The Impact of BCL11A Polymorphisms on Endometrial Cancer Risk Among Chinese Han Females.

Background: Endometrial carcinoma (EC) is one of the most common malignant gynecological malignancies. BCL11A gene may have a tumor-suppressor role in EC. Until now, no studies have reported the effect of BCL11A variants on EC predisposition in Chinese population.
Methods: Six BCL11A polymorphisms were genotyped using Agena MassARRAY system among 509 EC patients and 506 matched healthy women. Risk assessment of the BCL11A polymorphisms for EC risk was performed by calculating odds ratios (OR) with 95% confidence intervals (CI) through logistic regression models.
Results: We found that rs7581162 (OR = 1.29, p = 0.012), rs10189857 (OR = 1.26, p = 0.028), rs1427407 (OR = 1.30, p = 0.015), rs766432 (OR = 1.27, p = 0.025), and rs6729815 (OR = 1.32, p = 0.008) in BCL11A were associated with higher susceptibility to EC in Chinese Han women. Age and BMI stratified analysis displayed that the risk association between BCL11A variants and EC predisposition might be age- and BMI-dependent. Haplotype analysis revealed that Ars10189857Trs1427407 and Grs10189857Grs1427407 haplotypes were related to an increased risk of EC. MDR analysis indicated that rs1427407 was the most influential attributor on EC risk in the single-locus model, and the best combination was the two-locus model containing rs7581162 and rs766432.
Conclusion: Our study provided the first evidence that rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 in BCL11A were risk factors for EC in Chinese Han women. These findings add our understanding of the role of BCL11A gene in EC pathogenesis.

Introduction

Endometrial carcinoma (EC) occurs in the endometrium characterized by a thickening or mass of the endometrium.1 EC is one of the most common malignant gynecological tumors with an estimated age-standardized incidence risk of 13.1.2 In China, EC ranks fourth among female cancers in both incidence and mortality, and the incidence increases steadily with decreasing age at diagnosis.3,4 Risk factors for EC are unopposed estrogen, family history of EC, age, excess body weight, diabetes, hypertension, and Lynch syndrome.5 Although several risk factors have been identified, endometrial carcinogenesis remains poorly understood. Genetic variants such as single nucleotide polymorphisms (SNPs) are known to play important roles in cancer predisposition.6–8 Many SNPs related to EC susceptibility have been identified in previous genome-wide association studies (GWAS), but many polymorphic loci have not been reported.

The B-cell lymphoma/leukemia 11A (BCL11A) gene encodes a C2H2 type zinc finger protein transcription factor, and the role of BCL11A in tumors appears to be contextual. In some cancers, it has oncogenic effects,9 while in some cancers, it may act as a tumor suppressor.10 For example, BCL11A, as an oncogene, promoted tumor formation, cancer cell mobility and epithelial-mesenchymal transition by activating the Wnt/β-catenin signaling pathway in breast cancer carcinogenesis.11 Downregulation of BCL11A protein in colorectal cancer cells was related to enhanced radioresistance, supporting BCL11A as a tumor-suppressor role.12 The expression of BCL11A in ER-/PR-EC was higher than that in normal endometrium, atypical hyperplasia endometrium, and ER+/PR+EC.13 Moreover, the expression of BCL11A in EC was associated with age, menopause, EC classification, para-aortic lymph node metastasis, tumor differentiation, histological type, ER/PR expression and p53 expression.13 A targeted next-generation sequencing study displayed that BCL11A mutations were associated with endometriosis and tumor lesions.14 In EC, consistent with an anti-cancer function, credible variants in BCL11A were associated with reduced BCL11A expression in endometrial tumors.15 A GWAS study in an Australian population found that BCL11A polymorphism was associated with EC risk.16 Until now, there were no studies reported the effect of BCL11A genetic variants on EC predisposition in Chinese population.

Here, six SNPs (rs7581162, rs10189857, rs1427407, rs766432, rs6729815, and rs2556378) in BCL11A with a minor allele frequency (MAF) > 0.05 in dbSNP database () and the call rate >95% in our study population were randomly selected and genotyped to evaluate the potential association between BCL11A variants and EC risk among the Chinese Han women at single-SNP or combined SNPs interface.

Materials and Methods

Subjects

A total of 509 patients with EC and 506 matched healthy women were recruited from Hainan General Hospital (Table 1). All participants were genetically unrelated Han Chinese women. Patients were diagnosed with EC based on histopathologic biopsies using the guidelines of the International Federation of Obstetrics and Gynecology (FIGO) criteria. Patients who received preoperative chemotherapy, radiotherapy, or hormone therapy were excluded. Patients with immunological diseases or other cancers were also excluded. Age-matched healthy control underwent routine gynecologic examination at the same hospital. Selection criteria included no malignancy, no history of cancer and no chronic or acute disease. Demographical and clinical characteristics were collected from medical records. This study was approved by the ethics committee of Hainan General Hospital (Ethical approval No.: Med-Eth-Re [2018] 14), in compliance with the Declaration of Helsinki. Informed consent was obtained from all recruited participants.

