A Novel Role for the Notch Effector Rbpj in Tumorigenesis

A NOVEL ROLE FOR THE NOTCH EFFECTOR RBPJ IN TUMORIGENESIS

Iva Kulic

B.Sc., Simon Fraser University, 2004

A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF

THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

The Faculty of Graduate Studies

(Experimental Medicine)

THE UNIVERSITY OF BRITISH COLUMBIA

(Vancouver)

November 2010

The Notch signaling pathway, which converges on RBPJ, is deregulated in a number of malignancies. Following pathway activation, RBPJ, the DNA-binding component of the pathway, associates with Notch to activate transcription of target genes. In the absence of Notch activity, RBPJ acts as a transcriptional repressor by recruiting a co-repressor complex that must be displaced to reinitiate the cycle of activation. As RBPJ is a key regulator of Notch signaling and is constitutively expressed in normal cells, we set out to evaluate the effect of RBPJ loss in the context of human cancer. Frequent RBPJ loss was detected in human breast and lung tumors. Moreover, depletion of RBPJ in a human breast cancer cell line accelerated xenograft tumor growth, whereas over-expression of a mutated version of RBPJ (which allows retained function as a transcriptional repressor but prevents activation via Notch) reduced tumor growth in a mouse model. These findings were confirmed in a lymphoma knock-out cell line, where a complete loss of RBPJ strikingly increased tumor growth in mice. RBPJ-deficient tumor xenografts showed up-regulated expression of HEY family genes, which represent direct canonical RBPJ targets. Blockade of Notch activation had no effect on the magnitude of HEY gene derepression in the absence of RBPJ, indicating that Notch does not participate in deregulated signal activation resulting from loss of transcriptional repression. To identify other aberrantly induced genes that contribute to the oncogenic phenotype with RBPJ loss, we performed a global analysis. RBPJ removal led to enrichment of acetylated histone H4 at induced gene promoters. We therefore used this epigenetic mark as an indirect measure of promoter activity to identify processes that were differentially active in the RBPJ-depleted breast cancer cells compared to RBPJ-containing controls. RBPJ loss enriched for a Notch-like signal and increased acetyl marks at genes associated with cell survival. Indeed, resistance to cell death was observed in RBPJ-deficient breast cancer and lymphoma tumors both in vitro and in vivo. This work defines a new role for RBPJ as a tumor suppressor, the loss of which represents an alternate mechanism for deregulating Notch signaling in cancer.

PREFACE

None of the thesis work described here is published. However, a number of people collaborated in the generation of this data and their contributions are listed below. The experiments were conceptualized by Iva Kulic and Aly Karsan. This project was led by Iva Kulic who prepared all written work and figures, and performed experiments with the exception of those listed below:

William Lockwood (Wan Lam Lab) provided all array comparative genomic hybridization data and microarray expression data for the parental breast cancer cell lines, lung cancer cell lines and primary lung cancers.

Blake Gilks provided sections of tissue microarrays.

Bettina Kempkes provided DG75 RBPJ knock-out and control cells.

A team lead by Martin Hirst (Marco Marra Lab) performed all chromatin immunoprecipitation experiments and sequenced the acetylated histone H4 library.

Gordon Robertson (Steve Jones Lab) performed bioinformatics analysis of the acetylated histone H4 data.

Jennifer Baker (Andrew Minchinton Lab) performed immunofluorescent staining on whole tumor cryosections and analyzed the tumor mapping data.

Vennie Chou at PMI labs performed all immunohistochemistry on paraffin sections.

Megan Fuller and Winnie Mok assisted Iva Kulic with animal experiments and in vitro experiments respectively.

The use of mouse models and biohazardous materials was approved by the UBC Animal Care and Ethics Committee and the Biosafety Committee (Animal Care Certificate A07-0074 and Biohazard Approval Certificate B06-0186).

iii

TABLE OF CONTENTS

ABSTRACT ...... ii PREFACE ...... iii TABLE OF CONTENTS ...... iv LIST OF TABLES ...... vii LIST OF FIGURES ...... viii LIST OF ABBREVIATIONS ...... x ACKNOWLEDGEMENTS ...... xiii

CHAPTER 1: INTRODUCTION ...... 1 1.1 CANCER ...... 2 1.2 BREAST CANCER ...... 3 1.3 THE NOTCH SIGNALING PATHWAY ...... 6 1.4 NOTCH SIGNALING IN CANCER ...... 9 1.5 NOTCH SIGNALING IN BREAST CANCER ...... 10 1.6 A CLOSER LOOK AT RBPJ ...... 12 1.7 RBPJ IN MAMMARY DEVELOPMENT AND CANCER ...... 16 1.8 EPIGENETIC REGULATION ASSOCIATED WITH HISTONE MODIFICATIONS . 18 1.9 RBPJ NUCLEAR CYCLE ...... 19 1.10 NOTCH/RBPJ TARGET GENES AND THEIR FUNCTION ...... 25 1.11 NOTCH SIGNALING CROSS-TALK WITH OTHER PATHWAYS ...... 29 1.12 AIM OF PRESENTED STUDY ...... 30

CHAPTER 2: MATERIALS AND METHODS ...... 34 2.1 CELL CULTURE ...... 35 2.2 RNA INTERFERENCE AND GENE TRANSFER ...... 35 2.3 RNA ISOLATION AND qPCR ...... 36 2.4 IMMUNOBLOT ANALYSIS ...... 37 2.5 CO-CULTURE EXPERIMENTS ...... 37 2.6 CELL DEATH ASSAY ...... 39 2.7 CELL GROWTH ASSAY ...... 39 2.8 MICROARRAY DATA ...... 40 2.9 ARRAY COMPARATIVE GENOMIC HYBRIDIZATION (aCGH) ...... 40 iv

2.10 XENOGRAFT TUMOR GROWTH ...... 40 2.11 IMMUNOHISTOCHEMISTRY ON PARAFFIN BLOCKS ...... 41 2.12 WHOLE SECTION TUMOR STAINING ...... 41 2.13 ELECTROPHORETIC MOBILITY SHIFT ASSAYS (EMSAs) ...... 43 2.14 CHROMATIN IMMUNOPRECIPITATION, LIBRARY CONSTRUCTION AND SEQUENCING ANALYSIS ...... 43 2.15 INGENUITY PATHWAY ANALYSIS ...... 45 2.16 STATISTICAL ANALYSIS ...... 46

CHAPTER 3: RBPJ LOSS INCREASES TUMORIGENICITY ...... 47 3.1 INTRODUCTION ...... 48 3.2 RESULTS ...... 48 3.2.1 RBPJ expression in normal tissues ...... 48 3.2.2 Frequent RBPJ deficiency in human cancer cell lines ...... 49 3.2.3 Evidence for RBPJ loss in primary human breast and lung cancers ...... 53 3.2.4 RBPJ depletion increases xenograft tumor growth ...... 56 3.2.5 Evidence for RBPJ loss during breast cancer progression ...... 58 3.3 DISCUSSION ...... 59

CHAPTER 4: RBPJ DEFICIENCY CAUSES INDUCTION OF NOTCH TARGET GENES ...... 62 4.1 INTRODUCTION ...... 63 4.2 RESULTS ...... 63 4.2.1 HEY induction occurs with RBPJ removal and is Notch independent ...... 63 4.2.2 A negative correlation exists between HEY2 and RBPJ expression in human breast cancer ...... 66 4.2.3 The outcome of modified HEY transcript levels on xenograft tumor growth...... 68 4.2.4 Evidence of gene derepression in cells derived from Rbpj KO mouse ...... 70 4.3 DISCUSSION ...... 71

CHAPTER 5: RBPJ LOSS PROMOTES TUMOR CELL SURVIVAL ...... 73 5.1 INTRODUCTION ...... 74

5.2 RESULTS ...... 75 5.2.1 Depletion of RBPJ results in increased epigenetic marks of promoter activity at the RBPJ binding motif in the HEY1 and HEY2 promoter ...... 75 5.2.2 A Notch-like signal is generated in RBPJ deficient cells ...... 78 5.2.3 RBPJ depletion confers a survival advantage ...... 80 5.3 DISCUSSION ...... 87

CHAPTER 6: PERSPECTIVES, FUTURE DIRECTIONS AND IMPLICATIONS ...... 90 6.1 POTENTIAL MECHANISMS OF LOSS OF RBPJ TRANSCRIPTIONAL REPRESSION ...... 91 6.2 RBPJ LOSS IN THE CONTEXT OF NORMAL MAMMARY DEVELOPMENT AND BREAST CANCER ...... 93 6.3 ROLE OF HEY GENES IN ONCOGENESIS MEDIATED BY RBPJ DEFICIENCY 94 6.4 INDUCTION OF OTHER GENES FOLLOWING RBPJ REMOVAL ...... 96 6.5 RBPJ LOSS PROMOTES CELL SURVIVAL ...... 96 6.6 OTHER POTENTIAL CONSEQUENCES OF RBPJ LOSS ...... 98 6.7 FUTURE DIRECTIONS ...... 100 6.7 IMPLICATIONS OF WORK ...... 104

REFERENCES ...... 108 APPENDIX ...... 123 Appendix A. Previously published material ...... 123

LIST OF TABLES

Table 1.1 RBPJ nomenclature in different species ...... 13

Table 1.2 Target genes where direct RBPJ binding has been identified by ChIP, EMSA or luciferase reporter assays ...... 26

Table 2.1 List of primer sequences used ...... 38

Table 3.1 Features of breast cancer cell lines used for aCGH analysis at the RBPJ locus ... 52

Table 6.1 Summary of proposed experiment evaluating the requirement of HEY2 downstream of RBPJ loss in MDA-MB-231 cells ...... 102

vii

LIST OF FIGURES

Figure 1.1 Features of carcinogenesis ...... 3

Figure 1.2 Structure of the human mammary gland ...... 4

Figure 1.3 The Notch signaling cascade ...... 8

Figure 1.4 RBPJ recognition sequence, exon structure and knock-out strategies ...... 15

Figure 1.5 Schematic of the RBPJ nuclear cycle ...... 20

Figure 2.1 Exon position where hairpin constructs target the RBPJ transcript ...... 36

Figure 3.1 RBPJ protein is expressed in various normal human tissues ...... 49

Figure 3.2 RBPJ is frequently lost in human cancer cell lines ...... 50

Figure 3.3 RBPJ is lost in primary human breast cancers ...... 54

Figure 3.4 RBPJ is lost in primary human lung cancers ...... 55

Figure 3.5 RBPJ depletion increases MDA-MB-231 xenograft tumor growth ...... 56

Figure 3.6 Illustrated mode of action of mt-RBPJ ...... 57

Figure 3.7 RBPJ loss accelerates DG75 xenograft tumor growth ...... 58

Figure 3.8 More aggressive breast carcinomas have reduced RBPJ expression ...... 59

Figure 4.1 HEY gene induction occurs with RBPJ removal ...... 64

Figure 4.2 HEY gene induction resulting from RBPJ depletion is Notch independent ...... 65

Figure 4.3 Complete loss of RBPJ causes a Notch-independent HEY1 and HEY2

induction ...... 66

Figure 4.4 Evidence that RBPJ and HEY2 mRNA expression are negatively correlated in human breast cancers ...... 67

Figure 4.5 Effect of HEY2 on xenograft tumor growth ...... 69

Figure 4.6 Evidence for target gene induction in Rbpj KO cells derived from the Rbpj null mouse ...... 70

Figure 5.1 Changes in histone marks at the HEY1 and HEY2 promoter upon KD of RBPJ in MDA-MB-231 cells ...... 76

Figure 5.2 Correlation of acH4 enrichment with up-regulated gene expression in MDA-MB- 231 RBPJ KD cells ...... 79

viii

Figure 5.3 RBPJ target genes are enriched in the top 1000 ranked acH4 gene list in MDA- MB-231 cells ...... 81

Figure 5.4 Identification of differentially enriched cellular functions ...... 81

Figure 5.5 Validation of the cellular death function in vitro ...... 82

Figure 5.6 Staining and quantitative mapping of multiple markers in whole tumor

cryosections ...... 84

Figure 5.7 RBPJ deficient tumors have reduced cell death ...... 85

Figure 5.8 Differential effect of RBPJ removal on tumor cell proliferation in vitro and in vivo . 86

Figure 6.1 Schematic illustrating possible effects of RBPJ loss ...... 107

LIST OF ABBREVIATIONS aCGH array comparative genomic hybridization acH4 acetylated histone H4

ADAM a disintegrin and metalloprotease

ANK ankyrin domain

B-H Benjamini-Hochberg multiple testing correction bHLH basic helix-loop-helix

BrdU 5-bromo-2-deoxyuridine

ChIP chromatin immunoprecipitation

ChIP-seq ChIP followed by sequencing

CIR C promoter binding factor interacting co-repressor

CNS central nervous system

CoA co-activator

CoR co-repressor

CpG cytosine-phosphate-guanine DNA region

CtIP CtBP interacting protein

DAPT N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester

DCIS ductal carcinoma in situ dUTP 2´-deoxyuridine, 5´-triphosphate

E(spl) enhancer of split

EGF epidermal growth factor

EMSA electrophoretic mobility shift assay

ER estrogen receptor alpha

FLH1 four and a half LIM domains 1

GAPDH glyceraldehyde 3-phosphate dehydrogenase

GFP green fluorescent protein

H&E hematoxylin and eosin

H3K4me3 histone H3 trimethylated at lysine 4

HAT histone acetyltransferase

HC hematopoietic cell lines

HDAC histone deacetylase

HER2 human epidermal growth factor 2

HES hairy and enhancer of split

HEY HES related with a YRPW motif

HIF1α hypoxia inducible factor 1 alpha

HIV human immunodeficiency virus

HUGO human genome organization

IDC invasive ductal carcinoma

IHC immunohistochemistry

ILC invasive lobular carcinoma

IPA Ingenuity Pathway Analysis

IRES internal ribosomal entry site

KD knock-down

KO knock-out

LCIS lobular carcinoma in situ

LTR long terminal repeat

MAML mastermind-like 1

MIG MSCV-IRES-GFP

MMTV mouse mammary tumor virus

MSCV murine stem cell virus promoter mt mutant or mutated

MTG16 myeloid translocation gene 16

NcoR nuclear receptor co-repressor

NF-κB nuclear factor-kappaB

NOD/SCID non-obese diabetic/severe combined immunodeficient mice

NotchIC intracellular domain of Notch

NSCLC non-small cell lung carcinoma xi

oligo oligonucleotide

PI propidium iodide

PR progesterone receptor qPCR real-time quantitative polymerase chain reaction

RAM RBPJ-associated molecule

RBP2N dominant isoform of RBPJ

RBPJ recombination signal binding protein for immunoglobulin kappa J region

RBPJL RBPJ-like

ROS reactive oxygen species

SHARP SMRT and HDAC associated repressor protein shRNA short hairpin RNA

SKIP Ski-interacting protein

SMRT silencing mediator of retinoid and thyroid receptors

Su(H) suppressor of hairless

TBHP tert-butyl hydroperoxide

TDLU terminal ductal lobular unit

TFIIA transcripton factor II A

TMA tissue microarray

TNM tumor, node, metastasis

TSS transcription start site

TUNEL terminal deoxynucleotidyl transferase dUTP nick end labeling

WAP whey acidic protein

WRPW tryptophan-arginine-proline-tryptophan amino acid motif wt wild-type

YFP yellow fluorescent protein

YRWP tyrosine-arginine-proline-tryptophan amino acid motif

xii

ACKNOWLEDGEMENTS

I owe my deepest gratitude to my grandfather, Franjo Kalodera, who keeps a file with every one of my report cards. Without him this thesis would not have been possible. The stories he read to me about ocean life and our seaside sample-collecting trips inspired me to become a biologist. I thank my parents, Boris and Palma Kulic, for supporting me in everything I do. I am grateful to have my older sister, Dana Kulic, to relate to when it comes to science and my younger sister, Matea Kulic, to distract me from science.

A special thank you goes to my fiancé, James Heyes, for his knowledge and encouragement in all things, his infectious positive attitude and for being the foundation of my life. His dedication to reading my thesis while sailing in a storm will not be forgotten.

I am grateful to my supervisor, Dr. Aly Karsan, for seeing potential in an experiment ‘gone wrong’ and encouraging me to pursue this work at a time when I was only interested in studying angiogenesis. He has refined my “Incoherent Logic” and given me the opportunity to learn many interesting techniques.

Many thanks to the members of my supervisory committee, Drs. Carolyn Brown, Marco Marra and Calvin Roskelley for their interest in my project, helpful suggestions and guidance.

I would like to thank the many people who have contributed to this highly collaborative project, especially Gordon Robertson for his patience and enthusiasm when it came to introducing me to bioinformatics and Jennifer Baker for her staining expertise.

I would also like to thank all the members of the Karsan lab with whom I have had the pleasure of working. In particular Shauna Dauphinee for her advice and support and for making the shared experience of grad school the highlight of my PhD. I am also grateful to Megan Fuller, Fred Wong and Winnie Mok for contributing to this project, Nelson Wong for insightful conversations and feedback, and Jeremy Parker for his editing talents.

Financial support during my graduate studies was provided by a Master’s scholarship from Natural Sciences and Engineering Research Council of Canada, a junior graduate studentship from the Michael Smith Foundation for Health Research, a doctoral award from Canadian Institutes of Health Research, and a fellowship from the University of British Columbia.

xiii

CHAPTER 1: INTRODUCTION

1.1 CANCER

In Canada 45% of men and 39% of women will develop cancer during their lifetimes and 1 in 4 people will die of the disease1. Overall mortality rates for all cancers combined continue to decline because of improvements in early diagnosis and treatment1. However, cancer is not a single disease but rather a heterogeneous group of diseases. Even when these diseases occur in the same tissue type, such as cancers of the breast, they can have vastly different aberrations and outcomes2.

How normal human cells transform into malignant tumors has been in the spotlight of research since the early 1970s. Initial work established the foundation that tumors are caused by heritable genetic mutations in individual cells that result from viral infection and/or environmental exposure. Initiation due to a primary transforming event is followed by progression, during which additional heritable changes result in clonal expansion and evolution3,4. Changes can also be epigenetic as these too are heritable and therefore subject to natural selection, and can cause aberrant expression of genetic material. These transforming mutations occur in two types of genes: (1) oncogenes, which are produced by gain-of-function alterations that promote tumorigenesis, and (2) tumor suppressor genes which have loss-of- function alterations that impair their ability to oppose tumorigenicity. Most tumor suppressor genes are recessive such that expression of both alleles must be lost for the phenotype to emerge, whereas oncogene mutations are dominant.

On a cellular level, a number of key phenotypic alterations occur in tumor cells compared to normal counterparts, and these have been termed the hallmarks of cancers5. They are as follows: (1) insensitivity to anti-growth signals, (2) self-sufficiency in growth signals, (3) evasion of apoptosis, (4) limitless replicative potential (immortalization), (5) the presence of an inflammatory microenvironment, (6) sustained angiogenesis and (7) tissue invasion and metastasis5,6. The last of these, the capability of a benign tumor to progress to a malignant one, is perhaps the most decisive feature, as tissue invasion and metastasis are responsible for most cancer mortalities7. Additional hallmark features have been proposed, including increased glycolysis and resistance to subsequent microenvironmental acidification8. Figure 1.1 summarizes these hallmarks with an example protein that can participate in each of these processes.

Limitless replication Sustained angiogenesis Increased glycolysis Migration, invasion and resistance to anoikis

Evading apoptosis Insensitivity to anti-growth Hallmark Self-sufficient growth Inflammatory microenvironment Resistance to acidosis

Normal Hyper- Primary Meta- DCIS Carcinogeneisis Cell plasia Tumor stasis

Notch Cyclins HIF1α MMP Molecular strategy HER2 Myc Slug

Figure 1.1 Features of carcinogenesis (adapted from Gatenby and Gillies 20088 and Fang et al. 20089). Steps from development of a primary tumor to dissemination are shown in blue, and progress past hallmark obstacles is marked in red boxes. Strategies to overcome barriers may occur in a distinct order for different cancers. Anoikis is a specific form of apoptotic cell death that results following detachment from extracellular contacts. Proteins discussed in the text have been chosen as examples of potential molecules that can mediate progression when deregulated. These may play a role in multiple stages but may not be important in all cancers. HIF1α can up-regulate transporters that mediate resistance to an acidic microenvironment. DCIS, ductal carcinoma in situ.

1.2 BREAST CANCER

Breast cancer is the most commonly diagnosed cancer among Canadian women and incidence rates for this cancer type are similar to other developed countries but are lower in less developed countries1. One in nine Canadian women will develop breast cancer and a third of the women diagnosed will die from the disease, making it the leading cause of cancer-related death in women worldwide10. The most common site of breast cancer metastasis is bone, followed by lungs, liver and brain. In transgenic mouse models of breast cancer metastasis, the seeding organ is predominantly the lungs11. Breast cancer is 100 times less frequent in men than in women12.

Normal breast tissue consists of ducts and lobules, each of which has an inner epithelial lining (derived from luminal progenitors) and a supporting contractile myoepithelial layer found in the basal compartment where mammary stem cells are thought to reside13,14 (Figure 1.2). Milk gets produced in the clustered alveoli of the lobules and is secreted into the ducts that drain through the nipple. The epithelial layer expresses estrogen receptor alpha (ER) and

A B Lobules Lobes

Basal cell

Basement membrane

Nipple Luminal cell

Myoepithelial cell Ducts TDLU

Figure 1.2 Structure of the human mammary gland (adapted from Ali and Coombes 200215 and Cowin et al. 200516). (A) Ducts that drain at the nipple give rise to lobules, which are made up of alveolar sacs that support many clustered alveoli for milk production during pregnancy. Breast cancer arises in the terminal ductal lobular unit (TDLU)17,18. (B) Cell composition of a duct cross-section showing luminal epithelial cells that line the ducts surrounded by contractile myoepithelial cells. Basal cells represent the putative stem cell population14.

progesterone receptor (PR) at various degrees; however, the myoepithelial layer is hormone receptor negative19. Estrogen and progesterone signaling play an important role in the development and function of breast tissue, and high levels of these hormones are present during pregnancy20. However, inappropriate ER or PR signaling can promote tumor growth and progression21,22.

There are numerous ways to classify breast cancer that are useful for both diagnosis and for therapeutic strategies. They include: (1) histological subtypes, (2) histological grade, (3) tumor, node, metastasis (TNM) staging, (4) immunohistochemical markers, and (5) molecular classification.

The two major histologic subtypes, invasive ductal carcinoma (IDC) and invasive lobular carcinoma (ILC), represent the most common malignancies of the breast; these account for approximately 80% and 15% of all invasive breast tumors respectively23. IDC and ILC are characterized by having already penetrated the basement membrane. Earlier-stage tumors that are confined to the lumens are classified as in situ (lobular carcinoma in situ, LCIS or DCIS). 4

Although it is commonly stated that ILC begins in the lobules (the glands that produce milk) and ductal carcinoma originates from the draining milk ducts, both types actually arise in the TDLU17,18. Lobular tumor cells are distinguished by being loosely cohesive due to frequent loss of the cell-cell adhesion molecule E-cadherin24 and they tend to invade in a single file, while ductal tumors tend to form glandular structures23.

Histologic grading and TNM staging are both used as indicators of prognosis. Grading is based on the degree of differentiation, nuclear changes and mitotic cell number, whereas staging classifies the progression of a cancer by evaluating the size and spread of the primary tumor, lymphatic involvement and the presence (or absence) of distant metastases18.

Immunohistochemical markers, such as the presence of the receptors ER, PR and human epidermal growth factor receptor 2 (HER2, also known as ERBB2), can be used to determine the suitability of treatments. ER and PR positive receptor status is a good prognostic marker as these breast tumors are more likely to be responsive to anti-hormonal and chemotherapeutic treatment25, whereas HER2 gene amplification and over-expression is associated with an unfavorable prognosis although it also predicts response to therapy26. ILC are more often ER and PR positive23 where HER2 expression has been linked with DCIS26.

Five molecular classes have been identified by gene expression profiling27, although in the clinic immunohistochemical protein markers are used as surrogates for gene expression in subtyping breast cancer. The current molecular subtypes are luminal A (ER+ and/or PR+, HER2- and low grade), luminal B (ER+ and/or PR+, HER2+ but often high grade), HER2+ (ER- /PR-) basal-like (ER-/PR-, HER2-) and unclassified or normal breast-like (negative for all 3 markers but also negative for cytokeratins 5 and 6)28. These classes correspond to cancers that are believed to originate from different mammary progenitor cell populations29. The expression profiles for luminal subtypes is similar to normal lumen cells that line the breast ducts and glands28. HER2+ and basal-like cancers tend to grow rapidly, but like ER+ tumors, HER2+ cancers respond to treatment with specific pathway inhibitors30, whereas basal cancers are high-grade and represent a very aggressive subtype28. Of note, heterogeneity also exists in the tumor microenvironment and stromal subtyping can also provide useful prognostic information31. Despite these advances more work is needed in breast cancer research to elucidate factors involved in tumor progression, metastasis and recurrence so we may better be able to predict these events and overcome therapeutic resistance19.

1.3 THE NOTCH SIGNALING PATHWAY

Regulated activation of the Notch pathway is required for normal developmental processes such as differentiation, proliferation and apoptosis, but aberrant Notch signaling has been implicated in a growing number of human cancers, including breast cancer32-34. The Notch gene earned its name from studies in the early 1900s in Drosophila where loss of one gene copy resulted in wing notching32,35. In this system Notch signaling specifies the cells that organize the outgrowth of the wing36. A partial loss of Notch prevents cell fate specification causing a phenotype of notches in the wing. Conversely, ectopic Notch expression produces extra wing tissue36. The Notch gene encodes a transmembrane receptor and in mammals four paralogs have been identified, and named Notch1 to Notch4. These interact with one of 5 ligands that are also transmembrane proteins: Jagged1 and 2 and Delta-like1, 3 and 4. Loss-of-function studies in mice have revealed the essential roles of these proteins in development. Null mice for any Notch ligand show embryonic or postnatal lethality with severe abnormalities resulting from skeletal and vascular defects37-41. Notch342 and Notch443 knock-out mice are viable and fertile. However, Notch143,44 or Notch245 inactivation results in embryonic lethality with neuronal and vascular remodeling defects. A synergistic effect is observed with the Notch1 and Notch4 double knock-out with respect to abnormal vascular development43. This demonstrates that some Notch receptors have partially overlapping roles. Notch receptors and ligands differ from their respective family members predominantly by the number of epidermal growth factor (EGF)-like repeats within the extracellular domain of the proteins which are primarily involved in ligand-receptor binding46. Notch proteins are synthesized as inactive precursors and must undergo several cleavage events to become active.

The Notch signaling cascade is a direct paracrine signaling system. It is initiated by binding of a Notch receptor to its cognate ligand, present on the membrane of a neighboring cell, causing signal activation by proteolytic cleavage of the receptor (Figure 1.3). The extracellular cleavage is mediated by a disintegrin and metalloproteinase (ADAM10 or ADAM17)47 and occurs following endocytosis of the ligand (bound to the extracellular domain of Notch) into the ligand-expressing cell48. What remains of the Notch receptor is subsequently cleaved by a γ- secretase complex (either on the cell surface or in endosomal compartments) which liberates the intracellular domain of the receptor (NotchIC) from the membrane tether and allows its translocation into the nucleus47-49. Indeed, Notch has been described as a membrane-bound transcription factor whose release from the surface is controlled by ligand activation50. NotchIC

represents the functional portion of the receptor responsible for protein activity. Over-expression of this domain results in constitutively active Notch signaling35,51,52.

Since NotchIC lacks a DNA-binding domain, transcriptional activation of Notch target genes occurs via association with RBPJ (recombination signal binding protein for immunoglobulin kappa J region). This DNA-binding partner mediates signaling by all four Notch receptors in mammals. The RBPJ-associated molecule (RAM) domain, and seven consecutive ankyrin repeats of NotchIC are responsible for binding to RBPJ35,53. In the absence of NotchIC, RBPJ confers transcriptional repression via a co-repressor complex that includes SMRT, SHARP and histone deacetylases (HDACs)35. Displacement of these co-repressors, with concomitant recruitment of co-activators, such as Mastermind-like 1 (MAML) and histone acetyltransferases (HATs), is required for activation of RBPJ-mediated transcription35,48. The basic helix-loop-helix (bHLH) transcriptional repressors of the HES and HEY families are some of the best characterized Notch/RBPJ target genes54,55 and are commonly used as indicators of Notch pathway activity. Nuclear interactions between components of the Notch signaling pathway are discussed in further detail in subsequent sections of the introduction (section 1.9 and 1.10).

This highly conserved56 signaling cascade is direct with no amplification steps from activation of membrane bound receptor, to nuclear translocation of NotchIC, to induction of gene expression by RBPJ. This direct path allows for rapid signal transduction and tight control of signal intensity as well as fast signal attenuation. Indeed, NotchIC is degraded in the nucleus in a matter of minutes57,58. Alternately, the full receptor can be internalized from the surface for degradation via Numb59. Perturbation of the balance between transcriptional activation and repression can manifest in a number of disorders, including cancer4.

Although recent work has greatly contributed to our knowledge of Notch signaling, many pieces of the puzzle remain unresolved. Factors that control expression of Notch pathway components and determine which cells within tissues preferentially become the signal-sending or signal-receiving cells are questions yet to be fully answered. The supposition that following Notch binding, ligand processing in the ligand-expressing cell causes activation of a signaling cascade within that cell60-62, leading to bidirectional signaling, needs further corroboration. Identification and validation of direct target genes continues to add to the knowledge of how Notch signaling exerts its effects in different contexts. However, a more thorough knowledge of the role of secondary events, such as those that result from HES and HEY induction will further complete the picture.

Figure 1.3 The Notch signaling cascade. Activation of one of four Notch receptors found in mammals occurs by binding to one of five ligands of the Jagged (Jag) or Delta-like (Dll) families. This interaction initiates the signaling cascade, causing two cleavage events that free the active intracellular domain of the receptor (NotchIC) from the membrane and allow its translocation into the nucleus. DAPT, a highly specific γ-secretase inhibitor, is routinely used to block the second cleavage event preventing activation of the pathway. Once in the nucleus, NotchIC complexes with co-activators (CoA) and RBPJ which binds to the 5'-(C/T)GTGGGAA-3' consensus sequence on DNA33. This interaction results in displacement of co- repressors (CoR) activating transcription of a number of target genes including those of the HEY family. Pathway image created using proteinlounge freeware (www.proteinlounge.com). 8

1.4 NOTCH SIGNALING IN CANCER

Besides playing a key role in cell fate determination during embryonic development, the Notch pathway regulates homeostasis in multiple adult tissues through stem cell maintenance (in hematopoietic, neural, mammary and intestinal tissue, among others)56,63-65. Pathways that are important for such functions must be down-regulated to allow differentiation and often re- emerge during tumorigenesis, as they confer a plastic stem-like phenotype46,63.

In most human malignancies the deregulation of Notch signaling has oncogenic outcomes, although in some contexts such as the skin, Notch signaling can be tumor suppressive58,66. The observation that aberrant Notch signaling is associated with bone cancer67, brain cancer68, colorectal cancer69, prostate cancer70, ovarian cancer71, pancreatic cancer72, breast cancer, lung cancer, leukemia, melanoma and other malignancies33,46,58, highlights the importance of tight control of this pathway in many different tissues. In addition to aberrant activation of this pathway in the tumor cell population, Notch signaling also contributes to tumorigencity by affecting the stroma. A fine balance of Notch signaling is required in vascular development and the pathway plays a key role in tumor angiogenesis46,58. Thus far, not much is known about the contribution of Notch signaling to the inflammatory microenvironment of solid tumors but because of its role in the immune system34 and macrophage activation73, it is likely to be important.

The four Notch receptors are rarely mutated in human cancers74, with the exception of T- lineage acute lymphoblastic leukemia, where frequent Notch1 receptor gain-of-function mutations occur that result in ligand-independent pathway activation or cause an increase in the half-life of NotchIC in the nucleus75. However, aberrant expression of Notch pathway machinery is detected in many solid cancers33,49,74. Mutations may exist in factors that regulate the gene expression of Notch signaling components, enzymes that process the ligands and receptors, or proteins of the co-activator and co-repressor complex that control Notch target gene expression in the nucleus.

Several molecular opportunities exist for inhibition of the Notch signaling pathway for therapeutic purposes49. The clinical focus has been on agents that block Notch receptor cleavage by inhibition of γ-secretase enzyme activity, thereby preventing Notch activation46,58. A number of ongoing clinical trials using γ-secretase inhibitors for the treatment of cancer can be found at ClinicalTrials.gov. The chemical DAPT (N-[N-(3,5-difluorophenacetyl)-L-alanyl]-S- phenylglycine t-butyl ester), shown in Figure 1.3, is one such γ-secretase inhibitor. 9

1.5 NOTCH SIGNALING IN BREAST CANCER

Notch receptors and ligands show the highest expression in the luminal epithelium of normal breast tissue and are important in normal mammary gland function13,76. Notch signaling is thought to maintain cells of the luminal lineage in an undifferentiated state13,77. Specifically Notch3 may be critical in differentiation of human bipotent progenitors into the luminal lineage, whereas Notch4 is expressed most highly in the basal undifferentiated mammary progenitors78. Using in situ hybridization, Notch3 mRNA has been detected in normal luminal epithelial cells, while Jagged1 expression in the adjacent myoepithelial layer suggests interaction between the ligand/receptor pair79. In the mouse mammary gland Notch1, 2 and 3 and Jagged1 are preferentially expressed in the luminal subtypes (with Notch1 showing the highest expression) versus the basal mammary stem cells where their expression is still detected but at lower levels (although this subtype had increased expression of Delta-like1)13. Forced expression of Notch1IC in mouse mammary stem or progenitor cells results in hyperplastic luminal nodules that eventually progress to tumors13.

Work in recent years has defined a contributory role for Notch signaling in breast cancer46. Notch1 and Notch4 were identified as common mouse mammary tumor virus (MMTV) integration sites that resulted in a truncated and constitutively active Notch which caused development of mammary tumors in mice80. Indeed over-expression of a constitutively active form of Notch181,82, Notch383, or Notch484,85 in the mouse mammary gland driven by mammary specific promoters leads to arrest of alveolar development and mammary tumorigenesis86. Over-expression of a hyperactivate Notch176 or Notch487 in vitro is also able to transform human or mouse mammary epithelial cells respectively.

Blocking Notch signaling in a human-to-mouse MDA-MB-231 xenograft model of breast cancer (the same approach used in this thesis), inhibits xenograft tumor growth and metastases. Preventing endogenous ligand-induced Notch activation via a soluble ectodomain of Notch1 or Notch4 resulted in inhibited tumor growth and reduced the number and size of metastases88. Similarly, treatment of MDA-MB-231 tumor xenografts with a γ-secretase inhibitor that blocks Notch cleavage inhibited tumor growth and metastases compared to vehicle-treated control animals89,90, although recent work has demonstrated that the particular γ-secretase used in these studies (Z-LLNle-CHO) mediates proteosome inhibition in addition to down-regulating the Notch pathway91. Conversely, over-expression of mouse Notch4IC in human MDA-MB-231 cells greatly accelerates tumor growth92. Interestingly, Notch2 may oppose the oncogenic effect

of Notch1 and Notch4 in MDA-MB-231 cells and its expression has been correlated with increased survival in human breast cancers92,93. The mechanism of this tumor suppressor role of Notch2 is unknown but may occur though differential regulation of cytokines92.

Notch receptors, ligands and downstream targets are expressed in primary human breast tumors and have been associated with poor outcome. Expression of all Notch homologues has been detected in human breast cancer, along with ligands Jagged1, Jagged2, Delta-like1 and Delta-like490,94. Protein levels of these four ligands were increased in grade 3 IDC when compared to normal breast specimens, as was protein expression of Notch1, Notch2 (albeit slight) and Notch3; Delta-like1 and Notch3 were undetectable in normal breast tissue94. Elevated mRNA expression of Notch1, Notch3 or Jagged1 are associated with reduced overall survival in human breast tumors, with co-expression of Jagged1 and Notch1 predicting a very poor outcome79,95. Notch3 has been previously reported to be present at high levels in breast tumor vasculature32. Vascular expression of Delta-like4 may also be relevant in some breast cancer types96. Notch4 did not show expression changes between breast cancer and normal tissue but was the most highly expressed receptor in the normal tissue samples94.

In addition to expression of pathway components, there is now evidence that the Notch pathway is functionally active in breast cancer. Notch1 nuclear localization and Jagged1 membrane expression is significantly correlated with Ki67, a proliferation marker97. We have reported a positive expression correlation between Jagged1 and the Notch target genes HEY1, HEY2 and HEYL from analysis of two independent breast cancer microarray datasets88. Moreover, active Notch1IC protein co-stains with downstream targets HES1 and HES5 in two thirds of breast cancers analyzed compared to undetectable levels in normal tissue as measured by immunohistochemistry94. This co-expression was also present in hyperplasia and DCIS, suggesting Notch activation may be an early event in breast cancer progression94. Although normal breast tissue expresses Notch receptors and ligands, the absence of Notch activity in the majority of cells may be due to the presence of Numb, a negative regulator of Notch signaling. Loss of Numb occurs in 50% of human breast cancers and Numb protein expression is inversely correlated with NotchIC, an association that predicts more aggressive disease76,94,98.

Recent studies have begun to address the involvement of Notch in different breast cancer disease subtypes. Staining for Notch1, in particular, has been reported in primary human breast ductal carcinoma. Protein accumulation of Notch1IC and Notch4IC and the target HES1 in DCIS is associated with breast cancer recurrence57,94,99. However Notch1, Notch4, Jagged1 and 11

Delta-like1 staining was also detected in ILC, with Notch4 protein being present in the highest proportion of these tumors90. Notch signaling activity has been linked with basal-type breast cancers (ER, PR, and HER2-)89,90. These tumors stain for Notch1 protein, and in a microarray analysis, expression of Notch1 mRNA is associated with basal-like cancers over other molecular subtypes100. Notch transcriptional activity has been shown to be preferentially up- regulated in ER- breast cancer cell lines90. This finding was initially puzzling as in the absence of estrogen, Notch mediates nuclear localization of ER and cooperates in activating ER- dependent transcription101. However, estrogen acts in a feedback mechanism to inhibit this process. Although estrogen can increase Notch1 protein at the membrane, it inhibits Notch cleavage by an unknown mechanism, thus reducing transcriptional activity by down-regulating NotchIC90,97. Therefore, canonical estrogen signaling inhibits activation of the Notch pathway. HER2 also inhibits Notch signal transduction providing further evidence that Notch signaling may be active in the basal-like molecular subtype of breast cancer49. This has therapeutic implications as both pharmacological inhibition of estrogen and HER2 can reactivate Notch signaling49,90. Hence, Notch pathway components may be promising targets in ER- and/or HER2- tumors, or ER+ and HER2 over-expressing tumors in combination with inhibitors of ER and HER2 signaling. Jagged1 mRNA expression has also been associated with a basal subtype of breast cancer102. A discrepancy exists between the association of Notch activity and basal breast cancer and observations in normal mammary cells showing Notch specifies luminal cell fate and limits basal cell proliferation13,77. Recently, over-expression of Notch1 and Jagged1 was noted in the luminal subtypes of human breast cancer over the basal group103. Moreover, constitutively active Notch1 transformed mouse mammary stem cells into luminal-like tumors13. Hence the role of Notch signaling in a particular molecular subtype remains to be fully resolved. Perhaps aberrant Notch activity is present in multiple subtypes but in the triple negative basal- like cancers additional abnormalities prevent lineage specification.

1.6 A CLOSER LOOK AT RBPJ

The transcriptional regulator RBPJ was initially identified as a DNA binding protein specific to the immunoglobulin J (joining) type recombination signal of V(D)J recombination104,105. Enzymes involved in this process are generally referred to as recombinases105, and those of the integrase family contain a conserved 40 residue motif that forms a covalent link with DNA at the site of recombination106. RBPJ was presumed to have a role in this process; however, the binding sequence was later shown to be different from the recombination signals, and the 12

integrase-like motif of RBPJ was not essential for DNA binding activity107,108. Instead, the two surrounding regions were found to be important, especially the residues flanking the N-terminal side of the integrase-like motif108. Although RBPJ does not have activity related to recombination of antigen receptor genes107,109, the protein still retains its historic name.

RBPJ has been highly conserved through evolution with 84% amino acid identity between Drosophila and the human protein and 92% identity between mouse and human RBPJ53,110. Table 1.1 summarizes naming of RBPJ across various species. The initial characterization of RBPJ structure came from evaluation of the predominant transcript in mice (called RBP2N), which is one of 4 isoforms in this organism. The human gene has 7 transcripts described in the GRCh37 (hg19) UCSC Genome Browser assembly. There may be more splice variants in addition to those reported111, although the predominant transcript in humans is also RBP2N112. Both human and mouse splice variants differ in the N-terminal region, as a result of different transcriptional start sites and first exon sequences105,113. The functional diversity of the isoforms remains to be investigated. All isoforms contain a nuclear localization signal (exon 4)114, and the integrase-like motif (split by intron 6) with adjacent regions important for DNA binding105,107. Crystallography structures have revealed three structurally integrated RBPJ protein domains. The N-terminal Rel homology domain functions similarly to that of Rel transcription factors,

Table 1.1 RBPJ nomenclature in different species.

Abbreviated name Full Name Species Reference

CBF1 C promoter binding factor 1 Homo sapiens (Ling et al. 1993)115

General gene CSL CBF1, Su(H), LAG1 family name (Lai 2002)36

Caenorhabditis LAG1 lin-12 -and glp-1 elegans (Christensen et al. 1996)116

Mammalian recombination signal binding protein species* RBPJ for immunoglobulin kappa J region First identified in (Hamaguchi et al. 1989)104 Mus musculus Drosophila Su(H) Suppressor of hairless melanogaster (Furukawa et al. 1991)117

*HUGO gene nomenclature now recognizes RBPJ as the official gene name in humans.

binding the major groove of DNA. It also interacts with the ankyrin domain of Notch. The DNA- binding beta-trefoil domain is involved in binding the minor groove of DNA, the RAM domain of Notch (NotchIC interacts with RBPJ primarily though this region) and association with co- repressor proteins118. The C-terminal domain interacts with the co-activator MAML and the ankyrin domain of Notch53,119,120 (Figure 1.4A).

RBPJ is ubiquitously expressed in normal tissue110,121, having a promoter sequence characteristic of a housekeeping gene105. The protein shows nuclear localization111,122, and the canonical 5’-(C/T)GTGGGAA-3’ RBPJ binding sequence is now well validated33,108,109. We confirmed this consensus sequence using experimental data for 47 diverse binding regions from the literature to generate an RBPJ-like motif for use in subsequent high throughput computational motif discovery (Figure 1.4B shows the sequence logo). The number and orientation of RBPJ binding motifs can modify the strength of the transcriptional activation and presumably repression123. For example, at promoters of some Notch-responsive genes such as HES1, two Notch transcriptional complexes bind to tandem responsive elements on DNA and can dimerize (mediated by ankyrin-ankyrin interactions on NotchIC), resulting in more potent transcriptional activation124.

The RBPJ DNA binding element overlaps regulatory sequences recognized by other transcription factors. Ikaros, a lymphoid-cell-specific transcriptional repressor recognizes the consensus binding site GGGAA125, in addition to its alternate binding sequence (TGGGGGT)4, and can compete with RBPJ for DNA binding126. Ikaros has also been shown to bind simultaneously with RBPJ to promoter regions that contain a recognition sequence for each protein, strengthening RBPJ-mediated transcriptional repression125. The consensus motif of RBPJ can occur within degenerate nuclear factor-kappaB (NF-κB) elements and RBPJ has been shown to regulate gene expression from a proportion of NF-κB binding sites4. The E- Twenty-Six (Ets) family of transcription factors recognize a partial sequence (GGAA/T) in the RBPJ canonical motif and could also potentially regulate transcription from this site4.

RBPJ null mouse models and a human knock-out (KO) cell line have been generated (Figure 1.4C). The original complete mouse KO was made by insertional mutagenesis, resulting in a disruption of exon seven which is present in all RBPJ transcripts127. Growth retardation and major developmental abnormalities (incomplete turning of the body axis, microencephaly,

NotchIC-ANK NotchIC-RAM A B SKIP CIR, SMRT NotchIC-ANK, MAML

1 2 3 4 5 6 7 8 9 10 11

NLS DNA binding motifs

DNA binding motif flank integrase-like motif neo C 1 2 3 4 5 6 7 8 9 10 11

loxP loxP D 1 2 3 4 5 6 7 neo 8 9 10 11

E 1 2 3 4 5 6 7 8 9 10 11

Figure 1.4 RBPJ recognition sequence, exon structure and knock-out strategies. (A) RBPJ gene organization is highly conserved between mouse and human, with isoforms differing in the N-terminal region. Exons are shown as blue boxes, introns are displayed as separating lines, and important protein interaction domains are labeled in black with DNA binding domains and a nuclear localization signal (NLS) in red. The integrase-like motif is split by intron 6 and two separate regions to either side of this motif are required for DNA binding (especially the 5’ region); these represent some of the most highly conserved residues in the protein108. Not drawn to scale. Ank = ankyrin domain. (B) Sequence logo representation of the RBPJ DNA binding motif derived by us for computational work. RBPJ gene inactivation strategies are shown (C to E) each resulting in a frame shift (denoted by a dashed exon outline) following targeted gene disruption. (C) The original mouse Rbpj KO was generated by insertion of neomycin (neo) in the reverse orientation into the integrase motif causing a disruption of exon 7127. (D) The mouse conditional Rbpj KO results from deletion of exon 6 and exon 7 upon excision of the loxP-flanked sequence by tissue specific expression of Cre recombinase128. (E) The human RBPJ KO cell line has deletion of exon 4, resulting in retention of only three in- frame exons and a disruption of the downstream sequence113.

abnormal placental development, anterior neuropore opening and defective somitogenesis) were observed in 8.5 to 9.5 day embryos, resulting in lethality after 10.5 days of gestation127. The phenotype of these Rbpj KO mice is similar to, but more severe than, the Notch1 KO mice129. A conditional KO mouse was generated with a somewhat different gene inactivation strategy from the original KO. Conditional excision of exons 6 and 7 mediated by Cre expression generates a frameshift in exon 8, resulting in the removal of the full integrase-like motif as well as the primary NotchIC binding domain (Figure 1.4D)128. This mouse has been used to inactivate Rbpj in numerous tissue types, including the mammary gland86,130.

An alternate RBPJ KO made in an independent laboratory further shortens the resulting protein from the one potentially made in the KO mice113 (Figure 1.3E). The human Burkitt’s lymphoma cell line, DG75, is amenable to targeted gene inactivation and is reported to express RBPJ from a single functional allele113. The genomic fragment encompassing exon 4 of all RBPJ alleles was successfully deleted in these cells113. This strategy produces a frameshift mutation resulting in loss of all DNA binding sites and interaction domains with only half of the N-terminal Rel homology domain retained.

1.7 RBPJ IN MAMMARY DEVELOPMENT AND CANCER

Conditional deletion of Rbpj in the mouse mammary gland using Cre-loxP-mediated recombination has been evaluated in both a normal developmental context and a cancer model. RBPJ inactivation via the MMTV promoter (which is active in progenitor cells during mammary development) favored formation of myoepithelial over luminal cell populations during pregnancy77. Based on this phenotype, the study concluded that Notch signaling maintains the luminal cell fate and regulates alveolar development77. The finding was confirmed in a parallel system by conditional deletion of Pofut1 (O-fucosyltransferase 1), an enzyme required for Notch receptor activity which mediates post-translational modification of Notch EGF-like repeats77.

