The Roles of Atf3, an Adaptive-Response Gene, in Breast

THE ROLES OF ATF3, AN ADAPTIVERESPONSE GENE, IN BREAST CANCER DEVELOPMENT

DISSERTATION

Presented in Partial Fulfillment of the Requirements for

the Degree Doctor of Philosophy in the Graduate

School of The Ohio State University

Xin Yin, B. M.

*****

The Ohio State University 2008

Dissertation Committee: Approved by Dr. Tsonwin Hai, Advisor

Dr. James DeWille Advisor

Dr. Anthony Young The Ohio State Biochemistry Program

Dr. Mike Xi Zhu

ABSTRACT

During cancer progression, cells encounter many stress signals and all along they have builtin mechanisms to eliminate themselves. The successful cancer cells managed to foil this hardwired stress response. Emerging evidence indicates that some of the genes that normally function to eliminate the cells are coopted to become oncogenes. How the cellular (normal vs. cancerous) context determines that some genes undergo this “JekyllandHyde” conversion is an intriguing but largely unresolved issue in cancer biology.

Breast cancer is the most common malignancy among women in the United

States and is the second leading cause of cancer death in women, after lung cancer. The initiation and development of breast cancer rely on both cell autonomous (genetic and epigenetic) alterations as well as stromacancer interactions.

In this thesis work, I found ATF3, an ATF/CREB family transcription factor encoded by an adaptiveresponse gene, is a new regulatory molecule with a dichotomous role. It enhances apoptosis in untransformed cells, but protects the cells from stressinduced death and promotes cell motility in malignant cancer cells. In an in vivo xenograft mouse model, in addition to promoting primary tumor growth, ATF3 also increases lung metastasis. ii

To explore the potential mechanisms by which ATF3 promotes metastasis, I set out to address three major questions. First, as a transcription factor, will

ATF3 regulate the expression of some target genes involved in cell motility?

Second, as an adaptive response gene, will ATF3 be induced in cancer cells in response to stromal signals and whether the induction of ATF3 mediates cancer cells’ response to stromal signals? Third, will the induction of ATF3 in cancer cells feed back on stromal cells to affect stromacancer interaction?

By examining potential ATF3 target genes, I found ATF3 regulates a set of genes involved in cell motility, some of which were proved to be direct targets of

ATF3. As an adaptiveresponse gene, ATF3 was induced by multiple stromal signals, such as TGFβ, TNFα and IL1β. When focusing on one multifunctional cytokine, TGFβ, I found that ATF3 can mediate TGFβ effects on target gene expression and cell motility regulation. The interaction between ATF3 and

Smad2/3 offers a partial mechanistic understanding. Besides mediating the intracellular signaling of stromal factors, ATF3 is also involved in stromacancer interaction, as demonstrated by the effect of ATF3 expression in cancer cells on macrophage recruitment and angiogenesis.

Finally, by examining ATF3 status in human breast cancer, I found that wild type ATF3 gene is frequently amplified and its expression elevated in human breast tumors, suggesting a pathophysiological relevance of ATF3 to human cancer. Through the database analysis of microarray data generated by two other groups, I found that the higher expression of ATF3 correlates with worse outcome of breast cancer patients.

iii

The significance of this work is severalfold. First, it identified a novel mediator that allows the cells to interpret the exogenous signals and initiate dichotomous cellular processes, in a manner dependent on the malignancy of the cells.

Second, as a downstream mediator for TGFβ, another welldemonstrated stromal factor with paradoxical functions during cancer development, ATF3 constitutes a link in the stromal signaling and transcriptional networks during stromacancer interactions. Third, the dichotomous effect of ATF3 was demonstrated in isogenic breast cancer cells representing different stages of cancer development.

Since ATF3 can regulate some genes in an opposite direction, this cell system may provide a handle to elucidate the “cellular contexts” (such as interacting proteins or cofactors) that allow ATF3 to regulate the same promoters in an opposite manner. This would help to unravel the mysteries behind the dichotomy in cancer development. Fourth, about 50% human breast tumors show up regulation of wildtype ATF3. In combination with the functional consequences of

ATF3 in both cell system and mouse model, we have discovered a new oncogene in malignant breast cancer cells that is likely to have pathophysiological relevance to human cancer.

DEDICATION

To Mom, Dad, Wei and my little girl, Cynthia

ACKNOWLEDGMENTS

I would like to give my sincere thanks my advisor, Dr. Tsonwin Hai, for the tireless dedication and endless support at all levels throughout these years. I appreciate the trainings and challenges she offered to help me understand the logic, thought process and experimental design for biological sciences. She never lets me down when I need guidance and encouragement, in science or in daily life. I wouldn’t be even close to where I am now without her.

I am also grateful for my committee members, Dr. James DeWille, Dr.

Anthony Young and Dr. Mike Xi Zhu, for their encouragement and instructive comments on my research, as well as their generous help on my future career.

I would like to thank the current and past members of the Hai lab: Dr. Matthew

G. Hartman, Dr. Dan Lu, Milyang Kim, Dan Li and Shawn Behan, with special thanks to Erik Zmuda and Christopher Wolford. They help me on experimental details, project discussions and English communications.

VITA

April 28, 1977…………………………….….Born – Jing Dezhen, P. R. China

1994 – 2000 ………………………………..B.S. Clinical Medicine, China Medical University, Shenyang, P.R.China

2000 – 2002……………………………….…Assistant Editor, Chinese Medical Sciences Journal , Beijing, P.R.China

2002 – 2003……………………………….…Program Fellowship, The Ohio State University

2003 – present………………………………Graduate Research Associate The Ohio State University

PUBLICATIONS

1. X. Yin , JW. DeWille and T. Hai. (2008) A potential dichotomous role of ATF3, an adaptiveresponse gene, in cancer development. Oncogene. 27(15): 2118-27 .

2. D. Li, X. Yin , EJ. Zmuda, CC. Wolford, X. Dong, MF. White and T. Hai. (2008) The repression of IRS2 gene by ATF3, a stressinducible gene, contributes to pancreaticcell apoptosis. Diabetes. 57(3): 635-44.

FIELD OF STUDY

Major Field: Ohio State Biochemistry Program

vii

TABLE OF CONTENTS

Page

ABSTRACT ...... ii DEDICATION ...... v ACKNOWLEDGMENTS ...... vi VITA…...... vii PUBLICATIONS ...... vii FIELD OF STUDY ...... vii LIST OF TABLES ...... xi LIST OF FIGURES ...... xii LIST OF FIGURES ...... xii ABBREVIATIONS...... xiv INTRODUCTION ...... 1 1. INTRODUCTION ...... 1 1.1. An overview of the ATF/CREB family of transcription factors and ATF3...... 3 1.1.A. The ATF/CREB family of transcription factors...... 3 1.1.B. Cloning and Characterization of ATF3 ...... 5 1.1.C. ATF3 as an adaptiveresponse gene...... 8 1.1.D. Biological consequences of ATF3 expression ...... 11 1.1.E. Summary...... 15 1.2. Overview of mammary development and tumorigenesis ...... 16 1.2.A. Mammary gland anatomy and development ...... 16 1.2.B. Epidemiology of breast cancer...... 19 1.2.C. Causes, evolution and heterogeneity of breast cancer ...... 19 1.2.D. Classification of breast cancer ...... 22 1.2.E. The multistage development of breast cancer and cancer metastasis ...... 23 1.2.E. Stromalcancer interaction ...... 24 1.2.F. Breast cancer, adaptive response and “JekyllandHyde” conversion...... 26 1.2.G. Conclusion...... 28 2. A POTENTIAL DICHOTOMOUS ROLE OF ATF3, AN ADAPTIVE RESPONSE GENE, IN BREAST CANCER DEVELOPMENT ...... 46 2.1. ABSTRACT...... 46 2.2. INTRODUCTION ...... 47

viii

2.3. MATERIALS AND METHODS ...... 51 2.4. RESULTS ...... 54 2.4.1. ATF3 has opposite effects on the untransformed MI cells and the malignant MIV cells ...... 54 2.4.2. ATF3 is proapoptotic in untransformed breast epithelial cells .56 2.4.3. ATF3 protects aggressive breast cancer cells from stress induced cell cycle arrest and promotes their cell motility and invasiveness in vitro ...... 57 2.4.4. ATF3 modulates the expression of genes known to regulate cell motility ...... 59 2.4.5. Differential regulation of gene expression by ATF3 in MI versus MIV cells...... 60 2.4.6. Direct target genes of ATF3 ...... 60 2.4.7. ATF3 promotes primary tumor growth and lung metastasis in an orthotopic xenograft mouse model ...... 63 2.4.8. ATF3 gene is upregulated in human breast cancer cells...... 64 2.5. DISCUSSION ...... 64 3. ATF3 AS A MEDIATOR FOR TGFβ SIGNALING IN BREAST CANCER CELLS...... 81 3.1. ABSTRACT...... 81 3.2. INTRODUCTION ...... 82 3.3. MATERIALS AND METHODS ...... 85 3.4. RESULTS ...... 87 3.4.1. ATF3 is induced by multiple stromal signals...... 87 3.4.2. ATF3 expression in MIV cells induces a morphological alteration resembling epithelialmesenchymal transition ...... 87 3.4.2. TGFβ induces ATF3 through multiple signaling pathways...... 88 3.4.3. ATF3 is sufficient and necessary for TGFβinduced target gene expressions and cell migration ...... 88 3.4.4. ATF3 interacts with Smad2/3 ...... 89 3.4.5. ATF3 and Smad3 cooccupy the promoter of target genes and reciprocally regulate the cell motility induced by each other...... 90 3.5. DISCUSSION ...... 91 4. A POTENTIAL ROLE OF ATF3 IN STROMACANCER INTERACTIONS105 4.1. ABSTRACT...... 105 4.2. INTRODUCTION ...... 106 4.3. MATERIALS AND METHODS ...... 109 4.4. RESULTS ...... 112 4.4.1. ATF3 recruits more macrophages in both primary tumors and lung metastatic nodules...... 112 4.4.2. ATF3 enhances angiogenesis in both primary tumors and lung metastatic nodules ...... 113 4.4.3. An in vitro analysis of cancermacrophage interaction ...... 114 4.4.4. ATF3expressing cancer cells juxtapose macrophages and endothelial cells in human breast cancer...... 116

4.5. DISCUSSION ...... 116 5. RELEVANCE OF ATF3 TO HUMAN BREAST CANCER...... 127 5.1. ABSTRACT...... 127 5.2. INTRODUCTION ...... 128 5.3. MATERIALS AND METHODS ...... 130 5.4. RESULTS ...... 132 5.4.1. ATF3 is upregulated in approximately 45% of human breast cancers...... 132 5.4.2. ATF3 gene is amplified in approximately 80% of human breast cancers...... 134 5.4.3. ATF3 has no mutations in the open reading frame in human breast tumors ...... 135 5.4.4. Higher ATF3 steadystate mRNA levels correlate with worse clinical outcomes of breast cancer patients ...... 135 5.4.5. Higher ATF3 expression level correlates with higher wildtype p53 level and lower Ecadherin level in human breast cancer...... 136 5.5. DISCUSSION ...... 138 6. FUTURE PERSPECTIVES...... 151 REFERENCES...... 155

LIST OF TABLES

Table Page

1.1 A partial list of the mammalian ATF/CREB family of transcription factors ...... 30 1.2 A partial list of treatments that induce ATF3 expression ...... 31 2.1 Properties of MCF10Aderived cell lines ...... 69 2.2 A list of the primers used in this study for the indicated assays ...... 71 2.3 Classification of potential ATF3 target genes based on their differential expression in MI versus MIV cells ...... 71 5.1 1q amplification in several human cancers ...... 142 5.2 Mutational status of p53 ORF in human breast tumors ...... 143

LIST OF FIGURES

Figure Page

1.1 Dendrogram representation of sequence homology within the bZip domain of transcription factors ...... 32 1.2 Schematic representation of the mRNA and protein for ATF3 ...... 33 1.3 Schematic representation of the ATF3 gene, the mRNA species generated by alternative splicing and the corresponding protein structures predicted predicted ...... 34 1.4 Nucleotide sequence of the 5’flanking region of ATF3 ...... 35 1.5 Anatomy of human breast ...... 36 1.6 Drawings depicting the development of the human breast, the basic functional unit of which is the terminal ductal lobular unit (TDLU) ...... 38 1.7 Schematic (Aa–d) and wholemount (Ba–d) presentation of the different stages of mouse mammary gland development ...... 38 1.8 Comparison of human and mouse mammary glands ...... 40 1.9 Multistage development of breast cancer ...... 40 1.10 Schematic representation of the metastatic process ...... 41 1.11 Efficiency of specific steps in metastasis and their dependence on the degree of malignancy of the cells ...... 42 1.12 Protumoral functions of tumorassociated macrophages (TAM) and their interplay with tumor cells ...... 43 1.13 General overview of the main signaling networks regulated by TGFβ in cancer in cancer ...... 45 1.14 TGFβ switches from tumor suppressor in the premalignant stages of tumorigenesis to prooncogene at later stages of disease leading to metastasis ...... 45 2.1 Characterization of stable cells expressing ATF3 or the control vector .. 72 2.2 ATF3 is proapoptotic in untransformed mammary epithelial cells ...... 74 2.3 ATF3 is protective in malignant mammary epithelial cells ...... 74 2.4 ATF3 enhances cell mobility and invasiveness in malignant breast carcinoma cells ...... 75 2.5 ATF3 regulates cellmotilityrelated genes ...... 76 2.6 Potential ATF3 target promoters ...... 77

xii

2.7 FN1, TWIST1, Snail, and Slug are potential direct target genes of ATF3 ...... 78 2.8 ATF3 promotes tumor growth and metastasis in vivo ...... 79 2.9 ATF3 gene is upregulated in breast cancer cell lines ...... 80 3.1 ATF3 can be induced by multiple stromal factors ...... 96 3.2 ATF3 expression in MIV cells induces an EMT phenotype ...... 97 3.3 ATF3 is induced by TGFβ through multiple signaling pathways ...... 98 3.4 ATF3 is necessary for TGFβmediated cell migration ...... 99 3.5 ATF3 is sufficient to regulate TGFβ target genes ...... 100 3.6 ATF3 is necessary to regulate TGFβ target genes ...... 101 3.7 ATF3 interacts with Smad2/3 ...... 102 3.8 Potential ATF3Smad2/3 target promoters ...... 103 3.9 ATF3 cooperates with Smad2/3 to regulate target gene expression and cell motility ...... 104 4.1 Ectopic ATF3 expression in MIV cells correlates with more macrophages recruitment in the lung metastatic nodules ...... 120 4.3 Ectopic ATF3 expression in epithelial cancer cells correlates with more macrophages recruitment in both primary tumors and lung metastatic nodules ...... 122 4.4 Ectopic expression of ATF3 in epithelial cancer cells correlates with higher level of angiogenesis in both primary tumor and lung metastatic nodules ...... 123 4.5 Breast cancer cells with higher ATF3 expression closely associate with macrophages and endothelial cells in xenograft mouse model ...... 124 4.6 MIV/ATF3 cells promotes macrophage migration and upregulates several chemoattractive cytokines ...... 125 4.7 Breast cancer cells with higher ATF3 expression closely associate with macrophages and endothelial cells in human breast cancers ...... 126 5.1 ATF3 expression is upregulated in human breast tumors ...... 144 5.2 ATF3 gene is amplified in human breast tumors ...... 145 5.3 Schematic view of ATF3 gene structure and primers used to amplify and sequence its open reading frame in human breast tumors ...... 146 5.4 KaplanMeier survival curves for patients with low ATF3 expression in breast tumor (blue line) versus those with high ATF3 expression in breast tumor (red dashed line) ...... 147 5.5 ATF3 expression positively correlates with p53 expression in human breast tumors ...... 149 5.6 Schematic view of p53 gene structure and primers used to amplify and sequence its open reading frame in human breast tumors ...... 149 5.7 Correlation between ATF3 expression and Ecadherin in human breast tumors ...... 150 6.1 ATF3 differentially regulates the migration of various breast cancer cell lines ...... 154

xiii

ABBREVIATIONS

aa amino acid

AP1 activator protein 1

ATF activating transcription factor

BCG Bacillus CalmetteGuérin bp base pair

BW body weight bZip basic regionleucine zipper

C.M. conditioned medium

C/EBP CCAAT/enhancer binding protein

CAV1 caveolin1 cdc42 cell division cycle 42 cDNA complementary DNA

CGH comparative genomic hybridization

ChIP chromatin immunoprecipitation

CoIP Coimmunoprecipitation

COL4A2 collagen IV a2

CRE cAMP responsive element

CREB cAMP responsive element binding protein

DAXX deathassociated protein 6

DC dendritic cell

DCIS ductal carcinoma in situ

xiv

dH2O deionized water

ECM extracellular matrix

EGF epidermal growth factor

EGFR epidermal growth factor receptor eIF eukaryotic initiation factor

EMT epithelialmesencymal transition

ER endoplastic reticulum

ETD extralobular terminal duct

FGF fibroblast growth factor

FGFR2 fibroblast growth factor receptor 2

FISH fluorescence in situ hybridization

FL full length

FN fibronectin

GMCSF granulocytemacrophage colonystimulating factor

HGF hepatocyte growth factor

HMECs human mammary epithelial cells hr hour

IDC invasive ductal carcinoma

IF immunofluorescence

IFN interferon

IHC immunohistochemistry

IL interleukin

IP intraperitoneal

IR irradiation

IRB Institutional Review Board

ITD intralobular terminal duct

kb kilobase

LCM laser capture microdissection

LPC lysophosphatidylcholine

LPS lipopolysaccharide

LSP1 lymphocytespecific protein

MAP3K1 mitogenactivated protein kinase kinase kinase 1

MAPK mitogenactivated protein kinase

MCP1 monocyte chemotactic protein1

MCSF macrophage colony stimulating factor

MEM Minimum essential media

MMP matrix metalloproteinase

MMS methyl methanesulfonate

NFκB nuclear factor κB

No. Number

NSAID nonsteroidal antiinflammatory drugs

ORF open reading frame oxLDL oxidized low density lipoprotein

PAI1 plasminogen activator inhibitor1

PAK p21activated kinase

PAR6 partitioningdefective protein 6

PCR polymerase chain reaction

PDGF platelet derived growth factor

PI3K phosphatidylinositol 3kinase

PMA paramethoxyamphetamine

xvi

PP2A protein phosphatase 2A

PR progesterone receptor

ROCK1 Rhoassociated, coiledcoil containing protein kinase 1

ROS reactive oxygen species

RTPCR reverse transcriptionpolymerase chain reaction

SBE Smadbinding element

SCID severe combined immunodeficient

SLPI secretory leukocyte peptidase inhibitor

SMURF1 Smad ubiquitination regulatory factor 1

SNP single nucleotide polymorphism

TAK1 TGF betaactivated kinase 1

TAM tumorassociated macrophage

TDLU terminal ductal lobular unit

TGF transforming growth factor

TM transmembrane domain

TNF tumor necrosis factor

TNRC9 thymocyte selectionassociated high mobility group box 9

TPA tetradecanoylphorbol acetate

TPSR Tissue Procurement Shared Resource

TSS Transcription start site

TβRII type II TGFβ receptor uPA urokinasetype plasminogen activator

UTR untranslated region

UV ultraviolet

VEGF vascular endothelial growth factor

xvii

WT wildtype

xviii

CHAPTER 1

INTRODUCTION 1. INTRODUCTION

This thesis presents my research over the past five years in Dr. Tsonwin Hai’s laboratory. The major interest in our laboratory is to understand how physiological stress leads to human diseases. To understand this process, we set out to study the stressregulated responses on multiple levels. First, on the molecular level, how do the cells sense the stress signal(s), activate downstream signaling pathways, and initiate a variety of alterations in and out of the cells, such as target gene expression, proteinprotein interaction, production and secretion of intracellular molecules? Second, on the cellular level, in response to stress stimuli, what are the phenotypic alterations in cellular processes such as morphology, proliferation, death, motility and so on? Third, on the tissue and organism level, what are the functional consequences of the stress response, single celltypedependent or involving multiple cell types, singleorgan dependent or involving multiple organs? Our strategy is to focus on a stress inducable gene, activating transcription factor 3 (ATF3), and by examining the regulations and funcitions of ATF3 to gain insightful understanding on how cells respond to various physiological and pathological stresses.

Currently, we are working on two stressassociated disease models: diabetes and breast cancer. My work focuses on the functions of ATF3 expression in breast cancer, the potential underlying mechanisms and the relevance of ATF3 to human breast cancer. The first part of this chapter covers some background information related to ATF3, including an overview on the ATF/CREB family of transcription factors, the cloning and characterization of ATF3 gene, identification of ATF3 as an adaptiveresponse gene, and functional consequences of ATF3 in different biological processes. The second part of this chapter focuses on the current understanding of breast cancer, including the anatomy and normal development of mammary gland, the epidemiology, cause, evolution, classification and multistage development of breast cancer, the contribution of stromacancer interaction in breast cancer development, the relationship between breast cancer development and adaptive response, and the dichotomous phenomenon in cancer development. Following this introduction, chapter 2 presents evidence supporting a dichotomous function of ATF3 in breast cancer development; chapter 3 examines the role of ATF3 as a mediator of TGFβ signaling in breast carcinoma cells; chapter 4 explores the involvement of ATF3 in stromacancer interactions; chapter 5 investigates the relevance of

ATF3 to human breast cancer; and chapter 6 provides my future perspectives of the current study.

1.1. An overview of the ATF/CREB family of transcription factors and ATF3

1.1.A. The ATF/CREB family of transcription factors

In the late 1980’s, while studying the factor(s) that activate the transcription of

E1Ainducible adenovirus early promoters E2, E3 and E4, Lee and colleagues found that a factor with the ability to bind to a common core DNA sequence (5’

CGTCA3’) can activate the transcription of those promoters and they named this factor Activating Transcription Factor (ATF) (Lee et al., 1987). The same year, in an independent study, Montminy and Bilezikjian identified a factor that binds to the cAMP responsive element (CRE) of the somatostatin promoter and named it cAMP responsive element binding protein (CREB)(Montminy and Bilezikjian,

1987). Later on, ATF was found to bind to the consensus sequence (5’

TGACGTCA3’) necessary and sufficient for CREB binding and was required for the transcription for cAMPinduced somatostatin gene (Lin and Green, 1988).

Based on this result, the authors proposed that ATF and CREB are similar or identical (Lin and Green, 1988). In the years followed, this hypothesis was proved to be only partially correct. It was correct because ATF and CREB are similar, both having the capability to bind to the consensus sequence 5’

TGACGTCA3’ (ATF/CRE site). It was incorrect because both ATF and CREB are representing not just one factor, but multiple factors. Over the years, many cDNAs encoding proteins with the activity to bind to the consensus ATF/CRE site have been isolated (Hai and Hartman, 2001) and they constitute a family currently with almost 20 members – ATF/CREB family (Hai, 2006a). Table 1.1

lists some of the mammalian ATF/CRE proteins grouped based on the overall amino acid similarity (Hai and Hartman, 2001).

In addition to binding to the consensus ATF/CRE site, a common feature for all members in ATF/CREB family is that they all contain the basic regionleucine zipper (bZip) DNA binding domain. However, not all bZipcontaining proteins belong to the ATF/CREB family, due to the inability to bind to the consensus

ATF/CRE site. Currently, there are about 55 human bZip proteins, and based on the homology within this bZip domain, some ATF/CREB proteins are more similar to members of other bZipcontaining superfamilies than with members in the

ATF/CREB family (Hai, 2006a; Newman and Keating, 2003) (Figure 1.1).

Furthermore, accumulative evidence indicates that through the bZip domain,

ATF/CREB family member can form homodimers and selective heterodimers with each other, and also with other bZipcontaining proteins, such as activating protein 1 (AP1) and CCAAT/enhancer binding protein (C/EBP) family of proteins

(Hai et al., 1999 and references therein).

Even with a structurally homologous bZip domain, the proteins in ATF/CREB family exhibit rather diverse biological functions. One common theme for a subgroup members of this family, including ATF2, ATF3, ATF4 and ATF6, is that they are all involved in stress responses, though in different context and with different regulatory mechanisms (Hai, 2006b). ATF2 is ubiquitously expressed and is rapidly phosphorylated in response to various stress stimuli, such as DNA damaging agents, inflammatory cytokines, and osmotic stress. The phosphorylation of ATF2 increases its halflife and transactivaion activities.

ATF4 and ATF6 both play a role in endoplasmic reticulum (ER) stress response, with ATF4 upregulated at the translational level in response to the phosphorylation of eukaryotic initiation factor 2α (eIF2α) and general translational inhibition, while ATF6 activated from a precursor form by proteases in the Golgi

(Hai, 2006b). The next section of this chapter will focus on another important stressinduced ATF/CREB family member, ATF3, with more detailed information on its cloning, characterization, regulation and functions.

1.1.B. Cloning and Characterization of ATF3

The human ATF3 was first cloned in 1989 from a cDNA library derived from

HeLa cells treated with tetradecanoylphorbol acetate (TPA) using a DNA probe containing three tandem ATFbinding sites (Hai et al., 1989). Since then, ATF3 orthologs have been identified from other species, including LRF1 in rat (Hsu et al., 1991) and TI241 in mouse (Ishiguro et al., 1996), both of which share approximately 95% amino acid identity with the human ATF3. The human ATF3 gene is localized in chromosome 1q32.3, covering a region of more than 55 kilobases. Characterization of the human ATF3 gene structure revealed that the fulllength ATF3 is encoded by four exons, designated as exons A, B, C, and E

(Liang et al., 1996). Exon A is 167 bp in length and contains the 5’ untranslated region. Exon B contains the AUG initiation codon and codes for the first 80 amino acid of fulllength ATF3. Exon C codes for the 36 amino acids that primarily represent the basic region, and exon E encodes 65 amino acids that

make up the leucine zipper region and the 3’ untranslated region (Liang et al.,

1996) (Figure 1.2).

As a transcription factor, ATF3 can either activate or repress the transcriptions of target genes, depending on the cellular context. When forming a homodimer,

ATF3 functions as a transcriptional repressor and can repress the activity of its own promoter (Chen et al., 1994; Wolfgang et al., 1997; Wolfgang et al., 2000), presumably by recruiting inhibitory cofactors to the target promoters. In heterodimerizing with other bZip proteins, such as ATF2, cJun, JunB and JunD,

ATF3 was shown to either activate or repress target genes (Chu et al., 1994; Hai et al., 1999).

In addition to the fulllength ATF3 (Liang et al., 1996), several alternatively spliced isoforms have been identified from different studies.

In a largescale study to identify and characterize the putative alternative promoters of human genes, an alternative transcription start site (TSS) was mapped to approximately 50 kilobases upstream from exon B and localized in a new exon A’ (Kimura et al., 2006), generating a long isoform 1. This long isoform

1 utilizes the same splice acceptor site of exon B and therefore produces the same protein as the fulllength ATF3.

The ATF3Zip isoform was isolated serendipitously in the process of studying fulllength ATF3 (Chen et al., 1994). Due to the inclusion of an additional exon

D, which is located between exons C and E, ATF3Zip produces a truncated protein lacking the leucine zipper dimerization domain. The ATF3Zip isoform does not bind to consensus ATF/CRE site, but consistently stimulates

transcription from promoters with or without ATF sites. This transcription stimulating activity is postulated to result from sequestering the inhibitory cofactors away from the target promoters (Chen et al., 1994).

ATF3Zip2a and ATF3Zip2b isoforms were originally identified in HUVEC cells following the treatment with homocysteine (Hashimoto et al., 2002). Both isoforms utilize a novel splice acceptor site located 76 bases upstream of exon

D, designated as exon D’. Splicing between exons D and E occurs in

ATF3Zip2b in a similar manner as ATF3Zip; however, this 91 base pair fragment is not removed in ATF3Zip2a (Hashimoto et al., 2002). Both isoforms encode a 135amino acid (aa) protein, designated as ATF3Zip2, which is common to fulllength ATF3 and ATF3Zip in the Nterminal 1115 aa. At the C terminal, ATF3Zip2 deletes the leucine zipper domain, which was replaced by a novel sequence of 20 aa. Lacking the leucine zipper domain, ATF3Zip2 cannot bind to DNA and counteracts the transcriptional repression by fulllength ATF3.

Similar to fulllength ATF3, ATF3Zip2 can be induced by various stress stimuli in both HUVEC and Saos2 cells. But they are not expressed in several metastatic esophageal cancer cell lines, which express high levels of fulllength

ATF3 (Hashimoto et al., 2002).

ATF3Zip2c and ATF3Zip3 isoforms were first reported in HepG2 hepatoma cells following amino acid deprivation and endoplastic reticulum (ER) stress (Pan et al., 2003). Alignment of nucleotide sequences showed that Zip2c arises from a splicing event that removes 87 bp within exon B and therefore creates two exons designated as exons B1 and B2. The Zip2c retains both exons D’ and D

but lacks exon E and produces a protein lacking the leucine zipper dimerization domain (Pan et al., 2003). ATF3Zip3 has an intact exon B; however, no splicing event occurs between exons C and D’, resulting in the inclusion of additional sequence containing a stop codon. As to the protein sequence,

ATF3Zip2c protein is composed of 106 amino acids and differs from the

ATF3Zip2 protein by the deletion of 29 amino acids near the N terminus.

ATF3Zip3 encodes a truncated protein of 120 aa with a nearly complete loss of the leucine zipper dimerization domain (Pan et al., 2003).

The ATF3b isoform, encoding a 124amino acid protein that lacks the N terminal 57 amino acids, was originally identified from untreated glucagons expressing pancreatic α cells (Wang et al., 2003). mRNA encoding this isoform arises due to a splicing event that removes 106 nucleotides covering the ATG start codon found in exon B. ATF3b is derived from translation initiation at an alternative downstream start codon, producing a truncated protein with an intact bZip domain that was shown to have higher binding affinity to glucagon promoter than fulllength ATF3 and can stimulate proglucagon gene transcription (Wang et al., 2003). The schematic representations of the mRNA and protein structures of different ATF3 spliced isoforms are illustrated in Figure 1.3.

1.1.C. ATF3 as an adaptiveresponse gene

In the earlier studies examining ATF3 expression profile, we noticed that

ATF3 fits the definition of immediateearly genes in many aspects (Hai et al.,

1999 and references therein). First, ATF3 mRNA level is relatively low or non 8

detectable in most biological samples examined, but greatly increases in response to various stress stimuli. Second, the 3’ untanslated region (3’ UTR) of

ATF3 mRNA contains several AUUUA sequences, a characteristic of many immediateearly genes. Third, ATF3 paralogs include cfos and fosB, both of which are wellcharacterized immediateearly genes. Although ATF3 induction is immediate and transient in most scenarios, it can be delayed, sustained or biphasic (Hai et al., 1999).

Promoter analysis of approximately 2 kilobases (kbs) region of the ATF3 gene upstream from the transcriptional start site revealed the presence of a TATA box located at 30 relative to the transcriptional start site (Liang et al., 1996). Many transcription factor binding sites, such as those for ATF/CRE, NFκB, AP1, E2F,

Myc/Max, Smad, p53, and HIF, were identified (Liang et al., 1996) (Figure 1.4).

The presence of these sites confers the induction of ATF3 in response to various intracellular and extracellular stimuli. For example, the ATF/CRE sites at the –90 and –20 positions relative to the TSS can be bound by ATF3 itself and confers autorepression activity, partially explaining the transient nature of ATF3 expression in response to various stress stimuli (Wolfgang et al., 2000). In response to the TGFβ treatment, Smad3 binds to a region between 1850 to

1408 relative to the transcription start site (TSS) of ATF3 gene and mediates its induction (Kang et al., 2003). p53 was shown to bind to two consensus elements at –379 and –342 relative to the TSS of ATF3 gene and mediated ATF3 expression in response to ultraviolet (UV) and protease inhibitor MG132 treatment (Zhang et al., 2002). Thus, ATF3 induction during the stress

responses appears to be mediated, at least partially, at the transcriptional level.

The presence of Myc/Max and E2F sites within the ATF3 promoter suggests that

ATF3 expression may be regulated in a cell cycledependent manner, consistent with the observation that ATF3 expression is induced in Sphase (Cho et al.,

2001; van der Meijden et al., 2002).

By examining the signals that induce ATF3, we noticed that many of them are stress stimuli, i.e., they are the signals capable of damaging the cells or tissues.

Therefore, we originally characterized ATF3 as a stressinducible gene (Hai and

Hartman, 2001; Hai et al., 1999). With more indepth understanding of ATF3, however, we realized that this characterization might be oversimplified, since many signals that can induce ATF3 do not fit in with the conventional definition of stress stimuli, such as serum, growth factors, adipokines and Sphase transition.

