Introduction

Helicobacter pylori (H. pylori), a group I human gastric carcinogen by the International Agency for Research on Cancer (IARC) [1], is the strongest risk factor in the gastric cancer development, and persistent H. pylori infection is the first step towards gastric carcinogenesis [1]–[3]. In spite of numerous evidence that H. pylori plays a crucial role in gastric carcinogenesis, only a small portion of infected people develop gastric cancer. This implies that other factors involved in the pathogenic mechanism of H. pylori can modify individual susceptibility for gastric cancer. Our previous studies demonstrated H. pylori infection itself was not associated with the gastric cancer risk but specifically CagA positive H. pylori infection significantly increased risk for gastric cancer by 3.57-fold [4], [5].

Cytotoxin-associated gene A (CagA), an immunodominant protein secreted by H. pylori, appears to be one of the pathogenic modifying factors [6]–[8]. After infecting H. pylori into gastric epithelial cells, CagA acts as a major carcinogenic and virulent component through sequential CagA signal transduction pathway. The first step begins with the interaction between CagA and diverse proteins such as SRC, SHP2, CRK, CRKL, and CSK after phosphorylation and c-MET and GRB2 without phosphorylation [9]–[12]. Tyrosine phosphorylated CagA by SRC family kinases interacts with SHP2 tyrosine phosphatase (encoded by the PTPN11 oncogene), CRK, CRKL and CSK and induces cell scattering, dissociation and mortality connected to cancer development [8], [12]–[15]. Additionally, non-phosphorylated CagA interacts with c-MET and GRB2 which promotes oncogenic response including cell proliferation and morphological changes such as hummingbird formation [8], [16]–[19]. This CagA translocation and its cellular interaction with those proteins can be a crucial initiating step in gastric carcinogenesis [9]–[12].

Cellular alteration in the CagA positive H. pylori pathogenic mechanism appears to explain different susceptibility of gastric cancer among H. Pylori infected persons. Since cellular functions can be regulated by their host genes, genetic variants related to the CagA interacting molecules may be the key for individual gastric cancer susceptibility. Based on the putative genetic differences, we hypothesized that genes which encode CagA-interacting proteins may modify risk for gastric cancer. Moreover, we focused on phytoestrogens as an effect modifier in the CagA signal transduction process. Studies have reported that phytoestrogens with anti-inflammatory, anti-bacterial and anti-oxidant properties can inhibit H. pylori activity and gastric cancer cell growth and proliferation [20]–[22]. Especially, genistein, one of phytoestrogens and phosphotyrosine kinase inhibitors, is reported to be an effective blocker for CagA phosphorylation [23].

To evaluate the hypotheses, a two-stage genetic analysis that focused on genes which directly encode CagA-binding molecules, SRC, PTPN11, CRK, CRKL, CSK, c-MET and GRB2, was conducted that included: 1) the discovery phase that screened and identified single nucleotide polymorphisms (SNPs) with a significant genetic association on gastric cancer; 2) the extension phase that re-analyzed the most significant SNPs in the discovery phase. Additionally, in a sub-analysis, we evaluated the gene-environment interaction to determine whether phytoestrogen levels modify the association between gene polymorphisms which directly encode CagA-binding molecules and risk of gastric cancer.

Methods

Results

There was no significant difference between cases and controls according to sex, H. pylori infection, CagA/VacA seropositivity, smoking/drinking status and gastric ulcer history in the discovery and extension phases (p>0.05). CagA/VacA seropositivity and the proportion of current smokers were significantly higher among gastric cancer cases in the pooled data (p = 0.03, p<0.01, p = 0.02, respectively) (Table S1).

Of the 116 SNPs in the seven candidate genes encoding CagA interacting proteins analyzed in the discovery phase, 22 SNPs in three genes, SRC, c-MET, and CRK, were significantly associated with gastric cancer (p-LRT<0.05). SRC rs6122566 significantly increased risk for gastric cancer in the recessive models (OR = 4.90, [95% CI 1.19–14.2]). Thirteen SNPs that were rs41739, rs16945, rs41738, rs6566, rs10435378, rs41737, rs2023748, rs41736, rs41735, rs6951311, rs183642, rs2237717 and rs38859 in c-MET gene showed a significant gene-dose effect in the linear trend tests (p<0.05). CRK rs7208768 had a marginally significant gene-dose effect. 100,000 permutation tests in the single SNP analysis showed SRC rs6122566, c-MET rs41739 and CRK rs7208768 with the most significant permutated p-value in each gene (p_permutation = 0.00284, p_permutation = 0.00989, p_permutation = 0.01392, respectively). The marginal significance of the corrected permutated p-value was observed for SRC rs6122566 (p = 0.0918) but all FDR p-values in all genetic models were not significant (p>0.2) (Table S2).

