UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Role of notch signaling in angiogenesis and breast cancer Leong, Kevin G. 2005

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2005-104104.pdf [ 37.75MB ]
Metadata
JSON: 831-1.0099820.json
JSON-LD: 831-1.0099820-ld.json
RDF/XML (Pretty): 831-1.0099820-rdf.xml
RDF/JSON: 831-1.0099820-rdf.json
Turtle: 831-1.0099820-turtle.txt
N-Triples: 831-1.0099820-rdf-ntriples.txt
Original Record: 831-1.0099820-source.json
Full Text
831-1.0099820-fulltext.txt
Citation
831-1.0099820.ris

Full Text

ROLE OF NOTCH SIGNALING IN ANGIOGENESIS AND BREAST CANCER  by KEVIN G. LEONG B.Sc, The University of British Columbia, 1999  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF i  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Experimental Medicine  THE UNIVERSITY OF BRITISH COLUMBIA February 2005  © Kevin G. Leong, 2005  ABSTRACT  The Notch family of transmembrane receptors consists of four members in mammals, Notchl through Notch4. Upon ligand binding, Notch receptors become activated and participate in intracellular signaling pathways that regulate cell fate decisions. Notchl-3 are expressed on numerous cell types. Because Notch4 is primarily expressed on endothelial cells, we postulated that Notch4 activation would modulate cell fate decisions in an endothelial-specific manner. Angiogenesis, the sprouting of endothelial cells from pre-existing microvessels, requires modulation of the endothelial cell phenotype. We have identified a role for Notch4 activation in the regulation of angiogenesis. Expression of activated Notch4 inhibits endothelial sprouting in vitro and angiogenesis in vivo. Activated Notch4 does not inhibit endothelial cell migration through the extracellular matrix protein fibrinogen, whereas migration through collagen is inhibited. Activated Notch4 increases endothelial cell adhesion to collagen by modulating the affinity state of cell-surface collagen receptors belonging to the pi integrin family. Specifically, activated Notch4 converts pi integrin from an inactive, non-ligand-binding state to an active, high-affinity conformation. Our findings suggest that Notch4 activation in endothelial cells inhibits angiogenesis in part by promoting pi integrin-mediated adhesion to the underlying matrix. Although Notch signaling regulates normal cellular processes, increasing evidence suggests a role for aberrant Notch signaling in cellular transformation. During tumor progression, epithelial tumor  cells often  acquire  a mesenchymal  phenotype through  epithelial-to-  mesenchymal transition (EMT), a process that promotes invasion and dissemination of cancer cells. In human breast cancer, E M T directly correlates with downregulated expression of the adherens junction protein epithelial (E)-cadherin. Given that Notch pathway elements are  n  expressed at sites of epithelial-mesenchymal cell-cell interactions during embryogenesis and within primary human breast tumors, we investigated whether Notch signaling would modulate E-cadherin expression in human breast cells. Our studies identify activated Notch signaling as a novel mechanism for the downregulation of E-cadherin expression in normal human breast epithelial cells. In human breast tumor xenografts lacking E-cadherin expression, we show that a soluble inhibitor of Notch signaling attenuates E-cadherin promoter methylation and induces Ecadherin re-expression. This re-induction of E-cadherin in turn inhibits p-catenin nuclear accumulation, resulting in a marked reduction in breast tumor growth and metastasis.  iii  TABLE OF CONTENTS  ABSTRACT  ii  TABLE OF CONTENTS  iv  LIST O F F I G U R E S  vii  LIST O F T A B L E S  ix  LIST O F A B B R E V I A T I O N S  x  ACKNOWLEDGEMENTS  xiv  CHAPTER 1  1  INTRODUCTION  1.1 NOTCH 1.1.1 Overview of Notch functions 1.1.2 The Notch receptor and ligand families 1.1.2.1 Notch receptors 1.1.2.2 Notch ligands 1.1.3 Notch receptor activation 1.1.3.1 Ligand-mediated Notch activation 1.1.3.2 Constitutive Notch activation 1.1.4 Notch signaling 1.1.4.1 CBF1-mediated transcriptional repression 1.1.4.2 Notch-mediated CBF1 transcriptional activation 1.1.4.3 Notch target genes 1.2 THE NOTCH PATHWAY AND DISEASE PATHOGENESIS 1.2.1 Notch ligand-associated diseases 1.2.2 Notch receptor-associated diseases 1.2.2.1 Notch and murine breast cancer 1.2.2.2 Notch and human breast cancer 1.3 ANGIOGENESIS 1.3.1 Mechanism of angiogenesis 1.3.2 Angiogenic activators 1.3.3 Angiogenic inhibitors 1.3.4 Pathological angiogenesis 1.3.4.1 Tumor angiogenesis 1.3.4.2 Tumor metastasis 1.3.5 Notch and blood vessel development 1.4 CELL ADHESION 1.4.1 Overview of cell adhesion molecules 1.4.2 Integrins 1.4.2.1 Overview of integrin functions  IV  1 1 2 2 5 7 7 10 10 10 12 13 19 19 20 23 25 26 27 30 33 36 36 38 40 45 45 46 46  1.4.2.2 1.4.2.3 1.4.2.4 1.4.2.5  Integrin receptors Integrin ligands Integrin signaling Regulation of integrin-mediated adhesion 1.4.2.5.1 Integrin affinity 1.4.2.5.2 Integrin avidity 1.4.2.6 Endothelial integrins 1.4.3 Cadherins 1.4.3.1 Overview of the cadherin family 1.4.3.2 E-cadherin structure and function 1.4.3.3 E-cadherin downstream signaling 1.5 EPITHELIAL-TO-MESENCHYMAL TRANSITION (EMT) 1.5.1 Physiological EMT 1.5.2 Pathological EMT 1.5.2.1 Mechanisms of E-cadherin silencing 1.5.2.2 E-cadherin downregulation in breast cancer 1.5.3 Notch and EMT 1.6 AIMS OF THE STUDY... CHAPTER 2  MATERIALS AND METHODS  2.1 TISSUE CULTURE 2.1.1 Cell culture 2.1.2 Primary human breast epithelial cell isolation 2.1.3 Gene transfer 2.2 PROTEIN ANALYSIS 2.2.1 Immunoblotting 2.2.2 Immunofluorescence microscopy 2.2.3 Immunohistochemistry 2.2.4 Flow cytometry 2.3 ANGIOGENESIS ASSAYS .... 2.3.1 Endothelial sprouting assay 2.3.2 Chick chorioallantoic membrane (CAM) assay 2.3.3 Proliferation assay 2.3.4 Migration assay 2.3.5 Adhesion assay 2.3.6 Ligand binding assay 2.4 RIBONUCLEIC ACID (RNA) ANALYSIS 2.4.1 RNA isolation 2.4.2 Reverse transcriptase-polymerase chain reaction (RT-PCR).. 2.5 METHYLATION ASSAYS 2.5.1 Methylation-specific PCR (MSP) 2.5.2 Genomic bisulfite sequencing 2.5.3 Global genomic deoxyribonucleic acid (DNA) methylation analysis 2.6 TUMORIGENICITY ASSAYS  46 49 51 52 52 54 55 56 56 57 61 63 63 63 64 66 67 69 71  71 71 72 73 74 74 75 77 78 80 80 81 82 83 84 85 86 86 86 89 89 89 90 90  2.7 STATISTICAL ANALYSIS CHAPTER 3  .-  A C T I V A T E D N 0 T C H 4 INHIBITS A N G I O G E N E S I S : R O L E O F pl-INTEGRIN A C T I V A T I O N  3.1 ABSTRACT 3.2 RESULTS 3.2.1 Constitutively-active Notch4 inhibits endothelial sprouting in vitro 3.2.2 Constitutively-active Notch4 inhibits angiogenesis in vivo 3.2.3 Notch4 inhibition of endothelial sprouting in vitro cannot be explained by reduced endothelial cell proliferation 3.2.4 Notch4 inhibits endothelial cell migration through collagen but not fibrinogen 3.2.5 Notch4 promotes adhesion to extracellular matrix (ECM) proteins through p1 integrins 3.2.6 Increased p1 integrin-mediated adhesion plays a role in the Notch4 inhibition of endothelial sprouting 3.2.7 Activation of p1 integrins is sufficient to inhibit endothelial sprouting in vitro and angiogenesis in vivo 3.3 DISCUSSION CHAPTER 4  N O T C H S I G N A L INHIBITION A T T E N U A T E S B R E A S T TUMOR GROWTH BY REVERSING THE MESENCHYMAL PHENOTYPE  4.1 ABSTRACT 4.2 RESULTS 4.2.1 Activated Notch signaling inhibits E-cadherin expression in human breast epithelial cells 4.2.2 Notch signal inhibition reduces human breast tumor growth in vivo 4.2.3 Notch signal inhibition induces E-cadherin expression in human breast tumors in vivo 4.2.4 Expression of E-cadherin alone is sufficient to inhibit human breast tumor growth in vivo 4.2.5 Induction of E-cadherin expression is mediated by attenuated E-cadherin promoter methylation 4.3 DISCUSSION CHAPTER 5  SUMMARY, PERSPECTIVES, A N D FUTURE DIRECTIONS  91 92  92 93 93 95 98 100 102 107 109 111  116  116 117 117 121 131 133 135 142 146  REFERENCES  151  APPENDIX  198  vi  LIST O F F I G U R E S  CHAPTER 1  Figure Figure Figure Figure Figure Figure Figure  1.1 1.2 1.3 1.4 1.5 1.6 1.7  Figure 1.8  The mammalian Notch receptor family The mammalian Notch ligand family Ligand-mediated activation of Notch signaling Mechanism of angiogenesis Structure of an integrin a/p heterodimer Integrin a/p heterodimer combinations Structure of E-cadherin  3 6 9 28 47 48 58  The p-catenin signaling pathway  60  CHAPTER 2  No Figures CHAPTER 3  Figure 3.1  Activated Notch4 inhibits endothelial sprouting from gelatincoated microcarrier beads in vitro Figure 3.2 Activated Notch4 inhibits angiogenesis in the chick CAM in vivo Figure 3.3 Immunohistochemical analysis of Notch4IC expression in the CAM Figure 3.4 Activated Notch4 does not inhibit HMEC proliferation Figure 3.5 Activated Notch4 inhibits endothelial cell migration through collagen but not fibrinogen Figure 3.6 Activated Notch4 promotes endothelial cell adhesion to various extracellular matrix proteins through p1 integrins Figure 3.7 Activated Notch4 does not increase endothelial cell-surface expression of p1 integrins but enhances binding of soluble collagen Figure 3.8 Activated Notch4-expressing cells display p1 integrins in a high-affinity conformation Figure 3.9 Activated Notch4 does not inhibit endothelial sprouting from dextran-coated microcarrier beads in vitro Figure 3.10 Activation of p1 integrins alone, independent of Notch4 activation, is sufficient to inhibit endothelial sprouting in vitro and angiogenesis in vivo  94 96 97 99 101 103 105 106 108 110  CHAPTER 4  Figure 4.1  Activated Notch signaling inhibits E-cadherin expression in MCF-10A human breast epithelial cells  vii  118  Figure 4.2  Activated Notch signaling inhibits E-cadherin expression in primary human breast epithelial cells Figure 4.3 MDA-MB-231 and Lewis lung carcinoma parental tumor cells both express Notch ligands and receptors, with MDA-MB-231 cells expressing greater levels of HES1 Figure 4.4 Extracellular Notch4 (XNotch4) protein is secreted Figure 4.5 Notch signal inhibition reduces human breast tumor growth in vivo Figure 4.6 Notch signal inhibition reduces metastasis of human breast tumors in vivo Figure 4.7 Soluble XNotch4 inhibits Notch signaling specifically in the tumor cell compartment of MDA-MB-231 tumors in vivo Figure 4.8 Soluble XNotch4 does not affect vascular density in vivo Figure 4.9 Notch signal inhibition induces E-cadherin expression in xenografted human breast tumor cells in vivo Figure 4.10 Expression of E-cadherin alone, independent of Notch signal inhibition, is sufficient to inhibit human breast tumor growth in vivo Figure 4.11 The human E-cadherin proximal promoter Figure 4.12 Notch signal inhibition does not affect the expression of E-cadherin transcriptional repressor proteins Figure 4.13 Notch signal inhibition attenuates E-cadherin promoter methylation, but does not induce a generalized demethylation of the genome CHAPTER 5  No Figures  viii  120 122 124 125 127 128 130 132 134 136 138 140  LIST O F T A B L E S CHAPTER 1  No Tables CHAPTER 2  Table 2.1  PCR primer sequences  Table 2.2  PCR primer sets and conditions  CHAPTER 3  No Tables CHAPTER 4  No Tables CHAPTER 5  No Tables  ix  LIST O F A B B R E V I A T I O N S  Ab ADAM AMF Ang AP APC ARNT ATP bHLH BrdU C Ca CADASIL 2 +  CAM CBF CDH cDNA CHF CIR CpG DAPI deltaEF DGEA Dll DMEM dNTP DSL E E-cadherin ECM EDTA EGF EMT EndoMT EpCAM E-selectin FACS FAK FCS FGF FGF-R FITC  antibody a disintegrin and metalloprotease autocrine motility factor angiopoietin activator protein adenomatous polyposis coli aryl hydrocarbon receptor nuclear translocator adenosine triphosphate basic helix-loop-helix 5-bromo-2'-deoxyuridine carboxycalcium cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy chorioallantoic membrane C protein binding factor cadherin complementary deoxyribonucleic acid cardiovascular helix-loop-helix factor C protein binding factor 1-interacting corepressor cytosine-phosphate-guanine 4',6-diamidino-2-phenylindole delta-crystallin/E2-box factor aspartic acid-glycine-glutamic acid-alanine Delta-like Dulbecco's modified Eagle's medium 2'-deoxynucleoside 5-triphosphate Delta/Serrate/Lag-2 embryonic day epithelial cadherin extracellular matrix ethylenediaminetetraacetic acid epidermal growth factor epithelial-to-mesenchymal transition endothelial-to-mesenchymal transition epithelial cell adhesion molecule endothelial-selectin fluorescence-activated cell sorter focal adhesion kinase fetal calf serum fibroblast growth factor fibroblast growth factor-receptor fluorescein isothiocyanate  X  GAPDH GBP GFOGER GFP GSK3p GTP H&E HA HAT HAV HDAC HERP HES HESR HEY HEYL HGF HIF HMEC HRT HUVEC IAP ICAM IDAPS IRES kDa LEF LFA LRP L-selectin LTR mAb Maml MAP MASH MATH MDR MHC MIG MIY MMP MMTV mRNA  glyceraldehyde-3-phosphate dehydrogenase glycogen synthase kinase-binding protein glycine-phenylalanine-hydroxyproline-glycine-glutamic acidarginine green fluorescent protein glycogen synthase kinase 3p guanosine triphosphate hematoxylin and eosin hemagglutinin histone acetyltransferase histidine-alanine-valine histone deacetylase hairy/enhancer-of-split-related repressor protein hairy/enhancer-of-split hairy/enhancer-of-split-related hairy/enhancer-of-split-related with YRPW hairy/enhancer-of-split-related with YRPW-like hepatocyte growth factor hypoxia inducible factor human dermal microvascular endothelial cell hairy-related transcription factor human umbilical vein endothelial cell intracisternal A particle intercellular adhesion molecule isoleucine-aspartic acid-alanine-proline-serine internal ribosomal entry site kilodalton lymphoid enhancer factor lymphocyte function-associated low density lipoprotein receptor-related protein lymphocyte-selectin long terminal repeat monoclonal antibody Mastermind-like mitogen-activated protein mammalian achaete-scute homologue mammalian atonal homologue multidrug resistance major histocompatibility complex murine stem cell virus - internal ribosomal entry site - green fluorescent protein murine stem cell virus - internal ribosomal entry site - yellow fluorescent protein matrix metalloprotease Mouse Mammary Tumor Virus messenger ribonucleic acid  xi  MSCV MSP N NFKB NOD/SCID  Notchl IC Notch4IC NotchIC PBS PCR PDZ  PECAM PEST PI3K PKC PLC PIGF P-selectin QIDS  RAM RBPJK RGD  RIP RNA RNAi RNase RT-PCR SDS  SIP SIRT SKIP SMRT SNP SV40 TACE TAD  TAE T-ALL TAN TCF TCRp TE(V/I)GAF TFIIA TFIID TGFp  murine stem cell virus methylation-specific polymerase chain reaction aminonuclear factor KB non-obese diabetic/severe combined immunodeficient Notchl intracellular domain Notch4 intracellular domain Notch intracellular domain phosphate-buffered saline polymerase chain reaction PSD-95/Dlg/zonula occludens-1 platelet endothelial cell adhesion molecule proline-glutamine-serine-threonine phosphatidylinositol 3-kinase protein kinase C phospholipase C placenta growth factor platelet-selectin glutamine-isoleucine-aspartic acid-serine recombination signal binding protein vk-associated module recombination signal binding protein JK arginine-glycine-aspartic acid , regulated intramembranous proteolysis ribonucleic acid ribonucleic acid interference ribonuclease reverse transcriptase polymerase chain reaction sodium dodecyl sulfate Smad-interacting protein sirtuin Ski interacting protein silencing mediator of retinoid and thyroid hormone receptor single nucleotide polymorphism simian virus 40 tumor necrosis factor a converting enzyme transactivation domain tris acetic acid T-cell acute lymphoblastic leukemia translocation-associated Notch T-cell factor T-cell receptor p threonine-glutamic acid-(valine/isoleucine)-glycine-alaninephenylalanine transcription factor MA transcription factor IID transforming growth factor p  xii  Tie TIMP  TLE TNFa  uPA uPAR VCAM VE-cadherin VEGF VEGF-R VSMC vWF WAP WDR Wnt WRPW XNotch YFP YHSW YRPW ZO  tyrosine kinase with immunoglobulin and epidermal growth factor homology domain tissue inhibitor of metalloprotease transducin-like enhancer-of-split tumor necrosis factor a urokinase-type plasminogen activator urokinase-type plasminogen activator receptor vascular cell adhesion molecule vascular endothelial cadherin vascular endothelial growth factor vascular endothelial growth factor-receptor vascular smooth muscle cell von Willebrand factor whey acidic protein tryptophan-aspartic acid-arginine wingless-type tryptophan-arginine-proline-tryptophan extracellular Notch yellow fluorescent protein tyrosine-histidine-serine-tryptophan tyrosine-arginine-proline-tryptophan zonula occludens  xiii  ACKNOWLEDGEMENTS  Where to begin. First and foremost I would like to thank my parents, David and Peggy Leong, for supporting me both emotionally and financially. They have always believed in my abilities and have been an unwavering source of wisdom and love. To my older brother Terry Leong and my younger sister Kathryn Leong, I thank you for your honesty and kindness. Sharing with you has made my successes sweeter and my disappointments bearable. I also thank my sister-in-law Amy Leong, for introducing me to my niece Meaghan and my nephew Kyle. I am grateful to my grandparents for their care, and to my many aunts, uncles, and cousins for their words of encouragement. I would like to acknowledge my colleagues, both past and present, who have influenced my academic career. Thank you to Xiaolong Hu for laying the groundwork on which my research was built, to Ingrid Pollet for providing technical assistance with animal studies, to Fred Wong and Denise McDougal for performing flow cytometry, and to Kyle Niessen for generating cell lines. Thank you to the other members of my laboratory and my department for your support. A special thank  you to Maisie Lo. Your encouragement,  enthusiasm, and  companionship made my thesis writing experience enjoyable. To my graduate committee members, Drs. Peggy Olive, Michel Roberge, and Calvin Roskelley, thank you for your dedication and insight into my research project. To my research supervisor Dr. A l y Karsan, I am indebted to you for your mentorship. Thank you for providing me with the opportunity to learn and discover. Last but not least, thank you to the many mice that have given their lives in my pursuit of knowledge. Your sacrifices shall not be in vain. This thesis is dedicated to the memory of my family members who have lost their battles with cancer.  xiv  Financial support for my graduate career was provided by a Summer Studentship Award from the Heart and Stroke Foundation of British Columbia and the Yukon, a University Graduate Fellowship Award from the University of British Columbia, a Doctoral Research Award from the Canadian Institutes of Health Research, and a Predoctoral Fellowship Award from the United States Department of Defense.  I  xv  Chapter 1 INTRODUCTION  1.1  NOTCH  1.1.1  Overview of Notch functions  1  In 1917, the first documented case of a strain of Drosophila characterized by notches at the ends of their wing blades was reported (Radtke and Raj, 2003). These notches were caused by haploinsufficiency of an unknown gene, which was subsequently cloned in the mid-1980's and identified as the gene coding for the Drosophila transmembrane receptor Notch (Kidd et al., 1986; Wharton et al., 1985). Notch orthologues have since been identified in numerous organisms including mammals (Mumm and Kopan, 2000). The Notch signaling pathway was originally described as a mechanism for the inhibition of cell differentiation, and has been reported to inhibit neurogenesis (Baker, 2000), myogenesis (Kopan et al., 1994; Nye et al., 1994), granulocytic differentiation (Li et al., 1998b), and T cell development (Robey et al., 1996; Washburn et al., 1997). B y maintaining cells in an undifferentiated state, Notch signaling allows cells to respond to inductive cues at appropriate times and thus facilitates the generation of cell diversity (Artavanis-Tsakonas et al., 1995; Weinmaster, 1997). This view of Notch function, however, proved to be oversimplified as Notch signaling can also promote cell differentiation for instance during gliogenesis (Wang and Barres, 2000). Hence activated Notch signaling can either block cell differentiation or direct cells towards an alternate differentiation fate. In addition to regulating cell fate, Notch signaling has been shown to play a role in cell proliferation (Nicolas et al., 2003; Noseda et al., 2004a; Rangarajan et al., 2001b; Ronchini and  1  Capobianco, 2001) and apoptosis (Hamada et al., 1999; Han et al., 2000; Jehn et al., 1999; MacKenzie et al., 2004b; Miele and Osborne, 1999; Nair et al., 2003; Rangarajan et al., 2001a; Shelly et al., 1999), as well as three processes directly addressed in this thesis: cell adhesion, blood vessel development, and epithelial-to-mesenchymal transition (EMT).  1.1.2 The Notch receptor and ligand families  1.1.2.1 Notch receptors Notch is synthesized in the endoplasmic reticulum as a full-length unprocessed protein composed of extracellular, transmembrane, and intracellular domains (Blaumueller et al., 1997). Following transport through the secretory pathway to the trans-golgi network, Notch is cleaved within intracellular vesicles by a furin-like convertase (Logeat et al., 1998) at a site referred to as the SI cleavage site approximately 70 amino acids N-terminal to the transmembrane domain (Blaumueller et al., 1997; Logeat et al., 1998). Two subunits are thus generated: one consisting of the majority of the extracellular domain and the other consisting of the remainder of the extracellular domain and the complete transmembrane and intracellular domains. These two subunits associate noncovalently via a calcium (Ca )-coordinated bond (Rand et al., 2000), 2+  resulting in the cell-surface expression of a mature heterodimeric Type I transmembrane receptor with an amino (N)-terminal extracellular domain and a carboxy (C)-terminal intracellular domain (Blaumueller et al., 1997) (Figure 1.1). Notch proteins contain several structural motifs critical to receptor function (Figure 1.1).' Epidermal growth factor (EGF)-like repeats are cysteine-rich consensus sequences found in members of the E G F family as well as other growth factor and receptor molecules (Wharton et al., 1985). Within the Notch extracellular domain, numerous EGF-like repeats have been  2  Notchl  Notch2  Notch3  Notch4  extracellular  EGF-like repeats  domain  -  Lin-12 / Notch repeats  transmembrane J h RAM23 domain  domain  cdc 10 / ankyrin repeats intracellular domain PEST domain  Figure 1.1: The mammalian Notch receptor family. The mammalian Notch receptor family consists of four members: Notchl through Notch4. Following synthesis in the endoplasmic reticulum, full-length unprocessed Notch proteins are transported to the trans-golgi network where they are cleaved by a furin-like convertase (referred to as S1 cleavage), thus generating two subunits. These two subunits associate non-covalently, giving rise to a mature heterodimeric transmembrane receptor expressed on the cell surface. Notch proteins contain several conserved structural motifs. The extracellular domain contains a variable number of EGF-like repeats involved in ligand binding, and three Lin-12/Notch repeats involved in Notch heterodimerization. The intracellular domain contains a R A M 2 3 domain involved in binding Notch downstream signaling proteins, seven cddO/ankyrin repeats required for mediating downstream signaling, and a P E S T domain involved in Notch protein degradation.  3  identified and are involved in mediating ligand binding (Rebay et al., 1991). Also located within the Notch extracellular domain are three tandem copies of a Lin-12/Notch repeat essential for the formation of a stable, mature Notch heterodimer (Rand et al., 2000). Within the intracellular domain, Notch proteins contain a recombination signal binding protein J (RBPJic)-associated K  module-23 (RAM23) domain involved in binding Notch downstream signaling proteins (Hsieh et al., 1996; Roehl et al., 1996; Tamura et al., 1995). Also involved in mediating downstream signaling are seven tandem copies of a cdclO/ankyrin repeat (Lubman et al., 2004). Flanking the cdclO/ankyrin repeats are three potential nuclear localization signals (Mumm and Kopan, 2000). C-terminal to the cdclO/ankyrin repeats is a putative transactivation domain (TAD) involved in transcriptional activation of downstream Notch target genes (Mumm and Kopan, 2000), followed by a proline-glutamine-serine-threonine (PEST) domain that plays a role in Notch protein degradation (Rogers et al., 1986). The mammalian family of Notch receptors consists of four members: Notchl-4 (Ellisen et a l , 1991; Lardelli et al., 1994; Uyttendaele et al., 1996; Weinmaster et al., 1992) (Figure 1.1). Research presented in this thesis primarily focuses on Notch4 and the potential functional consequences of altered Notch4 signaling. Based upon evolutionary analysis, Notch4 is the most divergent gene in the Notch family (Kortschak et al., 2001). The Notch4 amino acid sequence is 60% similar and 43% identical to the other vertebrate Notch proteins (Uyttendaele et al., 1996). Indeed, Notch4 protein exhibits key structural differences from other Notch family members. Notch4 contains the fewest number of EGF-like repeats in its extracellular domain. Notch4 contains 29 EGF-like repeats, whereas Notchl and Notch2 both have 36 and Notch3 has 34 (Uyttendaele et al., 1996). The Notch4 intracellular domain is the shortest amongst the mammalian Notch family members (Uyttendaele et al., 1996). Within the intracellular domain,  4  the PEST domain of Notch4 is significantly shorter than that of the other Notch proteins (Uyttendaele et al., 1996). In addition to structural differences, Notch4 exhibits a relatively restricted expression pattern in normal cells compared to the other Notch proteins. Specifically, Notch4 is primarily expressed on endothelial and endocardial cells (Li et al., 1998a; Shirayoshi et al., 1997; Uyttendaele et al., 1996), although there is recent evidence of Notch4 messenger ribonucleic acid (mRNA) (Vercauteren and Sutherland, 2004) and protein (Dontu et al., 2004) expression in other cell types.  1.1.2.2 Notch ligands The Notch ligand family consists of five members in mammals: Jaggedl/2 and Delta-like (Dll)l/3/4 (Bettenhausen et al., 1995; Dunwoodie et al., 1997; Lindsell et a l , 1995; Rao et al., 2000; Shawber et al., 1996) (Figure 1.2). Similar to Notch receptors, Notch ligands are Type I transmembrane proteins (Mumm and Kopan, 2000). Within the extracellular domain, Notch ligands belonging to the Jagged subfamily contain a cysteine-rich region likely involved in the ' control of Notch receptor binding specificity, as well as a von Willebrand factor (vWF) Type C domain likely involved in ligand oligomerization (Fleming, 1998; Lissemore and Starmer, 1999). Ligands of the D l l subfamily, in contrast, do not contain these motifs (Fleming, 1998). Also located in the extracellular domain of all Notch ligands is a single Delta/Serrate/Lag-2 (DSL) domain which functions as a receptor binding site (Lissemore and Starmer, 1999; Tax et al., 1994), as well as a variable number of EGF-like repeats that may stabilize receptor binding (Lieber et al., 1992). Specifically, Jaggedl and Jagged2 both have 16 EGF-like repeats, D i l l and D114 have eight, and D113 has six (Fleming, 1998; Lissemore and Starmer, 1999). The intracellular domain of Notch ligands is required for normal ligand-mediated Notch signaling (Sun and Artavanis-Tsakonas, 1996). Although Notch ligand intracellular domains exhibit  5  Jagged! Jagged2  Dill  DII3  DII4  extracellular domain  transmembrane domain intracellular domain Type C domain  Figure 1.2: The mammalian Notch ligand family. The mammalian Notch ligand family consists of five  members: Jagged1/2 and DII1/3/4. Within the extracellular domain, Jagged family members contain a cysteine-rich region likely involved in the control of Notch receptor binding specificity, as well as a vWF Type C domain likely involved in ligand dimerization. These motifs are not present in DH family members. Extracellular motifs common to all Notch ligands include a single DSL domain involved in receptor binding, as well as a variable number of EGF-like repeats that may stabilize receptor binding.  6  relatively few structural motifs (Hock et al., 1998), several reports have highlighted the possibility of nuclear signaling mediated by the ligand intracellular domain (LaVoie and Selkoe, 2003; Sun and Artavanis-Tsakonas, 1996). The Jaggedl intracellular domain contains a PSD95/Dlg/zonula occludens-1 (ZO-1) (PDZ) domain which mediates intracellular signaling in Jaggedl-expressing cells (Ascano et al., 2003). Moreover, this P D Z domain has been shown to be required for Jaggedl-induced cellular transformation (Ascano et al., 2003).  1.1.3  Notch receptor activation  1.1.3.1  Ligand-mediated Notch activation  Notch expressed on the cell surface as a heterodimer is the mature, ligand-accessible form of the receptor (Blaumueller et al., 1997). Uncleaved full-length Notch receptors, however, have been reported to be found at the cell surface (Bush et al., 2001; Sakamoto et al., 2002). As well, Notch that has been cleaved within the EGF-like repeat domain and Notch receptors lacking C-terminal sequences have been found at the plasma membrane (Wesley and Mok, 2003; Wesley and Saez, 2000a; Wesley and Saez, 2000b). The functional significance of these alternate Notch receptor forms has not been reported. In the absence of ligand binding, heterodimeric Notch receptors are inactive (Lindsell et al., 1995; Rebay et al., 1993). The extracellular domain of Notch contains a motif located between the Lin-12/Notch repeats and the transmembrane domain, termed the negative control region, that functions to repress spontaneous Notch signaling (Greenwald, 1994; Kimble et al., 1998; Lieber et al., 1993). Also located within the extracellular domain of Notch receptors are EGF-like repeats 11 and 12, which are both necessary and sufficient to bind Notch ligand (Rebay et al., 1991). Notch ligand binding to Notch receptor on an adjacent cell induces a stress-  7  based conformational change in the Notch extracellular domain, resulting in the exposure of a proteolytic cleavage site within the Notch extracellular domain referred to as the S2 cleavage site (Mumm et al., 2000; Parks et al., 2000) (Figure 1.3). Following cleavage by the metalloprotease tumor necrosis factor a (TNFa) converting enzyme (TACE) (also known as a disintegrin and metalloprotease 17, A D A M 1 7 ) (Brou et al., 2000), the newly released Notch extracellular domain is transendocytosed into the ligand-expressing cell (Parks et al., 2000). This process has been identified in both Drosophila and zebrafish and may be mediated by dynamin, a GTPase involved in the generation of clathrin-coated endocytic vesicles that is required for Notch signaling (Le Borgne and Schweisguth, 2003; Seugnet et al., 1997). As a result of transendocytosis, the negative control region of the Notch receptor is removed and thus Notch receptor activation can occur (Mumm and Kopan,.2000). Following proteolysis at the S2 cleavage site, Notch receptors undergo regulated intramembranous proteolysis (RIP) at a conserved S3 cleavage site located within the transmembrane domain (Saxena et al., 2001). Hence transmembrane-bound Notch is no longer tethered to the cell membrane and the Notch intracellular domain (NotchIC) is released into the cytoplasm, which subsequently translocates into the nucleus to effect Notch signaling (Mumm and Kopan, 2000). S3 cleavage is mediated by a protease that possesses y-secretase activity (Mumm and Kopan, 2000). Presenilin, a multipass transmembrane protein required for RIP (Wolfe et al., 1999a; Wolfe et al., 1999b), is essential for y-secretase activity during S3 cleavage of Notch (Mizutani et al., 2001) and has been shown to complex with Notch at the cell membrane (Ray et al., 1999a; Ray et al., 1999b). Because a functional y-secretase requires a complex of four proteins consisting of presenilin, nicastrin, Pen-2, and Aph-1 (Edbauer et al., 2003), whether presenilin itself is the y-secretase or not remains to be determined.  