UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Regulatory mechanisms governing mammary epithelial and progenitor cell growth Burleigh, Angela 2011

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

Item Metadata

Download

Media
24-ubc_2011_fall_burleigh_angela.pdf [ 154.26MB ]
Metadata
JSON: 24-1.0071902.json
JSON-LD: 24-1.0071902-ld.json
RDF/XML (Pretty): 24-1.0071902-rdf.xml
RDF/JSON: 24-1.0071902-rdf.json
Turtle: 24-1.0071902-turtle.txt
N-Triples: 24-1.0071902-rdf-ntriples.txt
Original Record: 24-1.0071902-source.json
Full Text
24-1.0071902-fulltext.txt
Citation
24-1.0071902.ris

Full Text

REGULATORY MECHANISMS GOVERNING MAMMARY EPITHELIAL AND PROGENITOR CELL GROWTH by Angela Burleigh B.Sc., The University of British Columbia, 2005 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Pathology and Laboratory Medicine) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  June 2011 © Angela Burleigh, 2011  Abstract The processes involved in mammary gland development are intimately linked with those that drive breast oncogenesis. Regulation of growth, tissue polarity and genome stability are a few of the factors the maintain homeostasis in breast epithelium and prevent malignant progression. In this work, a series of clonal 184-hTERT cell lines were generated that modeled the in vitro growth characteristics of bi-potent mammary progenitor cells; they were dependent upon fibroblasts for low-density growth and formed dual-lineage acini in 3D culture. These lines were subsequently used in a genome-wide siRNA screen to identify the factors that regulate fibroblast-driven epithelial cell growth. Fibroblasts constitute the majority of cells within stroma, which plays a major role in supporting mammary progenitor cell growth. From this screen, 49 surface and secreted factors were identified that putatively transduce the signals emanating from the fibroblasts that are required for epithelial cell growth. These factors were more potent than any of the previously described growth factor receptors. When assessed in primary tissue, Gpr39, Scarb2, Ntn1, Efna4, Nptx1, and Ctnna1 were found to have the greatest effect on overall progenitor cell growth, while SerpinH1 differentially suppressed luminal progenitor cells, and Nkain4 and Kcnj5 differentially suppressed bi-potent progenitor cells. Further profiling of these lines identified the planar cell polarity protein Celsr1 as differentially regulated under fibroblast-dependent conditions. Silencing of Celsr1 increased the number of bi-potent progenitor cells detected in the colonyforming assay. Furthermore, it induced branching morphogenesis within normally spherical acini and disrupted the apical polarity of these structures in 3D culture. Within this system, Celsr1 is suspected of signaling through Shisa4. This is the first description of a noncanonical Celsr1 interactor. Finally, a curious variant line was identified amongst the collection of 184-hTERT cells generated for this work. This line harbours mitotic spindle and cell cycle checkpoint defects, and rapidly gains chromosome 20 during passaging. Through an elimination process, de novo promoter hypomethylation and subsequent overexpression of CENPI was identified as likely being responsible for this phenotype. This is the first description of CENPI deregulation and one of a few descriptions of gene promoter hypomethylation resulting in genome instability.  ii  Preface Co-authorship statement Chapter 3 The parental 184-hTERT cell line cloned in this chapter was generated by Dr. Carl Barrett (Laboratory of Molecular Carcinogenesis at the National Institute of Environmental Health Sciences). Adrian Wan and Alaeddin Tafech assisted with the Southern Blotting after I prepared the genomic DNA and probe. The automated immunohistochemical staining and the multiplex fluorescence in situ hybridization were performed at the Centre for Translational and Applied Genomics (BC Cancer Agency). The array-based comparative genomic hybridization was performed by NimbleGen after I extracted the genomic DNA. Renal capsule xenografting was performed with Dr. Peter Eirew (Terry Fox Laboratory, BC Cancer Agency). I conducted all of the remaining experiments and designed the experiments with intellectual input from Dr. Samuel Aparicio, Dr. David Hunstman and Dr. Connie Eaves. Chapter 4 Steven McKinney performed the statistical analysis for the Affymetrix gene expression studies and the RNA interference screening. Steven Poon optimized the InCell Profiler program and analyzed all of the images from the siRNA screens and catalogued the data for storage (with assistance from Ken Fong). John Fee programmed the robotic transfections and assisted in performing them. Primary mammary tissue was collected and processed into ‘A’ pellets by Darcy Wilkinson (Terry Fox Laboratory). Lentiviral shRNA constructs were packaged and titered by Adrian Wan and Viola Ng. They both assisted with the lentiviral infections after I prepared the cells. I conducted the post-infection processing outside of the Level III biocontainment facility. Affymetrix gene expression arrays were performed at the Centre for Translational and Applied Genomics after I prepared the RNA. The METABRIC data plots were generated by Dr. Samuel Aparicio. I conducted all of the remaining experiments. Steven McKinney authored a portion of section 4.2.5 discussing the  iii  statistical analysis of the siRNA screen data. Experiments were designed in consultation with Dr. Samuel Aparicio with feedback from Dr. Connie Eaves. Chapter 5 The primary mammary tissue used in this chapter was collected and processed into ‘A’ pellets by Darcy Wilkinson. Adrian Wan packaged and titered the lentiviral shRNA constructs used in this chapter, and conducted the infections after I prepared the cells. I conducted the post-infection processing outside of the Level III facility. The Affymetrix gene expression array hybridization was performed at the Centre for Applied Genomics (Hospital for Sick Children) after I prepared the RNA. Steven McKinney performed the statistical analysis for the Affymetrix arrays. The animals used in these experiments were housed at the Rederivation Facility at the University of British Columbia. Teresa Ruiz de Algara assisted with the animal husbandry and breeding. I conducted all of the remaining experiment and designed the experiments in consultation with Dr. Samuel Aparicio. Chapter 6 The whole-transcriptome shot-gun sequencing libraries and the methylation sequencing libraries were constructed and sequenced at the Genome Sciences Centre (BC Cancer Agency) by Yongjun Zhao and Dr. Martin Hirst. The transcriptome data was analyzed by Gillian Leung under the supervision of Sohrab Shah. The methylation data was analyzed by Drs. Allen Delaney, Martin Hirst, and Samuel Aparicio. The Affymetrix SNP 6.0 arrays were hybridized by AROS Applied Biotechnology after I extracted the genomic DNA. The SNP 6.0 data was analyzed by Gavin Ha. The Affymetrix gene expression arrays were hybridized at the Centre for Applied Genomics after I prepared the RNA. The data was analyzed by Dr. Samuel Aparicio. Fluorescence in situ hybridization after the drug treatments was performed by the Centre for Translational and Applied Genomics. The spheroid thickness plots from the 3D Matrigel data after 5-aza-2’deoxycytidine treatments were generated by Steven Poon. I conducted all of the remaining experiments. The sequencing experiments were designed in consultation with Dr. Samuel Aparicio and Dr. Martin Hirst. The remaining experiments were designed with Dr. Samuel Aparicio, with intellectual input from Dr. David Huntsman and Dr. Connie Eaves.  iv  Research ethics approval The work conducted in the course of this thesis was approved by the University of British Columbia Research Ethics Board. UBC Research Ethics Certificate numbers are H06-00210, A10-0009, A07-0524 and A06-0423.  v  Table of contents Abstract.................................................................................................................................... ii	
   Preface..................................................................................................................................... iii	
   Table of contents .................................................................................................................... vi	
   List of tables ............................................................................................................................ x	
   List of figures.......................................................................................................................... xi	
   Acknowledgements .............................................................................................................. xiv	
   Dedication .............................................................................................................................. xv	
   1 Introduction......................................................................................................................... 1	
   1.1	
   The human mammary gland ................................................................................................. 1	
   1.1.1	
   Mammary gland development........................................................................................... 1	
   1.1.2	
   Concept of mammary stem and progenitor cells............................................................... 2	
   1.1.3	
   Detection of multi-potent human mammary epithelial cells in vitro ................................ 4	
   1.1.4	
   Factors regulating mammary development ....................................................................... 6	
   1.1.5	
   Cell polarity in mammary development ............................................................................ 8	
   1.2	
   Breast cancer ......................................................................................................................... 10	
   1.2.1	
   History of breast cancer treatment and subtypes............................................................. 10	
   1.2.2	
   Tumour heterogeneity ..................................................................................................... 13	
   1.2.3	
   Cell of origin ................................................................................................................... 15	
   1.3	
   Epigenetics............................................................................................................................. 16	
   1.3.1	
   Molecular mechanisms of epigenetics............................................................................. 16	
   1.3.2	
   Epigenetics in the mammary gland ................................................................................. 18	
   1.4	
   Thesis objectives.................................................................................................................... 19	
    2 Materials and methods ..................................................................................................... 24	
   2.1 	
   Cell culture ........................................................................................................................... 24	
   2.1.1	
   Culture of epithelial cell lines.......................................................................................... 24	
   2.1.2	
   Preparation of irradiated NIH 3T3 feeder cells ............................................................... 25	
   2.1.3	
   Derivation of the 184-hTERT clonal lines ...................................................................... 25	
   2.1.4	
   3D Matrigel culture ......................................................................................................... 26	
    vi  2.2	
   Primary mammary tissue..................................................................................................... 26	
   2.2.1	
   Dissociation of primary mammary tissue........................................................................ 26	
   2.2.2	
   Colony-forming assays.................................................................................................... 27	
   2.2.3	
   Fluorescence activated cell sorting.................................................................................. 27	
   2.3	
   Molecular techniques............................................................................................................ 28	
   2.3.1	
   Nucleic acid extractions .................................................................................................. 28	
   2.3.2	
   Southern Blotting ............................................................................................................ 29	
   2.3.3	
   Protein extraction and Western Blotting ......................................................................... 29	
   2.3.4	
   Reverse transcription and quantitative polymerase chain reaction ................................. 30	
   2.3.5	
   Immunohistochemistry and immunofluorescence........................................................... 32	
   2.3.6	
   Fluorescent in-situ hybridization..................................................................................... 33	
   2.3.7	
   Vectorette PCR................................................................................................................ 33	
   2.3.8	
   5-bromo-2-deoxyuridine cell cycle profiling .................................................................. 35	
   2.3.9	
   Microarrays ..................................................................................................................... 35	
   2.4	
   RNA interference .................................................................................................................. 36	
   2.4.1	
   Transfection of cell lines with small-interfering RNAs .................................................. 36	
   2.4.2	
   Viral packaging and titering of lentiviral short-hairpin RNAs........................................ 36	
   2.4.3	
   Transduction of lentiviral short-hairpin RNAs ............................................................... 37	
   2.5	
   Animal experiments.............................................................................................................. 38	
   2.5.1	
   Renal capsule xenografts................................................................................................. 38	
   2.5.2	
   Celsr1 deleted mice ......................................................................................................... 39	
    3 Derivation of clonal cell lines with properties reminiscent of mammary progenitor cells ......................................................................................................................................... 40	
   3.1	
   Introduction........................................................................................................................... 40	
   3.2	
   Results .................................................................................................................................... 42	
   3.2.1	
   Generation of the 184-hTERT parental cell line and clonal derivatives ......................... 42	
   3.2.2	
   Identification of 14 independent derivative clones by integration site analysis.............. 44	
   3.2.3	
   184-hTERT lines have an antigen profile typical of myoepithelial cells........................ 44	
   3.2.4	
   Clonal derivatives are cytogenetically normal and non-tumourigenic............................ 45	
   3.2.5	
   Clonal derivatives form structures in 3D culture reminiscent of normal mammary acinar units ................................................................................................................................. 47	
   3.2.6	
   184-hTERT lines display a dose-dependent growth response to feeder cells ................. 48	
   3.3	
   Discussion .............................................................................................................................. 48	
    vii  4 Systematic identification of signal transducers functionally required for mammary progenitor cell growth .......................................................................................................... 62	
   4.1	
   Introduction........................................................................................................................... 62	
   4.2	
   Results .................................................................................................................................... 68	
   4.2.1	
   Low-density growth of 184-hTERT cells can be supported by bovine pituitary extract or feeder cells....................................................................................................................... 68	
   4.2.2	
   Growth signals provided by bovine pituitary extract are different than those provided by feeder cells....................................................................................................................... 70	
   4.2.3	
   184-hTERT cells are susceptible to siRNA knockdown while irradiated NIH 3T3 cells are not .............................................................................................................................. 73	
   4.2.4	
   Development of a co-culture assay for high-content screening ...................................... 76	
   4.2.5	
   High-content screening identifies 49 signal transducers required for mammary epithelial cell growth ....................................................................................................................... 80	
   4.2.6	
   Signal transducers have variegated effects on 3D morphogenesis.................................. 85	
   4.2.7	
   Bi-potent and luminal progenitor cells are differentially affected by signal transducers required for epithelial cell growth ................................................................................... 87	
   4.3	
   Discussion .............................................................................................................................. 89	
    5 Celsr1 regulates mammary progenitor cell growth and branching morphogenesis 127	
   5.1	
   Introduction......................................................................................................................... 127	
   5.2	
   Results .................................................................................................................................. 130	
   5.2.1	
   Celsr1 differentially regulates mammary progenitor cells ............................................ 130	
   5.2.2	
   Silencing of Celsr1 induces epithelial cell branching in 3D Matrigel culture .............. 131	
   5.2.3	
   Celsr1 is required to establish apical polarity in 3D Matrigel culture........................... 133	
   5.2.4	
   Celsr1 may coordinately regulate branching with Shisa4 ............................................. 134	
   5.2.5	
   Preliminary phenotype of Celsr1 loss in mouse mammary glands ............................... 135	
   5.3	
   Discussion ............................................................................................................................ 138	
    6 Epigenetic regulation of genome stability in non-transformed mammary epithelial cells ....................................................................................................................................... 157	
   6.1	
   Introduction......................................................................................................................... 157	
   6.2	
   Results .................................................................................................................................. 163	
   6.2.1	
   184-hTERT cells have the propensity to gain chromosome 20 .................................... 163	
   6.2.2	
   Rapid gainers of chromosome 20 have a gene expression profile dominated by genes involved in cell cycle regulation and DNA damage repair ........................................... 165	
   viii  6.2.3	
   Mitotic spindle and cell cycle checkpoint defects are present in rapid gainers of chromosome 20 ............................................................................................................. 166	
   6.2.4	
   Rapid gainers of chromosome 20 have disrupted 3D acinar morphogenesis................ 167	
   6.2.5	
   The rapid gainer phenotype is not caused by the integration of viral telomerase ......... 168	
   6.2.6	
   The genotype of the cell of origin does not lead to the rapid gainer phenotype ........... 170	
   6.2.7	
   Epigenetic alterations contribute to the rapid gainer phenotype ................................... 171	
   6.2.8	
   MRE-sequencing reveals regions of differential methylation in the rapid gainer cells 172	
   6.3	
   Discussion ............................................................................................................................ 174	
    7 Discussion and future directions .................................................................................... 203	
   7.1	
   Major contributions............................................................................................................ 203	
   7.2	
   Implications and future directions .................................................................................... 205	
   7.2.1	
   New mediators of mammary progenitor cell regulation................................................ 205	
   7.2.2	
   Improved progenitor cell isolation strategies ................................................................ 206	
   7.2.3	
   Novel signaling paradigm for planar cell polarity genes............................................... 206	
   7.2.4	
   Potential strategy for maintaining genome stability...................................................... 207	
   7.3	
   Concluding comments ........................................................................................................ 208	
    Bibliography ........................................................................................................................ 210	
   Appendices........................................................................................................................... 235	
   Appendix A – Gene level changes when 184-hTERT cells are cultured with...…….....235 Appendix B – Exon level changes when 184-hTERT cells are cultured with.…….......238 Appendix C – RT-QPCR validation of targets reproduced by the secondary screen.....242 Appendix D – RT-QPCR validation of 184-hTERT-L9 cell lines with stably…….…..248 Appendix E – RT-QPCR validation of shRNA lentiviral constructs completely …......250 Appendix F – Expression differences between 184-hTERT-L2 and 184-hTERT-L9....251 Appendix G – Primers used for RT-QPCR validation of expression differences in .....255 Appendix H – Expression values from RT-QPCR surrounding 184-hTERT-L2….......258 Appendix I – List of putative single nucleotide variants specific to the 184-hTERT....259 Appendix J – List of putative small insertions and deletions specific to the………….273 Appendix K – List of putative small insertions and deletions specific to the…………277 Appendix L – List of primers used for the validation of the putative genomic to …....279  ix  List of tables Table 3.1 - Classification of clonal lines based on telomerase integration site ...................... 60	
   Table 3.2 – Plating efficiency of clonal lines in 3D matrigel culture ..................................... 61	
   Table 5.1 – Program of coordinate gene regulation upon Celsr1 silencing.......................... 156	
   Table 6.1 - Structural and numerical karyotypic abnormalities in 184-hTERT-L2 ............. 202	
    x  List of figures Figure 1.1 – Architecture of the human mammary gland....................................................... 22	
   Figure 1.2 – Mammary epithelial cell hierarchy..................................................................... 23	
   Figure 3.1 - Generating the 184-hTERT cell lines ................................................................ 50	
   Figure 3.2 – Southern Blot Hybridization of the 184-hTERT clones..................................... 51	
   Figure 3.3 – 184-hTERT clonal lines express markers typically associated with myoepithelial cells .................................................................................................................... 52	
   Figure 3.4 – Status of clinically relevant markers in 184-hTERT clonal lines....................... 53	
   Figure 3.5 – 184-hTERT clonal lines are cytogenetically normal.......................................... 54	
   Figure 3.6 – 184-hTERT clonal lines are karyotypically normal ........................................... 55	
   Figure 3.7 – 184-hTERT clonal lines are non-tumourigenic.................................................. 56	
   Figure 3.8 – Retention of differentiation capacity in clonal 184-hTERT lines ...................... 57	
   Figure 3.9 – Colony formation changes relative to irradiated NIH 3T3 density .................... 59	
   Figure 4.1 – Differential response of 184-hTERT cells to varying growth medias ............... 96	
   Figure 4.2 – 184-hTERT cells require close contact with feeder cells for growth in colony forming assays ................................................................................................... 98	
   Figure 4.3 – Differential signaling response of 184-hTERT cells to fibroblasts and bovine pituitary extract .................................................................................................. 99	
   Figure 4.4 – RNA interference is ineffective in irradiated NIH 3T3 cells ........................... 101	
   Figure 4.5 – RNA interference is effective in 184-hTERT cells .......................................... 103	
   Figure 4.6 – Optimized plating density and order can reduce assay variability ................... 104	
   Figure 4.7 – Plate effects produce variability across 96-well plates..................................... 106	
   Figure 4.8 – Acquisition parameters affect the dynamic range of the assay ........................ 108	
   Figure 4.9 – Total nuclear count accurately reflects the 184-hTERT count in co-culture ... 110	
   Figure 4.10 – High-throughput screening workflow ............................................................ 112	
   Figure 4.11 – Primary screening identified 388 potential targets transducing the signals required for fibroblast-dependent epithelial cell growth ............................... 114	
   Figure 4.12 – Secondary screening identified 140 reproducible candidate signal transducers ........................................................................................................................ 116  xi  Figure 4.13 – Signal transducers that are functionally required for mammary epithelial cell growth ............................................................................................................ 117	
   Figure 4.14 – RIPK2 and ACE2 are differentially expressed in breast cancer subtypes...... 119	
   Figure 4.15 – Signal transducers required for fibroblast-dependent 2D growth have varying effects in 3D culture....................................................................................... 121	
   Figure 4.16 – Abrogating Edg7 blocks proliferation and activation of ERK1/2 signaling .. 123	
   Figure 4.17 – Growth signal transducers coordinately or differentially regulate luminal and bi-potent progenitor cells ............................................................................... 125	
   Figure 5.1 – Celsr1 is differentially expressed in fibroblast-dependent growth and in mammary progenitor cell subsets .................................................................... 142	
   Figure 5.2 – Celsr1 negatively regulates the colony-forming ability of bi-potent progenitor cells .................................................................................................................. 144	
   Figure 5.3 – Celsr1 regulates branching in 184-hTERT 3D Matrigel cultures .................... 145	
   Figure 5.4 – Celsr1 silenced through siRNA remains repressed in Matrigel ....................... 147	
   Figure 5.5 – Celsr1 regulates branching in 3D culture of primary mammary cells.............. 148	
   Figure 5.6 – Celsr1 regulates polarity in 184-hTERT 3D Matrigel culture.......................... 149	
   Figure 5.7 – Silencing of Celsr1 in MCF10A cells leads to luminal filling and hyperproliferation in 3D Matrigel culture........................................................ 150	
   Figure 5.8 – Silencing of Shisa4 phenocopies the regulation of branching observed with Celsr1 ................................................................................................................ 151	
   Figure 5.9 – Celsr1 deletion results in hypobranching of the epithelium within mouse mammary gland ............................................................................................... 152	
   Figure 5.10 – Silencing of Celsr1 affects the morphology of 3D structures in mouse cells 154	
   Figure 6.1 – 184-hTERT cells gain extra copies of chromosome 20 during passaging ....... 177	
   Figure 6.2 – 184-hTERT cells do no gain chromosome 8 or 17 during passaging .............. 179	
   Figure 6.3 – Global expression differences exist between 184-hTERT-L2 and 184-hTERTL9 cells.............................................................................................................. 180	
   Figure 6.4 – Biological Process Gene Ontology annotations for genes with increased expression in 184-hTERT-L2 cells.................................................................. 182	
   Figure 6.5 – Centrosome amplification and spindle multipolarity are seen in 184-hTERT-L2 cells .................................................................................................................. 184	
    xii  Figure 6.6 – 184-hTERT cells have a defective G2-M cell cycle checkpoint...................... 186	
   Figure 6.7 – 184-hTERT-L2 cells display increased sensitivity to irradiation..................... 187	
   Figure 6.8 – 184-hTERT-L2 cells form disorganized acinar structures in 3D Matrigel culture .......................................................................................................................... 188	
   Figure 6.9 – Viral integrated telomerase resides at 8q24.3 in 184-hTERT-L2 cells and 7q36.1 in 184-hTERT-L9 cells .................................................................................... 189	
   Figure 6.10 – Retroviral integration does not substantially interfere with normal gene regulation in 184-hTERT-L2 cells................................................................. 191	
   Figure 6.11 – Genetic variation is not detected between the 184-hTERT-L2 and 184-hTERTL9 cells........................................................................................................... 193	
   Figure 6.12 – Gain of chromosome 20 in 184-hTERT-L2 cells can be reversed with 5-aza-2'deoxycytidine................................................................................................. 195	
   Figure 6.13 – Gain of chromosome 20 in 184-hTERT-L2 cells is not affected by histone modification ................................................................................................... 197	
   Figure 6.14 - 5-aza-2'-deoxycytidine corrects the aberrant morphology of 184-hTERT-L2 cells in 3D Matrigel culture ............................................................................ 198	
   Figure 6.15 – Differential CpG island methylation with correlated gene expression changes in the 2 regions of the 184-hTERT-L2 .......................................................... 200	
