Epstein–Barr Virus-Induced VEGF and GM-CSF Drive Nasopharyngeal Carcinoma Metastasis via Recruitment and Activation of Macrophages

Microbial infection causally accounts for more than 20% of malignancies worldwide (1). Epstein–Barr virus (EBV) is a human gammaherpesvirus that infects approximately 95% of the global population and linked to various types of epithelial and lymphoid cancer, including nasopharyngeal carcinoma (NPC; ref. 2). Elevated EBV infection titers in serum, induced by environmental stress like smoking and other chemicals like phorbol-12-myristate-13-acetate (TPA; ref. 3) and sodium butyrate (4), is closely associated with NPC progression (5).

Tumor initiation and progression caused by viral infections reflect properties that are intrinsic to tumor cells, including expression of viral oncoprotein and nontranslated viral RNAs that directly contribute to tumor growth (6), metastasis (7), and drug resistance (8). Viral carcinogenesis can also occur indirectly. In many cases, although the immune response to pathogens that establish persistent infections is designed to promote host defense followed by tissue repair, it can also stimulate chronic inflammation and tumor growth (9). Mounting evidence has indicated that persistent viral infections can foster chronic inflammatory microenvironments that extrinsically promote tumor development (9, 10). For example, Hepatitis B and C viruses can lead to the chronic production of protumor cytokines via the induction of cell death (11). In addition, inflammation can also be induced by virus-encoding oncoproteins. For example, the oncoprotein encoded by human papillomavirus (HPV) in cervical epithelial cells can induce a hyperinflammatory response by promoting IL17A production (12). Although it has been well documented that EBV plays a central role in the pathogenesis of NPC, the interaction between EBV activity and the tumor-promoting inflammatory microenvironment in NPC remains elusive.

Macrophages are the most abundant recruited inflammatory cells in the tumor microenvironment, and they are closely associated with poor prognosis in various tumor types (13–16). Chemokine (C-C motif) ligand (CCL)18 has been shown to be exclusively secreted by myeloid-derived cells and is one of the most abundantly produced cytokines of tumor-associated macrophages (TAM) in various types of cancer, including breast, gastric, and ovarian cancer (17–19). TAM-derived CCL18 has been reported to promote cancer cell invasion by inducing the epithelial–mesenchymal transition (EMT; refs. 17, 20). However, the clinical significance of CCL18 and its underlying mechanism of action in NPC remain unknown.

In this study, we investigated the contribution of EBV replication in NPC cells to macrophage activation and the interaction between NPC cells and TAMs. The results demonstrated the underlying mechanism linking EBV replication and tumor metastasis.

We obtained 580 pathologically proven and nondistant metastatic paraffin-embedded nasopharyngeal carcinoma specimens for this study. A total of 321 specimens were obtained from the Sun Yat-sen University Cancer Center (Guangzhou, China) between Jan 1, 2010 and Dec 31, 2011, and 259 specimens were collected from the Sun Yat-Sen Memorial Hospital (Guangzhou, China) between Jan 1, 2009 and Dec 31, 2011. All samples were pathologically reassessed by two pathologists. None of the patients had received radiotherapy or chemotherapy before biopsy sampling. All MRI and CT scans were reassessed separately by two radiologists, and any disagreements were resolved by consensus. Clinical staging was reclassified according to the criteria of the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC)-staging manual (seventh edition). All patients were treated with 2.0 Gy per fraction with five daily fractions per week for 6–7 weeks using either two-dimensional (2D) radiotherapy (2DRT), IMRT, or three-dimensional (3D) conformal radio therapy (3DCRT), in accordance with the treatment policy adopted by each center. The cumulative radiation doses were 66 Gy or greater to the primary tumor and 60–66 Gy to the involved neck area. All potential sites of local infiltration and bilateral cervical lymphatics were irradiated to 50 Gy or greater. Patients with stage III to IV_B disease also received concurrent platinum-based chemotherapy. This study was approved by the institutional ethical review boards of both hospitals, and written informed consent was obtained from all patients. The studies were conducted in accordance with International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS).

