HIF-α/MIF and NF-κB/IL-6 Axes Contribute to the Recruitment of CD11b+Gr-1+ Myeloid Cells in Hypoxic Microenvironment of HNSCC^,

Abstract

CD11b+Gr-1+ myeloid cells have gained much attention due to their roles in tumor immunity suppression as well as promotion of angiogenesis, invasion, and metastases. However, the mechanisms by which CD11b+Gr-1+ myeloid cells recruit to the tumor site have not been well clarified. In the present study, we showed that hypoxia could stimulate the migration of CD11b+Gr-1+ myeloid cells through increased production of macrophage migration inhibitory factor (MIF) and interleukin-6 (IL-6) by head and neck squamous cell carcinoma (HNSCC) cells. Hypoxia-inducible factor-1α (HIF-1α)- and HIF-2α-dependent MIF regulated chemotaxis, differentiation, and pro-angiogenic function of CD11b+Gr-1+ myeloid cells through binding to CD74/CXCR2, and CD74/CXCR4 complexes, and then activating p38/mitogen-activated protein kinase (MAPK) and phosphatidylinositide 3-kinases (PI3K)/AKT signaling pathways. Knockdown (KD) of HIF-1α and HIF-2α in HNSCC cells decreased MIF level but failed to inhibit the CD11b+Gr-1+ myeloid cell migration, because HIF-1α/2α KD enhanced nuclear factor κB (NF-κB) activity that increased IL-6 secretion. Simultaneously blocking NF-κB and HIF-1α/HIF-2α had better inhibitory effect on CD11b+Gr-1+ myeloid cell recruitment in the hypoxic zone than individually silencing HIF-1α/2α or NF-κB. In conclusion, the interaction between HIF-α/MIF and NF-κB/IL-6 axes plays an important role in the hypoxia-induced accumulation of CD11b+Gr-1+ myeloid cells and tumor growth in HNSCC.

Introduction

A highly proliferating mass of tumor cells develops faster than the vasculature, and tumor cells meet up with an avascular microenvironment deficient in oxygen, i.e., hypoxia [1]. The oxygen pressure within solid tumors is heterogeneous, ranging from approximately 5% O₂ in well-vascularized regions to anoxia near necrotic regions but is, on average, in the hypoxic range (about 1% O₂) [2]. Such hypoxic zones have been postulated to be an incubator for malignant evolution, where more aggressive cancer cells are selected. Hypoxia induces numerous cellular adaptations during tumor progression, including a switch to anaerobic metabolism, increased genetic instability, promotion of angiogenesis, activation of invasive growth, and preservation of the stemness. In addition, hypoxic tumor cells also show increased resistance to radiotherapy and chemotherapy [1,3]. The major mechanisms mediating adaptive responses to hypoxia are the stabilization and activation of the hypoxia-inducible factors (HIFs), especially HIF-1α and HIF-2α. HIF-1α and HIF-2α trans-activate a set of genes that facilitate tumor growth, angiogenesis, and metastasis [4–6]. Although HIF-1α and HIF-2α have striking similarities in structure, function, and regulation, many lines of evidence suggest that these two HIF-α units play distinct and functionally overlapping roles in the tumor progression [6].

Recently, much attention has been paid to a population of myeloid cells, identified by expressing the cell surface markers CD11b and Gr-1 in mouse [7]. CD11b+Gr-1+ myeloid cells are a large group of myeloid cells consisting of immature macrophages, granulocytes, dendritic cells, and early myeloid progenitors [8]. CD11b+Gr-1+ myeloid cells are also termed myeloid-derived suppressor cells, related to their ability to suppress tumor immunity and to impede cancer immunotherapy [9]. In human, however, the corresponding cells are inadequately characterized because of the lack of uniform markers. In head and neck squamous cell carcinoma (HNSCC), it was reported for the first time that CD34+ myeloid cells have immune suppressor function in patients with HNSCC [10]. Since then, a growing body of evidence suggests that level of circulating CD34+ myeloid cells is correlated with lymph node metastasis and recurrence in patients with HNSCC [11].

Clinical data showed that circulating myeloid-derived suppressor cells correlated with cancer stage and metastatic tumor burden [12]. CD11b+Gr-1+ myeloid cells function by inhibiting CD4+ and CD8+ T cell proliferation, by blocking natural killer cell activation, by limiting dendritic cell maturation, and by polarizing immunity toward a type 2 phenotype [13]. In addition, CD11b+Gr-1+ myeloid cells have been implicated in a whole array of non-immunologic functions, such as the promotion of angiogenesis, tumor cell invasion, and metastases [14–17]. Despite the data defining the infiltration and functions of CD11b+Gr-1+ myeloid cells in tumor progression, the molecular mechanisms for their recruitment have not been well clarified.

More recently, Corzo et al. [8] demonstrated that hypoxia through HIF-1α dramatically alters the functions of CD11b+Gr-1+ myeloid cells in the tumor microenvironment and redirects their differentiation toward tumor-associated macrophages. In addition, HIF-2α modulated the tumor-associated macrophage infiltration in murine hepatocellular and colon carcinoma models through regulating the expression of cytokine receptor macrophage colony-stimulating factor receptor (M-CSFR) and the chemokine receptor CXCR4 [18]. Moreover, Bv8 [19] and stromal-derived factor-1 (SDF-1) [20], which might be related to hypoxia, are suggested to induce peripheral mobilization of CD11b+Gr-1+ myeloid [19,21]. These results provided a mechanistic link between CD11b+Gr-1+ myeloid cell function and hypoxic tumor microenvironment.

Here, we demonstrate that hypoxia may stimulate the accumulation of CD11b+Gr-1+ myeloid cells through increased production of migration inhibitory factor (MIF) in HIF-1α/HIF-2α-dependent ways. When HIF-1α/2α was artificially blocked, nuclear factor κB (NF-κB)/interleukin-6 (IL-6) axis would be increased to compensate for the loss of HIF-α/MIF under hypoxic condition. Simultaneously targeting HIF-α and NF-κB would be an attractive strategy for inhibiting recruitment, differentiation, and pro-angiogenesis of CD11b+Gr-1+ myeloid cells in hypoxic tumor microenvironment of HNSCC.

Materials and Methods

Human and animal studies have been approved by the Institutional Ethics Committee of State Key Laboratory of Oral Diseases, West China Hospital of Stomatology, Sichuan University (Chengdu, China). Detailed protocols are listed in the Supplementary Materials and Methods.

shRNA Cloning, Lentivirus Packaging, and Stable Selection

Lenti-X shRNA Expression System (Clontech, Mountain View, CA) was used for knocking down HIF-1α, HIF-2α, MIF, and IL-6 as described previously [6]. Stable clones were selected using 1.0 µg/ml puromycin.

Cell Culture, Hypoxia Treatment, and siRNA Transfection

Two human HNSCC cell lines Cal-27 and Tca8113 were cultured under 20% O₂ (normoxia) or 1% O₂ (hypoxia) conditions, in a three-gas incubator (Binder, Tuttlington, Germany). siRNAs were transfected by Lipofectamine 2000 (Invitrogen, Carlsbad, CA) at 30 nM.

Xenograft Tumor Model and Treatments

Cells were injected subcutaneously [s.c.; 5 x 10⁶ cells/200 µl phosphate-buffered saline (PBS) per mouse] on the back of nude mice. Tumor size was calculated by πls²/6 (l = long side, s = short side).

Flow Cytometry Analyses

Single-cell suspension was incubated with fluorescence label antibodies and analyzed on a Cytomics FC 500 MPL Flow Cytometer (Beckman Coulter, Brea, CA) using RXP software (Beckman Coulter). Data were analyzed using WinMDI 2.9 software.

