controls. Histogram gates were set using isotype controls.

Western blot analysis

Cells were lysed in Cell Signaling Lysis Buffer (Cell Signaling Technology) containing complete, Mini, EDTA-free Protease Inhibitor Tablet (Roche Diagnostics) and phosphatase inhibitors (Sigma-Aldrich P5726, P0044), resolved by electrophoresis, and transferred to Immobilon-P membranes (Millipore). Blots were incubated overnight at 4°C in the following antibodies: Hif-1α (1:1,000; Abcam, clone Ab51608), Hif-2α (Abcam Ab199), p53 (1:2,000; DO-1), and β-actin (1:20,000; Sigma A1978) prior to secondary detection.

Luciferase reporter assays

Reporter constructs for human and mouse VISTA promoters were purchased from Genecopoeia. The promoter sequences were subcloned into pGL4 luciferase vector and deletion of putative hypoxia response elements (HRE) was performed with the Quickchange XL kit (Stratagene). Mutagenesis primers are listed in Table 1. Forty micrograms of wild-type (WT) or ΔHRE construct vector and 10 μg of pGL4.75[hRluc/CMV] (Promega, cat. #E6931) control vector (NEB) were electroporated into THP-1 cells using the Nucleofector II (Amaxa) and the Human Monocyte Nucleofector Kit. Cells were cultured in normoxia or hypoxia and harvested at 48 hours after transfection. Cell lysis and luciferase detection were performed using the Dual Luciferase Assay System (Promega). Luciferase activity of test constructs was normalized to pGL4.75[hRluc/CMV] signal.

Chromatin immunoprecipitation

ChIP assays were performed on PBMCs cultured in hypoxia or normoxia for 48 hours in triplicate using 5 × 106 cells per IP. Cross-linking was done with 1.42% formaldehyde. After quenching with 125 mmol/L glycine for 5 minutes, cells were washed twice at 4°C with PBS. The remainder of the assay was performed using the ChIP-IT High Sensitivity kit (Active Motif) according to the manufacturer's instructions. Immunoprecipitation was performed using anti-HIF1α (1:1,000; Ab51608) or control rabbit IgG (ab171870). The purified immunoprecipitated DNA was quantified in triplicate by quantitative real-time PCR as detailed below and reported relative to control IgG. PCR primers are in Table 2.

Quantitative PCR and RT-PCR

Total RNA was isolated from 1 × 106 purified or cultured cells using the RNEasy mini kit (Qiagen); and 0.5 μg was reverse transcribed using the iScript supermix. Quantitative (real-time) PCR was performed in triplicate using 2% of cDNA per reaction with iQ SYBR Green Super Mix (Bio-Rad) and primers in Table 2. Relative mRNA concentrations were calculated by the ΔΔCt method relative to the geometric mean of GAPDH and PCNA (human) or Gapdh and Rn18s (mouse) mRNA concentrations. Results are reported normalized to the zero time point (set to a relative quantity of 1.0).

Immunofluorescence staining and analysis

Tumors were harvested and embedded in OCT (Tissue-Tek) and 10-μm sections taken on a Leica CM1860 cryostat. Tissue sections were fixed in acetone and blocked in 1% bovine serum albumin and 5% normal goat serum (BSA/NGS) and subsequently stained with primary [pimonidazole–FITC, CD11b–Alexa Fluor 647 (clone M1/70)], VISTA (MH5A) unconjugated or respective isotype then secondary antibody (goat anti-Armenian hamster TRITC) in BSA/NGS for 1.5 hours and 30 minutes, respectively, at room temperature. Sections were counterstained with DAPI, mounted, and imaged using Axio Observer.Z1 inverted microscope (Zeiss) at 25°C with a 20× Plan Apochromat lens (numerical aperture = 0.8) and acquired with a Photometrics CoolSnap HQ2 camera. Montages created using ZEN software. Image analysis and processing of single plane 10 μm slices of tumor were done using the ImageJ processing package (Fiji; ref. 24). Five sections were randomly selected from 100 consecutive tiles per sample. Total cell and CD11b+ VISTA+ cell subsets were quantified using the Colocalization Threshold and Analyze Particle functions of Fiji. Mean intensity of VISTA was performed by colocalizing VISTA with CD11b and subsequent pairing with pimonidazole positive and negative pixel regions using the BioVoxxel plug-in for Fiji (24). All images were normalized to isotype control prior to image analysis.

TCGA and Gene-Expression Omnibus (GEO) data extraction and analyses

Illumina HiSeq RNA-seq expression data for patients with colon or rectal adenocarcinoma were downloaded from the TCGA Data Portal at https://tcga-data.nci.nih.gov/tcga. Pearson rank-order correlation test was used to evaluate strength of correlation. Tumor purity values were calculated by PANCAN12 using ABSOLUTE and are publicly available for download (https://www.synapse.org/#!Synapse:syn1710466/version/2). Briefly, ABSOLUTE infers the purity of a tumor sample from somatic copy-number variation (25). TCGA sample genotype data from SNP arrays (SNP6) were used as input for ABSOLUTE, and all colon (COAD, dbGaP Accession phs000178) and rectal (READ, #phs000178) adenocarcinoma samples with high quality calls were used for downstream analyses involving purity. In total, both purity and HiSeq expression data were available for 283 patients.

