Nitric Oxide Production by Myeloid-Derived Suppressor Cells Plays a Role in Impairing Fc Receptor–Mediated Natural Killer Cell Function

Natural killer (NK) cells are large granular lymphocytes that participate in innate immune responses against virus-infected and neoplastic cells via receptors for major histocompatibility antigens, markers of cellular stress (NKG2D), and immunoglobulin G (FcR; ref. 1). NK cells are unique in that they constitutively express only a low-affinity, activating FcR (FcγRIIIa or CD16), enabling them to recognize antibody (Ab)-coated targets (1, 2). FcR-activated NK cells can kill antibody-coated targets and secrete cytokines such as IFNγ and chemokines (e.g., RANTES and MIP-1α) that inhibit tumor cell proliferation, enhance antigen presentation, and promote the chemotaxis of T cells (3). These properties of FcR-activated NK cells are an important component of the response to mAb therapy (4).

NK cells are an important component of immune surveillance against the development of malignancies and the control of established tumors. NK-cell infiltration of established tumors is associated with improved disease prognosis (5, 6). However, NK cells from patients with cancer have reduced cytotoxic function and cytokine production (7). The cause of this dysfunction is not fully understood, but has been attributed to factors secreted by or expressed on the surface of tumor cells (8, 9). Only a few studies have examined the role of immunosuppressive cells in mediating NK-cell dysfunction (10, 11) in the setting of cancer.

MDSCs are immature myeloid cells with immunosuppressive properties that expand in response to tumor and stroma-derived factors (12). The frequency of circulating MDSCs correlates with tumor burden and has prognostic value (13). In mice, MDSCs are identified by expression of Gr-1 and CD11b, and in humans, they are characterized as CD33⁺/CD11b⁺/HLA-DR^low/neg (12). MDSCs inhibit T cells by a number of mechanisms including the production of free radicals, expression of amino acid–catabolizing enzymes, and the secretion of suppressive cytokines (14). Studies in murine models indicate that disruption of MDSC function can improve antitumor immune responses and impair tumor growth (15). Given the ability of MDSCs to suppress antitumor immune responses, they have received significant interest as a potential biomarker and therapeutic target (16).

Studies have shown that MDSCs impair NK-cell cytokine production and MHC-I–dependent cytotoxicity through contact-dependent mechanisms (e.g., checkpoint ligand expression and reactive oxygen species production) and contact-independent mechanisms (arginase and TGFβ production; refs. 10, 11, 17–21). However, the impact of MDSCs on NK-cell FcR-mediated function is unknown. Given the prominence of antibody therapy in the treatment of cancer, elucidation of the effects of MDSCs on FcR-mediated NK-cell function could have important clinical implications. In particular, this approach could identify novel mechanisms of resistance to mAb therapy.

It was hypothesized that MDSCs could inhibit FcR-mediated NK-cell functions and antagonize mAb therapy. Utilizing MDSCs and autologous NK cells from patients with cancer and murine models of mAb therapy, we show that MDSC nitric oxide (NO) production inhibits NK-cell FcR-mediated signal transduction and downstream effector functions including antibody dependent cellular cytotoxicity (ADCC), cytokine production, and antitumor activity in vivo. To our knowledge, this is the first report to demonstrate that MDSC-derived NO can antagonize NK-cell FcR function and blunt the response to antibody therapy.

EMT6, 4T1, K562, HT-29, and Raw 264.7 cells were purchased from ATCC and maintained under the recommended cell culture conditions. Panco2-EGFR⁺ cells were generated as described previously (22). CT26-HER and EMT6-HER2 cells were gifts from the laboratory of Dr. P.T.P. Kaumaya (Ohio State University, Columbus, OH). Cells were cultured in RPMI, Iscove's modified Dulbecco's medium, or DMEM with 10% FBS and 1% antibiotic–antimycotic.

Peripheral blood from patients with cancer was obtained at the Ohio State University Comprehensive Cancer Center under Institutional Review Board approved protocols (OSU IRB Nos. 1999C0348, 2004C0096, and 2010C0036, respectively). NK cells and MDSCs were isolated by 30-minute incubation with the NK- or myeloid cell enrichment cocktail (Stem Cell Technologies) followed by Ficoll-Paque centrifugation (GE Healthcare). For MDSCs, cells were further selected using anti-HLA-DR magnetic beads for 15 minutes at 4°C and isolated using a MS-MACS column (23).

