Contribution of ZBTB20 Polymorphisms to Esophageal Cancer Risk Among the Chinese Han Population.

Background: ZBTB20 was overexpressed in esophageal cancer (EC). The study aimed to identify genotypes of ZBTB20 polymorphisms and their correlation with EC occurrence in a Chinese Han population.
Methods: Four single nucleotide polymorphisms (SNPs) in ZBTB20 were randomly selected for genotyping through Agena MassARRAY system among 525 EC patients and 522 healthy controls. Multiple genetic models were applied to assess the association of ZBTB20 polymorphisms with EC susceptibility by calculating odds ratios (ORs) with 95% confidence intervals (CIs).
Results: Rs10934270 was associated with lower EC susceptibility (OR = 0.64, p = 0.004) with statistical power >90% in overall analysis. Specifically, the correlation of rs10934270 with EC susceptibility was found in subgroups including patients with esophageal squamous cell carcinoma (ESCC), males, subjects aged ≤65 years, subjects with BMI ≤ 24 kg/m2, and smokers. Rs9841504 might be a risk-increasing factor for ESCC. Moreover, rs9288999 in subjects aged ≤65 years and rs73230612 in females were related to lower EC risk.
Conclusion: Our research is the first to report that ZBTB20 rs10934270 is associated with reduced EC susceptibility in the Chinese Han population. These data provide a scientific basis for understanding the influence of the ZBTB20 gene on EC occurrence.

Introduction

Esophageal cancer (EC) is a highly aggressive cancer of the digestive system with an increasing incidence and a 5-year survival rate of about 15–25%.1 EC is the eighth most common cause of cancer worldwide (604,100 new cases) overall, and the mortality of EC ranks sixth in malignancy-related mortality (544,076 new deaths).2 Esophageal squamous cell carcinoma (ESCC) accounts for 88.84% of all EC cases in China and is the main pathological type worldwide.3 There are obvious gender differences in EC, EC is 2 to 8 times more common in men than in women in most areas of the world because of the use of tobacco and alcohol, especially in the developed countries. However, the incidence of EC in male and female can be very close in some regions where smoking and drinking play only a minor role in EC development (eg, Huai’an in Jiangsu,4,5 and Taihang mountain region)6,7. EC is considered to be a complex disease caused by the interaction of multiple factors such as genetics and environmental factors. It is now generally accepted that unfavorable habits (tobacco and alcohol), poor nutritional status, caloric intake and obesity are the main risk factors for EC.8 Previous studies on the attributable risk of EC in China showed 46% of EC (51% in men and 33% in women) were attributable to tobacco smoking, alcohol drinking, and low vegetable and fruit intake. Tobacco smoking and alcohol use are major risk factors for squamous cell carcinoma.9 A negative correlation between overall obesity, as measured by body mass index (BMI) and risk of ESCC has been reported.10 Nevertheless, not everyone exposed to risk factors will eventually develop EC. Recently, the role of genetic polymorphisms in the occurrence of EC has been reported.11–13

Zinc finger and BTB domain containing 20 (ZBTB20, also named ZNF288), a dendritic cell-derived BTB/POZ zinc finger (DPZF), belongs to a family of transcription factors with BTB/POZ domain (N-terminal) and zinc finger domain (C-terminal).14 ZBTB20 is considered as a key transcriptional repressor, and its deficiency may lead to high expression levels of alpha-fetoprotein.15 ZBTB20 plays a role in many processes, including glucose homeostasis, and tumor progression.16 Studies have reported that the high expression of ZBTB20 is closely related to the prognosis of hepatocellular carcinoma.17 In addition, studies have also clarified that ZBTB20 promotes the migration and invasion of gastric cancer cells.18 ZBTB20 was overexpressed in glioblastoma, and ZBTB20 knockdown inhibited glioblastoma progression.19 These findings indicated that ZBTB20 acted as a tumor progression gene in tumor progression. Based on the GEPIA database (), ZBTB20 was overexpressed in EC. A study has evaluated the relation between ZBTB20 rs9841504 and ESCC susceptibility, but no association has been found.20 Furthermore, the contribution of other polymorphisms in ZBTB20 to EC has not been investigated.

In this study, SNPs (rs10934270, rs9288999, rs9841504, and rs73230612) in the intronic region of ZBTB20 were randomly selected for genotyping based on the following criteria: 1) minor allele frequency (MAF) >0.05 and r2 ≥ 0.8 in the Chinese Han population from the Chinese Han population in Beijing of the 1000 Genomes Project and the dbSNP database; 2) Hardy-Weinberg equilibrium (HWE) >0.05, and the call rate for genotyping >99.5%; 3) previous literatures.21–25 The genotype of ZBTB20 polymorphisms and its correlation with EC occurrence was evaluated in the Chinese Han population. Moreover, the genotypic-phenotypic analysis of cancer type and lymph node metastasis were investigated. Considering that age, sex, BMI, smoking, and alcohol consumption are confounding factors for EC, stratified analysis was also performed to evaluate the contribution of ZBTB20 SNPs to EC risk, which will provide important evidence for elucidating the pattern of association.

Patients and Methods

Characteristics of Subjects

The study consisted of 525 EC patients and 522 healthy controls from Hainan Cancer Hospital. All recruited subjects were Chinese Han nationality. Patients were newly diagnosed and histopathologically confirmed as primary EC according to the criteria of Manual of Clinical Oncology, Oesophagus established by the Union for International Cancer Control (UICC). Tissue sections were reviewed by two experienced pathologists to ensure that tumor cell purity was greater than 50% and to confirm histological type. Patients with prior cancer history, upper gastrointestinal diseases, and serious chronic diseases were excluded. Blood samples were collected from patients prior to any treatment. The age- and gender-matched controls were composed of randomly recruited healthy participants with no history of cancer or upper gastrointestinal diseases. Basic characteristics (age, gender, smoking and drinking, and BMI) and pathological data (subtypes, lymph node metastasis, and stages) were recorded via questionnaires and medical records, respectively. This study was conducted under the approval of the Ethics Committee of Hainan Cancer Hospital (No.: ZDKJ202005) according to the Helsinki Declaration. Written informed consent to participate in the study was obtained.