8  plasma membrane CYTOSOL EXTRACELLULAR SPACE Notch receptor -y S2 cleavage plasma membrane S3 cleavage  NotchIC  (C/T)GTGGGAA  i!T351 hes / hrt  Figure 1.3: Ligand-mediated activation of Notch signaling. Notch expressed on the cell surface as a heterodimer is the mature, ligand-accessible form of the receptor. In the absence of ligand binding, heterodimeric Notch receptors are inactive. When Notch ligand binds to Notch receptor on an adjacent cell, a series of proteolytic cleavages occur (referred to as S 2 and S 3 cleavages), resulting in release of NotchIC that subsequently translocates into the nucleus. In the absence of nuclear NotchIC, the transcription factor C B F 1 binds to the D N A sequence 5 ' - ( C / T ) G T G G G A A - 3 ' within Notch target gene promoters and represses transcription. W h e n Notch signaling is activated, nuclear NotchIC binds to C B F 1 and, following recruitment of the nuclear protein Maml, results in the formation of a ternary complex that functions as a transcriptional activator. Maml recruits the histone acetyltransferase protein p300, resulting in histone acetylation and conversion of the local chromatin structure to a form amenable to active transcription. Hence transcription of Notch target genes belonging to the H E S and H R T families occurs.  9  1.1.3.2 C o n s t i t u t i v e N o t c h a c t i v a t i o n  Although Notch proteolytic processing at the S1 cleavage site does not result in Notch activation, processing at both S2 and S3 cleavage sites results in activation of the receptor r  (Mumm and Kopan, 2000). Translocation of endogenous NotchIC to the nucleus is required for activation of Notch downstream signaling (Mumm and Kopan, 2000). In normal cells, detection of nuclear NotchIC by current biochemical and immunocytochemical techniques has proven extremely difficult as low levels are sufficient to produce a functional response (Rand et al., 2000; Schroeter et al., 1998). To circumvent this limitation, studies involving overexpression of exogenous NotchIC at easily detectable levels have been performed (Capobianco et al., 1997; Iso et al., 2001a). Furthermore, spontaneous processing of the Notch receptor can be achieved by expressing truncated Notch consisting of the intracellular and transmembrane domains without most or all of the extracellular domain (Greenwald, 1994; Mizutani et al., 2001). Such truncated Notch proteins, similar to NotchIC, do not require ligand binding for activation and hence function as constitutively active receptors. Much of our current understanding of Notch signaling has derived from studies utilizing exogenous NotchIC.  1.1.4  Notch signaling  1.1.4.1 C B F 1 - m e d i a t e d t r a n s c r i p t i o n a l  repression  Notch signaling involves the transcriptional activation of downstream target genes by nuclear NotchIC (Mumm and Kopan, 2000) (Figure 1.3). In the absence of nuclear NotchIC, target gene expression is repressed by the transcriptional repressor protein C protein binding factor 1 (CBF1; also known as  RBPJK  and KBF2) (Dou et al., 1994; Hsieh and Hayward, 1995).  CBF1 is a constitutively expressed transcription factor that binds as a monomer to the  10  deoxyribonucleic acid (DNA) consensus sequence 5'-(C/T)GTGGGAA-3' within Notch target gene promoters (Brou et al., 1994; Hamaguchi et al., 1992; Ling et al., 1994; Tun et al., 1994). Although the D N A binding domain of CBF1 has been mapped to the central third of the protein (Hsieh and Hayward, 1995; Tang and Kadesch, 2001), CBF1 does not contain a known D N A binding motif (Brou et al., 1994; Tun et al., 1994). Transcriptional repression by CBF1 is mediated via two distinct mechanisms. The first mechanism involves direct disruption of activated transcription (Olave et al., 1998). Transcription factor 1TD (TFIID) and transcription factor IIA (TFIIA) are two members of the basal transcription machinery that interact with each other to mediate transcriptional activation (Gill, 2001). CBF1 has been shown to interact with a subunit of TFIID involved in the binding of TFIIA, thus destabilizing their interaction and effectively inhibiting transcription (Olave et al., 1998). The second mechanism of CBF1 -mediated transcriptional repression involves the recruitment of corepressor proteins (Hsieh et al., 1999). The corepressor protein KyoT2 has been shown to compete with NotchIC for binding to CBF1 (Taniguchi et al., 1998). CBF1 can also bind to at least two corepressor complexes, whose individual components cooperate to inhibit transcriptional activation (Mumm and Kopan, 2000). Although the exact protein composition of these complexes is not known, several members have been identified. In one such complex, corepressor proteins include CBF1-interacting corepressor (CIR), histone deacetylase (HDAC)-2, and SAP30 (Hsieh et al., 1999; Zhou et al., 2000b). Another more well-defined corepressor complex is composed of silencing mediator of retinoid and thyroid hormone receptor (SMRT), NcoR, H D A C - 1 , SHARP, and Ski interacting protein (SKIP) (Kao et al., 1998; Oswald et al., 2002; Zhou et al., 2000a; Zhou et al., 2000b). When Notch is not activated, CBF1 directly interacts with SMRT, H D A C - 1 , SHARP, and SKIP resulting in transcriptional repression (Kao  11  et al., 1998; Oswald et al., 2002; Zhou et al., 2000a; Zhou et al., 2000b). SKIP may function as an adapter protein during Notch signal activation by mediating the interaction between CBF1 and nuclear NotchIC (Zhou et al., 2000a; Zhou et al., 2000b).  1.1.4.2 N o t c h - m e d i a t e d C B F 1 t r a n s c r i p t i o n a l a c t i v a t i o n  When Notch signaling is activated, nuclear NotchIC binds to CBF1 and converts CBF1 from a transcriptional repressor to an activator (Mumm and Kopan, 2000; Saxena et al., 2001) (Figure 1.3). Two distinct domains of NotchIC have been shown to interact with CBF1. These include the RAM23 domain of NotchIC which binds to the central third of CBF1, and the cdclO/ankyrin repeat domain of NotchIC which binds to both the N - and C-terminal regions of CBF1 (Hsieh and Hayward, 1995; Tani et al., 2001). NotchIC binding to CBF1 serves two functions: (i) to antagonize the interaction between CBF1 and corepressor proteins and hence alleviate transcriptional repression, and (ii) to recruit coactivator proteins that promote transcriptional activation (Bray and Furriols, 2001). A model for NotchIC-mediated conversion of CBF1 from a transcriptional repressor to an activator has been proposed (Mumm and Kopan, 2000). In addition to binding CBF1, SKIP can also bind to either SMRT or NotchIC in a mutually exclusive manner (Zhou et al., 2000a; Zhou et al., 2000b). Because SKIP has a higher affinity for NotchIC than SMRT, when NotchIC is present, SKJP-CBF1-SMRT complexes are converted to SKJP-CBF1 -NotchIC complexes (Mumm and Kopan, 2000). Furthermore, CBF1 binding to SHARP or NotchIC may be mutually exclusive (Oswald et al., 2002). NotchIC could therefore convert CBF1-SHARP complexes to CBF1-NotchIC complexes (Oswald et al., 2002). Hence nuclear NotchIC displaces corepressor complexes from CBF1.  12  Although NotchIC contains a transactivation domain, NotchIC itself cannot promote gene transcription (Fryer et al., 2002). Instead, NotchIC must recruit transcriptional coactivators. Histone acetyltransferases (HATs) are enzymes that acetylate histones thereby altering the structure of chromatin to a form amenable to active transcription (Peterson and Laniel, 2004). Several studies have reported a direct interaction between NotchIC and HATs (Kurooka and Honjo, 2000; Oswald et al., 2001). However, others have been unable to detect a direct interaction (Fryer et al., 2002). Mastermind-like (Maml), a nuclear protein that functions as a transcriptional activator, has been shown to be required for Notch signaling (Fryer et al., 2002; Petcherski and Kimble, 2000). Three Maml proteins are expressed in mammals: Mamll/2/3 (Kitagawa et al., 2001; Lin et al., 2002; Wu et al., 2000; Wu et al., 2002). The fact that the three Maml isoforms do not exhibit overlapping expression patterns suggests against functional redundancy (Wu et al., 2002). Maml forms a ternary complex with CBF1 -NotchIC via a direct interaction with NotchIC (Wu and Griffin, 2004). Specifically, Maml contains an N-terminal basic domain that binds to the fourth cdclO/ankyrin repeat of NotchIC (Petcherski and Kimble, 2000; Wu and Griffin, 2004). Maml contains two proposed T A D domains: TAD1 within the central region of the protein and TAD2 at the C-terminus (Wu and Griffin, 2004). TAD1 contains a binding site for the H A T protein p300/CBP (Wu and Griffin, 2004). Hence the ternary complex composed of CBFl-NotchlC-Maml functions as a transcriptional activator, resulting in Notch target gene transcription (Mumm and Kopan, 2000).  1.1.4.3 N o t c h t a r g e t g e n e s  Numerous genes have been identified as targets of the Notch signaling pathway. Among the primary targets are several genes belonging to the basic helix-loop-helix (bHLH) family of proteins (Iso et al., 2003). These proteins contain two common structural features: a basic  13  domain and an H L H domain (Murre et al., 1994). The basic domain mediates D N A binding specificity (Murre et al., 1994). The H L H domain serves two functions: (i) it contains hydrophobic residues that facilitate the formation of protein dimers (Murre et al., 1994) and (ii) it positions the basic domains of the dimer to allow for proper D N A binding (Blackwell and Weintraub, 1990; M a et al., 1994). Hence b H L H proteins bind specific D N A sequences as dimers and modulate gene transcription (Murre et al., 1994). In mammals, members of two families of b H L H proteins contain CBF1 binding sites in their promoters and thus are induced following Notch activation: (i) the hairy/enhancer-of-split (HES) family and (ii) the hairy-related transcription factor (HRT) family (Iso et al., 2003). Both the HES and H R T families function as transcriptional repressors (Iso et al., 2003). Seven mammalian HES family members have been identified to date (Akazawa et al., 1992; Bae et al., 2000; Bessho et al., 2001; Hirata et al., 2000; Ishibashi et al., 1993; KoyanoNakagawa et al., 2000; Pissarra et al., 2000; Sasai et al., 1992). However, only four are potential Notch target genes: HES1, HES4, HES5, and HES7 (Iso et al., 2003; MacKenzie et al., 2004a). HES family members repress transcription by three proposed mechanisms. One mechanism involves active repression. Specifically, HES proteins form homodimers and bind to specific D N A consensus sites, with subsequent recruitment of transducin-like enhancer-of-split (TLE) proteins (Kageyama et al., 2000; Paroush et al., 1994). Direct interaction between HES and T L E proteins is mediated by a conserved C-terminal tryptophan-arginine-proline-tryptophan (WRPW) tetrapeptide motif of HES and a C-terminal tryptophan-aspartic acid-arginine (WDR) domain of T L E (Fisher et al., 1996; Grbavec and Stifani, 1996; Jimenez et al., 1997). T L E in turn, via a conserved glycine/proline-rich domain, recruits H D A C s (Chen et al., 1999). HES has also been shown to directly bind to the H D A C protein sirtuin-1 (SIRT-1) (Takata and Ishikawa, 2003). As  14  a result of histone deacetylation by HDACs, the local chromatin structure is altered to a form that opposes gene transcription (Iso et al., 2003). A second mechanism of transcriptional repression by HES involves passive repression. HES can dimerize with b H L H proteins that normally function to activate transcription, thus effectively sequestering them in a non-functional heterodimer complex (Hirata et al., 2000; Sasai et al., 1992). A third mechanism for HESmediated transcriptional repression involves a repression domain of HES located C-terminal to the b H L H domain, termed the Orange domain (Dawson et al., 1995). Although the mechanism of transcriptional repression mediated by the Orange domain has not been well eludicated, this domain has been shown to be involved in the repression of specific transcriptional activators (Dawson et al., 1995) and may be essential for the repression of several Notch target genes (Castella et al., 2000). HRT proteins comprise a second family of b H L H proteins induced in response to Notch activation. Also known as hairy/enhancer-of-split-related  (HESR),  hairy/enhancer-of-split-  related with tyrosine-arginine-proline-tryptophan (YRPW) (HEY), HES-related repressor protein (HERP), cardiovascular H L H factor (CHF), and gridlock, the H R T family consists of three members  in  mammals:  (i)  HRT 1/HESR 1/HEY 1/HERP2/CHF2;  (ii)  HRT2/HESR2/HEY2/HERP1 /CHF 1 /gridlock; and (iii) HRT3/HESR3/HEY-like(HEYL)/HERP3 (Iso et al., 2003). Because all three H R T genes exhibit unique and dynamic expression patterns, HRT proteins may not be functionally redundant (Steidl et al., 2000). However, the fact that all three H R T proteins contain nearly identical basic domains raises the possibility that common D N A consensus sequences may be targeted (Nakagawa et al., 1999). Similar to HES proteins, HRT proteins function as transcriptional repressors (Iso et al., 2003). Two mechanisms of HRTmediated transcriptional repression have been proposed. The first mechanism involves active  15  repression. HES-mediated active repression utilizes the C-terminal W R P W tetrapeptide motif (Fisher et al., 1996; Grbavec and Stifani, 1996). In contrast, HRT-mediated repression activity does not require a C-terminal tetrapeptide motif, despite the fact that in humans both HRT1 and HRT2 have a C-terminal Y R P W motif and HRT3 has a tyrosine-histidine-serine-tryptophan (YHSW) variant motif (Iso et al., 2001b). Hence T L E proteins likely do not participate in HRTmediated active repression (Iso et al., 2001b). Instead, active repression activity resides primarily in the b H L H domain of H R T proteins (Iso et al., 2001b). H R T proteins have been shown to bind to the mSin3 complex, a corepressor complex composed of at least seven subunits (Iso et al., 2001b). This binding requires the b H L H domain of H R T proteins (Iso et al., 2001b). Among these subunits are mSin3A, a large protein that functions as a scaffold for the formation of the mSin3 complex, as well as the corepressor proteins SMRT, NcoR, H D A C - 1 , and HDAC-2 (Ayer, 1999; Knoepfler and Eisenman, 1999). Whereas HRT2 can directly interact with mSin3A and NcoR, association with HDACs is indirect (Iso et al., 2001b). HRTs can also repress gene expression via a second mechanism, passive repression. HRT2-mediated repression of vascular endothelial growth factor (VEGF) expression involves this mechanism (Chin et al., 2000). During hypoxia, the transcription factor aryl hydrocarbon receptor nuclear translocator (ARNT; also known as hypoxia inducible factor 1(3, HIFlp (Wang et al., 1995)) binds as a heterodimer with H I F l a to the V E G F promoter to induce V E G F expression (Maltepe et al., 1997). HRT2 can directly interact with A R N T , resulting in dissociation of A R N T from D N A and thus inhibition of ARNT/HIFla-dependent V E G F transcription (Chin et al., 2000). HES and H R T proteins exhibit sequence similarities (Nakagawa et al., 1999), as well as structural similarities such as the presence of b H L H and Orange domains (Iso et al., 2003). Furthermore, both HES and HRT proteins function to repress gene transcription (Iso et al.,  16  2003). Despite these common features, several key differences are evident. HES and HRT proteins exhibit different C-terminal tetrapeptide motifs (Iso et al., 2001b). C-terminal to the tetrapeptide motif, H R T proteins contain an additional conserved motif threonine-glutamic acid(valine/isoleucine)-glycine-alanine-phenylalanine (TE(V/I)GAF) that is not found in HES proteins (Iso et al., 2003). At a corresponding position within the basic domain, HES proteins have a proline residue whereas H R T proteins have a glycine residue (Iso et al., 2003). These respective residues are conserved from Drosophila to humans (Iso et al., 2003). These structural differences, together with the fact that HES and HRT proteins employ different mechanisms for transcriptional repression (Iso et al., 2003), suggest that HES and H R T proteins may regulate both common and different genes (Iso et al., 2001b). Indeed, HES1 binds D N A sequences such as the E box motif C A C G T G (Nakagawa et al., 2000), N box motifs ( C A C N A G ) (Takebayashi et al., 1994), class B sites (CANGTG) (Takebayashi et al., 1994), class C sites (CACGNG) (Kim and Siu, 1998), and a class C variant ( C A C G C A ) (Chen et al., 1997a). H R T proteins, on the other hand, preferentially bind to E box motifs. Both HRT1 and HRT2 bind to the E box motif C A C G T G (Nakagawa et al., 2000). Interestingly, HRT3 does not bind to the E box motif C A C G T G , raising the possibility that HRT3 may function differently from the other H R T proteins (Nakagawa et al., 2000). In order for D N A binding to occur, b H L H transcriptional repressors must dimerize (Vinson and Garcia, 1992). Cells that express only one of HES or H R T must therefore form homodimers to repress downstream genes (Chin et al., 2000; Kokubo et al., 1999; Leimeister et al., 2000a; Nakagawa et al., 1999; Sasai et al., 1992). In cells that express both HES and H R T proteins, however, heterodimer formation may occur. Studies have shown that HES and HRT proteins can heterodimerize in intact cells in vitro in the absence of D N A , and that repression  17  mediated by HES-HRT heterodimers is synergistic (Iso et al., 2001b; Leimeister et al., 2000a). Heterodimerization may therefore function as a strategy for the amplification of HES/HRTmediated transcriptional repression. Several genes repressed by Notch-induced b H L H proteins have been identified. The gene mammalian achaete-scute homologue-1 (mashl), which encodes for a protein that regulates neuronal differentiation (Casarosa et al., 1999; Cau et al., 1997), contains a class C variant site C A C G C A in its promoter (Chen et al., 1997a). HES1 has been shown to directly bind to this site and repress mashl gene expression (Chen et al., 1997a). Other target genes of HES1 include mammalian atonal homologue-1 (mathl) (Akazawa et al., 1995; Ben-Arie et al., 1997; Zine and de Ribaupierre, 2002), neurogenin (Fode et al., 1998; M a et al., 1998), CD4 (Kim and Siu, 1998), acid a-glucosidase (Yan et al., 2001), p21 (Castella et al., 2000), and hesl itself (Takebayashi et al., 1994). In vivo target genes of H R T proteins, in contrast, have not been reported (Iso et al., 2003). Interestingly, in cells that express both HES and H R T proteins, the mashl gene has been suggested to be a target of HES-HRT heterodimers (Iso et al., 2003; Iso et al., 2001b). Furthermore, in vitro experiments suggest that HRT2 may regulate its own expression (Nakagawa et al., 2000). In addition to b H L H proteins belonging to the HES and H R T families, consensus CBF1 binding sites are found in many genes. The identification of target genes mediated through CBF1 but independent of HES and HRT illustrates the complexity of the Notch signaling pathway. Examples of such genes include p21 (Tun et al., 1994) (although p21 has been suggested to be a target gene of HES1 (Castella et al., 2000)), cyclin D l (Ronchini and Capobianco, 2001),  NFKB2  (Oswald et al., 1998), c-jun (Rangarajan et al., 2001b), major histocompatibility complex (MHC) class I (Israel et al., 1989; Shirakata et al., 1996), CD23 (Ling et al., 1994), p-globin (Lam and  18  Bresnick, 1998), and erbB2 (Chen et al., 1997b). Adding to the complexity is the identification of Notch signaling pathways independent of CBF1 (Mumm and Kopan, 2000). Moreover, a growing number of modulators of Notch signaling have been described, including Fringe, Deltex, Notchless, and Numb (Haines and Irvine, 2003; Kadesch, 2000).  1.2  THE NOTCH PATHWAY AND DISEASE PATHOGENESIS  1.2.1  Notch ligand-associated  diseases  Aberrant expression of Notch ligands can result in disease phenotypes. Jaggedl mutations have been linked to the development of Alagille syndrome, an autosomal dominant disorder characterized by pleiotropic developmental defects that affect numerous organs including the heart, kidney, liver, eye, face, and skeleton (Li et al., 1997; Oda et al., 1997). Clinical features normally manifest within the first two years of age, with a mortality rate of 1520% (Emerick et al., 1999). Approximately 70% of patients with Alagille syndrome exhibit mutations in Jaggedl (Spinner et al., 2001), with 3-7% of patients exhibiting a complete deletion of the Jaggedl gene (Krantz et al., 1997). Among patients with mutated Jaggedl, 83% exhibit nonsense or splice-site mutations that result in the expression of a truncated Jaggedl protein (Ropke et al., 2003). The remaining 17% exhibit missense mutations that cluster around the extracellular domain of Jaggedl, in sequences encoding the D S L domain and the first EGF-like repeat (Ropke et al., 2003). Another disease associated with Jaggedl mutations is Familial tetralogy of Fallot, the most common form of complex congenital heart disease (Gridley, 2003). Upregulation of Jaggedl protein expression has been observed in a growing number of human malignancies, including cancers of the cervix (Gray et al., 1999) and prostate (Santagata et al., 2004). Upregulation of Jaggedl mRNA expression has been reported in human pancreatic cancer  19  (Miyamoto et al., 2003). Jagged2 protein overexpression has also been observed in human malignancies, including pancreatic cancer (Miyamoto et al., 2003) and multiple myeloma (Houde et al., 2004). Mutations in D113 have been identified as a causal factor in the development of spondylocostal dysostosis, a family of related diseases that affects the development of the vertebrae (Bulman et al., 2000). In the autosomal recessive form of this disease, 17 different mutations in D113 have been identified (Turnpenny et al., 2003). Structural consequences of these mutations include replacement of an essential amino acid, addition or deletion of cysteine residues within the EGF-like repeats, or truncation of the protein, and hence loss of D113 function (Turnpenny et al., 2003). Characteristics of this particular form of the disease include abnormal segmentation of the spine (Kusumi et al., 1998) and rib fusions (Bulman et al., 2000). D1I4 mRNA expression has been detected in the vasculature of human breast and kidney tumors (Mailhos et al., 2001).  1.2.2  Notch receptor-associated  diseases  Numerous diseases have been linked to the deregulated expression of Notch receptors. In humans, aberrant Notch 1 expression has been identified as a causative factor in the development of T-cell acute lymphoblastic leukemia (T-ALL) (Aster et al., 1994; Ellisen et al., 1991). Analysis of patient samples has revealed a chromosomal translocation, t(7:9)(q34;q34.3), that juxtaposes the C-terminal domain of Notchl starting from within EGF-like repeat 34 next to the T-cell receptor p (TCRp) locus (Ellisen et al., 1991). This results in the expression of a truncated constitutively active Notchl, commonly referred to as translocation-associated Notchl (TAN1) (Ellisen et al., 1991). Subsequent experiments have shown that only those forms of Notchl capable of activating a CBF1-dependent reporter can induce T - A L L (Aster et al., 2000). In  20  addition to chromosomal translocation, activating mutations in Notchl have been identified in more than 50% of human T-ALLs (Weng et al., 2004). Notchl has been linked to the development and progression of cervical cancer in humans. In contrast to T - A L L , activation of Notchl signaling in cervical cancer is primarily liganddependent (Rangarajan et al., 2001a). Whereas normal cervical tissues do not express Notchl protein in differentiated cervical epithelial cells (Zagouras et al., 1995), cervical carcinomas spontaneously overexpress Notchl (Daniel et al., 1997; Zagouras et al., 1995). Intense cytosolic and nuclear Notchl staining can be consistently detected by immunohistochemistry (Daniel et al., 1997; Zagouras et al., 1995). Interestingly, in late stage cervical cancers, Notchl protein expression is reduced (Talora et al., 2002). This downregulation of Notchl has subsequently been shown to play a role in the maintenance of malignant transformation (Talora et al., 2002). A proposed model is that Notchl expression in early-stage cervical cancer plays a tumor-promoting function whereas expression in late-stage cervical cancer is tumor-suppressive (Talora et al., 2002). Indeed, a tumor-suppressive function for Notchl has been demonstrated in mouse skin (Nicolas et al., 2003). In addition to T - A L L and cervical cancer, deregulated Notchl protein expression has been observed in human malignant melanoma (Nickoloff et al., 2003), as well as human cancers of the colon (Zagouras et al., 1995), lung (Zagouras et al., 1995), and pancreas (Miyamoto et al., 2003). Human breast cancers have been reported to express aberrant Notchl m R N A (Callahan and Egan, 2004) and protein (Parr et al., 2004; Weijzen et al., 2002). Similar to Notchl, upregulated Notch2 protein expression has been observed in human cervical cancer (Zagouras et al., 1995). However, in contrast to Notchl, downregulation of Notch2 is not required to maintain the malignancy of late-stage cervical cancers (Talora et al., 2002). Hence advanced cervical cancers display selective loss of Notchl with retention of  21  Notch2. Other studies have identified expression of activated Notch2 protein in human malignant melanoma (Nickoloff et al., 2003), as well as Notch2 protein overexpression in human cancers of the colon (Zagouras et al., 1995), pancreas (Miyamoto et al., 2003), and breast (Parr et al., 2004). Overexpression of Notch2 m R N A has been reported in human brain (Fan et al., 2004) and breast (Callahan and Egan, 2004) cancer. The human disease most closely related to deregulated Notch3 expression is cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) (Joutel et al., 2000; Joutel et al., 1996). C A D A S I L is an adult onset disease with a mean clinical presentation at 45 years of age (Chabriat et al., 1995). Typically, mortality occurs approximately 15-20 years post clinical diagnosis (Chabriat et aL, 1995). Symptoms include mood disorder, migraine, stroke, and progressive dementia (Chabriat et al., 1995). A common characteristic of this disease is cerebrovascular fragility (Joutel et al., 1996), which results from the loss of vascular smooth muscle cells (VSMC) that surround small cerebral arteries (Jung et al., 1995; Ragno et al., 1995; Tournier-Lasserve et al., 1993). In addition to providing structural support to blood vessels, V S M C s also play a role in the release of V E G F (Ruchoux et al., 2002). Hence V S M C loss results in reduced blood flow to the brain (Bruening et al., 2001). The cause of this disease has been attributed to mutations in Notch3. To date, more than 70 mutations in human Notch3 associated with C A D A S I L have been described (Hansson et al., 2004). These mutations which can be either missense or splice-site mutations (Joutel et al., 1997), occur exclusively within EGF-like repeats in the Notch3 extracellular domain (Hansson et al., 2004), and always involve either the addition or deletion of a cysteine residue (Joutel et al., 1997). EGF-like repeats normally contain six cysteine residues that participate in the formation of three disulfide bonds per repeat (Joutel et al., 1997). Mutated Notch3, in contrast, contains an odd number of cysteine  22  residues which inhibits normal disulfide bond formation (Joutel et al., 1997). Potential consequences include conformational changes in Notch3 protein, abnormal receptor proteolytic processing and oligomerization, as well as alterations in ligand binding, post-translational modifications, cell trafficking, and membrane clearance (Joutel et al., 2000; Joutel et al., 1997; Karlstrom et al., 2002). Accumulation of an extracellular Notch3 protein fragment has been detected within the brain vasculature of C A D A S I L patients (Joutel et al., 2000). Indeed, detection of this fragment within the skin vasculature is commonly used as a diagnosis for C A D A S I L (Joutel et al., 2001). Additional disease-related alterations in Notch3 include protein overexpression in human malignant melanoma (Nickoloff et al., 2003) and human pancreatic cancer (Miyamoto et al., 2003). Notch3 mRNA expression has been reported in human breast cancer (Callahan and Egan, 2004). Interestingly, Notch3 m R N A was found to be highly expressed in the vasculature of breast tumors, suggesting a role for Notch3 in breast tumor angiogenesis (Callahan and Egan, 2004). A role for Notch4 in the development of schizophrenia has been suggested (Imai et al., 2001). Recent studies, however, have failed to identify such a relationship (Glatt et al., 2005). Human malignant melanomas (Nickoloff et al., 2003) and pancreatic tumors (Miyamoto et al., 2003) have been shown to overexpress Notch4 protein. A recent report has identified Notch4 mRNA expression in human breast cancer (Callahan and Egan, 2004). A causal role for Notch4 activation in breast cancer development, however, has only been demonstrated in mice.  1.2.2.1  Notch and murine breast cancer  A role for Notch signaling in murine mammary gland tumorigenesis is well recognized. The first study to identify this relationship came from observations of Czech II mice infected with Mouse Mammary Tumor Virus ( M M T V ) (Gallahan et al., 1987; Jhappan et al., 1992). The  23  M M T V proviral genome contains enhancer sequences within its long terminal repeats (LTRs), and hence acts as an insertional mutagen by integrating into the genome of infected mammary epithelial cells to cause deregulated expression of adjacent genes (Nusse, 1988). For this reason, genes targeted for M M T V integration are referred to as "int" genes (Nusse, 1988). Czech II mice infected with M M T V develop mammary tumors (Gallahan et al., 1987; Jhappan et al., 1992). Among the mammary tumors that develop, 18% (9 of 45 tumors) exhibit M M T V integration into the gene int3, which coincides with the Notch4 gene locus (Gallahan et al., 1987). O f the tumors displaying M M T V integration into Notch4, 100% (9 of 9 tumors) exhibit integration within the region of Notch4 between the Lin-12/Notch repeats and the transmembrane domain (Gallahan et al., 1987). Hence int3 protein represents a truncated Notch4 protein consisting primarily of a constitutively-active form encoding the transmembrane and intracellular domains (Jhappan et al., 1992). Intracisternal A particles (IAPs) are murine retroviruses which, similar to M M T V , act as insertional mutagens to deregulate gene expression (Christy and Huang, 1988). LAP integration into the Notch4 locus, resulting in the expression of constitutively active Notch4, has been observed in spontaneous mammary tumors in both Czech II (Kordon et al., 1995) and Balb/c mouse strains (Lee et al., 1999). These studies clearly link aberrant Notch4 signaling to murine mammary gland tumorigenesis. To directly demonstrate that activated Notch4 signaling is oncogenic in the murine mammary gland, transgenic mouse studies have been performed. In wildtype mice, virgin mammary glands consist of a fat pad containing highly organized branching tubules that, during pregnancy, develop secretory lobules responsible for milk production (Daniel and Smith, 1999). Transgenic mice expressing activated Notch4 under the control of the M M T V L T R exhibit impaired mammary gland development and function. In virgin glands of activated Notch4  24  transgenic mice, the mammary epithelium fails to penetrate the mammary fat pad and therefore does not branch into an organized tree-like structure (Jhappan et al., 1992; Smith et al., 1995). During pregnancy, hormonal changes induce complete development of the ductal system in activated Notch4 transgenic mouse mammary glands; however, secretory lobules do not develop and hence lactation is disrupted (Jhappan et al., 1992). Importantly, these transgenic mice develop mammary carcinomas with 100% penetrance and subsequent metastasis to the lungs (Jhappan et al., 1992; Smith et al., 1995). Transgenic mice expressing activated Notch4 under the control of the whey acidic protein (WAP) promoter have also been studied. Because activity of the W A P promoter is restricted to the secretory mammary epithelium (Burdon et al., 1991), virgin glands of these animals develop normally (Gallahan et al., 1996). The mammary gland at pregnancy, however, exhibits disrupted secretory lobule development and lactation (Gallahan et al., 1996). Similar to M M T V L T R transgenic mice, WAP-Notch4 mice develop mammary carcinomas with 100% penetrance (Gallahan et al., 1996) as well as metastatic lung lesions (Gallahan et al., 1996). A role for Notchl in murine mammary tumorigenesis has also been reported. c-ErbB2 transgenic mice infected with M M T V develop mammary tumors, some of which exhibit M M T V integration into the Notchl gene and hence express constitutively activated Notchl (Dievart et al., 1999). Moreover, expression of either activated Notchl or activated Notch4 has been shown to induce transformation of mouse mammary epithelial cells in vitro (Dievart et al., 1999; Robbins et al., 1992; Soriano et al., 2000). Hence activated Notch signaling plays a causal role in murine mammary tumorigenesis.  