Functional FGFR4 Gly388Arg polymorphism contributes to cancer susceptibility: Evidence from meta-analysis

Fibroblast growth factor receptor 4 (FGFR4) is a member of receptor tyrosine kinase family. A functional Gly388Arg (rs351855 G>A) polymorphism in FGFR4 gene causes a glycine-to-arginine change at codon 388 within the transmembrane domain of the receptor. Although the FGFR4 rs351855 G>A polymorphism has been implicated in cancer development, its association with cancer risk remains controversial. Here, we have systematically analyzed the association between the rs351855 G>A polymorphism and cancer risk by performing a meta-analysis of 27 studies consisting of 8,682 cases and 9,731 controls. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to measure the strength of the association. The rs351855 G>A polymorphism was associated with an increased cancer risk under the recessive model (OR=1.19, 95% CI=1.01-1.41). Stratified analysis by cancer type indicated the rs351855 G>A polymorphism was associated with an increased risk of breast and prostate cancer, but a decreased risk of lung cancer. This meta-analysis demonstrates the FGFR rs351855 G>A polymorphism is associated with increased cancer risk and suggests it could potentially serve as a chemotherapeutic target or biomarker to screen high-risk individuals.


INTRODUCTION
Cancer represents an enormous economic burden on society in both developing and developed countries.Based on the GLOBOCAN 2012 estimates, there were about 14.1 million new cancer cases in 2012, and 8.2 million deaths [1].Cancer is a complex multifaceted disease that results from gene-environment interactions.Apart from the lifestyles associated with an increased risk of a number of common cancers, genetic variations, including single nucleotide polymorphism (SNP), have been known to affect cancer susceptibility.

Association between FGFR4 rs351855 G>A polymorphism and cancer risk
The main findings of the meta-analysis are shown in Table 2  A significant heterogeneity existed in the overall pooled analysis for the association of interest under all genetic models, as indicated by the P values of heterogeneity test (P het ) in the Table 2.Moreover, in the analyses stratified by cancer type, ethnicity, and source of control, significant heterogeneity was detected in prostate cancer, Asians, and in HB studies, respectively.Heterogeneity indicates diversity, which may result from differences in subjects, genotyping method, study design, sample size, ethnicity, and many other factors varying across studies.Therefore, the Mantel-Haenszel random effects model was used to decrease the effect of heterogeneity in these cases.
We assessed validity of the results by sensitivity analyses and examination of potential publication bias.Sensitivity analysis revealed that no individual study substantially altered the pooled results, showing the stability of the present meta-analysis (Supplementary Figure 1).The meta-analysis was free of publication bias, as indicated by symmetrical funnel plots (Figure 3), and by the Egger's linear regression test (homozygous: P=0.461; heterozygous: P=0.085; recessive: P=0.229; dominant: P=0.27; allele contrast: P=0.938).