We proposed that a more accurate view is to characterize ATF3 as an adaptive response gene, meaning that the induction of ATF3 will help the cells to adapt to various signals, either intracellular or extracellular (Hai, 2006a). Table 1.2 lists some of the signals that were demonstrated to induce ATF3 in either cell cultures or animal models (Hai et al., 1999 and references therein). The list indicates that the induction of ATF3 is nonspecific; it is not celltype specific or stimulus specific. There are two potential explanations for the nonspecific induction of

ATF3 in response to various stress stimuli. First, ATF3 is just a marker protein for stress response, without any other functions. Second, in addition to functioning as a marker, ATF3 is an active component of the common subset of

“adaptive response” genes that play important roles in a variety of cellular

responses (Hai, 2006a). Through years of study on ATF3, we believe the second explanation best fits the roles of ATF3 in stress responses and more evidence will be presented as below.

1.1.D. Biological consequences of ATF3 expression

As described above, ATF3 can be induced by a variety of signals in all different types of cells and tissues, suggesting that ATF3 expression may have various functional consequences. Currently, the least understood part of ATF3 biology is what biologically important functions are performed by ATF3 and how these functions are carried out. Below, I will briefly review several biological processes reported to be regulated by ATF3.

Apoptosis: Quite a few studies reported the involvement of ATF3 in apoptosis using either cell culture model or mouse model, but the conclusions are rather confusing and conflicting. When transiently or stably expressed, ATF3 markedly promoted apoptosis of HeLa cells treated with etoposide or camptothecin (Mashima et al., 2001). It also significantly enhanced the ability of curcumin, an anticancer compound to induce apoptosis (Yan et al., 2005a). In vascular endothelial cells, ATF3 was rapidly induced by tumor necrosis factor

(TNF)alpha, oxidized low density lipoprotein (oxLDL), and lysophosphatidylcholine (LPC) and its expression mediated the cell death induced by these agents (Nawa et al., 2002). Consistently, transgenic mice expressing ATF3 also show functional defects in the corresponding tissues. 11

Mice expressing ATF3 in the heart displayed atrial enlargement, as well as atrial and ventricular hypertrophy, and the isolated cardiac myocytes had reduced contractility (Okamoto et al., 2001). When expressing ATF3 in the liver and pancreatic ductal epithelium, mice showed liver dysfunction and defects in endocrine pancreas development (AllenJennings et al., 2001; AllenJennings et al., 2002). The islets isolated from ATF3 knockout mice were partially protected from stressinduced apoptosis (Hartman et al., 2004). All the above evidence points to the function of ATF3 as a proapoptotic gene. However, there are also reports suggesting a protective effect of ATF3. For example, ATF3 protected cell death in cJun activated PC12 and SCG cells (Nakagomi et al., 2003). ATF3 can also enhance cJunmediated neurite sprouting (Pearson et al., 2003). Even in cardiac myocytes and endothelial cells, where ATF3 was demonstrated to contribute to apoptosis, ATF3 was also found to have protective roles (Kawauchi et al., 2002; Nobori et al., 2002).

Cell proliferation: In addition to apoptosis, ATF3 was also associated with the regulation of cell proliferation. Similar to the situation in apoptosis, ATF3 was demonstrated to either promote or suppress cell cycle progression. In hepatoma cells, ectopic ATF3 expression induced DNA synthesis and cyclin D1 gene expression (Allan et al., 2001). ATF3 was shown to mediate the cell proliferation function of cmyc (Tamura et al., 2005). ATF3 can partially transform chick embryo fibroblasts by promoting proliferation under low serum concentration

(Perez et al., 2001). In HeLa cells, however, overexpression of ATF3

moderately suppressed cell cycle progression (Fan et al., 2002). In mouse embryonic fibroblasts, ATF3 expression promotes both cell death and cell cycle arrest, and suppress Rasmediated cell transformation and tumorigenesis (Lu et al., 2006b). In colorectal cancer cells, overexpression of fulllength ATF3 has antitumorigenic activities, as demonstrated from in vitro focus formation assay and in vivo tumorigenicity in a xenograft mouse model; while the expression of antisense ATF3 has the opposite effects (Bottone et al., 2005). Detailed analysis indicated that apoptosis is not affected by the overexpression of ATF3, suggesting other mechanisms, such as proliferation, are targeted by ATF3

(Bottone et al., 2005).

Metastasis: Emerging evidence indicates that ATF3 is involved in tumor metastasis. When the mouse ATF3 cDNA was first isolated by differential hybridization from mouse melanoma cell B16F10, transfection of ATF3 into the low metastatic clone F1 converted the parental cells from low into high metastatic cells (Ishiguro et al., 1996). Followup study by Tsuruo’s group using lossoffunction approach showed that for colon cancer cell HT29, ATF3 antisense oligonucleotide changed cell morphology to a rounder shape, inhibited the attachment, cell migration and invasion ability in vitro, and improved mouse survival in vivo (Ishiguro and Nagawa, 2001; Ishiguro et al., 2000). Consistently, by comparing cancer cell lines and surgically excised human colon cancer samples, they found ATF3 was expressed at higher levels in the cell lines derived from metastatic sites than in those from original tumor sites. Also ATF3

was expressed at higher levels in tumor specimens than in adjacent normal mucosa, especially in the tumors that had invaded the lymphatic ducts and/or the vessels, while there was not a significant relationship between ATF3 expression and the other pathological features (Ishiguro and Nagawa, 2000). Conversely, there are other reports suggesting a metastasisinhibitory effect of ATF3. For example, ATF3 was shown to repress matrix metalloproteinase 2 (MMP2) expression in different cell types either directly or indirectly in response to different stress signals (Chen and Wang, 2004; Stearns et al., 2004; Yan et al.,

2002). As an important metastasispromoting factor, the repression of MMP2 by

ATF3 suggests ATF3 may inhibit tumor metastasis. In consistent with these findings, when treating breast cancer cells HCC1395 with Genistein, one of the major isoflavones that potently inhibits the growth and metastasis of breast cancer, ATF3 was strongly induced while MMP2, MMP7, and CXCL12 were significantly downregulated (Lee et al., 2007). Bottone and colleagues found that in HCT116 human colorectal cancer cells, ATF3 was upregulated by treatment with nonsteroidal antiinflammatory drugs (NSAID), troglitazone, diallyl disulfide, and resveratrol. Overexpression of ATF3 inhibited invasion, whereas antisense ATF3 increased invasion (Bottone et al., 2005).

Angiogenesis : Before this thesis work, only a few studies implicated the involvement of ATF3 in angiogenesis. By applying the cDNA microarray analysis on nontubulogenic and tubulogenic endothelial cells in a variety of culture conditions to look for genes upregulated in tubulogenic conditions, Kobayashi et

al. found ATF3 as a highscore gene in the tubulogenic cells (Kobayashi et al.,

2004), suggesting ATF3 may contribute to angiogenesis. On the other hand, the repression of MMP2, a welldocumented angiogenesispromoting factor, by ATF3

(Chen and Wang, 2004; Stearns et al., 2004; Yan et al., 2002) implicates ATF3 may inhibit angiogenesis.

1.1.E. Summary

As discussed above, it seems difficult to conclude the functions of ATF3 in different biological processes, given all the confusing and conflicting results from current literature. One major goal of this thesis work is to resovle these contradictory results derived from vastly different cell lines and models based on the hypothesis that ATF3 may regulate the same biological process in opposite directions depending on the signals to induce ATF3 and the cellular context. To test this hypothesis, we choose to focus on one maladaptative disease model — breast cancer, the development of which involves multiple biological processes, including cell proliferation, apoptosis, cell motility, angiogenesis and metastasis.

Since ATF3 was implicated in all these biological processes, breast cancer provides a uniform platform to study the mechanisms and functions of ATF3.

The second half of this introduction will review some current knowledge on breast cancer, including the anatomy and normal development of mammary gland, the epidemiology, cause, evolution, classification and multistage development of breast cancer, the contribution of stromacancer interaction in breast cancer

development, the relationship between breast cancer development and adaptive response, and the dichotomous phenomenon in cancer development.

1.2. Overview of mammary development and tumorigenesis

1.2.A. Mammary gland anatomy and development

The mammary gland is a structurally dynamic organ, varying with age, menstrual cycle and reproductive status (Osborne, 2000). The major function of mammary gland is to produce milk to nourish the offspring. Structurally, it is a branched tubuloalveolar gland embedded within a heterogeneous connective tissue (Stingl et al., 2005). In human, the mature female breast tissue contains approximately 15 to 20 lobes with varying numbers of ducts and lobules surrounded by a fibrous connective tissue sheath. Each lobe connects to a lactiferous duct, several of which converge to form a lactiferous sinus or milk chamber. These sinuses empty into the nipple, which is surrounded by a pigmented area, the areola. A layer of adipose tissue surrounds the breast glands and constitutes the interlobular stroma. Beneath the breast tissue lies the muscle of chest wall and in between there is a layer of connective tissue known as fascia. The breast is attached to the chest wall by fibrous strands called

Cooper’s ligaments (Figure 1.5) (Osborne, 2000). Microstructurally, each lobule is composed of approximately 200 alveoli (acini) surrounded by a connective tissue sheath. Alveoli are saclike structures lined with a single layer of milk secreting luminal epithelial cells surrounded by contractile myoepithelial cells

(Figure 1.5). 16

The deveopment of human mammary gland is a multiphase process starting from the fourth week of gestation (embryonic phase) as a pair of milk lines in the epidermis of the thoracic region. The milk lines then grow downward into the underlying mesenchyme and gradually branch into a rudimentary ductal tree at the time of birth, which will maintain largely dormant till the onset of puberty. At the onset of puberty (adolescent phase), the changing of hormonal environment leads to rapid growth and development of the mammary gland in females. In male, testosterone acts on the mesenchymal cells to inhibit further growth of the mammary gland, which will not be further discussed in this thesis. The major changes occurring during puberty in female include a size increase of the gland mainly due to the deposition of interlobular fat, extension and branching of mammary ducts into the fibrous stroma, and formation of the lobule, also called terminal ductal lobular unit (TDLU). The mammary glands remain in this mature but inactive state until pregnancy, the next major hormonal change. During pregnancy (pregnancy phase), the mammary gland experiences its most rapid phase of proliferation, including the formation of prospective secretory luminal epithelial cells and contractile basal myoepithelial cells, differentiation of milk producing alveolar cells from terminal end buds and thinning of the adipose tissue and fibrous stroma. In the course of lactation (lactation phase), the functional gland architecture is maintained and the secretory alveolar cells become fully active, producing a large amount of milk. After lactation, the alveolar cells undergo massive apoptosis—involution (involution phase), and the ducts regress back to a resting state. The stromal fibroblasts reconstruct the

interlobular connective tissue and the gland is redeposited with adipose tissue.

With the onset of menopause (menopause phase) and reduction of circulating ovarian hormones, the mammary ductal elements degenerate and dense connective tissue replaces the intralobular loose connective tissue (Howard and

Gusterson, 2000; Russo and Russo, 2004). The major stages of human breast development are summarized in Figure 1.6 (Parmar and Cunha, 2004).

Over the decades, people have been using rodents, especially mice, as a model to study mammary gland development and carcinogenesis, and have acquired unparalleled insights. Similar to human mammary gland, the mouse mammary gland undergoes multistage development mostly during postnatal life

(Figure 1.7) (Hennighausen and Robinson, 2005). But there are also significant differences between human and mouse mammary tissues, sometimes making the extensive extrapolation of results problematic (Hovey et al., 1999). First, the number and location of mammary glands vary among different species. In mice, five pairs of mammary glands run in an anterioposterior direction ventrally between the fore and hind limbs, whereas only one pair develop in the thoracic region in humans (Veltmaat et al., 2003). Second, and more importantly, the cellular composition and architecture of mammary fat pad (or mammary stroma) are different. In postnatal mouse, the mammary fat pad consists primarily of adipocytes with small amounts of interspersed fibrous connective tissue; while in human, there is far more fibroblastic connective tissue compared to the white adipose tissue. As a consequence, the lobuloalveolar structures of mouse mammary gland are in close contact with fat tissue but those of human breast

are embedded in fibrous connective tissue, away from the adipose tissue (Figure

1.8)(Parmar and Cunha, 2004).

1.2.B. Epidemiology of breast cancer

Since the breast tissue undergoes constant remodeling and differentiation in response to the hormonal flucturations during the menstrual cycle and pregnancy, breast epithelial cell are at a higher risk of accumulating mutations that may contribute to breast cancer development. Currently, breast cancer is the most common cancer among women in the United States, and is the second leading cause of cancer death in women, after lung cancer (Jemal et al., 2008).

In year 2008, it is estimated that about 182,460 women in the United States will be diagnosed with breast cancer and 40,480 women will die from the disease, accounting for 15% of all female cancer deaths (Jemal et al., 2008). In addition, far from being a disease restricted to the elderly, breast cancer is the most common cause of death from cancer among women between the ages of 25 to

54 years old (Chodosh, 1999).

1.2.C. Causes, evolution and heterogeneity of breast cancer

The causes for breast cancer are not known exactly, though several risk factors have been associated with increased incidence of the disease, including gender (female), old age, a positive family history, race (Caucasian), drinking alcohol, smoking, and overweight/obese. On the molecular level, several alterations correlate with breast cancer development. First, the aggregation of 19

breast cancer in families suggests hereditary factors may contribute to cancer development. Germline mutations in highpenetrance cancer susceptibility genes, including BRCA1, BRCA2 and TP53, account for less than 25% of total breast cancers, indicating variations in other moderate or lowpenetrance genes may contribute to the majority of breast cancers (Sellers, 1997). Several large scale genomewide association studies have identified some genetic variations associated with higher risk of breast cancer, such as SNPs in fibroblast growth factor receptor 2 (FGFR2), thymocyte selectionassociated high mobility group box 9 (TNRC9), mitogenactivated protein kinase kinase kinase 1 (MAP3K1), lymphocytespecific protein 1 (LSP1) (Easton et al., 2007; Hunter et al., 2007), caspase 8 and transforming growth factor β1 (TGFB1) genes (Cox et al., 2007).

Another survey of all published cancer genes (including breast cancer) identified

291 genes for which there are molecularly characterized mutations and evidence of a causative role in tumorigenesis, representing more than 1% of the human genome (Futreal et al., 2004). Second, besides the genetic mutations, the overall genomic instabilities such as chromosome aberration, including translocations, deletions, amplifications and fusions, have been observed in and correlated with breast cancer (Venkitaraman, 2004) (Rooney et al., 1999). Third, although genetic alterations are the major driven force for breast cancer development, epigenetic regulations also play important roles in breast cancer development (Asch and BarcellosHoff, 2001; Mielnicki et al., 2001).

Although various alterations on the genetic, epigenetic and/or genomic levels correlate with and may contribute to breast cancer development, it is not clear

how breast cancer evolves from cells with these alterations. More importantly, as a result of the evolutionary process, breast cancer, similar to most other cancers, is not a single disease, but is highly heterogenous at both molecular and clinical levels (Perou et al., 2000; Sorlie et al., 2001). Two distinct, but not necessarily mutually exclusive hypotheses have been proposed to explain the evolutionary pathways and extreme heterogeneity of cancers: cancer stem cell hypothesis and clonal evolution hypothesis. The cancer stem cell hypothesis suggests only a small population of cancer cells within the tumor that have the capacity of self renewal and differentiation — cancer stem cells — can initiate and maintain the tumor development. The heterogeneity of the tumor is due to the various pathways that the initiating stem cells take, leading to aberrant differentiation.

While the clonal evolution hypothesis suggests that most cells can proliferate extensively and the cells with the right genetic and epigenetic alterations are selected to contribute to tumor progression. This hypothesis proposes that the heterogeneity of tumor comes from the competition and selection of cells with different phenotypes under various conditions encountered by the cells during cancer development (Polyak, 2007; Reya et al., 2001). Up to now, cells with stem cell activities have been successfully isolated from various human cancers, including leukemia (Bonnet and Dick, 1997), colon cancer (O'Brien et al., 2007), breast cancer (AlHajj et al., 2003), brain tumor (Singh et al., 2003), and heat andneck carcinoma. However, there are multiple evidence to support each model (Campbell and Polyak, 2007), and it seems that breast tumor heterogeneity might be caused by a combination of both mechanisms, i.e., the

cancer stem cells may undergo clonal selection in the process of selfrenewal and differentiation, with some progeny cells acquiring the genetic or epigenetic alterations to regain selfrenewal capability and other cancerpromoting traits.

These cells will dominate the tumor, and continued mutations and selections of cell populations may lead to heterogenous cancer cells that can selfrenew and differentiate into all tumor cell types (Campbell and Polyak, 2007).

1.2.D. Classification of breast cancer

The history of breast cancer research is also the one of developing various methods to classify breast tumors. With the advancement of research techniques, the classification methods have evolved from the traditional histopathology, to molecular pathology, genetic analysis, and gene expression profiling (Sims et al., 2007). Traditional histopathology classifies breast cancers into ductal carcinoma, lobular carcinoma, nipple and others, based on the site where the cancer starts (Morrow and Harris, 2004). Molecular pathology applies immunohistochemistry (IHC), immunofluorescence (IF) and fluorescence in situ hybridization (FISH) methods to detect various cellular antigens, such as estrogen receptor (ER), progesterone receptor (PR) and HER2. Genetic analysis examined the intracellular alterations on both the genomic level (chromosome) and individual gene level. Based on the comprehensive gene expression profiles of large sets of tumors, breast cancer is divided into five subtypes at molecular level: basallike, luminal A, luminal B, Her2+/ER and normal breastlike (Hu et al., 2006; Perou et al., 2000; Sorlie et al., 2006). Various classification methods 22

provide important information on tumor behavior (prognosis) and responsiveness to cancer therapies. For example, basallike tumors have the worst clinical outcomes and luminal Atype tumors the best (Sorlie et al., 2001); ER and/or PR positivity determines the likelihood of cancer to endocrine therapies.

1.2.E. The multistage development of breast cancer and cancer metastasis

Despite the extreme heterogeneity, in order for a cell to become cancerous, it needs to acquire six physiological traits in one way or another: selfsufficiency in growth signals, insensitivity to antigrowth signals, evasion of apoptosis, acquisition of a limitless replicative potential, sustained angiogenesis, and tissue invasion and metastasis (Hanahan and Weinberg, 2000). In the process of acquiring these cellular alterations, the breast cancers, as well as many other solid tumors derived from epithelial cells, exhibit morphologically distinct stages from normal epithelium to carcinoma in situ , invasive carcinoma and metastasis

(Figure 1.9).

As the final step of breast cancer development, metastases, rather than primary tumors, are responsible for most cancer deaths. Similar to cancer development, the metastatic process also consists of a series of steps, including angiogenesis within primary tumor, local invasion of cancer cells, intravasation into the circulatory blood/lymph system, arrest in secondary sites, extravastation from the circulation into the surrounding tissue, colonization and proliferation to give rise to clinically detectable metastasis (Figure 1.10)(Fidler, 2003). It has been known for a long time that metastasis is inherently an inefficient process 23

(Weiss, 1990). Accumulative studies using in vivo videomicroscopy technology in “experimental metastasis” and “spontaneous metastasis” model system have identified varying efficiencies of specific steps in metastasis (Figure 1.11). The intravasation, survival and growth of cells after extravasation are inefficient and the corresponding efficiency depends on the malignancy of the cancer cells; whereas the survival of cancer cells in the circulation and extravasation processes are quite efficient, independent of the malignancy of the cancer cells

(Chambers et al., 2000). Consistently, though breast cancer cells can be detected in blood, bone marrow or lymph node, especially with the advancement of various detecting techniques, the prognostic and clinical significance of these cells has been called into question (Allred and Elledge, 1999; Funke and

Schraut, 1998). Therefore, targeting the inherently inefficient, ratelimiting steps may significantly reduce cancer metastasis and improve patient outcome.

1.2.E. Stromalcancer interaction

The development of breast cancer is not only driven by intracellular alterations, such as those on the genetic and epigenetic levels, but also significantly affected by extracellular environment — the mammary stroma.

Breast tissue is composed of two compartments, epithelium and stroma. These two compartments do not exist in isolation from each other, but closely communicate and significantly affect the behavior of each other. This bi directional communication plays an essential role in various aspects of mammary gland physiology and pathology, including breast tumorigenesis (Howlett and 24

Bissell, 1993; Neville et al., 1998). Both in vitro and in vivo studies have demonstrated that cells composing the microenvironment, including fibroblasts, leukocytes, endothelial cells, adipocytes, and molecules in the extracellular matrix (ECM) can modulate the morphogenesis and development of normal breast, as well as the growth, survival, polarity, and invasive behavior of breast cancer cells (Bissell and Radisky, 2001). For this thesis work, I will briefly review the crosstalk between breast cancer cells and an important infiltrating leukocyte component, macrophages.

Macrophages are a terminally differentiated cell type of the mononuclear phagocyte system (Leek and Harris, 2002). Two functional phenotypes of macrophages have been proposed based on the secreted cytokine profiles and phagocytic ability: M1 and M2 (Mantovani et al., 2002). M1 macrophages are highly phagocytic and mainly secrete proinflammatory cytokines such as interleukin (IL)1, IL6, IL12 and tumor necrosis factoralpha (TNFα), while M2 macrophages secrete wound healing and antiinflammatory cytokines such as IL

10 and transforming growth factorbeta (TGFβ) (Mantovani et al., 2002).

In human breast cancers, macrophages are a major component of infiltrative leukocytes, and in some cases, they comprise as much as 50% of the tumor mass (Lewis and Pollard, 2006). Such macrophages that enter the tumor parenchyma are referred to as tumorassociated macrophages (TAMs). TAMs are derived from circulating monocytes and recruited to the tumor microenvironment by chemotactic factors produced mainly from the neoplastic cells, including cytokines such as macrophage colony stimulating factor (MCSF,

also known as CSF1), vascular endothelial growth factor (VEGF), platelet derived growth factor (PDGF) and chemokines such as CCL2, CCL5, CCL7,

CXCL12 (Mantovani et al., 2006). The differentiation of incoming monocytes to macrophages are mediated mainly through the tumorderived MCSF (Sica et al.,

2008). TAMs have been characterized as a M2polarized population (Biswas et al., 2006) and their major functions in tumor progression include the followings: promoting the proliferation of tumor cells by secreting mitogenic cytokines, such as EGF, PDGF, TGFβ and TGFα (Leek and Harris, 2002); enhancing migration and metastasis by producing ECMdegrading enzymes (Pollard, 2004); stimulating angiogenesis to help cancer survival in hypoxic conditions and metastasis (Dirkx et al., 2006); evading immune surveillance due to suppression of antitumor immunity (Leek and Harris, 2002) (Figure 1.12).

1.2.F. Breast cancer, adaptive response and “JekyllandHyde” conversion

As discussed above, the development of breast cancer can be viewed as a process of continuous interaction between neoplastic cells and stromal microenvironment, and persistent adaptations of neoplastic cells to the selective pressures they encounter. During cancer development, the cells encounter many stress signals, including genotoxic damages, inappropriate activation of oncogenes and inactivation of tumor suppressor genes, telomere erosion, and hypoxia and oxidative stress in the tumor microenvironment (Evan and Vousden,

2001). All along, the cells have builtin mechanisms to restrain or eliminate themselves (Hanahan and Weinberg, 2000). The successful cancer cells are 26

those well adapted to the environmental stress by developing multiple ways to breach the hardwired anticancer mechanisms within the cells and tissues

(Hanahan and Weinberg, 2000). One phenotype supporting the adaptive development of cancer is that some of the genes that under normal conditions would function to eliminate the cancer cells — the tumor suppressors — appear to be coopted to become oncogenes. Two welldocumented “JekyllandHyde” molecules in cancer development are tumor necrosis factor alpha (TNFα) and transforming growth factor beta (TGFβ). Here I will only briefly review the dichotomous conversion of TGFβ during cancer development, which is closely related to my thesis work.

TGFβ is a multifunctional cytokine that can potently regulate cell growth, differentiation and cell migration (Derynck et al., 2001; Derynck and Zhang, 2003;

Siegel and Massague, 2003). It signals through heteromeric cell membrane receptors to activate downstream Smaddependent and Smadindependent pathways (Figure 1.13)(Bierie and Moses, 2006). Functionally, it induces apoptosis or cell cycle arrest in normal or less transformed cells, but facilitates survival and metastasis in more advanced tumors (Figure 1.14) (Derynck et al.,

2001; Roberts and Wakefield, 2003; Siegel et al., 2003). A partial explanation for the functional alterations of TGFβ during cancer development is that many components in the TGFβ pathway are mutated, leading to the loss of growth inhibitory response to TGFβ (Levy and Hill, 2006). But for some cancers, such as breast cancer, mutations in TGFβ receptors or downstream signaling molecules are rare (Anbazhagan et al., 1999; Tomita et al., 1999). Smad3 has

been shown to at least partially mediate the dichotomous effect of TGFβ (Tian et al., 2003). Since TGFβ can elicit its effects through either tumorcellautonomous mechanisms or hosttumor interactions (Bierie and Moses, 2006), it is important to figure out how tumor cells interpret TGFβ signal differently from normal epithelial cells and how stromacancer interactions contribute to the “Jekylland

Hyde” conversion of TGFβ.

1.2.G. Conclusion

Breast cancer is a highly heterogeneous disease. Its development involves continuous adaptive responses of the mammary epithelial cells to various stress stimuli and stromal alterations. To understand how successful cancer cells manage to foil the hardwired stress responses that they normally use to eliminate themselves has been and will continue to be an important issue to be addressed in cancer research. Specifically, for dichotomous cytokines, such as TNFα and

TGFβ, what are the downstream signaling mechanisms in cancer cells that lead to altered functional readouts? To address these questions, we focused on an adaptive response gene—ATF3. By examining its functions and mechanisms in breast cancer, we hope to gain more insight on how adaptive responses evolve in tumor development and how we can utilize our understanding in cancer therapy. Chapter 2 of this thesis presents our finding on ATF3 as a dichotomous molecule in breast cancer development. Chapter 3 provides evidence that ATF3 may function to mediate TGFβ signaling in cancerous epithelial cells. Chapter 4 explores the potential involvement of ATF3 in stromacancer interaction, 28

specifically, TAMcancer interaction. Chapter 5 examines the physiological relevance of ATF3 to human breast cancer. The last chapter (Chapter 6) discusses some of the future perspectives I proposed for ATF3 research in cancer biology.

Subgroup Members Alternative Names CREB CREB CREB1, CREB327, ATF47 CREM ATF1 TREB36, TCRATF1, ATF43 ATF2, CREB2, HB16, TREB, TCRATF2, CREBP1 CREBP1 mXBP ATFa ATF7 CREBPa ATF3 ATF3 LRF1, LRG21, CRG5, TI241 JDP2 ATF4 ATF4 CREB2, TAXREB67, mATF4, C/ATF, mTR67 ATFx hATF5 ATF4L1 ATF6 ATF6 ATF6α CREBRP G13, ATF6β, CREBL1 BATF BATF JDP1 p21SNFT, DNAJC12

Table 1.1 A partial list of the mammalian ATF/CREB family of transcription factors. The subgroups were established based on homology of the protein sequences (Taken from Hai, 2006a).

1.1 A partial list of the mammalian ATF/CREB family of transcription factors

Animal Models Culutured Cells Tissue Treatment Cell Type Treatment Liver Partial hepatectomy, Hepatocyte Cycloheximide, EGF, HGF alcohol, carbon tetrachloride, ischemia, acetaminophen Heart Ischemia, ischemia Fibroblasts UV, IR, MMS, serum, reperfusion anisomycin, E1A Kidney Ischemiareperfusion Myeloid cells Fas antibody Brain Seizure Macrophages Cytokines, LPS, BCG, PMA Skin Wounding Neuroblastoma Forskolin, FGF Thymus AntiCD3ε Epithelial cells Cytokines, hypoxia, Peripheral Axotomy anisomycin, UV, IR, anti nerves cancer drugs, cycloheximide Leukemia cells doxorubicin

Table 1.2 A partial list of treatments that induce ATF3 expression. EGF: epidermal growth factor; HGF: hepatocyte growth factor; UV: ultraviolet; IR: irradiation; MMS: methyl methanesulfonate; LPS: lipopolysaccharide; BCG: Bacillus CalmetteGuérin; PMA: paramethoxyamphetamine; FGF: fibroblast growth factor (Derived from (Hai et al., 1999).

1.2 A partial list of treatments that induce ATF3 expression

Fig. 1.1 Dendrogram representation of sequence homology within the bZip domain of transcription factors (Taken from (Hai, 2006a).

1.1 Dendrogram representation of sequence homology within the bZip domain of transcription factors

1 167 411 519 717 1914 mRNA ABCE

CAAAUG GAG GUA AAA GAG AGCUAA M E VK ES * Protein Basic ZIP 1 80 116 181

Fig. 1.2 Schematic representation of the mRNA and protein structure for ATF3. Exons in mRNA are indicated by boxes labeled as A, B, C, and E. Functional domains within protein are indicated by boxes, with the basic region and leucine zipper (ZIP) domains labeled. The codons and amino acids at the border of each domain are indicated (Taken from Liang et al., 1996).

1.2 Schematic representation of the mRNA and protein for ATF3

A. B. ATG TAA Long Isoform 1 A’ B C E Basic Leucinezipper Fulllength Long isoform 1, ATF3 Fulllength 181 aa ATF3 A BCETAG Zip 118 aa Zip ADBCE 135 aa TGA Zip2a, b Zip2a 106 aa ADBCED’ E’ Zip2c TGA

Zip2b Zip3 120 aa A BCED’ D TGA ATF3b 124 aa Zip2c A B1B2 C D’ D TAG Zip3 ABC C’ D’ D ATG TAA ATF3b A B3B4 CE Fig. 1.3 Schematic representation of the ATF3 gene, the mRNA species generated by alternative splicing and the corresponding protein structures predicted. A. The genomic structures of different alternatively splicing isoforms compared with that of fulllength ATF3. Exons are indicated by boxes, and the translation start sites (ATG) and translation stop codons (TAA, TAG, TGA) for each isoform are shown. B. The predicted protein structures for each of the ATF3 isoforms are shown, with the basic and leucine zipper regions and total protein length in amino acids (aa) indicated (Hartman, 2005).

1.3 Schematic representation of the ATF3 gene, the mRNA species generated by alternative splicing and the corresponding protein structures predicted predicted

Myc/Max AAGCTTCACGTGTTCT

2040 CCCTCCTCTCCTTGCTTCACTTTATAATGGTGCTATTTATTCCAGAACAATCTATAAGTAGATAAAATAG

1970 CTAAGTAGAGATAACAAATAACTTCATTCAAATGCAAACACTCCTCCACCTAATCCCGCCCGGTGTCCGC 1920 CGGGCTGCTCCGACACGCCCGGGGTTTACCTGCGCGCACTCCAGCGGGAGGGCGGGTTGTGGAGGTGTGC 1850 TGAGCGGCGCGCGGGGGTGAGGGCGTGGAATCTGAGGGTGGGGCCCGGAGAGCCGTTACCAGGGCGAAAA Smad 1780 GTAAAGCGAAAACACCCGCCCTGCACTTCCCGCGCGACGCCGCTGGAAATCGGTTCAGGTCCAGAGCAGG 1710 ATCTCGGAGGATCCCGCGTGGAACCCCAGGGCTCCCGGGTCCGCCGGGGCGCAAAGACTTCCGAGGCCGC Myc/Max Myc/Max AP1 NF?B 1640 CCTCCGCGTGTTCCCAGGCCCGTGGAGAGGTGGGTGGTCTGAGTAAGGTCGGGCTTGGCGGCGAGGAACC NF?B Smad 1570 CCGGTGGGGGGAACTGGGGACTTCAAGTGAGACCCAGGATCCAGACACCTCTAGTTTCTACCCCAAATTA

1500 ACCAAACTGTGACCTTGGGCCGCTCTCTCCCAGAGGCAGGTGGAAAGAAGCAGGTGTTTCTGCCCTTCAC E2F 1430 CGTGCCCCCACACCCTGCGGCCGCGCAGGTCTCCCTCCCAGGCAGGTGCGAAAGTCCCAGGCCACACTTG AP1 1360 TCTACAAATAGTCATCCACGGGAGTAAGAAGGTTCCTTGGTTCTGCCGCTCTCTGAGCAGAAATTGTTGG

1290 GGTCGGGGAATAAGAACCAGGAAATCGTTTTTAAGGTTCAAACCCAGTTCTGCTGAGGTCTCAGCTCGAA E2F 1220 TCTCGGACCACGGGGCCCCGCCTTTCCCGCCACCCTGGCTTGAGGGCAGAGGGGATTTCTGCTGCGGGTT Myc/Max 1150 CCGCCTGTGGTCAGTGCGTCCCCATTCCGGGCCGTCCGGTCCCAGTCCAATCGGCTCTGGAGCGAGAGAC 1080 GTGAAAGCTGAAGATGGTTTTCCCTAAATATGCCTGAGACGGCACCCCAGGCCTGGGCAGGTTCGCGGAC

1010 CCCAAAGCACCTTCTTCTTTCCCCCTCCTCCTGGCCGCTGGCTTCCGCCCCCTCCTACCCTCCCACCGGG E2F 940 TTGCCTCTGATTCCTCCTGGACTCCGATCTTTTCACGCTCTTGTTGGTTTACTGACAGTTCTTGTCAATT 870 TCAAACGCTTTGTGATTGTAAAAAAAAAAAAATCGAACCGATACGGTCCTACCACTCGCCCTAGTTTCGG Myc/Max 800 AGCCCGGAGCTGTCCTGCGTGTGCGTCCATGTGGAGTGTCCGGGGCTGCGGGCTCGGGCGCACGGTGCCA

730 GCCGAGGGCTGCCCTCCGCTTTGTGTTAACCGGCGGGCTTCTCGGTCCCCGCCGCAGAGGTCACACCCGG AP1 660 CGGGTAACGGAGTGGATACACCGAAGGGTGACTTTGGACACCTTCCCCACACCACAGACTAACGCTTCTG

590 CCCCTACTCCGCCCCTGCTAGAGAAGTAGGAGGCCAATTGGGGAGGGGGTTATTTTCCTGAAGCTCCAGA

520 AAATGACCACGTATTTTAGAGAAAGGTTCGTGCCCGCTTCCCAGCCTCACCTAGTCTGGGCTGGGGCCGG

450 GACCCGCCTCCCCACCTTCCCCGCCCCCCCCGCTCTTCAACCTAGCGGAGGGACAGATGCCAGCGCGGTG p53 p53 380 GAGTCATGCCGCTGGCTTGGGCACCATTGGTCATGCCTGGAACACGCAGCAGCGAGTACGCACATCTGGC 310 GGCTATCCCGGGCGGCTCCGGTCCTGATATGGAGAGAGAGGGCGGGCTGGTGTGTGTCTCAGTGAGCGAG

240 GGCGGGGGAACGCGCCTGGGCTGGCTCCTCCCCGAACTTGCATCACCAGTGCCCCCTCTCTCCACCCCTT 170 CGGCCCCGCCTTGGCCCCTCCTCCACCCCCCTTCCTCCGCTCCGTTCGGCCGGTTCTCCCGGGAAGCTAT TATA ATF/CRE 100 TAATAGCATTACGTCAGCCTGGGACTGGCAACACGGAGTAAACGACCGCGCCGCCAGCCTGAGGGCTATA Box ATF/CRE 30 AAAGGGGTGATGCAACGCTCTCCAAGCCACAGTCGCACGCAGCCAGGCGCGCACTGCACAGCTCTCTTCT +1 +41 CTCGCCGCCG

Fig. 1.4 Nucleotide sequence of the 5’flanking region of ATF3. The TATA box and several transcription factorbinding sites are underlined and labeled. The arrow marks the transcription start site (+1) (Modified from Liang et al., 1996).