    xiii  Acknowledgements I am extremely fortunate to have been mentored by Dr. Samuel Aparicio. His passion and dedication for cancer research is infectious. As a supervisor, he is generous with his precious time and overwhelming knowledge. I have the greatest respect and admiration for him, and I will forever be indebted to him for nurturing my intellectual growth and scientific curiosity. Dr. Aparicio is a firm believer in the strength of collaborations, and as such, I have benefited from a close relationship with Dr. Connie Eaves and her laboratory. I am grateful to Dr. Eaves for her advice and mentorship over the years. I also would like to thank my supervisory committee members, Drs. Catherine Pallen, Poul Sorensen, and Keith Humphries, for their excellent guidance and dedication to training. Finally, I would like to thank Dr. David Huntsman for providing me with an outside source of experimental ideas and advice. I am grateful to have been exposed to a wonderful group of colleagues at the BC Cancer Agency. I would like to specifically thank Heidi Hare and Katie Meehan for their help during my early years at the BCCA. Leah Prentice for scientific advice, motivation and much needed distractions. Elizabeth Hajen for maintaining an orderly lab environment. Teresa Ruiz de Algara for her compassion and dedication to her work. Damian Yap and Jason Wong for challenging theoretical discussions. I would also like to extend my gratitude to Steve McKinney, Steven Poon, Guilsa Turashvili, Ursula Kortmann, Arusha Oloumi, Adrian Wan, Viola Ng, Darren Saunders, Cath Ennis, Courteney Lai, Peter Eirew, Afshin Raouf, Maisam Makarem, Darcy Wilkinson, Sohrab Shah, Gavin Ha, Gillian Leung, Ali Bashashati, Sarah Maines-Bandiera, Clara Salamanca, John Fee, Ken Fong, Jenny Song and Jenny Cromarty. I would be remiss not to thank the excellent staff at the Genome Sciences Centre, the Centre for Translational and Applied Genomics and the Terry Fox Laboratory Flow Cytometry Facility. I would like to thank my loving husband, Damien, for his support and unwavering faith in me. I am also grateful to my family, in particular my sister Lisa, for the continuing motivation they provide me with. Finally, thank-you to all of the breast cancer patients who inspire our questions and the people who fundraise for and generously contribute towards cancer research. xiv  Dedication  For my mother, the embodiment of courage and hard work. A never-ending source of inspiration, motivation, and love.  xv  1 1.1 1.1.1  Introduction The human mammary gland Mammary gland development The processes involved in human breast development have been elucidated by  gross and histological examination of normal tissue, cross-species comparisons of developmental processes, and the study of endocrine deficiency diseases. In a survey of 70 human embryos, the staging of mammary bud development during embryogenesis was determined based upon the size of the embryo [1]. Unfortunately, precise gestational age is hard to infer from size measurements due to variations in birth weight. When an embryo is between 7 and 8 mm, the nipple primordium becomes apparent. By 10 mm, a single layer of mesodermal cells develops beneath the nipple, which differentiates into 4 distinct layers of mesenchyme by 14 mm. By 15 mm, the epithelial cells located at the nipple primordium begin to form a nodule that sinks into the mesoderm. This nodule will continue to proliferate to form the mammary bud, which remains fully encased within the developing mesenchyme. By 150 mm, the primitive nipple has developed and subcutaneous adipose tissue deposits become visible. At 170 mm, epithelial outgrowths extend out from the mammary bud and invaginate through the first two mesodermal layers to reach the third layer, which is composed of loosely arranged connective tissue. By 180 mm, the epithelial extensions begin to branch and the central cells lyse to form hollow lumens. Finally, at 330 mm the canalized branches have formed several terminal branches and are clearly composed of 2 epithelial cell layers, which we now know correspond to the luminal and basal cells, surrounded by a layer of basement membrane [2]. At term, roughly 15-20 individual lobes of radially arranged branched epithelium are present that are partitioned by the connective tissue [3]. Until puberty, no identifiable differences exist in the morphology of male and female mammary glands [2]. From infancy until puberty, the mammary gland grows at a rate comparable to the rest of the body. At puberty, extensive remodeling begins in both the stroma and the epithelial cells in the female mammary gland. The epithelium undergoes dichotomous branching, with the distal leading ends of the branches composed of solid, club-shaped 1  terminal tips that are tightly surrounded by stromal fibroblasts [2]. This eventually leads to the formation of progressively branched lobules, which continue to arise during each ovulatory cycle until approximately age 35 [4]. The functional unit of the adult breast is termed the terminal duct lobular unit and is composed of roughly 80 acini per lobule (Figure 1.1A) [5]. Compared to the stroma found between lobules, the stroma within the terminal ductal lobular units is concentrated with fibroblasts that surround each acini (Figure 1.1B) [2, 6]. During pregnancy, the terminal ductal lobular units increase in complexity and size. Under the stimulation of lactogenic hormones, these acini become the secretory units of the breast. The luminal cells become distended as they begin to accumulate lipid vacuoles [4]. After parturition, milk is continually synthesized and released from the acinar luminal cells. This processes is aided by the contraction of the surrounding myoepithelial cells, which forcibly expunges milk through the ducts and to the sinuses beneath the areola where it becomes available to a suckling infant [7]. The accumulation of milk within the cytoplasm of the luminal cells upon the cessation of breastfeeding stimulates involution of the secretory acini. Upon the completion of this post-lactational regression, the mammary gland resembles that of the nulliparous adult with the exception that there is an overall increase in the amount of glandular epithelial tissue [4]. Involution will occur again after menopause whereby the number of lobules and ducts significantly deceases and the stroma is largely replaced by adipose tissue, effectively concluding the lifecycle of the normal mammary gland [2]. 1.1.2  Concept of mammary stem and progenitor cells A hierarchy of stem and progenitor cells exists within the mammary gland, akin to  what is seen in the hematopoietic system (Figure 1.2). Evidence for a mammary stem cell first came from the transplantation of epithelial fragments from donor mice into recipient fat pads that had been cleared of endogenous epithelium [8]. Upon transplantation, the donor epithelium will regenerate a branched ductal structure that fills the entire fat pad within 9-12 weeks and responds to endogenous hormones in a manner that parallels the normal mammary gland. This donor epithelium can retain this regenerative capacity for 5 to 8 serial transplantations, indicating a vast yet finite proliferative potential within these cells [9]. Regenerative potential is equivalent between  2  explants sampled from varying locations along the mammary ductal system and from donors of varying ages [10]. Whilst donor age is irrelevant, advanced age of the recipient fat pad (previously cleared at 3 weeks old) negatively affects regenerative potential [11]. The reduced capacity for ductal elongation of the donor tissue grafted into the fat pads of mice over 12 months of age was rescued by harvesting and re-transplanting the grafted donor epithelium into recipient fat pads of younger mice. This suggests that age-related changes in either the stroma or the hormonal environment greatly influence epithelial growth. It also suggests that stem cells can remain viable after a period of dormancy. A major advancement in the understanding of the mammary stem cell hierarchy came from the prospective identification of single cells with the capacity to repopulate cleared recipient fat pads [12, 13]. This was accomplished by dissociating donor glands into single cell suspensions and using fluorescence activated cell sorting to fractionate cells based upon their expression of the surface markers CD24 and α6-integrin or β1integrin. These cells fit the description of a bona fide adult tissue stem cell as they were able to differentiate into cells of both the luminal and myoepithelial lineage to form a functional gland and were able to self-renew over several serial transplantations [14]. These stem cells expressed markers consistent with a basal phenotype and were able to form colonies in 2D culture and solid spherical acini in 3D Matrigel culture composed of both basal and luminal cells [12, 13]. At the same time, a population of cells with a luminal phenotype that lacked the in vivo repopulating capacity but possessed the ability to form 2D colonies and hollow 3D spheres was also identified. The remaining epithelial cell fractions had no potential for 2D or 3D growth. This pointed to the existence of a stem cell hierarchy within the breast that has been further characterized to include luminal progenitor cells committed to either a ductal or alveolar fate [15]. Experiments within the human mammary system have been complicated by the lack of a truly synonymous transplantation assay to detect human mammary stem cells. Transplantation of dissociated human breast tissue into cleared fat pads of immunocompromised mice led to the formation of small, lackluster, spherical structures that failed to branch despite supplementation with exogenous human hormones [16]. Recently, two in vivo models have been developed that recapitulate important aspects of human mammary epithelial morphogenesis within immunocompromised mice. Within 3  these assays, branched ducts and acinar units are formed that morphologically resemble human tissue and express milk proteins upon stimulation with lactogenic hormones. The first model involves humanizing the cleared mammary fat pad of a recipient mouse with human mammary fibroblasts prior to engraftment of the human mammary epithelial cells [17]. The second model consists of embedding human epithelial cells and fibroblasts into collagen gels that are then grafted under the kidney capsule of immunocompromised mice [18]. While both assays have provided a significant step forward in identifying a human mammary stem cell, controversy exists around whether it is appropriate to label the structure forming cells within these assays as true stem cells. Adult tissue stem cells are generally regarded to be functionally undifferentiated and capable of proliferating, producing a large number of differentiated progeny, self-renewing to maintain their population, and capable of regenerating functional tissue [19]. As such, these structure forming cells do not technically satisfy all of the requirements set forth to define an adult stem cell as they are unable to reform an entire mammary gland, a task that is impossible to achieve or even address in humans. Nonetheless, these repopulating cells remain the most primitive detectable cell within the human mammary epithelial cell hierarchy that had been previously established through in vitro analysis. 1.1.3  Detection of multi-potent human mammary epithelial cells in vitro The formation of differentiated colonies in vitro has been used as a surrogate  method of assaying many types of stem cells and originated in the hematopoietic system [20-23]. The identification of distinct colony types arising from human mammary tissue occurred shortly after the description of techniques to clonally expand these cells [24, 25]. An early description of colonies arising from human breast milk identified three morphological colony types. From this, a differentiation hierarchy was inferred as colonies, now recognized as derived from bi-potent progenitor cells (‘open’), could give rise to the colonies now recognized as derived from luminal progenitor cells (‘elongated’ and ‘cuboidal’) after passaging [26]. The subsequent advent of monoclonal antibodies specific for luminal cells allowed for the first connection between colony morphology and immunophenotype [27]. From reduction mammoplasty tissue, two distinct colony types were routinely identified  4  and classified as either ‘compact’ or ‘spread’ colonies [28]. These corresponded to the ‘open’ and ‘elongated and cuboidal’ colony types, respectively, previously identified from breast milk. ‘Compact’ colonies never stained with 2 monoclonal antibodies derived against luminal cells whereas the ‘spread’ colonies always proved immunoreactive with these antibodies. Prospective isolation of luminal and myoepithelial cells by flow cytometry using epithelial membrane antigen (EMA) and CD10 (CALLA) antibodies, respectively, saw a replication of the same ‘spread’ and ‘compact’ colony types after clonal expansion [29]. 98% of the colonies arising from the EMA–positive luminal cells were of the ‘spread’ colony-type and stained positive for luminal markers cytokeratin 18 and cytokeratin 19, and negative for myoepithelial cell markers cytokeratin 10 and alpha-actin. The converse was seen for the CALLA-positive cells whereby 90% of the colonies were of the ‘compact’ phenotype and displayed the opposite immunophenotype. The expansion of available antibodies for prospective segregation of mammary cells allowed for the first systematic link between colony phenotype and the existence of corresponding progenitor cells [30]. Dual-labeled immunofluorescence for luminal and myoepithelial markers allowed for definitive discrimination between mixed and myoepithelial colony types thus segregating previously described morphologically ‘compact’ colonies into 2 definitive types. Thus, three distinct colony types were identified and proposed to derive from luminal-restricted, myoepithelial-restricted and bipotent progenitor cells based upon the cellular composition of the resultant colonies. Through an extensive search of surface antibodies and sorting parameters, α6-integrin and EpCAM were identified as markers to prospectively segregate luminal-progenitor and bi-potent progenitor cells [31]. Using these markers, prospective fractionation of luminal-progenitor enriched, bi-potent progenitor enriched, mature luminal, and stromal fractions revealed that only the progenitor enriched fractions are able to form 3D structures in Matrigel [32]. The luminal progenitor cells form primarily single-layered structures that are cytokeratin 18 positive. The bi-potent progenitor cells form dense, multi-layered, cytokeratin 5/6 positive spheres that can be induced to express milk proteins. Thus, the 2D progenitor cell classifications are maintained within 3D culturing. It is this bi-potent progenitor enriched cell fraction that contains the putative human 5  mammary stem cells identified through sub-renal xenotransplantation suggesting that all three assays may be detecting the same primitive cell [18]. 1.1.4  Factors regulating mammary development The mammary gland has historically been shown to be inherently sensitive to  changes within the hormonal environment. In humans, hormone deficiency diseases have aided in identifying regulators of mammary growth, but confounding holistic influences can complicate their interpretation. The clearest example of hormonal regulation in the mammary gland comes from the expansive growth of the mammary epithelium elicited by the onset of puberty in females. Experiments with primates identified maturation of the hypothalamus and subsequent pulsatile release of gonadotropin-releasing hormone as initiating the hormonal cascade that leads to the onset of puberty [33]. The delay or absence of puberty in patients with hypogonadotropic hypogonadism, associated with mutations in the gonadotropin-releasing hormone receptor and Gpr54, and Kallmann syndrome, resulting from failed migration of neurons to the hypothalamus, supports the role of the hypothalamic-pituitary-gonadal signaling axis in initiating puberty [34-38]. Gonadotropin-releasing hormone stimulates both the production and release of folliclestimulating hormone and luteinizing hormone from the pituitary gland. These gonadotrophins are then responsible for stimulating the production of the gonadal hormones. In an intricate feedback mechanism, the release of these gondal products in turn regulates the gonadotropin-releasing hormone [39]. Surges of luteinizing hormone occur before the onset of menarche and in the absence of ovulation, resulting in low levels of estrogen in the absence of cyclical progesterone production [40]. This is of particular interest as proliferative activity of both the mammary epithelium and the surrounding stromal cells is stimulated just prior to the onset of puberty [4]. This suggests that these unopposed fluctuations in estrogen levels may drive early adolescent mammary development despite the observation that mammary stem cells do not express the estrogen receptor. It has been shown that the estrogen receptor is not required on epithelial cells for their proliferation in the mammary gland. Rather, it is the presence of the estrogen receptor on the stromal cells that, upon stimulation with estrogen, mediates paracrine signaling events that drive the growth of the epithelium [41, 42]. Within  6  normal virgin human breast tissue, the proliferating epithelial cells are negative for the estrogen receptor and progesterone receptor. However, these proliferative cells exist in close proximity to estrogen receptor positive cells further supporting that the proliferative response of mammary stem cells to estrogen is mediated through paracrine signals [43]. Additionally, low levels of estrogen stimulate the expression of the progesterone receptor in luminal epithelial cells [44]. In progesterone receptor knockout mice, the mammary rudimentary ductal structure develops normally. However, the loss of the progesterone receptor in the epithelium inhibits side-branching and lobuloalveologenesis during pregnancy [45]. The growth response of epithelial cell to these hormones is particularly relevant in the case of hormone replacement therapy and its link to breast cancer. Patients receiving estrogen therapy alone had an increased proliferative index of their normal epithelium, an effect that was further enhanced by the addition of progestin in combination with estrogen [46]. Combination hormone replacement therapy significantly increased breast cancer incidence and mortality in a randomized prospective clinical trial, a risk that declined shortly after cessation of hormone therapy [47]. Recent investigations identified some of the paracrine signal that are responsible for regulating the steroid receptor negative mammary stem cells [48, 49]. These studies confirmed that mammary stem cells in the mouse are estrogen receptor and progesterone receptor negative. The removal of steroid hormones through ovariectomy decreased the proportion of luminal progenitor cells, but not stem cells, present within the mammary gland. Despite the lack of effect on population size, ovariectomy had a significant impact on the ability of these stem cells to repopulate a normal cleared mammary fat pad in an intact female [48]. This suggests that the stem cells hold a memory of their previous environment, which was in this case presumably deficient in paracrine stimulation from hormone responsive cells. Through the use of chemical inhibitors, this effect was found to be solely attributable to the loss of estrogen [48]. This effect was rescued after the recipients of transplanted cells underwent 4 menstrual cycles. Conversely, continual stimulation with estrogen and progesterone in combination, but not alone, lead to an increase in the number of luminal progenitors, the number of stem cells, and the repopulating ability of the mammary stem cells. This combination of steroids lead to the upregulation of Wnt4 and RankL in the luminal progenitor cells and the induction of their  7  cognate receptors Lrp5 and Rank in the stem cells [49]. This was not seen with estrogen treatment alone. This points to Wnt4 and RankL as paracrine effectors of progesteronemediated signaling. Treatment of mice with a RankL inhibitor partially suppressed the in vitro colony forming ability of mammary stem cells [48]. Additionally, loss of Lrp5 has previously been shown to diminish the numbers and in vivo repopulating activity of mammary stem cells in a knockout mouse model [50]. However, this is only part of the story, as other signaling axes are known to regulate mammary stem cell fate. More axes are likely to be discovered along with corresponding improvements in stem cell isolation procedures that increase the purities of the isolated cell populations. 1.1.5  Cell polarity in mammary development The remodelling that occurs in the mammary gland in response to puberty,  pregnancy, and lactation occurs in a highly organized fashion. During stasis, the mammary epithelium consists of a bi-layer of luminal and myoepithelial cells comprising the branched ductal network. The entire duct is encased by a thin layer of unbreached basement membrane [51]. Directly adjacent to the basement membrane is a discontinuous layer of myoepithelial cells upon which the luminal cells rest. The apical surface of the luminal cells is covered in microvilli, which function to effectively increase the surface area of the cells and thus maximize the release of secretory products into the open lumen during lactation. Luminal cells are bound to their lateral neighbors by tight junctions, adherens junctions, desmosomes and GAP junctions [52]. The tight junctions provide a diffusion barrier through which solute and ion diffusion can be controlled. Additionally, they demarcate the boundary between the apical and basolateral position in polarized cells and constrain membrane proteins to the appropriate location [53]. Adherens junctions are primarily composed of cadherin-catenin complexes and function to form and maintain intercellular adhesions [54]. While desmosomes aid in firming these intercellular adhesions, they only appear to be present between luminal cells in virgin glands [52]. Their marked absence in lactating glands and glands that have undergone involution suggests that the inter-luminal desmosomes do not survive the mechanical stress imposed upon these cells during lactation [52]. Desmosomes attach the basal side of the luminal cells to the underlying myoepithelial cells, which are  8  subsequently attached to the basement membrane by hemidesmosomes [55]. During branching morphogenesis, the leading cells that invade through the mammary fat pad are loosely arranged and lack these specialized intercellular junctions [56]. During this process, the establishment of apico-basal polarity in the newly formed epithelium has been shown to precede lumen formation, indicating that apical exposure to luminal space is unlikely to be required for polarization [57]. Rather, it is thought that the instructions directing luminal cell polarity stem from the myoepithelial cells. In vitro, this process has been shown to be dependent upon the production of the basement membrane constituent laminin-1 by myoepithelial cells [58, 59]. Here, the establishment of polarity is mediated through laminin-1 binding to α6-integrin in the luminal cells [59]. This interaction is also required for full functional differentiation of the luminal cells into milk-secreting cells [60, 61]. This process is likely to be considerably more complicated in vivo and presumably involves the integration of cues from diffusible signaling molecules, cell-to-cell contacts, and cell-to-matrix contacts. Despite the lack of clarity surrounding the mechanisms initiating the establishment of polarity in vivo, numerous pathways involved in establishing this response once initiated have been identified. During division, cells require an organized cytoskeleton. This pre-existing cytoskeletal polarity is harnessed to position the appropriate cues required for the establishment of apico-basal polarity. In Drosophila melanogaster, centrosomes are located above the nucleus after division occurs in polarized cells. Microtubules, which are anchored at the centrosomes, project down the lateral sides of the cells forming an inverted basket over the nucleus [62]. Through dynein-mediated transport, Par-3 is shuttled along these microtubules where it accumulates at the apical surface of the cells [62]. This asymmetric distribution of the core polarity protein Par-3 is the first step in establishing epithelial cell polarity [63]. Concurrently, actin-based microvilli found on the apical surface of cells promote the clustering of e-cadherin and β-catenin [64]. Par-3 repositions these complexes to the apico-lateral position where they subsequently form adherens junctions [63, 64]. Independently, Par-3 recruits Par-6 and atypical protein kinase C (aPKC) to the apical membrane where they form a complex. The physical interaction formed between Par-3 and aPKC is disrupted by aPKC phosphorylating Par-3 [65]. The polarity determinant 9  Crumbs out-competes phosphorylated Par-3 for binding to Par-6, thus releasing Par-3 from this complex. Crumbs, and its associated proteins, directly anchor the Par-6/aPKC complex to the apical membrane and help to maintain cell polarity [65]. This separation of Par-3 from Par-6 and aPKC rectifies the observation that, in polarized cells, Par-3 is located in the apico-lateral position, well beneath the apical Par-6/aPKC complex. The Scribble complex, located in the baso-lateral position, antagonizes Par-3 initiated polarity and represses apical identity along the lateral sides of the cells [66, 67]. In the mammary gland, depletion of Par-3 severely inhibits branching morphogenesis. When ducts do develop, they are disorganized, they do not possess the proper end bud structures, and they are often multi-layered and thus reminiscent of hyperplasia [68]. Similarly, depletion of Scribble leads to disorganized, multi-layered mammary ducts with no ductal space. Occasionally, tumours arose in Scribble-depleted murine mammary glands indicating that Scribble may act as a tumour-suppressor gene in human breast cancers [69]. It is anticipated that additional polarity proteins will be identified as crucial to mammary development as the discoveries in Drosophila are continually applied to models of mammalian development. 1.2 1.2.1  Breast cancer History of breast cancer treatment and subtypes At the turn of the 20th century, the accepted treatment for breast cancer was  limited to surgical excision of primary tumours and their recurrences when deemed operable. Early retrospective studies recognized the limitations of surgical intervention in preventing recurrence at all stages of disease, particularly when spread beyond the axillary glands of the breast [70]. The benefit of prophylactic treatment with x-ray radiation or radium in conjunction with surgery was also noted shortly after its introduction in the 1920s [70, 71]. The introduction of chemotherapeutic agents into clinical use following their advent for chemical warfare provided a bank of agents capable of inhibiting cell division [72]. Initial trials with topical application or intratumoural injection of mustard gas to carcinoma lesions lead to successful tumour regression [73]. Following this, trials ensued with additional cytotoxic agents, many of  10  which remain in use today [74].  Further advancement to the concept of  chemotherapeutic interventions came with the introduction of combination therapies, originally tested in acute leukemia, to effectively decrease the dose and thus toxicity of individual agents while maximizing therapeutic response [75]. In the late 19th century, the potential link between hormones and breast cancer began to emerge along with anecdotal evidence of spontaneous regression of breast cancer following menopause (reviewed in [76]). This lead to the limited belief that oophorectomy could be successfully used in the adjuvant therapy of breast cancer, although adoption of this practice was limited due to the high mortality associated with this procedure at that time [77-79]. A greater understanding of the hormonal signaling axes eventually lead to the supposition that metastatic breast cancer could be cured through hypophysectomy, or removal of the pituitary gland [80-82]. However, this failed to produce the desired outcome of ablated estrogen levels [83]. The next major advancement in the hormone treatment of cancer came with the development of the antiestrogenic compound tamoxifen. Shown in vitro to halt the proliferation of estrogenresponsive breast cells, early clinical trails demonstrated significant response rates with limited toxicity in patients with metastatic disease [84, 85]. At the same time, variegated expression of the estrogen receptor across tumours was noted and hypothesized to be predicative for patient response to endocrine therapy [86, 87]. Selecting patients for endocrine therapy based upon estrogen receptor expression improved success rates. However, a significant fraction of estrogen receptor positive tumours still existed that were resistant to endocrine therapy. The expression of progesterone receptor was added as a surrogate marker for a functional estrogen receptor as estrogen responsiveness is required to induce progesterone receptor expression [88, 89]. This resulted in a modest increase in predictive response to endocrine therapy from 60-70% with estrogen receptor alone, to 80% with the combination of receptors [90]. In addition to tamoxifen, aromatase inhibitors designed to block the final step of estrogen biosynthesis have joined oophorectomy and radiation-induced ablation of the ovaries as options for hormone therapy in breast cancer patients [91]. Fourteen years after tamoxifen was discovered, HER-2/neu gene amplification was identified as superior to all of other prognostic factors (barring lymph node status)  11  for shorter time to relapse and decreased overall survival provided a means through which patients could be further stratified [92]. The development of a monoclonal antibody targeting HER-2/neu that functionally blocks proliferation in cells overexpressing this proto-oncogene lead to the development of an effective targeted therapy [93-95]. Used in conjunction with chemotherapy, the humanized form of this antibody (known as trastuzumab) provides significant survival benefits for patients with both early-stage and metastatic HER-2/neu amplified breast cancer [96]. Patient prognosis has thus become intimately linked with the specific treatments available to subgroups of patients that can be provided in addition to the generalized management strategies such as surgical excision, radiation and chemotherapy. The advent of gene expression profiling has allowed for further stratification of breast cancers into 5 intrinsic subtypes termed luminal A, luminal B, HER-2 overexpressing, basal-like and normal-like [97-99]. Gene expression profiling has not been implemented as a routine diagnostic tool due to issues surrounding the cost and complexity of implementing such techniques. However, surrogate immunohistochemistry panels have been developed for these sub-types to aid in indicating prognosis, which ultimately guides treatment strategies [100, 101]. Briefly, the HER-2 overexpressing and the basal-like subtype are hormone receptor negative and are the subtypes with the poorest prognosis [97-99]. Whilst the HER-2 overexpressing subtype can be treated with trastuzumab, chemotherapy remains the mainstay of treatment for the basal-like breast cancers. The increased risk for local recurrence and reduced overall survival noted in basal-like breast cancer can be directly attributable to the lack of alternative therapeutic strategies available to these patients [102]. It is anticipated that further sub-classification of the basal-like breast cancers will occur, which will ultimately aid in the development of targeted therapeutics for this population of patients [103]. In contrast, the luminal breast cancer subtypes are hormone receptor positive and have a more favourable outcome than the Her-2 overexpressing and basallike breast cancers [97-99]. However, the luminal B subtype reflects a poorer outcome when compared to luminal A tumours, and can be distinguished from the latter through its distinctly proliferative signature. Recently, ZNF703 was identified as a novel oncogene in luminal B breast cancer and as such has the capacity to transform and drive  12  proliferation in non-malignant cells [104, 105]. Ultimately, targeted intervention of ZNF703 may lead to improved prognosis for luminal B cancers. One would predict that as more tumours are profiled, the number of breast cancer subtypes would increase and provide a clearer indication of the molecular insults leading to their origin (information that is perhaps currently masked by groupings that remain too heterogeneous). Once the molecular basis of each subtype is determined, targeted therapeutics can be developed and prognosis can be anticipated to improve on a subtype-by-subtype basis. 