1.2.2.2 N o t c h a n d h u m a n b r e a s t c a n c e r '  Several studies have highlighted a potential role for Notch signaling in human breast cancer development. Overexpression of constitutively active Notch4 in normal human breast  25  epithelial cells induces transformation in vitro (Imatani and Callahan, 2000). Among a panel of human breast cancer cell lines, seven of eight cell lines examined express elevated levels of a Notch4 RNA species corresponding to full-length Notch4 (Imatani and Callahan, 2000). Furthermore, two of these cell lines express an additional Notch4 RNA species encoding a truncated constitutively active Notch4 (Imatani and Callahan, 2000). In a study involving seven breast cancer specimens, Notchl protein expression was detected in all tumors examined, with normal breast tissue at the margins of tumor sections exhibiting little or no Notchl protein expression (Weijzen et al., 2002). In a second study involving 25 specimens, mRNAs for all four Notch receptors were expressed at varying frequencies (Callahan and Egan, 2004). A third study involving 97 specimens demonstrated that poorly-differentiated breast tumors were associated with elevated levels of Notchl protein and reduced patient survival (Parr et al., 2004). Interestingly, well-differentiated breast tumors were associated with elevated levels of Notch2 protein and increased patient survival (Parr et al., 2004). Taken together, these studies suggest an oncogenic role for Notchl and Notch4 and a tumor-suppressive role for Notch2 in human breast cancer development.  1.3  ANGIOGENESIS Angiogenesis is a process through which new blood vessels are formed from preexisting  vessels (Auerbach and Auerbach, 1994; Hanahan and Folkman, 1996; Risau, 1997). The human circulatory system is generally divided into (i) the macrovasculature, which consists of vessels with a diameter greater than 100 u.m and includes arteries and veins, and (ii) the microvasculature, which includes small diameter vessels such as arterioles, capillaries, and venules (Junqueira et al., 1992). Capillaries and postcapillary venules are responsible for the exchange of gases, macromolecules, waste products, and cells between blood and tissue  26  (Junqueira et al., 1992). Angiogenesis, however, initiates exclusively from postcapillary venules (Carmeliet, 2000). Blood vessels of all calibers are lined by endothelial cells, which undergo extensive functional modulations during angiogenesis (Auerbach and Auerbach, 1994; Hanahan andFolkman, 1996; Risau, 1997).  1.3.1  Mechanism of angiogenesis  Three types of angiogenesis can occur during the growth and remodeling of the vascular network. Intussusception, also known as non-sprouting angiogenesis, involves the splitting of a preexisting vessel into two daughter vessels (Carmeliet, 2000). Initiation of intussusception requires proliferation of endothelial cells within the vessel wall, resulting in widening of the vessel lumen (Risau, 1997). Subsequent formation of transcapillary pillars which divide the lumen in half, followed by invagination of the surrounding periendothelial cells and basement membrane, results in the generation of two functional vessels (Carmeliet, 2000). Another type of angiogenesis is bridging angiogenesis, in which an existing vessel is divided by transendothelial cell bridges into individual vessels (Carmeliet, 2000). The most common form of angiogenesis, however, is sprouting angiogenesis which involves the sprouting of new vessels from preexisting vessels (Carmeliet, 2000). The mechanism of sprouting angiogenesis has been well elucidated (Figure 1.4). Angiogenesis is controlled by a balance of activators and inhibitors (Hanahan and Folkman, 1996). When the concentration of activators exceeds that of inhibitors, angiogenesis is stimulated. In contrast, when inhibitors dominate over activators, angiogenesis is inhibited. Initiation of the angiogenic response occurs when an excess of angiogenic activators, produced by cells such as inflammatory cells, mast cells, or macrophages, diffuse into nearby tissues (Leek et al., 1994; Sunderkotter et al., 1994). These angiogenic factors bind to specific receptors on the  27  V ^  4  angiogenic activator  e n d o t h e l i a l cell  Figure 1.4: M e c h a n i s m of a n g i o g e n e s i s . Angiogenesis is controlled by a balance of activators and inhibitors. When the local concentration of activators exceeds that of inhibitors, angiogenesis is stimulated. Angiogenic activators bind to specific receptors on the surface of endothelial cells, resulting in endothelial cell activation. Activated endothelial cells produce proteases that degrade the local basement membrane, as well as nitric oxide that induces vascular permeability. A s a result of increased permeability, plasma proteins extravasate from the vessel into the tissue space and form a provisional matrix on which activated endothelial cells migrate. Migration is facilitated by the upregulation of cellE C M adhesion receptors belonging to the integrin family. Endothelial cells proximal to the migrating tip of the new sprout undergo proliferation, effectively increasing the length of the sprout. A s the sprout lengthens, additional proteases are secreted which degrade the tissue at the leading edge of the sprout. Maturation of the new sprout occurs following the recruitment of peri-endothelial cells.  28  surface of endothelial cells lining preexisting blood vessels, resulting in endothelial cell activation (Brooks, 1996). Activated endothelial cells in turn produce proteinases such as those belonging to the plasminogen activator and matrix metalloprotease (MMP) families, which degrade the local basement membrane thus releasing additional angiogenic factors sequestered within the extracellular matrix (ECM) (Coussens et al., 1999). Activated endothelial cells also produce nitric oxide which induces vasodilation of the existing vessel and an increase in vascular permeability (Carmeliet, 2000). In response to increased permeability, plasma proteins extravasate from the vessel into the tissue space, thus laying down a provisional matrix that will be used by migrating endothelial cells (Carmeliet, 2000). The provisional matrix is primarily composed of fibrin (Dvorak et al., 1995) and fibronectin (Clark et al., 1982). In addition to the production of proteases and nitric oxide, activated endothelial cells undergo a shape change and extend elongated processes such as pseudopodia into the surrounding tissue (Dvorak et al., 1995; Pepper, 1997). Furthermore, these cells upregulate the expression of adhesion molecules such as those belonging to the integrin family, which function to pull the endothelial cell forward toward the angiogenic stimulus (Bazoni et a l , 1999; Cockerill et al., 1995). Hence a rudimentary endothelial sprout migrating on the extravascular provisional matrix is formed. Endothelial cells located just proximal to the migrating tip of the sprout begin to proliferate, causing an increase in the length of the sprout (Auerbach and Auerbach, 1994; Hanahan and Folkman, 1996). As the sprout lengthens, additional proteases are produced which degrade tissue in front of the sprout (Stetler-Stevenson, 1999). Endothelial cells trailing the proliferative zone undergo another shape change, stop proliferating, adhere tightly to each other, and begin to form a lumen (Auerbach and Auerbach, 1994; Hanahan and Folkman, 1996). In addition, these endothelial cells deposit a new E C M composed primarily of laminin and collagen type IV (Carey, 1991; Iruela-Arispe et al.,  29  1991b; Risau and Lemmon, 1988). Both E C M proteins have been shown to promote endothelial cell morphogenesis (Carey, 1991; Iruela-Arispe et al., 1991b; Risau and Lemmon, 1988). Secondary sprouting from the migrating tip results in the formation of a capillary plexus, and the fusion of individual sprouts at their tips closes the loop and circulates blood into the vascularized area (Auerbach and Auerbach, 1994; Hanahan and Folkman, 1996; Risau, 1997). For maturation of new vessels to occur, peri-endothelial cells such as smooth muscle cells or pericytes must be recruited (Carmeliet, 2000). This process serves four functions: (i) to provide structural support to new vessels to prevent rupture or regression; (ii) to assist in the production of E C M proteins; (iii) to maintain endothelial cells in a quiescent state; and (iv) to provide contractile function to modulate vessel caliber (Benjamin et al., 1998; Carmeliet, 2000). Finally, pruning of the vasculature occurs in which vessels lacking peri-endothelial support undergo regression (Benjamin et al., 1998).  1.3.2  A n g i o g e n i c activators  Many growth factors and cytokines are known to stimulate angiogenesis. Two angiogenic activators used in this thesis are fibroblast growth factor-2 (FGF-2) and V E G F (Dvorak et al., 1995; Rak and Kerbel, 1997). Whereas FGF-2 exhibits pleiotropic effects and can stimulate numerous cells including endothelial cells and smooth muscle cells, V E G F is primarily an endothelial mitogen (Gospodarowicz et al., 1989; Klagsbrun and Moses, 1999; Leung et al., 1989). The F G F family consists of 23 members (Wiedlocha and Sorensen, 2004). FGF-2, also known as basic fibroblast growth factor, was the first angiogenic activator to be identified (Shing et al., 1984). FGF-2 is normally sequestered within the E C M through binding to heparan sulfate proteoglycans (Vlodavsky et al., 1991). Following digestion by heparinases, FGF-2 can be  30  released from the E C M and induce downstream signaling (Vlodavsky et al., 1991). Four FGFreceptor (FGF-R) family members have been identified: FGF-R1/2/3/4 (Jaye et al., 1992; Johnson and Williams, 1993). In addition, alternative splicing can result in the generation of FGF receptor variants (Jaye et al., 1992; Johnson and Williams, 1993). Disruption of either the FGFR l or the FGF-R2 gene in mice results in embryonic lethality (Deng et al., 1994; X u et al., 1998). In contrast, FGF-R3-null mice survive but exhibit skeletal abnormalities (Colvin et al., 1996). Gene disruptions for FGF-R4 have not been described (Cross and Claesson-Welsh, 2001). A l l four F G F receptors are receptor tyrosine kinases (Jaye et al., 1992; Johnson and Williams, 1993). Induction of angiogenesis by FGF-2 is mediated through FGF-R1 downstream signaling (Plotnikov et al., 1999). Indeed, FGF-R1 has been shown to be essential for the proper development and maintenance of the embryonic vasculature in mice (Lee et al., 2000). FGF-2null mice, however, exhibit normal embryonic vascular development (Dono et al., 1998). FGF-2 forms a dimer which binds and recruits two FGF-R1 receptor molecules, resulting in FGF-R1 dimerization and autophosphorylation (Klint and Claesson-Welsh, 1999; Plotnikov et al., 1999). Signaling pathways activated by FGF-R1 include the Ras, phosphatidylinositol 3-kinase (PI3K), phospholipase C (PLC), and Src family tyrosine kinase pathways (Cross and Claesson-Welsh, 2001). These pathways in turn induce endothelial cell proliferation, migration, and protease production (Christofori, 1996; Montesano et al., 1986). Moreover, FGF-2 also stimulates angiogenesis by inducing the recruitment of mesenchymal or inflammatory cells (Carmeliet, 2000). The V E G F cytokine family consists of seven members: V E G F - A , V E G F - B , V E G F - C , V E G F - D , V E G F - E , placenta growth factor-1 (P1GF1), and P1GF2 (Larrivee and Karsan, 2000; Migdal et al., 1998). V E G F - A , the first member to be discovered, was originally identified as  31  vascular permeability factor due to its ability to induce vascular leakage (Senger et al., 1983). VEGF-A also plays an essential role during vasculogenesis, the process of in situ differentiation of angioblasts into endothelial cells that assemble to form a primitive vessel network (Beck and D'Amore, 1997; Ray et al., 1999b; Risau and Flamme, 1995). Mice heterozygous for VEGF-A die at embryonic day (E)ll and exhibit growth retardation and abnormal blood vessel development (Carmeliet et al., 1996; Ferrara et al., 1996). Homozygous VEGF-A mutants exhibit more severe vascular abnormalities compared to heterozygous mutants (Carmeliet et al., 1996). Alternative splicing of the human VEGF-A gene can give rise to five isoforms: VEGFA121, VEGF-A145, VEGF-Ai , VEGF-Ai89, and V E G F - A 6 (Neufeld et al., 1996). Whereas the 65  20  former three isoforms are freely diffusible proteins, the latter two isoforms are primarily sequestered in the ECM (Neufeld et al., 1996). The VEGF-A isoform used to stimulate angiogenesis in this thesis is  VEGF-A165,  the most predominant (Cross and Claesson-Welsh,  2001) and biologically active (Neufeld et al., 1999; Neufeld et al., 1996) isoform of VEGF-A. VEGF ligands form disulfide-linked dimers that bind to members of the VEGF-receptor (VEGF-R) family (Cross and Claesson-Welsh, 2001). Five VEGF-R family members have been identified: (i) VEGF-R 1/fit-1; (ii) VEGF-R2/flk-1 /KDR; (iii) VEGF-R3/flt-4; (iv) neuropilin-1; and (v) neuropilin-2 (Larrivee and Karsan, 2000). VEGF-A can bind to VEGF-R 1, VEGF-R2, neuropilin-1, and neuropilin-2 (Larrivee and Karsan, 2000). VEGF receptors 1/2/3 all share a similar overall structure, which includes an N-terminal extracellular domain containing seven immunoglobulin (Ig)-like domains, a transmembrane domain, and a C-terminal intracellular domain with a two-part tyrosine kinase domain (Petrova et al., 1999). While the second Ig-like domain mediates ligand binding, both the first and third Ig-like domains are required for highaffinity binding (Barleon et al., 1997; Cunningham et al., 1997; Davis-Smyth et al., 1996; Fuh et  32  al., 1998). Following ligand binding, the fourth Ig-like domain mediates dimerization of VEGF receptors, resulting in receptor autophosphorylation and hence activation of downstream signaling pathways (Barleon et al., 1997). Mice deficient for VEGF-R2 die around E10, and are characterized by the absence of endothelial cells (Shalaby et al., 1995). Hence no organized blood vessels can be detected at any stage of embryonic development. VEGF-R1 -deficient mice also die at E10, and also exhibit vascular defects (Shalaby et al., 1995). However, unlike VEGF-R2 mutant mice, vascular defects in VEGF-R1 mutant mice arise not from a lack of vessels but from the formation of abnormally large and fused vessels (Shalaby et al., 1995). Indeed, VEGF-R1 mutant mice possess an overabundance of endothelial cells (Shalaby et al., 1995). This increased endothelial cell density has been postulated to prevent endothelial cell assembly into normal functional vessels. The fact that endothelial cells are absent in VEGF-R2- but not VEGF-Rl-null mice suggests that VEGFR2 plays a more important role than VEGF-R1 during mouse embryonic development. Indeed, almost all biologically-relevant VEGF signaling has been shown to be mediated by VEGF-R2 (Zachary and Gliki, 2001). VEGF-R2 activation can stimulate endothelial cell migration, proliferation, and survival (Zachary and Gliki, 2001). Signaling mediated by VEGF-R1, in contrast, has not been well elucidated (Zachary and Gliki, 2001). Instead, VEGF-R1 is thought to sequester VEGF and thus function as a negative regulator of VEGF-R2 (Zachary and Gliki, 2001).  1.3.3  Angiogenic inhibitors  Angiogenic inhibitors can be divided into two classes: (i) endogenous angiogenesis inhibitors and (ii) synthetic angiogenesis inhibitors. Among endogenous angiogenesis inhibitors, two of the best-studied are the tumor-derived inhibitors angiostatin and endostatin. Angiostatin is  33  a 38 kilodalton (kDa) fragment of the serine protease precursor plasminogen first isolated from the urine of mice bearing Lewis lung carcinoma tumors (O'Reilly et al., 1994). Intact plasminogen itself, in contrast to angiostatin, has no anti-angiogenic activity (Zetter, 1998). Angiostatin is not produced by tumor cells directly (Gately et al., 1996). Instead, tumor cells produce and secrete proteases into the circulation which cleave plasminogen to generate angiostatin (Gately et al., 1996). Hence in addition to inhibiting angiogenesis locally, angiostatin can function as a circulating angiogenesis inhibitor that suppresses angiogenesis at distant sites (O'Reilly et al., 1994). The mechanism of action of angiostatin involves interference with adenosine triphosphate (ATP) production (Moser et al., 1999). A T P is normally synthesized by the A T P synthase F l complex expressed in mitochondria (Alberts et al., 2002). Endothelial cells have been shown to express the ATP synthase F l complex on the cell surface, which functions to produce A T P that subsequently diffuses into the cell to facilitate numerous intracellular processes (Moser et al., 1999). Angiostatin can bind and inhibit the function of the A T P synthase F l complex, thus inhibiting endothelial cell proliferation and migration (Moser et al., 1999). Angiostatin can also increase tumor sensitivity to radiation (Gorski et al., 2003). Endostatin is a 20 kDa fragment of the E C M protein collagen XVIII, a collagen frequently found near blood vessels (O'Reilly et al., 1997). Intact collagen XVIII does not exhibit anti-angiogenic activity (Zetter, 1998). Initially purified from conditioned media of cultured hemangioendothelioma cells (O'Reilly et al., 1997), endostatin has been shown to inhibit endothelial cell function by attenuating growth factor signaling. Specifically, endostatin may downregulate the expression of angiogenic growth factor ligands or receptors, thus inhibiting endothelial cell proliferation (Tee and DiStefano, 2004). Moreover, endostatin can bind with  34  high-affinity to heparan sulfate proteoglycans on the surface of endothelial cells, thus acting as a competitive inhibitor of growth factor binding (Tee and DiStefano, 2004). In addition to angiostatin and endostatin, numerous other endogenous angiogenesis inhibitors have been identified. The E C M protein thrombospondin-1 has been shown to block endothelial cell proliferation, migration, and morphogenesis (Bagavandoss and Wilks, 1990; Iruela-Arispe et al., 1991a; Vogel et al., 1993). Tissue inhibitor of metalloproteases (TLMPs), which function to suppress E C M degradation by MMPs, can directly block endothelial cell proliferation and migration (Anand-Apte et al., 1997; Martin et al., 1996; Murphy et al., 1993). Several cytokines possess anti-angiogenic activity, including interleukin-12 (Hiscox and Jiang, 1997), leukemia inhibitory factor (Pepper et al., 1995), platelet factor 4 (Bikfalvi, 2004), and interferon-a/p (Lindner, 2002). A second class of anti-angiogenic factors consists of synthetic angiogenesis inhibitors. These inhibitors have been designed to block specific steps during the angiogenic cascade. V E G F signaling is a potent inducer of angiogenesis and hence is targeted by numerous synthetic angiogenesis inhibitors. Examples include V E G F antibodies (Abs) that bind to V E G F and prevent its association with V E G F receptors (Adamis et al., 1996; K i m et al., 1993; Presta et al., 1997), V E G F receptor Abs that prevent V E G F binding (Witte et al., 1998), soluble V E G F receptors that sequester V E G F and prevent endogenous receptor activation (Aiello et al., 1995; Goldman et al., 1998; Kendall et al., 1996), and V E G F receptor tyrosine kinase inhibitors that block receptor transphosphorylation (Fong et al., 1999). Synthetic M M P inhibitors that interact with the zinc-binding site of M M P s to block their function prevent MMP-mediated degradation of the E C M during endothelial sprouting (Sledge et al., 1995; Wojtowicz-Praga et al., 1998). Integrin antagonists such as Abs and inhibitor peptides block interaction of endothelial integrins  35  with E C M proteins (Brooks et al., 1994a; Brooks et al., 1994b; Gutheil et al., 2000). Because integrin ligation to E C M proteins induces cellular survival signals (Giancotti and Ruoslahti, 1999), integrin antagonists prevent these signals and thus induce endothelial cell apoptosis (Brooks et al., 1994b). Recent studies have demonstrated that some conventional chemotherapeutic drugs, when administered at low doses and on a more frequent schedule, can exhibit anti-angiogenic properties. Such drugs would therefore function not as anti-tumor agents but instead as antiendothelial agents (Browder et al., 2000; Klement et al., 2000). Examples of drugs that have demonstrated this property include adriamycin (Steiner, 1992), cyclophosphamide (Browder et al., 2000), paclitaxel (Belotti et al., 1996), and vinblastine (Vacca et al., 1999).  1.3.4  Pathological angiogenesis  Angiogenesis plays an important role in numerous physiological processes. These include embryonic development, wound healing, and development and repair of the female reproductive system (Cross and Claesson-Welsh, 2001; Klagsbrun and Moses, 1999). Because angiogenesis is controlled by a balance of angiogenic activators and inhibitors (Hanahan and Folkman, 1996), deregulated angiogenesis can result in the development of pathological disorders (Folkman, 1995; Isner, 1998). Diseases such as diabetic retinopathy, rheumatoid arthritis, and ischemia, as well as chronic inflammation, are all characterized by pathological angiogenesis. O f importance to this thesis is pathological angiogenesis in the context of tumor angiogenesis and metastasis.  1.3.4.1 T u m o r a n g i o g e n e s i s  Tumors with a diameter of 1-2 mm or smaller are avascular (Folkman, 1972). Without a blood supply, these tumors cannot grow larger due to rate-limiting diffusion of nutrients and  36  waste products and hence remain dormant (Folkman, 1971). Despite rapid tumor cell proliferation, a net change in tumor size does not occur due to a concomitant increase in tumor cell apoptosis (Hanahan and Folkman, 1996). In order to initiate blood vessel-dependent growth, avascular tumors engulf adjacent preexisting host organ blood vessels (Holash et al., 1999). Engulfed vessels allow avascular tumors to grow beyond a diameter of 1-2 mm (Denekamp, 1993). For continued tumor expansion to occur, however, tumor cells produce angiogenic activators that stimulate angiogenesis (Zetter, 1998). Host stromal cells recruited to the tumor such as macrophages, mast cells, and lymphocytes, also produce angiogenic activators that facilitate tumor growth (Freeman et al., 1995; Meininger and Zetter, 1992; Miguez et al., 1986). Typically more than one type of angiogenic activator is produced by a single tumor (Kerbel, 2000; Relf et al., 1997). Because the overall activity of newly-produced angiogenic activators exceeds that of locally-produced inhibitors, angiogenesis is activated and tumor growth occurs (Zetter, 1998). Angiogenesis-induced tumor growth does not occur as a result of increased tumor cell proliferation; instead, angiogenesis reduces the rate of tumor cell apoptosis, thus effecting a net expansion of the tumor mass (Holmgren et al., 1995). Tumor blood vessels are structurally and functionally different from normal blood vessels. Tumor blood vessels can be enlarged and contain abnormal tortuosities, blind ends, corkscrew structures, and atypical branching (Kerbel, 2000; Weidner et al., 1991). Functionally, tumor vessels are highly permeable due to the lack of proper basement membrane deposition and peri-endothelial cell recruitment (Dvorak et al., 1995). The mechanism by which recruitment of peri-endothelial cells is inhibited involves angiopoietin-2 (Ang-2), a secreted glycoprotein that binds to the tyrosine kinase with Ig and E G F homology domains-2 (Tie-2) receptor (Davis et al., 1996). Tie-2 is expressed on the surface of endothelial cells (Partanen and Dumont, 1999), and  37  normally signals to induce vessel maturation by recruiting peri-endothelial cells to newly-formed vessels (Dumont et al., 1995). Tumors have been shown to produce Ang-2, which binds to Tie-2 and antagonizes Tie-2 signaling, thus preventing tumor blood vessel maturation (Davis et al., 1996; Davis and Yancopoulos, 1999). Tumor blood vessels lack innervation and therefore exhibit abnormal responses to vasoactive stimuli (Mitchell et al., 1994). Blood flow within the tumor vasculature is chaotic (Vaupel et al., 1989). As a result of high tumor interstitial pressure, blood viscosity is increased within the tumor vasculature thus reducing the rate of blood flow (Vaupel et al., 1989). Intermittent flow can also occur, with prolonged periods of stasis or reverse flow (Jain, 1990; Vaupel et al., 1989). Arterio-venous shunting of blood has been reported in tumors (Vaupel et al., 1989). Tumor blood vessels of a mosaic composition have been described in which tumor vessels are lined by both endothelial and tumor cells (Maniotis et al., 1999). This feature of the tumor vasculature, termed vasculogenic mimicry, remains highly controversial (Folberg and Maniotis, 2004). In addition to providing a blood supply to the tumor, the tumor vasculature functions as a route for the spread, or metastasis, of cancer.  1.3.4.2 T u m o r m e t a s t a s i s  In general, cancers which do not metastasize do not kill (Miller and Sledge, 1999). Ninety percent of human cancer deaths are caused by metastases (Sporn, 1996). In order for metastases to be established, cancer cells must survive a series of steps (Poste and Fidler, 1980). Each step, however, is rate-limiting since deficiency at any step results in failure of the metastatic process (Poste and Fidler, 1980). Initiation of metastasis involves tumor cell dissemination at the primary tumor site. Tumor cells have been shown to modulate their cell-cell and cell-substrate interactions by modulating the expression and/or adhesive activity of cell adhesion proteins (Webb and Vande Woude, 2000). In addition, both tumor and infiltrating  38  stromal cells produce chemotactic and chemokinetic factors which enhance tumor cell migration and invasion (Levine et al., 1995; Negus and Balkwill, 1996). Some tumor cells secrete autocrine motility factor (AMF), which binds and activates the A M F receptor expressed on the same tumor cell thus promoting migration and invasion (Nabi et al., 1992). Stromal cells can produce hepatocyte growth factor (HGF), which binds to the Met receptor expressed on tumor cells to induce downstream signaling (Webb and Vaunde Woulde, 2000). H G F stimulation has been shown to induce tumor cell expression of both urokinase-type plasminogen activator (uPA) and its receptor uPAR (Jeffers et al., 1996). V E G F and FGF-2 can also induce uPA expression (Pepper et al., 1990; Pepper et al., 1991). The binding of uPA to uPAR induces the conversion of an inactive plasminogen zymogen into active plasmin, a broad specificity protease that degrades numerous E C M proteins including fibrin, fibronectin, gelatin, laminin, and vitronectin (Mignatti and Rifkin, 1996). Moreover, H G F can induce the expression of several M M P s (Hamasuna et al., 1999). Hence plasmin and M M P s degrade E C M proteins, effectively fragmenting the basement membrane surrounding tumor capillaries. Tumor cells adjacent to such capillaries enter the circulation via intravasation (Zetter, 1998). Both single tumor cells and tumor emboli have been observed to enter the circulation (Nime et al., 1977). When in circulation, tumor cell survival time is dependent upon the tumor cell's ability to avoid apoptosis and immune detection (Webb and Vande Woude, 2000). Tumor cells avoid apoptosis by inducing the expression of survival proteins, or by altering the expression of cell adhesion molecules belonging to the integrin family (Webb and Vande Woude, 2000). Mechanisms used by tumor cells to escape immune surveillance include masking of surface  antigens normally recognized by lymphocytes  (Marincola et al., 2000), as well as expression of Fas ligand which activates the Fas receptor on the surface of lymphocytes, resulting in lymphocyte apoptosis (O'Connell et al., 1996). Although  39  many tumor cells in circulation die, a significant proportion survive and attach to the endothelium of organ capillary beds at secondary sites (Zetter, 1998). Whereas initial tumor cellendothelial cell interactions are weak and are mediated by adhesion proteins of the selectin family (Tedder et al., 1995), strong interactions are mediated by integrins (Saiki et al., 1989). Tumor cell binding to the endothelium induces retraction of endothelial cells, thus exposing the sub-endothelial basement membrane (Nicolson, 1982). Tumor cells readily adhere to the subendothelial basement membrane (Abecassis et al., 1987) and therefore extravasate through the vessel wall (Zetter, 1998). Tumor cell extravasation is also facilitated by VEGF-induced disruption of the endothelial barrier in host organ blood vessels (Weis et al., 2004). Subsequent tumor cell proliferation at secondary sites results in the establishment of metastases.  