Metformin enhances gefitinib efficacy by interfering with interactions between tumor-associated macrophages and head and neck squamous cell carcinoma cells

Abstract

Background

Tumor-associated macrophages (TAMs) play an important role in drug resistance in many tumors, including head and neck squamous cell carcinoma (HNSCC). However, how TAMs interact with HNSCC cells to induce drug resistance, especially under hypoxic conditions, is unclear. In this study, we investigated the mechanism of TAM-induced gefitinib resistance in HNSCC cells and sought for novel therapeutic strategies.

Methods

The effects of hypoxia-treated HNSCC cells on the migration and polarization of macrophages were analyzed. Recombinant cytokine proteins and neutralizing antibodies were used as controls. In addition, we assessed the cytotoxic effects of gefitinib on HNSCC cells treated with M2-type macrophage conditioned medium, and carried out a cytokine antibody array analysis, thereby revealing the key factor CCL15. The relationship between serum CCL15 expression levels and prognosis in HNSCC patients was analyzed. In addition, we performed bioinformatic analyses to pursue the mechanisms of CCL15-induced gefitinib resistance. Finally, metformin was used to evaluate the sensitizing effects of gefitinib treatment on HNSCC cells in vitro and in vivo.

Results

We found that HNSCC cells recruited macrophages by secreting VEGF and polarized the macrophages to the M2 phenotype through IL-6. Conversely, we found that M2-type TAMs promoted HNSCC cell resistance to gefitinib through paracrine CCL15 signaling. The serum CCL15 levels in HNSCC patients showed a significant correlation with patient prognosis. Furthermore, we found that M2-type TAMs could suppress the sensitivity of HNSCC cells to gefitinib through the CCL15-CCR1-NF-κB pathway. In addition, we found that metformin not only inhibited CCL15 expression in M2-type TAMs enhanced by hypoxia, but also suppressed CCR1 surface expression in HNSCC cells. Encouragingly, we found that metformin sensitized HNSCC cells to gefitinib treatment in vitro and in vivo.

Conclusions

Based on our data we conclude that we have identified a novel interaction between M2-type TAMs and HNSCC cells that contributes to gefitinib resistance. We also found that metformin inhibited the cross-talk between macrophages and tumor cells, thereby eliciting therapeutic effects both in vitro and in vivo.

Electronic supplementary material

The online version of this article (10.1007/s13402-019-00446-y) contains supplementary material, which is available to authorized users.

Introduction

Currently, the treatment options for head and neck squamous cell carcinoma (HNSCC) are diverse, the most common methods being surgery, chemotherapy and radiotherapy. Molecular targeted therapy represents another effective means of supplemental or even replacement therapy for patients with advanced, inoperable HNSCC and for patients in poor physical condition who cannot undergo chemo-radiotherapy [1–4]. Vascular endothelial growth factor (VEGF)-targeted therapy is the most commonly used molecular targeted therapy. While it inhibits angiogenesis, it may promote the production of pro-angiogenic transcription factors, thereby accelerating tumor growth and recurrence after a short period of effectiveness [5–7], which limits its clinical use. In addition to VEGF, epidermal growth factor receptor (EGFR) is a key factor that promotes tumor growth by binding to EGF and other cytokines and activating downstream signaling pathways [8]. Small molecule inhibitors (e.g., gefitinib) and antibodies (e.g., cetuximab) are currently used in EGFR-targeted therapy [9, 10], but their therapeutic efficacy is limited due to drug resistance brought about by epithelial-mesenchymal transition (EMT), mutation of the KRAS gene, and activation of alternative signaling pathways [11].

It is well known that tumors are heterogeneous in nature and that different types of tumors and different tumor sites exhibit different mechanisms of resistance to anticancer drugs. Previously identified mechanisms of EGFR-targeted therapy resistance in HNSCC include co-expression of human epidermal growth factor receptor (HER)2 and HER3, and increased EMT and cyclin D1 expression [12]. A more recent understanding of tumor development includes a role of the tumor microenvironment - the surrounding physical and chemical environment, such as low oxygen and low pH - in addition to the tumor cells themselves. These factors have all been described as playing important roles in the mechanisms of anti-HNSCC drug resistance [13, 14]. Our previous studies have, for example, shown that hypoxia can activate interactions between inflammation and hypoxia and reduce the therapeutic effect of cisplatin in HNSCC [15, 16]. Our recent work also showed that hypoxia alone promotes the therapeutic effect of gefitinib on HNSCC [17]. In addition to the tumor cells themselves, a very important factor that should be noted is the large number of stromal cells that may be present, such as bone marrow stem cells, immune cells, fibroblasts, and macrophages, surrounding the tumor cells. It has also been reported that tumor-associated fibroblasts can promote HNSCC tolerance to cetuximab treatment through matrix metalloproteinases [18] and that tumor-associated macrophages (TAMs) can be associated with chemotherapy tolerance of various tumors [19–21]. With a deeper understanding of the tumor microenvironment, it has become clear that stromal cells can also be affected by tumor cell-associated chemical and physical factors [22]. It is generally understood that hypoxia is one of the most important characteristics of the HNSCC microenvironment [23]. Thus, it appears appropriate to study the effects of hypoxia on stromal cells, the subsequent tumor progression, and the resistance of tumors to related treatments. Our laboratory recently studied HNSCC-associated macrophages and found that these macrophages themselves are affected by hypoxic microenvironments. However, the underlying molecular mechanism is still unclear, although it is known that macrophages can under the influence of hypoxia promote tumor cell resistance to gefitinib. Here, we aimed to investigate a simple therapeutic regimen for HNSCC through exploring the relevant underlying mechanisms, thereby providing clinically translatable guidance. We found that HNSCC cells can recruit and polarize macrophages under the influence of a hypoxic microenvironment. Subsequently, we found that these macrophages secreted C-C motif chemokine ligand 15 (CCL15) through a hypoxia-inducible factor (HIF)-2α-dependent pathway, after which CCL15 reacted with C-C motif chemokine receptor 1 (CCR1) on the tumor cell surface to activate NF-κB signaling to promote gefitinib resistance in the tumor cells. More significantly, similar to some of our previous studies, we found that metformin showed diverse anticancer effects not only by inhibiting CCL15 secretion by M2-type TAMs, but also by suppressing CCR1 expression in HNSCC cells.