Haplotype blocks were identified by the LD plot (Figure S1). The largest block was constructed with the most significant SNPs including rs41739, rs6566, and rs41738, but the omnibus p-value was not significant (p>0.05). Four blocks defined by SRC and one block defined by CRK did not show statistical significance in the omnibus test. The results of the haplotype analysis did not present information beyond individual SNP results (data not shown).

In the extension phase, two SNPs, rs6122566 and rs6124914, in SRC gene and rs7208768 in CRK remained significantly associated with an increased risk for gastric cancer (OR = 4.01, [95% CI: 1.62–9.96]; OR = 1.30, [95% CI: 1.00–1.70]; OR = 1.33, [95% CI: 1.08–1.64], respectively). Associations between the SNPs in c-MET gene (rs41739 and rs41737) and gastric cancer risk were attenuated. In the combined analysis that included the discovery and extension phases, the risk estimate of SRC rs6122566 in the recessive model was significantly associated with gastric cancer in both the pooled and meta-analyses (OR = 3.96, [95% CI: 2.05–7.65]; OR = 4.59, [95% CI: 2.74–7.70], respectively). Moreover, SRC rs6124914, c-MET rs41739 and CRK rs7208768 showed significant gene-dose effects for gastric cancer in both analyses. There was no heterogeneity across the analyses (Cochran Q test, p>0.05) (Table S3).

Among a total of 823 subjects (406 cases and 417 controls) who were measured the plasma levels of the four phytoestrogen biomarkers, the overall concentrations of genistein, daidzein and enterolactone in cases were significantly lower than those of the controls (genistein 167.6 nmol/L in cases vs. 200.2 nmol/L in controls, p = 0.0004; daidzein 91.4 nmol/L in cases vs. 131.6 nmol/L in controls, p<0.0001; enterolactone 51.0 nmol/L in cases vs. 77.7 nmol/L in controls, p<0.0001). Overall plasma concentrations of equol, a daidzein metabolite, were lower in cases but not statistically significant (50.3 nmol/L for cases vs. 62.2 nmol/L for controls; p = 0.0977). In stratified analysis according to phytoestrogen biomarkers, a significant gene-environment interaction was observed in CRK. Risk allele of CRK rs7208768 had a significantly increased risk for gastric cancer at low phytoestrogen levels. Specifically, the A allele of rs7208768 was associated with a greater risk of gastric cancer at low genistein, daidzein, equol and enterolactone and statistically significant (OR = 1.91, [95% CI: 1.44–2.52] at low genistein; OR = 2.09, [95% CI: 1.46–3.01] at low daidzein; OR = 1.87, [95% CI: 1.26–2.78] at low equol; OR = 1.77, [95% CI: 1.10–2.85] at low enterolactone). The p-interaction was significant (p = 0.0001, p = 0.0013, p = 0.0147, p = 0.0404, respectively) (Table S4).

Though additional stratified analyses were also conducted to detect an interaction between CagA seropositivity and each gene effect for gastric cancer risk, interactions were not significant in any of the three genes, SRC, c-MET and CRK (data not shown).

Discussion

CagA-secreting H. pylori infection appears to play an important role in gastric carcinogenesis via sequential CagA signal transduction pathway. CagA initially binds to seven protein components to activate aberrant cellular responses that underlie the development of gastric cancer. Since function of the protein can be regulated by their host genes, genes that encode CagA interacting molecules may be able to modify risk for gastric cancer. To evaluate this hypothesis, we genotyped 137 SNPs in seven candidate genes and demonstrated that genetic variants of SRC (rs6122566 and rs6124914), c-MET (rs41739) and CRK (rs7208768) were significantly associated with gastric cancer risk. Additionally, an interactive effect of CRK genetic polymorphism, rs7208768, and four phytoestrogen biomarkers, genistein, daidzein, equol and enterolactone on gastric cancer risk were analyzed.