1.4 Nucleotide sequence of the 5’flanking region of ATF3

B. Lobule A.

(Taken from http://www.breasthealthy.com/, http://www.yoursurgery.com/, and http://classes.ansci.uiuc.edu/ansc43 8/Mamstructure/histology_2.html)

Fig 1.5 Anatomy of human breast. A. A drawing of human breast tissue with different structures indicated. B. A schematic view of a mammary lobule. C. A crosssection drawing of alveoli/acini within a lobule.

1.5 Anatomy of human breast

Fig. 1.6 Drawings depicting the development of the human breast, the basic functional unit of which is the terminal ductal lobular unit (TDLU). (A) Emerging of a TDLU from the extralobular terminal duct (ETD), ITD: intralobular terminal duct. (B) The histology of a normal TDLU in the resting nulliparous state, showing the terminal duct and associated acini. (C) The terminal branching structures from a 15yearold human breast. An extralobular duct is shown with rudimentary terminal branching structures. True lobules and alveoli/acini are not yet present. (D) TDLUs from a 22yearold nulliparous human mammary gland that well developed lobules and acini. (E) The breast structures of a 30yearold pregnant woman. Note well developed TDLUs. (F) The breast tissue of a 55 yearold parous menopausal woman. Ducts are dilated, and the TDLUs are atrophied (Taken from Parmar and Cunha, 2004). 37

1.6 Drawings depicting the development of the human breast, the basic functional unit of which is the terminal ductal lobular unit (TDLU)

Fig 1.7 Schematic (Aa–d) and wholemount (Ba–d) presentation of the different stages of mouse mammary gland development. A rudimentary ductal structure within the mammary fat pad is visible at birth (Aa). During puberty, the cyclical production of ovarian hormones promotes and accelerates ductal outgrowth (Ba). At this stage, conspicuous clubshaped structures (terminal end buds (TEB)) appear at the ductal tips. In the mature virgin, the entire fat pad is filled with a regularly spaced system of primary and secondary ducts, with side branches that form and disappear in each oestrous cycle (Ab, Bb). Hormonal changes that occur when pregnancy begins (the release of prolactin, placental lactogens and progesterone) increase cell proliferation and the formation of alveolar buds (Ac, Bc), which grow and differentiate into milk secreting alveoli at the end of pregnancy (Ad, Bd). During lactation, alveoli are fully matured and the luminal cells synthesize and secrete milk components into the lumina (Modified from Hennighausen and Robinson, 2005).

1.7 Schematic (Aa–d) and wholemount (Ba–d) presentation of the different stages of mouse mammary gland development

Fig. 1.8 Comparison of human and mouse mammary glands. (A) Hematoxylin & eosin (H&E) stained section of human breast tissue showing a terminal ductal lobular unit (TDLU) embedded in a fibrous connective tissue stroma. (B) Schematic representation of a human TDLU, emphasizing the intimate association of epithelial structures with interstitial fibrous connective tissue stroma and the more distant adipose tissue. (C) H&E stained section of the mouse mammary gland, showing ducts imbedded in a stroma composed of adipose tissue. (D) Schematic representation of the mouse mammary gland, displaying ducts in intimate contact with fibroblasts and adipocytes (Taken from Parmar and Cunha, 2004).

1.8 Comparison of human and mouse mammary glands

Normal In situ Invasive Metastasis

Fig. 1.9 Multistage development of breast cancer. Upper panel, schematic view of normal, in situ , invasive, and metastatic carcinoma. Normal breast ducts are composed of the basement membrane, a layer of luminal epithelial cells and a layer of myoepithelial cells. Cells composing the stroma include various leukocytes, fibroblasts, myofibroblasts, and endothelial cells. In in situ carcinomas the myoepithelial cells are epigenetically and phenotypically altered and their number decreases, while the number of stromal fibroblasts, myofibroblasts, lymphocytes, and endothelial cells increases. Loss of myoepithelial cells and basement membrane results in invasive carcinomas, in which tumor cells can invade surrounding tissues and can migrate to distant organs, eventually leading to metastases. Lower panel, hematoxylin and eosin staining of normal breast (A), ductal carcinoma in situ (DCIS, B), invasive ductal carcinoma (IDC, C) and brain metastatic lesions(D), respectively (Modified from Polyak, 2007 and Fujita et al., 2005).

1.9 Multistage development of breast cancer

Primary tumor Angiogenesis Local invasion Intravasation

Metastasis Extravasation Arrest Transport

Fig. 1.10 Schematic representation of the metastatic process. Within the primary tumor, the nutrients were initially supplied by simple diffusion. Angiogenesis must occur if a tumour mass is to exceed 1–2 mm in diameter. The synthesis and secretion of angiogenic factors establish a capillary network. Local invasion of the host stroma by some tumour cells occurs preceding the entry of tumor cells into the circulation (intravasation). Thinwalled venules, such as lymphatic channels, offer very little resistance to penetration by tumour cells. Within the circulation, some tumor cells are rapidly destroyed. The survived tumor cells become trapped in the capillary beds of distant organs by adhering either to capillary endothelial cells or to subendothelial basement membrane that might be exposed. Extravasation occurs next — probably by mechanisms similar to those that operate during intravasation. Proliferation within the organ parenchyma completes the metastatic process. To continue growing, the micrometastasis must develop a vascular network and evade destruction by host defenses (Modified from Fidler, 2003).

1.10 Schematic representation of the metastatic process

Fig. 1.11 Efficiency of specific steps in metastasis and their dependence on the degree of malignancy of the cells (taken from Chambers et al., 2000).

1.11 Efficiency of specific steps in metastasis and their dependence on the degree of malignancy of the cells

Fig. 1.12 Protumoral functions of tumorassociated macrophages (TAM) and their interplay with tumor cells. Tumorderived chemotactic factors such as CCL2, MCSF and VEGF actively recruit blood monocytes to the tumor site, where they differentiate to resident macrophages. Several TAM products directly stimulate the growth of tumor cells (e.g. EGF, FGF, cytokines). TAM also contribute to the angiogenic switch by releasing angiogenic factors (VEGF, FGF, TGF β, chemokines) and to the degradation and remodeling of the matrix with the production of metalloproteases, MMPs. Inhibition of antitumor responses is achieved by the secretion of immunosuppressive cytokines, like IL10 and TGF β, and by selective recruitment of naive T cells, via the chemokine CCL18, and of Th2 and Treg, via the chemokines CCL17 and CCL22 (Taken from Sica et al., 2008).

1.12 Protumoral functions of tumorassociated macrophages (TAM) and their interplay with tumor cells

Fig. 1.13 General overview of the main signalling networks regulated by TGFβ in cancer. TGFβ ligands that are activated in the extracellular matrix or at the cell membrane bind the extracellular domain (EC) of type II TGFβ receptor (TβRII) homodimers with high affinity. The ligandbound TβRII complex then binds and transactivates a type I TGFβ receptor (TβRI), which results in the phosphorylation of a glycine serinerich region termed the GS box (GS). In carcinoma cells the predominant T RI receptor is activinlike kinase 5 (ALK5). The activation of ALK5, a serine threonine kinase, results in the activation of downstream pathways. The level of signalling through each downstream pathway is context and celltypedependent. The resulting net activation of downstream pathways in each cell type determines the response to TGFβ in vivo and in vitro . CDC42, cell division cycle 42; DAXX, deathassociated protein 6; EMT, epithelial–mesenchymal transition; MAP3K1, mitogenactivated protein kinase, kinase, kinase 1; PAK, p21activated kinase; PAR6, partitioningdefective protein 6; PI3K, phosphatidylinositol 3kinase; PP2A, protein phosphatase 2A; ROCK1, Rhoassociated, coiledcoil containing protein kinase 1; SMURF1, Smad ubiquitination regulatory factor 1;TAK1,TGF activated kinase 1; TM, transmembrane domain (Taken from Bierie and Moses, 2006).

1.13 General overview of the main signaling networks regulated by TGFβ in cancer in cancer

Fig. 1.14 TGFβ switches from tumor suppressor in the premalignant stages of tumorigenesis to prooncogene at later stages of disease leading to metastasis. Progression to metastatic disease is generally accompanied by decreased or altered TGFβ responsiveness and increased expression or activation of the TGFβ ligand. These perturbations, along with other changes in genetic or epigenetic context of the tumor cell and its stromal environment, combine to alter the spectrum of biological responses to TGFβ (Roberts and Wakefield, 2003).

1.14 TGFβ switches from tumor suppressor in the premalignant stages of tumorigenesis to prooncogene at later stages of disease leading to metastasis

CHAPTER 2

A POTENTIAL DICHOTOMOUS ROLE OF ATF3, AN ADAPTIVERESPONSE

GENE, IN BREAST CANCER DEVELOPMENT

2. A POTENTIAL DICHOTOMOUS ROLE OF ATF3, AN ADAPTIVE RESPONSE GENE, IN BREAST CANCER DEVELOPMENT

2.1. ABSTRACT

Activating transcription factor 3 (ATF3) is a member of the ATF/cyclic AMP response elementbinding (CREB) family of transcription factors. Its corresponding gene is induced by a variety of stress signals, including carcinogens and hypoxia. Intriguingly, ATF3 gene is localized in the human chromosome at 1q32, within the 1q amplicon — the second most frequently amplified region of solid tumors. I present evidence that ATF3 has a dichotomous role in breast cancer development. By both gain and lossof function approaches, I found that ATF3 enhances apoptosis in the untransformed

MCF10A mammary epithelial cells, but protects the aggressive MCF10CA1a cells and enhances its cell motility. Array analyses indicated that ATF3 upregulates the expression of several genes in the tumor necrosis factor (TNF) pathway in the MCF10A cells, consistent with its proapoptotic functions in these cells. In the aggressive MCF10CA1a cells, ATF3 upregulates the expression of several genes implicated in tumor metastasis, including TWIST1, fibronectin

(FN)1, collagen IVa2 (COL4A2), plasminogen activator inhibitor1 (PAI1), urokinasetype plasminogen activator (uPA), caveolin1 (CAV1) and Slug, partially explaining its effect on enhancing cell motility. I also present evidence that ATF3 binds to the endogenous promoters and regulates the transcription of the TWIST1, FN1, Snail and Slug genes. Given the fact that some of the ATF3 upregulated genes, such as FN1, uPA, COL4A2, PAI1 and SLPI, encode secreted molecules, by conditioned medium experiments, I found that ATF3 has a paracrine/autocrine effect. The oncogenic feature of ATF3 in MCF10CA1a cells in vitro prompted us to test its function in vivo. In an orthotopic xenograft mouse model, ectopic ATF3 expression leads to faster growth of primary tumor and more metastastic lesions in the lung tissue. By surveying a panel of human breast carcinoma cell lines, I found ATF3 expression is elevated in ~90% (16/18) of the cell lines examined, providing a correlative argument that it is advantageous for the malignant cancer cells to express ATF3, consistent with its oncogenic roles suggested by the in vitro MCF10CA1a cell data and in vivo mouse model.

2.2. INTRODUCTION

During cancer progression, the cells encounter many stress signals, including genotoxic damages, inappropriate activation of oncogenes, telomere erosion, and hypoxia in the tumor microenvironment (Evan and Vousden, 2001). All along, the cells have builtin mechanisms to restrain or eliminate themselves

(Hanahan and Weinberg, 2000). However, the successful cancer cells manage 47

to foil the hardwired stress response to eliminate themselves. In fact, emerging evidence indicates that some of the genes that under normal conditions would function to eliminate the cancer cells — the tumor suppressors — appear to be coopted to become oncogenes. Two examples of this dichotomy are TGFβ and

TNFα. TGFβ induces apoptosis or cell cycle arrest in normal or less transformed cells, but facilitates metastasis in advanced tumors (Derynck et al.,

2001; Massague, 2000; Roberts and Wakefield, 2003). Similarly, TNFα, despite its proapoptotic activity, promotes cancer progression when chronically induced in the tumor (Balkwill, 2002; BenBaruch, 2003; de Visser et al., 2006). This

“JekyllandHyde” conversion is an intriguing but largely unresolved issue in cancer biology. In this chapter, we present evidence suggesting that ATF3 is a new dichotomous molecule in cancer development.

As reviewed in Chapter 1, ATF3 is a member of the ATF/CREB family of transcription factors, which share the basic region/leucine zipper DNA binding motif and bind to the CRE/ATF consensus sequence TGACGTCA (Hai and

Hartman, 2001; Hai et al., 1999). ATF3 mRNA level is not detectable in most cells, but greatly increased by a variety of stress signals, including anoxia (Ameri et al., 2007), hypoxia, DNA damage and carcinogens (Hai and Hartman, 2001; Hai et al., 1999). By examining the signals that induce ATF3 expression, we propose that ATF3 is an adaptiveresponse gene that participates in cellular processes to adapt to extra and/or intracellular changes (Hai, 2006a; Lu et al., 2006a).

Two major clues prompted us to hypothesize that ATF3 may have a dichotomous role in cancer development. First, ATF3 has been demonstrated to

play a role in apoptosis and proliferation, two cellular processes critical for cancer progression. However, the literature shows conflicting results: ATF3 can either promote or suppress these processes (Hai, 2006a). Second, Massagué and colleagues demonstrated that TGFβ induces the expression of ATF3; its gene product in turn interacts with Smad3, resulting in a functional repressor complex on the Id1 promoter (Kang et al., 2003). Since TGFβ is wellknown for its dichotomy, this in combination with the conflicting data on ATF3 in the literature promoted us to hypothesize that ATF3 has a dichotomous role in cancer progression.

One drawback of the literature on ATF3 is that it was derived from vastly different cell lines and models, making it difficult to deduce the reasons for the contradictory results. Thus, we selected an isogenic breast epithelial cell system developed by Miller and colleagues (Santner et al., 2001) to test the dichotomy hypothesis. This cell system contains four human breast epithelial cell lines:

MCF10A, MCF10AT1k, MCF10CA1h and MCF10CA1a. MCF10A is a spontaneously immortalized but nontransformed mammary epithelial line and exhibits many characteristics of normal breast epithelium (Debnath et al., 2003;

Soule et al., 1990). The MCF10A cells were transformed by oncogenic Ras, followed by selections via xenograft injection and passages in culture to give rise to the premalignant MCF10AT1k cells. The MCF10AT1k cells were then used to make the lowgrade malignant MCF10CA1h and highgrade malignant

MCF10CA1a cells by repeating several rounds of selection through in vivo xenograft injection and in vitro culture (Santner et al., 2001). We will refer to

MCF10A as MI cells, MCF10AT1k as MII, MCF10CA1h as MIII, and

MCF10CA1a as MIV cells, following the previous nomenclature (Tang et al.,

2003). Detailed characterization of each cell line suggested that they represent different stages of breast cancer development, therefore offering “a unique tool for the investigation of molecular changes during progression of human breast neoplasia and the generation of tumor heterogeneity on a common genetic background” (Santner et al., 2001). The properties of each line are summarized in

Table 2.1.

This chapter documents that ATF3 enhances apoptosis in untransformed cells but inhibits apoptosis and promotes cell motility in aggressive cancer cells. It also demonstrates that ATF3 regulates different target genes in untransformed cells versus in aggressive cancer cells, and that it has a paracrine/autocrine function by regulating genes encoding secreted factors. In an orthotopic xenograft mouse model, ATF3 not only promotes primary tumor growth, but also increases lung metastasis. Finally, ATF3 expression is elevated in ~90% (16/18) of the breast carcinoma cell lines examined, providing a potential pathophysiological relevance of ATF3. Taken together, this work is the first to suggest a dichotomous role of

ATF3 and the first to demonstrate that the degree of malignancy of the cells affects ATF3 function.

2.3. MATERIALS AND METHODS

Cell lines

MCF10A (MI), MCF10AT1k (MII), MCF20CA1h (MIII), and MCF10CA1a (M

IV) cells were purchased from the Barbara Ann Karmanos Cancer Institute

(Detroit, MI, USA) and cultured according to the instructions from the suppliers.

Retroviruses and siRNAs

Retroviruses were generated by introducing pBABEpuro or its derivative expressing ATF3 with VSVGexpressing plasmid into pantropic packaging cell line GP2293 (Clontech, CA, USA). Viral transduction was carried out in the presence of 8 g/ml polybrene, followed by antibiotic selection at 48 hr later for 1 to 2 wk to generate stable cells. The ATF3 siRNAs containing a pool of four siRNA duplexes and the control siRNA (Dharmacon, CA, USA) were transfected into the indicated cells according to the instructions from the manufacturers.

Cell viability assay, flow cytometry, immunoblot and immunofluorescence

Cell viability was measured by crystal violet assay as detailed previously (Lu et al., 2006b). Flow cytometry was carried out after fixing and staining the cells with propidium iodide, using the BD FACS Calibur TM system (BD Bioscience,

San. Jose, CA, USA ). Immunoblot was carried out using antiATF3 (Santa Cruz,

CA, USA), antiactin (Sigma, MO, USA), antiactivated caspase 3 and anti caspase 8 (Cell Signaling, MA, USA) antibodies. For immunofluorescence

analysis, cells were grown on coverslips, fixed with 4% paraformaldehyde and permeabilized with 0.2% Triton X100.

RNA isolation, RTqPCR and SuperArray analysis

Total RNAs were isolated using Trizol (Invitrogen, CA, USA) and qPCR carried out using the QuantiTect SYBR Green PCR Kit (Qiagen, CA, USA) in triplicate with ßactin as an internal control. PCR primer sequences are listed in the Table 2.1. For pathwayspecific array analysis, total RNAs were isolated using RNeasy kit (Qiagen, CA, USA) and analyzed using the Apoptosis (OHS

012) or Metastasis Microarray (OHS028) (SuperArray, MD, USA).

Conditioned medium, Boyden chamber migration and invasion assay

Cells (2.5 x 10 5) were seeded in growth media on 8m cell culture insert without coating (to assay migration) or coated with matrigel (to assay invasion)

(BD Biosciences, NJ, USA). TGFβ1 at 2.5 ng/ml (R&D systems, MN, USA), vehicle control, or conditioned medium was used in the lower chamber. To make the conditioned medium, 5 x 10 6 MIV/Vec or MIV/ATF3 cells were seeded on a

10cm plate and cultured in 6 ml of growth media for 48 h; the media were centrifuged at 500 x g for 5 min to remove the debris. At 24 hr after incubation in the Boyden chamber, cells on the upper side of the insert were removed by cotton swab and cells on the under side were stained with crystal violet and the

A595 measured.

Chromatin immunoprecipitation (ChIP) and in vivo transcription assay

ChIP assay was carried out as described previously (Lu et al., 2006b). For transcription assay, RNA Pol II occupancy on the genes of interest were analyzed by ChIP and quantified by qPCR. AntiATF3, antiPol II antibodies and control IgG (Santa Cruz, CA, USA) were used for the assay. PCR primer sequences are listed in Table 2.2.

Orthotopic xenograft mouse model

All experiments on mice were performed in accordance with the regulations of the Institutional Animal Care and Use Committee (IACUC) at the Ohio State

University. Female athymic nude mice with the age between 46 weeks were purchased from Harlan (Indianapolis, IN, USA) and randomly divided into two groups (n=15 for each group). The mice were anesthetized with intraperitoneal

(i.p.) injection of Avertin (Sigma, MO, USA) at a dose of 250 mg/kg body weight

(BW). 1 x 10 6 MIV (Vec vs. ATF3) cells were resuspended in 20 l of growth media, mixed with matrigel (BD Biosciences, San Jose, CA) in 1:1 (volume: volume) ratio, and injected into the No. 4 mammary fat pad of nude mice.

Monitor the primary tumor growth by by measuring the length (L) and width (W) of the tumor twice a week. The tumor size (Vt) is calculated as: Vt=0.5 x L x W 2.

When the primary tumor reaches an average diameter of 2 cm, remove the primary tumor and allow the mice to survive for another two months before euthanize the mice and fix the lung tissue with intratracheal injection of formalin followed by dissecting out the lung tissue. Count the lung metastatic nodules

visible to the naked eye, measure the length (L) and width (W) of each nodule and calculate the nodule size (Vn) as: Vn=0.5 x L x W 2.

Statistics

All quantitative data from viability, qRTPCR, and qPCR assays are expressed as mean ± S.D. Comparisons were made by the Student’s t test and p<0.05 is considered statistically significant.

2.4. RESULTS

2.4.1. ATF3 has opposite effects on the untransformed MI cells and the malignant MIV cells

I generated stable derivatives from the four isogenic cells developed by Miller, et al. (Santner et al., 2001) to express ATF3 or a control vector by retroviral transduction, referred as MI (MII, MIII, MIV)/Vec, and MI (MII, MIII, MIV)/ATF3 in the rest of this thesis. The protein level of the exogenous ATF3 in these cells was comparable to that of the endogenous ATF3 induced by TGFβ as shown by immunoblot (Figure 2.1.A). Immunofluorescence indicated that ATF3 was expressed in the majority of the cells, as demonstrated in MI and MIV cells respectively (Figure 2.1.B). To reduce the possibility of changes in cells due to prolonged passages, I used cells within 10 passages after establishment, and all results presented in this thesis were confirmed by at least three independent transductions. Ectopic expression of ATF3 did not result in statistically significant differences in cell growth, except in MIII cells (Figure 2.1.C). The ATF3 54

expressing MIII (MIII/ATF3) cells consistently had higher cell numbers than the control (MIII/Vec) cells under normal growth conditions (Figure 2.1.C). The reason for this difference is not clear. Serumwithdrawal reduced the numbers of vector control cells. Significantly, ectopic expression of ATF3 in MI cells further reduced the cell numbers upon serumwithdrawal, indicating that ATF3 is deleterious. In contrast, ectopic expression of ATF3 in MIV cells resulted in higher cell numbers than the vector control upon serumwithdrawal, indicating that

ATF3 is protective. The ATF3expressing MII and MIII cells showed a similar trend as the MIV counterpart: higher cell number than their corresponding vector control cells. However, the difference for MII was not statistically significant with the current number of repeats (N=3); the situation for MIII was complicated by the increased cell number of the ATF3expressing cells under normal growth condition. Thus, it is not possible to make definitive conclusions at this point.

However, our data clearly indicate that ATF3 expression has opposite effects on the untransformed MI cells and the malignant MIV cells. Although ATF3 has been demonstrated to have opposite functions in the literature (see introduction), our result is the first to demonstrate this dichotomy using isogenic cell lines, and the first to demonstrate that the degree of malignancy of the cells affects ATF3 function. To further analyze the effects of ATF3 on these isogenic cells, we focused on MI and MIV cells.

2.4.2. ATF3 is proapoptotic in untransformed breast epithelial cells

The deleterious effect of ATF3 on the MI cell number (under serum withdrawal, Figure 2.1.C) could be due to increased apoptosis, increased cell cycle arrest, or both. Flow cytometry analysis showed that serumwithdrawal induced apoptosis in the vector control cells as evidenced by the cell population with <2N DNA content (Figure 2.2.A) and immunoblot analyses showed increases in the activated caspase 3 and caspase 8 (Figure 2.2.B). Significantly, ATF3 enhanced the apoptotic response in MI cells by both assays (Figures 2.2.A and

2.2.B). Analysis of the cell morphology showed rounding up of the MI/ATF3 cells at 6 hours after serumwithdrawal (Figure 2.2.C). I also examined the cells for

DNA replication by BrdU incorporation upon serumwithdrawal. Under this condition, the cells did not show much labeling by BrdU, consistent with the FACS profile. No difference was observed between vector control and ATF3expressing cells (data not shown). Thus, the primary reason for the lower cell number of the

MI/ATF3 cells than the MI/Vec cells upon serumwithdrawal (Figure 2.1.C) was enhanced apoptosis. Since serumwithdrawal also induced the expression of the endogenous ATF3 gene (Figure 2.2.B), we asked whether ATF3 is necessary for serumwithdrawal to induce apoptosis in the control cells. As shown in Figure

2.2.D, ATF3 knockdown reduced caspase activation. Control siRNA did not affect

ATF3 expression and did not affect serumwithdrawal induced caspase activation.

To gain further insight on the effect of ATF3 on apoptosis, I examined the gene expressions in MI/ATF3 cells vs. MI/Vec cells using Apoptosis Microarray, which contains 112 genes involved in apoptosis. Several genes in the TNF pathway 56

were upregulated in the MI/ATF3 cells under normal growth conditions: TNF

SF7, TNFSF10, and TNFRSF12A. SYBR greenbased reverse transcription coupled with quantitative PCR (RTqPCR) confirmed their differential expression

(Figure 2.2.E). Members in the TNF family include paired ligands and their cognate receptors. TNFSF7 and TNFSF10 encode two ligands, and their cognate receptors are TNFRSF7 and TNFRSF10, respectively. TNFRSF12 is a receptor, which is bound by a ligand called TNFSF12. The presence of both ligands and receptors allows the binding and subsequent intracellular signaling.

Therefore, I also examined the genes encoding their cognate ligands and receptors by RTqPCR and found no upregulation by ATF3 (data not shown).

This result provides a partial explanation for the lack of obvious apoptosis in the unstressed MI/ATF3 cells.

2.4.3. ATF3 protects aggressive breast cancer cells from stressinduced cell cycle arrest and promotes their cell motility and invasiveness in vitro

To address whether ATF3 protected the aggressive breast cancer cells by inhibiting apoptosis, cell cycle arrest, or both, I analyzed the cells by FACS analysis and caspase activation as above. As shown in Figure 2.3.A., serum withdrawal increased G1 cell population in the MIV/Vec cells but did not increase the cell population with <2N DNA content. Thus, the aggressive MIV cells responded to serumwithdrawal differently than the untransformed MI cells.

Instead of an apoptotic response, the cells exhibited G1 arrest (albeit incomplete).

Importantly, ATF3 protected the cells from this response (Figure 2.3.A). I also 57

assayed apoptosis by caspase 3 and caspase 8 activation. As shown in Figure

2.3.B, serumwithdrawal did not induce their activation, consistent with the FACS analysis. This lack of apoptotic response in MIV cells was not due to the defects in their apoptotic machinery, since camptothecan, a DNAdamaging agent, induced caspase activation (Figure 2.3.B). Thus, ATF3 protected the MIV cells from serumwithdrawal induced reduction in viable cell numbers primarily by inhibiting cell cycle arrest.

I next examined the ability of ATF3 to affect cell motility using Boyden chamber and found that MIV/ATF3 cells had higher motility than the control cells either in the absence or presence of TGFβ (Figure 2.4.A). A representative picture of the motility assay is shown in 2.4.B. ATF3 also increased the cell invasion when matrigelcoated membrane was used (Figure 2.4.A). Since the endogenous level of ATF3 in the MIV cells was detectable (Figure 2.1.A), we asked whether ATF3 plays a necessary role in its basal motility. Significantly, siRNA knockdown of

ATF3 reduced the cell motility of MIV cells (Figure 2.4.C), indicating that ATF3 is partially necessary for the MIV cells to migrate. Taken together, both the gain and lossoffunction approaches indicated that ATF3 promotes cell migration in the

MIV aggressive breast cancer cells. For a comparison, I examined whether ATF3 affects the cell motility of the untransformed MI. As shown in Figure 2.4.D., ectopic expression of ATF3 also increased its cell motility. Since the basal expression level of ATF3 in MI cells was usually undetectable, I did not carry out the knockdown experiments. We note that ATF3 was demonstrated to increase cell motility in a melanoma line and a colon cancer line (Ishiguro et al., 2000;

Ishiguro et al., 1996). Thus, the ability of ATF3 to affect cell motility is not limited to breast epithelial cells.

2.4.4. ATF3 modulates the expression of genes known to regulate cell motility

To understand the mechanism by which ATF3 regulates cell motility, I carried out array analysis using the Metastasis Microarray, which contains 113 genes known to modulate cellcell or cellextracellular matrix (ECM) interactions — key processes affected in metastasis. Several genes were modulated by ATF3 in repeated array analysis: Fibronectin1 (FN1), secretory leukocyte peptidase inhibitor (SLPI), collagen IVa2 (COL4A2), and caveolin1 (CAV1). RTqPCR confirmed their differential expression (Figure 2.5.A.). In addition, we examined a few candidate genes known to modulate cell motility but are not included in the array, and found that urokinasetype plasminogen activator (uPA, also called

PLAU), plasminogen activator inhibitor1 (PAI1), TWIST1, and Slug are up regulated in the MIV/ATF3 cells (Figure 2.5.A.). ATF3 expression also up regulated the Snail steadystate mRNAs. However, the fold change was small

(1.6 fold) and not statistically significant ( p > 0.05). Either the number of experimental repeats (N=3) is not high enough or ATF3 by itself is not sufficient to upregulate Snail expression. All of these genes have been implicated in cancer cell motility and/or metastasis, providing a mechanistic explanation for the ability of ATF3 to increase cell motility. Intriguingly, several of these genes —

FN1, SLPI, uPA, PAI1 and COL4A2 — encode secreted factors and are 59

thought to play a role in remodeling the extracellular matrix. I thus carried out a conditioned medium (C.M.) experiment by incubating the C.M. from MIV/ATF3 cells with the MIII cells, the lowgrade carcinoma cells. Figure 2.5.B showed that the C.M. from MIV/ATF3 cell promotes the migration of MIII to a greater extent than the C.M. from MIV/Vec cells, indicating that ATF3 expression in the cancer epithelial cells can promote cell motility via a paracrine (or autocrine) effect.

2.4.5. Differential regulation of gene expression by ATF3 in MI versus MIV cells

To determine whether ATF3 differentially regulates gene expression in MI versus MIV cells, we carried out qRTPCR analyses of both MI and MIV cells for all the genes identified above. Table 2.3 summarizes the results and shows that these genes can be classified into three groups. (a) Group 1 are genes that are regulated similarly in MI and MIV cells: SLPI, PAI1, uPA, TWIST1, and

COL4A2. (b) Group 2 are genes downregulated in MI cells but upregulated in

MIV cells: FN1, Slug, and CAV1. (c) Group 3 are genes upregulated in MI cells but not changed in MIV cells: TNFRSF12A, TNFSF7, TNFSF10 and

Snail perhaps. The significance of this finding is discussed below.