Genotyping of ZBTB20 Polymorphisms

Peripheral blood (5 mL) from each participant was collected in EDTA tubes and was stored at 4°C. Genomic DNA was purified within 1 week of blood collection using GoldMag DNA Blood Mini Kit (GoldMag Co. Ltd. Xi’an, China). DNA concentration and purity was detected through NanoDrop 2000 (Thermo, Waltham, MA, USA). Four SNPs (rs10934270, rs9288999, rs9841504, and rs73230612) in the intronic region of ZBTB20 were randomly selected for genotyping based on the following criteria: 1) minor allele frequency (MAF) >0.05 and r2 of ≥0.8 in the Chinese Han population from the Chinese Han population in Beijing of the 1000 Genomes Project and the dbSNP database, 2) HWE > 0.05, and the call rate for genotyping >99.5%, 3) previous literatures.21–25 HaploReg v4.1 () was used to determine the frequencies of these SNPs in other populations. HaploReg v4.1, RegulomeDB () and GTEx Portal () are databases used to predict the potential functions of selected SNPs. Genotyping was determined through Agena MassARRAY system (Agena, San Diego, CA, USA) with built-in software. The MassARRAY platform is based on matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF MS).26–28 The general principle of the MassARRAY platform is based on the difference in primer masses caused by sequence changes. The primer sequences were shown in . All samples were genotyped using a double-blind model. For quality control, approximately 10% of randomly chosen samples were run in duplicate with 100% consistency.

Statistical Analysis

Student’s t-test or χ2 test was used to identify differences in baseline data between EC patients and healthy controls, as appropriate. A goodness-of-fit χ2 test was used for HWE analysis in controls. Multiple genetic models were applied to assess the association of ZBTB20 polymorphisms with EC susceptibility. Logistic regression analysis adjusted for age and gender was employed to calculate odds ratios (ORs) with 95% confidence intervals (CIs) by PLINK software. Furthermore, subgroup analyses stratified by histological, demographic, and behavioral data were also estimated. False-positive report probability (FPRP) analysis was applied to assess noteworthy associations by setting a threshold of 0.2 and a prior probability of 0.1. The SNP–SNP interactions were analyzed through multifactor dimensionality reduction (MDR) (version 3.0.2) software. SPSS 18.0 software (SPSS Inc., Chicago, IL, USA) was used for data analysis, and a p value < 0.05 indicated statistically different, whereas an adjusted p value < 0.05/4 was considered significant after Bonferroni correction.

Results

Participants’ Characteristics

A total of 1047 subjects were recruited, including 525 EC patients (63.92 ± 9.18 years) and 522 healthy controls (63.70 ± 7.07 years). The demographic and clinical characteristics of all subjects were shown in Table 1. The ratio of male to female in the two groups was 3:1. The distributions of age (p = 0.657) and sex (p = 0.956) were not statistically different. However, there were significant differences in BMI (p < 0.001), smoking (p = 0.040), and drinking (p < 0.001) between the two groups. Among 525 EC patients, 73.7% (N = 387) were confirmed to have ESCC. There were 177 cases of lymphatic metastasis and 189 cases in stage III/IV.

Table 1

The Information of the Patients with Esophageal Cancer and Healthy Controls

Characteristics	Case	Control	p
Number	525	522
Age (mean ± SD, years)	63.92 ± 9.18	63.70 ± 7.07	0.657
> 65	236 (45.0%)	191 (36.6%)
≤ 65	289 (55.0%)	331 (63.4%)
Gender			0.956
Male	390 (74.3%)	387 (74.1%)
Female	135 (25.7%)	135 (25.9%)	<0.001
BMI (kg/m2)
≤ 24	435 (82.9%)	162 (31.0%)
> 24	74 (14.1%)	193 (37.0%)
Missing	16 (3.0%)	167 (32.0%)
Smoking			0.040
Yes	245 (46.7%)	117 (22.4%)
No	274 (52.2%)	179 (34.3%)
Missing	6 (1.1%)	226 (43.3%)
Drinking			<0.001
Yes	125 (23.8%)	116 (22.2%)
No	354 (67.4%)	155 (29.7%)
Missing	46 (8.8%)	251 (48.1%)
Pathological type
Squamous cell carcinoma	387 (73.7%)
Other	138 (26.3%)
Lymph node metastasis
Yes	177 (33.7%)
No	162 (30.9%)
Missing	186 (35.4%)
Clinical stages
I+II	159 (30.3%)
III+IV	189 (36.0%)
Missing	177 (33.7%)

Note: p values were calculated by χ2 test or the Student’s t-test.

The Association Between ZBTB20 SNPs and EC Risk

Table 2 summarized the information on ZBTB20 four polymorphisms. The genotype distribution of these polymorphisms in the control group was in accordance with HWE (p > 0.05), indicating that the selected population had a good representativeness. The detection rate of genotyping was >99.5%. The MAFs of these SNPs in healthy controls were similar to those in Asians. The potential functions of these polymorphisms explored by HaploReg v4.1 and RegulomeDB were displayed in Table 2. The results from HaploRegv4.1 indicated that these selected SNPs were associated with the regulation of enhancer histones, changed motifs, DNase, and selected eQTL hits. Besides, rs10934270 might be related to transcription factor (TF) binding, any motif and DNase peak. Based on the GTEx Portal database, the genotypes of rs10934270 (p = 1.4e-12, ) and rs73230612 (p = 3.4e-6, ) were related to ZBTB20mRNA expression in whole blood.

