CD96 is a novel target for cancer immunotherapy shown to regulate NK cell effector function and metastasis. Here, we demonstrated that blocking CD96 suppressed primary tumor growth in a number of experimental mouse tumor models in a CD8+ T cell–dependent manner. DNAM-1/CD226, Batf3, IL12p35, and IFNγ were also critical, and CD96-deficient CD8+ T cells promoted greater tumor control than CD96-sufficient CD8+ T cells. The antitumor activity of anti-CD96 therapy was independent of Fc-mediated effector function and was more effective in dual combination with blockade of a number of immune checkpoints, including PD-1, PD-L1, TIGIT, and CTLA-4. We consistently observed coexpression of PD-1 with CD96 on CD8+ T lymphocytes in tumor-infiltrating leukocytes both in mouse and human cancers using mRNA analysis, flow cytometry, and multiplex IHF. The combination of anti-CD96 with anti–PD-1 increased the percentage of IFNγ-expressing CD8+ T lymphocytes. Addition of anti-CD96 to anti–PD-1 and anti-TIGIT resulted in superior antitumor responses, regardless of the ability of the anti-TIGIT isotype to engage FcR. The optimal triple combination was also dependent upon CD8+ T cells and IFNγ. Overall, these data demonstrate that CD96 is an immune checkpoint on CD8+ T cells and that blocking CD96 in combination with other immune-checkpoint inhibitors is a strategy to enhance T-cell activity and suppress tumor growth.
Tumor antigen–specific CD8+ T cells become dysfunctional in the tumor microenvironment (TME), compromising their ability to proliferate and reducing effector function such as cytokine production and cytotoxicity. Therapeutic strategies to evoke antitumor immunity are largely aimed at reversing these immunosuppressive pathways. Antibody blockade of T-cell coinhibitory receptors CTLA-4 and PD-1 or the immunosuppressive ligand PD-L1 has achieved impressive overall response rates in some cancer patients, in part, by reactivating tumor-specific CD8+ T cells (1). However, additional immunosuppressive signals originate from diverse sources in the TME, potentially circumventing PD-1/PD-L1 pathways and limiting the population of cancer patients who respond to current immunotherapies (2). The identification of additional immune-suppressive ligands and the coexpression of additional coinhibitory receptors on chronically activated T cells suggest that combined blockade of coinhibitory receptors may improve response rates in cancer patients.
Certain proteins of the nectin and nectin-like (Necl) family, including CD155 and CD112, have emerged as candidate immune-suppressive ligands that may circumvent immune reactivation after PD-1/PD-L1 blockade. These ligands can both activate lymphocyte function via interaction with the costimulatory Ig superfamily member DNAM-1/CD226 and, conversely, inhibit cell function through interaction with other Ig superfamily members, TIGIT and CD96 (reviewed in ref. 3). We have demonstrated that CD155 is expressed on tumor cells and tumor-infiltrating myeloid cells in both human and mouse tumors and can impair antitumor T lymphocytes and NK cell function via interaction with TIGIT and CD96 (4). Importantly, the increased antitumor immunity upon blockade of PD-1 or PD-1 and CTLA-4 is more effective in settings in which CD155 is limiting (4), suggesting a mechanistic rationale for cotargeting PD-L1 and CD155 function. Blockade of the coinhibitor receptors for CD155, TIGIT, and/or CD96 is one rational therapeutic approach for optimizing antitumor immunity.
Blockade of TIGIT in combination with anti–PD-L1 improves T-cell responses to tumors via an intrinsic effect on CD8+ T-effector cells leading to an increased production of IFNγ and TNFα (5). TIGIT is also enriched on tumor-infiltrating T-regulatory cells (Tregs) compared with peripheral Tregs, and TIGIT expression on Tregs suppresses antitumor immunity (6). The expression pattern of CD96 is broadly similar between mice and humans, and CD96 is present on a proportion of T-effector and Tregs, NK cells, and NKT cells. CD96 expression is generally low or absent in tissues without lymphocyte infiltrate (reviewed in ref. 3). Earlier investigations of CD96 function have focused on an observed inhibitory function for CD96 on NK cells in anticancer immunity. For instance, the abrogation of lung metastases in a range of spontaneous and experimental models observed in CD96–/– mice or upon CD96/CD155 blockade with monoclonal antibody treatment was due to NK cell function, IFNγ, and effectively counterbalanced by the action of CD226 (7, 8).
We have confirmed CD96 expression in human CD4+ and CD8+ T cells and showed that CD96 mRNA expression was correlated with T-cell markers in primary and metastatic human tumors (9). However, T-cell function for CD96 in antitumor immunity remains undefined. Here, we showed that coexpression of CD96 with TIGIT and/or PD-1 in mouse and human tumor-infiltrating lymphocytes (TIL) and that using antibodies that selectively block CD155/CD96 interaction alone and in combination with anti–PD-1/PD-L1 regulates T cell–mediated tumor control.
