Introduction

Gastric cancer (GC) is the fifth most common malignancy and the third leading cause of cancer death worldwide.1, 2 With a 5-year survival rate merely ranging from 25 to 35% in advanced GC,3, 4, 5 more than 1.1 million patients die from GC every year.6 Although the decrease of incidence and mortality rates in GC has been reported because of a lower prevalence of Helicobacter pylori (Hp), lower salt intake and a higher consumption of fresh fruits and vegetables,7 GC still remains a main clinical challenge as many cases are diagnosed in advanced stages with limited treatment options and poor prognosis.8 Although therapeutic methods are improving in surgical combined with radiotherapy and chemotherapy, the benefits for advanced GC are still not optimistic. The improvements in early diagnosis and the treatment of the GC may continue to be the most effective strategy for improving patient survival. Thus, seeking for more sensitive detecting approaches and effective drugs, particularly those targeting cancer progression mechanisms, is urgently needed.

A recent phase III randomized study (ToGA) revealed that the addition of trastuzumab to chemotherapy improved survival in patients with advanced GC with HER2 gene amplification.9 It not only laid the foundation of gene detection in the diagnosis and treatment of GC, but also indicates the potential effect of targeted therapy against gene copy number variations (CNVs) in GCs.10, 11

CNV refers to a form of genomic structural variation that results in abnormal or, for certain genes, a normal variation in the number of copies of one or more sections of the DNA.12 DNA CNVs include gene amplification, gain, loss and deletion. In addition to gene mutation, CNV has a significant role in tumorigenesis in many cancers, such as GC,13 ovarian cancer,14 hepatocellular carcinoma,15 testicular germ cell tumors,16 colorectal carcinoma,17 bladder cancer18 and so on. The accumulation of CNVs during gastric oncogenesis may be a result of preferential selection by which transforming cells gain evolutionary advantage. The copies of apoptosis effector genes are often lost during cancer development, in comparison with the frequent amplification of proliferation-related genes.19 A recent study on 183 primary GC samples suggested that some established or potential anticancer drug target genes exhibited high levels of CNVs, including HER2, TUBB3 and TOP2A.20 Given the indicative value of CNVs in deregulated signaling pathways, CNV may provide useful information for the molecular subtyping of GC and optimization of therapeutic strategies.21 Increasing evidence showed that CNV genes are promising biomarkers and therapeutic targets in GC, which can be detected by various methods including flourescence in situ hybridization, array comparative genomic hybridization and single nucleotide polymorphism arrays.

Although a number of review articles have focused on the roles of specific genes that are relevant to molecular pathogenesis or targeted therapy of GC,22, 23 a systematic review with respect to gene CNVs in GC has not been provided.24 In this review, we discuss the recent CNV researches that shed light on the molecular pathogenesis of GC, with a specific emphasis on CNVs and associated genes that display biomarker potentials in GC.

GC involves CNVs and other genetic aberrations

GC is a very complex and heterogeneous disease, which is a multistep process involving deregulation of many oncogenic pathways. Adenocarcinoma is the major histological type of GC, in possession of 90–95% of all gastric malignancies. According to Lauren classification, adenocarcinomas are divided into two distinct pathological entities, intestinal and diffuse types, while the latter has a more aggressive behavior and worse prognosis than the former.25 Gastric carcinogenesis is considered as a result of a complex interaction between inheritance and environmental factors. In addition to hereditary predispositions, GC is also associated with Hp infection,26, 27, 28 obesity, nutritional supplement29 and certain dietary structures, such as high salt diet, food content with nitrates and smoked meats.30 Exposure to those risk factors for a long time will eventually result in cancer through a multistep process.31 Of note, the infection of Hp is associated with certain features of CNVs in GC. As an example, the loss of 16p occurs in 10% of the Hp-negative samples compared with 0% in the Helicobacter-positive samples, whereas the gain or amplification in 16p gain can be found in up to 14.71% of Hp-positive samples but only in 3.33% of the Hp-negative samples.32

The initiation and progression of GC involve deregulation of different signaling pathways by genetic and epigenetic alterations.33, 34, 35 Genetic alterations, such as gene mutations, CNVs and chromosomal translocations, could also influence the expression of tumor-suppressor genes, oncogenes and other genes, ultimately contributing to gastric carcinogenesis.36 In recent years, many micro-RNA (miRNAs) have been supposed as oncogenes or tumor suppressors by altering the expression of target genes participated in multiple steps of primary and metastasis GC, which are related to gene deletions, mutations, promoter hypermethylation or histone acetylation as well as other mechanisms.37, 38