Methods

Cells and culture conditions

The human cell lines CAL27 and THP-1 were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China), whereas cell line JHU011 was obtained from Johns Hopkins University (Baltimore, USA) and cell lines SCC7, FaDu and SCC9 were purchased from the American Type Culture Collection (Manassas, VA, USA). The cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM; CAL27, JHU011, and SCC7), minimal essential medium (MEM; FaDu), DMEM/F-12 supplemented with 400 ng/ml hydrocortisone (Selleck, Houston, USA; SCC9) or RPMI-1640 medium (THP-1) supplemented with 100 μg/ml streptomycin, 100 U/ml penicillin (KeyGen, Nanjing, China) and 10% fetal bovine serum (FBS, Gibco, NY, USA) in a humidified incubator at 5% CO₂/20% O₂ and 37 °C. Confluent cells were trypsinized with 0.05% trypsin containing 0.02% ethylenediaminetetraacetic acid (EDTA) (Gibco, NY, USA).

Collection of cell culture supernatants

To collect the supernatant of M2-type macrophages, we first polarized THP-1 cells. Briefly, the cells were stimulated with 10 ng/ml phorbol 12-myristate 13-acetate (PMA; Sigma-Aldrich, MO, USA) for 72 h. Non-adherent cells were removed, and the remaining adherent cells were phenotypically M0-type macrophages. Following stimulation with 20 ng/ml IL-4 (PeproTech, NJ, USA) and 100 ng/ml IL-13 (PeproTech, NJ, USA) for 24 h, M2-type macrophages were obtained. After subsequent culturing under normoxic or hypoxic conditions for 48 h, the supernatants of the M2-type macrophages and CAL27 and JHU011 cells were collected and stored at −80 °C until use following centrifugation for 5 min at 1000 rpm/min.

Transwell migration assay

A Transwell migration assay was conducted using Costar Transwell 24-well plates (Corning, NY, USA) to assess cell migration. In brief, 1 × 10⁵ M0-type macrophages were seeded into the upper chamber with 0.2 ml serum-free medium. A 1:1 mixture of complete medium and culture supernatant from CAL27 or JHU011 cells, or medium with 10 ng/ml recombinant protein (PeproTech, NJ, USA), or 50 ng/ml anti-VEGF neutralizing antibody (BioVision, CA, USA) were added to the lower chamber. After 24 h at 37 °C with 5% CO₂, the cells remaining on the upper surface were completely removed. The cells that migrated through the membranes were fixed with 4% paraformaldehyde (Boster, Wuhan, China) for 20 min, stained with crystal violet (Beyotime, Shanghai, China) for 15 min, and counted under a microscope.

RNA extraction and RT-qPCR analysis

Total RNA was extracted using TRIzol reagent (Invitrogen, CA, USA) after which reverse transcription was performed using a PrimeScriptTM RT reagent kit with gDNA Eraser (Takara, Dalian, China) according to the manufacturer’s instructions. Quantitative RT-PCR (RT-qPCR) was performed on an Applied Biosystems ViiA™7 instrument using a ChamQ Universal SYBR qPCR Master Mix (Vazyme, Nanjing, China) according to the manufacturer’s protocol. β-actin was used as reference gene, and the results are expressed as the relative expression ratio of the target gene to the reference gene. Data were analyzed using the 2^−ΔΔCt method. For RT-qPCR, amplification of CCL15 and β-actin from the cDNA template was carried out with 2 × Taq Master Mix (Vazyme, Nanjing, China) according to the manufacturer’s instructions. A 1.5% agarose gel (Biowest, Nuaillé, France) was prepared after which 10 μl of each PCR product was electrophoresed. A JS-780 Gel Imaging System (Chincan, Zhejiang, China) was used to analyze the bands. The sequences for the primers used in this study are listed in Supplemental file 1: Table S1.

Immunofluorescence analysis

M0-type macrophages were cultured in 28.2 mm Glass Bottom Culture Dishes (Nest, Wuxi, China). Subsequently, the cells were treated according the indicated conditions, including 20 ng/ml IL-4/100 ng/ml IL-13, a 1:1 mixture of complete medium and culture supernatants from CAL27 or JHU011 cells, or medium supplemented with 10 ng/ml recombinant protein (PeproTech, NJ, USA), or 50 ng/ml anti-IL-6 neutralizing antibody (BioVision, CA, USA). After 24 h, the cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100. After the resulting cells were blocked in 3% BSA (Biosharp, Hefei, China) for 30 min, they were incubated with a rabbit anti-CD206 antibody (1:200; Arigo, Hsinchu City, China), rabbit anti-CD68 antibody (1:100; Abcam, Cambridge, UK), rabbit anti-NF-κB p65 antibody (1:1000; Abcam, Cambridge, UK), or rabbit anti-phosphoNF-κB antibody p65 (1:200; Abcam, Cambridge, UK) overnight at 4 °C, washed and then incubated for 30 min with a DyLight 594-conjugated AffiniPure Donkey Anti-Rabbit antibody (1:400; EarthOx, CA, USA) at room temperature in the dark. Finally, the cells were co-stained with 4′,6-diamidino-2-phenylindole (DAPI; Beyotime, Shanghai, China) to detect nuclei and examined and imaged using a laser scanning confocal microscope (Nikon, Tokyo, Japan).

