AP 2 regulates estrogen receptor mediated long range chromatin interactions and gene transcription

1 1.1 Physiological importance of estrogen 1 1.2 ER structure and mechanism of action 2 1.3 General co-regulators of ER-mediated transcription 8 1.4 Significance of ER in breast cancer 1

AP-2γ REGULATES ESTROGEN MEDIATED LONG-RANGE CHROMATIN

RECEPTOR-INTERACTIONS AND GENE TRANSCRIPTION

TAN SI KEE

B.Sc (Hons.), NTU

A THESIS SUBMITTED

FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

NUS GRADUATE SCHOOL FOR INTEGRATIVE

SCIENCES AND ENGINEERING NATIONAL UNIVERSITY OF SINGAPORE

2012

DECLARATION

ACKNOWLEDGEMENTS

First and foremost, I am thankful to my Ph.D supervisors, Dr Yong Eu Leong and

Dr Edwin Cheung, for accepting me as their student in the pursuit of my Ph.D study I owe my deepest gratitude to Dr Edwin Cheung who has supported me throughout the past four years I greatly appreciate his guidance and mentorship,

as well as the opportunities that he had provided me to work with great scientists within and outside of GIS, which have contributed tremendously to my learning experience in the scientific field

I am deeply indebted to my colleagues who have worked with me on this project Special thanks to Mr Chang Cheng Wei for his extensive efforts in guiding our group in the usage of computational programs so that we can be self-sufficient in performing basic analyses on whole genome dataset Thanks to Mr Lin Zhen Hua who had provided the data on reporter and interaction assays

I wish to acknowledge Dr Pan You Fu, Dr Liu Mei Hui, Dr Chuah Shin Chet,

Dr Tan Peck Yean and Mr Chng Kern Rei for their valuable advice and insights during the course of my study It is a great pleasure to have met my present and former colleagues in GIS who have always been encouraging and made my stay enjoyable My heartfelt thanks to the GTB sequencing group in GIS for their commitment in generating high quality sequencing data

This thesis would not have been possible without the sponsorship of my graduate study from NGS I also want to thank my TAC members and annual GIS graduate seminar committee for their scientific input and future directions

Finally, I am blessed with family members and boyfriend who have provided me with continuous support and understanding during my graduate study

TABLE OF CONTENTS

DECLARATION ii

ACKNOWLEDGEMENTS iii

TABLE OF CONTENTS v

SUMMARY viii

LIST OF TABLES x

LIST OF FIGURES xi

ABBREVIATIONS xii

PUBLICATION xvi

CHAPTER 1 INTRODUCTION 1

1.1 Physiological importance of estrogen 1 1.2 ER structure and mechanism of action 2 1.3 General co-regulators of ER-mediated transcription 8 1.4 Significance of ER in breast cancer 12 1.5 Therapeutic regimens for ERα-positive breast cancer 14 1.6 Comprehending ERα transcriptional network – before next-

generation sequencing era 15 1.7 Genome-wide panorama of ERα-mediated transcription – in the

next-generation sequencing era 17 1.8 Known collaborative factors of ERα 21 1.9 FoxA1, a pioneer factor of ERα in breast cancer cells 25 1.10 Activator protein-2 (AP-2) family in normal and cancer

developments 32 1.11 Roles of AP-2γ in breast cancer 36 1.12 Aims of the study 39 CHAPTER 2 MATERIALS AND METHODS 40

2.2 Cryogenic preservation and recovery of cells 40

2.5 Plasmid DNA extraction 42

2.15 Solexa sequencing and binding site determination 51

CHAPTER 3 RESULTS 54 3.1 Enrichment of AP-2 motifs at ChIA-PET ERBS 54 3.2 Identification of AP-2γ as a potential collaborative factor of ERα 59 3.3 AP-2γ is essential for efficient transcription of estrogen-regulated

3.4 Prediction of AP-2 motifs at ChIA-PET ERBS associated with AP-2γ-dependent estrogen up-regulated genes 66 3.5 Expression of RET, a direct target of ERα, is regulated by AP-2γ 70 3.6 AP-2γ directly binds to RET-associated ERBS in a ligand-

3.7 AP-2γ regulates ERα-mediated long-range chromatin interactions 78 3.8 Recruitment of ERα to ERBS is dependent on AP-2γ 83 3.9 Similarity in AP-2γ and FoxA1 binding profiles in the MCF-7

3.10 FoxA1 co-occupies with AP-2γ at the RET-associated ERBS 91 3.11 Preferential co-localization of AP-2γ and FoxA1 at ERBS across

3.12 Interdependence between AP-2γ and FoxA1 at ERBS 101 3.13 Preferential association of AP-2γ and FoxA1 with ERα-mediated

CHAPTER 4 DISCUSSIONS 109

4.1 Exploring potential determinants of ERα-mediated long range transcriptional regulation 109 4.2 AP-2γ is a transcriptional activator and repressor of estrogen-

regulated transcriptome 113 4.3 Combinatorial action of AP-2γ and FoxA1 in regulating the ERα cistrome 115 4.4 AP-2γ is a critical determinant of estrogen-mediated long-range chromatin interactions 117 CONCLUSION, FUTURE DIRECTIONS AND PERSPECTIVES 122

