Epidemiological studies of esophageal cancer in the era of genome-wide association studies

Abstract

Esophageal cancer (EC) caused about 395000 deaths in 2010. China has the most cases of EC and EC is the fourth leading cause of cancer death in China. Esophageal squamous cell carcinoma (ESCC) is the predominant histologic type (90%-95%), while the incidence of esophageal adenocarcinoma (EAC) remains extremely low in China. Traditional epidemiological studies have revealed that environmental carcinogens are risk factors for EC. Molecular epidemiological studies revealed that susceptibility to EC is influenced by both environmental and genetic risk factors. Of all the risk factors for EC, some are associated with the risk of ESCC and others with the risk of EAC. However, the details and mechanisms of risk factors involved in the process for EC are unclear. The advanced methods and techniques used in human genome studies bring a great opportunity for researchers to explore and identify the details of those risk factors or susceptibility genes involved in the process of EC. Human genome epidemiology is a new branch of epidemiology, which leads the epidemiology study from the molecular epidemiology era to the era of genome wide association studies (GWAS). Here we review the epidemiological studies of EC (especially ESCC) in the era of GWAS, and provide an overview of the general risk factors and those genomic variants (genes, SNPs, miRNAs, proteins) involved in the process of ESCC.

Key Words: Esophageal cancer, Epidemiology, Genome wide association study, Single nucleotide polymorphism, MicroRNA

Core tip: Epidemiological study methods advance as the science and technique progress. In the era of genome wide association studies (GWAS), human genome epidemiology (HuGE) provide a great chance for epidemiologists and clinical scientists to explore the cause of disease and evaluate genomic biomarkers for diagnosis or prognosis. More and more epidemiological studies use GWAS methods to analyze genomic variants and the association with esophageal cancer. Here we review epidemiological studies of esophageal cancer in the era of GWAS, and briefly introduce the case-control study and cohort study methods in HuGE studies.

INTRODUCTION

Esophageal cancer (EC) caused about 395000 deaths in 2010[1]. The incidence rate and mortality rate varied among different geographic and ethnic populations. China has the most cases of esophageal cancer. EC is the fourth leading cause of cancer associated death in China. Esophageal squamous cell carcinoma (ESCC) is the predominant histologic type (90%-95%), while the incidence of esophageal adenocarcinoma (EAC) remains extremely low in China.

Traditional epidemiological studies have identified that environmental carcinogens play a critical role in the process of EC. Molecular epidemiological studies revealed that susceptibility to EC is associated with both genomic and non-genomic factors and the interaction between genomic and non-genomic factors. Of all the factors, some are associated with ESCC and others with EAC. Human genome epidemiology (HuGE) is denoted as “an emerging field of inquiry that uses systematic applications of epidemiologic methods and approaches in population-based studies of the impact of human genetic variation on health and disease”[2]. HuGE emerged after the sequencing of human genome was accomplished[3,4]. The characteristic of HuGE is the techniques applied in studies, especially the technique of DNA microarray chips used in genome-wide association studies (GWAS). These techniques can compare millions of SNPs between genome DNA from cases and controls. In this review we focus on the epidemiological studies of EC in the era of GWAS.

GENERAL RISK FACTORS FOR EC

The incidence of EC is associated with age. More than 85% of EC patients were diagnosed at an age more than 55 years old. The incidence of EC in males is higher than that in females. Esophageal reflux disease (GERD) is a risk factor of EAC. GERD is also a risk factor for Barrett’s esophagus (BE), and BE is associated with an increased risk for EC. Asian, especially Chinese, are more like to have an onset of EC than other populations.

Tobacco use (tobacco smoking, tobacco chewing, etc.) is a predominant risk factor for EC, especially ESCC. Alcohol drinking can also increase the risk of EC. Alcohol drinking is more likely to increase the risk of ESCC. People exposed to both tobacco use and alcohol had the risk of EC much more than those exposed to smoking or drinking alone. The risk of ESCC increased as the quantity of alcohol intake increased. The association between alcohol drinking and an increased risk of EC was more likely observed in Asian populations than in others[5]. Alcohol consumption and cigarette smoking are risk factors for ESCC in China and Japan[6,7].

