IL21 Therapy Combined with PD-1 and Tim-3 Blockade Provides Enhanced NK Cell Antitumor Activity against MHC Class I–Deficient Tumors

Downregulation or loss of MHC class I expression in tumor cells is often associated with tumor progression (1–3). Despite the high susceptibility to immunosurveillance by activated NK cells, MHC class I–low/deficient tumor cells are still detectable in many advanced cancer patients (4). Studies have uncovered that MHC class I–deficient tumors induce a functional exhaustion of tumor-infiltrating NK cells through the high expression of inhibitory molecules, such as programmed cell death-1 (PD-1) and T-cell immunoglobulin and mucin domain 3 (Tim-3; refs. 5–9). However, whether the anti–PD-1/anti–Tim-3 checkpoint blockade is effective for eradicating MHC class I–deficient tumors through the recovery of NK cell function remains to be elucidated (5).

During the last decade, therapies targeting PD-1 and Tim-3 inhibitory receptors have shown clinical efficacy in a subset of patients with advanced cancers (10–12). However, many cancer patients do not respond to anti–PD-1 and anti–Tim-3 therapy (13, 14). Although the combination of anti–PD-1 and anti–Tim-3 therapies has shown additive effects in preclinical animal models, unmet needs in combination therapy targeting in cancer patients remain (14). It has long been thought that the antitumor effect of checkpoint inhibitors is attributed to the reinvigoration of cytotoxic T cells (CTL), although most MHC class I–deficient tumor cells readily escape immunosurveillance by the CTLs (3). Considering that NK cells also coexpress PD-1 and Tim-3 in MHC class I–deficient tumor-bearing mice with tumor progression, NK cells have emerged as another target for checkpoint blockade immunotherapy (6).

Cytokine therapies have been tested in preclinical studies and clinical trials to treat cancer by directly activating immune cells, including T cells (4, 15). Because NK cells express receptors for various cytokines, including IL2, IL7, IL12, IL15, IL18, and IL21, the addition of these cytokines also promotes the activation and maturation of NK cells (4, 16, 17). For example, we previously demonstrated that the administration of IL21 in MHC class I–deficient tumor-bearing mice induced robust antitumor immunity through the reactivation of exhausted NK cells (5). Although many cytokine therapies have shown promising clinical outcomes in cancer patients, potential toxicity at higher doses has been a hurdle to their clinical application (18–20). Therefore, the development of strategies for minimizing the adverse effect of cytokine therapy while maintaining its efficacy is required.

In the present study, we found that PD-1 and Tim-3 blockade restrained MHC class I–deficient tumor progression by reinvigorating the function of NK cells. PD-1 and Tim-3 blockade with IL21 also eradicated MHC class I–deficient tumors. Finally, we showed that IL21 enhanced the antitumor effect of PD-1 and Tim-3 blockade by inducing CXCR3-dependent infiltration of NK cells into tumor sites.

Female C57BL/6 mice were purchased from Charles River Laboratories. All mice were used at 6 to 10 weeks of age and were bred and maintained in the specific pathogen-free vivarium of the Seoul National University. All animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) at Seoul National University. The human tumor tissue and normal tissue specimens from patients with colorectal (n = 3), melanoma (n = 1), and bladder cancer (n = 1) were obtained from surgical specimens of patients. The matched normal tissues were at least 2 cm away from the edge of corresponding tumors. The tumor tissues were placed in in RPMI 1640 medium (GIBCO) that was supplemented with 10% FBS (GIBCO), 1% penicillin/streptomycin (Lonza), 1 mmol/L sodium pyruvate (Lonza), 0.1 mmol/L NEAA (Lonza), 55 mmol/L 2-mercaptoethanol (GIBCO), and 25 mmol/L HEPES (Lonza) on ice before the tumor dissociation. Tumor tissues less than 1 cm were excluded from this study. The collection of human samples was approved by the ethical committee of Severance Hospital.

