Introduction
The Wnt signaling pathway plays a crucial role in the development of cancers in humans [
1]. The multiprotein destruction complex on the Wnt signaling pathway is organized by glycogen synthase kinase (GSK-3β), adenomatous polyposis coli (APC), and axis inhibition (AXIN). AXIN, a master scaffold protein in the destruction complex, serves as a scaffold protein that facilitates the phosphorylation of β-catenin by GSK-3β and acts as a vital mediator in cellular signaling.
AXIN is widely considered a negative regulator gene of Wnt/β-catenin signaling and plays an architectural role in integrating incoming signals to downstream effectors, which in turn manifest biological functions [
1].
Previous studies indicated AXIN protein expression was correlated inversely with tumor size in breast cancer [
2] and increased in colorectal carcinoma (CRC) tissues [
3]. The AXIN homologue conduction, also known as AXIL or AXIN2, serves as a scaffolding component of the multiprotein complex and negatively regulates the Wnt/β-catenin pathway [
4]. The AXIN2 protein acts as a tumor suppressor in numerous cancers [
5,
6]. The
AXIN2 gene has been mapped at human chromosome 17q23-q24, which shows frequent loss of heterozygosity (LOH) in cancers, and mutations in the
AXIN2 gene are associated with colorectal cancer with defective mismatch repair [
7,
8]. Some studies focused on the associations between risk of cancer and single nucleotide polymorphisms (SNPs) of the
AXIN2 gene, such as rs3923086, rs3923087, and rs2240308 [
9,
10]. The
AXIN2 SNP, Pro50Ser (rs2240308, c.148G > A), results in an amino acid change from a proline to a serine, which is located at exon 1 148 of the
AXIN2 gene, has been widely observed in lung cancer, ovarian cancer and prostate cancer [
11‐
13]. The
AXIN2 rs2240308 polymorphism seems to influence AXIN expression. The function of this SNP is closely associated with Wnt/β-catenin signaling and thus affects carcinogenesis [
14].
However, previous literature about the associations between the
AXIN2 rs2240308 polymorphism and risk of cancer has provided inconsistent results. Significant associations have been found in prostate cancer [
15] and lung cancer [
11,
14,
16], but similar results were not found in ovarian cancer [
12], astrocytoma [
10], colorectal cancer, and head and neck cancer [
16]. Significant racial differences have also been observed in the association between the
AXIN2 rs2240308 polymorphism and the risk of prostate cancer [
13,
15]. The objective of this meta-analysis is to evaluate broadly the available evidence on the
AXIN2 rs2240308 polymorphism and risk of cancer, for deriving a more reliable assessment.
Materials and methods
This meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement, including the search strategy, selection criteria, data extraction, and data analysis (Additional file
1) [
17]. The Venice criteria were used to assess the credibility of the genetic associations [
18].
Identification of eligible studies
We used the following specific combinations of search terms: “axis inhibition protein 2” or “AXIN2” in combination with “polymorphism”, “mutation” or “variant” in combination with “cancer” or “carcinoma” in Embase, PubMed, and Cochrane Library up to Nov 30, 2014. Two investigators (Yifan Sun and Zhitong Wu) conducted an extensive literature search independently for all publications. Articles in reference lists were also hand-searched. Only English-language articles and human studies were searched.
Inclusion and exclusion criteria
The following criteria were used to choose studies for inclusion: (1) case-control or cohort design studies; (2) studies offering the ability to extract data for calculating the odds ratio (OR), 95 % confidence intervals (CIs), and Hardy-Weinberg equilibrium (HWE); and (3) the DNA genotyping method and the source of the cases and controls were stated in the study. The exclusion criteria were (1) review articles, letters, case reports, editorials, and conference abstracts and (2) family-based studies.
Two investigators (Yifan Sun and Zhitong Wu) independently extracted data from the eligible studies. The data extracted included the first author’s name, publication date, country, ethnicity, total sample size, cancer type, genotyping method, genotype frequencies of the cases and controls, source of the case group and control group, and the source of specimens of cases that determined genotypes; HWE was calculated from the study data. If the literature did not provide sufficient data, the investigators attempted to contact the author to get the original data.