Within the context of normal mammary development RBPJ function has also been modified using the mouse mammary reconstitution model. This in vivo assay evaluates regeneration of the functional mammary gland upon orthotopic transplantation of primary mammary epithelial cells131. Knock-down of RBPJ specifically in the putative mammary stem cell enriched population (defined by CD24 positivity and high expression of surface marker CD29/β1-integrin) caused expansion of the stem cell population and increased mammary outgrowth13. These

outgrowths had more basal cells that stained positive for myoepithelial markers13. In comparison, over-expression of Notch1IC in the mammary stem cell enriched population leads to formation of luminal hyperplastic nodules within mouse mammary glands13.

Mammary reconstitution has been used to over-express a Xenopus derived mutant RBPJ in the mouse mammary epithelium131. The mutant RBPJ binds NotchIC but not DNA, blocking Notch receptor signaling and allowing endogenous RBPJ to retain its transcriptional repressor function. Implantation of mutant-RBPJ-expressing mammary epithelium into cleared mouse mammary fat pads resulted in mammary outgrowths that were able to invade less than half of the area of the vector-only controls131. If both of these strategies are presumed to block Notch signaling, the finding with mutant RBPJ contradicts the observation that depletion of RBPJ in the mammary stem cell enriched population increases mammary outgrowth13. However, impregnated mice containing mutant RBPJ transplants showed reduced lobuloalveolar development which agrees with previous findings131.

RBPJ loss in a cancer context has been evaluated in a background of Notch hyperactivation. In this model of Notch-induced mammary tumors driven by constitutively activated Notch4 over-expression, RBPJ deletion was mediated by the whey acidic protein (WAP) mammary specific promoter, which is most active during pregnancy86. Conditional RBPJ deletion in this system rescued normal mammary development and lactation, processes that do not occur with over-expression of active Notch486. However, it was not able to inhibit the primary and secondary mammary tumors that still formed, albeit with a longer latency than tumors resulting from active Notch4 over-expression in RBPJ-containing glands86.

These phenotypes indicate that the effect of altered RBPJ function is context dependent. Inactivation of RBPJ can produce an outcome similar to blocked Notch signaling. However, in other instances the result is similar to what would be expected with Notch activation. NotchIC has been reported to have non-canonical RBPJ independent functions50 that may predominate in the absence of RBPJ. However, RBPJ depletion can also cause gene induction resulting from loss of transcriptional repression123. Furthermore, NotchIC can recruit RBPJ to promoters where it was previously not bound132. Therefore, derepression would only be expected to occur at gene promoters that require RBPJ for their repression. At other target gene promoters, RBPJ may be recruited solely for activation of transcription via NotchIC132. This implies that in certain contexts removal of RBPJ may indeed prevent expression of some Notch regulated genes. On the other hand, gene induction resulting from RBPJ depletion could also contribute to the observed differences, although this occurrence has generally been overlooked in the literature 17

in mammalian systems. The phenomenon of derepression is discussed in further detail in section 1.9.

To our knowledge, no one has evaluated variations in RBPJ levels in human cancers. As RBPJ is a universal transcriptional regulator that converts the Notch signal into changes in gene expression, aberrant expression of this protein would also be expected to have deleterious consequences. Of note, chromosome 4, where the RBPJ gene resides in humans, is frequently lost in breast cancer both as an entire chromosome (53% of breast cancers evaluated)133, or a smaller region encompassing RBPJ (4p15.1–15.3 deleted in 57% of breast cancers evaluated)134. Studies of other human tumor types (bladder, cervical, colorectal, hepatocellular, esophageal, and squamous cell carcinomas of the head and neck and of the skin) have reported allelic loss of one or both arms of chromosome 4, suggesting the presence of multiple tumor suppressor gene loci that are frequently inactivated by deletion134.

1.8 EPIGENETIC REGULATION ASSOCIATED WITH HISTONE MODIFICATIONS

Epigenetic regulation plays an important role in Notch/RBPJ target gene expression, the mechanisms of which are only now being discovered. Alteration of both histone acetyl and methyl marks has been demonstrated at target promoters in response to Notch activation132,135- 137.

The regulated control of the information encoded in DNA is essential to normal cell function, whereas aberrant epigenetic patterns are associated with cancer progression138. Epigenetics, meaning “above genetics”, is the study of mechanisms causing changes in gene expression that that are not present in the primary DNA sequence138. These mechanisms encompass covalent, but reversible modifications on either cytosine bases of DNA, where a cytosine nucleotide precedes a guanine (CpG), or on the histones that package the genome139.

Histones are positively charged proteins and their interaction with DNA is mediated by the negatively charged phosphate group in the DNA backbone. Two of each of the core histones, H2A, H2B, H3 and H4, form a histone octamer around which 147 base pairs of DNA wraps 1.65 times to make a nucleosome, the basic unit of packaged DNA140,141. The accessible amino- terminal tails of the core histones can undergo more than 100 different post-translational modifications including acetylation, methylation, phosphorylation and ubquitination139. These changes modulate the accessibility of genes for transcriptional activation and repression, and 18

are also involved in DNA replication and repair142. Acetylation and methylation of lysines represent the most common and most studied modification of histone tails35.

Histone acetylation is controlled by HATs and HDACs, which typically act as transcriptional co-activators and co-repressors respectively. Acetylation of the lysines on histone tails by HATs abolishes their positive charge, and weakens the association with the DNA backbone. This results in chromatin accessibility to molecules such as transcription factors and RNA polymerase II138, but this mark can also be a signal for protein recruitment139. Conversely, removal of acetyl groups by HDACs causes compaction of chromatin and thus repression of gene expression143.

Methyltransferases are some of the most specific enzymes that modify histones, acting on a single lysine present on a single histone, which can be mono-, di- or trimethylated144. Conversely, demethylases remove the methyl mark and are also selective for the methylation state141. Methylation can either activate or repress transcription139,141.

The outcome of specific histone modifications depends on many factors, including the location of the modification relative to the transcription start site (TSS) and the presence of other epigenetic marks. While not a rigid rule, active promoters are generally associated with acetylation of residues of histone H3 and H4 (although unacetylated regions do not denote inactive genes) and trimethylation of histone H3 at lysine 4145.

Modification-specific histone antibodies can be used for chromatin immunoprecipitation (ChIP) followed by genomic microarrays (ChIP-chip) or high throughput sequencing (ChIP-seq) to examine the distribution of the modification throughout the genome138.

1.9 RBPJ NUCLEAR CYCLE

RBPJ was initially identified as a transcriptional repressor in human cells112. However it was found capable of inducing gene expression in invertebrate model systems146. Its role in gene regulation was in fact shown to be dependent on the interacting proteins35,36. RBPJ binds DNA as a monomer via two of its domains (N-terminal Rel homology domain and central beta-trefoil domain)53, providing the first building block of complex assembly for recruitment of either co- activators or co-repressors. A schematic of the RBPJ nuclear cycle is shown in Figure 1.5 and discussed in detail below.

Figure 1.5 Schematic of the RBPJ nuclear cycle. Simplified complexes are shown and do not include all reported interactions. Co-repressor assembly on RBPJ is mediated by SHARP, resulting in silencing of transcription. SMRT also binds NcoR (which makes direct contacts with RBPJ) and Sin3A which are not shown in the schematic36. Co-repressors recruit HDACs, such as HDAC1, but histone deacetylases from other classes have also been reported. NotchIC directly competes for binding on RBPJ, and physically displaces co-repressors. MAML recognizes a composite surface of NotchIC and RBPJ and recruits HATs such as p300 and PCAF/GCN5. SKIP binds both RBPJ and NotchIC directly147 and may also be part of the co-repressor complex. Phosphorylation of NotchIC targets it for ubiquitination and proteasomal degradation, dissociating the co-activator complex and initiating a new cycle of transcriptional repression. Pathway image created using proteinlounge freeware (www.proteinlounge.com).

In the absence of Notch activation, RBPJ is a default transcriptional repressor and enters the nucleus precommitted to this function118. Over-expression of wild-type RBPJ inhibits target gene expression148. Transcriptional repression of the RBPJ co-repressor complex depends on HDAC activity135,136,149, and involves proteins such as HDAC1, HDAC3, HDAC4 and RPD347, resulting in condensation of chromatin and silencing of gene expression. The complete set of HDACs involved in repression is not known, however HDAC1 is the most studied, and it co- immunoprecipitates with RBPJ135.

A number of proteins have been found to be associated with co-repression. CIR (CBF1 interacting co-repressor) is a direct RBPJ-binding partner that interacts with HDAC complexes136. Like CIR, SMRT (silencing mediator of retinoid and thyroid receptors) binds RBPJ directly135 and shares the same binding site on RBPJ used by NotchIC118. SMRT also directly associates with the related protein NcoR (nuclear receptor co-repressor), another RBPJ binding partner135 and Sin3A as well as HDAC1150. These two proteins are essential for repressor complex function, as RBPJ-mediated transcriptional repression is lost when interactions with SMRT and CIR are disrupted118. SHARP (SMRT and HDAC associated repressor protein) facilitates co-repressor assembly by providing a crucial protein interaction platform35. Indeed, SHARP is a ubiquitously-expressed protein, loss of which causes embryonic lethality in mice151. SHARP has been proposed as a functional homologue of the fly Hairless protein which is key to Su(H) (the Drosophila homologue of RBPJ) mediated co-repression35. Hairless acts to link co- repressor proteins to Su(H) to inhibit transcription of target genes152,153. SHARP interacts directly with SMRT/NCoR and RBPJ, also utilizing the NotchIC binding site149. SHARP can recruit a number of other repressive proteins including CtIP (CtBP interacting protein)154. These studies have demonstrated that RBPJ cannot function as a transcriptional repressor by itself, but requires association with other co-repressors to inhibit target gene expression.

RBPJ has been shown to bind to transcription factor IIA (TFIIA), a protein of the general transcriptional machinery required by RNA Polymerase II to initiate transcription. This interaction competitively inhibits binding of TFIIA to TFIID and prevents transcriptional activation155. Two FLH1 (four and a half LIM domains 1) isoforms, KyoT2 and KyoT3, also interact with RBPJ and repress transactivation of RBPJ by NotchIC156,157.

Recently, in addition to HDACs, histone demethylases have also been shown to participate in the co-repressor complex by interacting directly with RBPJ. These include KDM1A (or LDS1, a lysine (K)-specific demethylase 1A)158 and KDM5A, which is essential for RBPJ target gene silencing137. 21

Conflicting evidence exists as to whether SKIP (Ski-interacting protein) is present in the co- activator complex only159, where it may assist with splicing160, or is also a part of the co- repressor complex, serving as an adaptor protein118,147. Co-repressor function is supported by findings that SKIP interacts with the co-repressors CIR and SMART, and SKIP also binds RBPJ directly promoting its nuclear localization and retention118.

Interestingly displacement of RBPJ has been reported to lead to alleviated transcriptional repression86,132,161. The following are instances where removal of RBPJ results in target gene induction. In the fly, RBPJ loss-of-function can partially rescue wing defects in γ-secretase mutants152. This γ-secretase mutation results in blocked Notch cleavage, thereby inhibiting pathway activation. Partial restoration of the normal phenotype is due to the derepression of target genes, resulting from RBPJ loss-of-function, that allow the wing development to occur152. Moreover, knock-down (KD) of RBPJ in a Drosophila system derepresses target genes in the E(spl) (Enhancer of split) complex which are related to the mammalian HES and HEY gene families132. This derepression was shown to depend on prior RBPJ promoter occupancy132. Depletion of RBPJ in mammalian cells can also cause gene derepression, although this occurrence is not well established and is generally disregarded. RBPJ is a repressor of human immunodeficiency virus (HIV) transcription. KD of RBPJ is accompanied by activation of latent proviruses and an increase in both RNA polymerase II and acetylated H3, but a decrease of HDAC1 occupancy at the HIV promoter161. Conversely, over-expression of RBPJ in this system re-establishes repressive chromatin structure. This corresponds to reduced RNA polymerase II, and decreased acetylation of histone H3 and H4. However, HDAC1 and co-repressors CIR and Sin3A are increased at RBPJ binding sites in the long terminal repeat (LTR) promoter of HIV161. By evaluating the effect of depleted and over-expressed RBPJ at the HIV-LTR this study shows a direct epigenetic link with loss and re-establishment of transcriptional repression respectively. In addition, RBPJ depletion by short hairpin RNA (shRNA) has also been noted to up-regulate several unspecified genes in HC11 mouse mammary epithelial cells86.

Inhibiting co-repressor function can also cause target gene derepression. Recently, a number of co-repressors have been shown to play an essential role in gene repression in addition to those in the classical complex discussed above. KD of either RBPJ or KDM5A, a histone demethylase directly associated with RBPJ at repressed promoters, caused derepression of the target gene Deltex1 in mouse T cells137. Furthermore, depletion of RBPJ and KDM5A in a model of Notch induced Drosophila eye tumors resulted in increased tumor growth and metastasis137. Another Drosophila study focused on Asf1, an H3/H4 histone 22

chaperone involved in nucleosome assembly and chromatin remodeling162. Asf1 binds RBPJ and directly interacts with other co-repressors contributing to inhibited transcription; KD of Asf1 resulted in derepression of E(spl) target genes162. An RBPJ-interacting protein, MTG16 (myeloid translocation gene 16), which is involved in chromosomal translocations in acute myeloid leukemia, has been shown to associate with co-repressors and histone deacetylases163. However, MTG16 also binds NotchIC, which likely removes MTG16 from RBPJ facilitating transition to transcriptional activation. Hematopoietic progenitors purified from the bone marrow of MTG16 KO mice show induction of Notch/RBPJ target genes (Hes1, Nrarp, Notch1)163.

Gene translocations that disrupt RBPJ co-repressor function also lead to aberrant RBPJ- mediated transcriptional activity independently of Notch in acute myeloid leukemia. Mutation of OTT (a co-repressor related to SHARP) by generation of the OTT-MAL fusion oncogene activates RBPJ and induces leukemia in a mouse model164. Moreover, ETO, another protein implicated in the RBPJ co-repressor complex, can be part of a fusion event with AML1, resulting in induction of Notch responsive promoters165.

Based on work in Drosophila it has been suggested that alleviation of transcriptional repression is adequate to indirectly cause activation of Notch target genes123. This permissive signaling model implies that at some promoters displacement of co-repressors from RBPJ by NotchIC is sufficient to activate gene expression. The gene derepression observed with disrupted co-repressor function supports this model, although the phenomenon is not well documented in the mammalian system. Alternately, at other promoters, NotchIC may be necessary for both removal of co-repressors and transcriptional activation via RBPJ123,166.

Classically, RBPJ was considered constitutively bound to DNA, with the exchange of co- activators and co-repressors resulting in a cycle of gene activation and gene repression47. Recent work has shown that the protein-DNA interactions are dynamic, with the co-activator complex forming a more stable association with DNA, and NotchIC recruiting RBPJ to promoters where it previously was not bound167. In the latter case, and in the absence of Notch activity, RBPJ is likely not essential for the transcriptional repression of these genes. However, the most highly-expressed genes upon Notch activation are those with a pre-existing RBPJ promoter occupancy167.

As NotchIC interacts with the same region on RBPJ required for co-repressor binding a physical displacement of inhibitory proteins must take place for activation118,168. NotchIC itself contains a transactivation domain35 but it also interacts with the co-activator MAML, which

recognizes a composite surface of RBPJ/NotchIC119. The ternary complex recruits a HAT169, p300, and/or another co-activator with histone acetylase activity, PCAF/GCN5, to facilitate target gene expression170,171. Indeed, histone acetylation is a required mechanism for transcriptional activation of Notch/RBPJ responsive promoters: a dominant-negative MAML that cannot recruit p300 inhibits Notch-mediated gene activation172. Notch signaling is accompanied by both an increase in acetylated H4 (acH4) and a loss of nucleosomes at active genes132. Histone H3 trimethylated lysine 4 (H3K4me3) marks have also been observed at target genes following Notch activation132. As gene activation needs to be carefully controlled, NotchIC seldom accumulates in the nuclei of normal cells as it is rapidly degraded97. MAML plays a role in this process recruiting the nuclear kinase Cyclin-C/CDK8159 which phosphorylates NotchIC targeting it for subsequent ubiquitination and proteasomal processing48.

Notch receptor signaling has been reported to occur independently of RBPJ. Of note, Notch can activate a signaling pathway that requires the cytoplasmic protein Deltex50. Deltex interacts with the ankyrin repeats of Notch50 and can positively or negatively regulate Notch receptor signaling173. Increased Deltex expression has been shown to deplete Notch from the cell surface and promote its endocytosis and trafficking though endosomal compartments173, resulting in activation of RBPJ-independent gene expression174.

Notch-independent RBPJ signaling has also been reported. A number of viruses have been shown to utilize RBPJ for their own propagation including Epstein-Barr virus and Kaposi's sarcoma-associated herpesvirus161,175,176. Viral proteins behave like NotchIC, binding to the same beta-trefoil domain on RBPJ and displacing co-repressors to induce gene activation of RBPJ-responsive cellular or viral promoters110,118,177. Indeed there is partial overlap between the target genes of NotchIC and viral cellular targets113. This implies that viruses co-opt RBPJ to activate a transcriptional program and produce cellular outcomes similar to Notch signaling. However, RBPJ can also hinder viral infection and functions as a key transcriptional repressor during establishment of HIV latency161,175.

Other proteins may compete with NotchIC for binding to RBPJ, such as PTF1a a pancreatic and neural-restricted bHLH factor178. RBPJ participates in the PTF1-J complex during the initial stages of pancreatic and neural development178. This complex is composed of three proteins PTF1a, an E-box partner, and RBPJ which act together to induce RBPJL (RBPJ-like) as development continues. RBPJL is a pancreas-restricted paralog of RBPJ and does not interact with NotchIC179. Once induced by the PTF1-J complex RBPJL replaces RBPJ in the complex, (which is now called PTF1-L complex) and acts to regulate transcription of pancreatic digestive 24

enzymes in the adult178,180. The RBPJ-containing PTF1-J complex is also important in cell specification in the nervous system independently of RPBJL181.

Interestingly, fungal species lack Notch but do have RBPJ genes, indicating that RBPJ functions independent of Notch in this system. The RBPJ proteins in yeast recognize a canonical RBPJ binding motif and play roles in regulating genes involved in adhesion and division182. It has been proposed that transcriptional repression was the original and sole function of RBPJ, as yeast RBPJ orthologs are speculated to recruit an HDAC complex36. In animals, RBPJ function as both a transcriptional repressor and activator allows immediate shut off of Notch signaling before and after pathway activation, sharpening the cellular sensitivity to the pathway36.

1.10 NOTCH/RBPJ TARGET GENES AND THEIR FUNCTION

It is becoming evident that the Notch pathway can induce expression of a large number of genes183, although until recently very few of these genes were validated as direct Notch/RBPJ target genes35,47. Table 1.2 lists published targets to date that have been identified as being directly regulated by RBPJ in human and mouse cells. Most target genes contain one or several RBPJ binding sites within the first few hundred base pairs of the TSS; however, there are some exceptions with targets having functional RBPJ recognition sequences thousands of base pairs upstream184,185. The induction of these genes depends on the signal strength, the composition of RBPJ recognition motifs and context.

The most extensively studied and best understood direct Notch/RBPJ targets are from the HES (hairy/enhance of split) and HEY (HES related with a YRPW motif) genes families55,186 of closely related and conserved bHLH transcription factors, namely HES1, HES5 and HES7 and HEY1, HEY2 and HEYL52,55,187. These proteins act as transcriptional repressors and are characterized by direct binding to DNA as dimers (recognizing an E-box or N-box sequence for HES and an E-box sequence for HEY)52,54,55. HEY proteins differ from HES proteins in possessing a glycine instead of a proline in their bHLH domain, a YRPW (or YXXW) C-terminal tetrapeptide motif instead of a WRPW motif, and an additional adjacent peptide with unknown function, which is absent in HES proteins54,55. When bound to DNA, HES proteins repress target gene transcription by recruiting the co-repressor Groucho/TLE via their WRPW motif, which in turn mediates the recruitment of HDACs. On the other hand HEY proteins bind NcoR and Sin3A

Table 1.2 Target genes where direct RBPJ binding has been identified by ChIP, EMSA or luciferase reporter assays.

Target gene Description Reference

ACTA2 alpha 2 smooth muscle actin (Noseda et al. 2006)188

BIRC5 baculoviral IAP repeat-containing 5 (Survivin) (Lee et al. 2008)89

CCND1 cyclin D1 (Ronchini and Capobianco 2001)189

CCND3 cyclin D3 (Joshi et al. 2009)190

CDKN1A cyclin-dependent kinase inhibitor 1A (p21, Cip1) (Rangarajan et al. 2001)191

DTX1 deltex 1 (Deftos et al. 1998)192, (Chadwick et al. 2009)193, (Liefke et al. 2010)137

EFNB2 ephrin B2 (Iso et al. 2006)194,(Grego-Bessa et al. 2007)195

ERBB2 avian erythroblastosis oncogene B2 (Chen et al. 1997)196 fms-related tyrosine kinase 4 (vascular endothelial FLT4 growth factor receptor 3) (Shawber et al. 2007)197

FOXP3 forkhead box P3 (Ou-Yang et al. 2009)198

GATA3 GATA binding protein 3 (Fang et al. 2007)199,(Amsen et al. 2007)200

GFAP glial fibrillary acidic protein (Ge et al. 2002)201 (Jarriault et al. 1995)202, (Maier and Gessler 2000)186, (Fryer et al. HES1 hairy and enhancer of split 1 2004)159

HES5 hairy and enhancer of split 5 (Nishimura et al. 1998)203

HES7 hairy and enhancer of split 7 (Bessho et al. 2001)204 (Maier and Gessler 2000)186, (Nakagawa at al. 2000)52, (Iso et al. HEY1 HES related with YRPW motif 1 2001)187 Maier and Gessler 2000)186, (Nakagawa at al. 2000)52, (Iso et al. HEY2 HES related with YRPW motif 2 2001)187, (Gustafsson et al. 2005)205, (Bouras et al. 2008)13

HEYL HES related with YRPW motif-like Maier and Gessler 2000)186, (Nakagawa at al. 2000)52

IL6 interleukin 6 (Kannabrian et al. 1997)206, (Palmieri et al. 1999)207 v-Myc avian myelocytomatosis viral oncogene MYC homolog (Weng et al. 2006)208, (Klinakis et al. 2006)209, (Palomero et al. 2006)210 nuclear factor of kappa light polypeptide gene NFKB2 enhancer in B-cells 2 (p49/p100) (Oswald et al. 1998)211, (Vilimas et al. 2007)212 nuclear factor of kappa light polypeptide gene NFKBIA enhancer in B-cells inhibitor, alpha (IKBA) (Oakley et al. 2003)213

NODAL mouse nodal homolog (Krebs et al. 2003)214

NRARP NOTCH-regulated ankyrin repeat protein (Pirot et al. 2004)184

PDGFRB platelet-derived growth factor receptor, beta (Jin et al. 2008)215

PTCRA pre T-cell antigen receptor alpha (Reizis et al. 2002)185

RELB v-rel reticuloendotheliosis viral oncogene homolog B (Vilimas et al. 2007)212

SKP2 S-phase kinase-associated protein 2 (p45) (Sarmento et al. 2005)216

SNAI1 Drosophila snail homolog 1 (Sahlgren et al. 2008)217

SNAI2 Drosophila snail homolog 2 (Leong et al. 2007)88,(Niessen et al. 2008)218

TNC tenascin C (Sivasankaran et al. 2009)219

directly via their bHLH domain and associate indirectly with HDAC155. Sirtuins, another type of histone deacetylase, may also be involved in transcriptional repression by HES1 and HEY2220.

Interestingly, rather than direct DNA binding, a more common mechanism of repression (especially by HEY proteins) is through protein-protein interactions that convert transcriptional activators into repressors54. HEY proteins can homodimerize with each other, but also heterodimerize with HES1, which results in a more stable interaction causing synergistic transcriptional repression and possible Notch signal amplification52,54. However, via this mechanism HES and HEY proteins can also inhibit their own expression resulting in a negative feedback loop, producing an oscillatory gene expression signal52,55. Therefore, in addition to direct DNA binding, repression can be indirectly mediated by formation of HES-HEY heterodimers. In addition to forming homo- or heterodimers with each other, HES and HEY proteins can also bind and sequester other nuclear proteins54,55. Interestingly both HES and HEY proteins can directly bind RBPJ. HES1 or HEY2 association with RBPJ does not disrupt the ability of RBPJ to bind DNA or even interact with NotchIC, but enhances its transcriptional repressor function inhibiting activation of HES and HEY expression221. Therefore, in addition to feeding back on their own transcription, HES and HEY proteins may play a role in dampening the Notch signal via RBPJ. Similarly, promoters with binding motifs for both RBPJ and HES proteins could be positively and negatively regulated by each, respectively, as has been reported for the FOXP3 promoter198.

The Notch cascade is well-known for utilizing feedback mechanisms. Signal activation results in expression of Notch receptors but down-regulation of ligands, which specifies the signal receiving cell52. Alternately, Notch signaling activity can lead to up-regulation of ligand mRNA expression, especially under conditions such as hypoxia, as has been observed for Delta-like4222. It is likely that Notch receptors directly activate their own transcription, although the literature thus far has only provided evidence of up-regulated expression rather than direct RBPJ binding. Notch1 and Notch3 are up-regulated by Notch signaling193,208 and Notch1 has also been reported to induce Notch4 expression57,90. Furthermore, Notch signaling can result in induction of not only a gene but also its negative regulator (a phenomenon termed Incoherent Network Logic) such that the ultimate outcome is context dependent167.

Target genes of the HES and HEY transcriptional repressors remain largely elusive, although HES and HEY act to maintain an undifferentiated phenotype and thus are implicated in inhibiting genes that cause differentiation52,205. Indeed, higher expression of HEY1 and HEY2 is seen in luminal progenitors when compared with mature luminal cells of the breast13. 27

KO mouse models have provided insight into HES and HEY functions and their indispensible roles during development. HES proteins are essential in the nervous system, T lymphocyte development and formation of muscle tissue54. Hes5 null mice are viable and fertile with central nervous system abnormalities54. However, loss of Hes7 results in defective somitogenesis223 and Hes1 KO mice die during gestation or just after birth due to severe abnormalities in neurogenesis224. HEY proteins are best known for their critical role in the cardiovascular system225. All three HEY genes can regulate endothelial to mesenchymal transition in the developing heart226 and also play a role in maintaining neural precursor cells227. Hey1 or HeyL KO mice are viable and fertile226,228; however, inactivation of Hey2 leads to cardiac defects, resulting in postnatal lethality229. The combined loss of both Hey1 and Hey2228, or Hey1 and HeyL226, leads to embryonic or neonatal death, resulting from vascular and cardiac abnormalities.

Notch signaling via RBPJ regulates a diverse set of additional target genes. A number of these could play a role in critical tumor hallmarks. Some examples using the genes from the direct targets list in Table 1.2 are provided below. Induction of genes such as Survivin (BIRC5), a member of the Inhibitor of Apoptosis family, could promote inappropriate cell survival. Autonomous growth stimulatory signals may be mediated by the oncogenic HER2 (ERBB2) growth factor receptor that is amplified in 30% of breast cancers230. Increased proliferation could be conferred by direct targets such as CyclinD1 and CyclinD3 (CCND1 and CCND3), which mediate progression through the cell cycle. Similarly, the S-phase kinase-associated protein 2 (SKP2) plays an oncogenic role in breast cancers by inducing degradation of p27, a cyclin dependent kinase inhibitor, thereby promoting entry into S-phase of the cell cycle231,232. Alternately, Myc, another well characterized proto-oncogene and Notch target, controls expression of genes involved in cell growth, but is also implicated in deregulated tumor metabolism233,234. Indeed, Notch1 and Myc proteins are co-expressed in a significant proportion of human breast cancers235. The importance of Notch signaling in angiogenesis is well established236, and proteins involved in blood vessel formation such as Ephrin B2 (EFNB2) and platelet derived growth factor beta (PDGFRβ) could facilitate tumor progression. Tenascin-C (TNC) is also implicated in angiogenesis as well as tumor cell invasion in different malignancies including breast cancer237. Additionally, data from our lab shows that Slug (SNAI2), a protein important in epithelial-to-mesenchymal transition, plays a key role in Notch-induced metastasis88. PDGFRβ expression in epithelial cells has also been linked to invasive human mammary carcinomas and epithelial-to-mesenchymal transition238. Other direct target genes may be inhibitory to tumor progression. GATA binding protein 3 (GATA3) promotes tumor cell 28

differentiation and plays an important role in luminal cell commitment and differentiation in the normal breast tissue239. Forkhead box P3 (FOXP3) is a putative tumor suppressor in breast cancer and directly represses the HER2240 and SKP2241 promoter. However, positive FOXP3 staining has also been associated with worse overall survival of breast cancer patients242. Interleukin 6 (IL6) can have tumor promoting or inhibitory effects in the context of breast cancer243. These target genes highlight how Notch pathway activation controls diverse cell fates.

1.11 NOTCH SIGNALING CROSS-TALK WITH OTHER PATHWAYS

Several signaling cascades in addition to those already discussed have been shown to cross-talk with the Notch pathway. A brief overview of other potentially interacting pathways is discussed, but is not inclusive of all pathways reported to cross-talk with Notch signaling.

Notch pathway components are up-regulated in response to Wnt signaling and cooperate in the transformation of human breast epithelial cells244. However, the two pathways play opposing roles in mouse mammary development with Wnt signaling promoting branching of epithelial ducts and Notch activity blocking morphological differentiation245.

Ras and Notch1 have been reported to act synergistically in breast tumorigenesis94. Ras increases the half-life of NotchIC in the nucleus and positively correlates with Notch in human breast cancers57. Phospho-Erk1/2, a downstream target of the Ras-MAPK pathway, is co- expressed with cleaved Notch1 and/or Hes5 in human breast cancers94.

Notch and the Phosphoinositide 3-kinase-AKT1 (PI3K-AKT) signaling pathway have been shown to synergize in tumor models246,247. In breast epithelial cells, AKT activation by Notch confers protection from apoptosis248. An antagonist of the pathway, phosphatase and tensin homolog (PTEN), is inhibited by Notch target gene HES1, allowing activation of the PI3K-AKT pathway246. Notch activation of AKT signaling, mediated by release of an autocrine factor, confers resistance to apoptosis in the breast epithelial cell line MCF10A248. Active AKT signaling can positively feedback by upregulating Notch1 transcription249.

Notch and the transforming growth factor beta (TGFβ) signaling pathways can be antagonistic or synergistic in the induction of Notch responsive genes250. Notch can also regulate TGFβ signaling activity by controlling the expression or activity of receptor activated SMADs, the key transcriptional mediators of the TGFβ pathway, and NotchIC has been shown to physically recruit SMAD3 to Notch target promoters251. 29

Notch also has an inhibitory or synergistic relationship with the NF-κB pathway, with each being able to induce components of the other pathway252. Notch receptors can physically interact with NF-κB signaling proteins252. Furthermore, an estimated 12.5% of NF-κB binding sites are thought to overlap RBPJ sites253 and NF-κB has been suggested as a major mediator of Notch1-induced transformation in leukemia212.

Recently, a direct relationship has been shown between hypoxia and Notch signaling: NotchIC and hypoxia inducible factor 1 alpha (HIF1α) physically interact, and this direct binding stabilizes NotchIC and leads to recruitment of HIF1α to Notch-responsive promoters for transcriptional activation of Notch target genes by RBPJ205,217. These targets include HES and HEY genes as well as SNAI1205,217. Hypoxia conditions results in up-regulation of Delta-like1, and Notch is essential in hypoxia-induced epithelial-to-mesenchymal transition and invasiveness217.

Notch signaling also positively regulates the JAK-STAT pathway (Janus Kinase Signal Transducers and Activators of Transcription) in the developing nervous system, mediated by protein-protein interactions with HES1 and STAT3254. Moreover, activation of STAT3 was observed in mammary hyperproliferative structures induced by NotchIC255. STAT3 plays an important role as a death factor in normal mammary development, initiating mammary gland involution upon cessation of lactation256. However in breast tumors, STAT3 is often constitutively activated and associated with lymph node metastasis257. Here STAT3 has been shown to up-regulate apoptosis inhibitors and other factors that contribute to cancer progression257,258.

This extensive potential for cross-talk highlights the importance of context in cellular outcomes resulting from Notch signal activation.

1.12 AIM OF PRESENTED STUDY

The volumes of reports on Notch signaling over recent years have greatly contributed to our knowledge of the pathway and the important roles it plays in development and disease. However, from a cancer perspective, it is still not completely clear how the Notch cascade is deregulated as the observed Notch signal activity in solid tumors cannot be explained by Notch receptor mutations alone.

The expression of full length or constitutively active Notch receptors, ligands, and target genes (primarily HES and HEY) has been evaluated in several different malignancies. However, RBPJ has not received much focus in this context, despite its key role in mediating Notch signaling and providing a crucial inhibitory role in the absence of Notch activation. The importance of RBPJ is supported by the observation that it is both ubiquitously and constitutively expressed in normal tissues, yet loss of the chromosomal region where the RBPJ gene resides frequently occurs in breast cancers and other cancer types134.

Not all Notch target genes require RBPJ for their repression. RBPJ can be recruited by NotchIC to its DNA binding motif in some gene promoters where it only contributes to activation132. Conversely, Notch transactivation may not be required for all RBPJ/Notch target genes, a phenomenon termed permissive signaling123. In this scenario, the only role of NotchIC is to displace co-repressors from RBPJ while subsequent activation is NotchIC independent123. If alleviation of repression is sufficient for gene induction then loss of RBPJ or co-repressors should confer the same effect, resulting in transcriptional activation.

Co-repressor proteins that facilitate RBPJ repressive function continue to be identified, and loss of this interaction has been reported to cause aberrant expression of target genes, in some cases propagating neoplastic phenotypes in Drosophila137. Indeed, removal of RBPJ itself, from genes with prior RBPJ occupancy can cause target gene derepression132. This derepression that occurs with RBPJ loss is sufficient to partially rescue a fly developmental phenotype cause by inhibited Notch signaling152. However, most of these observations have been made in the Drosophila system and have not been the primary focus of investigation in mammalian cells.

A tissue-specific protein, Ikaros, plays an important role in attenuating Notch signal in the lymphoid system. Ikaros is a transcriptional repressor that does not physically interact with NotchIC, but recognizes the canonical RBPJ consensus motif and directly competes for DNA binding with RBPJ to prevent gene activation though Notch/RBPJ4. Ikaros is essential for lymphoid development, but loss of Ikaros function has been implicated in human leukemogenesis where Ikaros functions as a tumor suppressor4. Indeed, loss of Ikaros leads to derepression of Ikaros-regulated Notch target genes and subsequent development of leukemia occurs even in the absence of NotchIC259. Since Ikaros is a lymphoid-restricted protein, the transcriptional repression of genes in other cell types, such as the epithelium, may rely on RBPJ alone.

This work initially arose from an experiment intended to evaluate the outcome of blocked Notch signaling using alternate approaches to inhibit the pathway. Knock-down of RBPJ has been used numerous times by others for this purpose13,86. The rational is that removal of RBPJ, the common denominator through which all Notch receptors relay signal in the nucleus, would prevent induction of Notch target genes. To our surprise however, depletion of RBPJ resulted in increased rather than slower tumor growth, when compared to Notch receptor inhibition strategies which prevented tumor growth. The larger RBPJ KD tumors also showed derepression of direct target genes of the HEY family. Our data imply that RBPJ depletion does not prevent the downstream gene induction as is generally thought and can cause a phenotype similar to Notch activation. A further survey of the literature which is highlighted in the introduction revealed that, although poorly understood and mostly studied in Drosophila, perhaps this was not such an unexpected outcome and prompted us to address the effect of RBPJ loss in a mammalian system.

Disruption of the fine balance between gene activation and repression can lead to cell transformation. Just as Notch signaling can be oncogenic due to aberrant induction of target genes, we propose that RBPJ loss may also represent an oncogenic event, resulting in a similar outcome from derepression of target genes. As Notch signaling is already implicated in breast cancer, we used this as a primary disease model and hypothesized that depletion of RBPJ in cancer cells would lead to derepression of a subset of RBPJ target promoters and increased oncogenicity. This thesis, therefore, aims to define a role for RBPJ as a novel tumor suppressor, and to evaluate the mechanism by which RBPJ loss promotes tumor growth.

To evaluate the relevance of RBPJ loss in tumor promotion, we pursued three related objectives that are presented in separate chapters of this thesis. First we looked for RBPJ expression changes in human cancers to determine whether RBPJ loss is potentially relevant to human disease. To test the oncogenic potential of RBPJ removal we modeled RBPJ loss in human-to-mouse xenograft studies and measured differences in tumor growth.

Our second aim was to evaluate how RBPJ deficiency affects target gene expression and whether gene derepression, which has been reported to occur in the fly model system, is detected in a human cancer cell context. Identifying potential aberrant gene induction would be informative with respect to what downstream effectors were mediating the effect of RBPJ loss. To determine if Notch receptor signaling functions independently of RBPJ in target gene up- regulation, we also assessed the contribution of NotchIC in transcriptional activation in the absence of RBPJ. 32

Finally, to evaluate the phenotypic outcome of RBPJ loss we looked at epigenetic changes taking place at promoters in RPBJ deficient cells, to facilitate identification of other potentially induced genes, in addition to canonical targets. This global analysis led to discovery of cellular processes that play a role in increased tumor growth in the absence of RBPJ and have contributed to our understanding of how RBPJ may oppose oncogenicity.

This thesis proposes a previously unidentified tumor suppressor function of RBPJ in the context of cancer, and provides an alternate mechanism of how the Notch signaling pathway can be deregulated. As a Notch-like signal is generated in the absence of RBPJ, this work cautions against using RBPJ knock-down as a means to block the Notch cascade as this could lead to misleading conclusions about RBPJ-independent Notch signaling. We also raise important implications for therapeutic approaches in the treatment of malignancies with aberrant Notch signaling.

CHAPTER 2: MATERIALS AND METHODS

2.1 CELL CULTURE

The human DG75 Burkitt’s lymphoma parental and RBPJ KO cell lines were provided by the Kempkes laboratory113. The HPB-ALL leukemia T-cell line was a gift from A.P. Weng (BC Cancer Research Centre). These cells were grown in RPMI 1640 medium (Sigma-Aldrich, St. Louis, MO) supplemented with 10% heat inactivated fetal bovine serum (HyClone, Logan, UT), 2 mM glutamine (Invitrogen, Carsbad, CA) and 100 U of each penicillin and streptomycin. MDA- MB-231 human breast carcinoma cells (a gift from C.D. Roskelley, University of British Columbia), mouse endothelial cell line SVEC4-10, lentiviral producer line 293T and a derivative for retroviral production (AmphoPhoenix, obtained from Gary Nolan, Stanford University, Palo Alto, CA) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Sigma) and supplemented as above. For RBPJ transcript and protein expression analysis, breast cancer cell lines were obtained from Wan Lam (BC Cancer Research Centre) and grown as described in Neve et al. 2006260, or if not specified here, in RPMI 1640 supplemented as above. These cells lines are as follows: BT-474, Sk-Br-3, HCC1187, HCC1143, HCC1937, HCC2218, HCC38, MCF-7, HCC1599, HCC1395. The normal breast epithelial cell line MCF10A was cultured in a 1:1 mixture of DMEM/F12 (Sigma) supplemented with 5% horse serum (Sigma), 2 mM glutamine (Sigma), 20 ng/ml of epidermal growth factor (Sigma), 100 ng/ml cholera toxin (Cedarlane, Burlington, ON), 10 µg/ml insulin (Sigma), 500 ng/ml hydrocortisone (Sigma), and 100 U/ml each of penicillin and streptomycin (Invitrogen). All of these cells were maintained at

37°C in a 5% CO2 atmosphere. Mouse Rbpj null (OT11) and wild-type (OT13) embryonic fibroblast cell lines were provided by T. Honjo (Kyoto University, Kyoto)261 and were grown in DMEM supplemented with 10% FBS and 100 units/ml of mouse IFN-γ (Stem Cell, Vancouver, BC) at 33°C (conditions which allow immortalization by the temperature sensitive simian virus 40 large tumor antigen).

2.2 RNA INTERFERENCE AND GENE TRANSFER shRNA-encoding DNA oligonucleotides (Integrated DNA Technologies, Coralville, IA) were cloned into the HpaI and XhoI site of the pLentilox3.7 green fluorescent protein (GFP) lentiviral vector (gift from L. Van Parijs, Massachusetts Institute of Technology, Cambridge, MA)262. The targeting sequences are as follows: shRandom 5’-GTTGCTTGCCACGTCCTAGAT-3’, shScrambled 5’GATTAGAACCCTCACGGTACG-3’, human shRBPJ sense (NM_015874) 5’- GAGTCTCAACCGTGTGCAT-3’, human shRBPJ-set2 sense 5’- GCATGGCACTCCCAAGATTGA-3’, mouse shRBPJm (NM_001080927) sense 5’-

GGACGTAGCAATGCTTGAACT-3’ (see Figure 2.1 for a schematic of where the hairpins target the RBPJ transcript), and human shHEY2 (NM_012259) sense 5’- GGATCATTTGAAGATGCTTCA-3’. These constructs were generated by Kyle Niessen218 or designed using BLOCK-iT RNAi Designer (Invitrogen). For over-expression the following retroviral vectors all encoding human genes were used: pLNCX (gift from A. D. Miller, Fred Hutchinson Cancer Research Centre, Seattle, WA), pLNC-FlagRBP2N, pLNC- FlagRBP2NR179H (the mutated RBPJ)263, MSCV-IRES-yellow fluorescent protein (YFP, MIY), pMIY-Jagged1, pMIY-Notch1IC33, MSCV-IRES-GFP (MIG), and MIG-HEY2. A total of 15 µg plasmid DNA was transfected using 45 µl TransIT-siQUEST transfection reagent (Mirus Bio Corporation, Madison, WI). Infectious lentiviruses were produced in 293T cells by co- transfection of 6 µg pLentilox shRNA vector, and 3 µg each of pVSVG, RSV-REV, and pMDL g/p RRE. Infectious retroviruses were produced in AmphoPhoenix packaging cells by transfection of 15 µg of pLNCX, pMIY or pMIG retroviral vectors. Viral supernatants produced in 293T or AmphoPhoenix cells were passed though a 0.45 µm filter, and used to transduce target cells in the presence of 8 µg/ml Polybrene (Sigma). A total of two rounds of infection was performed for the lentiviral system every 24 hrs. For retroviral constructs, infection was done four times approximately every 12 hours. Target cells were allowed to recover for at least 24 hrs and were purified based on GFP or YFP expression by flow sorting using a FACS-440 flow- sorter (Becton Dickinson Inc, Franklin Lakes, NJ). Cell lines expressing LNCX constructs were selected with 300 µg/ml G418 (Invitrogen). Knock-down and over-expresson were accessed by real-time quantitative polymerase chain reaction (qPCR) and immunoblotting.

RBPJ 1 2 3 4 5 6 7 8 9 10 11

shRBPJ shRBPJ-set2 shRBPJm

Figure 2.1 Exon position where hairpin constructs target the RBPJ transcript. shRBPJ and shRBPJ-set2 were used to deplete human RBPJ and shRBPJm was used to knock-down the mouse transcript. These short hairpin RNA sequences are predicted to target all RBPJ isoforms.

2.3 RNA ISOLATION AND qPCR

Total RNA was extracted using TRIzol (Invitrogen). cDNA was synthesized from 2.5 µg DNaseI treated total RNA with Superscript II Reverse Transcriptase and random hexamer primers (all from Invitrogen) in a 50 µl reaction volume. qPCR was performed using the ABI Prism 7900HT 36

detection system (Applied Biosystems, Foster City, CA) and the SYBR Green PCR Master Mix kit (Applied Biosystems), with gene-specific primer pairs in a 15 µl final volume. For this reaction, 2.5 µl of the 1:2.5-times diluted cDNA sample was used (equivalent to 20 ng total RNA). Primer sets were designed with the PrimerQuest software (Integrated DNA Technologies) and validated to ensure amplification of a single product. Cycling parameters were 95°C-5 min, 40x(95°C-30 sec, 56°C-30 sec, 72°C-30 sec), 95°C-15 sec, 60°C-15 sec, 95°C-15 sec. Data obtained from the qPCR reaction was analyzed using the comparative threshold cycle (Ct) method (User Bulletin No. 2, PerkinElmer Life Sciences) with glyceraldehyde 3-phosphate dehydrogenase (GAPDH) used for normalization as the reference gene. All primer sequences for gene expression analysis are listed in Table 2.1.

2.4 IMMUNOBLOT ANALYSIS

For immunoblotting cultured cells were lysed in a phosphate buffered saline (PBS)-based RIPA lysis buffer (0.5% sodium deoxycholate, 1% NP-40, 0.1% sodium dodecyl sulfate (SDS) all from Sigma) with addition of a fresh protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN). Tumor tissues were ground in liquid nitrogen using a mortar and pestle and lysed in 50 mM Hepes buffer (Sigma), pH 7.6, containing 2% Triton X-100, 5 mM EDTA and fresh protease inhibitor cocktail (Sigma). Total protein concentration was measured using the Bio-Rad DC Protein Assay (Bio-Rad Laboratories, Hercules, CA) and lysates were resolved by SDS- polyacrylamide gel electrophoresis (5% stacking gel and 7.5% or 10% separating gel), transferred to nitrocellulose membranes (Bio-Rad Laboratories, Hercules, CA) and developed by enhanced chemiluminescence (PerkinElmer Life Sciences, Boston, MA). Membranes were probed overnight with primary antibodies using the following dilutions: 1:1000 rat anti-RBPJ clone number 76719 (Institute of Immunology, Tokyo, Japan), 1:500 rabbit anti-cleaved Notch1 Val1744 (Cell Signaling Technology, Danvers, MA), 1:10,000 mouse anti α-tubulin (Sigma).

2.5 CO-CULTURE EXPERIMENTS

105 MDA-MB-231 shRandom or shRBPJ cells were co-cultured with 105 mouse SVEC4-10 endothelial cells transduced with MIY control vector or MIY-Jagged1. Cells were plated in 12- well tissue culture dishes in media either containing 10 µM DAPT (Calbiochem, Germany) or DMSO vehicle control (Invitrogen). Cells were harvested after 48 h of co-culture and qPCR was performed with human-specific primers to avoid amplification of mouse transcripts.

Table 2.1 List of primer sequences used.