To support the RNA findings, we used the correlation between VISTA and other HIF1α targets, as a measure of HIF1α protein activity. Because we found no ChIP-seq/ChIP-chip data sets for HIF1α in colon cancer cell lines in the ChEA database (26), we used the ChIP-seq data set in the breast cancer MCF-7 cell line to identify a list of candidate HIF1α target genes. Using the GSE39582 colon gene-expression data set (27), we calculated the Spearman correlation coefficients (SCC) of mRNA expression between HIF1α target genes and VISTA.

Overall survival and tumor mRNA expression of 585 patients who underwent surgery for primary colon adenocarcinoma (GSE40967) were downloaded from the GEO (27, 28). For survival analysis, 313 patients with overall survival data and who did not receive adjuvant chemotherapy were selected. The median normalized RNA expression of VISTA was used to stratify patients into VISTAhigh versus VISTAlow for survival comparison using Cox multivariate regression to identify independent predictors of survival.

Determination of the cell-lineage profile in TCGA data

Cell lineage scores (CLS) were calculated from tumor expression data using BASE (29). Briefly, gene-expression data from the Immunologic Genome Project (ImmGen) were processed to create normalized gene-expression profiles for 239 murine immune cell lineages. Genes with a 1-to-1 homology mapping to human were identified using Mouse Genome Informatics (http://www.informatics.jax.org/), and the resulting gene-by-lineage normalized expression matrix was provided as input to the BASE algorithm (30) along with gene expression for the tumor samples. The resulting CLS values for each lineage are an inferred approximation of the relative levels of lineage-specific immune infiltrate across the tumor samples. Vista was excluded from the analysis in order to rule out any potential bias when correlating CLS with VISTA expression in tumors.

Survival analysis

For survival analysis, 313 patients with overall survival data and who did not receive adjuvant chemotherapy were selected. X-Tile (31) was used to stratify patients into high/low expression of vista with respect to a cutoff (RNA normalized log2 value of 5.18, in the range 3.09–6.08) that optimizes survival stratification of VISTAhigh versus VISTAlow (17 samples vs. 296 samples; ref. 31). In order to conservatively correct for multiple cutoff testing and ensure the robustness of our findings despite unbalanced comparison arms, X-Tile computes an adjusted P value for the difference between survival curves from 1,000 Monte Carlo simulations using the software X-Tile in addition to a standard P value from the log-rank test (31). A log-rank test with subsequent Monte Carlo simulations was used to determine significance.

Statistical analysis

Graphs and statistical analysis were generated using Prism 6 (GraphPad Software, Inc.). Except for the TCGA and GEO data stated above, means were compared between groups using Student t test (two-tailed) or two-way ANOVA. Significance indicated by NS, P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001.

High expression of VISTA is associated with poor survival in colon cancer patients

Several lines of evidence led us to explore the potential role of VISTA in the colon cancer TME. We observed that VISTA expression is increased in the TME in multiple tumor models, including CT26 colon cancer (8). In clinical samples, published data implicate VISTA as a mechanism of resistance in melanoma (10, 12), prostate cancer (5), and oral cancer (11). We have previously shown increased VISTA expression in limited human colon cancer samples (32). High VISTA expression was observed in 28 resected colon cancer samples (33). Because of the unavailability of cohorts of colon cancer patients with VISTA protein data, we used VISTA mRNA expression as a surrogate to evaluate whether VISTA expression is associated with poor prognosis in colon cancer in a cohort of patient from TCGA data set GSE40967 (27). High VISTA expression (grouped by optimized expression cutoff) was significantly associated with poor prognosis compared with low VISTA expression (Supplementary Fig. S1). Although median cutoff for VISTA expression was not significantly associated with survival on univariate analysis, it was a significant independent predictor of decreased survival on multivariate analysis (Fig. 1A). The effect size of VISTA expression on survival was comparable with that of well-established parameters including mismatch-repair proficiency and tumor stage (Fig. 1A).

