Human Head and Neck Squamous Cell Carcinoma–Associated Semaphorin 4D Induces Expansion of Myeloid-Derived Suppressor Cells

One of the mechanisms by which malignancies can induce immune suppression is through the production of cytokines that affect the maturation and differentiation of inflammatory cells in the tumor microenvironment. Semaphorin 4D (Sema4D) is a proangiogenic cytokine produced by several malignancies, which has been described in the regulation of the immune system. In the present study, we examined the role of human head and neck squamous cell carcinoma (HNSCC)–secreted Sema4D on myeloid cell differentiation. CD33+ cells cultured in HNSCC cell line–derived conditioned medium differentiated into myeloid derived suppressor cells (MDSC) (CD33+CD11b+HLA-DR−/low). The addition of anti-Sema4D Ab to HNSCC conditioned medium significantly reduced the expansion of the MDSC population. Similarly, knockdown of Sema4D in an HNSCC cell line resulted in a loss of MDSC function as shown by a decrease in the production of the immune-suppressive cytokines arginase-1, TGF-β, and IL-10 by MDSC, concomitant with recovery of T cell proliferation and IFN-γ production following stimulation of CD3/CD28. Importantly, CD33+ myeloid and T cells cultured in conditioned medium of HNSCC cells in which Sema4D was knocked down promoted antitumor inflammatory profile, through recovery of the effector T cells (CD4+T-bet+ and CD8+T-bet+), as well as a decrease in regulatory T cells (CD4+CD25+FOXP3+). We also showed that Sema4D was comparable to GM-CSF in its induction of MDSC. Collectively, this study describes a novel immunosuppressive role for Sema4D in HNSCC through induction of MDSC, and it highlights Sema4D as a therapeutic target for future studies to enhance the antitumorigenic inflammatory response in HNSCC and other epithelial malignancies.

H ead and neck squamous cell carcinoma (HNSCC) is a malignancy of high morbidity and mortality, with 45,780 new cases and 8,650 estimated deaths of oral and pharyngeal cancer estimated to occur in the United States in the year 2015 (1). There is accumulating evidence indicating the immunomodulatory effects of HNSCC by which it can escape and/or suppress the immune system (2)(3)(4)(5)(6).
Myeloid-derived suppressor cells (MDSC) have been described in peripheral blood, draining lymphoid tissue, and tumor tissue of several malignancies (5,(7)(8)(9)(10). Circulating MDSC correlated with advanced stages of HNSCC (stages III and IV) as well as other carcinomas (8,10,11). MDSC represent a key player in immune regulation in the tumor microenvironment. It is generally agreed that they comprise a heterogeneous population of myeloid progenitor cells and immature myeloid cells that have a suppressive function on T cells (12,13). MDSC described in human malignancies have the phenotype of CD33 + , CD11b + , and non-lineage determined with poor Ag presentation abilities (HLA-DR 2/low ). They can have a progranulocytic phenotype expressing CD66b or CD15 (polymorphonuclear leukocyte-MDSC) or monocytic features expressing CD14 (10,14,15). MDSC induce their immunesuppressive effect mainly through production of arginase-1 and inducible NO synthase, which consume extracellular arginine and accordingly suppress T cell activation in an Ag-nonspecific manner in the tumor microenvironment. However, they mediate Ag-specific suppression by NADPH oxidase production of reactive oxygen and nitrogen species, particularly in peripheral lymphoid tissue, as well as by other mechanisms (12,(15)(16)(17). In addition to direct T cell suppression, recent evidence suggests a role for MDSC in the expansion of CD4 + CD25 + FOXP3 + regulatory T cells (Tregs) in the tumor microenvironment through both TGF-b-dependent and independent pathways (11,18). Although several mechanisms have been described by which tumor cells induce MDSC, the specific pathways by which HNSCC recruit, expand, and activate MDSC remain to be investigated (15,19,20).
Tumor cells overexpress several cytokines to manipulate their own microenvironment, among which are multiple semaphorins, which have the potential to act on different stromal cells (18). Semaphorin 4D (Sema4D; CD100) is a transmembrane glycoprotein belonging to the fourth group of the semaphorin family that can also be found in a soluble form following proteolytic cleavage. It was initially identified as an evolutionarily conserved chemorepellent protein that regulates axonal guidance in the developing nervous system (21). Later on, its interactions in other systems were emphasized, including the cardiovascular system and immune system. In the immune system, Sema4D is described as being expressed abundantly on resting T cells and weakly on resting B cells and APCs (22)(23)(24)(25)(26). Two opposing roles of Sema4D have been described in the immune system. One role is a proinflammatory response where, for example, in the humoral and cellmediated immune system, Sema4D acts on B cells and dendritic cells, respectively, promoting proinflammatory cytokines (25)(26)(27). Sema4D expressed by T cells and NK cells has also been implicated in their activation through a Sema4D-associated tyrosine kinase (28), and it has been shown to play a role in T cell priming and accordingly in the pathogenesis of autoimmune diseases (29). Alternatively, an anti-inflammatory role of Sema4D in the immune system has also been described. On monocytes and immature dendritic cells, Sema4D can act on plexin C1 and plexin B1, respectively, inhibiting their migration, but not that of mature dendritic cells, which can provide more interaction between immature myeloid cells and T cells (30,31). Furthermore, in vitro studies have shown that Sema4D can modulate cytokine production by monocytes and dendritic cells, and it can induce a significant increase in the anti-inflammatory cytokine IL-10 and a decrease in the proinflammatory cytokines IL-6, IL-8, and TNF-a (26,30,31).
The role of Sema4D produced by tumor cells in inducing tumor angiogenesis, invasiveness, and progression has been demonstrated in several malignancies, both in human and animal models, and correlates with poor prognosis (32)(33)(34)(35)(36)(37)(38). Sema4D overexpression has been described in a large number of human HNSCC tumor tissue samples and in a panel of primary and metastatic cell lines (32). In the tumor microenvironment, tumor-associated macrophages were reported to be the main producers of Sema4D in a mouse breast cancer model (39). Taken together, the hypothesis tested in the present study was that the net inflammatory profile orchestrated by Sema4D in the tumor microenvironment is protumorigenic through induction of immune-suppressive cells.
In this study, we investigated whether HNSCC-derived Sema4D plays a role in inducing MDSC and examined its effect on T cell phenotype and function. We found that Sema4D produced by the HNSCC HN6 and HN13 cell lines polarized myeloid cells into an MDSC phenotype, which corresponded with a reduction in T cell proliferation and IFN-g production. Furthermore, short hairpin RNA (shRNA) inhibition of Sema4D in HN6 resulted in a decrease in the production of the immunosuppressive factors arginase-1, TGF-b, and IL-10 by myeloid cells and the recovery of autologous T cell proliferation, as well as IFN-g production, specifically leading to an increase in the effector T cell population and decrease in Tregs. Our findings describe HNSCC-associated Sema4D production as one of the mechanisms by which the tumors induce MDSC, which subsequently mediate their immunosuppression effects on T cells.