Table 1

Basic Information of Endometrial Cancer and Health Controls

Characteristics	Cases (n = 509)	Control (n = 506)	p-value
Age (years, mean ± SD)	54.94 ± 8.85	54.61 ± 9.07	0.553
BMI (kg/m2)
≥ 24	199 (39.1%)	191 (37.7%)	0.699
< 24	310 (60.9%)	315 (62.3%)
CEA (ng/mL, mean ± SD)	11.20 ± 2.09	2.08 ± 2.22	< 0.001
AFP (ng/mL, mean ± SD)	8.96 ± 5.32	2.86 ± 1.17	< 0.001
CA199 (U/mL, mean ± SD)	17.87± 20.99	12.79 ± 10.45	< 0.001
CA125 (U/mL, mean ± SD)	24.22 ± 39.00	12.97 ± 9.51	< 0.001
Stage
I–II	261 (51.3%)
III–IV	93 (18.3%)
Missing	155 (30.5%)

Notes: p values were calculated using χ2 tests or Student’s t-test. Bold indicated that p < 0.05 meant the data was statistically significant.

Abbreviations: SD, standard deviation; BMI, body mass index; CEA, carcinoembryonic antigen; AFP, alpha fetoprotein; CA, carbohydrate antigen.

SNP Selection and Genotyping

Six SNPs (rs7581162, rs10189857, rs1427407, rs766432, rs6729815, and rs2556378) in the BCL11A gene with minor allele frequency (MAF) > 0.05 in the dbSNP database () and call rate > 95% were selected. The potential functions of these polymorphisms were evaluated through the HaploReg v4.1 (), RegulomeDB () and GTEx Portal databases ().

Venous blood (5 mL) was collected from each subject in an EDTA vacutainer tube. Genomic DNA was isolated using a commercially available GoldMag DNA Purification Kit (GoldMag Co. Ltd, Xi′an, China). Genotyping was performed by the Agena MassARRAY system (Agena, San Diego, CA, USA) as previously described. Primers design () and data interpretation were performed by the corresponding supporting software following the manufacturer’s instructions. The MassARRAY platform is based on MALDI-TOF (matrix-assisted laser desorption/ionization—time of flight) mass spectrometry. The analytical accuracy of MALDI-TOF MS is quite high, 0.1–0.01% of the determined mass.17,18 Furthermore, 5% of the samples were used for re-genotyping to quality control, and the consistency rate was 100%.

Statistical Analysis

Age and clinical characteristics were expressed as mean ± standard deviation (SD), and differences between EC patients and health controls were evaluated by Student’s t-test. The difference of body mass index (BMI) between cases and controls were analyzed by χ2-test Hardy–Weinberg equilibrium (HWE) was assessed to compare the genotype frequencies in controls using the goodness-of-fit chi-square test. Comparison of genotype and allele frequencies in cases and controls were performed by χ2-test. The major-type allele was used as a reference, and minor-type allele was used as a risk factor. Risk assessment of BCL11A polymorphisms for EC risk was performed by calculating odds ratios (OR) with 95% confidence intervals (CI) using logistic regression models. D’ values for pairwise linkage disequilibrium (LD) plots were generated by Haploview software (version 4.2). The relationship of BCL11A haplotypes with EC susceptibility was evaluated by χ2-test and logistic regression model. Multifactor dimensionality reduction (MDR) analysis was performed using MDR_3.0.2 software to identify high-order interaction models for EC predisposition. False-positive report probability (FPRP) analysis was used to evaluate the noteworthy associations and statistical power of the significant findings.19,20 We set 0.2 as a FPRP threshold and assigned a prior probability of 0.1 for an association with genotypes under investigation. Analysis of Variance (ANOVA) was performed for the correlation of BCL11A variants with CEA, AFP, CA199, CA125 of EC patients and healthy controls. Data analysis was performed using SPSS version 20.0 (SPSS Inc., Chicago, IL, USA) and PLINK version 1.0.7 software. p-value was two-tailed and p < 0.05 was defined statistical significance, whereas adjusted p < 0.05/5 was considered significant after Bonferroni correction.

Results

Subject Characteristics

In the study, we enrolled 509 EC patients (54.94 ± 8.85 years) and 506 healthy controls (54.61 ± 9.07 years), as shown in Table 1. Age and BMI distributions were not significantly different between cases and controls (p = 0.553, and p = 0.669, respectively). There were significant difference in the levels of CEA, AFP, CA199, and CA125 between two groups (p < 0.001). There were 261 cases of I–II stage and 93 cases of III–IV stage.

Relationships of BCL11A Polymorphisms with EC Predisposition

Six SNPs in BCL11A were genotyped for subsequent studies, and the call rates were > 99.5%, as summarized in Table 2. The genotype distribution of rs2556378 polymorphism was not inconsistent with HWE (p = 0.001), therefore, this variant was excluded from subsequent studies. The MAFs of rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 in EC patients were higher than those in healthy controls. BCL11A rs7581162-T (OR = 1.29, 95% CI: 1.06–1.57, p = 0.012), rs10189857-A (OR = 1.26, 95% CI: 1.03–1.54, p = 0.028), rs1427407-T (OR = 1.30, 95% CI: 1.05–1.60, p = 0.015), rs766432-C (OR = 1.27, 95% CI: 1.03–1.56, p = 0.025), and rs6729815-T (OR = 1.32, 95% CI: 1.07–1.62, p = 0.008) was associated with the increased risk of EC. The significance of rs6729815 still existed after Bonferroni correction. The potential function of these polymorphisms by HaploReg v4.1 database and RegulomeDB database was displayed in Table 2. Based on GTEx Portal database, the genotypes of rs1427407 (p = 0.00026) was related to the mRNA expression of BCL11A in cell ().

