Immune checkpoint therapies target tumor antigen-specific T cells, but less is known about their effects on natural killer (NK) cells, which help control metastasis. In studying the development of lung metastases, we found that NK cells lose their cytotoxic capacity and acquire a molecular signature defined by the expression of coinhibitory receptors. In an effort to overcome this suppressive mechanism, we evaluated NK cell responses to the immunostimulatory cytokine IL12. Exposure to IL12 rescued the cytotoxicity of NK cells but also led to the emergence of an immature NK cell population that expressed high levels of the coinhibitory molecules PD-1, Lag-3, and TIGIT, thereby limiting NK cell–mediated control of pulmonary metastases. Notably, checkpoint blockade therapy synergized with IL12 to fully enable tumor control by NK cells, demonstrating that checkpoint blockers are not only applicable to enhance T cell–mediated immunotherapy, but also to restore the tumor-suppressive capacity of NK cells. Cancer Res; 77(24); 7059–71. ©2017 AACR.
Metastasis—the spread of malignant cells from the tumor of origin to distant sites—represents the main cause of cancer-associated death (1). Metastasis formation is a tightly regulated multistep process involving the crosstalk between tumor cells and components of the surrounding local microenvironment, which together influence disseminating tumor cells. Especially, leukocytes have been demonstrated to either restrict or potentiate metastatic growth (2, 3).
As pivotal players of the innate immune defense, NK cells have shown the potential to recognize and reject tumor cells in various tumor models, being particularly effective in controlling metastatic dissemination (4, 5). To recognize malignant cells, NK cells are equipped with a diverse repertoire of activating and inhibitory receptors that tightly regulate their activity (6). Thus, whereas healthy cells avoid NK cell–mediated killing through the engagement of inhibitory receptors by MHC class I molecules, malignant cells may either lose this signal or upregulate activating receptors as a result of cellular stress (7), leading to tumor cell elimination by cytotoxic granules and death receptors (8).
Evidence for NK cell–mediated tumor surveillance has also been reported in cancer patients, with correlative studies associating high levels of tumor-infiltrating NK cells with favorable prognosis (9–12). However, due to the low NK cell numbers found within the tumor microenvironment (TME; refs. 13, 14) and the tumor-induced NK cell suppression (15–19), the extent to which they contribute to immune surveillance against primary tumors is unclear. Attempts to sustain antitumor NK cell responses have focused on their ex vivo expansion and activation using cytokines such as IL2, IL15, or IL12 (20). Although pilot studies showed a therapeutic potential of cytokine-activated NK cells, their antitumor activity is restricted by the TME (21), which comprises several suppressive immune cell types and tumor-secreted factors (22).
Checkpoint blockers, which target coinhibitory receptors or their ligands, are among the most promising approaches aimed to antagonize TME-mediated immune suppression and have by now yielded significant clinical benefit against several types of cancer (23). Even though checkpoint blockade has been mainly exploited to liberate tumor antigen-specific T cells, evidence for a role of immune checkpoints on other tumor-infiltrating leukocytes, such as NK cells, is only now emerging (24). This could be of particular importance in distal metastatic sites such as the lung, an organ where NK cells are abundant (25). However, the application of checkpoint blockade therapy to specifically target metastases requires a better understanding of the mechanisms regulating NK cell antitumor responses operating at late-stage disease.
In preclinical mouse models of metastasis, we show that lung NK cells are rendered dysfunctional during the course of cancer progression. Concomitant with their impaired ability to eliminate metastasizing tumor cells, lung NK cells displayed reduced levels of activating receptors, whereas the expression of coinhibitory receptors, classically described on T cells, was increased. We used IL12 to therapeutically boost lung NK cell cytotoxicity, but the antimetastatic activity was limited through a further dramatic increase of checkpoint receptors on the NK cell surface. Coadministration of checkpoint blockers against PD-1 or Lag-3 increased the efficacy of IL12 immunotherapy in an NK cell–dependent manner. Taken together, these data demonstrate that blocking immune checkpoints on NK cells can be exploited to complement NK cell–based antimetastatic therapies.
Materials and Methods
Female (6–10 weeks old) BALB/c and C57BL/6 mice were obtained from Janvier labs. MMTV-PyMT (FVB/N-Tg(MMTV-PyVT)634Mul/J) mice were obtained from The Jackson Laboratory. C57BL/6 Ncr1cre/wt mice (NKp46iCre) were provided by Eric Vivier and Rosa26iDTR mice by Ari Waisman (Mainz, Germany). Nkp46iCre mice were crossed to Rosa26iDTR mice to obtain NKp46iCre/wtRosa26iDTR/wt (Nkp46iCreR26RiDTR) or NKp46iCre/wtRosa26wt/wt controls (Nkp46iCre). Mice homozygous for the Tcrdtm1Mom targeted mutation (Tcrd−/− mice) were from The Jackson Laboratory (Stock #002120; ref. 26). All animals were kept in house according to institutional guidelines under specific pathogen-free conditions. All animal experiments were approved by the Swiss cantonal veterinary office (licenses 147/2012 and 142/2015).