Autologous NK cells and MDSCs were cocultured overnight with or without 2.5 mmol/L L-NIL. Healthy donor NK cells were treated with DMSO or 0.1 mmol/L SNAP (Sigma Aldrich) overnight. Target HT-29 tumor cells were labeled with ⁵¹Cr and then coated with cetuximab or IgG. Target cells were added to NK cells, and following a 4-hour incubation, supernatants were harvested and quantified using a gamma counter. Mean percent cell lysis was determined as described previously (2).

Ninety-six well flat-bottom plates were coated with 100 μg/mL of polyclonal human IgG in cold PBS overnight at 4°C. NK cells, MDSC, or NK cells plus MDSC were plated at 2 × 10⁵ cells/well as described previously (24). Where indicated, cells were treated with 2.5 mmol/L of the iNOS inhibitor L-NIL. Cell-free supernatants were harvested and analyzed for levels of IFNγ, MIP-1α, or TNF-α (ELISA; R&D Systems or eBioscience; ref. 24).

Autologous NK cells and MDSC were cocultured overnight. Where indicated, cells were also treated with 2.5 mmol/L of L-NIL. The cells were incubated with the 3G8 anti-CD16 antibody (0.2 mg/mL) on ice for 30 minutes and then cross-linked with a goat anti-mouse F(ab)₂ for 15 minutes. Cells were permeabilized, stained with phospho-Erk, CD56 antibodies, and analyzed by FACS (24).

NK cells were treated with 0.1 mmol/L S-Nitroso-N-acetylpenicillamine (SNAP; ref. 25). Alternatively, autologous NK cells and MDSC were cocultured overnight at a 1:1 ratio. Cells were washed with cold PBS, incubated in buffer A (Invitrogen), and permeabilized using cold methanol. Cells were stained with anti-CD16 and anti-nitrotyrosine antibodies in buffer B (Invitrogen) and analyzed by FACS.

NK cells were treated with SNAP as described above. Twenty micrograms of anti-nitrotyrosine beads (Millipore 16310) was added to 50 μg of protein lysate and incubated overnight at 4°C. The beads were washed with cold PBS and resuspended in 2× Laemelli buffer and boiled. The sample was loaded on a SDS gel for protein expression by immunoblot analysis. The membrane was probed with an anti-Erk antibody (Cell Signaling Technology; ref. 26).

NK cells isolated from healthy donor leukopaks (American Red Cross) were treated with various concentrations of SNAP or DMSO. NK cells were stained with Annexin V and propidium iodide (PI; BD Pharmingen; ref. 23).

Human MDSCs were stained with a panel of antibodies recognizing CD11b, CD33, HLA-DR, CD15, and CD14 (Beckman Coulter) as well as TIGIT ligands CD155 and CD112 (Biolegend; ref. 13). NK cells were stained using antibodies against CD16 and CD56 (BD Pharmingen). Mouse MDSCs were stained with antibodies against Gr-1 and CD11b (BD Biosciences; ref. 26). Data were acquired using a BD-LSRII and the data were analyzed using the FlowJo software.

MDSCs were cultured in 10% human AB (HAB; Sigma Aldrich) media for 24 hours. Equal amounts of supernatant were mixed with modified Greiss reagent (Sigma Aldrich). Absorbance at 550 nm was measured using a microplate reader. Nitrite concentrations were determined using a standard curve (23).

MDSCs were stained with 2.5 μmol/L DAFFM (Molecular probes, D-23841) for 30 minutes in serum-free media. Cells were then washed three times with serum-free media and incubated for an additional 20 minutes in RPMI complete media. Following incubation, MDSCs were washed and analyzed by FACS (27).

Balb/c mice received single intraperitoneal injections of gemcitabine (80 mg/kg) or 5-FU (50 mg/kg). MDSC depletion was also accomplished by intraperitoneal injection of anti- Gr-1 (250 μg) daily for 5 days. To inhibit MDSC function, Balb/c mice received intraperitoneal injections of the following agents for 5 consecutive days: L-NIL (20 mg/kg), anti-TGFβ Ab (200 μg, BioXcell, clone 1D11.16.8) (28), anti-IL10 Ab (250 μg, BioXcell, clone JES5-2A5), the arginase inhibitor nor-NOHA (20 mg/kg; ref. 29) or the IDO inhibitor, 1-Methyl-d-tryptophan via oral gavage (400 mg/kg, Sigma Aldrich; ref. 30). Mice were sacrificed 24 hours after the last treatment. NK cells were isolated from the spleen using a Mouse NK Cell Enrichment Kit (Stem Cell Technologies).