We used the 321 specimens from Sun Yat-sen University Cancer Center as a training set to obtain the optimal cut-off point of TAM count or CCL18⁺ TAM count per field for disease-free survival (DFS), overall survival (OS), and disease-specific survival (DSS). To assess whether TAM count or CCL18⁺TAM count also had prognostic value in different patient population, we used the 259 samples from Sun Yat-sen Memorial Hospital as an independent set to externally validate the results.

DSS was defined as the time from surgery to death from NPC or death due to treatment; patients known to be alive or lost to follow-up at the time of analysis were censored at their last follow-up and deaths from other causes without disease were treated as a competing risk. Overall survival (OS) was defined as the time from surgery to death from any cause; patients known to be still alive at the time of analysis were censored at the time of their last follow-up. DFS was defined as the time from surgery to first evidence of recurrent/metastatic disease or death from any cause; patients known to be alive without recurrent/metastatic disease or lost to follow-up at the time of analysis were censored at the time of their last follow-up.

A total of 5–10 mL of each patient's blood was collected. The VCA-IgA antibody titers of EBV were measured using a commercial kit (Zhongshan Bio-tech Co Ltd) based on standard immunoenzymatic techniques, which were described previously (5). To minimize bias, samples were blinded to laboratory personnel. Serial dilutions of quality control sera (1:10, 1:20, 1:40, and 1:80) were applied to every assay for the evaluation of intraset variability. To minimize experimental error, all tests were conducted in the same laboratory by the same technician staff.

Paraffin-embedded samples were sectioned at 4-μm thickness. Antigen retrieval was performed using a pressure cooker for 30 minutes in 0.01 mol/L citrate buffer (pH 6.0). For immunofluorescence staining, the specimens were incubated with antibodies specific for CD68 (catalog no. sc7083, Santa Cruz Biotechnology) and CCL18 (catalog no. MAB394, R&D Systems) overnight at 4°C, followed by incubation with Alexa Fluor 555 donkey anti-goat IgG (H+L) and Alexa Fluor 488 donkey anti-mouse IgG (H+L; Life Technologies). Cells stained with the indicated antibody were imaged using a confocal laser-scanning microscope (Carl Zeiss) with a core data acquisition system (Applied Precision) and calculated per field of view, and at least 10 fields of view per section were evaluated at 400× magnification.

All the NPC cell lines used in this study were kindly provided by Dr. Musheng Zeng from the Sun Yat-Sen University Cancer Center in 2014. All cell lines were passaged less than 10 times or 6 months after initial revival from frozen stocks. The cell lines were tested negative for mycoplasma contamination (catalog no. 201203, Qiagen). All cell lines were authenticated by short tandem repeat profiling prior to use. Two poorly differentiated NPC cell lines (CNE2 and HNE1) and a highly differentiated NPC cell line (HK1) were grown in RPMI1640 medium (Life Technologies) containing 10% FCS (Life Technologies) in a humidified 5% CO₂ incubator at 37°C. EBV⁺ NPC cells were generated from parental cell lines by coculturing of the surface IgG crosslinked EGFP-neo'EBV–infected Akata cells according to the previous report's protocol (21).

Fresh human cord blood was obtained from Sun Yat-Sen Memorial Hospital, according to guidelines approved by the ethics boards and the Clinical Research Committee at Sun Yat-Sen Memorial Hospital. The hematopoietic stem cells (HSC) were isolated by using direct CD34 Progenitor Cell Isolation Kit (Miltenyi Biotec) as described previously (20). All animal work was conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee at the medical college of Sun Yat-Sen University. Humanized mice were generated as described previously (20). Briefly, NOD/SCID mice (3–4 weeks old) were subjected to 200 cGy total body irradiation 12 hours before being injected with 2 × 10⁵ HSCs in 0.2 mL of medium via the tail vein. At 8 weeks after HSC transplantation, 5 × 10⁶ CNE2-EBV⁻, CNE2-EBV⁺, or CSE-treated CNE2-EBV⁺ NPC cells were subcutaneously injected into the back of mice. For antibody treatment, mice were injected with CCL18-specific neutralizing antibody (catalog no. ab9849; Abcam) and/or GM-CSF–specific neutralizing antibody (catalog no. MAB215; R&D Systems) and/or VEGF-specific neutralizing antibody (catalog no. ab9570; Abcam) via the tail vein at 10 mg/mouse twice a week after the xenografts became palpable (about 0.4 cm in diameter). Tumor growth was evaluated by monitoring tumor volume (length × width² × 0.5) every 3 days. The animals were sacrificed when the xenografts reached 1.5 cm in diameter. The tumor xenografts and lungs of the sacrificed mice were harvested for further investigation. The xenografts were collected to exam the state of EMT and macrophages infiltration, while the lungs were harvested to exam the metastasis by IHC staining as described before (22, 23). Similar animal experiments were performed with HNE1 cell line.