Murine tumor cell lines
4T1 cells were provided by M. Detmar (Institute of Pharmaceutical Sciences, ETH, Zurich, Switzerland), YAC-1 cells by M. van den Broek (Institute of Experimental Immunology, University Zurich, Zurich, Switzerland), and 4T1-mCherry cells by Nicola Aceto (Department of Biomedicine, University of Basel, Basel, Switzerland). LLCs and B16F10 melanoma cells were purchased from ATCC. All cancer cells were used between passages 2 and 6. Cell lines have been tested by Mycoscope PCR Detection Kit (Genlantis) and authenticated according to ATCC STR database.
Expression and purification of IL12Fc
IL12Fc expressed in 293T cells was purified from supernatant using a protein A column (1 mL, HiTrap, GE Healthcare). After elution with 0.1 mol/L citric acid, pH 3.0 using a purifier (ÄktaPrime) and dialysis for 40 hours in PBS, pH 7.4, the concentration and purity of IL12Fc were measured using the mouse IL12 (p40) ELISA kit (BD OptEIATM, 555165) and silver staining (Pierce Silver Stain Kit, Thermo Scientific), respectively.
Models of tumor metastasis
Cells (1 × 105 4T1) were injected in the second mammary fat pad of 6- to 10-week-old female BALB/c mice. Lung metastases were analyzed after approximately 23 days. To quantify NK cell ligands on 4T1 cells, 4T1-mCherry cells were injected. For experimental metastasis, Balb/c mice were injected intravenously with 1 × 105 4T1 cells into the tail vein and C57BL/6 WT and Tcrd−/− mice with 4 × 105 LLCs or 3 × 105 B16F10 melanoma cells. For the resection experiments, 1 × 105 4T1 cells were injected in the fourth mammary fat pad of 10-week-old female Balb/c mice. To resect primary tumors, mice were intraperitoneally (i.p.) injected with fluniximin (Biokema; 5 mg/kg body weight) before being anesthetized with 3% to 5% isoflurane (Minrad) in an induction chamber. Anesthesia was maintained at 3% isoflurane delivered through a nose adaptor. Tumors weighing about 0.5 to 0.8 g were removed by blunt dissection using sterilized instruments.
Administration of IL12 and checkpoint blockers
A total of 200 ng of IL12Fc or the IgG fragment as control (Ctrl) diluted in 25 μL PBS were intranasally (i.n.) administered per mouse. MMTV-PyMT mice were treated three times per week starting at week 6. Balb/c mice were treated one day before primary tumor resection and subsequently 3 times per week starting 1 day after resection. 4T1 tumor–bearing mice were treated i.n. with 50 ng of IL12Fc in combination with i.p. administration of 200 μg of anti–Lag-3 antibodies (C9B7W, BioXCell), anti–PD-1 antibodies (RMP1-14, BioXCell) or corresponding isotype IgG (2A3, BioXCell) at day 12 after tumor cell injection. Mice were treated three times per week and studies were terminated at day 25. In the experimental metastasis model, mice were treated with 150 μg of anti–Lag-3 antibodies (C9B7W, BioXCell), anti–PD-1 antibodies (RMP1-14, BioXCell), or corresponding isotype IgG (2A3, BioXCell) and 200 ng of IL12Fc.
Balb/c mice were treated i.p. with 50 μL of anti-asialo GM1 antibodies (Wako Pure Chemical Industries) or rabbit IgG (Sigma) at days 3, 7, 10, and 16 after tumor cell injection. MMTV-PyMT mice were treated i.p. with 50 μL of anti-asialo GM1 antibodies (Wako Pure Chemical Industries) or rabbit IgG (Sigma) twice a week starting at week 9 of age. Lung metastases in MMTV-PyMT mice were quantified at 14 weeks of age via India ink. In the LLC experimental metastasis model, 50 μL of anti-asialo GM1 antibodies (Wako Pure Chemical Industries) were administered i.p. at day −1 and day 7 after tumor cell injection. Anti-NK1.1 antibody (200 μg/mouse, clone PK136, BioXCell) or Diphtheria Toxin from Corynebacterium diphtheria (Calbiochem, 250 ng/mouse for initial depletion and 125 ng/mouse for the following injections) diluted in PBS were injected i.p. at days −1, 2, 5, 8, and 11 after tumor inoculation. Anti-CD4 (100 μg/mouse, GK1.5, BioXCell) was injected i.p. at day −1, 6, 12, and 17 after tumor inoculation. In Tcrd−/− mice inoculated with B16F10 melanoma cells, lung colonies were counted 21 days after tumor inoculation. Depletion efficiency was monitored during the studies by bleeding the mice and subsequent flow-cytometric analysis.