1.2.2  Tumour heterogeneity Breast cancer is a remarkably heterogeneous disease. Currently, stratification  based upon molecular profiling has identified 5 distinct tumour subtypes that have both prognostic and therapeutic implications [97-100]. Additionally, characterization of immunophenotypically separated subpopulations of breast cancer cells shows that several distinct molecular profiles exist within a single tumour [106]. Through next generation sequencing, intratumoural heterogeneity has been extended to include varying mutational profiles between tumour cells [107]. Nevertheless, the pattern of X-inactivation in human breast carcinomas reveals a predominately monoclonal origin of breast cancer [108, 109]. Two alternate theories have been put forth to explain intratumoural heterogeneity; the Cancer Stem Cell Hypothesis and the Clonal Evolution Model, although they are not necessarily mutually exclusive. Both models surmise that oncogenesis results from an initial transformation event in an individual cell that can then acquire additional mutations during progression. However, these models differ in the way in which they explain acquisition of heterogeneity and the mechanisms behind tumour propagation and therapeutic resistance [110, 111]. In the Clonal Evolution Model, an initial non-descript cell acquires a growth advantage through the accumulation of multiple mutations [112]. The enabling characteristic of genome instability allows for the production of daughter cells with additional mutations during cell division. These mutations may, or may not, provide a growth advantage over the other cells already existing within the tumour [113]. These varying subpopulations expand and contract due to selective pressures resulting in tumour heterogeneity. This is an on-going process and allows for any cell within the  13  tumour to gain the characteristics required for invasion, metastasis and therapeutic resistance. As every cell within the tumour thus has the ability to develop resistance to therapy, complete eradication of every tumour cell during therapy is required to eliminate the possibility of recurrence due to the clonal selection and outgrowth of drug resistant cells [110, 111, 114]. Alternatively, the Cancer Stem Cell Hypothesis is a hierarchical model that surmises that only a small subset of cells, termed the cancer stem cells, have the potential to form and propagate the tumour. A distinction in terms must be made here to avoid the assumption that the cancer stem cell theory is inherently referring to a normal stem cell as the cell of origin for the tumour. Rather, it is referring to the tumour-propagating capacity of a specialized cell within the tumour that has the functional capacity to regenerate the full spectrum of cells found within that tumour [115]. The first description of a prospectively identified cancer stem cell within a solid malignancy was in breast cancer [116]. By fractionating cells based upon their expression of CD24 and CD44, a population was identified that could continually reform the tumour upon serial transplantation in a xenograft model; the remaining tumour cells did not have this ability. This was followed by similar discoveries in brain tumours, head and neck cancers, colon cancer, and pancreatic cancer [117-120]. One cautionary note is that these assays involve the transplantation of human tumour cells into immunodeficient mice. Thus, the cells capable of forming tumours under these conditions may not be relevant to human cancers and may only reflect the ability of a sub-population of cells to adapt to and colonize a foreign environment. A high percentage cells isolated from a murine B-cell lymphoma model will successfully transplant into congenic mice, suggesting that species specific differences may be influencing the supposed rarity of human cancer stem cells [121]. Additionally, refinements to the transplantation techniques have improved the efficacy of detection of human melanoma-initiating cells to a point where 27% of unselected primary melanoma cells can form tumours from single cells when transplanted into mice [122]. It is not clear whether this will be applicable to other tumour types. Despite these caveats, this theory proffers that intratumoural heterogeneity arises from differentiation of the cancer stem cells into phenotypically divergent daughter cells that lack the capacity for limitless proliferation and are no longer able to reform all of the  14  cells types of the tumour. Cancer stem cells are proposed to be inherently resistant to chemotherapeutic and radiological interventions [123, 124]. From this, it is presumed that current therapeutic interventions preferentially eliminate the differentiated cells within a tumour leaving the cancer stem cells viable and capable of causing a recurrence. If this is case, strategies need to be designed to specifically target these cells to achieve disease-free survival [111]. 1.2.3  Cell of origin The presence of intertumoural heterogeneity across patient samples is thought to  arise through two main mechanisms, neither being mutually exclusive. The accumulation of distinct sets of genetic or epigenetic changes with the same target cell in different patients can presumably lead to varying tumour phenotypes derived from the same cell of origin. Alternatively, identical aberrations in different cells of origins can also potentially lead to diverse histopathology [115]. Extrinsic influences emanating from activated fibroblasts, extracellular matrix and inflammatory responses may further contribute to intertumoural heterogeneity [125]. Comparative expression profiling of breast cancer subtypes and fractionated populations of normal mammary cells led to the observation that, despite their name, basal-like breast cancers most closely resemble luminal progenitor cells [32]. Furthermore, normal mammary tissue taken from women harbouring BRCA1 mutations displayed an expanded population of luminal progenitor cells that possess different growth requirements in vitro than normal luminal progenitor cells [32]. Using population specific promoters, BRCA1 was deleted through cre-mediated recombination in either luminal progenitor cells or mammary stem cells in a p53 heterozygous deleted mouse [126]. When BRCA1 was deleted in the luminal progenitor cells, the resultant tumours phenocopied the basal-like tumours that normally arise in BRCA1 deficient patients. Deletion in the mammary stem cells resulted in the production of malignant adenomyoepitheliomas, a rare tumour in humans that is not correlated with BRCA1 deletion in human. Conversely, constitutive Notch activation in both mammary stem cells and luminal progenitor cells prior to implantation into a cleared mammary fat pad led to the development of identical hyperplastic nodules composed exclusively of luminal  15  cells [127]. Thus, it appears that the relative contribution of the cell of origin and specific genetic aberrations to tumour heterogeneity is context dependent and is likely a combinatorial process. It should be noted that continual stimulation of mice with estrogen and progesterone led to an increase in the number of both the luminal progenitors cells and stem cells [48, 49]. If stem and/or progenitor cells are revealed to be the cells of origin during oncogenesis, this fluctuation with hormone levels would explain the epidemiological finding that a lower number of menstrual cycles correspondingly results in lower breast cancer risk [128]. Additionally, the transient expansion of mammary stem cells during pregnancy would explain the short-term increased risk of breast cancer that occurs immediately following parity [48, 129]. 1.3 1.3.1  Epigenetics Molecular mechanisms of epigenetics Epigenetic traits are defined as heritable characteristics outside of the primary  DNA sequence that are transmissible through mitosis or meiosis [130]. They are involved in the temporal and spatial regulation of genes during development and thus must remain stable after lineage commitment of specialized cell types for multiple generations. There are 3 main mechanisms that govern epigenetic change: DNA methylation, histone modifications, and RNA interference through non-coding RNAs [131]. Together, these mechanisms comprise the epigenetic landscape that, upon disruption, can lead to a host of cellular changes including altered gene expression, genome instability, DNA recombination, and defects in DNA replication and repair [132135]. DNA methylation is the best-characterized epigenetic change in the context of malignant progression and consists of the addition of a methyl group onto the 5-carbon of cytosine. This occurs through the action of DNA methyltransferases, which utilize Sadenosyl methionine as the methyl group donor. De novo methylation of DNA occurs through the action of DNA methyltransferase 3a and 3b [136]. Maintenance of methylation patterns during replication is conducted by DNA methyltransferase 1 [137]. Often, this occurs in cytosines located directly 5’ of guanines, or CpG sites, which are  16  preferentially enriched in motifs known as CpG islands [138]. CpG islands are found within the promoters of 72% of genes and are largely hypomethylated within the germline [139]. Methylation around transcription start sites is negatively correlated with gene expression, whereas methylation around transcription termination sites and within genes is positively correlated with gene expression [140]. Aberrant DNA methylation is a gradual and reversible process that can occur during malignant progression and is commonly associated with a global decrease in gene expression compared to adjacent normal tissues. Since the discovery of promoter hypermethylation in the CpG island of the retinoblastoma tumour suppressor, methylation-induced silencing has been described in a host of other genes whose expression is required to prevent malignant progression [141]. In breast and colon cancer, an average of 11 DNA sequence mutations in protein coding genes are presumed to exist per tumour [142]. Conversely, hundreds of regions of de novo promoter hypermethylation exist resulting in a dramatic change in the transcriptional landscape of a tumour cell [143]. These discoveries have revived interest in the use of DNA demethylating agents, such as 5-aza-2’-deoxycytidine, in clinical cancer therapy [144, 145]. In addition to direct modification of the DNA through methylation, changes to the histones surrounding the DNA can also control gene expression through the regulation of chromatin structure. Differing degrees of methylation and acetylation of the lysine residues on histones alter the chromatin state of DNA and thus functionally lead to altered gene transcription [146]. Acetylation acts in part by neutralizing the positive charge of lysines, thus diminishing the ability of histones to bind DNA through electrostatic interactions and allowing for chromatin expansion. This allows regulatory proteins to access DNA and initiate transcription [147]. Additionally, it signals the recruitment of bromodomain-containing transcription complexes to the DNA providing an additional mechanism through which acetylated lysines increase gene expression [148]. Histones can be reversibly methylated at both lysine or arginine residues, with 3 known methylation states for lysine residues and 2 for arginine [149]. While complex combinations of marks are associated with chromatin states, 3 sites that are commonly associated with inactive chromatin configuration and transcriptional silencing are histone-H3-lysine-9, histone-H3-lysine-27, and histone-H4-lysine-20. Histone-H3-  17  lysine-4, histone-H3-lysine-36, and histone-H3-lysine-79 are often implicated in activation of transcription [150]. Whilst posttranslational histone modifications are proposed to extend the information encoded within DNA by regulating chromatin states, alternate functions aside from organizing nuclear architecture have emerged [151]. Histone lysine methylation has been directly implicated in coordinating the recruitment of DNA damage response proteins to sites of double-stranded breaks and plays a role in maintaining genome stability [152]. A distinct pattern of histone modifications is also used to mark the chromatin around centromeres, clearly distinguishing it from euchromatin or heterochromatin. This allows for proper kinetochore assembly and cell division [153]. With advances in methods to both detect and alter epigenetic states within cells, it is likely that additional roles for epigenetics in both normal development and disease will continue to emerge. 1.3.2  Epigenetics in the mammary gland Initial work characterizing epigenetic regulation during mammary development  focused around the ability of the gland to produce milk in response to lactogenic hormones. Hypomethylation of the β- and γ-casein genes in the lactating mouse mammary gland is observed and correlated with an up-regulation in gene expression compared to non-lactating glands and liver tissue [154]. Similarly, in the rat mammary gland, expression of the κ-casein gene during lactation or following prolactin treatment is associated with hypomethylation of the region surrounding the gene [155]. Continued investigation into the functional differentiation of mammary cells subsequently revealed roles for chromatin conformation and histone modifications in the expression of a complement of milk proteins [156-158]. DNA methylation patterns specific to mammary progenitor cells during lineage commitment have emerged as a potential mechanism regulating the associated global gene expression changes amongst these cells [159]. A selection of these methylation differences remained conserved when assessed in analogous cell types within malignant tissue. The interplay between these changes and known transcription factors that regulate self-renewal and lineage commitment remains  18  uncharacterized, but will likely emerge as the epigenetic programmes within these cells continue to be explored. In up to 50% of breast cancer cases, global DNA hypomethylation is evident in tumour tissue when compared to directly adjacent normal tissue [160, 161]. This is mostly observed in repetitive DNA sequences, which are normally methylated to prevent retrotransposons from reactivating and inserting elsewhere in the genome [162]. This can ultimately result in insertional mutagenesis and genome instability. Hypomethylation of promoter regions is relatively rare in breast cancer. A notable exception is the multidrug resistance 1 gene, whose promoter was hypomethylated in 47 of 100 patients with invasive ductal carcinoma [131]. More often, promoter hypermethylation is observed as a frequent and early event in breast cancer and has been implicated in the silencing of multiple tumour suppressor genes [163]. Levels of post-translational histone modifications have also recently been associated with breast cancer. Specifically, global loss of histone acetylation is an independent prognostic factor for shorter disease free survival [164]. Despite establishing a role for aberrant DNA methylation and histone modifications in breast cancer, the mechanisms driving these epigenetic changes remain unclear. 1.4  Thesis objectives Breast tissue homeostasis and specifically the factors regulating growth and  differentiation of mammary epithelium are not completely understood. The goal of this thesis is to address the global regulation of mammary epithelial cell growth and differentiation. This was accomplished by testing the following specific hypotheses. 1) Systematic RNA interference screening will identify undiscovered signaling pathways that must be activated to allow for mammary epithelial cell growth. 2) The planar cell polarity protein Celsr1, identified through gene expression profiling of fibroblast driven epithelial cell growth, plays a crucial role in mammary gland development. 3) Genome instability in normal mammary epithelial cells in culture can arise through inherent variations in the original cell population prior to cell immortalization.  19  These hypotheses were addressed through the use of a model system developed specifically for this work and through systematic, genome-wide methods for identifying the genes important in these processes. In what was initially an effort to uncover the role of fibroblasts in regulating mammary progenitor cell growth, a system was identified that allowed for the study of fibroblast-driven epithelial cell growth regulation, the role of polarity genes in mammary development and genome stability in non-transformed cells. These topics are unified in this thesis by their physiological relevance to the maintenance of breast tissue homeostasis. Chapter 3 begins with the characterization of a series of immortalized clonal mammary epithelial cell lines that were established to model the growth of primary mammary cells in culture. These lines were karyotypically normal at low passage numbers and expressed markers consistent with the myoepithelial cells of normal breast tissue. When placed into 3D culture in Matrigel, these lines produced polarized spherical structures resembling normal mammary lobular units. In these conditions, luminal markers were expressed in the inner ring of cells signifying that, like progenitor cells, these lines retain differentiation capacity. When plated at clonogenic densities, these lines exhibit fibroblast-dependency, a growth property similarly demonstrated by mammary progenitor cells. These cell lines provided a suitable alternative to primary mammary tissue and served as the basis for the work conducted in Chapter 4 and Chapter 6. Described within Chapter 4 is the development of a high-content screen to identify genes that regulate fibroblast-dependent mammary epithelial cell growth. The inability to replace fibroblasts in mammary epithelial cell growth with defined growth factors led to the hypothesis that a series of undiscovered signaling pathways must be activated to allow these cells to grow. Uncontrolled epithelial cell growth is a feature of breast cancer, and thus delineating these unknown signaling axes would have implications beyond understanding normal biology. Through systematic RNA interference screening, we identified 49 potential signal transducers in mammary epithelial cells with a greater effect on mammary cell growth than any of the previously described growth factor receptors. The applicability of these signal transducers to mammary progenitor cell growth was then evaluated. In doing so, differential and common regulators of both bi-potent and luminal progenitor cells were found.  20  Differential regulatory effects were confirmed in 2D and 3D growth, suggesting that any potential growth regulator must be assessed within multiple contexts to fully understand its role in normal physiology. In Chapter 4, Celsr1, a non-classical type cadherin with a central role in regulating planar cell polarity, was identified as up-regulated when 184-hTERT cells were grown in co-culture with fibroblasts. Based upon the imperativeness of stromalepithelial cell interactions in the mammary gland, and the importance of Celsr1 in other organ systems, it was hypothesized that Celsr1 would play a crucial role in mammary gland development. Celsr1 was discovered to regulate both bi-potent progenitor cell growth and branching morphogenesis in the mammary gland. This work also introduced the existence of a previously undiscovered signaling axis for Celsr1 by discovering Shisa4 as a potential partner for Celsr1 in controlling branching morphogenesis. While characterizing the cells lines generated in Chapter 3, a curious clonal variant was discovered that rapidly gains chromosome 20 during passaging. Several other defects in these cells constituted this rapid gainer phenotype. These lines had a single integration site for the viral telomerase used in their immortalization, which differed from the integration sites found in the other normal comparator lines. As these lines were otherwise identical, this led to the hypothesis that the rapid gainer phenotype was either due to a genetic disruption caused by the viral integration site or an inherent variation within the original cell population prior to immortalization. Through the investigations described within this chapter, an epigenetic change was identified in the promoter of CENPI, likely present in the original cell population, and was thought to be responsible for generating the rapid gainer phenotype. This not only implies a role for CENPI in malignant progression, but also describes a heterogeneous change in DNA methylation within normal mammary cells that leads to cell cycle checkpoint defects and ascendance to aneuploidy.  21  A  B  Figure 1.1 – Architecture of the human mammary gland A) Schematic drawing of an individual lobe of the adult nulliparous female showing the progressive branching of the ductal epithelium culminating in the terminal duct lobular units (adapted from review article [165]). B) Schematic drawing of a terminal duct lobular unit in cross section depicting the close association of the acinar units with intralobular fibroblasts. Epithelial cells are depicted as solid grey squares and fibroblasts are depicted as black spindles (adapted from review article [165]).  22  Figure 1.2 – Mammary epithelial cell hierarchy In the human mammary gland, a common or bi-potent progenitor cell exists that can give rise to the lineage-restricted progenitor cells. This cell type is either the human mammary stem cell or closely derived from a stem cell, a distinction that is not currently assayable (adapted from review article [166]).  23  2 2.1 2.1.1  Materials and methods Cell culture Culture of epithelial cell lines The 184-hTERT parental line was constructed in the Laboratory of Molecular  Carcinogenesis at the NIEHS by Dr. Carl Barrett [167]. Routine sub-culturing was conducted in serum-free mammary epithelial cell basal media (MEBM, Lonza) supplemented with mammary epithelial cell growth media SingleQuots (Lonza), 5 µg/ml transferrin (Sigma), and 10-5 M isoproterenol (Sigma) to make mammary epithelial cell growth media (MEGM). Cultures were maintained in a humidified incubator at 37°C with 2.5% carbon dioxide (CO2) [168, 169]. Media tested for 184-hTERT and NIH 3T3 co-culture experiments included EpiCult B plus 0.5 µg/ml hydrocortisone (Stem Cell Technologies), Dulbecco's Modified Eagle Medium (DMEM) supplemented with 5% fetal bovine serum (FBS, Gibco), serum-free 7 media made from Dulbecco's Modified Eagle Medium: Nutrient Mixture F-12 (D/F12, Stem Cell Technologies) with 1 mg/ml bovine serum albumin, 1 µg/ml insulin, 0.5 µg/ml hydrocortisone, and 10 ng/ml cholera toxin, and finally “Visvader media” made as per the serum-free 7 media with the exception of 5 µg/ml insulin [12, 13]. MCF10A cells were obtained from the American Type Culture Collection and cultured as previously described in D/F12 supplemented with 5% horse serum (Invitrogen), 20 ng/ml epidermal growth factor, 0.5 µg/ml hydrocortisone, 100 ng/ml cholera toxin, and 10 µg/ml insulin [170]. Comma D-beta cells (obtained from Dr. Connie Eaves) were routinely grown in D/F12 supplemented with 5% FBS, 10 µg/ml insulin and 5 ng/ml epidermal growth factor. For the treatment of cells with epigenetic modifiers prior to extended passaging or placement in 3D Matrigel culture, cells were plated 24 hours prior to treatment into 10 cm2 tissue culture treated dishes at 30% confluency. Treatments were conducted for 72 hours, with freshly made drug containing media replaced every 24 hours. After an additional 24 hours of recovery time in MEGM alone, cells were harvested for passaging or 3D Matrigel culture. For the passaging experiments, cells were treated with 10 µM 5aza-2'-deoxycytidine, 100 nM trichostatin A, 5 µM 3-deazaneplanocin A (National 24  Cancer Institute), or 1 µM BIX-01294. For the Matrigel experiments, cells were treated with 0.01, 0.05, 0.1, 0.5 and 1 µM of 5-aza-2'-deoxycytidine. All drugs were solubilized in dimethyl sulfoxide (DMSO). Dilutions were normalized so that the control and each experimental condition received the same volume of DMSO.  Except where otherwise  noted, all reagents and chemicals in this section and hereafter were purchased from Sigma. 2.1.2  Preparation of irradiated NIH 3T3 feeder cells NIH 3T3 cells (provided by Dr. Connie Eaves) were routinely maintained in a  humidified incubator at 37°C with 5% carbon dioxide in DMEM supplemented with 5% FBS. Cells were grown to approximately 70% confluence before harvesting. After harvesting by trypsinization, cells were re-suspended in Hank’s Balanced Salt Solution (HBSS, Stem Cell Technologies) with 2% FBS for irradiation at a density of 2.0x106 cells/ml. Suspensions were placed on ice and exposed to 40 Gy using 300KVp X-rays. Cells were either cryopreserved or used immediately. 2.1.3  Derivation of the 184-hTERT clonal lines 184-hTERT parental cells at passage 6 were obtained from Dr. Carl Barrett and  grown for 24 hours prior to seeding an average of 0.3 cells/well in four 96-well plates. Two of the 96-well plates were supplemented with 2500 freshly irradiated NIH 3T3 cells per well. For the plates without feeder cells present, every well was checked microscopically eight hours post-seeding to ensure the presence of only one cell per well; wells not meeting this criterion were rejected. In the co-cultures, every well was inspected 5 days post-plating and those with more than one small colony present were rejected. 8-days post-plating, individual wells were harvested with 0.25% trypsin in 1mM ethylenediaminetetraacetic acid, tetrasodium salt (EDTA, Gibco) and passaged into 24-well plates. After 10 days, growth capacity and morphology was assessed on a brightfield microscope and recorded. Upon reaching confluency, cells were sequentially passed into 6-well plates and then 75 cm2 flasks prior to cryopreservation.  25  2.1.4  3D Matrigel culture Matrigel cultures were performed in a similar fashion to the previously described  overlay assay [170]. Briefly, 50 µl of growth factor reduced Matrigel (BD Biosciences) was spread evenly onto the surface of an 8-well LabTek II Chamber Slide (Nunc) and allowed to solidify at 37°C for 30 min. To this, 1250 cells per well were added in 300 µl of media supplemented with 2% growth factor reduced Matrigel. 184-hTERT cells were grown in MEGM, primary mammary cells in serum-free 7 media with B-27 Serum Free Supplement (Invitrogen), and Comma D-beta cells in D/F12 supplemented with 5% FBS, 10 µg/ml insulin and 5 ng/ml epidermal growth factor. Media was changed every 4 days and the assays were terminated 21-days post-plating. When used for RNA, 400 µl of TRIzol reagent (Invitrogen) was added to each well and processed as per below. For immunofluorescence, cultures were fixed in 4% paraformaldehyde in phosphate buffered saline (Electron Microscopy Sciences) for 20 min at room temperature prior to staining. 2.2 2.2.1  Primary mammary tissue Dissociation of primary mammary tissue Discarded tissue was collected with informed consent from premenopausal  women (ages 19-40) undergoing reduction mammoplasty surgery as approved by the University of British Columbia Research Ethics Board. Post-collection processing was performed as previously described [171]. Briefly, tissue was transported from the operating room on ice and was minced with scalpels prior to dissociation in Ham’s D/F12 supplemented with 2% wt/vol Fraction V bovine serum albumin (BSA, Gibco), 300 U/ml collagenase and 100 U/ml hyaluronidase for 18 hours. Differential centrifugation at 80 xg for 4 min was used to collect an epithelial-cell rich pellet. This was cryopreserved in media containing 6% DMSO at –135°C until use. Single cell suspensions were subsequently prepared from freshly thawed pellets by sequentially treating the cells with 2.5 mg/ml trypsin with 1 mM EDTA (StemCell Technologies) and 5 mg/ml dispase (StemCell Technologies) with 100 µg/ml DNase1 (Ambion). Between treatments, cells were washed with cold HBSS supplemented with 2% FBS. Cell suspensions were  26  passed through a 40 um filter (BD Biosciences) to remove remaining cell aggregates and kept on ice for use within 1 hour. 2.2.2  Colony-forming assays Gridded 60 mm tissue culture dishes (Sarstead) were pre-coated with a 1:43  dilution of collagen solution in PBS (Stem Cell Technologies) for 1 h at 37°C. Plates were rinsed with PBS to remove any residual acid left over from the collagen preparation. Each dish was then seeded with 100 to 5000 epithelial cells in 4 ml of serum-free 7 media with 5% FBS and 1.6x105 freshly thawed NIH 3T3 cells, previously irradiated at 40 Gy. After 8-10 days, dishes were fixed and stained with 0.8% wt/vol methylene blue in methanol (Sigma). Colonies with 50 cells or more were visually scored under a dissecting microscope. 2.2.3  Fluorescence activated cell sorting For the separation of primary mammary epithelium into luminal and basal  fractions, cells were pre-blocked in HBSS supplemented with 2% FBS and 10% human serum (Jackson ImmunoResearch). An allophycocyanin-conjugated rat antibody to human CD49f (clone GOH3, R&D Systems) and a phycoerythrin-conjugated mouse antibody to human EpCAM (clone 9C4, Biolegend) were used for basal and luminal population discrimination as previously described [18]. Hematopoietic and endothelial cells were labeled with biotin-conjugated mouse antibodies to human CD45 (clone HI30, Biolegend) and human CD31 (cloneWM59, eBiosciences), respectively, followed by FITC–conjugated streptavidin (BD Biosciences). Either propidium iodide at 10 µg/ml or Dapi at 0.1 µg/ml (Sigma) was used for live/dead cell discrimination. Sorting was performed on a FACSDiva (Becton Dickinson) using gates that excluded 99.9% of events present in negatively stained control samples and events with very high forward and side light scatter profile.  27  2.3 2.3.1  Molecular techniques Nucleic acid extractions Material for DNA extraction (cell pellets and mouse ear notches) was digested  overnight at 55°C in a solution of 10 mM tris(hydroxymethyl)amino methane (pH8.0)(Tris, Bio Basic), 10 mM EDTA (pH 8.0), 50 mM NaCl (Bio Basic), 1% sodium dodecyl sulfate, and 50 µg/ml Proteinase K. An equal volume of phenol:chloroform:isoamyl alcohol 25:24:1 was added prior to rigorous shaking and centrifugation at 13,000 rpm for 1 min in a table-top micro-centrifuge. An equal volume of chloroform was added to the upper aqueous phase after it was transferred to a new tube. After shaking and centrifugation, the upper aqueous phase was again transferred to a new tube, mixed with an equal volume of isopropanol (Sigma), and left at -20°C for at least 90 min. Precipitated DNA was pelleted by centrifugation in a 4°C refrigerated tabletop microfuge at 13,000 rpm for 10 min, washed with 70% ethanol (Sigma), pelleted as before and air-dried prior to resuspension in Ultrapure Dnase/Rnase Free Distilled Water (Invitrogen). DNA concentration and purity (A260/A280) was assessed using a Nanodrop 2000 Spectrophotometer (Thermo Scientific). RNA extractions were performed using TRIzol reagent (Invitrogen) as per the manufacturers instructions. TRIzol was added either directly to tissue culture plates, or to sorted and pelleted primary cell fractions and incubated for 5 min prior to trituration. A volume of chloroform equal to 20% of the original volume of TRIzol was added and mixed vigorously for 1 min prior to centrifugation at 12,000 rpm for 1 min. Without disrupting the organic layer, the upper aqueous layer was transferred to a new centrifuge tube and precipitated by adding an equal volume of isopropanol and placing at -20°C for more than 90 min. Samples were centrifuged at 12,000 rpm for 10 min at 4°C. The pelleted RNA was washed with 70% ethanol and centrifuged again at 12,000 rpm for 5 min at 4°C. RNA pellets were air-dried before the addition of an appropriate volume of DNAse/RNAse free water (Invitrogen) and subsequent incubation for 10 min at 60°C. RNA concentration and purity was assessed using a Nanodrop 2000 Spectrophotometer prior to the removal of DNA with the TURBO DNA-free kit (Ambion) as per the manufactures instructions. 