Product Human qPCR Forward 5'-3' Reverse 5'-3' Size (bp)

BCL2 CTGCACCTGACGCCCTTCACC CACATGACCCCACCGAACTCAAAGA 119 FAM81A TGGTTGCAACAGGAACAAGAACGG ATCCCTTTCACTGGCTCCACTGTT 102 GAPDH GCAAATTCCATGGCACCGT TCGCCCCACTTGATTTTGG 106 GFP CACATGAAGCAGCACGACTT AGTTCACCTTGATGCCGTTC 233 GUCY1A3 TACAAGGTGGAGACCATTGGCGAT ATCCAGAGTGCAGTCCAATTCGCA 175 HES1 CGACACCGGATAAACCAAAGACAG CGCGAGCTATCTTTCTTCAGAGCA 145 HEY1 AGAGTGCGGACGAGAATGGAAACT CGTCGGCGCTTCTCAATTATTCCT 115 HEY2 TTGAAGATGCTTCAGGCAACAGGG TCAGGTACCGCGCAACTTCTGTTA 115 HEYL ATGCAAGCCAGGAAGAAACGCAGA AGCTTGGAAGAGCCCTGTTTCTCA 125 MMP1264 GGGCTTGAAGCTGCTTACGA TGTCCCTGAACAGCCCAGTAC 83 MMP3 TGAAGTTACTAGCAAGGACCTCGT AGGGTGTGGATGCCTCTTGGGTAT 105 MMP9265 TGACAGCGACAAGAAGTG CAGTGAAGCGGTACATAGG 143 NFKB2266 GGACTGGTAGGGGCTGTAGG CACATGGGTGGAGGCTCT 108 NOTCH1 CGCACAAGGTGTCTTCCAG AGGATCAGTGGCGTCGTG 87 NOTCH4 AGGCCACAGCAGGATCAC AGGTGGGGCCTTCAAAAC 66 PPM1E CCACACAAACCAGACAGAGAGGAT AACCGACAGACTTCCATTCACCCT 102 RBPJ GCTGACTTATGCATTGCCTCAGGA CCACTGCTGTGAACTGGCATGAAA 123 TREM1 AGGGTTCCGGTGTTCAACATTGTC AATGACCTCAGCGTGACAGCAAAC 98

Mouse qPCR Forward 5'-3' Reverse 5'-3'

GAPDH TGCAGTGGCAAAGTGGAGAT TTTGCCGTGAGTGGAGTCATA 96 HES1 GAGAGGCTGCCAAGGTTTTT ACATGGAGTCCGAAGTGAGC 195 HEY1 CACGCCACTATGCTCAATGT TCTCCCTTCACCTCACTGCT 198 HEY2 TTCTGTCTCTTTCGGCCACT TTTGTCCCAGTGCTTGTCTG 205 RBPJ AAAGTCTTCAGTTGAATGGCGGCG CTCCACATCTGTACATTGTTTCGGC 123

ChIP Primers Forward 5'-3' Reverse 5'-3'

HEY2 TSS CGGAAGGCCTGGGCCGGT AACTCTGCTCGGCCGCAAGAGCC 97 HEY2 -160 CAAGATTGCGCTCATTCTCGCCC CAAGGCTCCAGCCGAGGCGT 99

EMSA Primers Forward 5'-3' Reverse 3'-5'

HEY1+45 wt GGGGCTCAGCGTGGGAAAGGATGGTTGAGT GAGTCGCACCCTTTCCTACCAACTCAGGG 30 mer HEY1-378 wt GCGGCGCCGGCACTTTCCCACGGCCCG CGGCCGTGAAAGGGTGCCGGGCGCCGG 30 mer HEY2+41 wt GGAGTTGCGGCGTGGGAAAGAGCCGCTAGGA AACGCCGCACCCTTT CTCGGCGATCCTGGG 30 mer HEY2-160 wt GGGTCCGCGGCCGCGTTCCCACGCCTCGG AGGCGCCGGCGCAAGGGTGCGGAGCCGAGG 30 mer HEY1+45 mt GGGGCTCAGCGTGCTGCAGGATGGTTGAGT GAGTCGCACGACGTCCTACCAACTCAGGG 30 mer HEY1-378 mt GCGGCGCCGGCACTGCAGCACGGCCCG CGGCCGTGACGTCGTGCCGGGCGCCGG 30 mer HEY2+41 mt GGAGTTGCGGCGTGCTGCAGAGCCGCTAGGA AACGCCGCACGACGT CTCGGCGATCC TGGG 30 mer HEY2-160 mt GGGTCCGCGGCCGCGG CAGCACGCCTCGG AGGCGCCGGCGCCGTCGTGCGGAGCCGAGG 30 mer wt = wild-type, mt = mutant, underline highlights wt or mt RBPJ binding sites, +/- represents distance from the TSS.

2.6 CELL DEATH ASSAY

We used the following stresses to induce cell death: serum starvation, anoikis assays, or treatment with peroxide. For serum starvation, MDA-MB-231 were plated at 5 x 105 cells/well, allowed to adhere overnight and incubated in serum free media for 72 hr. For DG75 cell lines 5 x 105 cells/well were plated in serum free media and incubated for 48 h. Anoikis experiments were performed with MDA-MB-231 cells by plating 2.5 x 105 cells/well in Costar low-attachment plates (Stem Cell Technologies). Cells were then incubated in suspension for 48 hours. For treatment with 100 µM tert-butyl hydroperoxide (TBHP, Molecular Probes, Invitrogen), DG75 cells were plated at 7.5 x 105 cells/well and treated for 2 hr to induce death. Each assay was done in triplicate in 6-well plates using a 2 ml volume per well. Incubations were performed in a 37°C humidified incubator. To quantify the surviving fraction following these stress conditions, cells were harvested (for MDA-MB-231 cells the supernatant, PBS wash and adherent cells were pooled). Cells were pelleted, washed once with cold PBS and resuspended in 100 µl AnnexinV binding buffer (10 mM Hepes, 140 mM NaCl, 2.5 mM CaCl2, pH 7.4 all from Sigma). Following addition of 5 µl AnnexinV-488 to DG75 cells or AnnexinV-APC to MDA-MB-231 cells (both fluorescent conjugates from Invitrogen) and 5 µl propidium iodide (1mg/ml, Sigma) per tube, the reaction was incubated for 15 min. 500 µl ice cold AnnexinV binding buffer was added to each tube and flow cytometry analysis was performed immediately using the EPIC Elite flow cytometer (Beckman Coulter, Brae, CA) and FCS Express (De Novo Software, Los Angeles, CA). Data are expressed as the proportion of surviving cells (defined as AnnexinV/propidium iodide negative) relative to untreated cells within each group.

2.7 CELL GROWTH ASSAY

To determine cell growth all viable cell counts were performed manually by trypan blue exclusion. For each time point, cells were seeded in triplicate. Adherent MDA-MB-231 cell lines were plated in 6-well plates, 1 x 105 cells per well in a 2 ml final volume. The cells were trypsinized and counted on day 2, 4, 6, and 8. For DG75 cells extrapolated cell growth assays were performed as described by Weng et al. 2003172. Cultures were initially seeded at 7.5 x 105 viable cells per well in a 2 ml volume in 6-well dishes. Cells were re-seeded on day 2, 4, 6 and 8 at the same density. Extrapolated cell counts were calculated each day by using the formula [(current day’s cell concentration)/(3.75 x 105 cells/ml)] x (previous day’s extrapolated cell count).

2.8 MICROARRAY DATA

Published microarray data were obtained from the following sources: Gene Expression Omnibus (GEO), accession numbers GSE5364 (human breast cancer and adjacent non- malignant tissue, Yu et al. 2008267), GSE4824 (lung cancer cell lines, Lockwood et al. 2008268), GSE17768 (breast cancer cell lines MCF10A, BT-474, MCF-7, HCC1395, HCC2218, HCC1143, HCC1599, HCC38, HCC1937, Chari et al. 2010269), GSM183440/GSM183441 (breast cancer cell line Sk-Br-3, Vivanko et al. 2007270), GSM307015 (breast cancer cell line MDA-MB-231, Li et al. 2009271), GSE3141 (111 NSCLC tumors, Bild et al. 2006272) and the Oncomine database (normalized breast cancer microarray data, Ginestier et al. 2006273 and Zhao et al. 2004274, http://www.Oncomine.org). Expression data for HCC1187 is unpublished (Chari and Lam, unpublished data). In addition, data for 67 normal human bronchial epithelial cells analyzed with the Affymetrix U133 plus array platform and 49 lung cancer samples analyzed with a custom Affymetrix chip (both from Lockwood et al. 2010)275 were obtained from the System for Integrative Genomic Microarray Analysis (SIGMA) website (http://sigma.bccrc.ca/, Chari et al. 2006276). All microarray data are log2 transformed.

2.9 ARRAY COMPARATIVE GENOMIC HYBRIDIZATION (aCGH)

DNA copy number profiles were generated for 215 cancer cell lines, and 49 lung cancer samples using whole genome tiling path aCGH. Details of the genomic array, DNA extraction, labeling and hybridization, image analysis and normalization have been described previously (Ishkanian et al. 2004277, Lockwood et al. 2008268). Data are available through the SIGMA website (http://sigma.bccrc.ca/).

2.10 XENOGRAFT TUMOR GROWTH

All protocols involving mice were evaluated and approved by the University of British Columbia Animal Care and Ethics Committee. Female non-obese diabetic/severe combined immunodeficient mice (NOD/SCID, Animal Resource Centre of the British Columbia Cancer Research Centre) between 10 to 12 weeks of age were injected subcutaneously into the dorsa with human tumor cell lines. Tumor growth was monitored at the indicated times by external measurements with calipers, and tumor volume calculated using the formula (π/6 x length x width x height). For MDA-MB-231 tumor studies 5 x 106 stably transduced cells were injected in 100 µl serum free DMEM. For DG75 tumor studies 1 x 107 cells were injected in 200 µl PBS.

Individual tumors were split for either fixation in 4% paraformaldehyde or freezing, and then used for histology, immunostaining, RNA and protein collection.

2.11 IMMUNOHISTOCHEMISTRY ON PARAFFIN BLOCKS

The small breast tissue cancer microarray (07-002) and the normal control tissue microarray (07-009) sections were obtained from Dr. Blake Gilks at the Genetic Pathology Evaluation Centre (Jack Bell Research Centre, Vancouver, BC). These sections were cut from formalin- fixed, paraffin-embedded small-core biopsies placed in the same block. Xenograft tumors excised from mice were fixed overnight in 10% phosphate-buffered formalin, progressively dehydrated through gradients of alcohol, and embedded in paraffin. Further analysis was performed by PMI Labs (Vancouver, BC). Briefly, 4 μm sections were cut, deparaffinized in xylene, rehydrated, and then stained with hematoxylin and eosin. For immunohistochemical analysis, antigen retrieval was performed with a Pascal decloaker (Dako, Carpinteria, CA) using citrate buffer, pH 6 (Dako). The rat anti-RBPJ7A11 antibody (obtained from the Bettina Kempkes, Germany)113 and IgG2Bkappa control (BD Biosciences, San Jose, CA) were used as primary antibodies at a 1:10 dilution. The slides were stained using the avidin-biotin-peroxidase method with the EnVision Flex kit (Dako) as per manufacturer’s instructions, a 1:50 dilution of Biotin-SP-AffiniPure donkey anti-rat secondary (Medicorp Inc, Montreal, QC), and 1:500 peroxidase-conjugated streptavidin (Medicorp Inc) followed by the chromagen 3,3’- diaminobenzidine (Dako) to develop the immunostain. All sections were counterstained with hematoxylin. 1.5% sodium bicarbonate was used as a blueing agent. Sections were imaged using an Olympus BX41 light microscope. Tissue microarray slides were scanned with the BLISS imaging system (Bacus Laboratories Inc. Lombard, IL) and images were imported into the BLISS software for viewing and acquisition of micrographs. Immunophenotypic evaluation was performed independently by two evaluators who scored the tissue microarrays for presence or absence of RBPJ.

2.12 WHOLE SECTION TUMOR STAINING

Tumor staining study protocols were similar to those previously reported278,279. Mice were administered 1500 mg/kg 5-bromo-2-deoxyuridine (BrdU, Sigma) in 0.9% NaCl and 60 mg/kg pimonidazole (Hypoxyprobe, Burlington, MA) by intraperitoneal injection 2 h before tumor excision. Harvested tumors were frozen at -20°C on an aluminum block, covered in embedding medium (O.C.T.) and stored at -80°C until sectioning. Cryosections 10 µm thick were obtained using a Cryosar HM560 (Microm International GmbH, Thermo Scientific, Germany), dried 41

overnight and fixed in 50:50 acetone-methanol for 10 min at room temperature. Incorporated BrdU was detected using a monoclonal mouse anti-BrdU antibody (clone BU33, Sigma), followed by an anti-mouse peroxidase-conjugated antibody (Sigma) and a metal-enhanced 3,3’- diaminobenzidine substrate (Pierce, Thermo Scientific, Germany). Hypoxic cells labeled with pimonidazole were detected using a polyclonal rabbit anti-pimonidazole antibody (Hypoxiprobe) and visualized with a fluorescent Alexa-488 tagged anti-rabbit secondary antibody. Vasculature was stained using a rat anti-mouse monoclonal antibody to CD31 (clone MEC 13.3, BD Biosciences) and fluorescent Alexa-546 anti-rat secondary antibody (Invitrogen). Apoptotic cells were detected using TUNEL in situ cell death detection kit (Roche Diagnostics, Laval, QC) with a TMR red-tagged dUTP. Following imaging of fluorescence, slides were placed in distilled water for 10 min and treated with 2 mol/L HCl at room temperature for 1 h, followed by neutralization for 5 min in 0.1 mol/L sodium borate and rinsing in PBS. Slides were then counterstained with hematoxylin, dehydrated, and mounted using Permount (ThermoFisher Scientific) before imaging.

Images were acquired using a fluorescence microscope (Zeiss Imager Z1), a cooled, monochrome CCD camera (Retiga 4000R, QImaging), and motorized x-y stage (Ludl Electronic Products). This system allows tiling of adjacent fields of view to compile images of entire tumor cryosections at a resolution of 0.75 µm/pixel; sections were repeat-imaged at various points during staining. Using NIH-Image and user supplied algorithms, images of BrdU, CD31, pimonidazole, TUNEL and hematoxylin staining, all from the same tumor section, were overlaid and staining artifacts removed. Grayscale images of CD31 and TUNEL were inverted, thresholded, colorized (blue and red respectively) and overlaid onto images of BrdU and hematoxylin (black and grey respectively). Grayscale images of pimonidazole (green channel) were overlaid using Adobe Photoshop CS (version 8.0) and combined using the multiply mask. Tumor tissues were analyzed either as whole tumor sections, or as viable tissue only following crop and removal of confluent necrosis. TUNEL positive staining was identified by selecting pixels a minimum of 5 standard deviations above tissue background levels, and is reported as either the fraction of TUNEL positive staining for whole sections, or as the fraction of TUNEL positive staining in viable tissue. BrdU positive staining is reported as the fraction of positive pixels in viable tissue. Distribution analyses was performed in viable tissue only; each pixel in an image cropped to remove necrosis was sorted based on its distance relative to CD31- positive vasculature. For pimonidazole staining the average pixel intensity is reported as a function of distance from vasculature, and for BrdU and TUNEL analyses the fraction of pixels

meeting or exceeding the stain threshold is reported as a function of distance from nearest vasculature.

2.13 ELECTROPHORETIC MOBILITY SHIFT ASSAYS (EMSAs)

Nuclear lysates were collected from LNCX, a vector expressing a mutated form of RBPJ (LNC- mt-RBPJ), shRandom, shRBPJ and shRBPJ-set2 over-expressing MDA-MB-231 cells for the RBPJ EMSA assays. Briefly, the cells were washed two times with PBS, resuspended in 4 pellet volumes of buffer A (10 mM Hepes-KOH (pH 7.8), 10 mM KCl, 1.5 mM MgCl2, 5 mM with addition of a protease inhibitor cocktail all from Sigma), pelleted for 15 sec at 12,000 x g and resuspended in 375 µl buffer A containing 0.5% NP-40 (Sigma). Following mixing, cells were incubated at 4°C for 10 min to lyse the cells. The cells were centrifuged for 10 min at 12,000 x g at 4°C and the cytosolic extract was removed. The nuclei were washed once with buffer A, and resuspended in 3 pellet volumes of buffer B (50 mM Hepes-KOH (pH 7.8), 10% glycerol, 50 mM

KCl, 300 mM NaCl2, 0.1 mM EDTA, with addition of a protease inhibitor cocktail all from Sigma) and incubated at 4°C for 20 min. The nuclear lysates were collected after centrifugation for 10 min at 12,000 × g at 4°C and stored at -80°C. The oligonucleotides containing RBPJ wild-type or mutant motifs in the HEY1 and HEY2 promoter (shown in Table 2.1) were end-labeled with 32P dCTP using the Klenow fragment of DNA Polymerase I. The binding reaction (10 mM TrisHCl, 50 mM NaCl, 1 mM EDTA pH 8, 4% glycerol, 2 µg PolydI-dC binding buffer all from Sigma), and 10 μg of nuclear protein was performed by preincubating with either 50-fold excess wild-type or mutant188 nonradioactive duplex oligos for 15 min on ice, and then adding a 150,000-cpm 32P-labeled double-stranded probe and incubating for 30 min at room temperature. DNA-protein complexes were electrophoresed on 5% Tris-Borate EDTA gels, dried and exposed to a phosphorimager plate for 16 h.

2.14 CHROMATIN IMMUNOPRECIPITATION, LIBRARY CONSTRUCTION AND SEQUENCING ANALYSIS shRandom or shRBPJ ChIP samples were prepared from stable MDA-MB-231 cell lines grown to 70% confluence. Briefly, cells were cross-linked with 1% formaldehyde for 10 min at room temperature, the reaction was quenched by addition of 0.125 M glycine, and then the cells were washed in PBS, and harvested in the presence of protease inhibitors. ChIP was performed as described in Robertson et al. 2008280. Chromatin DNA was fragmented by sonication for 10 min using the Sonic Dismembrator 550 (cup horn, Fisher Scientific, Canada) to produce chromatin fragments ranging from 100 bp to 800 bp. The chromatin was pre-cleared with 40 µl of blocked 43

Protein A/G sepharose beads (Amersham, England) at 4oC for 2 hrs. Supernatants were removed from the beads following centrifugation, and transferred to fresh tubes. Sepharose beads were used for ChIPs and each immunoprecipitation was carried out with 100 µg of pre- cleared chromatin and 2.5 μg of anti-acetylated histone H4 antibody (H4K5/K8/K12/K16, product number 06-866, Upstate Biotechnology, Lake Placid, NY), 4 µl rabbit anti-histone H3 trimethyl lysine 4 clone MC315 (Millipore, Billerica, MA), or 4 µg Rabbit IgG from (Upstate) and incubated at 4oC for 1 hr. To each immunoprecipitation reaction, 20 µl of Protein A/G beads were added and incubated by rotating at 4oC overnight. Beads were recovered by centrifugation and washed twice with ChIP wash buffer (20 mM Tris-HCl pH 8.0, 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl) and once with ChIP final wash buffer (20 mM Tris-HCl pH 8.0, 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 500 mM NaCl). DNA-antibody complexes were eluted using 100 µl Elution Buffer (100 mM NaHCO3, 1% SDS), and 5 µg of DNAse-free RNAse (Roche, Mississauga, ON) was added and incubated at 68°C for 2 hours with shaking on a thermomixer to reverse the crosslink. The beads were pelleted by centrifugation and the supernatant was collected. Elution was repeated with the addition of 100 µl of Elution Buffer and incubation at 68oC for 5 min with shaking on a thermomixer. After pooling the two elutions, DNA was recovered from the eluate using the QIAquick PCR Purification kit (Qiagen, Germany). The immunoprecipitated DNA was validated by qPCR using 0.25 µM final concentration GAPDH primers (Forward Part#101221; reverse Part #101222 from Active Motif, Carlsbad, CA), HEY2-160, HEY2 TSS (see Table 2.1) and ZNF3-set2 as a negative control (data not shown). qPCR was set up on a 7900HT Fast Real-Time PCR System (Applied Biosystems) using 1 µl of immunoprecipitated eluate and the SYBR Green PCR Master Mix (Applied Biosystems). Cycling parameters were 95°C-10 min, 40x(95°C-10 sec, 59°C-30 sec, 72°C-30 sec), 72°C-5 min, 95°C-15 sec, 59°C-15 sec, 95°C-15 sec. Fold enrichment was calculated as 2^[(Ct IgG – Ct target)].

Sequencing libraries for the acH4 ChIP were prepared as described (Robertson et al. 2007)281. Briefly, ChIP DNA (roughly 50 ng) was run in 8% PAGE and the 100–300 bp fraction was excised, eluted overnight at 4°C in 300 μl of elution buffer (5:1, LoTE buffer (3 mM Tris-HCl, pH 7.5, 0.2 mM EDTA)-7.5 M ammonium acetate), and was purified using a Spin-X Filter Tube (Fisher Scientific), by ethanol precipitation. Libraries were prepared using the Illumina single- end (SE) library construction protocol. This involved DNA end-repair, and phosphorylation by T4 DNA polymerase, Klenow DNA Polymerase and T4 polynucleotide kinase, respectively, in a single reaction and subsequent 3’ A overhang generation by Klenow fragment (3’ to 5’ exo

minus) and ligation to Illumina SE adapters (with 5’ overhangs). Adapter-ligated products were purified on Qiaquick spin columns (Qiagen) and PCR-amplified using Phusion DNA polymerase in 10 cycles and the SE primer set (Illumina, San Diego, CA). The PCR product was purified using 8% PAGE gels and DNA quality was assessed and quantified using an Agilent DNA 1000 series II assay (Aligent, Santa Clara, CA) and Nanodrop 7500 spectrophotometer (Nanodrop, Wilmington, DE), and subsequently diluted to 10 nM. The final concentration was confirmed using a Quant-iT dsDNA HS assay kit and Qubit fluorometer (Invitrogen). Clusters were generated on the Illumina cluster station and sequenced to 75 bp reads using the Illumina 1G analyzer following the manufacturer’s instructions.

We identified genes with the 1000 highest and lowest differential enrichment ranks, as follows. We calculated differential acH4 enrichment ratios for RefSeq transcripts, considering regions that extended from 2500 bp 5’ to 1500 bp 3’ of each transcript’s TSS. Reviewing known Notch target genes in the UCSC hg18 genome browser suggested that such regions should include important changes in enrichment profiles. Each acH4 enrichment profile represented the per- nucleotide depth of 200 bp immunoprecipitated virtual DNA fragments (XSETs)281 that were indicated by aligned Illumina sequence reads; profile areas integrated this depth across a TSS region. Profile areas were adjusted by the ratio of total number of aligned sequence reads in the two libraries, so that enrichment ratios would be approximately normalized to equal numbers of aligned reads. RefSeq transcripts were ranked by the ratio, R, of the TSS region profile areas for knock-down (Akd) and control (Ac), using R=(Akd-Ac)/(Akd+Ac). The ratio varied from -1 to +1, corresponding to H4ac enrichment in the control and in the KD, respectively. Because TSS regions that had relatively low numbers of aligned reads would tend to have ratios that were more strongly affected by noise from the aligned-read background, we removed from analysis all TSS regions that had relatively low profile areas; setting a threshold to the 39th percentile profile area for each dataset. For genes with multiple transcripts we used the ratio from the transcript that had the largest absolute value ratio. From the resulting list of R-ranked genes, we retained for more detailed analysis the genes with the 1000 highest and lowest differential enrichment ranks.

2.15 INGENUITY PATHWAY ANALYSIS

Ingenuity Pathway Analysis (Ingenuity Systems, Redwood City, CA, www.ingenuity.com) was used to interpret global acH4 data by performing Core Analysis using default parameters in the context of biofunctions (molecular and cellular functions) and a custom Notch/RBPJ target gene 45

list (Table 1.2). The 1000 highest and lowest acH4 R-ranked genes were defined as value parameters for the analysis. Significance was tested by either Fisher Exact test or with the Benjamini-Hochberg correction for multiple testing using P ≤ 0.01.

2.16 STATISTICAL ANALYSES

Specific tests are noted in the text and figure legends. For pairwise comparisons if data were normally distributed, we used the Student's T-test. Otherwise, we used the Mann-Whitney U test. The parametric one-way ANOVA and nonparametric Kruskal-Wallis one-way analysis of variance tests were used to compare multiple groups. The association between mRNA expression of two genes was assessed by Pearson correlation test. GraphPad Prism 5 (GraphPad Software, La Jolla, CA) was used as the statistical analysis software package. P values of less than 0.05 were considered significant unless otherwise stated.

Box-and-whisker plots were generated with MATLAB (MathWorks, Natik MA) to depict the range and percentiles of values obtained. The box indicates the 75th (top), 50th (middle line - median), and 25th (bottom) percentiles of the values obtained. Observations were considered outliers (shown as pluses) if they exceeded 1.5 times the interquartile range.

CHAPTER 3: RBPJ LOSS INCREASES TUMORIGENICITY

3.1 INTRODUCTION

The critical Notch partner RBPJ is thought to be constitutively expressed in normal cells110,121, and in the absence of Notch activation functions as a transcriptional repressor47,118,148. To verify whether ubiquitous expression of RBPJ occurs, as has been reported the literature, we evaluated RBPJ protein levels in a number of normal human tissue specimen.

Evidence exists for aberrant Notch signaling activity in human breast cancer79,95,98, and other cancer types33,46, and inhibition of the Notch signaling pathway has been shown to impede mammary tumorigenesis88. We therefore asked whether expression of RBPJ, the key transcriptional effector of Notch, is altered in human tumor cells to determine if RBPJ loss is potentially relevant to human disease.

To further understand the function of RBPJ in a cancer context, we evaluated the effect of RBPJ removal on xenograft tumor growth in a non-obese diabetic (NOD)/severe-combined immunodeficient (SCID) mouse model. We also assessed whether enforcing the transcriptional repression function of RBPJ would reduce the oncogenic potential of xenografted breast tumors.

3.2 RESULTS

3.2.1 RBPJ expression in normal tissues

Invariant RBPJ expression has been reported in all normal mouse tissues and most human cell lines evaluated121 which has led to the conclusion that the gene is constitutively and ubiquitously expressed. To our knowledge, RBPJ expression has not been assessed in primary human tissue, other than in human amniotic fluid cells where the protein was detected in the nucleus111. We therefore stained a collection of various normal human tissues for RBPJ by immunohistochemistry (IHC). The tissues evaluated (human breast, lymph node, thymus, salivary gland, esophagus, stomach, small intestine, colon, pancreas, kidney, prostate, testis, ovary, fallopian tube, uterus and thyroid) stained positive for RBPJ protein. Figure 3.1 shows examples of tissues with some of the most intense RBPJ staining.

Figure 3.1 RBPJ protein is expressed in various normal human tissues. RBPJ was detected by immunohistochemistry (IHC) on paraffin sections using diaminobenzidine as a chromogen (in brown) with the tissues shown in the figure having the most intense staining of all samples evaluated. Haematoxylin was used as a nuclear counter stain (blue). In the breast tissue (derived from a reduction mammoplasty) epithelial cells in the ducts stain positive for RBPJ. Scale bar for all = 50 µm.

3.2.2 Frequent RBPJ deficiency in human cancer cell lines

The high RBPJ expression levels detectable in many normal tissues imply that RBPJ plays an essential role in regular cell function. To evaluate whether RBPJ is lost in cancer, we first analyzed RBPJ deletion status using array comparative genomic hybridization (aCGH) across a panel of 215 cancer cell lines, representing 15 different tissue types (Figure 3.2A). This technique uses microarray technology to evaluate DNA copy number changes for regions in the genome by co-hybridizing a differentially labeled pathological sample (tumor) and normal reference sample to genomic probes, and evaluating the ratio of fluorescent intensity282. For our analysis the reference genomic DNA was obtained from blood samples of normal unmatched healthy individuals (either a single male or female or a pool of normal males or females). A loss of at least 1 allelic copy of RBPJ occurred at an incidence of 35% across all cancer cell lines, with a 65% loss detected in the breast cancer cell line subset (n=34, Figure 3.2A and B). In contrast, RBPJ amplification was identified in only one of the 34 breast cancer cell lines evaluated (HCC1419, a cell line derived from a primary ductal carcinoma). Of the 22 breast cancer cell lines that showed RBPJ loss, the mutation occurred as part of a large deletion

Figure 3.2 RBPJ is frequently lost in human cancer cell lines. (A) RBPJ deletion status evaluated using aCGH across a panel of 215 cancer cell lines (CNS, central nervous system, HC, hematopoietic cell lines). At least one copy of RBPJ is lost at an overall frequency of 35%. (B) Breast cancer cell lines, which represent the largest group next to lung cancer cell lines, show a high frequency of RBPJ gene loss, which occurs as a deletion of chromosome 4p, compared to the cumulative loss observed in all 215 cancer cell lines examined. A single copy loss and homozygous deletions are scored the same and values are summarized across all the samples to obtain a frequency. Loss = -1 (in green), normal = 0 and gain = 1 (in red). (C) Paired RBPJ expression and aCGH analysis of 11 breast cancer cell lines normalized to a normal breast epithelial cell line (MCF10A). Cell lines with a genetic loss of RBPJ are depicted in green and those with normal RBPJ copy number in red. Microarray expression data (top), qPCR validation (middle) and immunoblot analysis (bottom) corroborate RBPJ expression with RBPJ copy number status. (D) MDA-MB-231 breast cancer cells show no loss at chromosome 4p, which encompasses the RBPJ locus. (E) Association of RBPJ copy loss with ER, PR and HER2 receptor status or molecular subtype in breast cancer cell lines. Cell lines (n=33) with neutral RBPJ copy number (retention, n=11) or RBPJ deletion (n=22) were sorted based on receptor status. Each bar of the graph represents proportion of cells showing RBPJ retention or loss within that group. ER, PR or HER2 status was individually considered, regardless of the presence or absence of the other two receptors. The cells were also sorted into following molecular subtypes; luminal A (ER+ and/or PR+, HER2-), luminal B (ER+ and/or PR+, HER2+), HER2+ (ER and PR-, HER2+) and triple negative (ER, PR and HER2-). The number of cell lines showing RBPJ deletion or retention is as follows: ER+ (6 deletion, 4 retention), ER- (16 deletion, 7 retention), PR+ (6 deletion, 2 retention), PR- (16 deletion, 9 retention), HER2+ (10 deletion, 1 retention), HER2- (12 deletion, 10 retention), luminal A subtype (5 deletion, 3 retention), luminal B subtype (2 deletion, 1 retention), HER2+ subtype (8 deletion, 0 retention) and triple negative subtype (7 deletion, 7 retention). (F) Boxplot representation of data from 66 lung cancer cell lines showing reduced RBPJ expression correlating with RBPJ copy loss. P value obtained from a Mann-Whitney test comparing RBPJ mRNA expression in lung cancer cell lines with RBPJ deletion to those with a normal copy number.

encompassing the whole short arm of chromosome 4 (Figure 3.2B). Table 3.1 names all of the 34 breast cancer cell lines along with RBPJ status, histological type, and ER, PR and HER2 receptor status.

To determine whether RBPJ genomic loss corresponds to reduced RBPJ expression, we further evaluated RBPJ levels in 11 breast cancer cell lines for which we had aCGH data and could obtain mRNA and protein. Genomic loss of at least one copy of RBPJ was accompanied by reduced transcript and protein levels (Figure 3.2C). Conversely, cell lines with normal RBPJ copy number such as MDA-MB-231 cells (Figure 3.2D) had high RBPJ expression (Figure 3.2C).

To access a possible association between RBPJ copy loss and ER, PR and/or HER2 receptor status or molecular subtype we evaluated 33 breast cancer cell lines which had neutral RBPJ copy number or RBPJ copy loss (omitting the HCC1419 cell line that showed RBPJ amplification, see Table 3.1). RBPJ copy loss was marginally more frequent in ER- over ER+ (60% and 70% respectively) and PR+ compared to PR- (75% and 64% respectively) cell lines (Figure 3.2E). However, RBPJ deletion was especially evident in HER2+ when compared to 51

Table 3.1 Features of breast cancer cell lines used for aCGH analysis at the RBPJ locus.

Cell line RBPJ status Type ER PR HER2 BT-474 Deletion IDC + + + BT-483 Retention IDC + + - BT-549 Deletion IDC - - - HCC1008 Deletion IDC - - + HCC1143 Deletion DC ------HCC1187 Deletion DC HCC1395 Retention DC - - - HCC1419 Amplification DC - - + HCC1428 Deletion Met AC + + -

HCC1500 Deletion DC + + - HCC1569 Deletion Met C - - + HCC1599 Retention DC - - -

HCC1806 Retention SC - - - HCC1937 Deletion DC - - - HCC1954 Deletion DC - - +

HCC202 Deletion DC - - + HCC2157 Deletion DC - + + HCC2218 Deletion DC - - +

HCC3153 Deletion DC - - - HCC38 Deletion DC - - - HCC70 Retention DC - - -

HCC712 Retention DC + + - HS-578T Retention IDC - - - MCF-7 Deletion Met AC + + -

MDA-MB-134-VI Retention IDC + - - MDA-MB-157 Retention MC - - - MDA-MB-175-V I I Deletion IDC + - -

MDA-MB-231 Retention Met AC - - - MDA-MB-435 Deletion IDC - - - MDA-MB-453 Deletion Met C - - +

Sk-Br-3 Deletion AC - - + T-47D Deletion IDC + + -

UACC-893 Deletion IDC - - +

ZR-75-30 Retention IDC + - +

IDC = invasive ductal carcinoma, DC = ductal carcinoma, Met AC = metastatic adenocarcinoma, Met C = metastatic carcinoma, SC = squamous carcinoma, AC= adenocarcinoma, MC = medullary carcinoma. Breast tumor type, ER, PR and HER2 status were obtained from Gazdar et al. 1998283, Meltzer et al. 1988284, Neve et al. 2006260 and Kao et al. 2009285.

HER2- cell lines (91% and 51% respectively) regardless of ER or PR status (Figure 3.2E). Rather than individually considering each receptor we sorted the cells into molecular subtype based on ER, PR and HER2 status. RBPJ deletion occurred in a similar proportion of luminal- type breast cancer cell lines (luminal A 63% and luminal B 67%, Figure 3.2E). The 8 cell lines with a HER2+ (ER and PR-) molecular subtype all showed RBPJ copy loss (Figure 3.2E). Breast cancer cell lines negative for all three markers (triple negative) had an equal proportionof RBPJ retention versus loss (Figure 3.2E). These data imply that RBPJ copy loss occurs more frequently in HER2+ breast cancer cell lines, although RBPJ deletion was common in cell lines of each subtype.

Although the incidence of RBPJ loss was not as frequent in human lung cancer cell lines (27%, n=86), a subset of these (n=66), for which we had paired aCGH and microarray mRNA expression data, confirmed that RBPJ loss (29%, n=19) corresponds to significantly decreased RBPJ transcript (Figure 3.2F). Interestingly, RBPJ gene amplification (21%, n=14) did not result in increased mRNA expression compared to that of lung cancer cell lines with a neutral RBPJ copy number (50%, n=33).

3.2.3 Evidence for RBPJ loss in primary human breast and lung cancers

To evaluate RBPJ expression in primary human breast cancers we used published microarray data from Yu et al. 2008267, accessed via the gene expression omnibus website (www.ncbi.nlm.nih.gov/geo/). This dataset includes 183 breast cancer samples and 13 normal adjacent non-malignant tissue samples that could be used for comparison. We found RBPJ expression was significantly down-regulated in breast cancer samples compared to normal breast specimens (Figure 3.3A). We further validated loss of RBPJ protein and mRNA in primary human breast and lung tumors using additional samples from the BC Cancer Agency. A small tissue microarray (TMA) of 38 breast carcinoma cases of unspecified histological type was analyzed by IHC and scored for presence or absence of RBPJ staining. For generation of the TMA small discs of tissue (cores) from separate paraffin blocks were placed in an array on a recipient paraffin block so they could be analyzed simultaneously. RBPJ protein expression was absent in 9 of these samples (24%), despite being detectable at high levels in normal human breast tissue (Figure 3.3B and Figure 3.1A). IHC staining of the small breast cancer TMA was also performed for ER and PR and using hematoxylin and eosin (H&E). This information confirmed that lack of RBPJ staining was due to absence of RBPJ protein and not the quality of the tissue sample, as some RBPJ-negative cores were positive for ER and/or PR (Figure 3.3C to F). Although receptor status was not obtainable for all samples due to loss of cores during the 53

Figure 3.3 RBPJ is lost in primary human breast cancers. (A) Data from the Yu el al. 2008 study267 was analyzed for RBPJ mRNA expression in human breast cancer samples or adjacent non-malignant tissue (normal breast). Boxplot representation of data shows under-expression of the RBPJ transcript in the breast cancer specimens. P value obtained from Mann-Whitney test comparing RBPJ expression in normal and cancerous tissue of the breast. (B) IHC staining using an RBPJ specific antibody in a human breast TMA composed of 38 tissue cores in one paraffin block. Tissue cores were scored for presence or absence of RBPJ stain by two independent evaluators with a similar outcome. (C and D) Example micrographs showing RBPJ staining results used to generate the graph in panel B. Scale bar = 200 µm. (C) An RBPJ positive (+) specimen with negative ER and PR staining. (D) An RBPJ negative (-) tumor core which is positive for ER and PR, indicating the lack of RBPJ staining is due to absence of RBPJ protein. (E and F) Magnified images of the tissue cores described in C and D respectively. Scale bare = 50 µm. H&E staining was used to confirm tissue integrity of cores. 54

staining procedure, absence of RBPJ staining was more common in ER+ versus ER- cores (60% versus 40%). Lack of positive RBPJ stain occurred in a similar proportion of PR+ and PR- cores (33% and 39% respectively). HER2 status was not evaluated in this dataset.

In addition to the human breast cancer samples, we were able to access RBPJ transcript levels in two lung cancer datasets. RBPJ mRNA expression was significantly reduced in non- small cell lung carcinomas (n=111) compared to bronchial epithelium (n=67) taken from healthy individuals (Figure 3.4A), providing further evidence for the relevance of RBPJ loss in human cancers compared to normal tissue. Analysis of another independent clinical lung cancer dataset, containing paired aCGH and microarray expression records for 49 tumor samples of mixed type, showed that genomic loss of at least one copy of RBPJ (29%, n=14) was consistent with reduced transcript levels (Figure 3.4B). Unlike in the lung cancer cell lines, amplification of RBPJ in the clinical lung cancer dataset (10%, n=5) corresponded to increased mRNA expression compared to samples with a neutral RBPJ copy number. However, RBPJ amplification was observed less frequently than deletion.

Figure 3.4 RBPJ is lost in primary human lung cancers. (A) Boxplot representation of decreased RBPJ mRNA expression in non-small cell lung carcinoma (NSCLC) tumors versus normal bronchial epithelium collected from healthy individuals. (B) Analysis of 49 lung cancer samples of mixed type with paired mRNA and aCGH data shows RBPJ deletion results in significant under-expression of transcript. Conversely, RBPJ amplification significantly increases RBPJ mRNA, but this is observed in fewer of the samples. P values for comparing RBPJ mRNA expression (between samples with RBPJ neutral versus deletion status and neutral versus amplification status) were obtained using a Mann-Whitney test.

3.2.4 RBPJ depletion increases xenograft tumor growth

To model the loss of RBPJ observed in primary tumors, lentiviral-expressed shRNA was used to knock-down (KD) RBPJ in the MDA-MB-231 human breast cancer cell line, which has a normal RBPJ copy number, expresses RBPJ at a high level (Figure 3.2C and D) and is tumorigenic in mice. When we subcutaneously injected these cells into immunocompromised NOD/SCID mice, RBPJ deficiency increased xenograft tumor growth compared to control tumors expressing a control non-specific shRNA, shRandom (Figure 3.5A). Knock-down of RBPJ was verified at the protein level by both immunoblot and IHC, showing a high level of

Figure 3.5 RBPJ depletion increases MDA-MB-231 xenograft tumor growth. (A) RBPJ KD (shRBPJ, n=16) accelerated tumor growth compared to tumors expressing a non-specific shRNA (shRandom, n=22). (B) Immunoblot analysis confirms KD is achieved in tumor lysates at the end of the experiment. (C) IHC showing diminished RBPJ staining at the tumor cell level. Scale bar = 50 µm. (D) Depletion of RBPJ using an alternate RBPJ targeting construct (set2) and a different silencing control (shScrambled) recapitulates original results (n=4 for each group). (E) Cells transduced with the LNCX vector control show increased tumor growth compared to cells expressing a mutant RBPJ construct (LNC-mt-RBPJ, n=8 for each group). (F) RBPJ protein levels were evaluated by immunoblot. P values obtained from Mann-Whitney tests for (A) and (E) and one-way ANOVA for (D) and data presented as mean ± SEM.

depletion in both tumor lysates and tumor cells respectively (Figure 3.5B and C). These findings were confirmed in an independent tumor experiment using an alternate RBPJ KD construct (set2) and a scrambled non-targeting shRNA control, shScrambled (Figure 3.5D). Forced Notch1IC expression in MDA-MB-231 cells also leads to increased tumor growth92. In contrast, forced expression of a mutated (mt)-RBPJ inhibited tumor growth (Figure 3.5E and F). The mutated RBPJ protein has a single amino acid substitution (arginine to histidine at position 179) which abolishes its ability to bind DNA but does not disrupt its interactions with NotchIC263, thereby blocking Notch activation but permitting endogenous RBPJ to exert its transcriptional repressor function (Figure 3.6 shows a schematic of how this construct functions). The inhibited tumor growth caused by mt-RBPJ is similar to that reported with direct blockade of Notch activation using various strategies88-90. Two of the studies that reported reduced MDA-MB-231 tumor growth in response to γ-secretase inhibitor Z-LLNle-CHO89,90 have been brought into question, due to off-target effects on the proteosome identified with this compound91. However, our previous data using a soluble ectodomain of Notch to block ligand mediated-activation88, and the results presented here with mt-RBPJ, confirm that NotchIC blockade in vivo inhibits tumor growth.

As RBPJ is ubiquitously expressed in normal tissue we sought to determine whether the tumor suppressor properties of RBPJ could be extended to other cell types. We used a genetic model of RBPJ loss in the B-cell lymphoma cell line DG75, in which RBPJ was inactivated by homologous recombination that deleted exon 4113 to generate RBPJ KO cells (for further details see Figure 1.4E in introduction section). Compared to the parental DG75 cells, growth of subcutaneously implanted RBPJ KO Burkitt lymphoma tumors was strikingly increased (Figure 3.7A). This finding, using an alternate strategy to lentiviral-delivered shRNA, confirms that loss

Figure 3.6 Illustrated mode of action of mt-RBPJ. The mt-RBPJ protein has an amino acid substitution in its beta- trefoil DNA-binding domain (just 5’ of the integrase-like motif), abolishing its interaction with DNA. However, NotchIC binding is still retained, enabling mt-RBPJ to sequester active Notch, while endogenous RBPJ retains transcriptional repression. Over-expression of mt-RBPJ therefore blocks Notch signaling. The human construct used by our lab was made on the basis of the mouse mutant108. Schematic generated at www.proteinlounge.com.

Figure 3.7 RBPJ loss accelerates DG75 xenograft tumor growth. (A) RBPJ KO Burkitt lymphoma tumors (n=10) were compared to DG75 parental tumors (n=9). P values obtained from Mann-Whitney test and data are presented as mean ± SEM. (B) Tumor lysates show absence of RBPJ protein. (C) RBPJ IHC confirms absence of RBPJ in the bulk of the tumor population in the KO xenograft compared to intermediate staining in parental tumors, which only have one functional copy of RBPJ. Positively stained cells in the KO tumor are presumed to be mouse stromal cells. Scale bar = 50 µm.

of RBPJ promotes tumor growth in multiple cancer types. Parental DG75 cells have only one functional copy of RBPJ113. This is evident in the IHC staining intensity of RBPJ protein in these tumor sections compared to the KO tumors, which showed a clear loss of RBPJ (Figure 3.7B and C).

Our work shows that depletion of RBPJ versus retained RBPJ transcriptional repression has opposite effects on xenograft tumor growth. A similar outcome is observed with constitutively active Notch signaling as with RBPJ loss, with both increasing tumor growth. In contrast restoration of RBPJ-mediated transcriptional repression or blockade of Notch receptor activation reduces tumorigenicity.

3.2.5 Evidence for RBPJ loss during breast cancer progression

Since loss of RBPJ in xenograft experiments increases the tumorigenicity of cancer cells, we evaluated whether reduced RBPJ expression correlated with more aggressive cancers. To perform this analysis, normalized expression data (from the Ginestier et al. 2006 study273) was retrieved using the Oncomine database of published microarray studies (www.Oncomine.org)286. Indeed, RBPJ expression was significantly down-regulated in grade 3 human breast carcinoma compared to grade 1 and grade 2 samples273 (Figure 3.8). As grade 3

Figure 3.8 More aggressive breast carcinomas have reduced RBPJ expression. Boxplot representation of mRNA expression in grade 1, grade 2 and grade 3 human breast carcinoma specimens shows RBPJ down-regulation in the most aggressive subset. Increasing grade represents well to poorly differentiated carcinomas. Data from Ginestier et al. 2006273 was acquired from the Oncomine database286. P value obtained from Kruskal-Wallis one-way analysis of variance test comparing RBPJ mRNA expression between the three breast cancer grades.

cancers are more poorly differentiated, this provides evidence for RBPJ loss during cancer progression.

3.3 DISCUSSION

The presence of RBPJ protein was detected by staining a collection of human tissues, agreeing with previously published findings that RBPJ is ubiquitously expressed in most cell types. Indeed, Notch signaling has been shown to control stem cell proliferation in the crypts of the intestine287 as well as the hematopoietic system64,288 and the mammary gland65, so it is no surprise that RBPJ protein is expressed here. However, whereas Notch receptor activation would be expected to occur in select cells, RBPJ should be present in most normal cells at constant levels to mediate transcriptional repression in the absence of signal.

RBPJ genomic loss was frequent in a large collection of human cancer cell lines. Partial or complete loss of chromosome 4 has been reported in breast cancer and several other cancer types133,134. The RBPJ gene is located at 4p15.2, and a smaller region of loss encompassing this location has also been observed in 66.7% (4p15.1-4p16.3) of breast cancer cell lines evaluated289 and in >30% (4p15.2 –p15.31) of breast cancer samples290. This implies the presence of multiple tumor suppressor loci in this region that remain to be identified. Furthermore, RBPJ genomic retention may not be indicative of the presence of RBPJ protein as epigenetic mechanisms can also lead to loss of RBPJ expression.

We could not segregate RBPJ loss to a breast cancer histological type as the vast majority of breast cancer cell lines are derived from ductal carcinomas and the pathology reports were

not available for the breast cancer TMA. However, Oncomine analysis suggests that RBPJ loss may be more common in lobular rather than ductal carcinomas, and the single lobular breast cancer cell line evaluated (MDA-MB-330) showed RBPJ loss at both the genomic and protein level (data not shown). Lobular carcinomas are more often ER and PR positive compared to ductal carcinomas23, however they tend to be HER2 negative291. None of the breast cancer cell lines in Table 3.1 are derived from lobular carcinomas although RBPJ copy loss is more frequent in HER2 expressing cell lines. For the breast cancer TMA cores, HER2 staining remains to be evaluated; however absence of RBPJ staining occurred more frequently in ER+ samples. Multiple datasets in Oncomine show significant down-regulation of RBPJ mRNA in hormone receptor positive breast cancers, including 6 studies showing association with ER+ breast cancers, 2 with PR+ status and 1 with HER2+ tumors (data not shown). RBPJ under- expression in ER, PR and HER2 positive breast tumors indicates RBPJ loss may be more common in luminal and HER2+ molecular subtypes of breast cancer, although RBPJ gene loss is also detected in a high proportion of ER, PR and/or HER2 receptor negative breast cancer cell lines. Breast cancer subtypes have been proposed to represent either a stage of developmental arrest following a transformation event that originates earlier in the hierarchy, or transformation of a cell type at one specific stage of mammary development29. Both luminal A, luminal B and HER2 expressing tumors show enrichment of a mature luminal cell gene expression signature29. Perhaps RBPJ inactivation occurs more frequently upon transformation of a differentiated cancer cell and contributes to acquisition of self-renewal mechanisms.