CNVs affect both protein-coding and non-coding genes

Previous researches have revealed numerous chromosomal DNA gains and losses in GC patients, with the former far more prevalent than the latter. Recent high-throughput studies identified gains of 3p22, 4q25, 8q24, 11p13 and 20q13, as well as losses of 1p36 and 9p21 where many cancer-related genes (CTNNB1, MYC, CDKN2A, TOP2A and so on) are located.39, 40 DNA CNVs are significant influential factors for gene expression, which may affect the activities of different oncogenic or tumor-suppressing pathways. This may explain the clinical relevance of many CNVs in GC. It has been reported that GC patients with lymph node metastasis have remarkably higher numbers of gains, losses and total CNVs than those cases without metastasis. In addition, another research indicated that frequent gains observed on chromosomes 1q, 5p, 7, 8, 13 and 20 and losses observed on chromosomes 1p, 3p, 4, 5q, 9p, 17p, 18q, 19p, 21 and 22.41 Although an increased number of CNV regions have been identified in GC,42 it still requires further investigation which of the affected genes may have functional roles in GC.13, 43

Most previous studies have focused on protein-coding genes in CNV regions, but it is increasingly likely that the expression levels of long non-coding RNAs and miRNAs are also influenced by CNVs. A recent study by Fang, Xu and colleagues44 revealed that one-third of aberrantly expressed long non-coding RNAs are associated with CNVs in the GC genome. Because long non-coding RNAs may have causative roles in oncogenesis,45 it deserves further investigation whether CNV-associated long non-coding RNAs may have oncogenic roles and present diagnostic or prognostic significance in GC. Moreover, recent studies have demonstrated that miRNA deregulation caused by CNVs may contribute to gastric oncogenesis.45 As discussed above, CNVs are clustered in different chromosomal regions and may affect the expression of different types of genes, therefore contributing to gastric oncogenesis. In the following paragraphs, we will discuss significant CNVs in GC according to their locations in the genome (from different chromosomes to mitochondrial DNA (mtDNA)).

Chromosome 3

PIK3CA gene, which is located on chromosome 3q26.3, is frequently amplified in GC.46 Importantly, the overexpression of PIK3CA resulted from gene amplification increased PI3-kinase activity and phosphorylated Akt level, contributing to aberrant cell proliferation and apoptosis which are directly associated with tumorigenesis (schematic representation in Figure 1).47 Furthermore, PIK3CA amplification notably influenced the overall prognosis in GC regardless of early or late stage tumors, suggesting that this genetic event has an important role in the multistep process of gastric carcinogenesis.48, 49 Taken together, PIK3CA may function as a GC driver with independent prognostic significance.

Figure 1
figure 1

An overview of CNV-affected genes in association with signaling pathways, pathological features and therapeutic targets in GC. The genes on CNV regions have been highlighted either in red (indicating gain or amplification) or green (loss or deletion). Their relationships with different signaling pathway have been shown with arrows. In addition, the relevance to disease features ('R'), targeted therapy ('T') and oncogenic functions ('F') have also been labeled for each gene. The signaling pathways related to the receptor tyrosine kinase genes (MET, EGFR, FGFR2 and HER2) have been labeled in the graph.

Full size image

In addition, Yoshida et al.50 detected that ribosomal protein S6 kinase2 (S6K2) amplification was associated with poor prognosis, and S6K2 may function as an upstream driver gene leading to deregulation of mammalian target of rapamycin (mTOR) in GC. Beyond that, Shinmura et al.51 suggested that TNK2 (locate on 3q29) amplification may be an independent indicator of poor prognosis in GC patients, contributing to an increase in the malignant potential.

Chromosome 5

The adenomatous polyposis coli (APC), a tumor-suppressor gene, located on 5q21-q22, has a critical role in several cellular processes including microtubule polymerization, signal transduction and cell adhesion.52 Its protein product negatively regulates WNT signaling and its inactivation leads to β-catenin accumulation and transcriptional activation of genes (MYC, cyclin D1) related to cell proliferation.53, 54 As reported, the chromosome locus of APC is frequently deleted in GCs,55, 56, 57, 58, 59 and its decreased copy number significantly associated with lymph node invasion and metastasis in GC patients.55 Moreover, APC deletion was principally found in advanced GCs, suggesting that it might be involved in the progression but not initiation of GCs. Furthermore, in a study concerning 131 sporadic gastric adenocarcinoma samples with matched normal tissues, Fang et al.60 demonstrated APC copy number deletions were found in a relatively high percentage (25.9%) and were associated with lymph node invasion or metastasis of GC.