Materials and Methods
C57BL/6 and BALB/c wild-type (WT) mice were purchased from the Walter and Eliza Hall Institute for Medical Research and bred in-house. C57BL/6.Rag2–/–γc–/–, C57BL/6.Batf3–/–, BALB/c.Batf3–/–, and C57BL/6 IL12p35–/– mice were maintained at QIMR Berghofer Medical Research Institute as previously described (7, 10). C57BL/6 Tigit–/– mice were kindly provided by Bristol-Myers Squibb. C57BL/6.Cd96–/– and Cd226–/– mice were obtained from Dr Marco Colonna (Washington University School of Medicine, St Louis, MO) and have already been described (7). All mice were bred and maintained at the QIMR Berghofer Medical Research Institute and used when more than 6 weeks of age. All experiments were approved by the QIMR Berghofer Medical Research Institute Animal Ethics Committee.
B16F10 melanoma (ATCC, 2007), MCA1956 fibrosarcoma (Robert Schreiber, Washington University School of Medicine, St Louis, MO, 2013), and CT26 colon carcinoma (Peter MacCallum Cancer Centre, 2012) were maintained for no more than 2-week culture, injected, and monitored as previously described (11, 12). All these cell lines express CD155 as previously described (4). Briefly, B16F10 cells were cultured in DMEM supplemented with 10% heat-inactivated fetal calf serum (Thermo Scientific), 1× glutamax, penicillin (50 U/mL), streptomycin (100 μg/mL), and 10 mmol/L HEPES (Sigma-Aldrich), whereas other cell lines were cultured in RPMI supplemented with 10% heat-inactivated fetal calf serum, 1× glutamax, penicillin (50 U/mL), streptomycin (100 μg/mL), 1 mmol/L sodium pyruvate from Gibco-Life Technologies and 10 mmol/L HEPES in 5% CO2. All mouse tumor cell lines were tested for Mycoplasma using the Lonza Mycoalert Mycoplasma Detection Kit, but cell line authentication was not routinely performed.
Tumor growth and treatments
Tumor cells, B16F10 (1 × 105), CT26 (2 × 105), and MCA1956 (1 × 106), were injected subcutaneously into the left flank of 6- to 20-week-old C57BL/6 or BALB/c WT mice or various C57BL/6 gene–targeted mice. Mice were treated with i.p. injections of 250 μg of anti-CD96 (3.3, rat IgG1, Bio X Cell), anti-TIGIT (4B1, mouse IgG1 D265A, or IgG2a, Bristol-Myers Squibb), anti–PD-1 (RMP1-14, Bio X Cell), anti–PD-L1 (10F.9G2, rat IgG2b, Bio X Cell), anti–CTLA-4 (UC10-4H10, hamster Ig), or control IgG (1-1, rat IgG1, Bio X Cell) in schedules as indicated. In some experiments, anti-CD8β (53.5.8, 100 μg), anti-DNAM-1 (480.1, 250 μg), anti-asGM1 (50 μg), or anti-IFNγ (H22, 250 μg) from Bio X Cell were injected i.p. immediately prior to and during immunotherapy treatment as indicated. Tumor size was measured every 2 to 4 days a week with a digital caliper as the product of two perpendicular diameters. Survival was measured by plotting the last day of ethical tumor size measurement as the time of sacrifice.
Adoptive transfer of CD8+ T cells
Splenocytes from WT and CD96−/− mice were stained with TCRβ–PerCP-Cy5.5 (H57-597), CD8-BV711 (53.6.7), and zombie aqua, and CD8+ T cells (live TCRβ+CD8+) were sorted on FACSAria II (BD Biosciences; >95% purity). WT or CD96−/− CD8+ T cells were injected intravenously into Rag2−/−γc−/− mice. After 10 days, blood was collected to check the equivalent reconstitution of CD8+ T cells by flow cytometry. B16F10 (1 × 105) melanoma cells were injected subcutaneously into C57BL/6 Rag2−/−γc−/− mice, and mice were monitored for the tumor growth.
Groups of 10 to 15 C57BL/6 WT male mice were inoculated s.c. in the hind flank with 300 μg of methylcholanthrene (MCA; Sigma-Aldrich) in 0.1 mL of corn oil as described previously (13). Mice were treated with cIg (1–1), anti-CD96 (3.3), anti–PD-1 (RMP1-14), anti-TIGIT (4B1, IgG2a), anti-TIGIT (4B1, D265A), or their combinations (100 μg each i.p., twice/week) for 6 weeks from the second palpable tumor measurement (0.1–0.4 cm2, days 84–147 relative to MCA inoculation). Mice were then monitored for fibrosarcoma development over 250 days, with measurements made with a caliper square as the product of two perpendicular diameters (cm2). Data were recorded as tumor size in cm2 of individual mice or tumor growth rate (mm2/day) relative to treatment initiation.