DISCUSSION
The current meta-analysis demonstrates that the FGFR rs351855 G>A polymorphism is associated with an increased cancer risk.Consistent with our study, a previous meta-analysis in 2010, which included 12 studies with 4,892 cases and 3,663 controls, suggested that the FGFR rs351855 G>A polymorphism conferred increased genetic susceptibility to cancer [40].The same meta-analysis also showed that the rs351855 G>A polymorphism significantly elevated the risk of prostate cancer and breast cancer [40].However, in the analysis stratified by cancer type, only three studies for breast cancer were included, and no study for lung cancer was reported [40].Since 2010, new studies have emerged.The current meta-analysis of 27 studies comprising 8,682 cases and 9,731 controls, confirmed the increased cancer risk of the rs351855 G>A polymorphism with increased statistical power.Moreover, our results indicate that the FGFR4 rs351855 G>A polymorphism increases the risk of prostate and breast cancer, but decreases the risk of lung cancer.The reasons why the FGFR4 rs351855 G>A polymorphism has opposite effects on different types of cancer are unclear, suggesting that other genetic and environmental factors might be involved, or that this polymorphism modifies cancer susceptibility in a tissuespecific manner.Alternatively, the association found with lung cancer might not be noteworthy, since only two studies were involved.Other two meta-analyses were performed regarding the FGFR rs351855 G>A polymorphism [41,42].However, one study focused on the association between the polymorphism and prognosis (nodal status and overall survival) [41], and the other analysis only included studies on prostate cancer [42].
Interestingly, the FGFR rs351855 G>A polymorphism genotypes are distributed differently among different ethnic groups [23,40].The current meta-analysis confirmed that the FGFR Arg 388 allele is the most prevalent in Asians (40.1%), and then Caucasians (30.4%), and Africans (11.7%).However, analysis stratified by ethnicity failed to find any ethnic-specific association between the FGFR rs351855 G>A polymorphism and cancer risk.That is inconsistent with the previous meta-analysis study [40], which reported a significant association in Asians, with almost half the sample size (11 studies vs. 6 studies).
The biological effect of the FGFR4 Arg388 expression remains unclear.However, several lines of evidence indicate the implication of the rs351855 G>A polymorphism in carcinogenesis.In vitro experiments with triple negative breast cancer cells (MDA-MB-231) demonstrated that cells expressing the FGFR4 Arg388 variant have increased motility compared with cells expressing the FGFR4 widetype counterpart (Gly388) [18].Wang et al. reported that compared to Gly388 cDNAs, transfection of FGFR-4 Arg388 cDNAs promoted migration and invasion of PNT1A cells (prostatic epithelial cell line) [23].
However, no correlation has been established yet between the FGFR4 rs351855 G>A genotype and FGFR4 protein levels.A previous study did not detect any correlation between the FGFR4 rs351855 G>A genotype and FGFR4 protein expression among 104 HNSCC patients [43].In addition, the Gly388Arg polymorphism did not alter FGFR4 protein expression in normal lung tissue [28].FGFR4 serves as a receptor tyrosine kinase.However, the rs351855 G>A polymorphism did not alter the tyrosine kinase activity of FGFR4 in breast cancer [18] and prostate cancer cells [23].Therefore, the SNP may increase cancer risk through other mechanisms, including altering FGFR4's ligand affinity, degradation, or its capacity to interact with downstream effectors.In this regard, Tateno et al. found that the FGFR4 polymorphism contributes to pituitary tumorigenesis through increasing phosphorylation of the mitochondrial STAT, resulting in increased cell growth [44].Ulaganathan and Ullrich have recently reported that replacement of the glycine residue with a charged arginine residue at codon 388 causes changes in the transmembrane region, which consequently exposes a membrane-proximal cytoplasmic STAT3 binding site Y390-(P)XXQ393 [45].Such de novo exposure of the STAT3 binding site facilitates STAT3 tyrosine phosphorylation, thereby stimulating cell proliferation [45].Collectively, the FGFR4 rs351855 G>A polymorphism may promote tumorigenesis by enhancing cell migration, invasion, and proliferation.More studies are needed to validate the association and investigate the underlying mechanisms.This SNP may hold a promise as a potential chemotherapeutic target and a biomarker to screen high-risk individuals.There were some limitations in our meta-analysis study.First, the moderate sample size might have reduced the statistical power, especially in the stratified analyses.Second, because of the unavailability of demographic and environmental information, confounding factors could not be adjusted for, including age, sex, smoking, and drinking.Since the strength of the association was measured by crude ORs, our results might be open to confounding bias.Third, publication bias may be inevitable since we were only able to acquire data from unpublished articles.Finally, the meta-analysis was associated with a significant heterogeneity, which might weaken reliability of the meta-analysis.
In conclusion, our study demonstrates the association between the FGFR4 rs351855 G>A polymorphism and overall cancer risk.In terms of cancer type, the FGFR4 rs351855 G>A polymorphism was found to modify susceptibility to breast, prostate, and lung cancer.

MATERIALS AND METHODS
The meta-analysis was performed according to the latest meta-analysis guidelines (PRISMA) [46].

Data sources
We retrieved all the published studies relating the FGFR4 rs351855 G>A polymorphism and cancer risk by searching PubMed and EMBASE databases.The combination of search terms were as follows: "FGFR4 or fibroblast growth factor receptor 4", "rs351855 or Gly388Arg", "polymorphism or variation or variant", and "cancer or carcinoma or tumor".The latest search was carried out in October 2016.We also examined the reference lists of the relevant original publications and review articles as well as previously published metaanalyses to maximize the coverage of the current metaanalysis.No language restriction was applied.

