Background
Despite great improvements in early detection, surgical techniques, chemotherapy, radiotherapy, biological treatment and multidisciplinary treatment in recent years, cancer is still a major public health problem globally, which is associated with high morbidity, mortality and economic burden [
1]. It is estimated that 1,735,350 new cancer cases and 609,640 cancer deaths are projected to occur in the United States in 2018 [
2]. Given the poor prognosis of cancer patients, numerous investigators have focused on searching for biomarkers that could predict prognosis of cancer. However, sensitivity and specificity of most cancer biomarkers widely used now are not yet satisfactory [
3]. Therefore, it is desperately needed to identify novel applicable prognostic biomarkers, not only improving poor prognosis but also providing novel therapeutic targets.
Chemokines are chemotactic cytokines that could regulate the migration of immune cells into damaged or diseased organs in response to pro-inflammatory stimuli [
4]. According to cysteine residues in the NH2-terminal part of the protein, chemokines can be classified into four highly conserved groups, namely C, CC, CXC, and CX3C [
5]. Chemokines and their receptors could bring about the transcription of target genes involved in cell invasion, motility, survival and interactions with the extracellular matrix, which can induce migration, chemotaxis and rearrangement of the cytoskeleton in the target cell, and therefore promote multiple physiological functions of cells, including cell growth, development, differentiation and apoptosis [
6‐
9]. Over the past few years, accumulating evidence has revealed that chemokines play pivotal roles in progression of tumor [
10]. Chemokines produced by tumor and stromal cells can induce the expression and distribution of tumor-associated leukocytes, trigger angiogenesis, contribute to the growth and metastasis of malignant cells and generate fiber keratinocytes [
6,
11,
12]. In addition, chemokines and their receptors are critical mediators of inflammation microenvironment of cancer, which has been proposed to represent the seventh hallmark of cancer [
13,
14]. Given the important roles of chemokines in cancer, abnormal expression of chemokines has been detected in many tumors, and several chemokines have been proven to be associated with poor prognosis of cancer patients [
15‐
17].
CXCL5, also known as epithelial-derived neutrophil-activating peptide 78 (ENA78), is originally discovered as a potent chemoattractant and activator of neutrophil function. Through binding to its receptor CXCR2, CXCL5 could induce the chemotaxis of neutrophils, promote angiogenesis, and remodel connective tissue [
18]. Accumulating evidence suggests that CXCL5 may participate in cancer-related inflammation, which is involved in many aspects of malignancy in cancer biology [
19]. Furthermore, abnormal expression of CXCL5 has been identified in many tumors. CXCL5 is overexpressed in gastric cancer, prostate cancer, endometrial cancer, squamous cell cancer, hepatocellular carcinoma and pancreatic cancer, and increased expression of CXCL5 is associated with advanced tumor stages, local invasion, neutrophil infiltration and metastatic potential [
20‐
25]. Recent studies have revealed that CXCL5 could serve as a potential prognostic biomarker for patients with cancer [
5,
19,
26,
27]. However, its prognostic value is still controversial owing to the fact that most studies reported so far are limited in discrete outcome and sample size. Therefore, we performed the current quantitative meta-analysis to elucidate the prognostic significance of CXCL5 expression in cancer patients.
Materials and methods
Study strategy
The present review was performed in accordance with the standard guidelines for meta-analysis and systematic reviews of tumor marker prognostic studies [
28,
29]. The database Web of Science, PubMed and Embase were independently searched by two researchers (Binwu Hu and Huiqian Fan) to obtain all relevant articles about the prognostic value of CXCL5 in patients with any tumor. The literature search ended on March 1, 2018. The search strategy used both MeSH terminology and free-text words to increase the sensitivity of the search. The search strategy was: “CXCL5 or CXC chemokine ligand 5 or ENA78 or epithelial cell derived neutrophil attractant 78” AND “cancer or tumor or carcinoma or neoplasm or malignancy” AND “prognostic or prognosis or survival or outcome”. We also screened the references of retrieved relevant articles to identify potentially eligible literatures. Conflicts were solved through group discussion.