Overweight or obesity is associated with a higher risk of EAC. A diet with more fruits or/and vegetables is reported to reduce the risk of EC. On the contrary some diet habit may raise the risk for EC. Drinking very hot liquids frequently may increase the risk of ESCC. Overeating is the risk factor for EAC.

Infection with human papillomavirus (HPV) is associated with a number of cancers. HPV infection has been observed in about one-third of EC patients in Asia and South Africa.

Risk factors for EC varied among different countries, which may explain in part by the social-economic difference. The risk factors for EC are different between high- and low-incidence areas[6]. A study in Kashmir[8] recruited 703 cases and 1664 controls and found an inverse association between tooth cleaning and ESCC risk. A study based on a network of Italian and Swiss case-control studies found that a family history of oral and pharyngeal cancer was associated with an increased risk for EC[9]. In China individuals with a family history of EC were found to have an increased risk for EC[10]. The Miyagi Cohort Study found that people who drink one or more cups of coffee per day compared with those who did not drink have a lower risk of EC and oral pharyngeal cancer[11]. The major risk factors for EC are summarized in Table 1.

Table 1 Major risk factors for esophageal cancer.

Risk factor	Ref.
Cigarette smoking (tobacco use)	Fan et al[12], Oze et al[7]
Alcohol drinking (alcohol consumption)1	Oze et al[13], Fan et al[12], Islami et al[5]
Drinking hot tea or soup at high temperature	Wu et al[14]
Food mutagens	Yokokawa et al[15], Zhang et al[16]
Family history	Turati et al[9], Gao et al[17]
Nutritional deficiency	Tran et al[18]
Poor oral hygiene/ESCC	Dar et al[8]
Coffee consumption2	Naganuma et al[11]
HPV infection	Li et al[19], Cui et al[20]
Obesity	Chen et al[21]

¹Alcohol consumption depends on the quantity of alcohol intake;

²Coffee consumption: reverse relation. ESCC: Esophageal squamous cell carcinoma.

GENERAL VIEW OF EPIDEMIOLOGY IN THE EAR OF GWAS

Epidemiology studies in the era of GWAS are characterized by large sample size and the use of the technique of microarray. HuGE has advanced to the stage of GWAS[22-26]. Table 2 shows the genomic variants identified to be associated with ESCC. Some of those genetic variants were confirmed in other populations and some others were not identified in other populations. GWAS in China showed that variants in several chromosome regions conferred an increased risk of EC, but only genetic variants in alcohol-metabolizing genes were risk factors for ESCC in Japanese[6,22-26]. A 2-step GWAS including 1070 cases and 2836 controls identified that single nucleotide polymorphisms (SNPs) rs671, rs1229984, alcohol consumption, and tobacco use were risk factors for ESCC[23].

Table 2 Genomic variants identified to be associated with esophageal squamous cell carcinoma.

