Inhibition of MDSC Trafficking with SX-682, a CXCR1/2 Inhibitor, Enhances NK-Cell Immunotherapy in Head and Neck Cancer Models

Subclones of tumor cells within heterogeneous carcinomas harbor genomic alterations that confer resistance to T-cell antitumor immunity. Altered genes encoding components of the IFN response pathway (1), antigen processing (1, 2), and antigen presentation (1–3) have been identified in tumors from patients that have failed T-cell–based immunotherapy. Such clonal cell populations may be irreversibly insensitive to detection or elimination by T cells, but may retain the ability to be detected by natural killer (NK) cells through antigen- and MHC-independent mechanisms. Engineered NK cellular therapies are under clinical investigation (NCT03027128 and NCT03387111).

Carcinomas such as head and neck squamous cell carcinoma (HNSCC) develop an immunosuppressive microenvironment through multiple mechanisms (4). Myeloid-derived suppressor cells (MDSC), a heterogeneous population of immunosuppressive myeloid cells, are pathologically expanded in mice and humans with cancer and are recruited into tumors through chemokine gradients (5–8). Both neutrophilic and monocytic subsets of MDSC suppress T-cell function primarily through production of arginase (7, 8). The effects that MDSC may have on endogenous NK cells or NK-cell–based immunotherapies are less understood.

Here, we demonstrated in an immunologically “cold” syngeneic model of oral cancer (mouse oral cancer 2, or MOC2) that Ly6G^hi neutrophilic-MDSC (PMN-MDSC) suppress NK-cell function through multiple independent mechanisms. Inhibition of CXCR2⁺ PMN-MDSC trafficking into MOC2 tumors with SX-682, a small-molecule dual inhibitor of CXCR1/2, enhanced tumor infiltration, activation, and the therapeutic efficacy of adoptively transferred NK cells. The mechanism of this enhancement was due to inhibition of PMN-MDSC trafficking and independent of direct antitumor effects. In addition, we verified the accumulation of CD15⁺CD14⁻ PMN-MDSC and CD15⁻CD14⁺HLA-DR^low monocytic-MDSC (M-MDSC) positive for both CXCR1 and CXCR2 in the circulation of patients with advanced HNSCC. PMN-MDSC and M-MDSC that trafficked into HNSCC tumors demonstrated enhanced suppressive capacity of healthy donor NK cells compared with circulating MDSC through multiple, independent mechanisms. Together, data from this study establish the role of MDSCs in abrogating NK-cell function in mice and humans, and provide the rationale for the clinical study of CXCR1/2 inhibitors in combination with adoptively transferred NK cells.

MOC2 cells were cultured as described previously (9). Genomically characterized cell stocks were used at low passage number (<30) for all experiments, verified free of Mycoplasma and murine-associated pathogens, and authenticated with CD44 and MHC expression. KIL murine NK cells were purchased from Kerafast and cultured as described previously (10). Four days prior to use, low passage number KIL (<10) were exposed to murine IL2 (20 ng/mL). In some experiments, GM-CSF (10 ng/mL) was added to media for MDSC culture (11). SX-682 was obtained under a Cooperative Research and Development Agreement with Syntrix Pharmaceuticals. Tumors were established by subcutaneous injection of MOC2 cells (1 × 10⁵ cells in PBS) in the flanks of wild-type C57BL/6 mice (Taconic). Tumor volumes were obtained three times weekly and calculated as (length × width²)/2. Control high fat chow and chow formulated with SX-682 (drug added to expose mice to 500 mg/kg body weight/day) was obtained from Research Diets, Inc. KIL were adoptively transferred via intraperitoneal injection. All human peripheral blood and HNSCC tumor biopsies were obtained under an IRB-approved biospecimen procurement protocol (NIH protocol number 18-DC-0051) conducted in accordance with the U.S. Common Rule after obtaining informed written consent.

For details of flow cytometry, see Supplementary Materials and Methods.