Quantification of lung metastases
Pulmonary metastases were quantified by intratracheal injection of India ink (15% India ink in PBS). India ink–injected lungs were fixed in Feket's solution (300 mL 70% ethanol, 30 mL 37% formaldehyde, and 5 mL glacial acetic acid) overnight. White lung metastases were counted under a dissection microscope. Metastatic index was calculated as the number of lung metastases divided by the primary tumor weight.
Flow-cytometric analysis of lungs from 4T1 tumor–bearing mice was performed at day 10 or day 20 after tumor cell injection. Lungs were harvested, digested with Collagenase IV (0.4 mg/mL) for 45 minutes at 37°C and erythrocytes were lysed with ACK (ammonium–chloride–potassium) lysis buffer. Tibiae and femora were flushed with PBS and erythrocytes were lysed subsequently. Cells were incubated for 20 minutes in Fc-blocking buffer (2.4G2) before being stained with the following antibodies: anti-CD45 (30-F11, BioLegend), anti-CD49b (Dx5, BioLegend), anti-NKp46 (29A1.4, eBioscience), anti-CD3 (17A2, eBioscience), anti–Gr-1 (6-8C5, BioLegend), anti-CD11b (M1/70, BioLegend), anti-CD27 (LG.3A10, BioLegend), anti-NKG2D (CX5, eBioscience), anti–DNAM-1 (10E5, BioLegend), anti-KLRG1 (2F1, eBioscience), anti-Ly49G2 (eBio4D11, eBioscience), anti-TIGIT (1G9, BioLegend), anti–Lag-3 (eBioC9B7W, eBioscience), anti–PD-1 (29F.1A12, BioLegend), anti-Ly6G (1A8, BioLegend), anti-SiglecF (E50-2440, BD), anti-F4/80 (CI:A3-1, AbD Serotec), anti-Ly6C (HK1.4, BioLegend), anti-CD11c (N418, BioLegend), anti-MHC II (10F.9G2, BioLegend), anti-H60 (205326, R&D), anti–Rae-1 (186107, R&D), anti-CD155 (690912, R&D), anti–PD-L1 (10F.9G2, BioLegend), anti-MHC I (34-2-12, BioLegend) anti-NK1.1 (PK136, Biolegend), anti-CD94 (18d3, Biolegend), anti-NKG2A/C/E (20d5, BD Bioscience) and anti-Ly49E/F (CM4, eBioscience). To exclude dead cells, we used the Zombie Aqua fixable viability kit (BioLegend). Doublets were excluded by FSC-A/FSC-H gating. For intracellular cytokine staining, cells were stimulated for 4 hours at 37°C and 5% CO2 in RPMI 1640 medium containing 10% FCS, 50 ng/mL PMA, 500 ng/mL ionomycin and 1 μL/mL GolgiPlug (BD Bioscience). For detection of intracellular IFNγ or granzyme B, cells were fixed after surface staining, permeabilized with Cytofix/Cytoperm (BD Biosciences) and stained with an anti-IFNγ mAb (XMG1.2, BD) or an anti-granzyme B mAb (GB11, BD Bioscience). Acquisition was performed on an LSRII Fortessa flow cytometer (BD Bioscience), and data were analyzed using FlowJo Version X (TreeStar). Absolute cell numbers were quantified using AccuCheck Counting beads (Life Technologies).
RNA from NK cells (CD45+CD3−Ly6G−NKp46+CD49b+) isolated at days 10 and 20 after 4T1 tumor cell injection was amplified using the SMART-Seq v4 Ultra Low Input RNA Kit (Clontech) at the New York Genome Center. After synthesis, cDNA was sheared to a size of approximately 350 bp, and libraries were generated using the KAPA Hyper DNA Library Prep according to the manufacturer's instructions. Final libraries were quantified using the KAPA Library Quantification Kit (KAPA Biosystems), Qubit Fluorometer (Life Technologies) and Agilent 2100 BioAnalyzer, and were sequenced on an Illumina HiSeq2500 sequencer (v4 chemistry) using 2 × 125 bp cycles aiming for 30 million reads per sample. Differential gene expression for NK cells isolated from day 10 versus day 20 was calculated using DESeq2, and differentially expressed genes with a P value of less than 0.05 were used for hierarchical clustering.