Inclusion and exclusion criteria
Studies included in this analysis had to meet the following inclusion criteria: (1) patients were pathologically diagnosed with any type of human cancer. (2) CXCL5 expression levels were determined in human tissues or plasma samples. (3) Patients were divided into two groups according to the expression levels of CXCL5, the relationship between CXCL5 expression levels with survival outcome was investigated. (4) Sufficient published data or the survival curve were provided to calculate hazard ratios (HR) for survival rates and their 95% confidence intervals (CI). Exclusion criteria were as follow: studies using non-human samples, studies without usable or sufficient data, laboratory articles, reviews, letters, case reports, non-English or unpublished articles and conference abstracts. All eligible studies were carefully screened by two researchers (Binwu Hu and Huiqian Fan), and discrepancies were resolved by discussing with a third researcher (Xiao Lv).
Two investigators (Binwu Hu and Huiqian Fan) extracted relevant data independently and reached a consensus on all items. For all eligible studies, the following information of each article was collected: author, year of publication, tumor type, samples detected, expression associated with poor prognosis, Newcastle–Ottawa Scale (NOS) score, method of obtaining HRs, characteristics of the study population (including country of the population enrolled, number of patients (high/low), follow up (month)), endpoints, assay method, cut-off value and survival analysis. For endpoints, overall survival (OS), disease-free survival (DFS), progression-free survival (PFS) and recurrence-free survival (RFS) were all regarded as endpoints. We employed HR which was extracted following a methodology suggested previously to evaluate the influence of CXCL5 expression on prognosis of patients [
30]. If possible, we also asked for original data directly from the authors of the relevant studies.
Quality assessment
Quality of all included studies was assessed independently by two researchers (Binwu Hu and Huiqian Fan) using the validated Newcastle–Ottawa Scale, and disagreements were resolved through discussion with another researcher (Songfeng Chen). This scale uses a star system to evaluate a study in three domains: selection of participants, comparability of study groups, and the ascertainment of outcomes of interest. We considered studies with scores more than 6 as high-quality studies, and those with scores no more than 6 as low-quality studies.
Statistical analysis
Statistical analysis was performed using Stata Software 14.0 (Stata, College Station, TX). Pooled HRs (high/low) and their associated 95% CIs were used to analyze the prognostic role of CXCL5 expression in various cancers. The heterogeneity among studies was evaluated using Cochran’s Q and I2 statistics. A p value less than 0.10 or an I2 value larger than 50% were considered statistically significant. The fixed-effect model was used for analysis without significant heterogeneity between studies (p > 0.10, I2 < 50%). Otherwise, the random-effect model was chosen. To explore the source of heterogeneity, subgroup analysis and meta-regression were preformed through classifying the included studies into subgroups according to similar features. We also conducted sensitivity analysis to test the effect of each study on the overall pooled results. The publication bias was evaluated by using both Begg’s test and Egger’s test. A p value less than 0.05 was considered statistically significant.
Discussion
CXCL5 is originally discovered as a potent chemoattractant and activator of neutrophil function [
33]. Through interacting with CXCR2 receptor, it could function both as a chemoattractant and as an angiogenic factor [
35,
38,
39]. Recently, CXCL5 has been shown to be able to promote the proliferation, migration and invasion of various tumor cells and play pivotal roles in the pathogenesis and progression of cancer [
27,
37]. It was reported that CXCL5 protein was higher in various lung cancer tissues, which was positively associated with tumor stage, lymph node metastasis, and worse survival [
26] [
19,
40]. Zhou et al. also reported that CXCL5 was overexpressed in intrahepatic cholangiocarcinoma cell lines and tumor samples, which could promote intrahepatic cholangiocarcinoma growth and metastasis by recruiting intratumoral neutrophils [
27]. Furthermore, CXCL5 could directly induce endothelial cell proliferation and invasion in vitro and promote tumor angiogenesis in non-small cell lung carcinoma and pancreatic cancer [
41‐
43]. Considering the important functions of CXCL5 in cancer, studies have demonstrated that CXCL5 could serve as a potential prognostic biomarker for cancer patients. However, the prognostic value of CXCL5 is still controversial. Because even in the same type of tumor, there are almost opposite conclusions about the prognostic value of CXCL5 [
34‐
36].
Here we performed the current comprehensive meta-analysis to systematically explore the prognostic value of abnormally expressed CXCL5 in cancer patients. We examined 15 independent studies including 19 cohorts and 5070 patients. Through systematic analysis, our results demonstrated that high expression level of CXCL5 was significantly associated with poor OS in cancer patients. Due to the significant heterogeneity across these studies, we performed subgroup analysis and meta-regression analysis to explore the sources of heterogeneity. The results of subgroup analysis suggested that sample size (fewer than 100 or more than 100) altered the significance of prognostic role of CXCL5 in OS (HR 1.60, 95% CI 0.81–3.17 vs HR 1.69, 95% CI 1.37–2.08). This indicated that difference in sample size might be the source of heterogeneity. However, meta-regression analysis failed to identify the source of the significant heterogeneity in above covariates. In addition, by combining HRs from Cox multivariate analysis, we found that CXCL5 was an independent prognostic factor of OS in cancer patients.