The antibodies for flow cytometry were purchased from BioLegend, eBioscience, and BD Biosciences. The antibodies against PD-1 (RMP1-14, cat: 1141), Tim-3 (RMT3-23, cat: 1197), CD3ϵ (145-2C11, cat: 1003), CD19 (6D5, cat: 1155), NK1.1 (PK136, cat: 1087), CXCR3 (CXCR3-173, cat: 1265), CD16(3G8, cat: 3020), CD56(HCD56, cat: 3183), and CD45.2 (104, cat: 1098) were purchased from BioLegend. The antibodies against IFNγ (XMG1.2, cat: 7311), granzyme B (NGZB, cat: 8898), PD-1 (MIH4, cat: 9969), Tim-3 (F38-2E2, cat: 3109), IFNγ (4S.B3, cat: 7319), and granzyme B (GB11, cat: 8899) were purchased from eBioscience. The antibodies for in vivo depletion of NK1.1 (PK136, cat: BE0036), CXCR3 (CXCR3-173, cat: BE0249), and CD8 (2.43, cat: BE0061) were purchased from Bio X Cell. The recombinant mouse IL2 (cat: 402-ML), mouse IL21 (cat: 594-ML), human IL2 (cat: 202-IL), human IL21 (cat: 8879-IL), and human anti-NKp46 (cat: MAB1850) were purchased from R&D Systems.

The MC38, TC-1, and CT26 cell lines were purchased from ATCC (MC38 and TC-1 were purchased in 2006, and CT26 was purchased in 2002), and we checked morphology, growth kinetics, and antigen expression of tumor cells to validate them. The cells were distributed in several vials (1 × 10⁶/vial) with culture media [DMEM (GIBCO), 10% FBS (GIBCO)] that was supplemented with 10% DMSO (Sigma-Aldrich) and stored in a liquid nitrogen tank. The cells were cultured in DMEM (GIBCO) that was supplemented with 10% FBS (GIBCO) and 1% penicillin–streptomycin. MHC class I, PD-1, and galectin-9 molecule genes were knocked out by CRISPR/Cas9 technology. The optimized single-guide RNA (sgRNA) constructs, targeting PD-L1 and galectin-9, and the Cas9 expression construct, pRGEN-Cas9-CMV, were obtained from ToolGen. The MC38, TC-1, CT26 MHC class I–deficient cell lines and MC38 MHC class I–deficient and PD-L1 or galectin-9 double knockout cell lines were generated by transfection with the indicated sgRNA construct [H2-k1 exon 3: AGCCGTCGTAGGCGTACTGCTGG, H2-d1 exon 2: AGTCACAGCCAGACATCTGC TGG, B2m exon2: TCACGCCACCCACCGGAGAATGG, Pdcd1 exon 2: GACTTGTACGTGGTGGAGTATGG; Lgal9 exon2: CCCTTTACTGGACCAATCCAAGG; and pRGEN-Cas9-CMV using Lipofectamine 2000 (Invitrogen)], according to the manufacturer's protocol. Each genome-edited cell line was sorted by a T7E1 assay (ToolGen) with single-cell selection or FACSAria III (BD Bioscience). For T7E1 assay with single-cell selection, we followed the manufacturer's protocol. For sorting with FACSAria III, we sorted the negative population after staining with knockout molecules in tumor cells. For verifying the purity of knockout tumor cell lines, we analyzed the expression of knockout molecules using FACSAria III (BD Bioscience) after stimulation for 24 hours with mouse IFNγ (10 ng/mL; R&D Systems) to induce knockout of molecules.

For primary cell culture (primary cells were lymphocytes taken from mouse/human tumor tissue), the mouse primary cells were cultured in RPMI 1640 (GIBCO) medium that was supplemented with 10% FBS (GIBCO) and 1% penicillin–streptomycin (Lonza). Human primary cells were cultured in X-VIVO15 medium (Lonza) that was supplemented with 1% penicillin–streptomycin (Lonza). For NK cell culture, a low dose (1–2 ng/mL) of rIL2 was added to the culture medium for cell survival. All cell lines were found to be negative for mycoplasma contamination, and these cell lines were used at passages 3 to 7 for all experiments after MHC class I–deficient cell lines generation.