SRC, a non-receptor protein tyrosine kinase (TK), appear to be essential in gastric carcinogenesis. Once injected into gastric epithelial cells, CagA undergoes tyrosine phosphorylation by the SRC family kinases [8], [12], [18]. The tyrosine phosphorylation of CagA is an integral step in determining the sequential cellular signaling mechanism. Because some CagA interacting molecules such as SHP-2, CRK and CSK are only able to respond with phosphorylated CagA, SRC can be more important in influencing other's cellular functions and inducing development of gastric cancer. Additionally, SRC has been reported to play a crucial role in tumor progression and mediate cancer development and metastasis [32]. Cellular activity of SRC appears to be altered by the host gene and our results indicate that SRC rs6122566 and rs6124914 can be risk modifiers in gastric carcinogenesis. SRC genetic variations that influence the cellular capacity in gastric epithelial cells are associated with gastric cancer risk.

Despite the attenuated significance in the extension analysis, c-MET which is synonymous with HGFR (hepatocyte growth factor receptor) may be an independent risk gene for gastric cancer. Numerous previous studies reported that c-MET, one of the receptor TKs, promotes invasive tumor growth, cell invasion, and mortality, and amplification and/or overexpression of c-MET was associated with various human carcinoma including gastric cancer [17], [33]–[36]. In terms of c-MET cellular mechanism, CagA plays a role as an adaptor protein, Gab, to mediate receptor TK signaling by controlling a cluster of downstream components at the activated receptor such as Grb2, PLCγ, and SHP-2 [37], [38]. By functionally mimicking the Gab adaptor protein, CagA might stimulate abnormal proliferation and mortality of gastric epitherial cells [39]. In the present study, a polymorphism of c-MET gene (rs41739) was significantly associated with gastric cancer risk and a possible genetic susceptible factor on gastric cancer. Consistent with the cellular importance and function, c-MET gene appears to modify the risk for gastric carcinogenesis through CagA signal transduction pathway.

CRK adaptor protein which has splicing isoforms, CRK-I (SH2-SH3) and CRK-II (SH2-SH3-SH3), binds to TKs and controls transcription and cytoskeletal reorganization modulating cellular activities [40]. Also, this adaptor protein integrates various cellular signals and its dysregulation is connected to the human carcinoma [41]. Interaction between CRK and phosphorylated CagA has been reported to be a biological prerequisite that leads to morphological change, cell scattering and deregulation of cell–cell adhesion in the gastric epithelium [14]. Several studies have indicated that overexpression of CRK is associated with various types of human cancers including lung, gastric and colon cancer [42], [43]. Our findings also support the genetic potential of CRK rs7208768 on the development of gastric cancer and both genetic and cellular magnitude of CRK.

More interestingly, significant interactions between the CRK genetic polymorphism and four phytoestrogen biomarkers, genistein, daidzein, equol and enterolactone, modified gastric cancer risk. Studies indicated protective effects of phytoestrogens on gastric cancer [20], [28] and particularly, genistein inhibited the ERK signal transduction cascade induced by H. pylori infection playing a role as a tyrosine kinase inhibitor [44]. Considering CRK is the major upstream molecule of ERK activation, the risky genetic variants of CRK to activate the ERK signaling can be blocked by phytoestrogens, and mediate the development of gastric cancer.

SRC, c-MET and CRK are also involved in the protein TKs that is a diverse multigene family which controls cellular signal transduction pathway mediating a range of downstream cellular processes and plays significant roles in the development of various clinical diseases [45], [46]. TKs are also known as oncogenes involved in human malignancies. SRC belongs to a non-receptor TK and c-MET is a receptor TK, while CRK is an adaptor protein which binds to TK-phosphorylated proteins and strengthens the main proteins in the signal transduction pathway [41]. These three molecules encoded by SRC, c-MET, and CRK genes can independently induce cell differentiation, adhesion, death and morphological changes by transmiting cell signals related to their TK activities regardless of interaction with CagA. Futhermore, genes related to TK action appear to play a crucial role as a susceptible factor for gastric cancer considering genistein that is a tyrosine kinase inhibitor can reduce gastric cancer risk [28]. This indicates that genetic susceptibilities of SRC, c-MET, and CRK in gastric carcinogenesis should be treated as independent risk factors that modify the cellular signal transduction in TK dependent manners because uninfected persons with CagA secreting H. Pylori can be at risk for gastric cancer depending on individual genetic variants of the three genes.