Tumors were cut into small pieces and digested in digestion medium containing RPMI with collagenase II (1 mg/mL) and DNAse (20 μg/mL) for 45 minutes. Tumors and spleen samples were filtered through 70-μm filter, washed in PBS, and red blood cells were lysed by ACK buffer (0.15 M NH4Cl, 10 mmol/L KHCO3, EDTA 0.1 mmol/L, pH 7.2–7.4). Single-cell suspensions were incubated for 15 minutes in Fc blocking buffer (2.4G2 antibody in 2% FBS and 1 mmol/L EDTA in PBS) before staining with the following antibodies in 2% FCS PBS: anti-mouse CD45.2-APCcy7 (104), TCRβ–PerCP-Cy5.5 (H57-597), CD4-BV605 (RM4-5), CD8-BV711 (53.6.7), CD96-PE (3.3), TIGIT-APC (Vstm3), PD-1–FITC (J43), and zombie yellow or zombie aqua for live-dead stain. All mAbs were purchased either from BioLegend or eBioscience. For AH1 (gp70) tetramer staining, H-2Ld MuLV gp70 Tetramer-SPSYVYHQF-PE or -APC was obtained from NIH Tetramer Core Facility, and samples were incubated with AH1 tetramer in complete RPMI medium at 37°C in 5% CO2 incubator for 2 hours and then washed with PBS before staining with other antibodies. For intracellular cytokine staining, samples were either stimulated with gp70 peptide (sequence SPSYVYHQF) in the presence of Brefeldin A (1,000×) or with Cell Stimulation Cocktail (plus protein transport inhibitors; 1000×; eBioscience) for 4 hours in complete RPMI medium at 37°C before staining with cell-surface antibodies as described above. Samples were fixed and permeabilized using fixation permeabilization buffer from eBioscience for 20 minutes and frozen at −80°C. Samples were thawed for 20 minutes, washed, and then stained with anti-mouse FoxP3-efluor450 (FJK-16S), IFNγ-APC (XMG1.2), and TNFα-BV605 (Mab11). To determine absolute counts in samples, liquid-counting beads (BD Biosciences) were added directly before samples were run on a flow cytometer. Samples were acquired on LSRFortessa IV Flow Cytometer (BD Biosciences), and data were analyzed on FlowJo V10 (TreeStar).
For analysis of melanoma single-cell RNA-sequencing (RNA-seq), data processed from Tirosh and colleagues (14) were obtained from the Broad Single-Cell Portal (https://portals.broadinstitute.org/single_cell). Visualization tools and generation of scatter plots of transcript levels from a single-cell RNA-seq data study of melanoma patients were done as described (15). RNA-seq data from The Cancer Genome Atlas (TCGA) project were obtained from the UCSC Cancer Genomics Hub. To determine the correlation between expression of the CD96 with other genes of interest, RNA-seq data were normalized using the edgeR package. Pearson rank correlation was estimated on the normalized counts.
Multiplexed immunohistofluorescence (mIHF), image processing, and analysis of human tumors
Archival colorectal cancer samples were obtained from Envoi Specialist Pathologists, and archival melanoma samples were obtained from Melanoma Institute Australia. The study protocols were approved by the QIMR Berghofer Human Research Ethics Committee (P1298 and P2125). Microsatellite-instability (MSI) status was assigned to colorectal cancer samples based on immunohistochemical absence of the mismatch-repair protein MLH1, which serves as a surrogate for MSI. A multispectral fluorescence imaging panel that examined CD8, PD-1, and CD96 with either SOX10 for melanoma or SATB2 for colorectal cancer was used to quantitate expression of CD96 and PD-1 on CD8+ T cells. DAPI was used as a nuclear stain. Briefly, specimens were sectioned at 4 μm onto superfrost+ microscope slides and stored under vacuum until mIHF was performed. Heat-induced antigen retrieval with EDTA target retrieval buffer (DAKO) was performed using a microwave and staining was run on an automated tissue stainer (DAKO). Primary antibodies were visualized using the OPAL multiplex TSA detection system (PerkinElmer) as per the manufacturer's instructions with heating for 20 minutes at 100°C using EDTA buffer between sequential staining rounds to strip prior bound antibody/HRP complexes. Primary antibodies, working dilutions, secondary detection HRP antibodies, and OPAL TSA dyes are listed in Supplementary Table S1. Fluorescence-stained slides were scanned using a Vectra imaging system (PerkinElmer). Whole slide scanning was done at 4× magnification using mixed fluorescence, and regions for 20× multispectral imaging were selected based on tumor marker and CD8 signals. For each sample, two to five 20 × multispectral imaging regions were selected to represent the tumor immune microenvironment. Multispectral images were spectrally unmixed followed by tissue and cell segmentation using Inform analysis software (PerkinElmer; v2.2.1). Nuclear expression of SATB2 or SOX10 by tumor cells was used to segment tumor and stroma tissue regions. Merged data files were preprocessed, and fluorescence thresholds were set using Spotfire image-mapping tools (Tibco Spotfire Analyst; v7.6.1) followed by segmented cell counting using Spotfire.
Statistical analysis was achieved using GraphPad Prism Software. Data were considered to be statistically significant when the P value was equal to or less than 0.05. Data were compared using a Mann–Whitney U test. Differences in survival were evaluated using a Mantel–Cox test.
