Natural killer (NK) cells play a critical role in tumor immunosurveillance. Multiple activating and inhibitory receptors (IR) regulate NK-cell–mediated tumor control. The IR T-cell immunoglobulin and ITIM domain (TIGIT) and its counter-receptor CD226 exert opposite effects on NK-cell–mediated tumor reactivity.
We evaluated the frequency, phenotype, and functions of NK cells freshly isolated from healthy donors and patients with melanoma with multiparameter flow cytometry. We assessed TIGIT and CD226 cell surface expression and internalization upon binding to CD155. We evaluated the role of IL15 and TIGIT blockade in increasing NK-cell–mediated cytotoxicity in vitro and in two mouse models.
NK cells are present at low frequencies in metastatic melanoma, are dysfunctional, and downregulate both TIGIT and CD226 expression. As compared with TIGIT− NK cells, TIGIT+ NK cells exhibit higher cytotoxic capacity and maturation, but paradoxically lower cytotoxicity against CD155+ MHC class I–deficient melanoma cells. Membrane bound CD155 triggers CD226 internalization and degradation, resulting in decreased NK-cell–mediated tumor reactivity. IL15 increases TIGIT and CD226 gene expression by tumor-infiltrating NK cells (TiNKs) and, together with TIGIT blockade, increases NK-cell–mediated melanoma cytotoxicity in vitro and decreases tumor metastasis in two mouse melanoma models. Specific deletion of TIGIT on transferred NK cells enhances the antimetastatic activity of IL15, while CD226 blockade decreases the effects of IL15 and TIGIT blockade.
Our findings support the development of novel combinatorial immunotherapy with IL15 and TIGIT blockade to promote NK-cell–mediated destruction of MHC class I–deficient melanoma, which are refractory to CD8+ T-cell–mediated immunity.
See related commentary by Pietra et al., p. 5274
Here, we show that membrane-bound poliovirus receptor/CD155 triggers CD226 internalization and degradation, resulting in decreased natural killer (NK)-cell–mediated tumor reactivity. We also show that IL15 increases T-cell immunoglobulin and ITIM domain (TIGIT) and CD226 gene expression by tumor-infiltrating NK cells, and together with TIGIT blockade, increases NK-cell–mediated melanoma cytotoxicity in vitro and decreases tumor metastasis in vivo. Collectively, our findings support the novel combinatorial therapy with IL15 together with TIGIT blockade to promote NK-cell–mediated destruction of MHC class I–deficient melanoma, which are refractory to CD8+ T-cell–mediated immunity and PD-1 blockade.
There is ample evidence that natural killer (NK) cells exhibit tumor killing capacity and play a critical role in mediating tumor immunosurveillance of primary tumors and controlling metastases (1). NK cells express multiple activating receptors (ARs) and inhibitory receptors (IRs) that regulate their function. Hence, therapeutic strategies to engage ARs and/or counteract NK-cell inhibition have the potential to promote NK-cell–mediated tumor reactivity (2). Solid tumors are usually poorly infiltrated by NK cells (3), thus the phenotypic and functional studies of tumor-infiltrating NK cells (TiNK) remain very challenging in humans. TiNKs in ovarian, breast, lung, and prostate tumors downregulate multiple ARs, including DNAM-1/CD226, CD16, NKG2D, NKp30, NKp46, and 2B4, and are dysfunctional (4–9). NK cells upregulate multiple IRs that are also expressed by activated T cells, including CD94/NKG2A (10) and the T-cell immunoglobulin and ITIM domain (TIGIT; ref. 11). TIGIT binds with high and low affinity to CD155 (poliovirus receptor, PVR) and CD112 (Nectin-2), respectively, which are expressed on monocytes, dendritic cells, and tumor cells, including melanoma, and can also bind to the adhesion protein CD113 (Nectin-3; ref. 12). TIGIT competes with its costimulatory counter-receptor CD226 (DNAM-1), which binds to CD155 with lower affinity (11, 12). CD226 competes with CD112R for binding to CD112 (Nectin-2; refs. 13, 14). In mouse bearing tumors and in humans, dual PD-1/TIGIT blockade potently augmented tumor antigen CD8+ T-cell functions and promoted tumor rejection (15, 16). TIGIT acts in regulatory T cells (Treg) to promote tumor growth (17), and CD226 opposes TIGIT to disrupt Treg stability in melanoma (18). Several lines of evidence support the critical role of the TIGIT/CD226 axis in regulating NK-cell–mediated antitumor activity. First, TIGIT is highly expressed by circulating human NK cells (cNK) and impedes NK-cell–mediated killing of tumor cells (11). Upon CD155 binding to TIGIT, the ITT-like motif is phosphorylated and binds to Grb2 to recruit the SH domain containing inositol-5-phosphatase (SHIP1), impeding PI3 and MAP kinase pathways, and NF-κB signaling (19, 20). Second, CD226 associates with LFA-1 and is recruited to the immunologic synapse to promote NK-cell–mediated tumor cytotoxicity (21, 22). In vivo, CD226 is involved in NK-cell–mediated tumor surveillance and control of melanoma metastases and NK-cell–mediated lysis of melanoma (23–26). Third, one recent study in mouse tumor models has suggested that TIGIT acted primarily in NK cells to regulate CD8+ T-cell–mediated tumor reactivity (27). Among the various cytokines that expand and activate NK cells, IL15 potently enhances NK-cell–mediated tumor killing and is being actively investigated in many clinical trials (1). NK-cell–based therapies represent a powerful approach to kill MHC class I–deficient tumors that may arise upon CD8+ T-cell–mediated immune destruction of MHC class I presenting tumor cells, and may therefore counteract some of the mechanisms of resistance to PD-1 blockade (28, 29). However, therapeutic strategies to potently reinvigorate NK cells in human tumors remain to be developed. Here, we show that membrane bound (mb) CD155 triggers CD226 internalization and degradation, resulting in decreased NK-cell–mediated tumor reactivity. We also show that IL15 with TIGIT blockade augments the functional capacities of TiNKs in vitro and decreases tumor metastasis in mouse melanoma models. Altogether, our findings provide the rationale for combining IL15 with TIGIT blockade to counteract melanoma-induced NK-cell dysfunction and promote NK-cell–mediated lysis of MHC class I–deficient melanoma, which may prove useful in patients refractory to anti–PD-1 (28, 29).