Immunofluorescence and Immunohistochemistry

To detect the CD11b+Gr-1+ myeloid cells, frozen sections were incubated with rabbit anti-mouse CD11b (1:50; NOVUS, Littleton, CO) and rat anti-mouse Gr-1 (1:50; eBioscience, San Diego, CA) and with secondary Alexa Fluor 488 goat anti-rabbit IgG (1:500; Invitrogen) and DyLight 594 goat anti-rat IgG (1:500; Jackson, Newmarket, United Kingdom). To detect the blood vessel, sections were stained with primary rat anti-mouse CD31 (1:50; eBioscience) and with secondary DyLight 594 goat anti-rat IgG (Jackson). The formalin-fixed paraffin-embedded sections were stained with rabbit anti-Ki-67 (1:50; Santa Cruz Biotechnology, Dallas, TX), biotinylated goat anti-rabbit IgG, and streptavidin-peroxidase.

CD11b+Gr-1+ Myeloid Cell Sorting

Spleen cells were incubated with anti-mouse fluorescein isothiocyanate (FITC)-CD11b and phycoerythrin (PE)-Gr-1 (eBioscience). CD11b+Gr-1+ and CD11b-Gr-1 cells were sorted with a MoFlo XDP cell sorter (Beckman Coulter).

Chemotaxis

CD11b+Gr-1+ myeloid cells were seeded onto the top chamber of a transwell insert (3 µM; Corning, Corning, NY) and were placed in a 24-well plate that contains cytokines or tumor-conditioned media. Migrated CD11b+Gr-1+ myeloid cells were stained by 0.1% crystal violet, followed by cell lysis, and measurement of optical density at 540 nm. Recombinant human IL-6 (1 µg/ml; R&D Systems, Minneapolis, MN), recombinant human MIF (rhMIF, 1 µg/ml; R&D Systems), recombinant human PAI-1 (1 µg/ml; Millipore, Darmstadt, Germany), and recombinant human intercellular adhesion molecule 1 (ICAM-1, 0.5 µg/ml; eBioscience) were used.

Cytokine Array and ELISA

Cytokines in the supernatant were measured using Human Cytokine Array Kit (R&D Systems). Membranes were scanned by a densitometer (Bio-Rad Laboratories, Hercules, CA). Interested cytokines were further validated by ELISA using Quantikine ELISA Kit (R&D Systems). The intensity was measured at 450 nm in a microplate reader (Thermo, Waltham, MA).

Quantitative Real-Time Reverse Transcription-Polymerase Chain Reaction and Western Blot

Polymerase chain reaction (PCR) amplification was performed using Thunderbird SYBR qPCR mix (TOYOBO, Osaka, Japan) on an ABI PRISM 7300 System (Applied Biosystems, Foster City, CA). Protein were separated on 8% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and sequentially stained with primary antibodies and HRP-labeled secondary antibodies. Bands were scanned using a densitometer (GS-700; Bio-Rad Laboratories).

Luciferase Reporter Assay

Cells were transfected with pNF-κB-luc vector (Clontech) and pRL-TK Vector (Promega) using the Lipofectamine 2000 (Invitrogen) and then were lysed and assayed using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI). Luciferase activity was measured in a luminometer. Results are expressed as relative NF-κB activity compared with controls after normalizing to Renilla luciferase activity.

CXCR Internalization Assays

CD11b+Gr-1+ myeloid cells were treated with buffer, MIF, and MIF plus anti-CD74, respectively, and were washed with acidic glycine buffer. Then, these cells were stained with antibodies to CXCR2 or CXCR4 and were analyzed by flow cytometry.

Statistics

Comparisons of means among groups were analyzed by one-way analysis of variance (ANOVA), and mean tumor volume was analyzed by repeated measures ANOVA. The Bonferroni test was further used to determine the differences between groups. The correlations between CD34+ myeloid cell frequency and the clinical pathologic characteristics of patients were analyzed by Pearson χ² test. All statistical analyses were performed using SPSS 13.0. P < .05 was considered statistically significant.

Results

Expansion of CD11b+Gr-1+ Myeloid Cells in Tumor-Bearing Mice Promoted the Growth and Angiogenesis of Tumor

The expansion of CD11b+Gr-1+ myeloid cells was well described in immunocompetent mice due to their role in suppressing host immunity and impeding cancer immunotherapy [9]. In immunodeficient mice, however, whether CD11b+Gr-1+ myeloid cells would be induced by tumors has not been fully elucidated. We therefore investigated whether CD11b+Gr-1+ myeloid cells would be expanded in the immunodeficient nude mice bearing HNSCC. Two HNSCC cell lines, Cal-27 and Tca8113, were s.c. injected to nude mice. Four weeks later, CD11b+Gr-1+ myeloid cells in spleens and peripheral blood were analyzed by flow cytometry. The percentage of CD11b+Gr-1+ myeloid cells in the spleens of tumor-free mice varied from 3.2% to 5.7% (mean = 4.24%, n = 5), while that in the tumor-bearing mice varied from 9.5% to 12.7% (mean = 11.12%, n = 5). The percentage of CD11b+Gr-1+ myeloid cells in spleens of tumor-bearing mice was significantly increased in comparison with tumor-free mice for both cell lines (P < .001; Figure 1A). Similarly, the percentage of CD11b+Gr-1+ myeloid cells in peripheral blood of tumor-bearing mice (23.5%-28.4%, mean = 25.8%, n = 5) was significantly (P < .001; Figure 1A) increased compared with tumor-free mice (9.5%–12.6%, mean = 10.8%, n = 5). In human, CD34+ immune suppressive myeloid cells are similar to the murine CD11b+Gr-1+ myeloid cells [10,22]. To validate the data in nude mice, we measured the CD34+ immune suppressive myeloid cells in the peripheral blood from 48 patients with HNSCC and 10 healthy volunteers using flow cytometry. The percentage of CD34+ myeloid cells in patients (7.47%-22.5%, mean = 15.54%) was significantly higher (P < .001) than that in healthy volunteers (4.3%-16.3%, mean = 10.21%; Figure 1B). The animal model and human data support the notion that CD11b+Gr-1+ myeloid cells would be expanded by HNSCC.

Figure 1.

Expansion of CD11b+Gr-1+ myeloid cells in tumor-bearing mice promotes tumor growth and angiogenesis. (A) Representative images of flow cytometry analysis for frequencies of CD11b+Gr-1+ myeloid cells in spleens and peripheral blood. Two HNSCC cell lines, Cal-27 and Tca8113, were injected s.c. on the back of nude mice. Four weeks later, the frequency of CD11b+Gr-1+ myeloid cells in spleens and peripheral blood was analyzed by flow cytometry. Tumor-bearing mice had significant higher percentage of CD11b+Gr-1+ myeloid cells in both spleens and peripheral blood (n = 5, P < .001). (B) CD34+ myeloid cells in the peripheral blood of healthy volunteers and patients with HNSCC. Upper panel: Representative images of flow cytometry analysis of CD34+ myeloid cells. Lower panel: Quantitative data. In comparison with healthy volunteers, patients with HNSCC have significantly higher frequency of CD34+ myeloid cell in the peripheral blood. Central solid line: median, box: middle 50% of data, error bar cap lines: 10th and 90th percentiles (*P < .001). (C) Cal-27 cells admixed with CD11b+Gr-1+ cells or CD11b-Gr-1- cells were injected s.c. on the back of nude mice. Tumors admixed with CD11b+Gr-1+ cells had significant larger volume (left panel) and weight (right panel) than that admixed with CD11b-Gr-1- cells (n = 5, *P < .05). (D) Proliferation rate of tumor was measured by immunohistochemical staining of the proliferation marker Ki-67. The Ki-67-positive rate in the tumors admixed with CD11b+Gr-1+ cells was significantly higher than that admixed with CD11b-Gr-1- cells (n = 5, *P < .05, inset: 400x, bar: 50 µm). (E) Blood vessels in the tumor were detected by immunohistochemical staining of CD31. MVD was evaluated by a microscope. Tumors admixed with CD11b+Gr-1+ myeloid cells had significantly higher MVD than that with CD11b-Gr-1- cells (n = 5, *P < .05, bar: 50 µm).