2.4.6. Direct target genes of ATF3

Although the above results show an altered expression of several genes in the ATF3expressing cells, they do not indicate whether these are direct target 60

genes of ATF3. To address this issue, I first analyzed their promoter sequences using the MotifScanner. As shown in Figure 2.6, all of them contain consensus

CRE/ATF or CRE/ATFlike sequences. However, for some promoters the sites are more than 2 kilobases (kb) upstream from the transcriptional start (+1) site.

Although this per se does not rule out the possibility that they are direct target genes of ATF3, we focused our efforts on FN1, TWIST1 and slug, which have the potential ATF3 binding sites within 200 basepair (bp) from the +1 site. We also analyzed Snail, which has a potential binding site at around 1.7 kb. As shown by chromatin immunoprecipitation (ChIP) assay, ATF3 bound to these promoters in both MI and MIV cells (Figure 2.7.A). The binding to the FN1 promoter was obvious even in the MIV/Vec cells, which do not have exogenous

ATF3. Presumably, this was due to the binding by the endogenous ATF3, which was detectable by immunoblot (Figure 2.7.A).

Since ATF3 is a transcription factor, its binding to these promoters suggests, but does not prove, that ATF3 regulates their transcriptional activity. To address the issue of transcription, I examined RNA polymerase II (Pol II) occupancy on these genes by ChIP using two primer sets (Figure 2.7.B). (i) Primer set 1 targets to a region downstream (around 2 to 5 kb) from the +1 site. Pol II occupancy in these regions would indicate active transcription of the corresponding genes (Komarnitsky et al., 2000). Primer set 2 targets to the proximal promoter region around the +1 site. ChIP signals from this primer set indicate the loading of Pol II on the promoters. I quantified ChIP signal by qPCR and compared ATF3expressing cells to vector control cells by defining the

signals from the control cells as 1. As shown by primer set 1, in MI cells, ATF3 upregulated TWIST1 and Snail gene transcription but downregulated FN1 and

Slug transcription (Figure 2.7.C). This pattern paralleled the effects of ATF3 on their steadystate mRNA levels (Table 2.3, qRTPCR data). Although the fold change was not exactly the same, the trend was the same. For MIV cells, ATF3 upregulated FN1, TWIST1, and Slug gene transcription, but did not affect snail gene (Figure 2.7.C). Again, this pattern paralleled the modulation of their steady state mRNA levels by ATF3 (Table 2.3). In contrast to primer set 1, primer set 2 showed similar Pol II loading on these promoters in ATF3expressing and control cells (Figure 2.7.D), a pattern different from the steadystate mRNA data (Table

2.3). This result was observed in both MI and MIV cells, and indicated that Pol

II loading on the promoters does not necessarily correlate with transcription. This lack of correlation between Pol II loading and gene transcription has been reported previously (Cheng and Sharp, 2003; Komarnitsky et al., 2000); in fact, the accumulation of “paused” Pol II in the proximal promoter was suggested to be a common feature of mammalian transcriptional regulation (Cheng and Sharp,

2003). As a control for Pol II ChIP, we examined Pol II binding to the αactin and

βactin promoters. The ChIP signals were found on the βactin but not αactin promoter (data not shown), consistent with the fact that αactin, a musclespecific gene, is not expressed in the epithelial cells. Taken together, our results indicate that FN1, TWIST1, Slug, and perhaps, Snail are direct target genes of ATF3.

Ectopic expression of ATF3 can modulate their expression in both MI and MIV cells, except for Snail in MIV cells. This lack of Snail modulation in MIV cells

suggests that some other factors are required to regulate its expression in that cell. We have not examined the rest of the potential ATF3 target promoters identified above (such as PAI1 and uPA) for their transcriptional activity or binding by ATF3; thus it is not clear whether they are direct or indirect targets.

2.4.7. ATF3 promotes primary tumor growth and lung metastasis in an orthotopic xenograft mouse model

The effects of ATF3 in protecting MIV cells from apoptosis, releasing cell cycle arrest and enhancing cell motility all suggest an oncogenic role of ATF3 in vitro . To further examine the in vivo functions of ATF3 in regulating cell survival and motility in the malignant mammary epithelial MIV cells, I used an orthotopic xenograft mouse model (Figure 2.8.A). I injected the MIV/ATF3 and MIV/Vec cells in a 1:1 volume ratio with matrigel at the #4 fat pad of female nude mice, and measured the tumors weekly by caliper. When the tumor reached ~ 4 cm 3

(~4 weeks for MIV/ATF3 cells and ~5 weeks for MIV/Vec cells, Figure 2.8.B), I removed the tumors and waited for 2 months before sacrificing them and analyzing their lung metastases. As shown in Figure 2.8.CF, ATF3 promotes lung metastasis as judged by 3 criteria: (a) higher incidence (percentage of mice with lung metastases, 7/11 for ATF3 group versus 6/13 for vector group), (b) higher number of lung nodules per mouse (2.6 versus 0.7), and (c) higher tumor burden per mouse (0.18 cm 3 versus 0.01 cm 3). I started 15 mice per group, but some mice died before they were sacrificed for analyses: 2 died in the vector group and 4 in the ATF3 group. I also analyzed bone, liver, and brain — the 63

other potential metastatic sites for breast cancer — and found no obvious metastases.

2.4.8. ATF3 gene is upregulated in human breast cancer cells

The above results from the MIV cell and animalbased systems strongly suggest that ATF3 plays an oncogenic role in malignant breast carcinoma cells.

Since both experiments were performed using MIV cells, to examine whether the oncogenic feature of ATF3 is only specific for MIV cells, I surveyed the expression of ATF3 by immunoblot in human breast cancer cell lines. Among the

18 breast cancer cell lines examined, 16 (~ 90%) expressed ATF3 at a level higher than the nontransformed breast epithelial MI cells (Figure 2.9).

2.5. DISCUSSION

Using an isogenic breast epithelial cell system, it is the first time, to our knowledge, that ATF3 was demonstrated to play a dichotomous role at different stages during cancer development. Since ATF3 is a transcription factor, we focused on examining the underlying mechanisms of ATF3 dichotomy by exploring the potential ATF3 target genes in this study. It is highly possible that

ATF3 has transcriptionindependent functions and they may also contribute to its dichotomous effects. As shown in Table 2.3, ATF3 differentially regulates several genes in the MI and MIV cells —the group 2 and group 3 genes. The significance of this finding is twofold. First, this is the first to demonstrate

differential gene regulation by ATF3 in cells with different degrees of malignancy, providing at least partial mechanistic explanations for the dichotomous role of

ATF3. Second, since the MI and MIV cells are isogenic, they share mostly the same genetic backgrounds, providing a tractable system to study the cellular contexts that allow ATF3 to differentially regulate gene expression. As shown in

Figure 2.7, ATF3 binds to the promoters of FN1 and Slug, two genes oppositely regulated by ATF3. It is possible that ATF3 recruits different factors to these promoters in MI versus MIV cells, resulting in differential regulation.

Although ATF3 exhibits opposite regulation on cell survival in response to stress signals, it enhances cell motility in both MI and MIV cells in vitro (Figure

2.4.D), suggesting the regulation of cell motility by ATF3 is not dependent on the malignancy of the cells.

Besides the in vitro effects of ATF3, I also examined the functions of ATF3 in vivo in an orthotopic xenograft mouse model. By monitoring the primary tumor growth, I found tumors injected with MIV/ATF3 cells grow faster than those with

MIV/Vec cells (Figure 2.8.B). Before and during the course of this thesis work,

ATF3 was demonstrated to be either a tumor suppressor or an oncogene in xenograft injection models using various cell lines (Bandyopadhyay et al., 2006;

Bottone et al., 2005; Ishiguro et al., 2000; Lu et al., 2006b). The inconsistency could be caused by the differences in cell types, the malignancy status of the cells and the ways of injection (orthotopic vs. subcutaneous).

In addition to promoting primary tumor growth, ATF3 also promotes lung metastasis, which is consistent with the study by Ishiguro et al., who used

“experimental metastasis” assay (Ishiguro et al., 2000; Ishiguro et al., 1996). In the “experimental metastasis” assay, the cancer cells (colon and melanoma cells in their reports) are injected into the tail vein; thus, it measures the ability of cancer cells to colonize the lung after being “filtered out” by the pulmonary capillaries, not a true metastasis assay. Another study by Bandyopadhyay et al. injected subcutaneously prostate cancer cells and examined the lung metastasis, without removing the primary tumors (Bandyopadhyay et al.,

2006). Compared to these studies, our model has two improvements: First, the cancer cells were injected at the sites where the tumors are supposed to develop, the mammary fat pad, in contrast to prostate cancer cells subcutaneously; second, the primary tumors were removed long before metastasis was examined, better mimicking the clinical situations. Thus, our result is significant: it demonstrates the ability of ATF3 to enhance metastasis in an improved and more relevant model; furthermore, it shows that the cells we engineered (with ATF3) will be useful for studying breast cancer metastasis.

Metastasis is a multistep process including local migration and invasion, intravasation, traveling and arrest in blood or lymph circulation, extravasation, and colonization in the secondary site to give rise to micro and macro metastasis (Fidler, 2003). It involves a broad spectrum of crosstalks between the cancer cells and its environment. Therefore, to understand how ATF3 enhances metastasis in vivo , we need to examine at which step(s) in the metastatic process that ATF3 exerts its actions, whether those functions are mediated only through the transcriptional regulation by ATF3 or include

transcriptionindependent regulations, and what is the nature of ATF3 functions, cancercellautonomous or interactive between cancer cells and stroma? Some of these questions will be further addressed in Chapter 3 and 4.

The endogenous ATF3 level in MIV cells are higher than that in MI cells

(Figure 2.1.A). Since ATF3 is prosurvival and prometastatic in the malignant breast epithelial MIV cells, it is beneficial for MIV cells to express a higher level of ATF3 than in MI cells, where ATF3 is proapoptotic. This gene expression difference also correlates with the status of chromosome 1q within each cell line, as MIV cells have 3 copies of chromosome 1q, while MI only 2 copies (Santner et al., 2001). Human ATF3 gene is located on chromosome 1q.32 region. The amplification of a chromosomal region where ATF3 resides, combined with the mechanistic and functional data we obtained from cell lines and mouse model, strongly supports the notion that ATF3 is an oncogene for breast cancer. In addtion, 16 out of 18 breast cancer cell lines we examined also have higher

ATF3 expression compared to nontransformed MI cells. However, there are some limitations for the information we obtained from cell lines and more informaiton from human breast cancer tissues are essential. First, it is not clear whether the higher expression of ATF3 in breast cancer cell lines is intrinsic to the original samples (human breast cancer tissues) or is a secondary effect during the establishment of cell line. To address this issue, a survey of human breast cancer tissues is necessary. Second, though gene amplification is a strong evidence for gene function in cancer, the crude resolution of the methods used to identify chromosome amplification makes it difficult to determine whether

ATF3 gene is indeed amplified. Third, in this study, we examined the functions of wildtype ATF3 in cell system and mouse model, which suggests an oncogenic feature of ATF3 in breast cancer. Without any knowledge of ATF3 status (wild type or mutated) in human breast cancer tissues or cell lines, we cannot make any conclusion on ATF3 as an oncogene or tumor suppressor gene, as the lesson we learned from p53. p53 was first identified as an oncogene due to its high expression in many different cancers but turns out to be a tumor suppressor gene since most p53 in tumors are mutated (Lane and Benchimol, 1990).

Fourth, given the fact that ATF3 promotes metastasis in the orthotopic xenograft mouse model and metastasis is the most common cause of cancerrelated death in human, it is important to examine the correlation of ATF3 to the outcome of patients with breast cancer. Such correlation analysis will help to determine whether ATF3 could be used as a biomarker for predicting patient outcome and as a potential target for treatment. Chapter 5 in this thesis describes the detailed analyses of ATF3 in human breast cancer tissues and provides answers for some of these important questions.

MCF10AT1k (M MCF10CA1h MCF10CA1a Cell Line MCF10A (MI) II) (MIII) (MIV) Stage of breast Spontaneously Oncogenically Lowgrade Highgrade cancer immortalized initiated carcinoma metastatic modeled but premalignant carcinoma untransfromed epithelium epithelium Description of Spontaneously Derived from MI Derived from a Derived from a line immortalized cells transfected xenograft xenograft line from non with activated H implant of MII implant of MII malignant RAS oncogene cells that cells that human breast and passaged progressed to progressed to epithelium through mice as carcinoma carcinoma persistent lesions with features of proliferative breast disease Tumorigenicity Non Initially form Form Form poorly on tumorigenic simple ducts that predominantly differentiated xenografting progress to well carcinomas; into nude mice benign differentiated metastatic to hyperplastic carcinomas lung in tail lesions and vein injection occasionally assay carcinomas

Table 2.1 Properties of MCF10Aderived cell lines (Tang et al., 2003).

2.1 Properties of MCF10Aderived cell lines

Assays Gene Forward Primer (lab ID) Reverse Primer (lab ID) RTqPCR ATF3 CTGCAGAAAGAGTCGGAG (TH511) TGAGCCCGGACAATACAC (TH512) βactin AAGATCAAGATCATTGCTCCTC (TH381) GGACTCATCGTACTCCTG (TH382) TNFR GTTCACCACCCCCATAGA (TH463) AGGCAGAGACTGGCTCTA (TH464) SF12A TNFSF7 TGCTTTGGTCCCATTGGT (TH469) GTGATTCAGCTGCAGCTC (TH470) TNFSF10 CCAGAGGAAGAAGCAACAC (TH467) GGAATGAATGCCCACTCC (TH468) FN1 GCAACTCTGTCAACGAAG (TH444) ACTTCCAAAGCCTAAGCA (TH445) SLPI CTGGAAAGTCCTTCAAAGCTG (TH455) CATTTGATGCCACAAGTGTC (TH456) PAI1 CATCATCAATGACTGGGTG (TH459) AGTAGAGGGCATTCACCA (TH460) uPA TACGGCTCTGAAGTCACC (TH457) CCAGTCAAAGTCATGCGG (TH458) COL4A2 GAGGACTTGGTTTCTACGG (TH453) ATCTGGGTGGAAGGTGAC (TH454) CAV1 GCAACATCTACAAGCCCA (TH442) CTTCAAAGTCAATCTTGACCAC (TH443) TWIST1 GAGACCTAGATGTCATTGTTTCC (TH533) GCCCTGTTTCTTTGAATTTGGA (TH534) Slug ACACATACAGTGATTATTTCCC (TH625) ATGAGGAGTATCCGGAAAGAG (TH626) Snail ACTCTAATCCAGAGTTTACCTTCC (TH623) AGGACAGAGTCCCAGATGAG (TH624) ChIP for FN1 TCCCTTTCCTCCCAGCCGCTTCC (TH539) AGCCGACCGCGCGCCGATTG (TH540) ATF3 binding TWIST1 GAGGGGGACTGGAAAGCGGAAA (TH560) TGGGCGAGAGCTGCAGACTTG (TH561) Slug AGGAGCTGAAATCTGAACCTCT (TH636) GTTGAAATGCTTCTTGACCAGG (TH637) Snail AAGGTCTGGGATGTTGGAGG (TH638) GTCCTCTCCTCAGCCAACTC (TH639) ChIP for FN1 (Set CTGAGAAGTGTTTTGATCATGC (TH594) TTCTAGAAGTGCAAGTGATGC (TH595) RNA Pol II 1) occupancy FN1 (Set TCCCTTTCCTCCCAGCCGCTTCC (TH539) AGCCGACCGCGCGCCGATTG (TH540) 2) TWIST1 GGCACCATCCTCACACCTCT (TH541) CGACCTCTTGAGAATGCATGC (TH542) (Set 1) TWIST1 GAGGGGGACTGGAAAGCGGAAA (TH560) TGGGCGAGAGCTGCAGACTTG (TH561) (Set 2) Slug (Set TTCATGTATGATTGGCAGCAGTA (TH649) GCAAATGCTCTGTTGCAGTGA (TH648) 1) Slug (Set AGGAGCTGAAATCTGAACCTCT (TH636) GTTGAAATGCTTCTTGACCAGG (TH637) 2) Snail (Set CTTTGGACCCTGGCTGTGTGT (TH646) CTTCTTGACATCTGAGTGGGTC (TH645) 1) Snail (Set CGTCAGAAGCGCTCAGACCA (TH742) ACTCCTCCGAGGCGGGGTT (TH743) 2) βactin CGGCCAACGCCAAAACTCTC (TH564) AACTTTCGGAACGGCGCACG (TH565) αactin GCAAACCCGCTAGGGAGACA (TH562) TGGGTGTTGGGCACTAGAGC (TH563)

Table 2.2 A list of the primers used in this study for the indicated assays. 70

2.2 A list of the primers used in this study for the indicated assays

Group Gene Name MI (Fold) MIV (Fold) SLPI 0.03 0.2 PAI1 9.5 5.7 1 uPA 3.2 4.2 TWIST1 9.6 6.6 COL4A2 2.6 2.3 FN1 0.4 8.4 2 Slug 0.1 5.3 CAV1 0.6 2.1 Snail* 7.1 TNFRSF12 3.0 3 No Changes TNFSF7 6.3 TNFSF10 2.1

Table 2.3 Classification of potential ATF3 target genes based on their differential expression in MI versus MIV cells. The numbers indicate the fold change: relative mRNA levels in ATF3 expressing cells to that in the corresponding vector control cells. >1: genes are upregulated in ATF3 cells vs. Vec cells; <1: genes are downregulated in ATF3 cells vs. Vec cells. Group 1: genes similarly regulated by ATF3 in MI and MIV cells; Group 2: genes down regulated by ATF3 in MI cells but upregulated in MIV cells; Group 3: genes up regulated by ATF3 in MI cells but not regulated in MIV cells. *: not statistically significant ( p > 0.05).

2.3 Classification of potential ATF3 target genes based on their differential expression in MI versus MIV cells

A. MI MII MIII MIV Stable TGFβ Stable TGFβ Stable TGFβ Stable TGFβ Vec ATF3 0hr 2hr Vec ATF3 0hr 2hr Vec ATF3 0hr 2hr Vec ATF3 0hr 2hr ATF3 Actin

B. ATF3 TOTO3 ATF3 TOTO3

MI/Vec MIV/Vec

MI/ATF3 MIV/ATF3

C. * 1.6 Vec * ATF3 1.2 * 0.8

0.4 Relative Cell No. Relative * 0 Serum withdrawal + + + + MI MII MIII MIV Fig. 2.1 Characterization of stable cells expressing ATF3 or the control vector. A. Stable cells derived from MI, MII, MIII, and MIV cells to express ATF3 or the control vector (Vec) were analyzed by immunoblot using the indicated antibodies. The levels of endogenous ATF3 expression induced by TGFβ were examined in parallel in the parental MI to MIV cells treated with TGFβ (2.5 ng/ml) for 2 hr and analyzed by immunoblot. B. MI/ATF3, MI/Vec, MIV/ATF3 and MIV/Vec cells were analyzed by immunofluorescence using the ATF3 antibodies (green). TOTO3 (red) was used to indicate the nuclei. C. ATF3expressing and the corresponding vector control cells derived from MI to MIV cells were seeded into normal growth media (serumwithdrawal ) or serum free media (serumwithdrawal +) and grown for 72 hr. Viable cells were analyzed by crystal violet assay. The values from vector control cells in normal growth media were arbitrarily defined as 1. Shown are the mean ± S.D. from three independent experiments (* p < 0.05, versus corresponding Vec cells). 2.1 Characterization of stable cells expressing ATF3 or the control vector

A. MI/Vec MI/ATF3 B. MI/Vec MI/ATF3 Stress (hr) 0 8 12 0 8 12 0.4% 0.3% ProCasp 8 0 hr *Casp 8 *Casp 3

Actin

2N 4N 2N 4N Short ATF3 73% Long 57% 24 hr C.

<2N 2N 4N <2N 2N 4N D. E. RTqPCR siCtrl siATF3 8 MI/Vec MI/ATF3 Stress (hr) 0 8 12 18 0 8 12 18 * ProCasp8 6 *Casp8 4 *Casp3 *

ATF3 2 * Relative mRNA Relative Level mRNA Actin 0 TNFR- TNF- TNF- SF12A SF7 SF10

Fig. 2.2 ATF3 is proapoptotic in untransformed mammary epithelial cells. A. MI/Vec and MI/ATF3 cells were cultured in serumfree media for the indicated times and analyzed by FACS to detect DNA content. Shown are representative data from three independent experiments. Numbers indicate the percentage of cells with <2N DNA content. B. MI/Vec and MI/ATF3 cells were cultured in serumfree media (Stress) as in panel A for the indicated times and analyzed by immunoblot. Two exposures (short and long) for ATF3 signals are shown. *Casp denotes activated caspase. C. MI/Vec and MI/ATF3 cells were seeded in serumfree media and photographed by phase contrast microscopy at 6 hr after seeding. D. MI cells were transfected with control (siCtrl) or ATF3 (siATF3) siRNAs for 72 hr and then cultured under serumfree media (Stress) for the indicated times, followed by immunoblot. *Casp denotes activated caspase. E. Total RNAs were isolated from MI/Vec and MI/ATF3 cells, and analyzed by RTqPCR for TNFRSF12, TNFSF7 and TNFSF10. The levels of each mRNAs in MI/Vec cells were arbitrarily defined as 1. Shown are the mean ± S.D. from three independent experiments (* p < 0.01, versus MI/Vec). 73

2.2 ATF3 is proapoptotic in untransformed mammary epithelial cells

Serum A. B. Stress: Withdrawal CPT MIV/Vec MIV/ATF3 MIV Cells: Vec ATF3 Vec ATF3 (hr) 0 8 12 0 8 12 8 8 0 hr *Casp 8 2N: 60% 58% *Casp 3

24 hr ATF3 Actin 2N: 72% 62%

Fig. 2.3 ATF3 is protective in malignant mammary epithelial cells. A. M IV/Vec and MIV/ATF3 cells were cultured and analysed by FACS as in Fig.2.2.A, with the percentage of G1 cells indicated. B. MIV/Vec and MIV/ATF3 cells were cultured in serumfree media and analysed by immunoblot. Camptothecan (CPT) at 30 mM was used to induce apoptosis as a control.

2.3 ATF3 is protective in malignant mammary epithelial cells

A. MIV/Vec B. MIV/Vec MIV/ATF3 3 MIV/ATF3 * * 2 *

1 TGFβ

0 Relative Migration Relative TGFβ +TGFβ TGFβ Migration Invasion

2.7 C. D. Vec * ATF3 siCtrl siRNA 1.0 siATF3 1.8 Ctrl ATF3 *

* 0.5 ATF3 0.9 Actin

0 Relative Migration

Relative Migration Relative Migration 0.0 MI MIV

Fig. 2.4 ATF3 enhances cell mobility and invasiveness in malignant breast carcinoma cells. A. MIV/Vec and MIV/ATF3 cells were analyzed by Boyden chamber assay for migration or invasion in the absence or presence of TGFβ (2.5 ng/ml) as indicated. Migrated cells were quantified by crystal violet staining as detailed in materials and methods section. A 595 reading from MIV/Vec cells in the absence of TGFβ was arbitrarily defined as 1. Quantitative data in all panels of this figure are the mean ± S.D. from three independent experiments. * p < 0.01, versus MIV/Vec. B. MIV/Vec and MIV/ATF3 cells were analyzed by Boyden chamber assay and the migrated cells were photographed under microscope after staining with crystal violet. C. MIV cells were transfected with control (siCtrl) or ATF3 (siATF3) siRNA for 72 hr and assayed for cell migration (left panel). * p < 0.05, versus MIV/Vec. The level of ATF3 protein after siRNA treatment was determined by immunoblot (right panel). D. MI/Vec, MI/ATF3, M IV/Vec and MIV/ATF3 cells were analyzed by Boyden chamber assay for migration. A 595 readings from the corresponding vector control cells were arbitrarily defined as 1. Shown are the mean ± S.D. from three independent experiments. * p < 0.01, versus Vec cells. 2.4 ATF3 enhances cell mobility and invasiveness in malignant breast carcinoma cells

A. B. RTqPCR 1.5 * * 9 MIV/Vec 1.0 MIV/ATF3 * 6 * * * 0.5 3 * * * RelativeMigration 0 0 MIV/Vec MIV/ATF3 Relative mRNA RelativemRNA Level FN-1 SLPI PAI-1 uPA COL4A2 CAV1 TWIST1 Slug Snail Conditioned Media

Fig. 2.5 ATF3 regulates cellmotilityrelated genes. A. Total RNAs were isolated from MIV/Vec and MIV/ATF3 cells and analyzed by RTqPCR for the indicated mRNAs. * p < 0.05, versus MIV/Vec. B. MIII cells were assayed for migration in the presence of conditioned media (C.M.) from MIV/Vec or M IV/ATF3 cells as indicated. A 595 reading from the cells incubated with the M IV/vector C.M. was arbitrarily defined as 1. * p < 0.05, versus C.M. from M IV/Vec.

2.5 ATF3 regulates cellmotilityrelated genes

CRE/ATF Site

183 Gene Position Sequence FN1 183 TGACGTCA 48 TWIST1 48 TCACGTCA 1 700 Snail 1 700 CA ATGTCA +20 Slug +20 TGACTTCA 2 930 1 975 SLPI 1 975 TGACCTCA 3 087 2 930 TGACCTCA 3 087 TGACCTCA 5 512 3 253 PAI15 512 TGAGGTCA 3 253 TGGCGTCA 4 708 4 399 3 697 2 017 uPA 4 708 TGACCTCA 4 399 TGACGACA 3 697 TGAGGTCA

2 111 2 017 TGAAGTCA COL4A2 2 111 TGAGGTCA 1 958 903 CAV1 903 TGACTG CA

CRE/ATF site 1 958 TGAGGCCA Consensus TGACGTCA Transcriptional start site

Fig. 2.6 Potential ATF3 target promoters. Schematic representations of the promoters are shown on the left, with the CRE/ATF sites indicated. The corresponding CRE/ATF sequences are indicated on the right, with the deviations from the consensus sequence underlined.

2.6 Potential ATF3 target promoters

A.MI MIV B. 183 +5 200 Input ChIP Input ChIP FN1 Cells: Vec ATF3 Vec ATF3 Vec ATF3 Vec ATF3 Set 2 Set 1 Ab: –– IgGA3 IgG A3 – –IgG A3 IgG A3 48 +1 740 TWIST1 FN1 Set 2 Set 1 1 700 200 +4 700 TWIST1 Snail Set 2 Set 1 Snail +20 +2 200 Slug Slug Set 2 Set 1 CRE/ATF site βactin Transcriptional start site C. D. 1 0 1 .6 Primer Set 1 * Vec Primer Set 2 8 ATF3 1 .2 6 * 0 .8 * 4 * * 0 .4 2

Relative PolRelative II Binding 0 * * 0 .0 FN1 TWIST1 Snail Slug FN1 TWIST1 Snail Slug FN1 TWIST1 Snail Slug FN1 TWIST1 Snail Slug MI MIV MI MIV

Fig. 2.7 FN1, TWIST1, Snail, and Slug are potential direct target genes of ATF3. A. MIV/ATF3 and MIV/Vec cells were analyzed by ChIP for ATF3 binding to the indicated promoters using the ATF3 (A3) or control (IgG) antibodies. βactin promoter was included as a negative control. Input: 5% of genomic DNA. Shown is the representative of three independent experiments.B. Schematic diagrams of the indicated genes: the transcriptional start site (+1), the potential CRE/ATF sites, and the positions of the primer target sites relative to +1 in basepairs (bp) are indicated. C. and D. MIV/ATF3 and MIV/Vec cells were analyzed by ChIP for Pol II binding to the indicated promoters. ChIP signals were quantified by qPCR. Pol II binding in the MIV/Vec cells (open bars) was arbitrarily defined as 1 to obtain the relative Pol II binding in MIV/ATF3 cells (close bars). Shown are the representative mean ± S.D. from three independent experiments (* p < 0.05, versus open bars).

2.7 FN1, TWIST1, Snail, and Slug are potential direct target genes of ATF3

9 A. D. 8 MIV/Vec MIV/Vec vs. 7 MIV/ATF3 MIV/ATF3 6 2 months 5 4 3 2 1 Cell Remove 1° Examine 0

injection tumor at ~2 cm lung mets. Nodules/Mouse of No. 1 diameter MIV/Ctrl MIV/ATF3 Incidence: (6/13) (7/11) Growth of Primary Tumors B. E. 10 MIV/Vec )

3 MIV/ATF3 ) 6 3 MIV/Vec 1 5 MIV/ATF3 10 1 4 10 2 3 10 3 10 4

2 (cm Size Nodule 1 10 5

Tumor size (cm size Tumor MIV/Vec MIV/ATF3 0 7 14 21 28 Total No.: (n=9) (n=29) Time after fat pad injection (days) Distribution of Nodule Size in Lung C. F. 10 MIV/Vec 9 MIV/ATF3 8 7 6 5 4 3

No. of Lung Mets. Lung No. of 2 1 MIV/Ctrl MIV/ATF3 0 5 4 3 2 1 3 10 10 10 10 >10 Size (cm ): 10 410 3 10 210 1

Fig. 2.8 ATF3 promotes tumor growth and metastasis in vivo . A. Schematic diagrams of the orthotopic xenograft mouse model. B. The growth curve of primary tumors derived from MIV/Vec and MIV/ATF3 cells after injection into the No. 4 mammary fat pad over a period of 28 days. C. Photographs of the lungs after fixing with formalin. D. Number (No.) of metastatic nodules in the lung tissues from mice injected with MIV/Vec ( ◆) or MIV/ATF3 ( ■) cells. E. Size of metastatic nodules in the lung tissues from mice injected with M IV/Vec ( ◆) or MIV/ATF3 ( ■) cells. F. Size distribution of metastatic nodules in the lung tissues from mice injected with MIV/Vec (□) or MIV/ATF3 ( ■) cells. “— ” in E and F indicates the mean value.

2.8 ATF3 promotes tumor growth and metastasis in vivo

SUM cells MI MIV 159 44 185 190 229 1315 52 102 149 225 ATF3

Actin

MI MIV HT 231 468 1937 SK BT MCF7 ATF3

Actin

Fig. 2.9 ATF3 gene is upregulated in breast cancer cell lines. ATF3 expression levels in various breast cancer cell lines were examined by immunoblot. Actin was used as an internal control. HT, HT1080 cells; 231, MDA MB231 cells; 468, MDAMB468 cells; 1937, HCC1937 cells; SK, SKBR3 cells; BT, BT468 cells. MI cells were used as negative control and MIV as positive control.

2.9 ATF3 gene is upregulated in breast cancer cell lines

CHAPTER 3

ATF3 AS A MEDIATOR FOR TGFβ SIGNALING IN BREAST CANCER CELLS

3. ATF3 AS A MEDIATOR FOR TGFβ SIGNALING IN BREAST CANCER CELLS

3.1. ABSTRACT

ATF3 is a member in the ATF/CREB family of transcription factors. Data presented in Chapter 2 indicates that ATF3 promotes cell migration/invasion in vitro and tumor metastasis in vivo. Given the importance of stromal factors in regulating tumor behaviors, I examined the potential involvment of ATF3 in downstream signaling of stromal factors in breast cancer MCF10CA1a cells. I found that ATF3 is induced by multiple stromal signals. When focusing on one important stromal factor, TGFβ, by both gainoffunction and lossoffunction approaches, I demonstrated that ATF3 contributes to the effects of TGFβ on target gene regulation and cell motility. The functions of ATF3 in TGFβ signaling are at least partially mediated through the interaction with Smad2/3, as supported by the cooccupancy of these proteins on the promoter of the same target gene and the reciprocal regulation of one protein on the cell motility induced by the other.

3.2. INTRODUCTION

Transforming growth factor β (TGFβ) superfamily consists of a large number of structurally related polypeptide growth factors that regulate many aspects of cellular processes, such as embryogenesis, proliferation, differentiation, adhesion, apoptosis, migration, and fibrosis (Massague, 1998). The dominant signaling system of TGFβ starts from ligand binding to type II and type I serine/threonine kinase receptors on the plasma membrane, which leads to a heteromeric receptor complex formation and activation of type I receptors by type

II receptors. The consequently activated type I receptors recruit and phosphorylate receptorregulated Smad proteins (RSmads). Following phosphorylaiton and activation, RSmads dissociate from the cell membrane receptors and form a complex with a common mediator Smad, Smad4. These

Smad complexes will then translocate into the nucleus, where they regulate the transcription of target genes (Derynck and Zhang, 2003). Among the RSmads,

Smad2 and Smad3 are specifically activated by TGFβ and activin type I receptors, while Smad1, Smad5 and Smad8 are substrates for BMP type I receptors. The Smad proteins can regulate transcription of target genes both positively and negatively in response to TGFβ signaling. Several studies have revealed the Smadbinding elements (SBEs, an optimal one for Smad3 is

CAGAC)(Shi et al., 1998; Zawel et al., 1998). However, the presence of SBEs alone in the promoter of some genes does not suffice the transcription regulation by Smads because, in isolation, the Smads bind DNA either with low affinity

(such as Smad3) or not at all (as in case of Smad2). The Smad proteins need to 82

cooperate with other sequencespecificbinding factors for efficient binding to the promoters of target genes (Moustakas et al., 2001) and those transcription factors that interact with Smad proteins determine the specific response to ligand in different cell types. Besides this canonical Smadmediated signaling pathway,

TGF β also activates other noncanonical signaling cascades, including mitogen activated protein kinase (MAPK) pathways, phosphatidylinositol3kinase (PI3K) pathway, and PP2A/p70s6K (Derynck and Zhang, 2003; Wakefield and Roberts,

2002).