1.3.5  Notch and blood vessel development  Since the start of this thesis project, numerous studies have linked Notch expression and downstream signaling with vascular development and angiogenesis. Mouse gene disruption studies have highlighted the importance of various Notch pathway elements in the development of the vasculature. Jaggedl-null mice are lethal at E10.5 and exhibit massive hemorrhaging associated with malformation of the vasculature (Xue et al., 1999). These mice also have defects in angiogenesis (Xue et al., 1999). Similarly, Dill-null mice die at E10.5 from hemorrhaging (Hrabe de Angelis et al., 1997). Mice homozygous for a hypomorphic Notch2 mutation exhibit disrupted vessel remodeling in multiple vascular beds (McCright et al., 2001). Presenilin-1-null mice are viable until birth, and suffer from severe brain hemorrhage (Shen et al., 1997; Wong et al., 1997). Presenilin-2-null mice are also viable but display only mild hemorrhage compared to presenilin-1-null mice (Herreman et al., 1999). Presenilin-l/presenilin-2 double homozygous deficient mice, however, die at E9.5 and exhibit a severe phenotype associated with delayed  40  vascularization and a complete lack of blood circulation (Herreman et al., 1999). This phenotype is similar, i f not identical, to that observed in Notchl-null mice (Herreman et al., 1999). Notchl gene disruption is lethal around E10 (Huppert et al., 2000) due to hemorrhaging, with extensive defects in angiogenic vascular remodeling observed in the embryo, yolk sac, and placenta (Krebs et al., 2000). Notch4 gene disruption, in contrast, results in viable mice with no apparent defects (Krebs et al., 2000). The lack of a mutant phenotype in these mice suggests that Notch4 is dispensable for normal embryonic development (Krebs et al., 2000). Interestingly, approximately 50% of Notch l/Notch4 double homozygous deficient mice exhibit more severe vascular defects than Notchl single null mice (Krebs et al., 2000). Hence a role for Notch4 in embryonic vascular development cannot be excluded. Interestingly, expression of activated Notch4 in the mouse embryonic vasculature, under the control of the VEGF-R2 promoter, results in vascular patterning defects (Uyttendaele et al., 2001). Therefore both increases and decreases in Notch4 signaling result in a common vascular phenotype, disrupted blood vessel development. The expression of various Notch ligand and receptor family members has been reported in endothelial cells. O f importance to this thesis is that Notch4 is primarily expressed on the endothelium (Uyttendaele et al., 1996). Endothelial-restricted Notch4 m R N A expression is first evident in mice during embryogenesis (Shirayoshi et al., 1997; Uyttendaele et al., 1996) and persists into adulthood (Favre et al., 2003; Johnson et al., 2001; Taichman et al., 2002). Examination of human tissues has revealed selective Notch4 m R N A expression in endothelial cells (Nijjar et al., 2001). B y in situ hybridization, Notch target genes of the H R T family have been shown to be highly expressed in blood vessels (Fischer and Gessler, 2003; Leimeister et al., 1999; Leimeister et al., 2000b; Nakagawa et al., 1999). In the murine embryonic vasculature, HRT1 m R N A exhibits an endothelial-restricted expression pattern (Fischer and Gessler, 2003).  41  HRT2 and HRT3 mRNAs, however, are predominantly expressed in the smooth muscle cell layer surrounding the endothelium, and to a lesser extent the endothelium itself (Fischer and Gessler, 2003). Interestingly, mRNA expression of HES family members is absent or barely detectable in the mouse embryonic vasculature (Kageyama et al., 2000; Sasai et al., 1992). Hence HRT genes, and not HES genes, may be the primary mediators of Notch signaling during vascular development. Recent studies have identified a role for Notch signaling in the determination of arterial versus venous fate of newly-formed blood vessels. In mouse embryos, m R N A expression of Dlll/3 and Notch2 cannot be detected in vessels (Villa et al., 2001). Jaggedl/2, D114, and Notchl/3/4 mRNAs, in contrast, are all specifically expressed in arteries and not veins (Villa et al., 2001). Interestingly, only D114 and Notch4 mRNAs have been shown to be expressed in capillaries (Villa et al., 2001). Arterial Jaggedl m R N A expression can be detected in both endothelial and smooth muscle cell types (Villa et al., 2001). Notch3, on the other hand, is not expressed in endothelial cells (Joutel et al., 2000). Instead, Notch3 protein has been shown to be expressed in vascular smooth muscle cells (Joutel et al., 2000). In the mouse umbilical cord, D114 mRNA expression can be detected in the umbilical artery but not in the adjacent umbilical vein (Shutter et al., 2000). During development of the mouse retina, D114 m R N A is expressed in an arterial-restricted pattern (Claxton and Fruttiger, 2004). Indeed, D114 is the first Notch ligand to be expressed in arteries (Krebs et al., 2000), and has since been shown to be required in a dosage-sensitive manner for normal arterial development in the mouse (Duarte et al., 2004; Gale et al., 2004). In zebrafish, mRNA expression of the Notch ligand DeltaC can be detected in endothelial cells prior to their acquisition of an arterial fate (Smithers et al., 2000). Endothelial cells that express gridlock (a zebrafish homologue of HRT2) m R N A incorporate exclusively into  42  arteries and not veins (Zhong et al., 2000). Accordingly, gridlock gene disruption adversely affects the formation of arteries, resulting in expansion of venous regions (Zhong et al., 2000). Hence Notch signaling may promote artery formation through repression of the venous fate. In a rat model of endothelial denudation, regenerating endothelial and smooth muscle cells in injured carotid arteries and aortae exhibit induced m R N A expression of Jagged 1/2 and Notch 1/2/3/4 (Lindner et al., 2001). Notch signaling may therefore play a role in the arterial vascular response to injury. It should be noted, however, that venous expression of Notch pathway elements has been reported in humans. For example, human umbilical vein endothelial cells (HUVECs) express Notchl/4 and D114 mRNAs (Liu et al., 2003) and D i l l protein (Han et al., 2000), and our laboratory has detected expression of Notch 1/2/3/4 and Jaggedl m R N A in H U V E C (Noseda et al., 2004a). Evidence of cross-talk between Notch and angiogenic growth factor signaling pathways has emerged. Human endothelial cells isolated from different vascular beds, when transduced with either FGF-2 or V E G F , exhibit m R N A induction of various Notch receptors and ligands (Liu et al., 2003). Interestingly, synergistic induction of Notch receptor and ligand m R N A expression in endothelial cells can be achieved by co-transduction of FGF-2 and V E G F (Liu et al., 2003). Stimulation of human arterial endothelial cells with soluble recombinant V E G F but not FGF-2 induces Notchl and D114 m R N A expression (Liu et al., 2003). Our laboratory has shown that Notch4IC can downregulate VEGF-R2 mRNA expression (MacKenzie et al., 2004a). HRT1 overexpression can also downregulate VEGF-R2 m R N A expression, thus attenuating the responsiveness of endothelial cells to V E G F (Henderson et al., 2001). Because the VEGF-R2 gene promoter contains E box binding motifs, activated Notch may act via HRT1 to repress VEGF-R2 expression (Henderson et al., 2001; MacKenzie et al., 2004a). V E G F signaling can  43  also be attenuated by HRT2-mediated passive repression of V E G F expression. Specifically, HRT2 has been shown to interfere with A R N T transcription factor binding to the V E G F promoter (Chin et al., 2000). Notch signaling plays a role in endothelial sprouting in vitro and angiogenesis in vivo. EHJVEC cultured in three-dimensional fibrin gels as endothelial tubes have been shown to upregulate D114, Notchl/4, and HRT1 mRNA expression (Nakatsu et al., 2003). Upregulation of HRT1 in endothelial tubes is postulated to play a role in the establishment of a mature endothelial cell network (Henderson et al., 2001). Low or absent HRT1 expression may be permissive to endothelial proliferation and migration whereas overexpression of HRT1 may block these processes, thus inhibiting the formation of new endothelial tubes (Henderson et al., 2001). Expression of activated Notchl or overexpression of HES1 has been shown to stabilize endothelial tube formation in Matrigel and induce endothelial cell cycle arrest in twodimensional cultures in vitro (Liu et al., 2003). Results from our laboratory demonstrate that Jaggedl-mediated Notch activation induces cell cycle arrest in primary endothelial cells (Noseda et al., 2004a). These studies suggest that activated Notch signaling inhibits endothelial sprouting in vitro. Accordingly, several studies have shown that inhibited Notch signaling promotes endothelial sprouting in vitro and angiogenesis in vivo. In a collagen gel assay, antisense oligonucleotides directed against Jaggedl enhance FGF-2-induced endothelial tube formation (Zimrin, 1996). Interestingly, H U V E C cultured on fibrin-coated plates and stimulated with FGF2 have been shown to upregulate expression of a Jaggedl m R N A corresponding to a soluble, non-transmembrane form of Jaggedl (Zimrin, 1996). Overexpression of a complementary D N A (cDNA) encoding this transcript has since been reported to antagonize Notch signaling (Small et al., 2001). In a chick chorioallantoic membrane ( C A M ) assay, soluble Jaggedl induces  44  angiogenesis in the chick C A M and hence functions as an angiogenic activator (Wong et al., 2000). Interestingly, expression of activated Notch4 in the rat brain endothelial cell line RBE4 has been shown to promote the formation of microvessel-like structures (Uyttendaele et al., 2000). Hence the effect of Notch signaling on angiogenesis may be dependent upon endothelial cell type. It should be noted, however, that RBE4 cells express the adenovirus E l A oncoprotein, which has been reported to mimic the effects of NotchIC by converting CBF1 from a transcriptional repressor to an activator (Ansieau et al., 2001).  1.4  CELL ADHESION  1.4.1  Overview of cell adhesion molecules  Cells use multiple molecular mechanisms to adhere to other cells as well as to the E C M . Proteins involved in cell-cell and cell-ECM adhesion can be divided into four families: Ig superfamily, selectins, integrins, and cadherins. The Ig superfamily consists of over 100 members that share a characteristic structural motif, the presence of one or more Ig-like domains commonly found in Abs (Alberts et al., 2002). These proteins mediate Ca -independent cell-cell adhesion (Alberts et al., 2002). Examples of Ig superfamily members include intercellular adhesion molecule-1 (ICAM-1), vascular cell adhesion molecule-1 (VCAM-1), lymphocyte function-associated protein-1 (LFA-1), and platelet endothelial cell adhesion molecule-1 (PECAM-1) (Alberts et al., 2002). Selectins are a family of cell-surface carbohydrate-binding proteins that mediate  Ca -dependent 2+  transient  adhesion between  blood cells and the  endothelium (Tedder et al., 1995). These interactions allow blood cells to extravasate from the bloodstream into tissues (Bevilacqua, 1993). Three members have been identified: lymphocyte (L)-selectin, platelet (P)-selectin, and endothelial (E)-selectin (Alberts et al., 2002). Similar to  45  selectins, cell-surface molecules belonging to the integrin and cadherin families mediate C a z+  dependent cell adhesion (Alberts et al., 2002). Studies presented in this thesis examine the role of integrins and cadherins in angiogenesis and breast cancer, respectively, and therefore these two adhesion protein families are described in detail below.  1.4.2  Integrins  1.4.2.1 O v e r v i e w o f i n t e g r i n f u n c t i o n s  Integrins constitute a large family of cell adhesion proteins expressed on virtually all cell types (Alberts et al., 2002). In addition to mediating cell-cell adhesion, integrins can mediate cell-ECM adhesion by functioning as transmembrane linkers, or integrators, that connect the cytoskeleton to the E C M (Alberts et al., 2002). Integrin ligation by E C M proteins results in the activation of intracellular signaling pathways that regulate numerous cellular processes including proliferation, migration, survival, and differentiation (LaFlamme and Auer, 1996). The phenotype of a cell is directly influenced by the repertoire of integrins expressed on the cell surface (Plow et al., 2000).  1.4.2.2 I n t e g r i n r e c e p t o r s  Integrins are heterodimeric receptors, each composed of an a and a p transmembrane glycoprotein subunit, that associate noncovalently to give rise to an asymmetric structure consisting of a single extracellular globular head and two flexible tails (Carrell et al., 1985; Hynes, 1992) (Figure 1.5). Thus far, 18 different a subunits and eight different p subunits have been identified, which associate to give rise to at least 24 distinct a/p heterodimers (Hynes, 2002) (Figure 1.6). Moreover, alternative splicing of integrin subunits has been documented thus contributing to integrin signaling versatility (Hynes, 1992; Sheppard, 2000). Both subunits are  46  Figure 1.5: Structure of an integrin a/p heterodimer. Each integrin receptor is composed of two transmembrane subunits, called a and p, that are held together by noncovalent bonds. Both a and p subunits consist of a large extracellular domain, transmembrane domain, and a short intracellular domain (with the exception of p4, which has a large intracellular domain). Alpha subunits contain four divalent cation binding sites. In some cases, the a subunit is initially synthesized as a single polypeptide chain, which is then cleaved to form a large extracellular domain and a smaller transmembrane/intracellular domain held together by a disulfide bond. Beta subunits contain a single divalent cation binding site, as well as four cysteine-rich repeats involved in intrasubunit disulfide bonding. Proximal to the divalent cation binding sites of both a and p subunits is a ligand binding domain.  47  Figure 1.6: Integrin <x/p heterodimer combinations. Integrin heterodimers are formed from one a and one (3 subunit. Thus far, 18 different a subunits and eight different p subunits have been identified, which associate to give rise to at least 24 distinct a/p heterodimers.  48  required for integrin expression on the cell surface (OToole et al., 1989). Alpha subunits range in size from 140-210 kDa and consist of a large extracellular domain followed by a single-pass transmembrane domain and a short intracellular domain (Hynes, 1992) (Figure 1.5). Within the extracellular domain, seven homologous repeating segments fold into a divalent cation binding motif containing four cation binding sites essential for proper integrin function (Tozer et al., 1996). In some cases, the a subunit is initially synthesized as a single polypeptide chain, which is then cleaved to form a large extracellular domain and a smaller transmembrane/intracellular domain held together by a disulfide bond (Alberts et al., 2002). Furin has been identified as an enzyme involved in the endoproteolysis of a subunits (Lehmann et al., 1996). Integrin p subunits typically range in size from 90-130 kDa and have an overall structure similar to that of a subunits (Hynes, 1992). One exception, however, is the p4 subunit which possesses an intracellular domain larger than that of the-other p subunits (Sonnenberg et al., 1991). The extracellular domain of all p subunits contains a single divalent cation binding motif and four cysteine-rich repeats involved in intrasubunit disulfide bonding (Calvete et al., 1989). Extracellular divalent cations are essential for the proper formation of integrin heterodimers (Tozer et al., 1996). Heterodimer formation does not involve the intracellular domains of a or p subunits but instead requires the extracellular domains of both subunits (Dana et al., 1991).  1.4.2.3 I n t e g r i n l i g a n d s  Integrins bind to a diverse set of proteins such as E C M proteins, other cell surface proteins, plasma proteins, and bacterial and viral proteins (Bouvard et al., 2001). Most integrins can recognize several E C M proteins, and individual E C M proteins can bind to several integrins (Hynes, 1987; Ruoslahti and Pierschbacher, 1987). The a and p subunits of an integrin heterodimer both contain several ligand binding sites located just proximal to the divalent cation  49  binding motif (Sanchez-Mateos et a l , 1996). These ligand binding sites are believed to act in a sequential manner. Specifically, one or several binding sites may mediate initial ligand binding which triggers a conformation change that exposes additional binding sites, thus resulting in stable integrin-ligand interactions (Hogg et al., 1994; Stuiver and O'Toole, 1995). Ligand binding is regulated by the presence of extracellular divalent cations (Tozer et al., 1996). The type of divalent cation, however, directly influences the strength and specificity of the interaction between integrin and ligandi (Alberts et al., 2002). Specificity is also mediated by amino acid recognition sequences presented on ligands. For example, the sequence arginineglycine-aspartic acid (RGD) can be found in the E C M proteins collagen, fibrinogen, fibronectin, and vitronectin, as well as vWF and thrombospondin (Plow et al., 2000). This R G D sequence is recognized by numerous integrins such as a3pi, a5pl, and ccvp3 (Plow et al., 2000). Integrin binding to ligands containing R G D sequences can be inhibited by treatment with R G D peptides (Andronati et al., 2004). The mere presence of an R G D sequence on a ligand, however, does not ensure recognition by corresponding integrin binding partners. Rather, the context in which an R G D sequence is presented determines whether productive integrin-ligand interactions will occur (Haas and Plow, 1994). These can include the three-dimensional presentation of the R G D sequence, residues flanking the R G D sequence, as well as features specific to the integrin such as the ligand binding motif (Haas and Plow, 1994). Additional examples of recognition sequences are glycine-phenylalanine-hydroxyproline-glycine-glutamic acid-arginine (GFOGER) expressed on collagen and recognized by a l p l and a2pl, aspartic acid-glycine-glutamic acid-alanine (DGEA) expressed on collagen and recognized by a2pl, isoleucine-aspartic acid-alanine-prolineserine (IDAPS) expressed on fibronectin and recognized by oc4pl, and glutamine-isoleucine-  50  aspartic a c i d - s e r i n e ( Q I D S ) expressed o n V C A M - 1 and r e c o g n i z e d b y cc4pl ( H y n e s , 1992; P l o w et a l . , 2000).  1.4.2.4 I n t e g r i n s i g n a l i n g I n response to E C M b i n d i n g , integrins transmit signals into the c e l l that activate v a r i o u s i n t r a c e l l u l a r s i g n a l i n g pathways. T h i s type o f s i g n a l t r a n s d u c t i o n , t e r m e d " o u t s i d e - i n s i g n a l i n g " , enables the E C M to d i r e c t l y i n f l u e n c e c e l l u l a r phenotype a n d b e h a v i o r ( G i a n c o t t i and R u o s l a h t i , 1999). S i g n a l i n g i s i n i t i a t e d through i n t e g r i n l i g a t i o n . B e c a u s e l i g a n d s a n d d i v a l e n t cations share a c o m m o n b i n d i n g p o c k e t o n the integrin g l o b u l a r head, l i g a n d b i n d i n g d i s p l a c e s divalent cations thus i n d u c i n g c o n f o r m a t i o n a l  changes i n i n t e g r i n structure ( D ' S o u z a  et a l . , 1994).  These  c o n f o r m a t i o n a l changes traverse the p l a s m a m e m b r a n e a n d m o d u l a t e the c o n f o r m a t i o n o f the i n t e g r i n i n t r a c e l l u l a r d o m a i n s ( D ' S o u z a et a l . , 1994), r e s u l t i n g i n i n t e g r i n c l u s t e r i n g w i t h i n the plane o f the p l a s m a m e m b r a n e ( G i a n c o t t i and R u o s l a h t i , 1999). Integrin a subunits have been s h o w n to b i n d to c a v e o l i n - 1 , a m e m b r a n e adapter p r o t e i n that mediates c l u s t e r i n g o f v a r i o u s transmembrane proteins w i t h i n the p l a s m a m e m b r a n e d u r i n g l i p i d raft f o r m a t i o n ( W a r y et a l . , 1996; W e i et a l . , 1999). C a v e o l i n - 1 m a y a i d i n i n t e g r i n c l u s t e r i n g s i n c e i n h i b i t i o n o f c a v e o l i n - 1 e x p r e s s i o n b l o c k s i n t e g r i n s i g n a l i n g ( W a r y et a l . , 1996; W e i et a l . , 1999). Integrin c y t o p l a s m i c tails are d e v o i d o f e n z y m a t i c a c t i v i t y a n d h e n c e integrins must associate w i t h adapter proteins f o r d o w n s t r e a m s i g n a l i n g t o o c c u r ( G i a n c o t t i a n d R u o s l a h t i , 1999). Integrin a 3 a n d a 6 subunits h a v e b e e n s h o w n to b i n d to C D 1 5 1 , a m e m b e r o f the tetraspanin f a m i l y c o n s i s t i n g o f proteins w i t h four transmembrane 2002).  CD151  m a y a c t as a transmembrane  linker between  d o m a i n s ( K a z a r o v et a l . ,  the i n t e g r i n  i n t r a c e l l u l a r s i g n a l i n g m o l e c u l e s i n v o l v e d i n regulating c e l l u l a r m o r p h o l o g y  a  subunit a n d  ( K a z a r o v et a l . ,  2002). R e c e n t l y , the i n t r a c e l l u l a r d o m a i n o f the integrin cdlb subunit has b e e n s h o w n to b i n d to  51  and regulate the signaling activity of protein phosphatase-1 (Vijayan et al., 2004). Following integrin clustering, integrin p-subunit cytoplasmic tails associate with the cytoskeletal proteins talin, a-actinin, vinculin, and paxillin (Lo and Chen, 1994). These proteins in turn induce the assembly of actin filaments, which reorganize into large stress fibres that cause more integrin clustering (Giancotti and Ruoslahti, 1999). The cytoskeletal proteins also recruit enzymes such as Src, protein kinase C (PKC), and focal adhesion kinase ( F A K ) (Lo and Chen, 1994). Specialized adhesion sites called focal adhesions, composed of integrins, cytoskeletal proteins, and kinases, are thus formed at sites of contact between the E C M and integrins (Lo and Chen, 1994). Intracellular signals originating from focal adhesions have been shown to promote numerous  cell processes including proliferation, migration, survival, and differentiation  (LaFlamme and Auer, 1996). In addition to "outside-in signaling", integrins can modulate their adhesive activity via "inside-out signaling", i.e. integrins are capable of bidirectional signaling (Giancotti and Ruoslahti, 1999).  1.4.2.5 R e g u l a t i o n o f i n t e g r i n - m e d i a t e d a d h e s i o n  Integrin-mediated adhesion is regulated by two processes, integrin affinity and integrin avidity (Sanchez-Mateos et al., 1996). Changes in integrin affinity and avidity are not mutually exclusive (Shattil et a l , 1998; Stewart and Hogg, 1996). Rather, cells can use both processes to modulate ligand binding (Shattil et al., 1998; Stewart and Hogg, 1996).  1.4.2.5.1 Integrin affinity Integrin affinity modulation refers to a change in the attraction of a single integrin for its ligand (Hughes and Pfaff, 1998). This change is rapid and reversible and involves a conformational change in the extracellular ligand binding site of the integrin (Hughes and Pfaff,  52  1998; Sanchez-Mateos et al., 1996). Integrins normally exist in a low-affinity, non-ligand binding state (Faull et al., 1994; Lollo et al., 1993). Following integrin activation, a transition from low to high-affinity occurs resulting in enhanced cell adhesion (Sanchez-Mateos et al., 1996). The integrin family can be divided into subfamilies based upon a common p subunit (Sanchez-Mateos et al., 1996). Increases in integrin affinity have been demonstrated in p 1/2/3/7 subfamilies (Altieri et al., 1988; Crowe et al., 1994; Faull et al., 1993; Shattil et a l , 1998). Numerous factors, acting from either inside or outside the cell, can induce activation of integrins. Modulation of integrin affinity from within the cell occurs via inside-out signaling. Although the exact mechanism of integrin affinity regulation by intracellular signaling pathways remains unknown, a model has been proposed. The a and p subunits of an integrin heterodimer each contain a conserved cytoplasmic membrane-proximal motif, which together form a hinge that locks the integrin in a low-affinity state (Hughes et al., 1996). Sequences C-terminal to the hinge motif also play a role in regulating integrin affinity (O'Toole et al., 1995; Williams et al., 1994). In response to intracellular signals from the mitogen-activated protein (MAP) kinase and PI3K pathways, intracellular proteins bind to the integrin cytoplasmic domains causing a change in the spatial relationship between the a and p subunit tails (Loftus and Liddington, 1997; Williams et al., 1994). This spatial rearrangement disrupts the membrane-proximal hinge (Loftus and Liddington, 1997; Williams et al., 1994), triggering a long-range conformational change that traverses the transmembrane domain to the extracellular region of the integrin thus exposing the ligand binding site (Du et al., 1993). Hence the integrin is activated and in a high-affinity state. Extracellular factors  that regulate integrin affinity include divalent cations and *  •  •  •  2+  2+  monoclonal Abs (mAbs). Three cations are involved in integrin affinity modulation: Ca , M g , and M n  2 +  (Sanchez-Mateos et al., 1996). The effect of C a  53  2+  on integrin affinity can be inhibitory  or stimulatory, depending on the specific integrin-ligand interaction (Longhurst and Jennings, 1998). In contrast, M g  2 +  and M n  2 +  both activate integrins, with M n  2 +  being the more potent  activator of the two cations (Sanchez-Mateos et al., 1996). A mechanism for cation-induced pi integrin activation has been proposed (Sanchez-Mateos et al., 1996). In the absence of extracellular divalent cations, pi integrins are in an inactive state that prevents ligand binding. Upon addition of C a , pi integrins undergo a transition from an inactive to a minimally-active 2+  state, but still cannot bind ligands. Upon further addition of M g , pi integrins are converted 2 +  from a minimally-active to a partially-active state. It is in this state that pi integrins can initiate low-affinity ligand binding. For maximum ligand binding to occur, the addition of M n removal of C a  2+  2 +  and the  must follow. These changes result in the conversion of pi integrins from a  partially-active state to a fully-active, high-affinity conformation. Similar to divalent cations, a group of mAbs called activating mAbs can induce integrin activation. Activating mAbs for p i , p2, and p3 integrin families have been described (Arroyo et al., 1993; Frelinger et al., 1991; Ortlepp et al., 1995). Several pi integrin activating mAbs bind to a conserved regulatory region located between two ligand binding motifs expressed on inactive pi integrins (Takada and Puzon, 1993). Because this regulatory region contains a structural motif that permits a sharp reversal of polypeptide chain orientation, activating mAb binding is proposed to induce a conformational change in the ligand binding motif that increases integrin affinity (Takada and Puzon, 1993).  1.4.2.5.2 Integrin avidity  Integrin avidity modulation refers to a change in the surface density of integrins that affects ligand binding (Hughes and Pfaff, 1998). Because changes in integrin distribution generally occur after ligand binding, integrin avidity is often described as a post-receptor  54  occupancy event (Sanchez-Mateos et al., 1996). Following ligand binding, integrin interaction with the cytoskeleton results in anchorage of the integrin at a given site, thus altering the integrin diffusion rate within the plasma membrane (Lub et al., 1995; Stewart and Hogg, 1996). Localization of additional integrins to these sites results in focal adhesion formation and integrin clustering (Lo and Chen, 1994). Hence integrin avidity is enhanced without a change in integrin affinity (Sanchez-Mateos et al., 1996). A mechanism of integrin clustering independent of ligand binding has been described. This mechanism involves inside-out signaling pathways that when activated, induce a redistribution of cytoskeletal proteins that interact with integrin cytoplasmic domains to effect integrin clustering (Sanchez-Mateos et al., 1996). Phorbol esters, for example, can activate PKC, which in turn induces actin cytoskeletal reorganization that increases pi and p2 integrin avidity, not affinity (Danilov and Juliano, 1989; Haverstick et al., 1992).  1.4.2.6 E n d o t h e l i a l i n t e g r i n s  Integrins play an important role in vascular development and angiogenesis (Brooks, 1996). Endothelial cells express avp3, avp5, a6p4, and pi integrins (ccl-6) (Bischoff, 1997; Hiran et al., 2003). Among these, avp3, avp5, and pi integrins have been shown to be required for angiogenesis (Bloch et al., 1997; Brooks, 1996; Eliceiri and Cheresh, 1999). Quiescent endothelial cells express low levels of ccvp3 (Eliceiri and Cheresh, 1998). In contrast, endothelial cells of sprouting blood vessels in vivo express upregulated avp3 (Brooks et al., 1994a; Brooks et al., 1994b). Both FGF-2 and VEGF can induce avp3 expression in endothelial cells (Brooks et al., 1994b). One function of avp3 during angiogenesis may be to promote endothelial cell survival. Function-blocking avp3 Abs induce apoptosis of endothelial cells within sprouting vessels in vivo (Brooks et al., 1994b). Furthermore, binding of avp3 to ligand increases expression of the anti-apoptotic protein Bcl-2, resulting in enhanced cell survival (Stromblad et  55  al., 1996). A second function of ocvp3 during angiogenesis may be to promote endothelial cell migration. Specifically, avp3 has been shown to bind (Brooks et al., 1996) and localize M M P - 2 to the endothelial cell surface (Varner et al., 1995). The role of ccvp5 during angiogenesis is less well known compared to that of avp3. Nonetheless, VEGF-induced angiogenesis, and to a lesser extent FGF-2-induced angiogenesis, may require avp5 integrins (Friedlander et al., 1995). Ligation of pi integrins to four different E C M proteins may modulate the angiogenic response. Specifically, pi integrins bind to (i) collagen type I, which may regulate endothelial cell adhesion and migration; (ii) fibronectin, which may induce endothelial expression of MMPs; (iii) collagen type IV, which may promote endothelial cell morphogenesis during lumen formation; and (iv) laminin, which may promote endothelial cell differentiation (Brooks, 1996).  1.4.3  Cadherins  1.4.3.1 O v e r v i e w o f t h e c a d h e r i n f a m i l y  Cadherins constitute a family of transmembrane proteins that primarily function to mediate cell-cell adhesion (Takeichi, 1991). A l l cadherin family members contain at least one cadherin repeat, a conserved sequence of approximately 110 amino acids that independently folds into an extracellular structural domain (Takeichi, 1990). Each cadherin repeat requires extracellular C a  2+  for proper folding, thus conferring rigidity upon the cadherin extracellular  domain (Koch et al., 1999) and protection from protease digestion (Grunwald et al., 1981). Cadherin family members can be divided into five subfamilies: (i) classical cadherins, which indirectly link the actin filaments of adjacent cells together in structures called adherens junctions; (ii) desmosomal cadherins, which indirectly link the intermediate filaments of adjacent cells together in structures called desmosomes; (iii) protocadherins, which mediate the formation  56  of synapses between neurons; (iv) Flamingo cadherins, which are seven-pass transmembrane proteins whose function in mammals remains largely unknown; and (v) cadherin-related proteins, which are unique cadherins that do not fall into the above defined subfamilies (Wheelock and Johnson, 2003). Classical cadherins can be further divided into type I and type II cadherins, based on the presence of an extracellular histidine-alanine-valine (HAV) tripeptide in type I cadherins and the lack thereof in type II cadherins (Wheelock and Johnson, 2003).  