We hypothesized that the expression of VISTA and the association with poor prognosis was related to tumor hypoxia. First, we tested the correlation between VISTA and HIF1α mRNA in human colorectal adenocarcinoma samples. Comparison of gene-expression data (RNA-seq V2 RSEM) of VISTA and HIF1A extracted from the TCGA revealed significant positive correlation between HIF1A and VISTA (r = 0.47, P < 0.0001; Fig. 1B). To delineate the contribution of tumor cells from that of tumor-infiltrating leukocytes, we dissected the relationship of VISTA and HIF1A expression using tumor purity scores based on the ABSOLUTE algorithm (34), which uses somatic DNA mutations from genotyping tumor samples to infer tumor purity (Fig. 1C). This analysis showed that tumor samples with the lowest purity scores (i.e., highly infiltrated tumors, Q1) had the strongest correlation between VISTA and HIF1A (r = 0.49), and the correlation was lost (r = 0.20, P = 1) in poorly infiltrated samples (highest purity quartile; Fig. 1D). These data suggested that the association is driven by tumor-infiltrating leukocytes. Cell-lineage analysis based on the method developed by the Cheng group (29) showed that VISTA expression was associated with increased myeloid infiltration, consistent with higher VISTA expression in the myeloid compartment (Supplementary Fig. S2). Although HIF1A expression as a surrogate for tumor hypoxia was demonstrated to be independently associated with poor survival in colorectal cancer patients (35), the association was not significant in the cohort we evaluated. Thus, differential survival due to VISTA expression unveiled a mechanism that could account for the negative prognostic impact of hypoxia in a subset of colon cancer patients.

Because HIF1α is tightly regulated at the protein level, we sought to confirm our findings using HIF1α target genes as surrogates for HIF1α protein activity and hypoxia. We found that HIF1α targets (3 of 251 genes) were 7 times more likely to be correlated with VISTA expression (SCC > 0.6) compared with all the other genes (32 of 19,069) in the GSE39582 colon gene-expression data set (27). We also applied an established computational method (30, 36) to calculate the HIF1α regulatory activity in each tumor sample based on the expression of its target genes. The resulting HIF1α activity was correlated with VISTA expression with an SCC = 0.28. Indeed, at the mRNA level, HIF1A activity was significantly correlated with VISTA with an SCC = 0.35 (P < 0.0001). Thus, VISTA expression was correlated with both HIF1A protein activity and mRNA expression. A similar analysis was performed by using the TCGA colon cancer data set, which provides the expression of genes measured by RNA-seq. In these data set, we found that HIF1α target genes were 2 times more likely to be correlated with VISTA expression (SCC > 0.6) compared with the other genes, and that HIF1α mRNA expression (SCC = 0.52) and protein activity (SCC = 0.39) correlated significantly with VISTA expression. In summary, these data suggest that tumor hypoxia drives VISTA expression through HIF1α. We thus sought to identify a causal role for this association using an established colon cancer model.

VISTA is preferentially upregulated in hypoxic areas of the CT26 tumor

We evaluated the expression of VISTA within the TME of the CT26 murine colon cancer model. We found that VISTA was expressed on all tumor-infiltrating leukocytes including CD4+ T cells (TCRβ+ CD4+FoxP3) CD8+ T cells (TCRβ+ CD8+), and TCRβ+ CD4+ FoxP3+regulatory T cells (Fig. 2A). VISTA was expressed at higher densities on infiltrating myeloid cells, namely, CD11bhigh CD11c+ dendritic cells, CD11bhigh F4/80+ macrophages, with highest expression on CD11bhighGr1+ MDSCs (Fig. 2B).

Because of our finding in the colon cancer data set, we next investigated whether hypoxia was driving VISTA expression in the TME. Using pimonidazole (pimo) to track hypoxic cells, we compared VISTA expression in the TME to that in secondary lymphoid organs. Compared with the spleen and draining lymph node (dLN), tumor-infiltrating leukocytes (TIL) stained almost three times higher (based on MFI) with pimo (Fig. 2C), consistent with hypoxia within the TME. Next, we compared VISTA expression in hypoxic TILs (pimohigh) to less hypoxic TILs (pimolow). This illustrated that even within the TME, hypoxic TILs express higher density of VISTA compared with less hypoxic TILs (Fig. 2D).

Immunofluorescent staining of tumor sections revealed heterogeneous pimo staining, indicating spatially delimited, rather than global, hypoxia (Fig. 2E). In agreement with flow staining, dense VISTA staining was seen only in focal hypoxic regions, suggesting that VISTA responds to hypoxia (Fig. 2E). Closer examination demonstrated that pimo primarily costained with VISTA in CD11b+ myeloid cells (Fig. 2E).

Hypoxia increases the expression of stromal-derived factor 1 (SDF1, CXCL12), driving a chemotactic gradient which attracts myeloid-derived cells from the periphery into the TME (37). Indeed, tumor-infiltrating CD11b+ VISTA+ myeloid cells were enriched in hypoxic regions with over 90% of CD11b+ VISTA+ cells costaining with pimo (Fig. 2F). To clarify whether recruitment on CD11b+ cells was the primary driver of increased VISTA, we limited our flow analysis to the CD11b+ compartment. CD11b+ myeloid cells that were hypoxic (pimo+) expressed significantly more VISTA than those that were not hypoxic (pimo; Fig. 2F). These findings are consistent with hypoxia driving the upregulation of VISTA on myeloid cells that migrated into low oxygen zones in the TME.