Materials and Methods
Study subjects and cell lines
Peripheral blood mononuclear cells (PBMCs) from 30 healthy donors and PBMCs and tumor samples from 30 patients with stage IV melanoma were obtained under the internal review board–approved protocols UPCI 05-140 and UPCI 96-099. Biopsies were obtained from 29 patients with stage IV melanoma, including 22 males and seven females, ranging from age 31 to 82. Metastatic sites included skin or soft tissue (20%), nodes (45%), lung (20%), and other visceral locations (20%). Ten of the 29 patients had received prior IFNα adjuvant therapy. The samples were collected before therapy for stage IV melanoma and more than 3 years after the end of IFNα adjuvant therapy. The melanoma cell lines were derived from metastatic lesions of patients with melanoma at the University of Pittsburgh (Pittsburgh, PA). FO-I is a β2m-deficient human melanoma cell line recognized by NK cells (30). K562 and L cells were purchased (ATCC). Human CD155/PVR transcript variant 1 Gene cDNA Clone (full-length ORF Clone, Sino Biological Inc.) was transfected into L cells using Lipofectamine 3000 (Invitrogen) according to the manufacturer's instructions. L-CD155 transfectants were selected upon exposure to Hygromycin (400 μg/mL), transfectants and L-CD155+ clonal cell lines were obtained with limiting dilution.
Phenotypic analysis and cell sorting
PBMCs were used for ex vivo flow cytometry analysis. We performed CD3-positive Beads Magnetic Separation (Miltenyi Biotec) from the single-cell suspensions obtained from metastatic melanoma and used the CD3-negative fraction for ex vivo flow cytometry. CD45+CD3−CD56+ TIGIT+ and TIGIT− NK cells were sorted under sterile conditions (FACSAria Cytometer, BD Biosciences). The following conjugated mAbs were used in flow cytometry experiments: TIGIT-PerCPeFluor710, TIGIT-PE, CD14-PacificBlue, and CD19-PacificBlue (Thermo Fisher Scientific), NKG2D-PE-Cy7, PD-1-PE-Cy7, and CD226-PE-Dazzle594 (BioLegend), CD96-APC and Tim-3-Alexa700 (R&D Systems), CD45-BV510, CD56-BUV395, CD8-BUV737, CD3-APC-Cy7, and CD3-FITC (BD Biosciences), HLA-DR-ECD and CD16-PE-Cy7 (Beckman Coulter), CD226-PE or CD226-biotin (DX11, Abcam) coupled with streptavidin-PE-TexasRed (Thermo Fisher Scientific), and/or IgG control mAbs. Viability was assessed using LIVE/DEAD violet, Aqua Kits (Thermo Fisher Scientific) or Zombie-NIR (BioLegend). Intracellular staining was performed as reported previously (18). NK cells were labeled for Granzyme-A-PacificBlue (BioLegend), Granzyme-B-APC (Thermo Fisher Scientific), and Perforin-FITC (BD Biosciences). Samples were acquired on a FACS LSR-II Machine (BD Biosciences) and analyzed using FlowJo Software v9 (Tree Star).
NK-cell stimulation and functional assays
NK cells were obtained from PBMCs and metastatic melanoma single-cell suspensions by a two-step cell separation. CD3-negative cells, obtained after CD3-positive selection with Magnetic Beads (Miltenyi Biotec), were used for CD56-positive selection (Magnetic Beads, Miltenyi Biotec) to isolate CD56+ NK cells. NK cells were stimulated for 16 hours with 100 U/mL IL2 and/or 10 ng/mL IL15 (PeproTech) prior to incubation for 4 hours with FO-I (1:1 ratio) +/− blocking aTIGIT 10D7.G8 (IgG4, BMS; ref. 15), aCD226 (DX11, Abcam), or aCD155 (D171, Thermo Fisher Scientific) mAbs. In some experiments, NK cells were stimulated (16 hours, 48 hours, 4 days, or 6 days) with IL2 (100 U/mL), IL15 (10 ng/mL), IL7 (10 ng/mL), or IL21 (10ng/mL; PeproTech) prior to 4-hour coculture with FO-I, or in presence or absence of STAT5 Inhibitor (CAS 285986-31-4, Sigma-Aldrich) or STAT3 Inhibitor VI, S3I-201 (CAS 501919-59-1, Sigma-Aldrich) prior to receptor expression analysis. In other experiments, NK cells were incubated for 48 hours with IgG- or CD155-Fc beads +/− blocking antibodies. Cells were washed in PBS 4 μmol/L EDTA and beads were removed with magnetic separation prior to the functional assays. NK-cell degranulation capacity and intracellular IFNγ expression were evaluated with flow cytometry using CD107a-PerCPcy5.5 and IFNγ-PE/Cy7 (BioLegend), as reported previously (15). NK-cell–mediated cytotoxicity (specific lysis) was evaluated using standard 51Chromium-release assay as reported previously (31). Briefly, 51Cr-(GE Healthcare) labeled FO-I cells were incubated for 4 hours at 37°C either alone (spontaneous release), with 2.5% Triton-X-100 (maximum release), or with NK (E:T = 30:1) +/− aTIGIT and/or aCD226 blocking mAbs in triplicate wells.