Loci associated with ESCC	Method/design	Case sample size	Control sample size	Ref.
PLCE1 (10q23 rs227422) and C20orf54 (20p13)	GWAS	1077	1733	Wang et al[22]
ALDH2 (4q21-23, rs671) and ADH1B (12q24, rs1229984)	GWAS	1070	2836	Cui et al[23]
PLCE1 (10q23 rs2274223)	GWAS	2115	3302	Abnet et al[24]
ALDH2 (4q23, rs671) and ADH1B (12q24.11–13, rs1229984)	GWAS	1071	2762	Tanaka et al[25]
5q11 (rs10052657) 21q22 (rs2014300), 6p21 (rs10484761), 10q23 (rs2274223), and 12q24 (rs2074356, rs11066280) CYP1A1 A2455G polymorphism(Ile/Val, rs1048943)	GWAS	2031 1881	2044 3786	Wu et al[26] Shen et al[27]
	Meta-analysis
	(13 case-control studies)
CYP1A1/CYP2E1 (MTHFR) C677T and A1298C polymorphisms with ESCC	Case-control study	565 /482 3213	468/466 4354	Wang et al[28] Fang et al[29]
	Meta-analysis
	(15 case-control studies)
rs1014867 polymorphisms in FAT4 gene	Case-control study	2139	2273	Du et al[30]
Interleukin 1B rs16944	Case-control study	380	380	Zheng et al[31]
CHRNA5-A3-B4 rs667282 TT/TG	Case-control study	866	952	Wang et al[32]
rs1494961, rs1229984 and rs1789924, and rs671	Case-control study	2139	2273	Gao et al[33]
Genetic variants in DNA repair pathway genes/(EGFR) signaling pathway	Case-control study	1942	2111	Li et al[34], Li et al[35]
Sex hormone metabolic genes	Case-control study	1026	1452	Hyland et al[36]
Chromosome 1 open reading frame 10 (C1orf10)	Case-control study	991	984	Zhang et al[37]

ESCC: Esophageal squamous cell carcinoma; GWAS: Genome wide association studies; PLCE1: Phospholipase C epsilon 1; C20orf54: Chromosome 20 open reading frame 54; ADH1B: Alcohol dehydrogenase; ALDH2: Acetaldehyde dehydrogenase.

Genetic polymorphisms can affect the susceptibility to EC. Cytochrome P450 1A1 (CYP1A1) enzyme is a member of the CYP superfamily and prone to mutation, and an association between CYP1A1 enzyme activity and the risk of developing EC was revealed[38]. A meta-analysis uncovered that the A2455G polymorphism (Ile/Val, rs1048943) was a risk factor for EC[27]. By combining the technique of DNA microarray and epidemiology data of EC patients living in North or South China, the polymorphisms of CYP1A1 and CYP2E1 were studied[28]. In South area there was a significant association between CYP1A1 m2 polymorphism and EC. In North area there were significant associations between CYP2E1 Pst I and CYP2E Rsa I polymorphisms and EC. A significantly increased risk of ESCC was identified for smokers with the methylenetetrahydrofolate reductase (MTHFR) 677T allele[29]. MTHFR 677T and MTHFR 1298C conferred an increased risk for ESCC in Chinese population than in other populations. Four SNPs (rs1014867, rs12508222, rs1039808 and rs1567047) in FAT4 as potential risk factors for EC were studied[30]. The T allele of rs1014867 (Pro4972Ser) was associated with a reduced risk for EC[30]. The functional IL1B rs16944G > A polymorphism might be associated with the risk of ESCC and IL3 rs2073506 G > A polymorphism was a risk factor for ESCC with higher TNM stages[31]. CHRNA5-A3-B4 rs667282 TT/TG genotypes were risk factors of ESCC in Chinese[32]. In China, a case-control study including 2139 cases and 2,273 controls was carried out to evaluate the associations of six reported SNPs (rs1494961, rs1229984, rs1789924, rs971074, rs671 and rs4767364) with risk of ESCC. Results indicate that rs1494961, rs1229984, rs1789924, and rs671 may be used as biomarkers for ESCC[33]. Based on the SNPs identified in GWAS, 25 SNPs, 4 non-genomic factors (sex, age, tobacco use and alcohol drinking) and their associations with ESCC risk were studied[39]. Results indicate that genomic factors, none-genomic factors and their interactions can predict who are at high risk for ESCC. In contrast to association with a risk of ESCC in Asians, the PLCE1 rs2274223 and RFT2 13042395 SNPs were not associated with a risk of EC in Dutch Caucasians[40]. GWAS also identified three SNPs (rs10419226 in CRTC1, rs11789015 in BARX1 and rs2687201 near FOXP1) that were associated with a risk of EAC and BE[41].

