Background
Colorectal cancer (CRC) is one of the most common cancers worldwide [
1]. Although many breakthroughs in the diagnosis and treatment of CRC have been made over the past few decades, CRC-related mortality remains high [
2]. As the major cause of death for most cancer patients, tumor metastasis is an important adverse factor in the treatment and prognosis of CRC patients [
3]. Tumor metastasis is a complicated and multistep process which requires the ability of the tumor to migrate and invade [
4,
5]. Despite an increase in our understanding of cell biology and the identification of many metastasis-related molecules [
6], the chemokine-related alterations in the tumor microenvironment that promote CRC metastasis remain largely unknown.
Chemokines are a multifunctional superfamily of small proteins that bind to G protein-coupled receptors on target cells. CXCL5 belongs to a subset of CXC chemokines and functions as the ligand for CXCR2 [
7]. By binding to CXCR2, CXCL5 mediates various cellular behaviors including neutrophil trafficking, tumor cell migration and invasion [
8]. Recent studies have reported that TNF-α-induced mesenchymal stromal cells (MSCs) can secrete CXCL5 to recruit CXCR2+ neutrophils that promote breast cancer metastasis [
9]. These findings indicate that CXCL5 can act as a protumoral molecule in a paracrine way through recruiting immune cells. However, the mechanisms underlying the role of CXCL5 on CRC remain unknown.
In our study, we examined the expression of several chemokines in CRC tissues and found that CXCL5 was highly expressed in CRC tissues. In addition, we investigated the influence of CXCL5 on the survival of CRC patients and showed that CXCL5 levels were negatively correlated with prognosis. Furthermore, we also revealed that CXCL5 secreted by tumor cells was able to promote CRC migration through ERK/Elk-1/Snail-mediated EMT (Epithelial mesenchymal transition) and invasion via the AKT/GSK3β/β-catenin/MMP7 pathway in a CXCR2-dependent manner. Our study conclusively demonstrates that the overexpression of CXCL5 in CRC is able to promote CRC metastasis and predicts a poor outcome in patients with CRC, indicating that CXCL5 may serve as a potential therapeutic target.
Discussion
Chemokines link the extra-tumoral microenvironment to the tumor and facilitate the progression of CRC [
20]. For this reason, we examined the expression of several chemokines in CRC and preliminarily verified the upregulation of CXCL5 in CRC. Previous studies have reported that CXCL5 produced by fibroblasts plays a prominent role in the progression, growth and spread of various types of cancer [
15,
21]. However, in our study, we revealed that CRC cells, rather than fibroblasts, are able to secrete CXCL5. This promotes liver metastasis in vivo and contributes to CRC cell migration via the ERK pathway through the induction of EMT, as well as invasion via the AKT pathway in vitro.
As described above, we first screened the expression profiles of several chemokines. The functions of the most upregulated chemokines have been studied in various cancers, including CRC. We primarily concentrated on CXCL5 in our study. Next, we employed two cohorts of clinical samples to determine the levels of CXCL5 expression. Our results demonstrated that the expression of CXCL5 was clearly up-regulated in CRC tissues. In the second cohort, we linked the results to clinical follow-up data. CXCL5 was significantly correlated with tumor size, Dukes’ stage, tumor invasion, lymph node localization, and liver metastasis. The upregulation of CXCL5 in tumor tissues was also positively correlated with a poor prognosis in CRC patients. CXCL5
high patients had lower OS and DFS rates compared with CXCL5
low patients. In contrast, Frank et al. demonstrated that the absence of CXCL5 expression in tumor tissues was correlated with poor prognosis in CRC, and they also found that CXCL5 was overexpressed in tumor tissues [
22]. They attributed these phenomena to intratumoral infiltration of T cells and neutrophils. However, according to our unpublished data, the neutrophils infiltrating the tumor region are more likely to be N2-type protumoral neutrophils, rather than classical N1-type antitumoral neutrophils. We suspect that the reason for this conflict may be that there were too many stage I and stage II patients (67.1% vs 61.5% in our study) and no stage IV patients, even though there was no significant difference in the cohort distribution. Perhaps the protumoral function of T cell infiltration is dominant only in the early stages of cancer progression [
23]. As the disease progresses, the influence of CXCL5 on tumor cells becomes the dominant cause of cancer. Furthermore, our results revealed that CXCL5 is an independent prognostic factor for DFS, but not for OS. This may be because not all of the patients died from cancer recurrence in the OS group. Some patients may have died from cardiovascular disease. Taken together, these data indicate that CXCL5 is highly expressed in CRC tissues and predicts a poor prognosis for CRC patients.
