Assessing both genetic variation (SNPs/CNVs) and gene–environment interactions may lead to personalized gastric cancer prevention

Technological advances with next-generation DNA-sequencing and microarrays have revolutionized research on the human genome [1–3]. Three more genomes, including the first one for a cancer patient, have been entirely sequenced and published in Nature[4,5]. Could this rapid evolution in assessing human genetic variation be incorporated into cancer clinical practice [6–10] and enable a true overall personalized medicine [11]?

Genetic variation describes the differences in both the coding and noncoding portions of our DNA and is what makes each of us unique. It can also contribute to our personalized susceptibility to disease. Exhaustive analysis of human SNPs has led to the identification of interesting SNP markers for certain disorders. Although less common than SNPs, copy number variations (CNVs) – gain or loss of segments of genomic DNA relative to a reference – have also been shown to be associated with several complex and common disorders such as cancer [12,13]. There has been an explosion in biomedical research for the identification of genetic variation-based personalized management of complex diseases. Over the last 2 years, genome-wide association studies (GWAS) have identified more than 100 new chromosomal regions at which more than 165 novel DNA variants influence risk of common human diseases and clinical phenotypes [2].

Given the poor prognosis of advanced gastric cancer, primary prevention represents a major research goal. If rational research could be translated successfully into targeted screening of the general population and individualized risk prediction-based preventive intervention, we could discuss for a revolution in the clinical management of gastric cancer. Potential highly effective, risk-guided prevention strategies could dramatically reduce incidence and mortality of the disease. In an era of rapid technological advance, based on the assessment of human genomic variation through whole-genome scans and personal genomics, how realistic is this expectation [1–3]?

Incidence & mortality

Despite declining incidence in the Western developed world, gastric cancer remains a serious health problem: over 900,000 new cases and 700,000 deaths occur annually worldwide [14]. Screening programs and standardized D2 surgery in Japan, have achieved the best global survival results: 5-year survival rates of 70% in advanced stages II/III and 60% overall [15]. However, the annual age-standardized incidence rate in Japan ranges from 50 to 80 per 100,000 among men and from 20 to 30 per 100,000 among women [16]. This high incidence rate demonstrates that even in Japan (a country well-reputed for its research and clinical practice for gastric cancer), no effective primary prevention strategy has been developed and established. This failure, in contrast to dramatic advances in early detection and treatment in Japan, reflects the challenges in developing effective prevention strategies.

Risk assessment & targeted prevention

Despite efforts, no effective primary prevention is available. Discrimination of the general population and individual persons into high-, intermediate- and low-risk categories for developing gastric cancer is a first essential step in designing prevention strategies. Apparently, more aggressive preventive interventions, such as prophylactic surgery and drugs, may be justifiable for high-risk people, while lifestyle modifications are sufficient for low-risk subjects. Age of starting and intensity of endoscopic screening also varies according to risk assessment. Balancing risks and benefits of targeted screening and preventive interventions is essential, as subjects in this scenario are healthy individuals and not patients.

Familial & sporadic gastric cancer

Carcinogenesis

Does heritability or the environment cause gastric cancer? Cancer arises from the accumulation of inherited and somatic mutations in cancer-causing genes. A large, population-based study in twins has demonstrated that although environmental factors play the predominant role, inherited genetic factors may account for approximately 30% of the relative risk (RR) for most cancer types, including gastric cancer [17]. Identifying environmental and genetic factors and understanding gene–gene and gene–environmental interactions are considered prerequisites for highly effective prevention.

Lauren classification

Approximately 40 years after the introduction of Lauren classification [18], a recent GWAS confirmed that the intestinal and diffuse types are two distinctly different subtypes of gastric cancer [19]. The intestinal type appears to be associated with Helicobater pylori infection that may lead to atrophic gastritis, intestinal metaplasia, dysplasia and invasive adenocarcinoma, and is particularly common in some high-risk geographic regions, including Japan. By contrast, the diffuse-type develops through a different pathway. The genetic background is more important than for the intestinal-type and appears unrelated to the presence of H. pylori. It has a much more uniform geographical distribution [20]. Genetic factors have an important role in tumorigenesis.