Following TRIzol extraction (Invitrogen) and RNeasy purification (Qiagen), total RNA was quantitated and reverse transcribed. cDNA was used to measure gene expression by Real-Time PCR using predesigned primer/probe sets and 2× TaqMan Universal PCR Master Mix with18s rRNA as an internal control (Applied Biosystems; ref. 23).

For ex vivo studies of the effect of depletion or inhibition of MDSC on NK-cell function, 4- to 5-week-old female Balb/c mice were injected with 1 × 10⁵ 4T1 mammary carcinoma cells in the mammary fat pad. For in vivo studies of the effect of depletion or inhibition of MDSC function on the response to antibody therapy, Balb/c mice were injected with 1 × 10⁶ EMT6-HER2 mammary carcinoma cells in the mammary fat pad. MDSC transfer studies utilized immune-deficient athymic nude mice (FOXN1^−/−) injected subcutaneously with 1 × 10⁶ Panco2-EGFR⁺ pancreatic carcinoma cells. For the generation of MDSCs in transfer studies, Balb/c mice were injected with 1 × 10⁵ 4T1 mammary carcinoma cells, while C57BL/6 or Nos2^−/− B6.129P2-Nos2^tm1/Lau/J mice were injected with 1 × 10⁵ B16F10 melanoma cells. MDSCs were isolated from the spleen using a Myeloid Derived Suppressor Cell Isolation Kit (Miltenyi Biotec). Tumor volume was measured three times weekly using digital calipers. These studies were conducted under a protocol approved by Ohio State University's Institutional Animal Care and Use Committee (IACUC 2009A0179-R2).

Three days prior to the start of treatment, a cohort of tumor-bearing mice were injected intraperitoneally with 50-μL anti-asialo GM1 polyclonal antibody (1 mg/mL) to deplete NK cells. NK-depleted mice were administered the depleting antibody every 4 days until the end of the study. Depletion of NK cells was confirmed via flow cytometry by staining splenocytes from control and anti-asialo GM1-treated mice with an antibody against DX5.

Statistical differences between treatment groups were determined using an ANOVA model and Student t test. For murine tumor studies, a linear mixed model was employed to model longitudinal tumor volume for mice under each treatment. Comparisons were done at each time point and averaged across all time points using t statistics. The Holm–Bonferroni method was used for adjusting raw P values for multiple comparisons across treatment groups.

Written informed consent was received from all the human participants prior to inclusion in the experimental studies under Institutional Review Board–approved protocols (OSU IRB nos. 1999C0348, 2004C0096 and 2010C0036). All mouse studies were conducted under a protocol approved by Ohio State University's Institutional Animal Care and Use Committee (IACUC 2009A0179-R2). Studies were conducted in accordance with ethical guidelines as detailed in the U.S. Common Rule.

Given the ability of MDSCs to interfere with the antitumor immune responses, it was hypothesized that MDSCs could inhibit the ADCC function of NK cells. Autologous MDSCs inhibited MHC-I–dependent NK-cell cytotoxicity (K562 cytotoxicity assay), which was dose dependent, while PBMC had no effect (Supplementary Fig. S1A and S1B). The effect of MDSCs on NK-cell ADCC was tested next. MDSCs isolated from the peripheral blood of patients with melanoma (Fig. 1A; Supplementary Fig. S1C), head and neck squamous cell carcinoma (HNSCC, Fig. 1B), and breast cancer (Fig. 1C) all inhibited NK-cell ADCC function. MDSC inhibition of NK-cell ADCC reached statistical significance in the study of four melanoma patients, whereas PBMCs had no effect (Fig. 1A; Supplementary Fig. S1C). Again, the ability of MDSCs to inhibit NK-cell ADCC was dose dependent (Supplementary Fig. S1D). A phenotype of CD33⁺, CD11b⁺, and HLA-DR^low was used to identify MDSCs from melanoma patient blood draws. Before isolation, the frequency of MDSCs was 33.15% among total PBMCs and after isolation the frequency was 76% representing a 2.3-fold enrichment (Supplementary Fig. S2A and S2B). Isolated MDSCs were able to suppress T-cell proliferation (Supplementary Fig. S2C).