GENOMIC VARIANTS IN PATHWAY GENES AND THEIR ASSOCIATIONS WITH EC

A GWAS aimed to explore the DNA repair pathway genes as risk factors for ESCC and GC was carried out[34]. One thousand six hundred and seventy-five SNPs were genotyped in cases (ESCC, GC) and controls from Shanxi and Linxian[34]. The DNA repair pathway genes were found to be risk factors for ESCC. CHEK2 was significantly associated with ESCC. Li et al[35] explored 3443 SNPs in genes involved in the EGFR signaling pathway in a study including 1942 ESCC cases, 1758 GC cases, and 2111 controls. Gene-level analyses found that GNAI3, CHRNE, PAK4, WASL, and ITCH were associated with a risk of ESCC[35]. A study analyzed 797 SNPs in 51 sex hormone metabolic genes in 1026 cases and 1452 controls[36]. Six genes including SULT2B1, CYP1B1, CYP3A7, CYP3A5, SHBG and CYP11A1 were identified as risk factors for ESCC[36]. Chromosome 1 open reading frame 10 (C1orf10), which is involved in heat shock and ethanol response, is either absent or down-regulated in ESCC tissues. Six strongly linked SNPs in a region of 7 kb were observed in a case-control study[37]. Compared with -1139GG, -1139CC genotype was a risk factor for ESCC[37].

The HuGE progressed form the discovery of novel genes or SNPs to the functional or mechanistic study of those genes or SNPs. Moreover, HuGE studies try to screen some of those genes, SNPs or miRNAs that are clinical treatment targets or biomarkers for diagnosis or prognosis. A low mtDNA copy number variant (CNV) was a risk factor for EAC[42]. A case-control study was carried out to analyze the relationship between SNPs (rs17417407, rs2274223 and rs22744224) in PLCE1 and susceptibility to ESCC[43]. Rs2274223G was identified to be a risk factor for ESCC, and rs2274224G was observed as a favorable factor for ESCC[43]. Phenotypes for rs17417407T, rs2274223G and rs2274224G were observed as risk factors for ESCC. Genomic polymorphisms in PLCE1 can affect the risk of ESCC in Chinese population exposed to tobacco smoking[43].

Zhang et al[37] found that there was an interaction between the -1139G/C genotype in C1orf10 and smoking, which increases the risk of ESCC. An HPV gene chip was used to detect HPV genotypes in 183 EC cases and 89 controls[20]. The frequency of seven HPV genotypes (16, 18, 35, 52, 6, 11, 43) in EC tissues was higher (31.7%) than that in controls (9.0%, P < 0.001), indicating that HPV infection was a risk factor for EC in Kazakh population. Moreover, heterozygote rs2274223 in PLCE1 was associated with an increased risk of HPV infection[20].

MICRORNAS AND THEIR ASSOCIATIONS WITH EC

MicroRNAs (miRNAs) are non-coding RNAs that modulate the translation of RNAs. MiRNAs have been involved in cancer initiation and development. Different miRNAs show differential expression levels in EC tissue or EC cell lines. The levels of miR-145 and miR-143 were decreased in ESCC tissues. An inverse association between miR-143 expression levels and cancer invasion or metastasis was identified[44]. Results showed that miR-143 may act as a suppressor in the process of ESCC. MiRNA microarray technique can be used to explore the profiles of miRNAs in ESCC cell lines. MiR-10a and MiR-205 were observed as potential specific biomarkers for ESCC (Table 3)[45].

Table 3 MicroRNA expression and their associations with esophageal squamous cell carcinoma[45].

MiRNA	Compared to normal esophageal tissue	Proved targets
miR-10a	Decreased	HOXA3, HOXB1, HOXB3
		HOXD4, HOXD10
miR-21	Increased	PCDCD4, NFIB, PTEN, TPM1
miR-93	Increased	FUSA, E2F1, TP53, INP1
miR-129	Increased	LATS2
miR-203	Increased/ decreased	ABL1, TP53INP1, SOCS3
miR-205	Decreased/increased	ZEB1, ZEB2, E2F5, HER3, ERBB3, PRKCE, LRP1
miR-375	Decreased	PDK1

Kan and Meltzer[46] reviewed miRNAs in BE and EAC. They surmised the following: (1) miRNA profiles were different between BE and EAC; (2) miR-196a is overexpressed in EAC tissues and is favorable to EAC cell survival; miR-196a might be a biomarker during the carcinogenesis from BE to EAC; and (3) the miR-106b-25 polycistron is involved in EC progression via suppression of p21 and Bim. The potential role of miRNAs in GC and EC and the mechanisms of action have been reviewed previously[47].