For some experiments, PMN-MDSC and M-MDSC were isolated from splenocyte single-cell suspensions using the Stemcell EasySep Mouse MDSC Isolation Kit per the manufacturer's recommendations and labeled with CellTracker Red CMTPX Dye (Thermo Fisher Scientific) per the manufacturer's recommendations prior to adoptive transfer. For other experiments, KIL were labeled with qTracker 585 (Thermo Fisher Scientific) per the manufacturer's recommendations prior to adoptive transfer. Tumors were digested into a single-cell suspension as described and assessed by flow cytometry or images of whole tumor digests were acquired on an EVOS Live Cell Imaging System (Thermo Fisher Scientific).

Peripheral murine NK cells were isolated from splenic single-cell suspensions using the mouse NK Cell Negative Isolation Kit on an AutoMACS Pro (Miltenyi Biotec) per the manufacturer's recommendations. Peripheral murine neutrophilic cells were isolated using the Ly6G Microbead Positive Selection Kit (Miltenyi Biotec). Peripheral human NK cells were isolated using the human NK Cell Isolation Kit (Miltenyi Biotec) on PBMCs isolated from CPT collection tubes. Peripheral human myeloid cells were isolated using the CD15 or CD14 Positive Selection Kits (Miltenyi Biotec) following an 80% Percoll gradient separation of whole blood collected in heparinized collection tubes. To isolate tumor-infiltrating myeloid cells, murine or human tumors were digested into single-cell suspensions as described and subjected to a 40/80% Percoll cell separation gradient. Murine neutrophilic cells were isolated from retrieved leukocytes using a Ly6G positive selection kit, and human myeloid cells were isolated using CD15 or CD14 positive selection kits.

MOC2 cells (5 × 10³ cell/well) were plated in the presence or absence of DMSO control or SX-682. Other experiments were designed to measure MOC2 cytotoxicity when exposed to sorted NK or KIL cells. For detailed methods, see Supplementary Materials and Methods.

Cells or whole-tumor digests were lysed and RNA was purified using the RNEasy Mini Kit (Qiagen). cDNA was synthesized utilizing a high capacity reverse transcription kit with RNAse Inhibitor (Applied Biosystems). A TaqMan Universal PCR master mix was used to assess the relative expression of target genes compared with GAPDH on a Viia7 qPCR analyzer (Applied Biosystems). Primers were purchased from Thermo Fisher Scientific.

Fresh sorted human healthy donor or HNSCC patient NK cells were stimulated with K562 cells at a K562-to-NK ratio of 1:1 with IFNα (1000 U/mL) as described previously (12). In some wells, fresh sorted healthy donor or HNSCC patient peripheral or tumor-infiltrating CD15⁺ or CD14⁺ myeloid cells were added at different myeloid cell-to-NK ratios. In other wells, TGFβ mAb (clone 1D11.16.8, 2 μg/mL), catalase (300 μmol/L), PD-L1 mAb (clone 29E.2A3, 2 μg/mL), L-NMMA (300 μmol/L), or nor-NOHA (300 μmol/L) were added. In coincubation assays with specific inhibitors, MDSCs were exposed to the inhibitors for 1 hour prior to coincubation with NK cells, and the inhibitors remained in the well for the duration of the experiment. IFNγ production was measured via ELISpot Assay (R&D Systems) per the manufacturer's recommendations and measured on an Immunospot ELISpot Plate Reader (Cellular Technology).

Tests of significance between pairs of data are reported as P values, derived using a Student t test with a two-tailed distribution and calculated at 95% confidence. Comparison of multiple sets of data was achieved with ANOVA with Tukey multiple comparisons. Survival analysis was determined by log-rank (Mantel-Cox) analysis. Error bars indicate SD. Statistical significance was set to P < 0.05. Analysis was performed using GraphPad Prism v7.

