Background
In most Western countries breast cancer is a leading cause of cancer related deaths in women. Breast cancer accounts for over one million of the estimated 10 million cancers that are diagnosed globally each year in both males and females and claimed about 375,000 deaths in the year 2000 [
1]. The use of screening mammography and MRI has led to a decline in breast cancer-related mortality. But these screening procedures can also lead to overdiagnosis and false positive cases leading to unnecessary treatment. Therefore there is a critical need to develop molecular biomarkers that can detect early stage disease so that treatment can begin promptly.
Genetic as well as epigenetic changes play crucial roles in tumour development and progression. Alterations in DNA methylation include both genome-wide hypomethylation and hypermethylation events. Promoter hypermethylation at specific gene loci is a major mechanism of epigenetic inactivation in cancer cells leading to gene silencing. A recent genome-wide study in breast and colon cancer led to identification of tumour suppressor genes that are inactivated by both genetic and epigenetic events and importantly reduced expression of a subset of these genes correlated with poor clinical outcome [
2]. Although the list of hypermethylated genes in breast cancer is growing, only a few show promise as biomarkers for early detection and risk assessment [
3,
4]. We and others identified
RASSF1A as one of the most highly methylated genes in many cancer types including breast cancer [
5,
6]. Identification of more genes of this type would facilitate the development of sensitive molecular markers across multiple malignancies.
In the present study we employed a genome-wide approach to identify methylated genes in breast cancer and further investigated these genes in other common epithelial cancers. For the high throughput genome-wide DNA methylation approach we utilized a very sensitive assay recently developed in our laboratory (Gerd Pfeifer lab). The methylated-CpG island recovery assay (MIRA) is based on the high affinity of the MBD2/MBD3L1 complex for methylated DNA and allows one to detect cell type-dependent differences in DNA methylation when used in combination with CpG island arrays [
7,
8].
Discussion
Using a high throughput genome-wide approach we have identified 5 genes showing frequent tumour acquired methylation in breast cancer. To the best of our knowledge methylation of
EMILIN2,
SALL1,
DBC1,
FBLN2, or
CIDE-A has not been reported previously in breast cancer. Furthermore we demonstrate that the methylation is cancer specific and leads to gene silencing. In addition we found that
EMILIN2,
FBLN2 and
SALL1 were also frequently methylated in other common epithelial cancers,
EMILIN2 and
SALL1 in colorectal cancer and
FBLN2 in prostate cancer. To the best of our knowledge
EMILIN2 and
CIDE-A methylation in primary tumours has not been previously reported,
SALL1 and
FBLN2 methylation has been previously reported in acute lymphocytic leukemia [
11,
12] and
DBC1 methylation has been reported in NSCLC, bladder cancer, oral cancer and in leukaemia [
9,
13‐
15].
EMILIN2 belongs to a family of extracellular matrix (ECM) glycoproteins containing an EMI domain at the N-terminus and a gC1q domain at the C-terminus. EMILIN2 has been shown to suppress growth of cancer cells [
16] and triggers apoptosis in cancer cells via the extrinsic apoptosis pathway following EMILIN2 binding to the trail receptors DR4 and DR5.
CIDE-A belongs to a novel family of cell death inducing DFF45 (DNA fragmentation factor-45) like effector proteins. In humans there are three family members (CIDE-A, CIDE-B, CIDE-C). In cell lines overexpression of CIDE-A leads to caspase-independent cell death associated with DNA fragmentation [
17]. In addition, mice lacking functional Cidea have higher metablic rates and lipolysis in the brown adipose tissue as well as being resistant to diet-induced obesity and diabetes as compared to their wildtype littermates [
18]. A recent report indicates that CpG methylation plays a critical role in determining tissue and cell specific expression of the CIDE-A gene [
10].