Western blot assay

Following the indicated treatments, cells were lysed on ice using RIPA buffer (Beyotime, Shanghai, China) supplemented with a protease inhibitor cocktail (Bimake, TX, USA) and phenylmethylsulfonyl fluoride (PMSF) (Beyotime, Shanghai, China). The resulting supernatants were collected by centrifugation at 12,000×g at 4 °C for 25 min, after which total protein concentrations were determined using a bicinchoninic acid (BCA) protein assay kit (KeyGen, Nanjing, China). Next, protein samples were mixed with 5× loading buffer (GenScript, Nanjing, China) and heated at 95 °C for 10 min, after which equal amounts of protein were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). The proteins were subsequently transferred to 0.45 mm nitrocellulose (NC) membranes (EMD Millipore, MA, USA), and blocked with 5% fat-free milk in TBST buffer (150 mM NaCl, 50 mM Tris-HCl, and 0.5% Tween 20, pH 7.6) at room temperature for 2 h. Next, the membranes were incubated with primary antibodies at 4 °C overnight and then with horseradish peroxide-conjugated secondary antibodies at room temperature for 2 h, followed by exposure to an ECL reagent (EMD Millipore, MA, USA). Images were captured using a Tanon 6200 Luminescent Imaging Workstation (Tanon, Shanghai, China). The following primary antibodies were used to detect proteins: mouse anti-ARG1 (1:4000; Proteintech, Wuhan, China), goat anti-CCL15 (1:1000; Arigo, Hsinchu City, China), mouse anti-HIF-1α (1:500; Abcam, Cambridge, UK), mouse anti-HIF-2α (1:200; Abcam, Cambridge, UK), rabbit anti-CCR1 (1:500; Novus, CO, USA), rabbit anti-NF-κB p65 (1:2000; Abcam, Cambridge, UK), rabbit anti-phosphoNF-κB p65 (1:1000; Abcam, Cambridge, UK), and rabbit anti-β-actin (1:2000; Proteintech, Wuhan, China). The secondary antibodies used in this study were goat anti-mouse IgG (H + L) HRP (1:4000; KeyGen, Nanjing, China), rabbit anti-goat IgG (H + L) HRP (1:3000; Fomacs, Nanjing, China), and goat anti-rabbit IgG (H + L) HRP (1:3000; Fomacs, Nanjing, China).

Histopathological and immunohistochemical analyses

Tissue samples were collected from patients and animal models. Tissue sections (4-μm) were deparaffinized, and subjected to antigen recovery using 100 mM citrate buffer target retrieval solution, pH 6.0, at 95 °C in a water bath for 20 min. Endogenous peroxidase activity was blocked by incubation with phosphate-buffered saline (PBS) and 3% hydrogen peroxidase for 30 min. After washing with PBS, the sections were incubated with mouse anti-HIF-1α (1:400; Abcam, Cambridge, UK), mouse anti-HIF-2α (1:500; Abcam, Cambridge, UK), rabbit anti-CD68 (1:8000; Abcam, Cambridge, UK), rabbit anti-CCR1 (1:200, Novus, CO, USA), rabbit anti-NF-κB p65 (1:1000; Abcam, Cambridge, UK), mouse anti-ARG1 (1:500; Proteintech, Wuhan, China) and mouse anti-Ki-67 (Typing, Nanjing, China) antibodies overnight at 4 °C, followed by the Envision Dual Link System HRP method (Dako, Glostrup, Denmark). All antibodies were diluted in Dako antibody diluents and reactions were revealed by incubating the sections with 3,3′-diaminobenzidine tetrahydrochloride (Dako, Glostrup, Denmark). In vivo, hypoxia was detected using Hypoxyprobe™-1 Plus Kits (Millipore, MA, USA) according to the manufacturer’s instructions. Briefly, 15 min before being sacrificed, mice were intra-peritoneally injected with a Hypoxyprobe™-1 (pimonidazole HCl) solution at a dose of 100 mg/kg body weight. Xenograft tumor tissues were removed for formalin fixation and paraffin embedding, followed by immunostaining with FITC-MAb1 (primary antibody, 1:100) and a peroxidase-conjugated anti-FITC secondary reagent (1:100). The subsequent steps were performed as mentioned above. The sections were observed and imaged using a laser scanning confocal microscope.

Cell counting kit-8 (CCK-8) assay

The cell viability of HNSCC cells following treatment with gefitinib and pyrrolidinedithiocarbamate ammonium (PDTC) (Selleck, Houston, USA), or the indicated conditioned media, was tested using a CCK-8 kit (Bimake, TX, USA) following the manufacturer’s instructions. Briefly, cells were seeded in 96-well plates at a density of 3000 cells/well in 100 μl complete culture medium. Blank control wells containing the same volume of complete culture medium were included in each assay. Next, the plate was incubated overnight at 37 °C to allow adherence of the cells, after which they were washed with PBS and incubated with specific reagents, which were serially diluted in complete culture medium before use, and incubated for 48 h. Next, CCK-8 (10 μl) was added to each well and the optical density (OD) of formazan at 450 nm was recorded every 0.5 h until the OD reached 1.0–2.0. Six wells were used corresponding to each concentration of the above reagents. Cell viability was calculated as follows: cell viability = ([OD] test–[OD] control)/([OD] control–[OD] blank)*100%.

Colony formation assay

For colony formation assessment, cells were plated in 6-well plates at a density of 500 cells per well. After growing for 10 days with the indicated treatments, the cells were stained with crystal violet (Beyotime, Shanghai, China) after which images were captured using a scanner (EPSON V330, Suwa City, USA). Each experiment was repeated three times.

5-Ethynyl-2′-deoxyuridine (EdU) incorporation assay

An EdU incorporation assay was performed according to the manufacturer’s protocol (RiboBio, Guangzhou, China). Briefly, cells were cultured in triplicate in 28.2 mm Glass Bottom Culture Dishes (Nest, Wuxi, China) for 24 h and then treated with 50 μM EdU for 2 h at 37 °C. After fixation in 4% formaldehyde for 30 min and permeabilization with 0.5% Triton X-100 for 10 min at room temperature, the cells were treated with a 1× Apollo reaction cocktail for 30 min. Subsequently, cell nuclei were stained with Hoechst 33342 and visualized under a laser scanning confocal microscope (Nikon, Tokyo, Japan). Each experiment included three replicates and was performed in triplicate.

Cytokine antibody array

A qualitative assessment of 80 cytokines in cell lysates was performed using a RayBio Human Cytokine Antibody Array 5 (RayBiotech, GA, USA) according to the manufacturer’s instructions. Briefly, membranes were blocked with a blocking buffer at room temperature for 30 min and incubated with the samples at room temperature for 2 h. Antigen-specific immuno-reactivity was detected using biotin-conjugated soluble antibodies and horseradish peroxidase-conjugated streptavidin. Densitometry of chemiluminescence-exposed X-ray film was used for quantification. The mean OD signal was determined for each spot. Positive controls were used to normalize the results obtained from different membranes. The chemokine profile was further analyzed using a dedicated microarray tool (Microsoft Excel software).