All statistical analyses were performed using SPSS for Windows version 13.0 (SPSS). We selected the optimum cut-off score for the count of CCL18⁺ TAM in nasopharyngeal carcinoma using X-tile plots based on the association with the patients' DSS. X-tile plots provide a single and intuitive method to assess the association between variables and survival, which can automatically select the optimum data cut-off point according to the highest χ² value (minimum P value). We did the X-tile plots using the X-tile software version 3.6.1 (Yale University School of Medicine, New Haven, CT). Spearman correlation and regression analysis was performed to assess the relationship between EBV titres in the serum of patients and CCL18⁺ counts, CD68⁺ counts, and CCL18⁺TAM counts in human nasopharyngeal tissue. Kaplan–Meier survival curves were plotted, and the log-rank test was performed. We used the Cox regression model to do the multivariable survival analysis. The mean level of serum EBV VCA-IgA was calculated by geometric mean. Groups of discrete variables were compared using the Mann–Whitney U test or Kruskal–Wallis nonparametric ANOVA. All cell culture experiments were performed independently in triplicate at least three times. In all cases, P values <0.05 were considered statistically significant.

A total of 580 patients with NPC were included in this study, including 321 patients from Sun Yat-sen Cancer Center as the training cohort and 259 patients from Sun Yat-sen Memorial Hospital as the validation cohort. The median follow-up time of the training cohort was 43 months (range, 8–58 months), among which 98 patients (30.5%) developed local recurrences and/or distant metastases. The median follow-up time of the validation cohort was 67 months (range, 7–79 months), among which 84 patients (32.4%) had tumor relapse. The baseline demographic and clinical characteristics among two groups showed no significant statistical difference (Supplementary Table S1).

To investigate the clinical significance of macrophage infiltration and CCL18 expression in NPC, we performed double immunostaining of CCL18 and CD68 in the tumor samples. We observed that CCL18 expression was predominantly in CD68⁺ TAMs (Fig. 1A), which is consistent with the previous reports in other cancer types (17, 24) and 63.49% ± 12.21% of TAMs in the NPC samples were positive for CCL18. Next, we correlated the total number of macrophages (CD68⁺ cells) and the number of CCL18⁺ TAMs (CCL18⁺CD68⁺ cells) with the prognosis of NPC patients. The cut-off points of total TAM count and CCL18⁺ TAM count for patient outcome were determined by X-tile analysis in the training set. Total TAM count cut-off points for DFS, OS, and DSS were 22.20, 23.20, and 22.00 cells per field, while CCL18⁺ TAM count cut-off points for DFS, OS, and DSS were 20.70, 22.30, and 20.70 cells per field, respectively (Supplementary Table S2). We selected the cut-off points for DSS (≤ versus > 22.00 cells/field for total TAM number; ≤ versus > 20.70 cells/field for CCL18⁺TAM count;) as the uniform cutoff points for DFS, OS, and DSS in both training cohort and validation cohort.