For the subgroup analysis, the cancer type, ethnicity, genotype method, and source of control were categorized according to the studies. If the data in a study came from different cancers, the study was treated as separate studies in our meta-analysis. To determine the accuracy of the extracted information, the data extracted by the two investigators should have been the same; they checked their data again if there was a dispute. If the two investigators could not reach an agreement, the dispute was submitted to a third reviewer (Hong Xu) to decide.
Quality score assessment
The quality of the selected studies was assessed by two investigators (Yifan Sun and Zhitong Wu) independently following the criteria predefined by Thakkinstian
et al. [
19]. The criteria were based on the sources of the cases and controls, the total sample size, the cases’ specimens, and the HWE of the controls (Table
1). According to the quality score assessment, a study that scored <10 was classified as “low quality,” a study that scored ≥10 was classified as “high quality”; the lowest score was 0, and the highest score was 15 [
20].
Table 1
Scale for quality assessment
Representativeness of cases | |
Selected from population or cancer registry | 3 |
Selected from hospital | 2 |
Selected from pathology archives but without description | 1 |
Not described | 0 |
Representativeness of controls | |
Population-based | 3 |
Blood donors or volunteers | 2 |
Hospital-based(cancer-free patients) | 1 |
Not described | 0 |
Specimens of cases determining genotypes | |
White blood cells or normal tissues | 3 |
Tumor tissues or exfoliated cells of tissue | 0 |
Hardy-Weinberg equilibrium in controls | |
Hardy-Weinberg equilibrium | 3 |
Hardy-Weinberg disequilibrium | 0 |
Total sample size | |
≥1000 | 3 |
≥500 but <1000 | 2 |
≥200 but <500 | 1 |
<200 | 0 |
Statistical analysis
We accessed the association between the AXIN2 rs2240308 polymorphism and risk of cancer by using different comparison models, including an allelic model (G vs. A), a co-dominant model (GA vs. AA, GG vs. AA), a dominant model (GA + GG vs. AA), and a recessive model (GG vs. GA + AA). We defined the GA and GG genotypes as “G carriers.” Unadjusted odds ratios (ORs) and the corresponding 95 % confidence intervals (CIs) were calculated according to the frequencies of genotypes but not by logistic regression because it is difficult to get the all original data from author.
Following the literature [
21,
22], heterogeneity was assessed with a chi-square Q test and I-square statistics. If
P
Q
< 0.1 or
I
2
> 50 %, we considered the heterogeneity significant, and a random-effects model was conducted using the DerSimonian and Laird method. Otherwise, the summary OR and the corresponding 95 % CI were calculated with the fixed-effects model (the Mantel-Haenszel method). We also carried out a subgroup analysis of ethnicity, genotyping method, source of the controls, hepatitis virus type, liver disease type, and quality assessment score. Galbraith plots analysis was performed for further exploration of the heterogeneity.
HWE in the controls was tested with the chi-square test for goodness of fit, and a P-value < 0.05 was considered out of HWE. Sensitivity analysis was conducted to examine such influence by removing studies one by one and recalculating the pooled OR and 95 % CI.
Begg’s funnel plot and Egger’s test were used to investigate the publication bias in the meta-analysis; P < 0.05 indicated that the result was statistically significant.
All the tests in this meta-analysis were conducted with STATA software (version 12.0; Stata Corporation, College Station, Texas, USA).
Discussion
This is the first meta-analysis of the association between the AXIN2 rs2240308 polymorphism and risk of cancer. Our results, based on 8 articles that included 10 case-control studies with 1,502 cases and 1,590 controls, suggested that the AXIN2 rs2240308 polymorphism increases the risk of cancer in a co-dominant model (GG vs. AA) and the recessive model, especially in lung cancer and Asian populations.