Interestingly, HER2, (ERBB2) is a direct Notch/RBPJ target196. However we could not detect changes in HER2 expression upon either over-expression of Notch1IC or RBPJ shRNA in MDA- MB-231 cells. Although MDA-MB-231 cells are ER, PR and HER2 negative we used them in our breast cancer xenograft model to evaluate the effect of RBPJ depletion on tumor growth. These cells have a normal RBPJ copy number and express RBPJ at high levels so we could access the outcome of RBPJ removal. Reconstitution of RBPJ in a HER2+ RBPJ-deficient breast cancer cell line would best be evaluated in the absence of Notch receptor activity. This would enable RBPJ to function as a transcriptional repressor so its ability to inhibit tumorigenicity could be tested.

In comparison Notch receptor signaling activity is associated with a basal type of breast cancer (ER, PR, HER-)89,90. However, Notch1IC and HEY1 protein are expressed in breast cancer cell lines such as MCF-7 (ER/PR+) and Sk-Br-3 (HER2+)76 that show RBPJ copy deletion. Furthermore, MCF-7 cells have high mRNA levels of target genes HES1 and HEY2

compared to the MCF10A breast epithelial cell line which is known to have no Notch activity due to high expression of Numb217. As RBPJ is constitutively expressed and Notch is the limiting factor in the nucleus, cells with partial RBPJ loss may have activation of RBPJ dependent HES and HEY promoters via NotchIC and/or though loss of RBPJ-mediated transcriptional repression. The ability of RBPJ to act as a transcriptional repressor would depend on nuclear levels of both RBPJ protein and active cleaved Notch.

In lung and breast cancer cells, RBPJ deletion consistently corresponded to reduced mRNA expression. Amplification did occur, although it was less frequent. However, in the smaller primary lung cancer dataset RBPJ amplification did correspond to significantly increased RBPJ expression over samples with normal RBPJ copy number. This was in contrast to the lung cancer cell lines where RBPJ amplification was not correlated with increased transcript, indicating elevated RBPJ expression levels are not favored. As over-expression of wild-type RBPJ has a repressive effect148, similar to forced expression of mt-RBPJ, a possible explanation for this up-regulation is to facilitate hyperactive Notch signaling potentially present in this cancer subset or as a negative feedback mechanism.

While Notch can function as an oncogene in the context of breast cancer and other cancers33, these findings imply that RBPJ may play a tumor suppressor role. RBPJ deficiency increased xenograft tumor growth of both an epithelial and hematopoietic cancer cell line. Moreover, evidence for RBPJ loss exists in primary human breast and lung cancer in addition to a number of other human cancer types (Oncomine data, not shown). Although this work focuses on breast cancer as a model system, RBPJ loss appears relevant across other malignancies.

CHAPTER 4: RBPJ DEFICIENCY CAUSES INDUCTION OF NOTCH TARGET GENES

4.1 INTRODUCTION

The previous chapter has shown a causative role for RBPJ loss in tumor promotion. As RBPJ is defined as a transcriptional regulator we wished to evaluate whether deregulated gene induction occurs upon RBPJ removal and contributes to oncogenicity.

In the absence of Notch interaction, RBPJ is thought to enter the nucleus precommitted to a transcriptional repressor function118. Over-expression of wild-type RBPJ inhibits target gene expression148 and RBPJ or co-repressor depletion has previously been reported to cause derepression at target gene promoters86,132,137. This indicates that although NotchIC may provide a transcriptional activation domain at some RBPJ-bound promoters, at other gene promoters transcriptional activation may occur indirectly, simply through blocked interaction of RBPJ with co-repressors50,123. To assess whether transcriptional changes contribute to increased tumor growth with RBPJ loss we evaluated expression of canonical RBPJ-target genes upon RBPJ removal.

The members of the HEY family of basic helix-loop-helix transcriptional repressors, HEY1, HEY2 and HEYL, are among the best characterized Notch/RBPJ target genes54,55. Indeed, our lab has shown positive expression correlations with the Notch ligand (Jagged1) and each of HEY1, HEY2, and HEYL in human breast cancer88. Since retention of RBPJ repressive function (via over-expression of mt-RBPJ) inhibited tumor growth, we hypothesized that RBPJ deficiency would abolish transcriptional repression at some target gene promoters, resulting in induction (or derepression) of target genes.

4.2 RESULTS

4.2.1 HEY induction occurs upon RBPJ removal and is Notch independent

To assess the effect of RBPJ removal on target gene promoters we evaluated expression of HEY1, HEY2, and HEYL. Knock-down of RBPJ in MDA-MB-231 tumor xenografts resulted in significantly up-regulated HEY gene expression compared to control tumors (Figure 4.1A). Forced expression of the mouse mt-RBPJ has previously been reported to inhibit HEY gene expression148. Similarly, over-expression of human mt-RBPJ (the counterpart to mouse mt- RBPJ) tended to decrease HEY gene expression in MDA-MB-231 xenograft tumors (Figure 4.1B). Not very high levels of mt-RBPJ were retained in endpoint tumors (<5-fold over-

Figure 4.1 HEY gene induction occurs with RBPJ removal. qPCR analysis of gene expression in MDA-MB-231 tumors. Data are expressed as fold-change relative to control which corresponds to 1 along the y axis. (A) Tumors expressing shRBPJ (n=15) had up-regulated HEY transcripts compared to shRandom tumors (n=14). RBPJ mRNA was depleted by 90% in KD tumors compared to controls at the endpoint. (B) LNC-mt-RBPJ tumors (that retain transcriptional repression) show a trend for decreased HEY gene expression compared to LNCX control tumors (n=8 for each group). P values obtained from Mann-Whitney tests and data are presented as mean ± SEM.

expressed by qPCR) which may be the reason expression of HEY gene is not significantly inhibited. These data show that removing RBPJ-mediated repression and restoring it via expression of the mutated protein has opposite effects on target gene expression; induction and inhibition of transcription respectively.

Next, we assessed whether Notch activity contributes to gene derepression in the absence of RBPJ by blocking Notch activity in vitro. We used two different strategies for pan-Notch inhibition: (1) DAPT, a highly specific γ-secretase inhibitor91 that prevents Notch receptor cleavage or (2) dominant-negative MAML, which binds NotchIC and RBPJ, but cannot recruit the co-activator p300, and thus inhibits transcriptional activation172. Neither of these approaches had an effect on target gene transcript levels, which were low initially in parental MDA-MB-231 cells. This indicates that MDA-MB-231 cells do not have active basal Notch signaling in culture (which is supported by Han et al.91), as is observed in vivo. However, the MDA-MB-231 cell line expresses Notch receptors88,90, and these can be activated by ligand stimulation. To induce endogenous Notch signaling, MDA-MB-231 cells were co-cultured with Jagged1 ligand- expressing mouse endothelial cells (the SVEC 4-10 cell line), and gene expression was measured in the breast cancer cells using human-specific primers (Figure 4.2A). In shRandom control MDA-MB-231 cells stimulated with ligand, the significant induction of HEY1, HEY2 and 64

Figure 4.2 HEY gene induction resulting from RBPJ depletion is Notch independent. (A) A co- culture system was used to induce endogenous Notch signaling in human MDA-MB-231 cells (MDA231) by growing them together with Jagged1 ligand expressing mouse SVEC cells and using human specific primers to probe for gene expression changes in MDA-MB-231 cells. (B) Ligand stimulation with Jagged1 up-regulates HEY1, HEY2 and HEYL and this induction is abolished by blocking Notch activation using 10 µM DAPT (n=5). (C) In RBPJ KD cells the magnitude of HEY gene induction is not affected by DAPT treatment. RBPJ was depleted by at least 80% compared to the shRandom control (n=5). Data expressed as fold change relative to shRandom + vehicle (DMSO). P values obtained from Student’s T-test and data are presented as mean ± SDM.

HEYL expression indicates that these genes are targets of Notch activation (Figure 4.2B). Consistent with this, HEY expression resulting from ligand stimulation was abolished upon treatment of these cells with DAPT (Figure 4.2B). Knock-down of RBPJ in ligand-stimulated cells caused a significant up-regulation of HEY1, HEY2 and HEYL (Figure 4.2C). However, DAPT treatment of unstimulated or ligand-stimulated RBPJ KD cells had no effect on the extent of HEY gene derepression (Figure 4.2C), indicating that Notch receptor activity does not play a role in RBPJ target gene induction in the absence of RBPJ.

We confirmed these findings in the DG75 lymphoma KO system. Similar to our observations in MDA-MB-231 xenografts, complete loss of RBPJ resulted in significantly increased HEY1 and HEY2 transcript levels in DG75 tumors compared to RBPJ-containing controls (Figure 4.3A). Culture of DG75 suspension cells in vitro showed RBPJ KO cells have significantly up-regulated HEY1 and HEY2 expression, and treatment of these cells with DAPT did not affect the

Figure 4.3 Complete loss of RBPJ causes a Notch-independent HEY1 and HEY2 induction. (A) DG75 RBPJ KO tumors (n=10) had increased expression of HEY1 and HEY2, compared to parental control tumors (n=9) to which data is normalized (represented by 1 along the y-axis). P values obtained from Mann-Whitney test and data are presented as mean ± SEM. (B) qPCR analysis of cultured DG75 cells treated with vehicle control (DMSO) or DAPT (n=5). Data are expressed as fold change relative to parental + vehicle. P values obtained from Student’s T-test and data are presented as mean ± SDM. (C) Immunoblots using an antibody that detects exposed valine 1744 of endogenous intracellular Notch1 produced upon γ-secretase cleavage292, but does not recognize other cleavage products or full length Notch receptors. Notch cleavage is not detected in the DG75 cell lines under any treatment condition. However, DAPT inhibited Notch cleavage in the HPB-ALL T-leukemia cell line, which has active Notch signaling.

magnitude of HEY gene induction (Figure 4.3B). Moreover, cleaved Notch1 was not detected even in vehicle control treated DG75 cells (Figure 4.3C), suggesting that gene derepression occurs independently of Notch activity.

4.2.2 A negative correlation exists between HEY2 and RBPJ expression in human breast cancer

Of all evaluated target genes, HEY2 consistently showed the biggest induction upon RBPJ removal. We therefore used Oncomine to analyze whether a negative relationship exists between RBPJ and HEY2 mRNA expression in published tumor microarray data. In the Ginestier et al. 2006273 study (described in Figure 3.8), RBPJ and HEY2 were significantly negatively correlated only in grade 3 but not grade 1 or grade 2 breast cancers (Figure 4.4A), indicating that in the most aggressive tumors, lower RBPJ transcript levels are associated with higher HEY2 expression. Using data from another study (Zhao et al. 2004274) we found RBPJ

was expressed at lower levels in an ILC versus IDC type of breast carcinoma (Figure 4.4B). Furthermore, only in the ILC, and not the IDC, histological group was RBPJ also negatively correlated with HEY2 (Figure 4.4C), providing further support for HEY2 as a potential marker of RBPJ loss.

Figure 4.4 Evidence that RBPJ and HEY2 mRNA expression are negatively correlated in human breast cancers. (A) Data from Ginestier et al. 2006273 shows that a significant RBPJ versus HEY2 negative correlation exists only in grade 3 but not grade 2 breast carcinomas. For grade 1 specimens the n=4 was too small for correlation analysis. Boxplots of these data are shown in Figure 3.8 in the previous chapter. (B) Boxplot representation of data from Zhao et al. 2004274. Although not significant, RBPJ expression is decreased in the ILC compared to IDC histological subtype. P value obtained from Mann- Whitney tests. (C) A negative correlation between RBPJ and HEY2 exists only in the ILC samples and not the IDC sample. Correlation P values in (A) and (C) obtained from the Pearson’s correlation test.

4.2.3 The outcome of modified HEY transcript levels on xenograft tumor growth

To determine the outcome of excessive HEY2 levels on xenograft tumor growth we used a green fluorescent protein (GFP) expression vector (MIG) to over-express HEY2 (MIG-HEY2) in MDA-MB-231 breast cancer cells. Analysis of gene expression in stably transduced cell lines shows very high levels of HEY2 are attained (~3500 fold), but this results in down-regulation of HEY1, HEYL and RBPJ expression (Figure 4.5A). Seeding of tumor cells with MIG-HEY2 over- expression into NOD/SCID mice inhibited rather than increased tumor growth (Figure 4.5B). However, this system is not directly comparable to the up-regulated endogenous HEY2 expression that occurs upon RBPJ depletion, as much higher levels of HEY2 result from HEY2 over-expression. Also, HEY2 was over-expressed in the presence of RBPJ rather than being up-regulated in its absence. HEY2 can form homodimers with other HEY family members, as well as heterodimers with RBPJ itself54,55. HEY2 association with RBPJ enhances the transcriptional repressor function of RBPJ221, negatively affecting HEY gene expression. Indeed, HEY1 and HEYL transcripts were both down-regulated upon over-expression of HEY2, supporting the occurrence of this feedback. Interestingly, expression of RBPJ itself is inhibited with MIG-HEY2 (Figure 4.5A). Although it has not been reported that RBPJ regulates its own expression, there are at least two predicted RBPJ binding motifs proximal to the TSS in the RBPJ promoter. Therefore, elevated HEY2 levels may cause HEY2/RBPJ dimer-mediated repression of RBPJ expression.

To better understand the role of HEY2 downstream of RBPJ loss, we depleted HEY2 transcript using shRNA in DG75 lymphoma RBPJ KO cells and compared them to HEY2- containing RBPJ KO cells. HEY2 expression was knocked-down to 10% of that present in KO cells with no resulting effect on HEY1 transcript levels (HEYL is not expressed in DG75 cells in vitro, Figure 4.5C). Upon subcutaneous implantation of these cells into immunocompromised mice no significant difference was observed in tumor growth (Figure 4.5D). However, analysis of gene expression in endpoint tumors showed KD of HEY2 was not retained in cells that were initially transduced with the shHEY2 construct. Furthermore, these tumors expressed significantly less GFP than DG75 RBPJ null controls transduced with a non-specific shRandom construct (Figure 4.5E), implying a potential selective advantage and outgrowth of cells that have lost KD of HEY2.

mFigure 4.5 Effect of HEY2 on xenograft tumor growth. (A) qPCR analysis of MDA-MB-231 cells over-expressing HEY2 (MIG-HEY2), normalized to vector control (MIG, which represents 1 along the y axis) prior to injection into NOD/SCID mice shows down-regulation of HEY1, HEY2 and RBPJ gene expression. (B) Forced HEY2 expression in MDA-MB-231 tumors inhibits tumor growth compared to vector control tumors (n=4 for each group). P value obtained from Student’s T-test. (C) HEY2 was knocked-down (shHEY2) in DG75 RBPJ KO cells to 90% of that in the DG75 RBPJ KO shRandom control to which data are normalized. (D). Depletion of HEY2 in DG75 RBPJ KO tumors does not affect tumor growth (n=6 for each group). (E) However, significant HEY2 KD was not maintained at the experimental endpoint (mean KD = 40%) and loss of GFP expressed from the shRNA construct implies RBPJ KO cells with HEY2 depletion are selected against. All bars represent mean ± SEM. 69

4.2.4 Evidence of gene derepression in cells derived from the RBPJ KO mouse

Complete and conditional Rbpj KO mice, each generated by the Tasuku Honjo group127 (see Figure 1.4C and D in the Introduction) are used in the literature with the interpretation that RBPJ removal blocks Notch signaling. From our data, RBPJ loss in cancer cells up-regulates the canonical Notch/RBPJ target genes of the HEY family. This results in a phenotype comparable to Notch signal activation, rather than impairment, despite being independent of NotchIC. We therefore obtained mouse embryonic fibroblast cells (MEFs) derived from Rbpj null 9.5 day embryos, and immortalized by simian virus 40 large tumor antigen261. This mouse Rbpj KO cell line, called OT11, has a disruption in the coding sequence beginning within the integrase-like motif in exon 7 (see Figure 1.4C in the Introduction), and was compared to its Rbpj-containing counterpart, cell line OT13 (Figure 4.6A). A comparison of cultured KO and wild-type (wt) MEFs showed increased HES1 expression in the KO cells, indicating that this particular family member may be derepressed with RBPJ loss in this context (Figure 4.6B). However, the lack of derepression of HEY1 and HEY2 in KO MEFs could be a result of subsequent adaptations that have occurred in response to RBPJ deficiency as they have been propagated. Knock-down of RBPJ in the wt MEFs, using an shRNA construct that targets mouse RBPJ (shRBPJm), resulted in derepression of not only HES1, but also HEY2 (Figure 4.6C). If derepression occurs in Rbpj KO MEFs and human RBPJ KO lymphoma cells, a similar phenomenon may be expected in

Figure 4.6 Evidence for target gene induction in Rbpj KO cells derived from the Rbpj null mouse. (A) Rbpj wt but not KO MEFs express RBPJ by immunoblot. (B and C) qPCR analysis of gene expression in wt and KO MEFs using mouse specific primers (n=2). (B) Complete loss of Rbpj results in elevated HES1 expression. (C) KD of RBPJ in Rbpj wt MEFs using a mouse specific targeting construct (shRBPJm) induces both HES1 and HEY2.

other contexts with RBPJ loss. Indeed, we have observed HEY gene derepression upon RBPJ depletion in several non-cancer cell lines (data not shown) in addition to the wt MEFs, although we have yet to test whether this occurs independently of Notch activation in these cells. However, HEY gene induction is most pronounced in MDA-MB-231 and DG75 tumor cell lines where other deregulated pathways may contribute to signal activation.

4.3 DISCUSSION

Just as aberrant Notch activation is oncogenic due to inappropriate target gene expression, RBPJ deficiency can produce a similar outcome by derepressing target genes. HEY genes are therefore useful markers not only of Notch activation, but also of potential RBPJ loss in human tumors. We are currently evaluating RBPJ and HEY gene expression in additional studies and have found evidence for negative RBPJ versus HEY expression correlations in some of the lung cancer data.

Our results show that gene derepression in the absence of RBPJ does not occur via Notch. This means that NotchIC does not interact with other factors to induce transcription. We therefore support the previously described permissive signaling model, in which Notch is required to displace co-repressors from some RBPJ-bound promoters, without being required for co-activation123. RBPJ loss, and the resulting disruption of transcriptional repression, leads to activation by a currently unknown mechanism. Gene derepression has also been reported upon loss of Ikaros, a protein which antagonizes Notch signaling by binding to the RBPJ consensus motif and inhibiting transcriptional activation4. In leukemia, Ikaros deficiency causes derepression of Ikaros-regulated Notch target genes, which occurs independently of Notch activity259,293. This agrees with our data and supports that an alternate factor provides transcriptional activation. In our experiments using cancer cells and other immortalized cell lines, aberrant activation of additional pathways may contribute to the severity of the phenotype seen with RBPJ removal.

Our experiments aimed at accessing the contribution of up-regulated HEY2 in driving tumorigenicity upon RBPJ loss were inconclusive. In fact, over-expression of HEY2 in the presence of RBPJ appears to be tumor suppressive and inhibits RBPJ target gene expression. This suggests that HEY2 is involved in a feedback loop and associates with RBPJ to dampen Notch signal, as has been reported by others221. RBPJ depletion would presumably prevent this negative feedback, via loss of RBPJ-HEY interactions. Further in vivo experiments are required 71

to access the loss of just HEY2 or both HEY2 and RBPJ in MDA-MB-231 xenografts for determining the involvement of HEY2 in tumor progression. In DG75 RBPJ null cells, depletion of HEY2 appears unfavorable, indicating that HEY genes have functions other than just negative feedback. It is speculated that HEY proteins contribute to both negative feedback, and propagating downstream signals. However, this hypothesis requires further study. As the Notch cascade subtly regulates expression of a large number of genes183, it is likely that HEY proteins alone are not mediating all the effects resulting from RBPJ depletion.

In human RBPJ null cells and cells derived from the Rbpj KO mouse gene derepression is observed. This has important implications in using the RBPJ KO system and making inferences about inhibited Notch signal. In some of these studies, conclusions about RBPJ independent Notch signaling could be explained by gene derepression13,86. Conversely, other studies have used the conditional Rbpj KO system, in addition to employing other methods to inhibit the pathway (such as deletion of O-fucosyltransferase177), and have found similar outcomes. Although we have observed that RBPJ depletion results in HEY gene derepression in a number of immortalized but non-transformed cell lines, the magnitude of derepression varies. In cells that rely on Notch receptor signal, removal of RBPJ may indeed hinder NotchIC-mediated activation of select promoters. Thus with RBPJ loss the magnitude of gene derepression, and the resulting effect, may be context dependent.

In our cancer model, RBPJ deficiency causes significant differences in tumor growth. As a transcription factor, RBPJ is predicted to bind to many promoter sequences and regulate genes in addition to those of the HEY family. In cancer cells where multiple abnormalities contribute to signal deregulation, even a modest derepression of a number of key genes could lead to tumor progression. To understand how RBPJ removal confers increased xenograft tumor growth a global analysis was necessary to better evaluate the effect of loss of transcriptional repression on cancer hallmarks.

CHAPTER 5: RBPJ LOSS PROMOTES TUMOR CELL SURVIVAL

5.1 INTRODUCTION

Although HEY genes are up-regulated in response to RBPJ loss independently of NotchIC, the contribution of these transcriptional regulators in propagating downstream oncogenic effects needs further examination. However, elimination of Drosophila E(spl) genes, which are related to mammalian HEY genes, does not mimic the wing notching phenotype caused by reduced Notch function, implying that other targets play an essential role in this process294. Likewise, in addition to HEY genes, we predicted that RBPJ depletion leads to transcriptional induction of a number of other factors which contribute to tumorigenicity.

In addition to the HEY gene family, Notch/RBPJ is known to regulate many genes183, some of which act in opposition167. This suggests that cellular outcomes are determined by complex balances in gene expression relationships, and that aberrant Notch pathway components contribute to tumorigenesis by deregulating a number of genes. Indeed, the combined action of several genes, rather than the effect of a single gene, better recapitulates tumorigenic phenotypes, such as breast cancer metastasis to lungs295. To discover mechanisms that play an oncogenic role upon RBPJ depletion, we set out to evaluate globally derepressed genes and connect them to functions that could mediate tumor growth, as the observed effects are likely not limited to HEY genes alone.

Microarrays and more recently transcriptome sequencing are methods that measure differential gene expression on a global scale. Microarray technology relies on nucleic-acid hybridization using specific probes and can miss detection of low-abundance signal296,297. Moreover, microarrays can significantly underestimate gene expression changes298. Direct sequencing of material removes experimental bias that is present when using probes and offers increased sensitivity299, however optimal analysis tools that best interpret this data are still under development.

Chromatin immunoprecipitation (ChIP) can be used to evaluate the distribution of histone modifications on DNA. The chromatin is first fragmented either by digestion with micrococcal nuclease or sonication if histone proteins are chemically cross-linked to DNA with formaldehyde (the latter approach was employed by us)299,300. Antibodies specific to the histone modifications are used to immunoprecipitate the associated DNA regions. The cross-links are then removed and the purified enriched DNA reflects where the modified histone protein was bound299. The immunoprecipitated genomic fragments can then be sequenced for identification (ChIP-seq)300.

This approach has previously been used to evaluate distribution of acetylated histone H4 (acH4), a mark associated with active promoters301.

5.2 RESULTS

5.2.1 Depletion of RBPJ results in increased epigenetic marks of promoter activity at the RBPJ binding motif in the HEY1 and HEY2 promoter.

Electrophoretic mobility shift assays (EMSAs) which measure protein-DNA interactions were used to evaluate whether the RBPJ binding sequences remain unbound in the absence of RBPJ protein (Figure 5.1A). RBPJ recognition sequences proximal to the TSS in the HEY1 and HEY2 promoters were used as labeled oligonucleotide (oligo) probes. A gel shift corresponding to RBPJ binding of the labeled probe occurred in LNCX, LNC-mt-RBPJ and shRandom MDA- MB-231 nuclear lysates, and this band was competed away using 50-fold excess unlabeled wild-type, but not mutated oligo. Depletion of RBPJ using two different targeting sequences abolished oligo binding altogether, indicating that RBPJ binding to its recognition sequence is sufficiently depleted and that the protein is not replaced by another transcription factor at this site in RBPJ KD cells (Figure 5.1A).

To evaluate transcriptional changes that occur with RBPJ depletion, microarray analysis of genome wide expression profiles were compared between triplicate experiments of cultured MDA-MB-231 RBPJ KD versus control cells (using the Affymetrix Human Exon 1.0 ST Array). However, this method was not sensitive enough to detect changes in gene expression resulting from the depletion of RBPJ protein. At a false discovery rate of 10% (which represents the percent of expected false predictions) only 9 genes were differentially expressed (including RBPJ, which was down-regulated in the KD cells as expected). Four of these genes were up- regulated only 1.5-3 fold in the KD cell by qPCR. The remaining 4 genes did not validate by qPCR, as they showed no difference in expression levels between groups. HEY genes were not identified by the microarray analysis, and upon closer evaluation it was concluded that the probes for these genes did not show a signal above the background.

Instead we evaluated how RBPJ removal affects epigenetic regulation at target promoters. Notch activation results in recruitment of histone acetyltransferases (HATs) to the RBPJ co- activation complex, and this leads to increased levels of acH4 at expressed genes132. Conversely, transcriptional repression by the RBPJ co-repressor complex depends on histone

Figure 5.1 Changes in histone marks at the HEY1 and HEY2 promoter upon KD of RBPJ in MDA- MB-231 cells. (A) EMSA showing binding of the RBPJ consensus motif is lost when RBPJ is depleted. Labeled oligos representing binding sites in the HEY1 and HEY2 promoter (HEY1: -378 to -385, and +45 to +52, HEY2: -160 to -167 and +41 to +48, also see red arrows in panel D for location of RBPJ consensus motifs) were used as probes. Nuclear lysates were incubated with labeled oligo alone, labeled oligo with 50-fold excess unlabeled mutated probe (+mt) or labeled oligo with 50-fold excess wild-type probe (wt). ChIPs for (B) and (C) were performed with the indicated antibodies and bound fragments were quantified by qPCR evaluating two sites in the HEY2 promoter: the closest upstream RBPJ binding site (-160 bp from the TSS) and the TSS. Bars represent mean ± SDM. Enrichment levels were normalized to the IgG isotype control within each group. Both acH4 (B) and H3K4me3 (C) show enrichment at the RBPJ binding site when RBPJ is depleted. (D) ChIP-seq profile graphics of acH4 enrichment in a region 2500 bp upstream and 1500 bp downstream from the TSS in the HEY1 and HEY2 promoter. acH4 peaks in KD cells are shown in purple, in the control cells in light green and overlap between the two peaks in dark green with red arrows representing high scoring RBPJ motif matches. An area ratio of enrichment (R) in RBPJ KD versus control was calculated (see text) and ranked from +1 to - 1, with positive R-values representing acH4 enrichment in the KD cells.

deacetylase (HDAC) activity135,136,149, which leads to chromatin condensation and silencing of gene expression. It has been suggested that gene repression is mediated by intermittent co- repressor activity, rather than constitutive presence of co-repressors302. In this model permissive signaling occurs when co-repressors cannot be recruited to condense chromatin302. We postulated that RBPJ removal, and therefore disruption of the co-repressor complex, would result in loss of HDACs and concomitant acetylation of histones, as has been reported for KD of RBPJ and induction of latent HIV161. We performed acH4 ChIP, followed by qPCR at the HEY2 promoter, to validate this hypothesis. Enrichment of acH4 occurred at the RBPJ binding site upstream of the HEY2 TSS in RBPJ depleted cells compared to RBPJ-containing control cells (Figure 5.1B). As histone H3 lysine 4 demethylation also plays an important role in RBPJ- mediated gene silencing137, and Notch activation increases the levels of the active chromatin mark H3K4me3 at target genes132, we evaluated H3K4me3 enrichment in RBPJ-KD MDA-MB- 231 cells. As with acH4, H3K4me3 was increased in RBPJ KD cells at the RBPJ binding site in the HEY2 promoter (Figure 5.1C).

These initial results imply that RBPJ deficiency results in accumulation of active marks at well characterized target promoters. To evaluate changes at other promoters, and identify genes that are potentially induced by RBPJ removal in MDA-MB-231 cells, we performed a global acH4 enrichment analysis using ChIP followed by Illumina sequencing (ChIP-seq) to identify genomic regions (and effected target genes) that are associated with this modification.

To discover target genes with differential acH4 enrichment in RBPJ KD versus control cells we evaluated regions that were 2500 bp upstream and 1500 bp downstream relative to the RefSeq TSS’s (-2500/+1500 bp). For each transcript we calculated a ratio of control to KD areas 77

under the acH4 enrichment profile using the formula R=(Akd-Ac)/(Akd+Ac), where Akd (area KD) and Ac (area control) were areas under the enrichment profiles for aligned Illumina reads that were directionally extended to a length of 200 bp. The ratio varied from -1 to +1, corresponding to H4ac enrichment in the control and in the KD, respectively. We ranked genes by the ratio of control to KD areas and for genes with multiple transcripts we retained the transcript with the largest absolute value ratio. To identify potentially induced genes, we evaluated genes with the 1000 highest and lowest ranks. HEY1 and HEY2, our positive controls, were ranked in the top 100 R-values and both showed acH4 enrichment close to the TSS in RBPJ KD MDA-MB-231 cells compared to controls (Figure 5.1D) confirming our previous ChIP-qPCR results (Figure 5.1B). We further validated our method by evaluating a number of other targets in the top 1000 ranked gene list to verify the analysis and correlate enrichment of acH4 peaks to increased gene expression. Figure 5.2 shows acH4 profile graphics and qPCR analysis of validated genes other than those of the HEY family. These data suggest that induced genes are marked by acH4. As some of these genes are known direct RBPJ targets, their induction is likely due to derepression caused by loss of RBPJ rather than as a secondary effect resulting from other up-regulated genes. As the acH4 mark is associated with active promoters we used the global acH4 ChIP-seq data as an alternate approach to microarrays to identify differentially active processes in RBPJ depleted cells.

5.2.2 A Notch-like signal is generated in RBPJ deficient cells

To evaluate whether direct Notch/RBPJ target genes were overrepresented in the top ranked acH4 gene list (associated with acH4 enrichment in KD cells), we used a manually compiled list of 31 direct Notch/RBPJ target genes published in the literature (listed in Table 1.2 of the Introduction) and the Ingenuity Pathway Analysis (IPA) software (Ingenuity® Systems, www.ingenuity.com). The IPA tool enables analysis of a gene list for insight into biological mechanisms, pathways and functions associated with the data. A right-tailed Fisher’s exact test was used to calculate a P value to determine the probability that enrichment of our custom list is due to chance alone. When evaluating the significance of a single feature in relation to the dataset, this uncorrected P value is an appropriate way to measure the likelihood of the observed result if the association is random. We found 7 direct Notch/RBPJ target genes (23%) were represented in the top 1000 ranked area ratios (P value = 0.00384) compared to 2 genes (7%) present in the bottom negative 1000 ranked list (P value = 0.399, Figure 5.3). Enrichment of RBPJ/Notch target genes indicates that a Notch-like signal is generated in cells when RBPJ is depleted, even though this derepression is independent of Notch.

Figure 5.2 Correlation of acH4 enrichment with up-regulated gene expression in MDA-MB-231 RBPJ KD cells. (A) ChIP-seq profile graphics of acH4 enrichment in a region -2500/+1500 bp from the TSS. Gene promoters with acH4 enrichment in the RBPJ KD over control MDA-MB-231 sample are shown. acH4 peaks in KD cells are shown in purple, in control cells in light green and overlap between the two peaks in dark green with red arrows representing high scoring RBPJ motif matches. The motif used for this work is show in Figure 1.4B of the Introduction. (B) cDNA was made from the same batch of cells as was used for acH4 ChIP analysis and qPCR was performed to evaluate gene expression of high ranking promoter regions. R-values were used to rank genes. For 9 of the 12 genes acH4 enrichment corresponded to increased mRNA expression greater than 1.5-fold that of the non-targeting shRandom control (to which data are normalized). For these experiments RBPJ was knocked-down by 90% compared to control cells.

5.2.3 RBPJ depletion confers a survival advantage

To discover how globally up-regulated genes (marked with acH4) contribute to increased tumorigenicity, we identified statistically significant biological functions by comparing the top and bottom ranked genes from the acH4 enrichment analysis in IPA. We used a P value of <0.01 with a Benjamini-Hochberg (B-H) multiple testing correction303 (accepting a 1% false discovery rate). This correction is relevant when performing multiple comparisons for a set of functions, in other words testing many hypotheses, as it is necessary to control for obtaining false positives for the entire collection of tests. The analysis returned nine significant cellular functions shown in Figure 5.4. Of these, we chose to focus on cell death and cellular growth and proliferation, which were third and sixth ranked respectively, as both of these processes could directly impact primary tumor volume. There is also evidence in the literature that Notch signaling contributes to both survival and proliferation of cancer cells304. As we are using a xenograft model in immunocompromised mice we did not pursue antigen presentation (top ranked). Preliminary findings for cell movement, the third most significant function, are covered in the discussion of this section.

For the cell death biofunction, the top R-positive acH4 group (enriched in RBPJ KD) had 200 genes within this function (B-H P value of 0.00298 to 0.07166), compared to the bottom R negative acH4 (enriched in control) group with 147 genes in this category (B-H p-value of 0.159 to 0.223, Figure 5.4). In the R-positive acH4 set 196 genes mapped to cellular growth and proliferation (B-H P value of 0.00513 to 0.0716), versus the bottom R-negative acH4 group which contained 146 genes in this category (B-H P value 0.0621 to 0.223, Figure 5.4). To validate whether cell death and growth and proliferation functions were altered in RBPJ- deficient cells, as predicted by IPA, we assessed these cancer hallmarks in vitro and in vivo.

Figure 5.3 RBPJ target genes are enriched in the top 1000 ranked acH4 gene list in MDA-MB-231 cells. Ingenuity analysis was performed using a custom gene list containing the 31 direct Notch/RBPJ target genes shown in Table 1.2. Top 1000 ranked genes (dark blue bars) represent genes with acH4 enrichment in RBPJ KD cells versus bottom 1000 ranked genes (light blue bars), which show acH4 enrichment in shRandom control cells. A threshold denoting a right-tailed Fisher’s exact P value of 0.01 is shown by the orange line. Seven direct RBPJ target genes overlap with acH4 enriched promoters in RBPJ KD cells (CCND3, HES1, HEY1, HEY2, NFKB2, NRARP, TNC) versus two that overlap with acH4 enriched promoters in shRandom control cells (EFNB2, PDGFRB).

Figure 5.4 Identification of differentially enriched cellular functions. Functions were evaluated for top (enriched in KD) and bottom (enriched in control) 1000 ranked acH4 gene lists (dark blue and light blue bars respectively) in MDA-MB-231 cells using IPA. The 9 significant cellular functions resulting from a threshold B-H P value of 0.01 (denoted by orange line) are shown. 81

Differences in cell death were evaluated by incubation of the adherent MDA-MB-231 epithelial cell lines in suspension for 48 h to induce anoikis. RBPJ KD cells were found to be markedly protected from cell death compared to control cells (Figure 5.5A). Because the DG75 lymphoma cell line normally grows in suspension culture, we used tert-butyl hydroperoxide (TBHP), which generates reactive oxygen species (ROS), to induce apoptosis in this system. RBPJ KO cells showed a significant resistance to death in the presence of this organic peroxide compared to parental cells (Figure 5.5B). Interestingly, both MDA-MB-231 RBPJ KD and DG75 RBPJ KO cells were sensitive to serum starvation, showing increased cell death under these

Figure 5.5 Validation of the cellular death function in vitro. (A) Anoikis was evaluated in MDA-MB- 231 cells following forced suspension for 48 h. The surviving fraction was identified by measuring the AnnexinV/Propidium Iodide (PI)-negative population by flow cytometry. Data are normalized to healthy attached cells within each respective group (n=4). (B) DG75 cells were treated with 100 µM TBHP, an organic peroxide, and the surviving fraction was evaluated by AnnexinV/PI staining. Data are normalized to untreated cells within each respective group (n=3). (C and D) Resistance to serum starvation was assessed by quantifying the AnnexinV/PI negative population using flow cytometry. Data are normalized to cells grown in serum containing medium within each group. (C) MDA-MB-231 RBPJ KO cells cultured without serum for 72 h show decreased survival (n=3). (D) RBPJ KO DG75 cells are also sensitive to serum free conditions following a 48 h incubation (n=3). P values for all panels were obtained from Student’s T-test and data are presented as mean ± SDM. 82

conditions (Figure 5.5C and D). This suggests that factors present in the serum (or blood in vivo) are required to mediate the pro-survival effects resulting from RBPJ depletion.

Next, we used terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) on whole tumor cryosections to assess whether the same survival advantage with RBPJ deficiency was present in vivo (Figure 5.6 shows representative staining images). MDA-MB-231 size- matched control tumors (collected when they reached a comparable size to the KD) showed significantly more TUNEL staining than both time-matched control tumors (harvested at the same time as the KD) and RBPJ-depleted tumors (Figure 5.6A and Figure 5.7A). Rather than evaluating TUNEL throughout whole sections, we cropped out TUNEL-positive necrotic areas prior to determining staining in viable MDA-MB-231 tumor tissue. Comparable results were obtained in this modified analysis, with both time- and size-matched control tumors showing significantly more TUNEL staining than RBPJ depleted tumors (Figure 5.7B). Moreover, we evaluated whether cell death was more prevalent in certain regions of the tumor tissue. We found that the difference in the proportion of TUNEL positive staining in RBPJ KD tumors was especially pronounced at further distances from blood vessels where hypoxia is highest (Figure 5.7C). Analysis of DG75 KO tumors is consistent with the findings in MDA-MB-231 xenografts. RBPJ null tumor cells showed significantly reduced tumor cell death (Figure 5.6B and Figure 5.7D), and the difference compared to DG75 parental tumors was more pronounced when TUNEL was evaluated in viable tumor tissue only, or at increasing distances from the vasculature (Figure 5.7E and F). These data suggest that with RBPJ loss, increased survival of tumor cells in vivo contributes to increased tumor volume.

Cellular growth and proliferation represents the sixth highest ranked function in the IPA analysis shown in Figure 5.4, implying that MDA-MB-231 RBPJ KD cells, from which this data are derived, may also have proliferative advantages. To evaluate cell growth in vitro we generated growth curves by manual cell counting. MDA-MB-231 cell lines did not show a difference in proliferation rates (Figure 5.8A). Furthermore, DG75 RBPJ KO cells actually grow significantly slower than DG75 parental controls in culture (Figure 5.8B). This is in contrast to substantially increased tumor growth evident in vivo.

To measure proliferation in tumor cryosections we stained for incorporated bromodeoxyuridine (BrdU), a marker of cell division. RBPJ KD MDA-MB-231 xenografts had significantly more BrdU positive staining than their time-matched or size-matched controls (Figure 5.8C), indicating that RBPJ KD tumors grow faster by upregulating both survival and growth in vivo. BrdU staining in RBPJ depleted breast tumors was only increased at distances 83

Figure 5.6 Staining and quantitative mapping of multiple markers in whole tumor cryosections. Dying cells are stained with TUNEL (red), hypoxic areas with Pimonidazole (green), S-phase proliferating cells with BrdU (black) and the vasculature with CD31 (blue). (A) MDA-MB-231 time-matched shRandom control tumors were harvested on the same day as shRBPJ tumors. Size-matched control tumors were collected once they reached a comparable size to RBPJ KD tumors. All scale bars = 500 µm. (B) Staining of whole DG75 tumor cryosections as described above. Parental tumors were harvested when they reached a comparable size to RBPJ KO tumors. All scale bars = 300 µm.

Figure 5.7 RBPJ deficient tumors have reduced cell death. Tumor cryosections shown in Figure 5.6 were analyzed by staining and quantitatively mapping multiple markers. Cell death was evaluated in MDA-MB-231 cells as the fraction of TUNEL positive pixels (n=7 for each group), and is reported either as fraction of total number of pixels in tumor sections, including confluent necrosis (A), or as a fraction of the viable tissue only, from sections where confluent necrosis had been cropped and removed (B). (C) TUNEL labeling of apoptosis and pimonidazole staining of hypoxic cells was evaluated as a function of the distance from nearest CD31 stained vasculature. Lines represent averaged data for each group. Solid lines are TUNEL staining and dashed line are pimonidazole data (n=7 for each group). Evaluation of cell death in DG75 xenograft tumor cryosections (n=7 for each group) via quantification of total TUNEL pixels (D), or TUNEL in viable tissue only (E). (F) TUNEL and pimonidazole distance from the vasculature charts were generated for DG75 tumors as described in C. For TUNEL data n=7 for each group. For pimonidazole data n=5 for RBPJ KO and n=4 for RBPJ parental. For (A), (B), (D) and (E) P values were obtained from Mann-Whitney tests and bars are presented as mean ± SEM.

Figure 5.8 Differential effect of RBPJ removal on tumor cell proliferation in vitro and in vivo. (A) No difference is observed in MDA-MB-231 proliferation between shRandom versus shRBPJ cells in culture. Data normalized to day 2 counts within each group (n=4). (B) DG75 RBPJ KO cells have reduced growth in culture (n=3). (C) Evaluation of BrdU positive staining in MDA-MB-231 viable tumor tissue (refer to Figure 5.6 stained images) indicates increased proliferation is present in RBPJ KD tumors (n=7 for each group). (D) DG75 tumors had comparable BrdU staining between groups (n=5 for RBPJ-containing parental tumors and n=4 for RBPJ KO tumors). P values for (A) and (B) obtained from Student’s T-test and data are presented as mean ± SDM. P values for (C) and (D) obtained from Mann-Whitney tests and data are presented as mean ± SEM.

proximal to the blood vessels and not in hypoxic tumor tissue (data not shown). Although the IPA analysis was performed with data generated from cultured MDA-MB-231 cells, our inability to detect differences in proliferation in vitro could be a result of growth factors whose up- regulation only confers an advantage within the 3D tumor environment. There was no difference in BrdU staining in DG75 tumors (Figure 5.8D). Indeed, from the in vitro growth curves it may be expected that RBPJ null cells actually show reduced proliferation in vivo. Similar to the MDA- MB-231 cells, this implies that the tumor environment provides important unknown factors that alter the response to RBPJ loss.

Using pathway analysis we identified cellular death and growth and proliferation among the significantly up-regulated functions in MDA-MB-231 RBPJ KD cells. We further validated these cancer hallmarks using cell culture assays and by staining for markers in xenograft tumor cryosections. RBPJ deficiency in both MDA-MB-231 and DG75 cells enables cancer cell survival in vitro and in vitro.

These data imply that protection from cell death is a contributor to increased tumor volume seen with RBPJ removal. However, RBPJ loss sensitizes cell to serum starvation. Furthermore, MDA-MB-231 RBPJ KD tumors showed a proliferative advantage in vivo, however, this phenotype was not observed in culture. In fact DG75 RBPJ KO cells had inhibited growth in vitro compared to parental RBPJ-containing controls. Taken together these data imply that stimuli present in the tumor microenvironment promote oncogenicity of these cells. Consistent with this we observed especially pronounced differences in tumor survival with RBPJ deficiency in hypoxic tumor tissue.

5.3 DISCUSSION

This chapter of the thesis provides further insight into how RBPJ deficiency enables larger tumor growth at the cellular level, with future work aimed at continued functional validation of specific genes that mediate this effect. At the molecular level, it can be inferred that as a result of RBPJ removal, co-repressor complex formation does not occur, preventing HDAC recruitment and condensation of chromatin. This leads to acH4 enrichment and transcriptional activation, presumably due to the presence of other bound transactivators.

RBPJ deficiency can also lead to reduced recruitment of histone demethylases, with a resulting increase in H3K4me3. KDM5A has been shown to directly associate with the RBPJ transcriptional repressor complex, and RBPJ KD abrogates recruitment of KDM5A to target promoters137. Depletion of KDM5A itself increases H3K4me3 at RBPJ-binding sites137. We chose to focus on histone acetylation rather than methylation due to the better validated connection between RBPJ transcriptional regulation and HDAC and HAT activity. However, it seems likely that gene induction with RBPJ loss is associated with increases in both histone methylation and acetylation.

We performed a genome-wide analysis of acH4, an epigenetic mark of promoter activity, in RBPJ KD versus control breast cancer cells, to determine how RBPJ deficiency impacts on 87

epigenetic modifications at target genes. We also used enrichment of acH4 as an indirect measure of promoter activity to identify processes that were differentially active in the RBPJ KD versus control breast cancer cells, and validate them across the RBPJ KO lymphoma system. In the case for cell death, RBPJ deficiency conferred protection from apoptotic stimuli in culture, and resulted in less death in tumor tissue. The exception was the sensitivity of both KD and KO cell lines to serum starvation, suggesting that a signal contained in serum bestows advantages to these cells. This is further supported by the observation that MDA-MB-231 RBPJ KD cells showed no growth advantage in vitro but had increased BrdU staining in the tumor population. Moreover, DG75 KO cells have a growth disadvantage in vitro but show a much more tumorigenic phenotype in vivo, pointing to an essential role of the microenvironment. Factors contained in the serum may activate pathways that synergize with RBPJ loss and enable gene induction in the absence of RBPJ.

Interestingly, we observed an especially pronounced difference in survival in hypoxic regions between control and RBPJ depleted tumor cells. Hypoxia can be caused by increased diffusion distances from the vasculature and is expected to occur at distances above 70 µm from a blood vessel wall305. In addition to receiving less oxygen, tumor cells that are further from the blood vessels also are depleted of nutrients. As RBPJ loss sensitizes cells to serum starvation we were surprised to observe that the most pronounced survival advantage occurred within regions that are distant from the blood supply. However, a factor present in both serum in vitro and hypoxic regions in vivo may enable RBPJ survival and provide the transcriptional activation required to induce genes in the absence of RBPJ. These data further supported the ability of RBPJ loss to protect cells from death under stress conditions such as the harsh tumor hypoxic environment that can select for survival of highly aggressive cells9. Our work showing RBPJ loss promotes cell survival is supported by published observations discussed in chapter 6.

The EMSA experiments demonstrate that the core RBPJ recognition motif remains unbound upon RBPJ depletion as all bands are out-competed by unlabelled wild-type (TGGGAA) and mutant (TGCTGC) oligos. Thus, at least at HEY genes promoters, binding by a transactivator occurs outside the RBPJ consensus. A global analysis of RBPJ target promoters was also attempted by chromatin Immunoprecipitation (ChIP), but was not successful due to lack of suitable antibodies.

An additional function pursued, but only tested in vitro, was migration. The IPA “cell movement” function encompasses both the process of migration and invasion. RBPJ depletion in MDA-MB-231 cells did not result in increased cell motility on plastic. Upon monitoring the 88

closure of a scratch wound in a confluent monolayer of cells, no advantage was observed between RBPJ KD and control cells (data not shown). A number of matrix metalloproteinases (MMPs) appeared among the genes in the listed cellular functions. MMPs are proteolytic enzymes that are involved in degradation of proteins, including those found in the extracellular matrix, and are best known for their role in invasion306. Although MMPs have not been validated as direct targets of Notch/RBPJ, an active Notch4 enables mammary epithelial cell invasion into a collagen gel. Use of a pan-MMP inhibitor suppressed matrix invasion of these cells in a dose- dependent manner87. This is consistent with reports that KD of Notch1 and Notch4 inhibit invasion of MDA-MB-231 cells though Matrigel, while over-expression of Notch1IC in the MCF7 breast cancer cell line enhances invasion in this assay90. These studies and the induction of MMPs point to a potential invasive advantage. However, we detected no difference in the ability of MDA-MB-231 cells to invade Matrigel in a Boyden chamber assay where cells were plated in serum free medium and serum containing medium was used as a chemoattractant. Further work is needed in this area as Matrigel may be the wrong substrate for these MMPs. Matrigel is predominantly composed of laminin307. As most of the MMPs identified by differential acH4 analysis are collagenases, collagen or gelatin, which is derived from collagen, may be a better basement membrane component for use in these experiments. In addition to elevated MMP expression we have yet to assess whether these enzymes are active.