In the training set, both high macrophage number and CCL18⁺ TAM count were correlated with early relapse, poor long-term overall survival and disease-specific survival (Fig. 1B; Supplementary Fig. S1A, top). Among them, the CCL18⁺ TAM counts showed higher prognostic value than total macrophage count, as indicated by the P value and HR (DFS: CCL18⁺ TAM: log-rank P < 0.001; HR: 2.037; 95% CI, 1.366–3.037 versus total TAM: log-rank P = 0.042; HR: 1.517; 95% CI, 1.008–2.283; OS: CCL18⁺ TAM: log-rank P = 0.001; HR: 2.062; 95% CI, 1.345–3.159 versus total TAM: log-rank P = 0.040; HR: 1.582; 95% CI, 1.021–2.452; DSS: CCL18⁺ TAM: log-rank P < 0.001; HR: 3.099; 95% CI, 1.904–5.043 versus total TAM: log-rank P = 0.022; HR: 1.738; 95% CI, 1.069–2.826). In addition, univariate Cox regression analysis showed that DSS was associated with T stage (tumor size), N stage (lymph node metastasis status), UICC stage, and CCL18⁺ TAM count (Table 1). Furthermore, adjusting for tumor size, node status, and UICC stage, CCL18⁺TAM counts remained an independent prognostic factor of survival in multivariate analysis (HR: 2.770; 95% CI, 1.669–4.595; P = 0.001, n = 321, Table 1).

To further validate the prognostic value of the TAM number and CCL18⁺ TAM count, we applied the above cut-off setting to an independent validation group with 259 cases. Similarly, patients with higher TAM number or CCL18⁺ TAM count had shorter DFS, OS, and DSS than those with lower TAM number or CCL18⁺ TAM count (Fig. 1B; Supplementary Fig. S1A, bottom). Again, the CCL18⁺TAM counts showed higher prognostic value than total macrophage count (DFS: CCL18⁺ TAM: log-rank P < 0.001; HR: 2.172; 95% CI, 1.414–3.337; versus total TAM: log-rank P = 0.018; HR: 1.928; 95% CI, 1.109–3.350; OS: CCL18⁺ TAM: log-rank P = 0.003; HR: 1.814; 95% CI, 1.209–2.720; versus total TAM: log-rank P = 0.022; HR: 1.708; 95% CI, 1.071–2.725; DSS: CCL18⁺ TAM: log-rank P < 0.001; HR: 3.115; 95% CI, 1.867–5.198; versus total TAM: log-rank P = 0.017; HR: 2.067; 95% CI, 1.247–3.425). In consistent to the training cohort, Cox regression multivariate analysis showed that CCL18⁺TAM counts were an independent prognostic factor for patient DSS in the validation cohort (HR: 2.177; 95% CI, 1.292–3.668; P = 0.003, n = 259, Table 1).

In addition, we investigated the association between the CCL18⁺ TAM count and the clinicopathological features in both cohorts. A higher CCL18⁺ TAM count was significantly associated with a higher histologic grade, more aggressive lymph node, distant metastases, and a more advanced cancer stage in both cohorts (Supplementary Table S3). Collectively, these data suggest that the TAM-derived-CCL18 plays an important role in the progression of human NPC.

Elevated EBV infection titer in serum is closely associated with NPC progression. Therefore, we investigated whether EBV activity is correlated with macrophage infiltration and their CCL18 expression in NPC patients. Our data showed that the total TAM count, CCL18⁺ cell count, and CCL18⁺ TAM count all significantly increased in the EBV⁺ group compared with the EBV⁻ group in both training cohort (Fig. 1C) and validation cohort (Supplementary Fig. S1B). Moreover, EBV infection titer in serum is positively correlated with the total TAM count, CCL18⁺ cell count, and CCL18⁺ TAM count in NPC patients (Fig. 1C; Supplementary Table S3). In addition, the percentages of CCL18⁺ TAMs (CCL18⁺CD68⁺) in total TAMs significantly increase in the EBV⁺ group compared with the EBV⁻ group in both cohorts (Supplementary Fig. S1C).

It has been reported that smoking is associated with an increased EBV seropositivity among NPC patients (5). In agreement, our data showed that the serum concentration of EBV VCA IgA in smokers (n = 309, geometric mean = 675.12, SE = 37.55) was higher than that in nonsmokers (n = 271, geometric mean = 536.78, SE = 24.14, P = 0.001 by using Mann–Whitney U test, Fig. 1D). Importantly, we observed that the total TAM count, CCL18⁺ cell count, and CCL18⁺ TAM count all significantly increased in the smokers compared with the nonsmokers in both training cohort (Fig. 1D) and validation cohort (Supplementary Fig. S1D). Collectively, these data indicated that tumor infiltration of macrophages and their CCL18 expression are closely associated with EBV infection, smoking, and poor prognosis of NPC patients.