Table 2

Basic Characteristics and Allele Model About Candidate SNPs in the BCL11A Gene

SNPs ID	Chr:Position	Allele (Minor/Major)	Call Rate	O(HET)	E(HET)	pa-value for HWE	MAF		Allele Model		Haploreg	Regulome DB
							Case	Control	OR (95% CI)	pb
rs7581162	2:60,477,349	T/A	99.9%	0.384	0.362	0.219	0.287	0.238	1.29 (1.06–1.57)	0.012	Promoter histone marks, Enhancer histone marks, Motifs changed, GRASP QTLhits	TF binding or DNase peak
rs10189857	2:60,486,100	A/G	99.8%	0.345	0.346	1.000	0.264	0.222	1.26 (1.03–1.54)	0.028	DNAse, Motifs changed, NHGRI/EBI GWAS hits	TF binding or DNase peak
rs1427407	2:60,490,908	T/G	99.8%	0.310	0.321	0.486	0.246	0.200	1.30 (1.05–1.60)	0.015	Enhancer histone marks, Proteins bound, Motifs changed, NHGRI/EBI GWAS hits, GRASP QTL hits	TF binding or DNase peak
rs766432	2:60,492,835	C/A	100%	0.338	0.335	0.895	0.254	0.212	1.27 (1.03–1.56)	0.025	Enhancer histone marks, DNAse, Motifs changed, NHGRI/EBI GWAS hits, GRASP QTL hits	TF binding + any motif + DNase Footprint + DNase peak
rs6729815	2:60,496,537	T/C	100%	0.330	0.335	0.790	0.262	0.212	1.32 (1.07–1.62)	0.008*	Enhancer histone marks, Motifs Changed, NHGRI/EBI GWAS hits	TF binding + DNase peak
rs2556378	2:60,535,367	T/G	100%	0.405	0.352	0.001	/	/	/	/	Promoter histone marks, Enhancer histone marks, DNAse, NHGRI/EBI GWAS hits, GRASP QTL hits	eQTL + TF binding/DNase peak

Notes: p-values for the HWE test were calculated using goodness of fit χ2 test. p were calculated using by Fisher’s exact test. Bold indicated that p < 0.05 meant the data was statistically significant. *p indicate that after Bonferroni correction (p < 0.05/5) means the data is statistically significant.

Abbreviations: SNP, single-nucleotide polymorphism; MAF, minor allele frequency; O(HET), observed heterozygosity; E(HET), expected heterozygosity; HWE, Hardy–Weinberg equilibrium.

Multiple genetic models were used to evaluate the relationship of BCL11A polymorphisms to EC predisposition. After adjusting for age, rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 were associated with higher EC susceptibility (Table 3). Specifically, the TT genotype of rs7581162 (genotype: TT vs AA, OR = 2.16, 95% CI: 1.27–3.69, p = 0.005; recessive: TT vs AA-AT, OR = 2.03, 95% CI: 1.20–3.42, p = 0.008; and additive: AA+AT+TT, OR = 1.30, 95% CI: 1.07–1.59, p = 0.010) and AA genotype of rs10189857 (genotype: AA vs GG, OR = 1.78, 95% CI: 1.05–3.01, p = 0.033; and additive: GG+GA+AA, OR = 1.26, 95% CI: 1.03–1.54, p = 0.027) contributed to increased EC risk. Rs1427407 had a risk-effect for EC under the genotype (GT vs GG: OR = 1.33, 95% CI: 1.02–1.74, p = 0.033), dominant (GT-TT vs GG: OR = 1.37, 95% CI: 1.06–1.76, p = 0.016) and additive (GG+GT+TT: OR = 1.30, 95% CI: 1.05–1.60, p = 0.014) models. Rs766432 was associated with increased EC predisposition (dominant: AC-CC vs AA, OR = 1.32, 95% CI: 1.03–1.70, p = 0.029; and additive: AA+AC+CC, OR = 1.28, 95% CI: 1.04–1.57, p = 0.022). Moreover, rs6729815 might have higher risk for the occurrence EC in the genotype (TT vs CC: OR = 1.94, 95% CI: 1.14–3.29, p = 0.015), dominant (CT-TT vs CC: OR = 1.32, 95% CI: 1.03–1.70, p = 0.030), recessive (TT vs CC-CT: OR = 1.79, 95% CI: 1.06–3.02, p = 0.028) and additive (TT+CC+CT:OR = 1.31, 95% CI: 1.07–1.61, p = 0.009) models. The significance of rs7581162 in the genotype and recessive models and rs6729815 in the additive model still existed after Bonferroni correction.

