Resolving of gastric cancer aetiology: an update in genetic predisposition

Abstract

Each year, gastric cancer accounts for nearly one million new cases and over 720,000 deaths worldwide. Prognosis is dismal as most patients are diagnosed with advanced gastric cancer. Gastric cancer outcomes will benefit from better ways to identify at risk individuals. An approach to target high-risk populations is to identify those who are genetically predisposed to gastric cancer as up to 15% of all patients report family history of the disease. Based on clinical manifestations, three gastric cancer syndromes have been described but recent findings suggest that the diagnosis of some of these syndromes is sub-optimal and it could benefit from genetic information. Using genome-wide association and next generation sequencing studies, the last decade has seen the identification of low penetrance and high-risk genes that have considerably increased our understanding of gastric cancer risk. PALB2 has emerged as new familial gastric cancer gene from these studies. Furthermore, recent genetic analyses in sporadic patients suggest that >10% of all cases have pathogenic mutations, a finding of great importance for cancer aetiology. In this article, we provide a comprehensive review on the role of genetics in gastric cancer aetiology and the implications of recent genetics findings for the prevention of this malignancy.

Introduction

With nearly one million new cases diagnosed per year, gastric cancer is the fifth most commonly diagnosed cancer worldwide.^1–3 After lung and liver cancer, gastric cancer is one of the most deadliest malignancies, each year accounting for >720,000 deaths.³ One third of all gastric cancers are diagnosed in the cardia, the mucosa originating about five centimetres from the gastro-oesophageal junction,⁴ and two thirds arise in the mucosa distal to the cardia (non-cardia cancers, Figure 1)^5,6 Histologically, gastric cancer is classified into intestinal cancers, which resemble the intestinal mucosa and have glandular formations with intercellular junctions; and diffuse cancers, where cell clusters lacking cell-cell cohesion and infiltrated stroma are seen⁷. Diffuse cancers include signet ring cell adenocarcinomas, where increased intracellular mucosa presses the nucleus to the cell periphery.⁷ These anatomic and histological classifications have associations with epidemiology. Helicobacter pylori (H. Pylori) infection is closely associated with intestinal non-cardia cancers while obesity or gastro-oesophageal reflux disease (GERD) has stronger effects on cardia cancers.^1,2 Due mostly to H. pylori eradication programs, non-cardia cancer incidence is decreasing while cardia cancer incidence is increasing; particularly in countries with high obesity rates.⁶

Figure 1:

Gastric cancer anatomical classification into cardia and non-cardia cancer and the role of environmental and lifestyle risk factors. Risk factors for both cardia and non-cardia are shown in the middle of figure. They are shown in italics if they have stronger effects for cardia tumours and in bold if they have stronger effects for non-cardia tumours

The limited disparity between incidence and mortality rates highlights the poor prognosis of gastric cancer. Only a small fraction of cases are detected at early stages and, as a consequence, gastric cancer has poor outcomes.^1,2,6 Only 5% of metastatic cancer patients are alive five years after diagnosis. ^1,2,6 These dismal survival rates underscore the need to identify at-risk individuals who can benefit from early interventions. Major gaps in gastric cancer prevention and treatment remain and it is hoped that recent genetic advances will provide new tools that lead to better outcomes. Here, we provide a review of gastric cancer aetiology, with an emphasis on the role genetic predisposition.

Risk Factors for Gastric Cancer

The major gastric cancer risk factors include age, H. pylori and Epstein-Barr virus (EBV) infection, race, sex, obesity, GERD, tobacco, alcohol and family history (Figure 1). ^1,2,6 Data from the U.S. Surveillance, Epidemiology and End Results program (https://seer.cancer.gov/statfacts/html/stomach.html) indicates that >80% of cases are diagnosed in individuals 55 years (y) or older and the median age of diagnosis is 68y.⁶ Approximately 89% of all non-cardia cancers are attributed to H. pylori⁸ and 9% to EBV.⁹ Race/ethnicity is another risk factor. In the United States, the gastric cancer incidence in individuals of Latino and Asian ancestry is about 2-fold higher than in Whites.¹⁰ However, when incidence is stratified by anatomic location, cardia cancers are more often diagnosed in Whites.^11,12 Indeed, cardia cancer incidence is increasing among White men as this group is enriched for risk factors including obesity and GERD ^11,12. Male sex increases cardia cancer risk by ~5-fold while a body mass index >30 and GERD increases risk by 82% ¹³ and 2–4-fold respectively. ^1,2,14,15 Interestingly, a recent study suggested an increase in corpus (non-cardia) cancers among young White women¹⁶. Multiple behavioural risk factors have been also identified for gastric cancer. Nearly 17% (95% CI: 10·5–29·5%) of all gastric cancers have been ascribed to smoking; with this risk factor having a stronger effect on cardia cancers (hazard ratio, HR=4·10, 95% CI: 1·76–9·57 for cardia vs. 1·94, 95% CI: 1·05–3·60, for non-cardia).¹⁷ Heavy alcohol consumption has an increased relative risk of 1·52 (95% CI: 0·80–2·88)¹⁸ and the combined effect of heavy smoking and heavy alcohol consumption increases risk by 4·9-fold (95% CI: 1·90–12·62).¹⁹

