A C-X-C Chemokine Receptor Type 2–Dominated Cross-talk between Tumor Cells and Macrophages Drives Gastric Cancer Metastasis

This article is featured in Highlights of This Issue, p. 3197

Gastric cancer is the second leading cause of cancer-related deaths worldwide (1). Most gastric cancer–related deaths result from cancer recurrence and metastasis (2, 3). Novel therapeutic targets for the treatment of gastric cancer metastasis are urgently needed. C-X-C motif chemokine receptor 2 (CXCR2) is a potent protumorigenic chemokine receptor that can induce inflammation in the tumor microenvironment (4, 5). CXCR2-targeted therapies can enhance the efficacy of immunotherapy in several solid tumors (4, 5) and increase the sensitivity of chemotherapy (6). Meanwhile, CXCR2-mediated recruitment of different stromal cells promotes progression of cancer cells (4–7). These studies hinted that the interaction between CXCR2 and tumor microenvironment was of critical importance for tumor progression.

As key regulators of the tumor microenvironment (8), macrophages comprise a large part of primary tumor mass (9). Macrophages can drive the metastasis of gastric cancer cells by increasing the activity of RhoA and Cdc42 in a p-Akt and p-ERK1/2–dependent manner (10). There is a robust correlation between CD163-positive macrophages infiltration and a prognostic gene signature named stromal-response cluster (SRC), which can predict the survival of gastric cancer and ovarian cancer (11). Tumor-associated macrophages (TAM) promote proliferation, invasion, and metastasis of tumor cells by producing a plethora of cytokines (12–15). The chemokines that interact with CXCR2 consist of CXCL1, CXCL2, CXCL3, CXCL5, CXCL6, CXCL7, and CXCL8. Studies have showed that CXCR2 ligands derived from tumor cells can promote metastasis and chemoresistance in breast cancer (6), pancreatic cancer (5), and rhabdomyosarcoma (4). CXCR2 ligands can also drive carcinogenesis and progression of gastric cancer (16–18). Mice that were infected with Helicobacter pylori had a higher level of CXCL1 and CXCL2, which would induce the development of dysplasia and gastric cancer (16). CXCL8-transgenic mice (IL-8Tg) presented with earlier initiation and progression of inflammation-associated gastric carcinogenesis and CXCL8 can induce metastasis of gastric cancer (17, 18). Previously, we found that lymphatic endothelial cells stimulated lymphangiogenesis, angiogenesis, and growth of gastric cancer by secreting CXCL1 (19, 20). CXCL1 expression is an independent prognostic factor in pancreatic ductal adenocarcinoma (PDAC; ref. 21). However, whether chemokines from macrophages can promote the metastasis of gastric cancer remains unclear.

Epithelial to mesenchymal transition (EMT) is the initial step of metastasis in epithelial tumors (22–24). During EMT, epithelial cells acquire the ability to efficiently invade and disseminate (25–27). EMT is often recognized as the result of gene mutation or as a response to the alteration of tumor microenvironment (28, 29). CXCR2 on cancer cells can induce EMT and metastasis of hepatic cancer (30). However, whether chemokines from macrophages contribute to gastric cancer metastasis via CXCR2-mediated EMT remains unknown.

In the current study, we constructed the coculture model to imitate the interaction between tumor cells and macrophages in the tumor microenvironment. We found that CXCR2 expression in cancer cells is a prognostic factor for gastric cancer. CXCL1 and CXCL5 can promote migration of gastric cancer cells by activating CXCR2. Interestingly, we discovered that macrophages are the major sources of CXCL1 and CXCL5 in the gastric cancer microenvironment. Macrophages contribute to tumor migration by activating a CXCR2/STAT3 feed-forward loop in gastric cancer cells. Gastric cancer cells secrete TNF-α to induce release of CXCL1 and CXCL5 from macrophages. Inhibiting the CXCR2 pathway of gastric cancer cells can suppress metastasis of gastric cancer in vivo. These findings indicate that CXCR2 may be a promising therapeutic target to suppress gastric cancer metastasis.