To study the function of CD11b+Gr-1+ myeloid cells in HNSCC-bearing nude mice, we sorted CD11b+Gr-1+ and CD11b-Gr-1- subpopulations from spleens of tumor-bearing mice using FACS. The purification of sorted CD11b+Gr-1+ myeloid cells was 89.27% as analyzed by flow cytometry (Figure W1). Cal-27 cells (4 x 10⁶ cells), admixed with CD11b+Gr-1+ myeloid cells (1 x 10⁶ cells) or CD11b-Gr-1- cells (1 x 10⁶ cells), respectively, were injected s.c. on the back of nude mice. The tumor size was monitored weekly. Tumors admixed with CD11b+Gr-1+ myeloid cells had significant larger volume and weight than that admixed with CD11b-Gr-1- cells (Figure 1C). The proliferation rate of tumor was measured by immunohistochemical staining of the proliferation marker Ki-67. In accordance with the growth curve, Ki-67-positive rate in the tumors admixed with CD11b+Gr-1+ myeloid cells was significantly higher than that admixed with CD11b-Gr-1- cells (P = .005; Figure 1D). Since nude mice lack thymus and unable to produce mature T cells, there should be more mechanisms, other than immune suppression, by which CD11b+Gr-1+ myeloid cells promote xenograft tumor growth in nude mice. Recently, CD11b+Gr-1+ myeloid cells were found to render tumors refractory to anti-vascular endothelial growth factor (VEGF) treatment [23] and contribute to the tumor angiogenesis [15,19,24]. We therefore measured the microvessel density (MVD) of xenograft tumors supplemented with CD11b+Gr-1+ or CD11b-Gr-1- cells. As shown in Figure 1E, the CD11b+Gr-1+ myeloid cell supplementation significantly increased the MVD (P = .004). These results indicate that CD11b+Gr-1+ myeloid cells may contribute to the growth of tumor through stimulating proliferation and angiogenesis.

MIF and IL-6 Mediated Hypoxic Chemotaxis of CD11b+Gr-1+ Myeloid Cells in HNSCC

Since hypoxia is a common feature of solid tumor, we sought to investigate whether hypoxic microenvironment would recruit more CD11b+Gr-1+ myeloid cells to the hypoxic tumor zone. Initially, we sorted CD11b+Gr-1+ myeloid cells from spleens of tumor-bearing mice using FACS and seeded cells in cell culture insets. Supernatants of normoxia- and hypoxia-treated cancer cells were added onto the lower wells and incubated for 6 hours. Compared with tumor-free medium, supernatant of normoxic cancer cells significantly increased the migration of CD11b+Gr-1+ myeloid cells (Figure 2A; Tca8113: P = .003; Cal-27: P = .018). The supernatant of hypoxia-treated cancer cells further increased the number of migrated CD11b+Gr-1+ myeloid cells in comparison with normoxic cancer cells (Figure 2A; Tca8113: P < .001; Cal-27: P = .001). We then measured the CD11b+Gr-1+ myeloid cells in xenograft tumors using immunofluorescence staining. CD11b+Gr-1+ myeloid cells were detected in the xenograft tumor, and interestingly, these cells tended to accumulate around the necrotic zone (Figure 2B). Since the area around the necrotic zone is suggested to be hypoxic [1], these results, combined with in vitro chemotaxis assay, indicate that hypoxic micro-environment may promote the chemotaxis of CD11b+Gr-1+ myeloid cells in HNSCC.

Figure 2.

MIF and IL-6 mediate hypoxic chemotaxis of CD11b+Gr-1+ myeloid cells in HNSCC. (A) Supernatants of Tca8113 and Cal-27 cells were added onto the lower wells, and CD11b+Gr-1+ myeloid cells were seeded on the upper chamber of a transwell assay. Compared with tumor-free medium, supernatant of normoxic (20% O₂) cancer cell significantly increased the chemotaxis of CD11b +Gr-1+ myeloid cells. The supernatant of hypoxia-treated (1% O₂) cancer cells further increased the number of migrated CD11b+Gr-1 + myeloid cells in comparison with normoxic cells. Experiments were carried out in triplicate (*P < .05). (B) Representative images of immunofluorescence staining of Gr-1 (green) and CD11b (red) in a xenograft tumor. Necrotic zone was marked by a dashed line. CD11b+Gr-1+ myeloid cells tend to accumulate around the necrotic zone (bar: 50 µm). (C) Cytokines in the supernatant of normoxic and hypoxic cancer cells were measured using human cytokine array kit. (D) Role of MIF (upper left), IL-6 (upper right), sICAM-1 (lower left), and PAI-1 (lower right) on the chemotaxis of CD11b+Gr-1+ myeloid cells. Recombinant cytokines or supernatants from siRNA-treated cancer cells were added on the lower chamber. IL-6 and MIF increased the chemotaxis of CD11b+Gr-1+ myeloid cells, while sICAM-1 and PAI-1 had nothing to do with CD11b+Gr-1+ myeloid cell chemotaxis. Experiments were carried out in triplicate (*P < .05). (E) Left panel: Representative flow cytometry analysis of CD11b+Gr-1+ myeloid cells in the tumor site. Right panel: Quantitative analysis of the flow cytometry data. MIF and IL-6 KD significantly impaired the recruitment of CD11b+Gr-1+ myeloid cells in the tumor site (n = 5, *P < .05, **P < .001).

Several tumor-secreted cytokines have been suggested to induce CD11b+Gr-1+ myeloid cell migration. We measured the cytokines in supernatant of cancer cells using a human cytokine array kit. MIF, IL-6, sICAM-1, and PAI-1 were detected in the normoxic supernatants and were increased in the hypoxic condition (Figure 2C). Quantitative analysis using ELISA confirmed these results in both HNSCC cell lines (Figure W2). We next investigated whether these hypoxia-induced factors, MIF, IL-6, sICAM-1, and PAI-1, could regulate the chemotaxis of CD11b+Gr-1+ myeloid cells. Recombinant sICAM-1, PAI-1, and siRNAs targeting sICAM-1 and PAI-1 had no significant effect on the chemotaxis of CD11b+Gr-1+ myeloid cells (Figure 2D, lower panel). Normoxic supernatants supplemented with recombinant MIF (1 µg/ml, P < .001) and IL-6 (1 µg/ml, P < .001) significantly increased the migration of CD11b+Gr-1+ myeloid cells by 2.33-fold and 1.65-fold, respectively (Figure 2D, upper panel). In addition, supernatant from MIF (P < .001) and IL-6 (P = .001) knockdown (KD) tumor cells that were cultured under hypoxia had significantly decreased migration of CD11b+Gr-1+ myeloid cells than that from control cells (Figure 2D, upper panel). The silencing efficiency of siRNAs was shown in Figure W3.