Tumor growth suppression by anti-CD96 requires CD8+ T cells and IFNγ
Although targeting of the CD96 immune checkpoint on NK cells has been validated as an approach to reduce both experimental and spontaneous metastases (7), the role of CD96 on CD8+ T cells in primary tumor immunity is largely unknown. Treatment of mouse CT26 colon carcinoma, B16F10 melanoma, and MCA1956 fibrosarcoma with anti-CD96 minimally reduced tumor growth (Fig. 1A–C). The antitumor activity of anti-CD96 therapy was independent of Fc-mediated effector function in the MCA1956 tumor model (Supplementary Fig. S1A). To determine the mechanism of action of anti-CD96, depleting antibodies for either NK cells or CD8+ T cells and neutralizing antibodies to IFNγ were used from the commencement of anti-CD96 therapy. In the MCA1956 fibrosarcoma model, depletion of CD8+ T cells and neutralization of IFNγ completely abolished the efficacy of anti-CD96 therapy, whereas NK cell depletion had no significant impact (Fig. 1D and E). CD8+ T cells also express perforin as a critical effector molecule. However, anti-CD96 therapeutic benefit was retained in perforin-deficient mice (Fig. 1E). The single-agent anti-CD96 efficacy observed in the MCA1956 fibrosarcoma (Fig. 1C) was reversed by the addition of blocking anti-CD226/DNAM-1 in WT mice or absent in DNAM-1–/– mice (Fig. 1F). Depletion of CD8+ T cells also significantly reduced the efficacy of anti-CD96 therapy in the B16F10 melanoma tumor model (Supplementary Fig. S1B). To define whether CD96 suppressed antitumor CD8+ T cells directly or indirectly, we utilized adoptive transfer of WT- or CD96-deficient (CD96–/–) T cells into immunodeficient Rag2–/–γc–/– mice bearing B16F10 tumors. Superior tumor control was observed when CD96-deficient CD8+ T cells were adoptively transferred compared with transfer of WT CD96-expressing CD8+ T cells (Supplementary Fig. S1C), suggesting that CD96 expression on CD8+ T cells directly limits their antitumor function. Overall, these results indicated that anti-CD96 therapy minimally controlled tumor burden as a monotherapy and required CD8+ T cells, CD226/DNAM-1, and IFNγ, but not NK cells and perforin.
Anti-CD96 therapy requires Batf3+ cross-presenting dendritic cells and IL12
Batf3+ cross-presenting dendritic cells (DCs) are major producers of IL12 and are critical regulators of tumor growth and metastasis (10, 16, 17). Immunotherapies such as anti–PD-L1 and the combination of anti–PD-1 and anti-CD137 require Batf3-dependent DCs (18, 19). The antitumor efficacy of anti-CD96 therapy was completely lost in Batf3-deficient mice and in IL12p35-deficient mice (Fig. 2A–D), suggesting that the mechanism of anti-CD96 is dependent on Batf3+ DCs, possibly through their IL12 secretion.
CD96 is coexpressed with PD-1 and TIGIT on human and mouse TILs
Given the obligate role for CD8+ T cells in the antitumor efficacy of anti-CD96 treatment, we evaluated the expression of CD96 on mouse and human TILs. Analysis of CT26 colon carcinoma tumors indicated the CD96 expression was predominantly on CD8+ T cells and was observed less frequently on CD4+FoxP3− and CD4+FoxP3+ T cells (Supplementary Fig. S2A–S2C). On CD8+ T cells, CD96 was coexpressed with the other inhibitory receptors PD-1 or TIGIT (∼20% CD8+T cells) and sometimes observed as CD96+TIGIT+PD-1+ cells (4.8% of CD8+ T cells). CD96, PD-1, and TIGIT expression on antigen-specific gp70+CD8+ TILs was relatively higher than on gp70-negative CD8+ TILs in CT26 tumors (Supplementary Fig. S3A).
Evaluation of gene-expression data from TCGA demonstrated that CD96 mRNA was correlated with T-cell markers, such as CD3E, CD4, and CD8A, across most tumor types represented (9). Here, we showed that expression of CD96 was also correlated with expression of PD-1, a marker of T-cell dysfunction upon chronic antigen exposure. CD96 expression was also correlated with other candidate coinhibitory Ig superfamily receptors TIGIT and PVRIG/CD112R and its complementary costimulatory receptor, CD226/DNAM-1 (Supplementary Fig. S3B) across most tumor types tested. Analysis of each of the 32 tumor types available from TCGA indicated that the correlation of CD96 with these markers was consistently observed, with correlations between CD96 and TIGIT observed (22/32 tumor types demonstrating a Pearson correlation coefficient > 0.75). Correlations between CD96 and PD-1 were also observed (13/32 tumor types demonstrating a Pearson correlation coefficient > 0.75; Supplementary Fig. S3B).