Table 2

The Information of Four Polymorphisms on the ZBTB20 Gene

SNP ID	Chr: Position	Alleles (Ref/Alt)	MAF		Call rate	HWE			Frequencya				HaploReg v4.1
			Cases	Controls		O(HET)	E(HET)	p	AFR	AMR	ASN	EUR
rs10934270	3:114384900	C/T	0.069	0.103	100.0%	0.176	0.186	0.239	0.05	0.39	0.11	0.43	DNAse, Motifs Changed, Selected eQTL hits
rs9288999	3:114429080	A/G	0.416	0.430	99.9%	0.488	0.490	0.929	0.67	0.41	0.43	0.24	Enhancer histone marks, Motifs changed
rs9841504	3:114643917	C/G	0.147	0.139	100.0%	0.251	0.239	0.359	0.34	0.17	0.16	0.08	Enhancer histone marks, Motifs changed
rs73230612	3:115131989	T/C	0.404	0.416	100.0%	0.475	0.486	0.652	0.81	0.83	0.40	0.89	Motifs changed
SNP ID	RegulomeDB		Pair of SNPs with r2 ≥ 0.8a
rs10934270	TF binding + any motif + DNase peak		rs6801183, rs9846724, rs62265723, rs13084997, rs1830095, rs1474426, rs1474425, rs2722004, rs7626635, rs10934269, rs2733405, rs12491672, rs2683792, rs2683791, rs2722007, rs2722006, rs2722005, rs6775754, rs9881173, rs9881461
rs9288999	Other		rs13090443, rs13091312, rs2722019, rs4580516, rs146634908, rs9840030, rs10934272, rs9882269, rs13067741, rs9816740, rs10470388, rs10470389, rs7630111
rs9841504	Motif hit		rs11294002, rs1290894, rs9820958, rs1274265, rs73857635, rs9878038, rs9841454, rs56298435, rs16823073, rs74848785, rs76819674, rs78502125, rs77352581, rs78889361, rs76373676, rs116319741, rs150495605
rs73230612	Other		rs77093417, rs10222496, rs16823443, rs7643617, rs6805723, rs1473580, rs9815319, rs9830947, rs2177039, rs16823508, rs12639377, rs13323268, rs12632241, rs73230612, rs66839906, rs199713828, rs9879646, rs114072304, rs7620646, rs73230620, rs141794376, rs56260350, rs9832181, rs980944, rs7653569, rs60015778, rs201752575, rs75352178, rs16823578, rs9822860, rs59774725, rs9790250, rs11712587, rs78572299, rs9867281, rs11718803, rs13318807, rs6768463, rs13326167, rs6806156, rs11706205, rs9811889

Notes: aData from Haploreg (). RegulomeDB ().

Abbreviations: SNP, Single nucleotide polymorphism; MAF, Minor allele frequency; HWE, Hardy-Weinberg equilibrium; O(HET), Observed heterozygosity frequency; E(HET), Expected heterozygosity frequency; AFR, African; AMR, American; ASN, Asian; EUR, European.

The allele and genotype frequencies of these polymorphisms are listed in Table 3. The T allele frequency of rs10934270 in EC patients was lower than that in controls. Based on multiple genetic models, rs10934270 was related to lower EC susceptibility under the allele (T vs C, OR = 0.64, 95% CI: 0.47–0.87, p = 0.004), genotype (CT vs CC, OR = 0.68, 95% CI: 0.49–0.96, p = 0.029), dominant (CT-TT vs CC, OR = 0.65, 95% CI: 0.46–0.91, p = 0.011), and log-additive (OR = 0.64, 95% CI: 0.47–0.88, p = 0.005) models. No effect of other SNPs on EC risk was found overall. After Bonferroni correction, the significant association of rs10934270 with EC risk in the allele, dominant and log-additive models still existed.

Table 3

Risk Analysis for ZBTB20 Polymorphisms and Esophageal Cancer in Different Genetic Models by Logistic Regression Analysis

SNP	Model	Genotype	Controls	Esophageal Cancer			Esophageal Squamous Cell Carcinoma
				Cases	OR (95% CI)	p-value	Cases	OR (95% CI)	p-value
rs10934270	Allele	C	936	978	1		719	1
		T	108	72	0.64 (0.47–0.87)	0.004*	55	0.66 (0.47–0.93)	0.017
	Genotype	CC	422	455	1		334	1
		CT	92	68	0.68 (0.49–0.96)	0.029	51	0.70 (0.48–1.02)	0.061
		TT	8	2	0.23 (0.05–1.10)	0.065	2	0.32 (0.07–1.50)	0.147
	Dominant	CC	422	455	1		334	1
		CT-TT	100	70	0.65 (0.46–0.91)	0.011*	53	0.67 (0.47–0.96)	0.031
	Recessive	CC-CT	514	523	1		385	1
		TT	8	2	0.25 (0.05–1.16)	0.077	2	0.33 (0.07–1.58)	0.167
	Log-additive				0.64 (0.47–0.88)	0.005*		0.67 (0.48–0.94)	0.020
rs9288999	Allele	A	594	613	1		462	1
		G	448	437	0.95 (0.79–1.12)	0.524	312	0.90 (0.74–1.08)	0.252
	Genotype	AA	170	180	1		137	1
		AG	254	253	0.94 (0.72–1.24)	0.665	188	0.92 (0.69–1.23)	0.572
		GG	97	92	0.90 (0.63–1.28)	0.551	62	0.79 (0.54–1.17)	0.249
	Dominant	AA	170	180	1		137	1
		AG-GG	351	345	0.93 (0.72–1.20)	0.578	250	0.88 (0.67–1.17)	0.387
	Recessive	AA-AG	424	433	1		325	1
		GG	97	92	0.93 (0.68–1.28)	0.654	62	0.84 (0.59–1.19)	0.314
	Log-additive				0.95 (0.80–1.13)	0.535		0.90 (0.74–1.08)	0.256
rs9841504	Allele	C	899	896	1		649	1
		G	145	154	1.07 (0.83–1.36)	0.611	125	1.19 (0.92–1.55)	0.180
	Genotype	CC	384	386	1		275	1
		CG	131	124	0.94 (0.71–1.25)	0.664	99	1.06 (0.78–1.43)	0.731
		GG	7	15	2.12 (0.85–5.26)	0.106	13	2.59 (1.02–6.59)	0.045
	Dominant	CC	384	386	1		275	1
		CG-GG	138	139	1.00 (0.76–1.32)	0.992	112	1.13 (0.84–1.52)	0.408
	Recessive	CC-CG	515	510	1		374	1
		GG	7	15	2.15 (0.87–5.33)	0.098	13	2.56 (1.01–6.47)	0.047
	Log-additive				1.06 (0.83–1.36)	0.630		1.19 (0.92–1.55)	0.183
rs73230612	Allele	T	610	626	1		450	1
		C	434	424	0.95 (0.80–1.13)	0.580	324	1.01 (0.84–1.22)	0.902
	Genotype	TT	181	184	1		127	1
		TC	248	258	1.03 (0.78–1.34)	0.853	196	1.13 (0.84–1.52)	0.416
		CC	93	83	0.88 (0.61–1.26)	0.475	64	0.98 (0.66–1.45)	0.920
	Dominant	TT	181	184	1		127	1
		TC-CC	341	341	0.99 (0.76–1.27)	0.907	260	1.09 (0.82–1.44)	0.549
	Recessive	TT-TC	429	422	1		323	1
		CC	93	83	0.86 (0.62–1.20)	0.378	64	0.91 (0.64–1.29)	0.606
	Log-additive				0.95 (0.80–1.13)	0.580		1.01 (0.84–1.22)	0.899

Notes: p values were calculated by logistic regression analysis with adjustments for age and gender. Bold p < 0.05 respects the data is statistically significant. *p indicate that after Bonferroni correction (p < 0.05/4) means the data is statistically significant.