NK cells are an important source of cytokines, such as IFNγ. Our group has shown in vitro and in phase I clinical trials that costimulation of NK cells via the FcγRIIIa and cytokines is a potent stimulus for the production of IFNγ and chemokines such as RANTES and MIP-1α (31). Therefore, the effect of MDSCs on NK-cell cytokine production was examined. Coculture of autologous MDSCs and NK cells from patients with melanoma significantly inhibited the production of IFNγ, whereas PBMC did not (Fig. 1D, P < 0.05; Supplementary Fig. S3A). This held for FcR-stimulated NK cells cultured with IL12 (Supplementary Fig. S3B). MDSC inhibition of IFNγ production was dose dependent, and a time-course experiment showed this effect was observable at 24 hours with maximal inhibition at 48 hours (Supplementary Fig. S3C and S3D). Coculture of NK cells with autologous MDSCs also significantly decreased the production of MIP-1α (Supplementary Fig. S3E, P < 0.01).

Erk activation is critical to NK-cell FcR-mediated effector functions and natural cytotoxicity (K562 killing). Given the impairment of these NK-cell functions in the presence of MDSCs, it was hypothesized that impaired Erk activation could lead to reduced NK-cell FcR-mediated functions following coculture with MDSCs (24). NK cells were stimulated via the FcR using the 3G8 anti-CD16 antibody and a cross-linking F(ab′)2 fragment. Measurement of p-Erk levels in CD56⁺ NK cells showed that coculture of melanoma patient NK cells and MDSCs resulted in a 40% decrease in p-Erk levels (Fig. 1E, P < 0.05 and representative dot plot; Supplementary Fig. S4). When NK cells were physically separated from MDSCs, levels of p-Erk in response to FcR stimulation were inhibited by an average of 28.3% (Fig. 1F, P < 0.05). When these cells were in direct contact, there was a small increase in the level of inhibition in comparison with the contact-independent condition (Fig. 1F). This result suggests that MDSC inhibition of NK-cell FcR-mediated signal transduction relies on diffusible substances with the potential for an additional contact-dependent mechanism to play a role.

MDSCs can promote immune suppression through several contact-independent mechanisms including expression of amino acid catabolizing enzymes, immunosuppressive cytokines, and production of nitric oxide (NO). To investigate the role of these factors in suppressing FcR-mediated NK-cell function, mice bearing 4T1 tumors were treated with neutralizing anti-IL10 (32) or anti-TGFβ (28) antibodies, or inhibitors targeting 2,3-indolamine dioxygenase (IDO; ref. 30), arginase (29), or inducible nitric oxide synthase (iNOS). NK cells were isolated from the spleen and used in ADCC assays against trastuzumab-coated CT26 cells expressing human HER2. Only inhibition of iNOS and arginase rescued NK-cell ADCC activity (Fig. 2A–C). Arginase and iNOS both use arginine as a substrate and MDSCs express high levels of both enzymes. This suggests that the arginase/iNOS arginine catabolism pathway in MDSCs plays an important role in regulating NK-cell function, and that manipulation of either pathway could impact NK-cell function. The arginase inhibitor produced a reduction in splenic MDSC frequency suggesting that the enhanced NK function in this group could reflect reduced MDSC accumulation (Supplementary Fig. S5A–S5C). Alternatively, as both arginase and iNOS utilize arginine as a common substrate, and arginine availability has been linked to NK-cell function, inhibition of either enzyme could improve NK-cell function. If this was the case, one could speculate that simultaneous inhibition of both enzymes would dramatically rescue NK-cell function. However, when this was tested, the inhibition of both enzymes was no more effective at rescuing NK-cell function than inhibition of either enzyme alone (Supplementary Fig. S5D). Together, these results suggest that the NO-arginase arginine catabolism pathway plays an important role in the regulation of NK-cell FcR-mediated functions and that iNOS is an important mediator of MDSC inhibition of NK-cell function. However, as it can be hard to draw firm conclusions from negative data obtained using neutralizing antibodies and chemical inhibitors, we concede that factors in addition to NO could still play an important role in the inhibitory effects of MDSCs.

To investigate the role of NO in MDSC inhibition of FcR-mediated NK-cell functions, NOS2 expression by MDSC was measured. MDSCs expressed high levels of NOS2 compared with NK and T cells (Supplementary Fig. S6A). NO production by MDSCs was detected by intracellular flow cytometry using a NO-sensitive dye 4-amino-5-methylamino-2′,7′-difluorofluorescein diacetate (DAF-FM; Supplementary Fig. S6B), and the iNOS inhibitor L-NIL significantly inhibited NO production as measured by the Greiss reagent (Supplementary Fig. S6C, P < 0.05).