Enzyme-linked immunosorbent assay (ELISA)

The levels of specific cytokines in cell culture supernatants or sera were detected using corresponding ELISA kits, including human VEGF, IL-6 (MultiSciences, Hangzhou, China), CCL15 (RayBiotech, GA, USA) and mouse CCL9 (Boster, Wuhan, China) kits, according to the manufacturers’ instructions. Each sample was analyzed in triplicate.

Flow cytometry

HNSCC cells were harvested and washed with PBS containing 3% bovine serum albumin. After washing, a monoclonal anti-human CCR1 PE-conjugated antibody (BioLegend, CA, USA) was added to the cells and incubated for 30 min at 4 °C in the dark. The resulting cells were washed three times with PBS and analyzed using a FACSVerse flow cytometer (BD Biosciences, CA, USA) to measure the expression of CCR1. The data obtained were analyzed using FlowJo software (TreeStar, Inc., OR, USA).

Target gene expression knockdown

CCL15 and HIF-2α small interfering ribonucleic acids (siRNAs) and a scrambled siRNA were synthesized by RiboBio (Guangzhou, China). M2-type macrophages were transiently transfected with the CCL15 and HIF-2α siRNAs using Lipofectamine 3000 (Invitrogen, CA, USA) according to the manufacturer’s instructions. At the same time, M2-type macrophages were transiently transfected with the scrambled siRNA as a control.

To knockdown the expression of CCR1 in HNSCC cells, a lentivirus encoding a short hairpin RNA (shRNA; GenePharma, Shanghai, China) targeting CCR1 was constructed. HNSCC cells were cultured at a density of 5 × 10⁴ cells /ml in culture medium. After 24 h, the cells were transfected with a lentiviral vector containing green fluorescent protein-light chain 3B (GFP LC3B). Twelve hours later, the culture medium was replaced with complete culture medium. At least 72 h after transfection, the cells were used for subsequent assays. Fluorescence microscopy was used to evaluate the infection efficiency. The sequences of the shRNAs used in this study were as follows: control shRNA (shNC), 5′- TTCTCCGAACGTGTCACGT -3′; CCR1 gene shRNA #1, 5′- CCCTACAATTTGACTATACTT -3′; CCR1 gene #2, 5′- GCCTTCACTTTCCTCACGAAA -3′; and CCR1 gene #3, 5′- CCTCTGTACTCCTTGGTATTT -3′.

Patient cohort

All 70 clinical serum samples included in the study were collected from patients diagnosed with HNSCC at the Department of Oral and Maxillofacial Surgery, Nanjing Stomatological Hospital, Medical School of Nanjing University between January 2013 and June 2013. The mean and median ages at diagnosis were 59.84 and 61 years, respectively (range, 26–82 years). Their clinico-pathological features are listed in Table 1. 12 HNSCC patients who received surgical treatment were included in the CCL15 expression analyses between the pre- and postoperative groups. Their clinico-pathological features are listed in Supplementary file 1: Table S2. All patients provided written informed consent.

Table 1.

Clinical and pathological characteristics of HNSCC patients included in the prognosis analysis

Characteristics		No. (%) of patients
Age	< 60	29 (41.43)
	≥ 60	41 (58.57)
Sex	Male	44 (62.86)
	Female	26 (37.14)
Tumor location	Lip	3 (4.29)
	Mouth floor	5 (7.14)
	Cheek	11 (15.71)
	Tongue	30 (42.86)
	Gingiva	11 (15.71)
	Palate	3 (4.29)
	Oropharynx	7 (10.00)
Tumor stage	T1	36 (51.43)
	T2	29 (41.43)
	T3	2 (2.86)
	T4	3 (4.29)
Nodal stage	N0	43 (61.43)
	N1	14 (20.00)
	N2	13 (18.57)
Metastatic stage	M0	70 (100.00)
	M1	0 (0.00)
Histological grade	Low Grade	10 (14.29)
	Intermediate Grade	39 (55.71)
	High Grade	21 (30.00)

KEGG pathway enrichment analysis and gene set enrichment analysis (GSEA)

RNA-sequencing data of HNSCC patient tumor samples were downloaded from the Cancer Genome Atlas (TCGA) database (https://cancergenome.nih.gov/). The HNSCC samples were divided into two groups based on the CCR1 expression level. The “limma” R package was used to screen for differentially expressed genes (DEGs) between the CCR1-high and CCR1-low HNSCC tissue samples. To better explore the biological significance of the DEGs, pathway enrichment was assessed using clusterProfiler, a package with analysis and visualization functions, to provide valuable information for the KEGG analyses. GSEA (http://software.broadinstitute.org/gsea/index.jsp) was conducted to verify whether the NF-kB pathway was enriched in the gene ranks derived from the DEGs between the two groups. A false discovery rate (FDR) < 0.05 and |log2 fold change (FC)| > 1 were applied.

In vivo assay

The C3H/HeNCrl mice used in this study were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd. (Beijing, China) and maintained under specific pathogen-free conditions. After knocking down CCR1 expression, SCC7 cells were injected subcutaneously into the dorsal flanks of each mouse (2 × 10⁷/ml, 0.2 ml/mouse; shCCR1 in the left flank and shNC in the right flank). After 5 days, the mice were orally gavaged every day with gefitinib (60 mg/kg). After 3 weeks of treatment, the mice were sacrificed, and the xenograft tumors were removed. The tumor volumes were measured, after which the tumors were fixed in formalin, embedded in paraffin and stained with hematoxylin and eosin (H&E). Immunostaining for CCR1 and p65 were also carried out.

In a sensitization assay, 0.2 ml SCC7 cells (2 × 10⁷/ml) was injected subcutaneously into the dorsal flank of C3H/HeNCrl mice. After 5 days, the mice were randomly assigned to four groups (n = 6) and orally gavaged every day with PBS, gefitinib (60 mg/kg), metformin (200 mg/kg, as reported by Koji Harad et al. [24]), or gefitinib plus metformin. After 21 days of treatment, the mice were euthanized, serum was collected for ELISA analysis, and the tumors were removed, fixed in formalin, embedded in paraffin, and stained with H&E. Immunostaining analyses for the ARG-1, CCR1, p65 and Ki-67 proteins were also carried out, and apoptosis was analyzed using a TUNEL Apoptosis Detection Kit (Sigma-Aldrich, MO, USA) according to the manufacturer’s instructions. Hypoxic areas in the tumors were detected using a Hypoxyprobe™-1 Plus Kit (Millipore, USA) as previously described. Tumor volumes were measured at the indicated time points with calipers and calculated using the formula: volume = long diameter × short diameter²/2. On the same days, blood glucose levels of the mice were measured using an Accu-Chek Active Blood Glucose Meter (Roche, Basel, Switzerland) at 9:00 a.m.