Our work shows that RBPJ depletion enriches for acH4 at promoters of induced genes, producing a Notch-like signal via up-regulation of direct target genes. This deregulated signal contributes to carcinogenesis by conferring a survival advantage to tumor cells with RBPJ loss.

CHAPTER 6: PERSEPCTIVES, FUTURE DIRECTIONS AND IMPLICATIONS

6.1 POTENTIAL MECHANISMS OF LOSS OF RBPJ TRANSCRIPTIONAL REPRESSION

Normal cells rely on constitutive RBPJ function to both mediate Notch receptor signal and immediately shut it off it in the absence of ligand activation. Our work demonstrates that RBPJ depletion in human cancer cells results in deregulated signal due to abolished transcriptional repression. Indeed, cancer cell lines show frequent genomic loss of RBPJ and there is evidence for RBPJ under-expression in primary tumors.

In addition to loss resulting from large genomic deletions (such as those we detected by aCGH), inactivation of RBPJ may occur by smaller deletions, insertions or point mutations. Based on our research and the work of others118, mutations occurring in key residues that alter the ability of RBPJ to bind DNA or recruit co-repressors could be expected to result in gene derepression.

Expression of RBPJ could also be lost as a result of epigenetic mechanisms, such as DNA methylation, histone modifications or microRNA-mediated gene silencing. We evaluated the levels of promoter methylation at a single CpG site (contained within a CpG island) 20 bp from the TSS in 15 breast cancer cell lines compared to that of the breast epithelial cell line MCF10A. No difference was found in methylation status between cells that expressed RBPJ protein and those that did not, with promoter methylation being low across the board (data not shown). However, this was a small sample set without complete coverage of the CpG island so this does not preclude blocked RBPJ transcription via promoter methylation or histone-mediated chromatin condensation. RBPJ mRNA may also be targeted by microRNAs prohibiting RBPJ translation. TargetScan (www.targetscan.org) predicts the binding of several microRNAs to the RBPJ transcript, and potential up-regulation of such microRNAs in cancers represents another mechanism to deplete RBPJ protein. Indeed one of these microRNAs, miR-223, has been identified as up-regulated in recurrent ovarian cancers308. However none of these microRNAs have been experimentally shown to target RBPJ.

Even in cells that harbor normal RBPJ, transcriptional repression could be abolished via loss of proteins involved in the co-repressor complex. These key interacting proteins mediate signal shut-off and prevent aberrant activation of RBPJ-dependent promoters. Derepression has been reported for the following RBPJ binding factors: the histone demethylase KDM5A137, the histone chaperone Asf1162, and co-repressor MTG16163. Loss of other co-repressors in the complex such as SHARP, SMRT or CIR would also result in derepression, and may even limit nuclear import of RBPJ, preventing its normal cellular localization35,118.

Finally the repressive function of RBPJ could be obstructed by mechanisms that sequester RBPJ or mediate its degradation. Interestingly strawberry Notch, a protein identified in Drosophila with mouse and human homologues309, has been linked to degradation of components of the RBPJ co-repressor complex resulting in gene derepression310. However, not much is known about the function of strawberry Notch in mammals311.

In the early characterization of RBPJ, it was noted that an RBPJ-specific antibody was able to detect RBPJ protein in nuclei of undifferentiated, but not differentiated cell lines, despite constant levels of protein in cell lysates122. The mouse embryonic carcinoma cell line, F9, lost its positive staining for RBPJ upon stimulation of differentiation. The authors show that the nuclei of undifferentiated F9 cells contain both free RBPJ (that the antibody is able to recognize), and a chromatin-bound form of RBPJ. In differentiated cells, all the RBPJ exists in the chromatin- bound form that is not detectable by immunostaining122. The large RBPJ-containing protein complex assembled on DNA presumably blocks the antibody from recognizing the epitope, therefore giving negative staining results in differentiated cells. This work implies that differentiated cells limit their plasticity by retention of the DNA bound form of RBPJ and co- repressor assembly. Undifferentiated cells may be able to induce a broader gene expression program via Notch activation of DNA-bound RBPJ, and/or derepression in the absence of DNA binding. A different RBPJ antibody was used in our studies for immunohistochemistry. This antibody is able to detect RBPJ in both MDA-MB-231 and DG75 cancer cell lines where we consistently observe nuclear staining. The antibody is sensitive to the dose of RBPJ with MDA- MB-231 cells showing stronger staining than DG75 cells, which only have one functional copy. Both the MDA-MB-231 and DG75 cell lines are characterized as undifferentiated312,313. However, we could also detect strong RBPJ staining in normal tissue sections containing differentiated cells, such as those present in the mammary epithelium. The antibody used to study RBPJ in the F9 cells could not stain stromal cells whatsoever122. Therefore, it is likely that our antibody is detecting chromatin bound RBPJ and that lack of staining is due to absence of the protein. This implies that loss of RBPJ-mediated repression may occur in a higher proportion of samples than what would be predicted by scoring presence or absence of RBPJ staining, as RBPJ protein could be present in the nuclei of cells but exist in a chromatin unbound form. Our tissue microarray data, therefore, may be underestimating RBPJ loss of function. What contributes to retention of RBPJ in the nucleus, but prevents it from associating with target promoters in undifferentiated cells, would be an interesting area to explore.

6.2 RBPJ LOSS IN THE CONTEXT OF NORMAL MAMMARY DEVELOPMENT AND BREAST CANCER

Notch activity is linked to stem cell maintenance and cancer cells themselves are characterized by being stem-like or de-differentiated63. Notch activation has been shown to increase mammosphere formation of primary breast cells65,94 and enhances the ability of the breast epithelial cell line, MCF10A, to grow in soft agar76. Mammosphere culture measures the ability of cells to proliferate in suspension as spherical structures which is thought to enrich for cells capable of self-renewal65. Comparably, soft agar evaluates the ability of cells to form colonies in anchorage independent conditions using semi-solid media. Anoikis assays can also be used to measure resistance to cell death in the absence of cell-matrix attachments. In preliminary experiments not shown here we evaluated the effect of RBPJ deficiency on normal cell function using MCF10A cells, which have a normal RBPJ copy number. Depletion of RBPJ using two targeting sequences caused derepression of HEY genes, although the magnitude of induction was smaller than that seen in MDA-MB-231 breast cancer cells. RBPJ KD in MCF10A cells increased mammosphere formation, conferred protection from anoikis and enhanced their ability to form colonies in soft agar. We are currently continuing experiments in this area to replicate these data. Despite only a modest HEY derepression seen in these cells RBPJ deficiency has measurable effects on MCF10A cell function. Moreover, the mouse embryonic fibroblasts derived from Rbpj null mice showed evidence for HES1 derepression compared to their RBPJ-containing counterparts. However, both the MCF10A and MEF cell lines are immortalized (and in addition MCF10A have an amplification of MYC)314. We have not evaluated how RBPJ loss impacts the normal functions of primary cells.

Published work has assessed the role of RBPJ in normal mammary development using the mouse mammary reconstitution model and two different strategies; a KD system to deplete RBPJ in the mammary stem cell enriched fraction13 or over-expression of a mutant RBPJ, similar to our construct, which prevents Notch activation but allows retained transcriptional repression though endogenous RBPJ131. Although different cell populations were transduced in the RBPJ KD versus mutant-RBPJ over-expression study, somewhat opposite outcomes resulted: One promoted mammary outgrowth and the other inhibited it. This is comparable to the opposing effects we see in our xenograft studies, with RBPJ depletion increasing tumor growth, and forced expression of the mutant-RBPJ decreasing it. If RBPJ removal and inhibition of Notch mediated activation are equivalent, we would expect a similar resulting phenotype. However, this is not observed in our studies. The differences can be explained by retention and

removal of transcriptional repression. In the Bouras et al. study13 the effect of RBPJ KD was interpreted as inhibited Notch signaling. However, as RBPJ depletion is capable of gene induction independent of Notch receptor activity the phenotype may not be due to a complete loss of Notch signal. In comparison, over-expression of Notch1IC in the mammary stem cell enriched population led to formation of hyperplastic nodules within mouse mammary glands13. In this context RBPJ loss likely causes a weaker global gene induction than Notch receptor signaling, as NotchIC can recruit RBPJ to promoters132 and additionally has non-canonical RBPJ independent functions50. Only genes that require RBPJ for their repression would be derepressed upon RBPJ loss. At other target genes, RBPJ may be recruited to promoters where it was previously not bound for activation of transcription via NotchIC132. Notch signaling would be expected to result in induction of a larger subset of genes; those already bound and repressed by RBPJ and those where RBPJ is recruited only for transcriptional activation.

Preliminary data suggest RBPJ loss in the context of breast cancer correlates with positive hormone receptor status and HER2 expressing breast cancer cell lines. This provides further incentive to validate RBPJ loss in a luminal or HER+ subtype of breast cancer. There is evidence for receptor driven Notch activation in basal-like cancers (ER, PR and HER2 negative)90,100. However, Notch1 transcript over-expression is also detected in luminal over basal subtypes of human breast cancer103. Furthermore, forced Notch1IC expression in the mouse mammary progenitor population gives rise to luminal hyperplastic nodules13. In normal mammary development Notch regulates luminal cell-fate commitment13,77,78. In breast cancer, aberrant Notch activity may be present in multiple tumor subtypes where normal lineage specification is disrupted. Further research into which breast cancer subtype is correlated with oncogenic Notch hyperactivation versus RBPJ loss could have important implications for future therapeutic strategies.

6.3 ROLE OF HEY GENES IN ONCOGENESIS MEDIATED BY RBPJ DEFICIENCY

Gene derepression has been reported by others to play a role in tumorigenesis. Derepression of cancer promoting genes, Claudin3 and Claudin4, results from the loss of repressive chromatin modifications, although the mechanism by which this occurs is unknown315. Loss of a transcriptional repressor has also been shown to cause induction of target genes. For example, down-regulation of DNMT1, leads to epigenetic derepression of its target gene p16/INK4A316. In the context of cancer more is known about transcriptional repression of tumor suppressor genes than is know about derepression of oncogenes315.

In our xenograft experiments, RBPJ loss caused a significant up-regulation of HEY gene expression. This is in contrast to a trend for inhibited HEY expression observed with the mt- RBPJ construct, which enables retained transcriptional repression. Furthermore, HEY gene promoters contain well-validated RBPJ recognition motifs, where binding of RBPJ is lost upon RBPJ depletion. Together this implies that the up-regulation of HEY expression is likely a direct result of gene derepression following RBPJ depletion.

HEY2 in particular showed the biggest expression changes in response to RBPJ removal. Our experiments, which examined the requirement of HEY2 in propagating oncogenic signal downstream of RBPJ loss, imply a partial role of HEY2 may be negative feedback to dampen the strength of Notch signal. HEY2 over-expression in RBPJ-containing MDA-MB-231 tumors down-regulated expression of other HEY family members and RBPJ and inhibited tumor growth. This effect likely occurs via direct association with RBPJ, resulting in transcriptional repression of target gene transcription221, although this remains to be validated in our system. Only a small number of direct target genes of HES and HEY proteins have been identified. Perhaps a partial role of these factors is to negatively target their own expression, which is supported by the oscillatory expression pattern observed for both HES and HEY52,55. In the absence of RBPJ, HEY proteins many not be able to exert their repressive effect. This may be the reason why a somewhat bigger induction of HEY genes is seen with RBPJ removal than with Notch activation in the MDA-MB-231 co-culture system.

We attempted to deplete up-regulated HEY2 in DG75 RBPJ null cells. Significant HEY2 KD was not maintained in tumor xenografts at the endpoint, when compared to HEY2 transcript levels in RBPJ KO tumors expressing a non-targeting sequence. The shRNA vector used to transduce these cells also expresses a GFP selection marker. There was significantly less GFP expression in the shHEY2 tumors compared to controls. This points to a selective outgrowth of cells that have lost HEY2 KD. If depletion of HEY2 combined with RBPJ loss is unfavorable to tumor cells, causing a selective pressure to not retain the HEY2 KD, then HEY genes have essential functions in addition to signal feedback. For the purpose of this thesis we used HEY genes as markers of deregulated signal. However, derepression of HEY genes may directly contribute to tumorigenicity by both abolishing a negative feedback mechanism and increasing HEY availability for transcriptional repression at other promoters. Further work is needed to address the above hypotheses regarding the role of HEY proteins in relation to RBPJ.

6.4 INDUCTION OF OTHER GENES FOLLOWING RBPJ REMOVAL

Although a microarray was attempted with MDA-MB-231 control versus RBPJ KD cells, as well as a chromatin immunoprecipitation directed against RBPJ itself, neither was successful. Knock-down of greater than 80% was achieved in all three biological replicate MDA-MB-231 shRBPJ samples used for expression profiling. However, many of the identified differentially expressed genes did not validate by qPCR. Microarrays have been known to significantly underestimate gene expression changes298 and their ability to detect small but consistent differences is hampered by noise297. In our hands, qPCR was more sensitive in identifying gene expression changes than the microarray analysis. We attempted to immunoprecipitate chromatin-bound RBPJ using several commercially available antibodies but did not detect enrichment of RBPJ at known recognition motifs in control compared to RBPJ KD MDA-MB-231 cells.

To identify other genes potentially induced upon RBPJ depletion, promoters marked with acetylated histone H4 in RBPJ KD MDA-MB-231 cells were used as a surrogate to global expression analysis. This method was successful for discovering new genes up-regulated by RBPJ deficiency. Validation of a subset of these genes showed enrichment of acH4 in promoter regions corresponds to increased mRNA expression. However, further work is required to evaluate whether these genes are direct RBPJ targets (and therefore derepressed following RBPJ removal) or secondarily up-regulated due to changes in expression of other genes. Differentially expressed genes that did not show changes in acH4 may have been missed as acH4 is an indirect estimator of gene activity compared to direct detection of transcript. Nevertheless, the acH4 data in MDA-MB-231 cells has allowed us to identify altered cellular functions contributing to deregulated signaling in RBPJ depleted cells. Ongoing work is aimed at identifying which up-regulated genes contribute to specific pathways that increased cell survival with RBPJ loss.

6.5 RBPJ LOSS PROMOTES CELL SURVIVAL

In vitro experiments corroborated that RBPJ deficiency increases cancer cell survival, as predicted by the computational analysis. Staining of whole tumor cryosections also confirmed that reduced cell death occurs in both RBPJ-deficient breast tumors and lymphoma tumors compared to control RBPJ-containing tumor xenografts.

Notch has been previously implicated in promoting cell survival. Increased Notch signaling in the breast epithelial cell line MCF10A protects cells from drug-induced apoptosis76. We have 96

reported that constitutively active Notch1 increases resistance of MCF10A cells to anoikis (a type of cell death resulting from abolished matrix attachment)88. Furthermore, blocking Notch signaling in MDA-MB-231 tumor xenografts increased tumor cell death as measured by staining for activated capsase 388.

A published study using the conditional mouse Rbpj KO strategy observed decreased levels of apoptosis in Rbpj null tumors86. The WAP mammary specific promoter (mostly active during pregnancy) was used in this system for two purposes: to drive Cre expression for conditional deletion of Rbpj, and for over-expression of a hyperactive Notch4. The resulting outcome is constitutively active Notch4 in the mammary glands of mice that lack Rbpj. WAP-Notch4 alone inhibits alveolar development and lactation, and causes development of mammary tumors that are transplantable86. Deletion of Rbpj in WAP-Notch4 mammary glands rescued mammary gland development and lactation, but still caused primary tumor development. These tumors had the ability to transplant into a recipient mammary gland, and occurred at the same frequency but with longer latency86. A proliferative difference was not observed in WAP-Notch4 Rbpj control versus WAP-Notch4 Rbpj KO tumors; however, the level of apoptosis was significantly lower in the KO tumors86. This agrees with our findings that RBPJ loss protects cells from death. The authors conclude that this effect is a consequence of RBPJ-independent Notch signaling. This is a possibility as Notch is known to have non-canonical functions outside of RBPJ50,317. However, their observations may also be due to gene derepression, resulting from Rbpj loss, which eventually leads to transformation.

Comparably, RBPJ depleted MDA-MB-231 cells acquired resistance to anoikis upon disruption of cell-matrix contacts. As DG75 cells are cultured in suspension we used peroxide as a stress, observing increased survival of RBPJ null cells under these conditions. Furthermore, both MDA-MB-231 RBPJ KD and DG75 RBPJ KO cells showed significantly reduced cell death as measured by TUNEL staining of tumor sections.

In vivo the most pronounced survival differences observed with RBPJ loss occurred at increasing distances from the vasculature in hypoxic tumor tissue. We could not detect differences in cell death upon culture of RBPJ depleted MDA-MB-231 cells in 1% oxygen in vitro (data not shown). However, the hypoxia-Notch signaling cross-talk requires physical interaction between NotchIC and HIF1α and recruitment of HIF1α to RBPJ-bound DNA for transcriptional activation205. Therefore, cross-talk with the HIF1α pathway may not be relevant in the absence

of RBPJ. However, another factor preferentially present in the hypoxic environment may promote survival of RBPJ deficient cells.

At this point it is not clear what specific factor promotes survival downstream of RBPJ loss. A number of genes that could mediate a pro-survival effect show a small but consistent up- regulation with RBPJ depletion. BCL2, a well characterized anti-apoptotic gene that is linked to cancer318, is up-regulated at the mRNA and the protein level in both RBPJ KD and KO cells. However this induction seems to be a secondary event. BCL2 does not have putative RBPJ binding motifs in its promoter and is not a reported direct target of RBPJ263. It is likely the effect of several genes, which contribute to multiple pathways, that act together to promote oncogenicity with RBPJ loss.

6.6 OTHER POTENTIAL CONSEQUENCES OF RBPJ LOSS

Interestingly, RBPJ-deficient cells show sensitivity to serum withdrawal and have different growth characteristics in vitro compared to in vivo. RBPJ-null lymphoma cells grow slower in culture but form tumors faster in mice. Comparably, RBPJ KD breast cancer cells have no growth differences in vitro compared to RBPJ containing controls. However, they also demonstrate increased tumor growth corresponding to higher BrdU staining, a marker of proliferation. This implies that RBPJ depletion is not sufficient and stimuli from the environment are required to confer advantages to these cells. Activation of an unknown signaling pathway by components in the serum may synergize with RBPJ loss; (1) as a result of induced genes, (2) via direct utilization of now vacant RBPJ binding sites and/or (3) through induction of genes that were previously repressed by RBPJ via alternate DNA binding sites.

A number of MMP gene promoters were identified as differentially acetylated by bioinformatics analysis. Of these we have validated three thus far and showed that MMP1 and MMP3 are predominantly up-regulated with MMP9 showing a smaller induction in RBPJ KD cells. These genes may represent direct RBPJ targets as each has at least one predicted high scoring RBPJ binding site within its promoter, although these have yet to be validated as functional RBPJ recognition sequences. Notch activity has previously been associated with MMP9. KD of both Notch1 and Jagged1 inhibited migration, and invasion corresponding to decreased MMP9 mRNA and protein expression and reduced MMP9 activation319,320. MMP9 may be involved in a positive feedback loop as it has been reported to proteolytically cleave and activate the Notch1 receptor at the membrane321,322. Our preliminary data in vitro could not

detect differences in Matrigel invasion but further work is needed to verify the outcome of up- regulated MMP expression.

Notch has been previously reported to activate NF-κB signaling, which in turn up-regulates MMP9 expression via binding to its consensus sequence in the MMP9 promoter319. In this instance MMP9 induction is mediated by NF-κB and represents a secondary, rather than a primary, target of NotchIC. A growing body of literature supports extensive cross-talk between NF-κB and Notch signaling252. The NF-κB pathway regulates numerous processes such as cell survival, proliferation, adhesion and migration323, and NF-κB signaling is implicated in a variety of human cancers324. Indeed NF-κB is a major mediator of Notch1 induced transformation in leukemia and NF-κB pathway components are direct Notch transcriptional targets212. Furthermore, one in eight NF-κB sites are also predicted to be RBPJ binding sites4,253. Since overlapping recognition sequences are partially shared, RBPJ deficiency may allow NF-κB access to DNA, allowing transcriptional activation mediated by NF-κB signal. Bioinformatics analysis of the HEY gene regulatory regions predicts that one RBPJ binding motif in HEY1 and HEY2 promoter and two RBPJ binding motifs in HEYL promoter could also be putative NF-κB sites. However our EMSA results imply that the RBPJ consensus is unbound in the absence of RBPJ and NF-κB would recognize an overlapping sequence. This does not preclude NF-κB playing a role at other induced promoters such as MMPs, or binding to its alternate recognition motifs that are divergent from the RBPJ consensus sequence. Loss of RBPJ may enable a transcription factor such as NF-κB accessibility to DNA even at sites distant to the RBPJ consensus.

Of note, depletion of RBPJ may lead to an excess of co-activators that are no longer utilized for Notch activation and are now available for functions in other pathways. Indeed MAML has recently been reported to have Notch-independent activities325. In addition to interaction with several other nuclear proteins, MAML positively regulates NF-κB signaling. MAML activates NF- κB dependent transcription (via RELA) and causes degradation of IκBα, the inhibitor of NF-κB signaling326. Both MAML and NF-κB pathway KO models show severe liver cell death326. These data provide a further mechanism by which RBPJ loss may bolster the activity of a transcriptional activator such as NF-κB. Identifying candidate pathways that mediate gene activation following RBPJ loss could have therapeutic implications for treatment of cancer, and further aid our understanding of how transcriptional activation occurs following loss of repression.

6.7 FUTURE DIRECTIONS

The work in this thesis has identified the importance of RBPJ in mediating transcriptional repression in the absence of Notch signaling and has evaluated outcomes of RBPJ loss in a cancer context. However, the observations reported here have also raised a number of interesting questions that should be addressed by future experiments. These include the following: how is RBPJ inactivated in cancers, what is the outcome of RBPJ loss in the context of normal cell function, is RBPJ deficiency associated with a certain subtype of cancer, what role do HEY genes play, how do other induced genes contribute to tumorigenicity and are these directly derepressed with RBPJ loss or secondary targets, which pathways contribute to the phonotype resulting from RBPJ removal, and what factor provides transcriptional activation that leads to gene induction upon RBPJ loss? Experiments are proposed below to address some of these topics.

We observed RBPJ loss as part of a large deletion occurring on the short arm of chromosome 4. To evaluate whether smaller gene mutations play a role in deregulating RBPJ, genomic DNA extracted from patient tumor samples could be used to sequence the RBPJ locus with primers targeting essential domains (DNA and co-repressors binding). Even in cells that retain RBPJ expression, its binding to DNA may be prevented by mutations in key residues.

The removal of functional non-mutated RBPJ from DNA could also result in derepression. RBPJ localization should be evaluated in cancer versus normal cells. Cell fractions can be separated into cytoplasmic, membrane and nuclear components. Nuclear proteins can be further fractionated using different salt concentrations to obtain DNA-bound versus unbound nuclear proteins122. Furthermore, loss of co-repressor assembly due to depletion of key co- repressor proteins would also abolish transcriptional repression. Relevance of co-repressor loss in human cancer can be probed using Oncomine, to search for evidence of co-repressor down- regulation in published microarray studies.

Whether RBPJ depletion is capable of inducing transformation in primary cells should be further explored. We have only begun to test how RBPJ loss impacts normal cellular function in vitro, and are planning on evaluating whether Rbpj KO mouse embryonic fibroblasts (OT11) enrich for colony formation in soft agar. However this work has focused on cell lines and not primary cells. In vivo, the mouse mammary gland represents a good model for studying any potential oncogenic effects resulting from inactivation of RBPJ tumor suppressor function. The mammary gland can completely regenerate upon orthotopic transplantation of primary

100

mammary epithelial cells131. Conflicting results were obtained with RBPJ KD13 versus mt-RBPJ over-expression131 in the mouse mammary gland, with one promoting and the other inhibiting mammary outgrowth respectively. A close examination of gene expression in RBPJ KD mammary transplants can resolve whether the resulting phenotype is due to loss of transcriptional repression or blocked signal altogether. Addressing these differences with mammary reconstitution experiments would provide further insight into how RBPJ function impacts normal cellular development. RBPJ depletion alone may not be enough to transform primary epithelial cells in the absence of other activating mutations but may lead to other abnormalities.

In the context of breast cancer RBPJ loss of expression may be associated with positive hormone or HER2 receptor status based on Oncomine expression data and RBPJ deletion in breast cancer cell lines. Primary human breast tumors of known type can be evaluated by IHC staining for the presence of RBPJ and NotchIC, combined with expression data using HEY genes as markers of signal activation. These studies would provide further insight into whether Notch hyperactivation or RBPJ loss segregate to a certain breast cancer subtype.

As HEY genes are well-validated direct RBPJ targets we used them as markers of aberrant signal activation with RBPJ loss. However it is not clear how their induction contributes to tumorigenicity. We are currently exploring the effect of HEY gene KD on survival of DG75 RBPJ KO cells following treatment with peroxide. These cells are protected against death compared to their RBPJ-containing counterparts. If HEY genes play an essential downstream role in RBPJ- null cells, any survival advantages conferred on these cells with RBPJ loss should be abolished.

We have focused our attention on HEY2 as it is most up-regulated HEY gene in MDA-MB- 231 and DG75 cells upon RBPJ loss. A competition experiment performed in vivo is needed to re-evaluate whether HEY2 depletion in cells with RBPJ loss is selected against and causes outgrowth of cells that have lost HEY2 KD. Co-injection of RBPJ null cells transduced with a GFP-labeled non-targeting shRNA, and RBPJ null cells expressing an mCherry fluorescent marker shHEY2 construct, would allow tracking of the two populations in vivo. If both cell lines are injected at the same concentration and have equal tumorigenic potential, there should be a one-to-one ratio of green to red cells upon endpoint sorting of the dissociated tumor. However, if green cells are overrepresented in the final tumor mass they either have a growth advantage over the red shHEY2 cells, or a loss of the mCherry vector (and therefore shHEY2 expression) occurs in cells transduced with this construct. Both scenarios would imply that HEY2 plays a

101

key role in tumor growth mediated by RBPJ loss and would warrant further experimentation. An inducible KD system may be required to achieve stable HEY2 depletion in tumors.

DG75 cells do not appear to have active Notch signaling in vitro (although it is unknown whether this is true in vivo). Our MDA-MB-231 cells appear to have no Notch signaling activity in vitro but blockade of Notch-mediated transcriptional activation in vivo inhibits tumor growth both in our previous experiments88 and in the work presented here. MDA-MB-231 cells may therefore be a good model system to study HEY function in vivo where Notch activity is present. The following proposed experiment is intended to dissect the involvement of HEY proteins in deregulated signal resulting from both Notch receptor activation and RBPJ loss, so their role in signal feedback and other downstream functions can be evaluated. Tumor growth and the resulting gene expression should be compared between control cells (which contain both RBPJ and HEY2), MDA-MB-231 RBPJ KD cells (HEY2 only), HEY2 KD cells (RBPJ only) or both RBPJ and HEY2 KD cells (no RBPJ or HEY2). The expected outcomes predicted based on our previous results are summarized in Table 6.1. It is not known if other HEY family members could compensate for the loss of HEY2.

Table 6.1 Summary of a proposed experiment evaluating the requirement of HEY2 downstream of RBPJ loss in MDA-MB-231 cells.

RBPJ KD HEY2 KD RBPJ + HEY2 KD RBPJ and HEY2 status Only HEY2 with no RBPJ Only RBPJ with no HEY2 No RBPJ or HEY2 Increased tumor growth if the major Inhibited tumor growth if HEY2 is Predicted outcomes Increased tumor growth. role of HEY genes is feedback to an essential effector downstream compared to control cells Implies RBPJ depletion is dampen Notch signal. of RBPJ. which contain both RBPJ stronger than Notch activation Decreased tumor growth if major role No effect if feedback is the major and HEY2 in this context. of HEY genes is to propagate aberrant function as RBPJ is also deficient. signal downstream.

In addition to HEY genes, other genes are up-regulated with RBPJ removal. We have investigated a number of these genes to see if acH4 enrichment corresponds with induced transcription, and are continuing with this work. Our aim is to elucidate specific mechanisms that cause increased tumorigenicity with RBPJ loss at the pathway level rather than the cellular functions level. We expect it is the combination of multiple induced genes that contribute to the phenotype observed with RBPJ loss. Although some of these genes have putative RBPJ binding motifs it is not know whether they are direct targets of RBPJ. Gene promoters with acH4 enrichment in MDA-MB-231 cells have to be individually evaluated using ChIP-qPCR or EMSA to verify whether they are directly bound by RBPJ and therefore derepressed upon RBPJ removal.

102

The DG75 cells would provide an alternate cellular context to probe for global gene expression changes for comparison with the MDA-MB-231 data to make generalized conclusions (across divergent cell lines), regarding RBPJ deficiency. DG75 RBPJ null cells have a complete loss of RBPJ, preventing undesired outcomes that are a consideration with shRNA depletion (e.g. off-target effects and incomplete KD). As global gene expression changes may be small but produce a synergistic effect, a technology other than microarray analysis such as qPCR arrays or whole transcriptome sequencing may be better suited to probing for differential transcript expression in these cells. In our lab, we have a Flag-tagged RBPJ construct that we have previously attempted to over-express in DG75 cells, without success, using a retroviral system. We are currently working on cloning Flag-RBPJ into a lentiviral system to achieve better transduction efficiency. Alternately, an inducible system where Flag-RBPJ expression is turned on selectively is also an option. Being able to transduce DG75 RBPJ KO cells with Flag-RBPJ will not only allow us to evaluate the direct effect of RBPJ reconstitution on target gene expression, but will also enable chromatin immunoprecipitation directed against the Flag-tag. Using DG75 RBPJ KO cells and DG75 RBPJ KO cells over- expressing Flag-RBPJ, gene expression profiling and ChIP-seq can be utilized to identify a global set of induced genes and provide information on which of these are direct RBPJ targets. Such experiments would identify directly derepressed genes versus secondary events that result from the action of derepressed genes.

In addition to focusing experiments on RBPJ, forced expression of Notch1IC in DG75 RBPJ- containing parental cells, compared to DG75 parental vector control cells, would allow comparison of gene expression changes as a result of constitutively active Notch versus RBPJ loss. This would be useful in evaluating the composition and strength of the signal produced by each, and overlapping versus non-overlapping transcript changes.

To look at a time course of primary and secondary events that occur upon RBPJ removal an inducible KO system would be ideal. The currently available mouse and human KO cell lines have inactivation of RBPJ in a different region of the gene, leaving five to six127,128 or three113 potentially functional exons respectively. In the mouse Rbpj KO construct at least one functional DNA binding domain is potentially retained. A human system using a conditional RBPJ KO strategy that removes all DNA binding domains in the potentially produced truncated protein would be worth generating.

Interestingly, the effect of RBPJ loss may be potentiated by an extracellular signal where a secreted factor present in the blood effects changes in RBPJ KD cells. Indeed RBPJ deficient 103

cells are sensitive to serum starvation. Fetal bovine serum, which is used to culture these cells, has been found to contain bioactive cytokines, hormones and growth factors327. To identify potentially interacting pathways, experiments should be performed in serum free conditions or charcoal stripped serum (which removes non-polar material including most growth factors, cytokines and hormones). A screen can be performed to test which supplemented growth factors restore viability of RBPJ deficient cells. Alternately, a screen could also evaluate the ability of various pathway inhibitors to abolish the survival advantage caused by RBPJ deficiency in serum containing media. Luciferase reporter assays could further test which of the potentially identified candidates enables activation of RBPJ-dependent promoters in the absence of RBPJ.

The expression of MMPs is up-regulated in RBPJ depleted cells, and these enzymes have known functions in invasion and metastasis306. MMP cleavage of numerous substrates can result in the release of previously latent signaling molecules328. These enzymes may explain the differences observed with RBPJ loss in the tumor microenvironment versus cells cultured in vitro. Preliminary in vitro studies could not detect potentially acquired invasive ability with RBPJ KD. However, we used Matrigel as a basement membrane matrix. A repeat of these experiments using an alternate substrate, such as gelatin or collagen, may reveal differences with cell invasion. To test whether the up-regulated MMPs mediate this effect, pan-MMP inhibitors, such as GM6001329 can be used to evaluate if any invasive advantages conferred by RBPJ KD are abolished. MMP enzyme activity will also need to be evaluated by zymography.

A link exists between NF-κB and Notch signaling and NF-κB is known to have antiapoptotic functions and can induce expression of MMPs319,320. It would be worth evaluating whether RBPJ loss results in up-regulated NF-κB signaling. Indeed the NFKB2 gene shows enrichment of acH4 and is induced 2-fold in RBPJ KD MDA-MB-231 cells compared to controls. NF-κB pathway inhibitors, such as BMS-345541 that prevent activation of the IKK complex212 can be used to block the cascade in vitro and test whether the increased survival advantage, resulting from RBPJ loss, is abolished.

6.6 IMPLICATIONS OF WORK

This study provides an alternate mechanism of how the Notch signaling pathway could be deregulated in cancers and offers evidence to support a role for RBPJ as a novel tumor suppressor in multiple cancer types. Studies using an RBPJ KD or KO system to evaluate the biological outcomes of blocking Notch signaling should be interpreted cautiously or reproduced 104

by inactivating alternate components of the pathway. In cells that require Notch receptor signal for their function, deletion of RBPJ may indeed cause a partial block of this signal due to inability of Notch to recruit RBPJ to specific promoters. However with RBPJ loss a subsequent gene induction should also be considered as a potential outcome.

Figure 6.1 proposes a model based on this thesis work, illustrating how the presence or absence of RBPJ affects activation at target gene promoters. With hyperactive Notch, excess nuclear NotchIC would drive RBPJ-mediated transcriptional activation (Figure 6.1A). As such retention of RBPJ would be selected for in some cancers. A partial function of induced HEY target genes may be signal feedback. Future studies should address the ability of HEY proteins to dampen Notch signal. Perhaps this is why basal Notch receptor signaling in vivo produces a weaker HEY induction and is less tumorigenic than RBPJ loss in MDA-MB-231 cells. However, in other contexts Notch activation can produce a stronger phenotype than RBPJ loss13,330 as NotchIC together with RBPJ may induce a larger subset of genes than require RBPJ for their repression.

Retention of one functional copy of RBPJ in the absence of active Notch may lead to weakened transcriptional repression or may be sufficient to block expression of target genes (Figure 6.1B). NotchIC does not play a role in gene induction in the absence of RBPJ in our experiments. However, in tumors that have one functional copy of RBPJ and active Notch signaling, both derepression due to depleted RBPJ levels and activation of promoter bound RBPJ by NotchIC may occur. It is tempting to speculate that the combination of constitutive Notch and RBPJ haploinsufficiency may synergize to produce the worst outcome. However, the presence of hyperactive Notch may already saturate all available RBPJ even when two functional copies exist. Indeed, mouse mammary tumors resulting from constitutive Notch signaling were phenotypically the same whether they had heterozygous inactivation of Rbpj or both functional gene copies86.

In cancers where RBPJ is lost, the resulting aberrant gene expression would represent an alternate mechanism of producing an inappropriate Notch-like signal (Figure 6C). Furthermore, γ-secretase inhibitors that inhibit Notch receptor cleavage would have no beneficial effect on tumors with RBPJ loss. An alternate therapeutic strategy would be required to restore transcriptional repression at RBPJ/Notch target genes. For example, upregulating Ikaros expression in cancer cells should inhibit constitutive Notch activity (since it does not bind NotchIC), but would restore transcriptional repression in the absence of RBPJ4.

105

Notch target genes have been shown to comprise both positive and negative regulators, such that the cells become poised for several outcomes167. Furthermore, the combination of transcription factors present within a cell can alter the response to Notch signal50. Whether through Notch receptor activation, or derepression via RBPJ loss, the composition of genes induced and the resulting signal strength may produce divergent effects. The severity of the phenotype produced by RBP loss is expected to be highly context dependent.

Our results indicate that depletion of RBPJ leads to derepression of a subset of RBPJ bound promoters and increased oncogenicity. In a recent study depletion of RBPJ was reported to enhance eye tumor growth and metastases in a Drosophila model system137, supporting our observations here. This work has defined a role for RBPJ as a novel tumor suppressor and implies that both Notch hyperactivation and RBPJ deficiency represent oncogenic events.

106

No Notch Active Notch

A: RBPJ function retained

B: Loss of one functional RBPJ copy

C: Complete loss of RBPJ

Figure 6.1 Schematic illustrating possible effects of RBPJ loss. (A) In cells where both functional copies of RBPJ are retained, target gene expression is inhibited in the absence of Notch receptor signaling and activated by NotchIC. Direct target genes of the HEY family such as HEY2 can bind RBPJ and dampen transcriptional activation without disrupting NotchIC/RBPJ interactions221. However, HEY transcription factors also exert their repressive effect by binding to other proteins and their target consensus sequence on DNA. (B) RBPJ haploinsufficiency may reduce the strength of repression in the absence of Notch receptor signaling. In the presence of nuclear NotchIC RBPJ would be utilized for Notch pathway activation. (C) With complete loss of RBPJ neither co-activator nor co-repressor assembly can take place. Loss of transcriptional repression results in gene activation by an unknown mechanism. Fewer genes may be derepressed with RBPJ loss as are directly induced by Notch activation, as only a subset of these genes require RBPJ for their repression. However, with RBPJ loss the magnitude of gene induction may be greater at promoters that require RBPJ for their repression as a potential feedback via HEY genes cannot occur, although this hypothesis requires further investigation. Schematic created using www.proteinlounge.com.

107

REFERENCES 1. Marrett, L.D., PhD, De, P., MHSc PhD, Airia, P., MD MSc, Dryer, D., MD & for the steering committee of Canadian Cancer Statistics Cancer in Canada in 2008. CMAJ 179, 1163-1170 (2008). 2. Vargo-Gogola, T. & Rosen, J.M. Modelling breast cancer: one size does not fit all. Nat Rev Cancer 7, 659-672 (2007). 3. Nowell, P.C. The clonal evolution of tumor cell populations. Science 194, 23-28 (1976). 4. Beverly, L.J. & Capobianco, A.J. Targeting promiscuous signaling pathways in cancer: another Notch in the bedpost. Trends Mol Med 10, 591-598 (2004). 5. Hanahan, D. & Weinberg, R.A. The hallmarks of cancer. Cell 100, 57-70 (2000). 6. Colotta, F., Allavena, P., Sica, A., Garlanda, C. & Mantovani, A. Cancer-related inflammation, the seventh hallmark of cancer: links to genetic instability. Carcinogenesis 30, 1073-1081 (2009). 7. Lazebnik, Y. What are the hallmarks of cancer? Nat Rev Cancer 10, 232-233. 8. Gatenby, R.A. & Gillies, R.J. A microenvironmental model of carcinogenesis. Nat Rev Cancer 8, 56-61 (2008). 9. Fang, J.S., Gillies, R.D. & Gatenby, R.A. Adaptation to hypoxia and acidosis in carcinogenesis and tumor progression. Seminars in Cancer Biology 18, 330-337 (2008). 10. Kamangar, F., Dores, G.M. & Anderson, W.F. Patterns of cancer incidence, mortality, and prevalence across five continents: defining priorities to reduce cancer disparities in different geographic regions of the world. J Clin Oncol 24, 2137-2150 (2006). 11. Ottewell, P.D., Coleman, R.E. & Holen, I. From genetic abnormality to metastases: murine models of breast cancer and their use in the development of anticancer therapies. Breast Cancer Res Treat 96, 101-113 (2006). 12. Ge, Y., et al. Immunohistochemical characterization of subtypes of male breast carcinoma. Breast Cancer Research 11, R28 (2009). 13. Bouras, T., et al. Notch signaling regulates mammary stem cell function and luminal cell-fate commitment. Cell Stem Cell 3, 429-441 (2008). 14. Shackleton, M., et al. Generation of a functional mammary gland from a single stem cell. Nature 439, 84-88 (2006). 15. Ali, S. & Coombes, R.C. Endocrine-responsive breast cancer and strategies for combating resistance. Nat Rev Cancer 2, 101-112 (2002). 16. Cowin, P., Rowlands, T.M. & Hatsell, S.J. Cadherins and catenins in breast cancer. Curr Opin Cell Biol 17, 499-508 (2005). 17. Wellings, S.R. A hypothesis of the origin of human breast cancer from the terminal ductal lobular unit. Pathol Res Pract 166, 515-535 (1980). 18. Sainsbury, J.R., Anderson, T.J. & Morgan, D.A. ABC of breast diseases: breast cancer. BMJ 321, 745-750 (2000). 19. Polyak, K. Breast cancer: origins and evolution. J Clin Invest 117, 3155-3163 (2007). 20. Jerry, D.J. Roles for estrogen and progesterone in breast cancer prevention. Breast Cancer Res 9, 102 (2007). 21. Carnevale, R.P., et al. Progestin effects on breast cancer cell proliferation, proteases activation, and in vivo development of metastatic phenotype all depend on progesterone receptor capacity to activate cytoplasmic signaling pathways. Mol Endocrinol 21, 1335-1358 (2007). 22. Pietras, R.J. & Marquez-Garban, D.C. Membrane-associated estrogen receptor signaling pathways in human cancers. Clin Cancer Res 13, 4672-4676 (2007). 23. Korkola, J.E., et al. Differentiation of Lobular versus Ductal Breast Carcinomas by Expression Microarray Analysis. Cancer Res 63, 7167-7175 (2003).

108

24. Moll, R., Mitze, M., Frixen, U. & Birchmeier, W. Differential loss of E-cadherin expression in infiltrating ductal and lobular breast carcinomas. Am J Pathol 143, 1731-1742 (1993). 25. Dunnwald, L.K., Rossing, M.A. & Li, C.I. Hormone receptor status, tumor characteristics, and prognosis: a prospective cohort of breast cancer patients. Breast Cancer Res 9, R6 (2007). 26. Ross, J.S. & Fletcher, J.A. The HER-2/neu Oncogene in Breast Cancer: Prognostic Factor, Predictive Factor, and Target for Therapy. Oncologist 3, 237-252 (1998). 27. Sorlie, T., et al. Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A 98, 10869-10874 (2001). 28. Carey, L.A., et al. Race, Breast Cancer Subtypes, and Survival in the Carolina Breast Cancer Study. JAMA 295, 2492-2502 (2006). 29. Prat, A. & Perou, C.M. Mammary development meets cancer genomics. Nat Med 15, 842-844 (2009). 30. Romond, E.H., et al. Trastuzumab plus adjuvant chemotherapy for operable HER2-positive breast cancer. N Engl J Med 353, 1673-1684 (2005). 31. Finak, G., et al. Stromal gene expression predicts clinical outcome in breast cancer. Nat Med 14, 518-527 (2008). 32. Callahan, R. & Egan, S.E. Notch signaling in mammary development and oncogenesis. J Mammary Gland Biol Neoplasia 9, 145-163 (2004). 33. Leong, K.G. & Karsan, A. Recent insights into the role of Notch signaling in tumorigenesis. Blood 107, 2223-2233 (2006). 34. Radtke, F. & Raj, K. The role of Notch in tumorigenesis: oncogene or tumour suppressor? Nat Rev Cancer 3, 756-767 (2003). 35. Borggrefe, T. & Oswald, F. The Notch signaling pathway: transcriptional regulation at Notch target genes. Cell Mol Life Sci 66, 1631-1646 (2009). 36. Lai, E.C. Keeping a good pathway down: transcriptional repression of Notch pathway target genes by CSL proteins. EMBO Rep 3, 840-845 (2002). 37. Xue, Y., et al. Embryonic lethality and vascular defects in mice lacking the Notch ligand Jagged1. Hum Mol Genet 8, 723-730 (1999). 38. Jiang, R., et al. Defects in limb, craniofacial, and thymic development in Jagged2 mutant mice. Genes Dev 12, 1046-1057 (1998). 39. Limbourg, A., et al. Notch ligand Delta-like 1 is essential for postnatal arteriogenesis. Circ Res 100, 363-371 (2007). 40. Dunwoodie, S.L., et al. Axial skeletal defects caused by mutation in the spondylocostal dysplasia/pudgy gene Dll3 are associated with disruption of the segmentation clock within the presomitic mesoderm. Development 129, 1795-1806 (2002). 41. Duarte, A., et al. Dosage-sensitive requirement for mouse Dll4 in artery development. Genes Dev 18, 2474-2478 (2004). 42. Krebs, L.T., et al. Characterization of Notch3-deficient mice: normal embryonic development and absence of genetic interactions with a Notch1 mutation. Genesis 37, 139-143 (2003). 43. Krebs, L.T., et al. Notch signaling is essential for vascular morphogenesis in mice. Genes Dev 14, 1343-1352 (2000). 44. Swiatek, P.J., Lindsell, C.E., del Amo, F.F., Weinmaster, G. & Gridley, T. Notch1 is essential for postimplantation development in mice. Genes Dev 8, 707-719 (1994). 45. Hamada, Y., et al. Mutation in ankyrin repeats of the mouse Notch2 gene induces early embryonic lethality. Development 126, 3415-3424 (1999). 46. Koch, U. & Radtke, F. Notch and cancer: a double-edged sword. Cell Mol Life Sci 64, 2746-2762 (2007).

109

47. Bray, S.J. Notch signalling: a simple pathway becomes complex. Nat Rev Mol Cell Biol 7, 678-689 (2006). 48. Ilagan, M.X. & Kopan, R. SnapShot: notch signaling pathway. Cell 128, 1246 (2007). 49. Rizzo, P., et al. Targeting Notch signaling cross-talk with estrogen receptor and ErbB-2 in breast cancer. Adv Enzyme Regul 49, 134-141 (2009). 50. Martinez Arias, A., Zecchini, V. & Brennan, K. CSL-independent Notch signalling: a checkpoint in cell fate decisions during development? Curr Opin Genet Dev 12, 524-533 (2002). 51. Mumm, J.S. & Kopan, R. Notch signaling: from the outside in. Dev Biol 228, 151-165 (2000). 52. Nakagawa, O., et al. Members of the HRT family of basic helix-loop-helix proteins act as transcriptional repressors downstream of Notch signaling. Proc Natl Acad Sci U S A 97, 13655- 13660 (2000). 53. Kovall, R.A. & Hendrickson, W.A. Crystal structure of the nuclear effector of Notch signaling, CSL, bound to DNA. EMBO J 23, 3441-3451 (2004). 54. Fischer, A. & Gessler, M. Delta-Notch--and then? Protein interactions and proposed modes of repression by Hes and Hey bHLH factors. Nucleic Acids Res 35, 4583-4596 (2007). 55. Iso, T., Kedes, L. & Hamamori, Y. HES and HERP families: multiple effectors of the Notch signaling pathway. J Cell Physiol 194, 237-255 (2003). 56. Gazave, E., et al. Origin and evolution of the Notch signalling pathway: an overview from eukaryotic genomes. BMC Evol Biol 9, 249 (2009). 57. Weijzen, S., et al. Activation of Notch-1 signaling maintains the neoplastic phenotype in human Ras-transformed cells. Nat Med 8, 979-986 (2002). 58. Rizzo, P., et al. Rational targeting of Notch signaling in cancer. Oncogene 27, 5124-5131 (2008). 59. McGill, M.A., Dho, S.E., Weinmaster, G. & McGlade, C.J. Numb regulates post-endocytic trafficking and degradation of Notch1. J Biol Chem 284, 26427-26438 (2009). 60. Ascano, J.M., Beverly, L.J. & Capobianco, A.J. The C-terminal PDZ-ligand of JAGGED1 is essential for cellular transformation. J Biol Chem 278, 8771-8779 (2003). 61. Bland, C.E., Kimberly, P. & Rand, M.D. Notch-induced proteolysis and nuclear localization of the Delta ligand. J Biol Chem 278, 13607-13610 (2003). 62. Hiratochi, M., et al. The Delta intracellular domain mediates TGF-beta/Activin signaling through binding to Smads and has an important bi-directional function in the Notch-Delta signaling pathway. Nucleic Acids Res 35, 912-922 (2007). 63. Strizzi, L., et al. Development and cancer: at the crossroads of Nodal and Notch signaling. Cancer Res 69, 7131-7134 (2009). 64. Radtke, F., Fasnacht, N. & Macdonald, H.R. Notch signaling in the immune system. Immunity 32, 14-27. 65. Dontu, G., et al. Role of Notch signaling in cell-fate determination of human mammary stem/progenitor cells. Breast Cancer Res 6, R605-615 (2004). 66. Nicolas, M., et al. Notch1 functions as a tumor suppressor in mouse skin. Nat Genet 33, 416-421 (2003). 67. Tao, J., Chen, S. & Lee, B. Alteration of Notch signaling in skeletal development and disease. Ann N Y Acad Sci 1192, 257-268. 68. Stockhausen, M.T., Kristoffersen, K. & Poulsen, H.S. The functional role of Notch signaling in human gliomas. Neuro Oncol 12, 199-211. 69. Qiao, L. & Wong, B.C. Role of Notch signaling in colorectal cancer. Carcinogenesis 30, 1979-1986 (2009). 70. Villaronga, M.A., Bevan, C.L. & Belandia, B. Notch signaling: a potential therapeutic target in prostate cancer. Curr Cancer Drug Targets 8, 566-580 (2008).