To evaluate the in vivo effects of IL-6 and MIF on CD11b+Gr-1+ myeloid cell recruitment, IL-6 and MIF in cancer cells were stably knocked down by lentiviral shRNAs. MIF KD or IL-6 KD cells were s.c. injected into nude mice. Four weeks later, the infiltration of CD11b+Gr-1+ myeloid cells in tumors was significantly reduced by both MIF KD (3.9% ± 1.2%, P < .001) and IL-6KD (5.6% ± 1.3%, P < .001) measured by flow cytometry (Figure 2E). These in vitro and in vivo results suggest that hypoxia-induced IL-6 and MIF may stimulate the chemotaxis of CD11b+Gr-1+ myeloid cells in tumor. IL-6 was reported to regulate CD11b+Gr-1+ myeloid cell migration by previous studies [13,25], while MIF, however, has not been well studied in the CD11b+Gr-1+ myeloid cell migration and function. We therefore further focused on how MIF regulate the chemotaxis of CD11b+Gr-1+ myeloid cells.

MIF/CD74/CXCR2/CXCR4 Complexes Mediated Hypoxia-Induced Migration of CD11b+Gr-1+ Myeloid Cells through p38/MAPK and PI3K/AKT Signaling Pathways

Extracellular MIF can bind to the cell surface receptor CD74 that forms signaling complexes with G-protein-coupled chemokine receptors (CXCR2 and CXCR4) to mediate MIF function [26,27]. We therefore detected the expression of CD74, CXCR2, and CXCR4 expression in CD11b+Gr-1+ myeloid cells sorted from spleens of tumor-bearing mice. Western blot showed that CD74, CXCR2, and CXCR4 were expressed in the CD11b+Gr-1+ myeloid cells (Figure 3A). We further investigated whether MIF regulated CD11b+Gr-1+ myeloid cell chemotaxis through interacting with CD74/CXCR2 and CD74/CXCR4 complexes. CD11b+Gr-1+ myeloid cells were seeded in cell culture insets. rhMIF was added into the lower wells with or without neutralizing anti-mouse CD74, CXCR2, and CXCR4, respectively. The neutralization of CD74, CXCR2, and CXCR4 significantly decreased the migration of CD11b+Gr-1+ myeloid cells (Figure 3B). To further confirm the binding of MIF to CXCR2 and CXCR4, we used an internalization assay that reports specific receptor-ligand interactions. Flow cytometry analysis of surface CXCR2 and CXCR4 on CD11b+Gr-1+ myeloid cells showed that MIF induced internalization of both CXCR2 and CXCR4. The internalization of CXCR2 and CXCR4 was attenuated by anti-CD74 antibody treatment. These results indicate that MIF stimulate the migration of CD11b+Gr-1+ myeloid cells through interacting with CD74/CXCR2 and CD74/CXCR4 complexes (Figure 3C).

Figure 3.

MIF/CD74/CXCR2/CXCR4 complexes contribute to migration of CD11b+Gr-1+ myeloid cells through p38/MAPK and PI3K/AKT signaling pathways. (A) Expression of CD74, CXCR4, and CXCR2 in the CD11b+Gr-1+ myeloid cells was measured by Western blot. Protein samples from CD11b+Gr-1+ myeloid cells of two different mice were loaded. (B) CD11b+Gr-1+ myeloid cells were seeded in cell culture insets. rhMIF was added into the lower wells with or without neutralizing anti-mouse CD74 (upper panel), CXCR4 (middle panel), or CXCR2 (lower panel). The neutralization of CD74, CXCR2, and CXCR4 significantly decreased the migration of CD11b+Gr-1+ myeloid cells. Crtl IgG: Control IgG. Experiments were carried out in triplicate (*P < .05). (C) Internalization assay: Flow cytometry analysis of membrane CXCR2 and CXCR4 on CD11b+Gr-1+ myeloid cells. MIF induced internalization of both CXCR4 (upper panel) and CXCR2 (lower panel). The internalization of CXCR2 and CXCR4 was attenuated by anti-CD74 antibody treatment. (D) The phosphorylation of AKT and p38 in the CD11b+Gr-1+ myeloid cells measured by Western blot. After a 15-minute treatment with rhMIF and hypoxia tumor supernatant (hypoxic SU), the levels of p-AKT and p-38 were significantly increased. In the presence of anti-CD74 neutralizing antibody or ISO-1, the increase in phosphorylation was attenuated. (E) CD11b+Gr-1+ myeloid cells were treated, respectively, with PI3K inhibitor and p38 inhibitor for 1 hour before stimulating with rhMIF. PI3K inhibitor and p38 inhibitor significantly decreased rhMIF-induced migration of CD11b+Gr-1+ myeloid cells by 33% and 40%, respectively. Experiments were carried out in triplicate (*P < .05).

It was reported that MIF modulates chemotaxis of neutrophil and monocytes in inflammation models through activating intracellular signaling pathways, such as p38/mitogen-activated protein kinase (MAPK) and phosphatidylinositide 3-kinases (PI3K)/AKT signaling pathways [28,29]. We therefore investigated whether MIF would activate these pathways in CD11b+Gr-1+ myeloid cells. CD11b+Gr-1+ myeloid cells were stimulated with rhMIF for 15 minutes. After rhMIF and hypoxic tumor supernatant treatment, levels of phosphor-AKT and phosphor-p38 were significantly increased. However, such increase was abrogated by anti-CD74 antibody and MIF antagonist ISO-1 (Figure 3D). To test whether the increased phosphorylation of PI3K/AKT and p38/MAPK contributes to increased CD11b+Gr-1+ myeloid cell chemotaxis, CD11b+Gr-1+ myeloid cells were treated, respectively, with PI3K inhibitor and p38 inhibitor for 1 hour before stimulating with rhMIF. PI3K inhibitor and p38/MAPK inhibitor significantly decreased rhMIF-induced CD11b+Gr-1+ myeloid cell migration by 35% and 46%, respectively (Figure 3E). These results suggest that hypoxia-induced MIF stimulates the chemotaxis of CD11b+Gr-1+ myeloid cells through activating PI3K/AKT and p38/MAPK signaling pathways.

MIF Regulated Differentiation and Angiogenic Potential of CD11b+Gr-1+ Myeloid Cells

We further investigated whether MIF could regulate the differentiation and function of CD11b+Gr-1+ myeloid cells in HNSCC-bearing nude mice. MIF KD cancer cells were s.c. injected to the nude mice. Four weeks later, we examined the granulocytic (CD11b+ Ly6G^highLy6C^low) and monocytic (CD11b+Ly6G^lowLy6C^high) phenotypes in the xenograft tumors using flow cytometry. The percentage of monocytic phenotype in CD11b+ population was significantly decreased in MIF KD tumors in comparison with control tumors (Figures 4a and W4). To study the effect of MIF on CD11b+Gr-1+ myeloid cell differentiation in vitro, CD11b+Gr-1+ myeloid cells were sorted from tumor-bearing nude mice and were cultured for 5 days in the presence of granulocyte-macrophage-colony-stimulating factor (GM-CSF). During the period of culture, cells were treated with recombinant MIF, tumor supernatant, or ISO-1. MIF, normoxic tumor supernatant, and hypoxic tumor supernatant induced differentiation of CD11b+Gr-1+ myeloid cells into monocytic phenotype (Figure 4B). Moreover, hypoxic tumor supernatant-treated CD11b +Gr-1+ myeloid cells had higher percentage of monocytic phenotype than that of normoxic tumor supernatant. In the presence of ISO-1, the induction of monocytic CD11b+Gr-1+ myeloid cells was eliminated (Figure 4B). These in vivo and in vitro data suggest that hypoxia-induced MIF may redirect the differentiation of CD11b+Gr-1+ myeloid cells toward monocytic phenotype.

Figure 4.