Statistical analysis

All of the data are expressed as the mean ± standard deviation (SD). Statistical differences were measured using an unpaired two-sided Student’s t test or one-way ANOVA for multiple comparisons when appropriate. All statistical analyses were performed using the SPSS 17.0 statistical software package (SPSS Inc., IL, USA). Statistical significance was set at p < 0.05. All of the graphs were created using GraphPad Prism 7 (GraphPad Software Inc., CA, USA).

Results

Hypoxia enhances HNSCC abilities to recruit and polarize macrophages

Previous studies have indicated that tumor cells can recruit monocytes and promote their differentiation into various types of macrophages [25–27], suggesting that the tumor microenvironment plays important roles during recruitment and polarization. Here, we found that human THP-1 monocytes adhered to culture dishes became CD68 positive after PMA stimulation, indicating that THP-1 monocytes were successfully induced to become macrophages (Supplementary file 1: Fig. S1A and B). In addition, we found using cell migration assays that conditioned media from CAL27 and JHU011 HNSCC cells cultured under hypoxic conditions were able to significantly promote macrophage migration (Fig. 1a). To further verify the underlying mechanism, we cultured 4 HNSCC cell lines (CAL27, FaDu, JHU011 and SCC9) under hypoxic conditions and used RT-qPCR to assess the expression of cytokines that are known to play a role in macrophage recruitment. We found that under these conditions, among all cytokines tested, VEGF exhibited the most significant increase in expression (Fig. 1b). This finding was further confirmed by ELISA (Supplementary file 1: Fig. S1C). In addition, we found that anti-VEGF neutralizing antibodies suppressed macrophage migration induced by conditioned media from CAL27 and JHU011 cells cultured under hypoxic conditions (Fig. 1c). Since TAMs can be divided into two groups, M1 and M2, and only M2-type macrophages exhibit tumor-promoting capacities [28, 29], we further tested whether HNSCC cells under hypoxic conditions can promote macrophage polarization into the M2 phenotype. Using immunofluorescence staining and Western blotting we found that the conditioned media from the CAL27 and JHU011 cells cultured under hypoxic conditions significantly promoted macrophage polarization into the M2 phenotype (Fig. 1d and e). Concordantly, we found using RT-qPCR that among the cytokines known to promote macrophage polarization into the M2 phenotype, interleukin (IL)-6 exhibited the most significant shift in expression in the 4 HNSCC cell lines tested (Fig. 1f). This result was again confirmed by ELISA (Supplementary file 1: Fig. S1D). Furthermore, we found that an anti-IL-6 neutralizing antibody could decrease the condition medium-induced M2 polarization, while a recombinant IL-6 protein could further promote this polarization (Fig. 1g and h). These results indicate that HNSCC cells under hypoxic conditions can promote macrophage recruitment and polarization through VEGF and IL-6 secretion, thereby eliciting tumor promoting effects.

Fig. 1.

Hypoxia enhances the ability of HNSCC to recruit and polarize macrophages. a A Transwell assay was used to evaluate M0 macrophages treated with complete medium or a 1:1 mixture of CAL27 or JHU011 conditioned medium. b After being cultured under hypoxia for 48 h, the expression of recruitment-related genes in HNSCC cells was analyzed by RT-qPCR. c A Transwell assay was used to evaluate M0 macrophages treated with complete medium, a 1:1 mixture of complete medium and CAL27 or JHU011 conditioned medium, VEGF or an anti-VEGF antibody. d, e After treatment with a 1:1 mixture of CAL27 or JHU011 conditioned medium, the expression of M2-type macrophage markers was detected by immunofluorescence (CD206) and Western blotting (ARG-1) (IL-4 and IL-13 were used as positive controls). f After culturing under hypoxic conditions for 48 h, the expression of polarization-related genes in HNSCC cells was analyzed by RT-qPCR. g, h After treatment with a 1:1 mixture of complete medium and CAL27 or JHU011 conditioned medium, IL-6 or an anti-IL-6 antibody, the expression of M2-type macrophage markers was detected by immunofluorescence (CD206) and Western blotting (ARG-1) (IL-4 and IL-13 were used as positive controls). Statistical analysis was performed using SPSS software. *, p < 0.05; **, p < 0.01; and ***, p < 0.001