BIBLIOGRAPHY 125

APPENDIX I 153

APPENDIX II 158

APPENDIX III 159

APPENDIX IV 159

APPENDIX V 161

APPENDIX VI 162

SUMMARY

Estrogen receptor α (ERα) is a key player in the development and progression of mammary tumorigenesis and factors collaborating with ERα within its network of transcription are likely to contribute to the pathological outcome Recently, the cistrome and interactome of ERα were mapped in breast cancer cell line, MCF-7, revealing the importance of spatial organization in estrogen-mediated transcription To unravel the relationship between ERα and other collaborative factors underlying such regulatory process, our genome-wide analysis on ERα binding sites (ERBS) identified from the Chromatin Interaction Analysis-Paired End DiTag (ChIA-PET) revealed a significant enrichment of AP-2 motifs Members of AP-2 transcription factor family are important regulators of vertebrate embryogenesis, required for proper formation of critical organs and body structures Moreover, their roles in adult tissues have been associated with multiple cancer types AP-2γ, earlier identified as estrogen receptor factor-1, was particularly involved in driving proliferation and tumor development in breast However, whether it works cooperatively with ERα at genomic level still remains unknown

In our study, we demonstrated that AP-2γ regulates nearly half of the mediated transcriptome It is also recruited to ERBS associated with their co-regulated genes in a ligand-independent manner, which is indicative of its early binding event Furthermore, perturbation of AP-2γ expression suggests its

estrogen-importance in ERα recruitment to chromatin and long-range chromatin interactions in response to estrogen Globally, we observed convergence of a large number of AP-2γ and ERα binding events across the genome The majority of these shared regions are also co-occupied by the pioneer factor, FoxA1, which shares similar genomic behavior with AP-2γ Our molecular studies further imply there is functional interplay between AP-2γ and FoxA1 at ERBS where they co-localize Finally, ERBS involved in long-range chromatin interactions are preferentially occupied by AP-2γ and FoxA1 Collectively, our findings suggest that AP-2γ is a novel collaborative factor of ERα to define higher-order chromatin structure for transcriptional regulation

LIST OF TABLES

Table 1 Summary of ChIP-Seq analysis 88

LIST OF FIGURES

Figure 1.1 Schematic diagram of ER functional domains and their homology 6

Figure 1.2 Classical and non-classical models of ER-mediated transcription 7

Figure 1.3 Domain structure of FoxA proteins 30

Figure 1.4 Distribution of lineage-specific H3K4 methylation dictates

FoxA1 recruitment sites 31

Figure 1.5 Domain structure of AP-2 proteins 35

Figure 3.1 AP-2 motifs are enriched at ERBS identified from ChIA-PET 58

Figure 3.2 AP-2γ is identified as a potential collaborative factor of ERα 61

Figure 3.3 Efficient transcription estrogen target genes require AP-2γ 65

Figure 3.4 AP-2 motifs are predicted at ChIA-PET ERBS associated with GREB1 and RET genes 69

Figure 3.5 AP-2γ is required for the expression of RET, which is a primary estrogen-regulated gene 73

Figure 3.6 Ligand-independent recruitment of AP-2γ at RET-associated ERBS 76

Figure 3.7 AP-2γ at RET-associated ERBS is required for long-range chromatin interactions 82

Figure 3.8 AP-2γ is required for efficient ERα binding 84

Figure 3.9 Global analysis of AP-2γ and FoxA1 binding events 90

Figure 3.10 Co-localization of FoxA1 and AP-2γ at RET-associated ERBS 95

Figure 3.11 AP-2γ, FoxA1 and ERα are co-localized at a large fraction of ERBS 100

Figure 3.12 Mutually dependent recruitment of AP-2γ and FoxA1 at ERBS 106

Figure 3.13 AP-2γ and FoxA1 are preferentially associated with ERBS

involved in long-range chromatin interactions 108

Figure 4.1 Proposed model for AP2γ, FoxA1 and ERα cooperation in long- range chromatin interactions 121

CARM1 Co-activator-associated arginine methyltransferase

CDKN1A Cyclin-dependent kinase inhibitor 1A

ChIA-PET Chromatin Interaction Analysis-Paired End DiTag

DREME Discriminative Regular Expression Motif

DTT Dithiothreitol

EBAG9 ER-binding site-associated antigen 9

ERBB2 v-erb-b2 erythroblastic leukemia viral oncogene

homolog 2

FAIRE Formaldehyde assisted isolation of regulatory

elements

GAPDH Glyceraldehyde-3-phosphate dehydrogenase GDNF Glial cell line-derived neurotrophic factor

GRO-Seq Global Nuclear Run-On and Sequencing

HER2 Human epidermal growth factor receptor 2

JMJD JuMonJi domain-containing histone demethylase

MEME Multiple EM for Motif Elicitation

M-MLV RT Moloney Murine Leukemia Virus Reverse

Transcriptase

PIAS Protein inhibitor of activated signal transducer and

activator of transcription

PPRE Peroxisome proliferator-activated receptor

response element

Runx1 Runt-related transcription factor 1

SERM Selective estrogen receptor modulator

SMRT Silencing mediator of retinoid and thyroid

receptor

SWI/SNF Switching/sucrose non-fermenting

TBST Tris-buffered saline containing Tween-20

TFBS Transcription factor binding sites

TGF-β Transforming growth factor-β

TRE Tetradecanoyl phorbol acetate response element

PUBLICATION

Tan SK, Lin ZH, Chang CW, Varang V, Chng KR, Pan YF, Yong EL, Sung WK, Cheung E (2011) AP-2gamma regulates oestrogen receptor-mediated long-range chromatin interaction and gene transcription The EMBO journal 30: 2569-2581