Table 3

The Effect of BCL11A SNPs on the Susceptibility to Endometrial Cancer

SNPs ID	Model	Genotype	Case	Control	Crude Analysis		Adjusted by Age
					OR (95% CI)	p	OR (95% CI)	p
rs7581162	Genotype	AA	261	288	1		1
		AT	204	194	1.16 (0.90–1.50)	0.259	1.16 (0.90–1.50)	0.257
		TT	44	23	2.11 (1.24–3.59)	0.006*	2.16 (1.27–3.69)	0.005*
	Dominant	AA	261	288	1		1
		AT-TT	248	217	1.26 (0.98–1.62)	0.066	1.27 (0.99–1.62)	0.063
	Recessive	AA-AT	465	482	1		1
		TT	44	23	1.98 (1.18–3.34)	0.010	2.03 (1.20–3.42)	0.008*
	Log-additive	—	—	—	1.30 (1.06–1.59)	0.011	1.30 (1.07–1.59)	0.010
rs10189857	Genotype	GG	280	305	1		1
		GA	189	174	1.18 (0.91–1.54)	0.209	1.18 (0.91–1.54)	0.206
		AA	40	25	1.74 (1.03–2.95)	0.038	1.78 (1.05–3.01)	0.033
	Dominant	GG	280	305	1		1
		GA-AA	229	199	1.25 (0.98–1.61)	0.076	1.26 (0.98–1.62)	0.072
	Recessive	GG-GA	469	479	1		1
		AA	40	25	1.63 (0.98–2.74)	0.062	1.66 (0.99–2.79)	0.054
	Log-additive	—	—	—	1.25 (1.02–1.53)	0.030	1.26 (1.03–1.54)	0.027
rs1427407	Genotype	GG	291	325	1		1
		GT	186	156	1.33 (1.02–1.74)	0.034	1.33 (1.02–1.74)	0.033
		TT	32	23	1.55 (0.89–2.72)	0.122	1.59 (0.91–2.78)	0.107
	Dominant	GG	291	325	1		1
		GT-TT	218	179	1.36 (1.06–1.75)	0.017	1.37 (1.06–1.76)	0.016
	Recessive	GG-GT	477	481	1		1
		TT	32	23	1.40 (0.81–2.43)	0.228	1.43 (0.82–2.49)	0.204
	Log-additive	—	—	—	1.29 (1.05–1.59)	0.016	1.30 (1.05–1.60)	0.014
rs766432	Genotype	AA	281	313	1		1
		AC	197	171	1.28 (0.99–1.67)	0.061	1.29 (0.99–1.67)	0.058
		CC	31	22	1.57 (0.89–2.77)	0.121	1.60 (0.90–2.84)	0.107
	Dominant	AA	281	313	1		1
		AC-CC	228	193	1.32 (1.02–1.69)	0.032	1.32 (1.03–1.70)	0.029
	Recessive	AA-AC	478	484	1		1
		CC	31	22	1.43 (0.81–2.50)	0.214	1.45 (0.83–2.55)	0.194
	Log-additive	—	—	—	1.27 (1.03–1.56)	0.025	1.28 (1.04–1.57)	0.022
rs6729815	Genotype	CC	283	315	1		1
		CT	185	167	1.23 (0.95–1.61)	0.120	1.23 (0.95–1.61)	0.119
		TT	41	24	1.90 (1.12–3.23)	0.017	1.94 (1.14–3.29)	0.015
	Dominant	CC	283	315	1		1
		CT-TT	226	191	1.32 (1.03–1.69)	0.031	1.32 (1.03–1.70)	0.030
	Recessive	CC-CT	468	482	1		1
		TT	41	24	1.76 (1.05–2.96)	0.033	1.79 (1.06–3.02)	0.028
	Log-additive	—	—	—	1.30 (1.07–1.60)	0.010	1.31 (1.07–1.61)	0.009*

Notes: p values were calculated by logistic regression analysis without or with adjustments for age. Bold indicated that p < 0.05 meant the data was statistically significant. *p indicate that after Bonferroni correction (p < 0.05/5) means the data is statistically significant.

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

Stratified Analysis for the Contribution of BCL11A Variant to EC Risk

We further explored the stratified analysis by age and BMI for the relationships of BCL11A variants with EC risk (Table 4). In the subjects with age > 55 years/BMI ≥ 24 kg/m2, no significant association between BCL11A polymorphism with the susceptibility to EC occurrence was observed. Among those aged 55 or younger and those had BMI < 24 kg/m2, we found that rs7581162-T, rs10189857-A, rs1427407-T, rs766432-C, and rs6729815-T were risk factors for the development of EC.

Table 4

Stratification Analysis by Age and BMI for the Effect of BCL11A SNPs on the Susceptibility to Endometrial Cancer