Prevention of rare hereditary diffuse gastric cancer

On the basis of family history, gastric cancer can be discriminated into familial cancer and sporadic cancer (no family history). The year 1998 saw a breakthrough: inherited mutations in the high-penetrance CDH1 gene caused hereditary diffuse gastric cancer [21]. Following this discovery and the currently available genetic test results, gastric cancer was classified into familial CDH1-positive and familial CDH1-negative diffuse gastric cancer. Individuals with inherited CDH1 mutations face a very high lifetime-risk of approximately 75% for gastric cancer and an additional 40% risk of lobular breast cancer in women [22]. The E-cadherin protein encoded by the CDH1 gene is important in the molecular connections between adjacent cells in the stomach, breast and other areas of the body. Loss of E-cadherin function, associated with CDH1 mutations, is observed in diffuse-type gastric cancer and also in invasive lobular breast cancer [23].

In approximately 70–90% of patients, gastric cancer is sporadic and 10–30% is familial. Among familial gastric cancer cases, only approximately 1–3% of all cases are carriers of mutations in CDH1[24].

Results of genetic CDH1-testing now increasingly guide decision-making for prevention. A characteristic example of the rapid incorporation of the discoveries in basic science, into clinic practise, is prophylactic total gastrectomy. It is thought to be a safe and highly effective preventive intervention, with an acceptable level of quality of life for CDH1-mutation carriers [25,26].

Prevention of familial non-CDH1 cancer & sporadic gastric cancer

As reported in less than 3% of all gastric cancer cases attributable to CDH1 mutations, guided primary prevention is currently feasible. For the remaining (more than 97%) cases, no effective targeted prevention strategy is available.

It is thought that interaction between H. pylori infection and dietary risk factors leads to gastric carcinogenesis. Therefore, considering these factors, the conduction of preventative, randomized trials in high-risk populations in certain regions of China and Japan would be effective [16]. However, such prevention trials with H. pylori eradication and/or nutritional interventions in high-risk populations were negative [27].

A series of reasons may explain the failure of prevention trials. Two key issues in the design of these trials were not considered: Lauren classification [18]; and genetic factors [17] and gene–environment interactions. Perhaps, before the conduction of randomized trials, several research steps are required. Table 1 summarizes these steps including identification of risk alleles and assessment of gene–gene and H. pylori/diet–genetic risk variants. The advent of new high-throughput technologies hold major promises for effective research directions.

Whole-genome scans

The completion of HapMap II project with the characterization of more than 3.1 million SNPs in the human gnome and the availability of genotyping platforms have enabled the performance of GWAS for identifying new susceptibility genes and variants. Indeed, genome-wide association-unbiased studies have identified several susceptibility alleles and many risk variants for breast, prostate and other common cancer types [28,29].

Recently, The Study Group of Millennium Genome Project for Cancer reported the results of a new GWAS with the identification of the PSCA gastric cancer susceptibility gene [19]. This two-stage GWAS used 85,576 SNPs on 188 patients with diffuse gastric cancer and 752 references in stage 1, and 2753 SNPs on 749 patients and 750 controls in stage 2 in Japan, and identified an association of a SNP (rs2976392) in the PSCA with diffuse gastric cancer. Resequencing of the PSCA region in the affected individuals revealed a number of SNPs that increase a person’s risk of diffuse gastric cancer, even if there is no family history (sporadic cancer). In a replication study in Korean populations, the authors also found significant association of PSCA SNPs with diffuse gastric cancer.

Unsurprisingly, the RR of all the genetic variants in PSCA in this study was smaller than two [17]. Indeed, this finding is consistent with most GWAS performed for cancer or other common diseases [2]. Although the flood of discoveries through whole-genome analysis is impressive, there is currently limited or no clinical implication due to the small effects of these risk variants [30].

Future: personalizing preventive interventions

It is felt that GWAS may help to achieve the major goal of this century: personalized risk prediction and response to preventive intervention for dramatic cancer-incidence reduction. However, there are too many challenges and hurdles to overcome.