Using the SMART-seq2 Amplification Kit (Clontech), RNA purified from NK cells (CD45+CD3−Ly6G−CD122+CD49b+) isolated from Ctrl- or IL12-treated lungs was converted into complementary DNA libraries, amplified and sequenced for 200 to 250 million reads using 50 bp paired-end read at the Quantitative Genomics Facility in Basel. FastQC was used to quality check the reads. Low-quality ends were clipped (3 bases from the start and 10 bases from the end). Trimmed reads were aligned to the reference genome and transcriptome (FASTA and GTF files were downloaded from the UCSC mm10 repository) using STAR version 2.3.0e_r291 with default settings. Distribution of the reads across genomic isoform expression was quantified using the R package GenomicRanges from Bioconductor (Version 3.0). Differentially expressed genes were identified using the R package edgeR from Bioconductor (Version 3.0). By setting a minimum mean of expression of 100 reads, very lowly expressed genes were filtered out. Sequencing information is available at the European Bioinformatics Institute (EBI; ENA: PRJEB15668).
Gene ontology analysis
Database for Annotation, Visualization, and Integrated Discovery (DAVID; http://david.abcc.ncifcrf.gov) was used to identify enriched biological functions and assess gene lists for enrichment of genes annotated with specific Gene Ontology Project (GO) biological process terms (http://www.geneontology.org). As input for DAVID ontology analysis, a list of 1,722 genes that were differentially expressed in NK cells day 10 compared to day 20 was created (P < 0.05). Heat maps corresponding to the different functional categories within this list were generated using the R package and an adjusted P value [false discovery rate (FDR)] < 0.1.
Lung NK cells (live, CD45+CD3−NKp46+CD49b+CD11b+CD27−) were sorted with a BD FACSAria III sorter. YAC-1 cells were stained with PKH26 Red Fluorescent Cell Linker Mini Kit (Sigma, MINI26-1KT). NK cells were incubated with YAC-1 target cells at effector:target ratios of 1:1, 3:1, and 6:1 for 5 hours at 37°C in 5% CO2. After removal of medium, Topro (0.8 μmol/L) was added to the cells and cells were acquired on an LSRII Fortessa flow cytometer (BD). Data were analyzed using FlowJo Version X (TreeStar). The percentage of specific lysis was calculated as follows: [(Experimental lysis − spontaneous lysis)/(maximum lysis − spontaneous lysis)] × 100%. Alternatively, sorted NK cells were added to 4T1-mCherry cells at effector:target ratios of 1:1, 5:1, and 10:1 and incubated for 10 hours at 37°C in 5% CO2. After removal of medium, mCherry fluorescence was detected with a microplate reader (Tecan Infinite M200 PRO, Männedorf, Switzerland). The percentage of lysis was calculated as followed: [100 − (Sample fluorescence × 100/4T1 alone fluorescence]. Cell supernatants were collected and subjected to IFNγ analysis by ELISA.
Quantitative real-time PCR
Lung NK cells (CD45+CD3−Ly6G−CD49b+NKp46+) from mice being treated with either Ctrl or IL12 together with either Ctrl IgG, anti–PD-1 or anti–Lag-3 mAb after i.v. injection of 4T1 tumor cells were purified with an AriaIII Sorter. RNA was isolated from NK cells using the Qiagen Micro Kit according to the manufacturer's protocol. Random primers (Invitrogen) were used for synthesis of complementary DNA and the following primers for quantitative real-time PCR using a CFX384 Cycler (Bio-Rad Laboratories): Prf1 5′-TCTTGGTGGGACTTCAGCTT-3′ and 5′-GAGCAGGGACAGGTCGTG-3′, Ifng 5′-GCATTC ATGAGTATTGCCAAG-3′ and 5′-GGTGGACCACTCGGATGA-3′, PolII 5′-CTGGTCCTTCGAATCCGCATC-3′ and 5′-GCTCGATACCCTGCAGGGTCA-3′, Pdcd1 5′-CGTCCCTCAGTCAAGAGGAG-3′ and 5′-GTCCCTAGAAGTGCCCAACA-3′, Cd226 5′-TCGCTCAGAGGCCATTACAG-3′ and 5′-CCCTGGGCTCTTTAAGTGGAA-3′, Csf2 5′-TGGAAGCATGTAGAGGCCATCA and 5′-GCGCCCTTGAGTTTGGTGAAAT-3′, Cxcr6 5′-TACGATGGGCACTACGAGGGAG-3′ and 5′-GCAAAGAAACCAACAGGGAGACCAC-3′, Cx3cr1 5′-GGTCTGGTGGGAAATCTGTTGG-3′ and 5′-GAAGAAGGCAGTCGTGAGCTTG-3′, Ccr5 5′-ACTGCTGCCTAAACCCTGTCA-3′ and 5′-GTTTTCGGAAGAACACTGAGAGATAA-3′, Cxcr4 5′-GAAGTGGGGTCTGGAGACTAT-3′ and 5′-TTGCCGACTATGCCAGTCAAG-3′, Il4ra 5′-ACACTACAGGCTGATGTTCTTCG-3′ and 5′-TGGACCGGCCTATTCATTTCC-3′, Gzmb 5′-TGGCCTCCAGGACAAGGCAG-3′ and 5′-GCCTCAGGCTGCTATCCTT-3′, Il10 5′-GGTTGCCAAGCCTTATCGGA-3′ and 5′-ACCTGCTCCACTGCCTTGCT-3′. RT2 qPCR Primer Assay (Qiagen) were used for Tigit (PPM68038A) and Lag3 (PPM03649A). Subsequent analyses were performed with Excel calculating the dCt values.