Downregulation of CD226 and TIGIT expression by NK cells
A total of 1 × 105 purified NK cells were cocultured for 48 hours with L cells, L-CD155, FO-I, K562, HLA-I+ melanoma cell lines, immature dendritic cells (ratio 1:1), or CD155-Fc–coated beads (ratio 1:4). In some additional wells, L-CD155 and CD155-Fc beads were incubated for 30 minutes with either blocking anti-CD155 mAbs (D171, Thermo Fisher Scientific), 20 μmol/L of the inhibitor of metalloproteases and ADAM17, TAPI-2 (Sigma-Aldrich), or 9 μmol/L of the specific inhibitor of ADAM10, GI254023X (Tocris Bioscience) prior to coculture with NK cells and flow cytometry using the following antibodies: CD56-PeCy7 and CD226-PE-Dazzle594 (BioLegend) and TIGIT-PerCPefluor710 (Thermo Fisher Scientific). Immature dendritic cells were obtained from CD14+ monocytes isolated from PBMCs of normal donors with Magnetic Separation (Miltenyi Biotec) and cultured for 6 days with 1,000 U/mL IL4 and GM-CSF (PeproTech). CD155-Fc beads were obtained by coating a chimeric CD155-Fc (BMS) on M-450 Tosylactivated Beads (Dynal) following the manufacturer's protocol. Cell viability was assessed with LIVE/DEAD violet.
A total of 5 × 105 NK cells were incubated for 48 hours either with L cells or L-CD155 +/− 10 ng/mL IL15. RNA was extracted with RNeasy Mini Kits (Qiagen) and cDNA was prepared by reverse transcription using M-MLV Reverse Transcriptase (Thermo Fisher Scientific). RT-PCR was performed with StepOne System (Thermo Fisher Scientific). All samples were analyzed and normalized to the expression of β-glucuronidase. CD226 and TIGIT expression were detected using previously described primers (18).
sCD155 levels in sera were evaluated by sandwich ELISA. High-binding 96-well plates were coated with 2 μg/mL aCD155 capture mAb (D171, Abcam) for 30 minutes at 37°C in sodium phosphate buffer (pH 7.6) and blocked for 1 hour at room temperature with blocking buffer (PBS 3% BSA and 0.05% Tween). Plates were washed three times (PBS 0.05% Tween). Human CD155-Fc and 1:10 diluted samples were plated at 100 μL for 2 hours at room temperature and washed three times. Plates were incubated with 100 μL of secondary rabbit polyclonal aCD155 antibody (2 μg/mL in blocking buffer, Lifespan) for detection, washed, then incubated for 30 minutes at room temperature with 100 μL HRP-conjugated goat anti-rabbit (1:1,000 in blocking buffer, Thermo Fisher Scientific), washed, and reacted with 100 μL of substrate solution (1:1 tetramethylbenzidine and hydrogen peroxide, BD Biosciences). The reaction was stopped after 5 minutes with 50 μL of 2 N sulfuric acid and absorbance was read at 450 nm. All values were determined in triplicates.
ImageStream flow cytometry
A total of 1 × 105 NK cells isolated from PBMCs were incubated for 1 hour at 37°C either with medium alone, L cells, L-CD155, or FO-I (1:1 ratio). Cells were washed in cold PBS and in refrigerated centrifuge to prevent further receptor internalization. Viability was assessed with the Zombie-NIR Kit (BioLegend). NK cells were stained with CD45-BV510 (BD Biosciences), CD56-PeCy7 (BioLegend), CD226-PE (DX11, Abcam), or TIGIT PerCPefluor710 (MBSA43, Thermo Fisher Scientific), and then permeabilized and stained with CD226-biotin (DX11, Abcam) and streptavidin PE-TexasRed, or with TIGIT-PE (MBSA43, Thermo Fisher Scientific) prior to analysis on ImageStreamX MARKII Imaging flow cytometer with INSPIRE Software (Amnis, EMD Millipore). The flow rate was set at minimum and the objective magnification was set at 60× for all samples. A multifluorophore-labeled sample was used to determine accurate laser settings and avoid oversaturation. Gradient RMS and aspect ratio versus area on the brightfield channel were used during acquisition to ensure collection of focused single cells. At least 5 × 103 live CD45+CD56+ NK cells were acquired per sample. Data analysis was performed using IDEAS Software (Amnis, EMD Millipore). CD226 and TIGIT internalization ratios were calculated using CD45 membrane expression as a mask (a region of interest) to determine the membrane and intracellular sections of the cells.
Intracellular receptor degradation
A total of 1 × 105 NK cells were incubated for 30 minutes either with or without 0.5 μmol/L bafilomycin A1 (Sigma-Aldrich), 25 ng/mL concanamycin-A (Sigma-Aldrich), or 1% DMSO as control, prior to 16- or 48-hour coculture with L cells, L-CD155, or FO-I (1:1 ratio) +/− IL2 (100 U/mL), IL15, IL7, or IL21 (10 ng/mL). Cells were washed with PBS and fixed with 1.3% paraformaldehyde. All cells were stained for surface CD45-BV510 (BD Biosciences) and CD56-PeCy7 (BioLegend). Cells were either stained for surface CD226 and TIGIT expression, or permeabilized and stained for total CD226 and TIGIT expression prior to flow cytometry. Experiments were repeated and performed in triplicates.
Mice and experimental metastasis models
C57BL/6 wild-type (WT) and C57BL/6 Rag2−/−γc−/− mice were bred in-house. C57BL/6 Tigit−/− mice were kindly provided by Bristol Myers Squibb. All mice were bred and maintained at the QIMR Berghofer Medical Research Institute (Brisbane, Queensland, Australia) and used when more than 6 weeks of age. No mice were excluded on the basis of preestablished criteria in all studies and no active randomization was applied to any experimental group. The investigators were not blinded to the group allocation during the experiment and/or when assessing the outcome. All experiments were approved by the QIMR Berghofer Medical Research Institute Animal Ethics Committee.