MIF regulates differentiation and angiogenic potential of CD11b+Gr-1+ myeloid cells. (A) Granulocytic (CD11b+Ly6G^highLy6C^low) and monocytic (CD11b+Ly6G^lowLy6C^high) phenotypes in the xenograft tumors were analyzed by flow cytometry. The percentage of monocytic phenotype in CD11b+ population was significantly decreased in MIF KD tumors in comparison with control tumors (n = 5, *P < .05). (B) CD11b+Gr-1+ cells were sorted from tumor-bearing nude mice and cultured for 5 days in the presence of GM-CSF. Cells were treated with rhMIF, ISO-1, and normoxic or hypoxic tumor supernatant (Su). Granulocytic and monocytic phenotypes were analyzed by flow cytometry. MIF, normoxic tumor supernatant, and hypoxic tumor supernatant increased the percentage of monocytic cells. ISO-1 significantly inhibited the induction of monocytic phenotype. Experiments were carried out in triplicate (*P < .05). (C) Immunofluorescence staining of CD31 in the xenograft tumor, and MVD was calculated under a microscope. The MVD in MIF KD tumor and ISO-1-treated tumor was significantly decreased (bar: 50 µm, n = 5, *P < .05). (D) Protein levels of VEGF in the supernatant of CD11b+Gr-1+ myeloid cells were measured by ELISA. MIF significantly increased the production of VEGF by CD11b+Gr-1+ cells. ISO-1, PI3K inhibitor, and p38 inhibitor significantly decreased MIF-induced production of VEGF by CD11b+Gr-1+ cells.

As mentioned above that supplement of CD11b+Gr-1+ myeloid cells increased MVD, we investigated the role of MIF on the MVD of the tumor. MIF KD tumors have decreased MVD compared with control tumors (Figure 4C). To study whether MIF could regulate the production of VEGF, CD11b+Gr-1+ myeloid cells were treated with rhMIF, ISO-1, PI3K inhibitor, or p38 inhibitor, respectively. VEGF in the supernatant was measured by ELISA. MIF significantly increased the production of VEGF by CD11b +Gr-1+ myeloid cells. ISO-1 attenuated the induction effects of rhMIF on VEGF expression. Moreover, the VEGF level was decreased by PI3K inhibitor and p38 inhibitor (Figure 4D). These results suggest that MIF may stimulate the angiogenic property of CD11b+Gr-1+ myeloid cells through activating PI3K/AKT and p38/MAPK signaling pathways.

HIF-α/MIF and NF-κB/IL-6 Axes Co-Regulated the Migration of CD11b+Gr-1+ Myeloid Cells under Hypoxia

HIFs are critical factors in regulating hypoxia adaptation and function [1,5]. We therefore cultured HIF-1α KD and HIF-2α KD HNSCC cells under hypoxic condition and collected the supernatant for cytokine array and CD11b+Gr-1+ myeloid cell chemotaxis assay. The knocking down efficiency of HIF-1α and HIF-2α was shown in Figure W5A. MIF level was significantly decreased by knocking down HIF-1α and HIF-2α measured by both cytokine array (Figure 5A) and ELISA (Figure W5B), indicating that hypoxic induction of MIF is regulated by both HIF-1α and HIF-2α in HNSCC cells. To further study the effects of HIF-1α and HIF-2α on the chemotaxis of CD11b+Gr-1+ myeloid cells, supernatant from hypoxia-treated HIF-1α KD and HIF-2α KD HNSCC cells was used for chemotaxis assay. The migration of CD11b+Gr-1+ myeloid cells was not significantly decreased by knocking down HIF-1α or HIF-2α (Figure 5B).

Figure 5.

HIF-α/MIF and NF-κB/IL-6 axes co-regulate the migration of CD11b+Gr-1+ myeloid cells under hypoxia. (A) Cytokines in the supernatant of control shRNA-transfected (ctrl shRNA), HIF-1α KD, and HIF-2α KD tumor cells were measured using a human cytokine array kit. (B) Supernatant from control shRNA-transfected (Ctrl shRNA), HIF-1α KD, and HIF-2α KD tumor cells were applied for chemotaxis assay. Neither HIF-1α nor HIF-2α KD reduced the migration of CD11b+Gr-1+ myeloid cells. Experiments were carried out in triplicate. (C) NF-κB activity was measured using NF-κB luciferase reporter assay. The NF-κB activity was significantly increased by HIF-1α and HIF-2α KD under hypoxia. Experiments were carried out in triplicate (*P < .05). (D) NF-κB activity in tumor cells was inhibited by siRNA targeting p65, PDTC, or BAY-11-7082. Then, supernatant of tumor cells was applied for chemotaxis assay of CD11b+Gr-1+ myeloid cells. CD11b+Gr-1+ myeloid cell migration was significantly inhibited by siRNA targeting p65, PDTC, or BAY-11-7082. Experiments were carried out in triplicate (*P < .05). (E) Combined silencing of NF-κB and HIF-1α/2α decreased the CD11b+Gr-1+ myeloid cell migration more than silencing NF-κB individually. Experiments were carried out in triplicate (*P < .05, **P < .001). (F) The frequency of CD11b+Gr-1+ myeloid cells in the xenograft tumors was measured by flow cytometry. Simultaneously targeting NF-κB and HIF-1α/2α had better inhibitory effect on the recruitment of CD11b+Gr-1+ myeloid cells in the tumor than targeting NF-κB and HIF-1α/2α individually (n = 5, *P < .05, **P < .001).

We showed above that MIF, as a common target of HIF-1α and HIF-2α, contributed to the chemotaxis of CD11b+Gr-1+ myeloid cells (Figure 2D). However, HIF-1α and HIF-2α KD did not significantly affect the migration of CD11b+Gr-1+ myeloid cells. Therefore, there should be some other mechanisms that compensated for the lost of MIF after HIF-1α and HIF-2α KD contribute to the migration of CD11b+Gr-1+ myeloid cells. When looked back into the cytokine array data, we noticed that HIF-1α and HIF-2α KD decreased the level of MIF, while increased the level of IL-6, IL-8, CXCL10, and CCL5 (Figure 5A). The increase of these cytokines was validated by ELISA (Figure W5C).

Interestingly, IL-6, IL-8, CXCL10, and CCL5 are all reported as targets of NF-κB [30]. We confirmed this in HNSCC cell lines by p65 shRNA that significantly inhibited the levels of IL-6, IL-8, CXCL10, and CCL5 in both HNSCC cell lines (Figure W5D). We therefore hypothesized that the increase of IL-6, IL-8, CXCL10, and CCL5 after HIF-1α and HIF-2α KD may be attributed to an increase of NF-κB activity. We measured NF-κB activity using NF-κB luciferase reporter assay. Hypoxia stimulated a 1.95 ± 0.26-fold increase in NF-κB luciferase activity (P = .005; Figure 5C). The NF-κB-dependent transcriptional activity was further significantly increased by HIF-1α (P = .005) and HIF-2α KD (P = .013; Figure 5C). These results support the hypothesis that NF-κB activity was increased by HIF-1α and HIF-2α KD.

We next investigated whether NF-κB contributed to the migration of CD11b+Gr-1+ myeloid cells in vitro. shRNAs targeting p65 and two NF-κB inhibitors, PDTC and BAY-11-7082, were used to inhibit the NF-κB activity in HNSCC cells. Supernatants were harvested and applied for CD11b+Gr-1+ myeloid cell migration assay. The inhibition of NF-κB pathway inhibited the hypoxic stimulation of CD11b+Gr-1+ myeloid cell migration (Figure 5D). Moreover, combined inhibition of NF-κB and HIF-1α/HIF-2α decreased the CD11b+Gr-1+ myeloid cell migration more than inhibition of NF-κB or HIF-1α/HIF-2α individually (Figure 5E).