First, the goal is to complete the genetic mapping, separately for intestinal and diffuse-type gastric cancer. According to the polygenic model, although a small risk is irrelevant, the combination of several low-risk alleles can lead to substantial risks, even in the absence of multiplicative statistical interactions [31]. For example, Vineis et al. estimated that one risk-allele confers a RR of 1.25; 10 risk-alleles confers a RR of 3.50; 20 risk-alleles confers a RR of 6.00, and 100 low-risk alleles confer a RR of 26.00. Therefore, the search for new risk-alleles is clinically important. It is logical to expect that future GWAS using modern genotyping platforms with more than 1 million SNPs or CNVs will be able to identify a dozen or more, further genetic-risk variants. The problem of false-positive and false-negative results in GWAS provides a challenge, but with large-scale GWAS using appropriate methodology, the numbers of false results can be reduced. Appropriate replication studies have already been integrated into GWAS and are considered essential for unbiased studies.

Second, once the main genetic effects have been concretely documented, the next step is how to proceed with the investigation of gene–gene and gene–environment interactions. New population-based prospective studies considering the intestinal and diffuse-type cancers separately will be required. These studies will evaluate the interactions between a large umber of low-risk alleles, such as PSCA and established environmental factors. Among classic risk factors beyond H. pylori infection, recent data from epidemiologic, experimental and animal studies indicate that diet plays an important role in the etiology of gastric cancer [16]. High intake of fresh fruits and vegetables, lycopene and lycopene-containing food products may reduce the risk of gastric cancer, while high intake of nitrosamines, processed meat products, salt and salted foods, and being overweight or obese are associated with an increased risk of gastric cancer [32].

It is possible that what really counts is not the main effect of genes or environmental risk factors, but their complex interactions. Finding and interpreting such interactions is not straightforward and no standard is available yet. For example, the reduction of the multifactor dimensionality is one of the algorithmic approaches proposed as data analysis but it has several limitations [8,9].

Finally, continuous updated integration of all evidence, from both old studies, current GWAS and future replication studies, and careful interpretation of the strength of the evidence are crucial to maximize transparency and lead to informative selection of the next steps of research in this field.

Major familial predisposition gene other than CDH1

At 10 years from the identification of the CDH1 gene causing hereditary diffuse gastric cancer, no other major familial susceptibility gene has been identified. Although the probability is limited, the existence of such a gene, for example, for familial intestinal gastric cancer cannot be excluded. Indeed, most GWAS have identified low-risk alleles. However, most recently, Mosse et al. demonstrated using a whole-genome scan, that heritable mutations of the anaplastic lymphoma kinase gene are the main cause of familial neuroblastoma [31], and that germ-line or acquired activation of this cell-surface kinase is a tractable therapeutic target for this lethal pediatric malignancy [33]. Therefore, it cannot be excluded that a high-penetrance gene exists in which inherited mutations cause hereditary cancer, for example inherited intestinal gastric cancer. If it is true, then future appropriate GWAS may discover such a gene.

Development of robust markers to predict risk and response to medical interventions in both prevention and treatment of cancer for clinical practice has been proved to be an extremely complicated and difficult translational research task. Recent technological advances, including next-generation DNA sequencing and microarrays, have revolutionized basic research. Although exciting findings are now provided by genetics, epigenetics, global gene-expression signatures, pharmacogenomics and, most recently, personal whole-genome scans [1–7,10,12,13,19,21,28–31,33–36], there are still numerous open challenges. Appropriate methodology, statistics and models are proposed; but at present, most of the exciting translational research findings cannot be integrated into clinical practise, personalizing prevention, or treatment decision [8,9,30].

Personal genome

The age of personal genomics is here [4]. Three more human genomes have been entirely sequenced; the first African, the first Asian and the first cancer patient [5], and published in Nature[4].

In order to explore the genetic underpinnings of cancer, Richard Wilson and colleagues sequenced genomes from both normal skin tissue and tumor tissue of a middle-aged woman who died of acute myelogenous leukemia [5]. They compared the DNA to determine what was different about the cancer cells. Approximately 97% of the 2.65 million SNPs found in the tumor cells also existed in the normal skin cell, suggesting they were not critical to the cancer process. The researchers also eliminated SNPs that had been previously identified elsewhere, as well as those that did not change the coding of a gene, ending up with ten SNPs unique to the tumor cells.