Mouse B16F10 melanoma cells were grown in DMEM supplemented with 10% FCS (Bovogen), 1% Glutamine (Gibco), 1% HEPES (Gibco), and 1% Penicillin/Streptomycin (Gibco). LWT1 melanoma cells were cultured in RPMI1640, supplemented with 10% FCS (Bovogen), 1% Glutamine (Gibco), and 1% Penicillin-Streptomycin (Gibco). All cell lines were maintained at 37°C, 5% CO2. Cell injection and monitoring procedures were described in previous studies (32–34). All cell lines were routinely tested negative for Mycoplasma, but cell line authentication was not routinely performed. B16F10 melanoma (5 × 105) or LWT1 melanoma (7.5 × 105) cells were injected intravenously into the tail vein of WT or Tigit−/− mice. On days 0 and 3 after tumor inoculation, some mice were treated intraperitoneally with PBS or IL15/IL15Ra Complexes (R&D Systems) or cIg or anti-mouse TIGIT (4B1) at the indicated doses. Some groups of mice were additionally treated with an aCD226 mAb (480.1) to block CD226. In some experiments, CD3− NK1.1+ NK cells were sorted by flow cytometry from spleens of WT or Tigit−/− mice to 95% purity and WT or Tigit−/− NK cells (2 × 105) were injected intravenously into Rag2−/−γc−/− mice. After 6 days, blood was collected to check the equivalent reconstitution of NK cells by flow cytometry, and B16F10 (5 × 105 or 1 × 104) melanoma cells were injected intravenously into Rag2−/−γc−/− mice. Lungs were harvested on day 14 and metastatic colonies on the surface of the lungs were counted using a dissecting microscope.
Statistical analyses were performed in Prism Software (GraphPad). The normality of each variable was evaluated using the Shapiro–Wilk test. In case of normally distributed data, the comparison was performed using unpaired or paired two-tailed t tests, one-way ordinary or repeated-measures ANOVA tests followed by Tukey multiple comparisons test to compare all data together or Dunnett multiple comparisons test to compare all data with control. Data that were not normally distributed were compared with Wilcoxon matched-pairs signed rank tests (two paired groups) or Kruskal–Wallis test followed by Dunn multiple tests (more than two groups, unpaired). Linear regressions were evaluated with Pearson correlation tests. Significant differences were indicated for each figure and defined as ns (nonsignificant), P > 0.05; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.
NK cells downregulate TIGIT and CD226 in metastatic melanoma
We first evaluated the expression of TIGIT and CD226 by NK cells (CD45+CD3−CD56+ cells, Supplementary Fig. S1A) in the periphery and at tumor sites in patients with melanoma. We measured lower NK-cell frequencies in metastatic melanoma than in PBMCs of patients with melanoma (Fig. 1A). In sharp contrast with CD8+ T cells (15), TiNKs exhibited lower TIGIT expression [frequency and mean fluorescence intensity (MFI)] than cNKs isolated from patients with melanoma and healthy donors (Fig. 1B and C). TiNKs exhibited lower CD226 expression (frequency and MFI) than cNKs, resulting in an increased TIGIT to CD226 expression ratio (Fig. 1D). We observed a positive correlation between the percentages and MFI of TIGIT expression in cNKs and TiNKs for individual donors. We also observed a positive correlation between the percentages and MFI of CD226 expression in TiNKs, which downregulate CD226 expression, but not in cNKs. (Supplementary Fig. S1B). CD56dim NK cells were enriched in metastatic melanoma and downregulated both CD226 and TIGIT expression as compared with the periphery (Supplementary Fig. S1C). CD56bright NK cells, which represent a minority of NK cells in the periphery and at tumor sites, expressed low TIGIT level (Supplementary Fig. S1C). TiNKs and cNKs exhibited similar levels of CD96 and CD57 expression and both expressed low level PD-1 (Fig. 1E). In addition, TiNKs displayed lower frequencies of activation/maturation markers such as Tim-3, NKG2D, and CD16 (Fig. 1E). Notably, TiNKs exhibited lower 2B4, NKp46, and NKp30 expression (frequencies) as compared with healthy donor cNKs, but not with cNKs from patients with melanoma (Fig. 1E).
Collectively, our findings show that TiNKs are present at low frequencies in metastatic melanoma and downregulate both TIGIT and CD226 but not CD96 as compared with cNKs.
TIGIT+ NK cells exhibit higher lytic potential but lower degranulation capacity than TIGIT− NK cells in melanoma
We next investigated the functional capacities of TIGIT+ and TIGIT− NK cells isolated from the periphery and tumor sites of patients with melanoma (Fig. 2A). The frequencies of granzyme A+TIGIT− NK cells were significantly lower than granzyme A+TIGIT+ NK cells, both for cNKs and TiNKs (mean frequency 70% ± SD: 9% vs. 90% ± 6%, respectively, for cNKs, P < 0.01; and 41% ± 14% vs. 62% ± 12%, respectively, for TiNKs, P < 0.0001). The frequencies of granzyme B+ cells were also significantly lower in TIGIT+ TiNKs as compared with TIGIT+ cNKs (mean frequency 62% ± 10% vs. 86% ± 6.3%, P < 0.01, respectively). Finally, the frequencies of perforin+ cells were also significantly lower in TIGIT+ TiNKs as compared with TIGIT+ cNKs (mean frequency 64% ± 11% vs. 81% ± 10%, P < 0.001). Perforin+ NK-cell frequencies positively correlated with TIGIT+ NK-cell frequencies both in the periphery and at tumor site (Fig. 2B). Also, TIGIT+ TiNKs displayed lower expression of granzymes and perforin than TIGIT+ cNKs. To evaluate the implication of these findings, cNKs and TiNKs were isolated from PBMCs and tumors of patients with melanoma, respectively (Supplementary Fig. S1A), before degranulation assay in the presence of the MHC class I–deficient melanoma cell line FO-I or K562, which express high-level CD155 and CD112 but low-level MICA/B and ULBPs (Supplementary Fig. S1D). TiNKs exhibited lower degranulation capacity than cNKs, and CD107a+ TiNK frequencies inversely correlated with the TIGIT/CD226 expression ratio, supporting the role of the TIGIT/CD226 axis in modulating NK-cell functions (Fig. 2C; Supplementary Fig. S1E). We then evaluated the lysis of FO-I by TIGIT− and TIGIT+ NK cells sorted from PBMCs of patients with melanoma. TIGIT+ NK cells exhibited lower lytic activity and CD107a expression than TIGIT− NK cells (Fig. 2D). Altogether, our findings show that TiNKs are more dysfunctional than cNKs and that TIGIT+ NK cells display higher lytic potential but paradoxically lower lytic activity than TIGIT− NK cells, against CD155+ MHC class I–deficient melanoma cells.