The NK cells were prepared from tumor-infiltrating lymphocytes from MHC class I–deficient, MHC class I–deficient plus PD-L1 knockout, MHC class I–deficient plus galectin-9 knockout tumor-bearing mice via FACSArea III sorting as effector cells. Next, NK cells were cocultured with ⁵¹Cr-labeled respective tumor cells as target cells. The effector-to-target cell ratio (E:T) was 1:1, 3:1, and 10:1, and NK cell cytotoxicity was calculated by ⁵¹Cr release in the culture supernatants through the specific lysis of the target cells, as measured by a Wallac 1480 Wizard automatic γ-counter (PerkinElmer). When NK cells were cocultured with tumor cells, the tumor cells were irradiated (5,000 rad) using a γ irradiator (GC 3000 Elan) at the National Center for Inter-University Research Facilities (NCIRF) at Seoul National University.

Conventional subcutaneous indicated tumors were generated by subcutaneous (s.c.) injection in the right flank. The tumor growth was measured by a metric caliper 2 to 3 times a week. For the therapeutic tumor model, 5 × 10⁵ tumor cells were subcutaneously injected into the left flank of mice on day 0. Recombinant IL21 (rIL21; 10 μg in 100 μL PBS) or PBS (100 μL) alone were injected intratumorally (i.t.) using a 31-gauge insulin syringe at the indicated time points. All depleting antibodies were administered every 3 days. The indicated checkpoint blockade antibodies (anti–PD-1: 300 μg/mouse; anti–Tim-3: 300 μg/mouse) or depleting antibodies (anti-CD8: 200 μg/mouse; anti-NK1.1: 200μg/mouse; and anti-CXCR3: 100 μg/mouse) were administered via intraperitoneal (i.p.) route, according to the manufacturer's instructions. The mouse survival rate was examined by actual survival, and statistical analysis was performed by log-rank Mantel–Cox test (conservative). Mouse whole tumor weight was measured by microbalance upon sacrifice at day 13 after tumor inoculation.

The cells (mouse and human primary cells: 1 × 10⁵ to 5 × 10⁵, respectively, and mouse tumor cells: 5 × 10⁴ to 1 × 10⁵) were stained with anti-mouse CD16/32 (BioLegend) for FcR blocking, and the specified surface antibodies in 50 μL of FACS buffer (PBSN plus 1% FBS). The cells were incubated on 4°C in the dark for 15 minutes, and then the cells were washed with FACS buffer. Dead cells were excluded by staining with Fixable Viability Dye (eBioscience), following the manufacturer's instruction. Intracellular staining for the indicated cytokines was performed after surface staining with the Cytofix/Cytoperm kit (BD Bioscience), according to the manufacturer's instruction.

To assay the cytokine release from human NK cells, 3 × 10⁴ – 1 × 10⁶ cells were stimulated in flat-bottomed, high-protein–binding plates (Corning) that were coated with hNKp46 antibodies for 5 hours in the presence of GolgiPlug (1 μg/mL; BD) before staining the cells for intracellular IFNγ, perforin, or granzyme B. The samples were acquired with a FACSCalibur or a FACSAria III instrument (BD Biosciences), and the data were analyzed with FlowJo software (TreeStar).

Mouse and human tumors were cut 2 to 5 mm sizes and placed C tube (Miltenyi Biotec) containing RPMI 1640 medium with 10% FBS containing Collagenase D (300 μg/mL; Roche), hyaluronidase (20 μg/mL; Sigma-Aldrich), and DNase I (20 μg/mL; Sigma-Aldrich) or reagents from a human tumor dissociation kit (Miltenyi Biotec, cat: 130-095-929). After, the tumors were dissociated using the gentleMACS Dissociator (Miltenyi Biotec, cat: 130-093-235). The dissociated tumors were incubated in 37°C for 30 minutes and washed in PBS. Next, lymphocytes were separated using lymphocyte separation medium (MP Biomedicals, cat: 0850494) and filtered in a 70-μm nylon mesh. Lymphocytes were counted and used for FACS staining tumor-infiltrating NK cell was calculated as |$\left( = { \frac{{{\rm{\% \ of\ CD}}{{45.2}^ + }{\rm{\ NK}}{{1.1} ^+} {\rm{\ cells\ }} \times {\rm{\ No}}.{\rm{\ of\ tumor - infiltrating\ lymphocyte}}}}{{{\rm{tumor\ weight\ }}( {\rm{g}} )}}}\right ).$|