TAMs promote HNSCC tolerance to gefitinib through CCL15 under hypoxic conditions

TAMs have been associated with tolerance to treatment in a variety of tumors [19]. However, the impact of the physical and chemical microenvironment on TAMs has rarely been reported. Using immunohistochemistry, we found that macrophages are extensively exposed to hypoxia in clinical HNSCC specimens and that positive CD68 staining highly overlapped with positive HIF-2α staining, but not positive HIF-1α staining, in these TAMs (Fig. 2a), suggesting that HIF-2α may play a key role. To subsequently test whether TAMs enhance HNSCC tolerance to gefitinib, TAM-conditioned medium was applied to HNSCC cell cultures. Using a CCK-8 assay, we found that the conditioned medium from M2-type TAMs attenuated the inhibitory effect of gefitinib on the 4 HNSCC cell lines tested. This inhibitory effect was further attenuated after the M2-type TAMs had been cultured under hypoxic conditions (Fig. 2b). Subsequent colony formation and EdU incorporation experiments confirmed the above observations (Fig. 2c; Supplementary file 1: Fig. S2A). Based on these results, we next set out to identify the exogenous components involved. To this end, a cytokine array analysis of M2-type TAMs after normoxic and hypoxic treatment was carried out. By doing so, we identified two cytokines exhibiting >1.5-fold changes: IL-13 and MIP-1δ (i.e., CCL15) with 1.86-fold and 1.52-fold changes, respectively (Fig. 2e). Since it has been reported that metformin can inhibit macrophage polarization towards the M2 phenotype [30], we next tested whether metformin can inhibit the promotion of drug resistance by M2-type TAMs under hypoxic conditions. Using a CCK-8 assay, we found that metformin treatment of M2-type TAMs under hypoxic conditions significantly reduced the ability of the TAMs to promote gefitinib tolerance (Fig. 2d). In addition, we found that the levels of MIP-1δ (CCL15), insulin-like growth factor binding protein 4 (IGFBP-4), and neurotrophic factor 4 (NT-4) were downregulated >1.5-fold, i.e., 1.76-fold, 1.61-fold, and 1.59-fold, respectively, after treating M2-type TAMs with metformin (Fig. 2e). Accordingly, CCL15 may be the most promising cytokine from M2-type TAMs that can promote HNSCC tolerance to gefitinib. This notion was substantiated by ELISA (Fig. 2f). Subsequent colony formation and EdU incorporation experiments confirmed that CCL15 may play an essential role in the promotion of drug resistance by M2-type TAMs (Fig. 2g; Supplementary file 1: Fig. S2B). Similar results were obtained after siRNA-mediated knockdown of CCl15 expression in M2-type TAMs (Supplementary file 1: Fig. S2C and D). Of note, we found that that CCL15 expression was more abundant in M2-type TAMs than in HNSCC cells (Supplementary file 1: Fig. S2E), confirming that CCL15 secreted by stromal cells (TAMs) plays an essential role in inducing gefitinib resistance in HNSCC cells. To further verify whether CCL15 can be used as a prognostic biomarker for HNSCC, pre-operative serum CCL15 levels of HNSCC patients admitted to our hospital between January and June in 2013 were measured by ELISA (Table 1). The results obtained suggest that these serum CCL15 levels were significantly correlated with patient prognosis (Fig. 2h, p < 0.001). Interestingly, we found that the post-operative serum CCL15 levels were significantly reduced in 12 HNSCC patients (Supplementary file 1: Table S2) who recently received surgical treatment (Fig. 2i, p = 0.024). Together, these results indicate that after surgical resection of HNSCC, CCL15 production was reduced, suggesting that CCL15 may play an important role in HNSCC progression.

Fig. 2.

CCL15 reduces the sensitivity of HNSCC cells to gefitinib. a H&E staining and immunohistochemical analysis of HIF-1α, HIF-2α and CD68 on serial sections of primary HNSCC samples. b The viability of HNSCC cells treated with different doses of gefitinib or supplied with conditioned medium from M2-type macrophages for 48 h was assessed using a CCK-8 assay. c Colony formation analysis was used to assess the effect of gefitinib at the corresponding IC50 value and the addition of conditioned medium from M2-type macrophages or not on HNSCC cells. d A CCK-8 assay was used to assess the effect of treatment with different doses of gefitinib or the addition of conditioned medium from M2-type macrophages for 48 h on HNSCC cells. e, f A cytokine antibody array and ELISA were used to evaluate the effects of treatment conditions (normoxia, hypoxia, and hypoxia with 100 μM metformin) for 24 h (red rectangular frame, CCL15) on M2-type macrophages. g The colony forming ability of HNSCC cells treated with gefitinib at the corresponding IC50 value and supplemented with conditioned medium from M2-type macrophages or not (1 ng/ml CCL15; 50 ng/ml anti-CCL15 antibody) was assessed. h Kaplan-Meier overall survival curves comparing HNSCC patients with low and high CCL15 expression levels in serum (n = 70; p < 0.001) are shown. i ELISA revealing that the CCL15 expression levels in the sera of HNSCC patients were significantly reduced after surgical treatment (n = 12; p = 0.024). Statistical analysis was performed using SPSS software. *, p < 0.05; **, p < 0.01; and ***, p < 0.001

M2 macrophages regulate CCL15 secretion in a HIF-2α-dependent manner under hypoxic conditions

Hypoxia-inducible factors, including HIF-1α and HIF-2α, are known to play important biological roles [31]. Using immunohistochemical analysis of clinical HNSCC specimens we found that the expression of the macrophage marker CD68 highly overlapped with that of HIF-2α, an observation that is consistent with previously reported findings [32]. Therefore, we hypothesized that the promotion of drug resistance by M2-type TAMs under hypoxic conditions may be regulated by HIF-2α. To test this hypothesis, siRNA was used to knock down HIF-2α expression in M2-type TAMs (Supplementary file 1: Fig. S3A-D). We found that after HIF-2α expression knockdown the mRNA level of HIF-1α was slightly upregulated in a compensatory manner, while no significant change in HIF-1α protein expression was observed. Instead, a significant reduction in CCL15 expression was found. Conversely, we found that HIF-2α expression was not changed after CCL15 expression knockdown (Fig. 3a). Subsequently we applied Roxadustat (FG-4592), a HIF-2α stabilizer, to the TAMS and found that CCL15 expression was upregulated after HIF-2α expression stabilization under normoxic conditions (Fig. 3b). Finally, using colony formation and EdU incorporation experiments, we found that HIF-2α expression knockdown inhibited the ability of M2-type TAMs to promote drug resistance (Fig. 3c). These results confirm that HIF-2α plays an important role in regulating CCL15 secretion by M2 macrophages.

Fig. 3.

Hypoxia promotes the expression of CCL15 in M2-type macrophages through HIF-2α. a Western blot analysis of the levels of the indicated proteins in M2-type macrophages under hypoxia for 24 h after being transiently transfected with CCL15- and HIF-2α-specific siRNAs. b Western blot analysis of the levels of the indicated proteins in M2-type macrophages under hypoxia or normoxia or supplied with 10 μM FG-4592 (Selleck, TX, USA) for 24 h. c Colony formation and EdU incorporation analyses of HNSCC cells treated with gefitinib at the corresponding IC50 value and supplemented with the indicated conditioned media from M2-type macrophages or not