28  2.3.2  Southern Blotting Ten micrograms of DNA was digested overnight at 37°C with the restriction  enzymes HindIII or BglII (New England Biolabs) and resolved on a 1% wt/vol agarose gel (Bio Basic) in a buffer of 40 mM Tris-acetate and 1 mM EDTA overnight at a current 20 eV. After denaturing for 30 min in 0.5 M sodium hydroxide (Sigma) with 1.5 M NaCl, gels were transferred to Zeta-probe GT Membranes (Bio-Rad) overnight in 10x saline-sodium citrate and baked at 80°C for 1 hour to fix the DNA to the membrane. The membrane was hybridized with a radiolabeled probe targeting the neomycin resistance gene following standard hybridization and washing procedures. The template for the probe was amplified from 184-hTERT genomic DNA using 5’caagatggattgcacgcaggttctccg-3’ and 5’-caagctcttcagcaatatcacgggtag-3’as primers in a standard PCR reaction using Platinum Taq (Invitrogen) and an annealing temperature of 55°C for 30 cycles. The probe was labeled using the Random Primed DNA Labeling Kit (Roche) according to the manufacturer’s instructions. 2.3.3  Protein extraction and Western Blotting Cells for protein extraction were washed with phosphate buffered saline (PBS)  prior to the addition of ice cold radioimmunoprecipitation assay lysis buffer (150 mM NaCl, 50 mM Tris HCl pH 7.5, 1 mM EDTA, 0.1% SDS, 1% Sodium deoxycholate, and 1% Triton-X 100) supplemented with Complete Protease Inhibitor Cocktail (Roche). Lysates were placed on ice on a rotating platform for 20 min prior to scraping the remaining adherent cells with a rubber spatula, or passing pelleted cells through a syringe fitted with a 26 gauge needle. After centrifugation at 12,000 rpm for 10 min at 4°C, the pelleted cell debris was removed and the protein concentration of the cleared lysates was assessed using the BCA Protein Assay Kit (Pierce). Size separation by gel electrophoresis was performed by loading 10 µg of protein lysate previously incubated with 4x NuPage LDS sample loading buffer and 10x Reducing Agent (Invitrogen) at 70°C for 10 min onto a 4-12% gradient NuPage Bis-Tris precast gel (Invitrogen). This was followed by a semi-dry transfer onto nitrocellulose using the iBlot system (Invitrogen) or a wet transfer onto nitrocellulose (Bio-Rad) following standard 29  procedures. Membranes were blocked for 1 hour at room temperature in Odyssey blocking buffer (Li-COR Biotechnology) and rinsed with PBS prior to overnight incubation at 4°C with primary antibodies targeting ERK 1/2(Cell Signaling), phosphoERK 1/2 (Cell signaling), p21 (Neomarkers), p16 (Neomarkers), p63 (Millipore), cytokeratin 5 (Covance), cytokeratin 18 (Progen), e-cadherin (VWR), β-actin (Santa Cruz), GAPDH (Santa Cruz), lamin (BD Biosciences), and Edg7 (AbCam). Three 5 min washes were conducted using PBS with 0.5% Tween prior to the addition of the secondary antibody for 1 hour at room temperature. Antibody detection was accomplished through either chemiluminescence or infrared imaging. Horseradish peroxidase conjugated secondary antibodies (Dako) were used with enhanced chemiluminescence (ECL) detection (Amersham), or infrared dye conjugated secondary antibodies (Rockland Immunochemicals) were used in conjuction with the Li-COR Odyssey infrared detection system (Li-COR Biotechnology). Three 5 min washes were conducted using PBS with 0.5% Tween prior to visualization of the probed membranes. For ECL detection, the SuperSignal detection kit (Pierce) was used prior to x-ray film (Kodak) exposure. 2.3.4  Reverse transcription and quantitative polymerase chain reaction RNA was DNAse treated with the Turbo DNA-free kit (Ambion) according to the  manufacturer’s instructions to eliminate contaminating genomic DNA from the sample. One microgram of total RNA, 250 ng of random hexamers, and 125 µM of each deoxynucleotide triphosphate (dNTP) were heated to 65°C for 5 min and then placed on ice. To this, 8 U of RNaseOUT and a final concentration of 1x First Strand Buffer and 0.01 M dithiothreitol were added and placed at 37°C for 2 min. Fifty units of M-MLV reverse transcriptase (RT) enzyme (Invitrogen) were added and incubated at 37°C for 50 min and then 70°C for 15 min. Gene specific primers (Integrated DNA Technologies) for reverse transcription quantitative polymerase chain reaction (RT-QPCR) were designed using the Roche Universal Probe Library (UPL) Design Program. Primers sequences and corresponding UPL probe numbers (Roche) are listed in the appendices with the following exceptions:  30  glyceraldehyde 3-phosphate dehydrogenase (GAPDH), UPL probe 60, 5' ctctgctcctcctgttcgac, 3' acgaccaaatccgttgactc; β-actin, UPL probe 64, 5' ccaaccgcgagaagatga, 3' tccatcacgatgccagtg; Celsr1 (C-terminal), UPL probe 34, 5’ atggatatctccaggcgtga, 3’ tgccagtgacaaggacaca; Celsr1 (N-terminal), UPL probe 80, 5’ catggaggtgtctgtgtctga, 3’ caggaacttctcctgggacat; Celsr2, UPL probe 84, 5’ tctgccacacaggacgtg, 3’ tggatcagctcccagtgc; Celsr3, UPL probe 8, 5’ gatcctcacacccatgtgc, 3’ gctgcttgtgggcagaac; Shisa4, UPL probe 50, 5' accatctgctgcttcctctg, 3' gcctgtcattggaatctcct; Gpr116, UPL probe 26, 5' cttcctgctgggtgattctc, 3' ggttgcctgggacttcatt; Zpld1, UPL probe 68, 5' aatggcacatttgtcagcac, 3' gctggttgtaggttgaatcgt; Fat4, UPL probe 11, 5' agcacaaggcatcctagatca, 3' catgggctgcgagtacaag; AdamTS1, UPL probe 25, 5' gaacaaaaccgacagaaagca, 3' tgtcacattccctcatcgtg; Trib2, UPL probe 17, 5' gtttttcgtgccgtgcat, 3' ggtagcagctgatatcaaacacc; SerpinG1, UPL probe 20, 5' taccccgcatcaaagtgac, 3' cccacacaggttaaggtcataag; SerpinB3, UPL probe 38, 5' gatctcacccttcattccaca, 3' ttcatggtgaactcgatgtga; Plat, UPL probe 36, 5' tgtgtggagcagtcttcgtt, 3' tcatctctgcagatcacttggt; Odz1, UPL probe 3, 5' cattttgaatcctcaaagtgga, 3' cccattatggttgatatgactgg; Ncrna0020454, UPL probe 54, 5' ttgcagagatgcaggtattgtc, 3' aacacttagggaggcacagg; IL1RL1, UPL probe 56, 5' ttgtcctaccattgacctctacaa, 3' gatccttgaagagcctgacaa; Gem, UPL probe 37, 5' cagcactgggattttctgg, 3' atcagctgggatgctcca; Fry, UPL probe 7, 5' cgggcttctttgagatcagt, 3' gaaccggaggcttgatgtaa; EpCam, UPL probe 3, 5' ccatgtgctggtgtgtgaa, 3' tgtgttttagttcaatgatgatcca; Elmod1, UPL probe 7, 5' gaaatgcagggatatcactaaagaa, 3' ccaattgccttatccatcctt; Csf3, UPL probe 1, 5' gagcaagtgaggaagatccag, 3' cagcttgtaggtggcacaca; Cdt1, UPL probe 10, 5' cggagcgtctttgtgtcc, 3' agcaggtgcttctccatttc; Bpil2, UPL probe 68, 5' ccagtgttggcctggttatt, 3' aaagcaaggcggaatctgt; ATP6V0D2, UPL probe 1, 5' gactatgtggagaatttgagtatttcc, 3' catcagcagaatcacattgtctatc. Amplification reactions were performed using 1x TaqMan Universal PCR Master Mix (Applied Biosystems), 1/20 of the above RT reaction, 0.1 µM UPL probe, and 0.8 µM of each primer. Standard cycling conditions on the Applied Biosystems (ABI) 7900HT Fast Real-Time System were used to amplify and detect mRNA transcripts (Applied Biosystems). GAPDH was used as an endogenous loading control prior to comparative normalization to one of the samples included in the analysis. 31  RQ Manager Software (Applied Biosystems) calculated the relative transcript expression using the equation RQ=2- Ct. Expression of β-actin was assessed in each sample to ΔΔ  ensure that GAPDH was a suitable endogenous control and that its expression always remained unchanged in comparison to another common housekeeping gene. 2.3.5  Immunohistochemistry and immunofluorescence Cell blocks for immunohistochemistry were prepared by immobilizing cells fixed  in 10% neutral-buffered formalin in 1% wt/vol agarose prior to paraffin-embedding. Four micrometer thick sections were cut and immunostained using a Ventana Discovery XT staining system (Ventana Medical Systems). Briefly, sections were baked at 60°C for 1 hour, deparaffinized in xylene, dehydrated with three alcohol changes, and washed in Ventana Wash solution. Endogenous peroxidase activity was blocked in 3% hydrogen peroxide and heat-induced antigen retrieval was performed using Cell Conditioning solution (CC1- Tris based EDTA buffer, pH 8.0, Ventana). Sections were incubated with antibodies against estrogen receptor (LabVision), progesterone receptor (Ventana), epidermal growth factor receptor (DAKO), erythroblastic leukemia viral oncogene homolog 2 (LabVision), cytokeratin 5/6 (Zymed Laboratories), e-cadherin (VWR), Ki67 (ThermoScientific), and p53 (DAKO). Secondary detection was performed using the pre-diluted Ventana Universal Secondary Antibody and DAB Map detection system. Finally, slides were counterstained with hematoxylin, dehydrated, cleared, and mounted. Immunofluorescence was performed on cells grown on glass coverslips or in 3D Matrigel cultures. Cells were routinely fixed in 4% paraformaldehyde (Electron Microscopy Sciences) for 20 min at room temperature and rinsed with PBS. Cells were washed twice with 0.1 M Glycine in PBS for 10 min and permeabilized with 0.1% Triton X-100 in PBS for 15 min at 4°C. Cells were rinsed once for 5 min with IF Buffer (130 mM NaCl, 7 mM Na2HPO4, 3.5 mM NaH2PO4 , 7.7 mM NaN3 , 0.1% bovine serum albumin, 0.2% Triton X-100, 0.05% Tween-20) and blocked for 1.5 hours in IF Buffer containing 10% serum from the species the secondary antibodies were raised in. Primary antibodies were diluted in IF buffer and incubated with the cells overnight at 4°C. Cells were washed three times for 20 min in IF buffer prior to the addition of the secondary antibodies, diluted in IF buffer, for 1 hour. Primary antibodies raised against pericentrin 32  (AbCam), α-tubulin (Sigma), GM130 (BD Biosciences), mucin-1 (StemCell Technologies) and α6-integrin (R&D Systems) were used. Secondary antibodies were all generated against the heavy and light chains of the primary antibody species and conjugated to Alexa Fluor 488, 546 or 680 dyes (Invitrogen). Cells were occasionally counterstained for 20 min with Oregon Green 488 phalloidin (Invitrogen). Nuclear staining for 5 min with Draq5 (Biostatus Limited) or propidium iodide was performed prior to imaging on a confocal laser scanning microscope (Nikon). 2.3.6  Fluorescent in-situ hybridization Trypsinized cell lines were exposed to 0.075 M potassium chloride prior to  progressive fixation with 3:1 methanol:acetic acid. Cells were dropped onto glass slides and air-dried prior to aging for 45 min in 2x SSC buffer (3.0 M Sodium Chloride, 0.3 M Sodium Citrate, pH 7.0). Slides were dehydrated by sequential changes in increasing concentrations of ethanol from 70% to 100%. After air-drying, probe was applied and co-denatured for 5 min at 75°C prior to hybridization for 18 hours at 37°C on a HYBrite (Vysis). Post-hybridization washing was performed at 73°C for 2 min in 0.4x SSC with 0.3% Nonidet P 40 Substitute (NP40) and room temperature for 1 min in 2x SSC with 0.1% NP40. Slides were dried and counterstained with DAPI. Interphase FISH signals were enumerated in over 150 morphologically intact and non-overlapping nuclei. SpectrumOrange labeled probes to MYC, ZNF217 (20q13.2) and HER-2/neu were used (Vysis). SpectrumGreen labeled probes to centromere 8, centromere 17 and the p-arm of chromosome 20 were used (Vysis). Multiplex FISH was performed using Spectra Vision probes (Vysis). 2.3.7  Vectorette PCR One microgram of genomic DNA was digested with EaeI or HaeII restriction  endonuclease (New England Biolabs) according to the manufacture’s instructions in a final volume of 50 µl. DNA was precipitated with 0.1 M sodium acetate and 65% ethanol at -20°C for 1 hour and pelleted by centrifugation at 12,000 rpm for 30 min at 4°C. Pelleted DNA was washed with 80% ethanol and centrifuged again for 5 min at 12,000 rpm at 4°C. DNA was air dried prior to ligation with the vectorette linkers. The 33  vectorette linkers were prepared by mixing 100 pmol each of the top and bottom linkers with 25 mM sodium chloride. This was heated to 99°C for 2 min followed by 65°C for 5 min and gradually cooled to room temperature. Synthetic oligonucleotide linker sequences (Integrated DNA Technologies) were as follows: EaeI top linker, 5’-ctctcccttctcgaatcgtaaccgttcgtacgagaatcgctgtcctctccttccggcc-3’; EaeI bottom linker, 5’ggaaggagaggacgctgtctgtcgaaggtaaggaacggacgagagaagggagag-3’; HaeII top linker, 5’ctctcccttctcgaatcgtaaccgttcgtacgagaatcgctgtcctctccttcagcgc-3’; HaeII bottom linker, 5’tgaaggagaggacgctgtctgtcgaaggtaaggaacggacgagagaagggagag-3’. Ligation was conducted by mixing 15 µl of the prepared linkers with 12 µl of digested DNA in 1x NEB T4 DNA ligase reaction buffer with 2000 U of T4 DNA ligase (New England Biolabs) and incubating overnight at room temperature. The first PCR reaction was performed with 10 µl of the ligation reaction in 1x PCR buffer, 1.5 mM magnesium chloride, 50 µM of each dNTP, 0.2 µM of each primer, and 1.25 U of Platinum Taq DNA polymerase (Invitrogen). After a 2 min 94°C hot start, 30 cycles were performed of 94°C for 30 sec, 65°C for 30 sec, and 72°C for 2.5 min, with a final extension of 72°C for 2 min. Two independent reactions were set up with vectorette primer 224, 5’cgaatcgtaaccgttcgtacgagaatcgct-3’, and either LTR-R-435, 5’-ggggttgtgggctcttttat-3’, or LTR-R-390, 5’-aacagaagcgagaagcgaac-3’. The second nested PCR was performed in the same manner with the exception that 1 µl of the first PCR reaction was used as template. The nested linker primer 5’-tacgagaatcgctgtcctctcctt-3’ was used with either LTR-R-390 (when LTR-R-435 was used in the 1st PCR reaction) or LTR-R-239, 5’tctggggaccatctgttctt-3’ (when LTR-R-390 was used in the first reaction). Reactions were resolved on a 1% wt/vol agarose gel and discrete bands were excised with a scalpel. DNA was purified using the QIAquick Gel Extraction Kit (Qiagen) as per the manufacture’s instructions and TA Cloned into a pCR2.1 vector (Invitrogen) following the manufacture’s recommendations. One Shot electrocompetent cells (Invitrogen) were transformed and plated onto bromo-chloro-indolyl-galactopyranoside Luria-Bertani agar plates with ampicillin following the manufacture’s recommendations. Three colonies were picked and grown in Luria-Bertani broth cultures with ampicillin and plasmids and then extracted using the QIAprep Spin Miniprep Kit (Qiagen) as per the manufacture’s instructions. Plasmids containing vectorette PCR products were Sanger sequenced 34  (Macrogen) using the M13F, 5’- gtaaaacgacggccagt-3’ and M13R, 5’gcggataacaatttcacacagg-3’ primers. 2.3.8  5-bromo-2-deoxyuridine cell cycle profiling 184-hTERT-L9 cells were plated at a density of 3125 cells/cm2 in 6-well plates 24  hours prior to treatments. For serum starvation, cells were rinsed three times in HBSS and then left in un-supplemented MEBM. For irradiation, plated cells were exposed to 4 Gy radiation using 300KVp X-rays prior to a media change into fresh MEGM. Drug treatments were diluted in MEGM as follows: 5 µM etoposide, 5 µg/ml Aphidicolin, 0.1 µg/ml colchicine and 1 µM mitomycin C. After 24 hours, 50 µM 5-bromo-2deoxyuridine (BrdU) was added for 30 min at 37°C. Cells were harvested by trypsinization, fixed in 100% cold ethanol for 30 min, treated with 2 N hydrochloric acid for 2 min, washed three times with PBS + 0.1% BSA and stained with an Alexa Fluor 488 conjugated BrdU antibody (Invitrogen). Cells were treated with 0.1  mg/ml RNAse A (Invitrogen) and 5  µg/ml propidium iodide prior to two-dimensional flow cytometry on a FACSCalibur (Becton Dickinson) to detect both Alexa Fluor 488 and propidium iodide. 2.3.9  Microarrays Array-based comparative genomic hybridization was performed using the Human  CGH 385K Whole-Genome Tiling Array v1.0 (NimbleGen) with DNA extracted from 184-hTERT-L9, 184-hTERT-L2, and 184-hTERT-gfp cells. Test samples were labeled with Cy3 and compared to Cy5 labeled normal human female genomic reference DNA (Promega). Sample processing, array hybridization, and data analysis was performed by NimbleGen as per standard protocols. Affymetrix gene expression arrays were hybridized at the Centre for Translational and Applied Genomics at the BC Cancer Agency. Experiments were performed in triplicate with DNA-free RNA extracted from 184-hTERT-L9 and 184-hTERT-L2 cells grown under standard conditions, or from 184-hTERT-L9 cells grown in low-density culture with either irradiated NIH 3T3 cells in MEGM (no bovine pituitary extract) or MEGM with bovine pituitary extract. Samples were prepared and hybridized to GeneChip Human Exon 1.0 ST Arrays (Affymetrix) as per the manufactures 35  recommendations. ArrayAssist version 5.1.0 (Stratagene) was used to analyze differential gene expression in both the gene-level and exon-level models, filtering for expression differences of greater than 1.5 fold in either direction and a p-value of less than 0.05. DNA extracted from the 184-hTERT-L2 or 184-hTERT-L9 cells grown in standard conditions was hybridized to Genome-Wide Human SNP 6.0 Arrays (Affymetrix) by AROS Applied Biotechnology as per the manufacturers instructions. Copy number status from the log ratio intensity data was predicted through HMMK11 modeling (available from http://compbio.bccrc.ca/?page_id=401). 2.4 2.4.1  RNA interference Transfection of cell lines with small-interfering RNAs All of the small-interfering RNAs (siRNAs) were purchased from the siGENOME  library (Dharmacon) as either pooled or de-convolved sets of 4 individual siRNAs and resuspended at 10 µM in 1x siRNA buffer (Dharmacon). Lipid-based transfection reagent Lipofectamine 2000 (Invitrogen), or DharmaFECT3 (Dharmacon) where noted, was diluted in MEBM and incubated for 5 min prior to mixing with an equal volume of pre-diluted siRNA in MEBM. Complexes were allowed to form for 20 min before they were added directly to cells that had been plated 24 hours prior in their respective growth media. Unless otherwise noted, siRNA were used at a final concentration of 30 nM and 0.3 µl of transfection reagent was used per well of a 96-well plate or 3ul was used per well of a 6-well plate. With the exception of the 184-hTERT cells where it was deemed unnecessary, media was changed to fresh media 24 hours post-transfection. 2.4.2  Viral packaging and titering of lentiviral short-hairpin RNAs Lentiviral particles were generated by transient transfection of the lentiviral  vectors along with the pCMV R8.91 packaging plasmid and pMD2 VSV-G envelope virus obtained from the RNAi Consortium (Broad Institute, MA). Briefly, HEK 293T cells (ATCC) were plated at a density of 5x105 cells/cm2 in DMEM with 10% FBS 24 hours prior to transfection. A ratio of 10:9:1 of vector to pCMV R8.91 and pMD2 VSVG was used for co-transfection using TransIT-LTI transfection reagent (Mirus Bio) as per  36  the manufacturer’s recommendations. The culture medium was replaced 18 hours posttransfection and viral supernatant was collected 24 hours and 48 hours later and passed through a 0.45 µm pore-size cellulose-acetate filter. Pooled supernatants were concentrated 100-fold by ultracentrifugation at 100,000 xg for 105 min. Pellets were resuspended by shaking in DMEM for at least 30 min and then stored in aliquots at -80°C until use. For lentiviral particles based on the pGIPZ backbone, viral titers were determined by flow cytometry in HEK 293T cells. For this, five-fold serial dilutions of viral stocks were prepared in HEK 293T growth media and added to the cells with 8 µg/ml polybrene (Santa Cruz Biotechnology). The transduction mix was removed from the cells 6 hours post-infection. The percentage of GFP positive cells was determined 72 hours post-transduction on a FACSCalibur (Becton Dickinson), after gating for viable cells identified through propidium iodide exclusion. For lentiviral particles based on the pLKO.1 backbone, viral titers were determined using the Lenti-Z p24 Rapid Titer Kit (Clontech) following the manufacture’s instructions. 2.4.3  Transduction of lentiviral short-hairpin RNAs 184-hTERT-L9 cells were plated at a density of 1.8x104 cells/cm2 in 6 cm cell  culture dishes 24 hours prior to transduction. Cells were infected with freshly thawed lentiviral particles at an estimated multiplicity of infection (MOI) of 5 with 8 µg/ml polybrene in the standard 184-hTERT-L9 culture media. After 18 hours at 37 ºC, cells were washed three times in HBSS with 2% FBS and returned to standard media. Sorted populations of primary human mammary epithelial cells were immediately infected with freshly thawed lentiviral particles at an estimated MOI of 10. Transduction was conducted in suspension with 8 µg/ml polybrene at a density of 5x105 cells in 100 µl of serum-free 7 media with 50 µg/ml GA-1000 (Lonza). After 18 hours at 37°C, cells were washed three times in HBSS with 2% FBS, re-suspended in 100 µl of fresh serumfree 7, and placed on ice. Cells were counted using an automated cell counter and equal numbers were plated into colony forming assays.  37  2.5 2.5.1  Animal experiments Renal capsule xenografts Concentrated rat’s tail collagen was prepared as previously described and  generously provided to us by Dr. Connie Eaves [172]. It was thawed and neutralized immediately prior to use as described [18]. Briefly, 78% concentrated collagen was mixed with 20% 5x DMEM and 2% 12 M sodium hydroxide. This was immediately added to a pellet of 3x105 184-hTERT cells and 2.2x105 irradiated CH3 10T1/2 cells (from Dr. Connie Eaves) for a final volume of 25 µl. This was pipetted as a crescent shape along the edge of a 24-well non-tissue culture treated plate and allowed to solidify at 37°C for 10 min. MEGM + 5% FBS was added to each well and a pipette tip used to dislodge the gel from the surface of the well. After an additional 50 min at 37°C, plates were placed on ice until they were transplanted. NOD/SCID/interleukin 2 receptor γ null mice were used for the subrenal xenografting [173]. Anesthetized mice were prepared for surgery by shaving their backs and swabbing the skin with 70% ethanol. A 1.5 cm anterior to posterior dorsal incision was made followed by a small incision in the peritoneal membrane above the kidneys. One at a time, the kidneys were exteriorized by applying gentle pressure on either side. Using a dissecting microscope, a small incision was made in the kidney capsule and the collagen gel was slid between the capsule and parenchyma of the kidney. After suturing the incisions on the peritoneal cavity membrane, the midline incision was sutured and a slow-release silicone pellet (NuSil Technology) containing 2 mg of β-estradiol and 4 mg of progesterone was inserted in the posterior position. Gels were harvested after 4 weeks and were fixed in formalin and embedded in paraffin. All mice were bred and maintained in microisolator cages and provided with sterile food, water and bedding. Animal procedures were carried out in accordance with the Canadian Council on Animal Care. Specific project and protocol approval was obtained from the University of British Columbia Animal Care Committee.  38  2.5.2  Celsr1 deleted mice Mice with loxP sites flanking the transmembrane region of the Celsr1 gene were  generously provided to us by Dr. Fadel Tissar [174]. These mice were crossed with pCX-NLS-Cre mice (from Dr. Fabio Rossi), a transgenic line that constitutively expresses nuclear-localized Cre recombinase [175]. Ear notches were obtained for genotyping and extracted using the Wizard SV Genomic DNA Purification Kit (Promega) as per the manufacture’s recommendations. pCX-NLS-Cre mice were genotyped using with the primers oIMR1084, 5’-gcggtctggcagtaaaaactatc-3’; oIMR1085, 5’-gtgaaacagcattgctgtcactt-3’; oIMR7338, 5’-ctaggccacagaattgaaagatct-3’; oIMR7339, 5’-gtaggtggaaattctagcatcatcc-3’. The PCR reaction was performed with 3 µl of genomic DNA in 1x PCR buffer, 1.67 mM magnesium chloride, 42 µM of each dNTP, 0.8 µM of each primer, and 1 U of Platinum Taq DNA polymerase (Invitrogen). After a 3 min 94°C hot start, 35 cycles were performed of 94°C for 30 sec, 52°C for 1 min, and 72°C for 1 min, with a final extension of 72°C for 4 min. The status of the Celsr1 allele was assessed with 2 independent PCR reactions using the 612/613 primer combination to detect the floxed and wild-type alleles, and the 612/615 primer combination to detect the deleted alleles. The primer sequences were 612, 5’-gaaagagactgttggtgagc-3’; 613, 5’ccactctgctaacggtagg-3’; 615, 5’-ctctgttgacttctgactgg-3’. Both of the PCR reactions were performed with 3 µl of genomic DNA in 1x PCR buffer, 3 mM magnesium chloride, 50 µM of each dNTP, 0.2 µM of each primer, and 1 U of Platinum Taq DNA polymerase (Invitrogen). After a 3 min 94°C hot start, 30 cycles were performed of 94°C for 30 sec, 60°C for 1 min, and 72°C for 1 min, with a final extension of 72°C for 10 min. Homozygous deleted Celsr1 female mice and their wild-type littermates were euthanized and matched by estrous cycle stage based upon the gross morphology of their uterus. The fourth inguinal mammary gland was resected from the matched pairs at 7 and 11 weeks of age and spread on a glass slide. Glands were fixed in Carnoy’s fixative (60% ethanol, 30% chloroform, 10% glacial acetic acid) for 2.5 hours. They were washed in 70% ethanol for 15 min prior to a gradual change into distilled water. Slides were stained with carmine alum (Stem Cell Technologies) overnight and washed for 15 min in 70%, 95% and 100% ethanol. Glands were imaged on a dissecting scope and stored in 100% ethanol. 39  3  Derivation of clonal cell lines with properties reminiscent of  mammary progenitor cells 3.1  Introduction There are several well-established model systems with which the study of normal  human mammary epithelial cells can be conducted. The most common method of reference in current use for studying normal mammary physiology is the use of discarded reduction mammoplasty tissue. While this tissue can be readily available when cooperative surgeons are within close proximity of a laboratory, the inherent nature of this tissue limits its usefulness in certain settings. Some considerations include the inability to culture these cells for extended periods of time, patient-to-patient variability, tissue heterogeneity, and limited transfection efficiency for plasmids and siRNA. In order to avoid these limitations, several cell lines have been generated from primary mammary tissue. However, none of the currently described lines encompass all of the qualities that are desirable when attempting to work within a model of normal physiology. Because of this, we sought to improve upon the available model systems and develop a cell line with the closest possible resemblance to normal mammary progenitor cells whilst still having properties amenable to manipulation for molecular analysis. Extensive analysis of the growth of human mammary epithelial cells (HMECs) in culture has been performed over the past 3 decades. Their growth can be divided into distinct phases defined by the presence of one to two growth barriers, dependent upon the culture conditions that must be overcome to achieve immortality. This is of interest as the ability to overcome senescence, and the loss of genomic integrity inherent to this process, are important obstacles in malignant progression. Initially, HMECs will grow in culture for up to 4 passages, at which point the first replication barrier is reached and growth in halted [176]. This stage has been referred to as senescence, stasis, or M0 (mortality stage 0) and is characterized by a heterogeneous mixture of actively growing cells amongst large, vacuolated, non-dividing cells [177-179]. Cells that emerge from this point have silenced p16INK4a expression, brought about by CpG island methylation of the promoter region [180]. This p16INK4a associated growth arrest can be avoided when HMECs are grown with a layer of fibroblasts, suggesting the arrest is a function of 40  the in vitro culture conditions and is not required for eventual cellular immortalization [181, 182]. Regardless of whether the first barrier is imposed or not, HMECs will continue to grow exponentially for an additional 10-20 passages until a crisis point is reached (also termed M1). This stage is defined by a delicate balance of cells undergoing active proliferation and apoptosis, with no net gain in cell number. Leading up to this growth plateau, rapid accumulation of gross chromosomal abnormalities and telomere shortening occurs concurrently with slowing proliferation [183]. Without intervention, spontaneously immortalized lines have not been reported to emerge after this point and the culture expires [179]. Immortalization can, however, be achieved by prolonged exposure to carcinogens, infection with HPV-16 E6/E7 or SV40 Large T-Antigen, overexpression of telomerase cDNA, or various oncogenes [184-189]. Aside from telomerase overexpression, these methods of immortalization have all been reported to result in cell lines with aberrant karyotypes and are not considered as suitable model systems for studying normal mammary epithelium development [187, 190-193]. In addition to cell lines derived from normal reduction mammoplasty tissue, several cell lines have been developed from benign tumours or tissues that are not considered pathologically normal. The most famous is the MCF-10 series of cell lines that spontaneously immortalized from a culture of tissue taken from a patient with fibrocystic disease [194, 195]. These lines are not cytogenetically normal and are heterogeneous in their karyotypes. The MCF-10F cell, derived cells floating within the original culture, were originally specified as 46,XX,Ip+,t(3;9)(pl3:p22). The MCF-10A cells, derived from the attached cells within the original culture, were 48,XX,3p,6p+,+8,9p+,+16. Since this original publication, the karyotype for the MCF-10A cells has been reported as increasingly abnormal [196]. A similar scenario is seen with the HMT-3352 cells derived from fibrocystic breast tissue and the Hs 578 Bst cells derived from normal tissue resected from adjacent carcinoma [197, 198]. While these cell lines have proven beneficial for other work, our goal of delineating the interactions of normal epithelial cells within breast tissue is dependent upon a model system that accurately parallels normal biology. The importance of diploid cells to achieve this has long been recognized, as has the need for clonality within model cell lines to reduce experimental variability [199].  41  The ideal cell line for replicating normal mammary physiology would be nontransformed, diploid, genomically stable, express markers of normal epithelium, and resemble primary epithelium in 2D and 3D culture. The cell line that comes closest to fitting these criteria is the 184-hTERT line generated at the National Institute of Health by Dr. C. Barrett. This line was generated through forced expression of telomerase cDNA in HMEC cells shortly after surpassing the first growth barrier [167]. However, this line is not clonally derived and thus the properties attributed to it may be a manifestation of a collection of cells rather than a single cell type. This is an important, yet subtle, point when studying a tissue composed of a hierarchical system of cells with varying capacities for differentiation and lineage restriction. Additionally, this line has a 48, XX, +20, +20 karyotype when assessed in cells that were passaged at least 30 times post-infection [200]. We analyzed the earliest available passage of 184-hTERT cells and found them to be diploid. Efforts were undertaken to generate clonal derivatives from these cells and assess their growth properties in relation to mammary epithelial cells, more specifically bi-potent progenitor cells. These cell lines are the closest model system we could find that are diploid and resemble the in vitro growth of HMEC cells. Whilst the use of primary tissue is the gold standard for studying mammary gland biology, specifically mammary progenitor cells, these lines may prove advantageous in certain scenarios when large quantities of homogeneous cells are desired or when ease of genetic manipulation is paramount. 