It was hypothesized that L-NIL could rescue FcR-mediated NK-cell functions in the presence of MDSC. Melanoma patient NK cells and MDSC were cocultured in the presence or absence of L-NIL, and L-NIL was found to rescue NK-cell ADCC (Fig. 2D, P < 0.01). These effects were also observed when examining the natural cytotoxicity function of NK cells (Supplementary Fig. S6D). Similarly, L-NIL rescued IFNγ production and Erk activation in the presence of MDSCs (Fig. 2E and F, P < 0.05). When NK cells and MDSCs were separated by a permeable membrane, L-NIL still rescued the generation of p-Erk supporting a predominant contact-independent mechanism (Supplementary Fig. S6E).

To examine the effect of NO on NK-cell function, NK cells from normal donors were incubated with a NO donor, S-Nitroso-N-Acetyl-penicillamine (SNAP, 0.1 mmol/L) or DMSO overnight in the presence or absence of IL12. SNAP significantly inhibited the ability of NK cells to lyse cetuximab-coated HT-29 colon cancer cells (Fig. 3A and B, P < 0.001). Similarly, SNAP significantly reduced the production of IFNγ (regardless of the presence of IL12; Fig. 3C, P < 0.05; Supplementary Fig. S7A). SNAP also inhibited Erk activation in FcR-stimulated NK cells (Fig. 3D, P < 0.05). Importantly, SNAP was not toxic to NK cells as the viability of SNAP-treated NK cells was >90% as measured by flow cytometry and Trypan blue exclusion (Fig. 3E and F; Supplementary Fig. S7B). In addition, SNAP did not affect the total number of NK cells remaining after treatment (Supplementary Fig. S7C). Finally, SNAP produced levels of NO that were equivalent to that produced by human MDSCs (data not shown). Together, these results suggest that NO can be a negative regulator of NK-cell FcR function.

Given the ability of MDSCs to inhibit Erk activation in NK cells and the ability of L-NILs to rescue this function, it was hypothesized that impaired Erk activation could be caused by nitration of signaling molecules downstream of the FcR. To investigate this, NK cells were treated with SNAP or DMSO, probed with anti-nitrotyrosine (NT) and anti-CD16 (FcR) antibodies, and analyzed by flow cytometry. SNAP treatment resulted in significant nitration of tyrosine residues on CD16⁺ NK cells compared with DMSO (Fig. 4A, P < 0.05). Coculture of MDSCs with NK cells also caused a significant increase in the tyrosine nitration of CD16⁺ NK cells (Fig. 4B, P < 0.01). Representative flow cytometry dot plots showing the change in nitrotyrosine-positive NK cells in the above two experiments are provided in Supplementary Fig. S8. Treatment of human NK cells with SNAP or coculture with MDSCs had no effect on CD16 expression (data not shown). Next, nitrotyrosine residues were immunoprecipitated from DMSO or SNAP-treated NK cells and probed for Erk. Immunoblot analysis showed that Erk was nitrated in SNAP-treated NK cells (Fig. 4C). This suggests that MDSC-derived NO results in the nitration of tyrosine residues on signaling proteins like Erk as one mechanism by which MDSC inhibit FcR-mediated NK-cell functions. With respect to the production of NO by murine MDSCs, our group has previously used IHC to show that large amounts of NO are produced by MDSCs present in the spleens of tumor-bearing mice (26). Also, it has been shown in the literature that murine MDSCs produce significant quantities of NO (12, 33).

Chemotherapeutic agents 5-fluorouracil (5-FU) and gemcitabine deplete MDSCs in tumor-bearing mice (34). Therefore, we investigated whether the elimination of MDSC from tumor-bearing mice could augment NK-cell function. NK cells were isolated from the spleen of mice with 4T1 mammary carcinomas 24 hours after treatment with 5-FU (50 mg/kg), gemcitabine (80 mg/kg), or PBS and used in an ADCC assay against trastuzumab-coated CT26-HER2 cells. NK cells from nontumor-bearing mice demonstrated cytotoxic activity in the range of 40% at the 25:1 effector-to-target (E:T) ratio, whereas NK cells from the PBS group exhibited cytotoxic activity of just 10%. Gemcitabine and 5-FU eliminated MDSCs from the spleens of tumor-bearing animals and this was associated with NK-cell ADCC function in the range of 30% lysis (Fig. 5A and B; Supplementary Fig. S9A and S9B). In contrast to other reports, treatment with an anti-Gr-1 antibody only modestly reduced MDSC frequency and this correlated with a slight increase in NK-cell ADCC function (Supplementary Fig. S9C and S9D). Together, these results demonstrate that MDSC in the spleen can impair NK-cell function and their elimination can augment NK-cell ADCC.