Abbreviations: SNP, single nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval.

Stratified Analysis by Histological Data

The stratified analysis by subtype, lymph node metastasis, and stage was performed to assess the association of candidateSNPs with histological data of EC patients. We found that rs10934270 was a protective factor against ESCC (allele: T vs C, OR = 0.66, p = 0.017; dominant: CT-TT vs CC, OR = 0.67, p = 0.031; and log-additive: OR = 0.67, p = 0.020), while rs9841504 was related to an elevated risk of ESCC (genotype: GG vs CC, OR = 2.59, p = 0.045; and recessive: GG vs CC-CG, OR = 2.56, p = 0.047, Table 3). However, no significant associations were detected between ZBTB20 SNPs and lymph nodes metastasis and staging in EC patients (data no shown).

Stratified Analysis by Demographic and Behavioral Data

Tables 4 and 5 summarized the results of subgroup analyses to explore the interaction of ZBTB20 variants and demographic data with EC risk. According to gender-stratified analysis (Table 4), rs10934270 was related to reduced EC risk in males under the allele (T vs C, OR = 0.64, p = 0.017), genotype (CT vs CC, OR = 0.66, p = 0.042), dominant (CT-TT vs CC, OR = 0.64, p = 0.025), and log-additive (OR = 0.65, p = 0.020) models. Among females, rs73230612 was a protective factor against EC in the allele (C vs T, OR = 0.70, p = 0.043), genotype (CC vs TT, OR = 0.42, p = 0.035), and log-additive (OR = 0.67, p = 0.034) models.

Table 4

Stratified Analysis by Gender and Age for the Associations Between ZBTB20 Polymorphisms and the Risk of Esophageal Cancer

SNP ID	Model	Genotype	Control	Case	OR (95% CI)	p–value	Control	Case	OR (95% CI)	p–value
Gender			Males				Females
rs10934270	Allele	C	699	730	1		237	248	1
		T	75	50	0.64 (0.44–0.93)	0.017	33	22	0.64 (0.36–1.12)	0.118
	Genotype	CC	317	342	1		105	113	1
		CT	65	46	0.66 (0.44–0.99)	0.042	27	22	0.76 (0.41–1.41)	0.381
		TT	5	2	0.37 (0.07–1.92)	0.237	3	0	/	/
	Dominant	CC/ CT-TT	317/70	342/48	0.64 (0.43–0.95)	0.025	105/30	113/22	0.68 (0.37–1.26)	0.219
	Recessive	CC-CT/ TT	382/5	388/2	0.39 (0.08–2.04)	0.267	132/3	135/0	/	/
	Log-additive				0.65 (0.45–0.93)	0.020			0.64 (0.36–1.13)	0.122
rs73230612	Allele	T	459	452	1		151	174	1
		C	315	328	1.06 (0.86–1.29)	0.588	119	96	0.70 (0.50–0.99)	0.043
	Genotype	TT	142	132	1		39	52	1
		TC	175	188	1.16 (0.84–1.58)	0.364	73	70	0.72 (0.42–1.22)	0.220
		CC	70	70	1.07 (0.71–1.61)	0.739	23	13	0.42 (0.19–0.94)	0.035
	Dominant	TT/ TC-CC	142/245	132/258	1.13 (0.84–1.52)	0.409	39/96	52/83	0.65 (0.39–1.08)	0.097
	Recessive	TT-TC/ CC	317/70	320/70	0.99 (0.68–1.42)	0.943	112/23	122/13	0.52 (0.25–1.08)	0.078
	Log-additive				1.05 (0.86–1.29)	0.605			0.67 (0.46–0.97)	0.034
rs9288999	Allele	A	437	467	1		157	146	1
		G	337	313	0.87 (0.71–1.06)	0.173	111	124	1.20 (0.85–1.69)	0.292
	Genotype	AA	124	143	1		46	37	1
		AG	189	181	0.83 (0.61–1.14)	0.251	65	72	1.38 (0.80–2.38)	0.251
		GG	74	66	0.78 (0.51–1.17)	0.225	23	26	1.41 (0.69–2.86)	0.347
	Dominant	AA/ AG-GG	124/263	143/247	0.82 (0.61–1.10)	0.179	46/88	37/98	1.39 (0.82–2.33)	0.219
	Recessive	AA-AG/ GG	313/74	324/66	0.86 (0.60–1.25)	0.434	111/23	109/26	1.15 (0.62–2.14)	0.657
	Log-additive				0.87 (0.71–1.07)	0.183			1.21 (0.85–1.71)	0.284
rs9841504	Allele	C	667	672	1		232	224	1
		G	107	108	1.00 (0.75–1.34)	0.990	38	46	1.25 (0.79–2.00)	0.342
	Genotype	CC	286	291	1		98	95	1
		CG	95	90	0.93 (0.67–1.29)	0.660	36	34	0.97 (0.56–1.68)	0.916
		GG	6	9	1.46 (0.51–4.15)	0.483	1	6	6.20 (0.73–52.44)	0.094
	Dominant	CC/ CG-GG	286/101	291/99	0.96 (0.70–1.32)	0.801	98/37	95/40	1.11 (0.66–1.89)	0.693
	Recessive	CC-CG/ GG	381/6	381/9	1.48 (0.52–4.21)	0.460	134/1	129/6	6.24 (0.74–52.59)	0.092
	Log-additive				1.00 (0.75–1.33)	0.987			1.25 (0.78–1.99)	0.350
Age, years			> 65				≤ 65
rs10934270	Allele	C	351	429	1		585	549	1
		T	31	43	1.14 (0.70–1.84)	0.607	77	29	0.40 (0.26–0.62)	3.25×10−5*
	Genotype	CC	161	194	1		261	261	1
		CT	29	41	1.19 (0.70–2.01)	0.519	63	27	0.40 (0.25–0.66)	3.29×10−4*
		TT	1	1	0.85 (0.05–13.81)	0.909	7	1	0.19 (0.02–1.54)	0.119
	Dominant	CC/ CT-TT	161/30	194/42	1.18 (0.70–1.98)	0.538	261/70	261/28	0.38 (0.24–0.62)	1.07×10−4*
	Recessive	CC-CT/ TT	190/1	235/1	0.83 (0.05–13.40)	0.893	324/7	288/1	0.21 (0.03–1.73)	0.147
	Log-additive				1.15 (0.70–1.90)	0.571			0.41 (0.26–0.64)	1.15×10−4*
rs9288999	Allele	A	225	266	1		369	347	1
		G	157	206	1.11 (0.84–1.46)	0.455	291	231	0.84 (0.67–1.06)	0.143
	Genotype	AA	65	78	1		105	102	1
		AG	95	110	0.93 (0.60–1.44)	0.751	159	143	0.87 (0.60–1.25)	0.440
		GG	31	48	1.21 (0.69–2.14)	0.505	66	44	0.56 (0.34–0.91)	0.019
	Dominant	AA/ AG-GG	65/126	78/158	1.00 (0.66–1.51)	0.995	105/225	102/187	0.77 (0.55–1.09)	0.146
	Recessive	AA-AG/ GG	160/31	188/48	1.26 (0.76–2.09)	0.363	264/66	245/44	0.61 (0.39–0.94)	0.025
	Log-additive				1.07 (0.82–1.41)	0.613			0.77 (0.60–0.97)	0.027
rs9841504	Allele	C	329	393	1		570	503	1
		G	53	79	1.25 (0.86–1.82)	0.250	92	75	0.92 (0.67–1.28)	0.635
	Genotype	CC	141	166	1		243	220	1
		CG	47	61	1.09 (0.7–1.71)	0.697	84	63	0.89 (0.61–1.30)	0.546
		GG	3	9	2.56 (0.67–9.76)	0.168	4	6	1.85 (0.50–6.81)	0.358
	Dominant	CC/ CG-GG	141/50	166/70	1.18 (0.77–1.82)	0.451	243/88	220/69	0.93 (0.64–1.35)	0.707
	Recessive	CC-CG/ GG	188/3	227/9	2.5 (0.66–9.49)	0.177	327/4	283/6	1.90 (0.52–6.99)	0.335
	Log-additive				1.23 (0.85–1.8)	0.275			0.99 (0.70–1.38)	0.933
rs73230612	Allele	T	223	283	1		387	343	1
		C	159	189	0.94 (0.71–1.23)	0.640	275	235	0.96 (0.77–1.21)	0.753
	Genotype	TT	62	87	1		119	97	1
		TC	99	109	0.80 (0.52–1.23)	0.309	149	149	1.19 (0.83–1.70)	0.353
		CC	30	40	0.92 (0.51–1.65)	0.781	63	43	0.83 (0.51–1.35)	0.459
	Dominant	TT/ TC-CC	62/129	87/149	0.83 (0.55–1.24)	0.365	119/212	97/192	1.08 (0.77–1.52)	0.650
	Recessive	TT-TC/ CC	161/30	196/40	1.05 (0.62–1.77)	0.862	268/63	246/43	0.75 (0.49–1.17)	0.204
	Log-additive				0.93 (0.70–1.23)	0.593			0.96 (0.76–1.21)	0.706