Hypoxia upregulates VISTA expression via HIF1α binding to the VISTA promoter

Studies were designed to distinguish the effects of hypoxia from those of other tumor-associated factors in upregulating VISTA on immune cell subsets. For this purpose, human peripheral mononuclear cells (hPBMC) were used instead of murine myeloid cells because of their superior in vitro survival. Human VISTA and mouse Vista share 90% sequence homology and exhibit indistinguishable expression patterns: low expression on αβ+ T cells and high expression on CD11bhigh myeloid cells (6, 38). We cultured hPBMCs under hypoxic (1% O2) or normoxic (standard culture conditions, 21% O2) conditions. VISTA was not induced on CD19+ B cells or CD3+ T cells in hypoxic culture (Fig. 3A). Consistent with a role for hypoxia, VISTA was upregulated on CD11b+ myeloid cells following culturing of total PBMC under hypoxia for 48 hours (Fig. 3A and B). These data substantiate a hypoxia-dependent myeloid-specific regulatory program for VISTA expression.

CD11b+ myeloid cells are a heterogeneous group of cells comprised of many subsets, which was also reflected in differential VISTA expression. Immunophenotyping was used to identify monocyte subsets CD11bhighCD14highHLA-DR+ and CD11bhighCD14low/−HLA-DR+ as well as monocytic MDSC subpopulations (HLA-DRlow/−; Fig. 3B). Both MDSCs and monocytes upregulated VISTA expression under hypoxic culture (Fig. 3C and D). Similarly, treatment of naïve mice with cobalt chloride (CoCl2), a commonly used hypoxia mimetic that chemically induces HIF1α, resulted in the upregulation of VISTA on CD11bhi Ly6G+ and CD11bhi Ly6C+ myeloid cells compared with PBS control treatment (Fig. 3E). These data demonstrated that hypoxia induces VISTA expression, suggesting a cell-intrinsic mechanism for regulation of VISTA in the TME.

The cellular response to hypoxia is complex, governed by multiple intracellular pathways and transcription factors. Although the hypoxia-inducible factors (HIF)-1α and 2α play a role as master transcriptional regulators in the adaptive response to low oxygen tension (39), other transcriptional and posttranscriptional programs also play a role. We sought to investigate the role for HIFs. Examination of the proximal VISTA promoter (approximately 1,300 bp upstream of the gene as annotated by active chromatin markers in the University of California Santa Cruz Genome Browser, http://genome.ucsc.edu/) identified two potential HREs at −827 and −530 bp from the transcriptional start site (TSS) in the human promoter and three putative sites in the mouse promoter at −1,151, −432, and −388. Because HRE sequences are abundant throughout the genome with less than <1% exhibiting hypoxia-dependent binding of HIFs (28), we designed experiments to assess the function of these HREs. First, we used VISTA promoter sequences to drive luciferase reporter assays in normoxia and hypoxia with WT and mutated human and mouse promoters. Targeted deletion of only the −530 site in human promoter and the two sites at −1,151 bp and −388 bp in the mouse promoter led to over 2-fold attenuation of hypoxia-dependent luciferase activity at each site, demonstrating that these HREs are responsive to hypoxia (Fig. 4A and B). To evaluate in vivo activity, we performed HIF1α chromatin immunoprecipitation (ChIP) on myeloid cells from hPBMCs cultured in hypoxia or normoxia. Although HIF1α occupancy at the VISTA promoter in normoxia was comparable with the negative control, necdin (NDN), there was over 20-fold enrichment under hypoxia (Fig. 4C). The elevated occupancy was comparable with HIF1α binding to the positive control VEGF promoter (19).

To evaluate the requirement of HIFs for the induction of VISTA, we used knockdown of HIF1α or HIF2α in hPBMCs. Human PBMCs were transduced with a stable shRNA construct targeting HIF1α, HIF2α, or a nonspecific shRNA. As expected, hypoxic culture induced HIF1α and HIF2α RNA and protein expression (Fig. 4D–F). Specific HIF1α knockdown by two distinct shRNAs prevented HIF1α protein accumulation under hypoxia (Fig. 4E) and VISTA mRNA and protein expression on CD11b+myeloid cells remained comparable with those under normoxic conditions (Fig. 4D and E). In contrast, knockdown of Hif-2α did not affect the hypoxia-dependent induction of VISTA (Fig. 4F). Collectively, these data reveal that VISTA is a target of HIF1α under hypoxia.