CCL15 promotes HNSCC tolerance to gefitinib through CCR1

CCR1 is the main receptor of CCL15 and its binding is known to play an important role in tumor progression [33, 34]. Here, we first assessed CCR1 expression in 4 HNSCC cell lines using flow cytometry (Supplementary file 1: Fig. S4A). Next, a lentivirus was used to transfect shRNA and knock down CCR1 expression in the respective HNSCC cell lines (Supplementary file 1: Fig. S4B-D). We found that CCR1 expression knockdown did not affect the proliferation of the HNSCC cell lines (Supplementary file 1: Fig. S4E). Additional colony formation and EdU incorporation experiments showed that the CCL15-mediated promotion of gefitinib resistance was significantly reversed after CCR1 expression knockdown. In addition, we found that gefitinib resistance was not induced by CCL15 in metformin-treated HNSCC cells (Fig. 4a). Subsequent Western blotting confirmed that metformin inhibited CCR1 expression in the HNSCC cell lines tested (Fig. 4b and c). These results indicate that metformin not only inhibited CCL15 secretion by macrophages, but also suppressed expression of the CCR1 receptor on the tumor cells. To substantiate the importance of CCR1 in drug resistance in vivo, its expression was knocked down in the murine HNSCC cell line SCC7 (Supplementary file 1: Fig. S4F), which was subsequently inoculated into C3H mice to establish a mouse allograft model (Fig. 4d). We found that the shCCR1 group exhibited significantly better therapeutic outcomes after gefitinib treatment than the shScramble group (Fig. 4e). Subsequent immunohistochemistry data confirmed that CCR1 expression in the shCCR1 group was significantly lower than that in the control group (Fig. 4f). Together, these results confirm that CCL15-CCR1 signaling plays an important role in promoting HNSCC resistance to gefitinib.

Fig. 4.

CCL15 reduces the sensitivity of HNSCC cells to gefitinib through CCR1. a Colony formation and EdU incorporation analyses were used to assess HNSCC cells treated with gefitinib at the corresponding IC50 value and supplemented with 1 ng/ml CCL15 or not. b, c Western blot analysis of CCR1 expression in CAL27 and JHU011 cells after treatment with different doses of metformin (Sigma-Aldrich, MO, USA) for 24 h or 10 mM metformin for different times. d, e and f CCR1 expression knock down enhanced the sensitivity of tumors to gefitinib and reduced the activation of p65 in vivo (p = 0.02)

The CCL15-CCR1-NF-κB signaling axis promotes HNSCC tolerance to gefitinib

Based on the above results, the identification of CCL15-CCR1 downstream signaling pathways involved in promoting HNSCC resistance to gefitinib was deemed appropriate. Using RNA-sequencing data of HNSCC tumor samples acquired from the TCGA database, DEGs were identified based on CCR1 expression (high versus low) (Supplementary file 1: Fig. S5A). Further enrichment analysis showed that the NF-κB signaling pathway was one of the major pathways in which the DEGs were enriched (Supplementary file 1: Fig. S5B and C), and GSEA analysis further confirmed that “the NF-κB pathway” was significantly enriched in the HNSCC samples that highly expressed CCR1 (Supplementary file 1: Fig. S5D-G). Subsequently, we conducted validation experiments using Western blotting and immunofluorescence staining, and found that NF-κB was activated by CCL15 (Fig. 5a and b), and that this activation was reversed after CCR1 expression knock down in HNSCC cells, suggesting that CCL15 activates NF-κB through CCR1 (Fig. 5c and d). Next, we set out to inhibit NF-κB in order to enhance gefitinib sensitivity using Pyrrolidine dithiocarbamate ammonium (PDTC), which is a small molecule inhibitor that inhibits NF-κB activation and nuclear translocation [35]. To explore the synergistic effect of PDTC, an appropriate concentration of PDTC that inhibits NF-κB activation without significantly affecting HNSCC cell proliferation was subsequently selected. Using a CCK-8 assay, Western blotting and immunofluorescence staining, we found that 10 μM PDTC exhibited no obvious cytotoxicity (Supplementary file 1: Fig. S5H), but significantly inhibited NF-κB activation and nuclear translocation (Supplementary file 1: Fig. S5I and J). Thus, 10 μM PDTC was selected for further experiments. Using colony formation and EdU incorporation experiments we found that 10 μM PDTC indeed suppressed CCL15-promoted gefitinib resistance (Fig. 5e). Finally, we used immunohistochemistry to evaluate the allograft mouse models established with SCC7 cells and found that the nuclear translocation of NF-κB was significantly reduced in the shCCR1 group (Fig. 4f).

Fig. 5.

CCL15 enhances the resistance of HNSCC cells to gefitinib by the CCR1-NF-κb pathway. a, b Western blot and immunofluorescence analyses of NF-κB p65 and p-p65 expression in CAL27 and JHU011 cells after treatment with different doses of CCL15 for 48 h. c, d Western blot and immunofluorescence analyses of NF-κB p65 and p-p65 expression in CAL27 and JHU011 cells with or without CCR1 expression knockdown stimulated with 1 ng/ml CCL15 for 48 h. e Colony formation and EdU incorporation analyses of HNSCC cells treated with gefitinib at the corresponding IC50 value and supplemented with 1 ng/ml CCL15 or 10 μM PDTC or not

Metformin and gefitinib synergistically inhibit HNSCC in vivo

Given the inhibitory effects of metformin on tumor growth as well as its therapeutic sensitization effects, the inhibition of M2-type TAM CCL15 secretion and the expression downregulation of the HNSCC cell surface receptor CCR1 by metformin, we next set out to study allograft mouse models to validate the mechanism by which metformin enhances gefitinib sensitivity (Fig. 6a). We found that the tumor volume of mice in the metformin plus gefitinib group was significantly lower than that of mice in the control group, the metformin group, and the gefitinib group (Fig. 6b and c). No significant differences in body weight were noted among the groups (Fig. 6d). In addition, we found that the spleen indices of all treatment groups were significantly better than those of the control group, suggesting that the antitumor immunity of the mice in the treatment groups was enhanced, with the strongest effect occurring in the metformin plus gefitinib group (Fig. 6e). Since mouse CCL9 is homologous to human CCL15 [36], we assessed mouse CCL9 serum levels using ELISA and found that they were reduced after metformin treatment (Fig. 6f). To exclude the possibility that metformin treatment may lower blood glucose levels in the mice, we assessed its levels and found that 200 mg/kg metformin per day did not lead to changes blood glucose levels between the treatment and control groups (Fig. 6g). Subsequent immunohistochemistry confirmed that the numbers of M2-type macrophages (ARG1-positive cells) in the metformin plus gefitinib group and the metformin only group were significantly reduced. These findings are consistent with those of a previous report [30]. Moreover, we found that the CCR1 expression level and the rate of NF-κB nuclear translocation in the metformin plus gefitinib group and the metformin group were significantly lower than those in the non-metformin groups. Ki-67 proliferation staining further confirmed the synergistic effect of metformin and gefitinib (Fig. 6h). The results underscore the efficacy and safety of metformin plus gefitinib in the in vivo treatment of HNSCC.