Notes: p values were calculated by logistic regression analysis with adjustments for age and gender. Bold p < 0.05 respects the data is statistically significant. *p indicate that after Bonferroni correction (p < 0.05/4) means the data is statistically significant.

Abbreviations: SNP, single nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval.

In age stratification, the median age (65-year) was set as the cut-off value for all subjects. In order to explore the effect of age on EC risk, we divided all subjects into two groups as ≤65 years and >65 years. The contributions of rs10934270 (allele: T vs C, OR = 0.40, p = 3.25 × 10−5; genotype: CT vs CC, OR = 0.40, p = 3.29 × 10−4; dominant: CT-TT vs CC, OR = 0.38, p = 1.07 × 10−4; and log-additive: OR = 0.41, p = 1.15×10−4) and rs9288999 (genotype: GG vs AA, OR = 0.56, p = 0.019; recessive: GG vs AA-AG, OR = 0.61, p = 0.025; and log-additive: OR = 0.77, p = 0.027) to a decreased EC risk (Table 4) were observed in subjects aged ≤65 years. The significant association of rs10934270 with EC risk in subjects aged 65 or younger still existed after Bonferroni correction.

In the subgroup with BMI ≤ 24 kg/m2 (Table 5), the risk-reducing association of rs10934270 with EC occurrence was found under the allele (T vs C, OR = 0.42, p = 2.11 × 10−5), genotype (CT vs CC, OR = 0.48, p = 0.002, and TT vs CC, OR = 0.13, p = 0.018), dominant (CT-TT vs CC, OR = 0.44, p = 3.64 × 10−4), recessive (TT vs CC-CT, OR = 0.15, p = 0.027), and log-additive (OR = 0.45, p = 1.35 × 10−4) models. The significant association of rs10934270 with EC risk remained after Bonferroni correction.

Table 5

Stratified Analysis by BMI and Smoking for the Associations Between ZBTB20 Polymorphisms and the Risk of Esophageal Cancer