Accumulating evidence suggests that TGF β is a “doubleedged sword” in carcinogenesis, i.e., both as a tumor suppressor and as a stimulator of tumor progression, invasion and metastasis (Akhurst and Derynck, 2001; Roberts and

Wakefield, 2003; Tang et al., 2003; Wakefield and Roberts, 2002). Using the forementioned MCF10A series of cells, Dr. Wakefield and colleagues first demonstrated that TGF β signaling can switch the roles from tumor suppression to promotion of metastasis during the course of cancer development (Tang et al.,

2003). Several mechanisms for the TGF β dichotomy have been proposed (Tian et al., 2003; Wakefield and Roberts, 2002). Based on the observation that many

Smad mediators downstream of the TGFβ receptors are also targets of mutation or epigenetic regulation in carcinogenesis except for Smad3 (de Caestecker et al., 2000), Dr. Roberts and colleagues hypothesized that the retention of Smad3 in fully malignant cell may provide some advantages and mediate key oncogenic events in the context of metastasis. To test their hypothesis, they manipulated the MII through MIV cells to stably overexpress either wildtype Smad3 or a C 83

terminally truncated dominant negative mutant of Smad3. Through the in vitro and in vivo tumorigenesis and metastasis analyses, they found that overexpression of Smad3 could inhibit tumorigenesis in the earlystage breast cancer cell lines (MII and MIII), but promote metastasis in the latestage one

(MIV). In contrast, the inhibition of endogenous Smad3 signaling by dominant negative Smad3 showed the opposite effects (Tian et al., 2003). The significance for this study is twofold. First, it identified a specific signaling molecule that can carry out the dichotomy of TGF β, i.e. , Smad3. Second, it demonstrated that the oncogenic activity of Smad3 is advantageous for the late stage cancers. Smad3 has been shown to interact with various transcription factors under different conditions to regulate target gene expression in response to TGFβ. Recently, ATF3 was found to be induced in the human mammary epithelial cell line, MCF10A, in response to TGFβ treatment and then associated with Smad3 to repress the expression of a TGFβtarget gene — Id1 (Kang et al.,

2003). A dominant negative form of ATF3 inhibits the downregulation of Id1 by

TGF β (Kang et al., 2003). This finding demonstrates the role of ATF3 as an integral component in the TGF β signaling. In combination with the previously demonstrated dichotomous role of ATF3 in carcinogenesis (Chapter 2) and

Smad3 as an important mediator of TGF β dichotomy in cancer development

(Tian et al., 2003), Kang’s study prompted us to hypothesize that ATF3, potentially through the cooperation with Smad3, can mediate the TGF β dichotomy in cancer. In this study, we focused on the potential involvement of

ATF3 in TGFβregulated cell migration, an important step in tumor metastasis, 84

since tumor metastases are responsible for most patient deaths related to breast cancer, and both ATF3 and TGFβ can promote tumor metastasis.

3.3. MATERIALS AND METHODS

Cell culture and treatments

COS1 cells were maintained in Minimum essential media (MEM) supplemented with 10% FBS. Transfection of COS1 cells was carried out using

Fugene 6 transfection reagent (Roche) following the manufacturer’s instruction.

MIV cell and its stable derivatives (MIV/Vec and MIV/ATF3) were established and cultured as described previously (Chapter 2). For treatment of the cells with various kinase inhibitors (Calbiochem), 2 х 10 5 cells were split into 6well plate.

24 hours (hrs) later, pretreat the cells with DMSO (vehicle control), SB202190 at a final concentration of 10 M, LY294002 at a final concentration of 10 M, JNK inhibitor II at a final concentration of 1 M, or PD98059 at a final concentration of

20 M for 2 hrs, then add TGFβ (R&D Systems) to a final concentration of 2.5 ng/mL for different time periods. ATF3, Smad2 and Smad3 siRNAs each containing a pool of four siRNA duplexes and control siRNAs (Dharmacon) were transfected into the MIV cells according to the instructions from the manufacturers.

Coimmunoprecipitation (CoIP) and immunoblot analysis

CoIP experiment was carried out as previously described (Kang et al., 2003).

Briefly, whole cell lysates were prepared in HKMG buffer (100 mM Hepes [pH 85

7.9], 100 mM KCl, 5mM MgCl 2, 0.1% NP40 and 10% glycerol) containing protease and phosphatase inhibitors (Sigma). After preclear with protein A sepharose beads (Sigma) twice, the whole cell lysate was incubated with indicated antibodies overnight at 4 °C with rocking followed by immunoprecipitation with protein A sepharose for 2 hrs. The protein complexes immunoprecipitated were washed with HKMG buffer 3 times followed by elution with 2% SDS. The antibodies used for immunoprecipitation and immunoblot include antiATF3 (Santa Cruz), antiactin (Sigma), antiEcadherin, antiSmad3, antiphosphoSmad3, antiphosphoErk, antiErk, antiphosphoATF2, antiATF2, antiphosphoAkt, antiAkt (Cell Signaling), and antiSmad2/3 (BD Biosciences) antibodies. Horse radish peroxidaselinked secondary antibodies (Cell Signaling) and LumiLight western blot substrate (Roche) were used to detect the signals according to the manufacturers’ instructions.

Chromatin immunoprecipitation (ChIP) and sequential ChIP analysis

ChIP analysis was carried out as previously described (Lu et al., 2006b). For sequential ChIP analysis, the proteinDNA complexes from the first immunoprecipitation was eluted with 10 mM DTT at 37 °C for 30 min, diluted with sonication buffer 40 times and proceed to standard protocol with the secondary antibody.

3.4. RESULTS

3.4.1. ATF3 is induced by multiple stromal signals

As shown in Chapter 2, ATF3 can be induced in MIV cells by TGFβ.

Consistent with the nature of ATF3 as an adaptive response gene, the induction of ATF3 is not unique to TGFβ. Other stromal factors, such as TNFα and IL1β, can also induce ATF3 in MIV cells (Figure 3.2.1).

3.4.2. ATF3 expression in MIV cells induces a morphological alteration resembling epithelialmesenchymal transition

During the process of establishing MIV cells stably expressing ATF3, I noticed that the MIV/ATF3 cells adopted a morphological change from tightly compacted, cobblestoneshaped epithelial cells to loosely associated, elongated, fibroblastlike cells (Figure 3.2.A). This morphological alteration resembles the epithelialmesenchymal transition (EMT), an essential developmental process involved in mesoderm and neural tube formation. EMT is characterized by loss of epithelial cell adherens junction, downregulation of Ecadherin, and increased cell motility (Thiery, 2002). Consistent with a morphological alteration resembling

EMT, MIV/ATF3 cells also have a lower Ecadherin level (Figure 3.2.B) and an enhanced migration compared with MIV/Vec cells (Figure 2.4.A). Since TGFβ is a key regulator of EMT and it can induce ATF3 in MIV cells (Figure 2.1.A), we further test the possibility that ATF3 may mediate the effect of TGFβ on regulating EMT and cell motility.

3.4.2. TGFβ induces ATF3 through multiple signaling pathways

TGFβ is known to signal through both the canonical Smaddependent pathway and several noncanonical Smadindependent pathways. To dissect the signaling pathways downstream of TGFβ to induce ATF3, I either inhibited the cononical pathway by siRNAs or noncanonical pathways by various inhibitors:

PI3K (LY294002), JNK (JNK Inhibitor II), p38 MAPK (SB202190) and MEK1

(PD98059). I examined the effect of those treatments individually on ATF3 induction in response to TGFβ. As shown in Figure 3.3.A to E, none of the above treatments significantly affect the induction of ATF3 by TGFβ, though they did effectively reduce the phosphorylation of corresponding targets, suggesting multiple pathways might be utilized to induce ATF3. Inhibition of any single pathway is not sufficient to affect the induction of ATF3 by TGFβ.

3.4.3. ATF3 is sufficient and necessary for TGFβinduced target gene expressions and cell migration

To test whether ATF3 plays a role in the ability of TGFβ to enhance motility, I first examined whether ATF3 can regulate cell motility in response to TGFβ by

Boyden Chamber Migration assay. As demonstrated earlier, ectopic expression of ATF3 in MIV is sufficient to promote cell migration and invasion, and the effect on cell migration was further enhanced by TGFβ treatment (Figure 2.4.A). Since endogenous ATF3 can be induced with TGFβ treatment in MIV cells, I also examined the effect of ATF3 knockdown on TGFβmediated cell migration. As shown in Figure 3.4.A, ATF3 knockdown significantly reduced cell migration in 88

the absence or presence of TGFβ. The reduction of cell migration by ATF3 knockdown in the presence of TGFβ was statistically significant, but to a minor extent. The lesser extent of reduction on cell migration by ATF3 knockdown in the presence of TGFβ — compared to in the absence — could be explained by the level of ATF3, which was higher in the presence of TGFβ (Figure 3.4.B).

Next, I examined the effect of ATF3 in regulating the expression of some motilityrelated target genes of TGFβ by both gainoffunction and lossof function approaches. As shown in Chapter 2, FN, Twist, Snail and Slug are four direct target genes for ATF3 in MIV cells. Importantly, these four genes are also known to be regulated by TGFβ and contribute to the EMT effect induced by

TGFβ (Moustakas and Heldin, 2007). In the gainoffunction approach, the ectopic expression of ATF3 further enhanced the ability of TGFβ to upregulate these genes at 2 hrs posttreatment except for Twist (Figure 3.5). In the lossof function approach, knockdown of endogenous ATF3 with siRNA significantly inhibited the induction of Snail, Slug and Twist by TGFβ (Figure 3.6), suggesting

ATF3 is necessary for TGFβ to upregulate these genes. The upregulation of

FN by TGFβ is not affected by knockdown of endogenous ATF3, consistent with previous finding that TGFβ stimulates the transcription of FN through multiple pathways (Kucich et al., 2000).

3.4.4. ATF3 interacts with Smad2/3

Kang et al. demonstrated that, in response to TGFβ, ATF3 was induced and then interacted with Smad3 to regulate the expression of Id1 gene (Kang et al., 89

2003). In this study, I wanted to test whether the interaction between ATF3 and

Smad3 also contributes to the regulation of TGFβ target genes and cell migration.

I first examined the interaction between ATF3 and different Smad proteins. As shown in Figure 3.7.A, ATF3 interacted with Smad2, in addition to Smad3, but not with Smad 1, 4 and 6. The data on Smad5 is inconclusive, because it was not expressed well for reasons unclear at this point. Then I examined the endogenous interaction between ATF3 and Smad2/3 in MIV cells treated with

TGFβ. With TGFβ treatment, the total level of ATF3 was increased, leading to more ATF3 coimmunoprecipitated by Smad2/3 (Figure 3.7.B).

3.4.5. ATF3 and Smad3 cooccupy the promoter of target genes and reciprocally regulate the cell motility induced by each other

Though the optimal Smadbinding element (SBE) was mapped as “CAGAC”, the Smad proteins have weak (such as Smad3) or no (such as Smad2) affinity to

SBE. It is the interaction between Smads and other sequencespecific transcription factors that confers efficient binding to the promoters of target genes, allowing specific response to ligand in different cell types (Moustakas et al., 2001). By analyzing the promoters of potential ATF3 target genes involved in cell motility, we found that close to the potential ATF/CRE site, there are optimal

Smadbinding sites (Figure 3.8), suggesting ATF3 may cooperate with Smad2/3 to regulate a set of target genes involved in cell motility. To test the potential interaction between ATF3 and Smad3 on target gene regulation, I examined the cooccupancy of these two proteins on the promoters of FN, Twist, Snail and 90

Slug by sequential ChIP. As shown in Figure 3.9.A, in response to TGFβ treatment, ATF3 and Smad3 cooccupied on the Twist and Snail promoters. I couldn’t detect the binding of these proteins on FN and Slug promoters by sequential ChIP, which is difficult to conclude at this point whether it’s due to no cooccupancy of both proteins on these promoters or due to technical problems.

To further investigate the importance of potential ATF3Smad2/3 interaction in regulating cell motility, I knocked down Smad2 and Smad3 in cells ectopically expressing ATF3 (MIV/ATF3). As shown in Figure 3.9.B, the knockdown of

Smad2+Smad3 reduced the motility of the MIV/ATF3 cells to the level comparable to that of the MIV/Vec cells. Although this result, by itself, does not prove that the increased cell motility observed in the MIV/ATF3 cells is due to the interaction between ATF3 and Smad2/3, this, combined with the data from the interaction assay, strongly supports this notion. In a reciprocal experiment, I knocked down ATF3 in cells ectopically expressing Smad3 (MIV/Smad3). In a previous study, Robert’s group showed that MIV/Smad3 cells had higher cell motility than MIV/Vec cells. The knockdown of ATF3 in these cells significantly reduced the cell migration (Figure 3.9.C), again supporting the notion that ATF3 cooperates with Smad3 to regulate cell motility.

3.5. DISCUSSION

TGFβ is a pleiotropic cytokine with important functions in maintaining homeostasis and regulating many biological processes, such as cell morphogenesis, cell differentiation, cell cycle regulation, apoptosis, cell adhesion,

cell migration, cytokine secretion, remodeling of extracellular matrix (ECM), antigen presentation, NK cell function and Tcell proliferation (Siegel and

Massague, 2003). During cancer development, TGFβ switches its function from a potent cytostatic and apoptotic factor in the normal epithelia to an oncogenic and metastasispromoting factor in malignant cancer cells — a “JekyllandHyde” conversion. The loss of responsiveness to the growth inhibition by TGFβ in various cancers can be partially attributed to the mutational inactivation of components in the TGFβ signaling pathway (Glasgow and Mishra, 2008). These mutations include mutational inactivation of type II TGFβ receptor (TβRII) in retinoblastoma (Kimchi et al., 1988), gastrointestinal cancer and colorectal cancer (Glasgow and Mishra, 2008), mutation of Smad2 in colorectal carcinoma

(Eppert et al., 1996), and mutation or deletion of Smad4 in approximately 50% pancreatic cancer and 10% colon cancer (Seoane, 2006; Takagi et al., 1996).

However, in some cancers, inactivation mutations of TGFβ signaling components are rare events, such as breast cancer. In these cancer cells, TGFβ also exhibits the “JekyllandHyde” conversion. What contributes to this conversion is not well understood. By using the isogenic MCF10 series of breast epithelial cells,

Roberts and colleagues demonstrated that Smad3 is capable of mediating the functional switch as seen for TGFβ (Tian et al., 2003). In this study, we found that in the malignant MIV cells, ATF3, a transcription factor with dichotomous function in breast cancer development (Chapter 2), can mediate the effect of

TGFβ on target gene regulation and cell motility. Since ATF3 can be induced in both nontransformed and malignant breast epithelial cells, and ATF3 can

promote apoptosis in nontransformed breast epithelial cells in response to stress

(Chapter 2), it is possible that ATF3 is a downstream mediator for TGFβ dichotomy during cancer development.

Though Smad2/3 are transcription factors known to promote cancer cell motility and tumor metastasis (Kang et al., 2005; Tian et al., 2003), due to its nature as a weak DNAbinding molecule, it is important to identify the other transcription factor(s) that confer(s) the stronger DNAbinding activity and promoter selectivity to carry out its motilitypromoting functions. In this study, we found that ATF3 might be such a transcription factor. ATF3 interacts with

Smad2/3 in response to TGFβ treatment. The protein complex involving both

ATF3 and Smad3 resides on the promoter of Twist and Snail genes after TGFβ treatment. Knocking down either ATF3 or Smad2/3 significantly reduces the cell migration induced by each other. Besides regulating target genes involved in cell motility, whether ATF3 cooperates with Sma2/3 to affect other cellular processes needs further investigation. In MI cells, Kang et al. showed that ATF3 and

Smad3 corepress the expression of Id1 gene, which promotes cell cycle progression, pointing to a tumorsuppressing activity of the ATF3Smad3 complex in mediating TGFβ signaling (Kang et al., 2003).

ATF3, as a transcription factor from the ATF/CREB family, was also reported to regulate tumor metastasis. When first isolating the mouse ATF3 cDNA from the melanoma cell line B16F10, Tsuruo and colleagues found that transfection of ATF3 converts the parental cells from low to highmetastatic cells (Ishiguro et al., 1996). Their followup study using lossoffunction approach showed that for

colon cancer cells, HT29, ATF3 antisense oligonucleotide changed cell morphology to a rounder shape, inhibited the attachment, cell migration and invasion ability in vitro, and improved mouse survival in a xenograft mouse model

(Ishiguro and Nagawa, 2001; Ishiguro et al., 2000). Consistently, by comparing various cancer cell lines, Ishiguro and colleagues found that ATF3 was expressed at higher levels in the cell lines derived from metastatic sites than in those from original tumor sites (Ishiguro and Nagawa, 2000). In addition, ATF3 was also expressed at higher levels in colon tumor tissues than in adjacent normal mucosa, especially in the tumors that had invaded the lymphatic ducts and/or the vessels, while there was not a significant relationship between ATF3 expression and the other pathological features (Ishiguro and Nagawa, 2000).

In contrast to the above findings, several reports suggest a metastasis inhibitory effect of ATF3, again, demonstrating the dichotomous nature of ATF3 functions. First, ATF3 was shown to repress matrix metalloproteinase 2 (MMP2) expression in different cell types either directly or indirectly in response to different stress signals (Chen and Wang, 2004; Stearns et al., 2004; Yan et al.,

2002). MMP2 is an important metastasis and angiogenesispromoting factor

(Gupta et al., 2007; Minn et al., 2005); thus the repression of MMP2 by ATF3 suggests ATF3 may inhibit tumor metastasis. Correlative evidence supporting these findings is that, when the breast cancer cells HCC1395 were treated with

Genistein, one of the major isoflavones that potently inhibits the growth and metastasis of breast cancer, ATF3 was strongly induced while MMP2, MMP7, and CXCL12 were significantly downregulated (Lee et al., 2007). Bottone and

colleagues found that in HCT116 human colorectal cancer cells, ATF3 was up regulated by treatment with nonsteroidal antiinflammatory drugs (NSAID), troglitazone, diallyl disulfide, and resveratrol. Ectopic expression of ATF3 inhibited invasion, whereas antisense ATF3 increased invasion (Bottone et al.,

2005). Given all these conflicting data, it is important to dissect out the mechanism(s) by which ATF3 regulates tumor metastasis. Chapter 4 will examine one potential mechanism by which ATF3 regulates metastasis using breast cancer as a model.

In summary, the significance of this study is threefold. First, it identifies that a new transcription factor that has a dual role may also contribute to the dichotomous switch of TGFβ during cancer development. Second, TGFβ is one of the most important multifunctional cytokines in the stromal microenvironment for breast cancer. Our study provides a mechanistic understanding on how cancer cells interpret the stromal signals to elicit their functions. Third, it offers a potential explanation on how ATF3 promotes tumor metastasis in vivo . That is, in response to stromal factors such as TGFβ, ATF3 is upregulated in cancer cells and cooperates with other transcription factors to regulate the expression of various genes involved in cell motility.

(min) 0 45 90 180

ATF3 TGFβ Actin

ATF3 TNFα Actin (min) 0 15 30 45 90

ATF3 IL1β Actin

Fig. 3.1 ATF3 can be induced by multiple stromal factors. MIV cells were treated with TGFβ (2.5 ng/mL), TNFα (60 ng/mL) or IL1β (10 ng/mL) for different time courses and assayed by immunoblot for ATF3 and actin. 3.1 ATF3 can be induced by multiple stromal factors

A. MIV/Vec MIV/ATF3

B. MIV Vec ATF3

Ecadherin Actin

Fig. 3.2 ATF3 expression in MIV cells induces an EMT phenotype. A. M IV/Vec and MIV/ATF3 cells were photographed under phase contrast microscopy. B. Ecadherin protein level was assayed by Western blot in M IV/Vec and MIV/ATF3 cells. Actin was used as an internal control.

3.2 ATF3 expression in MIV cells induces an EMT phenotype

JNK A.DMSO Inhibitor II D. DMSO PD98059 TGFβ (h) 0 1 2 0 1 2 TGFβ(h)0 1 2 0 1 2 ATF3 ATF3 pcJun pErk cJun Erk Actin Actin

B. E. DMSO LY294002 SiCtrl SiSmad2+3 TGFβ (h) 0 1 2 0 1 2 TGFβ (h) 0 1 2 0 1 2 ATF3 ATF3 pAkt Smad2+3 Akt Actin Actin

C. DMSO SB202190 TGFβ(h)0 1 2 0 1 2 ATF3 pATF2 ATF2 Actin

Fig. 3.3 ATF3 is induced by TGFβ through multiple signaling pathways. A to D. MIV cells were pretreated with specific inhibitor for JNK (JNK Inhibitor II, A), PI3K (LY294002, B), p38 MAPK (SB202190, C) and MEK1 (PD98059, D) for 1 hr, followed by TGFβ (2.5 ng/mL) treatment for different time courses. DMSO treatment was used as vehicle control. The expression of indicated proteins were examined by immunoblot. E. MIV cells were transfected with 100 nM Control or Smad3+Smad2 siRNAs for 72 hrs before treated with TGFβ for indicated time courses. The expression of indicated proteins were examined by immunoblot. Shown is representative of three independent experiments. 3.3 ATF3 is induced by TGFβ through multiple signaling pathways

A. B. 1.6 MIV/siCtrl MIV MIV/siATF3 * TGF β: – + 1.2 siRNA: Ctrl ATF3 Ctrl ATF3

0.8 * ATF3 0.4 Relative Migration Relative 0 Actin TGFβ +TGFβ

Fig. 3.4 ATF3 is necessary for TGFβmediated cell migration. A. MIV cells were transfected with either control or ATF3 siRNA for 72 hrs and assayed for migration by Boyden chamber in the presence or absence of TGFβ. A595 reading from MIV/siCtrl cells in the absence of TGFβ was arbitrarily defined as 1. Quantitative data in all panels of this figure are the mean ± S.D. from three independent experiments. * p < 0.05, versus MIV/siCtrl. B. Whole cell extract was made from MIV cells treated as in A and immunoblotted for indicated proteins. 3.4 ATF3 is necessary for TGFβmediated cell migration

20 4 MIV/Ctrl Level 16 MIV/ATF3 3 12 2 mRNA 8 1 4 0

Relative 0

0h 2h 0h 2h FN1 TWIST1

12 9

Level 8 6

mRNA 4 3

Relative 0 0 0h 2h 0h 2h Snail Slug

Fig. 3.5 ATF3 is sufficient to regulate TGFβ target genes. The steadystate mRNA levels of FN1, TWIST1, Snail and Slug was assayed by quantitative RT PCR in MIV cells treated with 2.5 ng/mL TGFβ for 0 hr and 2 hr, respectively. The levels of each mRNAs in MIV/Vec cells with 0 hr TGFβ treatment were arbitrarily defined as 1. 3.5 ATF3 is sufficient to regulate TGFβ target genes

100

5 2.5 4 2 3 1.5 2 1 1 0.5 0 0 0 45 90 180 (min) 0 45 90 180 (min) FN1 TWIST1 12 8

8 6 4 4 2 0 0 0 45 90 180 (min) 0 45 90 180 (min) Snail Slug

6 4 MIVsiCtrl 2 MIVsiATF3

0 0 45 90 180 (min) ATF3

Fig. 3.6 ATF3 is necessary to regulate TGFβ target genes. MIV cells were transfected with control and ATF3 siRNA for 72 hrs and treated with 2.5 ng/mL TGFβ for different time courses. Total RNAs were extracted and analyzed by quantitative RTPCR for indicated mRNAs. The levels of each mRNAs in M IV/Vec cells with 0 hr TGFβ treatment were arbitrarily defined as 1. 3.6 ATF3 is necessary to regulate TGFβ target genes

101

A. B. IP: Flag Input IP FlagSmads: S1 S2 S3 S4 S5 S6 (10%) IgG ATF3 Smad2/3 IB: ATF3 TGFβ 0 2 0 2 0 2 0 2 IB: Flag Smad3 5% Input ATF3 IB: ATF3

Fig. 3.7 ATF3 interacts with Smad2/3. A. COS1 cells were transfected with different Flagtagged Smad proteins as indicated, immunoprecipitated with Flag antibody and immunoblot for either ATF3 or Flag. B. MIV cells were treated with 2.5 ng/mL TGFβ for 0 hr and 2 hr respectively, whole cell lysate were made and immunoprecipitated (IP) with indicated antibodies and immunoblot for either ATF3 or Smad3. 10% whole cell lysate was saved as input. 3.7 ATF3 interacts with Smad2/3

102

CRE/ATF Site

199 Gene Position Sequence +1 FN1 199 TGACGTCA 1827 48 TWIST1 48 TCACGTCA +1 1700 1827 GGAGGTCA +1 Snail 1700 CA ATGTCA +20 Slug +20 TGACTTCA +1 29301975 SLPI 1975 TGACCTCA 2930 TGACCTCA 3087 +1 3087 TGACCTCA 5512 3253 +1 PAI15512 TGAGGTCA 3253 TGGCGTCA 4708 4399 3697 2017 uPA 4708 TGACCTCA +1 4399 TGACGACA 3697 TGAGGTCA

2111 2017 TGAAGTCA +1 COL4A2 2111 TGAGGTCA 1958 903 CV1 903 TGACTG CA +1 1958 TGAGGCCA CRE/ATF site Smad site (CAGACA) Consensus TGACGTCA Transcription starting site

Fig. 3.8 Potential ATF3Smad2/3 target promoters. Schematic representations of the promoters are shown on the left, with the CRE/ATF sites and Smad sites indicated. The corresponding CRE/ATF sequences are indicated on the right, with the deviations from the consensus sequence underlined (Modified from Figure 2.6 with potential Smadbinding sites indicated). 3.8 Potential ATF3Smad2/3 target promoters

103

A. Input Sequential ChIP –– Ab: IgG S+A IgG S+A TGF β (hr): 0 2 0 0 2 2 Snail

TWIST1

B. C. 1.8 2.5 2 1.2 1.5 * * 0.6 1 0.5 0 0 Migration Relative ATF3 + + Smad3 + +

Smad2/3 siRNA ATF3 siRNA + +

Fig. 3.9 ATF3 cooperates with Smad2/3 to regulate target gene expression and cell motility. A. MIV cells were treated with TGFβ for indicated time and the binding of ATF3 and Smad3 was examined by sequential ChIP. IgG: control rabbit IgG; S: Smad3 antibody; A: ATF3 antibody. B. MIV/Vec (dotted bar) or M IV/ATF3 (closed bars) cells were transfected with control () siRNA or siRNAs targeting both Smad2 and Smad3 (+) for 72 hrs, respectively. Cell motility was examined by Boyden chamber migration assay. C. MIV/Vec (dotted bar), M IV/Smad3 (closed bars) cells were transfected with control () siRNA or siRNAs targeting ATF3 (+) for 72 hrs, respectively. Cell motility was examined by Boyden chamber migration assay. The migration of MIV/Vec cells transfected with control siRNA was arbitrally defined as 1. * p < 0.05 compared to the corresponding cells transfected with control siRNA. 3.9 ATF3 cooperates with Smad2/3 to regulate target gene expression and cell motility

104

CHAPTER 4

A POTENTIAL ROLE OF ATF3 IN STROMACANCER INTERACTIONS 4. A POTENTIAL ROLE OF ATF3 IN STROMACANCER INTERACTIONS

4.1. ABSTRACT

Results in Chapters 2 and 3 suggest that ATF3 is a mestastasispromoting factor, since it promotes motility in cultured cells (Chapter 2) and tumor metastasis in a xenograft mouse model (Chapter 3) . In this chapter, I present data providing potential explanations for the mestastsispromoting effect of ATF3.

I found that in both the primary tumors and the lung metastatic nodules generated in the xenograft model, ectopic expression of ATF3 leads to higher levels of macrophages and vesselforming endothelial cells. Importantly, in vitro coculture experiments indicate that MIV/ATF3 cells recruit more macrophages than MIV/Vec cells, consistent with the in vivo data. Quantitative realtime RT

PCR showed that cancer cells with have higher expression of several genes that involved in macrophage recruitment, such as CCL5, TGFβ and TNFα. In summary, data presented in this chapter support the notion that ATF3 promotes metastasis, at least in part, by regulating stromacancer interaction, specifically, by increasing macrophage recruitment and blood vessel formation.

105

4.2. INTRODUCTION

The interactions between tumor cells and their microenvironment play important roles in regulating tumor growth and metastastis. The tumor microenvironment is composed of various stromal cells, extracellular matrix and secreted factors (BenBaruch, 2003). Among the various stromal cell types, macrophages are an abundant immune population within the tumor micro environment (Balkwill and Mantovani, 2001; Coussens and Werb, 2002) and are known as tumorassociated macrophages (TAMs). TAMs have been demonstrated to play a key role in promoting tumor growth, angiogenesis, metastasis and inhibiting adaptive immunity (Allavena et al., 2008). This chapter will focus on the interaction between TAMs and cancer cells, and I will briefly introduce some background information on TAMs as follows.

TAMs are derived from the monocytic precursors in the blood (Mantovani et al., 1992). The precursor monocytes are recruited to the tumor cells in response to several tumorderived chemoattractants, including CCL2, CCL5, vascular endothelial growth factor (VEGF), platelet derived growth factor (PDGF), TGFβ and macrophage colony stimulating factor (MCSF). Once in the tumor microenvironment, the same set of factors are also responsible for the survival and differentiation of the monocytes (Allavena et al., 2008).

In response to cytokines and microbial products, macrophages express specialized and polarized functional properties, referred to as classically activated M1 and alternatively activated M2 macropahges (Gordon, 2003;

Mantovani et al., 2004). M1 macrophages are normally induced by interferonγ 106

(IFNγ), granulocytemacrophage colonystimulating factor (GMCSF), tumor necrosis factor (TNF) and microbial products such as lipopolysaccharide (LPS).

They are characterized by the production of highlevel interleukin (IL)12, IL23 and lowlevel IL10 cytokines (IL12 high , IL23 high , IL10 low ). M2 macrophages are activated by signals other than those for M1, such as IL4, IL13, IL10 and glucocorticoids, and they generally share an IL12 low , IL23 low , IL10 high phenotype

(Mantovani et al., 2005). Functionally, M1 macrophages produce pro inflammatory cytokines and confer resistance against infectious pathogens and tumors cells, while M2 macrophages suppress the inflammatory responses and adaptive immunity, promoting wound healing, angiogenesis and tissue remodeling (Gordon and Taylor, 2005; Mantovani et al., 2005). TAMs have been characterized to have the phenotype and functions similar to M2 macrophages

(Mantovani et al., 2002).

Using a transgenic mouse model lacking the gene for MCSF, a crucial factor for differentiation of macrophages, Pollard and colleagues demonstrated a causal relationship between the infiltrations of TAMs into the primary tumor and the development of invasive and metastatic carcinoma (Lin et al., 2001). By in vivo imaging technique, the Pollard’s and Condeelis’ groups illustrate that macrophages close to the blood vessels significantly attract tumor cells and assist tumor cell intravasation (Wyckoff et al., 2007). Mechanistically, they found that there is a MCSF/EGF paracrine loop involved in the comigration and invasion of macrophages and tumor cells (Goswami et al., 2005). Macrophages express EGF, which promotes the cancer cells to invade and upregulate the

107

expression of various genes, including MCSF. Upon secretion by the cancer cells, MCSF, in turn recruites macrophages and promotes their differentiation and expression of EGF. The blockade of either signaling is sufficient to inhibit the migration and invasion of both cell types (Goswami et al., 2005). Recently, using a mouse model of breast cancer, Pollard and colleagues found that TAMs are necessary for turning on the “angiogenic switch,” which results in the establishment of a highdensity vessel network between tumor and host circulation (Lin et al., 2006). All these studies provide explanations to the well documented clinical correlations between TAMs, tumor angiogenesis and poor prognosis in human breast cancers (Leek et al., 1996).

Data presented in Chapter 2 showed that the ectopic expression of ATF3 in the malignant epithelial MIV cells led to more metastatic nodules in the lung tissue in an orthotopic xenograft model (Figure 2.8). By analyzing the functions of ATF3 in TGFβ signaling (Chapter 3), we provide evidence, that in cancer cells,

ATF3 may function as a downstream signaling molecule to carry out the effects of various stromal signals, including those on cell motility and tumor metastasis.