T-cell infiltration has been reported to be associated with a better prognosis in many tumors (25, 26). We evaluated T-cell infiltration in 580 NPC samples by CD3 immunostaining and observed both TAM count and CCL18⁺ TAM count were negatively correlated with T-cell counts (TAM count: R² = 0.199, P < 0.001; CCL18⁺TAM count: R² = 0.206, P < 0.001; Supplementary Fig. S1E), suggesting macrophage infiltration and their CCL18 expression may be associated with NPC immunosuppression.

We next investigated the interaction between macrophages and NPC cells with different EBV replication in vitro by using a Transwell coculture system. Freshly isolated human monocytes were added to the top chamber, and condition media (CM) of three GFP-EBV–infected NPC cell lines (CNE2-EBV⁺, HNE1-EBV⁺, and HK1-EBV⁺) or that of parental EBV⁻ cells was added to the bottom chamber of the transwell apparatus. We observed significantly more monocytes migrated toward CM of EBV⁺ NPC cells compared with that of EBV⁻ NPC cells and control medium (Fig. 2A). To investigate whether EBV⁺ NPC cells can activate monocytes to become TAM-like macrophages, we placed monocytes into the bottom transwell chamber to be cocultured with EBV⁻ NPC cells or EBV⁺ NPC cells for 5 days and evaluated surface markers (CD206^high/HLA-DR^low; ref. 20) and protumor cytokines associated with TAMs (CCL18, CCL17, CCL22, and IL10; refs. 17, 20, 27) in the cocultured macrophages. We found that only macrophages cocultured with EBV⁺ NPC cells, rather than those cocultured with EBV⁻ NPC cells exhibited a CD206^high/HLA-DR^low phenotype compared with untreated macrophages (Fig. 2B and C; Supplementary Fig. S2A). In consistent with the changes in surface markers, macrophages cocultured with EBV⁺ NPC cells produced significantly higher amounts of protumor cytokines compared with untreated macrophages and macrophages cocultured with EBV⁻ NPC cells (Fig. 2D).

Environmental stress like smoking can increase EBV replication. Indeed, we observed fluorescence intensity in GFP-EBV–infected NPC cell lines was enhanced in a dose-dependent manner following treatment with proportions of cigarette smoke extract (CSE) ranging from 5% to 20% (Supplementary Fig. S2B–S2D), which is consistent with a previous report (5). To further investigate whether EBV replication was associated with NPC cell's capacity to recruit and activate macrophages, we treated the monocyte/macrophages with the CM of EBV⁻NPC cells and EBV⁺NPC cells with/without CSE pretreatment. The chemotaxis assay results showed that the CM of CSE-treated EBV⁺NPC cells attracted more monocytes compared with that of untreated EBV⁺NPC cells. In contrast, there is no difference between the EBV⁻NPC cells with and without CSE treatment (Fig. 2A). In addition, macrophages cocultured with CSE-treated EBV⁺ cells exhibited a more TAM-like phenotype than those cocultured with untreated EBV⁺ cells, which was indicated by higher CD206 and lower HLA-DR expression and increased secretion of protumor cytokines. Similar to the chemotaxis assay, the activation state of macrophages has no difference between EBV⁻cells with and without the CSE treatment (Fig. 2B–D; Supplementary Fig. S2A). Collectively, these data suggest that EBV replication in NPC cells plays a crucial role in their capacity to recruit and activate macrophages.