A total of 155 archived human gastric cancer specimens were obtained from the gastric cancer database and tissue bank in the First Affiliated Hospital of Sun Yat-sen University. These tissues are from patients who underwent radical dissection for gastric cancer between 2006 and 2007 without preoperative chemotherapy or radiotherapy. The follow-up was maintained until December 2013. The studies were conducted in accordance with International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). The studies were performed after approval by the Institutional Review Board. We have obtained written consent from the subjects before the study.

Human gastric cancer cell lines (AGS, BGC-823, MGC-803, SGC-7901, HGC-27, and MKN-45), normal gastric epithelial cell line (GES-1), and human acute monocytic leukemia cell line (THP-1) were purchased from the ATCC in 2016. Cells were maintained and cultured according to routine procedures. CXCR2 inhibitor, CXCL1-specific neutralizing antibody, and CXCL5-specific neutralizing antibody were purchased from R&D Systems. CXCR2 ligands (CXCL1, 2, 3, 5, 6, 7, and 8) were purchased as recombinant proteins from PeproTech Inc.

A coculture model of macrophages and gastric cancer cells was established. A transwell apparatus with a 0.4 μm pore membrane 6-well plate was used for cancer cell/macrophage coculture. THP-1 differentiated into macrophages after stimulation with phorbol 12-myristate 13-acetate for 24 hours. Note that 10⁶ cells were seeded to the lower compartment of each well, and an equal number of coculture cells were added onto the upper compartment. After 24 hours, cells in the lower compartment were collected for RNA and protein measurement.

Human CXCR2 siRNA sequences were purchased from RiboBio as shown in Supplementary Table S1. AGS and BGC-823 cells were transfected with either 5 μmol/L of scrambled oligo control, si1, si2, or si3 using Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocols. The expression of CXCR2 was evaluated by Western blot assay (Abcam).

A total of 155 archived gastric cancer tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and then sectioned (5 μm), deparaffinized, rehydrated, and subjected to antigen retrieval. Tissue sections were incubated with CAS blocking buffer and subsequently incubated with primary antibodies (Supplementary Table S2) at 4°C overnight. Reactivity was detected using DAKO EnVision-HRP (Dako). The degree of immunostaining was scored by multiplying the percentage of positive cells (P) with intensity (I), according to the formula: H = P × I. The range for P is 0 to 4 (∼5% scores 0; 5%–25% scores 1; 25%–50% scores 2; 50%–75% scores 3; 75%–100% scores 4). The range for I is 0 to 3 (0, no staining; 1, weak; 2, moderate; 3, strong.). We calculated the staining index score by multiplying P with I and obtained a range from 0 to 12. A staining index score of 0 to 3 was defined as negative, and a staining index score of 4 to 12 was defined as positive. The number of TAMs was calculated from the mean number of CD163-positive cells in 5 random fields of 400× magnification. Two independent pathologists observed the 5 random fields of each tissue and scored each sample without knowledge of the patient outcome. The average value of these 2 scores was presented. A cutoff value of TAMs relating to the prognosis of patients with gastric cancer was calculated according to established procedures. The cutoff number for TAM (CD163 positive) is 14.9.

Human macrophages differentiated from THP-1 cells were cultured alone or in combination with gastric cancer cells for 24 hours. Total RNA was extracted using RNAplus reagent (Takara), according to the manufacturer's instructions. After treatment with RNase-free DNase, 1 μg of total RNA from each sample was used for cDNA synthesis using a Reverse Transcription kit (Takara). Quantitative reverse-transcriptase PCR (qRT-PCR) was performed with the 7900 HT (Applied Biosystems) using SYBRVR Green qPCR SuperMix (Invitrogen). Primers were designed and synthesized by Sangon Biotech. Sequences are shown in Supplementary Table S1. The expression of mRNA was normalized to the geometric mean of housekeeping gene GAPDH to control variability in expression levels. Results were expressed as the fold change using the 2^−ΔΔCt method.