Besides non-genetic risk factors, ~10–20% of gastric cancer patients show a familial aggregation of the disease and about 2–5% can be classified as hereditary cases. ^1,2,20 Twin studies suggested that 28% of the risk is accounted for by inherited factors.²¹ Despite the important role of genetic factors on risk, until very recently only one gene, E-cadherin (CDH1), ²² had been associated with inherited gastric cancer. The recent advances in high-throughput DNA analysis, however, have led to the identification of several genomic regions associated with an increased risk of gastric cancer.^23–32

Hereditary Gastric Cancer Syndromes

The three primary familial gastric cancer syndromes (See Table 1) that have been described in the literature include Hereditary Diffuse Gastric Cancer (HDGC), Familial Intestinal Gastric Cancer (FIGC) and Gastric Adenocarcinoma and Proximal Polyposis of the stomach (GAPPS). Gastric cancer is also reported in cancer syndromes such as Lynch Syndrome (LS), Li-Fraumeni Syndrome (LFS), Juvenile Polyposis Syndrome (JPS), Peutz-Jeghers Syndrome (PJS) and Hereditary Breast and Ovarian Cancer Syndrome (HBOCS). Below, we provide a brief overview of the genetics and clinical characteristics of the most common syndromes where gastric cancer cases are reported.

Table 1.

Genetic heritable syndromes associated with gastric cancer.

Syndromes			Associated Genes	Predominant Histology	Prevalence	Gastric Cancer Risk
Primary	Hereditary diffuse gastric cancer (HDGC)		CDH1 CTNNA1	Diffuse	1:1,000,000	56–70%
	Familial intestinal gastric cancer (FIGC)		APC	Intestinal	Unknown	Unknown
		Gastric Adenocarcinoma and proximal polyposis of the stomach (GAPPS)	APC	Intestinal	1:1,000,000	Unknown
Secondary		Familial adenomatous polyposis (FAPS)	APC	Intestinal	1:8,300	2.1–4.2%
	Hereditary breast and ovarian cancer (HBOC)		BRCA1 BRCA2	Diffuse	1:400 – 1:500	2.6–5.5%
	Juvenile-polyposis syndrome (JPS)		SMAD4 BMPR1A ENG	Intestinal	1:16,000 – 1:100,000	21%
	Li-Fraumeni syndrome (LFS)		TP53	Intestinal	1:10,000–1:25,000 (U.K.) 1:20,000 (U.S)	3.1–4.9%
	Lynch syndrome (LS)		MSH2 MSH6 MLH1 PMS2 EPCAM	Intestinal	1:370 to 1:2,000	11–30%
	Peutz-Jeghers syndrome (PJS)		STK11	Intestinal	1–9 / 1,000,000	29%

Hereditary Diffuse Gastric Cancer Syndrome (HDGC)

HDGC is a highly penetrant disease caused by germline mutations in CDH1, which encodes E-Cadherin. ²² E-cadherin is a transmembrane protein that plays an integral role in maintaining cell-cell cohesion, and binds intracellular catenin molecules which mediate cell growth pathways.³³ It is also involved in the development and metastases of multiple malignancies through its interactions with multiple signalling pathways.³³ Current guidelines classify families as HDGC if they meet one or more of the following criteria: 1) two or more gastric cancer cases, where at least one is the diffuse-type, 2) one case of diffuse gastric cancer before age 40y, and 3) personal or family history of diffuse gastric cancer or lobular breast cancer within a single individual before age 50y. Testing for CDH1 mutations should also be considered for: 1) families or persons with a history of lobular breast cancer under age 50y, 2) individuals with or family history of cleft lip/cleft palate and diffuse gastric cancer, or 3) individuals with signet ring cells in the stomach as diagnosed by an expert pathologist.³⁴

Forty percent of HDGC cases carry CDH1 mutations^6,35, which confer a gastric cancer lifetime risk of 70% (95% CI: 59–80%) in men and of 56% (95% CI: 44%−69%) in women.³⁶ Women with CDH1 mutations also have a 42% lifetime risk of developing lobular breast cancer.^35,36 The median age of gastric cancer diagnosis in CDH1 mutation carriers is 38y.^6,35 Due to poor prognosis, CDH1 mutation carriers are advised to have a prophylactic gastrectomy or to have frequent endoscopic surveillance following a specialised high-risk protocol, as endoscopy is not a fail-safe and may miss cancers.^34,35,37 In fact, 90% of prophylactic gastrectomies in CDH1 mutation carriers have signet ring cell gastric adenocarcinoma even after being cleared by endoscopy underscoring the need for a surgical intervention among CDH1 carriers.³⁸ In 2013, evidence for a second HDGC gene was provided when Majewski et al ³⁹ identified a CTNNA1 mutation in a CDH1-negative HDGC pedigree. However, CTNNA1 mutations are rare having not been consistently found in HDGC and not often screened for in the clinical setting.^{25,28,31,34,36,40–43}