To measure the production of the CCL5, CXCL9, and CXCL10 chemokines by tumor cells, MC38 and TC-1 MHC class I–deficient tumor cells were harvested from tumor-bearing mice, the dissociated tumor cells were maintained in RPMI 1640 medium supplemented with 10% FBS for 2 days prior to supernatant collection. The following chemokines were measured in the supernatant an ELISA kit according to the manufacturer's instructions: CCL5, CXCL9, and CXCL10 (R&D Systems).

Statistical comparisons were performed using the Prism (version 6.0) software (GraphPad Software). P values were determined as follows: (i) two-way ANOVA with a Bonferroni multiple comparisons test: Figs. 1B, 2B, 4B, 4E, Fig. 5J; Supplementary Fig. S4B; Supplementary Fig. S6B and S6C; (ii) unpaired two-tailed Student t test: Figs. 1C, 1D, 3A, 3B, 4D, 5B, 5D, 5E, 5F, 5I, 6E; Supplementary Figs. S3; S5; S7A; S7B, and S8; and (iii) a log-rank (Mantel–Cox) test (conservative): Figs. 2C and4C.

We previously reported that tumor-infiltrating NK cells are functionally exhausted and coexpress PD-1 and Tim-3 molecules on their surface in MHC class I–deficient tumor-bearing mice (5). To determine whether the PD-1/PD-L1 and Tim-3/galectin-9 axis transmit negative signals in tumor-infiltrating NK cells, we used the CRISPR/Cas9 system to genetically delete the Pdcd1 or Lgal9 gene in an MHC class I–deficient MC38 tumor cell line (Supplementary Fig. S1A). The proliferative capacities of the resultant PD-L1-deficient, MHC class I–deficient (MC38 M1&PD-L1 KO), and galectin-9-deficient, MHC class I–deficient (MC38 M1&Gal9 KO) cell lines were comparable with the MHC class I–deficient (M1 KO) control tumor cells in vitro (Supplementary Fig. S1B). When subcutaneously implanted in syngeneic C57BL/6 mice, each MHC class I–deficient tumor cell line comparably induced PD-1⁺Tim-3⁺ NK cells at the tumor site, regardless of PD-L1 or galectin-9 expression (Fig. 1A). To exclude the possible involvement of CD8⁺ T cells, CD8⁺ T cells were depleted by treating with anti-CD8 after MHC class I–deficient tumor inoculation (Supplementary Fig. S2). PD-L1 deficiency almost completely abrogated the progression of MHC class I–deficient tumor cells, whereas the tumor progression was partially inhibited by galectin-9 deficiency (Fig. 1B), and PD-L1 or galectin-9 deficiency increased intratumoral NK cell cytotoxicity against mice bearing each type of tumor (Fig. 1C). The frequencies of IFNγ⁺ and granzyme B (GrazB)⁺ NK cells were inversely correlated with tumor burdens, suggesting that both PD-1/PD-L1 and Tim-3/gal-9 axes contribute to the functional exhaustion of tumor-infiltrating NK cells in MHC class I–deficient tumor-bearing mice (Fig. 1D).