In vitro cell viability was evaluated by the Cell Counting Kit-8 (CCK-8) assay. Four thousand cells were seeded into the wells of a 96-well plate. After 16 hours of culture, the medium was replaced with a low serum medium. Then, cells were cultured with or without conditioned medium for 72 hours. Subsequently, the medium in each well was replaced with 100 μL RPMI-1640 medium containing 10-μL CCK-8 reagent (Dojindo Laboratories), and the absorbance was measured after 3 hours using a microplate reader at 450 nm.

Cell lysates and tumor lysates were collected using the whole protein extraction Kit (Keygen). Supernatants were recovered by centrifugation at 13,000 rpm for 15 minutes at 4°C. Protein concentrations were measured, and equal amounts of total protein were separated by SDS-PAGE. Proteins were transferred onto PVDF membranes (Merck Millipore), and the membranes were blocked for 1 hour in TBST. Then, the membranes were incubated overnight at 4°C with primary antibodies (Supplementary Table S2). After washing with TBST, the membranes were incubated with corresponding peroxidase-conjugated secondary antibodies for 1 hour at room temperature. Specific bands were detected using an enhanced chemiluminescence reagent (ECL; Perkin Elmer Life Sciences) on autoradiographic film.

The shCXCR2 RNAi lentivirus was constructed by Laura Biotech. BGC-823 cells (5 × 10⁵) were cultured in 25 cm² dishes. After culturing for 24 hours, cells were incubated with 5 mL of medium containing lentivirus and 5-μL polybrene. This procedure was repeated twice a day for 2 days. Infected cells were treated with 0.5 μg/mL of puromycin for 1 week. Cells stably expressing the shCXCR2 were isolated, and CXCR2 expression was determined by Western blot and q-PCR. The shCXCR2 sequence is shown in Supplementary Table S1.

Supernatant was collected from control cells (1 × 10⁶ AGS or BGC-823 or THP-1 cells) cultured in serum-free media in the presence or absence of cells (1 × 10⁶ AGS or BGC-823 or THP-1 cells). The concentration of CXCL1 and CXCL5 was measured by ELISA (R&D Systems).

A transwell migration was used to assess cell migration. AGS or BGC-823 cells (1 × 10⁵) were treated with mitomycin C (10 μg/mL, Sigma) for 2 hours before being seeded onto the upper compartment of Matrigel-coated transwell chambers (24-well insert, 8-μm pore size; BD Biosciences). Macrophages (1 × 10⁵) were cultured in the lower chamber. Tumor cells were allowed to migrate for prespecified times at 37°C. The number of cells that migrated to the underside of the transwell was counted in 5 high-power fields at 200× magnification. Data were normalized using a migration index.

Animal experiments were approved by Institutional Animal Care and Use Committee of the First Affiliated Hospital of Sun Yat-sen University. Six-week-old female Balb/c athymic mice were purchased from the Guangdong Medical Laboratory Animal Center. Mice were randomized into 4 groups (n = 5 per group). We constructed subcutaneous xenograft model by injecting gastric cancer cells alone (3 × 10⁶ cells/mouse) or combined with macrophages (1 × 10⁶ cells/mouse). Therefore, we have 4 groups: group A is shV gastric cancer cells, group B is shV gastric cancer cells plus macrophages, group C is shCXCR2 gastric cancer cells, and group D is shCXCR2 gastric cancer cells plus macrophages. Tumor size was measured twice a week, and tumor volume (V) was calculated by using the formula: V = 0.5 × (length × width²). After 19 days, all mice were euthanized with CO₂; and tumors were removed, weighed, and processed for IHC and Western blot. In the tumor metastasis experiment, nude mice (4 weeks, male) were obtained from the Shanghai Public Health Clinical Center (Shanghai). Mice were randomly assigned to 2 different groups (n = 6 per group). BGC-823-shCXCR2 or BGC-823 shV cells (3 × 10⁶/mouse) were injected subcutaneously into the footpad of the left hind limb of each mouse. Both of these 2 cell lines were stable cell lines which would express firefly mCherry and GFP. The mice were euthanized on day 28, and the primary tumors and popliteal lymph nodes were collected. Popliteal lymph nodes were evaluated for metastatic lesions by hematoxylin and eosin staining and IHC staining with anti-mCherry or anti-GFP antibodies.