3.2 3.2.1  Results Generation of the 184-hTERT parental cell line and clonal derivatives The 184-hTERT cells are an immortalized mammary epithelial cell line generated  through the forced expression of telomerase cDNA in cells derived from a reduction mammoplasty sample (patient number 184). This sample was obtained by Dr. M. Stampfer and was used extensively in her work characterizing the growth of normal, immortalized, and transformed human mammary epithelial cells, for which she generated at least 39 derivative lines from this patient alone [179, 192]. Some discrepancy exists regarding the age of the patient, with both 21 and 48 years cited, but the pertinent fact  42  that the sample was deemed free of any detectable pathology is unvaried [187, 201]. Following resection, the sample was enzymatically digested with collagenase and hyaluronidase to breakdown the stroma and basement membrane, respectively. The resultant epithelial organoids were crudely separated from the stomal cells through filtration and differential centrifugation [202]. Organoids were seeded in serum-free MCDB 170 media, which is the predecessor to the commercially available Mammary Epithelial Growth Media (MEGM) used throughout this work [176]. Under these conditions, epithelial cells flourish while the growth of any remaining fibroblasts is hindered. After approximately 4 passages, primary HMECs undergo a stress-associated senescence from which a subpopulation of cells arises with p16INK4a expression silenced through promoter methylation [203]. Whether p16INK4a silencing is an artifact of culture, or whether it is a variant population of cells from the original tissue containing this aberrant methylation that subsequently dominates the culture, remains controversial [204, 205]. It is at this stage that the 184-HMEC cells were obtained for immortalization with telomerase (Figure 3.1)[167]. The parental 184-hTERT cells generated by Dr. Carl Barrett contain multiple copies of Chromosome 20 [200]. The manner in which the cell line was generated leaves questions as to the mechanism behind this chromosomal gain. As the line was never clonally derived, this phenomenon may be driven through selection of an aneuploid cell during culturing or through an intrinsic mechanism inherent to the patient sample. Chromosome 20 contains a multitude of genes involved in regulating proliferation and preventing apoptosis and thus a growth advantage could conceivably be provided by its gain. We obtained the earliest available passage of 184-hTERT cells (passage 6 post telomerase infection derived from passage 11 184-HMECs). These cells, along with the original 184-HMEC line prior to infection, were diploid (data not shown) [190]. This compelled us to sub-clone a series of derivative lines to elucidate if a variant aneuploid line exists within the parental cells that subsequently dominates the culture (examined further in Chapter 6). This was accomplished by seeding the parental 184-hTERT cells at limiting dilution in the presence or absence of irradiated NIH 3T3 feeder cells. From an initial 103 wells where the presence of a single cell or colony (when feeder cells were used) was confirmed after seeding, 69 lines survived expansion to a 75cm2 flask. No  43  differences in cloning efficiency, growth rate, or cell morphology were observed between lines established with or without the presence of feeder cells (data not shown). These lines were assigned a letter and number to act as an independent identifier. The letters reflect the initial cellular morphology as assessed by bright-field microscopy whereby cells were grouped as spindle-like (assigned the letter S), cuboidal-like (assigned the letter L), intermediate between spindle and cuboidal-like (assigned the letter E), and small, clear, densely-packed cells (assigned the letter C). 3.2.2  Identification of 14 independent derivative clones by integration site analysis Southern blotting was used to confirm the presence of a single telomerase  integration site from the 69 184-hTERT clonal lines that were derived (Figure 3.2). This also served to group the lines on the basis of integration site and identify the lines by cell of origin stemming from the original 184-HMEC sample. As outlined in Table 3.1, 14 unique integration sites were identified amongst the lines. Two of the integration site groupings comprise 45% of the cell lines and were the only 2 detectable sites within the parental line, signifying that low-level heterogeneity can be difficult to assess through Southern Blotting. Seven lines did not have a detectable integration site and thus likely lack exogenous telomerase expression. Without telomerase, these cells have either undergone a rare spontaneous immortalization event to continue their growth, or have yet to reach the second HMEC crisis phase and would naturally senesce upon reaching this point in culture. 3.2.3  184-hTERT lines have an antigen profile typical of myoepithelial cells Within normal breast tissue, luminal and myoepithelial cells can be classified by  their morphology and position within the tissue architecture. However, upon dissociation and culturing we must rely upon surrogate markers to distinguish cell types. This can be done with an expanding panel of markers, the most common of which are the cytokeratins [51, 206-208]. Antibodies directed against cytokeratin 18 will exclusively react with luminal cells; antibodies to cytokeratin 5 will react with myoepithelial cells and some luminal cells [207, 209]. A panel of clonal lines was subjected to immunoblotting and found to be cytokeratin 5 positive and cytokeratin 18 negative  44  (Figure 3.3). This can be contrasted to a sample of freshly isolated and uncultured HMEC cells containing both luminal and myoepithelial cells (148-HMEC) that is positive for both cytokeratins. Also detectable at low levels is p63, a strict myoepithelial cell marker in the breast [210]. Although immunoblotting was not positive in all of the clonal lines, staining by immunofluorescence confirmed its expression in the remaining clonal lines (data not shown). Staining for E-cadherin confirms that these cells are epithelial, though the lack of E-cadherin staining in the 148-HMEC sample is curious and may be a function of the harsh enzymatic digestion protocol these cells were subjected to prior to lysis. Figure 3.3 also provides confirmation of p16 silencing in the clones, shows that p21 is expressed, and that ERK1/2 signaling is active in these cells. Although all myoepithelial cells can be considered basal cells, the reverse is not true despite the propensity within the breast literature to use these terms interchangeably [211]. Referring to cells as basal-like has become commonplace following the introduction of molecular sub-typing of breast cancers [97]. This has translated into the introduction of a panel of biomarkers for immunohistochemistry that further commits a subtype of the triple negative tumours to an even worse prognostic group, the basal-like cancers [100]. This was accomplished by the simple addition of EGFR and cytokeratin 5/6 positive staining to the current panel of antibodies used to define a triple-negative tumour (negative staining for estrogen receptor, progesterone receptor, and Her-2). 184hTERT cells are ER, PR and Her-2 negative, and EGFR and cytokeratin 5/6 positive (Figure 3.4). Additionally, these cells are negative for the androgen receptor and express wild-type p53. Whilst normal cells should not be classified as basal-like due to the confusion it can create by falsely implying a cell-type of origin in malignancy, the status of these biomarkers is relevant when designing model systems to study tumourigenesis. 3.2.4  Clonal derivatives are cytogenetically normal and non-tumourigenic As late passage 184-hTERT parental cells retain extra copies of chromosome 20,  it was imperative to survey the clonal lines for any cytogenetic anomalies that may exist. Two clonal lines were surveyed for gross cytogenetic or karyotypic changes using arraybased Comparative Genomic Hybridization (aCGH) and Multiplex Fluorescence In Situ Hybridization (M-FISH). Figure 3.5 shows the absence of any unbalanced DNA copy  45  number changes within the 184-hTERT-L2 and 184-hTERT-L9 clonal lines, signifying the absence of any whole or partial chromosomal deletions or amplifications. 184hTERT-gfp cells, a clonal derivative from the late passage parental line, were included in the aCGH comparison to survey the cells for any unreported numerical abnormalities and confirm the reported chromosome 20 aneuploidy [200]. M-FISH confirmed the absence of any balanced chromosomal rearrangements in the 184-hTERT-L9 cells (Figure 3.6A). One extra copy of chromosome 20 was detected in 2 of the 15 metaphases examined for the 184-hTERT-L2 cells. This finding necessitated further investigation into the chromosome 20 status of the clonal lines. The levels of chromosome 20 in 5 clonal lines was quantified by interphase Fluorescence In Situ Hybridization (FISH) using probes detecting the p and q arms of chromosome 20 (probe locations depicted in Figure 3.6B). Low-level changes in ploidy were detected in all of the clonal lines at an early passage number (Figure 3.6C). The gain and loss of chromosomes within diploid cells during in vitro culture has been recognized, as has the existence of ploidy mosacism within normal tissues [212-215]. In a meta-analysis of short-term culture of female human peripheral blood lymphocytes, an average of 5.5% of putatively normal cells were aneuploid [216]. As such, these clonal lines are within a range where they can be considered to be cytogenetically and karyotypically normal. Assessing tumourigenic potential is an important step in determining the relevance of these lines as a model for normal mammary epithelial cells. Traditionally, human breast cancer cells have been orthotopically grafted into the mammary fat pads of athymic nude mice [217]. While this is a convenient model system, the stromal environment does not resemble the human breast and the resultant histology of transplanted malignant cells does not resemble human tumours. Combining human cells with fibroblasts in collagen gels and subsequently engrafting these gels under the kidney capsule of immunocompromised mice has been a major advancement in the in vivo growth of normal and malignant breast cells [218]. Additionally, the introduction of recipient mice wholly incompetent for T, B, and Natural Killer cell function, has improved human breast cell engraftment [18, 219]. Using this system, we have shown that the 184-hTERT clonal lines do not form palpable tumours in mice (Figure 3.7). This is consistent with what has previously been seen with derivative lines of the 184 patient  46  sample immortalized through exposure to benzo[a]pyrene; cells did not form tumours upon injection into the mammary fat pads of nude mice [187]. 3.2.5  Clonal derivatives form structures in 3D culture reminiscent of normal  mammary acinar units When placed within a reconstituted basement membrane derived from Englebreth-Holm-Swarm tumour cells (Matrigel), primary HMECs are able to form spherical structures reminiscent of normal acinar structures found within normal breast tissue [220]. Prospective fractionation of primary HMECs into luminal-progenitor enriched, bi-potent progenitor enriched, mature luminal, and stromal fractions (using EpCam and CD49f staining) revealed that only the progenitor enriched fractions were able to form 3D structures in Matrigel [32]. The luminal progenitor cells form primarily single-layered structures that are cytokeratin 18 positive. The bi-potent progenitor cells form dense, multi-layered, cytokeratin 5/6 positive spheres that, upon close inspection of the representative images, contain some regions of squamous differentiation in the inner ring of cells. Under the same assay conditions, the 184-hTERT clonal lines form structures that are similar in morphology and immunophenotype to those generated by the bi-potent progenitor cells (Figure 3.8, Table 3.2). The characteristic squamous cell morphology of the inner ring of cells and the formation of keratin pearls within these structures clearly shows that squamous differentiation is occurring in the clonal lines and in the HMEC structures. This phenomenon has been previously described to occur in the 184-hTERT parental line [57]. However, we have found that this inner ring of squamous cells expresses luminal markers (cytokeratin 18, progesterone receptor and mucin-1). This suggests that these cells are somewhat capable of differentiating down the luminal lineage, albeit aberrantly, as expected from bi-potent progenitor cells. As this squamous differentiation can be seen with freshly dissociated HMECs, it is a property that likely reflects the artificial model system used rather than being an inherent property of the 184hTERT cell lines.  47  3.2.6  184-hTERT lines display a dose-dependent growth response to feeder cells The in vitro growth of mammary progenitor cells has been qualitatively described  as markedly enhanced by the presence of feeder cells [29, 31, 221]. To quantify this effect, bi-potent progenitor cell enriched fractions (CD49fhiEpCam-) were sorted from dissociated reduction mammoplasty tissue that had been depleted of hematopoietic and endothelial cells. When plated in colony forming assays along with an increasing density of irradiated fibroblasts, a dose-dependent growth response was observed (Figure 3.9). This response saturated at 160,000 irradiated NIH 3T3s per 60 mm dish, or 7400 cells per cm2. The 184-hTERT cells mimicked this dose-dependent growth response to feeder cells, with the response saturated at 8400 cells per cm2. 4200 cells per cm2 is a critical density for the 184-hTERT cells in this assay, as evidenced by 2 replicates having an intermediate growth response at this point and 2 replicates having little response. Fibroblasts are known to possess an instructive role in regulating mammary epithelial cells in normal development and oncogenesis [17, 218, 222]. As these interactions remain ill defined, the availability of a cell line that is reliant upon fibroblasts for lowdensity growth provides the opportunity to delineate this relationship in its most simplistic form. 3.3  Discussion The study of the human mammary gland is simplified by the availability of  discarded reduction mammoplasty tissue from non-diseased patients. Ready access to normal human tissue is advantageous to advance the understanding of both development and oncogenesis within a system. However, experimentation with primary tissue is not without limitations and therefore model systems that balance mimicry of the tissue of origin with ease of manipulation are required. We generated a series of clonal lines from the parental 184-hTERT cells that are diploid. This is the clear benefit of these cells over any other available cell line generated to date. While these cells express myoepithelial cell markers in 2D culture, they retain some capacity for differentiation and are able to express luminal markers when placed into 3D Matrigel culture. These structures resemble those formed by a fraction of cells enriched for bi-potent progenitor cells when placed into Matrigel [32]. This fraction also  48  contains the cells that are able to generate human ductal structures when transplanted into cleared mammary fat pads humanized with mammary fibroblasts. As this is not a pure population of cells, only an enriched fraction, it unclear whether the cells that form structures within Matrigel are the exact cells that repopulate the mammary gland. However, the ability of these cells to form structures reminiscent of normal mammary acini, and express markers from both cell lineages, allows them to be termed bi-potent progenitor cells regardless of their intersection or overlap with the mammary repopulating cells. The in vitro growth of bi-potent progenitor cells is reliant upon the presence of feeder cells. We showed a dose-dependent growth response of the bi-potent progenitor cells to an increasing density of irradiated feeder cells in the colony-forming assay. The 184-hTERT clonal lines mimicked this response and thus behave in a similar manner to bi-potent progenitor cells under colony-forming conditions. Interactions with fibroblasts have been crucial for the success of in vitro and in vivo assays used to detect the presence of mammary progenitor cells. This, combined with the role fibroblasts play in the developing mammary gland, led us to investigate this relationship further using a model system based upon the growth properties of the 184-hTERT clonal lines described here.  49  A  Reduction Mammoplasty Sample  Cultured in MEGM  184-HMECs passage 11 (Stampfer)  Infected with hTERT  184-hTERTs passage 1 (Barrett)  Cultured in MEGM  Clone from single cells at passage 6  Late passage heterogeneous 184-hTERTs (various)  184-hTERT Clonal lines (Aparicio)  HpaI  B  5 MoMuSV LTR  psi  HpaI  hTERT  IRES  Neor  3 MoMuSV LTR  Figure 3.1 - Generating the 184-hTERT cell lines A) Lineage of the 184-hTERT cell lines from the initial culturing of reduction mammoplasty tissue to sub-cloning. B) Telomerase cDNA construct used in the generation of the 184-hTERT cells.  50  HindIII digest Clone name  gfp  L2  L5  BglII digest L9  gfp  L2  L5  L9  8 kb  3 kb  1.6 kb  Figure 3.2 – Southern Blot Hybridization of the 184-hTERT clones Representative Southern Blot from 2 independent digests of 10ug DNA hybridized with a neomycin resistance cDNA probe. A clonal line derived from late-passage 184-hTERT cells, termed 184-hTERT-gfp, is included here along with 3 representative clonal lines termed the L2, L5 and L9 cells. L5 and L9 have the same, single, integration site; L2 has a differing integration site confirmed by 2 independent restriction enzyme digests.  51  Figure 3.3 – 184-hTERT clonal lines express markers typically associated with myoepithelial cells 10ug of protein was loaded onto a NuPAGE 4-12% gradient Bris-Tris gels and transferred to nitrocellulose prior to immunoblotting. The membrane was cut into strips according to molecular size, and stripped for re-probing where necessary. 148-HMEC refers to a freshly thawed primary reduction mammoplasty sample used to generate a control lysate for the blotting. 52  L2  L9  L2  EGFR  E-cad  Her2  CK5/6  ER  Ki67  AR  p53  L9  Figure 3.4 – Status of clinically relevant markers in 184-hTERT clonal lines Formalin-fixed paraffin-embedded cell blocks were generated from confluent cultures of L2 and L9 cells. Immunohistochemical analysis was performed using antibodies raised against markers currently used in the IHC-based definitions of breast cancer subtypes.  53  Segmentation for 184-hTERT-L2 (300000bp) 2  Log2 Ratio  1  0  1  2  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  19 21 X 18 20 22 Y  Segmentation for 184-hTERT-L9 (300000bp) 2  Log2 Ratio  1  0  1  2  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  19 21 X Y 18 20 22  17  19 21 X 18 20 22 Y  Segmentation for 184-hTERT-gfp (300000bp) 2  Log2 Ratio  1  0  1  2  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  Figure 3.5 – 184-hTERT clonal lines are cytogenetically normal Array-based comparative genomic hybridization was performed using the HG18 WG CGH Whole Genome Tiling Array. Log2 ratios of signal intensity for 184-hTERT cell lines compared to normal human female reference DNA are plotted in relation to chromosomal position. Clones L2 and L9 (passage 11) exhibit a normal profile devoid of any gross cytogenetic changes. The late passage clonal line 184-hTERT-gfp exhibits amplification of chromosome 20.  54  Figure 3.6 – 184-hTERT clonal lines are karyotypically normal A) Multiplex FISH of L2 cells (passage 11) and L9 cells (passage 12). L9 cells possess a normal, diploid karyotype, whereas two of the L2 metaphase spreads were 48XX+20. B) Location of the Chromosome 20 probes used for interphase FISH C) Distribution of chromosome 20 levels in 5 clonal lines enumerated by interphase FISH with probes targeting the p and q chromosomal arms (n=150 to 152).  55  Figure 3.7 – 184-hTERT clonal lines are non-tumourigenic Representative H&E stained sections from collagen gels transplanted under the kidney capsule of NODSCID IL2RKO mice with irradiated 10T1/2 feeder cells for 4 weeks. Palpable tumours did not form under these conditions.  56  A  HMEC  L2  L5  L9  H&E  CK 5/6  CK 18  PR  ER  57  Figure 3.8 – Retention of differentiation capacity in clonal 184-hTERT lines A) Formalin-fixed paraffin-embedded 21-day Matrigel cultures were stained for markers characteristic of luminal and myoepithelial cells. Keratin pearls (arrows) can be found within structures arising from the clonal lines and from some of the primary HMEC cells, a hallmark of squamous differentiation. While cells are mainly CK5/6 positive, some of the inner cells stain for luminal markers CK 18 and PR. B) Confocal microscopy of 21-day Matrigel cultures stained for F-actin (phallodin) confirms the squamous morphology of the inner cells and an acellular lumen. Staining for Muc-1 confirms that the inner cells express luminal markers.  58  Figure 3.9 – Colony formation changes relative to irradiated NIH 3T3 density A) Number of colonies formed from plating 500 prospectively sorted bi-potent progenitor cells with increasing numbers of irradiated NIH 3T3 cells plated per 60 mm dish (n= 3 patient samples). Colored crosses represent colony counts for individual patient samples. B) Number of colonies formed from 100 184-hTERT cells plated with increasing numbers of irradiated NIH 3T3 cells per 35 mm dish (n=4).  59  Integration site A B C  Clonal Lines C1, C2, C9, E8, E14, L5, L7, L8, L9, L10, L11, L15, S5, S6, S9, S29 C7, S2, S10, S14, S15, S17, S18, S21, S22, S23, S25, S30, S31, S34, S35 E1, E4, E6, E12 ,L4, L6, L16, S4, S33, gfp (late passage clonal line)  D  E9  F  C4, E3, E11, L14, S32  G  E13  H  E15  I  L1, L13  J  L2, C6, S1, S19  K  C5  L  S28  M  C8  N  E2  O  L3, S8, S13, S20  undetermined 2 sites  C3, L12, S11, S12, S26, S27, S28 E5  Table 3.1 - Classification of clonal lines based on telomerase integration site Single-site telomerase integration was confirmed through Southern Blotting and the restriction enzyme digested DNA patterns were used to group the clonal lines into 14 unique categories. In the parental 184-hTERT line, only integration sites A and B were visible through Southern Blotting. Integration sites were not detected in 7 lines, potentially signifying an absence of telomerase from these lines. One line contained 2 integration sites and thus was either not clonal, or contained 2 integrated copies of the telomerase cDNA construct.  60  Average Plating Clonal Line  Efficiency  95% CI  C2p17  5.60%  4.86% - 6.34%  L5p17  7.50%  5.28% - 9.72%  L8p19  10.80%  9.14% - 12.46%  L2p18  11.00%  5.73% - 16.27%  L9p17  7.70%  6.68% - 8.72%  Table 3.2 – Plating efficiency of clonal lines in 3D matrigel culture Percentage of cells forming 3D structures after 21 days when 1250 cells were plated in Matrigel coated 8-well chamber slides (n=7).  61  4  Systematic identification of signal transducers functionally required  for mammary progenitor cell growth 4.1  Introduction Mammary epithelial cells exist within a stroma of extracellular matrix proteins  and supporting cells, namely fibroblasts, adipocytes, endothelial cells and lymphocytes. In addition to structural support, the stroma contributes to mammary gland development and malignant progression. By embryonic day 12 in the mouse mammary gland, a layer of mesenchyme consisting mainly of fibroblasts surrounds the epithelial cells of the rudimentary mammary bud. A separate mesenchyme appears by embryonic Day 14 that is the precursor to the fat pad. By day 16, the epithelial cells and the two mesenchymes converge as the rudimentary bud grows towards and penetrates the fat pad. This becomes increasingly laden with fat and resembles adipose tissue by 2 days postpartum [223]. When embryonic mammary epithelial cells are transplanted with either of these 2 mesenchymes into the adult fat pad or under the kidney capsule, the fibroblast mesenchyme stimulates the formation of ductal hyperplasia and the fat pad mesenchyme supports typical ductal growth [223]. The influence of the mesenchyme on regulating epithelial cells also extends across tissues. When mammary epithelial cells are explanted and grown in combination with salivary gland mesenchyme, they form branched structures typical of what is seen with salivary gland explants, and vice versa [224]. This suggests that the mammary epithelial cells remain plastic and are heavily influenced by their stromal environment. The stroma of the mammary gland differs between species. Within the mouse, the mammary fat pad is predominately composed of adipocytes, with fibroblasts  62  interspersed throughout. A discontinuous layer of fibroblasts also lies directly adjacent to and in intimate contact with the basement membrane surrounding the ductal epithelium [225]. In humans, the terminal ductal lobular units are embedded within fibrous connective tissue stroma and are generally not found to be associated with the regions of adipose tissue found within the breast [226]. Fibroblasts secrete at least 3 types of collagen and contribute to the composition of the stroma within which they reside [227]. During puberty, the extending ducts are tightly encased by fibroblasts, whose proliferation precedes that of the epithelial cells [2]. Successful transplantation of normal human mammary cells into the mouse mammary fat pad will only occur if the surrounding human stroma is transplanted, or if the fat pad has been previously colonized by human fibroblasts [17, 228]. Alternatively, human mammary epithelial cells will form ductal structures when embedded within collagen and transplanted subcutaneously into athymic mice. In this system, mouse fibroblasts infiltrate through the gel, surround the ductal structures, and contribute to the formation of a basement membrane that envelopes the human epithelium [229]. Transplant take rates were improved by adding mammary fibroblasts to the collagen and placing the gels under the kidney capsule of immunocompromised mice to increase vascularization [218]. These transplantation models also support the growth of primary breast tumours, a task which has been otherwise difficult to achieve [17, 230]. Inclusion of irradiated NIH 3T3 cells along with marginally tumourigenic or non-tumourigenic cell lines form a variety of tumour sites in subcutaneous xenografts and resulted in accelerated tumour formation and shorter latency periods [231]. Taken together, this suggests that the stromal environment, namely the presence of fibroblasts, plays an important role in driving the growth of human mammary  63  epithelial cells during development and malignancy. Initial work in the cultivation of mouse mammary tissue identified a role for fibroblasts in directing the orientation and differentiation of epithelial cells. Adipocytes provide the fats required for the epithelial cells to produce milk during pregnancy [232, 233]. In human mammary cells, the ability to clone cells is markedly enhanced by the presence of fibroblast feeder layers [25]. Assays to quantify human mammary progenitor cells in 2D culture take advantage of this and rely upon the presence of fibroblasts for growth at clonal densities [30, 31]. In vitro growth of epithelial cells can also be accomplished in 3D. Initial studies aimed at maintaining the polarity and differentiation of mammary epithelial cells from a pregnant mouse used 3D floating collagen gels to accomplish this [234]. The addition of fibroblasts to these floating collagen gels seeded with human mammary cells induced branching morphogenesis within otherwise spherical acini [235]. Primary mouse and human epithelium can also be cultured in reconstituted basal lamina collected from Engelbreth-Holm-Swarm tumour cells (commercially available as Matrigel). Grown under these conditions, cells form hollow alveolar-like structures with apical-basal polarity and the ability to synthesize β-casein [236]. The cellular and extracellular compartments of the stroma seemingly contribute to different aspects of mammary gland development. The use of fibroblasts in epithelial cell cultures can be replaced through the addition of fetal bovine serum or bovine pituitary extract to defined media optimized for epithelial cell growth [176, 237, 238]. Attempts to develop a media without one of these supportive factors have failed to produce the same growth response and require the  64  addition of over 15 exogenous growth factors and hormones [239]. With that in mind, a variety of principal agents supporting mammary epithelial growth have previously been identified. However, no single factor can account for the complexity of responses seen when growth is supplemented with undefined components and/or extracellular matrix. Insulin and epidermal growth factor (EGF) are the most crucial known factors in epithelial cell growth as the elimination of either of these from defined media had the greatest effect on clonal growth [176]. Elimination of hydrocortisone, ethanolamine, phosphoethanolamine and prostaglandins had moderate growth effects. Other mitogens reported to increase epithelial cell proliferation in the mammary gland include the insulin-like growth factors, fibroblast-like growth factors, hepatocyte growth factor, growth hormone, and transforming growth factor alpha [240-244]. Conversely, TGF-β was shown to inhibit in vivo mammary epithelial cell growth and morphogenesis [245]. The addition of cholera toxin to epithelial cells grown in conditioned media and fetal bovine serum increased their intracellular cAMP levels and subsequent proliferation rate [246]. Any agent that activated intracellular cAMP was growth promoting when provided to epithelial cells in vivo, including cholera toxin, isoproterenol, dibutyryl cAMP, and forskolin [247]. From the response of the epithelium at puberty, it is obvious that the hormones estrogen and progesterone play a role in regulating epithelial cell growth. Estrogen stimulates mammary epithelial cell growth directly in vitro, but only in the presence of fibroblasts [248]. When injected into mice alone, estrogen and progesterone stimulate mammary cell proliferation, with additive effects upon coadministration [249]. Despite our broad knowledge of what regulates mammary epithelial cell growth, we are still unable to replace fibroblasts or undefined soluble  65  growth additives entirely. Here, using systematic RNA interference screening, we have identified previously unknown regulators of fibroblast-dependent mammary epithelial cell growth. The diversity amongst these newly identified growth regulators suggests that previous studies looking at the mechanisms of epithelial cell growth had only begun to uncover the complexity that exists in this process. The inability to identify the less obvious regulators of mammary epithelial cell growth suggested that a fresh approach to this question needed to be taken. Recent advancements in high-content screening made a genome-wide search for genes with a functional effect on fibroblast-dependent mammary epithelial cell growth through RNA interference possible. In normal mammary cells, genome-wide RNA interference screens have been used to identify mediators of Ras oncogene-induced senescence, suppressors of p16 gene expression, genes that regulate cell migration, and cell survival genes [250253]. These screens were conducted using either siRNA or lentiviral shRNA libraries. The latter format is advantageous in that it can be conducted by infecting large populations of cells with several thousand viral particles at once. By comparing the shRNAs present in the initial population of cells to those present after a selective pressure is applied, identification of the shRNAs negatively affected by the pressure is possible. This methodology is subject to biases within the viral packaging, infections, selection and final shRNA detection. High false negative rates limit the comprehensiveness of this approach. Nevertheless, they are relatively easy and inexpensive to accomplish. Genome-wide siRNA screening is limited in its implementation, as each targeted gene must be assessed individually. This mandates the use of liquid handling robotics to prepare and deliver complexed siRNAs, and high-throughput end-point detection  66  systems. The scale and expense of siRNA screens, or individually assessed shRNA screens, can be prohibitive. However, they provide the most comprehensive and least biased approach to RNA interference screening. The greatest limitation here is the identification of a reproducible model system that can be adapted to high-throughput screening and is amenable to RNA interference [254]. In Chapter 3, we identified a series of clonal cell lines that mimicked the fibroblast-dependence seen within primary mammary epithelial cells. At low densities, 184-hTERT cell growth can be supported by either the presence of fibroblasts or bovine pituitary extract. Gene expression profiling suggested that the growth signals provided under each condition were distinct. To specifically understand fibroblast driven growth, we developed a multi-well plate based assay whereby 184-hTERT cells in co-culture with irradiated NIH 3T3 cells were transfected with siRNAs targeting the entire genome. As we were interested in growth signals emanating from the fibroblasts, we focused the validation screen on cell surface and secreted targets whose silencing caused a statistically significant decrease in cell growth. After assessing reproducibility and eliminating off-target effects, we identified 49 genes that are functionally required for epithelial cell growth. These genes have varying effects on mammary progenitor cell growth and the formation of 3D acinir structures in Matrigel. The striking diversity of the genes found here suggests that functional screening will challenge our current perspective on regimented gene classification and functions.  