Association of ATM and BMI‐1 genetic variation with breast cancer risk in Han Chinese

Abstract We tested the hypothesis that genetic variation in ATM and BMI‐1 genes can alter the risk of breast cancer through genotyping 6 variants among 524 breast cancer cases and 518 cancer‐free controls of Han nationality. This was an observational, hospital‐based, case–control association study. Analyses of single variant, linkage, haplotype, interaction and nomogram were performed. Risk was expressed as odds ratio (OR) and 95% confidence interval (CI). All studied variants were in the Hardy‐Weinberg equilibrium and were not linked. The mutant allele frequencies of rs1890637, rs3092856 and rs1801516 in ATM gene were significantly higher in cases than in controls (P = .005, <.001 and .001, respectively). Two variants, rs1042059 and rs201024480, in BMI‐1 gene were low penetrant, with no detectable significance. After adjustment, rs189037 and rs1801516 were significantly associated with breast cancer under the additive model (OR: 1.37 and 1.52, 95% CI: 1.10‐1.71 and 1.14‐2.04, P: .005 and .005, respectively). In haplotype analysis, haplotypes A‐C‐G‐G (in order of rs189037, rs3092856, rs1801516 and rs373759) and A‐C‐A‐A in ATM gene were significantly associated with 1.98‐fold and 6.04‐fold increased risk of breast cancer (95% CI: 1.36‐2.90 and 1.65‐22.08, respectively). Nomogram analysis estimated that the cumulative proportion of 3 significant variants in ATM gene was about 12.5%. Our findings collectively indicated that ATM gene was a candidate gene in susceptibility to breast cancer in Han Chinese.

widely applied to genetic association studies. 12,13 Using this approach, we genotyped 2 candidate genes, HMGB1 and RAGE, and found a cumulative impact of multiple variants on breast cancer risk. 14 In this study, we focused on 2 cancer-predisposing genes, ATM and BMI-1, in association with breast cancer. The involvement of ATM and BMI-1 in carcinogenesis is biologically plausible. [15][16][17][18] ATM is an acronym for ataxia telangiectasia mutated. ATM is a serine/threonine kinase involving in DNA damage repair, cell arrest and apoptosis, and chromatin remodelling. 19,20 Genetic defects in ATM can cause multiple system dysfunctions and increase tumour susceptibility. [21][22][23][24] BMI-1 is an acronym for B-cell-specific Moloney murine leukaemia virus integration site 1. BMI-1 is a polycomb protein that plays an important role in tumour cell development and maintaining stem cell populations of many cell lineages. 25 We therefore develop a hypothesis that ATM and BMI-1 genes are candidate genes of breast cancer. To test this hypothesis, we genotyped 4 variants in ATM gene and 2 variants in BMI-1 gene in 524 breast cancer cases and 518 cancer-free controls from Heilongjiang province, China to see whether they can alter the risk of breast cancer.

| Study design
This is an observational, hospital-based, case-control, genetic association study, as previously reported. 14

| Study participants
Only female participants who were Han Chinese were recruited in this study. Recruitment was conducted in 4 hospitals of Heilongjiang province, China, that is, Daqing Oilfield General Hospital, The 2nd and 3rd Affiliated Hospitals of Qiqihar Medical University and Qiqihar Jianhua Hospital. This study was conducted during the period from January 2013 to August 2015.

| Ethical approval
The conduct of this study was approved by the Ethics Committee of Qiqihar Medical University. Written informed consent was obtained from all study participants at the time of recruitment. This study complied with the Declaration of Helsinki.

| Breast cancer diagnosis
A patient was recorded to have breast cancer if he or she had either newly diagnosed or histopathologically confirmed or previously untreated breast cancer.

| Eligibility criteria
Control participants had no clinical evidence of any types of cancer except non-melanoma skin cancer. All controls reported to have no family history of cancer. No restriction was placed on age and tumour stage at the time of recruitment.

| Sample size
In total, 1042 female participants were eligible for inclusion. There were 524 cases with breast cancer and 518 controls free of cancer.

| Information collection
The following information was collected from cases with breast cancer, including age of first onset, age of menarche, menopausal age, family history of cancer, invasion depth (T1-T4), tumour stage (I-III) and lymph node. From controls free of cancer, age at enrolment and age of menarche were recorded.

| Phlebotomy
Venous blood samples (3 mL) were drawn into 5-mL vacutainer tubes containing K 3 -EDTA tubes from each participant. Plasma was separated by centrifugation at 4°C and kept frozen in a deep freezer at À80°C until assayed.

| DNA extraction
Genomic DNA was extracted from leucocytes using the phenol-cholesterol method according to a standard procedure.