Cell lines and tissue culture
Human normal oral keratinocytes (NOK) and human-derived HNSCC cell lines WSU-HN6 and WSU-HN13 underwent DNA authentication (Johns Hopkins Genetic Resources Core Facility, Baltimore, MD) to ensure consistency in cell identity in comparison with their source (40). SCC-9 was obtained from American Type Culture Collection (Manassas, VA). The WSU-HN4 cell line was a gift from Dr. Silvio Gutkind (National Institute of Dental and Craniofacial Research, Bethesda, MD) (40). All HNSCC cell lines used were derived from primary carcinomas of the tongue. All cell lines were grown and maintained in DMEM supplemented with 10% FBS, 100 U/ml penicillin, 100 mg/ml streptomycin, and 250 ng/ml amphotericin B (Sigma-Aldrich, St. Louis, MO) at 37˚C in humidified air with 5% CO 2 . For treatment with conditioned medium, culture medium was collected from confluent HNSCC cells grown in 5 ml DMEM for 24 h. For myeloid and T cell coculture experiments, conditioned medium supplemented with 5% FBS was used.

Separation of myeloid and T cells from human peripheral blood
Mononuclear cell suspension was prepared from peripheral blood of healthy donors obtained from commercial vendors, that is, Biological Specialty (Colmar, PA) or New York Blood Center (New York, NY), using Ficoll-Paque Plus density gradient centrifugation (GE Healthcare Life Sciences, Pittsburgh, PA). CD33 + myeloid cells were separated from the PBMC using MACS CD33 microbeads (catalog no.130-045-501), and HLA-DR + cells were separated using anti-HLA-DR microbeads (catalog no.130-046-100; Miltenyi Biotec, Auburn, CA). Autologous T cells were enriched by negative selection with the human pan T cell isolation kit (catalog no. 19051; Stemcell Technologies, Vancouver, BC, Canada).