To validate this in vivo, p65 and HIF-1α/HIF-2α in HNSCC were knocked down individually or simultaneously by lentiviral shRNAs. Xenograft tumors were established by s.c. injecting the stable infected cells to nude mice. Four weeks later, CD11b+Gr-1+ myeloid cells in the tumor were analyzed by flow cytometry. In accordance with the in vitro data, there was a slight decrease of the CD11b+Gr-1+ myeloid cells in the HIF-1α and HIF-2α KD tumors compared with control shRNA tumors but did not quite achieve significance (Figure 5F ; P > .05). The silence of p65 significantly decreased the recruitment of CD11b+Gr-1+ myeloid cells in the tumor (Figure 5F ; P < .05). Moreover, simultaneously knocking down p65 and HIF-1α/HIF-2α had better inhibitory effect on CD11b+Gr-1+ myeloid cell recruitment in the tumor than individually knocking down HIF-1α/HIF-2α or p65 (Figure 5F; P < .001). The volume of xenograft tumor was monitored during the period of experiment. Knocking down HIF-1α, HIF-2α, and NF-κB individually inhibited the growth of tumor in comparison with the control group. The growth of tumor was further decreased by combined KD of HIF-1α/2α and NF-κB (Figure W6). These data indicate that hypoxia HNSCC may induce the recruitment of CD11b+Gr-1+ myeloid cells mainly through increased production of MIF in HIF-1α/2α-dependent ways. When HIF-1α/HIF-2α was artificially inhibited, NF-κB/IL-6 axis would be strengthened to compensate for the loss of HIF-α/MIF axis in regulating the recruitment of CD11b+Gr-1+ myeloid cells (Figure 6).

Figure 6.

Schematic cartoon illustrating how hypoxia regulates migration of CD11b+Gr-1+ myeloid cells. Hypoxia stimulates HNSCC cells to produce more MIF in an HIF-1α/2α-dependent way. MIF regulates CD11b+Gr-1+ myeloid cell migration, differentiation, and proangiogenic function through binding to CD74/CXCR2, and CD74/CXCR4 complexes, and then activating p38/MAPK and PI3K/AKT signaling pathways. NF-κB/IL-6 axis would be an alternative to compensate for the loss of HIF-α/MIF in regulating the migration of CD11b+Gr-1+ myeloid cells when HIF-1α/2α were inhibited.

Discussion

In the present study, we showed that CD11b+Gr-1+ myeloid cells contributed to the growth and angiogenesis of HNSCC xenograft tumor. Hypoxic microenvironment induced HNSCC cells to secrete more MIF and IL-6, both of which could stimulate the recruitment of CD11b+Gr-1+ myeloid cells. HIF-1α/2α-dependent MIF regulates CD11b+Gr-1+ myeloid cell migration, differentiation, and proangiogenic function through binding to CD74/CXCR2, and CD74/CXCR4 complexes, and then activating p38/MAPK and PI3K/AKT signaling pathways. KD of HIF-1α/2α decreased MIF but did not inhibit the CD11b+Gr-1+ myeloid cell migration, because it increased the level of IL-6 and CCL5 in an NF-κB-dependent way. Simultaneously silencing NF-κB and HIF-1α/HIF-2α had better inhibitory effect on CD11b+Gr-1+ myeloid cell recruitment in the tumor than individually knocking down HIF-1α/2α or NF-κB.

In this study, the athymic nude mice were used as host for the establishment of human HNSCC xenograft. The immunodeficient profiling of nude mouse makes it inappropriate for studying the immune suppressive function of CD11b+Gr-1+ myeloid cells. However, on the other side, it would exclude the interference of immune effect when studying the non-immunologic functions, i.e., angiogenesis. In this study, we showed that cancer cells co-injected with CD11b+Gr-1+ myeloid cells had higher MVD than that with CD11b-Gr-1- myeloid cells. In addition, inhibiting MIF decreased the recruitment of CD11b+Gr-1+ myeloid cells accompanied by decreased MVD and VEGF expression level. These results, along with previous reports [15,16], suggest that the infiltration of CD11b-Gr-1- myeloid cells may contribute to the angiogenesis of HNSCC.

Hitherto, there was no direct evidence to suggest that hypoxic microenvironment could induce migration of CD11b+Gr-1+ myeloid cells in HNSCC. In the present study, we found that CD11b +Gr-1+ myeloid cells tend to accumulate in the hypoxic zones in HNSCC xenograft tumors and that supernatant of hypoxia-treated HNSCC cells increased the CD11b+Gr-1+ myeloid cell migration in vitro. These results provided a direct link between hypoxia and migration of CD11b+Gr-1+ myeloid cells. We further measured the cytokines in supernatant of HNSCC cells and found that two cytokines, MIF and IL-6 secreted by HNSCC cells, were increased by hypoxia and contributed to the hypoxia-induced CD11b+Gr-1+ myeloid cell migration. SDF-1 was also found to induce CD11b+Gr-1+ myeloid cell migration in human and murine mammary tumors [14,20]. However, our data showed that SDF-1 is not an important cytokine that mediated the chemotaxis of CD11b+Gr-1+ myeloid cells by HNSCC cells. The reason may be that different types of cancer express different cytokines and respond differently to hypoxia.

MIF was originally identified as a T cell-derived factor responsible for the inhibition of macrophage migration in the mid-1960s. Extensive studies since have revealed its pleiotropic roles in immune and inflammatory responses [31,32]. Recently, a growing amount of evidence supports the notion that MIF exhibits pro-neoplastic activities, including tumor suppressor down-regulation, cyclooxygenase-2 (COX-2) and prostaglandin E₂ (PGE₂) up-regulation, potent induction of angiogenesis and enhanced tumor proliferation, growth, and invasiveness [32–35]. One of the most important among these is the modulation of hypoxic adaptation within the tumor microenvironment by interacting with HIF-1α [36]. Here, we showed that MIF level was increased by hypoxia in HNSCC cells, and this increase of MIF was abrogated by KD of either HIF-1α or HIF-2α, indicating that HIF-1α and HIF-2α play similar roles in the regulation of MIF. Intriguingly, we observed that HNSCC-derived MIF could bind to membrane receptor CD74, CXCR2, and CXCR4 on the CD11b +Gr-1+ myeloid cells and then activate p38/MAPK and PI3K/AKT signaling pathways. Through these effects, MIF may not only promote the chemotaxis of CD11b+Gr-1+ myeloid cells but also regulate the differentiation and pro-angiogenic function of CD11b+Gr-1+ myeloid cells. These results were consistent with recently published paper in which Simpson et al. [37] demonstrated that MIF increased the prevalence of CD11b+ myeloid cells within the tumor, leading to promotion of 4T1 breast cancer growth and metastasis. These results revealed novel functions of MIF on the regulation of CD11b+Gr-1+ myeloid cells, suggesting that MIF should be an attractive target for therapeutic strategy in human HNSCC.

As a target gene of HIF-1α/2α, MIF stimulate the migration and accumulation of CD11b+Gr-1+ myeloid cells. Intriguingly, HIF-1α/2α KD did not significantly inhibit the CD11b+Gr-1+ myeloid cell migration. It is possible that some other compensatory mechanisms exist after HIF-1α/2α KD in cancer cells. Our cytokine array showed that HIF-1α/2α KD increased the expression of IL-6, IL-8, CXCL10, and CCL5. Not incidentally, a previous study also reported that HIF-1α KD resulted in increased levels of IL-6, IL-8, and MCP-1 in human umbilical vein endothelial cells [38]. Since the four cytokines were reported as NF-κB target genes [30], we measured NF-κB activity using NF-κB luciferase reporter assay, which revealed a significant increase of NF-κB activity after HIF1α/2α KD. Given the importance of IL-6 in the hypoxia-induced migration of CD11b+Gr-1+ myeloid cells as our data has shown, one explanation is that the increase of NF-κB activity further induced IL-6 production, which mainly compensated for the loss of MIF after HIF-1α/2α KD to regulate the migration of CD11b+Gr-1+ myeloid cells (Figure 6).