To determine whether tumor-bearing hosts developed dysfunctional NK-cell immunity, splenic NK cells from wild-type mice bearing MOC2 tumors were studied. The fraction of NK cells within the splenocyte compartment was statistically decreased in tumor-bearing mice compared with naïve mice, as was the expression of the NK-cell–activating receptor NKG2D (Fig. 1A), but these changes were of unclear biologic significance. Functional analysis of sorted NK cells from MOC2 tumor–bearing mice revealed decreased capacity to kill MOC2 target cells at a range of effector:target ratios compared with NK cells from naïve mice (Fig. 1B), indicating suppressed function. This suppression of NK-cell function appeared to be durable as incubation of NK cells for 24 hours prior to assessing their ability to kill MOC2 tumor cells did not restore the reduced ability of NK cells from MOC2 tumor–bearing mice to kill (Fig. 1C). We next assessed peripheral myeloid cells. Both Ly6G^hiLy6C^int and Ly6G^lowLyC^hi myeloid splenocytes were significantly expanded in mice bearing MOC2 tumors compared with naïve mice (P < 0.001, P < 0.01, respectively; Fig. 1D). Ly6G^hiLy6C^int myeloid cells expressed high levels of CXCR2, Ly6G^lowLyC^hi myeloid cells expressed low CXCR2, and neither cell type expressed high levels of CXCR1. MOC2 tumors expressed the myeloid chemokines CXCL1/2/5 (Supplementary Fig. S1). To determine whether these CXCR2 ligands correlated with myeloid cell chemotaxis, MOC tumors were assessed for myeloid cell infiltration. This analysis revealed that greater than 70% of all immune CD45.2⁺ cells within day 10 MOC2 tumors were myeloid lineage, with the vast majority displaying a Ly6G^hiLy6C^int phenotype (Fig. 1E). These Ly6G^hi myeloid cells were sorted to high purity (Supplementary Fig. S2) and assessed for their ability to suppress NK-cell function in real-time impedance assays. Tumor-infiltrating Ly6G^hi myeloid cells abrogated the ability of sorted NK cells from wild-type mice to kill MOC2 tumor targets, validating these cells as PMN-MDSC (Fig. 1F). Together, these results demonstrated that mice bearing MOC2 tumors accumulated peripheral CXCR2⁺ PMN-MDSC that upon trafficking to tumor have the ability to suppress NK-cell function.

Sorted NK cells are challenging as a study tool due to their low frequency in the peripheral blood and spleens of mice. KIL cells are a culturable murine NK-cell line that can be expanded for in vitro and in vivo study (10, 13). KIL cells killed MOC2 tumor targets to a greater degree at different effector:target ratios compared with sorted NK cells from wild-type mice (Fig. 2A). We next determined the ability of PMN-MDSC to suppress KIL function. Tumor PMN-MDSC abrogated the ability of KIL to kill MOC2 targets to a greater degree than peripheral PMN-MDSC (Fig. 2B), and tumor PMN-MDSC inhibited KIL killing capacity in a dose-dependent fashion (Fig. 2C). These results validated KIL as a tool to study the effects of murine MDSC on NK cells.

To study the mechanisms by which tumor PMN-MDSC suppressed KIL effector function, coincubation experiments with neutralizing antibodies or specific inhibitors were performed. The addition of TGFβ mAb significantly abrogated the ability of tumor PMN-MDSC to suppress KIL killing of MOC2 cells (Fig. 3A). The addition of catalase, an inhibitor of extracellular hydrogen peroxide (H₂O₂), also reversed the ability of PMN-MDSC to suppress KIL effector function (Fig. 3B), although to a lesser degree than TGFβ mAb (Fig. 3C). The addition of PD-L1 mAb, the nitric oxide synthase (NOS) inhibitor L-NMMA, or the arginase inhibitor nor-NOHA did not alter the ability of PMN-MDSC to suppress KIL effector function (Supplementary Fig. S3A–S3C). Given data that tumor PMN-MDSC suppressed kill function to a greater degree than peripheral PMN-MDSC, expression of cell surface TGFβ and superoxide dismutase 1 and 2 (SOD1/2) was measured in these subsets. Tumor PMN-MDSC expressed greater TGFβ and SOD1 compared with peripheral PMN-MDSC (Fig. 3D and E). These data demonstrated that in the syngeneic MOC2 model, PMN-MDSC suppression of KIL effector function was mediated through TGFβ and H₂O₂.