Our conclusion is biologically plausible. First, as a tumor suppressor gene,
AXIN2 polymorphisms on the susceptibility of cancer have been found in lung cancer [
11], prostate cancer [
13], and other cancers [
24], while these cancers have all been linked to Wnt signaling [
1]. AXIN2 is a transcriptional target of the Wnt signaling pathway. AXIN2 can be unregulated by β-catenin, providing a negative feedback loop to restrain Wnt signaling [
25]. Second, a region of
AXIN2 is involved in the APC-binding site, and AXIN and APC are critical for β-catenin regulation [
26]. Truncated mutations in APC that eliminate the AXIN-binding site result in human cancer, suggesting that the binding avidity of the AXIN protein for APC affects carcinogenesis. Low expression of AXIN2 was found to be associated with β-catenin accumulation in lung cancer, and epigenetic silencing of AXIN2 leads to β-catenin nuclear accumulation in tumorigenesis [
27]. It has also been reported that AXIN2 mutations lead to increased β-catenin concentrations in cancer with defective mismatch repair systems [
8]. It is believed that the mutation of
AXIN2 rs2240308 may influence the expression of the AXIN protein, and then dysfunction in these signaling pathways, which play critical roles in carcinogenesis.
In the subgroup analysis by ethnicity, statistical significant associations were found in Asians but not in Caucasians. In fact, it is a widespread phenomenon that the same gene polymorphism plays different roles in risk of cancer among different ethnic subpopulations. Cancer is a complex multigenetic disease, and different genetic backgrounds may contribute to the discrepancy. Moreover, we observed statistically significant differences in allele and genotype frequencies of AXIN2 rs2240308 between Asians and Caucasians.
To investigate the association between the
AXIN2 rs2240308 polymorphism and risk of cancer in different cancers, we also performed a subgroup analysis by cancer type. The results indicated that the
AXIN2 rs2240308 polymorphism was significantly associated with an increased risk of lung cancer but not in other cancer types. Several possibilities may explain this result. First, in addition to acting as a tumor suppressor protein by negatively regulating Wnt signaling, AXIN also plays a positive role in TGF-β signaling and regulates TGF-β signaling by acting as an adaptor for Smad3, one of the TGF-β effectors [
28]. The TGF-β signaling pathway is considered a tumor suppressor and a cancer promoter [
29]. Decreased expression of AXIN has been detected in lung cancer tissues [
30], while the expression of AXIN2 increases in CRC [
3]. These factors suggest that AXIN plays a dual role in the development of different cancers. Second, gene-gene and gene-environmental interactions may lead to the formation of different cancers [
31‐
33]. Third, limited studies and small sample sizes may lead to insufficient statistical power; therefore, we should carefully interpret the negative results.
Since hospital-based controls cannot truly represent the general population, we performed subgroup analysis of the source of the controls. The results revealed that statistical significant associations were observed only in population-based studies, not in hospital-based studies, which suggested that selection bias and recall bias of the study population should be considered when explaining these inconsistent results. A previous study also indicated that hospital-based case-control studies have a high risk of producing unreliable results, and suggested that a methodologically preferable design with an appropriate representative population-based study is crucial to avoid selection bias [
34].
Heterogeneity is the most common problem when explaining the results of a meta-analysis. In this meta-analysis, we assessed heterogeneity by using different statistical methods. We found Mostowska A
et al.’s [
12] study was the source of heterogeneity in the allelic comparison model, due to significant higher A allele frequencies compared with other studies. However, the summary OR value did not change significantly after the study was excluded; in addition, all of the other comparison models had no significant heterogeneity (
P
Q
≥ 0.1). Therefore, we think that our results were statistically robust, and the sensitivity analysis confirmed this point.
Our meta-analysis had several limitations that must be considered when interpreting the results. Insufficient studies and small sample sizes were the biggest problem. The lack of association in other cancers may also be most likely because of insufficient studies. According to the Venice interim guidelines, the cumulative evidence for the association between AXIN2 polymorphisms and lung cancer is categorized as weak. Therefore, further studies in different ethnicities and cancers should be conducted to strengthen our results. Furthermore, because original data were lacking in the reviewed studies, a more precise analysis was not conducted if individual information including other covariates such as age, sex, and cancer stage became available. Finally, we must pay attention to publication bias when explaining the results even though it was not observed in statistical tests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
YS, ZW and HX conceived and designed the experiments YS and ZW performed the experiments. YS analyzed the data. CL and SZ contributed reagents/materials/analysis tools. YS and ST contributed to the writing of the manuscript. All authors read and approved the final manuscript.
Yifan Sun, Zhitong Wu and Shifu Tang are the co-first authors.