To identify the EBV⁺ NPC-derived cytokines that recruit and activate macrophages, the cytokine profiles of the CM from CNE2-EBV⁻, CNE2-EBV⁺, and CSE-treated CNE2-EBV⁺ cells were analyzed using a RayBio Human Cytokine Antibody Array. The gray intensity analysis showed that four cytokines were increased in the CM of CSE-treated EBV⁺ cells compared with the CM of EBV⁻ cells and that of EBV⁺ NPC cells; including GM-CSF, VEGF, GRO, and IL1α (Fig. 3A). The ELISA assay further confirmed the increases of those cytokines in EBV⁺ NPC cell lines compared with EBV⁻ ones, and further increase in CSE-treated EBV⁺ NPC cell lines, rather than CSE-treated EBV⁻ ones (Fig. 3B; Supplementary Fig. S3A). We then investigated which of those cytokines was responsible for monocyte recruitment and TAM activation. Using different concentrations of recombinant cytokines, we found that among these four cytokines, only VEGF recruited monocytes to a level comparable with that of CSE-treated EBV⁺ cell CM (Fig. 3C). Moreover, only GM-CSF activated macrophages to produce protumor cytokines to levels comparable with that of CSE-treated EBV⁺ cell CM (Fig. 3D). Furthermore, the addition of neutralizing GM-CSF antibody to the CM of EBV⁺ cells and CSE-treated EBV⁺ cells abrogated the induction of TAM-related surface marker expression (Fig. 3E; Supplementary Fig. S3B) and cytokine secretion (Fig. 3F; Supplementary Fig. S3C), whereas the addition of neutralizing VEGF antibody had no appreciable effect on surface marker expression or cytokine production (Fig. 3E and F). In comparison, neutralization of VEGF, rather than GM-CSF, markedly inhibited the monocyte chemotaxis induced by EBV⁺ cells and CSE-treated EBV⁺ cells (Fig. 3G; Supplementary Fig. S3D). These findings indicate that EBV-replicating NPC cells recruit monocytes and activate TAMs via VEGF and GM-CSF, respectively.

Recent studies show that EBV replication activates NF-κB in NPC cells via the interaction between latent membrane protein 1 and tumor necrosis factor receptor-associated factors (28). In addition, both VEGF and GM-CSF have been identified as target genes of NF-κB (20, 29). Thus, we examined NF-κB activation in NPC cells with different EBV replication. We found that EBV⁺ NPC cells exhibited enhanced NF-κB activation compared with EBV⁻ NPC cells, including IκB kinase (IKK) phosphorylation, IκB phosphorylation and degradation (Fig. 4A; Supplementary Fig. S4A), p65 nuclear translocation (Fig. 4C; Supplementary Fig. S4B); and NF-κB binding to its DNA consensus sequence (Fig. 4B). CSE treatment, which increased EBV replication, further promoted NF-κB activation in EBV⁺ cells. In contrast, CSE treatment could not activate the NF-κB pathway in EBV⁻ NPC cells (Fig. 4A–C; Supplementary Fig. S4A and S4B). To determine whether NF-κB activation in responsible for cytokine production in EBV⁺ cells, we suppressed NF-κB using the pharmacologic inhibitors 4-methyl-N1-(3-phenyl-propyl)-benzene-1,2-diamine (JSH-23; an inhibitor of NF-κB nuclear translocation) or BAY-117082 (an IKK inhibitor) or two small interfering RNAs (siRNAs) targeting p65. We found that NF-κB inhibition significantly decreased GM-CSF and VEGF production from CSE-treated EBV⁺ NPC cells (Fig. 4D; Supplementary Fig. S4C). These above data indicate that NF-κB is essential for cytokine production in NPC cells with active EBV replications.

Previous studies show that TAMs promote metastasis of various types of tumors by inducing EMT (17). Thus, we examined whether activated macrophages can induce EMT and increase the motility of NPC cells. We found that CNE2 cells cocultured with macrophages activated by GM-CSF changed from a rounded to an elongated morphology, decreased E-cadherin expression, increased vimentin expression and EMT-transcription factors (snail, twist, and ZEB1; refs. 30–34), indicating that NPC cells underwent EMT (Fig. 5A–C). As EMT is an essential step of metastasis, we tested the motility of the NPC cells in the absence or presence of macrophages in coculture. Consistently, the abilities of NPC cells to migrate and invade were significantly enhanced when cocultured with activated macrophages (Fig. 5D). Furthermore, compared with the negative control IgG antibody, the addition of neutralizing CCL18 antibody to the coculture system suppressed EMT and the motility induced by GM-CSF-activated macrophages. Moreover, CCL18 alone can induce EMT and enhance the migration and invasiveness of NPC cells to a comparable level as activated macrophages (Fig. 5A–E). Similar results were observed in other two NPC cell lines (Supplementary Fig. S5A–S5C). To investigate whether TAM-derived CCL18 induces EMT in clinical samples, we investigated the correlation of CCL18⁺ TAM count and ZEB1 expression of NPC cells in human tumor samples (Fig. 5E; Supplementary Fig. S5D). The numbers of nuclear ZEB1⁺ tumor cells, determined by double immunostaining for ZEB1 and cytokeratin increased in the high CCL18⁺ TAM group compared with the low one (Fig. 5F), and were positively correlated to CCL18⁺ TAM count (Pearson R² = 0.720, P < 0.001, n = 580, Fig. 5G). The above data indicated that CCL18 derived from TAMs promotes NPC EMT and aggressiveness.