The IRX1 tumor-suppressor gene is located on 5p15.33, a cancer susceptibility locus which is frequently deleted in GC.61 IRX1 expression suppresses cell proliferation, invasion, migration and oncogenesis both in vitro and in vivo. Guo et al.62 also confirmed the deletion of IRX1 gene and the crucial functions of IRX1 as a tumor suppressor in GC. In addition to gene copy number deletion, the expression level of IRX1 in GC also correlates with promoter methylation.61

Chromosome 7

MET locates on chromosome 7q21 that codes the hepatocyte growth factor receptor. The hepatocyte growth factor/MET pathway dysfunction has been observed in GC and many other human cancers.63 In GC, the activation of MET signaling is mainly caused by MET amplification,64 causing increased tumor cell growth, invasion and angiogenesis.63, 65, 66

The amplification of MET gene has been found in 0–23% of GCs, and is associated with advanced disease stages or worse clinical outcome.67, 68, 69, 70 In addition, a recent report showed MET gene amplification significantly associated with mRNA overexpression and poor GC patient survival.71 MET gene copy number amplification was also significantly associated with the depth of tumor invasion, metastasis and poor prognosis, suggesting it may be more valuable as a prognostic marker than protein overexpression.72 The clinical impact of MET amplification is highly consistent with the role of MET as a functional driver gene in GC. Recent studies have suggested that MET CNVs may serve as a selectable marker for MET inhibitor therapy. MET amplification labels a subgroup of GC patients who are susceptible to MET-TKIs.66, 73 In addition, the investigation of volitinib as a therapeutic option for GC patients points out a strong rationale for selecting patients harboring amplified MET.74 In future clinical studies, the effectiveness of MET amplification as a marker for treatment response should be further explored.69

Epidermal growth factor receptor (EGFR) resides on chromosome 7p12 and encodes a receptor tyrosine kinase ErbB and the other members are HER2, HER3 and HER4 (signaling related to receptor tyrosine kinases are shown in Figure 1).75 Inhibition of EGFR contributes to cell division, migration and apoptosis in GC.76, 77 EGFR copy number gains associate with an increased risk of invasion and metastasis in solid tumors including GC, suggesting its potential significance as a prognostic marker.78, 79, 80, 81 As reported, there is a strong relationship between EGFR gene copy number, protein expression and chromosome 7 polisomy.82 Oh et al.83 further observed that EGFR CNVs existed in a series of GC cases and discovered that it was associated with unfavorable prognosis. Another study based on 855 cases reported that gained EGFR gene copy can be found in 22.7% of GC patients, and it associates with poor disease outcome.84 These data collectively demonstrate that EGFR CNV may be a valuable biomarker for GC.

Chromosome 8

The c-MYC gene is located in this chromosome at the 8q24.1 band, encoding a transcriptional factor that regulates genes related to proliferation, differentiation and apoptosis. The deregulation of c-MYC has been considered as one of the main events in the pathogenesis of many cancers, including GC.85, 86 Many studies reported that a significant increase in MYC copy number can be detected in the carcinogenic process of GC and in gastric cell lines.87, 88, 89, 90, 91, 92 The amplification of MYC has also been suggested with independent prognostic value on overall survival.93 In addition, MYC CNVs has a tight connection with the clinicopathological features of GC. Wang et al.94 reported that gained copy number in MYC or TNFRSF11B (located at 8q24) genes strongly associated with the depth of lymph node metastasis, invasion and TNM stages. In another study, MYC amplification significantly correlated with MYC mRNA levels and MYC immunoreactivity, suggesting that MYC CNV indeed contributes to its overexpression in GC.95 In mucinous gastric carcinoma, c-MYC amplification was correlated with greater invasion depth and advanced tumor stage, while these differences were not found in non-mucinous gastric carcinomas, suggesting that c-MYC amplification in mucinous gastric carcinoma may be a genetic alteration contributing to the frequent presentation of advanced stage of MGC.96

The POU5F1B gene (POU domain class 5 transcription factor 1B, the OCT4 pseudogene), which is located on human chromosome 8q24 near MYC,97 is frequently amplified in GC. Hayashi et al.98 detected that POU5F1B copy number is amplified and overexpressed in GC, and it also promotes tumorigenicity and tumor growth. Therefore, POU5F1B amplification seems to be a GC-associated event that has oncogenic roles.