Gastric Adenocarcinoma and Proximal Polyposis of the Stomach (GAPPS), now a subset of Familial Adenomatous Polyposis Syndrome

GAPPS was a recently described syndrome characterized by an autosomal dominant inheritance in which patients have over 100 fundic polyps and gastric cancer.^6,35 Gastric cancers are associated with such polyps. Owing to its rarity, lifetime gastric cancer risk in GAPPS was unknown and its diagnostic criteria were not derived from international consensus.^44–47 The originally proposed GAPPS diagnostic criteria stated that cases with other polyposis syndromes such as Familial Adenomatous Polyposis Syndrome (FAPS), a condition caused by APC mutations, should be excluded.^45–47 However, recent studies identified APC mutations in seven families, which suggested that GAPPS is a subset of FAPS^44,45 and not a standalone gastric cancer syndrome.

Familial Intestinal Gastric Cancer (FIGC)

FIGC is characterized by intestinal-type gastric cancer and follows an autosomal dominant inheritance pattern.³⁵ Current FIGC diagnostic criteria ⁴⁸ is stratified on the basis of the population gastric cancer incidence. In high incidence countries, such as Portugal and Japan, FIGC is diagnosed if three criteria are met: 1) at least three relatives with intestinal gastric cancers and one of them is a first-degree relative of the other two, 2) gastric cancer patients occur in at least two generations, and 3) at least one patient is diagnosed with gastric cancer before age 50y. In low-incidence countries, such as the United States, FIGC is diagnosed if at least two first/second-degree relatives have had intestinal gastric cancer, one by age 50y, or if the family includes three or more first/second-degree relatives diagnosed with gastric cancer at any age.⁴⁹ Unlike HDGC or GAPPS (a subset of FAPS), the genetic basis of FIGC remains unknown. Since there is no clear evidence for the best approach to manage this syndrome, surveillance is recommended in a research environment and should involve regular endoscopy.⁵⁰

Other Syndromes where Gastric Cancer is Diagnosed

In addition to the primary gastric cancer syndromes discussed above, six cancer syndromes (Table 1), including four gastro-intestinal (GI) cancer syndromes (JPS, PJS, LS and FAPS) and two multi-organ/extra-GI cancer syndromes (LFS and HBOCS), have been associated with an increased gastric cancer risk.

Lynch syndrome (LS), the most common GI cancer syndrome, is an autosomal dominant inheritance condition caused by mutations in DNA mismatch repair (MMR) genes (MLH1, MSH2, MSH6, PMS1, PMS2) or by EPCAM deletions.³⁵ LS patients have a gastric cancer lifetime risk that ranges from 11–19%.^51–53 However, it is thought that incidence rates are decreasing in Europe while remaining higher in Asia suggesting an environmental component.⁵⁴ Currently, LS patients living in low risk areas are not recommended for gastric cancer surveillance because of low risk and an unproven benefit.^54–56 Furthermore, Barrow et al⁵⁵ found evidence for LS cohort effects and suggested that gastric cancer risk in those born after 1935, a cohort where incidence rates decreased, was not high enough to not justify surveillance. Juvenile Polyposis Syndrome (JPS) is an autosomal dominant disease caused by SMAD4, BMPR1A or ENG mutations ^35,57 and is associated with a lifetime gastric cancer risk of 21%.⁵⁸ Peutz-Jeghers Syndrome (PJS), caused by STK11 mutations^35,59, is another syndrome that has been associated with a gastric cancer lifetime risk of 29%.⁶⁰.Familial Adenomatous Polyposis Syndrome (FAPS) is an autosomal dominant syndrome caused by APC mutations.^35,57 About 2% of all APC mutation carriers develop gastric cancer.⁶¹

Gastric cancer is also often reported in individuals with Li-Fraumeni Syndrome (LFS) and with Hereditary Breast and Ovarian Cancer Syndrome (HBOCS). LFS follows autosomal dominant inheritance and is caused by TP53 mutations. ^35,57 TP53 mutation carriers have a gastric cancer lifetime risk of 2–3%.⁶² HBOCS is an autosomal dominant condition associated with BRCA1 and BRCA2 mutations ⁶³ and has a gastric cancer lifetime risk of ~5%.⁶⁴