SNP ID	Model	Genotype	Control	Case	OR (95% CI)	p–value	Control	Case	OR (95% CI)	p–value
BMI, kg/m2			> 24				≤ 24
rs10934270	Allele	C	356	135	1		277	812	1
		T	30	13	1.14 (0.58–2.26)	0.701	47	58	0.42 (0.28–0.63)	2.11×10−5*
	Genotype	CC	164	61	1		120	379	1
		CT	28	13	1.30 (0.62–2.73)	0.484	37	54	0.48 (0.30–0.77)	0.002*
		TT	1	0	/	/	5	2	0.13 (0.03–0.71)	0.018
	Dominant	CC/ CT-TT	164/29	61/13	1.26 (0.6–2.64)	0.534	120/42	379/56	0.44 (0.28–0.69)	3.64×10−4*
	Recessive	CC-CT/ TT	192/1	74/0	/	/	157/5	433/2	0.15 (0.03–0.81)	0.027
	Log-additive				1.21 (0.59–2.47)	0.605			0.45 (0.30–0.68)	1.35×10−4*
rs9288999	Allele	A	212	83	1		186	507	1
		G	174	65	0.95 (0.65–1.40)	0.810	136	363	0.98 (0.76–1.27)	0.874
	Genotype	AA	58	24	1		55	148	1
		AG	96	35	0.89 (0.47–1.67)	0.707	76	211	1.07 (0.71–1.61)	0.754
		GG	39	15	0.86 (0.39–1.90)	0.717	30	76	0.95 (0.56–1.61)	0.858
	Dominant	AA/ AG-GG	58/135	24/50	0.88 (0.49–1.59)	0.671	55/106	148/287	1.04 (0.70–1.52)	0.862
	Recessive	AA-AG/ GG	154/39	59/15	0.93 (0.47–1.85)	0.838	131/30	359/76	0.92 (0.57–1.47)	0.719
	Log-additive				0.92 (0.63–1.37)	0.695			0.99 (0.76–1.28)	0.937
rs9841504	Allele	C	334	125	1		281	741	1
		G	52	33	1.18 (0.69–2.01)	0.538	43	129	1.14 (0.78–1.65)	0.496
	Genotype	CC	141	54	1		120	318	1
		CG	52	17	0.81 (0.42–1.56)	0.529	41	105	0.99 (0.65–1.51)	0.963
		GG	0	3	/	/	1	12	4.96 (0.63–38.90)	0.127
	Dominant	CC/ CG-GG	141/52	54/20	0.96 (0.52–1.80)	0.907	120/42	318/117	1.08 (0.71–1.64)	0.713
	Recessive	CC-CG/ GG	193/0	71/3	/	/	161/1	426/12	4.98 (0.64–38.88)	0.126
	Log-additive				1.16 (0.66–2.05)	0.613			1.17 (0.81–1.70)	0.408
rs73230612	Allele	T	214	93	1		190	517	1
		C	172	55	0.74 (0.50–1.09)	0.122	134	353	0.97 (0.75–1.26)	0.807
	Genotype	TT	59	31	1		54	149	1
		TC	96	31	0.59 (0.32–1.09)	0.092	82	219	1.00 (0.67–1.50)	0.993
		CC	38	12	0.63 (0.28–1.41)	0.261	26	67	0.92 (0.53–1.61)	0.782
	Dominant	TT/ TC-CC	59/134	31/43	0.60 (0.34–1.07)	0.082	54/108	149/286	0.98 (0.67–1.44)	0.918
	Recessive	TT-TC/ CC	155/38	62/12	0.85 (0.41–1.77)	0.667	136/26	368/67	0.93 (0.56–1.52)	0.761
	Log-additive				0.75 (0.50–1.12)	0.159			0.97 (0.74–1.27)	0.816
Smoking			Yes				No
rs10934270	Allele	C	206	459	1		319	507	1
		T	28	31	0.50 (0.29–0.85)	0.009*	39	41	0.66 (0.42–1.05)	0.077
	Genotype	CC	91	216	1		144	233	1
		CT	24	27	0.43 (0.24–0.79)	0.006*	31	41	0.81 (0.49–1.36)	0.431
		TT	2	2	0.38 (0.05–2.72)	0.334	4	0	/	/
	Dominant	CC/ CT-TT	91/26	216/29	0.43 (0.24–0.77)	0.004*	144/35	233/41	0.72 (0.43–1.18)	0.188
	Recessive	CC-CT/ TT	115/2	243/2	0.43 (0.06–3.12)	0.406	175/4	274/0	/	/
	Log-additive				0.47 (0.28–0.81)	0.006*			0.65 (0.41–1.03)	0.069
rs9841504	Allele	C	214	420	1		301	464	1
		G	20	70	1.78 (1.06–3.01)	0.029	57	84	0.96 (0.66–1.38)	0.810
	Genotype	CC	97	181	1		124	199	1
		CG	20	58	1.51 (0.84–2.69)	0.166	53	66	0.77 (0.5–1.18)	0.229
		GG	0	6	/	/	2	9	2.69 (0.57–12.74)	0.211
	Dominant	CC/ CG-GG	97/20	181/64	1.65 (0.93–2.94)	0.086	124/55	199/75	0.84 (0.55–1.28)	0.413
	Recessive	CC-CG/ GG	117/0	239/6	/	/	177/2	265/9	2.89 (0.61–13.61)	0.179
	Log-additive				1.72 (1.01–2.95)	0.047			0.94 (0.65–1.37)	0.762
rs9288999	Allele	A	137	289	1		204	314	1
		G	97	201	0.98 (0.72–1.35)	0.912	154	234	0.99 (0.75–1.29)	0.925
	Genotype	AA	44	86	1		62	90	1
		AG	49	117	1.10 (0.66–1.83)	0.711	80	134	1.10 (0.72–1.70)	0.653
		GG	24	42	0.85 (0.45–1.62)	0.630	37	50	0.89 (0.52–1.54)	0.688
	Dominant	AA/ AG-GG	44/73	86/159	1.02 (0.64–1.64)	0.924	62/117	90/184	1.04 (0.69–1.55)	0.854
	Recessive	AA-AG/ GG	93/24	203/42	0.81 (0.46–1.44)	0.470	142/37	224/50	0.84 (0.52–1.36)	0.489
	Log-additive								0.97 (0.74–1.26)	0.794
rs73230612	Allele	T	130	275	1		215	345	1
		C	104	215	0.98 (0.71–1.34)	0.886	143	203	0.88 (0.67–1.16)	0.380
	Genotype	TT	38	79	1		63	104	1
		TC	54	117	1.02 (0.61–1.71)	0.949	89	137	0.91 (0.60–1.38)	0.658
		CC	25	49	0.93 (0.49–1.76)	0.831	27	33	0.73 (0.40–1.33)	0.302
	Dominant	TT/ TC-CC	38/79	79/166	0.99 (0.61–1.61)	0.971	63/116	104/170	0.87 (0.58–1.29)	0.484
	Recessive	TT-TC/ CC	92/25	196/49	0.92 (0.53–1.61)	0.781	152/27	241/33	0.77 (0.44–1.33)	0.350
	Log-additive								0.87 (0.65–1.15)	0.326

Notes: p values were calculated by logistic regression analysis with adjustments for age and gender. Bold p < 0.05 respects the data is statistically significant. *p indicate that after Bonferroni correction (p < 0.05/4) means the data is statistically significant.