Anti-Sema4D treatment and immune depletion
For target protein inhibition in the conditioned medium, anti-Sema4D (catalog no.610670) (BD Biosciences) was used at a concentration of 10 mg/ml. For immune depletion, immunoprecipitation of Sema4D or GM-CSF was carried out by overnight incubation of the conditioned media with Sema4D (catalog no. ab39710) (Abcam, Cambridge, MA) or GM-CSF (clone BVD2-21C11; catalog no.502301) (BioLegend; 10 mg/ml) mAbs using protein A beads, followed by precipitation.

Statistical analyis
Student paired t tests were performed as appropriate. All data analysis was presented with custom SD. The p values are *p # 0.05; **p # 0.01; ***p # 0.001.

Results
Sema4D expressed by HNSCC cell lines induces CD33 + , CD11b + , HLA-DR 2/low cells To examine the effects of HNSCC-derived condition medium on myeloid differentiation, total human PBMC or total myeloid cells (CD33 + ) isolated from PBMC were plated in conditioned medium from HNSCC cell lines HN4, HN6, HN13, and SCC-9. Total PBMC or separated myeloid cells cultured in HNSCC conditioned medium displayed markers characteristic of MDSC, namely CD33 + CD11b + HLA-DR 2/low compared with cells cultured in normal medium (NM) or NOK conditioned medium. The percentage of MDSC developed from total PBMC was ∼20% (Supplemental Fig. 1A), whereas that generated from CD33 + myeloid cells averaged 30% depending on the HNSCC cell line and density used (Fig. 1A,  Supplemental Fig. 1B). Moreover, titration of HN6-and HN13derived conditioned medium with NM corresponded with a dosedependent decrease in the MDSC population (Fig. 1B, 1C). Several groups have recently described an immunomodulatory role for tumor-produced proangiogenic factors (41). Given the immune modulatory role previously described for Sema4D in the immune system (27,30), we sought to investigate whether the proangiogenic factor Sema4D, secreted by HNSCC (32), plays a role in the induction of CD33 + CD11b + HLA-DR 2/low myeloid cells. Immunoblotting was performed on conditioned medium from HN6 and HN13 HNSCC cell lines, and their cellular lysates served as controls. We detected the cleaved form of Sema4D in the medium at ∼120 kDa ( Fig. 2A) as previously reported (42). We then neutralized Sema4D in the HN4-, HN6-, HN13-, and SCC-9 conditioned medium using anti-Sema4D Ab (clone 30/CD100) and then used the medium to culture myeloid cells. Blocking Sema4D in HNSCC conditioned medium resulted in an ∼50% reduction in the MDSC population compared with cells growing in HNSCC conditioned medium pretreated with the isotype control mAb (Fig. 2B, 2C, Supplemental Fig. 1B).

Sema4D inhibition rescues MDSC-induced T cell suppression
We next sought to ascertain whether HNSCC-derived Sema4D mediated its suppressive effects when myeloid cells were cultured with primary T cells. The induction of HLA-DR 2/low cells by HN6 conditioned media was still observed upon coculture of the CD33 + Cells were first gated for CD33 + then for CD11b + , followed by a third gate for CD11b + HLA-DR 2/low . CD33 cells were isolated using MACS from PBMC separated by centrifugation gradient of peripheral blood. All samples were run in duplicates. CM, conditioned medium. Fig. 2A). To determine whether Sema4D induction of MDSC induces T cell suppression, we constructed Sema4D-shRNA to knock down Sema4D in HN6 cells and accordingly in the conditioned medium (Supplemental Fig. 2B, 2C). Significant recovery of T cell proliferation was observed upon coculture of myeloid and T cells in the HN6 Sema4D-shRNA compared with the control (Ctl)-shRNA medium (Fig. 3A, 3B). T cells grown in HN6 Ctl-shRNA medium were suppressed upon coculture with the myeloid cells, but not when cultured in NM, indicating that the suppression is mediated by the induction of MDSC (Fig. 3B). Notably, the knockdown of Sema4D produced by HN6 resulted in significant recovery of the T cells at different myeloid/T cell ratios (Fig. 3A, 3B). This was paralleled with recovery of IFN-g production (Fig. 3C) as well as inhibition of IL-4 in the supernatant (Fig. 3D).

HN6-secreted Sema4D induces an immunosuppressive T cell phenotype
To investigate which T cell subset is rescued following blockade of Sema4D, a phenotypic analysis of the T cells was performed. We found that inhibition of Sema4D restored effector Th1 cells (CD4 + T-bet + ) and cytotoxic T cells (CD8 + T-bet + ) and significantly decreased Tregs (CD4 + CD25 + FOXP3 + ) at different myeloid/T cell ratios (Fig. 4, Supplemental Fig. 2D-F). These findings indicate that inhibition of Sema4D can rescue effector T cells while suppressing the Treg population in the presence of myeloid cells.