We next sought to examine whether the abrogation of either PD-1/PD-L1 or Tim-3/galectin-9 signaling by anti–PD-1 or anti–Tim-3 treatment, respectively, could inhibit the progression of MHC class I–deficient tumor cells. Our previous study revealed that Tim-3 expression on NK cells was induced a few days earlier than PD-1 expression. Based on this finding, anti–Tim-3 treatment started earlier than anti–PD-1 treatment, and to exclude the possible involvement of CD8⁺ T cells in response to checkpoint blockade, CD8⁺ T cells were depleted using anti-CD8 (Fig. 2A). We found that anti–PD-1 or anti–Tim-3 treatment inhibited tumor growth and improved survival compared with control IgG treatment (median survival: 32 days), and anti–PD-1 treatment inhibited tumor growth and improved the survival of the tumor-bearing mice (median survival: 46.50 days) to a greater extent than treatment with anti–Tim-3 (median survival: 36 days; Fig. 2B and C). We then tested whether a combination of anti–PD-1 and anti–Tim-3 could induce an additive increase in the antitumor effects against MHC class I–deficient tumors. Although tumor growth was not significantly inhibited by the combination therapy, the addition of an anti–Tim-3 antibody marginally extended the survival of the mice compared with the anti–PD-1 antibody alone (Fig. 2B and C).

Studies have shown that cytokine therapy restores the function of exhausted NK cells in MHC class I–deficient tumor microenvironments (5, 21, 22). We tested whether rIL2, rIL12, or rIL21 could increase the effector function of NK cells in MHC class I–deficient tumors. IFNγ⁺ NK cells were increased by rIL21 treatment in a dose-dependent manner, whereas either rIL2 or rIL12 treatment was less effective than rIL21 treatment at all tested doses (Supplementary Fig. S3). Our previous study demonstrated that rIL21 treatment restored the effector function of exhausted NK cells in MHC class I–deficient tumor-bearing mice through the PI3K–AKT–Foxo1 and STAT1 signaling pathways, while the rIL21-treated NK cells still expressed PD-1 and Tim-3 molecules on their surface (5). Based on this finding, we hypothesized that the blockade of PD-1 and Tim-3 has an additive effect on the rIL21-mediated reversal of NK cell function. Consistent with the decreased tumor progression with anti–PD-1/anti–Tim-3 treatment, IFNγ⁺ or granzyme B⁺ NK cells were increased in anti–PD-1/anti–Tim-3-treated MHC class-deficient tumor-bearing mice compared with those in the control-treated mice (Fig. 3A and B). The addition of exogenous rIL21 further increased IFNγ and granzyme B expression in tumor-infiltrating NK cells isolated from mice treated with anti–PD-1/anti–Tim-3 (Fig. 3A and B). Collectively, these results suggest that the combination of rIL-21 and anti–PD-1/anti–Tim-3 may foster antitumor immune responses against MHC class I–deficient tumors by enhancing the effector function of exhausted NK cells.

To evaluate whether checkpoint blockade could further enhance IL21-mediated antitumor immunity against MHC class I–deficient tumors, intratumoral rIL21 in combination with anti–PD-1/anti–Tim-3 was administered to M1 KO MC38, TC-1, and CT26 tumor-bearing mice. To exclude the possible involvement of CD8⁺ T cells in response to either IL21 or checkpoint blockade, CD8⁺ T cells were depleted using anti-CD8 (Fig. 4A; Supplementary Figs. S2A and S4A). Although both intratumoral rIL21 alone and anti–PD-1/anti–Tim-3 alone were comparably effective for restraining M1 KO MC38 and TC-1 tumor growth compared with control IgG, a combination of rIL21 and anti–PD-1/anti–Tim-3 therapies was more effective than each monotherapy in terms of tumor progression, as well as survival of the mice (Fig. 4B and C; Supplementary Fig. S4B and C). However, although intratumoral rIL21 and anti–PD-1/anti–Tim-3 alone were comparably effective for restraining M1 KO CT26 tumor growth compared with control IgG, the combination therapy was not more effective than each monotherapy in terms of tumor progression in mice bearing CT26 tumors (Supplementary Fig. S4D).