CCL18, the key cytokine released by human TAMs, does not have a mouse homolog. Therefore, to mimic the human tumor microenvironment, we evaluated the role of the interaction between EBV⁺ NPC cells and macrophages in a humanized mouse model as previously described (20, 35). CNE2-EBV⁻ cells, CNE2-EBV⁺ or CSE-treated CNE2-EBV⁺ cells were injected into the subcutaneous tissue of irradiated NOD/SCID mice without HSC transplantation and humanized mice, respectively. Consistent with the findings in vitro, injection of CNE-EBV⁺ cells into the subcutaneous tissue of humanized mice led to more CCL18⁺ macrophage infiltration in the xenografts, increased GM-CSF and VEGF production and exhibited EMT phenotype (decrease of E-cadherin, increase of vimentin, and ZEB1) compared with injection of EBV⁻ cells into the humanized mice. In addition, injection of CSE-treated CNE2-EBV⁺ cells further promoted the above effects. In sharp contrast with humanized mice, the xenografts of nonhumanized mice had no CCL18⁺ macrophage infiltration, low GM-CSF and VEGF production and absence of EMT phenotype regardless of EBV⁺ or not (Fig. 6A; Supplementary Fig. S6A–S6F). In addition, EBV replication in NPC cells led to increased metastasis in the lungs of the humanized mice, but not in those of nonhumanized mice, as determined by counting of metastatic colonies in the lungs (Fig. 6B) and quantification of human hypoxanthine phosphor-ribosyl transferase (HPRT) mRNA expression (Fig. 6C). Similar results were observed in the other NPC cell line HNE1 (Fig. 6B and C). To study the potential therapeutic values of targeting inflammatory cytokines in NPC, we treated the humanized mice bearing CSE-treated EBV⁺NPC xenografts with anti-GM-CSF, anti-CCL18, and anti-VEGF–neutralizing antibody. We found that the addition of either antibody could reduce lung metastases, while the combination use of these three neutralizing antibodies had the strongest effect on metastases inhibition (Fig. 6D and E; Supplementary Fig. S6G). The above data suggest that GM-CSF, VEGF, and CCL18 could represent therapeutic targets for inhibiting EBV-infected NPC metastasis.

More than 20% of human cancers are attributed to chronic microbial infection. However, how chronic inflammation promotes tumor progression remains incompletely understood. Our data indicate that NPC with different EBV replications has distinctive capacity to foster a tumor-promoting microenvironment. The NPC cells with active EBV replications recruit monocytes by VEGF and activate monocyte to TAMs by GM-CSF in an NF-κB-dependent manner. Reciprocally, TAMs promote tumor metastasis and further NF-κB activation by CCL18. Collectively, our study demonstrated a feedforward loop between tumor cells and macrophages, suggesting that the metastatic potential may evolve hand in hand with virus-induced chronic inflammation.