Chng KR, Chang CW, Tan SK, Yang C, Hong SZ, Sng NY, Cheung E (2012) A transcriptional repressor co-regulatory network governing androgen response in prostate cancers The EMBO journal 31: 2810-2823

CHAPTER 1 INTRODUCTION

1.1 Physiological importance of estrogen

In females, estrogens are produced mainly in the granulosa cells of mature ovaries, through aromatization of thecal cell-derived androgens by aromatase (Pettersson & Gustafsson, 2001) Estrogens are then secreted into the bloodstream

to act on distal target tissues Three different types of estrogens are synthesized in females from puberty to adulthood 17β-estradiol (E2) derived from testosterone

is the primary estrogen in premenopausal women; estriol (E3) is produced in large amount by the placenta during pregnancy; and estrone (E1), a weaker hormone than E2, is the principal source of estrogen produced by extragonadal tissues in postmenopausal women (Ali & Coombes, 2000; Simpson, 2003) Estrogens play

a key role in the development of female secondary sex characteristics such as stimulating the growth of mammary glands and endometrial during puberty and pregnancy (Couse & Korach, 1999) In addition, estrogens mediate other physiological functions which include regulation of metabolic homeostasis and lipid metabolism, protection of cardiovascular system and maintenance of bone

mass (Foryst-Ludwig & Kintscher, 2010; Imai et al, 2010; Mendelsohn & Karas,

discovered and the former ER was renamed as ERα (Kuiper et al, 1996)

Differential expression patterns of ERα and ERβ across species, genders and tissue types inevitably result in variable phenotypic outcomes (Couse & Korach, 1999)

1.2 ER structure and mechanism of action

ERα and ERβ, encoded by separate genes, are ligand-inducible transcription factors which belong to the superfamily of nuclear hormone receptors (NRs)

(Mangelsdorf et al, 1995) The ERα protein consists of 596 amino acids while

ERβ consists of 530 amino acids and this translates to a molecular weight of 66 kDa and 59 kDa respectively Like other steroid hormone receptors, they share common structural architecture as shown in Figure 1.1 The ER protein comprises six functional domains From the N-terminal A/B to the C-terminal E/F domain, each domain has variable species homology and confers different functions (Bai

& Gust, 2009; Nilsson et al, 2001) The transactivation function of ER is

governed by two critical domains The least conserved A/B domain harbors the ligand-independent transactivation function AF-1, and the E/F domain harbors the

ligand-dependent transactivation function AF-2 (Kraus et al, 1995; Tora et al,

1989) In the absence or low level of estrogen, site-specific phosphorylation within the AF-1 region enhances transcriptional activation of ER target genes

(Bunone et al, 1996; Joel et al, 1998; Kato et al, 1995; Tremblay et al, 1999) C

domain, the most highly conserved domain, has two zinc fingers which are responsible for ER dimerization and interaction with specific DNA sequences

Amino acids involved in DNA binding specificity of ER reside in the first zinc finger constituting the P-box (CEGCKA) that recognizes AGGTCA core

sequence (Pettersson & Gustafsson, 2001; Schwabe et al, 1993) D domain

contains the nuclear localization signal (NLS) and also acts as a hinge between the C and E/F domain Apart from ligand binding and transcriptional activation, E/F domain also interacts with co-regulators via the AF-2 region and contributes

to part of the dimerization interface with the C domain (Henttu et al, 1997;

Pettersson & Gustafsson, 2001) Moreover, differences in ERα and ERβ ligand binding domain confer their specificity and affinity for natural and synthetic

ligands (Barkhem et al, 1998)

ER dimerization results in formation of functional homo- or heterodimers that cooperatively bind to promoter and enhancer elements with consensus estrogen response element (ERE), which is typically made up of two core sequences arranged as palindromes with three-nucleotide spacer (AGGTCAnnnTGACCT)

(Cowley et al, 1997; Klinge, 2001; Pettersson et al, 1997) Since receptor

dimerization enhances DNA binding stability, ER may also bind to imperfect ERE, and thereby increasing the number of genomic regions that it can interact with (Kuntz & Shapiro, 1997)

ER can be activated by several pathways to mediate its transcriptional function (Figure 1.2) In the classical pathway, ER is inactive when complexed with the heat-shock protein 90 (HSP90) that acts as a chaperone to maintain the receptor in

a properly folded conformation Upon binding to estrogen, ER dissociates from HSP90 and undergoes conformational change for subsequent activation and dimerization After nuclear translocation, ER dimer binds to the regulatory elements of its target genes ER-DNA interaction may occur directly via EREs or indirectly by tethering to adaptor transcription factors such as activator protein-1 (AP-1), specificity protein 1 (Sp1) or the recently identified runt-related

transcription factor 1 (Runx1) (Paech et al, 1997; Porter et al, 1997; Qin et al, 1999; Stender et al, 2010; Webb et al, 1999) In a temporal- and spatial-specific

manner, ER then recruits general transcription factors including RNA polymerase

II (RNA Pol II) and co-regulators (co-activators and co-repressors) that ultimately leads to transcriptional activation or repression of its target genes (Sommer & Fuqua, 2001) (Figure 1.2)