Since ATF3 can regulate a wide range of target genes, including both intracellular and extracellular secreted molecules (Table 2.3), we hypothesized that the induction of ATF3 in cancer cells in response to stromal signals may also exerts feedback on stromal cells. Considering the importance of TAMs in regulating tumor angiogenesis and metastasis, we focuse on the TAM and examine whether ATF3 plays a role in cancerTAMs interactions.

108

4.3. MATERIALS AND METHODS

Tissue samples

Mouse tissue samples including both primary breast tumors and lung tissues are removed from the xenograft mouse model. Human breast tissues, including breast cancer tissue and nondiseased breast tissue from the same patient, were procured from female patients with breast cancer and acquired from the OSU

Comprehensive Cancer Center (TPSR) with the approval from the Institutional

Review Board (IRB). All tissues were fixed in formalin and embedded in paraffin.

Immunohistochemistry

Immunohistochemistry was performed following the protocol from Cell

Signaling Technology (MA, USA) with minor modifications. Briefly, tissue sections on slides were deparaffinized in xylene (10min, 2X), followed by rehydration in a diluted ethanol series (100%, 100%, 95%, 95%, 75%, 75%, 0%

5min each). After antigen unmasking, the sections were incubated in 3% hydrogen peroxide in deionized water (dH 2O) for 10 minutes at room temperature, washed in PBS three times (5 min each), and then blocked with avidin/biotin blocking kit followed by 10% normal goat serum (Vector

Laboratories, CA, USA) in a 0.1% NP40/PBS solution for 60 minutes at room temperature. The blocking solution was removed and the primary antibody diluted in the block solution was added and incubated overnight at 4 °C in a humidity chamber. After the primary antibody incubation, the sections were incubated with a biotinylated secondary antibody (1:200 dilution in the block 109

solution) for 30 minutes at room temperature in a humidity chamber. Following the secondary antibogy, the sections were incubated with the ABC elite reagent

(Vector Laboratories) for 30 minutes at room temperature in a humidity chamber and the color reaction was performed using the various substrate kits (Vector

Laboratories) according to the manufacturer’s recommendations. For double antigens labeling, after the first antigen staining and substrate development, the sections were blocked with avidin/biotin blocking kit again followed by the regular steps as for the first antigen. Antibodies used for immunohistochemistry include: antiATF3 (1:100, Atlas Antibodies, Stockholm, Sweden), antiF4/80 (1: 40,

Invitrogen, MD, USA), antiCD31 (1:50, Abcam, MA, USA), and antiCD68 (1:50,

Dako, CA, USA). The antigen unmasking methods used include 125 °C in pressure cooker for 10 min (antiATF3, antiCD31, antiCD68) and steaming for

60 min in a steamer (antiF4/80).

Quantification of immunohistochemistry signals

Five animals used for xenograft injection with either MIV/Vec or MIV/ATF3 cells and developed lung metastasis were used for immunohistochemistry analysis. For each antigen staining, at least 150 fields (x 100) for the primary tumors and 50 fields (x 100) for the lung metastases from each cellinjected group were quantified using NIH Image J software (Version 1.41h).

110

Collection of L929 conditioned medium

L929 cells were ordered from ATCC (Manassas, VA, USA) and maintained in

DMEM supplemented with 10% FBS (Invitrogen, MD, USA). To make conditioned medium, 0.8 x 10 6 L929 cells were plated into 15cm tissue culture plate in 20 mL of complete growth medium and cultred at 37 °C, 5% CO 2 for 6 days. The supernatant was collected and filtered through 0.2 µm filter.

Isolation and differentiation of bone marrow cells into macrophages

Bone marrow cells were isolated from the tibias and femurs of 5 to 6month old MMTVPyMT transgenic C57BL/6 mice (a generous gift from Dr. Clay B.

Marsh, Division of Pulmonary, Allergy, Critical Care & Sleep Medicine, the Ohio

State Univeristy). The bone marrow cells from one mouse were plated onto one

10cm tissue culture plate in DMEM medium with 10% lowendotoxin FBS and

20% conditioned medium from L929 cells, and allowed to differentiate for 6 days with medium changed every two days.

Coculture system

To coculture the macrophages and epithelial MIV cells, we used the transwell system from Corning (MA, USA). For gene expression and cytokine secretion analysis, the transwell membrane with 0.4 m pore size was used to allow exchange of soluble factors between the cells. For macrophage recuitment assay, the transwell membrane with 8.0 m pore size was used to allow migration of macrophage across the transwell membrane. For macrophage

111

recuitment assay, 1.5 to 2 million macrophages in 500 L media were loaded on top of the insert and 0.5 million MIV cells in 1 mL media were loaded on the bottom chamber. Migration of macrophages were assayed 24 hrs later.

Statistical analysis

Data comparision was done by the Student’s t test. p < 0.05 is considered statistically significant.

4.4. RESULTS

4.4.1. ATF3 recruits more macrophages in both primary tumors and lung metastatic nodules

In an orthotopic xenograft mouse model, I found that ectopic expression of

ATF3 in MIV cells leads to faster primary tumor growth and higher incidence of lung metastasis (Figure 2.8). To examine whether there is any correlation between the ectopic expression of ATF3 in MIV cells and macrophage recruitment, I carried out immunohistochemistry (IHC) staining of tissue sections from both primary tumors and lung metastatic nodules, which will be referred to as “MIV/Vec or MIV/ATF3 primary tumors and lung mets” accordingly. As illustrated in Figure 4.1, the MIV/ATF3 lung mets have more macrophages associated with them, as indicated by IHC staining for F4/80, a transmembrane protein primarily expressed in macrophages. The macrophages (F4/80) were quantified using the NIH Image J program as follows. Capture the IHC image

(Figure 4.2.A) at 100 x magnification, and subtract the background to facilitate 112

the analysis of foreground signals (Figure 4.2.B). Split the RBG images into red, green and blue channels, and select the channel with the best separation of the target signal (Figure 4.2.C). Adjust the threshold until the target signals are highlighted and convert the image into a “black and white” binary image (Figure

4.2.D). Then, count the pixel numbers of the signals and divide the number against the pixel number of the entire field to get the percentage of macrophage signals. Figure 4.3 shows the average percentage (%) signals of 5 primary tumors (left) and the average signals of at least 5 lung metastatic nodules (right,

N=5 for MIV/Vec group and N=7 for MIV/ATF3 group) derived from five mice per group. For each tumor, 4 sections were examined under 100X magnification;

50 representive fields for each primary tumor and at least 20 fields for each lung metastatic nodules were analzyed. As shown in Figure 4.3, there are significant more macrophages in both primary tumors and lung metastatic nodules derived from MIV/ATF3 cells than those from MIV/Vec cells (p < 0.001).

4.4.2. ATF3 enhances angiogenesis in both primary tumors and lung metastatic nodules

Since TAMs are thought to turn on “the angiogenic switch in breast cancer”

(Lin and Pollard, 2007), I next examined whether the tumors (both primary and metastatic) derived from MIV/ATF3 cells also associate with higher levels of vascular formation. I stained the tumor sections with CD31, a membrane glycoprotein predominantly expressed on the surface of endothelial cells, and quantified the CD31 signals using the NIH Imag J program as described above 113

(Figure 4.2). Tumors derived from MIV/ATF3 cells have more endothelial cells than those from MIV/Vec cells (Figure 4.4.A), and this difference is significant in both primary tumors and metastatic nodules ( p < 0.001, Figure 4.4.B).

To test whether macrophages and vessels are closely adjacent to the cancer cells with strong ATF3 expression, I examined the tumor sections by costaining for F4/80 (macrophages) and ATF3 or CD31 (vessels) and ATF3. Figure 4.5.A shows that macrophages are juxtaposed to cancer cells with strong ATF3 staining. This figure was taken at 400 x magnification and shows the intratumoral macrophages. It is different from Figure 4.1, which primarily shows peritumoral macrophages. Due to the lower magnification (64 x), the intratumoral macrophages are not as visible in Figure 4.1. Costaining for CD31 and ATF3 shows endothelial cells close to cancer cells with strong ATF3 expression (Figure 4.5.B). Taken together, these results, in conjunction with the literature on macrophage and angiogenesis, support the notion that cancer cells with strong ATF3 expression recruit macrophages, which in turn facilitats vessel formation.

4.4.3. An in vitro analysis of cancermacrophage interaction

To explore the potential mechanisms by which ATF3 expression in MIV cells may lead to more macrophages recruitment, I set up a coculture of MIV cells and macrophages in vitro using Boyden Chamber assay (Figure 4.6.A). To mimic the situation in mouse model, I isolated bone marrow cells from tumor bearing mice and differentiated them into macrophages. Five to sixmonth old 114

MMTVPyMT transgenic mice that have developed tumors at multiple mammary fat pads were used for bone marrow isolation. The differentiation was carried out using conditioned medium from mouse fibroblast L929 cells, which secrete macrophage colony stimulating factor (MCSF) (Genovesi et al., 1989). I first tested the differentiation efficiency of macrophages with the L929 conditioned medium. By staining the cells with F4/80, I found more than 90% of the cells are positive for F4/80 (Figure 4.6.B), suggesting a high differentiation efficiency. I then examined the migration of macrophage cells toward cancer cells. As shown in Figure 4.6.C, macrophages cocultured with MIV/ATF3 cells exhibited higher migration than those cocultured with MIV/Vec cells. This in vitro recruitment assay is consistent with the in vivo data that MIV/ATF3 primary tumors or lung metastatic nodules have higher number of macrophages than the MIV/Vec counterpart (Figures 4.1, 4.3 and 4.4).

Several cytokines and cytokine receptors, including CCL2 (also called monocyte chemotactic protein1, MCP1), CCL5, vascular endothelial growth factor (VEGF), platelet derived growth factor (PDGF), TNFα, TGFβ, EGF receptor (EGFR) and macrophage colony stimulating factor (MCSF), are suggested to promote macrophage recruitment to tumors (Allavena et al., 2008;

Goswami et al., 2005). By realtime RTPCR, I found that when cocultured with macrophages, the expression of CCL5, TGFβ1 and TNFα are significantly higher in MIV/ATF3 cells compared to MIV/Vec cells (Figure 4.6.D), suggesting that

ATF3 may promote the secretion of these chemoattractants and help the tumor cells to recruit more macrophages.

115

4.4.4. ATF3expressing cancer cells juxtapose macrophages and endothelial cells in human breast cancer

To examine whether the above findings have any relevance to human breast cancer, I examined the tumors from breast cancer patients using immunohistochemistry. As shown in Figure 4.7, the strong ATF3expression cancer cells closely associate with both macrophage (as indicated by CD68 staining) and endothelial cells (as indicated by CD31 staining), consistent with my finding in the xenograft mouse model and in vitro coculture system.

4.5. DISCUSSION

As discussed in Chapter 3, currently the literature provides both support and arguement on the function of ATF3 in promoting tumor metastasis. However, thus far, no reports have provided any mechanistic understanding. In work presented in Chapter 2, we found that ATF3 can promote the metastasis of breast cancer cells in a spontaneous metastasis model in nude mice. In this chapter, I examined the involvement of ATF3 in stromacancer interactions with a focus on macrophagecancer interactions. To our knowledge, this is the first demonstration that the upregulation of ATF3 in cancer cells associates with higher macrophage recruitment and endothelial cells presence. The in vitro co culture system suggests that one potential mechanism for ATF3 to facilitate this macrophage recruitment is through the upregulation of chemoattractant cytokines such as CCL5, TNFα and TGFβ. 116

During the process of this work, several studies examined the roles of ATF3 in metastasis, innate immune responses and angiogenesis. These studies have direct relevance to the work presented here and I will discuss them separately below.

Metastasis. Watabe and colleagues showed that, in prostate cancer, ATF3 expression was significantly inhibited by the induction of tumor metastasis suppressor gene Drg1 (Bandyopadhyay et al., 2006). When injected into the severe combined immunodeficient (SCID) mice subcutaneously, prostate cancer cells with ectopic expression of ATF3 significantly enhanced lung metastasis without affecting primary tumor growth. The immunohistochemical analysis of human prostate cancer specimens revealed that the expression of ATF3 has an inverse correlation with Drg1 expression and a positive correlation with metastases (Bandyopadhyay et al., 2006) . This study is consistent with our finding that ATF3 is a metastasispromoting factor. Since the studies are derived from different cancer types, prostate and breast cancer), they indicate that the function of ATF3 in tumor metastasis may not be specific to certain cancer types.

Innate immune response. The study presented in this chapter is on the function of ATF3 in cancer epithelial cells. However, as described in Chapter 1,

ATF3 is induced in a variety of cell types by stress signals. Relevant to the cancermacrophage interactions I presented in this chapter is the recent studies of ATF3 functions in innate immunity. Aderem and colleagues combined systems biology approaches with complementary DNA (cDNA) microarray to identify genes mediating Tolllike receptor (TLR) signaling in macrophages

117

(Gilchrist et al., 2006). Based on the cDNA microarrays in macrophages activated by TLR4 agonist bacterial lipopolysaccharide (LPS), the authors did cluster analysis and identified ATF3 as a key negative regulator of important pro inflammatory cytokines, such as IL6, IL12b and TNFα in LPSstimulated responses (Gilchrist et al., 2006). At about the same time, Williams and colleagues showed that ATF3 is induced by various TLRs in both macrophages and plasmacytoid dendritic cells (DCs) and negatively regulates TLR signaling

(Whitmore et al., 2007). Weinstein and colleague found ATF3 that negatively regulates CCL4 expression in murine macrophages to prevent excessive inflammation (Khuu et al., 2007). Thus, all three studies suggest that ATF3 dampens the LPSinduced innate immune response. Given the difference between LPSinduced M1 and alternatively activated M2 macrophages, the exact functions of ATF3 in TAMs, which are M2 macrophages, remains to be investigated. Although the work presented in this chapter focuses on cancer epithelium, we believe ATF3, as an adaptiveresponse gene, is also induced in macrophages and plays important roles.

Angiogenesis. Several reports explored the roles of ATF3 in endothelial cells and thus angiogenesis. By applying cDNA microarray analysis to nontubulogenic and tubulogenic cells in a variety of culture conditions to look for genes up regulated in tubulogenic conditions, ATF3 showed up with high scores

(Kobayashi et al., 2004). This correlative evidence was supported by the study of Maru’s group using both gainoffunction and lossoffunction approaches, where they demonstrated that ATF3 is upregulated by reactive oxygen species

118

(ROS) in endothelial cells and plays a role in diabetic angiopathy (Okamoto et al.,

2006). Maru and colleagues found that in endothelial cells ATF3 upregulates

EphA1, a receptor tyrosine kinase involved in neuronal pathfinding, boundary formation, and angiogenesis during development and tumor progression (Masuda et al., 2008). This study provides a potential mechanism by which ATF3 regulates angiogenesis (Masuda et al., 2008). In this chapter, we present another potential mechanism by which ATF3 enhances angiogenesis, i.e., through the upregulation of chemoattractive cytokines to recruit more macrophages, which in turn promotes angiogenesis.

All these studies, combined with our current work, suggest that ATF3 may function in different cell types including cancer epithelial cells, TAMs and endothelial cells, and regulate different sets of target genes to provide a nourishing microenvironment for tumor growth and metastasis.

119

Ab: ATF3 F4/80

MIV/ Vec T T

MIV/ T T ATF3

Fig. 4.1 Ectopic ATF3 expression in MIV cells correlates with more macrophages recruitment in the lung metastatic nodules. Lung tissues from MIV/Vec (upper) and MIV/ATF3 cells injected mice were stained by immunohistochemistry for ATF3 and F4/80 (brown signals). T: Lung metastatic nodules. 4.1 Ectopic ATF3 expression in MIV cells correlates with more macrophages recruitment in the lung metastatic nodules

120

A. B.

C. D.

Fig. 4.2 Quantification of immunohistochemistry signals. A. A picture of F4/80stained slide taken at the magnification of x 100. B. Using NIH Image J program, first subtract the background. C. Split the RGB image into red, green and blue channels and select the channel with the best separation of the target signal. D. Adjust the threshold till the target signals are highlighted and convert the image to binary image. Then measure the area (pixels) of the highlighted regions. 4.2 Quantification of immunohistochemistry signals

121

10 MIV/Vec MIV/ATF3 * 8 * 6 4

Signals F4/80 % 0

Primary Tumor Lung Mets.

Fig. 4.3 Ectopic ATF3 expression in epithelial cancer cells correlates with more macrophages recruitment in both primary tumors and lung metastatic nodules. Primary tumors (N=5 from each group) and lung metastatic nodules (N=5 for MIV/Vec group and N=7 for MIV/ATF3 group) were embedded in paraffin and 4 sections from each tumor were examined under 100 x magnification; 50 representive fields for each primary tumor and at least 20 fields for each lung metastatic nodule were analyzed for F4/80 signals. Quantification of F4/80 signals was carried out as described in Fig. 4.2 in both primary tumors and lung metastatic nodules (Lung Mets.) from MVI/Vec (□) or MIV/ATF3 (■) cells injected mice. * p < 0.001 compared to MIV/Vec cells. 4.3 Ectopic ATF3 expression in epithelial cancer cells correlates with more macrophages recruitment in both primary tumors and lung metastatic nodules

122

A. MIV/Ctrl MIV/ATF3

8 MIV/Vec MIV/ATF3 6 * *

% CD31 Signals % 0 Primary Tumor Lung Mets.

Fig. 4.4 Ectopic expression of ATF3 in epithelial cancer cells correlates with higher level of angiogenesis in both primary tumor and lung metastatic nodules. A. A representative figure of CD31 immunohistochemistry (brown signals) in lung metastatic nodules from MIV/Vec (upper) and MIV/ATF3 cellsinjected mice. B. Primary tumors (N=5 from each group) and lung metastatic nodules (N=5 for MIV/Vec group and N=7 for MIV/ATF3 group) were embedded in paraffin and 4 sections from each tumor were examined under 100 x magnification; 50 representive fields for each primary tumor and at least 20 fields for each lung metastatic nodule were analyzed for CD31 signals. Quantification of CD31 signals was carried out as described in Fig. 4.2 in both primary tumors and lung metastatic nodules (Lung Mets.) from MVI/Vec (□) or MIV/ATF3 (■) cells injected mice. * p < 0.001 compared to MIV/Vec cells. 4.4 Ectopic expression of ATF3 in epithelial cancer cells correlates with higher level of angiogenesis in both primary tumor and lung metastatic nodules

123

A. B.

Blue = ATF3 Blue = ATF3 Brown = F4/80 Brown = CD31

Fig. 4.5 Breast cancer cells with higher ATF3 expression closely associate with macrophages and endothelial cells in xenograft mouse model. A. Co staining of the MIV/ATF3 primary tumors for ATF3 (blue signal) and F4/80 (brown signal) (400 x). B. Costaining of the MIV/ATF3 primary tumors for ATF3 (blue signal) and CD31 (brown signal) (400 x).

4.5 Breast cancer cells with higher ATF3 expression closely associate with macrophages and endothelial cells in xenograft mouse model

124

A. B. C. Coculture Experiment Transwell 2 * Insert Migrating 1 MΦ * 0 Cancer Vec ATF3 Cells

Relative Migration (MΦ) Relative MIV D. 9 8 * 7 MIV/Vec 6 MIV/ATF3 5 4 * * 3 2 1 Relative mRNA Level mRNA Relative 0 CCL2 CCL5 MCSF EGFR TNFα TGFβ1 VEGF

Fig. 4.6 MIV/ATF3 cells promotes macrophage migration and upregulates several chemoattractive cytokines. A. A schematic representation of the set up for the coculture experiment. Macrophages were loaded on top of the transwell insert and the MIV cancer cells on the bottom. B. Immunostaining of macrophages differentiated from bone marrow precursor cells using F4/80 antibody (brown signal) and counterstained with hemotoxyline. C. Migration of macrophages toward MIV cells (Vec. and ATF3) was determined by Boyden chamber assay. D. MIV cells were cocultured with macrophages for 24 hrs and the expression levels of indicated target genes were examined by realtime PCR. The mRNA levels in MIV/Vec cells were arbitrally defined as 1. * p < 0.05 compared to MIV/Vec cells.

4.6 MIV/ATF3 cells promotes macrophage migration and upregulates several chemoattractive cytokines

125

A. B.

Blue = ATF3 Blue = ATF3 Brown = CD68 Red = CD31

Fig. 4.7 Breast cancer cells with higher ATF3 expression closely associate with macrophages and endothelial cells in human breast cancers. A. Co staining for ATF3 (blue signal) and F4/80 (brown signal)(400 x). B. Costaining for ATF3 (blue signal) and CD31 (dark red signal)(400x). 4.7 Breast cancer cells with higher ATF3 expression closely associate with macrophages and endothelial cells in human breast cancers

126

CHAPTER 5

RELEVANCE OF ATF3 TO HUMAN BREAST CANCER 5. RELEVANCE OF ATF3 TO HUMAN BREAST CANCER

5.1. ABSTRACT

As desribed in Chapters 2, 3 and 4, ATF3 is capable of enhancing in vitro cell survival, cell migration and cell invasion, mediating TGFβ signaling in the cancer cells, contributing to cancer cellmacrophage interaction, and promoting in vivo primary tumor growth and metastasis. However, since all the above data are derived either from cultured cells or from animal model, it is not clear whether

ATF3 is involved in human breast cancers. Work described in this chapter addresses this important question. I examined the status of ATF3 in breast tumors from patients. By comparing between breast cancer tissues and non diseased matching control tissues, I found that ATF3 expression is upregulated in approximately 50% (23/48), and ATF3 gene is amplified in about 80% (38/48) of human breast cancers. The Oncomine database analysis of ATF3 steady state mRNA level in breast tumors with clinical outcome data suggests that higher ATF3 expression correlates with worse outcome of the patients. These results support the notion that ATF3 is an oncogene in human breast cancer.

127

5.2. INTRODUCTION

Breast cancer, similar to all other cancer, develops as a result of accumulative genetic and epigenetic alterations. The cancer cells develop mutiple mechanisms to activate or upregulate oncogenes and delete or inhibit tumor suppressor genes. One common mechanism utilized by cancer cells is the alterations of chromosomes. By changing the copy number of specific chromosomal regions, genes residing within those regions can be gained or lost.

The genes within a chromosomal region that are always amplified together constitute an amplicon. Comprehensive cytogenetic analyses have been applied to understand chromosomal aberrations, define commone amplicons and help to identify functionally important genes associated with tumorigenesis (Vogel,

1979). For human breast cancer, comparative genomic hybridization (CGH) analyses in both primary breast tumors and breast cancer cell lines revealed that the most frequent gains are 1q and 8q amplicons, while the most frequent losses are 8p, Xp, 16q, Xq, 18q regions (Forozan et al., 1999; Tirkkonen et al., 1998).

However due the limited resolution of CGH (approximately 5 – 10 Mb, (Rooney et al., 1999), it is difficult to conclude on the gain or loss of individual genes within the chromosome regions. Although some genes important for breast cancer have been mapped to these chromosomal regions, such as MYC gene in chromosome 8q24.1 (Bieche and Lidereau, 1995), most positional candidate genes are not identified.

1q was identified as the second most frequently amplified region of many human solid tumors, based on metaanalysis of 100 papers describing 128

comparative genome hybridization (CGH) findings in > 2,000 solid tumors of 27 cancer types (Table 5.1) (Rooney et al., 1999). In another study, still using CGH technique, by analyzing 55 human primary breast carcinomas, Tirkkonen et al. found that 67% of the samples have 1q amplification, with a minimal amplicon covering 1q2432 region (Tirkkonen et al., 1998). This 1q gain shows no significant association with any clinicopathological features, such as tumor size, tumor grade, nodal status, ER/PR status, DNA index or Sphase fraction, nor does it associate with other chromosomal aberrations, such as 8q gain.

However, among earlystage carcinomas, 1q gain is the most common or the only genetic aberration, suggesting that it might be an early and probably important step in the development of human breast cancers (Tirkkonen et al.,

1998).

Intriguingly, ATF3 gene is localized in the human chromosome at 1q32.3, within the 1q amplicon. However, it is not clear whether ATF3 is simply a bystander or a functionally important gene for breast cancer, as suggested by its functions in the breast cancer cell lines and the orthotopic xenograft mouse model (Chapters 24). To address the functional importance/relevance of ATF3 to human breast cancer, in this study, I examined the status of ATF3 in human breast cancer, including its gene amplification status, expression level, mutation status, and correlation with patient outcome and other welldocumented breast cancer biomarkers.

129

5.3. MATERIALS AND METHODS

Human breast tissues

Human breast tissues, including breast cancer tissue and nondiseased breast tissue from the same patient, were procured from female patients with breast cancer and acquired from the OSU Comprehensive Cancer Center

(TPSR) with the approval from the Institutional Review Board (IRB).

Preparation of tissue lysate and immunoblot analysis

To prepare tissue lysate, 0.1 to 0.3 g of breast tissues was ground into powder in liquid nitrogen with a tissue pulverizer (ColeParmer, Vernon Hills,

Illinois). Powder was dissolved in RIPA buffer (50 mM TrisHCl [pH 7.4], 1% NP

40, 0.1% SDS, 150 mM NaCl, 1 mM EDTA and 0.25% sodium deoxycholate) with protease inhibitors (Sigma, MO, USA). To promote the sample lysis, sonication was used followed by centrifugation at 12 000 g at 4 °C for 15 min.

The supernatant (tissue lysate) was kept and protein concentration determined by Bradford protein assay (BioRad, Hercules, California, USA). Equal amount of total proteins (3050 g) were analyzed by immunoblot using antiATF3, antip53, antiEcadherin (Santa Cruz), and antiactin (Sigma) antibodies.

Gene amplification and mutation analysis

To examine the gene amplification and mutation status of ATF3 in human breast cancer samples, genomic DNA was isolated from the samples using

DNeasy Tissue Kit (Qiagen). 100 ng total genomic DNA was used for qPCR 130

analysis (for gene amplification) and regular PCR analysis (for gene mutation analysis). The regular PCR fragments were sent for sequencing at OSU Plant

Microbe Genomics Facility. The primers for gene amplification analyses are:

ATF3: upstream 5’ ATGCCCCAGAGCCCCTGAAA 3’, downstream 5’

TACCTGTGATAGAACCCACC3’; betaactin (internal control): upstream 5’

AAGATCAAGATCATTGCTCCTC 3’, downstream 5’

GGACTCATCGTACTCCTG 3’. The primers for gene mutation analyses are:

ATF3 exon B: upstream 5’ ATGCCCCAGAGCCCCTGAAA 3’, downstream 5’

TACCTGTGATAGAACCCACC 3’; ATF3 exons C to E: upstream 5’

GGTAGGTGGCATTTCTTACTGC 3’, downstream 5’

TCTCCAATGGCTTCAGGGTT 3’; p53 Exon 2 to 4: upstream 5’

GTTCCTTCTCTGCAGGCCCA –3’, downstream 5’

GGCCAGGCATTGAAGTCTCAT –3’; p53 Exon 5 to 6: upstream 5’

TGTTCACTTGTGCCCTGACT –3’, downstream 5’

CTGACAACCACCCTTAACCC –3’; p53 Exon 7 to 9: upstream 5’

CTTGCCACAGGTCTCCCCAA –3’, downstream 5’

GCCCCAATTGCAGGTAAAACAG –3’; p53 Exon 10 to 11: upstream 5’

TCTACTAAATGCATGTTGCT –3’, downstream 5’ ACCCAAAACCCAAAATGG

–3’.

131

Database analysis

ATF3 gene expression levels and clinical data from Sorlie’s study (Sorlie et al., 2001) and van 't Veer’s study (van 't Veer et al., 2002) were retrieved from

Oncomine database ( www.oncomine.org ) and analyzed using MedCalc software.

Statistical analysis

Data comparision was done by the Student’s t test or Fisher’s exact test. p <

0.05 is considered statistically significant.

5.4. RESULTS

5.4.1. ATF3 is upregulated in approximately 45% of human breast cancers

In Chapter 2, I surveyed a panel of human breast cancer cell lines and found that ATF3 is upregulated in approximately 90% of the cell lines examined. Due to the limitations of cell lines to represent the human breast cancer, it is necessary and important to examine the breast tumor tissues to gain more information on the potential relevance of ATF3 to human breast cancer

(discussed in Chapter 2). I first examined the expression of ATF3 by immunoblot in human breast tumors. To evaluate the expression level of ATF3, I used the nondiseased “normal” breast tissue from the same patient as control. The normality of the control tissue was determined based on pathological analysis. If the ATF3 protein level is higher in the tumor sample than in the control sample, I call it positive for ATF3; otherwise, I call it negative for ATF3. Among the 48 tumors examined, 23 (48%) expressed ATF3 at a level higher than their 132

matching nondiseased controls (ATF3 positive). Figure 5.1.A shows the results from the 23 ATF3positive tumors (Patient ID 123) and a representative result of a few negative samples (Patient ID 2427). One caveat of comparing tumors to normal breast tissues is that the predominant cells in the normal breast tissues are adipocytes but the predominant cells in the tumors are epithelial cells. Thus, it is possible that the elevated level of ATF3 is due to its differential expression in epithelium versus in adipocytes. To test this possibility, I examined proliferating primary human mammary epithelial cells (HMECs) (a generous gift of Dr. Martha

Stampfer at The Lawrence Berkeley National Laboratory). We used early passage cells that are at the prestasis stage (or before senescence). As shown in Figure 5.1.B, ATF3 was not detected in the primary breast epithelium. Six breast tumors and MIV cells were included for comparison. This result is important for two reasons. First, it indicates that the difference between the normal and tumor samples is not simply due to the higher percentage of epithelium in the tumors. Second, the HMECs at the passage number we used proliferate well in vitro , yet they do not express ATF3 to any appreciable degree.

Thus, the elevated expression of ATF3 in tumor reflects some features of the breast carcinoma, not simply their proliferation. In addition, the ATF3 protein levels increased upon the treatment of HMECs by TNFα or TGFβ (Figure 5.1.C), indicating that the programs to induce ATF3 expression are functional in the

HMECs.

133

5.4.2. ATF3 gene is amplified in approximately 80% of human breast cancers

The percentage of tumors with elevated ATF3 expression in our study is 48%, close to that of human breast tumors with 1q amplification (53%) as determined in Rooney’s study (Rooney et al., 1999). I thus examined the copy number of

ATF3 genes by qPCR in the tumors. As controls, I used MI cells, which have 2 copies of 1q, and MIV cells, which have 3 copies of 1q (Santner et al., 2001).

They allowed me to estimate the accuracy of the qPCR results. Only qPCR results with expected ratio (2:3) for MI vs. MIV cells were collected and a representative of three repeats is shown (Figure 5.2.A). Among the 23 tumors with elevated ATF3 protein levels (referred to as group A), 21 (>90%) had ≥ 3 copies of the gene, 16 had ≥ 5 copies and 7 had >10 copies (Figure 5.2.A upper panel and 5.2.B). Since the tumor samples contained both the stroma and cancer epithelia, we isolated the cancer epithelia by laser capture microdissection (LCM) and analysed the ATF3 gene copy number. As shown in

Figure 5.2.A inset, all four samples examined by this method showed ATF3 gene amplification. Although our LCM analysis was not comprehensive, our results — in the context of welldocumented DNA amplification in cancer cells — support the notion that the ATF3 gene is amplified in the breast cancer epithelia. I also examined the 25 tumors without elevated ATF3 expression (group B) and found

17 of them (68%) had ≥ 3 copies of the gene, 7 had ≥ 5 copies and 2 had >10 copies (Figure 5.2.A lower panel and 5.2.B).

134

5.4.3. ATF3 has no mutations in the open reading frame in human breast tumors

The above data indicate that ATF3 gene is frequently amplified and its expression is elevated in almost half of the breast tumors we examined.

However, they do not necessarily mean that ATF3 is an oncogene. It is well documented that p53 gene — a tumor suppressor — is mutated and its expression elevated in some tumors (1990; Finlay et al., 1989). Since ATF3 can be proapoptotic in the MI untransformed breast epithelium, it is important to find out whether the ATF3 gene is mutated in the breast carcinoma. Thus, I analyzed the intronextron structure of ATF3 gene, designed PCR primers to cover all the coding exons for ATF3 (Figure 5.3), and sequenced both strands of the open reading frame amplified by highfidelity PCR for all 23 tumors that had elevated expression of ATF3. No mutation was found.