Unfixed lungs were frozen in OCT Tissue Tek. Sections (8 μm thick) were fixed with ice-cold methanol before being stained with rat anti-mouse NKp46 (eBioscience, 1:200), rabbit anti-mouse Cytokeratin 8 (Abcam, 1:200) as well as the secondary antibodies goat anti-rat IgG Alexa 546 (Invitrogen, 1:300) and goat anti-rabbit IgG Alexa Fluor 488 (Life Technologies, 1:300), respectively. Images were acquired using a CLSM SP5 Leica microscope and ImageJ was used for further data analysis.
IFNγ levels were measured using a mouse IFNγ ELISA kit according to the user manual (BD OptEIATM, 555138).
P values were calculated using GraphPad statistical software (GraphPad Software Inc.). P values smaller than 0.05 were considered significant. *, P < 0.05; **, P < 0.01; ***, P < 0.001. If no asterisks are indicated, no statistically significant difference was found.
Lung NK cells become dysfunctional at advanced stages of breast cancer
Although much effort has been invested in elucidating the role of adaptive immunity in tumor immunosurveillance (27), the contribution of the innate immune system to this process is still poorly understood. Recently, we reported an important role of NK cells in control of lung metastases in the 4T1 mouse model of metastatic breast cancer (5). Consistent with the results obtained in Rag2−/−Il2rg−/− mice, anti-asialo GM1-mediated depletion of NK cells increased the metastatic burden in 4T1 tumor–bearing mice and in mammary tumor virus–polyoma middle T (MMTV-PyMT) transgenic mice of spontaneous mammary carcinoma (Fig. 1A and B). The critical role of NK cells in metastatic control was confirmed using anti-NK1.1 antibody or Nkp46iCreR26RiDTR mice, which resulted in higher metastatic load in mice bearing Lewis lung carcinoma (LLC; Supplementary Fig. S1A–S1C). In contrast, the depletion of CD4 T cells (Supplementary Fig. S1D) and the lack of T, B, and NKT cells in Rag1−/− mice (5) or γδ T cells in Tcrδ−/− mice (Supplementary Fig. S1E) did not alter lung metastatic burden.
Because metastases still arose in the presence of NK cells in control mice, we asked the question as to how these tumor cells escaped immunosurveillance. One potential mechanism is the exclusion of effector cells from the vicinity of cancer cells. However, whereas NK cells were sparse in primary tumors (data not shown), they were found to infiltrate lung metastases, decreasing only slightly during tumor progression (Fig. 1C and D). Cancers escape immune-mediated rejection by rendering NK cells dysfunctional (15, 18). Indeed, NK cells isolated from lungs at day 20 after tumor cell injection (4T1 d20), resembling a progressed tumor stage, were impaired in their lytic activity against YAC-1 and 4T1 cells when compared with lung NK cells from naïve mice or isolated at an earlier stage (4T1 d10; Fig. 1E and F). In addition, the production of IFNγ by these cells (4T1 d20) in coculture with tumor cells was severely compromised (Fig. 1G and H), although their capacity to produce IFN-γ after stimulation with PMA and ionomycin was not affected (Supplementary Fig. S1F). Concomitant with their impaired cytotoxicity, NK cells isolated from lungs at an advanced tumor stage acquired a less differentiated phenotype (CD11bhighCD27high; Fig. 1I).
Together, these findings highlight the crucial contribution of NK cells to metastatic control and further indicate that, despite their presence in the metastatic microenvironment, they fail to kill tumor cells over time, resulting in metastatic outgrowth.