The role PALB2 and homologous recombination DNA repair genes in gastric cancer

During the last three years, multiple studies have provided evidence for a causal role of PALB2 and other genes involved in homologous recombination DNA repair (HRDR) in gastric cancer risk.^{25,28,31,36,42,43} PALB2 is a breast and pancreatic cancer gene^65,66 and its protein, together with BRCA1, BRCA2, and RAD51C forms a complex that repairs chromosomal breaks during DNA replication.⁶⁷ Germline HRDR gene mutations generally associate with a somatic tumour phenotype called “BRCAness”, which is characterized by multiple chromosomal alterations. Cancers with BRCAness respond to certain chemotherapies and to poly (adenosine diphosphate [ADP]) ribose polymerase inhibitors (PARPi).⁶⁸

The first evidence of a role for HRDR genes in gastric cancer was provided by Hansford et al ³⁶, who examined 183 HDGC families for mutations in 55 cancer-associated genes. Of these, 19% (n=34) had CDH1 mutations and five of the 144 (3.5%) CDH1 mutation-negative families had mutations in three HRDR genes. ATM was mutated in three families, BRCA2 in one and PALB2 in another. Lu et al⁴³ used exome sequence data from The Cancer Genome Atlas study (TCGA) in 321 sporadic gastric cancers to identify 8% of such patients as having germline mutations in HRDR genes. Recurrently mutated HRDR genes in TCGA included PALB2 and ATM, mutated in four patients each, and BRCA1 and BRCA2, mutated in three patients each.

In our own recent exome sequencing study in 28 CDH1 mutation-negative HDGC cases we identified two patients with PALB2 mutations and one with a RAD51C mutation.²⁸ After independent replication, we identified a total of 11 patients with mutations in PALB2 (n=7), BRCA1 (n=3), or RAD51C (n=1). Our evaluation of gastric cancers from PALB2 and RAD51C mutation carriers found that they had the BRCAness mutational signature, which provided further causality evidence for these germline HRDR gene mutations.²⁸

Two recent sequencing studies provided further evidence of a role of HRDR genes in gastric cancer.^31,42 Slavin et al³¹ carried out targeted sequencing of 706 cancer-related genes in 43 familial cases and identified several individuals with mutations in the HRDR genes ATM, ATR, BRCA2, BRIP and FANCC. Fewings et al⁴², carried out exome sequencing in 22 CDH1 mutation-negative HDGC families and reported patients with mutations in PALB2 (n=1), ATR/NBN (n=1) and RECQL5 (n=2). More recently, Huang et al²⁵ evaluated exome sequence data from 443 TCGA cases, a sample which included all patients previously examined by Lu et al ⁴³, and reported several sporadic gastric cancer patients with HRDR gene mutations, including seven cases with ATM mutations, five with PALB2, four with BRCA2 and three with BRCA1.

Age of onset and histology among HRDR gene mutation carriers

The sequencing studies discussed above have independently identified mutations in six HRDR genes (Table 2). ^{25,28,31,36,42} PALB2 mutations have been identified in four studies^25,28,36,42, patients with ATM^25,31,36or BRCA2^25,31,36 mutations have been reported by three studies each, and BRCA1^25,28, BRIP1 ^25,31 and ATR^31,42 have been found to be mutated by two studies each. It now feels safe to conclude that PALB2 is a gastric cancer gene⁶⁹ and that ATM and BRCA1 also represent good candidates. The causal role for BRCA2, BRIP1 and ATR, however, will require further studies. Figure 2 shows box plots with the ages of onset for gastric cancer patients with mutations in these six genes. The average age of onset for the 14 patients that have been reported with PALB2 mutations is 63y (range 46–81y). ATM Mutations, reported in nine patients, are associated with an earlier age of onset (56y, range 41–70y). BRCA1 and BRCA2 mutations have been reported in seven and four patients, respectively, who on average have gastric cancer at ages 61y and 65y, respectively. BRIP1 and ATR mutations have been more rarely reported but seem to be associated with the youngest ages of onset.

Table 2.

Cancer gene panel and whole exome sequencing studies that have independently reported novel candidate causal HRDR genes for gastric cancer.

		Hansford36	Slavin31	Huang25	Fewings42	Sahasrabudhe28
Study Type		HDGC	Familial	Sporadic	HDGC	HDGC/Sporadic
Design		Gene Panel	Gene Panel	Gene Panel	WES	WES/Gene Panel
# Patients	Total	183	43	443	22	28/331
	CDH1	34	6	1	NA	NA
	Other	16	12	38	6	3/9
	PALB2	1	0	6	1	7
	ATM	3	1	7	0	0
	BRCA2	1	1	4	0	0
	BRIP1	0	1	3	0	0
	BRCA1	0	0	3	0	3
	ATR	0	1	0	1	0

Figure 2 (lower part) also shows the histology of the gastric cancers from HRDR gene mutation carriers. The histology of these cancers is heterogeneous and involves both cancers of intestinal and diffuse histology. For example, six patients with PALB2 mutations developed intestinal cancers, five diffuse cancers and one with a mixed histology. While two of the studies reporting PALB2 mutations had CDH1 mutation-negative HDGC ^31,42 as inclusion criteria, and hence all patients included had diffuse cancers, our study and the TCGA-based study^25,28 included sporadic gastric cancers, which represent most known mutation carriers with intestinal cancers. Notably, we found instances where patients with the same PALB2 mutation developed cancers with different histology.²⁸ Hence, HRDR-deficient gastric cancers do not fully conform to the histological criteria used for HDGC and current data suggest that they have multiple histological types.

