A novel diffuse gastric cancer susceptibility variant in E-cadherin (CDH1) intron 2: A case control study in an Italian population

Gastric cancer is a major cause of cancer-related mortality and is usually classified into two histological types, the intestinal and the diffuse form (Lauren classification [1]). The general incidence rates for stomach cancer are in a steady decline, largely due to decreasing rates of the intestinal cancer-type. This falling frequency is believed to be the result of improved nutrition and sanitary conditions. In contrast, the incidence of diffuse gastric cancer (DGC) alone appears more stable over the past few decades [1, 2]. Such a constant rate suggests a larger contribution of inherited genetic risk rather than environmental factors to the diffuse form of stomach cancer.

Owing to its early development underneath the gastric mucosal surface [3], DGC is usually diagnosed at an advanced stage and consequently associated with a poorer outcome [1]. Therefore, genetic DGC markers may facilitate the identification of individuals at risk and thereby contribute to an improvement in DGC diagnosis and therapy.

On a molecular level, DGC is distinguished from the intestinal type on the basis of its abnormal expression of the cell-cell adhesion molecule E-cadherin [4]. E-cadherin is the key component of the epithelial adherens junction and as such is required for functional intercellular adhesion within epithelial sheets [5]. In contrast to many other epithelial cancers, E-cadherin is downregulated very early during DGC development, suggesting a role in the initiation of this disease [3]. Mutation and promoter hypermethylation of the E-cadherin gene (CDH1) are the most consistent genetic alterations observed in sporadic DGC [6, 7]. Furthermore, CDH1 germline mutations predispose to hereditary DGC [8] consistent with an initiating function of E-cadherin deficiency in DGC. CDH1 germline mutations usually co-segregate with a dominant pattern of disease among affected families, and occasionally can be found in isolated DGC cases diagnosed at a young age (<45 y) [9]. However, they account for only about 1% of all DGC cases [9] and hence cannot explain the genetic aetiology postulated to contribute to apparent sporadic DGC cases. Genetic alterations other than CDH1 germline mutations are therefore likely to add to the risk of developing DGC in the absence of a clear family history or a young age at diagnosis.

Common allelic variants with a mild functional effect can influence the risk for sporadic disease. Indeed, a single nucleotide polymorphism (SNP) within the CDH1 promoter (-160C>A) has been associated with a significantly increased risk of sporadic DGC in certain high-incidence populations [10‐13]. Of the studied CDH1 SNPs, the -160A promoter allele is so far the only variant implicated in DGC risk but appears to act in combination with other CDH1 polymorphisms [10, 13].

Recently, a new CDH1 regulatory region has been described [14]. This region is contained within CDH1 intron 2, the largest non-coding CDH1 segment (66% of the total sequence) and has been shown to be required for both the initiation and maintenance of transcriptional CDH1 activity in differentiated epithelia. Importantly, intron 2 sequences are also necessary for normal CDH1 transcription during adult life, providing the possibility that variants within this region may affect diffuse gastric carcinogenesis.

Six conserved regions with a total size of 3.2 kbp were identified within CDH1 intron 2. Apart from the five known polymorphisms (CDH1163+14184ΔAGGG (rs10673765), 163+14384C>T (rs9932686), 163+37235G>A (rs1125557), 163+37276T>A (rs9282650), and 163+49526C>G (rs9931853)), no additional common polymorphisms specific for the Italian population under study were discovered by SSCP in the six regions.

Using restriction fragment length polymorphism and allelic discrimination assays, the relative frequencies of the genotypes resulting from the five variants were determined in the DGC cases and the controls. Sequencing of random samples confirmed the respective genotypes. All polymorphisms were in Hardy-Weinberg equilibrium for both cases and controls (p > 0.19). Table 2 summarises the genotype distributions and their differences between cases and controls.

Of the investigated variants, only the 163+37235G>A SNP was significantly associated with disease due to an overrepresentation of the A-allele among the DGC cases (56.7% vs. 38% in controls, χ² = 16.1, p < 0.0001; see Table 2). The 163+37235AA genotype was 2.8 times more frequent in cases compared to controls. The corresponding Odds Ratio (OR) suggested a significantly elevated risk of developing DGC for AA carriers relative to GG carriers (OR = 4.55, 95%CI = 2.09–9.93, p = 0.0002, power of association 100%; Table 2). No significant increase in risk was apparent from carrying the GA-genotype (OR = 1.38, 95%CI = 0.75–2.55, p = 0.3, power of association 41%; Table 2). The DGC risk for AA carriers remained significant, when ORs were adjusted for age, sex, alcohol intake and H. pylori infection (Table 3). In smokers, however, the associated risk was only of borderline significance (p = 0.089, Table 3). The risks associated with the other variants studied remained non-significant following adjustment (data not shown). No association was observed between the 163+37235G>A SNP and age at diagnosis (p > 0.16).