Dysfunctional NK cells express distinctive patterns of activating and inhibitory receptors
To reveal the molecular mechanisms by which NK cells become impaired during advanced stages of tumor progression, we performed deep transcriptome profiling comparing them with early-stage functional NK cells. Hierarchical clustering and principal component analysis displayed a close relationship among the samples within each time point as well as a distinct gene profile of dysfunctional compared to functional NK cells (Fig. 2A and B). Gene ontology analysis revealed a differential expression of genes involved in cell adhesion, secretion of chemokines and cytokines, cytotoxic mediators, and, most importantly, regulation of cell activation (Supplementary Fig. S2A; Fig. 2C). Within the latter category, the checkpoint inhibitors TIGIT (Tigit) and Lag-3 (Lag3), negative regulators of T cell function (28), were among the most differentially expressed genes in dysfunctional lung NK cells (Fig. 2C–E; Supplementary Table S1). In addition, lung NK cells at advanced stages of breast cancer upregulated inhibitory Ly49 receptors (Klra5 (Ly49E), Klra17, Klra7 (Ly49G2); Fig. 2C and D; Supplementary Fig. S2B; Supplementary Table S1), whereas the levels of the activating receptors DNAM-1 and NKG2D were clearly decreased (Fig. 2C and F; Supplementary Table S1). Other canonical NK cell receptors such as NKp46, CD94, NKG2A/C/E, and Ly49A remained unaltered (Supplementary Fig. S2C and S2D). Increased expression of inhibitory receptors and the concomitant loss of activating receptors were also detected on circulating and bone marrow NK cells of 4T1 tumor–bearing mice at advanced stages of disease (Supplementary Fig. S2E–S2L), indicating a systemic impairment of these cells during disease progression.
These results show that the diminished lytic activity that NK cells acquire during tumor development is associated with a differential expression of activating and inhibitory receptors. Therefore, restoration of NK cell function may improve the suppression of metastatic spread.
IL12 activates NK cells but also induces the expression of coinhibitory receptors
Cytokine therapy has been used to harness the cytotoxic potential of NK cells against cancer (21). Especially, IL12 can initiate NK cell–dependent antimetastatic responses in the 4T1 model when delivered directly to the lung via i.n. administration (5). Also in the transgenic mouse model of breast cancer metastasis (MMTV-PyMT) and in a postsurgical 4T1 breast cancer metastasis model, IL12 led to a potent reduction of pulmonary metastases (Fig. 3A and B). IL12 increased the expression of the activating receptor NKG2D, induced NK cell differentiation towards CD27lowKLRG1+ NK cells, boosted IFN-γ production and prevented the decline in the lytic activity of NK cells observed in Ctrl-treated mice (Supplementary Fig. S3A–S3D). To gain further insights into the effects of IL12 on NK cells, we performed genome-wide transcriptome analysis of lung NK cells from Ctrl- or IL12-treated mice. Analysis of differentially expressed genes revealed increased levels of cytotoxic mediators (Prf1, Gzmk), activating receptors (Tnfsf4, Cd160, Cd226), proinflammatory cytokines (Csf2, Ifng), adhesion molecules (Itgax, Icam1, Itgam), and chemokine receptors (Cxcr6, Cx3cr1, Ccr5) on lung NK cells upon IL12 treatment (Fig. 3C; Supplementary Table S2). The differential expression of these transcripts was confirmed in lung NK cells from 4T1 tumor–bearing mice by qPCR (Supplementary Fig. S3E).
Strikingly however, the beneficial effects of IL12 on NK cell activation were hindered by the concomitant induction of the immune inhibitory receptors PD-1 (Pdcd1), Lag-3 (Lag3), and TIGIT (Tigit) on lung NK cells (Fig. 3C; Supplementary Table S2), which was also observed in tumor-bearing mice (Supplementary Fig. S3E; Fig. 3D and E). Comparative analyses of the PD1+ versus the PD1− population showed that PD-1+ lung NK cells resemble an immature population of NK cells, indicated by an increase of CD27highKLRG1− cells, and a high expression of CD94 and NKG2A/C/E (Fig. 3F–H). Of note, PD-1+ NK cells were also expressing the highest amounts of Lag-3 (Fig. 3I) but low levels of granzyme B (Fig. 3J). Overall, IL12 partially restores the cytotoxic capacity of NK cells, but also leads to the emergence of a less mature NK cell population with elevated expression of coinhibitory receptors, which may limit the therapeutic efficacy of this cytokine.