SNPs ID	Model	Genotype	Case	Control	OR (95% CI)	p	Case	Control	OR (95% CI)	p
Age			> 55 years				≤ 55 years
rs7581162	Allele	A	333	326	1		393	444	1
		T	113	96	1.15 (0.84–1.57)	0.373	179	144	1.40 (1.09–1.82)	0.010
	Genotype	AA	117	123	1		144	165	1
		AT	99	80	1.30 (0.88–1.92)	0.185	105	114	1.06 (0.75–1.50)	0.746
		TT	7	8	0.92 (0.32–2.61)	0.872	37	15	2.84 (1.50–5.40)	0.001*
	Dominant	AA	117	123	1		144	165	1
		AT-TT	106	88	1.27 (0.87–1.85)	0.224	142	129	1.27 (0.91–1.76)	0.158
	Recessive	AA-AT	216	203	1		249	279	1
		TT	7	8	0.82 (0.29–2.30)	0.707	37	15	2.78 (1.49–5.18)	0.001*
	Log-additive	—	—	—	1.18 (0.84–1.64)	0.343	—	—	1.38 (1.08–1.78)	0.012
rs10189857	Allele	G	344	331	1		405	453	1
		A	102	89	1.10 (0.80–1.52)	0.551	167	135	1.38 (1.06–1.80)	0.016
	Genotype	GG	128	131	1		152	174	1
		GA	88	69	1.31 (0.88–1.94)	0.190	101	105	1.10 (0.78–1.56)	0.588
		AA	7	10	0.72 (0.26–1.94)	0.512	33	15	2.53 (1.33–4.85)	0.005*
	Dominant	GG	128	131	1		152	174	1
		GA-AA	95	79	1.23 (0.84–1.81)	0.291	134	120	1.28 (0.92–1.78)	0.142
	Recessive	GG-GA	216	200	1		253	279	1
		AA	7	10	0.65 (0.24–1.74)	0.389	33	15	2.44 (1.29–4.60)	0.006*
	Log-additive	—	—	—	1.11 (0.80–1.55)	0.540	—	—	1.36 (1.05–1.75)	0.020
rs1427407	Allele	G	353	342	1		415	464	1
		T	93	80	1.13 (0.81–1.57)	0.485	157	122	1.44 (1.10–1.89)	0.008*
	Genotype	GG	134	140	1		157	185	1
		GT	85	62	1.43 (0.96–2.15)	0.081	101	94	1.27 (0.89–1.80)	0.187
		TT	4	9	0.46 (0.14–1.54)	0.209	28	14	2.37 (1.21–4.67)	0.012
	Dominant	GG	134	140	1		157	185	1
		GT-TT	89	71	1.31 (0.89–1.94)	0.177	129	108	1.41 (1.01–1.97)	0.043
	Recessive	GG-GT	219	202	1		258	279	1
		TT	4	9	0.41 (0.12–1.35)	0.142	28	14	2.18 (1.12–4.23)	0.022
	Log-additive	—	—	—	1.14 (0.80–1.60)	0.472	—	—	1.41 (1.08–1.83)	0.011
rs766432	Allele	A	350	338	1		409	459	1
		C	96	86	1.08 (0.78–1.50)	0.653	163	129	1.42 (1.09–1.85)	0.010
	Genotype	AA	130	135	1		151	178	1
		AC	90	68	1.37 (0.92–2.04)	0.118	107	103	1.23 (0.87–1.74)	0.248
		CC	3	9	0.34 (0.09–1.30)	0.114	28	13	2.55 (1.27–5.09)	0.008*
	Dominant	AA	130	135	1		151	178	1
		AC-CC	93	77	1.25 (0.85–1.84)	0.254	135	116	1.37 (0.99–1.91)	0.059
	Recessive	AA-AC	220	203	1		258	281	1
		CC	3	9	0.30 (0.08–1.14)	0.078	28	13	2.35 (1.19–4.64)	0.014
	Log-additive	—	—	—	1.09 (0.77–1.53)	0.644	—	—	1.40 (1.08–1.83)	0.012
rs6729815	Allele	C	347	338	1		404	459	1
		T	99	86	1.12 (0.81–1.55)	0.490	168	129	1.48 (1.14–1.93)	0.004*
	Genotype	CC	131	135	1		152	180	1
		CT	85	68	1.29 (0.87–1.92)	0.212	100	99	1.20 (0.84–1.70)	0.316
		TT	7	9	0.80 (0.29–2.21)	0.668	34	15	2.70 (1.42–5.14)	0.003*
	Dominant	CC	131	135	1		152	180	1
		CT-TT	92	77	1.23 (0.84–1.81)	0.289	134	114	1.39 (1.00–1.94)	0.050
	Recessive	CC-CT	216	203	1		252	279	1
		TT	7	9	0.73 (0.27–2.00)	0.540	34	15	2.52 (1.34–4.74)	0.004*
	Log-additive	—	—	—	1.13 (0.81–1.58)	0.477	—	—	1.43 (1.11–1.85)	0.006*
BMI			≥ 24 kg/m2				< 24 kg/m2
rs7581162	Allele	A	293	287	1		433	483	1
		T	105	93	1.11 (0.80–1.53)	0.541	187	147	1.42 (1.10–1.83)	0.006*
	Genotype	AA	109	105	1		152	183	1
		AT	75	77	0.94 (0.62–1.42)	0.756	129	117	1.33 (0.95–1.85)	0.092
		TT	15	8	1.77 (0.71–4.37)	0.218	29	15	2.41 (1.24–4.68)	0.009*
	Dominant	AA	109	105	1		152	183	1
		AT-TT	90	85	1.01 (0.68–1.51)	0.948	158	132	1.45 (1.06–1.99)	0.022
	Recessive	AA-AT	184	182	1		281	300	1
		TT	15	8	1.82 (0.75–4.41)	0.188	29	15	2.14 (1.12–4.08)	0.022
	Log-additive	—	—	—	1.10 (0.79–1.53)	0.569	—	—	1.44 (1.11–1.86)	0.005*
rs10189857	Allele	G	296	289	1		453	495	1
		A	102	91	1.09 (0.79–1.52)	0.587	167	133	1.37 (1.06–1.78)	0.017
	Genotype	GG	114	108	1		166	197	1
		GA	68	73	0.88 (0.58–1.35)	0.564	121	101	1.43 (1.02–2.00)	0.037
		AA	17	9	1.76 (0.75–4.14)	0.196	23	16	1.75 (0.89–3.42)	0.104
	Dominant	GG	114	108	1		166	197	1
		GA-AA	85	82	0.98 (0.65–1.46)	0.917	144	117	1.47 (1.07–2.03)	0.018
	Recessive	GG-GA	182	181	1		287	298	1
		AA	17	9	1.85 (0.80–4.28)	0.153	23	16	1.52 (0.79–2.95)	0.212
	Log-additive	—	—	—	1.09 (0.79–1.50)	0.620	—	—	1.38 (1.06–1.78)	0.016
rs1427407	Allele	G	304	299	1		464	507	1
		T	94	81	1.14 (0.81–1.60)	0.442	156	121	1.41 (1.08–1.84)	0.012
	Genotype	GG	118	117	1		173	208	1
		GT	68	65	1.04 (0.68–1.59)	0.867	118	91	1.57 (1.12–2.21)	0.009*
		TT	13	8	1.56 (0.62–3.95)	0.344	19	15	1.57 (0.77–3.18)	0.214
	Dominant	GG	118	117	1		173	208	1
		GT-TT	81	73	1.09 (0.73–1.64)	0.664	137	106	1.57 (1.13–2.17)	0.007*
	Recessive	GG-GT	186	182	1		291	299	1
		TT	13	8	1.54 (0.62–3.85)	0.352	19	15	1.33 (0.66–2.68)	0.419
	Log-additive	—	—	—	1.13 (0.81–1.58)	0.478	—	—	1.41 (1.08–1.84)	0.012
rs766432	Allele	A	303	298	1		156	499	1
		C	95	84	1.11 (0.80–1.55)	0.533	164	131	1.37 (1.05–1.78)	0.019
	Genotype	AA	71	70	1		165	199	1
		AC	12	7	1.00 (0.65–1.52)	0.985	126	101	1.52 (1.09–2.12)	0.014
		CC	116	114	1.64 (0.62–4.34)	0.321	19	15	1.58 (0.78–3.21)	0.208
	Dominant	AA	83	77	1		165	199	1
		AC-CC	187	184	1.05 (0.70–1.58)	0.799	145	116	1.53 (1.11–2.10)	0.010
	Recessive	AA-AC	12	7	1		291	300	1
		CC	—	—	1.64 (0.63–4.29)	0.313	19	15	1.34 (0.67–2.69)	0.411
	Log-additive	—	71	70	1.11 (0.79–1.55)	0.559	—	—	1.39 (1.07–1.82)	0.015
rs6729815	Allele	C	300	193	1		451	504	1
		T	98	89	1.08 (0.77–1.49)	0.665	169	126	1.50 (1.15–1.95)	0.003*
	Genotype	CC	117	110	1		166	205	1
		CT	66	73	0.85 (0.56–1.30)	0.456	119	94	1.57 (1.12–2.21)	0.009*
		TT	16	8	1.84 (0.76–4.50)	0.179	25	16	1.99 (1.03–3.87)	0.042
	Dominant	CC	117	110	1		166	205	1
		CT-TT	82	81	0.95 (0.63–1.42)	0.802	144	110	1.63 (1.18–2.26)	0.003*
	Recessive	CC-CT	183	183	1		285	299	1
		TT	16	8	1.96 (0.81–4.72)	0.134	25	16	1.68 (0.88–3.23)	0.116
	Log-additive	—	—	—	1.07 (0.77–1.48)	0.694	—	—	1.49 (1.15–1.93)	0.003*