Next, it was investigated whether inhibition of iNOS could augment NK-cell function in vivo. Mice bearing 4T1 tumors were treated daily with PBS or the iNOS inhibitor L-NIL (20 mg/kg) for 1 week (35). NK cells isolated from the spleen were used in an ADCC assay against trastuzumab-coated CT26-HER2 cells. Treatment with L-NILs significantly improved the ADCC function of NK cells (Fig. 5C and D, P < 0.05). As seen in Supplementary Fig. S9E, there was minimal difference in the frequency of MDSCs between the PBS and NIL-treated groups.

NO generated by cancer cells has been postulated to dampen antitumor immune responses (36, 37). To delineate the role of tumor versus MDSC-derived NO, the above experiment was repeated using Nos2^−/−-deficient mice. NK cells from B16F10 tumor–bearing Nos2^−/− mice demonstrated normal ADCC activity compared with NK cells from wild-type B16F10 tumor-bearing mice (Fig. 5E). These results imply that host-derived and not tumor-derived NO is involved in the regulation of MDSC cell function in this tumor model.

mAbs like trastuzumab are widely used in the treatment of cancer, and there is evidence that NK cells participate in their efficacy (38, 39). Therefore, it was hypothesized that MDSCs could antagonize mAb therapy. To this end, we employed a well-characterized EMT6 mouse model of mammary carcinoma that overexpresses human HER2 (EMT6-HER2) and responds to antibody therapy in an NK-cell–dependent manner (40). Importantly, EMT6-HER2 cells show similar growth kinetics as EMT6 parental cells and drive the expansion of MDSCs (unpublished data). Once tumors were palpable (∼50 mm³), mice were treated with PBS/IgG, 5-FU (MDSC-depleting agent), trastuzumab, or 5-FU followed by trastuzumab. Tumors of mice receiving 5-FU alone or 5-FU in combination with trastuzumab exhibited a significant decrease in MDSCs (Supplementary Fig. S10A and S10B). There was only a modest reduction in tumor volume in mice treated with trastuzumab compared with the PBS/IgG-treated control group as might be expected in a model in which the MDSC population is so prominent. Mice treated with 5-FU showed a reduction in tumor volume compared with the PBS/IgG-treated group. Importantly, 5-FU markedly enhanced the effectiveness of trastuzumab compared with either agent alone (Fig. 6A, P < 0.01 and P < 0.001, respectively). It is possible that 5-FU had a direct antitumor effect in this tumor model, but IHC staining for Ki67 and caspase-3 showed no difference in proliferation or apoptosis between these treatment groups (Supplementary Fig. S11). These results are consistent with previous studies showing 5-FU did not induce apoptosis of tumor cells or affect the prevalence of immune cells (NK and T cells) within tumors (34). We propose that one major effect of 5-FU on tumor growth is due to its actions on the MDSC compartment.

Next the effect of NO inhibition on the response to trastuzumab was examined. EMT6-HER2 tumor-bearing mice were treated with PBS/IgG, L-NIL, trastuzumab or L-NIL followed by trastuzumab. No difference in tumor volume was observed between the PBS/IgG and L-NIL–treated groups, and trastuzumab modestly reduced tumor growth (Fig. 6B). However, there was a significant decrease in tumor volume when mice received NIL in combination with trastuzumab (P < 0.005, Fig. 6B). While there was a slight reduction of splenic MDSCs in treatment groups compared with the control, the frequency of tumor-infiltrating MDSC was similar indicating the observed differences in splenic MDSC frequency were likely driven by group differences in tumor volume (Supplementary Fig. S12). Importantly, the enhanced antitumor effect of the combination of trastuzumab and L-NIL was found to be NK-cell–dependent as depletion of NK cells with anti-asialo GM1 rescued tumor growth even in the setting of combination therapy (Fig. 6C; Supplementary Fig. S13A).

To confirm the ability of MDSC to antagonize mAb therapy of cancer, an MDSC adoptive transfer experiment was performed. Athymic nude mice were inoculated with a pancreatic cancer cell line that overexpresses human HER1 (Panco2-EGFR). The parental and HER1-expressing cell lines grow at identical rates in this mouse strain and there is minimal expansion of MDSCs (unpublished data). Following the establishment of tumors, mice were injected with PBS (mock injection), splenocytes from nontumor-bearing mice, or MDSCs from syngeneic mice bearing 4T1 mammary carcinoma tumors on day 7 prior to the initiation of IgG or cetuximab therapy on day 9. A second round of cell injections was conducted on day 14 (schema shown in Supplementary Fig. S13B). Mice receiving splenocytes and cetuximab showed a significant reduction in tumor volume compared with the control group. In contrast, MDSCs ablated the effect of cetuximab therapy (Fig. 6D and E; P < 0.05).