VISTA is induced by p53 in the context of apoptotic cells (40). Because p53 is also implicated in hypoxia signaling (39), we asked whether p53 also plays a role in the hypoxic upregulation of VISTA by knocking down p53 in hPBMCs. Background p53 expression increased slightly in control cells under hypoxia at 12 and 24 hours and returned to baseline at 48 hours (Fig. 4G). Despite knockdown of p53 using two different shRNA constructs, targeting p53 did not affect VISTA induction under hypoxia (Fig. 4G). Thus, although p53 regulates VISTA expression in the context of apoptosis, our data demonstrate that p53 does not play a role in hypoxia-mediated regulation of VISTA.

Hypoxia-induced VISTA suppresses T-cell activity

Mechanisms of MDSC suppression have focused on the roles of arginase-1, inducible nitric oxide synthase (iNOS), indoleamine-pyrrole 2,3-dioxygenase (IDO), and reactive oxygen species (17, 41). Considering the high density of VISTA on MDSCs and the report of PD-L1 as a mechanism of MDSC suppression (19), we evaluated whether hypoxia-induced VISTA on MDSCs was functionally relevant.

Using WT MDSCs or MDSCs isolated from Vista-knockout (Vista−/−) mice (6), we evaluated the function of VISTA expression on MDSCs in a T-cell suppression assay. Absence of VISTA expression on MDSCs under normoxic conditions did not affect T-cell suppression based on proliferation and the activation markers CD25 and CD44 (Fig. 5A). However, under hypoxic conditions, the absence of VISTA reduced MDSC suppression, resulting in increased T-cell activity by all measured parameters of proliferation and activation in both CD4+ and CD8+ T cells. Thus, these data suggest that the upregulation of VISTA under hypoxia contributes to MDSC-mediated suppression of T cells.

To address the concern that genetic ablation of Vista may interfere with expression of other checkpoints such as PD-L1, which is also hypoxia regulated (19), we used an antibody approach. We have previously demonstrated that treatment of tumor-bearing mice with monoclonal anti-VISTA enhances antitumor responses, resulting in tumor remission of CT26 and B16OVA melanoma (6, 8). Capitalizing on this, we tested the effect of anti-VISTA antagonism on MDSC-mediated T-cell suppression. Consistent with Vista−/− MDSC data, antibody blockade of VISTA led to minimal changes in T-cell proliferation and activation under normoxic culture (Fig. 5B). However, under hypoxia, blockade of VISTA led to significant increases in all measured parameters of CD4+ and CD8+ T-cell proliferation and activation, recapitulating the effects of genetic deletion of Vista on MDSCs (Fig. 5B) arguing for a role of VISTA. Overall, these experiments demonstrate that hypoxia-induced expression of VISTA on MDSCs contributes to their suppression of T-cell proliferation and activation. Distinct from other B7 family members, VISTA is highly expressed on myeloid cells and is predominantly hematopoietically restricted. Therefore, using anti-VISTA blockade to target the suppressive function of MDSCs may be a useful addition as a combination therapy with other immune-checkpoint inhibitors.

In tumors that exhibit infiltration by antigen-specific CD8+ T cells, immune escape and resistance to checkpoint inhibitor therapy is mediated by suppression of T-cell activation in the TME (1, 3). Expression of negative checkpoint regulators affects both primary and acquired resistance to therapy (1, 3, 4). VISTA, a negative regulator of T-cell function, was shown to be increased in the TME of mouse models of CT26 (colon) and B16 melanoma (6, 8). Several studies have demonstrated that VISTA is expressed in tumor-infiltrating immune cells in melanoma (10, 12), oral cancer (11), and prostate cancer (5). Increased VISTA expression is correlated with poor outcomes in oral squamous cell carcinoma (11) and cutaneous melanoma (12). Consistent with these reports, VISTA expression was correlated with HIF1A mRNA and HIF1α activity in a cohort of colon cancer patients from the TCGA. This correlation was strongest in patients whose tumors were highly infiltrated, suggesting that VISTA expression on immune cells was driving this correlation. Although HIF1A expression was not associated with survival in this cohort, high VISTA expression was an independent predictor of worse survival on multivariate analysis. These data are in agreement with a publication showing that VISTA, but not PD-L1 expression, was associated with the signature of genes involved tumor immune evasion and aggressiveness in a cohort of colon cancer patients (33). Collectively, these data demonstrate a role for VISTA in immunosuppression that is specific to the TME and likely driven by tumor hypoxia.