Next, we analyzed the expression of CD96 and its correlation with relevant markers in human melanoma single-cell RNA-seq data from Tirosh and colleagues (14). Gene expression was tested selectively in the CD45+CD3E+CD8A+CD19–ITGAM– population and stratified according to PDCD1low versus PDCD1high expression. CD96 mRNA was coexpressed with CD8A in melanoma TILs (39%; Supplementary Fig. S4), and this expression was detected in both PDCD1low (15%) and PDCD1high (46%) CD8+ T cells. Coexpression of CD96 with TIGIT mRNA was also observed in CD8+ melanoma TILs (54%), and again, this coexpression was observed in both PDCD1low and PDCD1high CD8+ T cells (Supplementary Fig. S4).
To confirm protein expression of CD96 by tumor-infiltrating T cells, we evaluated CD96 expression in colorectal cancer by multiplexed immunohistochemistry (mIHF; Fig. 3A). A cohort of 7 microsatellite unstable human primary colorectal cancers were selected (MSI-CRCs) and had higher immune cell infiltration and were amenable to immunotherapy checkpoint blockade (20).The number of CD8+ T cells per 20× high-power field (HPF) was counted. Consistent with the mRNA analysis, CD96+ lymphocytes were observed in colorectal cancers, with a subpopulation staining CD96 among PD-1+CD8+ T cells (Fig. 3B). PD-1+CD8+ T cells were significantly enriched within the tumor parenchyma versus tumor stroma (Fig. 3C). Although the overall frequency of CD96+PD-1+CD8+ T cells (median = 0.4 cells/HPF; range, 0.0–22.4) was less than CD96–PD-1+ (median = 16.4 cells/HPF; range, 2.0 – 90.4), a trend toward increased CD96+PD-1+CD8+ T cells was observed in the tumor parenchyma of patients with CD8+ T-cell–infiltrated tumors. Unfortunately, many efforts to detect TIGIT by IHF were unsuccessful. Flow cytometry analysis of 8 archival colorectal cancer samples revealed that CD96 was expressed on most PD-1high and PD-1low CD8+ T cells (Supplementary Fig. S5). TIGIT was expressed less often than CD96 but did trend for expression on PD-1high CD8+ T cells.
Similar results were observed by multiplex IHF of a cohort of 9 human metastatic melanoma samples (Fig. 4A–C). In melanoma CD96+PD-1+CD8+ T cells (median = 31 cells/HPF; range, 4.5–203.5) were more frequent than was observed for colorectal cancer, but in both tumor types, the cells were found at higher numbers in the tumor parenchyma. For melanoma, we identified a population of CD96+PD-1–CD8+ T cells (median = 17 cells/HPF; range, 0.5–287.5) that was not present in colorectal cancer. There was no difference in distribution of CD96+PD-1–CD8+ T cells between tumor stroma and parenchyma. These data suggested that tumor-infiltrating CD8+ T cells can coexpress CD96 and PD-1 and that these cells might localize to the tumor parenchyma in colorectal cancer and melanoma.
Anti-CD96 therapy enhances the efficacy of immune-checkpoint blockade (ICB)
Given that CD96 is frequently coexpressed with PD-1 and TIGIT in both mouse and human intratumor CD8+ T cells, we hypothesized that the combination of anti-CD96 with antibodies blocking the immune checkpoints PD-1/PD-L1 and TIGIT might enhance antitumor responses. We observed that concurrent treatment with anti-CD96 enhanced the efficacy of anti–PD-1, anti–PD-L1, and anti-CTLA4 in the CT26 colon carcinoma model (Fig. 5A and B). The combination of anti-CD96 with anti-TIGIT (IgG2a) decreased the growth of MCA1956 fibrosarcoma tumors and increased the number of tumor-free mice (Fig. 5C and D). These results suggested that anti-CD96 enhanced the antitumor efficacy of multiple immune-checkpoint inhibitors. The mechanism of combination anti-CD96 with anti-TIGIT was dependent on CD8+ T cells but not on NK cells, as the therapeutic efficacy of the anti-CD96/anti-TIGIT combination was lost in CD8+ T cell–depleted WT mice but retained in mice where NK cells were depleted with anti-asGM1 (Supplementary Fig. S6).
Given the superior activity of the anti–PD-1/CD96 dual combination in the CT26 tumor model, we next assessed the functional consequences of CD96 and PD-1 inhibition on TILs. Here, we demonstrated an increased frequency of intratumor CD8+ T cells, but not CD4+ T-effector cells or Tregs, in mice treated with the combination of anti–PD-1 and anti-CD96 (Fig. 6; Supplementary Fig. S7). Among CD8+ T cells, an increase in the frequency of IFNγ+ and IFNγ+ TNFα+CD8+ T cells was observed after treatment with the combination of anti–PD-1 and anti-CD96 compared with the control Ig or anti-CD96 or anti–PD-1 monotherapies (Fig. 6; Supplementary Fig. S7). These data suggested that blockade of CD96 and PD-1 selectively and synergistically enhanced the effector function of tumor-infiltrating CD8+ T cells, consistent with the coexpression of these receptors on TILs and the requirement for CD8+ T cells in anti-CD96/PD-1 treatment efficacy.