To elucidate the role of NO production in the inhibitory effects of transferred MDSCs, a similar study was conducted using MDSC from nos2^−/− deficient mice. As before, injection of wild-type MDSC abolished the antitumor effect of cetuximab therapy. However, nos2-deficeint MDSC were less antagonistic of cetuximab (Supplementary Fig. S13C). Taken together, these results demonstrate that MDSCs can dampen the antitumor response of mAb therapies and that NO production appears to play an important role in mediating this effect.

The current study provides evidence for a model in which MDSCs inhibit FcR-mediated NK-cell actions against antibody-coated tumor cells through the production of NO. The results suggest that nitration of NK-cell proteins on tyrosine residues resulted in impaired signal transduction and activation of NK cells following FcR stimulation. Elimination of MDSCs from tumor-bearing hosts or inhibition of NO production restored NK-cell effector functions and improved the anti-tumor effect of mAb therapy in vivo. Finally, transfer of wild-type MDSCs significantly antagonized mAb therapy. The results reported provide a mechanism by which MDSC-derived NO leads to impaired FcR mediated NK-cell function and reduced efficacy of mAb therapy for cancer.

NK cells play an important role in the response to mAb therapy through their ability to recognize and kill antibody-coated tumor cells via engagement of the low-affinity activating FcγRIIIa (FcR or CD16) and inhibiting metastatic spread of cancer. As a result, NK cells have received attention for their potential therapeutic and prognostic value (41). In both solid and hematologic malignancies, alterations in the frequency, maturity, and expression of activating and inhibitory receptors by NK cells have been reported that correspond with impaired NK-cell effector functions (42). While the mechanisms of this dysfunction are incompletely understood, reports have begun to explore these observations by analyzing the interaction between NK cells and cells with immunosuppressive function including MDSCs (10, 11, 18, 19, 21, 43–45).

This study provides evidence that MDSCs impair NK cell FcR-mediated functions via the production of NO and the nitration of protein tyrosine residues. Our group has previously shown that MDSC-derived NO results in nitration of key tyrosine residues on the STAT1 transcription factor resulting in impaired NK and T-cell response to IFN stimulation (26). Other groups have shown that MDSC-derived RNS can cause nitration of the T-cell receptor, thus impairing its interaction with the peptide MHC (46). In addition, RNS has been shown to cause posttranslational modification of chemokines resulting in impaired T-cell trafficking within tumors (47). Finally, tumor-derived NO has also been shown to promote immune dysfunction. Coculture of PBMCs with melanoma cell lines expressing high levels of NOS1 resulted in tyrosine nitration of T cells and monocytes. The findings that MDSC-derived NO resulted in the nitration of Erk and impaired FcR-mediated NK-cell function are the first to explore the effects of MDSCs on the NK-cell–mediated response to antitumor mAb therapy. These results further support the ability of NO to mediate immune suppression and negatively impact mAb-based treatment of cancer.

NK cells also appear to have a role in preventing metastatic spread of cancer and in the control hematologic malignancies. Notably, there are reports showing that immunosuppressive MDSC are increased in patients with leukemia (48). Our results showing that MDSCs can inhibit NK-cell natural cytotoxicity suggest that MDSCs could promote metastasis through impaired NK-cell function. Indeed, investigators have shown that MDSCs promote metastasis in the setting of breast cancer (49), but to date no report has implicated MDSC inhibition of NK-cell function in this process.

Erk is rapidly phosphorylated in primary human NK cells following FcR stimulation and this is critical to NK-cell cytokine production and ADCC (24, 50). Erk phosphorylation in FcR-stimulated NK cells was significantly inhibited by MDSCs, but this could be rescued in the presence of an iNOS inhibitor. Treatment of NK cells with the NO donor SNAP resulted in the nitration of Erk on tyrosine residues suggesting that this is one mechanism by which MDSCs inhibit Erk activation. Recently, it was identified that MDSC can inhibit p-Erk generation in NK cells through the inhibitory TIGIT receptor on NK cells (17). Analysis of NK cells and MDSCs from patients used in this work confirmed expression of these molecules. However, the ability of the iNOS inhibitor L-NIL to rescue p-Erk generation in NK cells cultured in direct contact with MDSC suggests that TIGIT may play an alternative role in this system.