TIGIT blockade alone failed to reverse NK-cell dysfunction in melanoma
We next evaluated the role of TIGIT and CD226 in regulating NK-cell–mediated reactivity against MHC class I–deficient melanoma (Fig. 2). NK cells isolated from PBMCs and tumor of patients with melanoma were stimulated with IL2 and IL15 prior to coculture with FO-I in the presence of blocking anti-(a)TIGIT and/or aCD226 as compared with IgG control mAbs. In-line with previously published findings (11, 27, 35), TIGIT blockade increased the frequencies of lytic cNKs, while CD226 blockade impeded NK-cell–mediated cytotoxicity and degranulation in the presence of FO-I (Fig. 2E; Supplementary Fig. S1F). The effects of TIGIT blockade on NK-cell–mediated cytotoxicity and degranulation capacity in the presence of FO-I were abrogated by CD226 blockade (Fig. 2E; Supplementary Fig. 1F). CD155 blockade had no significant impact on the degranulation capacity of cNKs against FO-I (Supplementary Fig. 1F). However, and in sharp contrast with the periphery, neither TIGIT blockade nor CD226 blockade significantly changed the low TiNK lytic activity against FO-I (Fig. 2F). TIGIT blockade modestly increased IFNγ production by cNK cells, but had no significant effect on TiNKs (Supplementary Fig. S1G).
Altogether, our findings show that, in sharp contrast with cNKs, TiNKs exhibit poor lytic function against MHC-deficient melanoma cells, which cannot be rescued upon TIGIT blockade.
Membrane bound CD155 induces CD226 internalization/degradation and NK-cell dysfunction
Because NK cells downregulate CD226 cell surface expression upon binding to membrane bound CD155 (7), we next investigated the role of membrane bound and soluble (s) CD155 in mediating the downregulation of CD226 and TIGIT by NK cells. To this end, cNKs were incubated in the presence of either CD155-Fc–coated beads, CD155-expressing cells including L cells transfected or not with human CD155 (L-CD155), K562, FOI, and dendritic cells, or sCD155 (Fig. 3). cNKs strongly downregulated CD226 expression and to a lesser extent TIGIT expression in presence of L-CD155, FO-I, K562, immature dendritic cells, and CD155-Fc–coated beads, but not L cells, resulting in an increased TIGIT/CD226 expression ratio (Fig. 3A and B). CD226 and TIGIT downregulation were abrogated in the presence of blocking aCD155 mAbs (Fig. 3A and B; Supplementary Fig. S2A), supporting that CD226 and TIGIT downregulation occurred in a CD155-mediated fashion. In contrast, NK cells did not significantly downregulate CD96 nor Tim-3 expression as control (Supplementary Fig. S2B). CD226 and TIGIT downregulation correlated with the level of CD155 expression (Supplementary Fig. S2C and S2D). Similarly, TiNKs showed decreased CD226 and increased TIGIT/CD226 expression ratio in correlation with the level of CD155 expression by the tumor cells in the tumor microenvironment (TME; Supplementary Fig. S2E). Notably, sCD155 or supernatants of melanoma cell lines containing high-level sCD155 (Supplementary Fig. S2C), did not induce CD226 downregulation by NK cells (Fig. 3C). CD155-induced CD226 downregulation by NK cells occurred at a posttranscriptional level because CD226 mRNA relative expression did not significantly change after 48-hour coculture with L-CD155 (Fig. 3D). CD226 downregulation by NK cells was not abrogated in the presence of protease inhibitors (TAPI-2 and GI254023X), suggesting that it did not occur upon ectodomain shedding (Supplementary Fig. S2F). Notably, NK cells expressed similar Annexin V expression in the presence of L-CD155 cells or CD155-Fc–coated beads, respectively, and compared with L cells or IgG beads, respectively (Supplementary Fig. S2G).
To investigate the fate of TIGIT and CD226 upon binding to mbCD155, including internalization, degradation, and recycling, cNKs were coincubated (1 hour) with FO-I, L-CD155, L cells, or no cells prior to intracellular and extracellular staining of TIGIT, CD226, and CD45 followed by ImageStream flow cytometry. In presence of L-CD155 and FO-I, but not L cells, the cell surface expression of CD226 and TIGIT strongly decreased together with an increased intracellular expression of both molecules (Fig. 3E and F), resulting in increased internalization ratios (Fig. 3G). To investigate whether CD226 and TIGIT undergo ligand-induced endocytosis followed by recycling or degradation (36), we next evaluated their surface and total cellular expression in cNKs incubated with L cells or L-CD155 with or without V-ATPases inhibitors (16 hours; Fig. 4). In the presence of L-CD155, but not L cells, total CD226 but not TIGIT expression in NK cells (MFI) sharply decreased as compared with L cells (Fig. 4A–C). V-ATPase inhibitors, which inhibit protein degradation by acidified lysosomes (37), abrogated the decrease in total cellular CD226 expression (Fig. 4B; Supplementary Fig. S3A and S3B).