Two SNPs occurred in genes previously linked to this leukemia. Eight SNPs led the researchers to new candidate acute myelogenous leukemia genes, including several tumor-suppressor genes and genes possibly linked to cell immortality. By sequencing the whole cancer genome, the authors believed that they captured what they did not know as well as what they did know regarding cancer genes. As new techniques develop, it is hoped that it would become faster and cheaper to perform entire genome sequencing with less than US$10,000. It is also hoped that in the next few years, the number of whole-personal cancer genomes for various cancer types including gastric cancer will increase dramatically, thus transforming our ability to understand cancer. In this wealth of information however, data-analysis synthesis and interpretation will require new complex mathematical models, methodologies and algorithmic approaches for robust conclusions.

Challenges regarding legal, ethical and the ability of this genomic research to translate into clinical utility are also addressed in a selection of news and opinion articles. The once fantastical notion that any person can walk into a doctor’s office with his or her genome on a hard drive, appears more and more of a reality. We cannot predict everything that will happen next, but we can be prepared [4].

Conclusions

Risk stratification is considered a major clinical utility for both targeted screening of the population and personalized preventive intervention in individual persons. However, at present, such a personalized strategic management for reducing the incidence of gastric cancer is unrealistic. A first step toward a person’s risk prediction has been taken with the discovery of the CDH1 gene. Individuals with heritable mutations in this major familial predisposition gene face a very high lifetime risk of diffuse gastric cancer and benefit from prophylactic total gastrectomy rather than gastroscopic screening. But familial CDH1 cancer accounts for only 1–3% of gastric cancer patients.

What are the perspectives in the prevention of the vast majority of the remaining patients? New GWAS using modern genotyping platforms with over 1 million SNPs/CNVs will identify a large number of low-risk alleles.

If there is another gene with large risk effects such as CDH1, for example for intestinal gastric cancer, it can be identified by GWAS with important clinical application. The next challenge is how to assess complex interactions between gene–gene and gene–H. pylori–dietary factors. Several sophisticated data analysis and statistical models are proposed but there is no standard approach. Finally, large-scale, prospective, population-based studies recording all classic clinicopathologic factors and new genetic factors identified by GWAS, may lead to risk-assessment, targeted screening of the population and personalized prevention.

Whether and when sophisticate algorithms combining genetic and environmental data analysis will be able to predict both a person’s risk of cancer or other common complex disorders such as cardiovascular disease and response to a given lifestyle modification or preventive intervention is difficult to predict. It is a subjective, rather than objective thought of an individual scientist’s biologically optimistic or pessimistic nature.

Table 1. Personal genomics, complex models for assessing interactions and challenges in designing large-scale prevention clinical trials.

Gastric cancer	Identified genetic variants in cancer-related genes	Research steps
		First : GWAS	Second: assessing interactions	Third: new prospective studies
Familial
Hereditary diffuse gastric cancer	CDH1: Rare, high-risk gene	Identification of novel low-risk modifier genes	CDH1-novel low-risk genes interaction
Non-CDH1 cancer	NA	✓ Identification of many low-risk alleles ✓ Although the probability is small, the identification of a new high-penetrance gene, for example for intestinal gastric cancer cannot be excluded	Combining available (MDR) and new sophisticated statistical models for assessing: ✓ Gene–gene multiplicative risk effects ✓ Gene–H. pylori–diet interactions	New, large-scale, population-based studies recording all clinical, medical, family history and new genetic data Data analysis for assessing interactions rather than genetic or environmental factors alone
Sporadic
Diffuse	PSCA: common low-risk gene	✓ Identification of multiple novel low-risk alleles	Combining available (MDR) and new sophisticated statistical models for assessing: ✓ Gene–gene multiplicative risk effects ✓ Gene–H. pylori–diet interactions	New, large-scale, with no family history and diffuse gastric cancer-based studies recording all clinical, medical, and new genetic data Data analysis for assessing interactions rather than genetic or environmental factors alone
Intestinal	NA		Combining available (MDR) and new sophisticated statistical models for assessing: ✓ Gene–gene multiplicative risk effects ✓ Gene–H. pylori–diet interactions	New, large-scale, with no family history and intestinal gastric cancer-based studies recording all clinical, medical, and new genetic data Data analysis for assessing interactions rather than genetic or environmental factors alone

GWAS: Genome-wide association studies; MDR: Multifactor dimensionality reduction.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.