Abbreviations: SNP, single nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval.

In smokers (Table 5), rs10934270 (allele: T vs C, OR = 0.50, p = 0.009; genotype: CT vs CC, OR = 0.43, p = 0.006; dominant: CT-TT vs CC, OR = 0.43, p = 0.004; and log-additive: OR = 0.47, p = 0.006) was associated with a decreased risk of EC, whereas rs9841504 (allele: G vs C, OR = 1.78, p = 0.029; and log-additive: OR = 1.72, p = 0.047) contributed to increased EC susceptibility. After Bonferroni correction, the significant association of rs10934270 with EC risk in smokers still existed.

When stratified by drinking, there was no correlation between ZBTB20 SNPs and EC risk in drinkers and non-drinkers (data no shown).

FPRP Analysis

FPRP analysis was performed to calculate positive findings, as shown in Table 6. At a prior probability level of 0.1, the significant association of rs10934270 remained noteworthy overall (FPRP = 0.040, and 0.055, and statistical power >90%). The correlation of rs10934270 with EC susceptibility was also positive in subgroups including ESCC patients, males, subjects aged ≤65 years, subjects with BMI ≤ 24 kg/m2, and smokers (FPRP < 0.2). Moreover, the association of rs9288999 with EC risk in subjects aged ≤65 years was still noteworthy (FPRP = 0.193). The low statistical power of subgroups may be related to the small sample size after stratification.

Table 6

False-Positive Report Probability Values for the Associations Between ZBTB20 Polymorphisms and Esophageal Cancer Susceptibility

Group/SNPs ID	Model	OR (95% CI)	p	Statistical Power	Prior Probability
					0.25	0.1	0.01	0.001	0.0001
Overall
rs10934270	Allele	0.64 (0.47–0.87)	0.004	0.942	0.014	0.040	0.315	0.823	0.979
	Genotype	0.68 (0.49–0.96)	0.029	0.960	0.081	0.210	0.745	0.967	0.997
	Dominant	0.65 (0.46–0.91)	0.011	0.906	0.091	0.232	0.768	0.971	0.997
	Log-additive	0.64 (0.47–0.88)	0.005	0.936	0.019	0.055	0.389	0.865	0.985
Esophageal squamous cell carcinoma
rs10934270	Allele	0.66 (0.47–0.93)	0.017	0.944	0.053	0.143	0.648	0.949	0.995
	Dominant	0.67 (0.47–0.96)	0.031	0.945	0.085	0.217	0.753	0.969	0.997
	Log-additive	0.67 (0.48–0.94)	0.020	0.955	0.060	0.162	0.679	0.955	0.995
rs9841504	Genotype	2.59 (1.02–6.59)	0.045	0.294	0.319	0.584	0.939	0.994	0.999
	Recessive	2.56 (1.01–6.47)	0.047	0.301	0.319	0.584	0.939	0.994	0.999
Males
rs10934270	Allele	0.64 (0.44–0.93)	0.017	0.902	0.060	0.161	0.679	0.955	0.995
	Genotype	0.66 (0.44–0.99)	0.042	0.910	0.128	0.306	0.829	0.980	0.998
	Dominant	0.64 (0.43–0.95)	0.025	0.890	0.083	0.213	0.749	0.968	0.997
	Log-additive	0.65 (0.45–0.93)	0.020	0.924	0.056	0.152	0.664	0.952	0.995
Females
rs73230612	Allele	0.70 (0.50–0.99)	0.043	0.971	0.119	0.288	0.817	0.978	0.998
	Genotype	0.42 (0.19–0.94)	0.035	0.336	0.237	0.483	0.911	0.990	0.999
	Log-additive	0.67 (0.46–0.97)	0.034	0.939	0.098	0.245	0.781	0.973	0.997
Age ≤ 65 years
rs10934270	Allele	0.40 (0.26–0.62)	3.25×10−5	0.159	0.001	0.002	0.025	0.207	0.724
	Genotype	0.40 (0.25–0.66)	3.29×10−4	0.191	0.005	0.016	0.148	0.637	0.946
	Dominant	0.38 (0.24–0.62)	1.07×10−4	0.136	0.002	0.007	0.072	0.440	0.887
	Log-additive	0.41 (0.26–0.64)	1.15×10−4	0.191	0.001	0.004	0.043	0.312	0.820
rs9288999	Genotype	0.56 (0.34–0.91)	0.019	0.676	0.079	0.204	0.738	0.966	0.996
	Recessive	0.61 (0.39–0.94)	0.025	0.816	0.084	0.216	0.752	0.968	0.997
	Log-additive	0.77 (0.60–0.97)	0.027	0.889	0.074	0.193	0.724	0.964	0.996
BMI ≤ 24 kg/m2
rs10934270	Allele	0.42 (0.28–0.63)	2.11×10−5	0.200	<0.001	0.001	0.013	0.121	0.579
	Genotype	0.48 (0.30–0.77)	0.002	0.433	0.016	0.046	0.348	0.844	0.982
		0.13 (0.03–0.71)	0.018	0.060	0.481	0.735	0.968	0.997	1.000
	Dominant	0.44 (0.28–0.69)	3.64×10−4	0.289	0.004	0.011	0.107	0.546	0.923
	Recessive	0.15 (0.03–0.81)	0.027	0.081	0.505	0.753	0.971	0.997	1.000
	Log-additive	0.45 (0.30–0.68)	1.35×10−4	0.308	0.001	0.004	0.046	0.327	0.829
Smokers
rs10934270	Allele	0.50 (0.29–0.85)	0.009	0.500	0.059	0.158	0.674	0.954	0.995
	Genotype	0.43 (0.24–0.79)	0.006	0.313	0.059	0.158	0.674	0.954	0.995
	Dominant	0.43 (0.24–0.77)	0.004	0.306	0.042	0.117	0.594	0.937	0.993
	Log-additive	0.47 (0.28–0.81)	0.006	0.412	0.046	0.125	0.612	0.941	0.994
rs9841504	Allele	1.78 (1.06–3.01)	0.029	0.668	0.124	0.298	0.823	0.979	0.998
	Log-additive	1.72 (1.01–2.95)	0.047	0.708	0.171	0.383	0.872	0.986	0.999

Notes: p values were calculated by logistic regression analysis with adjustments for age. Statistical power was calculated using the number of observations in the subgroup and the OR and p values in this table Bold prior probability < 0.2 (false-positive report probability threshold) respects the data is statistically significant.