Anti-CD96 enhances TIGIT and PD-1 blockade in a triple-combination therapy
To potentially improve the therapeutic benefit of ICB, we tested anti–PD-1/CD96/TIGIT in a triple-combination therapy. The triple combination of anti–PD-1/CD96/TIGIT was superior in reducing the tumor growth and improving the survival of mice in B16F10 melanoma and CT26 colon carcinoma compared with any monotherapy or dual-combination therapy (Fig. 7A and B; Supplementary Fig. S8A and S8B). The antitumor response achieved with each dual-combination treatment was significantly greater than monotherapy efficacies, with the dual combination of anti–PD-1 and anti-TIGIT (mouse IgG2a) having the best response (Fig. 7A and B).
TIGIT is expressed on intratumor Tregs (Fig. 3A), and TIGIT expression on Tregs has a reportedly prominent role in suppressing antitumor immunity compared with TIGIT expression on CD8+ T cells (6). Therefore, we questioned whether the enhanced antitumor efficacy of anti-TIGIT (IgG2a) with anti–PD-1 (IgG2a) or anti-CD96 (IgG1) in dual- or triple-combination therapy involved Fc receptor–mediated depletion of Tregs. To test this, we used anti-mouse TIGIT with different Fc isotypes (IgG1 vs. IgG2a vs. Fc-mutant D265A) in dual- or triple-combination therapy with anti-CD96 and anti–PD-1. Anti-TIGIT IgG2a was superior to anti-TIGIT IgG1 either as a monotherapy or in combination with anti-CD96 in reducing B16F10 melanoma tumor growth (Fig. 7C). Anti-TIGIT IgG2a also had a greater antitumor efficacy than Fc-mutant anti-TIGIT D265A as a monotherapy or when combined with anti-CD96 in dual or anti-CD96 and anti–PD-1 in triple-combination therapy (Fig. 7D; Supplementary Figs. S8C and S9A). The mechanism of triple combination of anti–PD-1, anti-CD96, and anti-TIGIT (G2a) was dependent on CD8+ T cells and IFNγ, as depletion of CD8+ T cells and neutralization of IFNγ almost completely abrogated the antitumor efficacy of triple therapy (Supplementary Fig. S9B and S9C). The triple-combination therapy of anti–PD-1/CD96/TIGIT(G2a) may require NK cells, in addition to CD8+ T cells, as the depletion of NK cells somewhat reduced the efficacy of the triple combination when used alone or in combination with CD8+ T-cell depletion (Supplementary Fig. S9B).
Triple ICB therapy is effective against de novo carcinogen-induced tumors
To test the efficacy of triple-combination therapy in a model of established fibrosarcomas induced in WT mice by s.c. injection of 300 μg MCA, mice were treated with anti-CD96, anti-TIGIT (G2a vs. D265A), and anti–PD-1 or control IgG over a 6-week period, either alone or as dual- or triple-combination therapy. Consistent with the results from the transplanted tumor models (Figs. 5 and 7), triple combination of anti–PD-1/CD96/TIGIT was superior at slowing tumor growth, resulting in more tumor-free mice (50%) compared with either monotherapy or dual-combination therapy (Fig. 8; Supplementary Fig. S10). The number of tumor-free mice in single-, dual-, or triple-combination therapy was greater when mice were treated with Fc active anti-TIGIT IgG2a, than with Fc-mutant anti-TIGIT D265A [single anti-TIGIT: IgG2a 3/12 (25%) vs. D265A 1/15 (6.6%); dual: anti–PD-1/TIGIT: IgG2a 5/12 (41.6%) vs. D265A 3/16 (18.7%); dual anti-CD96/TIGIT: IgG2a 3/12 (25%) vs. D265A 2/15 (13.3%); and triple anti–PD-1/CD96/TIGIT: IgG2a 6/12 (50%) vs. 8/24 (33.3%)]. Overall, these results suggest that the triple blockade of PD-1, CD96, and TIGIT was a superior therapeutic approach over the monotherapy or dual therapy combinations tested, and the Fc backbone of the anti-mouse TIGIT antibody plays a role in determining its antitumor efficacy.
The receptors CD96, TIGIT, and CD226/DNAM-1 comprise a critical regulatory system for lymphocyte activity and antitumor immunity. CD96, TIGIT, and CD226/DNAM-1 share the same ligands CD112 and CD155, and engagement of CD226 activates NK cell and T-cell activity, whereas TIGIT and CD96 are thought to counterbalance CD226-dependent lymphocyte activation. Prior work has demonstrated that CD96 is an intrinsic inhibitory receptor on NK cells, as the genetic deletion of CD96 or blockade using CD96-specific mAbs enhances NK cell IFNγ production and decreases spontaneous or experimental lung metastases (7, 8). Although it has been clearly demonstrated that TIGIT is an intrinsic T-cell–inhibitory receptor and can suppress CD226 activation of T cells (5, 6, 21–23), it has been unclear whether CD96 is intrinsically functional as an inhibitory receptor on T cells.