MiRNAs participate in the process of carcinogenesis by affecting the expression of genes to regulate cell apoptosis, proliferation and invasion. Some miRNAs have been proved to be associated with the characteristics of cancer or the survival time of patients, and those miRNAs might be valuable as biomarkers for diagnosis or prognosis prediction. A greater understanding of functions of miRNAs in EC could provide more details about the mechanisms of carcinogenesis (Table 4)[44,47,48].

Table 4 Common miRNA expression profiles in esophageal cancer[47].

ESCC		EAC
Up-regulated	Down-regulated	Up-regulated	Down-regulated
miR-21	Let-7c	miR-21	Let-7c
miR-155	miR-1	miR-28	miR-203
miR-93	MiR-99a	miR-3a-5p	miR-205
miR-129	miR-100	miR-143-145 cluster	miR-23a
	miR-133a	miR-192	miR-27a
	miR-143-145 cluster	miR-194	miR-27b
	miR-203	miR-215	miR-31
	miR-375		miR-99a
			miR-100

ESCC: Esophageal squamous cell carcinoma; EAC: Esophageal adenocarcinoma.

A study explored the expression of miRNAs in ESCC and found that 15 miRNAs were down-regulated[48]. Four miRNAs (miR-145, miR-30a-3p, miR-133a and miR-133b) were decreased in ESCC and might act as tumor suppressors. Three miRNAs (miR-133b, miR-133a and miR-145) can directly inhibit FSCN1 expression, which might decrease the risk for ESCC[48]. A hospital based case-control study including 380 cases and 380 controls was carried out to observe the association of SNPs in miRNAs with genetic susceptibility to ESCC[49]. Female individuals or people who never smoke or drink have a lower risk for ESCC if they carry MiR-196a2 rs11614913 T > C[49]. Zhang et al[50] reported that up-regulation of miR-203 in EC cells can significantly increase apoptosis and decrease miR-21 expression. MiR-203 overexpression can also inhibit cell invasion, migration and proliferation, and may act as a tumor suppressor in EC.

CLINICAL RESEARCH OF GENOMIC BIOMARKERS FOR EC

EC is a disease with a poor prognosis[51]. It is urgent to identify valuable biomarkers involved in the diagnosis, progress or therapy targets for ESCC. Qi[52] reviewed the proteins, identified by proteomics, which were associated with the process of ESCC. The mechanisms of action of the proteins identified by proteomics and involved in the progress of ESCC were also discussed[53].