110

71. Rose, S.L., Kunnimalaiyaan, M., Drenzek, J. & Seiler, N. Notch 1 signaling is active in ovarian cancer. Gynecol Oncol 117, 130-133. 72. Ristorcelli, E. & Lombardo, D. Targeting Notch signaling in pancreatic cancer. Expert Opin Ther Targets 14, 541-552. 73. Fung, E., et al. Delta-like 4 induces notch signaling in macrophages: implications for inflammation. Circulation 115, 2948-2956 (2007). 74. Lee, S.H., Jeong, E.G. & Yoo, N.J. Mutational analysis of NOTCH1, 2, 3 and 4 genes in common solid cancers and acute leukemias. APMIS 115, 1357-1363 (2007). 75. Weng, A.P., et al. Activating mutations of NOTCH1 in human T cell acute lymphoblastic leukemia. Science 306, 269-271 (2004). 76. Stylianou, S., Clarke, R.B. & Brennan, K. Aberrant activation of notch signaling in human breast cancer. Cancer Res 66, 1517-1525 (2006). 77. Buono, K.D., et al. The canonical Notch/RBP-J signaling pathway controls the balance of cell lineages in mammary epithelium during pregnancy. Dev Biol 293, 565-580 (2006). 78. Raouf, A., et al. Transcriptome analysis of the normal human mammary cell commitment and differentiation process. Cell Stem Cell 3, 109-118 (2008). 79. Reedijk, M., et al. High-level coexpression of JAG1 and NOTCH1 is observed in human breast cancer and is associated with poor overall survival. Cancer Res 65, 8530-8537 (2005). 80. Gallahan, D., Kozak, C. & Callahan, R. A new common integration region (int-3) for mouse mammary tumor virus on mouse chromosome 17. J Virol 61, 218-220 (1987). 81. Kiaris, H., et al. Modulation of notch signaling elicits signature tumors and inhibits hras1-induced oncogenesis in the mouse mammary epithelium. Am J Pathol 165, 695-705 (2004). 82. Dievart, A., Beaulieu, N. & Jolicoeur, P. Involvement of Notch1 in the development of mouse mammary tumors. Oncogene 18, 5973-5981 (1999). 83. Hu, C., et al. Overexpression of activated murine Notch1 and Notch3 in transgenic mice blocks mammary gland development and induces mammary tumors. Am J Pathol 168, 973-990 (2006). 84. Gallahan, D., et al. Expression of a truncated Int3 gene in developing secretory mammary epithelium specifically retards lobular differentiation resulting in tumorigenesis. Cancer Res 56, 1775-1785 (1996). 85. Jhappan, C., et al. Expression of an activated Notch-related int-3 transgene interferes with cell differentiation and induces neoplastic transformation in mammary and salivary glands. Genes Dev 6, 345-355 (1992). 86. Raafat, A., et al. Rbpj conditional knockout reveals distinct functions of Notch4/Int3 in mammary gland development and tumorigenesis. Oncogene 28, 219-230 (2009). 87. Soriano, J.V., Uyttendaele, H., Kitajewski, J. & Montesano, R. Expression of an activated Notch4(int-3) oncoprotein disrupts morphogenesis and induces an invasive phenotype in mammary epithelial cells in vitro. Int J Cancer 86, 652-659 (2000). 88. Leong, K.G., et al. Jagged1-mediated Notch activation induces epithelial-to-mesenchymal transition through Slug-induced repression of E-cadherin. J Exp Med 204, 2935-2948 (2007). 89. Lee, C.W., Raskett, C.M., Prudovsky, I. & Altieri, D.C. Molecular dependence of estrogen receptor-negative breast cancer on a notch-survivin signaling axis. Cancer Res 68, 5273-5281 (2008). 90. Rizzo, P., et al. Cross-talk between notch and the estrogen receptor in breast cancer suggests novel therapeutic approaches. Cancer Res 68, 5226-5235 (2008). 91. Han, J., Ma, I., Hendzel, M.J. & Allalunis-Turner, J. The cytotoxicity of gamma-secretase inhibitor I to breast cancer cells is mediated by proteasome inhibition, not by gamma-secretase inhibition. Breast Cancer Res 11, R57 (2009).

111

92. O'Neill, C.F., et al. Notch2 signaling induces apoptosis and inhibits human MDA-MB-231 xenograft growth. Am J Pathol 171, 1023-1036 (2007). 93. Parr, C., Watkins, G. & Jiang, W.G. The possible correlation of Notch-1 and Notch-2 with clinical outcome and tumour clinicopathological parameters in human breast cancer. Int J Mol Med 14, 779-786 (2004). 94. Mittal, S., Subramanyam, D., Dey, D., Kumar, R.V. & Rangarajan, A. Cooperation of Notch and Ras/MAPK signaling pathways in human breast carcinogenesis. Mol Cancer 8, 128 (2009). 95. Dickson, B.C., et al. High-level JAG1 mRNA and protein predict poor outcome in breast cancer. Mod Pathol 20, 685-693 (2007). 96. Jubb, A.M., et al. Expression of vascular notch ligand delta-like 4 and inflammatory markers in breast cancer. Am J Pathol 176, 2019-2028. 97. Yao, K., et al. Notch-1 and Notch-4 Receptors as Prognostic Markers in Breast Cancer. Int J Surg Pathol. 98. Pece, S., et al. Loss of negative regulation by Numb over Notch is relevant to human breast carcinogenesis. J Cell Biol 167, 215-221 (2004). 99. Farnie, G., et al. Novel cell culture technique for primary ductal carcinoma in situ: role of Notch and epidermal growth factor receptor signaling pathways. J Natl Cancer Inst 99, 616-627 (2007). 100. Lee, C.W., et al. A functional Notch-survivin gene signature in basal breast cancer. Breast Cancer Res 10, R97 (2008). 101. Hao, L., et al. Notch-1 activates estrogen receptor-alpha-dependent transcription via IKKalpha in breast cancer cells. Oncogene 29, 201-213. 102. Reedijk, M., et al. JAG1 expression is associated with a basal phenotype and recurrence in lymph node-negative breast cancer. Breast Cancer Res Treat 111, 439-448 (2008). 103. Adrover, E., et al. Notch and Jak/STAT signaling pathways in breast carcinoma of luminal and basal/triple-negative immunophenotypes. J Clin Oncol (Meeting Abstracts) 28, 618-. 104. Hamaguchi, Y., Matsunami, N., Yamamoto, Y. & Honjo, T. Purification and characterization of a protein that binds to the recombination signal sequence of the immunoglobulin J kappa segment. Nucleic Acids Res 17, 9015-9026 (1989). 105. Kawaichi, M., et al. Genomic organization of mouse J kappa recombination signal binding protein (RBP-J kappa) gene. J Biol Chem 267, 4016-4022 (1992). 106. Matsunami, N., et al. A protein binding to the J kappa recombination sequence of immunoglobulin genes contains a sequence related to the integrase motif. Nature 342, 934-937 (1989). 107. Honjo, T. The shortest path from the surface to the nucleus: RBP-J kappa/Su(H) transcription factor. Genes Cells 1, 1-9 (1996). 108. Chung, C.N., Hamaguchi, Y., Honjo, T. & Kawaichi, M. Site-directed mutagenesis study on DNA binding regions of the mouse homologue of Suppressor of Hairless, RBP-J kappa. Nucleic Acids Res 22, 2938-2944 (1994). 109. Tun, T., et al. Recognition sequence of a highly conserved DNA binding protein RBP-J kappa. Nucleic Acids Res 22, 965-971 (1994). 110. Liang, Y. & Ganem, D. Lytic but not latent infection by Kaposi's sarcoma-associated herpesvirus requires host CSL protein, the mediator of Notch signaling. Proc Natl Acad Sci U S A 100, 8490- 8495 (2003). 111. Jezierski, A., et al. Novel RBPJ Transcripts Identified in Human Amniotic Fluid Cells. Stem Cell Rev. 112. Dou, S., et al. The recombination signal sequence-binding protein RBP-2N functions as a transcriptional repressor. Mol Cell Biol 14, 3310-3319 (1994).

112

113. Maier, S., et al. A somatic knockout of CBF1 in a human B-cell line reveals that induction of CD21 and CCR7 by EBNA-2 is strictly CBF1 dependent and that downregulation of immunoglobulin M is partially CBF1 independent. J Virol 79, 8784-8792 (2005). 114. Amakawa, R., et al. Human Jk recombination signal binding protein gene (IGKJRB): comparison with its mouse homologue. Genomics 17, 306-315 (1993). 115. Ling, P.D., Rawlins, D.R. & Hayward, S.D. The Epstein-Barr virus immortalizing protein EBNA-2 is targeted to DNA by a cellular enhancer-binding protein. Proc Natl Acad Sci U S A 90, 9237-9241 (1993). 116. Christensen, S., Kodoyianni, V., Bosenberg, M., Friedman, L. & Kimble, J. lag-1, a gene required for lin-12 and glp-1 signaling in Caenorhabditis elegans, is homologous to human CBF1 and Drosophila Su(H). Development 122, 1373-1383 (1996). 117. Furukawa, T., et al. The Drosophila RBP-J kappa gene encodes the binding protein for the immunoglobulin J kappa recombination signal sequence. J Biol Chem 266, 23334-23340 (1991). 118. Zhou, S. & Hayward, S.D. Nuclear localization of CBF1 is regulated by interactions with the SMRT corepressor complex. Mol Cell Biol 21, 6222-6232 (2001). 119. Nam, Y., Sliz, P., Song, L., Aster, J.C. & Blacklow, S.C. Structural basis for cooperativity in recruitment of MAML coactivators to Notch transcription complexes. Cell 124, 973-983 (2006). 120. Wilson, J.J. & Kovall, R.A. Crystal structure of the CSL-Notch-Mastermind ternary complex bound to DNA. Cell 124, 985-996 (2006). 121. Hamaguchi, Y., et al. Biochemical and immunological characterization of the DNA binding protein (RBP-J kappa) to mouse J kappa recombination signal sequence. J Biochem 112, 314-320 (1992). 122. Sakai, T., et al. Loss of immunostaining of the RBP-J kappa transcription factor upon F9 cell differentiation induced by retinoic acid. J Biochem 118, 621-628 (1995). 123. Bray, S. & Furriols, M. Notch pathway: making sense of suppressor of hairless. Curr Biol 11, R217-221 (2001). 124. Nam, Y., Sliz, P., Pear, W.S., Aster, J.C. & Blacklow, S.C. Cooperative assembly of higher-order Notch complexes functions as a switch to induce transcription. Proc Natl Acad Sci U S A 104, 2103-2108 (2007). 125. Kathrein, K.L., Chari, S. & Winandy, S. Ikaros directly represses the notch target gene Hes1 in a leukemia T cell line: implications for CD4 regulation. J Biol Chem 283, 10476-10484 (2008). 126. Dumortier, A., et al. Notch activation is an early and critical event during T-Cell leukemogenesis in Ikaros-deficient mice. Mol Cell Biol 26, 209-220 (2006). 127. Oka, C., et al. Disruption of the mouse RBP-J kappa gene results in early embryonic death. Development 121, 3291-3301 (1995). 128. Tanigaki, K., et al. Notch-RBP-J signaling is involved in cell fate determination of marginal zone B cells. Nat Immunol 3, 443-450 (2002). 129. Conlon, R.A., Reaume, A.G. & Rossant, J. Notch1 is required for the coordinate segmentation of somites. Development 121, 1533-1545 (1995). 130. Yamamoto, N., Tanigaki, K., Han, H., Hiai, H. & Honjo, T. Notch/RBP-J signaling regulates epidermis/hair fate determination of hair follicular stem cells. Curr Biol 13, 333-338 (2003). 131. Welm, B.E., Dijkgraaf, G.J., Bledau, A.S., Welm, A.L. & Werb, Z. Lentiviral transduction of mammary stem cells for analysis of gene function during development and cancer. Cell Stem Cell 2, 90-102 (2008). 132. Krejci, A. & Bray, S. Notch activation stimulates transient and selective binding of Su(H)/CSL to target enhancers. Genes Dev 21, 1322-1327 (2007). 133. Tanner, M.M., et al. Genetic aberrations in hypodiploid breast cancer: frequent loss of chromosome 4 and amplification of cyclin D1 oncogene. Am J Pathol 153, 191-199 (1998). 113

134. Shivapurkar, N., et al. Multiple regions of chromosome 4 demonstrating allelic losses in breast carcinomas. Cancer Res 59, 3576-3580 (1999). 135. Kao, H.Y., et al. A histone deacetylase corepressor complex regulates the Notch signal transduction pathway. Genes Dev 12, 2269-2277 (1998). 136. Hsieh, J.J., Zhou, S., Chen, L., Young, D.B. & Hayward, S.D. CIR, a corepressor linking the DNA binding factor CBF1 to the histone deacetylase complex. Proc Natl Acad Sci U S A 96, 23-28 (1999). 137. Liefke, R., et al. Histone demethylase KDM5A is an integral part of the core Notch-RBP-J repressor complex. Genes Dev 24, 590-601. 138. Hirst, M. & Marra, M.A. Epigenetics and human disease. Int J Biochem Cell Biol 41, 136-146 (2009). 139. Bernstein, B.E., Meissner, A. & Lander, E.S. The mammalian epigenome. Cell 128, 669-681 (2007). 140. Luger, K., Mader, A.W., Richmond, R.K., Sargent, D.F. & Richmond, T.J. Crystal structure of the nucleosome core particle at 2.8 A resolution. Nature 389, 251-260 (1997). 141. Kouzarides, T. Chromatin modifications and their function. Cell 128, 693-705 (2007). 142. Huebert, D.J., Kamal, M., O'Donovan, A. & Bernstein, B.E. Genome-wide analysis of histone modifications by ChIP-on-chip. Methods 40, 365-369 (2006). 143. Vaissiere, T., Sawan, C. & Herceg, Z. Epigenetic interplay between histone modifications and DNA methylation in gene silencing. Mutat Res 659, 40-48 (2008). 144. Shi, Y. & Whetstine, J.R. Dynamic regulation of histone lysine methylation by demethylases. Mol Cell 25, 1-14 (2007). 145. Heintzman, N.D., et al. Distinct and predictive chromatin signatures of transcriptional promoters and enhancers in the human genome. Nat Genet 39, 311-318 (2007). 146. Waltzer, L., Bourillot, P.Y., Sergeant, A. & Manet, E. RBP-J kappa repression activity is mediated by a co-repressor and antagonized by the Epstein-Barr virus transcription factor EBNA2. Nucleic Acids Res 23, 4939-4945 (1995). 147. Zhou, S., et al. SKIP, a CBF1-associated protein, interacts with the ankyrin repeat domain of NotchIC To facilitate NotchIC function. Mol Cell Biol 20, 2400-2410 (2000). 148. Iso, T., Chung, G., Hamamori, Y. & Kedes, L. HERP1 is a cell type-specific primary target of Notch. J Biol Chem 277, 6598-6607 (2002). 149. Oswald, F., et al. SHARP is a novel component of the Notch/RBP-Jkappa signalling pathway. EMBO J 21, 5417-5426 (2002). 150. Alland, L., et al. Role for N-CoR and histone deacetylase in Sin3-mediated transcriptional repression. Nature 387, 49-55 (1997). 151. Kuroda, K., et al. Regulation of marginal zone B cell development by MINT, a suppressor of Notch/RBP-J signaling pathway. Immunity 18, 301-312 (2003). 152. Koelzer, S. & Klein, T. Regulation of expression of Vg and establishment of the dorsoventral compartment boundary in the wing imaginal disc by Suppressor of Hairless. Dev Biol 289, 77-90 (2006). 153. Nagel, A.C., et al. Hairless-mediated repression of notch target genes requires the combined activity of Groucho and CtBP corepressors. Mol Cell Biol 25, 10433-10441 (2005). 154. Oswald, F., et al. RBP-Jkappa/SHARP recruits CtIP/CtBP corepressors to silence Notch target genes. Mol Cell Biol 25, 10379-10390 (2005). 155. Olave, I., Reinberg, D. & Vales, L.D. The mammalian transcriptional repressor RBP (CBF1) targets TFIID and TFIIA to prevent activated transcription. Genes Dev 12, 1621-1637 (1998).

114

156. Taniguchi, Y., Furukawa, T., Tun, T., Han, H. & Honjo, T. LIM protein KyoT2 negatively regulates transcription by association with the RBP-J DNA-binding protein. Mol Cell Biol 18, 644-654 (1998). 157. Liang, L., et al. KyoT3, an isoform of murine FHL1, associates with the transcription factor RBP-J and represses the RBP-J-mediated transactivation. Biochim Biophys Acta 1779, 805-810 (2008). 158. Wang, J., et al. Opposing LSD1 complexes function in developmental gene activation and repression programmes. Nature 446, 882-887 (2007). 159. Fryer, C.J., White, J.B. & Jones, K.A. Mastermind recruits CycC:CDK8 to phosphorylate the Notch ICD and coordinate activation with turnover. Mol Cell 16, 509-520 (2004). 160. Nagai, K., et al. SKIP modifies gene expression by affecting both transcription and splicing. Biochem Biophys Res Commun 316, 512-517 (2004). 161. Tyagi, M. & Karn, J. CBF-1 promotes transcriptional silencing during the establishment of HIV-1 latency. EMBO J 26, 4985-4995 (2007). 162. Goodfellow, H., et al. Gene-specific targeting of the histone chaperone asf1 to mediate silencing. Dev Cell 13, 593-600 (2007). 163. Engel, M.E., Nguyen, H.N., Mariotti, J., Hunt, A. & Hiebert, S.W. Myeloid translocation gene 16 (MTG16) interacts with Notch transcription complex components to integrate Notch signaling in hematopoietic cell fate specification. Mol Cell Biol 30, 1852-1863. 164. Mercher, T., et al. The OTT-MAL fusion oncogene activates RBPJ-mediated transcription and induces acute megakaryoblastic leukemia in a knockin mouse model. J Clin Invest 119, 852-864 (2009). 165. Alcalay, M., et al. Acute myeloid leukemia fusion proteins deregulate genes involved in stem cell maintenance and DNA repair. J Clin Invest 112, 1751-1761 (2003). 166. Furriols, M. & Bray, S. A model Notch response element detects Suppressor of Hairless- dependent molecular switch. Curr Biol 11, 60-64 (2001). 167. Krejci, A., Bernard, F., Housden, B.E., Collins, S. & Bray, S.J. Direct response to Notch activation: signaling crosstalk and incoherent logic. Sci Signal 2, ra1 (2009). 168. Hsieh, J.J., et al. Truncated mammalian Notch1 activates CBF1/RBPJk-repressed genes by a mechanism resembling that of Epstein-Barr virus EBNA2. Mol Cell Biol 16, 952-959 (1996). 169. Oswald, F., et al. p300 acts as a transcriptional coactivator for mammalian Notch-1. Mol Cell Biol 21, 7761-7774 (2001). 170. Wallberg, A.E., Pedersen, K., Lendahl, U. & Roeder, R.G. p300 and PCAF act cooperatively to mediate transcriptional activation from chromatin templates by notch intracellular domains in vitro. Mol Cell Biol 22, 7812-7819 (2002). 171. Kurooka, H. & Honjo, T. Functional interaction between the mouse notch1 intracellular region and histone acetyltransferases PCAF and GCN5. J Biol Chem 275, 17211-17220 (2000). 172. Weng, A.P., et al. Growth suppression of pre-T acute lymphoblastic leukemia cells by inhibition of notch signaling. Mol Cell Biol 23, 655-664 (2003). 173. Luo, D., Renault, V.M. & Rando, T.A. The regulation of Notch signaling in muscle stem cell activation and postnatal myogenesis. Semin Cell Dev Biol 16, 612-622 (2005). 174. Wilkin, M.B. & Baron, M. Endocytic regulation of Notch activation and down-regulation (review). Mol Membr Biol 22, 279-289 (2005). 175. Hayward, S.D. Viral interactions with the Notch pathway. Semin Cancer Biol 14, 387-396 (2004). 176. Persson, L.M. & Wilson, A.C. Wide-scale use of Notch signaling factor CSL/RBP-Jkappa in RTA- mediated activation of Kaposi's sarcoma-associated herpesvirus lytic genes. J Virol 84, 1334- 1347. 177. Hsieh, J.J. & Hayward, S.D. Masking of the CBF1/RBPJ kappa transcriptional repression domain by Epstein-Barr virus EBNA2. Science 268, 560-563 (1995). 115

178. Beres, T.M., et al. PTF1 is an organ-specific and Notch-independent basic helix-loop-helix complex containing the mammalian Suppressor of Hairless (RBP-J) or its paralogue, RBP-L. Mol Cell Biol 26, 117-130 (2006). 179. Minoguchi, S., et al. RBP-L, a transcription factor related to RBP-Jkappa. Mol Cell Biol 17, 2679- 2687 (1997). 180. Masui, T., Long, Q., Beres, T.M., Magnuson, M.A. & MacDonald, R.J. Early pancreatic development requires the vertebrate Suppressor of Hairless (RBPJ) in the PTF1 bHLH complex. Genes Dev 21, 2629-2643 (2007). 181. Henke, R.M., et al. Neurog2 is a direct downstream target of the Ptf1a-Rbpj transcription complex in dorsal spinal cord. Development 136, 2945-2954 (2009). 182. Prevorovsky, M., et al. Cbf11 and Cbf12, the fission yeast CSL proteins, play opposing roles in cell adhesion and coordination of cell and nuclear division. Exp Cell Res 315, 1533-1547 (2009). 183. Margolin, A.A., et al. ChIP-on-chip significance analysis reveals large-scale binding and regulation by human transcription factor oncogenes. Proc Natl Acad Sci U S A 106, 244-249 (2009). 184. Pirot, P., van Grunsven, L.A., Marine, J.C., Huylebroeck, D. & Bellefroid, E.J. Direct regulation of the Nrarp gene promoter by the Notch signaling pathway. Biochem Biophys Res Commun 322, 526-534 (2004). 185. Reizis, B. & Leder, P. Direct induction of T lymphocyte-specific gene expression by the mammalian Notch signaling pathway. Genes Dev 16, 295-300 (2002). 186. Maier, M.M. & Gessler, M. Comparative analysis of the human and mouse Hey1 promoter: Hey genes are new Notch target genes. Biochem Biophys Res Commun 275, 652-660 (2000). 187. Iso, T., et al. HERP, a new primary target of Notch regulated by ligand binding. Mol Cell Biol 21, 6071-6079 (2001). 188. Noseda, M., et al. Smooth Muscle alpha-actin is a direct target of Notch/CSL. Circ Res 98, 1468- 1470 (2006). 189. Ronchini, C. & Capobianco, A.J. Induction of cyclin D1 transcription and CDK2 activity by Notch(ic): implication for cell cycle disruption in transformation by Notch(ic). Mol Cell Biol 21, 5925-5934 (2001). 190. Joshi, I., et al. Notch signaling mediates G1/S cell-cycle progression in T cells via cyclin D3 and its dependent kinases. Blood 113, 1689-1698 (2009). 191. Rangarajan, A., et al. Notch signaling is a direct determinant of keratinocyte growth arrest and entry into differentiation. EMBO J 20, 3427-3436 (2001). 192. Deftos, M.L., He, Y.W., Ojala, E.W. & Bevan, M.J. Correlating notch signaling with thymocyte maturation. Immunity 9, 777-786 (1998). 193. Chadwick, N., et al. Identification of novel Notch target genes in T cell leukaemia. Mol Cancer 8, 35 (2009). 194. Iso, T., et al. Dll4-selective Notch signaling induces ephrinB2 gene expression in endothelial cells. Biochem Biophys Res Commun 341, 708-714 (2006). 195. Grego-Bessa, J., et al. Notch signaling is essential for ventricular chamber development. Dev Cell 12, 415-429 (2007). 196. Chen, Y., Fischer, W.H. & Gill, G.N. Regulation of the ERBB-2 promoter by RBPJkappa and NOTCH. J Biol Chem 272, 14110-14114 (1997). 197. Shawber, C.J., et al. Notch alters VEGF responsiveness in human and murine endothelial cells by direct regulation of VEGFR-3 expression. J Clin Invest 117, 3369-3382 (2007). 198. Ou-Yang, H.F., et al. Notch signaling regulates the FOXP3 promoter through RBP-J- and Hes1- dependent mechanisms. Mol Cell Biochem 320, 109-114 (2009). 199. Fang, T.C., et al. Notch directly regulates Gata3 expression during T helper 2 cell differentiation. Immunity 27, 100-110 (2007). 116

200. Amsen, D., et al. Direct regulation of Gata3 expression determines the T helper differentiation potential of Notch. Immunity 27, 89-99 (2007). 201. Ge, W., et al. Notch signaling promotes astrogliogenesis via direct CSL-mediated glial gene activation. J Neurosci Res 69, 848-860 (2002). 202. Jarriault, S., et al. Signalling downstream of activated mammalian Notch. Nature 377, 355-358 (1995). 203. Nishimura, M., et al. Structure, chromosomal locus, and promoter of mouse Hes2 gene, a homologue of Drosophila hairy and Enhancer of split. Genomics 49, 69-75 (1998). 204. Bessho, Y., Miyoshi, G., Sakata, R. & Kageyama, R. Hes7: a bHLH-type repressor gene regulated by Notch and expressed in the presomitic mesoderm. Genes Cells 6, 175-185 (2001). 205. Gustafsson, M.V., et al. Hypoxia requires notch signaling to maintain the undifferentiated cell state. Dev Cell 9, 617-628 (2005). 206. Kannabiran, C., Zeng, X. & Vales, L.D. The mammalian transcriptional repressor RBP (CBF1) regulates interleukin-6 gene expression. Mol Cell Biol 17, 1-9 (1997). 207. Palmieri, M., et al. Interaction of the nuclear protein CBF1 with the kappaB site of the IL-6 gene promoter. Nucleic Acids Res 27, 2785-2791 (1999). 208. Weng, A.P., et al. c-Myc is an important direct target of Notch1 in T-cell acute lymphoblastic leukemia/lymphoma. Genes Dev 20, 2096-2109 (2006). 209. Klinakis, A., et al. Myc is a Notch1 transcriptional target and a requisite for Notch1-induced mammary tumorigenesis in mice. Proc Natl Acad Sci U S A 103, 9262-9267 (2006). 210. Palomero, T., et al. NOTCH1 directly regulates c-MYC and activates a feed-forward-loop transcriptional network promoting leukemic cell growth. Proc Natl Acad Sci U S A 103, 18261- 18266 (2006). 211. Oswald, F., Liptay, S., Adler, G. & Schmid, R.M. NF-kappaB2 is a putative target gene of activated Notch-1 via RBP-Jkappa. Mol Cell Biol 18, 2077-2088 (1998). 212. Vilimas, T., et al. Targeting the NF-kappaB signaling pathway in Notch1-induced T-cell leukemia. Nat Med 13, 70-77 (2007). 213. Oakley, F., et al. Basal expression of IkappaBalpha is controlled by the mammalian transcriptional repressor RBP-J (CBF1) and its activator Notch1. J Biol Chem 278, 24359-24370 (2003). 214. Krebs, L.T., et al. Notch signaling regulates left-right asymmetry determination by inducing Nodal expression. Genes Dev 17, 1207-1212 (2003). 215. Jin, S., et al. Notch signaling regulates platelet-derived growth factor receptor-beta expression in vascular smooth muscle cells. Circ Res 102, 1483-1491 (2008). 216. Sarmento, L.M., et al. Notch1 modulates timing of G1-S progression by inducing SKP2 transcription and p27 Kip1 degradation. J Exp Med 202, 157-168 (2005). 217. Sahlgren, C., Gustafsson, M.V., Jin, S., Poellinger, L. & Lendahl, U. Notch signaling mediates hypoxia-induced tumor cell migration and invasion. Proc Natl Acad Sci U S A 105, 6392-6397 (2008). 218. Niessen, K., et al. Slug is a direct Notch target required for initiation of cardiac cushion cellularization. J Cell Biol 182, 315-325 (2008). 219. Sivasankaran, B., et al. Tenascin-C is a novel RBPJkappa-induced target gene for Notch signaling in gliomas. Cancer Res 69, 458-465 (2009). 220. Takata, T. & Ishikawa, F. Human Sir2-related protein SIRT1 associates with the bHLH repressors HES1 and HEY2 and is involved in HES1- and HEY2-mediated transcriptional repression. Biochem Biophys Res Commun 301, 250-257 (2003). 221. King, I.N., et al. Hrt and Hes negatively regulate Notch signaling through interactions with RBP- Jkappa. Biochem Biophys Res Commun 345, 446-452 (2006). 117

222. Diez, H., et al. Hypoxia-mediated activation of Dll4-Notch-Hey2 signaling in endothelial progenitor cells and adoption of arterial cell fate. Exp Cell Res 313, 1-9 (2007). 223. Bessho, Y., et al. Dynamic expression and essential functions of Hes7 in somite segmentation. Genes Dev 15, 2642-2647 (2001). 224. Ishibashi, M., et al. Targeted disruption of mammalian hairy and Enhancer of split homolog-1 (HES-1) leads to up-regulation of neural helix-loop-helix factors, premature neurogenesis, and severe neural tube defects. Genes Dev 9, 3136-3148 (1995). 225. Fischer, A., et al. Hey bHLH factors in cardiovascular development. Cold Spring Harb Symp Quant Biol 67, 63-70 (2002). 226. Fischer, A., et al. Combined loss of Hey1 and HeyL causes congenital heart defects because of impaired epithelial to mesenchymal transition. Circ Res 100, 856-863 (2007). 227. Sakamoto, M., Hirata, H., Ohtsuka, T., Bessho, Y. & Kageyama, R. The basic helix-loop-helix genes Hesr1/Hey1 and Hesr2/Hey2 regulate maintenance of neural precursor cells in the brain. J Biol Chem 278, 44808-44815 (2003). 228. Fischer, A., Schumacher, N., Maier, M., Sendtner, M. & Gessler, M. The Notch target genes Hey1 and Hey2 are required for embryonic vascular development. Genes Dev 18, 901-911 (2004). 229. Gessler, M., et al. Mouse gridlock: no aortic coarctation or deficiency, but fatal cardiac defects in Hey2 -/- mice. Curr Biol 12, 1601-1604 (2002). 230. Slamon, D.J., et al. Studies of the HER-2/neu proto-oncogene in human breast and ovarian cancer. Science 244, 707-712 (1989). 231. Traub, F., Mengel, M., Luck, H.J., Kreipe, H.H. & von Wasielewski, R. Prognostic impact of Skp2 and p27 in human breast cancer. Breast Cancer Res Treat 99, 185-191 (2006). 232. Signoretti, S., et al. Oncogenic role of the ubiquitin ligase subunit Skp2 in human breast cancer. J Clin Invest 110, 633-641 (2002). 233. Wise, D.R., et al. Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis and leads to glutamine addiction. Proc Natl Acad Sci U S A 105, 18782-18787 (2008). 234. Yeung, S.J., Pan, J. & Lee, M.H. Roles of p53, MYC and HIF-1 in regulating glycolysis - the seventh hallmark of cancer. Cell Mol Life Sci 65, 3981-3999 (2008). 235. Efstratiadis, A., Szabolcs, M. & Klinakis, A. Notch, Myc and breast cancer. Cell Cycle 6, 418-429 (2007). 236. Phng, L.K. & Gerhardt, H. Angiogenesis: a team effort coordinated by notch. Dev Cell 16, 196- 208 (2009). 237. Hancox, R.A., et al. Tumour-associated tenascin-C isoforms promote breast cancer cell invasion and growth by matrix metalloproteinase-dependent and independent mechanisms. Breast Cancer Res 11, R24 (2009). 238. Jechlinger, M., et al. Autocrine PDGFR signaling promotes mammary cancer metastasis. J Clin Invest 116, 1561-1570 (2006). 239. Chou, J., Provot, S. & Werb, Z. GATA3 in development and cancer differentiation: cells GATA have it! J Cell Physiol 222, 42-49. 240. Zuo, T., et al. FOXP3 is an X-linked breast cancer suppressor gene and an important repressor of the HER-2/ErbB2 oncogene. Cell 129, 1275-1286 (2007). 241. Zuo, T., et al. FOXP3 is a novel transcriptional repressor for the breast cancer oncogene SKP2. J Clin Invest 117, 3765-3773 (2007). 242. Merlo, A., et al. FOXP3 expression and overall survival in breast cancer. J Clin Oncol 27, 1746- 1752 (2009). 243. Knupfer, H. & Preiss, R. Significance of interleukin-6 (IL-6) in breast cancer (review). Breast Cancer Res Treat 102, 129-135 (2007). 118

244. Ayyanan, A., et al. Increased Wnt signaling triggers oncogenic conversion of human breast epithelial cells by a Notch-dependent mechanism. Proc Natl Acad Sci U S A 103, 3799-3804 (2006). 245. Collu, G.M. & Brennan, K. Cooperation between Wnt and Notch signalling in human breast cancer. Breast Cancer Res 9, 105 (2007). 246. Palomero, T., et al. Mutational loss of PTEN induces resistance to NOTCH1 inhibition in T-cell leukemia. Nat Med 13, 1203-1210 (2007). 247. Efferson, C.L., et al. Downregulation of Notch pathway by a gamma-secretase inhibitor attenuates AKT/mammalian target of rapamycin signaling and glucose uptake in an ERBB2 transgenic breast cancer model. Cancer Res 70, 2476-2484. 248. Meurette, O., et al. Notch activation induces Akt signaling via an autocrine loop to prevent apoptosis in breast epithelial cells. Cancer Res 69, 5015-5022 (2009). 249. Bedogni, B., Warneke, J.A., Nickoloff, B.J., Giaccia, A.J. & Powell, M.B. Notch1 is an effector of Akt and hypoxia in melanoma development. J Clin Invest 118, 3660-3670 (2008). 250. Kluppel, M. & Wrana, J.L. Turning it up a Notch: cross-talk between TGF beta and Notch signaling. Bioessays 27, 115-118 (2005). 251. Fu, Y., et al. Differential regulation of transforming growth factor beta signaling pathways by Notch in human endothelial cells. J Biol Chem 284, 19452-19462 (2009). 252. Osipo, C., Golde, T.E., Osborne, B.A. & Miele, L.A. Off the beaten pathway: the complex cross talk between Notch and NF-kappaB. Lab Invest 88, 11-17 (2008). 253. Lee, S.H., Wang, X. & DeJong, J. Functional interactions between an atypical NF-kappaB site from the rat CYP2B1 promoter and the transcriptional repressor RBP-Jkappa/CBF1. Nucleic Acids Res 28, 2091-2098 (2000). 254. Kamakura, S., et al. Hes binding to STAT3 mediates crosstalk between Notch and JAK-STAT signalling. Nat Cell Biol 6, 547-554 (2004). 255. Mazzone, M., et al. Dose-dependent induction of distinct phenotypic responses to Notch pathway activation in mammary epithelial cells. Proc Natl Acad Sci U S A 107, 5012-5017. 256. Chapman, R.S., et al. Suppression of epithelial apoptosis and delayed mammary gland involution in mice with a conditional knockout of Stat3. Genes Dev 13, 2604-2616 (1999). 257. Hsieh, F.C., Cheng, G. & Lin, J. Evaluation of potential Stat3-regulated genes in human breast cancer. Biochem Biophys Res Commun 335, 292-299 (2005). 258. Walker, S.R., et al. Reciprocal effects of STAT5 and STAT3 in breast cancer. Mol Cancer Res 7, 966-976 (2009). 259. Chari, S., Umetsu, S.E. & Winandy, S. Notch Target Gene Deregulation and Maintenance of the Leukemogenic Phenotype Do Not Require RBP-J{kappa} in Ikaros Null Mice. J Immunol. 260. Neve, R.M., et al. A collection of breast cancer cell lines for the study of functionally distinct cancer subtypes. Cancer Cell 10, 515-527 (2006). 261. Kato, H., et al. Involvement of RBP-J in biological functions of mouse Notch1 and its derivatives. Development 124, 4133-4141 (1997). 262. Rubinson, D.A., et al. A lentivirus-based system to functionally silence genes in primary mammalian cells, stem cells and transgenic mice by RNA interference. Nat Genet 33, 401-406 (2003). 263. MacKenzie, F., et al. Notch4-induced inhibition of endothelial sprouting requires the ankyrin repeats and involves signaling through RBP-Jkappa. Blood 104, 1760-1768 (2004). 264. Wang, H., et al. Genetic and epigenetic mechanisms combine to control MMP1 expression and its association with preterm premature rupture of membranes. Hum Mol Genet 17, 1087-1096 (2008).

119

265. Chou, Y.T., Wang, H., Chen, Y., Danielpour, D. & Yang, Y.C. Cited2 modulates TGF-beta-mediated upregulation of MMP9. Oncogene 25, 5547-5560 (2006). 266. Saito, T., Sasaki, C.Y., Rezanka, L.J., Ghosh, P. & Longo, D.L. p52-Independent nuclear translocation of RelB promotes LPS-induced attachment. Biochem Biophys Res Commun 391, 235-241. 267. Yu, K., et al. A precisely regulated gene expression cassette potently modulates metastasis and survival in multiple solid cancers. PLoS Genet 4, e1000129 (2008). 268. Lockwood, W.W., et al. DNA amplification is a ubiquitous mechanism of oncogene activation in lung and other cancers. Oncogene 27, 4615-4624 (2008). 269. Chari, R., Coe, B.P., Vucic, E.A., Lockwood, W.W. & Lam, W.L. An integrative multi-dimensional genetic and epigenetic strategy to identify aberrant genes and pathways in cancer. BMC Syst Biol 4, 67. 270. Vivanco, I., et al. Identification of the JNK signaling pathway as a functional target of the tumor suppressor PTEN. Cancer Cell 11, 555-569 (2007). 271. Li, J., et al. An improved method for genome wide DNA methylation profiling correlated to transcription and genomic instability in two breast cancer cell lines. BMC Genomics 10, 223 (2009). 272. Bild, A.H., et al. Oncogenic pathway signatures in human cancers as a guide to targeted therapies. Nature 439, 353-357 (2006). 273. Ginestier, C., et al. Prognosis and gene expression profiling of 20q13-amplified breast cancers. Clin Cancer Res 12, 4533-4544 (2006). 274. Zhao, H., et al. Different gene expression patterns in invasive lobular and ductal carcinomas of the breast. Mol Biol Cell 15, 2523-2536 (2004). 275. Lockwood, W.W., et al. Integrative genomic analyses identify BRF2 as a novel lineage-specific oncogene in lung squamous cell carcinoma. PLoS Med 7, e1000315. 276. Chari, R., et al. SIGMA: a system for integrative genomic microarray analysis of cancer genomes. BMC Genomics 7, 324 (2006). 277. Ishkanian, A.S., et al. A tiling resolution DNA microarray with complete coverage of the human genome. Nat Genet 36, 299-303 (2004). 278. Kyle, A.H., Huxham, L.A., Baker, J.H., Burston, H.E. & Minchinton, A.I. Tumor distribution of bromodeoxyuridine-labeled cells is strongly dose dependent. Cancer Res 63, 5707-5711 (2003). 279. Huxham, L.A., Kyle, A.H., Baker, J.H., Nykilchuk, L.K. & Minchinton, A.I. Microregional effects of gemcitabine in HCT-116 xenografts. Cancer Res 64, 6537-6541 (2004). 280. Robertson, A.G., et al. Genome-wide relationship between histone H3 lysine 4 mono- and trimethylation and transcription factor binding. Genome Res 18, 1906-1917 (2008). 281. Robertson, G., et al. Genome-wide profiles of STAT1 DNA association using chromatin immunoprecipitation and massively parallel sequencing. Nat Methods 4, 651-657 (2007). 282. Davies, J.J., Wilson, I.M. & Lam, W.L. Array CGH technologies and their applications to cancer genomes. Chromosome Res 13, 237-248 (2005). 283. Gazdar, A.F., et al. Characterization of paired tumor and non-tumor cell lines established from patients with breast cancer. Int J Cancer 78, 766-774 (1998). 284. Meltzer, P., et al. Establishment of two new cell lines derived from human breast carcinomas with HER-2/neu amplification. Br J Cancer 63, 727-735 (1991). 285. Kao, J., et al. Molecular profiling of breast cancer cell lines defines relevant tumor models and provides a resource for cancer gene discovery. PLoS One 4, e6146 (2009). 286. Rhodes, D.R., et al. ONCOMINE: a cancer microarray database and integrated data-mining platform. Neoplasia 6, 1-6 (2004).

120

287. van Es, J.H., et al. Notch/gamma-secretase inhibition turns proliferative cells in intestinal crypts and adenomas into goblet cells. Nature 435, 959-963 (2005). 288. Maillard, I., Adler, S.H. & Pear, W.S. Notch and the immune system. Immunity 19, 781-791 (2003). 289. Saito, S., Morita, K. & Hirano, T. High frequency of common DNA copy number abnormalities detected by bacterial artificial chromosome array comparative genomic hybridization in 24 breast cancer cell lines. Hum Cell 22, 1-10 (2009). 290. Pierga, J.Y., et al. Microarray-based comparative genomic hybridisation of breast cancer patients receiving neoadjuvant chemotherapy. Br J Cancer 96, 341-351 (2007). 291. Gonzalez-Angulo, A.M., et al. Biologic markers in axillary node-negative breast cancer: differential expression in invasive ductal carcinoma versus invasive lobular carcinoma. Clin Breast Cancer 7, 396-400 (2006). 292. Schroeter, E.H., Kisslinger, J.A. & Kopan, R. Notch-1 signalling requires ligand-induced proteolytic release of intracellular domain. Nature 393, 382-386 (1998). 293. Chari, S. & Winandy, S. Ikaros regulates Notch target gene expression in developing thymocytes. J Immunol 181, 6265-6274 (2008). 294. Bray, S. & Bernard, F. Notch Targets and Their Regulation. in Current Topics in Developmental Biology, Vol. Volume 92 (ed. Raphael, K.) 253-275 (Academic Press, 2010). 295. Minn, A.J., et al. Genes that mediate breast cancer metastasis to lung. Nature 436, 518-524 (2005). 296. Shendure, J. The beginning of the end for microarrays? Nat Methods 5, 585-587 (2008). 297. Yao, B., et al. Accuracy of cDNA microarray methods to detect small gene expression changes induced by neuregulin on breast epithelial cells. BMC Bioinformatics 5, 99 (2004). 298. Yuen, T., Wurmbach, E., Pfeffer, R.L., Ebersole, B.J. & Sealfon, S.C. Accuracy and calibration of commercial oligonucleotide and custom cDNA microarrays. Nucleic Acids Res 30, e48 (2002). 299. Hurd, P.J. & Nelson, C.J. Advantages of next-generation sequencing versus the microarray in epigenetic research. Brief Funct Genomic Proteomic 8, 174-183 (2009). 300. Barski, A. & Zhao, K. Genomic location analysis by ChIP-Seq. J Cell Biochem 107, 11-18 (2009). 301. Ramsey, S.A., et al. Genome-wide histone acetylation data improve prediction of mammalian transcription factor binding sites. Bioinformatics 26, 2071-2075. 302. Perissi, V., Jepsen, K., Glass, C.K. & Rosenfeld, M.G. Deconstructing repression: evolving models of co-repressor action. Nat Rev Genet 11, 109-123. 303. Benjamini, Y. & Hochberg, Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological) 57, 289-300 (1995). 304. Purow, B.W., et al. Expression of Notch-1 and its ligands, Delta-like-1 and Jagged-1, is critical for glioma cell survival and proliferation. Cancer Res 65, 2353-2363 (2005). 305. Vaupel, P. & Harrison, L. Tumor hypoxia: causative factors, compensatory mechanisms, and cellular response. Oncologist 9 Suppl 5, 4-9 (2004). 306. Egeblad, M. & Werb, Z. New functions for the matrix metalloproteinases in cancer progression. Nat Rev Cancer 2, 161-174 (2002). 307. Flanagan, L.A., Rebaza, L.M., Derzic, S., Schwartz, P.H. & Monuki, E.S. Regulation of human neural precursor cells by laminin and integrins. J Neurosci Res 83, 845-856 (2006). 308. Laios, A., et al. Potential role of miR-9 and miR-223 in recurrent ovarian cancer. Mol Cancer 7, 35 (2008). 309. Majumdar, A., Nagaraj, R. & Banerjee, U. strawberry notch encodes a conserved nuclear protein that functions downstream of Notch and regulates gene expression along the developing wing margin of Drosophila. Genes Dev 11, 1341-1353 (1997). 121

310. Lai, E.C. Developmental signaling: shrimp and strawberries help flies make cones. Curr Biol 12, R722-724 (2002). 311. Baba, K., et al. A mouse homologue of Strawberry Notch is transcriptionally regulated by Reelin signal. Biochem Biophys Res Commun 350, 842-849 (2006). 312. Pervin, S., Singh, R. & Chaudhuri, G. Nitric oxide, N omega-hydroxy-L-arginine and breast cancer. Nitric Oxide 19, 103-106 (2008). 313. Huber, C., Sundstrom, C., Nilsson, K. & Wigzell, H. Surface receptors on human haematopoietic cell lines. Clin Exp Immunol 25, 367-376 (1976). 314. Worsham, M.J., et al. High-resolution mapping of molecular events associated with immortalization, transformation, and progression to breast cancer in the MCF10 model. Breast Cancer Res Treat 96, 177-186 (2006). 315. Kwon, M.J., et al. Derepression of CLDN3 and CLDN4 during ovarian tumorigenesis is associated with loss of repressive histone modifications. Carcinogenesis 31, 974-983. 316. Ohtani, N., Yamakoshi, K., Takahashi, A. & Hara, E. Real-time in vivo imaging of p16gene expression: a new approach to study senescence stress signaling in living animals. Cell Div 5, 1. 317. Shawber, C., et al. Notch signaling inhibits muscle cell differentiation through a CBF1- independent pathway. Development 122, 3765-3773 (1996). 318. Cory, S. & Adams, J.M. The Bcl2 family: regulators of the cellular life-or-death switch. Nat Rev Cancer 2, 647-656 (2002). 319. Wang, Z., et al. Down-regulation of notch-1 inhibits invasion by inactivation of nuclear factor- kappaB, vascular endothelial growth factor, and matrix metalloproteinase-9 in pancreatic cancer cells. Cancer Res 66, 2778-2784 (2006). 320. Wang, Z., et al. Down-regulation of Notch-1 and Jagged-1 inhibits prostate cancer cell growth, migration and invasion, and induces apoptosis via inactivation of Akt, mTOR, and NF-kappaB signaling pathways. J Cell Biochem 109, 726-736. 321. Garg, P., et al. Matrix metalloproteinase-9 regulates MUC-2 expression through its effect on goblet cell differentiation. Gastroenterology 132, 1877-1889 (2007). 322. Garg, P., et al. Matrix metalloproteinase-9 functions as a tumor suppressor in colitis-associated cancer. Cancer Res 70, 792-801. 323. Karin, M. & Greten, F.R. NF-kappaB: linking inflammation and immunity to cancer development and progression. Nat Rev Immunol 5, 749-759 (2005). 324. Karin, M. Nuclear factor-kappaB in cancer development and progression. Nature 441, 431-436 (2006). 325. McElhinny, A.S., Li, J.L. & Wu, L. Mastermind-like transcriptional co-activators: emerging roles in regulating cross talk among multiple signaling pathways. Oncogene 27, 5138-5147 (2008). 326. Jin, B., et al. The mastermind-like 1 (MAML1) co-activator regulates constitutive NF-kappaB signaling and cell survival. J Biol Chem 285, 14356-14365. 327. Zarzynska, J., Gajewska, M. & Motyl, T. Effects of hormones and growth factors on TGF-beta1 expression in bovine mammary epithelial cells. J Dairy Res 72, 39-48 (2005). 328. Page-McCaw, A., Ewald, A.J. & Werb, Z. Matrix metalloproteinases and the regulation of tissue remodelling. Nat Rev Mol Cell Biol 8, 221-233 (2007). 329. Simeone, A.M., et al. N-(4-Hydroxyphenyl)retinamide and nitric oxide pro-drugs exhibit apoptotic and anti-invasive effects against bone metastatic breast cancer cells. Carcinogenesis 27, 568-577 (2006). 330. Hofmann, J.J. & Iruela-Arispe, M.L. Notch signaling in blood vessels: who is talking to whom about what? Circ Res 100, 1556-1568 (2007).