1.4.3.2 E - c a d h e r i n s t r u c t u r e a n d f u n c t i o n  Epithelial (E)-cadherin is a type I classical cadherin that primary functions to mediate cell-cell adhesion between epithelial cells (Wheelock and Johnson, 2003). E-cadherin expression is not restricted to epithelial cells; expression in non-epithelial cells such as erythroid precursors and stromal cells has been reported (Corn et al., 2000). The functional significance of this expression, however, is unknown (Corn et al., 2000). E-cadherin possesses a large N-terminal extracellular domain, a single-pass transmembrane domain, and a C-terminal intracellular domain (Wheelock and Johnson, 2003) (Figure 1.7). The extracellular domain contains five tandem cadherin repeats, each bridged by a Ca  cation that plays a role in proper protein  folding. The binding specificity of E-cadherin has been mapped to the first cadherin repeat (Overduin et al., 1995). This cadherin repeat contains the H A V tripeptide, which has been shown to be required for cadherin-mediated cell-cell adhesion (Overduin et al., 1995). Cadherin molecules can interact with each other both in cis and in trans (He et al., 2003). Two types of trans-interactions have been described. Homophilic interactions involve the binding of a cadherin molecule on one cell with an identical cadherin molecule on an adjacent cell (Jiang, 1996). Heterophilic interactions involve the binding of a cadherin molecule on one cell with a different  57  extracellular cadherin repeats  domain  transmembrane domain  intracellular domain  catenin binding domain  F i g u r e 1.7: Structure of E-cadherin. E-cadherin is composed of an N-terminal extracellular domain, a single-pass transmembrane domain, and a C-terminal intracellular domain. The extracellular domain contains five tandem copies of a cadherin repeat, each bridged by a C a cation that plays a role in proper protein folding. The intracellular domain contains a catenin binding domain. 2 +  58  cadherin molecule on an adjacent cell (Jiang, 1996). In the epithelium, strong cell-cell adhesion is mediated by E-cadherin homophilic interactions (Jiang and Mansel, 2000). For effective epithelial cell-cell adhesion to occur, the cytoplasmic domain of E-cadherin must be linked to the actin cytoskeleton (Alberts et al., 2002). E-cadherin cannot directly interact with the actin cytoskeleton; rather, this interaction is mediated by a group of cytoplasmic proteins called catenins (Ozawa et al., 1989) (Figure 1.8). Three catenins have been shown to bind the cytoplasmic domain of E-cadherin: p-catenin, y-catenin, and pl20-catenin (Ozawa et al., 1989; Reynolds et al., 1994). p- and y-catenin bind to the catenin-binding domain present in the E-cadherin cytoplasmic domain in a mutually exclusive manner, and hence can substitute for each other during adherens junction formation (Nieset et al., 1997; Wheelock et al., 2001). In contrast, pl20-catenin does not appear to play a structural role in adherens junctions (Thoreson et al., 2000). A fourth catenin, called a-catenin, does not directly interact with E-cadherin but instead binds to p- or y-catenin (Wheelock and Johnson, 2003). a-catenin in turn interacts with actin filaments either directly or indirectly (via the binding of actin-binding proteins such as aactinin or vinculin), thus linking the E-cadherin-catenin complex to the cytoskeleton (Wheelock and Johnson, 2003). In addition to playing a structural role in cell-cell adhesion, catenins participate in downstream signaling pathways that modulate cellular responses. Hence Ecadherin has been linked to the regulation of numerous cellular processes including cell proliferation (Gottardi et al., 2001; Kantak and Kramer, 1998; Sasaki et al., 2000; St Croix et al., 1998; Stockinger et al., 2001) and cell survival (Day et al., 1999; Kantak and Kramer, 1998; St Croix et al., 1996; St Croix and Kerbel, 1997).  59  Figure 1.8: The p-catenin signaling pathway, p-catenin is localized to two cellular pools, at the cell surface and in the cytosol. At the cell surface, p-catenin binds to E-cadherin. E-cadherin-bound p-catenin in turn binds to a-catenin, which interacts with actin filaments either directly or indirectly (via the binding of actin-binding proteins such as a-actinin). Hence adherens junctions are formed, which effectively link the E-cadherin cytoplasmic domain to the actin cytoskeleton. p-catenin that is not bound to E-cadherin is found in the cytosol. In the cytosol, the stability of p-catenin is regulated by the Wnt signaling pathway. In the absence of Wnt signaling, p-catenin associates with a macromolecular complex composed of GSK3p, axin, and APC. This macromolecular complex promotes the phosphorylation (P) of p-catenin, resulting in p-catenin ubiquitination and subsequent degradation by the proteasome. In contrast, when Wnt ligand is present, Wnt binds to the receptor molecules Frizzled and LRP to form a ternary complex. This ternary complex activates Dishevelled, which in turn inactivates GSK3p. Hence GSK3P is unable to phosphorylate p-catenin, resulting in p-catenin stabilization and accumulation in the cytosol. Cytosolic p catenin eventually translocates into the nucleus and binds to the transcription factor TCF/LEF-1, resulting in gene transcription.  60  1.4.3.3 E - c a d h e r i n d o w n s t r e a m s i g n a l i n g  Signaling pathways mediated by p-catenin have been extensively studied, p-catenin is localized to two cellular pools: (i) at the cell surface, in adherens junctions that mediate cell-cell adhesion; and (ii) in the cytosol (Nusse, 1997; Papkoff et al., 1996) (Figure 1.8). A n equilibrium exists between cell surface and cytosolic pools of p-catenin (Eger et al., 2000; Sanson et al., 1996; Zhu and Watt, 1999). In the cytosol, the stability of p-catenin is regulated by the winglesstype (Wnt) signaling pathway (Nelson and Nusse, 2004). The mammalian Wnt protein family consists of 19 secreted ligands that control cell proliferation, migration, and morphology (Nelson and Nusse, 2004). In the absence of Wnt signaling, p-catenin associates with a macromolecular complex composed of glycogen synthase kinase 3p (GSK3p) and adenomatous polyposis coli (APC), held together by the scaffold protein axin (Ikeda et al., 1998; Kishida et al., 1998; Nelson and Nusse, 2004). GSK3p is a serine/threonine kinase that, when phosphorylated and in an active state, directly binds to and induces phosphorylation of p-catenin (Aberle et al., 1997; Orford et al., 1997). This results in p-catenin ubiquitination and subsequent degradation by the proteasome (Aberle et al., 1997; Orford et al., 1997). GSK3p can also bind and phosphorylate the tumor suppressor protein A P C (Rubinfeld et al., 1996). Phosphorylated A P C in turn exhibits increased binding to p-catenin, resulting in augmented GSK3p-mediated phosphorylation and degradation of p-catenin (Rubinfeld et al., 1996). Hence p-catenin not associated with cadherins at the cell membrane is rapidly phosphorylated and targeted for degradation (Nelson and Nusse, 2004). In the presence of Wnt signaling, p-catenin phosphorylation and degradation is inhibited (Nelson and Nusse, 2004). This involves the binding of a secreted Wnt ligand to two receptor molecules, Frizzled and low density lipoprotein (LDL) receptor-related protein (LRP), resulting in a Wnt-Frizzled-LRP ternary complex (Nelson and Nusse, 2004). Through an unknown  61  mechanism, this ternary complex activates the downstream signaling protein Dishevelled, which in turn directly dephosphorylates and inactivates GSK3p (Nelson and Nusse, 2004). Dishevelled can also recruit GSK-binding protein (GBP), which effectively inactivates GSK3p (Hatsell et al., 2003). Hence GSK3p is unable to phosphorylate p-catenin, resulting in p-catenin stabilization and accumulation in the cytosol (Jiang and Mansel, 2000). p-catenin accumulation in the cytosol eventually leads to its nuclear translocation (Jiang and Mansel, 2000). Within the nucleus, p-catenin binds to the transcription factor T-cell factor/lymphoid enhancer factor-1 (TCF/LEF-1) (Cadigan and Nusse, 1997). Nuclear p-catenin can also bind to the transcriptional coactivator Pontin52, which may promote the interaction between p-catenin and TCF/LEF-1 (Bauer et al., 1998). The p-catenin-TCF/LEF-1 complex, via transactivation domains provided by p-catenin and a sequence-specific D N A binding domain provided by TCF/LEF-1, initiates transcription of downstream target genes (Hsu et al., 1998; van de Wetering et al., 1997). These include the cell proliferation genes c-myc (He et al., 1998) and cyclin-Dl (Lin et al., 2000), invasion-related genes uPAR (Mann et al., 1999) and MMP-7 (Roose et al., 1999), multidrug resistance gene-1 (MDR-1) (Yamada et al., 2000), E-cadherin transcriptional repressor protein Slug (Conacci-Sorrell et al., 2003), adhesion-related proteins ZO-1 (Mann et al., 1999) and conductin (Jho et al., 2002), components of the transcription factor complex activator protein-1 (AP-1) (Mann et al., 1999), and the transcription factor TCF/LEF-1 (resulting in a positive feedback loop) (Hovanes et al., 2001).  62  1.5  E P I T H E L I A L - T O - M E S E N C H Y M A L TRANSITION (EMT)  1.5.1  Physiological EMT  During embryonic development, a fertilized egg divides repeatedly to give rise to numerous epithelial cells which cohere to form epithelial organs (Alberts et al., 2002). In addition to epithelial cells, however, a second cell type with a distinct shape and organization can be recognized (Thiery, 2002). These cells, called mesenchymal cells, exhibit reduced adherence and are found loosely embedded in the E C M which facilitates migration in the extracellular environment (Thiery, 2002). Developmental studies have determined that epithelial cells can undergo a morphological conversion into mesenchymal cells, via a process called E M T (Hay, 1995). Following E M T , mesenchymal cells depart their site of origin and migrate to new microenvironments, where they differentiate into diverse cell types that participate in organ development (Savagner, 2001). Hence E M T is a conserved and fundamental process that governs cellular morphogenesis in multicellular organisms (Thiery, 2002). E M T plays an essential role during embryogenesis,  including development of the heart, peripheral nervous  system,  musculoskeletal system, and most craniofacial structures (Thiery, 2002). In the adult, a partial E M T is required during wound healing, in which keratinocytes exhibit a transient migratory phenotype that facilitates re-epithelialization at the injured site (Paladini et al., 1996). Numerous cytokines have been shown to induce E M T , including EGF, HGF, FGF, and transforming growth factor p (TGFp) (Savagner, 2001).  1.5.2  Pathological E M T  By virtue of its ability to increase cellular invasiveness, E M T functions as a mechanism for the dissemination of cancer cells during tumor metastasis. Because E-cadherin expression  63  plays a role in the maintenance of the differentiated epithelial phenotype (Thiery, 2002), Ecadherin functions as a tumor suppressor protein (Berx et al., 1998). During E M T , epithelial cells undergo a dedifferentiation program associated with loss of E-cadherin expression (Thiery, 2002). Indeed, a direct correlation exists between loss of E-cadherin expression and loss of the epithelial phenotype (Behrens et al., 1989). Well-differentiated and less-invasive tumors generally express high E-cadherin levels (Gupta et al., 1997; Oka et al., 1993). Partial/complete loss of E-cadherin expression, in contrast, correlates with loss of differentiation, increased invasiveness and metastasis, increased tumor grade, and poor prognosis (Cadigan and Nusse, 1997; Daniel and Reynolds, 1997; Miller et a l , 1999; Polakis et al., 1999).  1.5.2.1 Mechanisms of E-cadherin silencing Silencing of E-cadherin expression can occur via genetic, epigenetic, and posttranslational mechanisms. One genetic mechanism involves mutation of the E-cadherin gene, cadherin (CDH)-l (Berx et al., 1998). The types of mutations reported include missense mutations, splice site mutations, and truncation mutations (Berx et al., 1998). Thus far, 74 distinct mutation sites in CDH-1 have been identified in various cancer cell lines and tumors (Berx et al., 1998; Berx et al., 1995a). CDH-1 mutation frequently occurs in combination with a second genetic mechanism of E-cadherin silencing, gene deletion, resulting in complete loss of E-cadherin expression (Berx et al., 1998). A third genetic mechanism for E-cadherin silencing involves single nucleotide polymorphism (SNP) (Li et al., 2000). The CDH-1 promoter contains a SNP site located 160 bases upstream of the transcription start site (Li et al., 2000). At this SNP site, the wildtype CDH-1 allele contains a cytosine whereas the polymorphic allele contains an adenine (Li et al., 2000). Although one study has reported a 70% reduction in E-cadherin  64  transcription in vitro due to the polymorphism (Li et al., 2000), another study has failed to find a relationship between the polymorphism and reduced E-cadherin expression (Cheng et al., 2001). Two epigenetic mechanisms of E-cadherin silencing directly addressed in this thesis include transcriptional repressor protein expression (Hennig et al., 1996) and promoter hypermethylation (Herman and Baylin, 2003). Transcriptional repressors that inhibit E-cadherin expression include Snail, Slug, Smad-interacting protein 1 (SIPT; also known as ZEB2), deltacrystalling-box factor 1 (delta-EFl; also known as ZEB1), E2A, and Twist (Hajra and Fearon, 2002; Yang et al., 2004). These proteins mediate active repression of E-cadherin expression by binding to E box motifs (CAGGTG) within the CDH-1 promoter. Nuclear (3-catenin-TCF/LEF-l complexes have also been reported to bind and repress the CDH-1 promoter (Jamora et al., 2003). D N A methylation inhibits gene transcription by inducing chromatin condensation, which prevents transcription factor access to gene promoters (Herman and Baylin, 2003). Methylation can occur on cytosine residues at CpG dinucleotides (cytosine and guanine nucleotides linked together by a phosphate bond), which are distributed throughout the genome but exhibit frequent clustering in small stretches of D N A called CpG islands (Herman and Baylin, 2003). The CDH-1 gene contains a CpG island within its promoter which, when hypermethylated, results in reduced E-cadherin transcription (Berx et al., 1995b). O f importance to this thesis is that both transcriptional repressor protein expression and promoter hypermethylation induce reversible downregulation of E-cadherin expression (Conacci-Sorrell et al., 2002), and thus represent targets for reversal of E M T . E-cadherin silencing can also be mediated by post-translational mechanisms such as Ecadherin shedding (Katayama et al., 1994) and endocytosis (Fujita et al., 2002), which both effectively remove E-cadherin from the cell surface during normal protein turnover.  65  1.5.2.2 E-cadherin downregulation in breast cancer One of the best markers of E M T associated with breast cancer progression is the loss of E-cadherin expression (Vincent-Salomon and Thiery, 2003). Two major histological subtypes of breast cancer are ductal and lobular carcinomas (Vincent-Salomon and Thiery, 2003). Although studies suggest a common molecular origin for ductal and lobular tumors, they are morphologically distinct (Roylance et al., 1999). Whereas ductal tumor cells grow in clusters as sheets or cords, lobular tumor cells remain isolated or form narrow cords (Vincent-Salomon and Thiery, 2003). This difference in morphology can be explained by differences in total E-cadherin expression. In ductal tumors, approximately 50% express E-cadherin protein at normal levels with the remaining 50% expressing reduced levels (Gamallo et al., 1993; M o l l et al., 1993). In contrast, 85% of lobular tumors exhibit a complete loss of E-cadherin protein expression (De Leeuw et al., 1997). The remaining 15% of lobular tumors that maintain E-cadherin expression nonetheless exhibit abnormal localization of the protein (De Leeuw et al., 1997). Hence E cadherin downregulation can occur in both ductal and lobular breast cancer, although complete loss of E-cadherin is more common in lobular tumors. Mechanisms to explain the downregulation of E-cadherin expression in ductal and lobular breast cancers have been identified. Many genetic factors play a role in E-cadherin silencing in breast cancer. Polymorphism in the CDH-1 promoter does not affect E-cadherin expression in ductal or lobular breast tumors, and hence does not confer an increased risk of developing either tumor type (Cheng et al., 2001; Lei et al., 2002). Gene deletion at the CDH-1 locus occurs frequently in both ductal and lobular breast cancer (Berx et al., 1996; Cleton-Jansen et al., 2001). However, an association between gene deletion and decreased E-cadherin expression is evident in lobular (Berx et al., 1996; Cleton-Jansen et al., 2001) but not ductal (De  66  Leeuw et al., 1997) breast tumors. CDH-1 gene mutations are more common in lobular than ductal breast cancer. In lobular breast tumors, the frequency of CDH-1 gene mutation ranges from 10-56% (Lei et al., 2002). In contrast, ductal breast tumors exhibit CDH-1 mutation frequencies from 0-5% (Kashiwaba et al., 1995; Lei et al., 2002). Genetic factors may work in conjunction with epigenetic factors to modulate E-cadherin expression in breast cancer. Upregulated expression of the E-cadherin transcriptional repressor protein Snail is sufficient to repress E-cadherin transcription (Hajra et al., 2002) and induce E M T in vitro (Cano et al., 2000), and has been reported in high-grade human breast tumors (Blanco et al., 2002). Normal breast epithelia do not display CDH-1 promoter methylation (Graff et al., 1995). In contrast, 50% of ductal tumors (Graff et al., 2000; Nass et al., 2000) and between 41-77% of lobular tumors (Droufakou et al., 2001; Sarrio et al., 2003) exhibit CDH-1 promoter methylation. Overall, these studies demonstrate that both genetic and epigenetic factors  contribute to E-cadherin  downregulation in breast cancer. 1.5.3 Notch and EMT Notch target genes are expressed at sites of epithelial-mesenchymal cell-cell interactions during embryogenesis (Mitsiadis et al., 1995; Nakagawa et al., 1999). In an immortalized human keratinocyte cell line, TGFp stimulation has been shown to induce E M T with an associated downregulation of E-cadherin protein expression (Zavadil et al., 2004). Treatment with antisense HRT1 or functional inactivation of Jaggedl, however, inhibits TGFp-induced downregulation of E-cadherin expression (Zavadil et al., 2004). Although the latter study suggests that Notch signaling may be required for TGFp-induced E M T in keratinocytes, a direct role for activated Notch signaling in the induction of E M T has not been reported. Recent studies have identified a role for Notch signaling in a specialized type of E M T , endothelial-to-mesenchymal transition  67  (EndoMT). During murine cardiac development, cardiac endothelial cells (endocardial cells) must undergo a transient EndoMT (Eisenberg and Markwald, 1995). This gives rise to mesenchymal cells that participate in the formation of the endocardial cushion, a specialized embryonic tissue from which the heart valves and septa develop (Eisenberg and Markwald, 1995). In wildtype mice, EndoMT occurs at E9.5 and coincides with Notchl m R N A expression in both endocardial and mesenchymal cushion cells (Timmerman et al., 2004). In Notchl-null mice, however, endocardial cells remain tightly associated and mesenchymal cells within the cardiac cushion are noticeably absent (Timmerman et al., 2004). A similar phenotype is observed in CBFl-null mice (Timmerman et al., 2004). Hence activated Notch signaling may be required for EndoMT during cardiac development. During EndoMT, endothelial cells can acquire a migratory phenotype by downregulating the expression of the cell-cell adhesion protein vascular endothelial (VE)-cadherin (Frid et al., 2002). Both VE-cadherin and E-cadherin belong to the classical cadherin subfamily (Wheelock and Johnson, 2003). In porcine aortic endothelial cells, overexpression of activated Notchl downregulates VE-cadherin expression and induces morphological transformation (Timmerman et al., 2004). Notch signaling may indirectly downregulate VE-cadherin expression via induction of the transcriptional repressor protein Snail. B y in vitro luciferase assays, activated Notchl induces Snail promoter activity, and Snail expression inhibits VE-cadherin promoter activity (Tan et al., 2001; Timmerman et al., 2004). A n inverse correlation between Notch and V E cadherin expression has also been described during murine cardiac development in vivo. In wildtype mouse embryos between E8.5-9.0, VE-cadherin m R N A expression is evident throughout the endocardium (Timmerman et al., 2004). At E9.5, which corresponds to the time of initiation of EndoMT, regions of the endocardium displaying increased Notchl mRNA  68  expression also display increased Snail mRNA expression, with a corresponding decrease in V E cadherin m R N A expression (Timmerman et al., 2004). In contrast, Notchl - or CBFl-null mice at E9.5 exhibit reduced endocardial expression of Snail mRNA and maintained expression of V E cadherin m R N A (Timmerman et al., 2004). These results suggest that Notch signaling may be required for the downregulation of endocardial cell-cell adhesion during EndoMT. Research from our laboratory has demonstrated that Notch activation can induce EndoMT in human endothelial cells (Noseda et al., 2004b). Activated Notch4 induces phenotypic changes consistent with EndoMT such as the downregulation of VE-cadherin and the upregulation of various mesenchymal markers (Noseda et al., 2004b). Moreover, cells expressing activated Notch4 exhibit increased migration towards platelet-derived growth factor, a known chemotactic factor for mesenchymal cells (Noseda et al., 2004b). Expression of activated Notchl or Jaggedl also induces EndoMT (Noseda et al., 2004b). Combined with the fact that Jaggedl, Notchl, and Notch4 mRNAs are all expressed in the embryonic murine heart at sites of EndoMT during cardiac development (Noseda et al., 2004b), these results raise the possibility that Jaggedl-Notch interactions may induce EndoMT in vivo.  1.6  AIMS OF T H E STUDY  It is well known that members of the Notch family of transmembrane receptors play an important role in the regulation of cell fate decisions and differentiation (Artavanis-Tsakonas et al., 1999). At the start of this thesis, several gene disruption studies from mice suggested a role for Notch and its ligands in the regulation of embryonic vascular development (Hrabe de Angelis et al., 1997; Huppert et al., 2000; Xue et al., 1999). Because Notch4 is primarily expressed on the endothelium (Li et al., 1998a; Uyttendaele et al., 1996), we investigated whether enforced expression of activated Notch4 in endothelial cells could inhibit endothelial sprouting in vitro  69  and angiogenesis in vivo. During the course of our experiments, two separate studies reported that both increases and decreases in Notch signaling result in a common vascular phenotype, disrupted embryonic blood vessel development (Krebs et al., 2000; Uyttendaele et al., 2001). Our studies demonstrate that increased Notch signaling disrupts endothelial sprouting in vitro and angiogenesis in vivo, and more importantly, provide a mechanism with which to explain the common effects of activated and inhibited Notch signaling on angiogenesis. Recent studies from our laboratory and others have shown that activation of Notch signaling promotes EndoMT, a specialized type of E M T (Noseda et al., 2004b; Timmerman et al., 2004). Although Notch pathway elements are expressed at embryonic sites of epithelialmesenchymal cell-cell interactions (Mitsiadis et al., 1995; Nakagawa et al., 1999), a direct role for activated Notch signaling in the induction of E M T has not been reported. Given that Ecadherin downregulation directly correlates with E M T in human breast cancer (Thiery, 2002), and that primary human breast cancers express elevated levels of Notch receptors (Callahan and Egan, 2004; Imatani and Callahan, 2000; Parr et al., 2004; Pece et al., 2004; Weijzen et al., 2002), we investigated whether activated Notch signaling would inhibit E-cadherin expression in human breast epithelial cells. We also sought to determine whether inhibition of Notch signaling would re-induce E-cadherin expression in human breast cancer cells previously lacking Ecadherin expression. Together, these experiments would better our understanding of the role of Notch signaling in E M T and breast cancer development. Overall, two major hypotheses were tested in this thesis: (1) Expression of activated Notch4 in endothelial cells inhibits angiogenesis in part by promoting pi integrin-mediated adhesion; (2) Inhibition of Notch signaling attenuates breast tumor growth by reversing the mesenchymal phenotype.  70  Chapter 2 MATERIALS AND METHODS  2.1  TISSUE CULTURE  2.1.1  Cell culture  The human dermal microvascular endothelial cell line HMEC-1 (referred to hereafter as H M E C ) , transformed with simian virus 40 (SV40) large T antigen (Ades et al., 1992), was provided by the Centers for Disease Control and Prevention (Atlanta, GA). H M E C cell lines were cultured in M C D B 131 medium (Sigma, St. Louis, MO) supplemented with 10% heatinactivated fetal calf serum (FCS; HyClone, Logan, UT), 10 ng/ml E G F (Sigma), and 100 units/ml each of penicillin and streptomycin (Gibco, Gaithersburg, M D ) . The avian retroviral packaging cell line Q2bn (gift of K. McNagny, University of British Columbia, Vancouver, B C , Canada) was cultured in Dulbecco's modified Eagle's medium ( D M E M ; Sigma) supplemented with 10% non-heat-inactivated FCS, 2.5% non-heat-inactivated chicken serum (Sigma), 60 ng/ml conalbumin (Sigma), 50 uM (3-mercaptoethanol (Sigma), 2 m M glutamine (Gibco), and 100 units/ml each of penicillin and streptomycin. The human breast epithelial cell line MCF-10A (gift of P. Sorensen, University of British Columbia) was cultured in a 1:1 mixture of DMEM/F12 (Sigma) supplemented with 5% horse serum (Sigma), 2 m M glutamine, 20 ng/ml EGF, 100 ng/ml cholera toxin (Cedarlane, Hornby, ON, Canada), 10 ng/ml insulin (Sigma), 500 ng/ml hydrocortisone (Sigma), and 100 units/ml each of penicillin and streptomycin. Primary human breast epithelial cells were cultured in Epicult™-B medium (StemCell Technologies, Vancouver, B C , Canada) supplemented with 5% FCS. The human breast carcinoma cell lines MDA-MB-231 (gift of C D . Roskelley, University  71  of British Columbia) and T47D (gift of J.T. Emerman, University of British Columbia) and the mouse Lewis lung carcinoma cell line (purchased from American Type Culture Collection, Manassas, V A ) were cultured in D M E M supplemented with 10% heat-inactivated calf serum (HyClone), 2 m M glutamine, and 100 units/ml each of penicillin and streptomycin. A l l cells were maintained at 37°C in an atmosphere of 5% C O 2 . Cells were passaged every 3 days by rinsing in phosphate-buffered  saline (PBS) and incubating in 0.25% trypsin/1 m M  ethylenediaminetetraacetic acid (EDTA; Sigma) for approximately 3 min, followed by replating into tissue culture dishes in new medium.  2.1.2  Primary human breast epithelial cell isolation  Primary human breast epithelial cells were obtained in collaboration with Dr. C. Eaves of the Terry Fox Laboratories (British Columbia Cancer Research Centre, Vancouver, B C , Canada). Discarded tissues from normal premenopausal  women undergoing reduction  mammoplasty surgeries were enzymatically digested and epithelial cell-enriched fractions were frozen until use, at which time single-cell suspensions were prepared (Stingl et al., 2001). Cells were co-cultured with 1.2 x 10 X-irradiated NFH3T3 mouse fibroblasts for 1 day in Epicult™-B 6  medium supplemented with 5% FCS. Following preparation of a single-cell suspension, epithelial cell adhesion molecule (EpCAM)-positive breast epithelial cells were magnetically I  separated using the human E p C A M selection cocktail EasySep™ (StemCell Technologies), and cultured at a density of 2 x 10 cells/plate in 35 mm plates. These 35 mm plates were pre-coated 5  with 2-3 ml Vitrogen™ (67 ug/ml in PBS; Cohesion Technologies, Palo Alto, CA) for 1 hour at 37°C, and washed with PBS prior to use.  72  2.1.3  G e n e transfer  H M E C overexpressing the Notch4 intracellular domain (HMEC-Notch4IC) and H M E C L N C X cells were constructed by retroviral transduction of a C-terminal hemagglutinin (HA)tagged human Notch4 c D N A (amino acids 1476-2003) or the empty p L N C X vector control, respectively (Li et al., 1998a). For retroviral transduction, constructs were transiently transfected into the retroviral packaging cell line AmphoPhoenix using Fugene™ 6 Transfection Reagent (Roche Diagnostic Corporation, Indianapolis, IN). Retroviral supernatants were used to transduce H M E C , and stable H M E C lines were obtained by selection in 300 ng/ml G418 (Gibco). Polyclonal H M E C lines were used to avoid artifacts due to retroviral integration site. Chicken retroviral expression vectors were constructed by inserting C-terminal HA-tagged human Notch4IC c D N A into the avian retroviral vector C K (gift of N . Boudreau, University of California, San Francisco, C A and M . Bissell, University of California, Berkeley, CA). Both C K Notch4IC and the empty vector were transiently transfected into Q2bn cells using Fugene™ 6 Transfection Reagent to generate producer lines. Retroviral vectors containing an internal ribosomal entry site (IRES) and either yellow fluorescent protein (YFP) or green fluorescent protein (GFP) were used to facilitate sorting of transduced cells. The MIYNotchllC construct was generated by inserting into the retroviral vector murine stem cell virus (MSCV)-IRES-YFP (MIY, gift of R.K. Humphries, Terry Fox Laboratories) the c D N A encoding human Notchl intracellular domain (NotchlIC; amino acids 1758 to 2556; gift of S. Artavanis-Tsakonas, Harvard Medical School, Cambridge, M A ) . The MIYNotch4ICHA construct was generated by inserting into the retroviral vector M I Y the cDNA encoding human Notch4IC (amino acids 1476 to 2003) tagged with a C-terminal H A epitope. The MIGXNotch4HA construct was generated by inserting into the retroviral vector M S C V -  73  LRES-GFP (MIG) the c D N A encoding the entire extracellular domain of human Notch4 (XNotch4; amino acids 1 to 1443) tagged with a C-terminal H A epitope. The MIYE-cadherin construct was generated by inserting into the retroviral vector M I Y the c D N A encoding fulllength human E-cadherin (gift of B . M . Gumbiner, University of Virginia, Charlottesville, V A ) . Cells were transduced with empty vector control or vector containing c D N A inserts, and transduced cells were sorted based on Y F P or GFP expression using a fluorescence-activated cell sorter (FACS® 440; Becton Dickinson, San Jose, CA).  2.2  2.2.1  PROTEIN ANALYSIS  Immunoblotting  Cultured cells were lysed in 50 m M Tris (Sigma), 150 m M NaCl (Sigma), 2% Triton® X 100 (Fisher Scientific, Suwannee, GA), 10 ng/ml soybean trypsin inhibitor (Sigma), and 200 u M phenylmethylsulfonyl fluoride (Sigma). Tumor tissues were homogenized on ice in 50 m M Hepes (Sigma) buffer, pH 7.6, containing 2% Triton X-100, 5 m M E D T A , and fresh protease inhibitor cocktail (Sigma). Protein concentration was determined using the Dc™ protein assay (Bio-Rad Laboratories, Hercules, CA). Total protein lysates (50 ng) were analyzed by sodium dodecyl sulfate membranes  (SDS)- polyacrylamide gel electrophoresis,  (Bio-Rad  Laboratories),  and  developed  by  transferred enhanced  to nitrocellulose chemiluminescence  (PerkinElmer Life Science, Boston, M A ) . Membranes were probed with mouse mAbs to H A (Sigma, 1:2000 dilution), E-cadherin (BD Transduction Laboratories, Mississauga, ON, Canada, 1:2500 dilution), total p-catenin (BD Transduction Laboratories, 1:1000 dilution), active pcatenin (Upstate, Lake Placid, N Y , 1:500 dilution), and a-tubulin (Sigma, 1:10,000 dilution). Protein expression was quantitated by densitometry. For E-cadherin and total p-catenin  74  expression, data are presented as the mean ± standard error from (i) 16 M I G tumors and 13 MIGXNotch4HA tumors, and (ii) 12 M I Y tumors and 12 MIYE-cadherin tumors. For active |3catenin expression, data are presented as the mean ± standard error from 14 M I G tumors and 11 MIGXNotch4HA tumors. Concentrated supernatant from cultured cells was obtained by filtering 3 day conditioned medium through a 100,000 kDa molecular weight cut-off Ultrafiltration Membrane (Millipore, Billerica, M A ) .  