Fig. 6.

Metformin enhances the sensitivity of HNSCC to gefitinib in vivo. a Scheme showing the in vivo metformin sensitization experiment. During the assay, b/c tumor volume, d body weight, e the spleen index (spleen weight/body weight), f the CCL9 level in the serum g and the blood glucose level were compared among different groups. h HE staining and IHC analyses of hypoxia, ARG-1, CCR1, Ki-67, TUNEL and p65 activation in different groups (100x and 400x). Statistical analysis was performed using SPSS software. *, p < 0.05; **, p < 0.01; and ***, p < 0.001

Discussion

In addition to surgery, chemo- and radiotherapy are the major treatment modalities for HNSCC. However, chemo- and radiotherapy are not target oriented and they affect not only tumor cells but also normal cells, such as hematopoietic cells. As a consequence, many cancer patients cannot adhere to the entire course of treatment. It is, therefore, warranted to explore individualized molecular targeted therapies in the era of precision medicine. As such, EGFR-targeted therapy holds promise. Several targeted drugs have been developed, such as gefitinib and erlotinib (firsts generation), afatinib and dacomitinib (second generation), and avitinib and rociletinib (third generation). Application of the first generation drugs results in EMT and cyclin D1 overexpression, which affects the therapeutic outcome of cancer treatment regimens [12]. Drug resistance may, however, occur and attempts to explain this resistance has turned out to be complicated. Despite some clinical trials that have been performed, such as the one by Perez and Van Waes et al. that used gefitinib for HNSCC treatment (www.clinicaltrial.gov), the USA Food and Drug Administration (FDA) has refrained from approving gefitinib as a first-line HNSCC treatment option. Thus, an in-depth investigation of how HNSCCs develop resistance to gefitinib is necessary to improve its treatment options and to obtain the best therapeutic effects.

The results of our current study are consistent with those of Laitala and Erler [37] indicating that HNSCC stromal cells may be affected by a hypoxic microenvironment. We found that the tumor cells themselves can recruit macrophages by secreting VEGF and polarize the macrophages into M2-type TAMs through IL-6, as has also been reported by Barleon et al. and Chen et al. [38, 39]. However, how these macrophages affect tumor cell responses to gefitinib has not been reported yet. Our cytokine array analysis showed that CCL15 may play a key role in how macrophages affect the response of HNSCC cells to gefitinib treatment. CCL15 is a chemokine that is mainly expressed in the intestine and liver [40, 41]. It has been reported that CCL15 may promote the migration and invasion of hepatocellular carcinoma cells through its receptor CCR1 [33]. We found that CCL15 is secreted by macrophages, and is not expressed by HNSCC cells. To further investigate the role of hypoxia on CCL15 secretion by macrophages, we found that CD68 expression highly overlapped with HIF-2α expression, which is consistent with a report by Beasley et al. [42], and suggests that HIF-2α rather than HIF-1α plays a key role in regulating the expression of CCL15 by TAMs. We also found that knocking down CCR1 in HNSCC cells partially reversed CCL15-promoted gefitinib resistance in these cells, confirming that CCL15-CCR1 signaling plays a key role. Through bioinformatic analysis we found that the downstream effect of CCL15-CCR1 signaling was activation of NF-κB signaling, which eventually generated gefitinib resistance (Supplementary file 1: Fig. S6).

Our results may be relevant for the clinical diagnosis and treatment of HNSCC. First, we found that the vast majority of CCL15 was secreted by M2-type TAMs and that HNSCC cells rarely secrete this cytokine. The serum CCL15 levels in HNSCC patients showed a significant correlation with prognosis. The postoperative serum CCL15 levels in the HNSCC patients were found to be significantly reduced, suggesting that CCL15 may serve as a prognostic marker for HNSCC that can be assessed using a serological assay. These observations are consistent with those reported by Li et al. [41]. Second, we report a new use for a traditional drug. Metformin is a drug that is routinely used to treat diabetic patients. Previous studies have shown that metformin may enhance the sensitivity to anticancer therapy by inhibiting EMT, targeting STAT3, and/or regulating antitumor immunity [43–45]. Our previous work has also shown that metformin may sensitize oral squamous cell carcinomas to cisplatin and gefitinib treatment [16, 17]. Here, we found that metformin alone can inhibit CCL15 expression in M2-type TAMs caused by hypoxia. In addition, we found that it can suppress CCR1 expression in HNSCC cells, thereby inhibiting the cross-talk between macrophages and tumor cells, and eliciting therapeutic effects both in vitro and in vivo.

Abbreviations

ARG-1

argininase 1

CCL2/5/15

C-C motif chemokine ligand 2/5/15

CCR1

C-C motif chemokine receptor 1

CSF-1

colony stimulating factor 1

EGFR

epidermal growth factor receptor

EMT

epithelial-mesenchymal transition

EOT

eotaxin

ET-2

endothelin 2

GSEA

gene set enrichment analysis

H-CM

hypoxic condition medium

HE

hematoxylin and eosin

HIF

hypoxia-inducible factors

HNSCC

head and neck squamous cell carcinoma

IL-4/6/10/13

interleukin 4/6/10/13

KEGG

Kyoto Encyclopedia of Genes and Genomes

NC

negative control

N-CM

normoxic condition medium

NF-κB

nuclear factor kappa-light-chain-enhancer of activated B cells

OSM

oncostatin M

PDTC

pyrrolidinedithiocarbamate ammonium

PMA

phorbol 12-myristate 13-acetate

SDF-1

stromal cell-derived factor 1

SEMA3A

semaphorin 3A

TAMs

Tumor-associated macrophages

TCGA

the Cancer Genome Atlas

TEFB

transcription factor EB

TGF

β: tumor growth factor β

Compliance with ethical standards

Conflict of interest

The authors declare no conflicts of interest.

Ethics statement

All experiments were approved by the Ethics Review Board of the Nanjing Stomatological Hospital (approval number: 2017-NKL012). All animal experiments and experimental protocols used were in accordance with the animal care and use committee of the medical school of Nanjing University.