Though PTPN11, CRKL, CSK, and GRB2 did not show any significant association with gastric cancer in the present study, their genetic effects should not be overlooked. At the cellular level, these molecules are significantly related to aberrant effects that underlie gastric carcinogenesis [12]. As one of the human proto-oncogenes, PTPN11 encodes cytoplasmic tyrosine phosphatase with SHP2 and can induce aberrant hyperactivation of the ERK signaling [47]. A study has also reported that a PTPN11 genetic variant increased the risk for gastric atrophy and cancer among CagA positive H. pylori infected people [48]. In the CagA signal transduction pathway, CRKL works quite similarly to CRK; CSK frustrates an activity of SRC family kinase and CagA-SHP2 signaling; and GRB2 acts as a trigger to activate the RAS/MEK/ERK pathway [14], [18], [49]. Further studies with a greater number of gastric cancer cases and wider coverage of genetic polymorphisms in these genes are warranted.

Gastric carcinogenesis induced by CagA positive H. Pylori infection can be infered from our study results and review of cellular mechanisms [16], [18], [47] (Figure S2). Once CagA is injected in gastric ephithelial cells, SRC initiates CagA phosphorylation that interacts with CRK adaptor protein and SHP2 to promote the ERK activation. Non-phosphorylated CagA mimics the Gab adaptor protein to potentiate the c-MET-HGF intracellular signaling and stimulate c-MET signals that also activates the ERK signal cascade. As a result, CagA binding molecules such as SRC, c-MET, CRK, SHP2, CRKL and GRB2 interacts with phosphorylated or non-phosphorylated CagA to stimulate downstream signals in the ERK activation inducing oncogenic effects on gastric cancer, whereas, CSK inhibits SRC family kinase functions and CagA-SHP2 signaling effects. In terms of genetic mechanism, significant genetic markers of SRC (rs6122566 and rs6124914), c-MET (rs41739) and CRK (rs7208768) are actively involved in gastric carcinogenesis. Interestingly, primary interacting molecules, SRC, c-MET, CRK and SHP2, contribute to mutual development and sequentially connect to operate carcinogenic effect. In particular, SRC can extensively influence the activity of the others from migration to signal transduction [11], [50].

We examined SNPs from our other study that used the Affymetrix 5.0 platform and the Korea Centers for Disease Control and Prevention (KCDC) study that used the Affymetrix 6.0 platform to find consistencies. In the KCDC study, SRC rs6122566 which was the most significant SNP in the present study showed a raw p-value of 0.0009 in the single SNP analysis. Unfortunately, many other significant SNPs were not included in the platforms due to different SNP selection methods (random SNPs vs. candidate SNPs according to MAF >0.05 in Asians) and target populations (Caucasian vs. Koreans). Nevertheless, our results might be applicable to most East-Asian populations because the minor allele frequencies of our significant SNPs showed similarities to other Asian populations such as Chinese and Japanese (http://www.hapmap.org) (Appendix S1).

Limitations should be noted. First, due to the restricted number of study subjects, we did not have sufficient statistical power and were not able to perform subgroup analyses for gastric cancer types such as histological (intestinal vs. diffuse) and anatomical subtypes (cardia vs. non-cardia). Second, we only focused on CagA interacting molecules and thus, secondary interacting molecules in CagA downstream signaling pathways such as RAS and ERK cascade were not included in this study. Third, in the extension phase, selection bias may be induced because hospital-based cases were matched to community-based controls. However, considering that 1) genetic traits are inborn and not easily changeable, 2) all cases were matched to controls according to the major covariates in the initial study design stage, 3) effects of confounding factors were considered by the use of multivariable models, and 4) no heterogeneity between hospital- and community-based cases in meta-analysis, the potential selection bias may be minimized.

In spite of the limitations, this is a two-phase genetic association study that provides evidence on the role of CagA cellular mechanism-related genes. Through the candidate approach of the discovery phase, the most significant SNPs were preliminarily screened and after, the SNPs were re-evaluated in the extension phase. Moreover, through intensive analyses that focused on gene and gene-environment interaction, conclusive evidence is provided to elucidate the etiology of gastric cancer.

This study shows SRC, c-MET, and CRK genetic variants can be susceptible genetic factors for the development of gastric cancer by controlling signals through the CagA transduction pathways. Moreover, an interaction between CRK genetic polymorphism and phytoestrogen biomarkers appear to play a role as risk modifiers in gastric carcinogenesis. Replication studies with a greater number of cases and more substantial genomic coverage of the genes will allow us to elucidate gastric cancer pathological mechanisms based on the CagA signal transduction.

Supporting Information