DUSP1 promoter methylation in peripheral blood leukocyte is associated with triple-negative breast cancer risk

DNA methylation is one of the most common epigenetic alterations, providing important information regarding cancer risk and prognosis. A case-control study (423 breast cancer cases, 509 controls) and a case-only study (326 cases) were conducted to evaluate the association of DUSP1 promoter methylation with breast cancer risk and clinicopathological characteristics. No significant association between DUSP1 methylation in peripheral blood leukocyte (PBL) DNA and breast cancer risk was observed. DUSP1 methylation was significantly associated with ER/PR-negative status; in particular, triple-negative breast cancer patients showed the highest frequency of DUSP1 methylation in both tumour DNA and PBL DNA. Soybean intake was significantly correlated with methylated DUSP1 only in ER-negative (OR 2.978; 95% CI 1.245–7.124) and PR negative (OR 2.735; 95% CI 1.315–5.692) patients. Irregular menstruation was significantly associated with methylated DUSP1 only in ER-positive (OR 3.564; 95% CI 1.691–7.511) and PR-positive (OR 3.902, 95% CI 1.656–9.194) patients. Thus, DUSP1 methylation is a cancer-associated hypermethylation event that is closely linked with triple-negative status. Further investigations are warranted to confirm the association of environmental factors, including fruit and soybean intake, irregular menstruation, and ER/PR status, with DUSP1 methylation in breast tumour DNA.

SCiEnTifiC RePoRTS | 7:43011 | DOI: 10.1038/srep43011 inactivation and gene silencing, and contributes to the tumorigenesis of several cancers [16][17][18] . It has been proposed that the methylation status of some CpG sites could be passed on from previous generations as an inherited marker 19 . Several studies have been conducted on the changes in DNA methylation in blood leukocyte DNA, and suggested a link of blood leukocyte DNA methylation with cancer susceptibility [20][21][22][23] . Ji-Yeob et al. found that leukocyte DNA hypomethylation is independently associated with the development of breast cancer 24 . Thus, peripheral blood leukocyte (PBL) DNA might be a potential surrogate biomarker for cancer risk assessment. In addition, epigenetic variation can arise as a consequence of environmental, dietary, and aging factors [25][26][27][28] . Different tissues may exhibit different responses to environmental factors, and methylation status in leukocytes may not fully reflect the changes in the target tissue 29 . Methylation-environment interactions may provide further explanations for the complexity of cancer development.
Genome-wide differing methylated regions have been detected by comparing breast tumour tissue DNA and adjacent normal tissue DNA using next-generation sequencing techniques 30 . Several promising methylated biomarkers have been identified from circulating cell-free DNA between breast cancer cases and controls [31][32][33][34] . Specific methylation patterns were proposed to correlate with distinct clinicopathological characteristics 35,36 and assist in identifying individuals who will respond to therapy and survive longer. Hence, with suitable assays and validation in large populations, such associations can be exploited in non-invasive diagnosis and personalized treatment decisions.
Given the lack of research on DUSP1 methylation in breast cancer in epidemiological studies, we first investigated the association between DUSP1 methylation in PBL DNA, interactions with environmental factors, and breast cancer risk. We also explored the correlation between clinicopathological characteristics and DUSP1 methylation in both tumour DNA and PBL DNA, as well as the effect of environmental factors on DUSP1 methylation in tumour tissue DNA.

Results
Association between DUSP1 methylation in PBL DNA and breast cancer risk. PBL DNA was extracted from 423 patients and 509 controls. Supplemental Table 1 shows the distribution of demographic characteristics in cases and controls. No significant difference was found for age (P = 0.276) and BMI (P = 0.154). However, there were significant differences for the distribution of marital status (P = 0.023), educational level (P = 0.002), occupation (P = 0.001), and family history of cancer (P = 0.000) between cases and controls. Hence, these four variables were adjusted in the subsequent multivariate analyses.
DUSP1 methylation was detected in 5.2% (22/423) breast cancer cases and 4.9% (25/509) controls in PBL DNA (Table 1). After adjusting for marital status, educational level, occupation, and family history of cancer, no significant difference in DUSP1 methylation was observed between cases and controls. Therefore, we cannot conclude any association between DUSP1 methylation in PBL DNA and breast cancer risk (OR 0.79, 95% CI 0.414-1.504, P = 0.472). Table 3 shows the univariate and multivariate logistic regression analyses for all associations between environmental factors and breast cancer risk. Several environmental factors, including the consumption of refined grains, vegetables, fruit, seafood, milk, smoked food, etc., were found to be associated with the development of breast cancer following adjustment for educational level, occupation, marital status, and family history of cancer. We analysed the interactions of DUSP1 methylation with all of the above significant environmental factors. However, no significant interaction was observed (as shown in Table 1). Therefore, we concluded that there was not enough evidence for DUSP1 methylation in PBL DNA as a biomarker for breast cancer risk assessment.