Notes: p values were calculated by Fisher’s exact test or logistic regression analysis with adjustments for age. Bold indicated that p < 0.05 meant the data was statistically significant. *p indicate that after Bonferroni correction (p < 0.05/5) means the data is statistically significant.

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

The results of age stratification (age ≤ 55 years) were as follows: in the allele (rs7581162, OR = 1.40, p = 0.010; rs10189857, OR = 1.38, p = 0.016; rs1427407, OR = 1.44, p = 0.008; rs766432, OR = 1.42, p = 0.010; and rs6729815, OR = 1.48, p = 0.004, respectively), genotype (rs7581162, OR = 2.84, p = 0.001; rs10189857, OR = 2.53, p = 0.005; rs1427407, OR = 2.37, p = 0.012; rs766432, OR = 2.55, p = 0.008; and rs6729815, OR = 2.70, p = 0.003, respectively), dominant (rs1427407, OR = 1.41, p = 0.043), recessive (rs7581162, OR = 2.78, p = 0.001; rs10189857, OR = 2.44, p = 0.006; rs1427407, OR = 2.18, p = 0.022; rs766432, OR = 2.35, p = 0.014; rs6729815, OR = 2.52, p = 0.004, respectively) and additive (rs7581162, OR = 1.38, p = 0.012; rs10189857, OR = 1.36, p = 0.020; rs1427407, OR = 1.41, p = 0.011; rs766432, OR = 1.40, p = 0.012; rs6729815, OR = 1.43, p = 0.006, respectively) models. The significance of rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 among those aged 55 or younger still existed after Bonferroni correction.