Inclusion and exclusion criteria
Inclusion and exclusion criteria were predetermined as described previously [47][48][49].To be included, studies had to: (i) be case-control or cohort studies, (ii) assess the association between the FGFR4 rs351855 G>A polymorphism and cancer risk, and (iii) provide adequate data for calculating odd ratios (ORs) and 95% confidence intervals (CIs).We discarded case reports, case only studies, review articles, and conference abstracts.Additionally, we excluded studies that departed from Hardy-Weinberg equilibrium (HWE) (P HWE < 0.05) in controls, if there was no extra evidence from another FGFR4 polymorphism satisfying HWE.When duplicate studies occurred, only the most recent or the largest study was included.

Data extraction
Two investigators (Fen Feng and Jianqun Ma) obtained data from individual studies independently following a standardized data extraction form.Data were extracted by name of the first author, year of publication, country where the study was conducted, ethnicity, genotype counts of cases and controls for the FGFR4 rs351855 G>A polymorphism, source of controls (controls were chosen from the general population or a hospital), genotype method, and the P-value of HWE in controls.The subgroups were defined by cancer type, ethnicity, and source of control [hospital-based (HB) and population-based (PB)].Publications containing different ethnicities, cancer types or different regions, were separated into different categories.The extracted information was imported into excel worksheets and examined for inconsistency.Conflicts were resolved by discussion between the two authors and a consensus would be reached ultimately.

Quality assessment
The quality of studies was scored based on the following criteria: representativeness of case, representativeness of control, ascertainment of cancer case, control selection, genotyping examination, HWE, and total sample size as described in previous study [50].The final quality score ranged from 0 (lowest) to 15 (highest).A score of 9 was used as a cutoff value (high quality: > 9; low quality: ≤ 9).

Statistical analysis
A goodness-of-fit chi-square test was run to check departure from HWE for the FGFR4 rs351855 G>A polymorphism in controls.The common measure of association across studies, the OR with 95% CI, was applied to the meta-analysis.The strength of the association between the FGFR4 rs351855 G>A polymorphism and cancer risk was determined under the four genetic models (homozygous: AA vs. GG; heterozygous: AG vs. GG; recessive: AA vs. AG/GG; dominant: AG/AA vs. GG) and allele contrast was performed (A vs. G).We performed standard chi-square tests and the I 2 statistic to quantify between-study heterogeneity [51].If significance betweenstudy heterogeneity was detected, summary ORs (95% Cls) were computed using a Mantel-Haenszel random effects model to decrease the effect of heterogeneity.If not, fixed effects model would take effect.Cancer type, ethnicity, and source of control were specified as study level characteristics for evaluating the source of heterogeneity using stratified analysis [52].In order to explore the impact of each individual study on pooled risk estimates, we also carried out sensitivity analyses, which were fulfilled by leaving out one study at a time and recalculating ORs for the remaining studies.Publication bias was assessed through Begg's funnel plots [53] and Egger's linear regression test [54].All analyses were conducted using STATA version 12 (Stata Corp, College Station, TX).All statistical analyses were two-sided, and a value of P <0.05 was considered significant.

Figure 2 :
Figure 2: Forest plot of the association between the FGFR4 rs351855 G>A polymorphism and cancer risk under the recessive model.

Figure 3 :
Figure 3: Funnel plots of FGFR4 rs351855 G>A polymorphism.Each point represents a separate study for the indicated association.(A) homozygous model; (B) heterozygous model; (C) dominant model; (D) recessive model.

Table 1 : Characteristics of studies included in the current meta-analysis Surname Year Cancer type Country Ethnicity Design Genotype method Case Control MAF HWE Score GG AG AA All GG AG AA All
Abbreviations: MAF, minor allele frequency; HWE, Hardy-Weinberg equilibrium, CRC, colorectal cancer; HNSCC, head and neck squamous cell carcinoma; HCC, hepatocellular carcinoma; OSCC, oral squamous cell carcinoma; NHL, non-Hodgkin lymphoma; PB, population based; HB, hospital based; PCR-RFLP, polymorphism chain reaction-restriction fragment length polymorphism; PCR-SSCP, polymorphism chain reaction-single-stranded conformation polymorphism.a data was provided for AG/AA.