Non-classical pathways which activate ER transcriptional activity may be dependent or -independent One of the pathways involves the membrane-bound

ligand-ER (mligand-ER) that responds to extracellular ligand and triggers downstream signaling Pathways independent of estrogen may be triggered by growth factors such as epidermal growth factor (EGF), insulin and transforming growth factor-β (TGF-β); neurotransmitters such as dopamine; and second messengers such as cyclic AMP (cAMP) These signals subsequently activate downstream protein kinases signaling cascades which include the mitogen-activated protein kinase (MAPK) and protein kinase A (PKA) pathways The activation signals eventually converge at ER to enhance its transcriptional activity through phosphorylation of

specific serine residues (Aronica & Katzenellenbogen, 1993; Bunone et al, 1996; Chen et al, 1999b; Gangolli et al, 1997; Thomas et al, 2005; Tremblay et al,

1999) Apart from phosphorylation, other post-translational modifications have

been identified as events modifying ER transcriptional activity (Le Romancer et

al, 2011) For example, Lys266 and Lys268 were identified as sites of acetylation

and sumoylation in ERα, which both are modified by p300 and protein inhibitor

of activated signal transducer and activator of transcription 1 (PIAS1) and PIAS3 respectively These modifications require the presence of estrogen and can result

in increased DNA binding affinity and transcriptional activity of ERα (Kim et al, 2006; Le Romancer et al, 2011; Sentis et al, 2005)

Figure 1.1 Schematic diagram of ER functional domains and their homology

ERα and ERβ are homologous proteins that are structurally characterized by six functional domains AF-1 and AF-2 within the A/B and E/F domains are responsible for transactivation in ligand-independent and -dependent manner, respectively The highly conserved C domain and E/F domain are essential for receptor dimerization and allows ER to associate with ERE within the regulatory elements of its target genes (Bai & Gust, 2009)

Figure 1.2 Classical and non-classical models of ER-mediated transcription

In the classical model, (i) cytosolic ER responds to estrogen which diffuses through the nuclear membrane In the non-classical models, (ii) mER, (iii) growth factor receptors and (iv) G-protein coupled receptors (GPCR) respond to extracellular signals and trigger downstream phosphorylation cascades, thereby transmitting the activation signals to ER Activated ER dimerizes and binds ERE, followed by recruitment of the basal transcriptional machinery to stimulate gene transcription (Bai & Gust, 2009)

1.3 General co-regulators of ER-mediated transcription

Similar to other NRs, transcriptional regulation by ER is a multistep process that involves 1) ER association with the regulatory elements of its target genes; 2) recruitment of co-regulators to the ER binding sites which may alter chromatin structure; 3) assembly of basal transcriptional machinery; and 4) termination of transcription events Depending on the cell type and ER subtype, different co-regulators may be recruited Different ligands can also result in preferential recruitment of certain co-regulators by inducing different ER structural

rearrangements (Paige et al, 1999) In general, co-regulators contain a highly

conserved LxxLL motif, also known as NR-box, which interacts with the AF-2 domain of ER (Glass & Rosenfeld, 2000; McKenna & O'Malley, 2002) In addition, co-regulators may interact with other domains of ER, such as the AF-1 and hinge region, or indirectly interact with ER via other anchoring co-regulators

To date, more than 100 co-regulators of ER have been documented (http://www.nursa.org/)

Broadly, co-regulators are classified into co-activators and co-repressors which activate and repress gene expression respectively ER is known to interact with p160/SRC family of co-activators, mainly steroid receptor co-activator-1 (SRC-1) and SRC-3/Amplified in breast cancer 1 (AIB1), for transcriptional activation in

the presence of estrogen (Azorsa et al, 2001; Henttu et al, 1997; Onate et al,

1995) SRC-1 has also been shown to mediate ligand-independent activation of

ER Moreover, overexpression of SRC-1 can enhance transcriptional activity of

ERα bound by anti-estrogen, tamoxifen (Smith et al, 1997; Tremblay et al,

1999) Transcriptional activation is generally preceded by chromatin modulation that involves post-translational modification of the histone tails and nucleosome remodeling Therefore, chromatin modifiers are usually further recruited by ER

Histone acetyltransferases (HATs), eg CREB-binding protein (CBP) and p300, are important ER co-activators that serve to acetylate lysine residues within the N-

terminal tails of core histones H3 and H4 (Chen et al, 1997; Hanstein et al, 1996; Ogryzko et al, 1996; Spencer et al, 1997) For instance, acetylation of histone H3 lysine 9 (H3K9ac) is linked to transcriptional activation (Cheng et al, 2006; Roh

et al, 2005) In addition, another ER-associated factor with HAT activity known

as the p300/CBP-associated factor (PCAF) works synergistically with p300 in mediating H3K9ac and H3K14ac to contribute to open chromatin structure

(Kouzarides, 2007; Santos-Rosa et al, 2003) SRC-1 and SRC-3 are also known to possess HAT activity (Chen et al, 1997; Spencer et al, 1997) Besides neutralizing

the positive charges on lysine residues to reduce the histone-DNA interaction, acetyl-lysine moiety also generate specific docking sites recognized by bromodomain proteins (Yang, 2004) Examples of bromodomain proteins include CBP, p300, PCAF and switching/sucrose non-fermenting (SWI/SNF) chromatin remodeling complex that hydrolyzed ATP to locally disrupt histone-DNA

interaction (Belandia et al, 2002; Chatterjee et al, 2011; Garcia-Pedrero et al, 2006; Manning et al, 2001; Mujtaba et al, 2007; Zeng et al, 2008) The

recruitment of bromodomain proteins to acetylated histones may therefore further modulate the chromatin state for transcriptional activation