The metastatic TME shows differential expression patterns of NK cell ligands
Tumor cells and their surrounding environment have been reported to express stress-induced ligands that modulate NK cell activity (18, 29). We used fluorescent mCherry-tagged 4T1 tumor cells to characterize NK cell–ligand expression in tumor cells and the TME during cancer progression (Fig. 4A and B). The NKG2D ligands H60 and Rae-1 as well as CD155 (ligand to both DNAM-1 and TIGIT) were mainly expressed by 4T1 cells, with higher expression of H60 and Rae-1 on metastasized compared with primary tumor cells (Fig. 4C and D). Albeit at lower amounts, CD155 was also expressed on several populations of myeloid cells (monocytes, macrophages, and granulocytes) from primary tumors or naïve and metastatic lungs (Fig. 4D), indicating that cancer-associated myeloid cells could modulate NK cell immune responses. Even though several tumor cell types have been shown to express PD-L1 (30, 31), 4T1 tumor cells are essentially PD-L1 negative. PD-L1 was, however, prominently expressed by macrophages both in primary tumors and metastatic lungs (Fig. 4E). The inhibitory capacity of the metastatic TME was further supported by elevated MHC I and MHC II (the main ligand for Lag-3, ref. 32) on metastasized tumor cells compared with the primary tumor (Fig. 4F and G). Interestingly, i.n. administration of IL12 not only reduced Rae-1 and H60 on tumor cells, it also increased the expression of PD-L1 and MHC II (Fig. 4C, E and G). Together, these “side” effects of IL12 likely limit the capacity of NK cells to eliminate cancer cells.
Overall, the differential expression pattern of NK cell ligands within the metastatic microenvironment suggests that mainly tumor cells but also myeloid cells are potentially equipped to impede NK cell function. Hence, we propose that blocking the interaction of NK cell receptors with their inhibitory ligands could improve the efficacy of IL12 immunotherapy.
IL12 treatment synergizes with checkpoint blockade in NK cell–mediated metastasis control
Recently, immune checkpoints including PD-1, Lag-3, and TIGIT were revealed to exert a negative effect on NK cell activity in vitro (33–35). Given that IL12 treatment enhances the expression of these coinhibitory molecules on NK cells, we assessed whether their neutralization could increase the antimetastatic effects of IL12. Notably, the combined treatment of IL12 with either anti–Lag-3 (α-Lag-3) or anti–PD-1 (α-PD-1) antibodies significantly reduced lung metastases when compared with monotherapy, while primary tumor weight was unaltered regardless of the therapeutic intervention used (Fig. 5A).
It is feasible that the combination therapy could influence the primary tumor or the dissemination of cancer cells. To dismiss this possibility, we injected 4T1 cells directly i.v., which leads to their predominant accumulation in the lung. Also, in the absence of the primary tumor, coadministration of IL12 and α-Lag-3 or α-PD-1 antibodies was far superior to either monotherapy (Fig. 5B). Importantly, whereas checkpoint blockade traditionally targets T cells (36), we could here demonstrate the dependence on NK cells, because NK cell depletion alone reversed the antimetastatic effect of both combinatorial therapies (Fig. 5C). Moreover, the combinatorial treatment was also shown to efficiently reduce the number of pulmonary metastases in Lewis lung carcinoma (LLC; Fig. 5D), a preclinical model of lung cancer in which checkpoint inhibitors were also induced on lung NK cells upon IL12 administration (Supplementary Fig. S4A–S4C). Mechanistically, the coadministration of IL12 and α-Lag-3 or α-PD-1 slightly altered total numbers of proliferating NK cells (Fig. 6A and B), but most notably resulted in increased frequencies of highly differentiated (CD11bhighCD27lowKLRG1+) NK cells (Fig. 6C and D), producing higher amounts of granzyme B, perforin and also IFNγ (Fig. 6E–G), a central mediator of IL12-mediated antimetastatic immune responses in the 4T1 model (5). Hence, the specific blockade of immune checkpoints on cytokine-activated NK cells reactivates their antitumor properties, leading to a potent control of metastasis.
The potential capacity of NK cells in targeting disseminating tumor cells is undisputable (4), and therapeutic approaches directed to increase their antitumor activity are of great clinical value. A successful utilization of these cells to control metastatic disease can, however, be achieved only by a detailed understanding of their functions at advanced stages of cancer, when facing the complexity of tumor–stromal interactions in distant sites. Here, we report an impaired capacity of lung NK cells to eliminate metastatic tumor cells in preclinical models of metastasis. Evidence of defective cytotoxicity and altered expression of NK cell receptors was previously observed in tumor-associated NK cells (15, 18, 37, 38). In patients with non–small cell lung carcinoma, for example, the TME in the lung led to a local downregulation of an array of activating receptors on NK cells (39). We here demonstrate that dysfunctional lung NK cells not only downregulate activating receptors, but acquire high expression of coinhibitory receptors such as TIGIT and Lag-3, but not PD-1 (Supplementary Fig. S5). Given that a tightly regulated balance between activating and inhibitory receptors determines the outcome of NK–target cell interactions (6), the vast array of receptors with negative regulatory function overexpressed on lung NK cells during disease progression may explain their inability to clear disseminated tumor cells.