Loss of chromosome 19p is one of the most frequent allelic imbalances in ESCC. Down-regulation of DIRAS1 was associated with a poor survival rate. About 50% of ESCC cases had down-regulation of DIRAS1, and this down-regulation was associated with unfavorable clinical characteristics such as lymph node metastasis and low survival rate[53]. A GWAS observed the relationship between SNPs and the survival of ESCC patients[54]. Results showed that SLC39A6 overexpression was associated with a shorter period of survival, which indicated that SLC39A6 might be a target for ESCC therapy[54]. HOTAIR, a well-known long non-coding RNA, has been reported to associate with ESCC. It was found that HOTAIR was overexpressed in ESCC compared to normal esophageal tissues[55,56]. Overexpression of HOTAIR was associated with poorer prognosis. The HOTAIR ⁄WIF-1 axis was identified to play an important role in cell metastasis and might be a target for ESCC therapy. PIK3CA mutations in ESCC are associated with longer survival, suggesting its role as a prognostic biomarker[57]. Proteomic methods were used to evaluate proteins as potential biomarkers for ESCC[58], and 33 proteins overexpressed and 14 proteins down-regulated in ESCC were identified[58]. The expression of fos related antigen 1 (Fra-1) was identified as an unfavorable factor for prognosis[59]. The effect of SNPs of long intergenic non-coding RNAs on ESCC was studied by Wu et al[60]. 52 SNPs were studied in 1493 ESCC cases and 1553 controls in China[60]. Compared with the AA genotype of rs11752942, AG and GG reduced the risk of ESCC. Rs11752942G allele could significantly down-regulate the expression level of lincRNA-uc003opf.1[60]. These results indicated that rs11752942 in lincRNAuc003opf. 1 exon was a biomarker for susceptibility to ESCC. Sakai et al[61] reviewed the most recent studies on miRNAs in EC and/or BE. Four miRNAs were identified as diagnostic biomarkers and five miRNAs were supposed to be valuable biomarkers for diagnosis and prognosis. The progress in miRNAs identified in EC is exciting, but there is still a lot of work to be done before those miRNAs can be used as biomarkers for diagnosis, efficacy evaluation or prognosis prediction.

EPIDEMIOLOGICAL STUDY DESIGN IN THE ERA OF GWAS

The advantages and disadvantages of case-control and cohort studies in the era of GWAS have been previously discussed in detail[62]. The great majority of GWAS conducted to date have used the case-control design, in which genome or SNPs were compared between tissues from esophageal cancer patients or esophageal cancer free controls[22,26]. Other risk factors for EC were also investigated and analyzed to search for the genetic and environmental factors influencing EC. Case-control design not only allows to study multiple factors that might associate with disease, but also permits a more detailed evaluation of risk factor exposure, such as tobacco use, alcohol drinking, occupational, HPV infection, family history of EC or dietary history. However, there are several biases that are related with the selection of cases and controls. If cases can be representative of all persons who develop EC, the bias from case selection in a case-control study is limited. However, cases in most of the case-control studies are often hospital based, typically through review of medical records, and those with early death have great chance not to be included, leading to survival bias. Theoretically, controls should be representative of all persons at risk for EC. In fact, selecting controls in a case-control study is the most difficult aspect. The evaluation of risk factor exposure should avoid bias, which is related to measuring exposures. Case-control studies are often easier and cheaper to conduct than cohort studies.

The major merit of the cohort study is that recall bias is controlled by collecting exposure prior to disease outcome. Cases identified in cohort are incident and free of survival bias. Results of cohort studies can be used to explain the cause of disease. The disadvantages of cohort studies include the requirement of large sample size if the incidence of disease is low, expensive cost for genomic test, and long term follow-up[63]. Due to reasons of cost and efficiency fewer GWAS use cohort study design. More and more case-control studies were carried out with large sample sizes, to explore the genomic and environmental risk factors for EC[23,25].

GWAS use high-throughput microarray technologies to analyze genetic SNPs, miRNAs or proteins and evaluate their association with disease or with clinical utilities (biomarkers for diagnosis or prognosis). Since 2005, more than 100 loci for more than 40 diseases have been discovered and confirmed. Many SNPs were first observed to be associated with disease risk. GWAS have some advantages in identifying genetic variants associated with disease. GWAS also have some limitations, including type I and type II errors and biases due to poor representative of participants. Two step or multi-step GWAS are recommended in epidemiological case-control studies.

CONCLUSION

The flood of GWAS findings from case-control studies has led to the increasing need for subsequent confirmation and functional studies in experimental systems to identify the biological mechanisms of the association between genomic variants and EC. Epidemiological studies of EC in the era of GWAS have explored the genomic variants affecting signaling, epigenetic regulation, RNAs, proteins and pathways involved in cell proliferation or invasion. However, much work remains to be done including identifying the biomarkers for screening, efficacy evaluation and prognosis prediction. In the future, more and more epidemiological studies will take the advantages of population-based, very large sample-sized GWAS.