HATs and p160 co-activators have also been shown to work synergistically with protein arginine N-methyltransferases (PRMTs) such as PRMT1 and PRMT4/co-

activator-associated arginine methyltransferase 1 (CARM1) (Koh et al, 2001; Ma

et al, 2001; Strahl et al, 2001) PRMT1 and CARM1 mediate transcriptional

activity by catalyzing the addition of methyl groups to specific arginine residues

on histone tails and both interact with ER indirectly through p160 co-activators

(Chen et al, 2000; Chen et al, 1999a; Teyssier et al, 2002) Studies have shown

that methylation of histone H4 arginine 3 (H4R3) specifically by PRMT1

facilitate subsequent acetylation of H4 tail by p300 (Strahl et al, 2001; Wang et

al, 2001) Such modification by PMRT1 is found at the promoter of trefoil factor

1 (TFF-1)/pS2 in the presence of estrogen and is necessary for gene transcription (Wagner et al, 2006) CARM1 also methylates H3R17 at TFF-1/pS2 promoter in

an estrogen-dependent manner (Bauer et al, 2002) Moreover, CARM1 was

reported to directly methylate other H3 arginine residues such as H3R2 and

H3R26 during hormone-dependent transcriptional activation (Ma et al, 2001; Schurter et al, 2001) Besides its role as a methyltransferase, CARM1 also serves

as an adaptor protein to interact with brahma-related gene 1 (BRG1) which is a

component of the SWI/SNF chromatin remodeling complex (Nie et al, 2000; Xu

et al, 2004) Both CARM1 and BRG1 have been shown to be recruited to ER

target genes in an estrogen-dependent manner to cooperatively activate gene

transcription (Xu et al, 2004)

On the other hand, methylation on histone H3 lysine 9 and 27 (H3K9 and H3K27) functions as repressive marks on non-active enhancers Lysine demethylases such

as lysine specific demethylase 1 (LSD1) and JuMonJi domain-containing histone demethylase 2B and 3 (JMJD2B and JMJD3) that respectively remove H3K9 and H3K27 methylation have been shown to mediate estrogen-regulated transcription

(Garcia-Bassets et al, 2007; Kawazu et al, 2011; Svotelis et al, 2011) Together,

these chromatin modifying enzymes induce an open chromatin structure at gene regulatory elements that is competent for recruitment of basal transcription machinery and transcriptional activation

Nuclear receptor co-repressor (N-CoR) and silencing mediator of retinoid and thyroid receptors (SMRT) are the best characterized and structurally related co-

repressors of nuclear receptors (Chen & Evans, 1995; Horlein et al, 1995) To

repress gene transcription, ER has been shown to associate with the co-repressors, which further recruit mammalian Sin3 (mSin3)/histone deacetylases (HDACs)

complexes to the chromatin (Alland et al, 1997; Heinzel et al, 1997; Laherty et al, 1998; Nagy et al, 1997) There are eighteen HDACs (HDAC 1-11 and SIRT 1-7)

identified in human HDACs generally play opposing roles to HAT by catalyzing the removal of acetyl groups from lysine residues in histone tails to induce

chromatin condensation (Bolden et al, 2006) For example, HDAC 1, 2, and 7

have been shown to be involved in ER-mediated transcriptional repression

(Laherty et al, 1998; Malik et al, 2010) Hence, combinatorial recruitment and

exchange of co-regulators fine tune ER activity and are fundamental in generating appropriate transcriptional outcomes

1.4 Significance of ER in breast cancer

Breast cancer is the most frequently occurring malignancy in women The disease accounts for 1 in 8 women being diagnosed in United States and nearly half a million deaths in 2008 worldwide (Maxmen, 2012) Advancements in diagnostic methods including mammography screening have facilitated regular breast checks Therefore, early detection and improved treatment regimens developed from better understanding of breast cancer biology has contributed to declines in

breast cancer mortality from 1990 to 2007 (DeSantis et al, 2011) More than a

century ago, ovarietomy performed on patients suffering from advanced breast cancer resulted in tumor regression indicate the importance of estrogen in driving breast cancer cells growth (Beatson, 1896) In normal breast tissue, ERβ is highly expressed in the luminal epithelium of the duct and the surrounding stroma On the other hand, ERα is expressed at low levels and only in 15% of the luminal epithelial cells However, ERα is the predominant ER in breast cancers with about 70% of all cases expressing ERα and progesterone receptor (PR) which an ERα-regulated gene In comparison, the expression of ERβ is typically down-regulated

in breast cancers (Fox et al, 2008; Hartman et al, 2009) Noteworthy, several

studies have reported the inhibitory effects of ERβ on ERα activity and

estrogen-stimulated proliferation of ERα-positive cell lines (Behrens et al, 2007; Chang et

al, 2006; Paruthiyil et al, 2004; Williams et al, 2008) Therefore, it is believed

that ERα is the master regulator of initial proliferation of cancer cells and tumor growth, rendering it an ideal drug target for treatment of the majority of breast cancers

Using the microarray platform, gene expression profiling on patient-derived tumor samples allows stratification of breast cancers into distinct subtypes These include normal-like, luminal A, luminal B, human epidermal growth factor

receptor 2 (HER2/ERBB2) amplified and basal-like subtypes (Cheang et al, 2008;