Hypoxia is a mediator of tumor immune escape and resistance to therapy, through its effect on tumor cells, the TME, and infiltrating hematopoietic cells (13–18, 35). However, its role in controlling expression of immune-checkpoint pathways is only emerging. Two studies have shown direct regulation of PD-L1 and 4-1BB by hypoxia (19, 42). In this report, we uncovered a regulatory network in which TME hypoxia, through HIF1α binding to conserved hypoxia responsive elements in the VISTA promoter, upregulates VISTA on MDSCs. Increased VISTA expression, in turn, contributes to MDSC-mediated T-cell suppression under hypoxia conditions. This finding is in agreement with the previous report showing that blocking PD-L1 only under hypoxia relieves MDSC-mediated T-cell suppression (19) and adds VISTA to the list of mediators of MDSC activity. Our data with Vista−/− MDSCs, which have normal PD-L1 expression, suggest that VISTA and PD-L1 nonredundantly contribute to MDSC-suppressive activity. This is also consistent with their nonredundant roles in controlling tumor immunity in vivo (9). We further showed this with anti-VISTA. Direct targeting of VISTA on T cells by anti-VISTA to relieve MDSC-mediated suppression is another possible mechanism for our observation. However, the data with Vista−/− MDSCs support the conclusion that anti-VISTA is working at least in part through targeting VISTA on MDSCs. Targeting VISTA on T cells would thus represent an additional mechanism of anti–VISTA-mediated enhancement of T-cell responses.

Our studies expand the current understanding of hypoxia-induced immune escape and checkpoint resistance. Our results suggest that therapeutic strategies targeting VISTA may synergize with current immunotherapies due to the ability of VISTA to target both T cells and MDSCs.

D.A. Pechenick is a scientist at ImmuNext. R.J. Noelle is CSO at, reports commercial research grant from, has ownership interest (including stock, patents, etc.) in, and is a consultant/advisory board member for ImmuNext. I. Le Mercier has ownership interest (including stock, patents, etc.) in several patent applications related to therapeutic use of VISTA antagonists. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Deng, C.H. Lowrey, R.J. Noelle, R. Mabaera

Development of methodology: J. Deng, J. Li, D.A. Leib, R. Mabaera

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Deng, J. Li, J.L. Lines, Y.-C. Lee, I. Le Mercier, D.A. Leib, R. Mabaera

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Deng, D.C. Qian, D.A. Pechenick, R. Manivanh, F.S. Varn, C. Cheng, D.A. Leib, R. Mabaera

Writing, review, and/or revision of the manuscript: J. Deng, D.C. Qian, R. Manivanh, F.S. Varn, C. Cheng, D.A. Leib, R.J. Noelle, R. Mabaera

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Deng, J. Li, R. Manivanh, R. Mabaera