2.2.2  Immunofluorescence microscopy  H M E C lines (5 x 10 ) were cultured on coverslips for 48 hours, fixed and permeabilized 4  in cold methanol for 5 min. Non-specific binding was blocked by incubating with PBS containing 5% goat serum (Sigma) and 0.1% Tween  20 (Fisher Scientific) for 30 min.  Following incubation with primary antibody (Ab) (rabbit anti-HA polyclonal Ab, Covance, Princeton, NJ, 1:100 dilution) for 60 min and secondary Ab (Texas Red™-conjugated goat antirabbit IgG, Molecular Probes, Eugene, OR, 1:200 dilution) for 30 min, coverslips were mounted on glass slides using an anti-fading solution (FluoroGuard™ Antifade Reagent, Bio-Rad Laboratories) containing 100 ng/ml Hoechst 33258 (Sigma) to stain the nuclei. MCF-10A cell lines (1 x 10 ) were cultured in chamber culture slides for 7 days. Primary 5  breast epithelial cells (1 xlO ) were cultured in chamber culture slides pre-coated with 5  Vitrogen™ (67 ug/ml in PBS) for 11 days. Cells were washed twice with PBS, fixed in 4% paraformaldehyde (Sigma) in PBS for 10 min, and blocked in PBS containing 10% goat serum for 10 min. Following incubation with mouse mAb to extracellular E-cadherin (10 ng/ml, Chemicon, Temecula, C A ) at room temperature for 60 min, chambers were washed twice with PBS and incubated with goat anti-mouse AlexaFluor® 594 (Molecular Probes, 1:100 dilution) at  75  room temperature in the dark for 30 min. Chambers were washed twice with PBS, nuclei counterstained with 4',6-diamidino-2-phenylindole (DAPI; Sigma) for 5 min, and slides mounted with 50% glycerol. Tumor cryosections (7 ^m thick) were immunostained for E-cadherin, p-catenin, and CD31. For E-cadherin immunofluorescence microscopy, tissues were hydrated in PBS, fixed in 100% ice-cold methanol for 2 min, rinsed in PBS, permeabilized in methanol/acetone (1:1 v/v) for 2 min, rinsed in PBS, blocked in PBS containing 5% goat serum for 45 min, and stained with mouse mAb to E-cadherin (T.500 dilution). For p-catenin immunofluorescence microscopy, tissues were hydrated in PBS, fixed in 4% paraformaldehyde in PBS for 15 min, rinsed in PBS, permeabilized in 0.1% Triton® X-100 for 10 min, rinsed in PBS, blocked in PBS containing 5% goat serum for 45 min, and stained with mouse mAb to p-catenin (1:100 dilution). For CD31 immunofluorescence microscopy, tissues were hydrated in PBS, fixed in 4% paraformaldehyde s  in PBS for 15 min, rinsed in PBS, blocked in PBS containing 5% goat serum for 45 min, and stained with rat mAb to mouse CD31 (BD Pharmingen, Mississauga, O N , Canada, 1:100 dilution) for 30 min. Following two washes with PBS, tumor sections were incubated with the fluorochrome-conjugated secondary Abs goat anti-mouse AlexaFluor® 594 (1:100 dilution) or goat anti-rat AlexaFluor® 594 (Molecular Probes, 1:100 dilution), nuclei counterstained with DAPI, and slides mounted with Vectashield® (Vector Laboratories, Burlingame, C A ) . Immunofluorescence was detected with an Axioplan™ II imaging microscope (Carl Zeiss Canada, Toronto, O N , Canada), and images were captured with a 1350EX digital camera (Qlmaging, Burnaby, B C , Canada). For quantitation of the percentage CD31 stained area, at least six random fields at 200x magnification were analyzed per tumor using Northern Eclipse™ software (Empix Imaging,  76  Mississauga, ON, Canada). 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 mm , 2  entire tumor sections were analyzed using Northern Eclipse™ software. For MDA-MJ3-231 tumors, data are expressed as the average percentage CD31 stained area ± standard error from 14 control tumors and 12 XNotch4 tumors, and the average number of vessels per mm ± standard errorfrom13 control tumors and 16 XNotch4 tumors. For Lewis lung carcinoma tumors, data are expressed as the average percentage CD31 stained area ± standard errorfromseven control tumors and six XNotch4 tumors, and the average number of vessels per mm ± standard error 2  from 14 control tumors and 16 XNotch4 tumors. For quantitation of percentage nuclear p-catenin staining, at least six random fields at 200x magnification were analyzed per tumor using Northern Eclipse™ software. Data are presented as the average percentage nuclear p-catenin staining (total number of nuclear pcatenin-positive cells / total number of cells) ± standard errorfromfour control tumors and four XNotch4 tumors.  2.2.3  Immunohistochemistry  CAMs treated with transfected Q2bn cell lines were harvested from E12 embryos, and processed for histological analysis. For hematoxylin and eosin (H&E) staining, CAMs were fixed in formalin overnight at room temperature, dehydrated, and embedded in paraffin. 6 nm sections were cut and stained with H&E. For immunohistological analysis, CAMs werefrozenin Tissue-Tek® optimal cutting temperature compound (Somagen Diagnostics, Edmonton, AB, Canada), and 10 um sections cut and fixed in acetone for 10 min. Sections were hydrated and incubated in 1.5% hydrogen peroxide solution for 5 min to quench endogenous peroxide activity. Non-specific binding was blocked by incubating in goat serum (1:20 dilution) for 20 min. For  77  vWF staining, sections were incubated with primary Ab ( D A K O , Mississauga, O N , Canada, 1:200 dilution), a biotinylated secondary Ab, followed by an avidin conjugate. For H A staining, sections were incubated with primary Ab (mouse anti-HA mAb, 1:500 dilution), secondary Ab (biotinylated goat anti-mouse IgG, D A K O , 1:200 dilution), followed by horseradish peroxidaseconjugated streptavidin (DAKO). Sections were developed with a diaminobenzidine-hydrogen peroxide reaction (Sigma), counterstained with hematoxylin, dehydrated, cleared, and mounted. Tumor cryosections (7 um thick) were stained with H & E or rabbit polyclonal Ab to H A (BAbCo, Richmond, C A ) . A biotinylated goat anti-rabbit IgG (Vector Laboratories, 1:100 dilution) followed by horseradish peroxidase-conjugated streptavidin (Vector Laboratories) were used, nuclei counterstained with hematoxylin, and slides mounted with Permount® (Fisher Scientific). Immunohistochemistry was examined using an Axioplan™ II imaging microscope and images captured using either a Coolpix™ 990 (Nikon, Tokyo, Japan) or a 1350EX digital camera.  2.2.4  Flow cytometry  Cells were detached by incubation in PBS based enzyme-free cell dissociation buffer (Invitrogen, Carlsbad, C A ) for 20 min at 37°C. To reduce nonspecific binding, cells were incubated in 10% heat-inactivated iron-supplemented calf serum (Sigma) in PBS for 30 min at 37°C. The following primary Abs were added to the cells and allowed to incubate at 37°C for 30 min: LM609 (Chemicon, 10 ug/ml), B44 (gift of J. A . Wilkins, University of Manitoba, Winnipeg, M B , Canada, 10 ng/ml), K20 ( A M A C Inc., Westbrook, M E , 10 ng/ml), E-cadherin (10 ng/ml). Mouse IgG2a Ab (Sigma, 10 ng/ml) was used as a control. Cells were washed twice in cold PBS, and secondary Ab (goat anti-mouse IgG- fluorescein isothiocyanate (FITC), Sigma,  78  1:64 dilution; goat anti-mouse AlexaFluor® 594, 1:100 dilution) was added and allowed to incubate an additional 60 min in the dark on ice. After washing cells twice in cold PBS, cells were fixed in 4% paraformaldehyde. Samples were run on an EPICS® ELITE-ESP flow cytometer (Beckman Coulter, Fullerton, CA), and data were analyzed with WinList™ version 2.0 (Verity Software House Inc., Topsham, M E ) or FCS Express™ V 2 (De Novo Software, Thornhill, ON, Canada). Histograms presented for H M E C lines are from one experiment and are representative of at least three independent experiments. Histograms for MCF-10A cells are from one experiment and are representative of two independent experiments. For the flow cytometric analysis of cell cycle distribution, H M E C lines were serum starved in 2% FCS in M C D B for 48 hours, and incubated for an additional 24 hours in medium alone (2% FCS in M C D B ) or medium supplemented with FGF-2 (R&D Systems, Minneapolis, M N , 15 ng/ml) or  VEGF165  (R&D Systems, 15 ng/ml). For the last 2 hours of incubation, cells  were incubated with 10 nM 5-bromo-2'-deoxyuridine (BrdU; Sigma) at 37°C. Cells were harvested by trypsinization and fixed in 70% ethanol at 4°C for 30 min. After washing in PBS, cells were incubated in 2 M HC1 for 30 min to denature D N A , followed by neutralization in serum-free medium. Cells were blocked and permeabilized in 0.5% Triton® X-100 plus 4% calf serum, then incubated with anti-BrdU-FITC conjugated Ab (BD Pharmingen) for 1 hour. Washed cells were stained with 1 ng/ml DAPI in PBS containing 0.5% Triton® X-100. Samples were run on an EPICS® ELITE-ESP flow cytometer, and data were analyzed with WinList™ version 2.0 software. Data are presented as the mean ± standard error of three independent experiments.  79  2.3  ANGIOGENESIS ASSAYS  2.3.1  Endothelial sprouting assay  Endothelial sprouting was assessed by a modification of the method of Nehls and Drenckhahn (Nehls and Drenckhahn, 1995). Briefly, microcarrier beads coated with gelatin (Cytodex™ 3, Sigma) or positively-charged, cross-linked dextran (Cytodex™ 2, Sigma) were seeded with H M E C lines. When the cells reached confluence on the beads, equal numbers of HMEC-coated beads were embedded in fibrin gels in 96-well plates. For preparation of fibrin gels, bovine fibrinogen (Sigma) was dissolved in PBS at a concentration of 2.5 mg/ml. Aprotinin (Sigma) was added at a concentration of 0.05 mg/ml and the solution filtered through a 0.22 um filter. Fibrinogen solution was supplemented with FGF-2 (15 ng/ml) or V E G F (15 ng/ml). As a control, fibrinogen solution without angiogenic factor was used. Following transfer of the fibrinogen solution to 96-well plates, HMEC-coated beads were added at an approximate density of 50 beads/well, and clotting was induced by the addition of thrombin (Sigma, 1.2 units/ml). After clotting was complete, gels were equilibrated with 2% FCS in M C D B at 37°C. Following 60 min of incubation, the overlying medium was changed for all wells. M C D B plus 2% FCS alone or containing FGF-2 (15 ng/ml) or V E G F (15 ng/ml) was added to the wells. After 3 days of incubation with daily medium changes, the number of capillary-like tubes formed was quantitated by counting the number of tube-like structures per microcarrier bead (sprouts/bead). Only sprouts greater than 150 nm in length and composed of at least 3 endothelial cells were counted. Data are presented as the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. For coating of Cytodex™ 2 beads with collagen, beads were resuspended in 1 mg/ml collagen type I (Sigma), allowed to dry overnight on petri dishes, and resuspended in PBS. For  80  coating of Cytodex™ 2 beads with Abs (IgG2a, 1:1000 dilution; 8A2, gift of J. Harlan, University of Washington, Seattle, W A , 1:1000 dilution; LM534, Chemicon, 1:1000 dilution), beads were incubated with Ab at 37°C for 2 hours, washed twice with PBS, and resuspended in PBS. After incubating the Ab-coated beads with cells for 3 days, the beads were placed in fibrin gels supplemented with the appropriate Ab at 1:1000 dilution.  2.3.2  Chick chorioallantoic membrane (CAM) assay  Fertilized White Leghorn chicken eggs (Gallus gallus domesticus) were incubated at 37°C under conditions of constant humidity. A l l chick eggs were handled according to institutional animal care procedures. On E6, the developing C A M was separated from the shell by opening a small circular window at the broad end of the egg above the air sac. The embryo was checked for normal development, the window sealed with Parafilm®, and the egg returned to the incubator for 2 more days. On E8, transfected Q2bn cell lines were trypsinized and washed in PBS, and 3 x 10 cells resuspended in 15 ^1 of D M E M supplemented with 30 ng/ml V E G F were 6  placed onto nylon meshes (pore size 250 ^m, Sefar America, Depew, N Y ) on the C A M . The cells distribute throughout the mesh and secrete control virus or virus containing Notch4IC. Meshes treated with vehicle alone (15 \A D M E M ) were used as negative controls, whereas , meshes treated with V E G F (30 ng/ml in 15 ul D M E M ) were used as positive controls. Eggs were resealed and returned to the incubator. On E l 2 , images of the C A M s were captured digitally using an Olympus™ SZX9 stereomicroscope (Olympus America, Melville, N Y ) equipped with a Spot™  RT digital  imaging system  (Diagnostic Instruments,  Sterling  Heights, MI).  Neovascularization was quantitated for each C A M by counting the number of vessels that entered the mesh area, and dividing by the perimeter of the mesh (vessels/mm). Northern Eclipse™ software was used for manual vessel counting and mesh perimeter measurements. Data  81  are presented as the mean ± standard error of three independent experiments each done in replicates of 4 to 6 eggs. Following photography, C A M s were harvested and processed for further studies. For C A M s treated with anti-integrin Abs, fertilized White Leghorn chicken eggs were prepared as described above. Mouse anti-avian pi integrin Abs T A S C (9D11; function-activating pl integrin Ab, gift of L. F. Reichardt, University of California, San Francisco, CA), V2E9 (nonfunction-modifying pi integrin Ab, Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), and W1B10 (function-blocking p l integrin Ab, Sigma) were prepared at 10 ng/ml in PBS supplemented with 30 ng/ml V E G F . On E8, 20 \A of each Ab preparation was loaded onto 2-ram gelatin sponges (Gelfoam®, Pharmacia Upjohn, Kalamazoo, MI), which were 3  then placed on the surface of the developing C A M . Sponges containing vehicle alone (20 \xl of PBS) were used as negative controls, whereas sponges containing 20 ul of 30 ng/ml V E G F in PBS were used as positive controls. C A M s were also treated with a function-blocking mouse anti-human avp3 A b LM609 (which cross-reacts with avian avp3 integrin) prepared at 10 ng/ml in PBS containing 30 ng/ml V E G F . LM609 has previously been shown to attenuate V E G F induced angiogenesis in the C A M (Friedlander et al., 1995), and thus serves as a positive control for angiogenesis inhibition. Eggs were resealed and returned to the incubator. On E10, digital images of the C A M s were captured and analyzed for neovascularization as described above. Data are presented as the mean ± standard error of two independent experiments each done in replicates of 3 to 5 eggs.  2.3.3  Proliferation a s s a y  Proliferation of endothelial cells in response to FGF-2 and V E G F was determined by neutral red uptake. Confluent plates of H M E C lines were serum starved in 2% FCS in M C D B  82  for 48 hours, and cells plated in 96-well plates at a density of 5 x 10 cells/well. After 4 hours 3  incubation to allow cells to bind, overlying medium was removed and the cells treated with 2% FCS in M C D B supplemented with FGF-2 (15 ng/ml) or V E G F (15 ng/ml). As a control, cells were treated with 2% FCS in M C D B alone. Cells were incubated for 0, 24, 48, and 72 hours with daily medium changes. After each time point, wells were emptied and incubated with 100 nl neutral red dye (0.0025% neutral red (Sigma) in M C D B supplemented with 2% FCS). Empty wells were also incubated with neutral red dye for background absorbance correction. After 4 hours of incubation, wells were aspirated and neutral red dye solubilized with 100 nl/well 1% acetic acid in 50% ethanol. Absorbance was determined at 570 nm. Data are presented as the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments.  2.3.4  Migration assay  The ability of endothelial cells to migrate towards FGF-2 or V E G F was measured using a Transwell™ filter assay (Corning Costar, Rochester, N Y ) . Polycarbonate filters (8.0 [iM pores) of the upper chamber were coated with 50 \A fibrinogen (2.5 mg/ml) or collagen type I (1 mg/ml) in PBS, and allowed to dry overnight. Confluent plates of H M E C cell lines were trypsinized, washed twice with 10 ng/ml soybean trypsin inhibitor, and resuspended in serum-free M C D B medium. 3.5 x 10 H M E C cells were placed in the upper chamber, and M C D B medium 4  supplemented with FGF-2 (15 ng/ml) or V E G F (15 ng/ml) was placed in the lower chamber. As a control, M C D B medium without added chemotactic factor was placed in the lower chamber. Following 16 hours of incubation at 37°C, filters were washed in PBS, fixed in 4% paraformaldehyde, and stained with 0.5% crystal violet (Sigma). After removing adherent cells from the upperside of the filter using a cotton swab, all cells that had migrated and adhered to the  83  underside of the filter were counted using an inverted microscope. Data are presented as the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments.  2.3.5 A d h e s i o n a s s a y  High binding 96-well plates (Corning Costar) were coated with 100 ul/well of the following E C M proteins at 20 ng/ml: fibrinogen, fibronectin (Sigma), collagen type I, collagen type IV (Sigma), and vitronectin (Sigma). Control wells were coated with poly-L-lysine (Sigma) at 20 ng/ml. After incubation for 1 hour at 37°C, all wells were aspirated and blocked with 4% B S A in PBS for 30 min at room temperature, followed by washing with PBS. Single cell suspensions were prepared by washing confluent cells once with PBS based enzyme-free cell dissociation buffer, and incubating the cells in the same buffer for 20 min at 37°C. Following resuspension in a mixture of P B S : D M E M (4:1 v/v), 100 nl of the cell suspension at 6 x 10  5  cells/ml was added to each well and incubated at 37°C for 20 min. Plates were then gently washed with PBS, fixed in 4% paraformaldehyde, and stained with 0.5% crystal violet. Following solubilization of dye in 1% SDS in PBS, absorbance was quantitated in an enzymelinked immunosorbent assay plate reader at 570 nm, with background absorbance subtracted at 630 nm. Data are presented as the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. For adhesion modulation studies, H M E C lines were incubated with Abs on ice for 20 min. Mouse IgG2a was used at 1:500 dilution, mouse monoclonal function-blocking anti-human pi integrin Ab P4C10 was used at 1:100, 1:250, and 1:500 dilutions (Gibco) or 2.5 ng/ml (Sigma), mouse monoclonal function-blocking anti-human avp3 integrin Ab LM609 was used at 10 ng/ml. and mouse monoclonal function-activating anti-human p i integrin Ab 8A2 was used at  84  1 ng/ml. For cells treated with both LM609 and P4C10, 10 ng/ml and 1:100 dilution were used, respectively. For adhesion studies using P4C10, cells were seeded into wells coated with collagen type I and/or collagen type IV. For adhesion studies using 8A2, cells were seeded into wells coated with collagen type I.  2.3.6  Ligand binding assay  The binding of soluble collagen type I to H M E C - L N C X and HMEC-Notch4IC cell lines was examined. H M E C lines were detached by incubation in PBS based enzyme-free cell dissociation buffer for 20 min at 37°C. 5 x 10 cells were incubated with FITC-conjugated 5  collagen type I (16.9 molecules of FITC per molecule of collagen, Molecular Probes) in a volume of 100 nl at the following concentrations (ng/ml): 0, 0.1, 1, 10, 100, 500, 1000. After 10 min binding at 37°C, cells were washed 3 times in 1 ml PBS, and fixed in 4% paraformaldehyde. Samples were run on an EPICS® ELITE-ESP flow cytometer, and data were analyzed with WinList™ version 2.0 software. In order to determine the number of molecules of FITC-collagen type I bound per endothelial cell, a fluorescence standard curve was constructed using j  Quantum™ 24 premixed microbeads (Bangs Laboratories Inc., Fishers, IN). Taking the molecular weight of FITC and collagen type I to be 390 Da and 300 kDa, respectively, and given an FITCxollagen ratio of 16.9, a curve of bound collagen (molecules) versus concentration of FITC-collagen type I conjugate (nM) was generated. Data are presented as the mean ± standard deviation of two independent experiments.  85  2.4  R I B O N U C L E I C ACID (RNA) A N A L Y S I S  2.4.1  R N A isolation  Total RNA isolation from cultured cells in vitro or tumor tissue in vivo was performed using TRIzol® Reagent (Invitrogen) or an RNeasy® kit (Qiagen, Mississauga, ON, Canada) according to manufacturer's recommendations.  2.4.2  Reverse transcriptase-polymerase chain reaction (RT-PCR)  First strand cDNA was synthesized using 50 \xl reactions containing 2.5 jig total RNA and 200 units of Superscript™ II reverse transcriptase (Invitrogen). Following ribonuclease (RNase) H treatment (2 units/reaction; Invitrogen), PCR was performed. Control reactions omitting reverse transcriptase were performed in each experiment. Primer sequences and annealing temperatures are described in Tables 2.1 and 2.2. Entire PCR samples were assessed in 1.5% tris acetic acid (TAE)-agarose gels containing ethidium bromide and quantitated by densitometry. Data are presented as the mean ± standard errorfromthree independent experiments. For semi-quantitative RT-PCR, individual reactions along with glyceraldehyde-3 phosphate dehydrogenase (GAPDH) control reactions were terminated after the designated number of PCR cycles. Control human cDNA was generated from pooled total RNA isolated from the following human cells: HMEC; vascular smooth muscle cells; cervical cancer cells, SiHa; colon cancer cells, WiDr; kidney epithelial cells, 293T. Control mouse cDNA was generated from pooled total RNA isolated from the following mouse cells: Lewis lung carcinoma cells; endothelial cells, SVEC4-10; fibroblasts, NLH3T3. Data are presented as the mean ± standard error from five control tumors and five XNotch4 tumors.  86  Table 2.1: PCR Primer Sequences Species . Specificity Primer Gene Identifier Human Mouse  Primer Sequences (5' to 3')  Reference  Forward:  CTATGATGAGGGGGATGCT  (Noseda et al., 2004a)  Forward:  AATGGAGACTCCTTCACCTGT  (Noseda et al., 2004b)  Reverse:  CGTCCATTCAGGCACTGG  (Noseda et al., 2004b)  Forward:  TGGGATGCCTGGCACA  This thesis This thesis  3  Jaggedl  A  X X  B C Jagged2  Delta-likel  Delta-like3  A  X X  B  X  Forward:  CAGGGCACGCGGTGT  C  X  Reverse:  CCGGCAGATGCAGGA  This thesis  Forward:  GAGGGAGGCCTCGTGGA  This thesis  TG GTTCTCTCAG AGTTAG C AGAG  This thesis  A  X X  B  X  Forward:  C  X  X  Reverse:  AGACCCGAAGTGCCTTTGTA  This thesis  X  X  Forward:  CGGATGCACTCAACAACCT  This thesis  Reverse:  GAAGATGGCAGGTAGCTCAA  This thesis  X  Reverse:  ATAGATGTCTCTGGGGAGATGA  This thesis  X  Forward:  GCATTGTTTACATTGCATCCTG  This thesis  Reverse:  GCAAACCCCAGCAAGAGAC  This thesis  Reverse:  GTAGCTCCTGCTTAATGCCAAA  Forward:  CACTGTGGGCGGGTCC  A B  X  C Delta-like4  A  X  B  X X  C Notchl  A  X  B  X  C Notch2  D  X  A  X  B Notch3  C  X  A  X  B Notch4  C  X  A  X  B HES1  C  X  A  X  B  GGCCACCTCTTCACTGCTTC  (Noseda et al., 2004b)  X  Reverse:  CCGGAACTTCTTGGTCTCCA  (Noseda et al., 2004b)  Forward:  AATCCCTGACTCCAGAACG  X  Forward:  AACTGGAGAGTCCAAGAAACG  (Noseda et al., 2004b)  X  Reverse:  TGGTAGACCAAGTCTGTGATGAT  (Noseda et al., 2004b)  Forward:  TGAGACGCTCGTCAGTTCTT  (Noseda et al., 2004a)  X  Forward:  CACCTTGGCCCCCTAAG  (Noseda et al., 2004b)  X  Reverse:  TGGAATGCAGTGAAGTGAGG  (Noseda et al., 2004b)  Forward:  TAGGGCTCCCCAGCTCTC  (Noseda et al., 2004a)  X  Forward:  CAAGCTCCCGTAGTCCTACTTC  (Noseda et al., 2004b)  X  Reverse:  GGCAGGTGCCCCCATT  (Noseda et al., 2004b)  Forward:  AGGCGGACATTCTGGAAATG  Reverse:  CGGTACTTCCCCAGCACACTT  (Muller et al., 2002) (Zheng et al., 2000) (Zheng et al., 2000)  X  This thesis  ( M u l l e r e t a l . , 2002)  X  Forward:  X  Reverse:  GCAAATTGGCCGTCAGGA  Forward:  CAGCCAGTGTCAACACGACAC  Reverse:  TCGTTCATGCACTCGCTGAA  This thesis (Fujita et al., 2003)  X X  X X  (Zine et al., 2001)  A  X  Forward:  CAGCACGTACACAGCCCTAA  B  X  Reverse:  ACCTGAGGCTTTGGATTCCT  (Herman e t a l . , 1996)  (Fujita et al., 2003)  C  X  Forward:  TTAGGTTAGAGGGTTATCGCGT  D  X  Reverse:  TAACTAAAAATTCACCTACCGAC  (Herman e t a l . , 1996)  Forward:  TAATTTTAG GTTAGAG G GTTATTGT  (Herman et al., 1996)  X  Reverse:  CACAACCAATCAACAACACA  (Herman e t a l . , 1996)  X  Forward:  GTTTAG TTTTGGGGAGGGGTT  X  Reverse:  ACTACTACTCCAAAAACCCATAACTAA  (Graff et al., 2000)  A  X  Forward:  AGATGCATATTCGGACCCAC  (Fujita et al., 2003)  B  X  Reverse:  CCTCATGTTTGTGCAGGAGA  (Fujita et al., 2003)  A  X  Forward:  AATCGGAAGCCTAACTACAGCGAG  (Okubo et al., 2001)  B  X  Reverse:  CCTTGGCCTCAGAGAGCTGG  ( O k u b o e t a l . , 2001)  X  Forward:  ACACCCCTGGCACAACAA  This thesis  X  Reverse:  GTGTCACTGCGCTGAAGGTA  This thesis  F G H  SIP1  GTTGTATTGGTTCGGCACCAT  Forward:  D  E  Snail  (Shou et al., 2001)  Reverse: X  C  F  Slug  This thesis (Shou et al., 2001)  GGAGAGGCTGCCAAGGTTTT  E E-cadherin  X  A B  X  87  (Graff e t a l . , 2000)  GAPDH  a  A  X  Forward:  GGACCTGACCTGCCGTCTAGAA  (Decary et al., 2002)  B  X  Reverse:  GGTGTCGCTGTTGAAGTCAGAG  (Decary et al., 2002)  C  X  Forward:  AATGTGTCCGTCGTGGATCT  (Liu et al., 1999)  D  X  Reverse:  CCCTGTTGCTGTAGCCGTAT  (Liu et al., 1999) (MacKenzie et al., 2004b) (MacKenzie et al., 2004b)  E  X  X  Forward:  CCCATCACCATCTTCCAG  F  X  x  Reverse:  ATGACCTTGCCCACAGCC  P r i m e r Identifier is a letter assigned to each individual primer. Various combinations of two primers (see Table 2.2 - Primer Set) are  used to amplify the target gene of interest.  Table 2.2: PCR Primer Sets and Conditions  Primer Set  Annealing Temperature (°C)  human J a g g e d l  A/C  53  507  mouse J a g g e d l  B/C  53  383  human Jagged2  A/C  53  550  mouse Jagged2  . B/C  58  550 448  Gene  a  3  -  Prod*uct|SJ|e|J^(*  human Delta-likel  A/C  55  mouse Delta-likel  B/C  55  409  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 N o t c h l  A/B  55  85  mouse N o t c h l  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  mouse Notch4  B/C  53  486  human H E S 1  A/B  55  103  mouse H E S 1  C/D  53  62  human and mouse H E S 1  E/F  53  307  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  human Snail  A/B  50  400  human SIP1  A/B  53  234  human G A P D H  A/B  53  142  mouse G A P D H  C/D  53  256  human and mouse G A P D H  E/F  53  446  ..  P r i m e r Set is a Primer Identifier pair (see Table 2.1 for individual Primer Identifiers) used to amplify the target gene of interest.  88  2.5  METHYLATION A S S A Y S  2.5.1  Methylation-specific P C R (MSP) G e n o m i c D N A w a s i s o l a t e d f r o m c u l t u r e d c e l l s o r t u m o r tissue u s i n g a DNeasy® T i s s u e  K i t ( Q i a g e n ) a c c o r d i n g to manufacturer's r e c o m m e n d a t i o n s . O n e m i c r o g r a m o f g e n o m i c D N A w a s b i s u l f i t e m o d i f i e d u s i n g a CpGenome™ D N A M o d i f i c a t i o n K i t ( C h e m i c o n ) a n d eluted i n 25 ul T r i s - E D T A b u f f e r , a c c o r d i n g to manufacturer's r e c o m m e n d a t i o n s . M S P w a s p e r f o r m e d u s i n g b i s u l f i t e - m o d i f i e d D N A ( - 1 2 0 n g ) , 5' a n d 3' p r i m e r s (each at 4 0 0 n M ) , 2 ' - d e o x y n u c l e o s i d e 5 ' triphosphates ( d N T P s ; Invitrogen) (each at 0.2 m M ) , I X P C R buffer, a n d 0.625 units o f HotStarTaq® D N A P o l y m e r a s e ( Q i a g e n ) , a c c o r d i n g to manufacturer's r e c o m m e n d a t i o n s . P r i m e r sequences a n d a n n e a l i n g temperatures are d e s c r i b e d i n T a b l e s 2.1 a n d 2 . 2 . E n t i r e M S P reactions w e r e assessed i n 2 % T A E - a g a r o s e gels c o n t a i n i n g e t h i d i u m b r o m i d e . B a n d s c o r r e s p o n d i n g to methylated  a n d unmethylated  P C R products w e r e quantitated b y densitometry.  D a t a are  expressed as a ratio o f m e t h y l a t e d o v e r u n m e t h y l a t e d P C R p r o d u c t s ( M / U ratio), a n d represent the m e a n ratio ± standard error from f i v e c o n t r o l t u m o r s a n d f i v e X N o t c h 4 tumors.  2.5.2  G e n o m i c bisulfite s e q u e n c i n g B i s u l f i t e - m o d i f i e d g e n o m i c D N A isolated from t u m o r tissue w a s a m p l i f i e d b y P C R u s i n g  primers  spanning  the E - c a d h e r i n  proximal  promoter.  Primer  sequences  a n d annealing  temperatures are d e s c r i b e d i n T a b l e s 2.1 a n d 2.2. P C R products w e r e p u r i f i e d , c l o n e d into the p D r i v e c l o n i n g v e c t o r ( Q i a g e n ) , a n d i n d i v i d u a l c l o n e s sequenced. F i v e c o n t r o l t u m o r s (a total o f 35 M I G clones) a n d three X N o t c h 4 t u m o r s ( a total o f 2 2 X N o t c h 4 c l o n e s ) w e r e a n a l y z e d , a n d data expressed as the percentage m e t h y l a t i o n p e r C p G site (total n u m b e r o f m e t h y l a t e d c l o n e s / total n u m b e r o f clones).  89  2.5.3  Global g e n o m i c deoxyribonucleic acid (DNA) methylation analysis  Genomic D N A (500 ng) was digested with the restriction enzymes Hpall (New England Biolabs, Pickering, O N , Canada), Mspl (New England Biolabs), or McrBC (New England Biolabs). Hpall is unable to digest CpG-methylated D N A , whereas its isoschizomer Mspl is not sensitive to CpG methylation and thus is a positive control for D N A digestion. McrBC digests CpG-methylated D N A only in the presence of guanosine triphosphate (GTP). For McrBC digestion, reactions were performed with GTP (McrBC-plus-GTP) or without GTP (McrBCminus-GTP). Entire reaction mixtures were assessed in 2% TAE-agarose gels containing ethidium bromide. Hpall, Mspl, and McrBC-plus-GTP digested D N A products between 2 and 7 kilobases were quantitated by densitometry. For McrBC-minus-GTP reactions, undigested D N A products corresponding to the uppermost D N A band were quantitated by densitometry. Data are expressed as a ratio of Hpall over Mspl digested D N A products (Hpall / Mspl densitometric ratio), or McrBC-plus-GTP  over McrBC-minus-GTP  D N A products  (+GTP / -GTP  densitometric ratio), and represent the mean ratio ± standard error from five control tumors and five XNotch4 tumors.  2.6  TUMORIGENICITY A S S A Y S  Female non-obese diabetic/severe combined immunodeficient (NOD/SCLD) mice were obtained from the Animal Resource Centre of the British Columbia Cancer Research Centre. For MDA-MB-231 tumor generation, 5 x 10 cells were injected subcutaneously into the dorsa of 6  mice. Once tumors were palpable, tumor volume (0.523 x length x width x height) was measured weekly using calipers. For each time point, MDA-MB-231 tumor data are presented as the mean tumor volume ± standard error from (i) 12 M I G tumors and 11 MIGXNotch4HA tumors, and (ii)  90  12 M I Y tumors and 12 MIYE-cadherin tumors. Tumor growth curves for one experiment are presented and are representative of three independent experiments. For Lewis lung carcinoma tumor generation, 2 x 10 cells were injected subcutaneously into the dorsa of mice. Once tumors 6  were palpable, tumor volume was measured every two days using calipers. For each time point, Lewis lung carcinoma tumor data are presented as the mean tumor volume ± standard error from six control tumors and seven XNotch4 tumors. Tumor growth curves for one experiment are presented and are representative of three independent experiments. Following sacrifice of mice, gross images of tumors attached to the dorsal skin flap were captured with an Olympus™ SZX9 stereomicroscope equipped with a Spot™ RT digital camera. Mice bearing MDA-MB-231 tumors (MIG and MIGXNotch4HA, or M I Y and MIYE-cadherin) were sacrificed at the same time and the total number and total weight (mg) of metastases determined. Data are presented as the average number of metastases per mouse ± standard error, and the average weight of each metastatic nodule (mg) ± standard error. Metastasis data were determined by analyzing (i) 25 M I G tumors and 27 MIGXNotch4HA tumors, and (ii) 12 M I Y tumors and 12 MIYE-cadherin tumors. Animal experiments were approved by the University of British Columbia Institutional Animal Care and Ethics Committee, and all animals were handled according to institutional animal care procedures.  