Reis-Filho & Pusztai, 2011) Luminal subtype tumors are ERα/PR-positive, low grade and show better prognosis because they are more likely to respond to anti-

hormonal therapy if treated at early stage (Ibrahim et al, 2009; Sorlie et al, 2001) HER2/ERBB2 subtype tumors generally grow faster and are more aggressive than

luminal subtypes due to overexpression of HER2 Intriguingly, even though 50%

of the HER2/ERBB2 subtype tumors are ERα-positive, they do not respond to

endocrine therapy Therefore, HER2-amplified breast cancers are usually treated with trastuzumab (Herceptin®) which is a monoclonal antibody that binds and neutralizes HER2 activity, and it has been shown to reduce the risk of disease recurrence (Fiszman & Jasnis, 2011; Prat & Baselga, 2008) Basal-like subtype is defined by tumors that do not express ERα, PR and HER2 (triple-negative) but express basal markers such as keratins and laminin They are often high grade

tumors and exhibit very poor prognosis (Ibrahim et al, 2009; Ihemelandu et al,

2008) Hence, advanced breast cancers generally show less dependency on ERα to manifest their aggressive phenotype

1.5 Therapeutic regimens for ERα-positive breast cancer

Despite more than two-thirds of breast cancers being ERα-positive, only 40% of them display responsiveness to anti-hormonal therapy The first anti-estrogen developed to treat ERα-positive breast cancer is tamoxifen (TAM), which was later reclassified under selective estrogen receptor modulators (SERMs) due to its selective agonistic and antagonistic activities in certain tissues Tamoxifen was successful in treating hormone-responsive breast cancer as it functions as an anti-estrogen solely in the breast, but acts like an estrogen in the uterus, bone and

cardiovascular system (Osborne, 1998; Osborne et al, 2000) Till today,

tamoxifen treatment has been the standard adjuvant therapy, although its

drawbacks includes greater risk for endometrial cancer (Fisher et al, 1994)

Another SERM, raloxifene (RAL), offers better protection against endometrial cancer by functioning as an antagonist in the breast and uterus, while exhibiting agonistic effects in the bone and cardiovascular system (Balfour & Goa, 1998;

Fuchs-Young et al, 1995; Levenson et al, 1998) Pure anti-estrogen, ICI 182,780

(fulvestrant), is used as an alternative treatment for breast cancer when tamoxifen therapy is ineffective Regardless of tissue type, it functions as an antagonist of

ERα by targeting the receptor to proteasome-mediated degradation (Howell et al,

2000; Robertson, 2001) Furthermore, studies have shown that treatment with another class of drug, known as aromatase inhibitors (AIs), has led to better

prognosis in post-menopausal women with hormone-responsive breast cancer by inhibiting the conversion of androgen to estrogen (Conte & Frassoldati, 2007; Osborne & Tripathy, 2005)

Although the anti-estrogen therapeutic approach was proven effective, a proportion of patients who were initially sensitive to tamoxifen eventually

become resistant to it (Clarke et al, 2003) The molecular mechanisms underlying

the acquirement of tamoxifen insensitivity still remains unclear However, it has been postulated that this could be partly due to ERα loss or aberrant change in

expression levels of co-regulators in the resistant cells (Kuukasjarvi et al, 1996)

Studies have also reported that up-regulation of co-activators, such as SRC-1 and SRC-3, or down-regulation of co-repressors, such as N-CoR, may result in

acquired resistance to tamoxifen (Anzick et al, 1997; Hurtado et al, 2008; Lavinsky et al, 1998; Xu et al, 1998)

1.6 Comprehending ERα transcriptional network – before next-generation sequencing era

Given the occurrence and importance of ERα in breast cancer, much emphasis has been placed on defining the transcriptional network of ERα in human breast cancer cell models Using gene expression profiling on microarray, many estrogen-regulated genes were identified in breast cancer cells In early studies of ERα-mediated gene transcription, proximal promoters of the estrogen-regulated genes which contain putative functional ERE or binding motifs of ERα-associated

transcription factors were assessed for promoter activity using reporter assay For

example, ERE is found within the proximal promoters of TFF-1/pS2, ER-binding site-associated antigen 9 (EBAG9) and Cathepsin D (CTSD), while promoters of other ERα target genes such as c-Myc, insulin-like growth factor-1 (IGF-1) and Cyclin D1 (CCND1) are devoid of ERE but contain AP-1 and Sp1 binding sites that may interact with ERα indirectly (Augereau et al, 1994; Berry et al, 1989; Dubik & Shiu, 1992; Ikeda et al, 2000; Shiozawa et al, 2004; Umayahara et al,

1994) Despite being able to identify and validate functional elements,

transcriptional studies based on in vitro measurement of ERα binding to the

histone-free DNA and transient transfection are insufficient to accurately

recapitulate the complexity of the genome in vivo

Chromatin immunoprecipitation (ChIP) is a powerful technique developed for

studying in vivo protein-DNA interaction and allows detection of genomic regions

that are bound by the protein of interest (Kuo & Allis, 1999) Besides revealing further insights on the genomic behavior of ERα and its associated co-regulators, ChIP assay also permits recognition of post-translational modifications on histone tails accompanying the network of transcription ChIP studies in breast cancer cells have demonstrated that ERα and its co-regulators are recruited and

assembled in sequential- and cyclical-manner at the promoters of TFF-1 and CTSD in response to estrogen, and this is necessary to continuously remodel the associated chromatin to achieve transcription competency (Metivier et al, 2003; Shang et al, 2000) However, ChIP assay coupled with semi-quantitative

polymerase chain reaction (PCR) or quantitative PCR (qPCR) is capable of assessing only a limited number of putative genomic binding sites To better understand the global transcriptional regulation by ERα, it is important to identify where and how it binds within the entire genome