Having established that immunosuppressive CXCR2⁺ PMN-MDSC infiltrate MOC2 tumors, we studied whether SX-682, an orally bioavailable small-molecule inhibitor of CXCR1/2, could alter PMN-MDSC trafficking and tumor accumulation. Mice bearing MOC2 tumors were treated with control chow or chow containing SX-682 for 1 week and assessed for PMN-MDSC accumulation by flow cytometry. SX-682 significantly decreased PMN-MDSC trafficking into MOC2 tumors (Fig. 4A). SX-682 treatment of tumor-bearing mice did not alter the viability of tumor leukocytes (Fig. 4B) and ex vivo exposure of sorted PMN-MDSC to SX-682 for 24 hours did not alter PMN-MDSC viability (Supplementary Fig. S4). Furthermore, SX-682 treatment did not alter the proliferation of tumor PMN-MDSC in tumor-bearing mice (Supplementary Fig. S5), suggesting that the primary mechanisms of SX-682 was inhibition of trafficking of PMN-MDSC into MOC2 tumors. To provide more definitive assessment of the ability of SX-682 to alter tumor trafficking of MDSC, fluorescently labeled MDSC were adoptively transferred into mice treated with control or SX-682 chow. SX-682 treatment abrogated the trafficking of adoptively transferred PMN-MSC and M-MDSC into MOC2 tumors (Fig. 4C).

Inhibition of tumor PMN-MDSC accumulation alone with SX-682 did not alter tumor accumulation of endogenous NK cells, tumor-infiltrating NK-cell activation status, or tumor infiltration of T lymphocytes (Supplementary Fig. S6A–S6C). We next assessed whether inhibiting PMN-MDSC tumor trafficking altered tumor accumulation of adoptively transferred KIL cells beyond that observed in control mice. Using fluorescently labeled KIL cells measured in whole-tumor digests by fluorescent imaging and flow cytometry, pretreatment of mice bearing MOC2 tumors with SX-682 significantly enhanced tumor accumulation of adoptive transferred KIL (Fig. 4D). Of note, KIL cells expressed CXCR1 but very low levels of CXCR2 (Fig. 4E). Furthermore, upon leukocyte isolation and stimulation, SX-682–treated tumor-infiltrating NK lineage cells expressed IFNγ (Fig. 4F) and granzyme B (Fig. 4G) to a greater degree than control, which may correlate with increased killing capacity. Following in vivo SX-682 treatment, tumor PMN-MDSC expression of cell surface TGFβ or superoxide dismutase 1/2 genes, responsible for the generation of H₂O₂, was not significantly altered (Supplementary Fig. S7A and S7B). Similarly, the suppressive capacity of PMN-MDSC was not significantly altered following SX-682 treatment of tumor-bearing mice or upon ex vivo exposure of sorted PMN-MDSC to SX-682 (Supplementary Fig. S7C and S7D). These results revealed that although SX-682 increased the tumor accumulation and activation status of adoptively transferred KIL cells primarily through inhibition of PMN-MDSC trafficking and not through direct alteration of PMN-MDSC function.