Chromosome 17

HER2 is located on human chromosome 17q21 and is a member of the EGFR family. Once bounded to its ligand, Her-2 is phosphorylated and it functions as a tyrosine kinase that promotes cell proliferation.99 The reported HER2 amplification in patients with GC ranged from 6 to 23%, and amplified HER2 gene was mostly associated with poor outcome.100, 101 HER2 gene amplification was significantly correlated with the depth of invasion, lymphatic metastasis and the TNM stage.102, 103 A recent report demonstrated H. pylori CagA may induce overexpression of the Her2 protein by increasing HER2 DNA copy number,104 thus adding to the connection between environmental factors and genomic aberrations. As high incidence of intratumoral HER2 heterogeneity has been reported, it is important to detect HER2 gene status with larger tissue samples.42, 105 Many methods for detecting HER2 CNVs, including flourescence in situ hybridization, in situ hybridization, chromogenic in situ hybridization and silver in situ hybridization, have been demonstrated as effective approaches.42, 101, 106 Moreover, the HER2:chr17 (chromosome 17) ratio may be an additional index to eliminate incorrect HER2 status determination in GC.107

The utilization of molecularly targeted therapeutics against HER2 has emerged as a significant strategy for advanced GCs. Trastuzumab, a monoclonal antibody targeting HER2, induces cellular cytotoxicity and inhibits HER2-mediated signaling pathways.108 This HER2 antibody has been tested by a randomized clinical trial (ToGA), which reported prolonged survival time of patients with advanced GC after combined with chemotherapy.9, 109 The European Medicines Agency has also approved trastuzumab in association with chemotherapy for the treatment of metastatic gastric adenocarcinoma with minor modifications of the ToGA trial criteria.110

TP53 gene mapped on 17p13.1 encodes a master regulator of genomic stability.111, 112 In response to DNA damage, the p53 protein triggers multiple cellular responses, including cell cycle arrest, DNA repair and apoptosis, cellular differentiation, metabolism, angiogenesis and the immune response.113, 114, 115 In the carcinogenic progression of GC, loss of the TP53 locus is one of the most common mechanisms involved in this gene dysfunction and is frequently found in GC.87, 88, 90, 116 A significant correlation has been found between loss of TP53 and gastric precancerous lesions, suggesting that TP53 CNV may be an early event in gastric carcinogenesis.117

Chromosome 20

The centrosome-associated kinase aurora A (AURKA) gene is located on chromosome locus of 20q13, encoding the Aurka protein that is ubiquitously expressed and regulates cell cycle events emerging from late S-phase through the M phase.118 In addition, AURKA overexpression results in the activation of several carcinogenic pathways including PI3K/AKT, β-catenin, NF-kB and JAK2-STAT3.119 Accumulating data revealed that AURKA was frequently amplified and overexpressed in GC,36, 120, 121, 122 and AURKA amplification associated with significantly worse survival.123 A recent study showed copy number gains of AUKRA were detected in a relative high percentage of GC samples (30.5%). A positive connection has been found between AURKA amplification and tumor progression,124 suggesting that AURKA may have prognostic significance.

The amplification and gain of C20orf11 gene at 20q13.33 almost discriminated moderately differentiated GC from poorly differentiated type,125 and C20orf11 CNV is correlated with TNM stages and histological subtypes of GC. It is helpful in highlighting this interesting gene as a potential marker for the differentiation status of GC.

CNVs on other chromosomes

The octamer transcription factor 1 (OCT1) gene locates on human chromosome 1, encoding Oct1 protein that belongs to the POU homeodomain family of transcription factors.126 Interestingly, OCT1 shares similar downstream target genes as the OCT4 pluripotent factor, and OCT1 has been reported as a determinant of somatic and cancer stem cells.127 High expression of OCT1 could activate synbindin, which promotes ERK phosphorylation on the Golgi apparatus.128, 129 An recent study revealed that OCT1 overexpression by amplification triggers synbindin-mediated ERK signaling and increases the aggressiveness of GC cells. The amplification, mRNA and protein overexpression of OCT1 were consistently correlated with poor survival of GC patients.130 These findings suggest that OCT1 may function as an oncogenic driver in GC, and it may be a promising diagnostic and prognostic marker for this deadly disease.

Kang et al.131 successfully identified the AMY2A gene as a 1p21.1 homozygous deletion target in GC. It is considered that the AMY2A gene may function as a tumor suppressor. Gastrokine 1 (GKN1), located on chromosome 2p13.3, has been found as a potent tumor suppressor that regulates gastric epithelial cell growth.132 Loss of GKN1 gene copy number has been frequently observed in GC,133 suggesting that GKN1 inactivation may be involved in GC development.