An additional layer of regulation of NK cell receptors arises from their own ligands, which are often induced in tumor cells upon cellular stress (40). In metastatic lungs, we detected the NKG2D ligands Rae-1 and H60 exclusively on tumor cells. Whether this is translated into enhanced NK cell functions is not clear, because sustained expression of NKG2D ligands or their shedding from the cell surface has been shown to cause NK cell desensitization (41–43). Tumor-derived CD155 and MHC II as well as PD-L1 were also found at high amounts within the metastatic microenvironment. This is of special interest, because TIGIT can bind CD155 with greater affinity than DNAM-1 (28) and potentially counteract signaling through this activating receptor. The engagement of the Lag-3/MHC II and PD-L1/PD-1 pathways on lung NK cells may also constitute mechanisms of immune evasion at this particular metastatic site.
External control of NK cell activity is orchestrated through proinflammatory cytokines such as IL2, IL18, IL15, and IL12 (21, 44). We selected IL12 to reactivate dysfunctional NK cells. Despite the inefficiency of this cytokine to reverse NK cell anergy in MHC I+ lymphomas (45), we could show that NK cell functions can be regulated by IL12 in the 4T1 model. Remarkably, the stimulating effects of IL12 on lung NK cells coincided with the induction of coinhibitory receptors, such as TIGIT, PD-1, and Lag-3. Within the NK cell population, the expression of these receptors was found to be higher in a unique subset of immature IL12-induced NK cells, closely resembling the previously described emergency NK cells (eNK cells; ref. 5). Hence, we suggest that the high expression of checkpoint blockers on cytokine-activated NK cells arrests these cells at a stage that prevents them from excessive activation, hindering the release of their full cytotoxicity. There is now mounting evidence from in vitro studies that coinhibitory receptors can be induced in NK cells. For instance, the engagement of PD-1 on NK cells from multiple myeloma patients reduced their cytotoxic potential (34). Also, CD96 and TIGIT, two inhibitory receptors that counteract the activity of DNAM-1 (6), were shown to be expressed on resting and activated NK cells, respectively (46). Even though CD96, rather than TIGIT, was shown to play an important role in the control of lung metastasis (24), we could not independently verify this, due to the lack of commercially available reagents.
However, whereas individual blockade of PD-1 or Lag-3 failed to exhibit any clinical benefit, the combination of checkpoint blockade with IL12 exerted a potent antimetastatic effect. The profound synergy between anti–PD-1 or anti–Lag-3 with IL12 in preclinical models of breast cancer was entirely dependent on NK cells. To translate this to human cancers, where checkpoint blockade activates tumor antigen dependent T-cell responses (23), the combinatorial treatment proposed here may concomitantly boost NK cell immunity against metastasis.
Collectively, our data demonstrate that a suitable immunomodulation of NK cells in metastatic lungs can be exploited for successful antimetastatic therapy. So far, the adoptive transfer of ex vivo expanded NK cells has not led to objective clinical responses in patients with solid tumors (47). We thus propose that a break release on these cells via blocking specific checkpoint receptors can be beneficial not only to boost cytokine-activated NK cells, but also to reactivate the endogenous NK cell compartment at the metastatic sites.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: I. Ohs, P. Kulig, B. Becher, S. Tugues
Development of methodology: I. Ohs, J. Marinho, S. Tugues
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Ohs, L. Ducimetière, P. Kulig, S. Tugues
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Ohs, L. Ducimetière, J. Marinho, P. Kulig, S. Tugues
Writing, review, and/or revision of the manuscript: I. Ohs, L. Ducimetière, B. Becher, S. Tugues
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Ohs, J. Marinho, S. Tugues
Study supervision: B. Becher, S. Tugues
We thank Sabrina Nemetz, Mirjam Lutz, Nicole Burkhalter, and the Flow Cytometry Facility, University of Zurich for assistance. We also thank Iva Lelios for advice on the analysis of Next Generation Sequencing data.
This work was supported by grants from the Swiss National Science Foundation (310030_146130 and 316030_150768 to B. Becher and CRSII3_136203 to B Becher), the University Research Priority Project “Translational Cancer Research” (B. Becher and S. Tugues), the Forschungskredit of the University of Zurich (grant no. FK-14-024 to I. Ohs), the European Union FP7 project TargetBraIn 279017, NeuroKine316722, and ATECT602239 (to B. Becher).
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.