On the basis of these data, we next studied whether treatment of mice bearing established MOC2 tumors with SX-682 could enhance the therapeutic efficacy of adoptively transferred KIL cells. Treatment with SX-682 alone did not alter tumor progression (Fig. 5A). Adoptive transfer of KIL cells (5 × 10⁶ cells/treatment for six treatments) delayed primary tumor growth in a subset of tumors and induced rejection of one of 20 (5%) tumors. Combination SX-682 treatment and KIL adoptive transfer consistently delayed primary tumor growth during the treatment period with significantly decreased tumor volume at day 21 compared with control or monotherapy groups (Fig. 5B). Although the majority of MOC2 tumors rebounded following completion of treatment, combination SX-682 and KIL adoptive transfer led to significantly prolonged survival (Fig. 5C) and induced complete tumor rejection in 4 of 19 (21%) mice. Despite low expression of CXCR1/2, exposure to SX-682 in vitro did not significantly alter MOC2 tumor cell proliferation or cell-cycle progression, viability, expression of NKG2D ligands, or sensitivity to KIL killing (Supplementary Fig. S8A–S8F). Accepting murine KIL cells as a model of adoptively transferred cell therapy, these data demonstrated that partial abrogation of CXCR2⁺ myeloid cell trafficking into tumors with SX-682 enhanced the therapeutic efficacy of adoptively transferred NK cells. The mechanism appeared to be because of inhibition of PMN-MDSC trafficking alone, and not due to tumor cell–specific changes following SX-682 treatment.

The role of peripheral and tumor-infiltrating myeloid cells in mediating T-lymphocyte immunosuppression within patients with HNSCC is well established (7, 8). The role that both neutrophilic and monocytic myeloid cells may play in mediating NK-cell immunosuppression in patients with HNSCC is less well understood. To clarify this role, we studied NK cells and myeloid cells from patients with newly diagnosed, advanced stage human papillomavirus–negative HNSCC. The NK-cell fraction in circulation was not significantly different in patients with HNSCC compared with healthy donors (Fig. 6A). However, when stimulated with K562 cells in the presence of IFNα, NK cells from patients with HNSCC were significantly less activated compared with NK cells from healthy donors, demonstrating reduced activation capacity (Fig. 6B). We next measured circulating putative MDSC based upon cell surface markers (14). Both patients with HNSCC and healthy donors have high levels of circulating CD11b⁺CD33⁺CD15⁺CD14⁻ total neutrophilic cells (Fig. 6C), but patients with HNSCC have greater fraction of CD11b⁺CD33⁺CD15⁻CD14⁺HLA-DR^low monocytic cells compared with healthy donors (Fig. 6D). These CD14⁺HLA-DR^low cells have been shown to be MDSC, whereas CD14⁺HLA-DR^hi myeloid cells more likely represent macrophages (8). Although CD15⁺ cells expressed the greatest levels of CXCR1 and CXCR2, CD14⁺HLA-DR^low cells also expressed greater levels of these chemokine receptors compared with CD14⁺HLA-DRhi myeloid cells or other immune cells such as lymphocytes (Fig. 6E).

We next measured myeloid cell accumulation in freshly digested HNSCC tumors. Within HNSCCs, CD11b⁺CD33⁺ myeloid cells accounted for 38%–55% of all viable CD45⁺ tumor leukocytes (Fig. 6F). Tumors accumulated greater numbers of CD15⁺ neutrophilic myeloid cells compared with CD14⁺HLA-DR^low monocytic myeloid cells (Fig. 6G). Tumors accumulated low numbers of NK cells (Fig. 6H). Sorted peripheral blood or tumor CD15⁺ or CD14⁺ myeloid cells were next assessed for their ability to suppress healthy donor NK-cell activation. A validated IFNγ production assays (ELISpot) was chosen due to its reproducibility and low cell requirement. Although both suppressed NK-cell activation, tumor-infiltrating CD15⁺ and CD14⁺ myeloid cells displayed greater immunosuppressive capacity compared with peripheral CD15⁺ and CD14⁺ myeloid cells. Tumor-infiltrating CD14⁺ myeloid cells suppressed NK-cell activation to a greater degree than tumor-infiltrating CD15⁺ myeloid cells. To study the mechanisms by which tumor-infiltrating myeloid cells suppressed healthy donor NK effector function, coincubation experiments with neutralizing antibodies or select inhibitors were performed (Fig. 6I). Similar to PMN-MDSC in mice, suppression of NK cells by CD15⁺ myeloid cells from HNSCC tumors was reversed in the presence of TGFβ mAb. Coincubation with catalase to inhibit extracellular H₂O₂ had no obvious effect. Conversely, suppression of NK cells by CD14⁺ myeloid cells from HNSCC tumors was reversed in the presence of the NOS inhibitor L-NMMA, but not other inhibitors. These data validated tumor-infiltrating CD15⁺ and CD14⁺ myeloid cells as MDSC capable of suppressing NK-cell function, and demonstrated the nonredundant mechanisms by which different subsets of tumor-infiltrating MDSC can suppress effector cell function in patients with HNSCC. Cumulatively, these results support the inhibition of myeloid cell trafficking into tumors as a means of enhancing the therapeutic efficacy of adoptively transferred NK cells.