Common genetic variants conferring a high risk of gastric cancer

With the advent of the whole genome genotyping arrays, genome-wide association studies (GWAS) have been used to identify single nucleotide polymorphisms (SNPs) associated with gastric cancer risk. So far, GWAS have identified eight gastric cancer-associated genomic regions (Table 3).^{23,24,26,27,29,30,32} Below, we briefly describe the main characteristics of these regions.

Table 3.

Gastric cancer susceptibility risk variants discovered by genome-wide association studies (GWAS).

Region	Gene	SNP	P	OR (95% CI)	Cancer Type	Population	Ref
1q22	MUC1	rs4072037-A	6·28E-17**	1·35 (1·27–1·45)	1,4,5	Asian; European	23,24,27,29,32
3q13.31	ZBTB20	rs9841504-C	2·00E-09	1·32 (1·20–1·45)	1	Asian	30
5p13.1	PRKAA1	rs13361707-C	8·00E-29**	1·41 (1·32–1·49)	1,2,5	Asian; European	30,76
5q14.3	lnc-POLR3G-4	rs7712641-C	1·21E-11	1·19 (1·14–1·25)	1	Asian	32
6p21.1	UNC5CL	rs2294693-C	2·50E-08	1·18 (1·12–1·26)	1	Asian	76
	LRFN2	rs2494938-A	5·00E-09	1·18 (1·12–1·25)	1	Asian	26
8q24.3	PSCA	rs2294008-T	1·00E-20**	1·33 (1·25–1·41)	1,3,4	Asian; European	29,32
10q23.33	PLCE1	rs2274223-G	1·00E-20**	1·57 (1·49–1·65)	2	Asian	23
11q22.3	ATM	rs758081262-T rs768362387-A rs777499935-G	8·00E-10	4·74	3,4,5	European	24

Cancer Type: 1) Non-cardia 2) Cardia 3) Diffuse 4) Intestinal 5) Total

^**Reporting values from largest genome-wide significant study

Chromosome 1q22 (MUC1)

The 1q22 region, tagged by SNP rs4072037-A, was first identified as a diffuse gastric cancer susceptibility variant in Japan²⁹, independently replicated in two East Asian studies. ^23,27 rs4072037-A is a variant in the second exon of MUC1, a gene encoding a cell surface mucin that is highly expressed in the gastric epithelium and is commonly mutated in gastric cancers.⁷⁰ This risk allele causes a shift in MUC1 splicing isoforms from variant 2 to variant 3, which differs by the removal of 9-amino acids from the tandem repeat (TR) domain of the N-terminal signal peptide, due to the reduction of stem-loop stability.⁷¹ This change may affect its cellular localization; however, localization of these splice variants have not yet been investigated.⁷² The MUC1 extracellular domain has been shown to protect the murine stomach epithelial lining from H. pylori colonization.⁷³ Evidence suggests that MUC1 is an oncogene implicated in multiple malignancies through interactions with its intracellular and extracellular domains, as reviewed by Horm et al.⁷⁴ Recent studies have found that rs4072037-A also increases the risk of non-cardia cancers.^24,32

Chromosome 3q13.31 (ZBTB20)

Shi et al ³⁰ identified rs9841504-C as a non-cardia risk variant (OR=1·32), located in an intron of ZBTB20. ZBTB20 encodes a zinc finger protein which regulates alpha-fetoprotein, a tumour marker.³⁰ Little is known about how the rs9841504-C increases cancer risk, but a study showed that ZBTB20 protein expression is associated with poor prognosis in hepatocellular carcinoma.³⁰

Chromosome 5p13.1 (PRKAA1)

rs13361707-C was discovered as a non-cardia gastric cancer susceptibility SNP by Shi et al.³⁰ It increases risk by 41% and maps to a predicted promoter region in the first intron of PRKAA1, a gene that encodes the alpha 1 catalytic subunit of 5’ adenosine monophosphate-activated protein kinase (AMPK), a key regulator of the ATP-consuming biosynthetic pathways and cellular metabolism.⁷⁵ In another study, Hu et al ⁷⁶ found that rs10074991-G, a SNP in perfect linkage disequilibrium with rs11361707-C and also located in the first intron of PRKAA1, is associated with the risk of both cardia and non-cardia cancers. Neither rs13361707-C nor rs10074991-G have been functionally characterized and the effect of these variants on PRKAA1 remains unverified. As an AMPK subunit, PRKAA1 plays a key role in cancer development and proliferation as reviewed by Ross et al. ⁷⁷ TCGA data shows PRKAA1 is highly expressed in gastric cancers.⁷⁰