Association of DUSP1 methylation in PBL DNA and environmental factors on breast cancer risk. Supplemental
Differences in DUSP1 methylation frequency between tumour DNA and PBL DNA in breast cancer patients. Genomic DNA from 326 breast tumour tissue samples was detected for DUSP1 methylation: the positive frequency of DUSP1 methylation was 59.2% (193/326). We successfully detected DUSP1 methylation in both PBL DNA and tumour DNA from 155 breast cancer patients. As shown in Table 2, a total of 83 tumour DNA were methylated in 155 tumour samples with a methylation frequency of 53.55%; in contrast, only five (3.23%) PBL DNA was methylated among the same patients. The P-value (0.000) from the McNemar Test indicates that there was a significant difference for the DUSP1 methylation frequency between the DNA samples from these two tissue types.
Effect of exposure to environmental factors on DUSP1 methylation in tumour tissue DNA. Supplemental Table 2 shows the distribution of demographic characteristics between patients with methylated and unmethylated tumour DNA. No statistically significant difference was observed for any demographic characteristic (all P > 0.05).  Table 1. Association of DUSP1 methylation in peripheral blood leukocyte DNA and environmental factors on breast cancer risk. OR crude , odds ratio generated by univariate logistic regression; OR adj , odds ratio generated by multivariate logistic regression; OR i , odds ratio generated by multivariate logistic regression for the interaction of DUSP1 methylation and environmental factors; 95%CI, 95% confidence interval.
We analysed the associations between environmental factors and DUSP1 methylation in tumour DNA in a multivariate analysis. Among all the environmental factors, there was no significant effect on DUSP1 methylation status from menopause, breast massage, breast hyperplasia, breast disease, smoking, alcohol etc. (Supplemental Table 4). However, the consumption of fruit and soybean, and irregular menstruation significantly correlated with the DUSP1 methylation status of tumour DNA in both univariate and multivariate analyses ( Table 4). Individuals with a lower fruit intake (≤ 1000 g/week) had a higher proportion (67.9%) of methylated DUSP1. Individuals with a higher soybean intake (>1 times/week) and irregular menstruation had higher proportions of methylated DUSP1 with OR of 1.955 (P = 0.006) and 2.000 (P = 0.020), respectively.
Since soybean intake and irregular menstruation are two hormone-related factors, the associations between soybean intake, menstrual cycle and DUSP1 methylation were further examined by dividing the subjects according to ER and PR status. As show in Table 5, we found significant data for soybean intake and irregular menstruation amongst the subgroups. Soybean intake (>1 times/week) was significantly correlated with increased DUSP1 methylation only in patients with ER-negative (OR 2.978, 95% CI 1.245-7.124) and PR-negative (OR 2.735, 95% CI 1.315-5.692) breast cancer. Meanwhile, irregular menstruation was significantly linked with increased DUSP1 methylation only in patients with ER-positive (OR 3.564, 95% CI 1.691-7.511) and PR-positive (OR 3.902, 95% CI 1.656-9.194) breast cancer.