Bioinformatics tools, such as Dragon, were developed to predict ERBS at gene

promoter regions based on the presence of ERE (Bajic & Seah, 2003; Bajic et al,

2003) Although such analysis is in concordance with the theoretical view that EREs are more likely to occur within the promoters of ERα target genes, this approach is biased by assuming that ERα genomic occupancy can be entirely

accounted by gene promoters (Kamalakaran et al, 2005; Lin et al, 2004) In fact, studies on the transcriptional regulation of β-globin gene have demonstrated that

distant regulatory elements situated more than 10 kb away from target genes can

control gene expression (Forrester et al, 1990; Kim et al, 1992; Sawado et al,

2003) Furthermore, unbiased computational screening for ERE in the entire

genome results in 14% true predictions upon validation (Vega et al, 2006) This implies that the presence of ERE alone is inadequate to define genuine in vivo

throughput next-generation sequencing that is based on different principles from Sanger-based sequencing allows interrogation and visualization of the genomic properties and functions from a panoramic view In recent years, such technology

is commonly utilized for genomic analyses as generation of sequences at unprecedented scale in a rapid and affordable manner is made possible

(Marguerat et al, 2008; Metzker, 2010) The wealth of information obtained from

powerful and wide applications of next-generation sequencing enable researchers

to tackle complex biological questions that were difficult to address before

Defining the cistrome of ERα

To globally map ERBS in an unbiased manner, ChIP-on-chip approach that couples ChIP with microarray technology was earlier utilized by a few groups From using Affymetrix tiled oligonucleotide microarrays that cover the entire non-repetitive sequence of chromosomes 21 and 22 to entire non-repetitive human

genome sequence, majority (more than 90%) of in vivo ERBS identified were

mapped to genomic regions that are distal from promoters of putative target genes

(Carroll et al, 2005; Carroll et al, 2006; Liu et al, 2008) Such findings support

the view that distal regions are important for gene transcription by functioning as enhancer elements

With advancements in the sequencing technology, limitations imposed by the array probes including specificity and design issues were eventually overcome ChIP-Paired End DiTag (ChIP-PET) and ChIP-Sequencing (ChIP-Seq) powered

by high-throughput sequencing have identified more than 10,000 ERBS with high

reproducibility (Fullwood et al, 2009b; Joseph et al, 2010; Lin et al, 2007; Innes et al, 2010; Welboren et al, 2009) Publically available bioinformatics tools

Ross-have also facilitated analyses of these genome-wide datasets For instance, scanning for non-ERE transcription factor motifs that are over-represented within ERBS have become increasingly common to predict transcription factors collaborating with ERα (from hereon called collaborative factors) Some examples of these web-based tools include Cis-Elements Annotation System (CEAS), CENTDIST and Discriminative Regular Expression Motif Elicitation

(DREME) (Bailey, 2011; Ji et al, 2006; Zhang et al, 2011) In addition,

association of ERBS with estrogen-regulated genes and/or genomic landscape of RNA Pol II and other transcription factors have provided valuable insights on the

mechanistic details underlying the ERα cistrome (Carroll et al, 2006; Welboren et

al, 2009) Furthermore, with maturation of the sequencing technology, scientists

are beginning to challenge from performing ChIP-Seq in breast cancer cell lines

to tumor samples to associate the genomic properties of ERα with clinical

outcomes (Ross-Innes et al, 2012)

Defining the interactome of ERα

Numerous studies have shown that majority of ERBS are situated far from gene promoters This raises the question as to which distal binding sites are functional

in transcriptional regulation Given that the chromatin is densely packaged and organized in higher-order three-dimensional structures, this implies that distant

DNA elements may be brought close spatially for functional interactions Chromosome conformation capture (3C) is a technique to detect long-range

chromatin interactions between regulatory elements (Dekker et al, 2002; Hagege

et al, 2007) For example, distal enhancer and proximal promoter ERBS that are involved in transcriptional activation of TFF-1 gene were shown to interact with 3C assay (Carroll et al, 2005; Pan et al, 2008) Other techniques to improve the

study of long-range chromatin interaction such as ChIP-3C, 3C-on-chip, 4C, 5C

and 6C have evolved from the original 3C assay (Cai et al, 2006; Dostie et al, 2006; Simonis et al, 2006; Tiwari et al, 2008; Zhao et al, 2006) However, these

techniques are limited to detection of chromatin interactions between a few genomic loci or a portion of the genome

To globally capture de novo long-range chromatin interactions, techniques such as Hi-C and ChIA-PET were developed (Fullwood et al, 2009a; Lieberman-Aiden et

al, 2009) Hi-C combines 3C and high-throughput sequencing to capture all interactions within the genome (Lieberman-Aiden et al, 2009) In comparison,

ChIA-PET integrates ChIP, 3C, paired-end cloning and high-throughput sequencing to capture genome-wide interactions between DNA fragments that are

brought into close proximity by protein factors of interest (Fullwood et al, 2009a) This technique was utilized to obtain a global view of the de novo long-range

chromatin interactions mediated by ERα Analysis of ChIA-PET data yields information on both the cistrome and interactome associated with ERα which indicates their functionality The data suggest that a proportion of distal ERBS

may act as bona fide transcriptional enhancers to control transcription via chromatin looping