We next assessed the functional implication of CD155-mediated CD226 degradation on NK-cell functions. To this end, cNKs isolated from patients with melanoma were incubated in the presence of CD155-Fc- or IgG-coated beads and/or aTIGIT and/or aCD226 mAbs prior to stimulation with IL2 and IL15, and incubation with FO-I. CD155-Fc–treated cNKs exhibited lower CD107a expression as compared with IgG-treated cNKs (Fig. 4D). TIGIT blockade partially reversed the effect of CD155-Fc on cNKs, while CD226 blockade had no significant effect (Fig. 4E; Supplementary Fig. S3C). As controls, IgG-coated beads and sCD155 had no significant impact on cNK cytolytic activity.
Altogether, our data show that mbCD155, but not sCD155, induced both TIGIT and CD226 internalization, and CD226, but not TIGIT, degradation. They also show that mbCD155-mediated CD226 degradation promoted NK-cell dysfunction that was only partially reversed upon TIGIT blockade.
IL15 together with TIGIT blockade reverses CD155-mediated NK-cell exhaustion and impedes experimental melanoma metastasis in vivo
IL15 upregulates CD226 expression by NK cells (38, 39). In addition, IL15 together with IL12 also upregulates TIGIT expression by cNKs isolated from patients infected with the human immunodeficiency virus (40). We next investigated whether IL15, which increases NK-cell–mediated antitumor activity, regulates CD226 and TIGIT expression by cNKs and TiNKs as compared with other γ-chain cytokines including IL2, IL15, IL7, and IL21. IL15 induced higher CD226 and TIGIT expression after 6-day stimulation than IL2, IL7, or IL21 (Supplementary Fig. S4A). We then incubated cNKs in the presence of IL15 and L cells or L-CD155 prior to flow cytometry and RT-PCR for CD226 and TIGIT expression. IL15 increased CD226 and TIGIT cell surface expression that was abrogated in the presence of STAT3 or STAT5 inhibitors (Fig. 5A). IL15 also increased Tim-3 and NKG2D expression by NK cells (Supplementary Fig. S4A) as reported previously (41–43). IL15 increased CD226 and TIGIT gene expression with or without L-CD155 (Fig. 5B). To evaluate whether IL15 reversed the downregulation of CD226 cell surface expression upon mbCD155, cNKs were incubated in the presence of IL15 and L cells, L-CD155, or FOI prior to flow cytometry and ImageStream. In the presence of L-CD155 or FO-I, IL15 significantly increased both cell surface and total CD226 expression by cNKs (Fig. 5C) in contrast with the other γ-chain cytokines (Supplementary Fig. S4B), but did not significantly impede CD226 internalization (Fig. 5D). Notably, IL15 also increased the expression of both CD226 and TIGIT by TiNKs (Fig. 5E).
We next assessed whether IL15 alone or together with TIGIT blockade increased TiNK-mediated antitumor activity and counteracted CD155-mediated NK-cell dysfunction. To this end, cNKs preincubated with CD155-Fc beads or IgG beads, or TiNKs were treated with IL15 (16 hours or 6 days) prior to evaluating their level of granzyme and perforin expression as well as CD107a expression in the presence of FO-I with or without aTIGIT mAbs. Prolonged IL15 (6 days) increased cNK and TiNK degranulation in the presence of FO-I (Fig. 5F) as well as perforin, granzyme A, and granzyme B expression (Supplementary Fig. S4B and S4C) as compared with overnight stimulation with IL15. Prolonged IL15 stimulation alone (mean fold change: 1.49 ± SD 0.32), but not TIGIT blockade (1.17 ± 0.15), significantly increased TiNK degranulation in the presence of FO-I as compared with overnight stimulation with IL15. Prolonged IL15 together with TIGIT blockade further increased the degranulation of TiNK in the presence of FO-I (1.83 ± 0.48; Fig. 5G; Supplementary Fig. S4E).
To support the relevance of our findings in vivo, we investigated whether IL15 and TIGIT blockade promoted NK-cell–mediated control of metastatic tumors in two mouse melanoma models (Fig. 6A). Therapy with aTIGIT mAbs induced antimetastatic activity against established B16F10 and LWT1 lung metastases only in mice treated with IL15/IL15Ra complexes (mean B16F10 metastasis numbers: 72 ± SD: 23, 144 ± 40, 324 ± 43, and 344 ± 36, respectively, for IL15/IL15R + aTIGIT mAbs, IL15/IL15R, aTIGIT mAbs, and IgG + PBS, respectively; Fig. 6B; Supplementary Fig. S5A and S5B). IL15 was more effective in Tigit−/− mice compared with WT mice, despite Tigit−/− mice having no significant reduction in metastasis compared with WT mice, as published previously (Fig. 6B; Supplementary Fig. S5A; refs. 32, 33). When NK cells from Tigit−/− and WT mice were transferred into immunodeficient Rag2−/−γc−/− mice (Fig. 6A, C, and D), they reduced B16F10 lung metastases by half as compared with no NK-cell transfer (393 ± 38 vs. 186 ± 35 and 192 ± 42, respectively), but IL15/IL15Rα complexes were more effective in reducing metastases in mice transferred with Tigit−/− as compared with WT NK cells (31 ± 17 and 100 ± 36, respectively; Fig. 6D). Similar findings were observed in mice with the LWT1 lung metastases (Supplementary Fig. S5C). In-line with previous findings, CD226 blockade significantly increased the number of lung metastases in mice. Interestingly, CD226 blockade abrogated the effects of IL15 alone or in combination with TIGIT blockade (Fig. 6E; Supplementary Fig. S5B). In addition, NK cell but not CD8+ T-cell depletion abrogated the antitumor effects of combined IL15/IL15R and TIGIT blockade in both B16F10 and LWT1 models. These findings support the role of IL15/IL15R and TIGIT blockade in promoting direct NK-cell–mediated tumor reactivity against lung metastases in vivo (Supplementary Fig. S5D).