2.7  STATISTICAL ANALYSIS  To determine statistical significance, a one-way analysis of variance with a Tukey test for multiple comparisons was used in all experiments. To assess a correlation between E-cadherin and p-catenin expression, Pearson correlation coefficients were calculated using the statistics program Statistical Package for Social Scientists version 11.0 (SPSS Inc., Chicago, IL). Statistical significance was taken at a P value of < 0.05.  91  Chapter 3  A C T I V A T E D N 0 T C H 4 INHIBITS A N G I O G E N E S I S : R O L E O F pi-INTEGRIN ACTIVATION  3.1  ABSTRACT  Angiogenesis involves modulation of the endothelial cell phenotype. Because activation of Notch in various cell systems has been shown to regulate cell fate decisions, and Notch4 is primarily expressed on endothelial cells, we postulated that Notch4 may be involved in regulating angiogenesis. To answer this question, we expressed the truncated, constitutivelyactive Notch4IC in endothelial cells. Our studies indicate that activated Notch4 inhibits endothelial sprouting in vitro and angiogenesis in the chick C A M in vivo. Activated Notch4 does not inhibit endothelial cell proliferation, nor does it inhibit endothelial cell migration through fibrinogen towards FGF-2 and V E G F . Migration through collagen, however, was inhibited. We demonstrate that the decreased sprouting of Notch4IC cells from collagen-coated beads is due in part to enhanced pl integrin-mediated adhesion to collagen. Although endothelial cells expressing Notch4IC do not show increased surface expression of p 1 integrins, we show that the pl integrins are in a high-affinity, active conformation. We also show that activation of pl integrins with function-activating pl integrin mAbs, independent of Notch4 expression, is sufficient to inhibit endothelial sprouting in vitro and angiogenesis in vivo. Thus, our results suggest that Notch4 activation in endothelial cells in vivo may inhibit angiogenesis in part by promoting pl integrin-mediated adhesion to the underlying matrix.  92  3.2  RESULTS  3.2.1  Constitutively-active Notch4 inhibits endothelial sprouting in vitro  We  generated H M E C lines expressing constitutively-active human Notch4IC. A  previously described endothelial sprouting assay which mimics the formation of capillary-like tubes in fibrin gels in vitro was used to evaluate the role of Notch4 in angiogenesis (Koblizek et al., 1998; Nehls and Drenckhahn, 1995). Microvascular endothelial cells are seeded as a confluent monolayer onto gelatin-coated beads, which are subsequently embedded in a fibrin gel. Following stimulation by angiogenic factors, endothelial cells migrate off the beads and into the fibrin matrix to form sprouts. Using this in vitro assay, we found that activated Notch4 inhibited spontaneous endothelial sprout formation on gelatin-coated beads, as well as sprouting in response to FGF-2 and V E G F (Figure 3.1A,B). Moreover, the sprouts that formed from Notch4IC cell lines were noted to be shorter than those derived from cells transduced with the empty vector. Notch4IC protein was expressed in HMEC-Notch4IC cells (Figure 3.1C). We typically achieved transduction efficiencies between 50-80%. As expected with polyclonal cell lines, HMEC-Notch4IC cells displayed heterogeneity in staining for Notch4IC protein by immunofluorescence microscopy (Figure 3.ID). The majority of the Notch4IC protein localized to the nucleus of HMEC-Notch4IC cells, which is typical of constitutively-active Notch proteins (Furriols and Bray, 2000). Because H M E C are a transformed endothelial cell line, we repeated the endothelial sprouting assay using primary H U V E C transduced with the Notch4IC construct or the empty vector. Similar to H M E C , activation of Notch4 in H U V E C inhibited endothelial sprouting (data not shown).  93  Figure 3.1: Activated Notch4 inhibits endothelial sprouting from gelatin-coated microcarrier beads in vitro. (A) Sprouting of H M E C - L N C X or HMEC-Notch4IC cells from gelatin-coated microcarrier beads embedded in fibrin gels supplemented with either F G F - 2 (15 ng/ml) or V E G F (15 ng/ml). Bars represent 100 um. Arrows indicate endothelial sprouts of sufficient length to be counted. (B) Endothelial sprout formation quantitated after 3 days incubation by counting the number of tube-like structures per microcarrier bead (sprouts/bead). Data shown are the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. *Control: P s 0.05. * F G F - 2 : P < 0.01. * V E G F : P < 0.001. (C) Immunoblot for expression of HA-tagged Notch4IC in H M E C lines. (D) Immunofluorescence of H M E C - L N C X and HMEC-Notch4!C cells stained with Hoechst 33258, as well as anti-HA primary A b and Texas Red-conjugated secondary A b to detect HA-tagged Notch4IC protein. Original magnification 40X.  94  3.2.2  Constitutively-active Notch4 inhibits angiogenesis in v i v o  To determine whether activation of Notch4 would inhibit angiogenesis in vivo, we used the C A M assay. The C A M functions as a respiratory structure for gas/nutrient exchange and undergoes intense  vascularization (Brooks et al., 1999), thus providing an excellent  microenvironment for assessing angiogenesis. Exogenous factors can be placed on the surface of the developing C A M in the presence or absence of a known angiogenic factor to assess antiangiogenic or pro-angiogenic activity, respectively. As previously described (Boudreau et al., 1997; Jiang et al., 2000), we generated avian retroviral packaging cell lines (Q2bn) transfected with the empty C K vector or CK-Notch4IC. On E8, these C K producer lines, in the presence of V E G F , were placed onto meshes on the chick C A M surface and incubated for an additional 4 days. The cells distribute throughout the mesh and secrete control virus or virus containing Notch4IC which infect the surrounding proliferating cells, the majority of which are endothelial. C A M s transduced with the empty vector demonstrated normal angiogenesis in response to V E G F , whereas angiogenesis was markedly inhibited by the expression of Notch4IC (Figure 3.2A,B). Expression of Notch4IC protein was detected in transfected Q2bn cells (Figure 3.2C). Histological analysis was performed on sections of harvested C A M s . For H & E stained sections, areas of the C A M s proximal to the Q2bn-containing mesh were analyzed. H & E staining of C K vector-transduced C A M s revealed the presence of numerous blood vessels in the subchorionic mesenchyme (Figure 3.3A). In contrast, C A M s transduced with CK-Notch4IC exhibited  a marked reduction in blood vessels  close to  the  mesh  (Figure 3.3B).  Immunohistochemistry was also performed on C A M sections, and areas proximal to the mesh examined. Staining for the endothelial-specific marker vWF (Sehested and Hou-Jensen, 1981) confirmed the presence of blood vessels in C K vector-transduced C A M s (Figure 3.3C).  95  control  VEGF  Q2bnQ2bn-CK Notch4IC + VEGF + VEGF  Figure 3.2: Activated Notch4 inhibits angiogenesis in the chick CAM in vivo. The avian retroviral packaging cell line Q2bn was transfected with the empty vector CK (Q2bn-CK) or CKNotch4IC (Q2bn-Notch4IC). On day 9, transfected Q2bn cell lines were placed onto nylon meshes on the CAM surface in the presence of VEGF (30 ng/ml). The grafted cells distribute throughout the mesh and secrete control virus or virus containing Notch4IC. Control CAMs were treated with medium alone or medium supplemented with VEGF. Images of the CAMs were captured on day 12. (A) CAMs treated with Q2bn-CK or Q2bn-Notch4IC cell lines in the presence of VEGF. Arrows indicate the edges of the nylon mesh. Bars represent 1 mm. (B) Vascular density quantitated after 4 days incubation by counting the number of vessels that entered the mesh area and dividing by the perimeter of the mesh (vessels/mm). Data shown are the mean ± standard error of three experiments each done in replicates of 4-6 eggs. *CK versus Notch4IC: P < 0.00001. (C) Immunoblot for expression of HA-tagged Notch4IC in Q2bn cell lines.  96  Q2bn-CK  Q2bn-Notch4IC  B sz 03 03 E  03  D  E  X o  JZ 03 © E E c  H  f*  c  "So  > C  H3  Figure 3.3: Immunohistochemical analysis of Notch4IC expression in the CAM. Q2bn packaging cells transfected with the vector control (Q2bn-CK) or Notch4IC (Q2bn-Notch4IC) were placed onto nylon meshes on the C A M surface. Treated C A M s were harvested on day 12 and sections prepared. (A-D) C A M sections proximal to mesh. H&E staining of Q 2 b n - C K treated C A M (A) or Q2bn-Notch4IC treated C A M (B). Anti-vWF staining of Q 2 b n - C K treated C A M (C) or Q2bn-Notch4IC treated C A M (D). (E-H) C A M sections distant from mesh. Anti-HA staining of Q 2 b n - C K treated C A M (E) or Q2bn-Notch4IC treated C A M (F). AntivWF staining of Q 2 b n - C K treated C A M (G) or Q2bn-Notch4IC treated C A M (H). Original magnifications: 40X (A,B,C,D,F,H) and 63X (E,G).  97  Notch4IC-transduced CAMs, in contrast, showed minimal staining for vWF (Figure 3.3D), confirming inhibition of blood vessel formation. To assess expression of the HA-tagged Notch4IC protein in endothelial cells, serial sections were stained with Abs against HA and vWF. Because Notch4IC-transduced CAMs were nearly devoid of small vessels proximal to the mesh, colocalization of staining was examined in vessels distant from the mesh. As expected for CK vector-transduced CAMs, vessels distant from the mesh did not stain for HA (Figure 3.3E) but did for vWF (Figure 3.3G). Analysis of Notch4IC-transduced CAMs demonstrated that vessels distant from the mesh exhibited costaining for HA (Figure 3.3F) and vWF (Figure 3.3H). Our findings suggest that expression of Notch4IC in vessels that feed the area of the mesh inhibits VEGF-induced endothelial sprouting and angiogenesis. In order to elucidate a possible mechanism(s) by which activated Notch4 inhibits endothelial sprouting in vitro and angiogenesis in vivo, we investigated the effects of Notch4IC expression on endothelial cell functions related to the angiogenic process using various in vitro assays.  3.2.3  N o t c h 4 inhibition of endothelial s p r o u t i n g in vitro c a n n o t be explained by r e d u c e d endothelial cell proliferation  Endothelial cell proliferation enables newly formed sprouts to increase in length and extend into the surrounding matrix. To determine whether reduced proliferation was a possible reason for the decreased sprouting of Notch4IC endothelial cells, we performed neutral red proliferation assays. When plated on normal tissue culture substrata, Notch4IC cells and control cells exhibited similar proliferation rates over 72 hours (the incubation time for the endothelial sprouting assay) (Figure 3.4A). hi fact, proliferation rates were the same when cells were grown in serum-containing medium, or medium supplemented with FGF-2 or VEGF. Proliferation on  98  A  0.12  o  0.10  OHMEC-LNCX • HMEC-NotCh4ICl  TP  m • 0.08 O  f  c 0.06 o  B  §  0.04  r  2 0.02  CL  0.00  0  Control  24  48  72  VEGF  FGF-2 0  24  48  72  0  24  48  72  Incubation time (hr)  B  HMEC-LNCX  -  D  B  6  DAPI Go/G,  1023  DAPI  H MEC-LNCX G /M S 2  HMEC-Notch4IC G /M Go/G, S  Control 44 ± 1 % 29 ± 4 % 25 ± 3% 36 ± 4 % 30 1 4 % FGF-2 37 ± 4 % 34 ± 2 % 24 ± 1 % 35 ± 3% 30 ± 5% VEGF 41 ± 1 % 33 ± 2% 26 ± 5% 38 ± 6% 32 ± 3%  2  30 ± 7% 33 ± 3% 27 ± 6%  Figure 3.4: Activated Notch4 does not inhibit HMEC proliferation. (A) Neutral red assay for proliferation. HMEC-LNCX and HMEC-Notch4IC cell proliferation was assayed over 72 hours in medium alone, or medium supplemented with FGF-2 (15 ng/ml) or VEGF (15 ng/ml). Data shown are the mean absorbances from a single experiment done in triplicate and are representative of at least three independent experiments. (B) Cell cycle distribution for HMEC-LNCX and HMEC-Notch4IC cells stimulated with VEGF. Cells were stained with DAPI for total DNA content and pulse-labeled with BrdU to detect DNA synthesis, and analyzed by flow cytometry. (C) Cell cycle distributions for HMEC-LNCX and HMEC-Notch4IC cells cultured in medium alone or medium supplemented with FGF-2 (15 ng/ml) or VEGF (15 ng/ml). Data shown are the mean + standard error of three independent experiments.  99  fibrinogen- and collagen-coated surfaces was also investigated, and was found to be equivalent for Notch4IC cells and control cells (data not shown). To confirm that Notch4IC does not affect H M E C cell cycle kinetics, we performed flow cytometry on H M E C lines pulse-labeled with BrdU and co-stained with DAPI. Both control and Notch4IC cells exhibited similar cell cycle distributions and BrdU incorporation in the absence or presence of growth factor (see Figure 3.4B for representative samples of VEGF-stimulated H M E C - L N C X and HMEC-Notch4IC, and Figure 3.4C for distribution percentages). Overall, the proliferation studies performed demonstrate that control and Notch4IC cells proliferate at similar rates. Hence the inhibition of HMEC-Notch4IC cell sprouting in vitro cannot be explained by a decrease in Notch4IC cell proliferation.  3.2.4  N o t c h 4 inhibits endothelial cell m i g r a t i o n t h r o u g h c o l l a g e n but not fibrinogen  In order for capillaries to sprout, endothelial cells need to migrate toward a stimulus. To examine whether defective migration could explain the Notch4-induced inhibition of sprouting, we performed chemotaxis assays using Transwell filters coated with either fibrinogen or collagen. When filters were coated with fibrinogen, both control cells and Notch4IC-expressing cells exhibited similar degrees of chemotaxis towards FGF-2 and V E G F (Figure 3.5). Migration through collagen-coated filters towards FGF-2 or V E G F , however, was reduced for Notch4IC cells compared to control cells (Figure 3.5). These data suggest that activated Notch4 does not affect the intrinsic motility of H M E C cells, but influences endothelial cell migration in a matrixdependent manner.  100  140 120  Fibrinogen HMEC-LNCX HMEC-Notch4IC  Collagen  100 E 80 0  Z5  C  - 60 CD  O 40 20 0 Control  FGF-2  VEGF  Control  FGF-2  VEGF  Figure 3.5: Activated Notch4 inhibits endothelial cell migration through collagen but not fibrinogen. Migration of H M E C - L N C X and HMEC-Notch4IC cells towards control medium, or medium supplemented with F G F - 2 (15 ng/ml) or V E G F (15 ng/ml) was assayed using Transwell filters coated with fibrinogen or collagen type I. Following 16 hours of incubation, cells that had migrated and adhered to the underside of the filter were stained and counted. Data shown are the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. * F G F - 2 : P < 0.01. * V E G F : P S 0.05.  101  3.2.5  N o t c h 4 p r o m o t e s a d h e s i o n t o E C M p r o t e i n s t h r o u g h p1 i n t e g r i n s  Modulation of both cell surface integrin levels and integrin affinity are crucial events throughout the course of capillary tube formation (Bloch et al., 1997; Grant et al., 1989) and cell migration (Lauffenburger and Horwitz, 1996). Therefore, to explain the matrix-specific inhibition of HMEC-Notch4IC migration, we investigated whether Notch4 activation affects endothelial cell adhesion to E C M proteins. Notch4IC-expressing cells exhibited increased adherence to various E C M proteins (Figure 3.6A). In contrast, when adhesion was mediated by charge interactions alone (on poly-L-lysine-coated wells), Notch4IC cells and control cells both adhered to the same degree (Figure 3.6A). Regulation of avp3 and pi integrins is required for angiogenesis (Bloch et al., 1997; Eliceiri and Cheresh, 1999). Because activation of Notch4 promoted adhesion to fibronectin, collagen type I, and collagen type IV, all of which are p 1 integrin substrates (Figure 3.6A), we postulated that the pattern of increased adhesion was due to effects of Notch4 activation on pi integrin expression or function. Using the function-blocking pi integrin Ab P4C10 (Carter et al., 1990), we confirmed that the majority of the increased HMEC-Notch4IC cell adhesion to collagen type I (Figure 3.6B,C) or collagen type TV (Figure 3.6C) was mediated by pi integrins. A function-blocking Ab directed against avp3 integrin (LM609) (Cheresh, 1987), however, did not affect HMEC-Notch4IC cell adhesion to collagen type I (Figure 3.6B). LM609 concentrations of up to 20 ng/ml were tested, with no effect on collagen type I adhesion (data not shown). Interestingly, when LM609 and P4C10 were used in combination, the inhibition of HMEC-Notch4IC cell adhesion to collagen type I was less effective than P4C10 used alone (Figure 3.6B). Although the reason(s) for the attenuated blocking is not clear, this may be due in part to steric hindrance.  102  0.9  o  0.8  * • HMEC-LNCX HMEC-Notch4IC  H  CO CO 0.7 I  o 0.61^ir> 0 . 5 Q O 0.4c= o 0.3 CO 0> 0.2  sz  <  0.1 0.0  LIILL FBG  B  FN  C O L I C O L IV  0.7-  o  co  - o  <  C  li  0.30.2  i  0.1 • 0.0  9 0.8 -  o3  CO  o  0) 0.2  <  0.1 0.0  V V  I  0.3 -  cn -C  if  'IgG P4C10  0.7 -  o 0.6 r— in 0.5 Q O 0.4 -  c  POLY  • HMEC-LNCX HMEC-Noteh4IC  0.6-  o 0.5f>» cn Q 0.4 O o  VN  -  HMECLNCX  HMECNotcMIC  Collagen  Collagen IV  Figure 3.6: Activated Notch4 promotes endothelial cell adhesion to various ECM proteins through P1 integrins. (A) Adhesion of H M E C - L N C X and HMEC-Notch4IC cells to E C M proteins. Plates were coated with the following proteins: F B G , fibrinogen; F N , fibronectin; C O L I, collagen type I; C O L IV, collagen type IV; V N , vitronectin; P O L Y , poly-L-lysine. Adherent cells were fixed, stained, and solubilized, and absorbance read at 570 nm with background absorbance subtracted at 630 nm. * F N : P < 0.001. * C O L I: P < 0.001. * C O L IV: P < 0.001. * V N : P < 0.05. (B) Adhesion of H M E C - L N C X and H M E C Notch4IC cells in the presence of function-blocking Abs against avp3 and p i integrins. Adhesion assays were pedormed on plates coated with collagen type I. H M E C - L N C X and HMEC-Notch4IC cells were preincubated with l g G 2 a (1:100 dilution), anti-avp3 Ab (LM609, 10 ug/ml), and anti-pi Ab (P4C10, 1:100, 1:250, and 1:500 dilutions). For cells treated with both avp3 and pi Abs, 10 ug/ml and 1:100 dilution were used, respectively. * Anti-pi 1:100 dilution versus IgG: P < 0.01. (C) Adhesion of H M E C - L N C X and HMEC-Notch4IC cells in the presence of a function-blocking p i integrin Ab. Adhesion assays were performed on plates coated with collagen type I or collagen type IV. Cells were pre-incubated with l g G 2 a (1:500 dilution) or anti-pi Ab (P4C10, 2.5 ug/ml). Adhesion data shown are the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. *Collagen I L N C X : P < 0.001. *Collagen I Notch4IC: P < 0.05. 'Collagen IV L N C X : P < 0.01. 'Collagen IV Notch4IC: P < 0.001.  103  We next tested whether Notch4IC affected the expression levels of av(33 and pl integrins at the cell surface. Using flow cytometry, we demonstrated that neither <xvp3 nor pl integrin levels were upregulated on the surface of HMEC-Notch4IC cells compared to controls (Figure 3.7A). In fact, in most experiments, there was decreased pl integrin, but not ccvp3, on the surface of HMEC-Notch4IC cells. Non-ligand-binding (inactive or low-affinity) integrins can be converted to a ligand-binding (active or high-affinity) state via inside-out signaling (Hynes, 1987). Our findings of increased pl integrin-mediated adhesion without increased pl integrin expression suggest that Notch4IC may participate in an inside-out signaling process that promotes pl integrin affinity. To test this hypothesis, we performed ligand binding assays using FITC-conjugated collagen type I. Binding of soluble collagen type I to HMEC-Notch4IC cells was greater than that for H M E C control cells (Figure 3.7B). Because HMEC-Notch4IC cells exhibit increased binding to soluble collagen type I (Figure 3.7B) without a corresponding increase in total pl integrin expression (Figure 3.7A), our findings suggest that HMEC-Notch4IC cells display a greater number of pl integrins in a high-affinity, active conformation, compared to control cells. To confirm the increased proportion of active pl integrin on the surface of H M E C Notch4IC cells, Notch4IC cells and control cells were stained with Abs that specifically recognize active pl integrin (B44) (Ni et al., 1998; Wilkins et al., 1996) or total pl integrin (K20) (Takada and Puzon, 1993), and mean fluorescence ratios (active pl / total pl) were determined by flow cytometry. Notch4IC cells, compared to control cells, displayed a greater proportion of pl integrin receptors in a high-affinity state (Figure 3.8A). Based on our findings, we reasoned that i f pl integrins expressed on HMEC-Notch4IC cells were already in a high-affinity state, we would not be able to further increase pl integrin-mediated adhesion to collagen. Using a  104  CD  E 3  HMEC-LNCX  HMEC-LNCX  HMEC-Notch4IC  HMEC-Notch4IC  C ©  o  CD .>  00  CD  or  IgG control  •  K20 (total B1)  • •  IgG control L M 6 0 9 (total txv83)  Fluorescence Intensity  B CD CO  •  14 12 T  _  X  - o -  - • -  H M E C - L N C X  HMEC-Notch4IC  10 -  8 %8 -o  3  «= CD  g o oo E  6 4 2  500  1000 1500 2000 2500 3000 3500 FITC-Collagen (nM)  Figure 3.7: Activated Notch4 does not increase endothelial cell-surface expression of p i integrins but enhances binding of soluble collagen. (A) Expression of av[33 and p i integrins on the surface of H M E C - L N C X and HMEC-Notch4IC cells. Cells were incubated with Abs (IgG-control, K20-total p i , LM609-total avp3) and analyzed by flow cytometry. Histograms shown are representative of at least three independent experiments. (B) Binding curve of soluble collagen to H M E C - L N C X and HMEC-Notch4IC cells. FITC-conjugated collagen type I was incubated with cells at the indicated concentrations, and the samples analyzed by flow cytometry. Binding data shown are the mean ± standard deviation of two independent experiments.  105  A  a  o.30  experiment #1  experiment #2  J° P  I Sg  o  0.25 0.20  0 * 0.15  § <D 0.10  »= > £ «  CD CD 1  B  0.05 0.00 HMEC- HMECLNCX Notch4IC  HMEC- HMECLNCX Notch4IC  0.50  0.00  HMEC-LNCX  Figure 3.8: Activated Notch4-expressing  HMEC-Notch4IC  cells display  p1 integrins in a high-affinity  conformation. (A) Mean fluorescence ratios of active pi to total pi on the surface of H M E C - L N C X and HMEC-Notch4IC cells. Cells were incubated with A b s (B44-active pi, K20-total pi) and assessed by flow cytometry. Data shown are from two independent experiments. (B) H M E C Notch4IC cell adhesion to collagen cannot be increased by function-activating pi integrin A b s . H M E C - L N C X and HMEC-Notch4IC cells pre-incubated with the function-activating pi integrin Ab 8A2 were added to collagen type l-coated wells. Adherent cells were fixed, stained, and solubilized, and absorbance read at 570 nm with background absorbance subtracted at 630 nm. Data shown are the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. * P £ 0.05.  106  function-activating B l integrin Ab (8A2) (Kovach et al., 1992), we found that whereas H M E C L N C X cell adhesion to collagen type I could be increased, 8A2 was unable to increase H M E C Notch4IC cell adhesion to collagen type I (Figure 3.8B). ,Taken together, our findings demonstrate that Notch4IC cells already display p l integrins in a fully-active conformation.  3.2.6  I n c r e a s e d p1 i n t e g r i n - m e d i a t e d a d h e s i o n p l a y s a r o l e i n t h e N o t c h 4 inhibition of endothelial sprouting  Our data suggest that the inhibited sprouting of Notch4IC cells in vitro may be explained in part by an increased affinity to gelatin (denatured collagen)-coated beads, and that this highaffinity adhesive state (which presumably cannot be "turned o f f due to the constitutive activation of Notch4) prevents the Notch4IC cells from migrating off the beads and into the fibrin gel. This would suggest that i f Notch4IC cells were seeded onto beads by charge interaction rather than p l integrin-mediated adhesion, the ability to form sprouts would be restored. To test this hypothesis, HMEC-Notch4IC cells were seeded onto dextran-coated microcarrier beads, and the beads were embedded into fibrin gels. In this assay, we noted that HMEC-Notch4IC cells formed sprouts to a similar extent as H M E C - L N C X cells (Figure 3.9). Hence in the absence of a p 1 integrin substrate with which to interact, Notch4IC cells are capable of forming sprouts. As a control, dextran-coated beads were further coated with collagen type I, and then seeded with HMEC-Notch4IC cells. Similar to results shown in Figure 3.IB, sprouting of HMEC-Notch4IC cells from these collagen-recoated beads was inhibited (Figure 3.9).  107  0.7 0.6 CO CD  HMEC-LNCX  HMEC-Notch4IC  0.50.4  Io  0.3  a  02 0.1 0.0  control  VEGF  Dextran coated  control  VEGF  Collagen coated  Figure 3.9: Activated Notch4 does not inhibit endothelial sprouting from dextran-coated microcarrier beads in vitro. Dextran-coated microcarrier beads were seeded with H M E C - L N C X or HMEC-Notch4IC cells. Equal numbers of beads were embedded in fibrin gels containing control medium or medium supplemented with V E G F (15 ng/ml). Endothelial sprout formation was quantitated after 3 days incubation. A s a control, dextran-coated beads were coated with collagen type I, which were then seeded with H M E C cell lines. Data shown are the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. 'Control: P < 0.01. * V E G F : P < 0.01.  108  3.2.7  A c t i v a t i o n o f p1 i n t e g r i n s i s s u f f i c i e n t t o i n h i b i t e n d o t h e l i a l s p r o u t i n g i n vitro a n d a n g i o g e n e s i s in v i v o  Our findings described thus far demonstrate that expression of activated Notch4 in endothelial cells inhibits angiogenesis both in vitro and in vivo, in part by promoting p 1 integrin activation. To determine whether activation of pi integrins alone (independent of constitutivelyactive Notch4 expression) was sufficient to inhibit angiogenesis, we performed in vitro and in vivo experiments using Abs that specifically activate pi integrins. Using the in vitro sprouting model with untransduced parental H M E C cells, we investigated the effect of the function-activating pi integrin Ab 8A2 on VEGF-induced sprouting. Dextran-coated beads were preincubated with either 8A2 or LM534, a non-functionmodifying pi Ab (Takada and Puzon, 1993). These beads were subsequently seeded with parental H M E C , and incubated for three days to allow parental H M E C to produce and secrete their own matrix proteins (many of which are pi integrin substrates) onto the bead surface (Kramer et al., 1985). Whereas V E G F was able to induce sprouting of parental H M E C from beads coated with LM534, sprouting from 8A2-coated beads was reduced (Figure 3.1 OA). Hence pl integrin activation was sufficient to inhibit VEGF-induced endothelial sprouting in vitro. Using the in vivo chick C A M assay, VEGF-induced angiogenesis was examined in the presence of various anti-avian pl integrin Abs (Figure 3.10J3,C). C A M s were treated with TASC (a function-activating p l integrin Ab) (Cruz et al., 1997; Neugebauer and Reichardt, 1991), V2E9 (a non-function-modifying pl integrin Ab) (Hayashi et al., 1990), or W1J310 (a functionblocking p l integrin Ab) (Cruz et al., 1997). Whereas V E G F was able to induce angiogenesis in C A M s treated with V2E9 and W1J310, C A M s treated with T A S C exhibited decreased angiogenesis (Figure 3.10B,C). In fact, angiogenesis in TASC-treated C A M s was reduced to a level similar to that of C A M s treated with LM609 (Figure 3.10B,C), a function-blocking avp3  109  + VEGF  Figure 3.10: Activation of pi integrins alone, independent of Notch4 activation, is sufficient to inhibit endothelial sprouting in vitro and angiogenesis in vivo. (A) In vitro sprouting of parental H M E C from microcarrier beads coated with anti-p1 integrin Abs. Dextran-coated microcarrier beads were pre-incubated with IgG control Ab, a function-activating pi integrin A b (8A2), or a non-function-modifying B1 integrin Ab (LM534). Data shown are the mean ± standard deviation of a single experiment done in triplicate and are representative of at least three independent experiments. *8A2 versus IgG: P s 0.05. *8A2 versus LM534: P 5 0.05. (B) Angiogenesis in the chick C A M in the presence of anti-p1 integrin Abs. The following Abs were used: T A S C (function-activating pi integrin Ab), V 2 E 9 (non-function-modifying pi integrin Ab), W 1 B 1 0 (function-blocking pi integrin Ab), LM609 (function-blocking <xvp3 integrin Ab). Abs (10 ug/ml) plus V E G F (30 ng/ml) were loaded onto gelatin sponges, and the sponges placed on the C A M s of day 8 embryos. A s controls, sponges containing P B S or V E G F (30 ng/ml) were placed on C A M s . Angiogenesis was quantitated on day 10. Data shown are the mean ± standard error of two experiments each done in replicates of 3-5 eggs. * T A S C versus V E G F : P s 0.05. * T A S C versus V2E9: P < 0.0001. (C) C A M s treated with V 2 E 9 , W1B10, T A S C , and LM609 A b s , all in the presence of V E G F . C A M s shown are representative of two independent experiments. Arrows indicate the corners of the sponges. Bars represent 1 mm.  110  integrin Ab previously shown to attenuate VEGF-induced angiogenesis in the C A M (Friedlander et al., 1995). Taken together, our findings demonstrate that activation of pl integrins, and hence increased adhesion through p l integrins, is sufficient to inhibit VEGF-induced endothelial sprouting in vitro and angiogenesis in vivo.  3.3  DISCUSSION  Quiescent endothelial cells are normally anchored by their abluminal surface to a collagen-rich matrix (Carey, 1991; Iruela-Arispe et al., 1991b; Risau and Lemmon, 1988). At the initiation of angiogenesis, the mature collagen-containing matrix is degraded and replaced by a provisional matrix composed of fibrin (Dvorak et al., 1995) and fibronectin (Clark et al., 1982) upon which endothelial cells migrate and proliferate. The endothelial sprouting assay used in our studies mimics angiogenesis in vivo. Specifically, microvascular endothelial cells are seeded as a monolayer onto gelatin-coated beads, and are then induced by angiogenic factors to migrate into a fibrin matrix to form sprouts. We report that endothelial cells expressing Notch4IC exhibit inhibited sprouting in vitro (Figures 3.1 and 3.9), and that this inhibition can be explained in part by an increase in Notch4IC cell adhesion to collagen (Figures 3.6, 3.7, and 3.8). B y enhancing cell adherence to collagen-coated beads, activated Notch4 prevents migration of the cells into the fibrin matrix. This is in accordance with our migration studies, in which Notch4IC cell migration through collagen, but not fibrinogen, is inhibited (Figure 3.5). Proliferation rates, on the other hand, are similar in FfMEC-Notch4IC and control cells (Figure 3.4). Our in vivo studies demonstrate that Notch4IC expression in the chick C A M inhibits VEGF-induced angiogenesis (Figures 3.2 and 3.3). Based on our in vitro findings, the inhibition of angiogenesis in vivo may be due in part to enhanced endothelial cell adhesion to matrix proteins, thereby inhibiting vascular remodeling in the C A M . Since the completion of this thesis, results from