67  4.2 4.2.1  Results Low-density growth of 184-hTERT cells can be supported by bovine  pituitary extract or feeder cells The long-term subculture of mammary epithelial cells has been previously characterized to require either the presence of bovine pituitary extract or irradiated feeder cells [176, 221]. In chapter 3, the 184-hTERT clonal lines were shown to mimic the growth of bi-potent progenitor cells when plated with increasing densities of irradiated NIH 3T3 feeder cells in the well-defined colony-forming assay. This assay is not amenable to high-content screening and thus we developed a co-culture model whereby epithelial cells were plated at low-densities with irradiated NIH 3T3 cells in a 96-well plate format. Initial assay optimizations were performed with green fluorescent protein labeled 184-hTERT cells (184-hTERT-gfp), clonally derived from late-passage parental 184-hTERT cells by the laboratory of Dr. Connie Eaves [200]. Within this assay, the epithelial cells remain dependent on fibroblasts for growth. Figure 4.1A outlines the growth characteristics of these cells in 4 common media conditions (a detailed description of the media compositions is provided in section 2.1.1). Irrespective of the presence of fibroblasts, DMEM + 5% FBS does not support the growth of 184-hTERTgfp cells. Serum-free 7, and its commercial counterpart Epicult B, are both designed for detecting the presence of mammary progenitor cells within the colony-forming assay. In the absence of fibroblasts, these media formulations did not support 184-hTERT-gfp cell growth. With the addition of fibroblasts, the number of epithelial cells steadily increased relative to the number of fibroblasts plated within the assay. When fibroblasts are present, these two medias are comparable at supporting growth. MEGM, which contains  68  bovine pituitary extract, is routinely used to sub-culture the 184-hTERT cells. It supports epithelial cell growth regardless of the presence or absence of fibroblasts, confirming that low-density growth of 184-hTERT-gfp cells can be supported by the presence of either bovine pituitary extract or fibroblasts. Interestingly, including both did not increase epithelial cell growth beyond what is seen in either condition alone. This suggests that they either stimulate growth through the same mechanism or that the growth signals being provided within this assay are saturating and epithelial growth capacity is already maximized. Serum-free 7 media provides a means through which we can analyze fibroblastdependent epithelial cell growth. However, upon further investigation, the cholera toxin contained within this media was toxic to all of the 184-hTERT lines that we tested following lipid-mediated transfection (data not shown). To identify an alternate growth media, we analyzed the components of MEGM to determine if a modified version of this media would support fibroblast-dependent but not independent growth. For these experiments, we used the 184-hTERT-L9 cells, which are not labeled with GFP, and thus could not be discriminated from irradiated NIH 3T3s during imaging (Figure 4.1B). As the fibroblasts are growth arrested, their numbers remain constant. The un-supplemented basal media for MEGM does not support epithelial cell growth (commercially available as MEBM). Using this condition, the background number of fibroblasts within our cell counts is discernable by comparing the 0 and 2000 irradiated fibroblast conditions. This background count is always found to remain consistent and is thus largely ignored in this assay. Base media refers to the MEBM basal media with the addition of hydrocortisone, transferrin, and isoproterenol. The addition of these components did not stimulate  69  epithelial cell growth (Figure 4.1B). The addition of EGF alone or insulin alone to the base media minimally stimulated growth in both conditions, as does the addition of bovine pituitary extract alone (Figure 4.1B). An exponential increase in epithelial cell growth occurs in the fibroblast-dependent condition when the base media is supplemented with both EGF and insulin; minimal stimulation occurs in the fibroblast independent condition (Figure 4.1B). This effect is comparable to that previously seen with serum-free 7 (Figure 4.1A). “Visvader media” was included here as a comparator as it is similar to serum-free 7 media with the exception of containing 5 times the amount of insulin (1 µg/ml versus 5 µg/ml, respectively); the same amount as found within the MEGM media [12]. MEGM media without the addition of bovine pituitary extract, henceforth referred to as assay media, can be used to evaluate fibroblast-dependent growth and is not toxic to cells during lipid-mediated transfection.  4.2.2  Growth signals provided by bovine pituitary extract are different than those  provided by feeder cells Growth of the 184-hTERT cell lines at low-densities can be supported in culture by feeder cells or bovine pituitary extract and does not respond additively or synergistically to the addition of both. This suggests that either the same signaling pathways are activated upon growth stimulation, or that the cells are dividing to their maximum capacity in both conditions and the addition of further growth stimulation cannot produce any additional growth. The addition of bovine pituitary extract clearly provides soluble growth factors; the mechanism of signaling from the feeder cells is unknown. To determine if the feeder cells are providing a soluble growth factor to the epithelial cells, a central strip was scraped off from an adhered layer of irradiated NIH  70  3T3 cells prior to the addition of 184-hTERT-L9 cells. No colonies formed on the cleared plastic. The requirement for close proximity to the feeder cells suggests that they are proving a contact-mediated or short-range signaling molecule to the epithelial cells. The absence of feeder cells on the cleared plastic, the distinct fibroblast-plastic border, and representative colonies are presented in Figure 4.2. While we cannot rule out the possibility that the 184-hTERT –L9 cells were simply unable to adhere to the cleared plastic, this is unlikely as these cells are routinely sub-cultured on the same tissue culture plastic without issue. To understand the signaling pathways involved in regulating epithelial cell growth under these two distinct conditions, we grew 184-hTERT cells with either fibroblasts or bovine pituitary extract and assessed their gene expression profiles with Affymetrix GeneChip Human Exon 1.0 ST Arrays. Appendix A lists the gene level changes and their RT-QPCR verified result; Appendix B lists the exon level changes and their RTQPCR verified result. RT-QPCR primer and probe sets were tested against cDNA prepared from irradiated NIH 3T3 cells to ensure that amplification of mouse transcripts was not occurring. When amplification did not occur, primer and probes were redesigned and retested. From the Affymetrix expression analysis, 63% of the gene level changes validated by RT-QPCR, with 12% of the RT-QPCRs failing due to non-amplification of the gene in either growth condition with two unique primer/probe sets. Similarly, 62% of the exon level changes validated by RT-QPCR, with 19% failing due to nonamplification. The RT-QPCR validated gene changes were analyzed using the “Core Analysis” function in Ingenuity Pathways Analysis platform (www.ingenuity.com). This interpreted the expression data in relation to interaction networks and predominant  71  canonical pathways. The top associated network functions of the differentially expressed genes are depicted in Figure 4.3 (cancer, cellular movement, cell-to-cell signaling and interaction). Note that only 11% of the RT-QPCR validated gene level changes are upregulated when the cells are grown with bovine pituitary extract; this is reflected in Figure 4.3 as this group of genes is poorly represented in all of the networks. Identifying the up-regulation of the fibroblast growth receptor (FGFR1) in the co-culture conditions validates this approach as fibroblast growth factors are secreted by fibroblasts. However, FGFs are not sufficient to stimulate mammary epithelial cell growth [255]. Several upstream molecules that feed into the NFκB and Mapk signaling pathways are upregulated in response to fibroblast-dependent growth. Of particular note is Kit ligand (Kitlg); known to induce proliferation in a myriad of cell types. Kitlg and its tyrosine kinase receptor, c-KIT, are found within normal mammary tissue, but their levels decline during malignant progression [256]. The exception is in triple-negative tumours, where c-Kit IHC staining is positive in 45% of cases and predicts reduced overall survival [257]. Myoepithelial cells in normal breast express Kitlg but not c-Kit, whereas luminal cells display variegated expression of both [258]. C-Kit has been recently described to serve as a marker for differentiated luminal cells within the breast [259]. While not explored further here, this may be an important signaling axis involved in luminal differentiation that is initiated from the stroma. Another molecule of particular interest in the mammary gland is matrix metalloproteinase-2 (Mmp-2), which is required for the initial invasion of the mammary epithelium into the fat pad at the onset of puberty, a point at which prolific expansion of the epithelium occurs [260]. Although the mechanism is not clear, it is suggested that  72  Mmp-2 is required for cell survival during branching morphogenesis. Another gene whose expression is induced by feeder cells is the planar cell polarity gene Celsr1, discussed further in chapter 5. Ultimately, it appears that several pathways are activated in epithelial cells grown with feeders and that these pathways do not necessarily overlap with those stimulated by bovine pituitary extract.  4.2.3  184-hTERT cells are susceptible to siRNA knockdown while irradiated NIH  3T3 cells are not To address how fibroblasts are regulating epithelial cell growth, we have the ability to manipulate the expression of complementary signaling molecules on either cell type and judge their effect. Gene expression can either be increased or decreased, with the latter providing a more direct assessment of the function of an individual signaling pathway component. As we do not know what signaling pathways are involved in regulating fibroblast driven 184-hTERT cell growth, we must take a systematic approach to identifying them. Reducing gene expression within cells can be accomplished through the use of siRNA, which is readily amenable to high-content screening. This will allow us to individually interrogate the function of a myriad of genes, but requires that the model we use is susceptible to siRNA mediated gene silencing. Cell lines are generally quite susceptible to RNA interference with lipid-based transfection of siRNA complexes. NIH 3T3 cells are often used for these studies as they are readily transfectable and siRNA produces robust gene silencing. Fibroblasts proliferate faster than normal mammary epithelial cells and thus when they are used as a feeder layer they must be growth arrested by irradiation to prevent them from overtaking the co-culture. Figure 4.4 addresses the efficacy of the use of siRNAs in irradiated NIH  73  3T3 cells. Using 4 different lipid-based transfection reagents, we were unable to effectively silence β-actin expression with a siRNA pool composed of 4 individual siRNAs targeting β-actin (Figure 4.4A). A limited reduction in β-actin expression is seen when higher levels of DharmaFECT 3 were used and so we further examined the use of this reagent. In a 96-well plate format, we transfected irradiated NIH 3T3 cells with increasing concentrations of both DharmaFECT3 and siTOX, a proprietary siRNA control from Dharmacon that is toxic to cells upon delivery (Figure 4.4B). The percentage of viable cells 48 hours post-transfection was assessed by live/dead discrimination with images collected using the InCell Analyzer. Live cell counts were determined by treating cells with calcein-AM, a cell permeable non-fluorescent compound that is cleaved by intracellular esterases to form non-permeable fluorescent calcein in live cells. Dead cells were marked with ethidium homodimer-1, a DNA dye that is only able to enter cells with compromised membranes. siTOX did not significantly affect the viability of the irradiated NIH 3T3 cells suggesting that siRNAs are not functional within these cells. RNA interference can fail due to the inability of the siRNAs to enter into the cells, or the inability of the siRNAs to silence their intended target once inside. To address if siRNAs can enter into the irradiated NIH 3T3 cells, we transfected the cells with increasing concentrations of DharmaFECT3 and siGLOnuclear, a fluorescently labeled oligonucleotide that localizes to the nucleus and indicates effective uptake of siRNAs. By enumerating the number of fluorescent siGLO granules within the cells, we are able to confirm that siRNAs efficiently enter into irradiated NIH 3T3s (Figure 4.4C and 4.4E). There was minimal toxicity associated with siGLO-nuclear transfection (Figure 4.4D). This result should, however, be interpreted with caution as  74  we did not rule out the possibility that the transfection reagent complexed siGLO may be sticking to the surface of the cell creating the false impression of internalization or nonlinearity in the result [261]. We confirmed that the β-actin siRNA pool and siTOX perform as intended in other cell lines and thus their inability to silence their intended targets is not questioned [262]. These experiments were repeated comparing irradiationinduced growth arrested human fibroblast cell lines IMR-90, HFF-1 and MRC-5 with unmanipulated cells. We confirmed that the fibroblasts were only amenable to siRNA induced gene silencing when they were not irradiated (data not shown). To assess if 184-hTERT-L9 cells are susceptible to RNA interference, we transfected them with a de-convolved siRNA pool targeting e-cadherin. Two of the 4 individual siRNAs, as well and the siRNA pool, silence e-cadherin expression (Figure 4.5A). In addition to the extent of knockdown produced by siRNAs, we are also interested in determining the percentage of cells that are transfected with any particular siRNA. To address this, we transfected a siRNA targeting GFP into the 184-hTERT-gfp cells and enumerated the number of cells with silenced GFP expression using the InCell Analyzer (Figure 4.5B). In the control condition, 94% of the cells are GFP positive suggesting that the unlabeled irradiated fibroblasts within this assay only comprise 6% of the total count. Using Lipofectamine 2000 to deliver the siRNA leads to 16% of the cells expressing GFP, suggestive of a transfection efficiency close to 80% when the contributing fibroblasts are considered. When DharmaFECT 3 is used, the GFP positive fraction decreases to 18%. However, this is confounded by the total number of cells dramatically decreasing due to the inherent toxicity of the transfection reagent. Lipofectamine 2000 produced less toxicity than DharmaFECT 3 showing that toxicity is  75  partially dependent upon the transfection reagent employed. Nevertheless, 184-hTERT cells are readily transfectable with siRNAs that are capable of producing robust gene silencing.  4.2.4  Development of a co-culture assay for high-content screening As the signaling pathways required for fibroblast driven epithelial cell growth are  not know, a genome-wide screening approach was taken to functionally identify the crucial pathways that are involved in an unbiased manner. Having developed a coculture model system through which we can interrogate this and a method to modulate gene function within this system, we were able to optimize an assay for use in highcontent screening. A major consideration in high-content screening is assay variability. This can be influenced by the order in which the cells within this assay are plated (Figure 4.6A). Four variations of plating order and media changes were evaluated. We see the least amount of variability when the fibroblasts are plated first in serum-containing media, followed by the 184-hTERT-L9 cells in the assay media. Serum is normally included to aid in cell adherence and appears to be more important for fibroblasts. Plating the 184-hTERT-L9 cells directly in the serum-free assay media does not increase variability as it eliminates a media change, which in itself reduces the potential for cell loss. Plating the 184-hTERT-L9 cells first provides an advantage in that the 184-hTERTL9 cells can be transfected with siRNAs prior to the addition of the fibroblasts. This would eliminate any confounding effects from the siRNAs silencing genes in the fibroblasts. Since it has been adequately proven that silencing does not occur in the irradiated NIH 3T3 cells, this is not truly a concern here. Therefore, for this screen, the  76  irradiated NIH 3T3 cells are plated with serum 24 hours prior to the addition of the epithelial cells in serum-free assay media. The density of cells plated within this assay is influenced by the addition of transfection reagent and complexed siRNAs due to their associated non-specific toxicities. Therefore, the number of irradiated NIH 3T3 and 184-hTERT-L9 cells per well in a 96-well plate must be optimized under these conditions. In Figure 4.6B and 4.6C, we utilized transfection reagent alone as a control condition and the transfection of a putatively neutral non-targeting control siRNA (siCONTROL3), respectively, to determine the number of cells that should ideally be plated per well. Plating 500 184hTERT-L9 cells gave the maximum growth response with the least number of cells and was chosen as the optimal plating density. For the irradiated NIH 3T3s, we are searching for an epithelial response to growth signals emanating from these cells and thus we must be cognizant of not saturating the system. Maximum growth of the epithelial cells is not necessarily the desired outcome. Rather, we are aiming to select a point within the doseresponse curve where the cells could still received further growth stimulation but retain an adequate dynamic range relative to the fibroblast-null condition. We chose 3000 irradiated NIH 3T3 cells as our screening density as it was within the dose-response curve, allowed for a small buffer to accommodate subtle plating variability, and provided a 6.15 fold difference between the fibroblast-dependent and independent conditions. Despite consistent plating numbers, variation in final cell counts can result in identically treated wells due to differences in temperature and liquid evaporation that can occur across the plate. To demonstrate this, plates transfected entirely with Lipofectamine 2000 alone, siCONTROL2, siCONTROL3, and Plk1 were assayed and heat maps were  77  generated from the final cell counts (Figure 4.7). Throughout the screen, we periodically included full control plates and normalized the results of each well to these controls to account for positional plate effects. Control wells included on each individual plate allowed us to assess variability over time and acted as the comparator for the siRNA treated wells. Image acquisition parameters can greatly affect the outcome of a screen and the feasibility of completing it on a genome-wide scale. The ability to detect a greater number of cells during imaging will provide a bigger range within the assay. However, the number of cells per well is inherently constrained by its surface area and is further reduced by an assay end-point that catches cells still within their exponential growth phase. Image acquisition parameters can influence the number of cells detected within a well. For imaging, we utilized the InCell Analyzer and InCell Developer software to collect and analyze images. This is an automated microscope designed for high-content screening and its correlated analysis software. We tested the 4x and 10x objective of the InCell Analyzer, with or without CCD pixel binning (Figure 4.8A). Collected images were batch-analyzed using standard algorithms from within the software. All images were flat field-corrected from a blank reference image and smoothed using 5 iterative passes of an isotropic diffusion filter. The nuclear centre from the 4',6-diamidino-2phenylindole (DAPI) stained channel was detected using intensity and shape information, and the seeded watershed algorithm was applied to detect the nuclear boundaries. Intensity information from the GFP channel was used to discriminate GFP positive and negative cells (Figure 4.8C). The 4x objective (numerical aperture of 0.2) is able to collect a maximum of 4 fields of view from a 96-well plate, ensuring no overlap between  78  fields. The lower numerical aperture limits the amount of light that can be collected from the objective; with lower signal intensity the background noise became an issue and the detection of nuclei was occasionally problematic. This led to lower cell counts with the 4x objective than with the 10x objective (numerical aperture 0.45), where 12 fields were collected covering approximately 50% of the surface area of each well (effectively less cells captured per well than with the 4x objective). Binning functions to improve the signal-to-noise ratio of an image and increase the image frame rate whilst reducing the resolution of the image. Using 4x4 pixel binning with the 10x objective, GFP positive cell counts were even further increased due to improvements in the image segmentation related to higher GFP signal intensities. A final consideration with the 10x objective is the number of fields to acquire during imaging; again related to balancing the image acquisition time with the dynamic range of the assay. We collected either 12 or 21 fields per well from a test-subset of siRNAs targeting the insulin signaling pathway run under screening conditions. Lipofectamine 2000 alone was included as the comparator, Plk1 as a transfection control, and siCONTROL 3 as a non-targeting control. The distribution of effects under both acquisition parameters is plotted in Figure 4.8B. In addition to increasing the overall GFP positive cell counts, the dynamic range between the median values of the comparator and Plk1 positive control also increased when 21 fields were acquired during imaging. The individual siRNAs and controls from the 21-field acquisition data set is plotted in ranked order in Figure 4.9. The relative order of these siRNAs does not change when total cell count is plotted alongside the GFP positive cell count. The GFP negative fraction (irradiated NIH 3T3 cells) remains consistent regardless of the treatment. Thus, total cell count is an acceptable surrogate for GFP  79  positive cell counts within this assay. Eliminating the need to image a second fluorescence channel decreases the imaging time by more than two-thirds (considering individual channel acquisition time and filter wheel rotations). This also allows us to used unlabelled 184-hTERT-L9 cells for screening, rather than the 184-hTERT-gfp cell line. Through careful optimization, we have established assay conditions that will allow us to robustly assess the signaling molecules required for fibroblast driven epithelial growth with minimal imaging acquisition time.  4.2.5  High-content screening identifies 49 signal transducers required for  mammary epithelial cell growth To identify the signaling pathways that drive epithelial cell growth in the presence of fibroblasts, we performed an initial genome-wide screen for genes that functionally affect growth, and continued to screen for reproducibility and off-target effects in a subset of these genes. We focused on genes located on the plasma membrane, or in the extracellular space, as they would putatively transduce the signals emanating from the fibroblasts. Candidate genes were carried forward and screened in primary mammary tissue to gauge their effect on the growth of mammary progenitor cells (see Figure 4.10 for the screening work-flow). The initial genome-wide screen used to identify factors regulating fibroblastdriven epithelial cell growth was conducted using the siGENOME library from Dharmacon. This library targets over 21,000 unique human genes annotated in the NCBI RefSeq database. It is composed of pools of 4 individual siRNAs per gene target, which reduces the effective concentration of each individual siRNA and provides more potent knockdown with less off-target effects [263-266]. The 184-hTERT-L9 cells were  80  screened in duplicate. An image segmentation and post processing software work-flow was developed using Cell Profiler [262]. Cell counts per well produced from the Cell Profiler image analysis were further analyzed using statistical linear regression models, comparing cell counts under the siRNA condition to cell counts under the control condition and adjusting for technical artifacts of well position and plate effect, thereby yielding plate normalized growth effect estimates for each siRNA. Well position effects were assessed by screening additional plates containing the same control condition in all wells, allowing estimation of well position effects such as poorer cell growth in the edge rows of the plates due to uneven conditions encountered in incubators and other processing steps. Plate effects were assessed by using multiple plates for each condition, thereby allowing for adjustment of plate-to-plate variability in the statistical model. The model-estimated cell count under the siRNA condition was then divided by the modelestimated cell count under the control condition to produce a measure of relative effect. The plate normalized growth effect for each individual siRNA relative to the Lipofectamine 2000 control is graphed in ranked order in Figure 4.11A. Highlighted within the graph are the positions of the receptors for insulin, EGF, transferrin, and isoproterenol, which are all defined additives in the assay media. 2337 of the targeted genes, or 11% of the original library, displayed a statistically significant decreased in growth to 25% or less of the control conditions (greater than 75% growth inhibition). Note that this cut-off was stringent enough to exclude the receptors for insulin and EGF, molecules with known biological signaling effects in mammary cells. Therefore, these are putatively less potent than the candidate genes selected for further validation. As we were interested in identifying genes that transduce the signals required for mammary  81  epithelial cell growth, we categorized the targets from the primary screen based upon their cellular location (Figure 4.11B). We focused upon targets located on the plasma membrane and in the extracellular space (included to account for potential autocrine signaling) and carried these forward to a secondary screen. From the candidate gene list, 388 (16.6%) were located on the plasma membrane or in the extracellular space, which is comparable to the 19.4% of genes in the whole genome library that are annotated to these locations. To assess the reproducibility of these 388 candidate target genes, their ability to re-capitulate the growth effects from the primary screen was tested in two clonal lines, the 184-hTERT-L9 and 184-hTERT-E11. False positives due to non-specific toxicity can result from errors in liquid handling at any point within the assay. Of the candidate genes, 140 (36%) were considered reproducible as they were able to reduce the growth of either clonal cell line by 75% or more compared to the control condition (Figure 4.12). In addition to non-specific toxicity, off-target effects stemming from the siRNA targeting unintended mRNA transcripts can produce false-positive results. A complementary region as small as 8 nucleotides between a sense or antisense strand of a siRNA and an unintended transcript can cause gene silencing [267]. Off-target effects from genomewide siRNA screening have been assessed through a variety of means. Employed strategies have included re-screening with siRNA pools designed using a different algorithm, re-screening with the 4 individual siRNAs contained within the original siRNA pool to ensure at least 2 reproduce the desired effect, correlating observed functional effects with gene expression profiles of the original cells, and ensuring silencing of the intended transcript by the siRNA pools through RT-QPCR [268-272].  82  The 140 reproducible candidate genes was still a large number to assess through deconvolution of the original siRNA pools. Therefore, we opted to ensure effective silencing of the intended targets from the original siRNA pools to gauge target specificity prior to deconvolution. To ensure the candidate siRNAs are targeting the mRNA transcript of interest, we assessed relative expression levels by RT-QPCR 48 hours after siRNA transfection (Appendix C). On-target gene silencing was only evident in 35% of the siRNAs tested. 21% showed no silencing and are thus likely off-target effects. 