Altogether, our findings show that IL15 increased CD226 and TIGIT expression by NK cells in a STAT3/5-dependent fashion. Prolonged IL15 stimulation together with TIGIT blockade increased the TiNK degranulation capacity and lysis of MHC class I–deficient melanoma. TIGIT blockade and TIGIT loss in NK cells were effective against tumor metastasis only in the presence of IL15. The antitumor activity of IL15 and TIGIT blockade were abrogated by CD226 blockade, supporting that TIGIT and CD226 act antagonistically to regulate the antitumor effector function of NK cells in vivo. Interestingly, the antitumor effects of Tigit−/− NK cells in Rag2−/−γc−/− mice with melanoma suggest that TIGIT depletion in NK cells enhanced NK-cell–mediated antitumor reactivity, independently of CD8+ T cells.
In this study, our findings support the development of combinatorial immunotherapy with IL15 and TIGIT blockade to reinvigorate TiNKs against MHC class I–deficient melanoma. We observed that NK cells were present at low frequencies in human metastatic melanoma, were more dysfunctional, and downregulated both TIGIT and CD226 as compared with cNKs. CD226 expression is downregulated in TiNKs as compared with cNKs and correlates with the levels of CD155 expression by melanoma cells in the TME. Noteworthy, CD226 appears to regulate T-cell responses to PD-1 blockade and combinatorial therapy in two mouse tumor models (44). Upon PD-1 blockade, T cells upregulate CD226, which is the substrate for dephosphorylation by SHP2 upon PD-1 engagement (44). These findings in mouse T cells may not be relevant to human NK cells for several reasons. First, these data were obtained in mouse tumor models that are responsive to PD-1 blockade with no significant CD226 downregulation by T cells in the TME, unlike solid human tumors (15). Second, and most importantly, the reported CD226 effects are mediated by PD-1 signaling in T cells (SHP2), which upregulates PD-1 in the TME. In sharp contrast with these findings, we show that human NK cells, unlike mouse NK cells, do not upregulate PD-1 expression in the periphery nor at tumor sites. TiNKs displayed decreased expression of multiple activation/maturation markers including CD16 and NKG2D. Our findings are reminiscent of previous studies supporting that the downregulation of multiple ARs by TiNKs is associated with NK-cell dysfunction and decreased tumor lysis capacity in many human tumors, including breast, ovarian, lung, and prostate cancers (4–7, 9). Furthermore, TiNKs exhibit decreased degranulation capacity that inversely correlated with the percentages of invading melanoma tumor cells (26), supporting the role of tumor cells in driving NK-cell dysfunction. The phenotypic features of TiNKs strongly contrasts with those of human CD8+ tumor-infiltrating lymphocytes, which upregulate TIGIT as well as others IRs like PD-1 and Tim-3 through Prdm1 and c-Maf activation downstream of T-cell receptor (TCR) activation (45), while they downregulate CD226 expression in the TME, resulting in an imbalance of TIGIT/CD226 expression. Strikingly, our findings in melanoma sharply contrast with those recently published in colon cancers, suggesting increased TIGIT expression by intratumor NK cells as compared with peritumor NKs (27). In this study including 19 colon tumors, it is unclear how the investigators precisely isolated low-frequency intratumor and peritumor NK cells for flow cytometry. One may also wonder whether the pathogens in the gut microbiome, which act on innate and adaptive immunity (46), may critically influence the activation and phenotype of NK cells in colon tumors. In addition, and in contrast to humans, mouse cNKs express very low-level TIGIT (19, 47), which makes comparative studies of TIGIT expression between mice and humans very difficult.
Several lines of evidence support the role of the TIGIT/CD226 axis in regulating NK-cell–mediated tumor killing capacity. As compared with TIGIT− NK cells, TIGIT+ NK cells, both in the periphery and at tumor site, upregulated multiple activation/effector markers including granzymes and perforin, supporting that TIGIT is a marker of NK-cell activation (48). Paradoxically, TIGIT+ NK cells exhibited lower killing activity against CD155+ MHC class I–deficient melanoma FO-I as compared with TIGIT− NK cells. Such findings are in-line with previously published findings (48, 49). They are also reminiscent of exhausted CD8+ T cells, which upregulate multiple IRs, are dysfunctional, and display increased levels of perforin and granzymes despite lower degranulation capacities in the presence of target cells as compared with nonexhausted CD8+ T cells (50, 51). The degranulation capacity of cNKs and TiNKs against CD155+ MHC class I–deficient melanoma FO-I inversely correlated with cell surface TIGIT/CD226 ratio, supporting the role of the TIGIT/CD226 axis in regulating NK-cell effector functions. Our findings showed that mbCD155, but not sCD155, induced both CD226 and TIGIT internalization with CD226, but not TIGIT, degradation. Multiple ARs and IRs, which regulate NK-cell activation, are downregulated upon binding to their respective ligands, including NKG2D and 2B4 (52, 53). Receptor downregulation upon ligand binding contributes to the regulation of receptor signaling in NK cells (54). In sharp contrast with NKG2D and soluble MIC-A/MIC-B (52), sCD155, which is present in the serum of patients with advanced cancers (55), did not induce CD226 endocytosis. In addition, CD226 downregulation correlated with the levels of mbCD155 expression, supporting that CD226 downregulation electively occurs in the TME. Although, CD226 shedding by metalloproteases has been previously reported in vitro (56), metalloprotease inhibitors failed to prevent CD155-mediated CD226 downregulation by cNKs. Notably, CD96, which also binds to CD155, was not significantly downregulated by NK cells in presence of mbCD155 in vitro. Collectively, our findings support that mbCD155 acts as a master regulator of the TIGIT/CD226 axis to limit NK-cell tumor killing capacities in the TME. This observation adds to the previous findings supporting that the downregulation of ARs, such as NKG2D, NKp46, and NKp30, impedes the tumor killing capacity of TiNKs (4–7). The nature of the intracellular motifs that differentially drive the internalization, recycling, or degradation of TIGIT, CD226, and CD96 remain to be identified.
