The Epigenetic Landscape of Promoter Genome-wide Analysis in Breast Cancer

Breast cancer is a heterogeneous disease due to its clinico-pathological features and response to therapy. The classification of breast tumors based on their hormone receptor status and pathologic features. Post-translational histone modifications come into prominence for regulation of gene expression in cancer pathogenesis. Here, we analyzed dysregulation of H3K9ac and H3K27me3-enriched subtype-specific genes using ChIP-on-chip assay in breast cancer tumors and matched normal tissue samples. Breast cancer tumors were classified according to St Gallen Consensus 2013. Our results indicated that the promoter regions of genes modified by H3K9ac epi-mark are commonly associated with tumors with HER2-positive and TNBC subtype. H3K27me3-enriched genes were comprised of Luminal A and B1 subtypes. We constructed a network structure to elicit epigenetically regulated genes related with breast cancer progression. The central genes of the network (RUNX1, PAX3, GATA4 and DLX5) were subjected for epigenetically dysregulation in association with different breast cancer subtypes. Our study submits epigenetic mechanisms are crucial to elicit subtype-specific regulation in breast cancer and ChIP-on-chip assay provides a better understanding for breast tumorigenesis and new approaches for prevention and treatment.

A large number of studies have focused on genetic basis of breast cancer, whereas recently knowledge about the impact of epigenetic mechanisms on breast cancer development and prognosis has been increasing 11 . Post-translational histone modifications are involved in regulation of gene expression in cancer pathogenesis. Furthermore, diversity of breast cancer subtypes are linked to dysregulation of gene expression associated with histone modifications 12 . To our best knowledge, investigations on modifications in breast cancer tumors are very limited.
Nowadays, chromatin immunoprecipitation (ChIP) is coupled with promoter DNA microarrays to evaluate the mechanisms of human gene regulation on a genome-wide scale. ChIP-on-chip technology could be used to investigate the alterations of global gene expression in tumorigenesis. In this study, we aimed to elicit differently regulated genes, associated with modified histone 3 lysine 27 trimethylation (H3K27me3) and histone 3 Lysine 9 acetylation (H3K9ac) in breast cancer tumors by ChIP-on-chip method. For this purpose, Agilent SurePrint G3 400k Human Promoter Microarrays were used to scan gene promoters in 15 breast tumors with their matched normal tissue samples. Breast tumor samples were classified according to St Gallen Consensus 2013 to identify the impact of epigenetic patterns on the diversity of breast cancer subtypes.

Results
Classification of Breast Cancer Subtypes. Breast cancer tumor samples have been subdivided into five molecular subtypes according to St Gallen International Expert Consensus 2013: luminal A (ER and PGRpositive, human Ki-67 protein <14%, histological grade 1 or 2 and HER2-negative), luminal B1, HER2-negative (ER-positive, PGR <10% or negative, Ki-67 > 14% and high grade), luminal B2 HER2-positive (ER-positive and PGR-negative), HER2-positive non-luminal (ER and PGR-negative) and TNBC (ER, PGR and HER2-negative). St Gallen classification is based on clinic-pathological factors of early invasive breast cancer. Luminal disease is responsive to endocrine therapy and usually has a more favorable prognosis. TNBC and HER2-positive diseases have poor clinical outcome. On the other hand, Luminal A subtype shows better prognosis with more endocrine sensitivity and differs from Luminal B tumors, which are more aggressive with less endocrine sensitivity.
Chromatin Immunoprecipitation. ChIP-on-chip technology could be used to identify the global levels of epigenetically dysregulated genes in tumorigenesis. Thereby, we investigated differentially regulated genes associated with modified H3K9ac and H3K27me3 in breast cancer subtypes. According to their effects on gene regulation, H3K9ac is generally associated with transcriptional activity, while methylation of H3K27me3 is associated with repression 13,14 . The performance of ChIP experiments was confirmed by using the genes TSH2B as a methylated positive control and C-FOS as an acetylated positive control. High levels of methylation for TSH2B and acetylation for C-FOS exhibited the efficiency of immunoprecipitation.
Epigenetic Dysregulation of Breast Cancer Subtype-Specific Genes. We determined contribution of differentially regulated gene promoters in relation with breast cancer subtypes using Agilent SurePrint G3 Human Promoter microarrays. Venn diagrams were constructed to elucidate the unique and overlapping gene promoters, which were found epigenetically dysregulated between each subtypes (Fig. 1). The aberrant gene regulation associated with H3K9ac modification predominantly was observed in basal-like subtypes; TNBC and HER2-positive tumors. Especially, 1016 modified gene promoters were specific for TNBC population and 479 were unique for HER2-positive tumor samples. H3K9ac epi-mark was down-regulated on gene promoters in HER2-negative tumors (70.1, 76.4 and 58.1% in TNBC, Luminal A and B1 subtype, respectively), beside the percentage of up-regulation found high in HER2-positive tumors (78.5 and 55.2% in HER2-positive, Luminal B2 subtype, respectively, Fig. 2). The level of H3K27me3 was down-regulated on large proportion of promoters in TNBC tumors (78.2%). The enrichment of H3K27me3 epi-mark found higher in tumors with Luminal B1 and HER2-positive subtypes (60.6 and 57.8% in Luminal B1, HER2-positive subtypes, respectively, Fig. 2). According to our results, epigenetic regulations displayed subtype-specific profiles. Notably, the aberrant regulations of genes by H3K9ac epi-mark were common in non-luminal and basal subtype, HER2-positive and TNBC, while modifications in luminal disease, particularly Luminal A and B1, were dysregulated by H3K27me3 epi-mark.
Breast Cancer Related Network Analysis. In this study, an agglomerative method was used: each observation started in its own cluster, and pairs of clusters were merged as one moves up the hierarchy. Two elements were merged in a cluster when their distances were the smallest. The goal of this analysis was to have homogeneous clusters. We identified the prominent dysregulated genes in different breast cancer subtypes using hierarchical clustering data (Fig. 3). The 79 differently regulated genes for H3K9ac epi-mark and 37 genes for H3K27me3 epi-mark were subjected for hierarchical clustering analyses. According to the results of our analyses, the network structure of the genes involved in signaling pathways was generated using Cytoscape program (Fig. 4). Gene Ontology Analysis. We performed Gene Ontology analysis on subtype-specific genes to investigate associated biological process (Table 1). We discovered that H3K9ac modification mainly regulated the biological process of cell proliferation, cell migration, regulation of cell differentiation, negative regulation of programmed Scientific REPoRTS | 7: 6597 | DOI:10.1038/s41598-017-06790-z cell death and cell-cell signaling. Otherwise, genes regulated by H3K27me3 modification were enriched mostly in cell-cycle regulation, such as negative regulation of cell-cycle, negative regulation of cell-cycle phase transition and cell-cycle arrest. In addition, negative regulation of apoptotic process and negative regulation of apoptotic signaling pathway were remarkable biological process for genes regulated by H3K27me3 epi-mark. The commonly involved biological process revealed that histone 3 modification patterns showed impact on dysregulation of breast cancer-related genes.