To determine whether the CDH1 163+37235A-allele confers DGC risk independently or in combination with the other intron 2 variants, haplotypes resulting from the five polymorphisms were reconstructed and their frequencies were estimated in the cases and controls. The two 5'-variants were not informative and were thus excluded. The intron 2 haplotypes showed a global association with disease (df = 7, χ² = 24.09, p < 0.002). In general, haplotypes containing the 163+37235A-allele were more frequent in cases compared to controls, while three of the four haplotypes with the G-allele were more frequent among the controls (Table 4). The strongest association with disease was observed for the AAG and the ATC haplotypes (with 163+37235A at position 1). Conversely, the GAG and the GTC haplotypes showed the strongest protection. Linkage disequilibrium analysis indicated largely absent LD between the investigated variants (r² < 0.03; Table 5), suggesting that the CDH1 163+37235G>A SNP may confer an increased susceptibility towards DGC independently of the other variants investigated.

Eighteen of the healthy controls related to the DGC patients were children. A few of them may develop DGC later in their life. We therefore repeated the analysis excluding all 18 children from the controls. The associations between the 163+37235A-containing genotypes/haplotypes with disease were similar to those obtained without exclusion. As expected, however, all associations were more significant and corresponding risk estimates increased (data not shown).

Inherited genetic risk is believed to be a crucial factor contributing to the incidence of sporadic DGC. Little is known, however, about susceptibility loci that may confer DGC risk without evoking an apparent family history. CDH1, coding for the epithelial adhesion molecule E-cadherin, is one of the few known genes to have an etiologic role in DGC. So far, only one polymorphism located within the CDH1 promoter has been implicated in the sporadic DGC risk of certain populations [10‐13]. In this study, we sought to assess a possible contribution of CDH1 intron 2 variants to disease risk, as this region has recently been shown to be essential for normal CDH1 transcription during adult life, similar to the CDH1 promoter.

Our present results suggest a role of CDH1 intron 2 alleles in the risk of developing sporadic DGC and identify the 163+37235G>A SNP as a putative susceptibility variant. Both individual genotype data and corresponding haplotype data are consistent with a contribution of the CDH1 163+37235A-allele to DGC risk that is independent of the other four investigated CDH1 intron 2 variants.

A strength of our study is the inclusion of healthy relatives into the control group. Unaffected relatives are expected to share more genetic variants with their related patients compared to unrelated, matched controls, resulting in a reduction of background genetic noise. While the genetic relation may decrease the significance level of an association, the detected genetic differences are likely more robust. Consistent with this, exclusion of relatives too young to have disease strengthened the association of the 163+37235A-allele with DGC. For all controls, the ratio of the relative frequencies of the A- and G-alleles (A:G) was 0.613, which is higher than reported for other Caucasian populations (0.38, GenomeBrowser [16]). However, the A:G ratio for unrelated controls only was 0.215, suggesting that the higher ratio is due to an enrichment of the disease allele among the relatives of DGC patients as one would expect for a high risk population. Somewhat unusual was the lack of any significant LD among the SNPs investigated. Intermarker values between the 163+37235G>A and 163+37276T>A variants were available for other populations on the HapMap Browser [21]. Strong LD was also absent in four different ethnic groups (r² range = 0.12–0.52), suggesting independent segregation of these variants may be common. Together, the data are consistent with a direct association of the 163+37235A-allele with an increased susceptibility to DGC.

A limitation of this study is its small sample size. The study was designed to detect at the 5% level of significance an OR of 2.0 for relatively common variants with power >90%. Thus, the observed lack of significant associations may well be due to inadequate power to detect variants of weaker effect, particularly with rarer alleles. Larger sample sizes will be required to conclusively assess the impact of these variants on disease risk. However, our aim was to identify common susceptibility loci that confer a relatively strong risk and hence might contribute to a significant number of apparent sporadic DGC cases. Another potential limiting factor is that the observed association is due to LD with other disease variants not investigated here. A candidate variant may be the CDH1 -160A-allele. Genotype data for the -160 SNP were available for a subgroup of the cases studied. Preliminary analysis suggested absence of LD between the 163+37235 and the -160 SNP (r² = 0.008; BH, unpublished results), consistent with an independent contribution of the intron 2 variant to disease. This idea is supported by the consistent association of 163+37235A-containing haplotypes with DGC (Table 4). Additionally, the observation that in mice the promoter and intron 2 are both independently required for CDH1 activity [14] is further evidence for an autonomous role of the intron 2 variant in DGC susceptibility. Moreover, the risk associated with the intron 2 variant remained significant following the adjustment for potential confounding factors, in support of a direct association. An exception was smoking, which reduced the associated risk to borderline significance. Of note, the risk estimates increased in both non-smokers and H. pylori-negative patients. While these trends may be due to the relatively low numbers of subjects in stratified subgroups, they may be consistent with genetic DGC risk factors being more important in individuals not exposed to environmental risks.