This study provides evidence that inhibition of MDSC trafficking with a CXCR1/2 small-molecule inhibitor can enhance the therapeutic efficacy of adoptively transferred NK cells. In the syngeneic oral cancer model MOC2, which harbors few genomic alterations and forms non-T-cell–inflamed tumors that are refractory to T-cell–based immunotherapy (5, 15, 16), CXCR1/2 blockade with SX-682 decreased tumor infiltration of PMN-MDSC and increased tumor infiltration and activation of adoptively transferred murine NK cells. Mechanistically, PMN-MDSC abrogated murine NK-cell function through production of TGFβ and H₂O₂. Arginase, NOS, and PD-L1 did not appear to play dominant roles in NK-cell suppression in this model. In addition, SX-682 did not directly alter tumor cell proliferation, survival, or sensitivity to NK-cell killing and did not alter the suppressive capacity of the PMN-MDSC themselves. Together, these data support that the function and therapeutic efficacy of adoptively transferred NK cells can be enhanced primarily through the mechanism of reduced trafficking of immunosuppressive myeloid cells into the tumor microenvironment.

Our work validates previous reports demonstrating the ability of MDSC to suppress NK-cell function, but adds translational relevance with the use of an orally bioavailable, clinical stage chemokine receptor inhibitor. Stiff and colleagues demonstrated nitric oxide–dependent suppression of NK-cell–mediated antibody-dependent cell-mediated cytotoxicity (ADCC) by Gr-1⁺ myeloid cells in multiple murine cancer models (17). Although neutrophilic and monocytic myeloid cells cannot be differentiated with the Gr-1 marker, this work revealed globally that Gr-1⁺ MDSC can suppress another important NK-cell function, ADCC, in addition to direct cytotoxicity. Our murine work revealed that NK-cell direct cytotoxicity of tumor cells is inhibited through PMN-MDSC TGF-β and H₂O₂. Li and colleagues demonstrated the ability of total (Gr-1⁺) MDSC to suppress NK-cell function through membrane bound TGFβ, similar to our results (18). Additional work in monocytic MDSC isolated from patients with hepatocellular carcinoma revealed suppression of NK-cell function through MDSC interaction with and blockade of NKp46, an activating coreceptor present on NK cells (19). MDSC suppress T-cell function primarily through arginase, but this did not appear to be the case for suppression of NK cells (6, 8). One interpretation of these data is that MDSC possess multiple mechanisms by which they can suppress the effector functions of NK cells. Inhibition of MDSC trafficking with a chemokine receptor inhibitor such as SX-682 to prevent their tumor infiltration altogether may be a more effective therapeutic strategy compared with attempts to block multiple, individual suppressive mechanisms.