Sema4D enhances GM-CSF plus IL-6-mediated induction of MDSC
To better assess the role of HNSCC-derived Sema4D on myeloid differentiation, we compared Sema4D with another well-described MDSC inducer, GM-CSF (14). GM-CSF has been shown to be produced by several MDSC-inducing human solid tumors, including HNSCC, and to best induce human CD33 + suppressive cells upon combination with IL-6 (14,16,19,20). In fact, we tested the HN6 HNSCC cell line using ELISA and it showed production of both GM-CSF and IL-6 (data not shown). To compare the effect of tumor-secreted Sema4D on the induction of MDSC to GM-CSF-mediated induction, CD33 + cells were cocultured with T cells in HN6 medium with immune depletion of Sema4D versus immune depletion of GM-CSF. Immune depletion of Sema4D resulted in a significant reduction in the MDSC population that was comparable to that observed upon depletion of GM-CSF in HN6 conditioned medium (Fig. 5A, 5B). The reduction in MDSC corresponded with a significant increase in T cell proliferation (Fig. 5C, 5D) and IFN-g production (Fig. 5E). These data strongly support a role for Sema4D produced by HN6 cells in the induction of MDSC and its subsequent T cell suppression. To further study the role of Sema4D in conjunction with other tumor-produced cytokines in MDSC induction, we grew CD33 + cells in NM to which the three recombinant proteins (Sema4D, IL-6, and GM-CSF) were added. The three-cytokine induction mixture of Sema4D plus GM-CSF plus IL-6 resulted in a 30% increase in MDSC induction compared with the GM-CSF plus IL-6 mixture. This indicated that Sema4D acts synergistically with GM-CSF plus IL-6 to increase the repertoire of the MDSC in the HNSCC tumor microenvironment (Fig. 5F, 5G).

Sema4D promotes the production of immunosuppressive mediators by MDSC
To investigate whether HNSCC-associated Sema4D has an effect on MDSC viability, a WST cell proliferation assay and a CFSE proliferation tracking experiment were performed using CD33 + cells grown in conditioned medium from the HN6 cells in which Sema4D has been knocked down using shRNA. The loss of Sema4D resulted in a significant decrease in the proliferation of total CD33 + cells, as well as that of the CD33 + HLA-DR 2/low population compared with conditioned medium from controls (Fig. 6A, 6B). Interestingly, reconstitution of the Sema4D-shRNA HN6 conditioned media with human recombinant Sema4D protein recovered the CD33 + HLA-DR 2/low proliferation as shown by CFSE analysis (Fig. 6C). Taken together, these findings showed that Sema4D produced by HNSCC plays a significant role in the proliferation and viability of MDSC.
The myeloid cells mediate their suppressive effects by various mechanisms. Among these, arginase-1 and NO play a major role in the MDSC suppressive effect. To investigate whether Sema4D produced by HNSCC-HN6 can affect arginase-1 production, CD33 + -enriched myeloid cells were grown in conditioned media from Sema4D-shRNA HN6 and then checked for arginase-1 expression. A significant reduction in arginase-1 production by the CD33 + cells was observed when grown in Sema4D-shRNA HN6 conditioned media (Fig. 6D). To check for the effect of Sema4D knockdown on NO production by myeloid cells, CD33 + cells were grown in HN6 Sema4D-shRNA-versus Ctl-shRNA conditioned media, and then the media were collected after 72 h to check for nitrite concentration. Interestingly, there was a significant reduc-tion in nitrite concentration produced by the myeloid cells in the media upon Sema4D knockdown in HN6 (Fig. 6E).
MDSC can also indirectly induce a suppressive effect through production of other immune-suppressive cytokines such as IL-10 and expansion of Tregs in a TGF-b-dependent pathway (11,18). These cytokines were assessed by ELISA, specifically to measure IL-10 and TGF-b produced by MDSC (43). Interestingly, there was a significant reduction in both IL-10 and TGF-b production by CD33 + cells upon inhibition of Sema4D in HN6 (Fig. 6F, 6G).