122

APPENDIX

Appendix A. Previously published material

This previous publication was a collaborative effort resulting in shared co-first authorship with lab members Kevin Leong and Kyle Niessen. The major findings of the publication and my contributions to this manuscript are detailed below.

Kevin G. Leong*, Kyle Niessen*, Iva Kulic*, Afshin Raouf, Connie Eaves, Ingrid Pollet, Aly Karsan. Jagged1-mediated Notch activation induces epithelial-to-mesenchymal transition through Slug-induced repression of E-cadherin. Journal of Experimental Medicine 204, 2935-48 (2007).

This peer-reviewed article, featured on the cover of the Journal of Experimental Medicine, identifies the protein Slug (SNAI2) as a novel direct target of Notch signaling, drawing from in vitro evidence, animal experiments and expression correlations in human breast cancers. As Slug plays a key role in cancer cell survival and migration, these findings implicate Notch in inhibition of anoikis and in eliciting an epithelial-to-mesenchymal transformation in tumor cells. The work provides a mechanism for how Notch receptor signaling mediates its oncogenic affects in a breast cancer model.

I created Illustrator files of the figures for publication, and directly generated the data, and wrote the results and material and methods sections for the following figures:

Figure 2. Slug is a direct target of Notch signaling.

Figure 7. The Notch-Slug axis inhibits anoikis of human breast cells.

123

The Journal of Experimental Medicine Published November26,2007 o.24 o 2 Nvme 6 07 9524 www.jem.org/cgi/doi/ 2935-2948 November 26,2007 Vol. 204,No. 12, $30.00 JEM ©TheRockefeller University Press fl trichostatin A;YFP,yellow small interferingRNA;TSA, short hairpinRNA;siRNA, quantitative RT-PCR;shRNA, tracellular domain;qPCR, butyrate; NotchIC,Notchin- specifi deacetylase; MSP,methylation- glutinin; HDAC,histone mal transition;HA,hemag- EMT, epithelial-to-mesenchy- cytosine-phosphate-guanine; mentary DNA;CpG, 5-azacytidine; cDNA,comple- Abbreviations used:5AZA, [email protected] Aly Karsan: CORRESPONDENCE uorescent protein. uorescent cPCR;NaBu,sodium

Connie Eaves, The online version of this article contains supplemental material. Theonlineversionof thisarticlecontainssupplementalmaterial. this work. K.G. Leong, K.Niessen,andI.Kuliccontributedequallyto receptor expressedonanadjacentcell( gand interactswithaNotchtransmembrane Notchsignalingisinitiatedwhenali- mesenchymal transitionandinhibitionofanoikis. induction ofSlug,promotestumorgrowthandmetastasischaracterizedbyepithelial-to- active ated withdown-regulatedSlugexpression,reexpressionofE-cadherin,andsuppression promoted apoptosisandinhibitedtumorgrowthmetastasis.Thisresponsewasassoci- induced NotchsignalinginxenograftedSlug-positive/E-cadherin – negative breasttumors which resultedin various humancancers.SlugwasessentialforNotch-mediatedrepressionofE-cadherin, and showthatelevatedlevelsofSlugcorrelatewithincreasedexpressionJagged1in prognosis isunknown.We identifySlug,atranscriptionalrepressor, asanovelNotchtarget cancer. However, thereasonthatJagged1and/orNotchoverexpressionportendsapoor Aberrant expressionofJagged1andNotch1areassociatedwithpooroutcomeinbreast 4 1 Kevin G. Leong, Slug-induced repression ofE-cadherin epithelial-to-mesenchymal transitionthrough Jagged1-mediated Notchactivation induces reported inhumanbreasttumors atvarying Expression ofallfourNotchreceptors hasbeen naling inhumanbreastcancer development. also highlightedapotentialroleforNotchsig- static breasttumors( 3, 4 ). Recent studieshave in miceleadstotheformationofaggressive,meta- tively activeNotch1IC,Notch3IC,orNotch4IC ing onthetumortype( 2 ). an oncogeneoratumorsuppressorgenedepend- development ofcancer,Notchmayactaseither proliferation, andapoptosis.Interestingly,inthe ous cellularprocesses,includingdiff stream targetgenesandtherebyregulatingvari- derepression andcoactivationofNotchdown- to thetranscriptionalrepressorCSL,resultingin the nucleus.Withinnucleus,NotchICbinds main (NotchIC),allowingittotranslocateinto tions thatreleasestheNotchintracellulardo- interaction triggersaseriesofproteolyticdiges- University ofBritishColumbia,Vancouver, BritishColumbiaV6T2B5,Canada British ColumbiaCancerAgency, Vancouver, BritishColumbiaV5Z1L3,Canada Experimental MedicineProgram, Department ofMedicalBiophysics, Mammary-specifi ␤ -catenin. Ourfi ndingssuggestthatligand-induced Notchactivation,throughthe 3,4,5,6 ␤ -catenin activationandresistancetoanoikis.Inhibitionofligand- c overexpressionofconstitu- 10.1084/jem.20071082 1,2,4 Pollet, Ingrid Kyle Niessen, 5 Department ofPathology andLaboratoryMedicine, 2 Department ofPathology andLaboratoryMedicine, erentiation, erentiation, http://jem.rupress.org/content/suppl/2007/11/20/jem.20071082.DC2.html http://jem.rupress.org/content/suppl/2007/11/19/jem.20071082.DC1.html Supplemental Material canbefoundat: 1 ). This ). This 1,2,5 1,2,4 and Aly Karsan

Iva Kulic, poor outcome( mRNA defi level coexpressionofJagged1andNotch1 cancer hasrecentlybeenreported,withhigh- and reducedsurvivalinpatientswithbreast mRNA expressionoftheNotchligandJagged1 Interestingly, anassociationbetweenincreased tein levelsandreducedpatientsurvival( tumors areassociatedwithelevatedNotch1 frequencies ( ( epithelial cellgrowthorinhibitingapoptosis ute totumorigenesisbypromotingmammary receptors ( 8, 9 ). Notchsignalingmaycontrib- through theinductionofNotchligandsand/or Wnt inthepromotionofmammarytumors ported tobeactivateddownstreamofRasand ing normalheart developmentandthatNotch naling inducesaspecializedtype ofEMTdur- Recent studieshavesuggested thatNotchsig- malignant cells,particularlyinbreast cancer( promoting theinvasionand dissemination of epithelial-to-mesenchymal transition (EMT)in mediated oncogenesis. about themolecularmechanismsofNotch- 10, 11 ). However,muchremainstobelearned Numerousreportshaveindicatedarolefor 1,2,4 ning asubsetofpatientswithvery 5 ). Poorlydifferentiatedbreast Afshin Raouf, Raouf, Afshin 1,2,4,5 7 ). Notchhasalsobeenre- 6 Department ofMedicalGenetics, 3 Terry FoxLaboratory,

3 ARTICLE

pro- 2935 12 ).

6 ).

Downloaded from from Downloaded on July 22, 2010 22, July on jem.rupress.org Published November 26, 2007

up-regulates Snail in endothelial cells to promote mesenchy- positive human breast cancers. Jagged1-mediated activation mal transformation ( 13, 14 ). However, there is no direct or of Notch in breast epithelial cells induces EMT through even correlative in vivo data that Notch regulates EMT in induction of Slug and subsequent repression of the cell– cell epithelial cancers. adhesion protein E-cadherin. Because Slug can be induced In this paper, we identify Slug, a zinc-fi nger transcrip- by factors other than Notch, we identify Notch downstream tional repressor functionally linked to human breast cancer target genes of the HEY family as potential markers of pri- progression and metastasis (15 ), to be a direct downstream target mary human breast tumors that have activated the Jagged1– gene of Notch that is up-regulated in Jagged1- and Notch1- Notch– Slug signaling axis. In Slug-positive/E-cadherin– negative Downloaded from jem.rupress.org on July 22, 2010

Figure 1. Notch activation inhibits E-cadherin expression in human breast epithelial cells through the induction of Slug. (A) Immunofl uorescent staining for E-cadherin (red), YFP (green), and DAPI (blue) in primary human breast epithelial cells transduced with MIY, MIYNotch1IC, or MIYNotch4IC. Bar, 50 ␮ m. (B) qPCR for expression of E-cadherin, Slug, Snail, and Twist1 in MCF-10A cells transduced with MIY or MIYNotch1IC. Data are expressed as the relative gene expression level, with MIY control as the comparator, and are from three independent experiments (mean + SEM). *, P ≤ 0.05. (C) Immuno- fl uorescent staining for Slug (red), YFP (green), and DAPI (blue) in MCF-10A cells transduced with MIY or MIYNotch1IC. Bar, 50 ␮ m. (D) qPCR for expression of Slug and E-cadherin in MCF-10A cell lines (MIY, MIYNotch1IC, or MIYSlug) transiently transfected with siRandom or siSlug. Data from two independent experiments are shown and are expressed as the relative gene expression level with MIY control as the comparator. n.d., not detectable.

2936 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ARTICLE

human breast cancer xenografts, inhibition of ligand-induced each of the labeled oligonucleotides was competed away with Notch signaling inhibits growth of the primary tumor and unlabeled wild-type, but not mutated, double-stranded oligo- distant metastases, which correlates with reduced Slug expres- nucleotide (Fig. 2, A and B). In addition, lentiviral-delivered sion and reexpression of E-cadherin. E-cadherin reexpression, short hairpin RNAs (shRNAs) targeting two distinct sites of either through Notch inhibition or enforced expression, is CSL (Fig. 2 C ) also prevented the gel shift, indicating that these associated with relocalization of ␤ -catenin from the nucleus sites in the Slug promoter bind endogenous CSL and, thus, to the plasma membrane and reversal of ␤ -catenin activation confi rming that Slug is a direct target of Notch/CSL. These in xenografted breast tumors. Our fi ndings suggest a critical results, combined with our data demonstrating the ability of Slug role for induction of EMT and inhibition of anoikis in pro- alone (independent of Notch1IC) to induce a spindle-shaped moting an aggressive phenotype in tumors exhibiting ligand- morphology and down-regulate E-cadherin expression in nor- induced Notch signaling. mal breast epithelial cells (Fig. S2, available at http://www.jem .org/cgi/content/full/jem.20071082/DC1), identify Slug as a RESULTS key player in the mechanism of Notch-induced EMT. Notch activation inhibits E-cadherin expression in human breast epithelial cells through the induction of Slug Jagged1/Notch1 correlates with Slug expression in human Down-regulation of E-cadherin is one of the best markers of breast cancers EMT in human breast cancer ( 12). To determine whether To determine whether Jagged1-triggered Notch activation Notch activation induces EMT in human breast epithelial could induce Slug expression, Jagged1-expressing mouse en- Downloaded from cells as manifested by repression of E-cadherin, the E-cadherin – dothelial cells (from the SVEC 4-10 cell line) were co-cultured positive normal human breast epithelial cell line MCF-10A was transduced with a retroviral vector (MIY) linking yellow fl uorescent protein (YFP) to activated Notch1 (Notch1IC) or activated Notch4 (Notch4IC). Hence, cells that express Notch1IC or Notch4IC also express YFP. Expression of either Notch1IC or Notch4IC caused this normal breast epithelial jem.rupress.org cell line to down-regulate E-cadherin, dissociate cell– cell contacts, and acquire a spindle-shaped morphology, consistent with mesenchymal transformation (Fig. S1, A– C, available at http://www.jem.org/cgi/content/full/jem.20071082/DC1). A similar ability of activated Notch to down-regulate E-cad- on July 22, 2010 herin was demonstrated in primary human breast epithelial cells (Fig. 1 A ). To identify a potential mechanism of Notch-mediated E-cadherin silencing, expression of three known E-cadherin repressors that initiate EMT in breast cancer— Slug, Snail, and Twist1 (16 ) — was assessed in a normal breast epithelial cell line by quantitative RT-PCR (qPCR). In contrast to what has been reported in endothelial cells (13 ), Snail mRNA expression was not detected in either Notch1IC-expressing or control cells. Twist1 mRNA levels did not diff er between control and Notch1IC-expressing cells (Fig. 1 B). Slug mRNA expression, however, was signifi cantly increased in Notch1IC cells, which was associated with a decrease in E-cadherin mRNA expression (Fig. 1 B). Notch1IC-induced expression of Slug protein was confi rmed by immunofl uorescence microscopy ( Fig. 1 C ). Knockdown of Slug, achieved by delivering small interfering RNA (siRNA) targeting Slug, into cells trans- Figure 2. Slug is a direct target of Notch signaling. EMSA of duced with either Notch1IC or Slug was suffi cient to restore nuclear lysates from MDA-MB-231 human breast cancer cells transduced E-cadherin expression (Fig. 1 D ). with nonspecifi c shRNA (shRand) or two different shRNAs targeting CSL To determine whether Slug is a direct target of Notch/ (shCSL1 and shCSL2). CSL consensus binding sites in the human Slug CSL, we examined the human Slug promoter and identifi ed promoter (A, − 846 to −853 [TATGGGAA]; and B, − 1686 to − 1679 [TGTGGGAA]) relative to the transcriptional start site were used as the 32 P- two potential CSL-binding consensus motifs ( −846 to −853 labeled probe, and either nonradioactive wild-type or mutated (mt) oligo- and − 1686 to − 1679 relative to the transcriptional start site). nucleotides were used as competitors in 50-fold excess. The CSL – DNA EMSAs using double-stranded oligonucleotides spanning these protein complex and the free DNA probe are identifi ed by arrows. sites showed a clear gel shift of MDA-MB-231 human breast (C) shRNA-mediated knockdown of CSL in MDA-MB-231 cell lines was carcinoma nuclear lysates (Fig. 2, A and B). The binding of verifi ed by RT-PCR analysis.

JEM VOL. 204, November 26, 2007 2937 Published November 26, 2007

with normal parental MCF-10A cells, and qPCR using human- or the Notch ligand Jagged1 ( 7 ). To examine whether increased specifi c primers was used to measure Slug expression in the Slug expression would be found in breast cancers showing human breast epithelial cells. Jagged1-induced Notch activa- increased Jagged1 or Notch1 expression, we accessed a com- tion, as demonstrated by induction of the target gene HEY1 pendium of 132 independent gene expression datasets repre- (Fig. S2 E), resulted in increased levels of Slug mRNA levels senting >10,000 microarray experiments in the Oncomine with a concomitant decrease in E-cadherin mRNA expres- database (17 ). Expression of both Jagged1 and Notch1 corre- sion in these normal breast epithelial cells (Fig. 3 A). These lated positively with Slug in two independent breast cancer fi ndings show that Jagged1 expression can activate Notch sig- datasets ( Fig. 3, B and C ). Of signifi cance, a positive correlation naling in a juxtacrine manner to induce Slug expression and between the expression of Jagged1 and Slug, but not Snail, mesenchymal transformation of breast epithelial cells. was observed in numerous cancers other than the breast, which Poor prognosis for breast cancer patients has been shown was consistent with our fi nding that Jagged1 is capable of to correlate with elevated levels of expression of either Slug (16 ) inducing Slug but not Snail (Fig. 3 D). Downloaded from jem.rupress.org on July 22, 2010

Figure 3. Jagged1 and Slug expression are correlated in human breast epithelial cells and in primary human breast cancer. (A) qPCR for expression of Slug and E-cadherin in MCF-10A parental cells co-cultured with mouse endothelial cells transduced with MIY vector control ( − Jagged1) or MIYJagged1 (+Jagged1). Human-specifi c primers were used to avoid amplifi cation of mouse transcripts and limit analysis to MCF-10A cells. Data are expressed as the relative gene expression level, with the empty vector control co-culture ( − Jagged1) as the comparator, and are from three independent experiments (mean + SEM). Slug: *, P ≤ 0.05; E-cadherin: *, P < 0.0001. (B – D) Expression correlations in primary human cancers. Pearson correlation coeffi - cients were obtained from microarray datasets deposited in the Oncomine database. For each dataset, individual correlations between two genes of interest (as well as all possible correlations in cases where replicate probes were present in the microarray) are represented by open circles. Bars represent the mean Pearson correlation coeffi cient. The numbers of normal and cancer specimens included in the correlation analysis for each dataset are indicated. (B) Expression correlations between Jagged1 and Slug in primary human breast cancer. Jagged1 expression correlations with Snail are also shown. *, P < 0.01; **, P < 0.001. (C) Expression correlations between Notch1 and Slug in primary human breast cancer. Notch1 expression correlations with Snail are also shown. *, P < 0.01; **, P < 0.001. (D) Expression correlations between Jagged1 and Slug in various primary human cancers. Where available, expression correlations between Jagged1 and Snail are also shown. *, P ≤ 0.05; **, P < 0.01; ***, P < 0.001. n.d., not determined.

2938 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ARTICLE

HEY genes are potential markers of human breast cancers Although these fi ndings suggest that repression of E-cad- that exhibit activation of the Jagged1– Notch – Slug herin is mediated by Slug and not HES/HEY, HEY genes signaling axis may represent surrogate markers of activation of the Jagged1– In response to Notch activation, transcriptional repressors of Notch– Slug signaling axis in human breast cancer. Examination the HES and HEY families of basic helix-loop-helix proteins of two independent breast cancer microarray datasets confi rmed are induced (18 ). Cell type and context determine which positive expression correlations between Jagged1 and HEY1, members are induced (18 ). qPCR was used to analyze whether HEY2, and HEYL, respectively (Fig. 4 C). Moreover, the HES1 and HEY1/2/L were up-regulated in response to Notch expression of HEY1, HEY2, and HEYL were all positively activation in breast epithelial cells. Although HES1 levels did correlated with the expression of Slug but not Snail (Fig. 4 D). not diff er between vector control and Notch1IC cells, all three Thus, the HEY genes may be potential markers of human breast HEY genes were absent in control cells and up-regulated in the cancers that exhibit Notch activation and may, therefore, presence of Notch1IC (Fig. 4 A ). classify a subset of breast cancer patients that would benefi t Because HEY proteins, similar to Slug, silence gene ex- from therapeutics specifi cally designed to target the Jagged1 – pression by binding to E-boxes in target gene promoters Notch– Slug pathway. ( 18), we determined whether enforced expression of any one of the HEY proteins was sufficient to down-regulate Inhibition of ligand-induced Notch activation blocks growth E-cadherin. In contrast to Slug (Fig. S2, B-E), none of the three and metastasis of breast tumors in vivo HEY proteins altered E-cadherin transcript levels in these To determine whether blockade of Notch ligand – receptor Downloaded from cells (Fig. 4 B ). interaction would reverse Slug-induced EMT and inhibit breast jem.rupress.org on July 22, 2010

Figure 4. HEY genes are potential markers of human breast cancers that exhibit activation of the Jagged1– Notch– Slug signaling axis. (A) qPCR for Notch target genes (HES1, HEY1, HEY2, and HEYL) in MCF-10A MIY and MIYNotch1IC cell lines. Data shown are the mean threshold cycle num-

ber (CT ) + SEM from three independent experiments. n.d., not detectable. (B) qPCR for gene expression in MCF-10A cells transduced with MIY, MIYHEY1, MIYHEY2, or MIYHEYL. Data show relative gene expression level or threshold cycle number (CT ). n.d., not detectable. (C and D) Expression correlations in primary human breast cancers. Pearson correlation coeffi cients were obtained from microarray datasets deposited in the Oncomine database. For each dataset, correlations between two genes of interest (as well as all possible correlations in the event of replicate genes in the microarray) are represented by open circles. Bars represent the mean Pearson correlation coeffi cient. The number of normal and cancer specimens included in the correlation analysis for each dataset are indicated. n.d., not determined. (C) Expression correlations between Jagged1 and each of the HEY target genes in primary human breast cancer. *, P ≤ 0.05; **, P < 0.001. (D) Expression correlations between each of the HEY target genes and Slug in primary human breast cancer. Where available, expression correlations between each of the HEY target genes and Snail are also shown. *, P < 0.01; **, P < 0.001. (E) Semiquantitative RT-PCR for Notch target genes in MDA-MB-231 MIG and MIGXNotch4HA tumor xenografts. Data are expressed as the relative gene expression level with MIG control tumor as the comparator (mean + SEM). *, P ≤ 0.05. n.d.

JEM VOL. 204, November 26, 2007 2939 Published November 26, 2007

tumor invasion and metastasis, we used the Slug-positive/ content/full/jem.20071082/DC1). These fi ndings show that E-cadherin– negative MDA-MB-231 human breast carcinoma inhibition of ligand-induced Notch signaling in breast tumor cell line. MDA-MB-231 cells possess a wild-type E-cadherin cells can inhibit tumor growth and metastasis. gene and, thus, exhibit reversible E-cadherin silencing (19 ), and they also express Notch4 (20 ). We confi rmed Notch4 Inhibition of ligand-induced Notch signaling restores expression and also found that MDA-MB-231 cells expressed E-cadherin expression and inactivates ␤ -catenin in breast multiple other Notch receptors and ligands, thus providing tumors in vivo these cells with the potential to activate Notch signaling Having demonstrated an antitumor eff ect of XNotch4 on through juxtacrine/autocrine ligand– receptor interactions breast tumor growth in vivo, we sought to determine whether (Fig. S3 A, available at http://www.jem.org/cgi/content/ inhibition of Notch signaling would reinduce expression of full/jem.20071082/DC1). E-cadherin in the tumor cells. Lysates from XNotch4 tumor To target the Notch pathway in vivo, we chose to use xenografts exhibited E-cadherin protein expression in contrast a soluble Notch receptor, which has been shown to block to tumors lacking XNotch4, which remained E-cadherin neg- ligand-induced Notch signaling (21 ), rather than ␥ -secretase ative (Fig. 5 C). Because the antibody used recognizes both inhibitors that can directly increase the expression of E-cad- human and mouse E-cadherin, RT-PCR was performed with herin by preventing its proteolysis ( 22). MDA-MB-231 cells human E-cadherin– specifi c primers to confi rm E-cadherin re- were retrovirally transduced with the soluble ectodomain of expression in the XNotch4 tumor cells ( Fig. 5 D). Functional human Notch4 (XNotch4) to block ligand-induced Notch reexpression of E-cadherin at the plasma membrane of the Downloaded from activation. Secretion of the soluble protein was confi rmed by xenografted tumor cells in response to Notch inhibition was immunoblotting the concentrated medium in which the cells demonstrated by immunofl uorescent microscopy ( Fig. 5 E ). were cultured (Fig. S3 B). The ability of XNotch4 to in- ␤ -Catenin contributes to breast tumorigenesis by regulat- hibit tumor growth in vivo was tested in a xenograft model ing the expression of genes involved in proliferation, inva- by implanting XNotch4-secreting MDA-MB-231 cells sub- sion, and EMT (27 ). When ␤ -catenin is bound to E-cadherin, cutaneously into immunodefi cient mice. Expression of XNotch4 signaling-competent nuclear ␤ -catenin levels diminish, and protein in the xenografts was confi rmed by immunohisto- cell proliferation and invasion are suppressed (28 ). To deter- jem.rupress.org chemistry (Fig. S3 C). mine whether surface E-cadherin reexpression in XNotch4 We assessed the mRNA levels of HEY genes in this tumors would aff ect ␤ -catenin activity, we immunoblotted model to determine whether any of the HEY genes were tumor cell lysates with an antibody specifi c for active ␤ -catenin down-regulated and to confi rm inhibition of the Notch and found signifi cantly reduced ␤ -catenin activity in tumors pathway. Interestingly, of the three Notch target genes as- where Notch activation was blocked (Fig. 5 C). Consistent on July 22, 2010 sessed, only HEYL exhibited a decrease in mRNA expression with this fi nding, ␤ -catenin relocated from the nucleus to similar to Slug in XNotch4 tumors (Fig. 4 E and see Fig. 6 A). the plasma membrane in tumors where Notch was inhibited The reason for this is not clear, but given that some HEY genes ( Fig. 5, E and F). are also downstream targets of other signaling pathways, To verify that activated Notch is able to activate ␤ -catenin such as TGF- ␤ ( 23), it is possible that in our model HEYL function, we examined expression of the ␤ -catenin target is up-regulated solely by the Notch pathway, whereas other genes in Notch-activated MCF-10A cells by qPCR. Interest- factors may also be up-regulating HEY1/2. Hence, inhibition ingly, Notch activation induced Axin2 and APCDD1 but not of Notch would only block HEYL expression, but would not Lef1 (Fig. S5 A, available at http://www.jem.org/cgi/content/ abrogate expression of HEY1/2. full/jem.20071082/DC1). In parental MCF-10A cells, Wnt3a Tumor growth was signifi cantly inhibited using XNotch4 stimulation also induced Axin2, but not Lef1 (Fig. S5 B), sug- to inhibit ligand-induced Notch signaling ( Fig. 5 A). Inhibi- gesting that the active ␤ -catenin levels correlate with tran- tion of tumor growth was also achieved by enforced expression scriptional activation. However, conditioned medium from of soluble extracellular Notch1 or dominant-negative CSL, Notch-activated cells did not induce Axin2 (Fig. S5 C), sug- confi rming that direct inhibition of Notch signaling within gesting that the observed ␤ -catenin activation is independent the tumor cells attenuates growth in MDA-MB-231 tumors of a secreted factor. Further, inhibition of Wnt activation by (unpublished data). In addition to inhibiting the growth of the Dkk1 did not block activation of the ␤ -catenin target Axin2 primary implant, attenuation of Notch signaling also reduced in Notch-activated cells, whereas Dkk1 did block Wnt3a- the number and size of metastases, as would be expected with induced Axin2 (Fig. S5, C and D), thereby confi rming Wnt- inhibition of EMT (Fig. 5 B ). independent activation of ␤ -catenin by Notch. Because aberrant Notch signaling results in disrupted To determine whether enforced expression of E-cadherin blood vessel development (24 – 26 ), we determined whether was suffi cient to reproduce the phenotype induced by blockade attenuated tumor growth could be explained by an anti- of Notch signaling, we transduced MDA-MB-231 cells with angiogenic eff ect of XNotch4. We did not observe diff erences in E-cadherin complementary DNA (cDNA) and implanted vascular density in the implanted tumors, suggesting that angio- vector- or E-cadherin– expressing cells onto the backs of immuno- genesis inhibition is not the mechanism of tumor suppression defi cient mice. Enforced expression of E-cadherin (independent in this model (Fig. S4, available at http://www.jem.org/cgi/ of Notch inhibition) was suffi cient to inhibit tumor growth and

2940 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ARTICLE Downloaded from jem.rupress.org on July 22, 2010

Figure 5. Inhibition of ligand-induced Notch activation blocks breast tumor growth and metastasis, restores E-cadherin expression, and inactivates ␤ -catenin in vivo. (A) Tumor growth curves for MDA-MB-231 cells transduced with MIG or MIGXNotch4HA grown as xenografts on the dorsa of immunodefi cient mice. Data are presented as the mean ± SEM of the tumor volumes. *, P < 0.01. (B) Quantitation of metastases in MDA-MB-231 MIG and MIGXNotch4HA tumor-bearing mice. Data shown represent the mean number of metastases per mouse + SEM and the mean weight of each metastatic nodule + SEM. *, P ≤ 0.05. (C) Immunoblots for expression of XNotch4HA, E-cadherin, active ␤ -catenin, and ␣ -tubulin in MDA-MB-231 MIG and MIGXNotch4HA tumors. Protein expression was quantitated by densitometry and normalized to ␣ -tubulin. Data shown represent mean + SEM. *, P ≤ 0.05. (D) RT-PCR for expression of human E-cadherin in MDA-MB-231 MIG and MIGXNotch4HA tumors. Human-specifi c primers were used to avoid amplification of mouse E-cadherin. (E) Immunofluorescent staining for E-cadherin (red), ␤ -catenin (red), and DAPI (blue) in MDA-MB-231 MIG and MIGXNotch4HA tumors. Bar, 15 ␮ m. (F) Quantitation of the proportion of cells exhibiting nuclear ␤ -catenin staining in MDA-MB-231 MIG and MIGXNotch4HA tumors. Data shown represent mean + SEM. *, P ≤ 0.05. (G) Tumor growth curves for MDA-MB-231 cells transduced with MIY or MIYE- cadherin grown as xenografts on the dorsa of immunodeficient mice. Data are presented as the mean ± SEM of the tumor volumes. *, P < 0.001. (H) Quantitation of metastases in MDA-MB-231 MIY and MIYE-cadherin tumor-bearing mice. Data shown represent the mean number of metastases per mouse + SEM and the mean weight of each metastatic nodule + SEM. *, P < 0.001. (I) Immunoblots for expression of E-cadherin and ␣ -tubulin in MDA- MB-231 MIY and MIYE-cadherin tumors. (J) Immunofl uorescent staining for E-cadherin (red),␤ -catenin (red), and DAPI (blue) in MDA-MB-231 MIY and MIYE-cadherin tumors. Bar, 15 ␮ m.

JEM VOL. 204, November 26, 2007 2941 Published November 26, 2007 Downloaded from jem.rupress.org on July 22, 2010

Figure 6. Restoration of E-cadherin expression by Notch inhibition is associated with Slug down-regulation and attenuation of E-cadherin promoter methylation. (A) qPCR for expression of Slug, Snail, and Twist1 in MDA-MB-231 MIG and MIGXNotch4HA tumors. Data are expressed as the relative gene expression level with MIG control tumors as the comparator (mean + SEM). *, P ≤ 0.01. (B) Immunofl uorescent staining for Slug (red), GFP (green), and DAPI (blue) in MDA-MB-231 MIG and MIGXNotch4HA tumor xenografts. Yellow represents the overlap of GFP and Slug immunostaining.

2942 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ARTICLE

metastasis with concomitant relocalization of ␤ -catenin to the Treatment of NotchIC- or Slug-expressing breast epithelial plasma membrane (Fig. 5, G– J). These data suggest that in- cells with the HDAC inhibitors trichostatin A (TSA) and hibition of ligand-induced Notch activation is tumor sup- sodium butyrate (NaBu), either alone or together, was suf- pressive, in part because of reinduction of surface E-cadherin fi cient to reverse the down-regulation of E-cadherin ex- expression. This in turn retains ␤ -catenin at the plasma mem- pression by Notch1IC or Slug (Fig. 6, F and G). A current brane, thereby attenuating ␤ -catenin nuclear activity and in- molecular model for transcriptional gene silencing suggests hibiting tumor growth and metastasis. that histone deacetylation is a primary event involved in the initiation of chromatin compaction, and that DNA methyla- Restoration of E-cadherin expression by Notch inhibition tion functions subsequent to histone deacetylation to estab- is caused by Slug down-regulation and attenuation lish a permanent state of gene inactivation (30 ). NotchIC- or of E-cadherin promoter methylation Slug-expressing cells were thus treated with the DNA methyl- To determine whether inhibition of ligand-induced Notch transferase inhibitor 5-azacytidine (5AZA), which also elic- signaling restored E-cadherin expression secondary to repres- ited reinduction of E-cadherin expression. However, 5AZA sion of Slug, expression of the genes encoding Slug, Snail, and TSA/NaBu together did not produce an additive eff ect and Twist1 was assessed by qPCR in the breast tumor xeno- on E-cadherin reexpression, suggesting that DNA methylation grafts. Although transcript levels of Snail and Twist1 did not may occur secondary to histone deacetylation, as previously diff er between control and XNotch4 tumors, Slug expression suggested (30 ). was signifi cantly reduced when Notch signaling was inhibited The E-cadherin promoter contains numerous cytosine- Downloaded from ( Fig. 6 A). XNotch4 also blocked expression of Slug protein, phosphate-guanine (CpG) sites, which result in E-cadherin because Slug was only present in areas of the tumor where silencing when methylated on the corresponding cytosine XNotch4 was absent, as determined by immunofl uorescent residue (31 ). Given that DNA methylation has been reported staining of breast tumor xenografts (Fig. 6 B). to be dominant over histone deacetylation in mediating gene To directly demonstrate that Jagged1 could induce Slug silencing (32 ), as well as our data suggesting that methyla- in breast tumor cells, MDA-MB-231 parental cells were co- tion may occur as a later step in silencing the E-cadherin cultured with mouse stromal cells transduced with either promoter, we sought to determine whether reduced Slug jem.rupress.org Jagged1 or empty vector, and human Slug transcript levels expression and reexpression of E-cadherin in our breast tumor were quantitated by RT-PCR using human-specifi c Slug xenografts correlated with reduced DNA methylation at the primers. In response to Jagged1-induced Notch signaling, E-cadherin promoter. Using two independent methods, Slug transcripts in MDA-MB-231 cells were increased approxi- methylation-specifi c PCR (MSP) and genomic bisulfi te se- mately fi vefold ( Fig. 6 C ). To prove that Slug was responsible quencing, E-cadherin promoter methylation in XNotch4 on July 22, 2010 for E-cadherin repression in these cells, Slug was targeted using tumors was found to be reduced compared with control lentiviral-delivered shRNA in breast tumor cells. Knockdown tumors ( Fig. 6, H and I ). Because XNotch4 tumors did not of Slug was confi rmed by qPCR and immunoblotting, and display generalized hypomethylation of the genome (Fig. S6, was found to be suffi cient to restore expression of E-cadherin A and B, available at http://www.jem.org/cgi/content/full/ ( Fig. 6, D and E). jem.20071082/DC1), Notch inhibition likely results in de- Because Slug recruits histone deacetylase (HDAC) com- methylation at specifi c promoters regulated by Notch/Slug. plexes to mediate transcriptional silencing (29 ), we deter- These results suggest that inhibition of Notch signaling re- mined whether HDAC activity was required for the ability of induces E-cadherin by repressing Slug and reversing E-cadherin Notch-induced Slug to down-regulate E-cadherin expression. promoter methylation.

Bar, 100 ␮ m. (C) qPCR for gene expression in MDA-MB-231 parental cells co-cultured with mouse endothelial cells transduced with MIY vector control ( − Jagged1) or MIYJagged1 (+Jagged1). Human-specifi c primers were used to assay Slug specifi cally in MDA-MB-231 cells and avoid amplifi cation of mouse transcripts. Data are expressed as the relative gene expression level, with the empty vector control co-culture ( − Jagged1) as the comparator, and are from three independent experiments (mean + SEM). *, P ≤ 0.05. (D) qPCR for expression of Slug and E-cadherin in MDA-MB-231 cells transduced with shRan- dom or shSlug. Data from two independent experiments are shown and are expressed as the relative gene expression level with shRandom control as the comparator. (E) Immunoblot for expression of E-cadherin, Slug, and ␣ -tubulin in MDA-MB-231 cells transduced with shRandom or shSlug. (F) Quantita- tion of the proportion of MCF-10A cells transduced with MIY or MIYNotch1IC that are positive for both YFP and E-cadherin. Cells were treated with TSA (100/500/1000 nM), NaBu (2/5/10 mM), or 5AZA (1/5/10 ␮ M) alone or in combination, and E-cadherin expression was assessed 3 d after treatment by immunofl uorescent microscopy. Data shown represent mean + SEM of at least three independent experiments. *, P ≤ 0.05. (G) Quantitation of the proportion of MCF-10A cells transduced with MIY or MIYSlug that are positive for both YFP and E-cadherin. Cells were treated with TSA (100/500/1000 nM), NaBu (2/5/10 mM), or 5AZA (1/5/10 ␮ M) alone or in combination. Data shown represent mean + SEM of at least three independent experiments. *, P ≤ 0.05. (H) MSP to assess methylation status of the E-cadherin promoter in MDA-MB-231 tumors. Primers specifi c for methylated (M) or unmethylated (U) E-cadherin promoter were used. Amplifi ed M and U products were quantitated by densitometry and expressed as the M/U ratio. Data shown represent mean + SEM. *, P < 0.001. (I) Genomic bisulfi te sequencing to assess methylation status of the E-cadherin promoter in MDA-MB-231 tumors (MIG,n = 35 clones from fi ve tumors; MIGXNotch4HA,n = 22 clones from three tumors). A total of 22 CpG sites within the E-cadherin proximal promoter ( − 104 to +118) were analyzed. Data shown represent the percentage of methylation observed at each CpG site.

JEM VOL. 204, November 26, 2007 2943 Published November 26, 2007

The Notch– Slug signaling axis inhibits anoikis of human breast cells During cancer progression and dissemination, malignant cancer cells enter the circulation or lymphatic system and then must survive in the vasculature until they successfully extrav- asate into a tissue site. Cell death associated with abolished matrix-initiated integrin signaling has been termed anoikis. For a tumor cell to metastasize to a distant site, it needs to overcome anoikis. Activated Notch inhibits apoptosis in response to various triggers in diff erent cell types, but it has also been reported to induce apoptosis in other situations ( 11, 33, 34). We thus asked whether breast epithelial cells expressing Notch1IC or Slug would exhibit a survival advantage compared with control cells when maintained under conditions that prevented adhesion. For this assay, cells were placed in suspension cultures, rather than soft agar (11 ), to mimic transit through the bloodstream. Both NotchIC- and Slug-expressing breast epithelial cells exhibited protection against anoikis Downloaded from compared with control cells ( Fig. 7 A). Conversely, breast tumor xenografts in which Notch signaling was inhibited showed reduced Slug expression (Fig. 6, A and B ), with a concomitant increase in cell death as measured by activated Caspase 3 levels (Fig. 7 B ). Figure 7. The Notch– Slug signaling axis inhibits anoikis of human breast cells. (A) Anoikis assay to assess cell death in MCF-10A

cell lines (MIY, MIYNotch1IC, and MIYSlug). The fraction of hypodiploid jem.rupress.org DISCUSSION cells was determined by fl ow cytometry, and the data representing the Although mouse models have indicated that mammary-spe- mean + SEM of three independent experiments are expressed as the fold cifi c overexpression of a truncated constitutively active Notch change in hypodiploid cells relative to MIY control. MIYNotch1IC: *, P < can result in breast tumors, human data suggest that both 0.01; MIYSlug: *, P ≤ 0.05. (B) Immunohistochemical staining for activated Notch ligands and receptors are up-regulated in a proportion caspase 3 in implanted MDA-MB-231 MIG and MIGXNotch4HA tumors. of breast cancers and that this expression is correlated with Data representing the mean + SEM are shown as the proportion of the on July 22, 2010 poor outcome. The implication is that juxtacrine or auto- activated caspase 3 – stained area compared with the total tumor area. ␮ crine Notch activation, rather than activating Notch muta- *, P = 0.012. Bar, 25 m. tions, may be responsible for an aggressive tumor phenotype. This is in marked contrast to T cell acute lymphoblastic leuke- ings may be generalized to other tumors in which Notch is mia, where more than half of the cases have activating muta- activated by ligand. tions of Notch1 ( 35). In this paper, we have shown that Jagged1 Expression of the Notch target gene HEYL has previously activation of Notch up-regulates the transcriptional repressor been reported to be absent in the normal breast epithelium Slug to promote carcinogenesis by two cellular mechanisms: but present in the tumor cell compartment of invasive breast (a) by facilitating cancer cell metastasis through initiation of cancers (37 ). These fi ndings fi t with our results, which show EMT, and (b) by enhancing cell survival in the absence of cell that HEYL mRNA is undetectable in control breast epithe- matrix adhesion. Importantly, we have demonstrated that, in lial cells but is up-regulated in cells exhibiting activation human breast cancers, expression of Jagged1 and Notch1 cor- of the Notch– Slug signaling axis. Importantly, our analysis relates positively with Slug expression. of primary human breast cancers has identifi ed HEY genes, In breast cancer patients, increased expression of either in particular HEYL , as potential markers of human breast Jagged1 or Notch1 is predictive of poor overall survival (7 ). cancers that exhibit activation of the Jagged1– Notch – Slug If both Jagged1 and Notch1 are increased, there is a further signaling axis. substantial reduction in overall survival (7 ). Our data provide The Snail gene has previously been reported to be a di- an explanation for the reported dose-dependent relationship of rect target gene of Notch in endothelial cells (13 ). However, Jagged1 expression and negative outcome in breast cancer (7 ), our data suggest that this is not the case in epithelial cells, be- suggesting that juxtacrine or autocrine Jagged1 – Notch inter- cause we did not observe a positive correlation between acti- actions induce Slug, which initiates EMT and inhibits anoikis, vated Notch signaling and Snail induction. Rather, we observed thereby promoting tumor invasion and metastasis. Interest- a positive correlation between Notch activation and Slug ex- ingly, Jagged1 expression has also been associated with prostate pression both in human breast epithelial cells in vitro and in cancer metastasis and recurrence (36 ), and our demonstration primary human breast cancers. Hence, our fi ndings indicate of positive correlations between Jagged1 and Slug expression that Slug, but not Snail, is a downstream target gene of Notch in a wide variety of tumors raises the possibility that our fi nd- in epithelial and other tumors. Interestingly, increased levels of

2944 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ARTICLE

Slug, but not Snail, have been associated with tumors from play a role in maintaining the tumor stem cell, as well as in patients with metastatic disease or disease recurrence (16 ). permitting these cells to survive the metastatic process. Although it has been reported that Slug has antiapoptotic Tumor cell metastasis is the primary cause of mortality in activity in some cell types, the current study is the fi rst dem- a vast majority of cancer patients. As such, there is a compel- onstration that Slug can inhibit anoikis. To initiate EMT, re- ling need to elucidate the mechanisms of tumor cell metastasis pression of E-cadherin permits loss of cell– cell cohesion, thereby and to develop rationally designed therapeutics to specifi cally disrupting apical-basal polarity and promoting cell migration. target the metastatic process. Our fi ndings highlight the po- However, given that the loss of E-cadherin renders cells sus- tential use of inhibitors of ligand-induced Notch signaling as ceptible to anoikis, Slug must independently promote cell viable and eff ective agents to block EMT and tumor metasta- survival in this context to allow local tumor invasion and distant sis in the treatment of human cancers that exhibit activation metastases. Thus, our fi ndings that Notch blockade inhibits of the Jagged1– Notch– Slug signaling axis. By specifi cally tar- growth of the primary tumor by triggering apoptosis and re- geting ligand-induced activation, there is the potential of reduces distant metastases is concordant with this model of Slug- ducing the more widespread side eff ects potentially evoked by dependent EMT and cell survival. less-specifi c therapeutics, such as the ␥ -secretase inhibitors. Both Slug and Notch have been shown to inhibit p53 function, in part through the inhibition of Puma expression, which may be one mechanism of the antianoikis eff ect ( 11, 38 ). MATERIALS AND METHODS Notch has also been shown to activate the phosphatidylinositol Cell lines. The human breast epithelial cell line MCF-10A was cultured in Downloaded from Ј a 1:1 mixture of DMEM/F12 (Sigma-Aldrich) supplemented with 5% horse 3 -kinase– Akt antiapoptotic pathway, but whether there is serum (Sigma-Aldrich), 2 mM glutamine (Sigma-Aldrich), 20 ng/ml of epi- cross talk with Slug remains to be seen (39, 40 ). The ability of dermal growth factor (Sigma-Aldrich), 100 ng/ml cholera toxin (Cedarlane), Notch to suppress c-Jun N-terminal kinase activation and 10 ␮ g/ml insulin (Sigma-Aldrich), 500 ng/ml hydrocortisone (Sigma- activate phosphatidylinositol 3Ј -kinase– Akt may explain the Aldrich), and 100 U/ml each of penicillin and streptomycin (Invitrogen). inhibition of p53 by Notch (11, 33, 39). Both Notch and Slug The human breast carcinoma cell lines MDA-MB-231 and T47D and the have also been shown to induce Bcl-2 (33, 41); thus, there are mouse endothelial cell line SVEC4-10 were cultured in DMEM supplemented with 10% heat-inactivated calf serum (HyClone), 2 mM glutamine, jem.rupress.org multiple potential anoikis– apoptosis pathways that Notch may and 100 U/ml each of penicillin and streptomycin. All cells were maintained

regulate through Slug. at 37 ° C in an atmosphere of 5% CO2 . It is also possible that Slug-mediated repression of E-cadherin, which we have shown results in activation of ␤ -catenin, Isolation of primary human breast epithelial cells. Normal human may be responsible for an antianoikis eff ect. Indeed, even breast tissue was obtained in accordance with guidelines approved by the University of British Columbia from anonymized discarded material from modest overexpression of active ␤ -catenin has been shown to ␤ normal premenopausal women undergoing reduction mammoplasty sur- on July 22, 2010 prevent anoikis (42 ). Of interest, -catenin has been reported geries. A crude epithelial cell– enriched cell suspension was obtained enzy- to induce Slug promoter activity (43 ). Hence, Notch induc- matically and cryopreserved. As required, cells were thawed, and single-cell tion of Slug is potentially enhanced by a feed-forward loop. suspensions were prepared as previously described (45 ). Cells were co- Initially, Notch/CSL would directly induce Slug. Subsequent cultured with 1.2 × 10 6 X-irradiated NIH3T3 mouse fi broblasts for 1 d in Slug-mediated E-cadherin repression, via interaction with Epicult-B medium (StemCell Technologies Inc.) supplemented with 5% FCS. Cells were harvested, and epithelial cell adhesion molecule (EpCAM)– positive E2-boxes in the E-cadherin promoter (12 ), would then release ␤ breast epithelial cells were magnetically separated using the human EpCAM -catenin from the plasma membrane, which would accu- selection cocktail EasySep (StemCell Technologies Inc.). These selected cells mulate in the nucleus and further activate the Slug promoter. were cultured in Epicult-B medium at a density of 2 × 105 cells per dish in However, the human Slug promoter also contains several cis 35-mm dishes precoated with Vitrogen (Cohesion Technologies). To precoat elements predicted to bind the HEY / HES family of Notch- the dishes, 2 – 3 ml of Vitrogen (67 ␮ g/ml in PBS) was used for 1 h at 37 ° C induced transcriptional repressors (unpublished data). Thus, and washed with PBS before use. these factors may act to attenuate the positive feedback loop Plasmid constructs and gene transfer. Retroviral vectors (MIY and described. Of note, our data showed that enforced expression MIG) containing an internal ribosomal entry site and either YFP (MIY) or of E-cadherin mimicked the eff ects of blocking Notch signal- GFP (MIG) were used to facilitate the sorting of transduced cells. cDNA ing in vivo, suggesting that a major function of Notch-induced constructs encoding the human Notch1IC, C-terminal hemagglutinin Slug in this model may be to repress E-cadherin. (HA)-tagged human Notch4IC, full-length C-terminal Flag-tagged human Recent evidence has demonstrated that breast cancers may Slug (a gift of E. Fearon, University of Michigan, Ann Arbor, MI), full- length human Jagged1, full-length human E-cadherin (a gift of B.M. be initiated in and propagated by a minority cell population Gumbiner, University of Virginia, Charlottesville, VA), and N-terminal myc- that have been designated tumor stem cells (44 ). Notch signal- tagged full-length human HEY1/2/L (a gift of D. Srivastava, University of ing appears to play an important role in normal mammary stem California, San Francisco, San Francisco, CA) were subcloned into MIY. The cell self-renewal (44 ). Thus, the Notch pathway may be selec- cDNA construct encoding the entire extracellular domain of human Notch4 tively activated in more primitive cancer stem cells. A mouse (XNotch4; amino acids 1– 1,443) tagged with a C-terminal HA epitope transgenic model of activated Notch4-driven breast cancer in- was subcloned into MIG. The pLentilox3.7-shRandom and pLentilox3.7-shSlug constructs were generated by inserting shRNAs targeting the sequences dicates that these tumors are highly metastatic (3 ). Interestingly, 5 Ј -GTTGCTTGCCACGTCCTAGAT-3Ј (Random) and 5 Ј -GCATTT- Slug has been shown to have antiapoptotic activity in hema- GCAGACAGGTCAAAT-3Ј (Slug) into pLentilox3.7 (a gift of L. Van Parijs, topoietic progenitor cells (38 ); thus, the Notch– Slug axis may Massachusetts Institute of Technology, Cambridge, MA). Cells were transduced