Study supervision: J. Deng, R.J. Noelle, R. Mabaera

Other (assisted in techinical work/performing some experiments): A. Sarde

The authors would like to thank the DartLab Core facility, the Mark Israel laboratory, and Dr. Christopher Amos for their support. This research was supported by NIH grant 1R01AI098007 (PI: R.J. Noelle) and The Dartmouth Clinical and Translational Science Institute, under award number KL2TR001088 (PI: Alan Green) from the National Center for Advancing Translational Sciences (NCATS) of the NIH (R. Mabaera). These studies were also supported by a research contract from ImmuNext and philanthropic support from the DF fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Syn
NL
,
Teng
MWL
,
Mok
TSK
,
Soo
RA
. 
De-novo and acquired resistance to immune checkpoint targeting
.
Lancet Oncol
2017
;
18
:
e731
41
.
2.
Spranger
S
,
Spaapen
RM
,
Zha
Y
,
Williams
J
,
Meng
Y
,
Ha
TT
, et al
Up-regulation of PD-L1, IDO, and T(regs) in the melanoma tumor microenvironment is driven by CD8(+) T cells
.
Sci Transl Med
2013
;
5
:
200ra116
.
3.
Shayan
G
,
Srivastava
R
,
Li
J
,
Schmitt
N
,
Kane
LP
,
Ferris
RL
. 
Adaptive resistance to anti-PD1 therapy by Tim-3 upregulation is mediated by the PI3K-Akt pathway in head and neck cancer
.
Oncoimmunology
2017
;
6
:
e1261779
.
4.
Koyama
S
,
Akbay
EA
,
Li
YY
,
Herter-Sprie
GS
,
Buczkowski
KA
,
Richards
WG
, et al
Adaptive resistance to therapeutic PD-1 blockade is associated with upregulation of alternative immune checkpoints
.
Nat Commun
2016
;
7
:
10501
.
5.
Gao
J
,
Ward
JF
,
Pettaway
CA
,
Shi
LZ
,
Subudhi
SK
,
Vence
LM
, et al
VISTA is an inhibitory immune checkpoint that is increased after ipilimumab therapy in patients with prostate cancer
.
Nat Med
2017
;
23
:
551
5
.
6.
Wang
L
,
Rubinstein
R
,
Lines
JL
,
Wasiuk
A
,
Ahonen
C
,
Guo
Y
, et al
VISTA, a novel mouse Ig superfamily ligand that negatively regulates T cell responses
.
J Exp Med
2011
;
208
:
577
92
.
7.
Flies
DB
,
Wang
S
,
Xu
H
,
Chen
L
. 
Cutting edge: a monoclonal antibody specific for the programmed death-1 homolog prevents graft-versus-host disease in mouse models
.
J Immunol
2011
;
187
:
1537
41
.
8.
Le Mercier
I
,
Chen
W
,
Lines
JL
,
Day
M
,
Li
J
,
Sergent
P
, et al
VISTA regulates the development of protective antitumor immunity
.
Cancer Res
2014
;
74
:
1933
44
.
9.
Liu
J
,
Yuan
Y
,
Chen
W
,
Putra
J
,
Suriawinata
AA
,
Schenk
AD
, et al
Immune-checkpoint proteins VISTA and PD-1 nonredundantly regulate murine T-cell responses
.
Proc Natl Acad Sci U S A
2015
;
112
:
6682
7
.
10.
Kakavand
H
,
Jackett
LA
,
Menzies
AM
,
Gide
TN
,
Carlino
MS
,
Saw
RPM
, et al
Negative immune checkpoint regulation by VISTA: a mechanism of acquired resistance to anti-PD-1 therapy in metastatic melanoma patients
.
Mod Pathol
2017
;
30
:
1666
76
.
11.
Wu
L
,
Deng
WW
,
Huang
CF
,
Bu
LL
,
Yu
GT
,
Mao
L
, et al
Expression of VISTA correlated with immunosuppression and synergized with CD8 to predict survival in human oral squamous cell carcinoma
.
Cancer Immunol Immunother
2017
;
66
:
627
36
.
12.
Kuklinski
LF
,
Yan
S
,
Li
Z
,
Fisher
JL
,
Cheng
C
,
Noelle
RJ
, et al
VISTA expression on tumor-infiltrating inflammatory cells in primary cutaneous melanoma correlates with poor disease-specific survival
.
Cancer Immunol Immunother
2018
;
67
:
1113
21
.
13.
Vaupel
P
,
Mayer
A
. 
Hypoxia in cancer: significance and impact on clinical outcome
.
Cancer Metastasis Rev
2007
;
26
:
225
39
.
14.
Hasmim
M
,
Noman
MZ
,
Messai
Y
,
Bordereaux
D
,
Gros
G
,
Baud
V
, et al
Cutting edge: hypoxia-induced Nanog favors the intratumoral infiltration of regulatory T cells and macrophages via direct regulation of TGF-beta1
.
J Immunol
2013
;
191
:
5802
6
.
15.
Labiano
S
,
Palazon
A
,
Melero
I
. 
Immune response regulation in the tumor microenvironment by hypoxia
.
Semin Oncol
2015
;
42
:
378
86
.
16.
Facciabene
A
,
Peng
X
,
Hagemann
IS
,
Balint
K
,
Barchetti
A
,
Wang
LP
, et al
Tumour hypoxia promotes tolerance and angiogenesis via CCL28 and T(reg) cells
.
Nature
2011
;
475
:
226
30
.
17.
Corzo
CA
,
Condamine
T
,
Lu
L
,
Cotter
MJ
,
Youn
JI
,
Cheng
P
, et al
HIF-1alpha regulates function and differentiation of myeloid-derived suppressor cells in the tumor microenvironment
.
J Exp Med
2010
;
207
:
2439
53
.
18.
Barsoum
IB
,
Smallwood
CA
,
Siemens
DR
,
Graham
CH
. 
A mechanism of hypoxia-mediated escape from adaptive immunity in cancer cells
.
Cancer Res
2014
;
74
:
665
74
.
19.
Noman
MZ
,