We also observed that CXCL5 was primarily expressed in tumor lesions, rather than the mesenchyme of cancer tissue, and CRC cell lines were capable of secreting CXCL5. This result indicates that the effect of CXCL5 may likely depend on an autocrine signaling pathway in CRC, rather than being fibroblast-dependent as reported in other types of cancer [
15]. To further understand this process, we constructed CRC cell lines that over-expressed or under-expressed CXCL5 and employed in vivo nude mouse models to elucidate the effects of CXCL5 on CRC liver metastasis. Upregulated expression of CXCL5 increased CRC-derived liver metastasis in the nude mouse model, which was dependent on CXCR2. In line with these results, in vitro experiments also showed that ectopic expression of CXCL5 promoted CRC cell migration and invasion.
EMT is characterized as a dynamic and reversible biological behavior [
19,
24]. When cancer cells escape from the primary site to metastasis, subsets of cancer cells undergo a morphological conversion from an epithelial-like phenotype to a mesenchyme-like phenotype [
25]. This process is characterized by the loss of epithelial cell junction proteins, such as E-cadherin, and an upregulation of mesenchymal markers, such as Vimentin and N-cadherin [
26]. In this study, we found that CRC cells with high levels of CXCL5 expressed low levels of E-cadherin and ZO-1 and high levels of N-cadherin and Vimentin. The inhibition of CXCR2 was able to reverse this process. CXCR2 was initially recognized as a G protein-coupled transmembrane chemokine receptor involved in a variety of pathways, such as the ERK, PI3K/AKT, JNK and STAT3 pathways [
27]. In our study, we verified that both the ERK and AKT pathways can be activated by CXCL5/CXCR2. We next investigated that inhibition of the ERK pathway instead of AKT pathway significantly leads to the alteration of EMT markers. These results indicate that CXCL5/CXCR2 may induce EMT through the ERK pathway. It has been reported that ERKs can activate several downstream transcription factors through phosphorylation to control the expression of specific genes [
28]. Elk-1 is one of those downstream transcription factors [
29]. In our study, we confirmed that inhibition of the ERK pathway leads to the downregulation of pElk-1, which is accompanied by the suppression of Snail.
Snail is an important transcription factor that acts as a repressor of E-cadherin expression and an inducer of EMT in various types of cancer [
30]. Here, we show that Snail is upregulated upon activation of the CXCL5/CXCR2 axis. Furthermore, the knock-down of Snail can reverse CXCL5/CXCR2-induced EMT in CRC cells. Additionally, inhibition of the ERK pathway, rather than the AKT pathway, eliminates the effects of the CXCL5/CXCR2 signaling axis on Snail. These data indicate that CXCL5/CXCR2 induces EMT via the ERK/Elk-1/Snail pathway in CRC cells.
Another important finding of our study was that the activation of the AKT pathway by the CXCL5/CXCR2 axis may be involved in promoting CRC cell invasion. We found that the CXCL5/CXCR2 axis was capable of promoting CRC cell invasion, but the mechanisms behind this were unclear. MMPs are overexpressed in many cancers and are capable of mediating invasion and metastasis via ECM degradation [
31]. It has been reported that MMP7 is a critical factor in the regulation of CRC cell invasion [
32]. In our investigation, we revealed that CXCL5/CXCR2 was responsible for the potentiation of MMP7 expression and the inhibition of the AKT pathway, and the inhibitor LY294002 instead of U0126 interfered with this process. GSK3β is a downstream effector of AKT, and pAKT can suppress the function of GSK3β by phosphorylating its serine residue [
24]. GSK3β is a key component of the destruction complex that facilitates the phosphorylation of β-catenin in CRC [
33]. Phosphorylated β-catenin can be recognized by β-TrCP, which ubiquitinates pβ-catenin, and the ubiquitinated pβ-catenin is then degraded by proteasomes [
34]. Here, we report that activation of the AKT pathway enhances the phosphorylation of downstream GSK3β and the expression of total β-catenin in CRC cells, perhaps because the phosphorylation of GSK3β results in the separation of the destruction complex and stabilization of total β-catenin. Additionally, we also found that the CXCL5/CXCR2 axis is capable of enhancing the translocation of β-catenin from the cytoplasm to the nucleus where β-catenin can bind to several promoters, including MMP7. Moreover, knock-down of β-catenin in CRC cells decreases the number of invading cells. Altogether, these results demonstrate that the CXCL5/CXCR2 axis promotes CRC cell invasion through the AKT/GSK3β/β-catenin/MMP7 pathway.