33% were not detected by two independent RT-QPCR primer/probe sets and were also not expressed in the whole transcriptome shotgun sequencing libraries prepared from the 184-hTERT-L9 cells (Chapter 6). This suggests that either we were unable to design a functional RT-QPCR assay for these target genes or that they are not actually expressed in the 184-hTERT-L9 cells and thus the functional effects from the screen were off-target effects. One caveat is that we assessed the growth effects in the screen in co-culture and the RT-QPCR analysis was performed on cells transfected in bulk culture. The transcriptome sequencing libraries were also generated from cells in bulk culture. Therefore, this may reflect differences in the growth conditions rather than reflecting true off-target effects. Additionally, 11% were not detected by two QPCR primer/probe sets, but were detected in the transcriptome sequencing libraries and were considered to be failed RT-QPCR assays. Ultimately, we were able to identify 49 putative signal transducers with diverse functions that were reproducible and silenced the intended target of interest. This broadly categorized into genes involved in neuronal signaling, cell adhesion, cardiac or hematopoietic function, development, matrix interactions, catalytic activities, and ion  83  sensing or transport (Figure 4.13). When the original siRNA pools for these 49 targets were deconvolved and re-screened, 39% had 2 or more individual siRNAs produce statistically significant effects on growth (data not shown). When growth following the stable introduction of independently designed lentiviral shRNA constructs for these 49 targets was assessed, 52% either completely abrogated or decreased 184-hTERT-L9 cell growth (Appendix D and Appendix E). When combined, the growth effects of 40 (82%) of these targets validated by either siRNA deconvolution or through independent assessment by an alternate RNA interference method and sequence design. This signifies that the methods currently used for validating off-target effects are not particularly reproducible across platforms and must be carefully considered when utilized as the sole means for narrowing down target selection. As the 49 putative signal transducers identified through the siRNA screening are required for normal growth regulation, we surveyed their expression within breast cancers to see if they may be involved in malignancy. Our laboratory initiated the METABRIC study, an international collaboration to survey the genetic alterations and expression profiles of over 2000 tumours in an effort to improve prognostic and therapeutic classifications. By stratifying these samples using their PAM50 profiles, we were able to assess the expression pattern of our genes of interest within breast cancer subtypes in relation to normal tissue. PAM50 is a subset of 50 genes whose expression profiles are sufficient to delineate intrinsic breast cancer subtypes into prognostic groupings [273]. While the majority of genes were unchanged, we identified Ripk2 and Ace2 as up-regulated in basal-like and Her2 cancers, respectively (Figure 4.14). Ripk2 is a member of the receptor-interacting protein family of serine/threonine protein kinases  84  and is a potent activator of the NFκB signaling pathway [274]. NFκB signaling is predominately activated in the basal-like subtype and has previously been suggested to regulate epithelial-stromal crosstalk in breast cancer [275, 276]. This work has identified a potential mechanism through which NFκB signaling may be differentially activated in basal-like breast cancers. If confirmed, we have identified a surface protein that can viably be targeted for therapeutic intervention in this sub-type of breast cancer, which is in particular need of targeted therapeutics. Ace2 is a newly identified homologue of the angiotensin-converting enzyme Ace1 [277] and converts angiotensin II to angiotensin I. Alterations in the renin-angiotensin system have previously been implicated in the growth of normal mammary epithelial cells and in breast cancer progression [278, 279]. The differential expression of Ace2 in the Her2 over-expressing subtype may reflect a differential reliance upon the renin-angiotensin system. If it is required for Her2 over-expressing breast cancer, Ace2 is an interesting potential therapeutic target given the success of Ace1 inhibitors in unrelated clinical trials and the current availability of an Ace2 inhibitor shown to be effective in animal trials for gastrointestinal disorders [280, 281].  4.2.6  Signal transducers have variegated effects on 3D morphogenesis Growth signals within a 3D environment differ dramatically from what is  experienced by cells in a 2D environment. To address how the genes required for 2D growth influence growth in 3D, we selected 8 genes and qualitatively assessed the structures formed when these genes are silenced in 3D Matrigel. 184-hTERT-L9 cells were infected with the previously validated lentiviral shRNA constructs targeting these genes prior to seeding in Matrigel. Although not formally quantified, the number of 85  structures in each condition decreased relative to non-silencing control condition. Silencing of Robo3, Sema3C and PlxnA2 did not alter the morphology of the 3D acini formed. Silencing of Ntn1 led to irregular shaped acini, whilst silencing of Efna4, Lgals1 and Procr led to disorganized structures with filled lumens (Figure 4.15). Targeting Edg7 resulted in the complete abrogation of 3D growth. When Edg7 is temporally silenced with siRNA prior to plating in Matrigel, the formation of structures is inhibited until Day 8. Upon the gradual re-expression of Edg7, structures begin to emerge within Matrigel and are fully formed by the assay end-point on Day 21 (Figure 4.16A). This signifies that loss of Edg7 is not lethal to the cells and that they remain viable but quiescent within the Matrigel until Edg7 re-expression prompts their expansion. The Edg7 siRNA pool was de-convolved and used to assess proliferation and viability of the 184-hTERT cells when Edg7 is decreased. 3 of the 4 individual siRNAs robustly silence Edg7, which corresponds to their ability to suppress 184-hTERT growth in co-culture with irradiated NIH 3T3 cells (Figure 4.16B and 4.16C, respectively). The 3 siRNAs that suppress growth decrease the percentage of BrdU positive cells without affecting cell viability (Figure 4.16D). This shows that silencing of Edg7 blocks cell proliferation. Edg7 has been previously shown to signal through Erk1/2. Levels of phospho-Erk1/2 decrease upon Edg7 protein silencing, suggesting a pathway through which it may ultimately control cell proliferation (Figure 4.16E). Note that the silencing of Edg7 by the individual siRNAs at the mRNA transcript level does not translate into efficient silencing at the protein level; this was only accomplished with the siRNA pool (Figure 4.16C and 4.16E, respectively). Recently, overexpession of Edg7 using the MMTV promoter in a mouse model was described. These mice formed late-onset  86  mammary carcinomas [282]. Edg7 thus plays an important role in regulating the growth of both normal and malignant mammary cells.  4.2.7  Bi-potent and luminal progenitor cells are differentially affected by signal  transducers required for epithelial cell growth Our overarching goal is to translate the results from the siRNA screen to the system of relevance, mammary progenitor cells. We have not been able to silence gene expression within primary mammary cells through lipid-mediated siRNA transfections (data not shown). Therefore, we had to optimize a new set of reagents before exploring the relevance of these genes in progenitor cell growth. Using lentiviruses, we have successfully manipulated gene expression within mammary progenitor cells in the past [104]. Therefore, we decided to use short hairpin RNA (shRNA) within this system to mediate RNA interference. We collected 1 to 3 lentiviral shRNA constructs per gene of interest and tested their efficacy in the 184-hTERT-L9 cells. Initially, we infected the 184-hTERT-L9 cells with each individual shRNA and selected for stable integration of the construct using the puromycin resistance marker present on the viral backbone. Not surprisingly, we were unable to propagate 39 of the 95 shRNA constructs tested, presumably due to their established growth effects. When re-infected into cells and harvested after 48 hours, gene silencing of the shRNA target was achieved in 23 of the 24 constructs tested by RT-QPCR (15 were not tested in this secondary approach)(Appendix E). Of the lines that were established after puromycin selection, only 15 of the 56 had effective gene silencing (Appendix D). The lentiviral shRNA constructs that abrogated 184-hTERT growth and silenced expression in the stable cell lines were carried forward for screening in mammary progenitor cells. For this, reduction mammoplasty tissue was  87  enzymatically dissociated into a single cell suspension and infected overnight with lentiviral particles prior to plating in the colony-forming assay. A multiplicity of infection (MOI) of 3 and 10 was tested using the shRNA constructs in the primary mammary tissue. Effective gene silencing in the primary cells as assessed by RT-QPCR was only noted at an MOI of 10 and thus this MOI was consequently used for all of the infections (data not shown). At this MOI, the transfection efficiency for the pGIPZ shRNA constructs is approximately 28% as assessed by FACS in primary cells infected and plated in bulk culture for 5 days (data not shown). After 2 weeks, the resultant colony types were morphologically discriminated into luminal or mixed/myoepithelial colonies and compared to non-silencing shRNA controls (Figure 4.17). The numbers of progenitors were inferred from the colony types that arose, with luminal colonies originating from luminal progenitor cells and mixed/myoepithelial colonies deriving from bi-potent and myoepithelial progenitors (the latter representing a very small proportion of the total progenitor cell population). Three patient samples were evaluated in technical duplicates. Statistical significance was not reached within the majority of the shRNAs tested due to extreme patient-to-patient variability. The largest growth effects seen within both progenitor cells types were when Gpr39, Scarb2, Ntn1, Efna4, Nptx1 or Ctnna1 were silenced. With SerpinH1 silencing, a greater effect was seen in the luminal progenitor cells, with negligible effects in the bi-potent progenitors. When either Nkain4 or Kcnj5 were silenced, an increase in luminal progenitors was observed whilst bi-potent progenitor cell growth was diminished. The ability to differentially affect the growth response of one progenitor cell type signals that these genes may be involved in differentiation within the mammary gland. However, further validation is required for  88  the majority of these targets to confirm that they are robustly affecting progenitor cell growth.  4.3  Discussion In this chapter, we took a systematic and unbiased approach to uncover the  mechanisms regulating fibroblast-dependent mammary epithelial cell growth. In doing so, we revealed a surprisingly diverse category of genes with novel roles in growth regulation. Furthermore, we identified genes that were more potent in functionally affecting growth than currently known regulators of epithelial cell growth, namely insulin and EGF. One limitation of our assay was the need to assess signal transducers in the epithelial cells, rather than molecules originating from the fibroblasts. This was due to a technical inability to silence genes within irradiated feeder cells. As a result, we cannot definitively pinpoint the ligands for all of these signal transducers, as many are currently unknown. Also, the targets may represent previously uncharacterized transducers for the media components included within the defined assay media and thus may not be related to the fibroblasts. Nevertheless, we were able to identify 49 signal transducers that have not been previously implicated in mammary epithelial cell growth alluding to the complexity that exists within this seemingly simple process. The screening approach taken here has limitations. The initial screen was performed in duplicate, with false positives eliminated through a secondary screen for reproducibility. We did not address the false negative results that undoubtedly exist and can result from liquid handling errors or transfection with ineffective siRNAs. Additionally, false negatives can arise from incomplete silencing of a transcript leaving  89  residual expression sufficient for gene function and proteins with a long half-life may remain intact despite effective silencing of their transcript. Another related concern surrounding the incomplete detection of all factors required for growth is the cut-off point of biological effects chosen for validation screening. We chose to proceed with statistically significant effects with a 75% or more inhibition of growth. This was guided by the size of a feasible rescreening effort and did not include the known growth receptors for insulin and EGFR. We were aware that some targets required for epithelial cell growth would be excluded with this cut-off, but were conversely encouraged that all of the signal transducers identified would have a greater biological effect than these known growth modulators. Guidelines for target selection in screens of this magnitude have not yet been established; common methods currently include selecting an arbitrary cut-off point or using a given number of standard deviations away from the mean as selection tools [268-272]. Once we had implemented our threshold for target selection, we segregated the results based upon cellular location. As we were interested in identifying possible signal transducers, we selected targets present on the plasma membrane or in the extracellular space. Any misclassification or incomplete data within the Gene Ontology annotations used for segregation would lead to improper inclusion or exclusion of targets [283]. Of the 388 putative signal transducers carried forward for validation screening, 140 (36%) had reproducible effects. The reproducibility of a RNA interference molecule does not signify that the effect it affords is related to the silencing of the intended mRNA target. Several nonspecific and sequence related effects have been characterized that could generate false positive results within the validated data set. When an individual siRNA is used within a  90  cell, a host of expression changes can occur that are related to the siRNA sequence and not necessarily the intended target gene [284]. The critical region for producing sequence-specific off target effects is complementarity of the first 2 to 8 nucleotides of the antisense siRNA strand to the 3’ untranslated region of an unintended target [285]. This was considered during the design of Dharmacon’s siGENOME library, but is a complication that cannot be easily eliminated through sequence selection alone. Modifying the second nucleotide of the antisense strand through 2′-O-methyl ribosyl modification helps to minimize this effect but cannot eliminate it [286]. The sense strand has also been modified to impede its entry into the RISC complex. Sequence-dependent effects that are unrelated to non-specific gene silencing are another consideration when using siRNAs. Small sequence motifs have been identified that produce toxic effects to cells entering the RISC complex and are related to the RNA interference mechanism within cells rather than gene silencing [287]. Other sequence motifs exist that can induce the intracellular production of cytokines in response to the presence of that particular siRNA [288, 289]. The delivery of siRNA into the cells is also a process that can produce unintended effects. While the same delivery system is used across cells and effects should be considered consistent, Lipofectamine 2000 can alter insulin signaling by activating and subsequently down-regulating the expression of its receptor [290]. This will undoubtedly complicate the result of any siRNAs acting within this signaling pathway and could further result in unintended growth effects. To control for off-target effects, we assessed the ability of each siRNA pool to silence their intended mRNA transcript. In doing so, we eliminated candidate siRNA pools that may be affecting growth through an unknown mechanism. On-target gene  91  silencing was only evident in 35% of the siRNAs tested. Of these, 82% validated either through siRNA deconvolution or through independent assessment by an alternate RNA interference method and sequence design. This suggests that despite careful siRNA design parameters implemented by Dharmacon, off-target effects remain a significant problem in siRNA screening. Utilizing the knowledge that the first few nucleotides of an siRNA can unintentionally target 3’ untranslated regions, a re-evaluation of off-target effects from 2 siRNA screens aimed at identifying genes regulating TRAIL-induced apoptosis lead to the discovery of endogenous miRNAs that act within this pathway [291]. Nevertheless, this screen identified a broad spectrum of signal transducers implicated in regulating fibroblast driven epithelial cell growth and provided a novel functional description for the majority of these genes. While our original interests lay in the fibroblast-dependent growth regulation of mammary progenitor cells, it was not feasible to perform a genome-wide screen using these cells. Some of the limitations included the inability to silence gene expression within these cells using siRNA, limited material per patient sample, extreme patient-topatient variability, and limited transferability of the colony-forming assay to detect progenitor cell growth to a multi-well format. Utilizing the 184-hTERT cell line as a surrogate to study this phenomena comes with its own limitations. Namely, the forced expression of telomerase and its implications may confound some of the growth effects seen within this assay. It is likely that the signaling molecules driving progenitor cell growth could differentially affect immortalized cells and may not be detected within this screen. Furthermore, isolating a pure population of progenitor cells is not possible and thus cells that have a similar phenotype to the progenitor cells but are unable to form a  92  colony will ultimately be present during culturing. Any potential contribution of these cells to progenitor growth is not accounted for when using the 184-hTERT cell line. In unfractionated primary tissue, matters are further complicated by the presence of 3 distinct progenitor cell populations that are all capable of proliferation and can all potentially influence each other. Despite this, we screened the 49 signal transducers in primary mammary tissue and gauged their ability to modulate progenitor cell growth in the colony-forming assay. For this, we established a validated set of lentiviral shRNA constructs targeting these genes and individually assessed their effect on progenitor cell growth in three independent patient samples. While trends in growth inhibition and differential progenitor cell responses were evident, extreme variability in the three patient samples tested for each shRNA posed a significant problem in the analysis of this data. Variable responses to the non-silencing control shRNA amongst patient samples, potentially related to off-target effects, further complicated the analysis. Finally, different proportions of progenitor cell types present within each patient sample made the analysis of differential progenitor cell responses difficult. Using the initial progenitor cell responses as a guide, re-analyzing a reduced subset of these signal transducers on more patient samples will likely provide statistically significance within a larger number of tested genes. Signal transducers with the greatest effects on overall progenitor cell growth include Gpr39, Scarb2, Ntn1, Efna4, Nptx1 and Ctnna1. SerpinH1 differentially suppressed luminal progenitor cell growth whereas the sodium and potassium channel modulators Nkain4 and Kcnj5 differentially suppressed bi-potent progenitor cell growth. Ntn1 has previously been implicated in regulating ductal branching morphogenesis  93  within the mouse mammary gland in conjunction with Slit2 [292]. While Efna4 has no described role within the mammary gland, its homologue Efna2 is essential for both epithelial cell proliferation and ductal branching in the mouse [293]. Gpr39 is a zinc receptor that has been implicated in insulin signaling and wound repair in keratinocytes [294, 295]. Both processes are intrinsically relevant to mammary epithelial cells. Ctnna1 is an adhesion molecule associated with e-cadherin whose expression is lost in 81% of breast cancers [296]. While this may seem contradictory, loss of e-cadherin mediated cell contacts can lead to the up-regulation of transcription factors that drive cell proliferation [297]. To assess the relevance of this gene set to breast cancer, we assessed their expression across breast cancer subtypes. We identified Ripk2 and Ace2 as up-regulated in basal-like and Her2 cancers, respectively. Ripk2 is a potent activator of the NFκB signaling pathway and regulates myoblast differentiation in both skeletal muscle development and in rhabdomyosarcoma [274, 298]. Ace2 is an enzyme that cleaves a variety of vasoactive peptides and thus modulates vasodilation [277]. Activity of Ace2 is reduced in vivo by administration of estrogen [299]. While neither gene has a described role in the mammary gland or in breast cancer, their biological functions provide potential mechanisms through which they may be exerting their effects in malignancy. This unbiased approach to identifying regulators of mammary epithelial cell growth uncovered a variety of genes with both novel and known roles in regulating mammary epithelium. It has revealed alternative functions for a handful of genes that would otherwise not be expected to regulate epithelial cells and has confirmed the enormous  94  complexity involved in a process that is as seemingly well understood as growth stimulation.  95  0 i3T3s  A  1000 i3T3s  2500 i3T3s  5000 i3T3s  3500  GFP positive cell count  3000 2500 2000 1500 1000 500 0  DMEM + 5% FBS 1 2 Epicult B  B  0 i3T3s  Total nuclear count  2500  MEGM  3 Serum-free 74  5  2000 i3T3s  2000 1500 1000 500 0 1  2  3  4  5  6  7  8  9  96  Figure 4.1 – Differential response of 184-hTERT cells to varying growth medias A) Growth of 184-hTERT-gfp cells in 4 media formulations when plated at low density in 96-well plates with increasing numbers of irradiated NIH 3T3 cells per well. GFP positive epithelial cells were enumerated after 5 days of growth using the IN Cell Analyzer and IN Cell Developer Software. DMEM + 5% FBS does not support the growth of 184-hTERT-gfp cells. Epicult B and Serum-free 7 allow for dose dependent fibroblast driven growth. MEGM supports low density growth of 184-hTERT cells irrespective of the presence of fibroblasts. Bars denote 95% confidence intervals (n=3). B) Effect of individual Clonetics media components on 184-hTERT-L9 growth when plated at low density with or without irradiated NIH 3T3 cells. Total nuclear count was determined from paraformaldehyde-fixed cells stained with Dapi after 5 days growth using the IN Cell Analyzer and IN Cell Developer Software. Basal Media is MEBM, Base Media refers to MEBM with the addition of hydrocortisone, transferring, and isoproterenol. Individual components were subsequently added at concentration used in full MEGM growth media. “Visvader media” is included as a comparator as it is identical to Serum-free 7 media with the exception of containing 5 times more insulin, a level comparable to the amount of insulin found in MEGM. Addition of BPE negates the requirement for fibroblasts for low-density growth. Base media with EGF and Insulin is henceforth referred to as Assay Media. Bars denote 95% confidence intervals (n=3).  97  Figure 4.2 – 184-hTERT cells require close contact with feeder cells for growth in colony forming assays Irradiated feeder cells were allowed to settle in a 60 mm dish for 24 hours at a density of 30,000 cells/cm2. A cell scraper was used to clear a central strip from the dish prior to plating 150 184-hTERT-L9 cells. After 8 days, cells were fixed with cold acetonemethanol (1:1) and stained with Wright-Giemsa solution. In the picture of the 60 mm dish (taken on an Alpha Imager), a lighter band of cleared plastic is evident, upon which no colonies formed. Colony formation was not inhibited in the areas where fibroblasts were present, and along the border of the cleared plastic. Bright field images from the regions outlined in the black boxes were taken with a 10x objective on a Leica inverted microscope.  98  Figure 4.3 – Differential signaling response of 184-hTERT cells to fibroblasts and bovine pituitary extract 184-hTERT-L9 cells were grown in the presence of either irradiated NIH 3T3 cells or bovine pituitary extract. Resultant gene expression profiles were assessed with Affymetrix GeneChip Human Exon 1.0 ST Arrays and validated by RT-QPCR. The top associated network from an Ingenuity Pathway Analysis of the validated differential expression consists of genes implicated in cancer, cellular movement, and cell-to-cell signaling and interaction. Green indicates genes that are up-regulated in the 184-hTERT cells grown with fibroblasts, red represents genes that are up-regulated in the cells grown with bovine pituitary extract (or down-regulated when cells are grown with fibroblasts), and white indicates no change in expression. Direct interactions are exhibited with solid arrows, and indirect interactions (executed through an intermediary molecule) with 99  dashed arrows. Interactions indicate physical association, induction/activation, or repression/inactivation of one gene product by the other gene product.  100  A  B  0 nM siTOX  20 nM siTOX  50 nM siTOX  Percent viable cells  100 80 60 40 20 0 0.3  0.6  1  1.6  Volume Dharmafect per well (+l) 20 nM siGLO-N  D  50 nM siGLO-N  100  20 nM siGLO-N  50 nM siGLO-N  100  Percent viable cells  Percent cells with granules  C 80 60 40 20 0  0.3  0.6  1  Volume Dharmafect per well (+l)  1.6  80 60 40 20 0 0  0.3  0.6  1  1.6  Volume Dharmafect per well (+l)  E  101  Figure 4.4 – RNA interference is ineffective in irradiated NIH 3T3 cells A) Irradiated NIH 3T3 cells were plated at a density of 10,000 cells/cm2 in 35 mm dishes and transfected with 50 nM of pooled siRNAs targeting beta-actin complexed with varying transfection reagents. After 96 hours, 10ug of protein was loaded onto a NuPAGE 4-12% gradient Bris-Tris gels and transferred to nitroceullulose prior to immunoblotting. Beta-actin and Lamin C levels were detected usig the LI-COR Odyssey imaging system. Although beta-actin is not effectively silenced in any condition, levels decrease slightly with the use of DharmaFECT 3. B) Irradiated NIH 3T3 cells were plated at a density of 30,000 cells/cm2 in 96-well plates and transfected with increasing concentrations of siTOX and DharmaFECT3. After 48 hours, live-dead discrimination was performed using Calcein-AM/ethidium homodimer staining and enumerated using the IN Cell Analyzer and IN Cell Developer Software. Toxicity is not observed after the application of siTOX suggesting that RNA silencing is not occurring. Error bars represent standard deviation (n=2). C) Irradiated NIH 3T3 cells were plated at a density of 30,000 cells/cm2 in 96-well plates and transfected with siGLO-nuclear and increasing concentrations of DharmaFECT3. After 48 hours, cells were stained with Calcein-AM/ Hoechst 3342 and the number of siGLO-nuclear granules per well was enumerated using the IN Cell Analyzer and IN Cell Developer Software. Fluorescently labeled siRNAs are able to enter cells at a high efficiency, with a range of 2-5 granules per cell. Error bars represent standard deviation (n=2). D) Cell viability, defined by Calcein-AM positivity, is not affected by transfection with siGLO-nuclear as described above. Error bars represent standard deviation (n=2). E) Representative image showing the distribution of siGLO-nuclear in irradiated NIH 3T3 cells transfected with 50 nM siGLO-nuclear and 0.3 µl of DharmaFECT3 per well. Images were captured with the 10x objective on the IN Cell analyzer.  102  100 nM siRNA pool  50 nM siRNA pool  100 nM siRNA #4  50 nM siRNA #4  100 nM siRNA #3  50 nM siRNA #3  100 nM siRNA #2  50 nM siRNA #2  100 nM siRNA #1  50 nM siRNA #1  L2K only  Untransfected  A  E-cadherin  Cell count  Untransfected  Lipofectamine 2000 + 30 nM gfp siRNA  total count  gfp+ count  total count  gfp+ count  2000 1800 1600 1400 1200 1000 800 600 400 200 0  total count  B  gfp+ count  Actin  DharmaFECT 3 + 30 nM gfp siRNA  Figure 4.5 – RNA interference is effective in 184-hTERT cells A) 184-hTERT-L9 cells were plated at a density of 4200 cells per cm2 in 35 mm dishes and transfected with Lipofectamine 2000 and either 50 nM or 100 nM of siRNAs targeting E-cadherin. After 72 hours, 10ug of protein was loaded onto a NuPAGE 4-12% gradient Bris-Tris gels and transferred to nitroceullulose prior to immunoblotting. Expression of E-cadherin is silenced with 2 of the 4 siRNAs and the siRNA pool. B) 184-hTERT-gfp cells were plated in Assay Media at a density of 1500 cells per cm2 in co-culture with 10,000 irradiated NIH 3T3s per cm2 plated 24 hours prior in 96-well plates. After an additional 24 hours, cells were transfected with either Lipofectamine 2000 or DharmaFECT3 and 30 nM of siRNA targeting GFP. 5 days later, GFP positive cells were enumerated using the IN Cell Analyzer and IN Cell Developer Software. Lipofectamine 2000 is less toxic to the cells than DharmaFECT3. Both conditions effectively silence GFP expression (n=6).  103  A  0 i3T3s  1000 i3T3s  2500 i3T3s  5000 i3T3s  3500  Total nuclear count  3000 2500 2000 1500 1000 500 0 Day 1  Plate i3T3s in DMEM + FBS  Plate i3T3s in DMEM + FBS  Plate 184Plate 184Day 2 hTERTs in assay hTERTs in assay media media + FBS  4500 4000  0 i3T3s  Plate i3T3s in assay media  Plate i3T3s in assay media + FBS Change to assay media  Change to assay media  Day 3  B  Plate 184Plate 184hTERTs in assay hTERTs in assay media + FBS media + FBS  2000 i3T3s  3000 i3T3s  4000 i3T3s  5000 i3T3s  6000 i3T3s  Lipofectamine 2000  Total nuclear count  3500 3000 2500 2000 1500 1000 500 0 0  C  4000 3500  400  0 i3T3s 2000 i3T3s Lipofectamine 2000 + 30 nM siCONTROL3  450 500 Number of L9 hTERTs plated 3000 i3T3s  4000 i3T3s  550  5000 i3T3s  600  6000 i3T3s  Total nuclear count  3000 2500 2000 1500 1000 500 0 0  400  450 500 Number of 184-hTERT-L9 s plated  550  600  104  Figure 4.6 – Optimized plating density and order can reduce assay variability A) The plating order of the 184-hTERT-L9 and irradiated NIH 3T3 cells affects the variability of the co-culture assay. Either 184-hTERT-L9 cells or an increasing density of irradiated NIH 3T3s cells were plated on Day 1 with the alternate cell type plated on Day 2 with or without FBS. Error bars represent standard deviation (n=2). B) Plating density of 184-hTERT-L9 cells and irradiated NIH 3T3 cells for high-content screening is affected by non-specific toxicity associated with Lipofectamine 2000. 184hTERT-L9 and irradiated NIH 3T3s were plated at increasing densities in 96-well plates prior to treatment with 0.3 µl of uncomplexed Lipofectamine 2000. Total nuclear count was determined from paraformaldehyde-fixed cells stained with Dapi after 5 days growth using the IN Cell Analyzer and IN Cell Developer Software. Error bars represent standard deviation (n=2). B) Plating density of 184-hTERT-L9 cells and irradiated NIH 3T3 cells for high-content screening is affected by toxicity associated with Lipofectamine 2000 complexed siRNAs. 184-hTERT-L9 and irradiated NIH 3T3s were plated at increasing densities in 96-well plates prior to treatment with 30 nM siCONTROL 3 complexed with 0.3 µl Lipofectamine 2000. Total nuclear count was determined from paraformaldehyde-fixed cells stained with Dapi after 5 days growth using the IN Cell Analyzer and IN Cell Developer Software. Error bars represent standard deviation (n=2).  105  L2K only  A1  1653  1031  H12 siCONTROL 2  A1  409 186  98  H12 siCONTROL 3  A1  9 922  493  H12 PLK1  A1  63 80  41  H12  1  106  Figure 4.7 – Plate effects produce variability across 96-well plates Entire 96-well plates were seeded with 3000 irradiated NIH 3T3 cells 24 hours prior to the addition of 550 184-hTERT-gfp cells. Plates were transfected with Lipofectamine 2000 alone or complexed with siCONTROL 2, siCONTROL 3 or siRNAs targeting Plk1. After 5 days, GFP positive cells were enumerated using the IN Cell Analyzer and IN Cell Developer Software. Heatmaps were generated from GFP positive counts in a 96-well plate format using JColorJar, with legends indicating highest (red), median (white) and lowest (blue) cell counts (generated from 12 fields of view from a 10x objective). The variability encountered signified the need to normalize for plate effects throughout the screen, which was achieved through the intermittent inclusion of entire control plates. Note the use of different scales, which disallows cross-plate comparisons. This was done to ensure maximum visual range on individual plates in order to highlight plate effects.  107  gfp + cell count  2000 1800 1600 1400 1200 1000 800 600 400 200 0  gfp + cell count  2000 1800 1600 1400 1200 1000 800 600 400 200 0  gfp + cell count  A  2000 1800 1600 1400 1200 1000 800 600 400 200 0  4x objective without binning  1  2  3  4  5  6  7  8  9  10  11  12  4  5  6  7  8  9  10  11  12  4  5  6  7  8  9  10  11  12  10x objective without binning  1  2  3  10x objective with binning  1  2  B  3  12000  21 acquisition fields per well  12 acquisition fields per well  gfp + cell count  10000 8000 6000 4000 2000 0 L2K  siCON3  PLK1  siRNAs  L2K  siCON3  PLK1  siRNAs  C 184-hTERT-gfp and i3T3 co-culture  Nuclear image mask  Thresholding to identify GFP+ cells  108  Figure 4.8 – Acquisition parameters affect the dynamic range of the assay A) GFP positive cell counts resulting from images acquired with a 4x or 10x objective, with or without binning, were compared across the columns of a 96-well plate (horizontal axis). Columns 1-2 do not contain fibroblasts, column 2 to 12 contain 3000 irradiated NIH 3T3 cells plated with 184-hTERT-gfp cells and imaged after 5 days (n=8). Use of a 10x objective with binning generates the greatest dynamic range between wells with and without fibroblasts present. 