Chromosome 5q14.3 (lnc-POLR3G-4)

The 5q14.3 variant (rs7712641-C) is associated with a 19% increased risk of non-cardia gastric cancer³² and located within the intron of lnc-POLR3G-4, a long non-coding RNA (lncRNA). To explore possible effects of rs7712641-C on lnc-POLR3G-4, Wang et al³² carried out expression analyses in 75 matched gastric tumours/normal gastric tissues and found that while lnc-POLR3G-4 was under-expressed in cancers, there was no association between rs7712641 genotypes and lnc-POLR3G-4 expression. These analyses therefore did not support a functional role of rs7712641-C in gastric cancer. As no further functional studies have been carried out to characterize rs7712641-C, the mechanisms by which this allele increases risk remains unknown. Besides being under-expressed in gastric tumours (which suggest a tumour suppressor role), little is known about the lnc-POLR3G-4 role in tumorigenesis.

Chromosome 6p21.1 (UNC5CL, LRFN2)

The 6p12.1 region harbours two independent risk variants. Jin et al²⁶ found that rs2494938-C increases non-cardia gastric cancer risk by 18%. rs2494938 is located in the first intron of the LRFN2 gene, which encodes the leucine rich repeat and fibronectin type III domain containing 2 protein.²⁶ Located 200Kbp upstream, a second independent variant (rs2294693-A) was discovered in Asians.⁷⁶ rs2294693 is located in the first intron of UNC5CL, a gene that encodes Unc-5 family C-terminal like protein. Neither rs2494938-C nor rs2294693-A have been functionally characterized and hence the mechanisms by which they increase cancer risk remain unknown. However, UNC5CL inhibits NF-kappa-B-dependent (NF-kB) transcription, a nuclear factor that plays a central role in inflammation and cancer progression.⁷⁸ Little is known about the role of LRFN2 in tumorigenesis.

Chromosome 8q24.3 (PSCA)

Sakamoto et al²⁹ first discovered an association between diffuse gastric cancer and two markers (rs2976392-G and rs2294008-T) in PSCA, a gene encoding the prostate stem cell antigen that is often silenced in gastric cancers. The association of rs2294008-G was discovered in non-cardia cancer, while rs2976392-T was discovered in intestinal cancers.²⁹ rs2294008-G increases risk by 33% and was originally reported as a frameshift mutation in the first methionine leading to an alternate start site. ²⁹ Current PSCA gene models, however, indicate that rs2294008-G occurs 26bp upstream of the start site and creates a binding site for YY1, a transcriptional repressor, causing PSCA promoter methylation, which leads to lower PSCA transcript levels. A recent study found that PSCA expression is reduced in oesophageal squamous cell carcinoma (OSCC). In normal oesophageal cells, PSCA binds, stabilizes, and translocate the retinoblastoma 1-inducible coiled-coil 1 protein, a key regulator of cell cycle arrest and differentiation, to the nucleus thereby promoting cell proliferation.⁷⁹ Hence the lower PSCA levels associated with rs2294008-G likely lead to increased proliferation. Given the strong effect on PSCA expression, rs2294008-G is therefore the most likely functional 8q24.3 variant.

Chromosome 10q23.33 (PLCE1)

In OSCC and gastric cancer GWAS, Abnet et al discovered a risk association with rs2274223-G.²³ Interestingly, when the study stratified cases based upon anatomical location, rs2274223 was only significant in cardia cancers.²³ rs2274223-G changes a Histidine to Arginine in the phospholipase C Epsilon 1 (PLCE1) protein.²³ The exact mechanism by which rs2274223-G affects PLCE1 still needs to be elucidated. However, phospholipases are activated by tyrosine kinase or G-protein coupled receptors in response to hormones and growth factors. Activated phospholipases hydrolyse phosphoinositides to increase the intracellular level of Ca²⁺ and produce diacylglycerol, an important mediator for signal transduction. ⁸⁰

Chromosome 11q22.3 (ATM)

Using an Icelandic cohort, Helgason et al ²⁴ discovered three variants (rs758081262-T, rs768362387-A, rs777499935-G) in the multi-cancer HRDR ATM gene (see above) that were associated with gastric cancer risk. These variants were located in different haplotypes and each had similar effect sizes (OR=4·7).