Desantis
G
,
Janji
B
,
Hasmim
M
,
Karray
S
,
Dessen
P
, et al
PD-L1 is a novel direct target of HIF-1alpha, and its blockade under hypoxia enhanced MDSC-mediated T cell activation
.
J Exp Med
2014
;
211
:
781
90
.
20.
Choi
BJ
,
Park
SA
,
Lee
SY
,
Cha
YN
,
Surh
YJ
. 
Hypoxia induces epithelial-mesenchymal transition in colorectal cancer cells through ubiquitin-specific protease 47-mediated stabilization of Snail: a potential role of Sox9
.
Sci Rep
2017
;
7
:
15918
.
21.
Wohlleben
G
,
Hauff
K
,
Gasser
M
,
Waaga-Gasser
AM
,
Grimmig
T
,
Flentje
M
, et al
Hypoxia induces differential expression patterns of osteopontin and CD44 in colorectal carcinoma
.
Oncol Rep
2018
;
39
:
442
8
.
22.
Harris
AL
. 
Hypoxia—a key regulatory factor in tumour growth
.
Nat Rev Cancer
2002
;
2
:
38
47
.
23.
Vatner
RE
,
Formenti
SC
. 
Myeloid-derived cells in tumors: effects of radiation
.
Semin Radiat Oncol
2015
;
25
:
18
27
.
24.
Schindelin
J
,
Arganda-Carreras
I
,
Frise
E
,
Kaynig
V
,
Longair
M
,
Pietzsch
T
, et al
Fiji: an open-source platform for biological-image analysis
.
Nat Methods
2012
;
9
:
676
82
.
25.
Yoshihara
K
,
Shahmoradgoli
M
,
Martinez
E
,
Vegesna
R
,
Kim
H
,
Torres-Garcia
W
, et al
Inferring tumour purity and stromal and immune cell admixture from expression data
.
Nat Commun
2013
;
4
:
2612
.
26.
Lachmann
A
,
Xu
H
,
Krishnan
J
,
Berger
SI
,
Mazloom
AR
,
Ma'ayan
A
. 
ChEA: transcription factor regulation inferred from integrating genome-wide ChIP-X experiments
.
Bioinformatics
2010
;
26
:
2438
44
.
27.
Marisa
L
,
de Reynies
A
,
Duval
A
,
Selves
J
,
Gaub
MP
,
Vescovo
L
, et al
Gene expression classification of colon cancer into molecular subtypes: characterization, validation, and prognostic value
.
PLoS Med
2013
;
10
:
e1001453
.
28.
Edgar
R
,
Domrachev
M
,
Lash
AE
. 
Gene Expression Omnibus: NCBI gene expression and hybridization array data repository
.
Nucleic Acids Res
2002
;
30
:
207
10
.
29.
Varn
FS
,
Andrews
EH
,
Mullins
DW
,
Cheng
C
. 
Integrative analysis of breast cancer reveals prognostic haematopoietic activity and patient-specific immune response profiles
.
Nat Commun
2016
;
7
:
10248
.
30.
Cheng
C
,
Yan
X
,
Sun
F
,
Li
LM
. 
Inferring activity changes of transcription factors by binding association with sorted expression profiles
.
BMC Bioinformatics
2007
;
8
:
452
.
31.
Camp
RL
,
Dolled-Filhart
M
,
Rimm
DL
. 
X-tile: a new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization
.
Clin Cancer Res
2004
;
10
:
7252
9
.
32.
Lines
JL
,
Sempere
LF
,
Broughton
T
,
Wang
L
,
Noelle
R
. 
VISTA is a novel broad-spectrum negative checkpoint regulator for cancer immunotherapy
.
Cancer Immunol Res
2014
;
2
:
510
7
.
33.
Xie
S
,
Huang
J
,
Qiao
Q
,
Zang
W
,
Hong
S
,
Tan
H
, et al
Expression of the inhibitory B7 family molecule VISTA in human colorectal carcinoma tumors
.
Cancer Immunol Immunother
2018
;
67
:
1685
94
.
34.
Carter
SL
,
Cibulskis
K
,
Helman
E
,
McKenna
A
,
Shen
H
,
Zack
T
, et al
Absolute quantification of somatic DNA alterations in human cancer
.
Nat Biotechnol
2012
;
30
:
413
21
.
35.
Baba
Y
,
Nosho
K
,
Shima
K
,
Irahara
N
,
Chan
AT
,
Meyerhardt
JA
, et al
HIF1A overexpression is associated with poor prognosis in a cohort of 731 colorectal cancers
.
Am J Pathol
2010
;
176
:
2292
301
.
36.
Khaleel
SS
,
Andrews
EH
,
Ung
M
,
DiRenzo
J
,
Cheng
C
. 
E2F4 regulatory program predicts patient survival prognosis in breast cancer
.
Breast Cancer Res
2014
;
16
:
486
.
37.
Zou
W
,
Chen
L
. 
Inhibitory B7-family molecules in the tumour microenvironment
.
Nat Rev Immunol
2008
;
8
:
467
77
.
38.
Lines
JL
,
Pantazi
E
,
Mak
J
,
Sempere
LF
,
Wang
L
,
O'Connell
S
, et al
VISTA is an immune checkpoint molecule for human T cells
.
Cancer Res
2014
;
74
:
1924
32
.
39.
Cummins
EP
,
Taylor
CT
. 
Hypoxia-responsive transcription factors
.
Pflugers Arch
2005
;
450
:
363
71
.
40.
Yoon
KW
,
Byun
S
,
Kwon
E
,
Hwang
SY
,
Chu
K
,
Hiraki
M
, et al
Control of signaling-mediated clearance of apoptotic cells by the tumor suppressor p53
.
Science
2015
;
349
:
1261669
.
41.
Condamine
T
,
Ramachandran
I
,
Youn
JI
,
Gabrilovich
DI
. 
Regulation of tumor metastasis by myeloid-derived suppressor cells
.
Annu Rev Med
2015
;
66
:
97
110
.
42.
Labiano
S
,
Palazon
A
,
Bolanos
E
,
Azpilikueta
A
,
Sanchez-Paulete
AR
,
Morales-Kastresana
A
, et al
Hypoxia-induced soluble CD137 in malignant cells blocks CD137L-costimulation as an immune escape mechanism
.
Oncoimmunology
2016
;
5
:
e1062967
.