Furthermore, we evaluated the prognostic value of CXCL5 in certain types of cancer. We found that high CXCL5 expression was associated with reduced OS in intrahepatic cholangiocarcinoma and hepatocellular carcinoma, which was consistent with previous studies. However, there was no significant association between expression level of CXCL5 with the OS of lung cancer and colorectal cancer. For lung cancer, results from Oksana et al. were contrary to others greatly [
32]. The reason might be that they only evaluated the prognostic value of CXCL5 in early stage non-small cell lung cancer (stages I and II) [
32]. Similarly, for colorectal cancer, the results from Speetjens et al. also conflicted with others because they did not include stage IV patients [
35,
36]. Therefore, we may speculate that CXCL5 might have different prognostic roles in different tumor stage and larger-scale, multicenter studies including all stage patients are needed to verify our hypothesis.
DFS, PFS and RFS are all important parameters reflecting the progression of tumor. Our results demonstrated that higher expression level of CXCL5 was significantly associated with shorter PFS and RFS in cancer patients. However, high or low expression of CXCL5 made no difference in predicting the DFS of cancer patients. In addition, because only three studies respectively were included to evaluate the association between CXCL5 expression levels with DFS, PFS and RFS, more studies are necessary to explore the relationship between CXCL5 with tumor progression.
Mechanisms underlying the regulatory role of CXCL5 in tumorigenesis and tumor progression have been extensively investigated. CXCL5 could activate multiple signaling pathways to promote the progression of cancer. Dai et al. found that overexpression of CXCL5 markedly upregulated the activity of the JNK, ERK and p38 MAPK signaling pathways, which may contribute to the promoting effects of CXCL5 on the proliferation and migration of glioma cells [
18]. In bladder cancer, CXCL5 was found to be significantly upregulated and the CXCL5/CXCR2 axis could promote the migration and invasion of bladder cancer cells by activating the PI3K/AKT-induced upregulation of MMP2/MMP9 [
19,
40]. The CXCR2/CXCL5 axis was also found to enhance epithelial-mesenchymal transition of hepatocellular carcinoma cells through the activation of the PI3K/AKT/GSK-3β/Snail signaling [
44]. Furthermore, Hsu et al. demonstrated that progression of breast cancer induced by TAOB-derived CXCL5 was associated with increased Raf/MEK/ERK activation and mitogen- and stress-activated protein kinase 1 (MSK1) and Elk-1 phosphorylation, as well as Snail upregulation [
44]. In addition, CXCL5 was shown to have potent effects on neutrophil recruitment in cancer [
45,
46]. Meanwhile, neutrophils could potentiate cancer cell migration, invasion and dissemination by secreting immunoreactive molecules such as hepatocyte growth factor, oncostatin M, b2-integrins or neutrophil elastase, which might be another mechanism for CXCL5 promoting cancer progression [
27,
47,
48]. What’s more, it has been reported that stem cells could produce CXCL5, and Zhao et al. demonstrated that CXCL5 secreted by adipose tissue-derived stem cells could promote breast tumor cell proliferation [
49]. Thus, we could speculate that CXCL5 might be the indicator of the presence of putative cancer stem cells, which have been shown to be associated with the metastasis and poor prognosis of cancer patients [
50,
51].
However, the current meta-analysis had some limitations. First, the cut-off value of high and low CXCL5 expression was different among studies, which might lead to the bias of the results. Second, some HRs could not be directly obtained from the publications. Thus, calculating them through survival curves might not be precise enough. Third, differences of paper quality and sample size across the studies might cause bias in the meta-analysis, although meta-regression did not show the paper quality or sample size as the resource of heterogeneity. Therefore, larger-scale, multicenter, and high-quality studies are desperately necessary to confirm our findings.
Authors’ contributions
BH, HF and XL collected, extracted and analyzed the data, wrote the paper; XL and SC performed quality assessment and analyzed the data. ZS conceived and designed this study. All authors read and approved the final manuscript.