Discussion
Several studies have suggested that individual variation in the epigenome in blood is associated with aging and environmental factors encountered throughout life 37,38 , with consequent risk of breast 39 , ovarian 40 , bladder 20 , and small-cell lung cancer 41 . The potential utility of specific methylation biomarkers in PBL DNA as novel markers of cancer susceptibility has been proposed. In this study, we first explored the value of DUSP1 methylation in PBL DNA for the risk assessment of breast cancer, but we failed to find any association between DUSP1 methylation in PBL DNA with breast cancer risk, or for the interactive effects of DUSP1 methylation and environmental factors. However, we did find significant correlations of triple-negative status with DUSP1 methylation in both tumour DNA and PBL DNA, and significant associations among soybean intake, irregular menstruation, ER/PR status, and DUSP1 methylation in tumour DNA.
Increasing numbers of studies have identified tissue-specific differential methylation 42 , which will provide important novel insights into normal and pathogenic mechanisms, as well as help in identifying markers of carcinogenesis and future epigenetic therapies. In this study, we analysed differences in DUSP1 promoter methylation between PBL DNA and tumour DNA. A significantly higher DUSP1 methylation frequency was observed in tumour DNA than in PBL DNA. In the research of Chen et al., they identified a normal breast cell line (M10) that was completely unmethylated while several breast cancer cell lines (MCF7, MDA-MB-231, SKBR3, and BT474) exhibited 100% methylation; unmethylation was dominant (86.2%) in benign breast tumours whereas methylation was dominant (57.2%) in invasive breast tumours 14 . Together, we concluded that DUSP1 methylation is a cancer-associated hypermethylation event. The biological significance of DUSP1 hypermethylation in breast cancer should be addressed in future in vitro studies. Given the very low (3.23-5.2%) methylation frequency in PBL DNA, DUSP1 methylation has potential as a biomarker in non-invasive breast cancer diagnosis if we can detect DUSP1 methylation from circulating cell-free DNA in plasma that is released by breast tumour cells.
Some genes that exhibit special methylation status in tumours are correlated with ER/PR status. ER-positive tumours were found to be more frequently methylated on RASSF1A than ER-negative tumours 43 . PR had a positive correlation with DUSP1 expression in 30 human breast cancer cell lines by binding to two progesterone response elements downstream of the DUSP1 transcriptional start site to upregulate DUSP1 promoter activity 44 . A critical novel finding of this study was the linkage of ER/PR-negativity with methylated DUSP1 in both breast tumour DNA and PBL DNA, which might account for the lower MKP-1 expression.
It is well known that negative results for oestrogen receptor (ER-), progesterone receptor (PR-), and HER2 (HER2-) expression in breast cancer cells signifies that the cancer is triple-negative cancer. These negative results indicate that the growth of the cancer is not supported by the oestrogen and progesterone hormones, or by the presence of too many HER2 receptors. Therefore, triple-negative breast cancer does not respond to hormonal therapy (such as tamoxifen or aromatase inhibitors) or therapies that target HER2 receptors. In contrast, ER-positive and PR-positive tumours are associated with improved response to hormonal therapy and with a longer disease-free interval and improved survival 45 48 . Epidemiological experimental evidence implicated that increased concentrations of endogenous oestrogen level or exogenous oestrogen intake may induce aberrant DNA methylation 49 . In this study, we found high intake of fruit was correlated with decreased DUSP1 methylation, while high intake of soybean and irregular menstruation were correlated with increased DUSP1 methylation.
Soybean is a unique food because it contains large amounts of isoflavones 50,51 . Isoflavones have a chemical structure that is very similar to the hormone oestrogen 52 . To the best of knowledge, breast cancer is a heterogeneous disease and biological differences in subtypes depend on the expression of receptors, including ER, PR, and HER2. Because of the ability of isoflavones to bind oestrogen receptors, the varied associations between soybean intake and breast cancer risk by the hormone receptor status of tumours have been suggested in eight published epidemiological studies [53][54][55][56][57][58][59][60] . However, few studies have researched the modified effect of soybean intake on DNA methylation, which may have pivotal functions in relation to tumour suppression, apoptosis, etc. in breast cancer. Harlid et al. used an Illumina Human Methylation450 BeadChip to evaluate epigenome-wide DNA methylation in vaginal cells from soy formula-fed and cow formula-fed girls; the results indicated that girls fed soy formula had altered DNA methylation in their vaginal cell DNA 61 . Only one study has been published that provides evidence on the potential effects of two naturally occurring isoflavones, genistein and daidzein, on the    62 . Our study is the first to explore the putative effects of soybean consumption on DUSP1 promotor methylation in breast cancer. We found that soybean intake was significantly associated with methylated DUSP1 in tumour DNA in ER/PR-negative patients. Messina and colleagues reviewed substantial epidemiological data from observational studies and cell culture data, and noted that the current limited knowledge regarding the effect of soybean on breast cancer-related issues suggested that clinicians should be careful of what they prescribe for patients 63,64 . Although information regarding soybean consumption was provided as retrospective data for the patients' dietary habit prior to cancer diagnosis, given the significant correlation between soybean intake and DUSP1 methylation we observed, prolonged or excessive consumption of soybean in ER/PR-negative patients is not recommended. In addition, progesterone balances oestrogen and in doing so minimises the negative effects of oestrogen-progesterone imbalances 65 . When these hormones become unbalanced it is usually because of oestrogen dominance, which means too much oestrogen compared with the levels of progesterone. Irregular menstruation flow is an oestrogen-dominant symptom. In the subgroup analyses in our study, irregular menstruation was significantly correlated with increased DUSP1 methylation in ER/PR-positive patients, which suggested that irregular menstruation may correlate with DUSP1 methylation through the indirect effect of oestrogen. Further in vitro studies are needed to validate this inference.
There are some limitations with the interpretation of the present results. First, some studies have suggested cell-specific variation in DNA methylation 66,67 ; in our case-control study design, we only focused on the differential methylation of DUSP1 in leukocyte DNA between cases and controls, and we did not further explore the cell type composition difference of PBL between the cases and controls. Second, our conclusion for the association between ER/PR status and DUSP1 methylation in breast tumour DNA was generated based on a population study without an experimental validation in vitro study. Third, the small number of subjects in the stratified analysis limited the statistical power to evidence the conclusion; further studies with larger sample sizes are encouraged to verify the relationship between environmental factors, ER/PR status, and DUSP1 methylation.