SPSS 17.0 (SPSS Inc.) software was used for statistical analysis. Quantitative data were presented as mean ± SD, and the Student t test was applied. The χ² test was utilized to assess the associations of the expression of CD163, CXCR2, CXCL1, and CXCL5 with clinicopathologic features. The relationship among the expression of proteins was analyzed by the Fisher exact test. Survival curves and overall survival (OS) rates were determined by Kaplan–Meier and log-rank methods. Two-sided P values <0.05 were considered statistically significant.

IHC analyses were performed on gastric cancer tissues. Among the 155 gastric cancer tissues, 56.1% (87/155) displayed strong expression of CXCR2 (Supplementary Fig. S1A; Supplementary Table S3). Positive CXCR2 expression was correlated to inferior prognosis in gastric cancer (Fig. 1A). We found that CXCR2 expression was higher in human gastric cancer tissue than that in paired normal gastric tissue by Western blot (Supplementary Fig. S1B). Accordingly, CXCR2 expression was higher in 6 different gastric cancer cell lines compared with that in a normal gastric epithelium cell line (GES-1; Supplementary Fig. S1C). Then we examined whether knocking down CXCR2 can suppress the progression of gastric cancer. We found that knocking down CXCR2 can suppress the migration of gastric cancer cells (Fig. 1B). Meanwhile, knocking down CXCR2 can significantly suppress cell proliferation (Fig. 1C).

Because CXCR2 has several ligands, we further investigated which ligands can promote the progression of gastric cancer. The expression of all CXCR2 ligands (including CXCL1, CXCL2, CXCL3, CXCL5, CXCL6, CXCL7, and CXCL8) in gastric cancer was analyzed in The Cancer Genome Atlas (TCGA) datasets (http://cancergenome.nih.gov) and 2 public datasets from GEO databases (GSE27342 and GSE29272). Gene expression was calculated by subtracting expression values of adjacent normal gastric epithelial tissue from that of the paired gastric cancer tissue. The results showed that most of gastric cancer tissue had higher level of CXCR2 ligands than the adjacent normal tissue (Supplementary Fig. S1D). We also analyzed the mRNA level of CXCR2 ligands in 8 pairs of gastric cancer tissue and adjacent tissue. The result showed that the level of CXCR2 ligands was higher than in gastric cancer tissue (Fig. 1D; Supplementary Fig. S2A), which was consistent with the data in the public databases (Supplementary Fig. S1D). To address the relative contribution of different CXCR2 ligands with respect to migration and EMT in gastric cancer, cells were treated with human recombinant proteins of CXCL1, CXCL2, CXCL3, CXCL5, CXCL6, CXCL7, or CXCL8. Among all the CXCR2 ligands, we observed that CXCL1 and CXCL5 significantly increased gastric cancer cell migration (Fig. 1E; Supplementary Fig. S2B). We examined whether expression of CXCL1 or CXCL5 in tumor tissue is associated with prognosis. The result showed that positive expression of CXCL1 can predict inferior prognosis in gastric cancer (5-year OS: 65.9% vs. 40.6%, P = 0.002). Similar result was also found in CXCL5 (5-year OS: 71.8% vs. 26.9%, P<0.001; Fig. 1F). Further study showed that CXCL1 and CXCL5 expression was positively associated with tumor metastasis and tumor–node–metastasis (TNM) stage (Supplementary Table S4). We also analyzed data from public databases (https://hgserver1.amc.nl/cgi-bin/r2/main.cgi). The results also showed that high expression of CXCL1 and CXCL5 in tumor tissue was associated with poor prognosis in lung cancer, neuroblastoma, and PDAC (Supplementary Fig. S3).