Abbreviations: SNP, single nucleotide polymorphism; OR, odds ratio; 95% CI, 95% confidence interval.

MDR Analysis

MDR analysis was used to detect the relationship between higher order interactions and EC risk (Table 7 and Figure 1). Rs10934270 was the most influential attribution factor for EC risk in the single-locus model (testing balanced accuracy of 0.5297, and cross–validation consistency of 10/10), which was consistent with the logistic analysis results. The combination of rs10934270, rs9288999 and rs73230612 (testing balanced accuracy of 0.5211, and cross–validation consistency of 9/10) was the best multi-locus model. The dendrogram (Figure 1A) presented that rs10934270 and rs9841504 exhibited strong redundancy effects on EC susceptibility. The Fruchterman-Reingold (Figure 1B) revealed that rs9288999 and rs73230612 had synergistic interaction with the positive information gain (0.21%) of EC.

Table 7

SNP–SNP Interaction Models of Candidate SNPs Analyzed by the MDR Method

Model	Training Bal. Acc.	Testing Bal. Acc.	CVC	p
rs10934270	0.5287	0.5297	10/10	0.0119
rs9288999, rs73230612	0.5412	0.4933	4/10	0.0187
rs10934270, rs9288999, rs73230612	0.5629	0.5211	9/10	<0.0001
rs10934270, rs9288999, rs9841504, rs73230612	0.5810	0.5086	10/10	<0.0001

Notes: p values were calculated using χ2 tests. Bold indicate that p < 0.05 indicates statistical significance.

Abbreviations: MDR, multifactor dimensionality reduction; Bal. Acc., balanced accuracy; CVC, cross–validation consistency; OR, odds ratio; CI, confidence interval.

Figure 1

The dendrogram (A) and fruchterman Rheingold (B) of ZBTB20 SNP-SNP interaction for EC risk. (A) Short connections among nodes represent stronger interactions. (B) Positive percent entropy indicates synergistic interaction.

Discussion

The ZBTB20 gene, located on chromosome 3q13.31, is reported to be involved in the proliferation, migration and invasion of cancer.18,19 It has been found that ZBTB20 expression is increased in EC by silico analyses. Previous studies have revealed that ZBTB20 polymorphisms are related to many diseases, such as cognitive aging,29 systemic lupus erythematosus,30 autism spectrum disorders,31 and gastric cancer.21 Here, a hospital-based study of 525 EC patients and 522 healthy controls explored the relationship between four SNPs (rs10934270, rs9288999, rs9841504, and rs73230612) in ZBTB20 and EC occurrence among the Chinese Han population. The results demonstrated for the first time that rs10934270-T was associated with lower EC susceptibility with statistical power >90% in overall analysis. However, there are no reports on rs10934270. Moreover, we also found that the rs9841504 GG genotype might be a risk-increasing factor for ESCC. Nevertheless, a previous study has shown no significant relationship between rs9841504-GG and ESCC risk,32 such inconsistencies in these studies might be due to different behaviors or sample sizes. Based on the GTEx Portal database, genotypes of rs10934270 were related to ZBTB20 mRNA expression in whole blood. These results suggest that rs10934270 may be involved in EC carcinogenic by affecting the expression or function of ZBTB20. This may be new biological findings in the development of EC; however, further experimental confirmation is still required.

It is well known that genetic, environmental, and behavioral risk factors may affect EC development.33 According to reports, the risk of EC increases with age, and the incidence of EC is higher in men than in women.34 In age stratification, the associations of rs10934270 T allele and rs9288999 GG genotype with decreased EC risk were observed in subjects aged ≤65 years, but not in subjects aged >65 years. According to the gender-stratified analysis, rs10934270-T was associated with reduced EC risk in males, and rs73230612-C was a protective factor against EC in females. These results indicated that the association between ZBTB20 polymorphisms and EC susceptibility appeared to be age- and gender-dependent. Obesity, cigarette-smoking, and alcohol-drinking are known risk factors for EC.35 Higher BMI levels also increase the risk of EC.36 Additionally, the risk-reducing association of rs10934270-T with EC occurrence was found in the subgroup with BMI ≤ 24 kg/m2. Smokers have a 2.21–3.73 fold increased risk of EC compared with non-smokers.37 In smokers, rs10934270-T was related to a decreased EC risk, whereas rs9841504-G contributed to increased EC susceptibility. These results need to be verified in a larger population. Alcohol consumption is associated with increased EC occurrence.38 When stratified by drinking, no association between ZBTB20 SNPs and EC risk was found. These findings suggested that gene-behavioral habit interactions might play a certain role in the carcinogenesis of EC. However, the results should be interpreted with caution due to the relatively small sample size in stratified analyses.

Furthermore, exploring intragenic SNP–SNP interactions can also help us discover potential risk factors for the onset of EC.39 The results of MDR showed that rs10934270 was the most influential attribution factor for EC risk in the single-locus model and the combination of rs10934270, rs9288999 and rs73230612 was the best multi-locus model.

Inevitably, this study has some limitations. First, the hospital-based research has selection bias, and all participants are Chinese Han population, so these findings may not be applicable to other populations. Second, only four SNPs were chosen to explore the effect of ZBTB20 variants on EC occurrence, and other loci in ZBTB20 were not investigated. The association between other SNPs in ZBTB20 and the risk of EC requires further evaluation in the future. Third, research on the functions of these SNPs and their association with the expression level of ZBTB20 should be conducted, which will further confirm the results of our study. The potential mechanisms and functions of these SNPs hidden behind the association need to be further explored in detailed experiments. Fourth, there was a limited sample size in the stratification analysis. Hence, a larger sample size is needed to verify our findings. Fifth, given that EC is a complex multifactorial disease that may be influenced by genetic and environmental factors, the role of environmental factors in the association of ZBTB20 variants with EC risk should be considered. In the future we will enlarge the cohort of subjects to explore the interaction between ZBTB20 variants and environmental factors on EC risk.

Conclusion

Taken together, our study is the first to report that ZBTB20 rs10934270-T is associated with lower EC susceptibility in the Chinese Han population. These data provide scientific evidence for understanding the influence of ZBTB20 on the occurrence of EC. However, it is still necessary to conduct functional studies to clarify the molecular mechanisms of EC behind these associations.