CDH1 intron 2 sequences are vital both to initiate transcriptional activity and to maintain E-cadherin expression in differentiated epithelia of mice [14]. Given the conserved function of E-cadherin in different species, it is very probable that intron 2 is also essential for CDH1 activity in human epithelia. This is supported by the presence of highly conserved regions within intron 2, suggesting that the conserved elements may be binding sites for transcription factors participating in CDH1 regulation. A search for putative binding sites revealed that the CDH1 163+37235 position lies within a recognition motif of the human nuclear factor I/X (NFIX, OMIM #164005). According to the TESS web page [17], the 163+37235G position (TGGCA) is the most conserved nucleotide within the NFIX recognition sequence, suggesting the 163+37235A-allele may alter the affinity of this transcription factor to the cis-regulatory element. Whether NFIX indeed is able to regulate E-cadherin expression in gastric tissue remains to be determined. In mice, however, NFIX is essential for embryonic development [22] and is expressed in many adult tissues including epithelial ones [23]. Genes regulated by NFIX have been identified and include repression of pro-angiogenic PDGFA [24] and p21, where repression surprisingly leads to growth inhibition [25]. NFIX may also slow down cell growth via downregulation of the adenine nucleotide translocase-2 gene (ANT2) [26]. In addition, NFIX has been shown to confer resistance towards transformation by nuclear but not cytoplasmic oncogenes [27]. The reported findings are compatible with a tumour suppressing role of NFIX, where reduced binding of this transcription factor to its recognition sequence may favour tumorigenic events. It is thus of some interest that the chromosomal region 19p13.3, the location of NFIX, appears frequently deleted in some [28, 29], however not all [30] series of gastric cancers.

To our knowledge, this is the first study to report an association between an intronic CDH1 variant and sporadic cancer. Similar studies will be required to confirm an association with DGC in other populations/ethnicities and, using a larger sample size, to determine whether the 163+37235A variant may also be a disease allele in cohorts from low DGC risk regions. Furthermore, case-control studies on other carcinoma types, including other gastric cancer histotypes, should address whether the intron 2 variant is specifically associated with DGC. Regarding the CDH1 promoter SNPs, the -160A-allele has not only been implicated in DGC in Italians [10, 11], but also in a Mexican [12] and a Japanese [13] population (here included in disease haplotypes). Not all reports have found significant associations, suggesting population-specific effects or differences in study design [31, 32]. To address this heterogeneity, Wang et al. have evaluated data from 11 case-control series and concluded that the -160A-allele is a gastric cancer susceptibility allele in European, but not Asian populations [33]. However, their meta-analysis did not include very recent studies reporting positive associations [11‐13] and therefore may underestimate the contribution of the -160A-allele to gastric cancer. Further positive associations have been observed with sporadic carcinomas at other sites such as prostate [34, 35], urether [36], bladder [37], breast [38], colorectum, and endometrium [11]. Another CDH1 promoter variant, -347insA, has not been associated with sporadic DGC so far, but appears to increase susceptibility to colorectal cancer [39] and to contribute to oesophageal/cardiac cancer risk [40]. Such observations imply that functional CDH1 variants may be involved in the susceptibility towards a broad range of epithelial cancers, consistent with the expression pattern of E-cadherin in tissues. It appears likely that, in many instances, CDH1 variants will require the presence of other etiologic factors to increase disease risk, as E-cadherin deficiency is usually associated with cancer progression rather than initiation [41]. Furthermore, CDH1 polymorphisms might also affect the differentiation degree of tumours; preliminary results from our laboratory suggest an overrepresentation of specific CDH1 haplotypes in poorly differentiated lung carcinomas (Emily Wilson, HSc-thesis). Therefore, the role of the intronic CDH1 163+37235A variant as a disease allele may not be limited to sporadic DGC risk.

We report the identification of a new putative susceptibility variant for DGC located within conserved sequences of CDH1 regulatory intron 2. Both individual genotype and haplotype data suggest a contribution of the CDH1 163+37235A-allele to sporadic DGC risk that is independent of other CDH1 variants within conserved intron 2 regions. A larger confirmatory study involving other populations and including complete CDH1 haplotypes will be required to assess the population-specificity and the relative contribution of the 163+37235A-allele to disease. The establishment of heritable DGC risk factors will be helpful particularly with respect to the difficult and often delayed diagnosis of diffuse-type stomach cancer. Given the universal expression of E-cadherin in epithelial tissues, this exploratory study may also provide a basis to investigate the role of the CDH1 163+37235G>A SNP in the incidence and progression of tumours other than DGC.