Future Directions

The role of a genetic predisposition in gastric cancer has been largely under-appreciated. This concept likely stems from the fact that gastric cancer has two strong infection-related risk factors (H. pylori and EBV) and that a limited number of familial gastric cancer syndromes have been genetically characterized. However, the recent analyses in TCGA suggest that an important fraction of patients develop the disease due to inherited factors. In the study of Huang et al²⁵, the prevalence of pathogenic and likely pathogenic mutations in 443 sporadic gastric cancer patients was 13%. This is remarkable given the fact that the average age of onset of these patients was 66y; hence, the study was not enriched by likely genetic cases. Even among mutation carriers, the average age of onset was 61y, an age not typically associated with inherited risk. Should the study of Huang et al be independently replicated, particularly in high-risk regions like China, where most cases are diagnosed⁸¹, the implications for gastric cancer epidemiology and prevention will be enormous. At-risk relatives of thousands of patients, of the one million newly diagnosed cases each year, could potentially benefit from genetically informed preventive interventions through cascade genetic counselling and testing that guide early detection and cancer prevention. Therefore, an important priority in future gastric cancer research should be to fully examine the population prevalence of inherited mutations, particularly in high-risk regions such as Asia and Latin America. An examination of mutation prevalence, however, will require well-designed and statistically powered studies with appropriately matched cases and population controls. Even though Huang et al²⁵,found that PALB2 and ATM were enriched (at false discovery rate, FDR, <0.05) for pathogenic variants, FDR significance for BRCA1 and BRCA2 (the other two recurrently mutated genes in TCGA) was not achieved. BRCA1 and BRCA2 were however mutated at 0.6% and 1.9% frequencies, respectively, which is higher than the carrier frequency in EXAC controls (0.2% for BRCA1 and 0.3% for BRCA2). Neither TCGA (a heterogeneous collection of patients from Asia, Europe and North America) nor EXAC (a dataset with individuals from founder populations where BRCA½ carrier frequency is higher than in the general population) are the right populations to compare mutation frequencies and to test for possible associations between BRCA½ mutations and gastric cancer risk. Hence, an effect of BRCA½ on gastric cancer risk in the Huang et al²⁵ study may have been missed by the lack of statistical power (443 cases and 443 well matched controls only provide 12% power to detect an odds ratio=2). More studies are therefore needed to further assess if the prevalence of BRCA½ mutation in gastric cancer patients is significantly higher than that in the general population.

The past decade has seen the identification of several gastric cancer GWAS variants. These studies have been carried out in Asian and European populations, but regions of high incidence such as Latin America have not been examined. Latino populations have a unique genetic demography and gastric cancer epidemiological profiles, which may facilitate the discovery of additional gastric cancer variants.^10,82 Hence, additional populations should be included in future studies. This is particularly important as it is clear that many risk variants remain to be identified. Furthermore, only few of the known risk alleles are functional variants (i.e. rs4072037-A/MUC1, rs2294008-T/PSCA and rs2274223-G/PLCE1) and hence, future studies should functionally characterize other risk variants and genes for which little is known about their biology. Functional evaluation of these risk variants will increase our knowledge of gastric tumorigenesis, which will ultimately lead to improved prevention and treatment strategies.

With regards to recent advances in familial genes, another area of great importance in future studies is to determine the risk associated with HRDR gene mutations, particularly among carriers of mutations in PALB2, a gene that represents a bona fide gastric cancer gene. ⁶⁹ For now, appropriate gastric cancer surveillance among PALB2 mutation carriers remains uncertain. On the basis of the few published data (see Figure 2), patients with family history and other risk factors could be offered, in a research setting, endoscopic surveillance after the third decade of life (i.e. about a decade earlier than the youngest patient reported). Clearly, larger studies will likely need international collaboration that helps provide gastric cancer risk estimates for PALB2 mutation carriers.

The role of non-genetic and genetic (GWAS SNPs) risk factors in gastric cancer is considerable and it is possible that gene-gene and gene-environment interactions mediate risk. For PALB2, BRCA1 and BRCA2 mutation carriers, the average age of gastric cancer diagnosis seems to be around a decade later than reported for breast cancer among carriers of mutations in the same genes.^63,65 This suggests that gastric cancer penetrance among HRDR gene mutations could be mediated by additional factors. Although the role of H. Pylori in gastric cancer causality among PALB2 carriers has been previously ruled out, such studies remain small. ^28,42 Investigating interactions between inherited mutations and known risk factors should therefore represent another research priority.

Another consensus that is emerging from recent studies of HDGC and other familial syndromes is that these conditions are better defined by genetic rather than by clinical criteria, as rightly suggested by Hansford et al.³⁶ GAPPS should now be classified as FAPS. HDGC should be defined as carrying CDH1 or in CTNNA1 mutations, as the studies by Hansford et al³⁶ and Fewings et al⁴², which included cases fulfilling HDGC clinical criteria, found that several of them had mutations in genes associated with PJS (STK11), LS (MSH2), HBOCS (BRCA1, BRCA2) or with other hereditary conditions (PALB2, ATM, TP53, BRIP1, SDHB). ^36,42 Given the histological heterogeneity reported among the carriers of mutations in these genes (Figure 2), FIGC is likely an amalgamation of cases with mutations in these recently discovered HRDR genes. From what has been reported and in the absence of a clearly associated intestinal gastric cancer-only/predominant syndrome, FIGC now seems unlikely to represent a unique condition.