Here, we showed that CD96 was highly expressed in both human and mouse tumor-infiltrating CD8+ T cells and was observed along with other coinhibitory receptors such as PD-1 and TIGIT, suggesting an association with increased dysfunction of CD8+ T cells. Pharmacologic blockade of CD96–CD155 interactions using CD96-specific mAbs significantly controlled subcutaneous tumor growth in multiple mouse models, via a mechanism that was independent of NK cells and, instead, required CD8+ T cells and IFNγ. The antitumor efficacy of anti-CD96 was dependent on the complementary costimulatory receptor CD226/DNAM-1, and blockade of CD96 enhanced the antitumor efficacy of anti–PD-1/PD-L1, which correlated with an enhancement of tumor-infiltrating CD8+ T-cell effector function. These data validate CD96 as an inhibitory receptor that controls antitumor CD8+ effector functions and suggest that targeting CD96 may supplement the foundation strategy of PD-1/PD-L1 inhibition to improve therapeutic responses.
Blockade of CD96 enhanced not only PD-1/PD-L1 inhibition, but also the antitumor effect of anti-TIGIT, leading to increased tumor control and regression. Given that TIGIT can uniquely bind CD112, whereas both CD96 and TIGIT bind CD155, the superior combination efficacy of saturating concentrations of blocking TIGIT and CD96 antibodies cannot be explained by different ligand binding interactions of each inhibitory receptor. The superior combination efficacy of anti-CD96 plus anti-TIGIT was dependent on CD8+ T cells but independent of NK cells, suggesting that the T-cell function of these inhibitory receptors is nonredundant.
The nonredundant mechanisms of CD96 and TIGIT blockade can be explained in multiple ways: (i) selective effects of each pathway on different cell types or distinct anatomic or tissue-specific compartments; (ii) distinct functional response of each receptor on the same cell; or (iii) distinct temporal expression of the CD96 and TIGIT on T cells. Although CD96 and TIGIT are coexpressed on a subfraction of human and mouse tumor-infiltrating T cells, including those marked by increased PD-1 expression, TIGIT expression is distinctly enriched in Tregs compared with CD96. Kurtulus and colleagues show that TIGIT function in T-effector cells is not as critical as the immunosuppressive effect of TIGIT on Tregs in tumor control (6). Conversely, analysis of anti-CD96 monotherapy efficacy or the improved tumor control upon adoptive transfer of CD96 gene–targeted CD8+ cells (compared with transfer of CD96+CD8+ cells) in this current study indicated a role for CD96 on dysfunctional T-effector cells, whereas evidence for CD96 function on Tregs in antitumor immunity is still lacking. To address the cell type–selective effects of TIGIT and CD96 mAbs, we utilized different Fc variants of the TIGIT mAb and showed the greatest antitumor efficacy when anti-TIGIT mIgG2a was used in combination with anti-CD96. We cannot exclude distinct functional activities of the two inhibitory receptors on CD8+ T-effector cells as enhanced antitumor control was also observed with combination anti-CD96 plus Fc-mutated anti-TIGIT mIgG1 D265A.
Optimal clinical efficacy can theoretically be achieved by combination immunotherapies targeting novel checkpoint receptors to improve tumor-specific immune responses. We demonstrated that the checkpoint inhibitor CD96 was expressed in parallel with PD-1 and TIGIT in human and mouse tumor–infiltrating CD8+ T cells. Indeed, anti-CD96 treatment enhanced tumor control in combination with anti-TIGIT plus anti–PD-1, not only in the transplanted CT26 model, but also in the de novo MCA-induced fibrosarcoma model. Altogether, these data suggest that CD96, TIGIT, and PD-1 represent nonredundant mechanisms of tumor-induced immune suppression and T-cell dysfunction and can be cotargeted to achieve superior antitumor responses.
We have demonstrated that anti-CD96 enhances the antitumor efficacy of multiple methods of ICB and that CD96 is coexpressed with PD-1 in human tumor–infiltrating CD8+ cells. These findings suggest that a therapeutic strategy for cotargeting these receptors in human cancers is tractable. The current study regarding CD96 and TIGIT expression and function and the prior observation that limiting CD155 will enhance anti–PD-1 responsiveness in a preclinical setting (4) together suggest that pretreatment expression of CD155 and/or CD96 and TIGIT may define a patient population refractory to PD-1/PD-L1 monotherapy. Correlation of immunotherapy response and baseline CD155 in tumors or baseline and/or on-treatment CD96 and TIGIT expression in TILs may provide insight into the clinical application of these approaches.