Next, we explored the major sources of CXCR2 ligands in gastric cancer tissue. We detected the mRNA level of CXCR2 ligands in gastric cancer cell lines and GES-1 cells by qPCR. The results showed that the mRNA level of most CXCR2 ligands did not increase in gastric cancer cell lines significantly (Supplementary Fig. S4A). These findings indicated that increased CXCR2 activity in gastric cancer cells was more likely the result of paracrine activation by ligands secreted from stromal cells in the gastric cancer microenvironment. However, whether macrophages can promote tumor progression via secreting CXCR2 ligands remains elusive. The clinical relevance of these findings was evaluated by determining the expression of CXCL1 and CXCL5 in cancer tissues in a cohort of 155 patients with gastric cancer. CXCL1 expression was high in 56.1% (87/155) patients with gastric cancer, whereas CXCL5 expression was high in 45.2% (70/155) patients with gastric cancer (Fig. 2A; Supplementary Table S4). CD163 expression positively correlated with CXCL1 expression (r = 0.340, P < 0.001) and CXCL5 expression (r = 0.300, P < 0.001; Fig. 2B). We detected the mRNA level of CXCL1 and CXCL5 in macrophages cocultured with normal gastric epithelial cells or gastric cancer cells by qPCR. To our surprise, the mRNA levels of CXCL1 and CXCL5 in macrophages increased dramatically when cocultured with gastric cancer cells (Fig. 2C). Meanwhile, the mRNA level of other CXCR2 ligands besides CXCL1 and CXCL5 also increased dramatically in macrophages cocultured with gastric cancer cells (Supplementary Fig. S4B). These results indicated that macrophages may also secrete high levels of CXCL1 and CXCL5 when cocultured with gastric cancer cells. To validate this hypothesis, we performed ELISA analysis to detect the levels of CXCL1 and CXCL5 in the supernatant of gastric cancer cells and macrophages when they were cultured alone or in the coculture system. We found that the levels of CXCL1 and CXCL5 in the supernatant of gastric cancer cells and macrophages increased in the coculture system (Fig. 2D and E). Intriguingly, we found that the level of CXCL1 and CXCL5 in the supernatant of macrophages increased more dramatically than that of gastric cancer cells. These results indicate that TAMs have higher level of CXCL1 and CXCL5 than that in gastric cancer cells.

Then we examined whether macrophages can promote the progression of gastric cancer via CXCL1 and CXCL5. The migration of gastric cancer cells increased when cocultured with macrophages (Fig. 3A). CXCR2 inhibitor SB225002 and CXCL1- or CXCL5-neutralizing antibodies can inhibit macrophage-induced migration of gastric cancer cells (Fig. 3B and C; Supplementary Fig. S6A). We investigated how CXCL1 and CXCL5 promote the migration and progression of gastric cancer. Previously, we found that CXCL1 promoted the migration and progression of gastric cancer via activating STAT3/VEGF pathway (20). So we detected the phosphorylation of STAT3 in gastric cancer cells treated with CXCL1 and CXCL5. As expected, CXCL1 and CXCL5 can increase the phosphorylation of STAT3 (Fig. 3D). Interestingly, when we knocked down STAT3, CXCR2 expression decreased (Supplementary Fig. S6E), and when we overexpressed CXCR2, phosphorylation of STAT3 increased (Fig. 3F). Furthermore, knocking down STAT3 can reverse CXCL1-induced upregulation of CXCR2 (Fig. 3E). These results indicated that there is a feed-forward loop between CXCR2 and STAT3. On the one hand, CXCR2 can increase the phosphorylation of STAT3. On the other hand, STAT3 can induce the upregulation of CXCR2.