| Variant selection
Four variants in ATM gene were selected, including rs189037 (A-111G), rs3092856 (1380His>Tyr or C4138T), rs1801516 (1853Asp>Asn) and rs373759 (126713G>A). Two variants were selected from BMI-1 gene, including rs1042059 (18Cys>Tyr) and rs201024480 (310Ser>Asn). The selection of these variants was based on published papers [25][26][27][28][29][30] and the NCBI-Gene website analysis (https://www.ncbi.nlm.nih.gov/gene/). In this study, no restriction was placed on the cut-off point of minor allele frequency of selected variants, as available linkage analyses only identified some rare variants of genes involved in DNA repair, including ATM gene under investigation, and these variants were associated with a moderate risk of breast cancer. 1,31

| Genotyping
Genomic sequences of 6 variants in ATM and BMI-1 genes were amplified by polymerase chain reaction. Genotyping was determined using ligase detection reaction method. 32 For each allele of a single variant, a specific probe was synthesized, and an additional common probe capped with 6-carboxy-fluorescein at the 3 0 end and with horylated at the 5 0 end was also synthesized. The primers and probes are available upon reasonable request.
Using the same method, 60 randomly selected samples were regenotyped, and the results were completely identical.
Genotyping was performed by laboratory workers in a manner blind to the case-control status and related characteristics of study participants.

| Statistical analyses
Continuous variables expressed as mean (standard deviation) were compared between cases and controls using t test or Mann-Whitney U test depending on its distribution. Categorical variable expressed as percentage were compared using the v 2 test. Testing Hardy-Weinberg equilibrium in control participants, as well as genotype and allele differences between cases and controls was completed by the v 2 test or Fisher's exact test. Breast cancer risk conferred by genetic variants was calculated using logistic regression analysis after adjusting for age and age of menarche. Effect size was indexed as odds ratio (OR) and 95% confidence interval (95% CI). In addition, given the limited sample size of this study, an internal validation was performed by evenly and randomly splitting all study participants into 2 groups, viz. training group (n = 512) and testing group (n = 512). Risk prediction for breast cancer was, respectively, conducted in both groups.
Linkage analysis was performed using the HaploView software Release 4.2 available at https://www.broadinstitute.org/haploview/ haploview, and linkage magnitude was expressed as D prime (D').
Haplotype analysis was performed using the HAPLO.STATS program after controlling for age and age of menarche. This program was realized using the R Project for Statistical Computing Release 3.4.3 available at https://www.r-project.org/.
Interaction analysis was performed using the open-source multifactor dimensionality reduction (MDR) software Release 3.0.2 available at http://www.multifactordimensionalityreduction.org/.
A risk nomogram calculator was produced based on baseline information and significant variants identified. The nomogram was depicted using regression modelling strategies (rms) program. This program also was realized using the R Project for Statistical Computing Release 3.4.3. Specifically, a nomogram is a 2-dimensional diagram that allows the approximate graphical computation of a mathematical function, and its accuracy can be justified by the concordance index (C-index). 33 The C-index measures the magnitude of concordance between predicted probabilities and the actual chance of having breast cancer.
Unless otherwise indicated, analyses were performed with the STATA software Release 13.0 (StataCorp LP, College Station, TX, USA).

| Baseline characteristics
This study involved 524 cases with breast cancer and 518 controls free of cancer. Table 1 shows the comparison of baseline characteristics between cases and controls. Mean age and mean age of menarche differed significantly between the 2 groups (both P < .001). In cases, mean menopausal age was 50.19 years and only a small proportion of cases had a family history of cancer (5.95%). In case of invasion depth, 49.69%, 42.77%, 3.77% and 3.77% of cases were in T1, T2, T3 and T4, respectively. Tumour stages II (49.27%) and III (46.10%) accounted for the majority of breast cancer cases, and 42.47% of cases had positive lymph nodes.

| Hardy-Weinberg equilibrium test
All studied variants were consistent with the distributions predicted by the Hardy-Weinberg equilibrium in controls (all P > .01).

| Linkage analysis
Linkage analysis indicated that the 4 variants in ATM gene and the 2 variants in BMI-1 gene were not correlated ( Figure S1). Table 2 shows the genotype and allele distributions/frequencies of 6 variants in ATM and BMI-1 gene. The mutant allele frequencies of rs1890637 (A), rs3092856 (T) and rs1801516 (A) were significantly higher in cases than in controls (P = .005, <.001 and .001, respectively). The genotype distributions differed remarkably significantly for rs3092856 and rs1801516 between cases and controls (both P < .001). The 2 variants, rs1042059 and rs201024480, in BMI-1 gene belonged to low-penetrance mutations, and there was no hint of significance with breast cancer risk. The genotype and allele comparisons of 6 studied variants between breast cancer cases, respectively, stratified by tumour stage, invasion depth, lymph node and controls are presented in Table S1.

| Haplotype analysis
Haplotype analysis was, respectively, conducted in ATM gene and BMI-1 gene, as they are mapped on different chromosomes. The derived haplotype frequencies and risk estimates for breast cancer are presented in Table 4. In ATM gene, haplotype G-C-G-G (alleles arranged by order of rs189037, rs3092856, rs1801516 and rs373759, with the same hereafter) was the most common, and its frequency was significantly higher in controls than in cases (30.82% vs 20.74%, P < .001). A low-penetrance haplotype A-C-A-A was overrepresented in cases than in controls (3.98% vs 0.95%, P = .001).
When taking the most common haplotype (G-C-G-G) as the ref- Other haplotypes carried no significance in association with breast cancer.

| Interaction analysis
The interaction of 6 studied variants is illustrated in Figure S2. The contribution of rs1801516 to breast cancer risk was the largest

| Nomogram presentation
A nomogram calculator is presented in Figure 1