SX-682 is an allosteric, small-molecule inhibitor of CXCR1/2 currently undergoing clinical evaluation (NCT03161431). Previous work has demonstrated the ability of SX-682 to reduce tumor accumulation of CXCR2⁺ MDSC in mice to enhance the efficacy of T-cell–based immunotherapy, including immune checkpoint blockade and adoptive T-cell transfer (20, 21). Our work is the first to demonstrate that a similar therapeutic strategy can be used to enhance the efficacy of adoptively transferred NK cells. Of note, SX-682 treatment did not enhance the accumulation or activation of endogenous NK cells. This suggests that approaches to increase the available pool of circulating effector immune cells such as adoptive cell transfer therapies may be needed in tumors that harbor intrinsic defects effector immune cell recruitment (5). Human NK cells express CXCR1, which brings up the possibility that SX-682 could abrogate trafficking of NK cells into tumors. However, KIL also express moderate levels of CXCR1, and tumor infiltration of adoptively transferred KIL was enhanced with SX-682 treatment, suggesting that CXCR1 expression alone is not sufficient to predict the ability of SX-682 to abrogate trafficking. This will need to be carefully studied in clinical trials.

Subclones of tumor cells within a heterogeneous tumor may acquire genomic alterations that disrupt IFN responses or antigen processing and presentation (1–3). These can confer resistance to T-cell–based immunotherapy and represent a significant mechanism of treatment failure. NK cells kill tumor cell targets via antigen- and MHC-independent mechanisms. NK-cell–based immunotherapy, such as adoptive transfer of ex vivo expanded clonal NK cells, may overcome these barriers and serve as an effective monotherapy or adjunct to T-cell–based immunotherapy. Newly developed NK-cell–based cellular therapies are currently under clinical study alone or in combination with other immune-oncology agents (NCT03027128 and NCT03387111). This work provides the preclinical rationale for the combination of such NK-cell therapies with a CXCR1/2 inhibitor designed to abrogate myeloid cell infiltration into the tumor microenvironment.

This work has also definitively demonstrated the presence of CXCR1/2⁺ neutrophilic and monocytic MDSC in the circulation of patients with HNSCC that traffic into the tumor microenvironment. The majority (>50%) of total tumor-infiltrating immune cells were CD15⁺ or CD14⁺ MDSC, validated with functional data demonstrating suppression of NK-cell activation by the sensitive ELISpot assay. Although others have demonstrated tumor infiltration of immunosuppressive CD14⁺HLA-DR^low monocytic MDSC (8) or the presence of CD14⁺ monocytic or CD66b⁺ neutrophilic MDSC in the peripheral blood of patients with HNSCC (7, 22), our work is the first to assign different mechanisms of suppression based on neutrophilic or monocytic differentiation. In our dataset, tumor-infiltrating MDSC were more suppressive than their circulating MDSC counterparts, and tumor-infiltrating CD14⁺ M-MDSC demonstrated greater NOS-dependent immunosuppressive capacity compared with TGFβ-dependent suppression mediated by CD15⁺ PMN-MDSC on a cell-by-cell basis. Other reports characterizing the HNSCC tumor immune microenvironment have failed to consistently document high levels of circulating or tumor-infiltrating neutrophilic cells (22–25), possibly due to the use of cryopreserved samples, which can lead to the loss of neutrophilic cell populations (26, 27).

Clear differences existed between murine and human MDSC that may be of translational importance. First, murine Ly6G^hi PMN-MDSC suppressed NK-cell function through TGFβ and H₂O₂. Human CD15⁺ PMN-MDSC suppressed through TGFβ, but the H₂O₂ inhibitor catalase had no effect on suppressive capacity. Whether direct TGFβ inhibition alone would prove effective in reversing PMN-MDSC–mediated suppression of NK-cell function requires further study. Human M-MDSC patently inhibited NK-cell function through NOS function, but we were unable to comment on whether murine M-MDSC possess a similar mechanism due to the lack of accumulation of these cells in MOC2 tumors. Second, although murine PMN-MDSC expressed high levels of CXCR2 and low levels of CXCR1, both PMN-MDSC and M-MDSC in the circulation of patients with HNSCC expressed greater levels of both CXCR1 and CXCR2 compared with other circulating immune cells. Whether dual inhibition of CXCR1 and CXCR2 with a small-molecule inhibitor such as SX-682 inhibits tumor trafficking of MDSC to a greater degree than more specific CXCR2 inhibitors also requires further study.