Next, we determined how CXCR2/STAT3 activation can promote the migration and progression of gastric cancer. We performed gene set enrichment analysis (GSEA) using data from TCGA (http://cancergenome.nih.gov), and found that CXCR2 levels correlate with STAT3-activated gene signatures and EMT-associated gene signatures (Supplementary Fig. S5). These observations indicate that CXCR2/STAT3 signaling may induce EMT in gastric cancer. We examined whether CXCL1 and CXCL5 can induce EMT of gastric cancer via activating CXCR2/STAT3. We detected the expression of EMT biomarkers via Western blot. The results showed that knocking down CXCR2 in gastric cancer cells decreased phosphorylation of STAT3 and suppressed EMT (Fig. 4A). Consistently, we found that coculturing with macrophages can increase the phosphorylation of STAT3 and induce EMT in gastric cancer cells, whereas CXCR2 inhibitor and CXCL1- or CXCL5-neutralizing antibodies can reverse it (Fig. 4B and C). Later, we found that knocking down CXCR2 or pretreating gastric cancer cells with STAT3 pathway inhibitor AG490 (2 μmol/L) for 4 hours can reverse macrophage-induced proliferation and migration of gastric cancer cells (Fig. 4D; Supplementary Figs. S6B and S8B). These results indicate that macrophages derived CXCL1 and CXCL5 can activate CXCR2/STAT3 pathway in gastric cancer cells, resulting in the progression of gastric cancer cells.

We constructed shCXCR2 stable cell line (Supplementary Fig. S6C and S6D). BGC-823 cells expressing a specific shRNA for CXCR2 or shV were implanted subcutaneously with or without macrophages in nude mice. Mean tumor weight was 1.3196 ± 0.3342 g, 1.1484 ± 0.14617 g, 0.5118 ± 0.1589 g, and 0.5528 ± 0.1555 g in the shV, shV+TAMs, shCXCR2, and shCXCR2+TAMs group, respectively (Fig. 5A and B). shCXCR2 suppressed tumor growth significantly (P < 0.05 on day 19; Fig. 5C). Furthermore, knocking down CXCR2 decreased expression of p-STAT3 and EMT markers Snail (Fig. 5D). These results revealed that the function of the CXCR2/STAT3 signaling pathway was responsible for the induction of tumor growth and EMT in gastric cancer. To further evaluate the role of CXCR2 in tumor metastasis, we established the lymph node metastasis mice model. We found that shCXCR2 decreased lymph node metastasis compared with the shV group. The lymph node metastasis rate was 83.3% in the shV group and 33.3% in the shCXCR2 group (Fig. 5E). In addition, we also constructed CXCL1-overexpressed gastric cancer cell line. CXCL1-7901 (5 × 10⁶) and GFP-7901 cells (5 × 10⁶) were subcutaneously injected into nude mice. Tumor growth was measured every 2 to 3 days for 25 days. The results showed that CXCL1 increased local tumor growth and the expression of Snail in xenograft nude mice (Supplementary Fig. S7A and S7B).

We explored whether cytokines secreted from gastric cancer could stimulate release of CXCL1 and CXCL5 from macrophages. First, we tested whether TGF-β secreted from gastric cancer cells can increase the expression of CXCL1 and CXCL5 in macrophages. However, we did not find that TGF-β increased the expression of CXCL1 and CXCL5 in macrophages (data not shown). Interestingly, we found that TNF-α levels increased in gastric cancer cells when cocultured with macrophages (Fig. 5F). TNF-α can induce the upregulation of CXCL1 and CXCL5 in macrophages (Fig. 5G). Blocking the downstream pathway of TNF-α by p38 inhibitor and p65 inhibitor can decrease the level of CXCL1 and CXCL5 in macrophages (Fig. 5G). These results indicated that gastric cancer cells can interact with macrophages via secreting TNF-α, and macrophages promote gastric cancer progression through CXCL1 and CXCL5.

Expression of p-STAT3, Snail, and CD163 was analyzed in 155 gastric cancer tissues by IHC (Figs. 2A and 6A). CXCL1 and CXCL5 expression positively correlated with the expression of Snail, p-STAT3, indicating CXCL1 and CXCL5 can drive EMT of gastric cancer cells (Fig. 6B and C). Meanwhile, the expression of CD163 and CXCR2 was also positively associated with the expression of Snail (Fig. 6D), suggesting macrophages are responsible for the activation of CXCR2 pathway in gastric cancer cells. Macrophages and the expression of Snail in gastric cancer cells predicted inferior prognosis (Supplementary Fig. S8A). These results strongly suggest that macrophages can promote EMT and progression of gastric cancer via activating CXCR2 pathway (Fig. 6E). CXCR2 expression is an independent prognostic factor for gastric cancer patients (Supplementary Table S5).