Given the existence of gastric cancer patients who carry mutations in these multi-organ genes, another important future research area should be to estimate gastric cancer risk among carriers of germline mutations in such genes. Through genetic testing, ascertained in families with breast cancer, thousands of BRCA1, BRCA2 and PALB2 mutation carriers have been identified.^63,65 While having gastric cancer family history may aid in refining gastric cancer risk among such mutation carriers, there is clearly a need to identify gastric cancer risk modifiers in such mutation-positive families.

Many of the genes discussed in this review are part of clinical gene panels. Although several gaps must be filled in before most of these candidates can be used in the clinic, a potential gastric cancer panel could include genes with variable levels of causality evidence. In Table 4 we present a list of candidate gastric genes and stratified their causality evidence based on the pre- and post-next generation sequencing (NGS) studies. For pre-NGS evidence, these genes were assessed based on their involvement in cancer syndromes and in gastric cancer-related syndromes in particular. The post-NGS evidence included their mutation status (mutated and/or recurrently mutated) and by belonging to candidate pathways (HRDR or CDH1 pathway). Based on these evidence lines, Tier 1 genes should include CDH1 and PALB2; Tier 2 genes should include CTNNA1, MSH2, TP53, STK11, ATM, BRCA1, BRCA2, BRIP1 and ATR; and Tier 3 genes should include APC, SMAD4, BMPR1A, ENG, MSR1, SDHA, SDHB, RAD51C, NBN, RECQL5 and FANCC. We emphasize that only CDH1 and PALB2 could be used in the clinic now. The remaining genes could be used, within research protocol settings, when gastric cancer patients seeking genetic counselling report personal or family history of gastric cancer associated syndromes (for example, BRCA½ testing could be useful in patients with HBOC history). These genes should also represent the priority list for future gastric cancer genetic studies of mutation prevalence and penetrance. Further, the s clinical genetics community should be encouraged to report the characteristics of gastric cancer cases carrying mutations in these genes.

Table 4.

Genes with evidence for gastric cancer causality

Gene	Evidence for causality	Familial Cancer Syndrome Evidence		Evidence from next generation sequencing studies
		Cancer syndrome	Gastric cancer related syndrome?	Mutated?	Recurrently mutated?	Gastric cancer related pathway?
Tier 1
CDH1	Strong	HDGC	Yes	NA	NA	CDH1
PALB2	Strong	Breast/ Pancreas cancer	No	Yes25,28,36,42	Yes	HRDR
Tier 2
CTNNA1	Moderate	HDGC	Yes	Yes 36	Yes	CDH1
MSH2	Moderate	LS	Yes	Yes 31,42	Yes	No
TP53	Moderate	LFS	Yes	Yes 31	No	No
STK11	Moderate	PJS	Yes	Yes 36	Yes	No
ATM	Moderate	Breast cancer	No	Yes 25,31,36	Yes	HRDR
BRCA1	Moderate	HBOC	Yes	Yes 25,28	Yes	HRDR
BRCA2	Moderate	HBOC	Yes	Yes 25,31,36	Yes	HRDR
BRIP1	Moderate	Breast cancer	No	Yes 25,31	Yes	HRDR
ATR	Moderate	None	No	Yes 31,42		HRDR
Tier 3
APC	Moderate/low	FAPS	Yes	No	No	No
SMAD4	Moderate/low	JPS	Yes	No	No	No
BMPR1A	Moderate/low	JPS	Yes	No	No	No
ENG	Moderate/low	JPS	Yes	No	No	No
MSR1	Moderate/low	Oeasophageal	No	Yes 42	Yes	No
SDHB	Moderate/low	Paraganglioma- Pheochromocytoma	No	Yes 36	No	SDHA/B
SDHA	Moderate/low	Paraganglioma- Pheochromocytoma	No	Yes 25	No	SDHA/B
RAD51C	Moderate/low	Breast cancer	No	Yes 28	No	HRDR
NBN	Moderate/low	Ovarian cancer	No	Yes 42	No	HRDR
RECQL5	Moderate/low	Breast cancer	No	Yes 42	No	HRDR
FANCC	Moderate/low	Breast cancer	No	Yes 31	No	HRDR

The fact that BRCAness has been found in gastric tumors^28,83 represents an opportunity to develop new PARPi therapies. Even though a recent trial⁸⁴ failed to show significant survival improvement with Olaparib (a PARPi) in ATM-negative gastric cancers, it does not necessarily mean a lack of PARPi potential for the treatment of this malignancy. Better biomarkers for the identification of HRDR tumours are however needed, as ATM immunohistochemistry, which was used in the recent trial, is a sub-optimal stratifier.⁸⁴ As more histological and molecular data in HRDR-deficient gastric cancers is reported, more precise ways to identify such cancers will be developed.

In conclusion, several important advances have been achieved in gastric cancer genetics. These findings will lead to the identification of high-risk groups that can be targeted for early interventions and that will improve our understanding of tumorigenesis. This will ultimately lead to improved outcomes for this common malignancy.