The associated results of BMI stratification (BMI < 24 kg/m2) were as follows: rs7581162 in the allele (OR = 1.42, p = 0.006), genotype (OR = 2.41, p = 0.009), dominant (OR = 1.45, p = 0.022), recessive (OR = 2.14, p = 0.022), and additive (OR = 1.44, p = 0.005) models; rs10189857 in the allele (OR = 1.37, p = 0.017), genotype (OR = 1.43, p = 0.037), dominant (OR = 1.47, p = 0.018), and additive (OR = 1.38, p = 0.016) models; rs1427407 in the allele (OR = 1.41, p = 0.012), genotype (OR = 1.57, p = 0.009), dominant (OR = 1.57, p = 0.007), and additive (OR = 1.41, p = 0.012) models; rs766432 in the allele (OR = 1.37, p = 0.019), genotype (OR = 1.52, p = 0.014), dominant (OR = 1.53, p = 0.010), and additive (OR = 1.39, p = 0.015) models; rs6729815 in the allele (OR = 1.50, p = 0.003), genotype (OR = 1.57, p = 0.009, and OR = 1.99, p = 0.042), dominant (OR = 1.63, p = 0.003), and additive (OR = 1.49, p = 0.003) models. The significance of rs7581162, rs1427407, and rs6729815 among those with BMI < 24 kg/m2 still existed after Bonferroni correction.

The association of these SNPs with the stage of EC patients was assessed (). However, no significant association was detected between BCL11A SNPs and at age in EC patients.

The Association Between BCL11A Haplotypes and EC Susceptibility

Moreover, haplotype analysis was performed to estimate the association between BCL11A haplotypes and EC susceptibility. As shown in Figure 1, rs10189857 and rs1427407 are in linkage disequilibrium. Haplotype analysis revealed that Ars10189857Trs1427407 (adjusted OR = 1.30, 95% CI: 1.05–1.60, p = 0.016) and Grs10189857Grs1427407 (adjusted OR = 1.26, 95% CI: 1.03–1.54, p = 0.023) haplotypes were related to the increased EC risk (). Furthermore, AT (adjusted OR = 1.42, 95% CI: 1.09–1.85, p = 0.010) and GG (adjusted OR = 1.36, 95% CI: 1.05–1.75, p = 0.019) haplotypes also conferred higher EC predisposition among subjects with age ≤ 55 years.

Figure 1

LD plots of six SNPs in the BCL11A gene. The number in the diamond indicates the D’ value of pairwise LD between SNPs.

MDR Analysis for the Effect of BCL11A SNP-SNP Interaction on EC Risk

MDR analysis was applied to explore the effect of BCL11A SNP-SNP interaction on EC risk. The dendrogram (Figure 2A) and Fruchterman Reingold (Figure 2B) showed that the BCL11A SNP-SNP interaction had a strong redundant effect. The best single-locus and multi-locus models of BCL11A SNPs for EC susceptibility were summarized in Table 5. In the single-locus model, rs1427407 was the most influential attributor for EC risk (testing accuracy = 0.5356, cross-validation consistency (CVC) = 10/10). In the multi-locus model, the best combination was a two-locus model containing rs7581162 and rs766432 (testing accuracy= 0.5336, CVC= 5/10).

Figure 2

The dendrogram (A) and Fruchterman Reingold (B) of BCL11A SNP-SNP interaction for EC risk. (A) Short connections among nodes represent stronger redundant interactions. Green and blue color indicated weak interactions. (B) This graphical model describes the percent entropy explained by each SNP. Values in nodes represent the information gains of individual attribute (main effects). Values between nodes are information gains of each pair of attributes (interaction effects). Positive percent entropy indicates synergy whereas the negative percent entropy indicates redundancy.

Table 5.

SNP–SNP Interaction Models of the BCL11A Gene the Predisposition of Endometrial Cancer

Model	Training Bal. Acc.	Testing Bal. Acc.	CVC	OR (95% CI)	p
rs1427407	0.5356	0.5356	10/10	1.35 (1.05–1.74)	0.0277
rs7581162, rs766432	0.5412	0.5336	5/10	1.42 (1.08–1.86)	0.0106
rs7581162, rs10189857, rs766432	0.5447	0.5287	4/10	1.44 (1.12–1.86)	0.0047
rs7581162, rs10189857, rs1427407, rs766432	0.5470	0.5208	4/10	1.47 (1.14–1.90)	0.0029
rs7581162, rs10189857, rs1427407, rs766432, rs6729815	0.5479	0.5178	10/10	1.46 (1.14–1.88)	0.0029

Notes: p values were calculated using χ2 tests. Bold indicated that p < 0.05 meant the data was statistically significant.

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

FPRP Analysis for the Significant Findings

FPRP analysis was carried out to interrogate whether the significant findings were deserving attention (). At the prior probability level of 0.1, the significant association for rs7581162, rs10189857, rs1427407, rs766432 and rs6729815 remained noteworthy in the overall analysis (FPRP < 0.2) and statistical power > 85% under the allele model. In the subgroup at age < 55 years, significant findings remained noteworthy for these variants.

The Relationship of BCL11A SNPs with Characteristics Among EC Patients/Healthy Controls

The relationship of BCL11A SNPs with CEA, AFP, CA199, and CA125 among EC patients/healthy controls were assessed, as displayed in . However, no statistically association was observed.

Discussion

In this study, we found that rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 in BCL11A were associated with EC susceptibility in Chinese Han women, especially in subjects aged ≤ 55 years and BMI <24 kg/m2. Haplotype analysis showed that rs10189857 and rs1427407 were in linkage disequilibrium, and the Ars10189857Trs1427407 and Grs10189857Grs1427407 haplotypes were related to the increased EC risk. MDR analysis indicated that rs1427407 was the most influential attributor on EC risk in the single-locus model, and the best combination was a two-locus model containing rs7581162 and rs766432. Our study is the first to report that BCL11A variants were risk factors for EC in Chinese Han females.