TIGIT blockade alone increased cNK cell killing of CD155+ MHC class I–deficient melanoma, which was abrogated upon CD226 blockade, but had no significant effect on dysfunctional TiNKs, which downregulate both TIGIT and CD226 and exhibit lower killing capacities. While IL15 increased NK-cell–mediated killing capacity, it also increased CD226 and TIGIT expression by NK cells in a STAT3/5-dependent fashion. Prolonged IL15 stimulation together with TIGIT blockade increased the human TiNK–mediated lysis of MHC class I–deficient melanoma. TIGIT blockade or TIGIT loss in NK cells only decreased tumor metastasis in two lung metastasis mouse models in the presence of IL15. Interestingly the effects of IL15 and TIGIT blockade on NK cells in vivo were abrogated upon CD226 blockade. These findings support that TIGIT and CD226 exert antagonistic effects to regulate the antitumor effector function of NK cells. They also suggest that combinatorial therapy with IL15 and TIGIT blockade promotes CD226 engagement of CD155 on NK cells to augment their effector functions that may occur through phosphorylation-mediated inactivation of transcription factor FOXO1 (57). We cannot exclude that the upregulation of other ARs like NKG2D by NK cells upon IL15 plays a role in NK-cell–mediated tumor reactivity. This will need to be further investigated. Interestingly, sustained IL15 stimulation of NK cells in vitro and in vivo appears to promote initial proliferation and maturation, followed with NK-cell exhaustion with impaired activation, cytotoxicity, and proliferative capacity (58, 59). Therefore, IL15 dosage and administration schedule will need to be carefully designed to avoid NK-cell exhaustion.
In a recent study in mouse tumor-bearing models with lung metastases (27), the therapeutic effects of PD-1, TIGIT, or dual PD-1/TIGIT blockade were reported to act primarily on NK cells to enhance antitumor activity mediated by CD8+ T cells. The mechanisms used by NK cells to regulate adaptive immunity upon TIGIT blockade have not yet been elucidated. Our findings in mice and in humans do not support these conclusions for several reasons. First, TIGIT blockade and TIGIT loss in NK cells were not effective against lung melanoma metastasis in the absence of IL15. Second, the antitumor effects of Tigit−/− NK cells in Rag2−/−γc−/− mice with melanoma showed that TIGIT depletion in NK cells enhanced NK-cell–mediated antitumor reactivity, but only in the presence of exogenous IL15, and independently of CD8+ T cells. Whether NK cells participate in the environmental signals guiding CD8+ T-cell priming, development of CD8+ effector T cells into CD8+ memory T cells, or CD8+ memory T-cell maintenance remains to be evaluated. Additional mechanistic studies are needed to thoroughly investigate these questions.
In summary, this study shows that mbCD155 triggers CD226 internalization and degradation by NK cells, resulting in increased cell surface TIGIT/CD226 expression ratio and decreased NK-cell–mediated tumor reactivity. IL15 together with TIGIT blockade reinvigorates TiNK-mediated killing of melanoma cells in vitro and in vivo in a CD226-dependent fashion. Altogether, our findings may support the development of novel combinational immunotherapy with IL15 and TIGIT blockade to promote NK-cell–mediated killing of MHC-deficient tumors that are refractory to CD8+ T-cell–mediated immunity.
Disclosure of Potential Conflicts of Interest
D. Davar reports grants from BMS, Checkmate Pharmaceuticals, CellSight Technologies, Merck, GlaxoSmithKline/Tesaro, and MedPacto, and personal fees from Immunocore and Array Biopharma outside the submitted work. J.M. Kirkwood reports grants and personal fees from Amgen, grants, personal fees, and non-financial support from BMS, Checkmate, and Novartis, and grants from Castle Biosciences, Immunocore, and Iovance outside the submitted work. R.J. Johnston reports employment with Bristol Myers Squibb, a company which develops drugs for profit. A.J. Korman reports other financial interests from BMS (shareholder) outside the submitted work, as well as a patent for tigit abs. M.J. Smyth reports grants from Bristol Myers Squibb (scientific research agreement) during the conduct of the study, as well as grants from Tizona Therapeutics (scientific research agreement), and personal fees from Tizona Therapeutics (scientific advisory board) and Compass Therapeutics (scientific advisory board) outside the submitted work. H.M. Zarour reports grants from NIH/NCI (R01CA228181 and P50CA121973), Bristol-Myers Squibb (research contract), and Cancer Research Institute (CVC team grant) during the conduct of the study, as well as grants from Bristol-Myers Squibb (research contract) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.
J.-M. Chauvin: Formal analysis, investigation, writing-original draft, writing-review and editing. M. Ka: Formal analysis, investigation. O. Pagliano: Formal analysis, investigation. C. Menna: Investigation, project administration. Q. Ding: Formal analysis, investigation. R. DeBlasio: Investigation, project administration. C. Sander: Resources. J. Hou: Formal analysis, investigation. X.-Y. Li: Formal analysis, investigation, writing-original draft. S. Ferrone: Resources. D. Davar: Resources. J.M. Kirkwood: Resources. R.J. Johnston: Resources. A.J. Korman: Resources. M.J. Smyth: Formal analysis, investigation, writing-original draft, writing-review and editing. H.M. Zarour: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, writing-original draft, writing-review and editing.
This work was supported by NIH/NCI grants R01CA228181 and R01CA222203 (to H.M. Zarour), a research grant by Bristol-Myers Squibb BMS (to H.M. Zarour), a cancer vaccine collaborative clinical strategy team grant (to H.M. Zarour), and NCI grant P50CA121973 (to J.M. Kirkwood). M.J. Smyth was supported by a National Health and Medical Research Council of Australia (NH&MRC) Senior Principal Research Fellowship (1078671), a NH&MRC Program grant (1132519), a NH&MRC Project grant (1124784), a CLIP award from the Cancer Research Institute (New York, NY), and a Project Grant from the Cancer Council of Queensland (1140251). This work benefited from ImageStreamX MARKII grant NIH 1S10OD019942-01.
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.