Discussion
In this study, we propounded that distinct breast cancer subtypes show unique epigenetic patterns and epigenetic marks H3K9ac and H3K27me3 figure the post-transcriptional regulation of tumorigenesis processes in breast cancer. We classified 15 breast cancer tumors according to St Gallen Consensus 2013 6 . The epigenetic profiles of H3K9ac and H3K27me3-enriched genes were investigated using ChIP assay coupled with promoter microassay analysis. In particular, the promoter regions of genes modified by H3K9ac epi-mark appeared in HER2-positive and TNBC tumors. The H3K9ac modification induced the down-regulation of the majority of related genes in HER2-amplified tumors. Moreover, H3K9ac-enriched genes were commonly down-regulated in TNBC and Luminal A subtypes. H3K27me3 mark enriched in Luminal A and B1 subtypes and involved in dysregulation of subtype-specific gene expression. The impact of histone modifications on gene regulation patterns has been reported in different breast cancer subtypes [19][20][21] . However, due to difficulty of working with patient tissue sample, the majority of investigations have been carried out in breast cancer cell lines 21,22 . Results of these studies asserted that epigenetic modifications are mainly subtype-specific and dysregulation of targeted gene expression profiles result in induction the development of different types of tumor. The high levels of H3K27me3 mark at the promoters of ER, PR and ERBB2 genes have been demonstrated in MDA-MB-436 cell lines baring basal-like features 19 . H3K27me3 enrichment was correlated with Luminal A subtype and good prognosis. Likewise, in our study the majority of genes enriched by H3K27me3 epi-mark was included in Luminal A and B1 subtypes. However, Healey et al. 23 reported that global levels of H3K27me3 has been associated with the pathology of Luminal A tumors, no association found with Luminal B subtype, we indicated that H3K27me3 mark was involved in epigenetic regulation of breast tumors with Luminal B1 subtype. Moreover, Holm et al. 20 identified the expression profiles of H3K27me3 in distinct subtypes and the levels of H3K27me3 mark were found diminished in Luminal B, HER2-positive and TNBC tumors, while levels were high in Luminal A tumors. In addition, low abundance of H3K27me3 was propounded as predictor for poor survival. There is limited number of investigations focused on the role of H3K9ac modification on subtype-specific gene expression patterns in breast cancer 22,24 . Elsheikh et al. 24 have demonstrated the correlation between high levels of H3K9ac mark with better disease-free survival and metastatic-specific survival, and breast cancer-specific survival.
The RUNX family genes are transcription factors and take part in hematopoiesis, osteogenesis and neurogenesis 25 . RUNX1 involves in tumorigenesis as a key regulatory factor, particularly in various epithelial cancers 26,27 . According to our results, H3K27me3 enrichment was up-regulated on RUNX1 gene in HER2-positive and Luminal A subtype, however down-regulation was determined in tumors with Luminal B1 and B2 subtype. A previous study indicated that RUNX1 conducted as oncogene in TNBC and the expression levels of RUNX1 was correlated with the poorest prognosis 28 . It is propounded that the effect of RUNX1 on tumor progression was conducted with its ability to diminish ER signaling. We did not found any correlation between RUNX1 regulation by H3K27me3 mark and TNBC subtype.
PAX3 gene is acting as an oncogene in breast tumorigenesis and the expression profile of PAX3 gene is regulated by epigenetic mechanisms. Recently, Zhao et al. 16 asserted that reduced level of H3K9me2 epi-mark in primary breast epithelial cells was involved in regulation of PAX3 expression. In our study, we demonstrated that H3K9ac enrichment increased on PAX3 oncogene in HER2-positive tumors. Hence, non-luminal HER2-positive tumors had relatively low five-year related survival and poorer prognoses than Luminal A, the epigenetic dysregulation of PAX3 by H3K9ac come forward as a pathological prognostic factor for HER2-positive tumors. The differently expression of transcription factor GATA4 has been reported in ERBB2-amplified breast tumors 29 . The role of GATA4 in epigenetic mechanisms is not clarified, however it has been shown that GATA4 possessed ERBB2 gene expression through transcriptional repression. The epigenetic regulation of GATA4 gene by H3K9ac modification has been shown in Luminal B and HER2-amplified tumors. On the other hand, we could not contribute to the H3K9ac levels with non-luminal HER2-positive tumors.