The BCL11A gene, located on 2p16, spans 102 kb, which is considered to be an oncogene in malignant haematological diseases, and is first detected in B-cell chronic lymphocytic leukaemia.21 BCL11A can cause transcriptional repression of mammalian target genes by binging to DNA motifs and promoting the deacetylation of H3/H4 histone.22 BCL11A may participate in cell cycle and cell growth by inhibiting of P21 induction.23 In addition, BCL11A is involved in cell apoptosis by upregulating the expression of BCL2, BCL2-xL, and MDM2and inhibiting the activity of P53.24 These studies support the possible involvement of BCL11A in tumorigenesis. Studies have found that BCL11A is abnormally expressed in EC tissues, which is related to EC classification, lymph node metastasis, tumor differentiation, histological type, ER/PR expression.13 Targeted next-generation sequencing evaluated BCL11A mutations in patients with endometriosis.14 These lines of evidence have led us to formulate the hypothesis that BCL11A could be of pathogenic importance in EC.

Genetic variations of BCL11A gene may be associated with gene expression, thereby affecting the occurrence and development of disease. To date, BCL11A polymorphisms have been reported to be associated with various cancer, such as pancreatic cancers, laryngeal squamous cell carcinoma, and chronic lymphocytic leukemia.25–27 Previous study have revealed that BCL11A rs7579014 is a risk locus for EC occurrence and is related to the expression of BCL11A in endometrial tumors.15 BCL11A rs148261157 has been reported to increase the risk of EC.16 However, the association between rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 in BCL11A and EC susceptibility has not been reported. Our study displayed that the MAFs of these five SNPs in BCL11A were higher in EC patients than in healthy controls, and had the increased risk effect on EC predisposition. To date, the functions of these SNPs have not been reported. In bioinformatics analysis, results from HaploReg v4.1 database displayed that these SNPs may be associated with promoter/Enhancer histone marks, DNAse, proteins bound, motifs changed, NHGRI/EBI GWAS hits, and/or GRASP QTL hits.28 Previously, rs766432 in the intronic regions of BCL11A gene may affect the binding of protein to this region.29 Rs1427407 in the second intron of the trans-acting element BCL11A is associated with the expression of BCL11A.30 Based on GTEx Portal database, the genotypes of rs1427407 was related to the mRNA expression of BCL11A in cell. These results suggest that these loci may be involved in EC carcinogenic by affecting the expression or function of BCL11A, which requires further experimental confirmation.

Age is a risk factor for the development of EC.31 A study reported that median age at diagnosis of endometrial cancer was 55 years.32 Besides, the mean ages of EC patients and the controls were respectively 54.94 ± 8.85 years and 54.61 ± 9.07 years in the study. In order to explore the contribution of age, we have divided the cases and controls into two groups as ≤ 55 years and > 55 years. The results of age-stratified analysis displayed that rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 were associated with EC susceptibility in subjects aged ≤ 55 years, but not in subjects aged > 55 years, suggesting that the risk association between BCL11A variants and EC predisposition might be age-dependent. Obesity is a risk factor for endometrial cancer risk and mortality.33 Using body mass index (BMI) as a measure of obesity, we assessed the association of BCL11A variants with BMI. Among those aged 55 or younger and those had BMI < 24 kg/m2, we found that rs7581162-T, rs10189857-A, rs1427407-T, rs766432-C, and rs6729815-T were risk factors for the development of EC. The underlying mechanism of this correlation awaits further study. Haplotype-based analysis may be more effective than single-locus analysis when there is linkage disequilibrium among SNPs.34 Haplotype analysis revealed that rs10189857 and rs1427407were in linkage disequilibrium, and the Ars10189857Trs1427407 and Grs10189857Grs1427407 haplotypes were related to an increased risk of EC. MDR analysis was used to identify specific combination effects of genetic variants with high-order interactions.35 In this study, we found that rs1427407 was the most influential attributor for EC risk in the single-locus model, and the best combination was the two-locus model incorporating rs7581162 and rs766432 in BCL11A.

Although our results provided evidence on the relationship between BCL11A polymorphisms and EC predisposition, several limitations should not be neglected. First, our subjects were recruited from a hospital that may contribute to selection bias. Second, due to the lack of adequate personal information, this study failed to assess the impact of gene-environment interactions on EC susceptibility and the association of genetic variants with clinicopathological data of EC patients. In the future, we would like to enlarge sample size and complete the clinicopathological data to evaluate the relationship. Third, although our findings suggested that BCL11A variants were associated with increased the risk of EC, the potential mechanisms and functions of these SNPs underlying the association have not been revealed. Fourth, the expression data of BCL11A is missing. In subsequent research, we would design detailed experiments to further explore the expression data of BCL11A and the potential mechanisms and functions of these SNPs in EC.

In conclusion, our study first provides evidence that rs7581162, rs10189857, rs1427407, rs766432, and rs6729815 in BCL11A are risk factors for EC occurrence in Chinese Han women. These findings increased our understanding of the role of BCL11A gene in EC pathogenesis. However, the expression data of BCL11A and the association between SNPs and BCL11A expression need further explore in the more detailed experiments.