The tumor microenvironment contributes to tumorigenesis, progression, and dissemination (31–33). TAMs promote EMT and metastases in several tumors including breast cancer and pancreatic cancer (13, 34, 35). CXCR2 is a key regulator for the interaction between tumor cells and stromal cells. There is emerging literature that suggests that CXCR2 is fundamental for driving tumor metastasis, and blockade of CXCR2 can significantly disrupt the stromal–tumor interaction in several solid tumors (5, 21). Targeting CXCR2 can suppress immune evasion of tumor cells, thus increasing the sensitivity of immunotherapy (4, 5). However, the role of CXCR2-mediated interaction between macrophages and tumor cells remains uncharacterized. Previously, we have shown that CXCL1/CXCR2 signaling pathway is important for the progression of gastric cancer (20, 36). In this study, we sought to characterize the role of CXCR2 in gastric cancer metastasis and unravel the mechanism of CXCR2-mediated cross-talk between macrophages and tumor cells. In this study, we found that macrophages promote EMT and metastases of gastric cancer cells through the release of CXCL1 and CXCL5, which activates a positive feed-forward loop between CXCR2 and STAT3 in gastric cancer cells. Most importantly, the elements involved in this cascade showed a substantial prognostic correlation in a cohort of patients with gastric cancer. Knocking down CXCR2 decreased EMT and metastasis in vivo, highlighting the clinical significance of these findings and suggesting that CXCR2 may serve as a rational target for gastric cancer treatment.

Through the analysis of chemokine profiles of TAMs, we found that CXCL1 and CXCL5 are critical for gastric cancer metastasis and associated with patient prognosis. TAMs secrete a number of cytokines like CXCL1, IL6, TGF-β, and VEGF, which promote tumor growth and metastasis (15, 37–39). Of all the cytokines secreted by macrophages, we mainly focus on CXC chemokine because there is growing evidence suggesting that CXC chemokine can drive inflammation and treatment resistance, which will result in the progression of cancer (40, 41). However, the role of CXC chemokine derived from macrophages remains uncharacterized. A recent study showed that macrophage-secreted CXCL1 can promote the metastasis of breast cancer by activating SOX4 signaling pathway (42). Our previous studies have suggested that CXC chemokine receptors might serve as potential therapeutic targets in gastric cancer (36). Thus, we sought to identify which CXC chemokine are responsible for driving the metastasis of gastric cancer. In prostate cancer, TAMs induce EMT through activating the CCL2/CCR2-STAT3 signaling (15). Based on the GSEA plot of the EMT process, we quantified Snail as an EMT marker for gastric cancer. We demonstrated that TAMs induced EMT of gastric cancer cells in a CXCR2-dependent manner. Then, we screened all CXCR2 ligands and found that CXCL1 and CXCL5 from TAMs can induce EMT and promote metastasis. Clinical data also showed that the expression of CXCL1 and CXCL5 was associated with tumor metastasis and TNM stage, which may serve as prognostic biomarkers for patients with gastric cancer. Thus, our findings revealed a novel clinically relevant mechanism in gastric cancer metastasis, which suggested that TAMs induced EMT and metastasis of gastric cancer by activating the CXCL1/5–CXCR2 axis.