1.8 Known collaborative factors of ERα

Genomic analyses of ERBS determined by high-throughput ChIP-based technologies such as ChIP-on-chip, ChIP-PET and ChIP-Seq have accelerated the discovery and understanding of the function of ERα collaborative factors Considering the complexity of transcriptional regulation, it is likely that there are other transcription factors required for coordinating ERα function which currently remain unidentified In this section, I will highlight some ERα collaborative factors that were previously identified from these genome-wide data

al, 2005) This is in agreement with a human promoter-based ChIP-on-chip study

that was published concurrently showing that FKH motifs are enriched within the

promoters bound by ERα (Laganière et al, 2005) Hepatocyte nuclear factor 3α (HNF3α) or FoxA1, a member of FKH family, is earlier known to interact with

ERα in a ligand-dependent manner and is involved in regulation of a few estrogen

target genes, eg TFF-1 and Vitellogenin B1 (Beck et al, 1999; Robyr et al, 2000;

Schuur et al, 2001a) However, this was the first time that the role of FoxA1 in

assisting ERα-mediated transcription was studied and it was found to occupy the

ERBS before and after estrogen stimulation (Carroll et al, 2005) More

importantly, FoxA1 is necessary for ERα recruitment and modification of chromatin structure at the gene regulatory regions to mediate transcriptional

competency, hence establishing its role as a pioneering factor of ERα (Carroll et

al, 2005; Eeckhoute et al, 2009)

c-Jun, Oct1, C/EBPα, Nkx3.1 and LEF-1

By extending the ChIP-on-chip experiment to cover the entire genome, Carroll et

al observed over-representation of other binding sequences from scanning 3,665

identified ERBS with position-specific score matrices (PSSM) from TRANSFAC

and JASPAR (Carroll et al, 2006) These include AP-1, octamer-binding

transcription factor (Oct), CCAAT/enhancer-binding protein (C/EBP) motifs Since specific member of each transcription factor family eg c-Jun, Oct1 and C/EBPα are highly expressed in MCF-7 cells and have been associated with ERα

pathway, their binding to selected ERBS was examined (Boruk et al, 1998; Cicatiello et al, 2004; Kushner et al, 2000) Indeed, they co-localize with ERα at a

subset of ERBS studied Moreover, unlike FKH, Oct and C/EBP, pairwise analysis showed that AP-1 and ERE motifs exhibit negative correlation in their co-occurrences at ERBS This coincides with the fact that AP-1 factors function

to tether ERα while the others may serve as collaborative factors through binding adjacent to ERα Furthermore, using the same dataset, screening for motifs within

ERBS which are highly conserved across sixteen species revealed an enrichment

of Nkx and TCF/LEF consensus binding sequences (Holmes et al, 2008) Nkx3.1

and LEF-1 are transcription factors that have been reported for their relevance in

prostate and mammary biology respectively (Bhatia-Gaur et al, 1999; El-Tanani

et al, 2001) In the presence of estrogen, both factors display reduced

co-localization with ERα on the chromatin In addition, knockdown and overexpression of Nkx3.1 and LEF-1 led to an increase and decrease in the expression of estrogen-activated genes respectively, suggesting that they may act

as transcriptional repressors of ERα (Holmes et al, 2008)

PAX2

In another study to replicate the whole genome-based ChIP-on-chip experiment in MCF-7 cells, 8,525 ERBS were identified using a lower threshold of 5% false

discovery rate (FDR) (Hurtado et al, 2008) Sequence analysis of these ERBS

with CEAS program revealed an enrichment of PAX motif PAX2 is a transcription factor that was previously implicated as a tamoxifen-regulated effector in endometrial cancer Co-occupancy of PAX2 and ERα was observed

within an intronic region of the HER2/ERBB2 gene in response to estrogen and tamoxifen treatment (Hurtado et al, 2008; Muratovska et al, 2003; Wu et al,

2005) Tamoxifen-induced binding of ERα and PAX2 at the intronic region prevents the recruitment of SRC-3 co-activator, which is necessary for mediating

transcriptional repression of ERBB2, and this renders MCF-7 cells sensitive to

tamoxifen Lower expression of PAX2 in tamoxifen-resistant cells is outcompeted

by SRC-3 This results in activation of ERBB2 transcription and thus driving

overexpression of HER-2 Furthermore, ERα-positive breast tumors with higher expression of PAX2 show better prognosis with lower recurrence rate in SRC-3-

negative patients, compared to their positive counterparts (Hurtado et al, 2008)

Therefore, this study suggests that collaborative factors in ERα gene regulation may be essential in determining response to endocrine therapy

RARα

Using the ChIP-on-chip and ChIP-Seq platforms, two studies have reported that ERα and retinoic acid receptor-α (RARα) share a large number of genomic

binding sites (Hua et al, 2009; Ross-Innes et al, 2010) RARα is a member of the

nuclear hormone receptor superfamily which gene expression is induced by

estrogen (Laganiere et al, 2005; Petkovich et al, 1987) Both studies illustrated an

interesting behavior of RARα for being able to positively and negatively influence ERα transcriptional activity under different circumstances Using transgenic MCF-7 cells line that stably expresses GFP-tagged RARα, the earlier study reported an inhibitory effect of RARα on estrogen signaling Treatment with retinoic acid agonist results in misregulation of a significant number of estrogen-regulated genes as ERα recruitment to their common binding sites is abolished

(Hua et al, 2009) On the other hand, the latter study demonstrated that in the

presence of estrogen alone, endogenous RARα can interact and cooperate with ERα to drive efficient gene transcription and cell proliferation in MCF-7 cells However, co-treatment with estrogen and retinoic acid led to similar observation