To date, the direct association between NF-κB and HIF-1α has been controversial. In neutrophil, NF-κB was demonstrated a hypoxia-regulated and HIF-1α-dependent target [39]. However, NF-κB was suggested as a critical transcriptional activator of HIF-1α in cultured cells and in the liver and brain of hypoxic animals [40]. Contrarily, NF-κB was also reported to block HIF-1α activity by direct competition of p300 [41]. In addition, the increased HIF-1α stabilization and transcriptional activity may be concurrent with decreased NF-κB activity [42]. In this study, we show that HIF-1α/2α KD increased NF-κB activity under hypoxia. However, whether HIF-1/2α KD resulted in a shift of p300 from HIF-1α/2α to NF-κB has not yet been well clarified and needs further investigation.

Although HIF-1α and HIF-2α have striking similarities in structure, function, and regulation, many lines of evidence suggest that the roles of HIF-1α and HIF-2α in response to hypoxia vary among cell types, and their biologic functions may be different [6,43]. In this paper, however, we show that knocking down HIF-1α and HIF-2α in HNSCC cells had similar effects on MIF level, NF-κB activity, and migration of CD11b+Gr-1+ myeloid cells. Our results revealed for the first time that HIF-2α plays equal important roles as HIF-1α in the regulation of CD11b+Gr-1+ myeloid cell migration and function.

We have shown that HNSCC cells admixed with CD11b+Gr-1+ myeloid cells had significantly higher volume and weight than that with CD11b-Gr-1- cells. We further showed that combined KD of HIF-1α/2α and NF-κB remarkably inhibited the recruitment of CD11b+Gr-1+ myeloid cells in the xenograft tumor mass. This process is accompanied by reduced tumor growth. These results suggest that the inhibition of tumor growth by HIF-1α/2α and NF-κBKD might be, at least partially, attributed to the decreased recruitment of CD11b+Gr-1+ myeloid cells.

The majority of previous studies have been focusing on the immune suppressive role of CD11b+Gr-1+ cells on growth of tumor [37,44] Here, we investigated the contribution of CD11b+Gr-1+ cells to HNSCC growth in immunodeficient nude mice. Our results suggest that CD11b+Gr-1+ cells could promote the tumor growth due to mechanisms other than antitumor immunity suppression. These results are consistent with a recently published paper, in which Park et al. [45] demonstrated that recruitment of CD11b+Gr-1+ cells to tumor tissue contributed to prostate cancer angiogenesis and growth in an immunodeficient mouse model.

In conclusion, we demonstrated that, in hypoxic microenvironment of HNSCC, HIF-α/MIF axis regulates the migration, differentiation, and pro-angiogenic function of CD11b+Gr-1+ myeloid cells that further contribute to the growth of tumor. When HIF-1α/2α were artificially inhibited, NF-κB/IL-6 axis would be increased to compensate for the loss of HIF-α/MIF. As such, therapeutic outcome in terms of inhibiting CD11b+Gr-1+ myeloid cells may benefit from combined targeting HIF-α/MIF and NF-κB/IL-6 axes.

Supplementary Materials and Methods

shRNA Cloning and Lentivirus Packaging

Lenti-X shRNA Expression System (Clontech) was used for the shRNA-mediated inhibition of HIF-1α, HIF-2α, MIF, and IL-6 according to the manufacturer's instruction. Short pairs of sense and antisense DNA oligo encoding a sense-loop-antisense sequence to target genes were synthesized by TAKARA (Dalian, China). The shRNA target sequences for HIF-1α [1], HIF-2α [1], MIF [2], IL-6 [3], p65 [4], and negative control [1] have been described previously. The complementary DNA oligos were annealed and cloned to the BamHI/EcoRI-digested pLVX-shRNA1 vectors (Clontech). The recombinant vectors were purified and cotransfected with Lenti-X HT Packaging Mix (Clontech) into HEK 293T packaging cells. The virus-containing cell culture supernatants were collected 48 hours after transfection, passed through a 0.45-µm filter, and stored at -80°C. Tumor cells were transduced with recombinant lentivirus with 5 µg/ml polybrene, and stable clones were selected using 1.0 µg/ml puromycin 24 hours after infection.

Cell Culture, Hypoxia Treatment, and Lenti-X shRNA Transduction

Two human HNSCC cell lines Tca8113 and Cal-27 were obtained from the State Key Laboratory of Oral Disease, Sichuan University. Cells were cultured in RPMI 1640 (Gibco, Grand Island, NY) supplemented with 10%FBS (Hyclone, Logan, UT), 2 mM l-glutamine, 25 mM 2-(4-(2-hydroxyethyl)-1-piperazinyl) ethanesulfonic acid (Hepes), and 100 units/ml penicillin and 100 µg/ml streptomycin in a humidified 5% CO₂ atmosphere. The cells were cultured under 20% O₂ (normoxia) or 1% O₂ (hypoxia) conditions, balanced with N2 in a three-gas incubator (Binder).

For siRNA transfection, cells were passaged and reseeded in six-well plates at a density of 2 x 10⁵ cells per well. Twenty-four hours later, siRNA and Lipofectamine 2000 (Invitrogen) were diluted by Opti-MEM I (Gibco), mixed, and then added to each well. The final concentration of siRNA was 30 nM. The siRNA sequences for MIF [2], IL-6 [3], PAI-1 [5], and ICAM-1 [6] used in this study have been described previously.

Xenograft Tumor Model

The nude mice (6 weeks of age) were obtained from the Laboratory Animal Center of Sichuan University. Tumor cells were injected s.c. (5 x 10⁶ cells/200 µl PBS per mouse) on the back of nude mice. The tumor size was monitored by measuring diameters using Vernier caliper weekly and was calculated as πls²/6, where l = long side and s = short side as described previously [7].

Flow Cytometric Analyses

Peripheral blood, spleens, and tumors were harvested, and single-cell suspensions were prepared. Red cells were removed using ammonium chloride lysis buffer. Freshly prepared cells (1 x 10⁶) were incubated at 4°C for 20 minutes with rabbit anti-mouse CD11b conjugated to FITC, rat anti-mouse Gr-1 conjugated to PE, or rat anti-human CD34-PE. Cells were washed with PBS, resuspended in 500 µl of 1% paraformaldehyde, and analyzed on a Cytomics FC 500 MPL Flow Cytometer (Beckman Coulter) using RXP software (Beckman Coulter). Acquired data were analyzed using WinMDI 2.9 software.

To analyze the granulocytic (CD11b+Ly6G^highLy6C^low) and monocytic (CD11b+Ly6G^lowLy6C^high) phenotypes, CD11b+Gr-1+ myeloid cells were stained with FITC-CD11b, PE-Ly6G, and PerCp/Cy5.5-Ly6C. Fluorescence-labeled antibodies were purchased from eBioscience. To analyze the CD34+ myeloid cell in the peripheral blood from patients and healthy volunteers, 1 ml of peripheral blood was stained with FITC mouse anti-human CD34 (BD Biosciences, San Jose, CA) at 4°C for 15 minutes. Red cells were lysed by ammonium chloride lysis buffer. To analyze the expression of CXCR2 and CXCR4 expression, CD11b+Gr-1+ myeloid cells were stained with anti-mouse PerCp/Cy5.5-CXCR2 (BioLegend, San Diego, CA), anti-mouse PE-CXCR4 (eBioscience), or isotype control IgG. Samples were then analyzed on a Cytomics FC 500 MPL Flow Cytometer.