our laboratory  111  have determined that Notch4IC-induced inhibition of endothelial sprouting requires the cdclO/ankyrin repeats of Notch4, and involves both CBF1-dependent and -independent pathways (MacKenzie et al., 2004a). Hence enhanced Notch4IC cell adhesion to collagen likely involves signaling mediated by the cdclO/ankyrin repeats. Cell migration requires the coordinated activation and deactivation of integrins (Lauffenburger and Horwitz, 1996). As a cell migrates across a matrix, integrins at the leading edge of the cell adhere to the substrate (Huttenlocher et al., 1996). At the same time, integrins at the trailing edge of the cell detach from the substrate to allow the cell to progress forward (Palecek et al., 1998). Thus during the sprouting process of angiogenesis, integrin affinity states are constantly being modulated. Although avB3 integrin has been shown to play a critical role in angiogenesis, several studies also delineate the essential contribution of Bl integrins to endothelial morphogenesis (Bloch et al., 1997; Eliceiri and Cheresh, 1999). Our data show that activated Notch4 increases endothelial cell adhesion (Figure 3.6), and that enhanced pl integrin affinity plays a role in this increased adhesion (Figures 3.7 and 3.8). Hence the ability of Notch4 to inhibit endothelial sprouting in vitro and angiogenesis in vivo may be related in part to its ability to increase the ligand-binding affinity of pl integrins. Other potential mechanisms, however, may act in concert with pl integrin activation to mediate the observed Notch4 effect. There is much evidence demonstrating that suppression of integrin activation is a physiological mechanism with which to control integrin-dependent cell adhesion and migration (Hughes and Pfaff, 1998). In addition, regulation of integrin activation has been reported to precede differentiation in several cell types. Regulation of pl integrin activity has been reported in neurogenic and myogenic differentiation, two processes that are also modulated by Notch (Boettiger et al., 1995; Neugebauer and Reichardt, 1991). In a baboon model, it has previously  112  been shown in uninjured saphenous arteries that endothelial cells and vascular smooth muscle cells express an epitope characteristic of (31 integrins in a high-affinity state (Koyama et al., 1996). However, six weeks following balloon injury, regenerating endothelial cells did not express this ligand-induced epitope, although there was no decrease in the expression of total p l integrin (Koyama et al., 1996). In the same study, activation of p l integrins with the functionactivating pl integrin A b 8A2 inhibited the migration of endothelial cells in vitro (Koyama et al., 1996). Together, these findings suggest that activated pl integrin is required to maintain endothelial cells in a quiescent state, but in order to repair arteries and possibly to allow neovascularization, dyshesion by downregulating p l integrin affinity is required. In fact, activation of p l integrins on human endothelial cells has been shown to inhibit capillary tube formation in collagen gels in vitro (Gamble et al., 1999). We report that activation of pl integrins on endothelial cells, independent of Notch4 activation, inhibits endothelial sprouting in vitro (Figure 3.1 OA). We demonstrate that pl integrin activation can inhibit angiogenesis in the chick C A M in vivo (Figure 3.10B,C). In a previous study using function-blocking Abs directed against specific a integrin subunits, a combination of a 1-blocking and a2-blocking Abs was shown to inhibit VEGF-induced angiogenesis in a mouse Matrigel plug assay (Senger et al., 1997). These findings suggest that blocking a l p l and a2pi integrin function can inhibit VEGF-induced angiogenesis (Senger et al., 1997). Although these results may seem contradictory to our data demonstrating that blocking p l integrin function does not inhibit VEGF-induced angiogenesis in the chick C A M (Figure 3.10B,C), it is important to note that the effect of functionblocking/activating Abs directed against the pl integrin subunit in the Matrigel plug assay was not reported. Because numerous a-pl integrin heterodimers are implicated in angiogenesis  113  (Bauer et al., 1992; Davis et al., 1993), blocking the function of only the al and a2 subunits may result in a different phenotype from that seen when the function of all Bl integrins is blocked. Alternatively, the different results may reflect intrinsic differences in the experimental models used. Indeed, the function-blocking pl integrin Ab C S A T has been reported to disrupt vascular development and lumen formation when micro-injected into quail embryos (Drake et al., 1992), whereas the same C S A T Ab does not affect FGF-2- or TNFa-induced angiogenesis in the chick C A M (Brooks et al., 1994a). Our work demonstrates that constitutive Notch4 activation inhibits vascular remodeling. Importantly, our studies provide a possible mechanism with which to explain the common vascular defects observed in mutant mice with either increased (Uyttendaele et al., 2001) or decreased (Krebs et al., 2000) Notch signaling. Because Notch plays a role in cell fate decisions, Notch signaling must be precisely regulated and hence requires cessation of receptor signaling at certain times (Artavanis-Tsakonas et al., 1999; Milner and Bigas, 1999; Weinmaster, 2000). Similarly, because cell adhesion influences cell functions such as migration and cell phenotype, modulation of cell adhesion must be strictly regulated (Bloch et al., 1997; Gamble et al., 1993; Lauffenburger and Horwitz, 1996; Palecek et al., 1997; Ruoslahti and Engvall, 1997). Therefore, it is possible that knocking out Notch4 and Notchl results in a loss of cell-ECM adhesion and hence inhibited vascular remodeling, whereas constitutive Notch4 activation results in excessive cell-ECM adhesion, thereby effectively fixing the cells in place. Taken together, our studies as well as the studies of Krebs et al. (Krebs et al., 2000) and Uyttendaele et al. (Uyttendaele et al., 2001) reveal that altered Notch4 signaling results in disrupted blood vessel development. Because Notch4 expression is restricted to the endothelium (Uyttendaele et al., 1996), and because Notch4 is the only Notch receptor expressed in capillary endothelium (Villa et al.,  114  2001), our findings implicate selective activation of Notch4 as a possible method with which to inhibit angiogenesis in pathological contexts. However, because our studies involve a constitutively-active, overexpressed form of Notch4 in endothelial cells, the physiological relevance of the data must be interpreted with caution. Further studies using ligands specific for Notch4 will be important to determine whether ligand-induced activation of Notch4 also inhibits angiogenesis.  115  Chapter 4  N O T C H S I G N A L INHIBITION A T T E N U A T E S B R E A S T T U M O R G R O W T H BY REVERSING THE M E S E N C H Y M A L PHENOTYPE  4.1  ABSTRACT  During tumor progression, tumor cells of epithelial origin often acquire a mesenchymal phenotype through E M T , a process that promotes invasion and dissemination of cancer cells (Vincent-Salomon and Thiery, 2003). Recent studies from our laboratory and others have shown that activation of Notch signaling promotes a specialized type of E M T , EndoMT (Noseda et al., 2004b; Timmerman et al., 2004). Given that Notch pathway elements are expressed at sites of epithelial-mesenchymal cell-cell interactions during embryogenesis (Mitsiadis et al., 1995; Nakagawa et al., 1999) and within primary human breast tumors (Callahan and Egan, 2004; Imatani and Callahan, 2000; Parr et al., 2004; Pece et al., 2004; Weijzen et al., 2002), and that loss of E-cadherin directly correlates with E M T in human breast cancer (Behrens et al., 1989), we determined whether Notch signaling would modulate E-cadherin expression and hence breast tumorigenesis. We have identified activated Notch signaling as a novel mechanism for the downregulation of E-cadherin expression in human breast epithelial cells. We demonstrate that inhibition of Notch signaling attenuates E-cadherin promoter methylation and induces E cadherin re-expression in human breast tumor cells previously lacking E-cadherin expression, resulting in restricted B-catenin nuclear accumulation and a marked reduction in tumor growth and metastasis.  116  4.2  RESULTS  4.2.1  Activated N o t c h signaling inhibits E-cadherin e x p r e s s i o n in h u m a n epithelial cells  breast  We transduced the E-cadherin-positive human breast epithelial cell line MCF-10A with a retroviral vector (MIY) containing Y F P linked to Notch 1IC or Notch4IC through an internal ribosomal entry site. Expression of Notchl IC or Notch4IC induced cytologic changes consistent with E M T , such as cell scattering and the acquisition of a spindle-shaped morphology (Figure 4.1 A). Notch 1IC or Notch4IC expression also reduced E-cadherin expression (Figure 4.1B,C). Because MCF-10A are an immortalized human breast epithelial cell line (Tait et al., 1990), we determined whether Notch 1IC or Notch4IC expression would inhibit E-cadherin expression in primary breast epithelial cells. Similar to MCF-10A, primary human breast epithelial cells transduced with activated Notchl or Notch4 downregulated E-cadherin expression (Figure 4.2).  117  phase contrast  MIY  MIYNotchllC MIYNotch4IC  E-cadherin expression Figure 4.1 (Part 1): Activated Notch signaling inhibits E-cadherin expression in MCF-10A human breast epithelial cells. (A) Phase contrast images of M C F - 1 0 A cells transduced with MIY, M I Y N o t c h l l C , or MIYNotch4IC. Bar, 100 nm. (B) Flow cytometry histograms for surface Ecadherin expression in M C F - 1 0 A cell lines. Data shown are from one experiment and are representative of two independent experiments. (C) (see next page).  118  Figure 4.1 (Part 2): Activated Notch signaling inhibits E-cadherin expression in MCF-10A human breast epithelial cells. (C) Immunofluorescence for E-cadherin (red), Y F P (yellow), and DAPI (blue) in M C F - 1 0 A cell lines. Bar, 50 um. Insets represent close-up magnifications of 6-8 cells.  E-cadherin  YFP  DAPI  Merged  Figure 4.2: Activated Notch signaling inhibits E-cadherin expression in primary human breast epithelial cells. Immunofluorescence for E-cadherin (red), Y F P (yellow), and DAPI (blue) in primary human breast epithelial cells transduced with MIY, M I Y N o t c h l l C , or MIYNotch4IC. Bar, 100 nm.  4.2.2  N o t c h s i g n a l inhibition r e d u c e s h u m a n breast t u m o r g r o w t h in v i v o  Having demonstrated that activated Notch signaling is associated with a decrease in E cadherin expression in human breast epithelial cells, we sought to determine whether inhibition of Notch signaling would induce re-expression of E-cadherin and inhibit tumor growth in M D A MB-231 human breast tumors. MDA-MB-231 cells are E-cadherin-negative despite possessing one wildtype E-cadherin gene, and thus exhibit reversible E-cadherin silencing (van de Wetering et al., 2001). We also performed experiments with mouse Lewis lung carcinoma cells, which express barely detectable levels of E-cadherin protein (Foty and Steinberg, 1997). Both cell types expressed multiple Notch receptors and ligands (Figure 4.3A); however, MDA-MB-231 cells exhibited approximately a 5-fold greater expression of the key Notch target gene HES1 (Figure 4.3B).  121  A  MDA-MB-231 parental  f#  ^  £>  t&  #  N  'b  * <# e  ^  ^  ^^y  ^  ^j.  ^  K O ^  ^  ^  »5  ^  Lewis lung carcinoma parental  //////////  B  HES1 GAPDH m MDA-MB-231 parental • Lewis lung carcinoma parental  Figure 4.3: MDA-MB-231 and Lewis lung carcinoma parental tumor cells both express Notch ligands and receptors, with MDA-MB-231 cells expressing greater levels of HES1. (A) R T - P C R for expression of human Notch ligands and receptors in MDA-MB-231 parental cells and mouse Notch ligands and receptors in Lewis lung carcinoma parental cells. Human-specific primer sets do not recognize mouse targets, and mouse-specific primer sets do not recognize human targets. (B) R T - P C R for expression of the Notch target gene HES1 in MDA-MB-231 and Lewis lung carcinoma parental cells. Primers were designed to recognize both human and mouse H E S 1 . Data shown represent mean ± standard error from three independent experiments. * P < 0.01. 122  The most commonly used inhibitors of Notch signaling are the y-secretase inhibitors, which block proteolytic processing of transmembrane Notch (Das et al., 2004). However, these inhibitors are not specific for Notch signaling as they also block signaling by other receptors such as ErbB4 (Lee et al., 2002). To specifically inhibit Notch signaling, we generated a retroviral construct expressing the soluble extracellular domain of human Notch4 (XNotch4). We transduced XNotch4 c D N A into MDA-MB-231 or Lewis lung carcinoma cells and confirmed secretion  of the  soluble protein  (Figure  4.4).  When implanted  subcutaneously  into  immunodeficient mice, XNotch4 inhibited MDA-MB-231 but not Lewis lung carcinoma tumor growth (Figure 4.5A). We confirmed in vivo expression of XNotch4 protein in both tumor types (Figure 4.5B). In addition to reduced growth of the primary tumor, inhibition of Notch signaling reduced both the number and average weight of metastatic nodules in axillary and subiliac lymph node regions in XNotch4 mice (Figure 4.6). Neither control nor XNotch4 mice exhibited metastases in internal organs (data not shown). B y semi-quantitative RT-PCR using two primer sets, one specific for human HES1 and the other for mouse HES1, we confirmed in M D A - M B 231 xenografts that XNotch4 inhibited Notch signaling specifically in the human tumor cells and not in the murine stromal cell compartment (Figure 4.7A,B). Also, HES1 m R N A levels were not altered between Lewis lung carcinoma tumors expressing vector or XNotch4 (Figure 4.7C). These findings may explain the lack of tumor growth inhibition in Lewis lung carcinoma tumors expressing XNotch4 (Figure 4.5A).  123  MDAL e w i s lung MB-231 carcinoma <  < X  X  cell lysate cell supernatant  IGXNol  o  s —"  IGXNol  1  U  o  o  •*»  HA  F i g u r e 4.4: X N o t c h 4 protein is s e c r e t e d . Immunoblot for expression of HA-tagged XNotch4 in MDAMB-231 and Lewis lung carcinoma cell lines transduced with MIG or MIGXNotch4HA. Concentrated supernatant from cultured cells was obtained by filtering 3 day conditioned medium through a 100,000 kDa molecular weight cut-off Ultrafiltration Membrane.  124  Lewis lung carcinoma In vivo  Figure 4.5 (Part 1): Notch signal inhibition reduces human breast tumor growth in vivo. (A) Tumor growth curves for MDA-MB-231 and Lewis lung carcinoma cell lines grown as xenografts in the dorsa of mice. MDA-MB-231 and Lewis lung carcinoma tumor volumes were measured weekly and every two days, respectively. Representative macroscopic images of tumors at the time of sacrifice are shown. Bar, 5 mm. * P < 0.01. (B) (see next page).  125  B  MDA-MB-231 In vivo MIGXNotch4HA  MIG  LU  i * ^/ 4  .1.  •  Figure 4.5 (Part 2): Notch signal inhibition reduces human breast tumor growth in vivo. (B) Immunostaining for XNotch4HA expression in MDA-MB-231 and Lewis lung carcinoma tumors. H & E staining is also shown. Bar, 50 u.m.  126  MDA-MB-231 Tumor Metastases  «2 D) Q)  E  t  3 ° v £  H L  120  MIG (n=25) MIGXNotch4HA (n=27)  Figure 4.6: Notch signal inhibition reduces metastasis of human breast tumors in vivo. Quantitation of metastases in MDA-MB-231 tumor-bearing mice. Control and XNotch4 mice were sacrificed at the same time, mice were examined for metastases, and the total number and total weight of metastases determined. Data shown represent the average number of metastases per mouse ± standard error, and the average weight of each metastatic nodule ± standard error. *P < 0.05. Metastasis data were determined by analyzing 25 MIG tumors and 27 MIGXNotch4HA tumors.  127  A  B  M D A - M B - 2 3 1 In v i v o 1 1 10 m 9 8 o  o  7  6 6 4  in a> i_ Q. x  LU  Human HES1  M D A - M B - 2 3 1 In v i v o  L e w i s l u n g c a r c i n o m a In v i v o in H  e  £  <  A  I  c  JZ  *  2 5  3 2 1  ••  0  MIG MIGXNotcMHA  hHES1  mHES1  mHES1  hGAPDH  mGAPDH  mGAPDH  hHES1  mHES1  hGAPDH 26 29 32 P C R cycle n u m b e r  MIGXNotch4HA  mGAPDH 29 32 35 P C R cycle number  mHES1  MIGXNotch4HA  mGAPDH 26 29 32 P C R cycle number  Figure 4.7: Soluble XNotch4 inhibits Notch signaling specifically in the tumor cell compartment of MDA-MB-231 tumors in vivo. Total R N A was harvested from MDA-MB-231 and Lewis lung carcinoma tumors, and reverse-transcribed into c D N A . For semi-quantitative R T - P C R , individual reactions along with G A P D H control reactions were terminated after the designated number of P C R cycles. (A) Semi-quantitative R T - P C R for expression of the Notch target gene H E S 1 specifically in the tumor cell compartment of MDA-MB-231 tumors. Data shown represent mean ± standard error. * P < 0.0001. (B) Semi-quantitative R T - P C R for H E S 1 expression in the host cell compartment of M D A - M B 231 tumors. Data shown represent mean ± standard error. (C) Semi-quantitative R T - P C R for H E S 1 expression in Lewis lung carcinoma tumors. Data shown represent mean ± standard error.  Although data from this thesis and other studies (Krebs et al., 2000; Uyttendaele et al., 2001) demonstrate that aberrant Notch signaling results in disrupted in vivo embryonic blood vessel development, we did not observe differences in vascular density in implanted tumors in response to XNotch4 (Figure 4.8A,B). This suggests that inhibition of angiogenesis is not the mechanism by which XNotch4 inhibits MDA-MB-231 tumor growth. However, because angiogenesis inhibitors can induce tumor regression without altering vascular density (Hlatky et al., 2002), vascular density alone may not provide a good evaluation of anti-angiogenic activity. Hence an anti-angiogenic role for XNotch4 cannot be completely ruled out.  129  M D A - M B - 2 3 1 In vivo MIGXNotch4HA  MIG  • i M I G (n=14) • MIGXNotch4HA(n=12)  B  • •  M I G (n=13) MIGXNotch4HA(/7=16)  i n  L e w i s lung c a r c i n o m a In vivo  MIGXNotch4HA  MIG  O O  2500  •  M I G (/7=7)  O  MIGXNotch4HA  H •  (n=6)  M I G (n=14) MIGXNotch4HA(n=16)  Figure 4.8: Soluble XNotch4 does not affect vascular density in vivo. (A) Vascular density in MDAMB-231 tumors in vivo. Immunostaining for the endothelial marker CD31 (red) and DAPI (blue). Bar, 50 um. Vascular density was quantitated by expressing the CD31 stained area as a percentage of the total tumor area, and by number of vessels per m m . Data shown represent mean ± standard error. (B) Vascular density in Lewis lung carcinoma tumors in vivo. Immunostaining for CD31 (red) and DAPI (blue). Bar, 50 um. Vascular density was quantitated by expressing the CD31 stained area as a percentage of the total tumor area, and by number of vessels per m m . Data shown represent mean ± standard error. 2  2  130  4.2.3  N o t c h signal inhibition induces E-cadherin e x p r e s s i o n in human t u m o r s in v i v o  breast  We examined MDA-MB-231 tumor xenografts to determine whether Notch inhibition induced re-expression of E-cadherin. XNotch4 but not control tumors expressed E-cadherin protein (Figure 4.9A). Because the Ab used recognizes both human and mouse E-cadherin, RTPCR was performed with human E-cadherin-specific primers. Human E-cadherin transcripts were detected only in tumors in which Notch was inhibited (Figure 4.9B), suggesting transcriptional reversal of E-cadherin repression specifically in the human tumor xenografted cells. B y immunofluorescence microscopy, E-cadherin staining was detected at the plasma membrane in response to Notch inhibition (Figure 4.9C), indicating a functional re-expression of E-cadherin. p-catenin contributes to breast tumorigenesis by translocating into the nucleus to modulate the expression of genes involved in cell proliferation, invasion, and E M T (ConacciSorrell et al., 2003; Crawford et al., 1999; He et al., 1998; L i n et al., 2000). When bound to E cadherin, however, signaling-competent  nuclear p-catenin levels diminish resulting in  suppression of cell proliferation and invasion (Wong and Gumbiner, 2003). Because p-catenin binding to cadherins stabilizes catenin expression (Wheelock et al., 2001), we determined whether E-cadherin re-expression in XNotch4 tumors modulated p-catenin expression. Whereas control tumors expressed low levels of p-catenin protein, levels were increased in XNotch4 tumors, which correlated positively with E-cadherin expression (Figure 4.9A). Importantly, however, p-catenin was present at the plasma membrane when Notch was inhibited, but exhibited cytoplasmic/nuclear localization in control tumors (Figure 4.9C). Indeed, we detected a significantly lower proportion of nuclear p-catenin-positive tumor cells in XNotch4 tumors compared to that of control tumors (Figure 4.9D), which was confirmed by immunoblot analysis  131  MDA-MB-231 In vivo p-catenm *  E-cadherin  <  mm MIG (n=16) C3 MIGXNotCh4HA(n=13)  g  3  2  Mi mm  10 20 30 E-cadherin expression (arbitrary units)  B  MDA-MB-231 MIG In vivo  A?  MDA-MB-231 MIGXNotch4HA In vivo  40  cortrc s  /i  A?  E  MDA-MB-231 In vivo  <  O  >-  i  2  C  o a 1.4 g ^ 1.2 | f 10 0.8 o 0.6 I I 0.2 0.4 active |Vcatenin =| 0.0 n-ti ihi ilin f.'.-tuuUMI. £ HMIG(n=14) O MIGXNotCh4HA (n=11)  £S I te  1 ill  Figure 4.9: Notch signal inhibition induces E-cadherin expression in xenografted human breast  tumor cells in vivo. (A) Immunoblot for expression of XNotch4HA, E-cadherin, B-catenin, and tubulin in MDA-MB-231 tumors. Total protein lysates were prepared from tumor tissue. Data shown represent mean ± standard error. * P £ 0.05. Positive correlation, P < 0.01. (B) R T - P C R for expression of human E cadherin in MDA-MB-231 tumors. Total R N A was isolated from MDA-MB-231 tumors, and reversetranscribed into c D N A . Primers recognize human but not mouse E-cadherin. (C) Immunofluorescence for E-cadherin (left panels; red), B-catenin (right panels; red), and DAPI (blue) in MDA-MB-231 tumors. Bar, 15 um. (D) Quantitation of nuclear p-catenin staining in MDA-MB-231 tumors. Data shown represent mean ± standard error. * P < 0.05. (E) Immunoblot for expression of active p-catenin and tubulin in M D A MB-231 tumors in vivo. Data shown represent mean ± standard error.  132  of tumor lysates with an Ab specific for active p-catenin (Figure 4.9E). These results suggest that inhibition of Notch signaling reverses the mesenchymal phenotype by re-inducing expression of E-cadherin, which in turn inhibits B-catenin nuclear translocation to elicit an inhibition of tumor growth. 4.2.4  Expression of E-cadherin alone is sufficient to inhibit human breast tumor growth in vivo To determine whether re-expression of E-cadherin alone (independent of Notch signal  inhibition) was sufficient to inhibit MDA-MB-231 tumor growth, we transduced MDA-MB-231 cells with human E-cadherin c D N A and grew xenografts in mice. E-cadherin overexpression had a similar effect on MDA-MB-231 tumors as that of XNotch4-induced E-cadherin expression. Overexpression of E-cadherin inhibited tumor growth (Figure 4.1 OA). E-cadherin expression also inhibited MDA-MB-231 metastasis formation (Figure 4.1 OB). We confirmed E-cadherin expression and p-catenin induction in E-cadherin-overexpressing tumors (Figure 4.IOC), and observed a positive correlation between E-cadherin and p-catenin expression (Figure 4.IOC). B y immunofluorescence microscopy, we detected plasma membrane staining for both E-cadherin and p-catenin (Figure 4.10D). Taken together with our XNotch4 tumor data, our results show that XNotch4-induced re-expression of E-cadherin plays a causal role in the inhibition of M D A MB-23 1 tumor growth and metastasis.  133  (arbitrary units)  Figure 4.10: Expression of E-cadherin alone, independent of Notch signal inhibition, is sufficient to inhibit human breast tumor growth in vivo. (A) Tumor growth curves for MDA-MB-231 cells transduced with MIY or MIYE-cadherin grown a s xenografts in mice. Tumor volumes were measured weekly. * P < 0.001. (B) Immunoblot for expression of E-cadherin, p-catenin, and tubulin in MDA-MB-231 tumors. Total protein lysates were prepared from tumor tissue. Data shown represent mean ± standard error. *E-cadherin: P < 0.0000001; *p-catenin: P < 0.0001. Positive correlation, P < 0.001. (C) Immunofluorescence for E-cadherin (left panels; red), p-catenin (right panels; red) and DAPI (blue) in MDA-MB-231 tumors. Bar, 15 um. (D) Quantitation of metastases in MDA-MB-231 tumor-bearing mice. After sacrifice, mice were examined for lymph node metastases and the total number and total weight of metastases determined. Data shown represent mean + standard error. * P < 0.001.  134  4.2.5  Induction of E-cadherin e x p r e s s i o n is mediated by attenuated E-cadherin promoter methylation  Our findings demonstrate that inhibited Notch signaling attenuates MDA-MB-231 tumor growth by inducing E-cadherin re-expression. To elucidate a possible mechanism(s) by which inhibited Notch signaling re-induces E-cadherin expression, we investigated the effects of Notch signal inhibition on known mechanisms of E-cadherin silencing. E-cadherin expression can be repressed by several mechanisms including gene mutation, gene deletion, shedding, endocytosis, transcriptional repression, and promoter hypermethylation (Berx et al., 1998; Fujita et al., 2002; Hennig et al., 1996; Herman and Baylin, 2003; Katayama et al., 1994). The latter two mechanisms have been reported to play a role in E-cadherin silencing in MDA-MB-231 cells (Graff et al., 1995; Hajra et al., 2002). The E-cadherin proximal promoter, containing binding sites for E-cadherin transcriptional repressors and sites of promoter methylation, is shown in Figure 4.11.  135  U primer (forward) M primer (forward)  1 5'GACCCTAGCAACTCCAGGCTAGAGGGTCACCGCGTCTATGCGAGGCCGGGTGGGCGG i  i  -189  -133  U primer (reverse) g e n o m i c bisulfite s e q u e n c i n g primer (forward)  I M primer (reverse)  GCCGTCAGCTCCGCCCTGGGGAGGGGTCCGCGCTGCTGATTGGCTGTGGCCGG^B C p G site —> |  -132  -76  BAACCCTCAGCCAATCAGCGGTAGGGGGGGCGGTGCTCCC-GGGCTM^B  G C T G C  -75  -19 |—> transcription start site  AGCCACGCACCCCCTCTCAGTGGCGTCGGAACTGCAAAG  8  .  - -i  -TGAGCTTGCGGA  9 10  11  -18  +39  AGTCAGTTCAGACTCCAGCCCGCTCCAGCCCGGCCCGACCCGACCGCACCCGGCGCC  13 +  4  °  14  15  g e n o m i c bisulfite s e q u e n c i n g  16 +96  primer (reverse)  T G C C C T C G C T C G G C G T C C C C G G C C A G C C A T G G G C C C T T G G A G C : G C A G C C T C T C G G C 3'  19  20 21  |—> translation start site  +153  +97  Figure 4.11: The human E-cadherin proximal promoter. The human E-cadherin gene from positions 189 to +153 relative to the transcription start site is shown. Three E-boxes ( H H H I ) recognized by Ecadherin transcriptional repressor proteins are indicated. For M S P , primer recognition sequences for methylated (M) and unmethylated (U) primer sets are shown. For genomic bisulfite sequencing, primer recognition sequences are shown, a s are the 22 C p G sites (CG) analyzed within the E-cadherin promoter from positions -104 to +118 ( | ' | ) relative to the transcription start site. The translation start site is also indicated. :  136  E-cadherin transcriptional repressor proteins function by binding to the E-cadherin promoter and blocking gene transcription (Hennig et al., 1996). We assessed the expression of three E-cadherin transcriptional repressor proteins, Slug/Snail/SIPl, in control and XNotch4 tumors by semi-quantitative RT-PCR. Expression of these three transcriptional repressor proteins did not differ between control and XNotch4 tumors (Figure 4.12). Thus E-cadherin re-expression in XNotch4 tumors cannot be explained by inhibited expression of Slug, Snail, or SIP1.  137  M D A - M B - 2 3 1 In v i v o  II  C  A  20  1  S  il  controls  H  hSIug hSnail hSlP1 hGAPDH mGAPDH  hSIug  MIG  hGAPDH MIGXNotch4HA  hSIug hGAPDH P C R cycle number  MIGXNotch4HA  hSnail hGAPDH hSnail hGAPDH P C R c y c l e number  MIG MIGXNotch4HA  hSIP1 hGAPDH hSIP1 hGAPDH P C R cycle number  Figure 4.12: Notch signal inhibition does not affect the expression of E-cadherin transcriptional repressor proteins. Semi-quantitative R T - P C R for expression of human Slug, Snail, and SIP1 in MDA-MB-231 tumors. Total R N A was harvested from MDA-MB-231 tumors and reverse-transcribed into cDNA. For semi-quantitative R T - P C R , individual reactions along with G A P D H control reactions were terminated after the designated number of P C R cycles. Primers recognize human but not mouse target genes. Data shown represent mean ± standard error.  The E-cadherin promoter contains numerous CpG sites which, when methylated on the corresponding cytosine residue, can result in E-cadherin silencing (Berx et al., 1995b). In a previous study of MDA-MB-231 cells in vitro, nearly every CpG site analyzed within the E cadherin proximal promoter was found to be methylated (Graff et al., 2000). We assessed the status of E-cadherin promoter methylation within MDA-MB-231 tumor xenografts using two methods: M S P (Herman et al., 1996) and genomic bisulfite sequencing (Graff et al., 1997). Both methods require bisulfite modification of genomic D N A , which converts unmethylated cytosines to uracils while methylated cytosines remain unchanged. M S P utilizes primers designed against D N A sequences containing one or more CpG sites (Figure 4.11); thus amplification of a P C R product is dependent on the methylation status of the target gene. Genomic bisulfite sequencing utilizes primers designed against D N A sequences that are relatively free of CpG sites, but which amplify a region of D N A containing a high density of C p G sites (Graff et al., 1997) (Figure 4.11). Hence the methylation status of the target gene can be inferred from the D N A sequence. Importantly, C p G methylation in this region of D N A has been shown to be tightly associated with loss of E-cadherin expression (Graff et al., 1997). B y both M S P and genomic bisulfite sequencing, E-cadherin promoter methylation in XNotch4 tumors was found to be reduced compared to control tumors (Figure 4.13A,B). In a previous genomic bisulfite sequencing study examining the same E-cadherin promoter region as that analyzed in this thesis, reduced E cadherin promoter methylation at 9 of 21 CpG sites was associated with E-cadherin reexpression in MDA-MB-231 cells (Graff et a l , 1997). We detected reduced E-cadherin promoter methylation at 16 of 22 CpG sites examined (Figure 4.13B). These results suggest that XNotch4 re-induces E-cadherin expression by attenuating E-cadherin promoter methylation.  139  A 0.25  E-cadherin promoter primers  .2  0  2  0  1 0.15  - 0.10 MDA-MB-231 MIG Tumor MDA-MB-231 MIGXNotch4HA Tumor 0.05 M control 0.00 U control m MIG (n=5) E=3 MIGXNotch4HA (n=5) 5  J  B  MDA-MB-231 In vivo 60  • i MIG (n=35 clones) • MIGXNotch4HA (n=22 clones)  50 c  g 03  |  40 30 20 10 0  J  ll... ..I.i lllll  1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 CpG site Figure 4.13 (Part 1): Notch signal inhibition attenuates E-cadherin promoter methylation, but does not induce a generalized demethylation of the genome. (A) M S P to assess the methylation status of the E-cadherin promoter in MDA-MB-231 tumors. Genomic DNA was isolated from M D A - M B 231 tumors, bisulfite-treated, and P C R performed using primers specific for methylated (M) or unmethylated (U) E-cadherin promoter. Amplified M and U products were quantitated by densitometry and expressed as an M/U ratio. Data shown represent mean ± standard error. * P < 0.001. (B) Genomic bisulfite sequencing to assess methylation status of the E-cadherin promoter in MDA-MB-231 tumors. Bisulfite-treated ge