FBLN2 encodes for an extracelluar matrix protein, belonging to a 6 member family of fibulins. FBLN2 binds various extracellular ligands and calcium. It acts as a scaffold protein in the ECM by binding to a variety of ligands including type IV collagen, fibronectin, fibrinogen, fibrillin, laminins, aggrecan and versican. It was recently shown to be downregulated in breast cancer cell lines and primary breast tumours. Reintroduction of FBLN2 into breast cancer cell lines lacking FBLN2 reduced cell motility and invasion in vitro but had no effect on cell growth and adhesion. Whilst, loss of FBLN2 expression increased cell migration and invasion [
19]. Recently another member of the Fibulin gene family,
FBLN3 (
EFEMP1) was shown to be frequently methylated in breast tumours [
20].
DBC1 (deleted in bladder cancer 1) was identified from a common region of loss of heterozygosity at 9q32-q33 in bladder cancer. DBC1 has been shown to be frequently methylated in bladder cancer, NSCLC, oral cancer and acute lymphocytic leukemia [
9,
13‐
15]. Exogenous expression of DBC1 in NSCLC and in bladder cancer cell lines lacking DBC1 expression inhibited cell growth [
9,
21]. It has also been demonstrated that DBC1 plays a role in cell cycle control [
21].
SALL1 is a human homologue of the Drosophila region-specific homeotic gene spalt (sal). The SALL1 gene product is a zinc finger protein thought to act as a transcription factor and maybe part of the NuRD histone deacetylase complex. Mice lacking Sall1 die in the perinatal period from kidney agenesis [
22]. Townes-Brocks syndrome affecting limb, ear, kidney and heart development is caused by defects in the SALL1 gene [
23]. SALL1 was recently demonstrated to be hypermethylated in acute lymphocytic leukemia [
11].
It has been demonstrated that methylated genes often reside in regions of chromosomal loss [
2].
SALL1 resides at 16q12.1- a region demonstrated to undergo loss of heterozygosity (LOH) in breast, prostate, ovarian cancer and in retinoblastoma [
24,
25].
EMILIN2 is located at 18p11.3 in a region showing LOH in NSCLC, CRC and breast cancer [
26].
FBLN2 is located at 3p25.1 in a region well documented for allelic loss in renal cell carcinoma and breast cancer [
27,
28].
CIDE-A is located at 18p11.2 in a region showing allelic loss in esophageal squamous cell carcinoma [
29] and DBC1 was identified from a region of frequent loss of hetrozygosity at 9q32-33 in bladder cancer [
30].
Breast cancer is a hormone dependent cancer. In patients with breast cancer estrogen receptor (ER) status is an important treatment and prognostic factor. The response of breast cancer patients to endocrine therapy is guided by the expression of estrogen and/or progesterone receptors. Breast cancers that are positive for estrogen receptor and/or progesterone receptor respond better to endocrine therapy compared to receptor negative breast cancers. Our findings indicate that methylation of EMILIN2, CIDE-A and FBLN2 are associated with positive estrogen and or progesterone receptor status, this may help in further refining breast cancers that are more likely to benefit from endocrine therapy.
In our cohort of breast cancer cases, EMILIN2 methylation also correlated with lymph node metastases, relapse and poor survival, hence EMILIN2 methylation is associated with less favourable prognosis. It will be of interest to analyse larger sample numbers to determine if EMILIN2 methylation status can be utilised as a prognostic marker in breast cancer.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
VKH did the experiments involving the breast cancer data (bisulphite modification, COBRA, sequencing, expression and bioinformatics analysis) and analysis of lung, colorectal and prostate cancer cell lines. LBH and RLW carried out analysis in colorectal tumours. T Dansranjavin and RD carried out analysis in prostate tumours. AD did bioinformatics analysis on the MIRA data. IB, DE and CL provided breast tumour DNA samples and NHMEC. IB and SV carried out real-time RT-PCR. ST and GPP performed the MIRA assay and CpG island array hybridisation. T Dobbins did statistical analysis. DG did tissue culture. JM provided breast tumour cell lines and N/T paired lung DNA samples. ERM contributed to the concept of the study and statistical analysis. FL conceived the studies, oversaw the experimental work, wrote the manuscript and established all the collaborations. All authors have read and approved the final manuscript.