5.4.4. Higher ATF3 steadystate mRNA levels correlate with worse clinical outcomes of breast cancer patients

Next, I examined whether it is possible to use ATF3 as a biomarker for predicting the outcome of patients with breast cancer in clinic. Several large scale comprehensive gene expression analyses have been carried out to characterize the correlation between gene expression patterns to clinical outcomes for breast cancer (Sorlie et al., 2001; van 't Veer et al., 2002; Wang et al., 2005). These studies are invaluable tools for predicting or analyzing functionally important genes in breast cancer (Lauss et al., 2008). In the present 135

work, I retrieved the ATF3 gene expression levels from these two microarray studies and analyzed its correlation with clinical outcomes. From Figure 5.4. AD, we can see that higher ATF3 expression from Sorlie’s and van’t Veer’s studies significantly correlates with worse clinical outcomes, including fiveyear survival

(Figure 5.4.A and C), relapsefree survival (Figure 5.4.B) and tenyear survival

(Figure 5.4.D). However, the higher ATF3 expression level from Wang’s study does not significantly correlate with worse relapsefree survival of patients ( p =

0.156)(Figure 5.4.E). One factor that may account for the difference between

Wang’s study and the other two studies is that all breast cancer patients from

Wang’s study are negative for lymphnode metastasis, suggesting early stage of cancer development. Since ATF3 has proapoptotic activity in untransformed MI cells, ATF3 may also function to inhibit early development of breast cancer.

Therefore, higher ATF3 expression in the early stage may correlate with better patient outcome, in contrast to its function in latestage breast cancer.

5.4.5. Higher ATF3 expression level correlates with higher wildtype p53 level and lower Ecadherin level in human breast cancer

As an alternative approach to analyze the functional significance of ATF3 in human breast cancer, I examined the correlation of ATF3 expression to two other welldocumented breast cancer biomarkers: p53 and Ecadherin. p53 is the first tumor suppressor gene identified and is mutated in approximately 20% of human breast cancer (Mazars et al., 1992; Pharoah et al., 1999). p53 mutation is associated with more aggressive disease and worse overall patient survival 136

(Mazars et al., 1992; Sjogren et al., 1996). Ecadherin is a transmembrane glycoprotein involved in cell adhesion and important for epithelial morphogenesis

(Takeichi, 1991). Lower expression of Ecadherin was associated with poorly differentiated and metastatic breast carcinoma (Gamallo et al., 1993; Oka et al.,

1993; Shiozaki et al., 1991). Another important clue that led us to focus on E cadherin is that ATF3 upregulates several transcriptional repressors for E cadherin, such as Twist, Snail and Slug, and the ectopic expression of ATF3 in

MIV cells leads to lower Ecadherin expression (Figure 2.5.A and 3.1.B).

To examine the p53 expression in breast tumors, I applied immunoblot analyses and compared the p53 protein level in tumor sample to the matching nondiseased control sample. Using the same criterion for defining ATF3 positivity, if the p53 level in the tumor saple is higher than in the control sample, I call it positive. Otherwise, I call it negative. I found that 14 out of 23 breast tumor samples that are positive ATF3 are also positive for p53, while 24 out of 28 breast tumor samples negative for ATF3 are negative for p53 also, indicating a significant positive correlation between these two protein levels in human breast cancer (Figure 5.5)( p < 0.001). Since p53 is a tumor suppressor gene, I examined the p53 mutation status by sequencing both strands of the open reading frame (ORF) (Figure 5.6 and Table 5.2). Out of the 20 tumor samples with the whole p53 coding exons sequenced, there are 3 point mutations (aa47,

123 and 138) and 2 SNPs (aa35 and 213). The most common alteration seen is the polymorphism of aa72 Pro to Arg. Among the 14 tumor samples that are positive for both ATF3 and p53, 8/14 have aa72 Arg homozygotes and 2/14

137

having heterozygotes (Table 5.2). Similar immunoblot analysis was used to examine the correlation between ATF3 and Ecadherin (Figure 5.7), which showed a trend of negative correlation but not significant yet ( p = 0.26).

5.5. DISCUSSION

In summary, work presented in this chapter describes the ATF3 status in human breast cancer, including the expression level, gene amplification, mutation and correlation with clinical outcome or other breast cancer biomarkers.

Based on the ATF3 protein level compared between tumor samples (T) and non diseased control samples (C), we divided our tumor samples into two groups:

Sapmles in Group A have higher ATF3 in T than in C, while those in Group B do not. Gene amplification analysis showed that 17 samples in Group B have an

ATF3 copy number higher than the normal diploid. However, the distribution was more toward the low copy numbers (35), in contrast to the tumors in Group A that had distribution toward higher copy numbers (Figure 5.2.B). These results suggest that gene amplification is potentially a contributing factor for the elevated

ATF3 protein levels in Group A. However, some tumors in Group B, which had no increase in ATF3 protein level, showed increased gene copy number (>2), suggesting gene amplification is not the only controlling factor. Clearly, the levels of protein can be regulated by various mechanisms, and the mechanisms for the elevated ATF3 expression remain to be determined.

We note that some of the Group B breast tumors had ATF3 gene loss rather than gain: No.26 (zero copy), 25, 47 and 48 (one copy). The actin gene control 138

was normal in these tumors. Although we have not ruled out the possibility that the ATF3 gene was rearranged so that the primers we used did not hybridize, this result is consistent with a dichotomous role of ATF3 in cancer cells. It is conceivable that some breast tumors were derived from cells that had deleted

ATF3 (a tumor suppressor), whereas others from cells that had “evolved” to co opt ATF3 (to become an oncogene). This speculation fits with the observations that cancer cells are heterogenous in their genetic and epigenetic alterations

(Albertson, 2006; Fidler, 2003; Loeb, 2001). Although frequently amplified, 1q was documented to be lost in some tumors (Rooney et al., 1999), such as renal carcinoma, rhabdomyosarcoma and gastrointestinal tumors. In all these tumors, both 1q gain and 1q loss have been found, again supporting the notion that different paths can be taken by the cancer cells to succeed.

As described in the Introduction section, the CGH technique has limited resolution. Even though 1q amplicon is frequently amplified in breast cancer and

ATF3 gene is localized within this region, it was not clear whether the ATF3 gene is indeed amplified. Therefore, this result is the first to demonstrate the amplification of ATF3 gene in breast tumors. We note that the ERBB2

(HER2/NEU) gene, a wellknown oncogene for breast cancer, is amplified in

25 –30% of breast tumors (Pauletti et al., 1996; Pollack et al., 1999), a number lower than that for ATF3 (~80%). Although data in this chapter are descriptive in nature, they, in combination with the results from the cultured cells described previously in this thesis, strongly suggest that ATF3 is an oncogene, rather than merely a bystander gene in the amplicon. Importantly, consistent with the 139

causative roles of ATF3 in breast cancer cell lines and mouse model, analysis of two largescale microarray data generated by Sorlie’s and van’t Veer’s groups using primary human breast cancer samples revealed that higher ATF3 expression in primary breast tumors correlates with worse clinical outcomes with statistical significance, suggesting the gene expression levels of ATF3 in tumor samples might be taken as a biomarker or predictor for patient prognosis.

p53 is the most commonly mutated gene in human carcinomas, with a frequency of about 50% (Gasco et al., 2002). In breast cancer, however, the frequency of p53 mutation is only about 20%, much lower than in other solid tumors (Pharoah et al., 1999). Those breastcancerassociated p53 mutants have been extenstively studied, reported to confer more than lossofwildtype p53 function and be predictive of worse patient survival (Mazars et al., 1992;

Sjogren et al., 1996). For the majority (~ 80%) of breast cancers without p53 mutations, multiple mechanisms have also been suggested for inactivating wild type p53, including alterations in its localization, upstream regulators, downstream target genes, and coactivators (Gasco et al., 2002). Interestingly, in the tumor samples I examined, the p53 expression significantly and positively correlates with the ATF3 expression ( p < 0.001) and the majority of the tumor samples positive for both ATF3 and p53 (12/14) have wildtype sequence for p53 in the open reading frame. Both ATF3 and p53 are transcriptional factors, and

Yan et al. reported that ATF3 binds p53 Cterminal region and reduces the ubiquitination and nuclear export of p53 (Yan et al., 2005b). We speculate that

ATF3 may also function as a cofactor for p53, regulating its transcriptional

140

selectivity and/or activity. Another interesting phenomenon is that 8 out of 14 tumor samples that are positive for both proteins have a homozygous polymorphism of Arg instead of Pro at aa72, which was associated with an increased risk of several human cancers, including cervical carcinoma (Storey et al., 1998) and breast cancer (Sjalander et al., 1996). How ATF3 inactivates wild type p53 function and whether this relates to the polymorphism of p53 at codon

72 or potentially other positions remain to be determined.

In Chapter 2, I presented evidence that ATF3 could upregulate Snail, Twist and Slug, all of which are transcriptional repressors for Ecadherin (Arias, 2001).

Consistently, I found ATF3 downregulates Ecadherin protein level in MIV cells

(Figure 3.1.B). In our survey of human breast tumor samples, we noticed a potentially correlation between ATF3 expression and reduced Ecadherin level ( p

=0.26). Given the limited number of samples examined (N=51), it is necessary to collect more samples to further evaluate the correlation between ATF3 and E cadherin. Since the downregulation is a hallmark for epithelialmesenchymal transition (EMT), an important process regulating cell motility and cancer metastasis (Thiery and Sleeman, 2006), it is possible that ATF3 may contribute to breast cancer metastasis, as supported by the xenograft mouse model and

Oncomine gene expression analysis, through the downregulation of Ecadherin and induction of EMT.

141

Breast Endo Cervical MCC Liver Sarcoma Lung % Rank % Rank % Rank % Rank % Rank % Rank % Rank 52.9 1 54.5 1 46.6 2 66.6 2 67.4 2 13.8 1 33.8 3

Table 5.1 1q amplification in several human cancers. Rooney et al. evaluated the results of CGH analyses in 2100 solid tumors of 27 cancer types to identify chromosomal regions of loss or gain. This table summarized the 1q amplification in a subset of cancers. %, percentage of the cancer samples showing 1q amplification by CGH analysis; Rank, the rank of 1q amplification among all chromosomal gains for that type of cancer; Endo, endometrial carcinoma; MCC, Merkel cell carcinoma (Rooney et al., 1999).

5.1 1q amplification in several human cancers

142

ATF3/p53 +/+ (N=14) +/ (N=6) /+ (N=3)* / (N=7)* status aa35 SNP, aa47 Pro to aa72 Pro to aa72 Pro to hetero (N=1); Thr, hetero Arg, hetero Arg, hetero aa123 Thr to (N=1); (N=1); homo (N=2); homo Arg, homo aa72 Pro to (N=2); (N=2); (N=1); Arg, hetero (N=2); homo aa213 SNP, (N=2); Alteration(s) in hetero (N=1); p53 ORF aa138 Ala to Val, hetero (N=1); aa72 Pro to Arg, hetero (N=2); homo (N=8);

Table 5.2 Mutational status of p53 ORF in human breast tumors. ATF3 and p53 status was determined based on Figure 5.3. “+” means the protein expression in the tumor sample is higher than in the matching nondiseased control sample, while ““ means not. aa: amino acid; SNP: single nucleotide polymorphism; hetero: heterozygote; homo: homozygote; * only Exon 2 to 4 of the p53 gene for these samples were sequenced; the whole p53 open reading frame was sequenced for all the other samples. 5.2 Mutational status of p53 ORF in human breast tumors

143

A. Patient ID 1 2 3 4 5 6 7 8 9 10 ATF3 Actin C T C T C T C T C T C T C T C T C T C T Patient ID 11 12 13 14 15 16 17 18 19 20 ATF3 Actin C T C T C T C T C T C T C T C T C T C T

Patient ID 2122 23 Patient ID 39 42 47 36 ATF3 ATF3 Actin Actin C T C T C T C T C T C T C T B. C. TNFα TGFβ MIV HMEC 3 4 7 11 13 23 (hr)0 1 2 4 1 2 4 ATF3 ATF3 Actin Actin

Fig. 5.1 ATF3 expression is upregulated in human breast tumors. A. Human breast tumors (T) and the matching adjacent breast controls (C) were analyzed by immunoblot. Samples 1–23: tumors with elevated ATF3 protein levels compared to their corresponding controls (group A); samples 24–27: examples of the 25 tumors without elevated ATF3 expression (group B). B. Proliferating primary human mammary epithelial cells (HMECs) were analyzed by immunoblot, with MIV cells and a few breast tumors for comparison. C. HMECs were treated with TGFβ (2.5 ng/mL) or TNFα (60 ng/mL) for the indicated times and analyzed by immunoblot. 5.1 ATF3 expression is upregulated in human breast tumors

144

A.30 B. LCM ATF3 Group Group Group A: 25 20 (Gene Copy #) A B Elevated ATF3 15 10 30 (N=23) 5 0 02 2 8 25 11 14 21 23 34 5 10 20 15 510 9 5 10 > 10 7 2 5 Total 23 25 0 ATF3 (Gene Copy #) (Gene ATF3 I IV 1 2 3 4 5 6 7 8 9 1011121314151617181920212223

30 Group B: Non 25 elevated ATF3 20 (N=25) 15 10 5 0 ATF3 (Gene Copy ATF3 #) I IV2425262728 29 30313233343536373839404142434445464748

Fig. 5.2 ATF3 gene is amplified in human breast tumors. A. Genomic DNAs from human breast tumors were analyzed by qPCR to determine the ATF3 gene copy number. The signals from MI cells (I) are defined as 2 (dashed lines) and that from MIV cells (IV) as 3. Inset shows the ATF3 gene copy number of the cancer epithelia from the indicated tumors after isolation by LCM. B. A summary of the ATF3 gene copy numbers in group A and B tumors. 5.2 ATF3 gene is amplified in human breast tumors

145

ATG TAA

Exon A Exon B C Exon E

248bp 5618bp 244bp 4748bp 108bp 695bp 1384p

Primer Primer Set 1 Set 2

Fig. 5.3 Schematic view of ATF3 gene structure and primers used to amplify and sequence its open reading frame in human breast tumors. Exons were represented as slashed rectangles and introns as dashed line. Primers (Set 1 and 2) for highfidelity PCR and sequencing are indicated as arrows.

5.3 Schematic view of ATF3 gene structure and primers used to amplify and sequence its open reading frame in human breast tumors

146

Fiveyear survival Relapsefree survival

A. p < 0.01 B. p < 0.05

N=43 N=43

N=33 N=33

C.Fiveyear survival D. Tenyear survival p < 0.02 p < 0.02

N=52

N=52 N=45 N=45

E. Relapsefree survival p = 0.156

N=72

N=108

Fig. 5.4 KaplanMeier survival curves for patients with low ATF3 expression in breast tumor (blue line) versus those with high ATF3 expression in breast tumor (red dashed line). Xaxis: patient survival time in months; Yaxis: % survival. High and low ATF3 expressions were defined based on the gene expression values from Oncomine database ( www.oncomine.org ) so that each group has similar number of patients. A. and B. Data analysis from Sorlie’s study (Sorlie et al., 2001). C. and D. Data analysis from van’t Veer’s study (van 't Veer et al., 2002). E. Data analysis from Wang’s study (Wang et al., 2005). Patient number (N) in each group is indicated.

5.4 KaplanMeier survival curves for patients with low ATF3 expression in breast tumor (blue line) versus those with high ATF3 expression in breast tumor (red dashed line)

147

A. Patient ID 1 2 3 4 5 6 7 8 9 10 p53 ATF3 Actin

p53 status + + + + +

Patient ID 11 12 13 14 15 16 17 18 19 20 p53 ATF3 Actin p53 status + + + + + + + Patient ID 21 22 23 C T C T C T C T C T C T C T p53 ATF3 Actin p53 status + + C T C T C T B. Patient ID 24 25 26 27 28 29 30 31 32 33 p53 ATF3 Actin p53 status + Patient ID 34 35 36 37 38 39 40 41 42 43 p53 ATF3 Actin p53 status + + Patient ID 44 45 46 47 48 49 50 51 C T C T p53 ATF3 Actin p53 status + C T C T C T C T C T C T C T C T C. p53+ p53 ATF3+ 14 9 ATF3 4 24

p 0.00095

Fig. 5.5 ATF3 expression positively correlates with p53 expression in human breast tumors. A. Human breast tumors (T) and the matching adjacent breast controls (C) were analyzed by immunoblot for indicated proteins. A. Samples 1–23: tumors with elevated ATF3 protein levels compared to their corresponding controls; B. samples 24–51: 28 tumors without elevated ATF3 expression. p53 status was arbitrarily defined as follows: “+” means higher p53 protein level in tumors compared to corresponding controls; “” means no higher p53 in tumors compared to controls. C. Correlation between ATF3 and p53 expression levels in human breast cancer as determined by Fisher’s exact test.

148

5.5 ATF3 expression positively correlates with p53 expression in human breast tumors

ATG TGA

1 2 3 4 5 6 7 8 9 10 11

Primer Primer Primer Set 1 Set 3 Set 4 Primer Set 2

Fig. 5.6 Schematic view of p53 gene structure and primers used to amplify and sequence its open reading frame in human breast tumors. Exons were represented as slashed rectangles and introns as line. Primers (Set 1 to 4) for highfidelity PCR and sequencing are indicated as arrows.

5.6 Schematic view of p53 gene structure and primers used to amplify and sequence its open reading frame in human breast tumors

149

Patient ID 1 4 5 7 8 21 22 ATF3

Ecadherin

Actin Ecadherin status Patient ID 25 31 36 32 33 39 41 ATF3

Ecadherin Actin Ecadherin status + + + + + + + C T C T C T C T C T C T C T B. Ecad+ Ecad

ATF3+ 8 15

ATF3 15 13

p 0.26

Fig. 5.7 Correlation between ATF3 expression and Ecadherin in human breast tumors. A. Examples of immunoblot for indicated proteins in human breast tumors (T) and the matching adjacent breast controls (C). Ecadherin status: “+” means higher Ecaherin protein level in tumors compared to corresponding controls; “” means no higher Ecadherin in tumors compared to controls. B. Correlation between ATF3 and Ecadherin expression levels in human breast cancer as determined by Fisher’s exact test.

5.7 Correlation between ATF3 expression and Ecadherin in human breast tumors

150

CHAPTER 6

FUTURE PERSPECTIVES

6. FUTURE PERSPECTIVES

Work presented in this thesis suggests a role of ATF3 in the development of human breast cancer. Data presented in Chapter 2 reveals a dichotomous function of ATF3 during breast cancer progression. Chapter 3 supports the notion that ATF3 functions as a downstream signaling molecule for stromal factor

TGFβ to regulate cancer cell behavior. Chapter 4 investigates the involvement of

ATF3 in stromacancer interaction. Data from Chapters 3 and 4 offer mechanistic understanding on how ATF3 promotes tumor metastasis in vivo .

Chapter 5 examines the status of ATF3 in human breast cancers, providing strong correlative evidence to support an oncogenic role of ATF3. Although the work covers various aspects of ATF3 in breast cancer biology, much more work remains to further elucidate the mechanisms and functions of ATF3 in human cancer. Here I propose several future perspectives on ATF3 research in breast cancer biology.

First, using the malignant and metastatic breast epithelial cell line MIV, we demonstrated ATF3 can promote cell migration in vitro and tumor metastasis in vivo . While surveying several breast cancer cell lines, we did notice that ectopic expression of ATF3 in some cell lines does not lead to enhanced cell migration, 151

such as in SUM159 cells and MDAMB231 cells (Figure 6.1). Since the endogenous ATF3 levels within these two cell lines are relatively low compared to MIV cells (Figure 2.9), the inability of ectopic ATF3 expression to further enhance the migration is probably not due to the endogenous ATF3 saturating the level of cell migration. My hypothesis is that the ability of ATF3 to regulate cell motility depends on cellular context. To test this hypothesis, we may start by comparing the regulation of motilityrelated genes in MIV cells by ATF3 to that in

MDAMB231 or SUM159 cells by ATF3. This could be achieved by either literaturebased candidate search or systematic search for components in cell motilityregulating network that are differentially regulated by ATF3 in MIV cells vs. in other cells. With potential candidates, we will have a handle to examine how ATF3 regulates these candidates differently between different cell lines.

Second, the “JekyllandHyde” conversion of some genes during cancer development is an intriguing but largely unresolved issue in cancer biology. In addition to identifying a new dichotomous molecule, ATF3, in breast cancer development, this study also suggests a potential mechanism on how ATF3 carries out its dual role, i.e., through the differential regulation of the same gene(s) depending on the cellular context. The nextlevel question would be how cellular context (normal vs. malignant) determines the differential regulation.

Since ATF3 is a transcription factor, to examine the transcriptional machineries brought by ATF3 to the promoter of differentially regulated gene(s) would offer important clues on how this is achieved.

152

Third, cellular context is important for determining ATF3 functions not only in the canceroriginating epithelial cells, but also in different stromal cells. Given the importance of stromacancer interaction in cancer development, it is crucial to investigate how ATF3 is regulated in other stromal cells, such as fibroblasts, endothelial cells, adipocytes and leukocytes, and if ATF3 is induced in these cells, how it affects tumor microenvironment and cancer cell behavior.

Fourth, several studies showed that ATF3 is induced by a variety of compounds with anticancer activities, including adriamycin (Mallory et al., 2005;

Yu et al., 1996), paclitaxel (Peters et al., 2007), nonsteroidal antiinflammatory drugs (Bottone Jr et al., 2003), the phosphatidylinositol 3kinase inhibitor

LY294002 (Yamaguchi et al., 2006), progestin (Syed et al., 2005), and dietary polyphenols such as curcumin from the tumeric plant (Yan et al., 2005a) and catecins in green tea (Baek et al., 2004). Some of these compounds have been used in clinical settings. The oncogenic activity of ATF3 in advanced cancer cells implies that these antitumorigenic compounds may have undesired side effects. Thus, elucidating the molecular context that determines whether ATF3 functions as a tumor suppressor or an oncogene will be helpful for future rational designs of anticancer treatments, thus changing the clinical practice in the future.

153

2.5 * Vec 2 * ATF3 1.5 *

0.5 RelativeMigration 0 MIV SUM159 MDA MDA SKBR3 MB231 MB468

Fig. 6.1 ATF3 differentially regulates the migration of various breast cancer cell lines. ATF3 (black bar) or control vector (Vec, white bar) was stably expressed in different breast cancer cell lines. Cell migration was examined by Boyden chamber migration assay. A595 reading from the matching Vec cells was arbitrarily defined as 1. Quantitative data in all panels are the mean ± S.D. from three independent experiments. * p < 0.05, versus Vec.

6.1 ATF3 differentially regulates the migration of various breast cancer cell lines

154

REFERENCES

(1990). The p53 tumor suppressor and gene product, In Genetic basis for carcinogenesis: Tumor suppressor genes and oncogenes, J. A. G. Knudson, ed. (Tokyo, Japan; London, UK: Japan Sci. Soc. Press; Taylor & Francis Ltd.), pp. 221230.

Akhurst, R. J., and Derynck, R. (2001). TGFbeta signaling in cancera double edged sword. Trends Cell Biol 11 , S4451.

AlHajj, M., Wicha, M. S., BenitoHernandez, A., Morrison, S. J., and Clarke, M. F. (2003). Prospective identification of tumorigenic breast cancer cells. Proc Natl Acad Sci U S A 100 , 39833988.

Albertson, D. G. (2006). Gene amplification in cancer. Trends Genet 22 , 447 455.

Allan, A. L., Albanese, C., Pestell, R. G., and LaMarre, J. (2001). Activating transcription factor 3 induces DNA synthesis and expression of cyclin D1 in hepatocytes. J Biol Chem 276 , 2727227280.

Allavena, P., Sica, A., Solinas, G., Porta, C., and Mantovani, A. (2008). The inflammatory microenvironment in tumor progression: the role of tumor associated macrophages. Crit Rev Oncol Hematol 66 , 19.

AllenJennings, A. E., Hartman, M. G., Kociba, G. J., and Hai, T. (2001). The roles of ATF3 in glucose homeostasis. A transgenic mouse model with liver dysfunction and defects in endocrine pancreas. J Biol Chem 276, 2950729514.

AllenJennings, A. E., Hartman, M. G., Kociba, G. J., and Hai, T. (2002). The roles of ATF3 in liver dysfunction and the regulation of phosphoenolpyruvate carboxykinase gene expression. J Biol Chem 277 , 2002020025.

155

Allred, D. C., and Elledge, R. M. (1999). Caution concerning micrometastatic breast carcinoma in sentinel lymph nodes. Cancer 86 , 905907.

Ameri, K., Hammond, E. M., Culmsee, C., Raida, M., Katschinski, D. M., and Wenger, R. H. (2007). Induction of activating transcription factor 3 by anoxia is independent of p53 and the hypoxic HIF signalling pathway. Oncogene 26 , 284 289.

Anbazhagan, R., Bornman, D. M., Johnston, J. C., Westra, W. H., and Gabrielson, E. (1999). The S387Y mutations of the transforming growth factor beta receptor type I gene is uncommon in metastases of breast cancer and other common types of adenocarcinoma. Cancer Res 59 , 33633364.

Arias, A. M. (2001). Epithelial mesenchymal interactions in cancer and development. Cell 105 , 425431.

Asch, B. B., and BarcellosHoff, M. H. (2001). Epigenetics and breast cancer. J Mammary Gland Biol Neoplasia 6 , 151152.

Baek, S. J., Kim, J. S., Jackson, F. R., Eling, T. E., McEntee, M. F., and Lee, S. H. (2004). Epicatechin gallateinduced expression of NAG1 is associated with growth inhibition and apoptosis in colon cancer cells. Carcinogenesis 25 , 2425 2432.

Balkwill, F. (2002). Tumor necrosis factor or tumor promoting factor? Cytokine Growth Factor Rev 13 , 135141.

Balkwill, F., and Mantovani, A. (2001). Inflammation and cancer: back to Virchow? Lancet 357 , 539545.

Bandyopadhyay, S., Wang, Y., Zhan, R., Pai, S. K., Watabe, M., Iiizumi, M., Furuta, E., Mohinta, S., Liu, W., Hirota, S. , et al. (2006). The tumor metastasis suppressor gene Drg1 downregulates the expression of activating transcription factor 3 in prostate cancer. Cancer Res 66 , 1198311990.

BenBaruch, A. (2003). Host microenvironment in breast cancer development: inflammatory cells, cytokines and chemokines in breast cancer progression: reciprocal tumormicroenvironment interactions. Breast Cancer Res 5 , 3136.

156

Bieche, I., and Lidereau, R. (1995). Genetic alterations in breast cancer. Genes Chromosomes Cancer 14 , 227251.

Bierie, B., and Moses, H. L. (2006). Tumour microenvironment: TGFbeta: the molecular Jekyll and Hyde of cancer. Nat Rev Cancer 6 , 506520.

Bissell, M. J., and Radisky, D. (2001). Putting tumours in context. Nat Rev Cancer 1 , 4654.

Biswas, S. K., Gangi, L., Paul, S., Schioppa, T., Saccani, A., Sironi, M., Bottazzi, B., Doni, A., Vincenzo, B., Pasqualini, F. , et al. (2006). A distinct and unique transcriptional program expressed by tumorassociated macrophages (defective NFkappaB and enhanced IRF3/STAT1 activation). Blood 107 , 21122122.

Bonnet, D., and Dick, J. E. (1997). Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat Med 3 , 730 737.

Bottone, F. G., Jr., Moon, Y., Kim, J. S., AlstonMills, B., Ishibashi, M., and Eling, T. E. (2005). The antiinvasive activity of cyclooxygenase inhibitors is regulated by the transcription factor ATF3 (activating transcription factor 3). Mol Cancer Ther 4 , 693703.

Bottone Jr, F. G., Martinez, J. M., Collins, J. B., Afshari, C. A., and Eling, T. E. (2003). Gene modulation by the cyclooxygenase inhibitor, sulindac sulfide, in human colorectal carcinoma cells: possible link to apoptosis. J Biol Chem 278 , 2579025801.

Campbell, L. L., and Polyak, K. (2007). Breast tumor heterogeneity: cancer stem cells or clonal evolution? Cell Cycle 6 , 23322338.

Chambers, A. F., Naumov, G. N., Vantyghem, S. A., and Tuck, A. B. (2000). Molecular biology of breast cancer metastasis. Clinical implications of experimental studies on metastatic inefficiency. Breast Cancer Res 2 , 400407.

Chen, B. P., Liang, G., Whelan, J., and Hai, T. (1994). ATF3 and ATF3 delta Zip. Transcriptional repression versus activation by alternatively spliced isoforms. J Biol Chem 269 , 1581915826.

157

Chen, H. H., and Wang, D. L. (2004). Nitric oxide inhibits matrix metalloproteinase2 expression via the induction of activating transcription factor 3 in endothelial cells. Mol Pharmacol 65 , 11301140.

Cheng, C., and Sharp, P. A. (2003). RNA polymerase II accumulation in the promoterproximal region of the dihydrofolate reductase and [lcub]gamma[rcub] actin genes. Mol Cell Biol 23 , 19611967.

Cho, R. J., Huang, M., Campbell, M. J., Dong, H., Steinmetz, L., Sapinoso, L., Hampton, G., Elledge, S. J., Davis, R. W., and Lockhart, D. J. (2001). Transcriptional regulation and function during the human cell cycle. Nat Genet 27 , 4854.

Chodosh, L. A. (1999). Breast Cancer: Molecular genetics, pathogenesis, and therapeutics. N Engl J Med 341 , 1244.

Chu, H. M., Tan, Y., Kobierski, L. A., Balsam, L. B., and Comb, M. J. (1994). Activating transcription factor3 stimulates 3',5'cyclic adenosine monophosphatedependent gene expression. Mol Endocrinol 8 , 5968.

Coussens, L. M., and Werb, Z. (2002). Inflammation and cancer. Nature 420 , 860867.

Cox, A., Dunning, A. M., GarciaClosas, M., Balasubramanian, S., Reed, M. W., Pooley, K. A., Scollen, S., Baynes, C., Ponder, B. A., Chanock, S. , et al. (2007). A common coding variant in CASP8 is associated with breast cancer risk. Nat Genet 39 , 352358. de Caestecker, M. P., Piek, E., and Roberts, A. B. (2000). Role of transforming growth factorbeta signaling in cancer. J Natl Cancer Inst 92 , 13881402. de Visser, K. E., Eichten, A., and Coussens, L. M. (2006). Paradoxical roles of the immune system during cancer development. Nat Rev Cancer 6 , 2437.

Debnath, J., Muthuswamy, S. K., and Brugge, J. S. (2003). Morphogenesis and oncogenesis of MCF10A mammary epithelial acini grown in threedimensional basement membrane cultures. Methods 30 , 256268.

158

Derynck, R., Akhurst, R. J., and Balmain, A. (2001). TGFbeta signaling in tumor suppression and cancer progression. Nat Genet 29 , 117129.

Derynck, R., and Zhang, Y. E. (2003). Smaddependent and Smadindependent pathways in TGFbeta family signalling. Nature 425 , 577584.

Dirkx, A. E., Oude Egbrink, M. G., Wagstaff, J., and Griffioen, A. W. (2006). Monocyte/macrophage infiltration in tumors: modulators of angiogenesis. J Leukoc Biol 80 , 11831196.

Easton, D. F., Pooley, K. A., Dunning, A. M., Pharoah, P. D., Thompson, D., Ballinger, D. G., Struewing, J. P., Morrison, J., Field, H., Luben, R. , et al. (2007). Genomewide association study identifies novel breast cancer susceptibility loci. Nature 447 , 10871093.

Eppert, K., Scherer, S. W., Ozcelik, H., Pirone, R., Hoodless, P., Kim, H., Tsui, L. C., Bapat, B., Gallinger, S., Andrulis, I. L. , et al. (1996). MADR2 maps to 18q21 and encodes a TGFbetaregulated MADrelated protein that is functionally mutated in colorectal carcinoma. Cell 86 , 543552.

Evan, G. I., and Vousden, K. H. (2001). Proliferation, cell cycle and apoptosis in cancer. Nature 411 , 342348.

Fan, F., Jin, S., Amundson, S. A., Tong, T., Fan, W., Zhao, H., Zhu, X., Mazzacurati, L., Li, X., Petrik, K. L. , et al. (2002). ATF3 induction following DNA damage is regulated by distinct signaling pathways and overexpression of ATF3 protein suppresses cells growth. Oncogene 21 , 74887496.

Fidler, I. J. (2003). The pathogenesis of cancer metastasis: the 'seed and soil' hypothesis revisited. Nat Rev Cancer 3 , 453458.

Finlay, C. A., Hinds, P. W., and Levine, A. J. (1989). The p53 protooncogene can act as a suppressor of transformation. Cell 57 , 10831093.

Forozan, F., Veldman, R., Ammerman, C. A., Parsa, N. Z., Kallioniemi, A., Kallioniemi, O. P., and Ethier, S. P. (1999). Molecular cytogenetic analysis of 11 new breast cancer cell lines. Br J Cancer 81 , 13281334.

159

Funke, I., and Schraut, W. (1998). Metaanalyses of studies on bone marrow micrometastases: an independent prognostic impact remains to be substantiated. J Clin Oncol 16 , 557566.

Futreal, P. A., Coin, L., Marshall, M., Down, T., Hubbard, T., Wooster, R., Rahman, N., and Stratton, M. R. (2004). A census of human cancer genes. Nat Rev Cancer 4 , 177183.