However, given that functional response to anti-TIGIT or anti-CD96 requires CD226 function (refs. 5 and 24 and this study, respectively), any strategy to restore CD8+ T lymphocyte reactivity against cancer by inhibiting ligand binding to TIGIT or CD96 should consider the contribution of CD226. Although we demonstrated that CD96 and CD226 transcripts were correlated in tumor tissue, protein coexpression on tumor-infiltrating CD8+ T cells remains to be demonstrated, and there is evidence that CD226/DNAM-1 expression may be downmodulated upon chronic antigen stimulation. For instance, downregulation of CD226 has been observed on virus-specific PD-1+Lag3+CD8+ T cells in both HIV and LCMV chronic infections (25) and, in the cancer setting, CD226 is reduced in chronically stimulated NK cells in a mouse multiple myeloma model (26) and CD8+ TILs from melanoma patients (22). Conversely, increased expression of CD226/DNAM-1 on TILs corresponded with a greater response to anti–PD-1/PD-L1 or anti–CTLA-4 therapies (4). A lower ratio of activating (CD226):inhibitory (CD96, TIGIT) receptors will potentially limit lymphocyte responses, and these ratios might be individually regulated in different cell types. The expression of CD96 or TIGIT in patient tumor–infiltrating CD8+ T cells may not be sufficient to predict response to anti-TIGIT or anti-CD96 and functional expression of CD226/DNAM-1 should also be considered.
The strategy of targeting novel checkpoint receptors in combination may simultaneously block nonredundant immunosuppressive mechanisms to further increase clinical efficacy, as exemplified by the increased clinical response in patients with advanced melanoma receiving concurrent blockade of CTLA-4 and PD-1 compared with patients receiving monotherapy (27). However, this combination strategy may come at the cost of increased frequency and severity of immune-related adverse effects (irAE), as has been observed upon combination treatment with anti–PD-1 with anti–CTLA-4 (28). We have generated gene-targeted mice that were double deficient for either PD-1 and CD96 or TIGIT and CD96 and observed no long-term immune-related toxicities or autoimmunity (29). These data suggest that cotargeting these pathways can enhance tumor efficacy without inducing serious immune-related toxicities.
In summary, our data revealed that CD96 is expressed on infiltrating CD8+ T cells in mouse and human tumors where it functions as an intrinsic inhibitory receptor and that blockade of CD96 either as a monotherapy or in combination with blockade of another Ig superfamily member, TIGIT, or checkpoint inhibitors leads to enhanced antitumor immunity. Along with prior data indicating an inhibitory role for CD96 on NK cells, our study provides the rationale for clinical evaluation of CD96 antagonists targeting multiple immune compartments as a strategy for cancer immunotherapy.
Disclosure of Potential Conflicts of Interest
G. Long has received speakers bureau honoraria from Bristol-Myers Squibb, Novartis, Roche, Merck, and Incyte and is a consultant/advisory board member for Bristol-Myers Squibb, Novartis, Roche, Amgen, Pierre Fabre, Array, Merck, and Incycte. N.O. Siemers, A.J. Korman, and R.J. Johnston have ownership interest in Bristol-Myers Squibb. M.W.L. Teng has received speakers bureau honoraria from Bristol-Myers Squibb. W.C. Dougall reports receiving commercial research funding from Bristol-Myers Squib, has received speakers bureau honoraria from Amgen Inc., and is a consultant/advisory board member for Cascadia Drug Development Group and Omeros Corp. M.J. Smyth reports receiving commercial research funding from Bristol-Myers Squibb, Tizona Therapeutics, Corvus Pharmaceuticals, and Aduro Biotech, and is a consultant/advisory board member for Tizona Therapeutics and Nektar Therapeutics. No potential conflicts of interest were disclosed by the other authors.
Conception and design: G.V. Long, A. Korman, W.C. Dougall, M.J. Smyth
Development of methodology: A.R. Aguilera, S.J. Blake, G.V. Long, R. Lan, M.J. Smyth
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Mittal, A. Lepletier, A.R. Aguilera, K. Stannard, S.J. Blake, V.L.J. Whitehall, C. Liu, M.L. Bettington, G.V. Long, R.A. Scolyer, R. Lan, A. Korman, M.W.L. Teng, M.J. Smyth
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Mittal, A. Lepletier, J. Madore, A.R. Aguilera, G.V. Long, R. Lan, N. Siemers, R.J. Johnston, W.C. Dougall, M.J. Smyth
Writing, review, and/or revision of the manuscript: D. Mittal, A. Lepletier, V.L.J. Whitehall, C. Liu, M.L. Bettington, R.A. Scolyer, A. Korman, M.W.L. Teng, R.J. Johnston, W.C. Dougall, M.J. Smyth
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Stannard, K. Takeda
Study supervision: W.C. Dougall, M.J. Smyth
The project was funded by a National Health and Medical Research Council of Australia (NH&MRC) Program Grant (1132519), Project Grant (1098960) and Development Grant (1093566), a Cancer Council of Queensland Project Grant (1083776), a Cancer Research Institute CLIP grant, and a research agreement with Bristol-Myers Squibb. M.J. Smyth was supported by a Senior Principal Research Fellowship (1078671). M.W.L. Teng was supported by a CDF1 Fellowship and NH&MRC Project Grant (1098960). R.A. Scolyer is supported by a NH&MRC Practitioner Fellowship.
The authors wish to thank Liam Town and Kate Elder for genotyping and maintenance and care of the mice used in this study. We would also like to thank the Bristol-Myers Squibb team for support in producing the anti-TIGIT used in this project. The authors would like to acknowledge the specimen donors and research groups associated with tumor samples acquisition and analysis associated with the Cancer Genome Atlas Research Network.
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.