Immunofluorescence and Immunohistochemistry

Xenograft tumors were cut into two parts, one was placed in liquid nitrogen and then frozen at -80°C, and the other was fixed by formalin and then embedded by paraffin. Frozen tissues were embedded in optimal cutting temperature (Sakura Finetek, Tokyo, Japan), frozen at -80°C, and cut (6 mm) in a cryostat (Leica Microsystem, Bannockburn, IL). Frozen sections were dried at 20°C for 1 hour, fixed in acetone at 4°C for 20 minutes, air-dried for 10 minutes at 20°C, and incubated in 20% goat serum at 37°C for 30 minutes to block nonspecific binding. To detect the CD11b+Gr-1+ myeloid cells, slides were incubated overnight with rabbit anti-mouse CD11b (1:50; NOVUS) and rat anti-mouse Gr-1 (1:50; eBioscience). The second day, slides were washed and incubated for 1 hour with Alexa Fluor 488 goat anti-rabbit IgG (1:500; Invitrogen) and DyLight 594 goat anti-rat IgG (1:500; Jackson). To detect the blood vessel, sections were stained with primary rat anti-mouse CD31 (1:50; eBioscience) and secondary DyLight 594 goat anti-rat IgG (Jackson). After washing, coverslips were counterstained with 4′, 6-diamidino-2-phenylindole (1 µg/µl) and examined by a Leica DMI6000 B fluorescence microscope (Leica Microsystem) using Leica FW4000 V1.0 software. MVD was assessed by a microscope as described previously [8].

The formalin-fixed paraffin-embedded sections were deparaffinized, rehydrated, endogenous peroxidase blocked, and antigen retrieved. Then, slides were incubated at 37°C with rabbit antihuman Ki-67 (1:500; Santa Cruz Biotechnology) for 2 hours. Slices were then incubated with biotinylated goat anti-rabbit IgG or goat anti-mouse IgG for 1 hour and streptavidin-peroxidase for 30 minutes. The 0.02% diaminobenzidine tetrahydrochloride was used as a chromogen, and the slides were counterstained with hematoxylin. The percentage of positive cells was estimated using an image analysis system (Leica, Wetzlar, Germany).

CD11b+Gr-1+ Myeloid Cell Sorting

Spleens from mice were harvested, and single-cell suspensions were prepared. Red cells were removed using ammonium chloride lysis buffer. Cells were resuspended in PBS and incubated with FITC-conjugated anti-mouse CD11b and PE-conjugated anti-mouse Gr-1 (eBioscience). CD11b+Gr-1+ cells were sorted with a MoFlo XDP cell sorter (Beckman Coulter) and cultured in RPMI 1640 with 10% FBS and 20 ng/ml GM-CSF.

Chemotaxis Assay

CD11b+Gr-1+ myeloid cells (5 x 10⁴ cells per well) were seeded onto the top chamber of a transwell insert (3 µM; Corning) and were placed in a 24-well plate that contains cytokines or tumor-conditioned media. The plates were incubated for 6 hours at 37°C with 5% CO₂. Migrated CD11b+Gr-1+ myeloid cells were stained by 0.1% crystal violet, followed by cell lysis, and measurement of optical density at 540 nm.

To investigate the role of interested cytokines on the in vitro migration of CD11b+Gr-1+ myeloid cells, recombinant human IL-6 (1 µg/ml; R&D Systems), rhMIF (1 µg/ml; R&D Systems), recombinant human PAI-1 (1 µg/ml; Millipore, Billerica, MA), and recombinant human ICAM-1 (0.5 µg/ml; eBioscience) were added to the lower well.

Cytokine Array and ELISA

Tumor cells were assigned to the following four groups: wild-type cells cultured at 20% O₂, wild-type cells cultured at 1% O₂, HIF-1α KD cells cultured at 1% O₂, and HIF-2α KD cells cultured at 1% O₂. Cytokines in the supernatant were measured using Proteome Profiler Human Cytokine Array Kit, Panel A (ARY005; R&D Systems) according to the manufacturer's protocol. Membranes were scanned using a densitometer (GS-700; Bio-Rad Laboratories), and quantification was performed using Quantity One 4.4.0 software. Interested cytokines were further validated by ELISA using Quantikine ELISA Kit (R&D Systems) following the manufacturer's instructions. The intensity of color was measured at 450 nm in a microplate reader (Thermo). The concentration of protein in culture media was determined in triplicate wells and was normalized to standard curves generated for each set of samples assayed.

Quantitative Real-Time RT-PCR and Western Blot

Total RNA was isolated with TRIzol reagent (Invitrogen) and reverse-transcribed using a RevertAid First-Strand cDNA Synthesis Kit (Fermentas, Vilnius, Lithuania) with random hexamer primer. PCR amplification of the cDNA template was performed using Thunderbird SYBR qPCR mix (TOYOBO) on an ABI PRISM 7300 Sequence Detection System (Applied Biosystems). Reactions were run in triplicate, and results were averaged. Each value was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as the housekeeping gene to control. The sequences of PCR primers for HIF-1α [8], HIF-2α [8], VEGF [8], PAI-1 [8], p65 [9], IL-6 [10], MIF [11], and ICAM-1 [12] have been described previously.

Total proteins were isolated with a total protein extraction kit (Keygen, Nanjing, China). Thirty micrograms of proteins was separated on 8% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred electrophoretically to polyvinylidene difluoride (PVDF) membranes (Millipore, Boston, MA). Membranes were incubated for 2 hours, respectively, with mouse anti-HIF-1α (Abcam, Cambridge, MA), rabbit anti-HIF-2α (Abcam), mouse anti-β-actin (Santa Cruz Biotechnology), mouse anti-CD74 (M-B741, Santa Cruz Biotechnology), rabbit anti-CXCR2 (H-100, Santa Cruz Biotechnology), rabbit anti-CXCR4 (Abcam), rabbit anti-phospho-AKT (Ser⁴⁷³; Cell Signaling Technology, Danvers, MA), rabbit anti-AKT (Cell Signaling Technology), rabbit anti-phospho-p38 (Thr¹⁸⁰/Tyr¹⁸²; Cell Signaling Technology), and rabbit anti-p38 (Cell Signaling Technology). HRP-conjugated anti-mouse or anti-rabbit IgG was used as a secondary antibody (1:5000). Bands were scanned using a densitometer (GS-700; Bio-Rad Laboratories), and quantification was performed using Quantity One 4.4.0 software.

Luciferase Reporter Assay

Tumor cells (1 x 10⁵/ml) were seeded into each well of a 12-well plate, incubating at 37°C for 24 hours. Then, cells were transiently transfected with 1 µg of pNF-κB-luc vector (Clontech) and 1 µgof pRL-TK Vector (Promega) using the Lipofectamine 2000 (Invitrogen). Six hours later, transfection media were changed, and cells were cultured under either normoxia (21% O₂) or hypoxia (1% O₂) conditions for an additional 24 hours. Cells were lysed in passive lysis buffer (Promega) for 15 minutes, and 20 µl of lysates was assayed using the Dual-Luciferase Reporter Assay System (Promega). Luciferase activity was measured in a luminometer. Experiments were carried out in triplicate, and results are expressed as relative NF-κB activity compared with controls after normalizing to Renilla luciferase activity.

Statistics

Comparisons of means among groups were analyzed by one-way ANOVA, and mean tumor volume was analyzed by repeated measures ANOVA. The Bonferroni test was further used to determine the differences between groups. The correlations between CD34+ myeloid cell frequency and the clinical-pathologic characteristics of patients were analyzed by Pearson χ² test. All statistical analyses were done using the SPSS 13.0. P < .05 was considered statistically significant.