Discussion
Understanding the specific mechanisms underlying tumormediated immunosuppression is important for the development of effective antitumor immunotherapy (44). Myeloid progenitor cells have been shown to mediate T cell suppression in patients with HNSCC (8,45,46), and even inducing differentiation of these immature myeloid cells can indirectly activate tumorinfiltrating T cells (47). The link between angiogenesis and immune suppression has been emphasized recently to play a crucial role in tumor development, progression, and metastases. In this study, we investigated Sema4D, a cytokine expressed by several epithelial malignancies and known to induce tumor angiogenesis, for its role in induction of MDSC, in the tumor microenvironment of HNSCC (24,32,33). We show Sema4D produced by HNSCCinduced MDSC and its immune-suppressive cytokines, with subsequent suppression of T cell proliferation and IFN-g production.
These findings are supported by others, who reported that Sema4D acts on immature myeloid cells and dendritic cells by inhibiting their migration and inducing significant increases in the immunesuppressive profile (26,30,31). They are also in good agreement with a recent report showing that Sema4D modulates the inflammatory cytokine milieu in a murine colon and mammary carcinoma model (48).
Several clinical trials showed that advanced cancers can be controlled by immunotherapy. However, these therapies are often compromised by the immune-suppressive tumor microenvironment (49,50). Interestingly, the composition of T cells reported in  human tumor-infiltrating lymphocytes in a cohort of colorectal cancer liver metastases and ovarian cancer indicated that whereas the CD8 + T cells showed signs of proliferation and activation, still there was no tumor rejection (50). This was attributed to several factors, including, but not limited to, presence of MDSC and Tregs. In this context, our data show promising results, in which inhibition of Sema4D in HN6 cells cocultured at several ratios of myeloid/T cells resulted in enhanced IFN-g production (Fig. 3C) and effector CD4 and CD8 T cells (Fig. 4B, 4C), concomitant with a significant decrease in both Tregs (Fig. 4D) and MDSC (Fig. 5A).
Our data indicate that inhibition of Sema4D produced by the tumor cells may recover T cell activation to Ag-nonspecific stimuli and avoid T cell anergy. This would be of importance in adoptive immunotherapy in which the transferred activated T cells are Ag specific, yet the presence of MDSC in the tumor microenvironment might induce anergy (51). This is further supported by others showing that although Sema4D knockout mice display severe impairments in the Ag priming of T cells, interestingly, Sema4Ddeficient T cells still retained the ability to respond normally to either mitogens or anti-CD3 stimulation, and, more importantly, the development of dendritic cells is not defective in cells lacking Sema4D (25,52).
Intriguingly, the opposing cellular responses played by Sema4D in the immune system (53), either as an anti-inflammatory or a proinflammatory cytokine, seem to be dependent on several factors, including, but not restricted to, its concentration, the specific cell type present in the microenvironment, and its stage of maturation upon which Sema4D acts (27,30,37,54). Also, note that in the studies showing the immunosuppressive effect and inhibition of myeloid cell migration, Sema4D recombinant protein was effective at ,10 ng/ml and ranged up to 80 ng/ml (30,31), and the concentrations of Sema4D used to show its proangiogenic effect in tumor microenvironment were in the range of 400 ng/ml (55). In contrast, concentrations used to show increased activation of the immune system, induction of proinflammatory cytokines, and maturation and activation of dendritic cells were in the micromolar range of soluble Sema4D (4-20 mg/ml), suggesting that the Sema4D mechanism of action postulated in other pathological conditions, such as autoimmune-mediated diseases (52,56), may differ from those in the tumor microenvironment, depending on its concentration and presence of tumor-educated cells in the microenvironment.
MDSC have proangiogenic activities that can mediate refractoriness to antiangiogenic therapies (49,63,64). Our data suggest an important role for Sema4D in the continuous positive feedback mechanism between MDSC induction and angiogenesis that is to an extent comparable to GM-CSF and reminiscent of the angiogenic role ascribed to vascular endothelial growth factor in the tumor microenvironment (41). Of particular note is the ability of MDSC to differentiate further into the immune-suppressive M2 macrophage, which has also been reported to produce Sema4D (39). This has significant implications for tumor resistance to antiangiogenic and immune therapy (41,(63)(64)(65)(66) and indicates that Sema4D inhibition might be required to disrupt this positive feedback loop between angiogenesis and immune suppression.
MDSC represent a cell population that contributes to tumor immune suppression. It responds to a variety of proinflammatory and anti-inflammatory stimuli, which drive their recruitment and activation. Understanding how the inflammatory milieu favors immune escape through the accumulation of MDSC could be very useful to improve the efficacy of cancer immunotherapy (67). In addition to the well-described proangiogenic role of Sema4D in the tumor microenvironment, in the present study we describe, to our knowledge for the first time, its immunosuppressive effect through the induction of MDSC. This novel role described for Sema4D may open new avenues for understanding and reversing tumor-associated immunosuppressive mechanisms, with the ultimate goal of increasing T cell activation and enhancing antitumorigenic immune responses in HNSCC, either as a component of immune-based therapy or as a complement to conventional treatment regimens.