DLX5 gene is mainly expressed in developing bones and controls osteoblastogenesis 30 . In addition, the overexpression has been observed in more aggressive tumors and increased risk of metastases in bone or lung 18 . The H3K9ac-enrichment on DLX5 gene was identified in Luminal B2 subtype in this study. The epigenetic dysregulation of DLX5 gene could be suggested as an indicator for aggressiveness of breast tumors.
According to GO term analysis on subtype-specific genes, H3K9ac modification was commonly involved in biological process of cell proliferation, cell migration, regulation of cell differentiation, negative regulation of programmed cell death and cell-cell signaling. On the other hand, the genes differently regulated by H3K27me3 mark were associated with biological process of negative regulation of cell-cycle, negative regulation of cell-cycle phase transition and cell-cycle arrest. Moreover, negative regulation of apoptotic process and negative regulation of apoptotic signaling pathway were remarkable biological processes for genes regulated by H3K27me3 epi-mark. It is not surprising that all these biological processes involve in the hallmark of cancer. However, shared biological processes vary across histone marks.
In this study, epigenetic modifications H3K9ac and H3K27me3 were analyzed to identify the role of epigenetic mechanisms on diversity of subtype-specific gene regulation in breast cancer. Tumor size, histological subtype and grade, lymph node status and expression of ER, PGR and HER2 are routinely used for classification of breast cancer tumor. However, these parameters are restrictive to predict individual survival and response to therapy. Our results may provide knowledge about subtype-specific epigenetic regulations in breast cancer. Mainly, expression profiles demonstrate diversity in HER2 positive and basal-like breast tumors. Working with tumor samples restricted our study because of difficulty of sample enrichment, cell heterogeneity and individual    their clinical profiles and hormone receptor status 6 . All experiments and methods were performed in accordance with relevant guidelines and regulations.
Chromatin Immunoprecipitation (ChIP). DNA extraction and shearing procedures were performed as indicated previously 32,33 . ChIP was carried out with SX-8G IP-Star ® Compact Automated System (Diagenode).
Microarray Hybridization and ChIP-on-chip. Human Promoter Microarrays were provided by Agilent Technologies to determine the histone methylation or acetylation of ChIP-enriched samples (SurePrint G3 400k Human Promoter microarrays). The array contains ~21,000 of the best-defined human gene regions. Hybridization and microarray procedures were performed in accordance with the manufacturer's instructions. Each microarray analysis was duplicated for one tumor and matched normal tissue samples. A High Resolution Microarray Scanner (Agilent Technologies) was utilized for scanning fluorescent intensities.

Microarray Data Analysis.
ChIP-on-chip data analyses were carried out using the package R/Bioconductor Ringo software 1.26.1. The selection criteria for the enriched regions or genes were: the region require a certain 250 bp probe spacing; contain at least 3 probes/region and smoothed intensities of probes mapped to this region need to exceed a threshold defined at 1.5.
Network Analysis. The network associated with genes regulated by epigenetic modifications among different subtypes was constructed using Cytoscape 14 (version 3.2.0). Particularly, cancer-related target genes and nearest-neighbor genes were considered to build the network structure.
Functional Annotation. Gene Ontology biological process terms analysis in breast cancer subtype-specific histone modification associated genes were determined using DAVID Functional Annotation Tool 34 .