Gamallo, C., Palacios, J., Suarez, A., Pizarro, A., Navarro, P., Quintanilla, M., and Cano, A. (1993). Correlation of Ecadherin expression with differentiation grade and histological type in breast carcinoma. Am J Pathol 142 , 987993.

Gasco, M., Shami, S., and Crook, T. (2002). The p53 pathway in breast cancer. Breast Cancer Res 4 , 7076.

Genovesi, E. V., Knudsen, R. C., Gerstner, D. J., Card, D. M., Martins, C. L., Quintero, J. C., and Whyard, T. C. (1989). In vitro induction of swine peripheral blood monocyte proliferation by the fibroblastderived murine hematopoietic growth factor CSF1. Vet Immunol Immunopathol 23 , 223244.

Gilchrist, M., Thorsson, V., Li, B., Rust, A. G., Korb, M., Roach, J. C., Kennedy, K., Hai, T., Bolouri, H., and Aderem, A. (2006). Systems biology approaches identify ATF3 as a negative regulator of Tolllike receptor 4. Nature 441 , 173178.

Glasgow, E., and Mishra, L. (2008). Transforming growth factorbeta signaling and ubiquitinators in cancer. Endocr Relat Cancer 15 , 5972.

Gordon, S. (2003). Alternative activation of macrophages. Nat Rev Immunol 3 , 2335.

Gordon, S., and Taylor, P. R. (2005). Monocyte and macrophage heterogeneity. Nat Rev Immunol 5 , 953964.

Goswami, S., Sahai, E., Wyckoff, J. B., Cammer, M., Cox, D., Pixley, F. J., Stanley, E. R., Segall, J. E., and Condeelis, J. S. (2005). Macrophages promote the invasion of breast carcinoma cells via a colonystimulating factor1/epidermal growth factor paracrine loop. Cancer Res 65 , 52785283.

160

Gupta, G. P., Nguyen, D. X., Chiang, A. C., Bos, P. D., Kim, J. Y., Nadal, C., Gomis, R. R., ManovaTodorova, K., and Massague, J. (2007). Mediators of vascular remodelling coopted for sequential steps in lung metastasis. Nature 446 , 765770.

Hai, T. (2006a). The ATF transcription factors in cellular adaptive responses, In Gene Expression and Regulation, M. J, ed. (Beijing, China and New York, USA: Springer New York), pp. 322333.

Hai, T., and Hartman, M. G. (2001). The molecular biology and nomenclature of the activating transcription factor/cAMP responsive element binding family of transcription factors: activating transcription factor proteins and homeostasis. Gene 273, 111.

Hai, T., Lu D, Wolford CC (2006b). Transcription factors: ATF., In Encyclopedia of Respiratory Medicine: Volume 4., G. J. L. a. S. D. Shaprio, ed. (Oxford, UK: Elsevier), pp. 257260.

Hai, T., Wolfgang, C. D., Marsee, D. K., Allen, A. E., and Sivaprasad, U. (1999). ATF3 and stress responses. Gene Expr 7 , 321335.

Hai, T. W., Liu, F., Coukos, W. J., and Green, M. R. (1989). Transcription factor ATF cDNA clones: an extensive family of leucine zipper proteins able to selectively form DNAbinding heterodimers. Genes Dev 3 , 20832090.

Hanahan, D., and Weinberg, R. A. (2000). The hallmarks of cancer. Cell 100 , 57 70.

Hartman, M. G. (2005) THE ROLES OF ATF3 IN STRESSREGULATED SIGNAL TRANSDUCTION AND CELL DEATH IN PANCREATIC BETACELLS The Ohio State University.

Hartman, M. G., Lu, D., Kim, M. L., Kociba, G. J., Shukri, T., Buteau, J., Wang, X., Frankel, W. L., Guttridge, D., Prentki, M. , et al. (2004). Role for activating transcription factor 3 in stressinduced betacell apoptosis. Mol Cell Biol 24 , 57215732.

Hashimoto, Y., Zhang, C., Kawauchi, J., Imoto, I., Adachi, M. T., Inazawa, J., Amagasa, T., Hai, T., and Kitajima, S. (2002). An alternatively spliced isoform of 161

transcriptional repressor ATF3 and its induction by stress stimuli. Nucleic Acids Res 30 , 23982406.

Hennighausen, L., and Robinson, G. W. (2005). Information networks in the mammary gland. Nat Rev Mol Cell Biol 6 , 715725.

Hovey, R. C., McFadden, T. B., and Akers, R. M. (1999). Regulation of mammary gland growth and morphogenesis by the mammary fat pad: a species comparison. J Mammary Gland Biol Neoplasia 4 , 5368.

Howard, B. A., and Gusterson, B. A. (2000). Human breast development. J Mammary Gland Biol Neoplasia 5 , 119137.

Howlett, A. R., and Bissell, M. J. (1993). The influence of tissue microenvironment (stroma and extracellular matrix) on the development and function of mammary epithelium. Epithelial Cell Biol 2 , 7989.

Hsu, J. C., Laz, T., Mohn, K. L., and Taub, R. (1991). Identification of LRF1, a leucinezipper protein that is rapidly and highly induced in regenerating liver. Proc Natl Acad Sci U S A 88 , 35113515.

Hu, Z., Fan, C., Oh, D. S., Marron, J. S., He, X., Qaqish, B. F., Livasy, C., Carey, L. A., Reynolds, E., Dressler, L. , et al. (2006). The molecular portraits of breast tumors are conserved across microarray platforms. BMC Genomics 7 , 96.

Hunter, D. J., Kraft, P., Jacobs, K. B., Cox, D. G., Yeager, M., Hankinson, S. E., Wacholder, S., Wang, Z., Welch, R., Hutchinson, A. , et al. (2007). A genome wide association study identifies alleles in FGFR2 associated with risk of sporadic postmenopausal breast cancer. Nat Genet 39 , 870874.

Ishiguro, T., and Nagawa, H. (2000). Expression of the ATF3 gene on cell lines and surgically excised specimens. Oncol Res 12 , 181183.

Ishiguro, T., and Nagawa, H. (2001). ATF3 gene regulates cell form and migration potential of HT29 colon cancer cells. Oncol Res 12 , 343346.

162

Ishiguro, T., Nagawa, H., Naito, M., and Tsuruo, T. (2000). Inhibitory effect of ATF3 antisense oligonucleotide on ectopic growth of HT29 human colon cancer cells. Jpn J Cancer Res 90 , 833836.

Ishiguro, T., Nakajima, M., Naito, M., Muto, T., and Tsuruo, T. (1996). Identification of genes differentially expressed in B16 murine melanoma sublines with different metastatic potentials. Cancer Res 56 , 875879.

Jemal, A., Siegel, R., Ward, E., Hao, Y., Xu, J., Murray, T., and Thun, M. J. (2008). Cancer statistics, 2008. CA Cancer J Clin 58 , 7196.

Kang, Y., Chen, C. R., and Massague, J. (2003). A selfenabling TGFbeta response coupled to stress signaling: Smad engages stress response factor ATF3 for Id1 repression in epithelial cells. Mol Cell 11 , 915926.

Kang, Y., He, W., Tulley, S., Gupta, G. P., Serganova, I., Chen, C. R., Manova Todorova, K., Blasberg, R., Gerald, W. L., and Massague, J. (2005). Breast cancer bone metastasis mediated by the Smad tumor suppressor pathway. Proc Natl Acad Sci U S A 102 , 1390913914.

Kawauchi, J., Zhang, C., Nobori, K., Hashimoto, Y., Adachi, M. T., Noda, A., Sunamori, M., and Kitajima, S. (2002). Transcriptional repressor activating transcription factor 3 protects human umbilical vein endothelial cells from tumor necrosis factoralphainduced apoptosis through downregulation of p53 transcription. J Biol Chem 277 , 3902539034.

Khuu, C. H., Barrozo, R. M., Hai, T., and Weinstein, S. L. (2007). Activating transcription factor 3 (ATF3) represses the expression of CCL4 in murine macrophages. Mol Immunol 44 , 15981605.

Kimchi, A., Wang, X. F., Weinberg, R. A., Cheifetz, S., and Massague, J. (1988). Absence of TGFbeta receptors and growth inhibitory responses in retinoblastoma cells. Science 240 , 196199.

Kimura, K., Wakamatsu, A., Suzuki, Y., Ota, T., Nishikawa, T., Yamashita, R., Yamamoto, J., Sekine, M., Tsuritani, K., Wakaguri, H. , et al. (2006). Diversification of transcriptional modulation: largescale identification and characterization of putative alternative promoters of human genes. Genome Res 16 , 5565.

163

Kobayashi, S., Ito, E., Honma, R., Nojima, Y., Shibuya, M., Watanabe, S., and Maru, Y. (2004). Dynamic regulation of gene expression by the Flt1 kinase and Matrigel in endothelial tubulogenesis. Genomics 84 , 185192.

Komarnitsky, P., Cho, E. J., and Buratowski, S. (2000). Different phosphorylated forms of RNA polymerase II and associated mRNA processing factors during transcription. Genes Dev 14 , 24522460.

Kucich, U., Rosenbloom, J. C., Shen, G., Abrams, W. R., Hamilton, A. D., Sebti, S. M., and Rosenbloom, J. (2000). TGFbeta1 stimulation of fibronectin transcription in cultured human lung fibroblasts requires active geranylgeranyl transferase I, phosphatidylcholinespecific phospholipase C, protein kinase C delta, and p38, but not erk1/erk2. Arch Biochem Biophys 374 , 313324.

Lane, D. P., and Benchimol, S. (1990). p53: oncogene or antioncogene? Genes Dev 4 , 18.

Lauss, M., Kriegner, A., Vierlinger, K., Visne, I., Yildiz, A., Dilaveroglu, E., and Noehammer, C. (2008). Consensus genes of the literature to predict breast cancer recurrence. Breast Cancer Res Treat 110 , 235244.

Lee, K. A., Hai, T. Y., SivaRaman, L., Thimmappaya, B., Hurst, H. C., Jones, N. C., and Green, M. R. (1987). A cellular protein, activating transcription factor, activates transcription of multiple E1Ainducible adenovirus early promoters. Proc Natl Acad Sci U S A 84 , 83558359.

Lee, W. Y., Huang, S. C., Tzeng, C. C., Chang, T. L., and Hsu, K. F. (2007). Alterations of metastasisrelated genes identified using an oligonucleotide microarray of genisteintreated HCC1395 breast cancer cells. Nutr Cancer 58 , 239246.

Leek, R. D., and Harris, A. L. (2002). Tumorassociated macrophages in breast cancer. J Mammary Gland Biol Neoplasia 7 , 177189.

Leek, R. D., Lewis, C. E., Whitehouse, R., Greenall, M., Clarke, J., and Harris, A. L. (1996). Association of macrophage infiltration with angiogenesis and prognosis in invasive breast carcinoma. Cancer Res 56 , 46254629.

164

Levy, L., and Hill, C. S. (2006). Alterations in components of the TGFbeta superfamily signaling pathways in human cancer. Cytokine Growth Factor Rev 17 , 4158.

Lewis, C. E., and Pollard, J. W. (2006). Distinct role of macrophages in different tumor microenvironments. Cancer Res 66 , 605612.

Liang, G., Wolfgang, C. D., Chen, B. P., Chen, T. H., and Hai, T. (1996). ATF3 gene. Genomic organization, promoter, and regulation. J Biol Chem 271 , 1695 1701.

Lin, E. Y., Li, J. F., Gnatovskiy, L., Deng, Y., Zhu, L., Grzesik, D. A., Qian, H., Xue, X. N., and Pollard, J. W. (2006). Macrophages regulate the angiogenic switch in a mouse model of breast cancer. Cancer Res 66 , 1123811246.

Lin, E. Y., Nguyen, A. V., Russell, R. G., and Pollard, J. W. (2001). Colony stimulating factor 1 promotes progression of mammary tumors to malignancy. J Exp Med 193 , 727740.

Lin, E. Y., and Pollard, J. W. (2007). Tumorassociated macrophages press the angiogenic switch in breast cancer. Cancer Res 67 , 50645066.

Lin, Y. S., and Green, M. R. (1988). Interaction of a common cellular transcription factor, ATF, with regulatory elements in both E1a and cyclic AMPinducible promoters. Proc Natl Acad Sci U S A 85 , 33963400.

Loeb, L. A. (2001). A mutator phenotype in cancer. Cancer Res 61 , 32303239.

Lu, D., Chen, J., and Hai, T. (2006a). The regulation of ATF3 gene expression by mitogenactivated protein kinases. Biochem J 401 , 559567.

Lu, D., Wolfgang, C. D., and Hai, T. (2006b). Activating transcription factor 3, a stressinducible gene, suppresses Rasstimulated tumorigenesis. J Biol Chem 281 , 1047310481.

Mallory, J. C., Crudden, G., Oliva, A., Saunders, C., Stromberg, A., and Craven, R. J. (2005). A novel group of genes regulate susceptibility to antineoplastic drugs in highly tumorigenic breast cancer cells. Mol Pharmacol 6 , 17471756. 165

Mantovani, A., Bottazzi, B., Colotta, F., Sozzani, S., and Ruco, L. (1992). The origin and function of tumorassociated macrophages. Immunol Today 13 , 265 270.

Mantovani, A., Schioppa, T., Porta, C., Allavena, P., and Sica, A. (2006). Role of tumorassociated macrophages in tumor progression and invasion. Cancer Metastasis Rev 25 , 315322.

Mantovani, A., Sica, A., and Locati, M. (2005). Macrophage polarization comes of age. Immunity 23 , 344346.

Mantovani, A., Sica, A., Sozzani, S., Allavena, P., Vecchi, A., and Locati, M. (2004). The chemokine system in diverse forms of macrophage activation and polarization. Trends Immunol 25 , 677686.

Mantovani, A., Sozzani, S., Locati, M., Allavena, P., and Sica, A. (2002). Macrophage polarization: tumorassociated macrophages as a paradigm for polarized M2 mononuclear phagocytes. Trends Immunol 23 , 549555.

Mashima, T., Udagawa, S., and Tsuruo, T. (2001). Involvement of transcriptional repressor ATF3 in acceleration of caspase protease activation during DNA damaging agentinduced apoptosis. J Cell Physiol 188 , 352358.

Massague, J. (1998). TGFbeta signal transduction. Annu Rev Biochem 67 , 753 791.

Massague, J. (2000). How cells read TGFbeta signals. Nat Rev Mol Cell Biol 1 , 169178.

Masuda, J., Usui, R., and Maru, Y. (2008). Fibronectin Type I Repeat Is a Nonactivating Ligand for EphA1 and Inhibits ATF3dependent Angiogenesis. J Biol Chem 283 , 1314813155.

Mazars, R., Spinardi, L., BenCheikh, M., SimonyLafontaine, J., Jeanteur, P., and Theillet, C. (1992). p53 mutations occur in aggressive breast cancer. Cancer Res 52 , 39183923.

166

Mielnicki, L. M., Asch, H. L., and Asch, B. B. (2001). Genes, chromatin, and breast cancer: an epigenetic tale. J Mammary Gland Biol Neoplasia 6 , 169182.

Minn, A. J., Gupta, G. P., Siegel, P. M., Bos, P. D., Shu, W., Giri, D. D., Viale, A., Olshen, A. B., Gerald, W. L., and Massague, J. (2005). Genes that mediate breast cancer metastasis to lung. Nature 436 , 518524.

Montminy, M. R., and Bilezikjian, L. M. (1987). Binding of a nuclear protein to the cyclicAMP response element of the somatostatin gene. Nature 328 , 175178.

Morrow, M., and Harris, J. (2004). Ductal carcinoma in situ and microinvasive carcinoma, In Diseases of the Breast, J. H. e. al., ed. (Philadelphia: Lippincott Williams and Wilkins), pp. 521537.

Moustakas, A., and Heldin, C. H. (2007). Signaling networks guiding epithelial mesenchymal transitions during embryogenesis and cancer progression. Cancer Sci 98 , 15121520.

Moustakas, A., Souchelnytskyi, S., and Heldin, C. H. (2001). Smad regulation in TGFbeta signal transduction. J Cell Sci 114 , 43594369.

Nakagomi, S., Suzuki, Y., Namikawa, K., KiryuSeo, S., and Kiyama, H. (2003). Expression of the activating transcription factor 3 prevents cJun Nterminal kinaseinduced neuronal death by promoting heat shock protein 27 expression and Akt activation. J Neurosci 23 , 51875196.

Nawa, T., Nawa, M. T., Adachi, M. T., Uchimura, I., Shimokawa, R., Fujisawa, K., Tanaka, A., Numano, F., and Kitajima, S. (2002). Expression of transcriptional repressor ATF3/LRF1 in human atherosclerosis: colocalization and possible involvement in cell death of vascular endothelial cells. Atherosclerosis 161 , 281 291.

Neville, M. C., Medina, D., Monks, J., and Hovey, R. C. (1998). The mammary fat pad. J Mammary Gland Biol Neoplasia 3 , 109116.

Newman, J. R., and Keating, A. E. (2003). Comprehensive identification of human bZIP interactions with coiledcoil arrays. Science 300 , 20972101.

167

Nobori, K., Ito, H., TamamoriAdachi, M., Adachi, S., Ono, Y., Kawauchi, J., Kitajima, S., Marumo, F., and Isobe, M. (2002). ATF3 inhibits doxorubicin induced apoptosis in cardiac myocytes: a novel cardioprotective role of ATF3. J Mol Cell Cardiol 34 , 13871397.

O'Brien, C. A., Pollett, A., Gallinger, S., and Dick, J. E. (2007). A human colon cancer cell capable of initiating tumour growth in immunodeficient mice. Nature 445 , 106110.

Oka, H., Shiozaki, H., Kobayashi, K., Inoue, M., Tahara, H., Kobayashi, T., Takatsuka, Y., Matsuyoshi, N., Hirano, S., Takeichi, M., and et al. (1993). Expression of Ecadherin cell adhesion molecules in human breast cancer tissues and its relationship to metastasis. Cancer Res 53 , 16961701.

Okamoto, A., Iwamoto, Y., and Maru, Y. (2006). Oxidative stressresponsive transcription factor ATF3 potentially mediates diabetic angiopathy. Mol Cell Biol 26 , 10871097.

Okamoto, Y., Chaves, A., Chen, J., Kelley, R., Jones, K., Weed, H. G., Gardner, K. L., Gangi, L., Yamaguchi, M., Klomkleaw, W. , et al. (2001). Transgenic mice with cardiacspecific expression of activating transcription factor 3, a stress inducible gene, have conduction abnormalities and contractile dysfunction. Am J Pathol 159 , 639650.

Osborne, M. (2000). Breast anatomy and development, In Diseases of the breast, H. JR, ed. (Lippincott Williams & Wilkins), pp. 113.

Pan, Y., Chen, H., Siu, F., and Kilberg, M. S. (2003). Amino acid deprivation and endoplasmic reticulum stress induce expression of multiple activating transcription factor3 mRNA species that, when overexpressed in HepG2 cells, modulate transcription by the human asparagine synthetase promoter. J Biol Chem 278 , 3840238412.

Parmar, H., and Cunha, G. R. (2004). Epithelialstromal interactions in the mouse and human mammary gland in vivo. Endocr Relat Cancer 11 , 437458.

Pearson, A. G., Gray, C. W., Pearson, J. F., Greenwood, J. M., During, M. J., and Dragunow, M. (2003). ATF3 enhances cJunmediated neurite sprouting. Brain Res Mol Brain Res 120 , 3845.

168

Perez, S., Vial, E., van Dam, H., and Castellazzi, M. (2001). Transcription factor ATF3 partially transforms chick embryo fibroblasts by promoting growth factor independent proliferation. Oncogene 20 , 11351141.

Perou, C. M., Sorlie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S., Rees, C. A., Pollack, J. R., Ross, D. T., Johnsen, H., Akslen, L. A. , et al. (2000). Molecular portraits of human breast tumours. Nature 406 , 747752.

Peters, C. M., JimenezAndrade, J. M., Jonas, B. M., Sevcik, M. A., Koewler, N. J., Ghilardi, J. R., Wong, G. Y., and Mantyh, P. W. (2007). Intravenous paclitaxel administration in the rat induces a peripheral sensory neuropathy characterized by macrophage infiltration and injury to sensory neurons and their supporting cells. Exp Neurol 203 , 4254.

Pharoah, P. D., Day, N. E., and Caldas, C. (1999). Somatic mutations in the p53 gene and prognosis in breast cancer: a metaanalysis. Br J Cancer 80 , 1968 1973.

Pollard, J. W. (2004). Tumoureducated macrophages promote tumour progression and metastasis. Nat Rev Cancer 4 , 7178.

Polyak, K. (2007). Breast cancer: origins and evolution. J Clin Invest 117 , 3155 3163.

Reya, T., Morrison, S. J., Clarke, M. F., and Weissman, I. L. (2001). Stem cells, cancer, and cancer stem cells. Nature 414 , 105111.

Roberts, A. B., and Wakefield, L. M. (2003). The two faces of transforming growth factor beta in carcinogenesis. Proc Natl Acad Sci U S A 100 , 86218623.

Rooney, P. H., Murray, G. I., Stevenson, D. A., Haites, N. E., Cassidy, J., and McLeod, H. L. (1999). Comparative genomic hybridization and chromosomal instability in solid tumours. Br J Cancer 80 , 862873.

Russo, J., and Russo, I. H. (2004). Development of the human breast. Maturitas 49 , 215.

169

Santner, S. J., Dawson, P. J., Tait, L., Soule, H. D., Eliason, J., and Mohamed, A. N. (2001). Malignant MCF10CA1 cell lines derived from premalignant human breast epithelial MCF10AT cells. Breast Cancer Res Treat 65 , 101110.

Sellers, T. A. (1997). Genetic factors in the pathogenesis of breast cancer: their role and relative importance. J Nutr 127 , 929S932S.

Seoane, J. (2006). Escaping from the TGFbeta antiproliferative control. Carcinogenesis 27 , 21482156.

Shi, Y., Wang, Y. F., Jayaraman, L., Yang, H., Massague, J., and Pavletich, N. P. (1998). Crystal structure of a Smad MH1 domain bound to DNA: insights on DNA binding in TGFbeta signaling. Cell 94 , 585594.

Shiozaki, H., Tahara, H., Oka, H., Miyata, M., Kobayashi, K., Tamura, S., Iihara, K., Doki, Y., Hirano, S., Takeichi, M., and et al. (1991). Expression of immunoreactive Ecadherin adhesion molecules in human cancers. Am J Pathol 139 , 1723.

Sica, A., Allavena, P., and Mantovani, A. (2008). Cancer related inflammation: The macrophage connection. Cancer Lett.

Siegel, P. M., and Massague, J. (2003). Cytostatic and apoptotic actions of TGF beta in homeostasis and cancer. Nat Rev Cancer 3 , 807821.

Siegel, P. M., Shu, W., Cardiff, R. D., Muller, W. J., and Massague, J. (2003). Transforming growth factor beta signaling impairs Neuinduced mammary tumorigenesis while promoting pulmonary metastasis. Proc Natl Acad Sci U S A 100 , 84308435.

Sims, A. H., Howell, A., Howell, S. J., and Clarke, R. B. (2007). Origins of breast cancer subtypes and therapeutic implications. Nat Clin Pract Oncol 4 , 516525.

Singh, S. K., Clarke, I. D., Terasaki, M., Bonn, V. E., Hawkins, C., Squire, J., and Dirks, P. B. (2003). Identification of a cancer stem cell in human brain tumors. Cancer Res 63 , 58215828.

170

Sjalander, A., Birgander, R., Hallmans, G., Cajander, S., Lenner, P., Athlin, L., Beckman, G., and Beckman, L. (1996). p53 polymorphisms and haplotypes in breast cancer. Carcinogenesis 17 , 13131316.

Sjogren, S., Inganas, M., Norberg, T., Lindgren, A., Nordgren, H., Holmberg, L., and Bergh, J. (1996). The p53 gene in breast cancer: prognostic value of complementary DNA sequencing versus immunohistochemistry. J Natl Cancer Inst 88 , 173182.

Sorlie, T., Perou, C. M., Tibshirani, R., Aas, T., Geisler, S., Johnsen, H., Hastie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S. , et al. (2001). Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 98 , 1086910874.

Sorlie, T., Wang, Y., Xiao, C., Johnsen, H., Naume, B., Samaha, R. R., and BorresenDale, A. L. (2006). Distinct molecular mechanisms underlying clinically relevant subtypes of breast cancer: gene expression analyses across three different platforms. BMC Genomics 7 , 127.

Soule, H. D., Maloney, T. M., Wolman, S. R., Peterson Jr, W. D., Brenz, R., and McGrath, C. M. (1990). Isolation and characterization of a spontaneously immortalized human breast epithelial cell line, MCF10. Cancer Res 50 , 6075 6086.

Stearns, M. E., Kim, G., Garcia, F., and Wang, M. (2004). Interleukin10 induced activating transcription factor 3 transcriptional suppression of matrix metalloproteinase2 gene expression in human prostate CPTX1532 Cells. Mol Cancer Res 2 , 403416.

Stingl, J., Raouf, A., Emerman, J. T., and Eaves, C. J. (2005). Epithelial progenitors in the normal human mammary gland. J Mammary Gland Biol Neoplasia 10 , 4959.

Storey, A., Thomas, M., Kalita, A., Harwood, C., Gardiol, D., Mantovani, F., Breuer, J., Leigh, I. M., Matlashewski, G., and Banks, L. (1998). Role of a p53 polymorphism in the development of human papillomavirusassociated cancer. Nature 393 , 229234.

Syed, V., Mukherjee, K., LyonsWeiler, J., Lau, K. M., Mashima, T., and Tsuruo, T. (2005). Identification of ATF3, caveolin1, DLC1, and NM23H2 as putative 171

antitumorigenic, progesteroneregulated genes for ovarian cancer cells by gene profiling. Oncogene 24 , 17741787.

Takagi, Y., Kohmura, H., Futamura, M., Kida, H., Tanemura, H., Shimokawa, K., and Saji, S. (1996). Somatic alterations of the DPC4 gene in human colorectal cancers in vivo. Gastroenterology 111 , 13691372.

Takeichi, M. (1991). Cadherin cell adhesion receptors as a morphogenetic regulator. Science 251 , 14511455.

Tamura, K., Hua, B., Adachi, S., Guney, I., Kawauchi, J., Morioka, M., Tamamori Adachi, M., Tanaka, Y., Nakabeppu, Y., Sunamori, M. , et al. (2005). Stress response gene ATF3 is a target of cmyc in seruminduced cell proliferation. Embo J 24 , 25902601.

Tang, B., Vu, M., Booker, T., Santner, S. J., Miller, F. R., and Anver, M. R. (2003). TGFbeta switches from tumor suppressor to prometastatic factor in a model of breast cancer progression. J Clin Invest 112 , 11161124.

Thiery, J. P. (2002). Epithelialmesenchymal transitions in tumour progression. Nat Rev Cancer 2 , 442454.

Thiery, J. P., and Sleeman, J. P. (2006). Complex networks orchestrate epithelialmesenchymal transitions. Nat Rev Mol Cell Biol 7 , 131142.

Tian, F., DaCosta Byfield, S., Parks, W. T., Yoo, S., Felici, A., Tang, B., Piek, E., Wakefield, L. M., and Roberts, A. B. (2003). Reduction in Smad2/3 signaling enhances tumorigenesis but suppresses metastasis of breast cancer cell lines. Cancer Res 63 , 82848292.

Tirkkonen, M., Tanner, M., Karhu, R., Kallioniemi, A., Isola, J., and Kallioniemi, O. P. (1998). Molecular cytogenetics of primary breast cancer by CGH. Genes Chromosomes Cancer 21 , 177184.

Tomita, S., Deguchi, S., Miyaguni, T., Muto, Y., Tamamoto, T., and Toda, T. (1999). Analyses of microsatellite instability and the transforming growth factor beta receptor type II gene mutation in sporadic breast cancer and their correlation with clinicopathological features. Breast Cancer Res Treat 53 , 3339.

172

van 't Veer, L. J., Dai, H., van de Vijver, M. J., He, Y. D., Hart, A. A., Mao, M., Peterse, H. L., van der Kooy, K., Marton, M. J., Witteveen, A. T. , et al. (2002). Gene expression profiling predicts clinical outcome of breast cancer. Nature 415 , 530536. van der Meijden, C. M., Lapointe, D. S., Luong, M. X., PericHupkes, D., Cho, B., Stein, J. L., van Wijnen, A. J., and Stein, G. S. (2002). Gene profiling of cell cycle progression through Sphase reveals sequential expression of genes required for DNA replication and nucleosome assembly. Cancer Res 62 , 32333243.

Veltmaat, J. M., Mailleux, A. A., Thiery, J. P., and Bellusci, S. (2003). Mouse embryonic mammogenesis as a model for the molecular regulation of pattern formation. Differentiation 71 , 117.

Venkitaraman, A. R. (2004). Preface: chromosomal instability and breast cancer pathogenesis. J Mammary Gland Biol Neoplasia 9 , 217220.

Vogel, F. (1979). Genetics of retinoblastoma. Hum Genet 52 , 154.

Wakefield, L. M., and Roberts, A. B. (2002). TGFbeta signaling: positive and negative effects on tumorigenesis. Curr Opin Genet Dev 12 , 2229.

Wang, J., Cao, Y., and Steiner, D. F. (2003). Regulation of proglucagon transcription by activated transcription factor (ATF) 3 and a novel isoform, ATF3b, through the cAMPresponse element/ATF site of the proglucagon gene promoter. J Biol Chem 278 , 3289932904.

Wang, Y., Klijn, J. G., Zhang, Y., Sieuwerts, A. M., Look, M. P., Yang, F., Talantov, D., Timmermans, M., Meijervan Gelder, M. E., Yu, J. , et al. (2005). Geneexpression profiles to predict distant metastasis of lymphnodenegative primary breast cancer. Lancet 365 , 671679.

Weiss, L. (1990). Metastatic inefficiency. Adv Cancer Res 54 , 159211.

Whitmore, M. M., Iparraguirre, A., Kubelka, L., Weninger, W., Hai, T., and Williams, B. R. (2007). Negative regulation of TLRsignaling pathways by activating transcription factor3. J Immunol 179 , 36223630.

173

Wolfgang, C. D., Chen, B. P., Martindale, J. L., Holbrook, N. J., and Hai, T. (1997). gadd153/Chop10, a potential target gene of the transcriptional repressor ATF3. Mol Cell Biol 17 , 67006707.

Wolfgang, C. D., Liang, G., Okamoto, Y., Allen, A. E., and Hai, T. (2000). Transcriptional autorepression of the stressinducible gene ATF3. J Biol Chem 275 , 1686516870.

Wyckoff, J. B., Wang, Y., Lin, E. Y., Li, J. F., Goswami, S., Stanley, E. R., Segall, J. E., Pollard, J. W., and Condeelis, J. (2007). Direct visualization of macrophageassisted tumor cell intravasation in mammary tumors. Cancer Res 67 , 26492656.

Yamaguchi, K., Lee, S. H., Kim, J. S., Wimalasena, J., Kitajima, S., and Baek, S. J. (2006). Activating transcription factor 3 and early growth response 1 are the novel targets of LY294002 in a phosphatidylinositol 3kinaseindependent pathway. Cancer Res 66 , 23762384.

Yan, C., Jamaluddin, M. S., Aggarwal, B., Myers, J., and Boyd, D. D. (2005a). Gene expression profiling identifies activating transcription factor 3 as a novel contributor to the proapoptotic effect of curcumin. Mol Cancer Ther 4 , 233241.

Yan, C., Lu, D., Hai, T., and Boyd, D. D. (2005b). Activating transcription factor 3, a stress sensor, activates p53 by blocking its ubiquitination. Embo J 24 , 2425 2435.

Yan, C., Wang, H., and Boyd, D. D. (2002). ATF3 represses 72kDa type IV collagenase (MMP2) expression by antagonizing p53dependent transactivation of the collagenase promoter. J Biol Chem 277 , 1080410812.

Yu, R., Shtil, A. A., Tan, T. H., Roninson, I. B., and Kong, A. N. (1996). Adriamycin activates cjun Nterminal kinase in human leukemia cells: a relevance to apoptosis. Cancer Lett 107 , 7381.

Zawel, L., Dai, J. L., Buckhaults, P., Zhou, S., Kinzler, K. W., Vogelstein, B., and Kern, S. E. (1998). Human Smad3 and Smad4 are sequencespecific transcription activators. Mol Cell 1 , 611617.

174

Zhang, C., Gao, C., Kawauchi, J., Hashimoto, Y., Tsuchida, N., and Kitajima, S. (2002). Transcriptional activation of the human stressinducible transcriptional repressor ATF3 gene promoter by p53. Biochem Biophys Res Commun 297 , 13021310.

175