The frequency of IFNγ⁺ intratumoral NK cells was significantly increased after the combination of rIL21 and anti–PD-1/anti–Tim-3 therapies, suggesting that combination therapy additively restored the function of exhausted NK cells (Fig. 4D; Supplementary Fig. S4C). In contrast, the frequency and absolute number of intratumoral NK cells were significantly increased by the combination therapy but were indistinguishable from those with the rIL21 treatment alone, indicating that IL21 was responsible for the recruitment of NK cells into the tumor site (Supplementary Fig. S5). The antitumor effect of combination therapy was dependent on NK cells, as NK cell depletion by anti-NK1.1 treatment abrogated the tumor growth inhibition by combination therapy (Fig. 4E).

To verify whether the combination of rIL21 with checkpoint blockade enhances antitumor effects in MHC class I–deficient tumors, we administered checkpoint blockade alone, rIL21 alone, or rIL21 with checkpoint blockade in M1&Gal9 KO or M1&PD-L1 KO MC38 tumor models (Supplementary Fig. S6A). Although both intratumoral rIL21 alone and anti–PD-1 alone were comparably effective for restraining M1&Gal9 KO MC38 tumor growth compared with control IgG, the combination of rIL21 with anti–PD-1 was more effective than each monotherapy in terms of tumor progression (Supplementary Fig. S6B), and although anti–Tim-3 alone or intratumoral rIL21 alone did not significantly inhibit M1&PD-L1 KO MC38 tumor growth compared with control IgG, the combination of rIL21 with anti–Tim-3 significantly suppressed M1&PD-L1 KO MC38 tumor growth (Supplementary Fig. S6C). Collectively, these results suggest that IL21 and anti–PD-1/anti–Tim-3 combination therapy improves antitumor immunity by enhancing tumor infiltration and the effector function of NK cells.

As we described earlier, we found that the combination of rIL21 and checkpoint blockade therapies or rIL21 alone increased the tumor infiltration of NK cells (Supplementary Fig. S5). It is possible that intratumoral administration of rIL21 increases the recruitment of NK cells to the tumor site. To address this possibility, rIL21 was administered intratumorally in MHC class I–deficient tumor-bearing mice. After 3 administrations of rIL21, the frequency and absolute number of tumor-infiltrating NK cells were significantly increased compared with the vehicle treatment (Fig. 5 A and B). CXCR3 expression on tumor-infiltrating NK cells and CXCL9 and CXCL10 production by MC38 and TC-1 tumor cells were significantly enhanced by rIL21 treatment, suggesting that CXCR3-dependent chemotaxis is associated with the increased recruitment of NK cells at the tumor site by rIL21 (Fig. 5C–E; Supplementary Fig S7A and B). However, CCL5 production by each tumor cell was not significantly enhanced by rIL21 treatment (Fig. 5F; Supplementary Fig. S7C). Consistent with this notion, CXCR3 blockade abrogated the IL21-mediated increase in NK cell infiltration at the tumor site (Fig. 5G–I) and completely abolished the tumor growth inhibition by rIL21 treatment (Fig. 5J). Altogether, these results clearly indicated that CXCR3-dependent recruitment of NK cells to the tumor site was essential for the IL21-mediated antitumor immune responses.

Studies have demonstrated that the coexpression of PD-1 and Tim-3 molecules marks functionally exhausted intratumoral NK cells in patients with different cancers and that the antibody-mediated blockade of these coinhibitory molecules restores the function of NK cells (6–8, 23, 24). We also previously found that rIL21 treatment was effective in restoring the function of exhausted PD-1⁺Tim-3⁺ NK cells isolated from tumor tissues of cancer patients (5). To evaluate whether the combination of rIL21 treatment and checkpoint blockade could enhance the functional reversal of exhausted NK cells in cancer patients, we treated normal tissues or tumor tissues isolated from cancer patients with rIL21 alone, anti–PD-1/anti–Tim-3 alone, or both. rIL21 treatment significantly increased IFNγ production of PD-1⁻Tim-3⁻ NK cells, whereas checkpoint blockade did not affect IFNγ production of PD-1⁻Tim-3⁻ NK cells from normal tissues (Supplementary Fig. S8). When we analyzed the IFNγ expression of intratumoral PD-1⁺Tim-3⁺ NK cells as a readout for NK cell function, it was significantly augmented by treatment with rIL21 alone, whereas checkpoint blockade was less effective. The combination of rIL21 and the checkpoint blockade increased IFNγ production by intratumoral PD-1⁺Tim-3⁺ NK cells (Fig. 6A and B). Taken together, these results suggest that a combination of rIL21 and checkpoint inhibitors can enhance antitumor immunity in cancer patients by additively increasing the effector function of NK cells.