FGFR2 gene locates on 10q26, and amplifications of FGFR2 (reported in 4–10% of GC) associates with poor prognosis in diffuse type GC.134, 135 FGFR2 encodes a receptor tyrosine kinase regulating cell growth and development.136 FGFR2 amplification was detected in 4.1% GC using formalin-fixed paraffin-embedded samples, which associated with poorer outcome. In addition, the development of FGFR2 inhibitors for the treatment of GC in consideration of FGFR2 amplification has been proposed.137

The variously sized 11q13.3 amplicon containing cyclin D1 (CCND1) and oral cancer overexpressed 1 (ORAOV1) are among the most frequent amplification events in GC.138, 139 The oncogenic role of CCND1 in GC is supported by its function on gearing the cell cycle from G1 phase to S phase.140 Stahl et al.141 found that CCND1 amplification often represents an early event during tumor development. Another study also suggest that the CCND1 gene may have a critical role in the development or progression of GC.142 Kang et al.142 reported that the ORAOV1 gene at the 11q13.3 region is associated with lymphatic metastasis, suggesting that ORAOV1 may have prognostic significance in GC.

Interestingly, it has been reported that miR-23a in amplified 19p13.13 loci targets metallothionein 2 A (MT2A) and promotes growth in GC cells.143 By integrating CNV and miRNA profiles in the same samples, the authors identified eight miRNAs (miR-1274a, miR-196b, miR-4298, miR-181c, miR-181d, miR-23a, miR-27a and miR-24-2) that were located in the amplified regions and were upregulated in GC. The amplification of miRNAs were confirmed by real-time PCR and in situ hybridization assays. Knockdown of miR-23a expression neutralized the effect of CNV and inhibited GC cell proliferation, suggesting potential therapeutic value of CNV-associated miRNAs in GC.

Mitochondrial DNA (mtDNA)

Human mtDNA is a 16.6 kb double-stranded circular DNA molecule, with a range of few hundreds to several thousands copies of mtDNA present in each cell, encoding 13 polypeptides of respiratory enzyme complexes, transfer RNAs and 2 ribosomal RNAs required for protein synthesis in mitochondria.144 Several somatic mutations in the mtDNA have been observed in GCs, including a very large deletion of 4977 bp and mutations in the D-loop region.145 Mitochondrial dysfunction by mtDNA somatic mutations and CNVs might have an important role in the malignant progression of GC owing to its important roles in energy production, cell metabolism and apoptosis.146 MtDNA copy number losses and point mutations are the two most common type of mtDNA alterations in GCs.147 Significant efforts have been made to develop therapeutic strategies by targeting mitochondria in cancers.148, 149

Interestingly, Fernandes et al.150 discovered that mtDNA quantification approaches by blood sampling would allow an early detection of GC, suggesting the diagnostic potential of mtDNA CNV. In addition, Wen et al.151 demonstrated that the mtDNA copy number deletion may be particularly notable in ill-defined GCs of clinicopathological stages III and IV.

Methods for discovering cancer-driver genes with CNVs

One major challenge in genome-wide CNV research is to identify cancer-driver genes that cause functional abnormalities. In recent studies, efforts have been made to pinpoint cancer-driver genes by combining cancer genomic data including gene CNVs, mutations and expression levels. The dominant effects of cancer-driver genes (DEOD) algorithm has been developed based on a partial covariance selection approach, which builds a gene network based on the above-mentioned data types.152 In comparison, the DawnRank algorithm was designated to identify personalized driver genes in cancer.153 Along with the expansion in cancer genomic datasets and continuous improvement of data-mining methods, the identification of cancer-driver genes may bring enormous therapeutic opportunities in the future.

Conclusions

GC is a complex, multistep process that involves aberrant CNV events in different genomic regions. Frequently occurring CNVs in GC result from preferential selections that favor the oncogenic process, thus the CNV-associated genes should be further characterized for their roles in gastric oncogenesis. Given the rich information that CNVs may provide in regard to disease signaling patterns and clinicopathological features, future studies in this field would provide enomous mechanistic insights and facilitate the development of novel biomarkers for this deadly disease. Challenges in accurate detection of CNV in the presence of intratumoral heterogeneity should be further tackled to obtain highly confident CNV information for guiding GC diagnosis, prognosis and targeted therapy.