A limitation of our study was how we isolated CD15⁺ and CD14⁺ myeloid cells from the peripheral blood and tumors of patients with HNSCC. Following Percoll gradients used to isolate total leukocytes from whole blood or tumor single-cell digestions, positive magnetic selection was used to isolate total CD15⁺ or CD14⁺ populations. Subsets of myeloid cells certainly exist within each of these populations, including HLA-DR high or low CD14⁺ cells and high- or low-density CD15⁺ cells differentially expressing CD16, LOX-1, or fatty acid transport protein 2 (28–30). Functional characterization of these individual cell subsets could yield important insights into MDSC biology in HNSCC, but in our experience, this type of isolation requires FACS that is traumatic to myeloid cells and disrupts their gene expression and viability. Conclusions from our work must be limited to whole CD15⁺ or CD14⁺ cell populations in the periphery or tumor microenvironment of patients with HNSCC.

In conclusion, we have established the importance of MDSC in mediating an NK-cell–suppressive tumor microenvironment, and demonstrated that inhibition of PMN-MDSC trafficking with a dual CXCR1/2 small-molecule inhibitor can enhance the therapeutic efficacy of adoptively transferred NK cells. NK-cell–based cell therapies are under clinical study, and, similar to T-cell–based immunotherapy strategies, combination with agents designed to enhance their efficacy through modulation of the tumor microenvironment will likely be required. NK-cell–based immunotherapy may be particularly useful for patients with tumors that exclude T cells or harbor genomic alterations in IFN response or antigen processing and presentation genes. Human HNSCCs are heavily infiltrated with immunosuppressive PMN- and M-MDSC that possess multiple independent mechanisms of NK-cell suppression. Cumulatively, these data support the clinical investigation of CXCR1/2 inhibition in combination with NK-cell–based immunotherapeutic strategies in patients with HNSCC.

N.C. Schmitt reports receiving commercial research grants from Astex Pharmaceuticals, and is an advisory board member/unpaid consultant for Checkpoint Surgical. D.Y. Maeda is an employee/paid consultant for Syntrix Pharmaceuticals. J.A. Zebala is an employee/paid consultant for and holds ownership interest (including patents) in Syntrix Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Greene, W.K. Mydlarz, J. Schlom, P.E. Clavijo, C. Allen

Development of methodology: W.K. Mydlarz, P.E. Clavijo, C. Allen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Greene, Y. Robbins, W.K. Mydlarz, N.C. Schmitt, L.A. Horn, C. Palena, C. Allen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Greene, W.K. Mydlarz, J. Schlom, C. Allen

Writing, review, and/or revision of the manuscript: S. Greene, W.K. Mydlarz, N.C. Schmitt, L.A. Horn, C. Palena, J.A. Zebala, P.E. Clavijo, C. Allen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.P. Huynh, J. Friedman, J. Schlom, J.A. Zebala

Other (provided SX-682 and SX-682 medicated chow for in vitro and in vivo experiments): D.Y. Maeda

S. Greene, Y. Robbins, W.K. Mydlarz, A.P. Huynh, N.C. Schmitt, J. Friedman, P.E. Clavijo, and C. Allen were supported by the Intramural Research Program of the NIH, National Institute on Deafness and Other Communication Disorders, project number ZIA-DC000087. S. Greene was supported through the NIH Medical Research Scholars Program, a public-private partnership supported jointly by the NIH and contributions to the Foundation for the NIH from the Doris Duke Charitable Foundation (DDCF grant no. 2014194), the American Association for Dental Research, the Colgate-Palmolive Company, Genentech, Elsevier, and other private donors. The authors thank Drs. Christopher Silvin for assistance with flow cytometry, and James Hodge and Jason Redman for their critical review of this article.

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.