Advances in cancer immunotherapy are not only changing standard therapies for cancer but are also expected to yield complete responses in patients with advanced cancer (25–27). Although various cancer immunotherapies, such as checkpoint inhibitors, immunostimulatory cytokines, tumor vaccines, and CAR-T cells, have shown clinical efficacy, a significant proportion of cancer patients still do not respond to any of these immunotherapies (28–30). Most immunotherapies have mainly focused on their effect on T cells, whereas NK cells are one of the most important effector arms, especially against MHC class I–deficient tumors, whose frequency is closely associated with the malignancy of the cancer (4, 5, 31, 32).

PD-1 and Tim-3 expressed on exhausted T cells are well known targets for cancer immunotherapy, and various ongoing clinical trials are targeting these molecules (11, 33, 34). Several studies have shown that PD-1 and Tim-3 expression on NK cells results in deceased cytokine production, degranulation, and cytotoxicity, indicating that these coinhibitory molecules not only act on T cells but also on NK cells (6–9, 24). In our previous study, we also show that PD-1⁺Tim-3⁺ NK cells with defective effector function are found in MHC class I–deficient tumor-bearing mice and cancer patients (5), and rIL21 treatment restores the effector function of exhausted NK cells. Interestingly, the PD-1 and Tim-3 expression levels on NK cells were still maintained even after rIL21 treatment (5). This finding led us to evaluate whether anti–PD-1/anti–Tim-3 checkpoint blockade enhances rIL21-mediated antitumor immunity against MHC class I–deficient tumors. Checkpoint blockade in combination with intratumoral rIL21 treatment further increased the effector function of NK cells, leading to the significant regression of established tumors. Whereas both rIL21 and checkpoint blockade contributed to the enhancement of NK cell effector function, the combination of the two was more effective than each single therapy. These results suggest that the two different therapies may enhance NK cell effector function via distinct mechanisms of action.

It has been demonstrated that IFNγ upregulates CXCL9 and CXCL10 chemokine expression in tumors (35–37). We also found that rIL21 administration augmented CXCL9 and CXCL10 production by tumors. However, whether IL21 directly induces CXCL9 and CXCL10 expression by tumor cells or whether increased IFNγ production by IL21-activated NK cells indirectly contributes to the chemokine production remains to be elucidated. Whereas the anti–PD-1/anti–Tim-3 checkpoint blockade was found to be less effective in NK cell recruitment to the tumor site, intratumoral rIL21 administration significantly increased the NK cell infiltration into the tumor. These results provide evidence that the two therapies act at different stages during development of the antitumor immune response.

Our present study has focused on the antitumor effects of NK cells induced by IL21 and checkpoint blockade combination therapy against MHC class I–deficient tumors. However, it is possible that the combination therapy may be beneficial for the eradication of MHC class I–sufficient tumors through enhancing the effector functions of exhausted T cells. Previous preclinical and clinical studies have suggested that IL21 enhances and sustains antitumor immune responses of cytotoxic CD8⁺ T cells (38–40), and targeting the PD-1 and Tim-3 pathways is effective in restoring antitumor immunity by reversing exhausted T cells (34, 41). Therefore, the combination of rIL21 and checkpoint blockade could foster the antitumor immune responses of exhausted T cells, as well as those of exhausted NK cells.