Although the function of CXCL5/CXCR2 on tumor metastasis is confirmed in vivo. however, it is difficult to translate the data acquired in vivo to human totally. Because there are at least two chemokines, CXCL5 and CXCL6, which are able to bind to CXCR2 in human [
35]. While CXCL5 and CXCL6, also called LIX or GCP-2, are the same chemokine in mice. Human CXCL5 and CXCL6 are homologically related to CXCL5 in mice [
36]. In addition, mice don’t have CXCL8 but CXCL8 and human CXCL6 are the only ligands to bind to both CXCR1 and CXCR2 in human [
37,
38]. Thus upregulating CXCL5 expression in mice may have a more drastic effect than upregulation of only one of the three chemokines in human.
Methods
Chemokine microarray analysis
The profiles of chemokines were examined using a Human Chemokine Antibody Array C1 kit (RayBiotech, USA). This experiment was performed according to the manufacturer’s instructions and supported by the BioTNT Corporation, Shanghai. The levels of certain chemokines in the tumor tissues were more than 2.0-fold higher, which we defined as “significant upregulation”. We also defined a 1.5- to 2.0-fold change as “insignificant upregulation”, a 0.67- to 1.5-fold change as “no difference”, a 0.5- to 0.67-fold change as “nonsignificant downregulation”, and a less than 0.5 change as “significant downregulation”.
Patients and follow-up
Two independent cohorts of colorectal cancer patients are enrolled in our study. All these patients were diagnosed specifically by pathology as CRC (before or after surgery) and treated with laparoscopic surgery in Minimally Invasive Surgery Centre, Ruijin Hospital, Shanghai Jiaotong University. Ethics approval for use of human specimen was obtained from the Biomedical Ethics Committee of Ruijin Hospital. All the tumor tissues and paired peritumoral normal tissues contained in Cohort 1 were collected from patients undergoing operation between Nov.2015 and Mar.2016. Tissues of Cohort 2 were collected from patients undergoing operation from 2010 to 2011. Clinical and pathological data were collected. All the patients who had received preoperative treatment such as radiation or chemotherapy were excluded. Pathological staging of CRC tumor was performed in accordance to the TNM classification [
39]. The follow-up data were acquired at 2-month intervals through outpatient visits, telephone calls, or office visits. The follow-up data were ceased in August 2015.
Cell lines
The human CRC cell lines used in our study were purchased from the American Type Culture Collection (ATCC, USA) and preserved in liquid nitrogen by the Shanghai Digestive Surgery Institute and routinely tested for mycoplasma contamination using PCR. These cell lines were authenticated prior to being used in our study. The SW480 cell line was cultured using Leibovitz’s L-15 medium, and the HCT116 cell line was cultured using McCoy’s 5A medium with 10% FBS, penicillin (107 U/L), and streptomycin (10 mg/L) at 37 °C with 5% CO2 in an incubator.
Animal experiments were performed in accordance with the ethical guidelines issued by the Ethics Committee of Shanghai Jiaotong University. Male BALB/c nu/nu mice (4 weeks old) were purchased from the Chinese Academy of Sciences, Shanghai and raised in a specific-pathogen-free environment. For the construction of the liver metastasis model, 6 nude mice were included in each group (HCT116CXCL5, HCT116shCXCR2, HCT116CXCL5-shCXCR2 and HCT116vector groups as well as SW480shCXCL5, SW480shCXCR2, and SW480shCXCL5-shCXCR2, and SW480shRNA groups were injected into the spleens of the mice). After being anesthetized with an intraperitoneal injection of 1% pentobarbital sodium (50 mg/kg), abdominal surgeries were performed to expose the spleen and slowly inject it with 1×107 cells suspended in 150 μL PBS. The spleen was then returned to the abdominal cavity, and the abdomen was closed. The cells were then transported by the circulatory system to the liver to spontaneously form metastatic lesions. The mice were euthanized by cervical decapitation 6 weeks after injection to examine the liver metastases of tumor cells. Liver metastases were confirmed by HE staining.
Statistical analysis
All of the statistical analysis were performed using SAS 8.0, SPSS 16.0 or R software 3.1.2 (R Core Team). The Pearson χ
2 test and Fisher’s exact probability method were used to analyze the relationship between CXCL5 and clinical features. Overall Survival and Disease Free Survival curves were plotted using the Kaplan-Meier method, and differences between the two groups were determined using a log-rank test. Univariate and multivariate analyses were performed using the Cox regression model. A nomogram was formulated based on the results of the multivariate analysis and plotted using R software. All experiments were performed in triplicate. P < 0.05 was considered to be statistically significant.
Acknowledgements
The authors would like to thank Wolfgang. E. Thasler and Tobias. S. Schiergens (General, Visceral, Transplantation, and Vascular Thoracic Surgery Hospital of University of LMU Munich) for their excellent theoretical and technical assistance. We are very grateful to Cellular and Molecular Biological Institute of Shanghai Jiaotong Univeristy School of Medicine for their technical assistance of confocal microscope. This work was supported in part by a grant from CSC (China Scholarship Council, 201506230091).