The cross-talk between tumor cells and macrophages also exists in other cancers. CCL18 released by TAMs enhanced EMT and metastasis of breast cancer cells (34). The CXCL5/CXCR2 axis contributes to the EMT of hepatic cancer cells (30). CXCL1/2 attracts CD11b(+)Gr1(+) myeloid cells to enhance breast cancer cell survival and chemoresistance (6). In addition, cancer cells also secrete cytokines to attract macrophages into the tumor, which in turn contributes to EMT (15, 29). Here, we found that gastric cancer cells–secreted TNF-α can stimulate the release of CXCL1 and CXCL5 from macrophages into the tumor microenvironment. CXCR2 can activate PI3K-Akt, NF-kB, and MAPK signaling pathways (6, 30, 43). CXCL1 promoted proliferation, and migration of gastric cancer cells is partially dependent on N-Ras (44). We previously demonstrated that the CXCL1/CXCR2 axis increased angiogenesis and tumor growth through activating STAT3/VEGF pathway in gastric cancer (20). Here, we found that CXCR2-mediated STAT3 pathway can be activated by CXCL1 and CXCL5, which increases expression of Vimentin, N-cadherin, and Snail, thus leading to EMT and metastasis. STAT3 is considered to be a putative downstream effector of the CXCL1/CXCR2 pathway (45), which can induce EMT (46) and promote invasiveness (47). Previously, we have demonstrated that STAT3 binds to the promoter of CXCR2 and increases CXCR2 expression (36). Thus, a positive feed-forward loop between CXCR2 and STAT3 forms a cascade to continuously induce EMT. Interestingly, a recent study showed that knocking down Snail can significantly decrease the expression of CXCR2 ligands, indicating that there may be a feed-forward loop between Snail and CXCR2 ligands (48). Consistent to our finding, a recent study showed that TAMs derived CXCL1 can drive metastasis of breast cancer (42). Anti-CXCR2 treatment can also increase the sensitivity of immunotherapy in several tumors (4, 5, 49). CSF-1 is a critical regulator for the infiltration and differentiation of macrophages (49). Reducing TAMs by blocking CSF-1 can enhance the antitumor efficacy of immune checkpoint blockade (50). Combination of CSF-1 inhibitor and CXCR2 inhibitor can dramatically increase the antitumor effect of PD-1 blockade in lung carcinoma and melanoma mice models (49). Targeting CXCR2 can also enhance the efficacy of anti–PD-1 therapy by disrupting the infiltration of myeloid-derived suppressor cells (4). Anti-CXCR2 therapy seems to be a promising strategy for those with a positive expression of CXCR2 in gastric cancer. Thus, further studies to evaluate therapeutic potential of CXCR2 inhibition in gastric cancer are warranted. Meanwhile, our study may also advance our understanding on the interaction between immune cells and cancers, which may be helpful for developing novel immunotherapy strategies.

In conclusion, we revealed a previously uncharacterized cross-talk between gastric cancer cells and macrophages. Gastric cancer cells secrete TNF-α to induce release of CXCL1 and CXCL5 from macrophages. CXCL1 and CXCL5 secreted by macrophages could activate CXCR2-mediated positive feed-forward loop to induce EMT and tumor metastasis in gastric cancer, thus representing promising therapeutic targets in gastric cancer.

No potential conflicts of interest were disclosed.

Conception and design: G. Xia, Y. He, M. Li, C. Zhang

Development of methodology: Z. Xiang, J. Zhu, K.-M. Fung, Y. He

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Z. Zhou, G. Xia, Z. Xiang, Z. Wei, X. Sun

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Z. Zhou, G. Xia, Z. Wei, J. Yan, N. Awasthi, K.-M. Fung, M. Li

Writing, review, and/or revision of the manuscript: Z. Zhou, M. Liu, W. Chen, N. Awasthi, X. Sun, K.-M. Fung, M. Li, C. Zhang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Liu, W. Chen, X. Sun

Study supervision: M. Liu, M. Li, C. Zhang

The authors would like to thank Professor Wen Li (Department of Surgery, the First Affiliated Hospital of Sun Yat-sen University) for valuable comments on this article and Professor Qiao Su (Laboratory Animal Center, the First Affiliated Hospital of Sun Yat-sen University) for technical assistance. This study was supported by the National Natural Science Foundation of China (grant No. 81272643 and grant No. 81272637), the “3 & 3” project of The First Affiliated Hospital of Sun Yat-sen University, the National Natural Science Foundation of Guangdong Province (grant No. c15140600000016), the higher education basic research foundation of Sun Yat-sen University (grant No. 17ykjc11), and International Program for Ph.D. Candidates, Sun Yat-Sen University.

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.