Conclusions
The results of this study indicated that DUSP1 methylation in PBL DNA and its interaction with environmental factors was not associated with breast cancer risk. This is the first study to report the modified effects of soybean consumption on DNA methylation in tumour by ER/PR status, and we provide preliminary evidence on potential epigenetic changes through DUSP1 methylation in triple-negative patients. Further validation of the association of environmental factors, including fruit and soybean intake, and irregular menstruation, with DUSP1 methylation by hormone receptor status in breast cancer should be undertaken.

Materials and Methods
Study subjects. We carried out this study after obtaining informed written consent from study subjects and approval from the Human Research and Ethics Committee of Harbin Medical University. All experiments including all relevant details were performed in accordance with relevant guidelines and regulations.
A case-control study was designed to assess the role of DUSP1 methylation and interactions with environmental factors on breast cancer risk. All breast cancer patients were newly diagnosed cases recruited from the Third Affiliated Clinical Hospital of Harbin Medical University from 2010 to 2014. Controls were recruited from patients admitted to the Orthopaedic and Ophthalmology of the Second Affiliated Hospital of Harbin Medical University, and volunteers from the Xiangfang community of Harbin City within the same time period. Any individual with a history of benign breast disease or any other cancer was excluded from the control group. Approximately 5 ml of peripheral venous blood was obtained from all cases either before surgery for the patients and at enrolment for the controls.
A case-only study was designed to explore the difference in DUSP1 methylation between breast tumour DNA and PBL DNA. Tumour tissues specimen were collected during surgery and rapidly frozen in liquid nitrogen after removal, then returned to the lab and stored at −80 °C immediately. We analysed the correlation between  Then, 0% M (universal unmethylated DNA standards) served as the cut-off value to distinguish methylation and non-methylation of DUSP1. We analysed methylation as a qualitative variable, methylated (any methylated status with methylation level higher than 0%M) and unmethylated. We duplicated sample DNA, two blank controls, and gradient methylated DNA standards in each plate. Figure 1(a,b) showed the profile of fluorescence obtained at the melting temperature for serial dilutions of methylated DNA (100%, 20%, 10%, 5%, 1%, and 0%). Figure 1(c,d) showed the melting profiles of a methylated breast tumour DNA and a unmethylated sample.  Immunohistochemical assay. The presence of oestrogen receptor (ER), progesterone receptor (PR), and HER2 in breast tumour tissue was tested by immunohistochemical (IHC) assay; further verification using fluorescence in situ hybridization (FISH) was needed if the results of IHC assays showed HER2-positivity.

Statistical analysis.
Categorical and continuous variables were tested by chi-square test and two-sample t-test, respectively. Univariate and multivariate logistic-regression analyses were used to calculate the crude and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) for the association of environmental factors, DUSP1 methylation in PBL DNA, and their association with breast cancer risk. Correlation between clinicopathological characteristics and DUSP1 methylation status in tumour DNA and PBL DNA was evaluated using odds ratios (ORs) and 95% CIs derived from unconditional logistic regression. The effect of environment factors on DUSP1 methylation in tumour DNA was calculated using unconditional univariate and multivariate logistic regression.
All statistical analyses were performed using SAS version 9.2, with P-values of < 0.05 considered statistically significant.