Another study has demonstrated that PD-1 expression of tumor-associated macrophages (TAM) gradually increase over time in tumor-bearing mice and in cancer patients with increasing disease stages (42, 43). These authors also show that PD-1/PD-L1 blockade inhibits tumor growth in various mouse tumor models by enhancing the phagocytosis of TAMs (42, 43). Another study shows that the intratumoral delivery of IL21 can skew TAM polarization from the M2 phenotype to the M1 phenotype, which is known to inhibit tumor progression (44). Based on these findings, a combination of IL21 and checkpoint blockade might have affected the function of the macrophages. In this study, NK cell depletion did not completely abrogate the tumor growth inhibition by IL21 and checkpoint blockade combination therapy, suggesting that other immune cells, such as macrophages, could be involved in enhancing the antitumor effect of combination therapy. Given that macrophages also contribute to the rejection of MHC class I–deficient tumors, further studies are needed to identify whether the enhancement of TAM phagocytosis is involved in the additive increase of antitumor immunity by the combination of IL21 and checkpoint blockade.

Although our study has examined the role of PD-L1, galectin-9, CXCL9, and CXCL10 expression by tumor cells, these molecules are also expressed or secreted by various immune cells. Myeloid-derived suppressor cells populating the tumor microenvironments express both PD-L1 and galectin-9 (45–47). CXCL9 and CXCL10 can be secreted by M1 macrophages (48). Their expression or secretion by various immune cells might be crucial for regulating antitumor immune responses. Therefore, further studies are needed to identify the roles of PD-L1, galectin-9, CXCL9, and CXCL10 expression by other immune cells.

Previous attempts to treat cancer with cytokines have provided dramatic clinical efficacy in some preclinical mouse models but have yielded disappointing results in human clinical trials with high systemic toxicity (18, 19). However, our combination regimen of rIL21 and checkpoint blockade did not induce any evident side effects, such as weight loss. Considering that rIL21 has been proven safe in clinical trials for treating metastatic melanoma patients or renal cell carcinoma patients, rIL21 might be one of the most promising candidates as a combination partner for checkpoint blockade immunotherapy (49, 50). Although rIL21 was intratumorally injected in the present study to avoid possible systemic detrimental effects of the cytokine, systemic delivery of IL21 could potentially induce antitumor effects in clinical trials (20).

In this study, although we showed that combination therapy enhanced antitumor immune responses in MHC class I–deficient mouse tumor models by an NK cell-dependent mechanism, not all of the tumor cells in cancer patients are MHC class I–deficient. Given that human cancer microenvironments are more complex than mouse tumor models, particularly in end-stage cancer patients with heterogeneous tumors, the direct application of these findings to the human clinical setting may be limited. However, combining IL21 and checkpoint blockade could enhance antitumor responses of NK cells, as well as CD8⁺ T cells, for eradicating heterogeneous tumors. In summary, our study showed that the PD-1/PD-L1 and Tim-3/galectin-9 axes transmitted negative signals on NK cells, and checkpoint blockade, such as anti–PD-1 or anti–Tim-3, significantly inhibited tumor growth, and intratumoral administration of rIL21 recruited NK cells to the tumor site through a CXCR3-dependent mechanism. The combination of IL21 with anti–PD-1/anti–Tim-3 was more effective than the monotherapy in eradicating MHC class I–deficient tumors. The improved therapeutic efficacy of combination of IL21 with checkpoint blockade warrants the initiation of a clinical trial with this regimen.

No potential conflicts of interest were disclosed.

Conception and design: H. Seo, C.-Y. Kang

Development of methodology: H. Seo, E.-A. Bae, C.-Y. Kang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Seo, E.-A. Bae, B.S. Min, Y.D. Han, S.J. Shin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Seo, B.-S. Kim, C.-Y. Kang

Writing, review, and/or revision of the manuscript: H. Seo, B.-S. Kim, E.-A. Bae, S.J. Shin, C.-Y. Kang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Seo, S.J. Shin, C.-Y. Kang

Study supervision: C.-Y. Kang

This work was supported by grants from the Basic Science Research Program (NRF-2015R1A2A1A10055844) and the Bio and Medical Technology Development Program (NRF-2016M3A9B5941426).

The authors thank the staff of the National Center for Inter-University Research Facilities (NCIRF) at Seoul National University for assistance with the cell irradiation by the gamma irradiator (GC 3000 Elan).

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.