Few studies have investigated the relationship between NO and NK-cell function. Earlier studies revealed that autocrine production of NO by NK cells could have a positive effect on NK-cell function, and that human NK cells appear to express endothelial NO synthase but not iNOS (51, 52). The amount of NO produced by activated NK cells is reported to be in the nanomolar range, which is substantially lower than the 2 to 20 μmol/L range of NO produced by MDSCs. This suggests that exposure to higher levels of NO produced by MDSC is detrimental to NK-cell activation and function.

In addition to producing NO, MDSCs mediate immune suppression through expression of arginase and 2,3-indoleamine dioxygenase (IDO) and inhibitory cytokines (TGFβ and IL10; ref. 12). Treatment of tumor-bearing mice with neutralizing antibodies against TGFβ and IL10 or the IDO inhibitor 1-methyl-d-tryptophan did not improve NK-cell ADCC. Nevertheless, based on studies in liver and lung cancer models, it is likely that these immunosuppressive molecules can contribute to impaired NK-cell function in vivo in other murine models and in humans (19). Notably, the arginase inhibitor nor-NOHA modestly improved NK-cell ADCC function. This finding is consistent with reports showing that depletion of L-arginine results in impaired NK-cell function (20). Arginase and iNOS utilize L-arginine as a substrate and MDSCs are known to express high levels of arginase and iNOS (12). As a result, inhibition of either could increase the availability of arginine and decrease NO generation resulting in improved NK-cell function.

The presence of immunosuppressive cells like MDSCs in the tumor microenvironment can be a major impediment to immune-based therapies (53). There are over a dozen mAbs approved by the FDA for the treatment of cancer. These agents appear to function, in part, through the activation of NK cells (4). The finding that MDSC can inhibit FcR-mediated activation and function of NK cells suggests that eliminating MDSCs or inhibiting their immunosuppressive function could also enhance mAb therapy. Recently sunitinib was found to reduce the frequency of monocytic MDSCs and their expression of arginase (54). In addition, imatinib and dasatinib were found to reduce the frequency of suppressive myeloid cells and increase the frequency of NK cells (55). Our group recently demonstrated that ibrutinib, a Bruton's tyrosine kinase inhibitor, reduced MDSCs in tumor-bearing mice and MDSC NO production (23). These results suggest that combining antibody therapy with kinase inhibitors active against MDSC could be an effective strategy. In general, strategies to deactivate or deplete MDSCs and other immunosuppressive cells could enhance the efficacy of immune-based therapies.

In summary, this report provides experimental evidence for an MDSC-mediated mechanism of action for impairment of FcR-medatied NK-cell function. MDSCs produce NO and are associated with inhibition of NK-cell ADCC, cytokine production, FcR signal transduction, and response to mAb therapy. These results suggest that mAb therapy in combination with agents targeting MDSCs could be a successful therapeutic strategy and provide a rationale for the development of clinical trials to test such combinations.

No potential conflicts of interest were disclosed.

Conception and design: A. Stiff, P. Trikha, M.A. Caligiuri, W.E. Carson

Development of methodology: A. Stiff, P. Trikha, B. Benner, N. Muthusamy, J.C. Byrd, W.E. Carson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Stiff, P. Trikha, T.A. Mace, K. Kendra, A. Campbell, D. Abood, M.C. Duggan, R. Wesolowski, M. Old, J.H. Howard, T. Olencki, E.L. McMichael, S. Gautam, I. Landi, V. Hsu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Stiff, P. Trikha, B.L. Mundy-Bosse, T.A. Mace, B. Benner, M.C. Duggan, M. Old, L. Yu, M.A. Caligiuri, W.E. Carson

Writing, review, and/or revision of the manuscript: A. Stiff, P. Trikha, B.L. Mundy-Bosse, T.A. Mace, B. Benner, A. Campbell, M.C. Duggan, R. Wesolowski, J.H. Howard, L. Yu, N. Muthusamy, J.C. Byrd, M.A. Caligiuri, W.E. Carson, S. Tridandapani

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. Stiff, P. Trikha, J.H. Howard, N. Stasik

Study supervision: P. Trikha

This work was supported by NIH grants P01 CA95426, K24 CA93670 (to W.E. Carson), T32 CA90338-27, P30 CA016058, and OSUCCC Translational Therapeutics Award TT142. This work was also supported by the Pelotonia Fellowship Program.

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.