Pathogen-induced chronic inflammation plays a central role in various types of tumors. Persistent microbial infections elicit inflammatory cytokine storm. This overactive immune response causes tissue damage and DNA mutation, which subsequently promote tumor progression in various malignancies, such as hepatocellular, cervical, and gastric carcinoma (36, 37). Similarly, it has been reported TAM infiltration is tightly associated with poor prognosis in NPC with EBV infection (38), suggesting the interaction between EBV activity and macrophages may have a role in NPC development. However, despite it is well-documented that EBV is associated with NPC, how EBV-induced inflammation promotes NPC progression remains largely unknown. Our data showed the EBV replication in NPC cells can induce an inflammatory microenvironment by recruiting and activating macrophages. This is consistent with the previous reports on other tumor types. However, the mechanism here seems to be different—through the NF-κB pathway directly activated by EBV replications in tumor cells. Indeed, a number of oncoproteins encoded by EBV can activate NF-κB. For example, Latent membrane protein 1 (LMP1), a major EBV oncoprotein, interacts with TRAFs, and activates both noncanonical and canonical NF-κB through its functional domains termed CTAR1/2/3 (39, 40). Another viral protein termed EBNA1, which enables the replication and segregation of EBV genomes in proliferating infected cells, can increase cell migration via recruitment of the transcription factors Ap1, Sp1, and NF-κB to upregulate matrix metalloproteinase-9 (MMP-9; ref. 41). It is possible that upregulation of viral oncoproteins like LMP1 and EBNA by EBV replications induces NF-κB activation and sequence cytokine production.

GM-CSF and VEGF have individually been identified to be involved in TAM formation (20, 42). Moreover, GM-CSF has also been proved to play a role in the formation of tumor immunosuppressive microenvironment (43). However, GM-CSF and VEGF have not previously been considered for dual targeting. Emerging evidence has validated the concept that inhibition of key signaling pathways of TAM could elicit potent antitumor activities in preclinical tumor models and cancer patients (16). For example, the RG7155 antibody, a humanized anti-human CSF1 receptor mAb, specifically depletes macrophages and exhibits antitumor efficacy in patients of various malignancies (44, 45). However, the presence of tumor-derived GM-CSF–induced RG7155 resistance of TAMs, indicating the therapeutic potential for targeting GM-CSF (44). GM-CSF neutralization is a rapidly expanding market with multiple agents in phase I–II development. These agents are well tolerated and demonstrate efficacy in various chronic inflammatory diseases, including rheumatic arthritis (46). However, the role of GM-CSF neutralization in virus-induced tumor progression has not been reported. In addition, increased VEGF-A expression was observed in EBV⁺ NPC tumors and related to lymph node metastasis and poor long-term survival (47), which is in consistent with our data. Bevacizumab, VEGF-neutralizing antibody, has already become the standard treatment for various cancer types (48, 49). Our study demonstrated that neutralization of GM-CSF and VEGF significantly reduced NPC metastasis in vivo, highlighting the rationales to investigate the therapeutic values of GM-CSF and VEGF-neutralizing antibodies in NPC patients in future clinical trials.

In addition, EBV is also etiologically linked to a wide range of other human tumors, including several B-cell malignancies, extranodal lymphomas of T/NK-cell origin and a distinct subset of gastric carcinomas (50). Our study demonstrated a positive feed-forward loop between EBV-infected tumor cells and macrophages play a crucial role in NPC progression. Whether this vicious cycle in tumor microenvironment also mediates the progression of other malignancies related to EBV infection warrants further investigation.

In summary, our study suggests an important role for the interaction between NPC cells and any step of this interaction, including GM-CSF, CCL18, and VEGF, could be a potential therapeutic target for NPC treatment.

No potential conflicts of interest were disclosed.

Conception and design: S.-J. Song, S.-C. Su

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Huang, S.-J. Song, Z.-Z. Wu, W. Wu, J.-N. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Huang, S.-J. Song, M.-S. Zeng

Writing, review, and/or revision of the manuscript: D. Huang, S.-J. Song

Study supervision: X.-Y. Cui

This work was supported by grants from the Natural Science Foundation of China (81472468, 81672614, 81622036, 81490750, 81372816, 81621004, 81230060, and 81442009), the National Key Research and Development Program of China (2016YFC1302300), Science Foundation of Guangdong Province (2016A030306023, 2014A0303130940, S2012030006287, and 2016B030229004), China Postdoctoral Science Foundation (2014M550449), Guangzhou Science and Technology Project (201710010083), Translational medicine public platform of Guangdong Province (4202037), Tip-top Scientific and Technical Innovative Youth Talents of Guangdong special support program (No. 2016TQ03R553), Guangdong Department of Science & Technology Translational Medicine Center grant (2011A080300002) and Sun Yat-sen University (16ykpy07).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.