JEM VOL. 204, November 26, 2007 2945 Published November 26, 2007

with empty vector control or vector containing cDNA inserts, and trans- annealing temperatures are described in Tables S1 and S2. Entire MSP reactions duced cells were sorted based on YFP or GFP expression using a cell sorter were assessed in 2% TAE-agarose gels containing ethidium bromide. Bands (FACS 440; Becton Dickinson). corresponding to methylated and unmethylated PCR products were quantitated by densitometry. Data are expressed as a ratio of methylated over un- Immunoblotting. Cultured cells or tumor tissue were lysed and analyzed methylated PCR products (M/U ratio) and represent the mean ratio + SE from by SDS-PAGE and immunoblotting with rabbit polyclonal antibody to Slug fi ve control tumors and fi ve XNotch4 tumors. For genomic bisulfi te sequenc- (Santa Cruz Biotechnology, Inc.), or mouse monoclonal antibodies to HA ing, bisulfi te-modifi ed genomic DNA isolated from tumor tissue was amplifi ed (Sigma-Aldrich), E-cadherin (BD Biosciences), ␤ -catenin (BD Biosciences), by PCR using primers spanning the E-cadherin proximal promoter. Primer active ␤ -catenin (Millipore), and ␣ -tubulin (Sigma-Aldrich). Protein expression sequences and annealing temperatures are described in Tables S1 and S2. PCR was quantitated by densitometry. products were purifi ed and cloned into the pDrive cloning vector (QIAGEN), and individual clones were sequenced. Five control tumors (a total of 35 MIG RNA isolation, RT-PCR, and qPCR. Total RNA isolation was per- clones) and three XNotch4 tumors (a total of 22 XNotch4 clones) were formed using TRIzol reagent (Invitrogen) or an RNeasy kit (QIAGEN), analyzed, and data were expressed as the percentage of methylation per CpG according to the manufacturers’ recommendations. First-strand cDNA was site (total number of methylated clones/total number of clones). synthesized using Superscript II reverse transcriptase (Invitrogen). After ribo- nuclease H treatment (Invitrogen), PCR was performed. Control reactions Immunostaining. Breast epithelial cells were stained with mouse mono- omitting reverse transcriptase were performed in each experiment. Primer clonal antibody to extracellular E-cadherin (Chemicon) or rabbit polyclonal sequences and annealing temperatures are described in Tables S1 and S2 (avail- antibody to Slug (Santa Cruz Biotechnology, Inc.). 7-␮ m-thick tumor cryo- able at http://www.jem.org/cgi/content/full/jem.20071082/DC1). For RT- sections were stained with rabbit polyclonal antibody to Slug (Santa Cruz Biotechnology, Inc.), mouse monoclonal antibodies to E-cadherin (BD Bio-

PCR, entire PCR samples were assessed in 1.5% TAE-agarose gels containing Downloaded from ethidium bromide. Control human cDNA was generated from pooled total sciences) or ␤ -catenin (BD Biosciences), and rabbit monoclonal antibody to RNA isolated from the following human cells: human mammary epithelial activated caspase-3 (BD Biosciences). For immunohistochemistry, a biotin- cells; vascular smooth muscle cells; cervical cancer cells, SiHa; colon cancer ylated goat anti– rabbit antibody (Vector Laboratories) followed by horseradish cells, WiDr; and kidney epithelial cells, 293T. Control mouse cDNA was peroxidase– conjugated streptavidin (Vector Laboratories) were used, and generated from pooled total RNA isolated from the following mouse cells: nuclei were counterstained with hematoxylin. For immunofl uorescence, the Lewis lung carcinoma cells; endothelial cells, SVEC4-10; and fi broblasts, fl uorochrome-conjugated secondary antibodies goat anti– rat Alexa Fluor NIH3T3. For qPCR, reactions were run on a real-time PCR system (ABI 594 (Invitrogen) and goat anti– mouse Alexa Fluor 594 (Invitrogen) were Prism 7900; Applied Biosystems). Gene expression was detected with SYBR used, and nuclei were counterstained with DAPI (Sigma-Aldrich). Immuno- jem.rupress.org green (Applied Biosystems), and relative gene expression was determined by fl uorescence was detected with an imaging microscope (Axioplan II; Carl ⌬ normalizing to GAPDH using the CT method. Zeiss, Inc.), and images were captured with a digital camera (1350EX; QImaging). To quantify the proportion of cells exhibiting nuclear ␤ -catenin EMSA. Nuclear lysates were collected from shRandom, shCSL1, and staining, at least six random fi elds at 200 × magnifi cation were analyzed per shCSL2 overexpressing MDA-MB-231 cells for the CSL EMSA assays. tumor using Northern Eclipse software. Data are presented as the mean per- The binding reaction (10 mM TrisHCl, 50 mM NaCl, 1 mM EDTA centage of nuclear ␤ -catenin staining (total number of nuclear ␤ -catenin – ␮ ␮ positive cells/total number of cells) + SEM from four control tumors and

[pH 8], 4% glycerol, 2 g PolydI-dC binding buff er, and 10 g of nuclear on July 22, 2010 protein) was performed by preincubating with either 50-fold excess wild-type four XNotch4 tumors. To quantify the proportional area of tumors showing (Slug 850, [forward] GGGCCCTTTTTCCCATAAAAAAAAAG and [reverse] activated caspase 3, entire tumor sections were analyzed using Northern GGGAAAAAGGGTATTTTTTTTTCGGG; Slug 1680, [forward] TGTGT- Eclipse software. Data are presented as the mean percentage of the activated GTTTTGTGGGAAATGGAG and [reverse] CTCCATTTCCCACAAAA) or caspase 3– stained area (total area positive for activated caspase 3/total area of mutant (Slug 800, [forward] GGGCCCTTTGCAGCATAAAAAAAAAAG tumor section) + SEM from 14 control tumors and 12 XNotch4 tumors. and [reverse] GGGAAACGTCGTATTTTTTTTTTCGGG; Slug 1600, [forward] TGTGTGTTTTGTGCTGCATGGAG and [reverse] CTC- siRNA transfection. MCF-10A MIY and MIYNotch1IC cells were fl ow CATGCAGCACAAAA) nonradioactive duplex oligos for 15 min on ice, sorted, and YFP-positive cells were transiently transfected with 100 nM and then adding a 150,000-cpm 32 P-labeled double-stranded probe and incu- siRNA targeting the sequences 5Ј -GTTGCTTGCCACGTCCTAGAT-3Ј bating for 30 min at room temperature. Binding reactions were run on 5% (siRandom) or 5Ј -GCATTTGCAGACAGGTCAAAT-3Ј (siSlug) using Tris-Borate EDTA gels and exposed to a phosphorimager plate for 12– 16 h. DharmaFECT 1 transfection reagent (Dharmacon Inc.). Cells were harvested 4 d after transfection. Inhibition of histone acetylation and DNA methylation. MCF-10A cell lines were fl ow sorted, and YFP-positive cells were plated into fourwell Co-culture. 10 6 parental human cells (MCF-10A or MDA-MB-231) were chamber slides at 7 × 105 cells per well and allowed to adhere and grow for 48 h. co-cultured with 106 mouse SVEC4-10 endothelial cells transduced with Cells were then treated with TSA (100/500/1000 nM), NaBu (2/5/10 mM), MIY control vector or MIYJagged1 and plated in 100-mm tissue culture and/or 5AZA (1/5/10 ␮ M) for 72 h. After immunofl uorescent staining for dishes. Cells were harvested after 3 d of co-culture. qPCR was performed E-cadherin, at least six random fi elds at 200 × magnifi cation were analyzed with human-specifi c primers to avoid amplifi cation of mouse transcripts. per well using Northern Eclipse software (Empix Imaging). Data are presented as the percentage of YFP/E-cadherin double-positive cells + SEM. Anoikis assay. MCF-10A cell lines were fl ow sorted, and YFP-positive cells were plated into 60-mm tissue culture plates (8 × 105 cells per plate) Methylation assays. Genomic DNA was isolated from cultured cells or coated with 1% agarose. After 24 h of incubation at 37 ° C, cells were fi xed/ tumor tissue using a DNeasy tissue kit (QIAGEN) according to the manu- permeabilized in 70% ethanol and stained with propidium iodide (Sigma- facturer ’ s recommendations. 1 ␮ g of genomic DNA was bisulfi te modifi ed Aldrich), and the proportion of cells with hypodiploid DNA content was using a CpGenome DNA modifi cation kit (Chemicon) and eluted in 25 ␮ l determined by fl ow cytometry. Data are expressed as the proportion of hy- Tris-EDTA buff er, according to the manufacturer ’ s recommendations. podiploid cells relative to MIY control. ng of bisulfi te-modifi ed DNA, 400 nM of 120 ف MSP was performed using 5Ј and 3Ј primers, 0.2 mM of 2’ -deoxynucleoside 5Ј -triphosphates (Invitrogen), Tumorigenicity assays. Female nonobese diabetic/severe combined 1 × PCR buff er, and 0.625 U of HotStarTaq DNA polymerase (QIAGEN), immunodefi cient mice were obtained from the Animal Resource Centre of according to the manufacturer’ s recommendations. Primer sequences and the British Columbia Cancer Research Centre. For MDA-MB-231 tumor

2946 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ARTICLE

implantation, 5 × 10 6 cells were injected subcutaneously into the dorsa of mice. REFERENCES Once tumors were palpable, tumor volume (0.523 × length × width × 1 . Mumm , J.S. , and R. Kopan . 2000. Notch signaling: from the outside in. height) was measured weekly using calipers. For each time point, MDA- Dev. Biol. 228 : 151 – 165 . MB-231 tumor data are presented as the mean tumor volume ± SEM from 2 . Leong , K.G. , and A. Karsan . 2006 . Recent insights into the role of (a) 12 MIG tumors and 11 MIGXNotch4HA tumors, and (b) 12 MIY tu- Notch signaling in tumorigenesis. Blood . 107 : 2223 – 2233 . mors and 12 MIYE-cadherin tumors. Tumor-growth curves for one experi- 3. Gallahan , D. , C. Jhappan, G. Robinson, L. Hennighausen, R. Sharp, E. ment are presented and are representative of three independent experiments, Kordon, R. Callahan, G. Merlino, and G.H. Smith. 1996 . Expression of a with 8– 16 mice per tumor group for each experiment. Mice bearing MDA- truncated Int3 gene in developing secretory mammary epithelium specifi - MB-231 tumors (MIG and MIGXNotch4HA, or MIY and MIYE-cadherin) cally retards lobular diff erentiation resulting in tumorigenesis. Cancer Res. were killed at the same time, and the total number and total weight (in mil- 56 : 1775 – 1785. 4 . Hu , C. , A. Dievart , M. Lupien , E. Calvo , G. Tremblay , and P. Jolicoeur . ligrams) of metastases were determined. Data are presented as the mean 2006 . Overexpression of activated murine Notch1 and Notch3 in trans- number of metastases per mouse + SEM and the mean weight of each meta- genic mice blocks mammary gland development and induces mammary static nodule (in milligrams) + SEM. Metastasis data were determined by an- tumors. Am. J. Pathol. 168 : 973 – 990 . alyzing (a) 25 MIG tumors and 27 MIGXNotch4HA tumors, and (b) 12 MIY 5. Callahan , R. , and S.E. Egan . 2004 . Notch signaling in mammary devel- tumors and 12 MIYE-cadherin tumors. Animal experiments were approved opment and oncogenesis. J. Mammary Gland Biol. Neoplasia . 9 :145 – 163. by the University of British Columbia Institutional Animal Care and Ethics 6. Parr , C. , G. Watkins , and W.G. Jiang . 2004 . The possible correlation of Committee, and all animals were handled according to institutional animal Notch-1 and Notch-2 with clinical outcome and tumour clinicopatho- care procedures. logical parameters in human breast cancer. Int. J. Mol. Med. 14 : 779 – 786. 7 . Reedijk, M. , S. Odorcic , L. Chang , H. Zhang , N. Miller , D.R. Microarray data analysis. Pearson correlation coeffi cients were obtained McCready , G. Lockwood , and S.E. Egan . 2005 . High-level coexpression Downloaded from from publicly available microarray datasets deposited in the Oncomine data- of JAG1 and NOTCH1 is observed in human breast cancer and is as- base (available at www.oncomine.org): breast 1 ( 46 ), breast 2 ( 47 ), adrenal ( 48), sociated with poor overall survival. Cancer Res. 65 : 8530 – 8537 . brain (49 ), endocrine (50 ), gastric (51 ), lung (52 ), ovarian (53 ), renal (54 ), 8 . Ayyanan , A. , G. Civenni , L. Ciarloni , C. Morel , N. Mueller , K. Lefort , salivary gland (55 ), sarcoma 1 ( 56 ), and sarcoma 2 ( 57 ). For each dataset, A. Mandinova , W. Raff oul , M. Fiche , G.P. Dotto , and C. Brisken . individual correlations between two genes of interest (as well as all possible 2006 . Increased Wnt signaling triggers oncogenic conversion of hu- correlations in the event of replicate genes in the microarray) were deter- man breast epithelial cells by a Notch-dependent mechanism. Proc. Natl. mined, as well as the mean Pearson correlation coeffi cient. Microarray data Acad. Sci. USA . 103 : 3799 – 3804 . 9 . Weijzen , S. , P. Rizzo , M. Braid , R. Vaishnav , S.M. Jonkheer , A. accession numbers are shown in Table S3 (available at http://www.jem Zlobin , B.A. Osborne , S. Gottipati , J.C. Aster , W.C. Hahn , et al. 2002 . jem.rupress.org .org/cgi/content/full/jem.20071082/DC1). Activation of Notch-1 signaling maintains the neoplastic phenotype in human Ras-transformed cells. Nat. Med. 8 : 979 – 986 . Statistical analysis. To determine statistical signifi cance, a one-way analysis 10 . Pece , S. , M. Serresi , E. Santolini , M. Capra , E. Hulleman , V. Galimberti , of variance with a Tukey test for multiple comparisons was performed using S. Zurrida , P. Maisonneuve , G. Viale , and P.P. Di Fiore. 2004. Loss of the statistics program Statistical Package for Social Scientists (version 11.0; negative regulation by Numb over Notch is relevant to human breast SPSS Inc.). Statistical signifi cance was taken at P ≤ 0.05. carcinogenesis. J. Cell Biol. 167 : 215 – 221 .

11 . Stylianou , S. , R.B. Clarke , and K. Brennan . 2006 . Aberrant activation on July 22, 2010 Online supplemental material. Tables S1 and S2 provide all primer se- of notch signaling in human breast cancer. Cancer Res. 66 : 1517 – 1525 . quences used in this study. Table S3 provides microarray accession number 12 . Vincent-Salomon , A. , and J.P. Thiery . 2003 . Host microenvironment information. Fig. S1 shows Notch-induced E-cadherin repression in MCF- in breast cancer development: epithelial-mesenchymal transition in breast 10A cells. Fig. S2 shows Notch-induced Slug expression in MCF-10A cells. cancer development. Breast Cancer Res. 5 : 101 –106 . Fig. S3 demonstrates the expression of Notch ligands and receptors in MDA- 13 . Timmerman , L.A. , J. Grego-Bessa , A. Raya , E. Bertran , J.M. MB-231 cells, and secreted XNotch4HA protein. Fig. S4 shows a lack Perez-Pomares, J. Diez , S. Aranda , S. Palomo , F. McCormick , J.C. of a vascular eff ect by XNotch4HA. Fig. S5 demonstrates Notch-induced Izpisua-Belmonte, and J.L. de la Pompa. 2004 . Notch promotes epithelial-mesenchymal transition during cardiac development and oncogenic ␤ -catenin activation independent of Wnt activation. Fig. S6 reveals that transformation. Genes Dev. 18 : 99 – 115 . inhibition of Notch does not cause a generalized hypomethylation of the 14 . Noseda , M. , G. McLean , K. Niessen , L. Chang , I. Pollet , R. Montpetit , genome. Online supplemental material is available at http://www.jem.org/ R. Shahidi , K. Dorovini-Zis , L. Li , B. Beckstead , et al . 2004 . Notch cgi/content/full/jem.20071082/DC1. activation results in phenotypic and functional changes consistent with endothelial-to-mesenchymal transformation. Circ. Res. 94 : 910 – 917 . We thank F. Wong and D. McDougal for assistance with fl ow cytometry and cell 15. Come , C., V. Arnoux , F. Bibeau , and P. Savagner . 2004. Roles of the tran- sorting. We also thank P.L. Olive for critical review of the manuscript. scription factors snail and slug during mammary morphogenesis and breast This research was supported by grants to A. Karsan from the National Cancer carcinoma progression. J. Mammary Gland Biol. Neoplasia. 9 :183 – 193. Institute of Canada with funds from the Canadian Cancer Society and the Cancer 16 . Martin , T.A. , A. Goyal , G. Watkins , and W.G. Jiang . 2005. Expression Research Society, and to C. Eaves from Genome Canada. K.G. Leong was supported of the transcription factors snail, slug, and twist and their clinical signifi - by a Doctoral Research Award from the Canadian Institutes of Health Research and cance in human breast cancer. Ann. Surg. Oncol. 12 : 488 – 496 . a Predoctoral Fellowship Award from the Department of the Army (DAMD17-01-1- 17 . Rhodes , D.R. , J. Yu , K. Shanker , N. Deshpande , R. Varambally, 0164). The U.S. Army Medical Research Acquisition Activity was the awarding and D. Ghosh , T. Barrette , A. Pandey , and A.M. Chinnaiyan . 2004 . administering acquisition offi ce. K. Niessen is the recipient of a Senior Graduate ONCOMINE: a cancer microarray database and integrated data-mining Studentship from the Michael Smith Foundation for Health Research, and I. Kulic platform. Neoplasia . 6 : 1 – 6 . is supported by a Postgraduate Scholarship from the Natural Sciences and 18. Iso , T., L. Kedes, and Y. Hamamori. 2003 . HES and HERP families: multi- Engineering Research Council of Canada and a Junior Graduate Studentship from ple eff ectors of the Notch signaling pathway. J. Cell. Physiol. 194 : 237 – 255. the Michael Smith Foundation for Health Research. A. Raouf is supported by a 19 . van de Wetering, M. , N. Barker , I.C. Harkes , M. van der Heyden , Post-doctoral Research Award from the Canadian Institutes of Health Research. N.J. Dijk , A. Hollestelle , J.G. Klijn , H. Clevers , and M. Schutte . 2001 . A. Karsan is a Senior Scholar of the Michael Smith Foundation for Health Research. Mutant E-cadherin breast cancer cells do not display constitutive Wnt The authors declare that they have no competing fi nancial interests. signaling. Cancer Res. 61 : 278 – 284 . 20 . Imatani , A. , and R. Callahan . 2000 . Identifi cation of a novel NOTCH- Submitted: 30 May 2007 4/INT-3 RNA species encoding an activated gene product in certain Accepted: 16 October 2007 human tumor cell lines. Oncogene . 19 : 223 – 231 .

JEM VOL. 204, November 26, 2007 2947 Published November 26, 2007

21. Small , D., D. Kovalenko , D. Kacer , L. Liaw , M. Landriscina, C. Di 40 . Androutsellis-Theotokis , A. , R.R. Leker , F. Soldner, D.J. Hoeppner , Serio , I. Prudovsky , and T. Maciag . 2001 . Soluble Jagged 1 represses the R. Ravin , S.W. Poser , M.A. Rueger, S.K. Bae , R. Kittappa , and R.D. function of its transmembrane form to induce the formation of the Src- McKay . 2006 . Notch signalling regulates stem cell numbers in vitro and dependent chord-like phenotype. J. Biol. Chem. 276 : 32022 – 32030 . in vivo. Nature . 442 : 823 – 826 . 22. Marambaud , P. , J. Shioi, G. Serban, A. Georgakopoulos, S. Sarner, V. Nagy, 41 . Bermejo-Rodriguez , C. , M. Perez-Caro, P.A. Perez-Mancera, M. L. Baki , P. Wen , S. Efthimiopoulos , Z. Shao , et al . 2002 . A presenilin-1/ Sanchez-Beato, M.A. Piris , and I. Sanchez-Garcia . 2006 . Mouse cDNA gamma-secretase cleavage releases the E-cadherin intracellular domain and microarray analysis uncovers Slug targets in mouse embryonic fi bro- regulates disassembly of adherens junctions. EMBO J. 21 : 1948 – 1956. blasts. Genomics . 87 : 113– 118 . 23. Zavadil , J. , L. Cermak , N. Soto-Nieves , and E.P. Bottinger . 2004 . 42. Weng , Z. , M. Xin , L. Pablo , D. Grueneberg , M. Hagel , G. Bain , T. Integration of TGF-beta/Smad and Jagged1/Notch signalling in epithe- Muller, and J. Papkoff . 2002 . Protection against anoikis and down-regula- lial-to-mesenchymal transition. EMBO J. 23 : 1155 – 1165 . tion of cadherin expression by a regulatable beta-catenin protein. J. Biol. 24 . Krebs , L.T. , Y. Xue , C.R. Norton , J.R. Shutter , M. Maguire , J.P. Chem. 277: 18677 – 18686. Sundberg, D. Gallahan , V. Closson , J. Kitajewski , R. Callahan , et al. 43 . Conacci-Sorrell , M. , I. Simcha , T. Ben-Yedidia, J. Blechman , P. 2000 . Notch signaling is essential for vascular morphogenesis in mice. Savagner , and A. Ben-Ze ’ ev. 2003 . Autoregulation of E-cadherin ex- Genes Dev. 14 : 1343 – 1352 . pression by cadherin– cadherin interactions: the roles of ␤ -catenin sig- 25 . Leong , K.G., X. Hu, L. Li , M. Noseda , B. Larrivee , C. Hull , L. Hood , naling, Slug, and MAPK. J. Cell Biol. 163: 847 – 857 . F. Wong , and A. Karsan . 2002 . Activated Notch4 inhibits angiogenesis: 44 . Liu , S. , G. Dontu , and M.S. Wicha . 2005 . Mammary stem cells, self- role of beta 1-integrin activation. Mol. Cell. Biol. 22 : 2830 – 2841 . renewal pathways, and carcinogenesis. Breast Cancer Res. 7 : 86 – 95 . 26. Zeng , Q. , S. Li , D.B. Chepeha , T.J. Giordano , J. Li , H. Zhang , P.J. 45. Stingl , J. , C.J. Eaves , I. Zandieh, and J.T. Emerman . 2001 . Characterization Polverini , J. Nor, J. Kitajewski , and C.Y. Wang . 2005 . Crosstalk between of bipotent mammary epithelial progenitor cells in normal adult human tumor and endothelial cells promotes tumor angiogenesis by MAPK breast tissue. Breast Cancer Res. Treat. 67 : 93 – 109. activation of Notch signaling. Cancer Cell . 8 : 13 –23 . 46 . van ‘ t Veer, L.J. , H. Dai , M.J. van de Vijver, Y.D. He , A.A. Hart , M. Downloaded from 27. Hatsell , S. , T. Rowlands , M. Hiremath, and P. Cowin . 2003 . Beta-catenin Mao , H.L. Peterse , K. van der Kooy, M.J. Marton , A.T. Witteveen , et al . and Tcfs in mammary development and cancer. J. Mammary Gland Biol. 2002 . Gene expression profi ling predicts clinical outcome of breast cancer. Neoplasia. 8 : 145 – 158. Nature . 415 : 530 – 536 . 28. Wong , A.S. , and B.M. Gumbiner . 2003 . Adhesion-independent mech- 47 . Zhao , H. , A. Langerod , Y. Ji , K.W. Nowels , J.M. Nesland , R. Tibshirani, anism for suppression of tumor cell invasion by E-cadherin. J. Cell Biol. I.K. Bukholm , R. Karesen , D. Botstein , A.L. Borresen-Dale, and S.S. 161: 1191 – 1203 . Jeff rey . 2004 . Diff erent gene expression patterns in invasive lobular and 29 . Tripathi , M.K. , S. Misra , S.V. Khedkar , N. Hamilton , C. Irvin-Wilson , ductal carcinomas of the breast. Mol. Biol. Cell . 15 : 2523 – 2536 .

C. Sharan , L. Sealy , and G. Chaudhuri. 2005 . Regulation of BRCA2 48 . Giordano , T.J. , D.G. Thomas, R. Kuick , M. Lizyness , D.E. Misek , A.L. jem.rupress.org gene expression by the SLUG repressor protein in human breast cells. Smith , D. Sanders , R.T. Aljundi , P.G. Gauger , N.W. Thompson , et al. J. Biol. Chem. 280 : 17163 – 17171 . 2003 . Distinct transcriptional profi les of adrenocortical tumors uncov- 30 . Bachman , K.E. , B.H. Park , I. Rhee , H. Rajagopalan, J.G. Herman , S.B. ered by DNA microarray analysis. Am. J. Pathol. 162: 521 – 531 . Baylin , K.W. Kinzler , and B. Vogelstein . 2003 . Histone modifi cations 49 . Freije , W.A. , F.E. Castro-Vargas , Z. Fang , S. Horvath , T. Cloughesy , and silencing prior to DNA methylation of a tumor suppressor gene. L.M. Liau , P.S. Mischel , and S.F. Nelson . 2004 . Gene expression profi l- Cancer Cell . 3 : 89 – 95 . ing of gliomas strongly predicts survival. Cancer Res. 64 : 6503 – 6510 . 31. Berx , G. , K. Staes , J. van Hengel , F. Molemans , M.J. Bussemakers , A. van 50 . Jain , S. , M.A. Watson , M.K. DeBenedetti , Y. Hiraki , J.F. Moley , and on July 22, 2010 Bokhoven, and F. van Roy . 1995 . Cloning and characterization of the hu- J. Milbrandt . 2004. Expression profi les provide insights into early malig- man invasion suppressor gene E-cadherin (CDH1). Genomics. 26 : 281 – 289. nant potential and skeletal abnormalities in multiple endocrine neoplasia 32. Cameron , E.E. , K.E. Bachman, S. Myohanen , J.G. Herman, and S.B. type 2B syndrome tumors. Cancer Res. 64 : 3907 – 3913 . Baylin. 1999 . Synergy of demethylation and histone deacetylase inhibition 51 . Chen , X. , S.Y. Leung , S.T. Yuen , K.M. Chu , J. Ji , R. Li , A.S. Chan , S. in the re-expression of genes silenced in cancer. Nat. Genet. 21 : 103 – 107. Law , O.G. Troyanskaya, J. Wong, et al . 2003 . Variation in gene expres- 33. MacKenzie , F. , P. Duriez , F. Wong , M. Noseda , and A. Karsan . 2004 . sion patterns in human gastric cancers. Mol. Biol. Cell . 14 : 3208 – 3215 . Notch4 inhibits endothelial apoptosis via RBP-Jkappa-dependent and 52 . Garber , M.E. , O.G. Troyanskaya , K. Schluens , S. Petersen , Z. Thaesler , -independent pathways. J. Biol. Chem. 279 : 11657 – 11663 . M. Pacyna-Gengelbach , M. van de Rijn , G.D. Rosen , C.M. Perou , 34. Zweidler-McKay , P.A. , Y. He , L. Xu , C.G. Rodriguez , F.G. Karnell , R.I. Whyte , et al. 2001 . Diversity of gene expression in adenocarci- A.C. Carpenter , J.C. Aster , D. Allman , and W.S. Pear . 2005. Notch noma of the lung. Proc. Natl. Acad. Sci. USA . 98 : 13784 – 13789 . signaling is a potent inducer of growth arrest and apoptosis in a wide 53 . Schaner , M.E. , D.T. Ross , G. Ciaravino , T. Sorlie , O. Troyanskaya , range of B cell malignancies. Blood . 106 : 3898 – 3906 . M. Diehn , Y.C. Wang , G.E. Duran , T.L. Sikic, S. Caldeira , et al. 2003 . 35. Weng , A.P. , A.A. Ferrando , W. Lee , J.P. Morris IV, L.B. Silverman , Gene expression patterns in ovarian carcinomas. Mol. Biol. Cell. 14 : C. Sanchez-Irizarry , S.C. Blacklow , A.T. Look , and J.C. Aster . 2004 . 4376 – 4386 . Activating mutations of NOTCH1 in human T cell acute lymphoblastic 54 . Higgins , J.P. , R. Shinghal , H. Gill , J.H. Reese , M. Terris , R.J. Cohen , leukemia. Science . 306 : 269 – 271 . M. Fero , J.R. Pollack , M. van de Rijn, and J.D. Brooks . 2003 . Gene 36. Santagata , S. , F. Demichelis , A. Riva , S. Varambally , M.D. Hofer , J.L. expression patterns in renal cell carcinoma assessed by complementary Kutok , R. Kim , J. Tang , J.E. Montie , A.M. Chinnaiyan , et al. 2004 . DNA microarray. Am. J. Pathol. 162 : 925 – 932 . JAGGED1 expression is associated with prostate cancer metastasis and 55 . Frierson , H.F. , Jr ., A.K. El-Naggar , J.B. Welsh , L.M. Sapinoso , A.I. Su , recurrence. Cancer Res. 64 : 6854 – 6857 . J. Cheng , T. Saku , C.A. Moskaluk , and G.M. Hampton . 2002 . Large 37. Parker , B.S. , P. Argani , B.P. Cook , H. Liangfeng, S.D. Chartrand , scale molecular analysis identifies genes with altered expression in sali- M. Zhang , S. Saha , A. Bardelli , Y. Jiang , T.B. St Martin, et al. 2004 . vary adenoid cystic carcinoma. Am. J. Pathol. 161 : 1315 – 1323 . Alterations in vascular gene expression in invasive breast carcinoma. Cancer 56 . Linn , S.C. , R.B. West , J.R. Pollack , S. Zhu , T. Hernandez-Boussard , Res. 64 : 7857 – 7866 . T.O. Nielsen , B.P. Rubin , R. Patel , J.R. Goldblum , D. Siegmund , et al. 38 . Wu , W.S. , S. Heinrichs , D. Xu , S.P. Garrison , G.P. Zambetti , J.M. 2003 . Gene expression patterns and gene copy number changes in der- Adams , and A.T. Look . 2005 . Slug antagonizes p53-mediated apoptosis matofi brosarcoma protuberans. Am. J. Pathol. 163 : 2383 – 2395 . of hematopoietic progenitors by repressing puma. Cell . 123 : 641 – 653 . 57 . Segal , N.H. , P. Pavlidis, W.S. Noble , C.R. Antonescu , A. Viale , U.V. 39. Mungamuri , S.K. , X. Yang , A.D. Thor , and K. Somasundaram . 2006 . Wesley , K. Busam , H. Gallardo , D. DeSantis , M.F. Brennan , et al. Survival signaling by Notch1: mammalian target of rapamycin (mTOR)- 2003 . Classifi cation of clear-cell sarcoma as a subtype of melanoma by dependent inhibition of p53. Cancer Res. 66 :4715 – 4724 . genomic profi ling. J. Clin. Oncol. 21 : 1775 – 1781 .

2948 JAGGED – NOTCH SIGNALING INDUCES EMT THROUGH SLUG | Leong et al. Published November 26, 2007 ONLINE SUPPLEMENTAL MATERIAL

Leong et al., http://www.jem.org/cgi/content/full/jem.20071082/DC1

Supplemental materials and methods Immunoblotting. Concentrated supernatant from cultured cells was obtained by filtering 3-d-conditioned medium through a 100,000-kD cutoff ultrafiltra- tion membrane (Millipore).

Flow cytometry. MCF-10A cell lines were detached by incubation in PBS-based enzyme-free cell dissociation buffer (Invitrogen). Cells were stained with mouse monoclonal antibody to surface E-cadherin (Chemicon), followed by goat anti–mouse Alexa Fluor 594 (Invitrogen) secondary antibody. Samples were run on a flow cytometer (EPICS ELITE-ESP; Beckman Coulter), and data were analyzed with WinList software (Verity Software House Inc.).

Methylation assays. For global genomic DNA methylation analysis, 500 ng of genomic DNA was digested with the restriction enzymes HpaII, MspI, or McrBC (all obtained from New England Biolabs, Inc.). HpaII is unable to digest CpG-methylated DNA, whereas its isoschizomer MspI is not sensitive to CpG methylation and, thus, is a positive control for DNA digestion. McrBC digests CpG-methylated DNA only in the presence of GTP. For McrBC digestion, reactions were performed with GTP (McrBC-plus-GTP) or without GTP (McrBC-minus-GTP). Entire reaction mixtures were assessed in 2% TAE- agarose gels containing ethidium bromide. For HpaII and McrBC-plus-GTP reactions, digested DNA products above 6 kb were quantitated by densitometry. For control and McrBC-minus-GTP reactions, undigested DNA products corresponding to the topmost DNA band were quantitated by densitometry. Data are expressed as a ratio of digested over undigested DNA products (digested/undigested ratio) and represent the mean ratio + SEM from five control tumors and five XNotch4 tumors.

Immunostaining. 7-μm-thick tumor cryosections were stained with hematoxylin and eosin, rabbit polyclonal antibody to HA (BAbCo), and rat monoclonal Downloaded from antibody to mouse CD31 (BD Biosciences). For quantitation of the percentage of CD31-stained area, at least six random fields at 200× magnification were analyzed per tumor using Northern Eclipse software. Vascular density was quantitated by expressing the CD31-stained area as a percentage of the total tumor area. For quantitation of the number of vessels per square millimeter, entire tumor sections were analyzed using Northern Eclipse software. Data are expressed as the mean percentage of CD31-stained area + SEM from 14 control tumors and 12 XNotch4 tumors, and the mean number of vessels per square millimeter + SEM from 13 control tumors and 16 XNotch4 tumors.

REFERENCES 1. Noseda, M., L. Chang, G. McLean, J.E. Grim, B.E. Clurman, L.L. Smith, and A. Karsan. 2004. Notch activation induces endothelial cell cycle arrest and jem.rupress.org participates in contact inhibition: role of p21Cip1 repression. Mol. Cell Biol. 24:8813–8822. 2. Noseda, M., G. McLean, K. Niessen, L. Chang, I. Pollet, R. Montpetit, R. Shahidi, K. Dorovini-Zis, L. Li, B. Beckstead, R.E. Durand, P.A. Hoodless, and A. Karsan. 2004. Notch activation results in phenotypic and functional changes consistent with endothelial-to-mesenchymal transformation. Circ. Res. 94:910–917. 3. Shou, J., S. Ross, H. Koeppen, F.J. de Sauvage, and W.Q. Gao. 2001. Dynamics of notch expression during murine prostate development and tumorigenesis. Cancer Res. 61:7291–7297. 4. Muller, P., S. Kietz, J.A. Gustafsson, and A. Strom. 2002. The anti-estrogenic effect of all-trans-retinoic acid on the breast cancer cell line MCF-7 is dependent on HES-1 expression. J. Biol. Chem. 277:28376–28379. on July 22, 2010 5. MacKenzie, F., P. Duriez, B. Larrivee, L. Chang, I. Pollet, F. Wong, C. Yip, and A. Karsan. 2004. Notch4-induced inhibition of endothelial sprouting requires the ankyrin repeats and involves signaling through RBP-Jkappa. Blood. 104:1760–1768. 6. Fujita, N., D.L. Jaye, M. Kajita, C. Geigerman, C.S. Moreno, and P.A. Wade. 2003. MTA3, a Mi-2/NuRD complex subunit, regulates an invasive growth pathway in breast cancer. Cell. 113:207–219. 7. Herman, J.G., J.R. Graff, S. Myohanen, B.D. Nelkin, and S.B. Baylin. 1996. Methylation-specific PCR: a novel PCR assay for methylation status of CpG islands. Proc. Natl. Acad. Sci. USA. 93:9821–9826. 8. Graff, J.R., E. Gabrielson, H. Fujii, S.B. Baylin, and J.G. Herman. 2000. Methylation patterns of the E-cadherin ′5 CpG island are unstable and reflect the dynamic, heterogeneous loss of E-cadherin expression during metastatic progression. J. Biol. Chem. 275:2727–2732. 9. Okubo, T., T.K. Truong, B. Yu, T. Itoh, J. Zhao, B. Grube, D. Zhou, and S. Chen. 2001. Down-regulation of promoter 1.3 activity of the human aroma- tase gene in breast tissue by zinc-finger protein, snail (SnaH). Cancer Res. 61:1338–1346. 10. van Doorn, R., R. Dijkman, M.H. Vermeer, J.J. Out-Luiting, E.M. van der Raaij-Helmer, R. Willemze, and C.P. Tensen. 2004. Aberrant expression of the tyrosine kinase receptor EphA4 and the transcription factor twist in Sezary syndrome identified by gene expression analysis. Cancer Res. 64:5578–5586. 11. Decary, S., J.T. Decesse, V. Ogryzko, J.C. Reed, I. Naguibneva, A. Harel-Bellan, and C.E. Cremisi. 2002. The retinoblastoma protein binds the promoter of the survival gene bcl-2 and regulates its transcription in epithelial cells through transcription factor AP-2. Mol. Cell Biol. 22:7877–7888. 12. Liu, L., M. Zeng, and J.S. Stamler. 1999. Hemoglobin induction in mouse macrophages. Proc. Natl. Acad. Sci. USA. 96:6643–6647. 13. MacKenzie, F., P. Duriez, F. Wong, M. Noseda, and A. Karsan. 2004. Notch4 inhibits endothelial apoptosis via RBP-Jkappa-dependent and -independent pathways. J. Biol. Chem. 279:11657–11663.

Table S1. PCR primer sequences

Species specificity Gene Primer Human Mouse Primer sequences (5′ to 3′) Reference identifiera Jagged1 A X Forward: CTATGATGAGGGGGATGCT 1 B X Forward: AATGGAGACTCCTTCACCTGT 2 C X X Reverse: CGTCCATTCAGGCACTGG 2 Jagged2 A X Forward: TGGGATGCCTGGCACA This study B X Forward: CAGGGCACGCGGTGT This study C X X Reverse: CCGGCAGATGCAGGA This study Delta-like1 A X Forward: GAGGGAGGCCTCGTGGA This study Published November 26, 2007B X Forward: TGGTTCTCTCAGAGTTAGCAGAG This study C X X Reverse: AGACCCGAAGTGCCTTTGTA This study Delta-like3 A X X Forward: CGGATGCACTCAACAACCT This study B X Reverse: GAAGATGGCAGGTAGCTCAA This study C X Reverse: ATAGATGTCTCTGGGGAGATGA This study Delta-like4 A X X Forward: GCATTGTTTACATTGCATCCTG This study B X Reverse: GCAAACCCCAGCAAGAGAC This study C X Reverse: GTAGCTCCTGCTTAATGCCAAA This study Notch1 A X Forward: CACTGTGGGCGGGTCC 3 B X Reverse: GTTGTATTGGTTCGGCACCAT 3 C X Forward: GGCCACCTCTTCACTGCTTC 2 D X X Reverse: CCGGAACTTCTTGGTCTCCA 2 Notch2 A X Forward: AATCCCTGACTCCAGAACG This study B X Forward: AACTGGAGAGTCCAAGAAACG 2 C X X Reverse: TGGTAGACCAAGTCTGTGATGAT 2 Notch3 A X Forward: TGAGACGCTCGTCAGTTCTT 1 B X Forward: CACCTTGGCCCCCTAAG 2 C X X Reverse: TGGAATGCAGTGAAGTGAGG 2 Notch4 A X Forward: TAGGGCTCCCCAGCTCTC 1 B X Forward: CAAGCTCCCGTAGTCCTACTTC 2 C X X Reverse: GGCAGGTGCCCCCATT 2 HES1 A X Forward: AGGCGGACATTCTGGAAATG 4 B X Reverse: CGGTACTTCCCCAGCACACTT 4 HEY1 A X Forward: GGAGAGGCGCCGCTGTAGTTA 5 B X Reverse: CAAGGGCGTGCGCGTCAAAGTA 5 Downloaded from C X X Forward: GAGAAGCAGGGATCTGCTAA This study D X X Reverse: CCCAAACTCCGATAGTCCAT This study HEY2 A X X Forward: ACAGGGGGTAAAGGCTACTTTG This study B X Reverse: CTGCTGCTGCTGCGTTT This study C X X Reverse: GAAGGACAGAGGGAAGCTGTGTG 5 HEYL A X X Forward: TCCCCACTGCCTTTGAG This study B X Reverse: CTGCTGGGGGCGACA This study C X X Reverse: GGCACTCTTCCCAGGAT This study jem.rupress.org E-cadherin A X Forward: CAGCACGTACACAGCCCTAA 6 B X Reverse: ACCTGAGGCTTTGGATTCCT 6 C X Forward: TTAGGTTAGAGGGTTATCGCGT 7 D X Reverse: TAACTAAAAATTCACCTACCGAC 7 E X Forward: TAATTTTAGGTTAGAGGGTTATTGT 7 F X Reverse: CACAACCAATCAACAACACA 7

G X Forward: GTTTAGTTTTGGGGAGGGGTT 8 on July 22, 2010 H X Reverse: ACTACTACTCCAAAAACCCATAACTAA 8 Slug A X Forward: AGATGCATATTCGGACCCAC 6 B X Reverse: CCTCATGTTTGTGCAGGAGA 6 Snail A X Forward: AATCGGAAGCCTAACTACAGCGAG 9 B X Reverse: CCTTGGCCTCAGAGAGCTGG 9 SIP1 A X Forward: ACACCCCTGGCACAACAA This study B X Reverse: GTGTCACTGCGCTGAAGGTA This study Twist1 A X Forward: CACTGAAAGGAAAGGCATCA 10 B X Reverse: GGCCAGTTTGATCCCAGTAT 10 GAPDH A X Forward: GGACCTGACCTGCCGTCTAGAA 11 B X Reverse: GGTGTCGCTGTTGAAGTCAGAG 11 C X Forward: AATGTGTCCGTCGTGGATCT 12 D X Reverse: CCCTGTTGCTGTAGCCGTAT 12 E X X Forward: CCCATCACCATCTTCCAG 13 F X X Reverse: ATGACCTTGCCCACAGCC 13 aThe primer identifier is a letter assigned to each individual primer. Various combinations of two primers (Table S2, Primer set) are used to amplify the target gene of interest.

Published November 26, 2007

Table S2. PCR primer sets and conditions Gene Primer Annealing Product size seta temperature (°C) (bp) Human Jagged1 A/C 53 507 Mouse Jagged1 B/C 53 383 Human Jagged2 A/C 53 550 Mouse Jagged2 B/C 58 550 Human Delta-like1 A/C 55 448 Mouse Delta-like1 B/C 55 410 Human Delta-like3 A/B 55 338 Mouse Delta-like3 A/C 55 329 Human Delta-like4 A/B 60 456 Mouse Delta-like4 A/C 55 473 Human Notch1 A/B 55 85 Mouse Notch1 C/D 60 529 Human Notch2 A/C 53 589 Mouse Notch2 B/C 53 583 Human Notch3 A/C 53 667 Mouse Notch3 B/C 60 449 Human Notch4 A/C 60 486 Downloaded from Mouse Notch4 B/C 53 486 Human CSL A/B 57 123 Human HES1 A/B 55 103 Human HEY1 A/B 57 428 Human and mouse HEY1 C/D 53 137 Human HEY2 A/B 57 574 Human and mouse HEY2 A/C 53 531 jem.rupress.org Human HEYL A/B 53 583 Human and mouse HEYL A/C 53 391 Human E-cadherin A/B 53 159 Human E-cadherin-M C/D 57 116 Human E-cadherin U E/F 53 97 Human E-cadherin–promoter sequencing G/H 50 270 Human Slug A/B 53 258 on July 22, 2010 Human Snail A/B 50 400 Human SIP1 A/B 53 234 Human GAPDH A/B 53 142 Mouse GAPDH C/D 53 256 Human and mouse GAPDH E/F 53 446 aThe primer set is a primer identifier pair (see Table S1 for individual primer identifiers) used to amplify the target gene of interest.

Published November 26, 2007

Table S3. Gene accession numbers associated with Oncomine microarray data sets Gene Accession no. Breast 1 Breast 2 Adrenal Brain Endocrine Gastric Lung Ovarian Renal Salivary Sarcoma 1 Sarcoma 2 Jagged1 NM_000214 X Jagged1 AA933616 X Jagged1 R70684 X X X X X X Jagged1 U77914 X X X X X Jagged1 AI378220 X X Jagged1 U73936 X Jagged1 U61276 X Jagged1 AI457817 X Jagged1 T96855 X Slug AI572079 X Slug H57309 X X X X X X Slug N91754 X X X X X X Slug U69196 X X X X Slug NM_003068 X Snail NM_005985 X

Snail AA465052 X X X Downloaded from Notch1 NM_017617 X Notch1 AA903201 X Notch1 H18865 X HEY1 NM_012258 X HEY1 R61374 X HEY2 NM_012259 X

HEY2 AI299482 X jem.rupress.org HEYL AL040198 X HEYL AA969508 X HEYL R27319 X X indicates accession numbers associated with the particular data set. on July 22, 2010