12 fields of view were collected when using the 10x objective, 4 fields of view were collected when using the 4x objective. Fields were collected in a fixed position for each well, evenly distributed in a non-overlapping manner across the entire well. B) The acquisition of 12 versus 21 fields using the 10x objective with binning were compared for effects on dynamic range using a subset of siRNAs targeting the insulin signaling pathway (see figure 4.9). Control conditions of Lipofectamine 2000 alone (n=6), siCONTROL 3 (n=5), and Plk1 (n=5) are plotted along with the compiled siRNAs (n=80). Acquisition of 21 fields produced a larger range between conditions. C) Representative image segmentation from IN Cell Developer of a co-culture sample to discriminate GFP positive cells (184-hTERT-gfp) from GFP negative cells (fibroblasts). Fields were collected in a fixed position for each well, evenly distributed in a nonoverlapping manner across the entire well.  109  110  Figure 4.9 – Total nuclear count accurately reflects the 184-hTERT count in co-culture Comparison of total nuclear count and GFP positive count in co-cultures treated with a siRNA library targeting cell cycle genes showed that the total count can be used as a surrogate for the GFP positive count. Three thousand irradiated NIH 3T3 cells were plated with 400 184-hTERT-gfp cells in 96-well plates prior to transfection with 0.3 µl of Lipofectamine 2000 and 30nM of siRNA per well. 21 fields of view were acquired after 5 days of growth using a 10x objective on the IN Cell Analyzer. The GFP negative count was obtained by subtracting the GFP positive count from the total nuclear count and represents the number of irradiated fibroblasts present in each well. The trend in effect is similar between the two counts and thus total nuclear count was used for screening due to the timesaving provided during imaging.  111  Primary Screen  Transfect with Dharmacon siRNA pools targeting 21,121 genes in duplicate Culture for 4 additional days  184-hTERTs plated onto a layer of i3T3s  Targets affecting cell growth identified through automated microscopy  2337 targeted genes demonstrate a statistically significant abrogation of growth to less than 25% of the control condition  Filter for the 388 targets located on the plasma membrane or in extracellular space  Secondary Re-screen  Transfect with siRNA pools against genes located on the plasma membrane and in the extracellular space in 2 clonal cell lines in duplicate Culture for 4 additional days 184-hTERTs plated onto a layer of i3T3s  Targets affecting cell growth identified through automated microscopy  49 targets decrease the mRNA transcript of interest as assessed by RT-QPCR 48 hours post transfection with siRNA pools  140 targets re-capitulate the growth effects seen in the primary screen  Specificity of 1-3 lentiviral shRNAs per target assessed by RT-QPCR after stable integration into 184-hTERT cells Mixed/myo colony  Luminal colony  Infect primary tissue with validated shRNA constructs; 3 patient samples per shRNA Plate in the Colony Forming Assay for 8-12 additional days in duplicate Dissociated primary reduction mammoplasty tissue in suspension culture  Targets affecting progenitor cell growth identified through morphological discrimination of resultant colony types  112  Figure 4.10 – High-throughput screening workflow Assay design and number of targets from primary screening, secondary screening and validation in primary tissue. Placement of controls for the primary and secondary screening are indicated in the plate maps with yellow denoting Lipofectamine 2000 alone, pink indicating siCONTROL 3 and green indicating Plk1. 184-hTERTs refers to the 184-hTERT-L9 line for the primary screen and the 184-hTERT-L9 or 184-hTERTE11 line for the secondary screen.  113  A  Ratio of plate normalized cell growth over control  <25%  INSR EGFR !2-adrenergic receptor Transferrin receptor  Ranked Effect of siRNA pools  B  114  Figure 4.11 – Primary screening identified 388 potential targets transducing the signals required for fibroblast-dependent epithelial cell growth A) Distribution of the ranked growth effects for siRNA pools utilized in the siRNA screen normalized to the Lipofectamine 2000 alone control condition. Receptors to components of the assay media are indicated in red. siRNA pools with a statistically significant effect that also reduced growth to 25% or less of the control condition are highlighted in blue (n=2). B) Cellular distribution of genes targeted by siRNA pools with a statistically significant effect that also reduced growth to 25% or less of the control condition. Targets located on the plasma membrane or in the extracellular space were selected for secondary screening to assess reproducibility.  115  I5&35&SBUJPPGQMBUFOPSNBMJ[FEDFMMHSPXUI I5&35-SBUJPPGQMBUFOPSNBMJ[FEDFMMHSPXUI  Figure 4.12 – Secondary screening identified 140 reproducible candidate signal transducers Covariate plotting of the ranked effects of the 388 candidate siRNAs carried forward from the primary screen normalized to the Lipofectamine 2000 alone control condition. Secondary screening was conducted on 2 of the clonal lines, 184-hTERT-L9 (from the primary screen) and 184-hTERT-E11. siRNA pools with a statistically significant effect that also reduced growth to 25% or less of the control condition in one or both cell lines are highlighted in pink (n=2).  116  NTN1 NTN2L FLJ30634 FLOT2 NPTX1 ROBO3 SEMA3C SLC6A4  Axon guidance/Neuronal Signalling netrin 1 netrin 3 plexin A2 flotillin 2 neuronal pentraxin I roundabout, axon guidance receptor, homolog 3 semaphorin-3C solute carrier family 6  Matrix Interactions COL9A3 collagen, type IX, alpha 3 LGALS1 lectin, galactoside-binding, soluble, 1 MMP24 matrix metallopeptidase 24 MMP28 matrix metallopeptidase 28 LTBP3 latent transforming growth factor beta binding protein 3 SERPINH1 serpin peptidase inhibitor, clade H TUFT1 tuftelin 1  CTNNA1 PCDHB13  Cell Adhesion catenin (cadherin-associated protein), alpha 1 protocadherin beta 13  PARD3 EFNA4 FZD2  Development par-3 partitioning defective 3 homolog (C. elegans) ephrin-A4 frizzled homolog 2  ACE2 ADMR BDKRB2 RHCE TMEM14C SAA1 PDCD1 PROCR  Cardiac or Hematopoietic function angiotensin I converting enzyme 2 G protein-coupled receptor 182 bradykinin receptor B2 Rh blood group, CcEe antigens transmembrane protein 14C serum amyloid A1 programmed cell death 1 protein C receptor, endothelial  C20ORF58 KCNJ5 SLC7A7 GPR39  Ion sensing/transport Na+/K+ transporting ATPase interacting 4 potassium inwardly-rectifying channel, subfamily J solute carrier family 7 G protein-coupled receptor 39  ADCY4 RIPK2 FIT1 PLA2G2F HSD17B2  Catalytic or kinase activity adenylate cyclase 4 receptor-interacting serine-threonine kinase 2 fat storage-inducing transmembrane protein 1 phospholipase A2, group IIF hydroxysteroid (17-beta) dehydrogenase 2  OPRS1 C11ORF15 CD79A GPR80 SCARB2 SNN MDS010 BST1 EDG7 PLUNC TUBA1C  Other/Ill-defined sigma non-opioid intracellular receptor 1 TMEM9 domain family, member B Ig-alpha protein of the B-cell antigen component oxoglutarate (alpha-ketoglutarate) receptor 1 scavenger receptor class B, member 2 stannin KTEL (Lys-Tyr-Glu-Leu) containing 1 bone marrow stromal cell antigen 1 lysophosphatidic acid receptor 3 palate, lung and nasal epithelium associated tubulin, alpha 1c  117  Figure 4.13 – Signal transducers that are functionally required for mammary epithelial cell growth Description and classification of genes identified through RNA interferences screening as functionally required for 184-hTERT cell growth. Classifications were based upon Gene Ontology annotations and literature searches for previously described functions.  118  RIPK2 3  Z score expression  2 Ɣ  Ɣ Ɣ  0  Ɣ  Ɣ í  í  Basal  Her2  LumA  LumB  Normal  ACE2  Z score expression  8  6  4  2  0  Ɣ  Ɣ Ɣ  Basal  Her2  LumA  Ɣ  LumB  Ɣ  Normal  119  Figure 4.14 – RIPK2 and ACE2 are differentially expressed in breast cancer subtypes The expression of the signal transducers identified through screening was assessed in over 2000 tumours. Tumours samples were stratified using their PAM50 profiles into basal-like, Her2, luminal A, luminal B and normal breast tissue. Z-score expression is plotted, which represents the number of standard deviations of the level of expression of each gene in each sample away from the mean level of expression of that gene across all the samples within the data set. Expression of RIPK2 is increased in the basal-like subtype and ACE2 displays higher expression in the Her2 subtype.  120  Phallodin  Nuclei  GFP  Merge  Non-si  EDG7  EFNA4  LGALS1  NTN1  PROCR  30 +m 121  Figure 4.15 – Signal transducers required for fibroblast-dependent 2D growth have varying effects in 3D culture 184-hTERT-L9 cells with stably integrated shRNAs targeting the genes required for fibroblast-dependent 2D growth were cultured for 21 days in Matrigel. Confocal microscopy reveals effects such as complete growth abrogation, disrupted organization and luminal filling.  122  Day 8  Day 21 CD49f Nuclei GM130  Relative Expression (Log 2)  Normalized % Growth  80% 60% 40% 20% 0%  0 -1 -2 -3  ol Po  A4 N  A3  Edg7 siRNA  si R  2  R N  si  si  R  N  A-  AN R  si  LF  2K  1  -4  BrdU +ve Cells  E  L2K  siCON  #1  #2  #3  #4  pool  EDG7  % Viable Cells 100%  10%  80%  8%  60%  6% 40%  4%  20%  2% 0%  % Viable Cells  12% % BrdU +ve Cells  D  1  Pool  C  100%  siRNA-4  B  Day 21  Edg7 siRNA  siRNA-3  Day 8  Edg7 siRNA  siRNA-2  LF2K Only  siRNA-1  LF2K Only  LF2K  A  GAPDH Edg7 siRNA LFK  siCON  #1  #2  #3  #4  pool  PhosphoERK1/2  0% LF2K  siRNA-1 siRNA-2 siRNA-3 siRNA-4  Pool  GAPDH  123  Figure 4.16 – Abrogating Edg7 blocks proliferation and activation of ERK1/2 signaling A) 184-hTERT-L9 cells were transfected with pooled Edg7 siRNAs 48 hours prior to plating in 3D Matrigel culture. After 8 days in culture, silencing of Edg7 has suppressed growth in Matrigel. As RNA silencing by siRNA is temporary, structures begin to arise between Day 8 and Day 21, albeit aberrantly, signifying that the cells were able to remain viable during their dormancy. B) Growth after application of the deconvolved Edg7 siRNA pool under co-culture assay conditions normalized to the Lipofectamine 2000 alone control (n=1). C) Relative expression (Log2) in 184-hTERT-L9 cells after application of the deconvolved Edg7 siRNA pool (n=3). D) Percentage of BrdU positive 184-hTERT-L9 cells after 4 days of co-culturing under assay conditions. Error bars denote standard error (n=5). E) 184-hTERT-L9 cells were plated at a density of 4200 cells per cm2 in 35 mm dishes and transfected with Lipofectamine 2000 and either 30 nM of siRNAs targeting Edg7. After 72 hours, 10ug of protein was loaded onto a NuPAGE 4-12% gradient Bris-Tris gels and transferred to nitroceullulose prior to immunoblotting. Levels of phosphorylated ERK1/2 decrease in relation to Edg7 levels.  124  GPR39     Luminal Mixed/Myo  í                   Luminal Mixed/Myo  í         NKAIN4           Luminal Mixed/Myo  í                       Luminal Mixed/Myo        .&1-              í                       Luminal Mixed/Myo  SERPINH1    í   CTNNA1           Luminal Mixed/Myo  í  Luminal Mixed/Myo  NPTX1  EFNA4  í                         Luminal Mixed/Myo  í   Mean colony count with target shRNA  171í%    6&$5%  Luminal Mixed/Myo  í         Mean colony count with non-silencing shRNA  125  Figure 4.17 – Growth signal transducers coordinately or differentially regulate luminal and bi-potent progenitor cells Reduction mammoplasty samples from 3 patients were dissociated into single cell suspensions and individually infected with lentiviral shRNA constructs targeting the 49 growth signal transducers identified from the screen. After 24 hours, cells were plated in the colony-forming assay along with irradiated NIH 3T3 cells. The number of luminal and mixed/myoepithelial colonies was scored after 10 days using morphological discrimination under a dissecting microscope. Plots show the mean colony count for the shRNAs targeting the indicated genes of interest compare to a non-silencing shRNA construct for each patient assessed. The minimum and maximum counts for each technical replicate is shown as error bars; vertical for the targeting shRNA range and horizontal for the non-silencing shRNA range. The dotted black line represents the 45 degree line, intercept 0, slope 1. The green dashed line extends from 0 through the mean of the luminal means. The red dashed line extends from 0 through the mean of the myoepithelial means. If a coloured line falls below the black line, the shRNA is presumably decreasing the colony formation ability of the cells. If a coloured line falls above the black line, the shRNA is presumably increasing the colony formation ability of the cells.  126  5  Celsr1 regulates mammary progenitor cell growth and branching  morphogenesis 5.1  Introduction Celsr1 is a member of a core group of conserved genes that regulate planar cell  polarity in a variety of model organisms and tissues. Planar polarity refers to the orientation of cells within a sheet or plane (orthogonal to their apical-basal axis) that is presumably conferred through an underlying gradient of positional information [300]. Studies in Drosophila melanogaster have been the driving force behind many of the advancements in understanding planar cell polarity. Celsr1, known as flamingo in Drosophila, was first identified as a cell autonomous regulator of planar cell polarity in the Drosophila wing [301]. Here, flamingo was apically located and bilaterally distributed on the proximal and distal sides of cells. Along the distal edge of the cell, flamingo forms a complex with frizzled and dishevelled/diego [302]. On the opposing proximal cell edge, flamingo complexes with strabismus (also known as Van Gogh) and prickle [303]. Between cells, flamingo interacts homotypically with itself and strabismus interacts heterotypically with frizzled [304]. The intracellular region of flamingo binds frizzled on the distal side of a cell. It then signals to its neighboring cell to recruit strabismus to its proximal cell edge through the extracellular domains of the interacting flamingo molecules [305]. This implies that flamingo is able to signal in a bi-directional manner and that the intracellular signals are asymmetric, a phenomenon that has not previously been reported for other cadherin molecules. Flamingo is essential for the initial recruitment of frizzled to the cell surface, but the loss of the downstream mediators of frizzled signaling (dishevelled and diego) blocks the subsequent asymmetric distribution of flamingo, frizzled and strabismus [306-308]. Despite understanding the propagation of polarity signals through a plane of cells, the manner through which the establishment of polarity is initiated remains debated [300]. The first description of Celsr1 playing a role in establishing planar cell polarity within mammals came about from the identification of two independent mouse mutants with point mutations in Celsr1 generated through an N-ethyl-N-nitrosourea (ENU) mutagenesis screen [309]. These mice exhibited a head-shaking phenotype resulting 127  from the loss of polarity within the sensory cells lining the organ of Corti. This link between Celsr1 and the stereocilia of the inner ear has remained conserved for ciliogenesis across other systems and other polarity genes [310-312]. Cilia formation in mesenchymal cells has been shown to be required for hair follicle development in adjacent epithelium [313, 314]. Not surprisingly, studies with Celsr1 mutant mice revealed a role for both Celsr1 and planar cell polarity in the organization and development of hair follicles [315]. Interestingly, deletion of Celsr2 and Celsr3 within mice disrupts cilia formation in the ependymal cells lining the cerebral ventricles. This impedes the flow of cerebrospinal fluid, leading to progressive hydrocephaly [316]. Despite hydrocephaly not being observed in Celsr1 mice, a similar finding in Van Gogh mutated mice suggests that this is related to planar cell polarity and not an alternative function of Celsr2 or Celsr3 [317]. It also suggests that some redundancy may exist between these homologues in certain situations. Celsr1 deficient mice exhibit severe neural tube defects whereby the neural tube remains open from the hindbrain down the spine, a common defect when the core planar cell polarity genes are disrupted [174, 309, 318]. Failed neural tube closure is thought to arise from defects in convergent extension, a process that requires both cell rearrangement and coordinated migration [319]. A direct association between cell migration, Celsr1, and planar cell polarity has been shown during epidermal wound repair [320]. Celsr1 has also been shown to specify the direction of facial branchiomotor neuron migration, which is required for the proper innervation of the muscles responsible for facial expression [321]. This function is thought to result from a non-cell autonomous effect whereby Celsr1 is required within the neuroepithelial cells to guide migration of neurons along their outer surface [322]. A final described role for Celsr1 and planar cell polarity is the regulation of branching morphogenesis within the murine lung [323]. Hypobranching of the lung epithelium occurs in the Celsr1 mutants, with lack of septation resulting in fewer epithelial cell branches. Within these branches, the epithelium is disorganized and the lumens are narrower than usual. Additionally, cells are found within the normally clear luminal space. Similar effects on kidney branching morphogenesis have been reported in Van Gogh mutated mice and are again related to its  128  role in planar cell polarity [324]. Because of this, it was not surprising to find here that Celsr1 is required for branching morphogenesis within the mammary gland. Here, we investigated the role of Celsr1 in mammary gland development. Celsr1 is differentially expressed in mammary progenitor cell populations, with an increase in expression corresponding to further commitment to the luminal cell fate. When placed within the colony-forming assay, silencing of Celsr1 leads to an increase in the number of detectable bi-potent progenitor cells with no effect on the luminal progenitor cells. This suggests that Celsr1 may control self-renewal of bi-potent progenitors, or may stimulate the de-differentiation of a non-progenitor cell back into a primitive cell fate. Further to this, Celsr1 prevents branching when 184-hTERT and primary human mammary cells are cultured in 3D Matrigel assays. Surprisingly, the opposite effect was seen when the transmembrane domain of Celsr1 was constitutively deleted in the mouse through Cre/lox-mediated excision of exons 26–29 [174]. In Celsr1 homozygous deleted females, the epithelial tree fails to penetrate through the entire mammary fat pad and hypobranching of the ductal structure is seen. The deletion of factors as diverse as the parathyroid hormone-related protein, vitamin D3 receptor, Adam17, and the progesterone receptor in mammary epithelial cells will impair mammary ductal branching [325-327]. Similarly, deleting growth hormone, Gli2, or activins from the stromal cells will also impair epithelial cell branching [328-330]. Thus, both stromal and epithelial cell effects are known to regulate mammary branching. In order to understand the differences between the in vitro human and in vivo murine systems, we must further investigate the role of the stroma in Celsr1 deficient branching morphogenesis. Finally, within the in vitro system, we identified Shisa4 as coordinately regulated with Celsr1 and similarly functioning to suppress branching in Matrigel. This may represent a novel signaling axis through which Celsr1 may act. Whilst alternate signaling pathways for Celsr1 have been proposed to exist, potential interactors have yet to be identified. Shisa4 is a promising candidate for this within the mammary gland.  129  5.2 5.2.1  Results Celsr1 differentially regulates mammary progenitor cells Celsr1 was initially identified in chapter 4 whilst investigating the fibroblast-  dependent growth regulation of mammary cells. When grown with fibroblasts, Celsr1 and its homologues, Celsr2 and Celsr3, are up-regulated in 184-hTERT cells relative to growth with bovine pituitary extract (Figure 5.1A). We were immediately interested in this gene as it is found to be co-regulated with the estrogen receptor in a subset of primary breast tumours, in a manner similar to the luminal fate determinant Gata-3 [331]. An estrogen receptor binding site is located directly adjacent to the Celsr1 gene suggesting that its expression may be controlled by estrogen in the mammary gland [332]. This prompted our investigation into the expression status of Celsr1 within mammary progenitor cell populations and subsequently in branching morphogenesis, both of which are heavily influenced by estrogen signaling during development. Figure 5.1B shows the relative expression of Celsr1 in 4 patient samples that were prospectively fractionated into bi-potent progenitor, luminal progenitor, differentiated luminal and differentiated myoepithelial/stromal cell enriched populations. Expression of Celsr1 increased as cells were progressively committed to the luminal lineage, from bi-potent progenitor cells to differentiated luminal cells. Expression was decreased in the differentiated myoepithelial/stromal fraction relative to the bi-potent progenitor cells, however this is a heterogeneous fraction with little known about the proportions of each cell type. Nonetheless, this confounds the expression results, as the lower expression of Celsr1 in the bi-potent progenitor cell fraction cannot be solely attributable to the primitive state of the cells and may be related to their overall basal phenotype. This expression pattern mimics what is seen with the Gata-3 transcription factor in these same cell populations [32]. The differential expression of Celsr1 within fractionated mammary cells led us to assess if a functional relationship exists between Celsr1 and mammary progenitor cells. To accomplish this, we silenced Celsr1 in primary human mammary epithelial cells obtained from reduction mammoplasty tissue prior to plating these cells in colony-  130  forming assays. After 10 days, luminal and mixed/myoepithelial colonies were enumerated through morphologic discrimination under a dissecting microscope. An increase in the number of mixed/myoepithelial colonies was observed when cells were infected with 2 of the 3 lentiviral shRNA constructs targeting Celsr1. Detection of pure myoepithelial colonies is rare in this assay since myoepithelial progenitor cells comprise only a small percentage of total colony forming cells [31]. Thus, the grouping of mixed and myoepithelial colony types refer mainly to mixed colonies and thus roughly quantifies the presence of bi-potent progenitor cells. Therefore, silencing of Celsr1 leads to an increase in the number of detectable bi-potent progenitor cells within this assay. No change was observed in the number of luminal colonies after infection and thus silencing of Celsr1 has no effect on luminal progenitor cells (Figure 5.2A). The expression pattern of Celsr1 in luminal progenitor and differentiated luminal cells suggests that Celsr1 may be involved in the specification of the luminal lineage. If this is true, the increase in the number of bi-potent progenitor cells seen within the colony assay in Figure 5.2A could be attributable to de-differentiation of luminal cells. To assess this possibility, we prospectively fractionated reduction mammoplasty samples into populations that were enriched for bi-potent progenitor cells. If de-differentiation of luminal cells was responsible for the increase in mixed/myoepithelial colonies, no change in colony numbers would be observed once luminal cells had been depleted. Figure 5.2B shows that this is not the case and that there is an even greater increase in the number of bi-potent progenitor cells upon fractionation. One hypothesis explaining this phenomenon is that Celsr1 prevents self-renewal of cells, and thus silencing its expression allows for expansion of the progenitor cell compartment. Alternatively, silencing of Celsr1 may confer colony-forming abilities on phenotypically similar cells found within the same fractionated compartment as the bi-potent progenitor cells. Further investigations are required to pinpoint the manner through which Celsr1 regulates bi-potent progenitor cell behavior in the colony-forming assay. 5.2.2  Silencing of Celsr1 induces epithelial cell branching in 3D Matrigel culture A role for Celsr1 and its homologues in cell migration and branching has begun to  emerge in the last few years. In the mammary gland, branching morphogenesis is a  131  complex process that can be affected by manipulating both the epithelial and the stromal cells. This phenomenon can be studied in vitro using 3D Matrigel cultures. Typically, human mammary epithelial cells will form spherical structures when seeded into a thick layer of artificial basement membrane (Matrigel). Branching within these structures can be induced by the addition of epidermal growth factor, hepatocyte growth factor, keratinocyte growth factor, fibroblast growth factor 7 and basic fibroblast growth factor 2 [333]. In light of the effect on bi-potent progenitor cells and the current function of Celsr1 within other organ systems, we used this assay to address potential effects of Celsr1 on branching morphogenesis within the mammary gland. 184-hTERT cells form structures reminiscent of normal mammary acini in Matrigel culture (Chapter 3). To establish the effect of silencing Celsr1 on this system, 184-hTERT cells were transfected with siRNAs prior to harvesting and seeding within Matrigel. 184-hTERT cells cannot be directly transfected within Matrigel and the lipidcomplexed siRNAs cannot penetrate through the gel. Figure 5.3 shows representative structures from 184-hTERT cells transfected 48 or 96 hours prior to seeding in Matrigel. Images were taken at day 21, but these elaborate branched structures had formed by Day 8. Silencing of e-cadherin, an unrelated surface molecule, does not affect the acini, nor does the non-targeting control siCONTROL3. As siRNAs only produce temporary gene silencing, we assessed Celsr1 transcript levels in Matrigel after 10 and 21 days in culture (Figure 5.4). Relative to the transfection reagent only control, the siRNA pool reduces Celsr1 transcript levels by 3.8 fold 48 hours after transfection. Transcript levels rise during culture, with relative expression almost fully restored by day 21. The presence of the branched structures by day 8 suggests that the spherical acini do not need to fully form before branching occurs. When placed in Matrigel in defined media with B27 supplement, only the progenitor cells from primary mammary tissue will form structures [32]. Differentiated cell fractions were non-proliferative in this system, similar to what is seen when they are placed in the 2D colony-forming assay. Under these conditions, the structures from both the bi-potent and luminal progenitor cells are spherical and do not branch. However, when primary cells are transduced with shRNAs targeting Celsr1 prior to seeding in Matrigel, branching occurs from these otherwise spherical structures (Figure 5.5). With 2  132  of the 3 shRNA constructs, the spherical acini appear to be fully formed with hollow lumens and established apico-basal polarity. However, from these acini, branched protrusions extend out as a multi-cellular layer. One of the three shRNAs completely inhibited 3D growth, and was the same shRNA construct that showed no effect in the 2D progenitor cells assays described above. 5.2.3  Celsr1 is required to establish apical polarity in 3D Matrigel culture Celsr1 is best known for its role in regulating planar cell polarity. While planar  cell polarity has not yet been described in the mammary gland, apico-basal polarity plays a major role in mammary gland morphogenesis and in 3D culturing. Recently, the lines between these two types of polarity have been blurred upon the realization that apical localization of the planar cell polarity determinants is required [334]. Notably, the apicobasal polarity gene Scribble has been implicated in establishing planar cell polarity within the eye and wing of Drosophila. This is partly mediated by directing the localization of the planar cell polarity determinants to the apical side of the cell [335]. Within mammary epithelial cells, specifically MCF10A cells, Scribble is required to establish apical polarity but does not disrupt basal cell polarity [336]. We evaluated the establishment of apico-basal polarity in 184-hTERT cells with silenced Celsr1 in Matrigel. Figure 5.6 shows that basal polarity, marked by CD49f expression, is retained throughout the structures. However, apical polarity, assessed through staining with the golgi marker GM130, was disrupted. As seen in the Lipofectmaine 2000 alone control, when apical polarity is established the Golgi will align above the nuclei in an organized manner parallel to the basement membrane. When Celsr1 is silenced, the spatial organization of GM130 is disorganized relative to the basement membrane. Thus, it appears that Celsr1 is required for the establishment of apical but not basal polarity in 184-hTERT cells. An important model for evaluating 3D morphogenesis in mammary cells is the growth of MCF10A mammary epithelial cells in Matrigel. While they are not an ideal model system, their widespread use has lead to the thorough characterization of the processes involved in their formation and their deregulation [337]. When Celsr1 is silenced in these cells prior to 3D culturing, they form hyper-proliferative structures with filled lumens (Figure 5.7). Unlike the 184-hTERT cells, the apical polarity in the outer  133  ring of cells in contact with the basement membrane is not disrupted. No polarity exists within the inner cells, which is as expected since these cells are not in contact with extracellular matrix. Normally, these inner cells would undergo apoptosis in response to the loss of this signaling axis [338]. Additionally, each individual structure is significantly larger when Celsr1 is silenced and composed of an exponentially greater number of cells. It is not yet known whether this is due to aggregation of smaller structures, or the enhanced proliferative capacity of clonal cells. To rigorously address this, single cells from each condition will need to be seeded within Matrigel and the structures compared. Irrespective of how they form, these structures are reminiscent of what is seen when ErbB2 is constitutively activated in MCF10A cells, a process that involves both the stimulation of proliferation and resistance to apoptosis [339]. This hints that, like Scribble, Celsr1 may be a potential tumour suppressor gene [340]. Downregulation of Celsr1 was associated with poor prognosis in breast cancer patients upon meta-analysis the expression profiles of 1004 breast cancer patients [341]. These observations warrant further investigation into the role of Celsr1 in malignant progression. 5.2.4  Celsr1 may coordinately regulate branching with Shisa4 Celsr1 has been primarily characterized to function through its role in canonical  planar cell polarity signaling. However, Celsr1 has been shown to possess direct signaling potential in murine neuroepithelium through homotypic interactions between Celsr1 molecules in adjacent cells [342]. The mechanism for this is not yet known, but it does not involve the planar cell polarity pathway. To investigate how Celsr1 may function in mammary epithelium, we silenced Celsr1 using siRNAs in 184-hTERT cells and assessed the resultant changes in gene expression. A striking program of coordinated down-regulation in several known polarity genes was identified (Table 5.1). If any of these genes were acting in conjunction with Celsr1, one would expect that upon their silencing, their phenotype would mimic that of Celsr1 silencing. Focusing on the branching phenotype, we transfected 184-hTERT cells with siRNAs targeting a selection of these genes prior to seeding the cells in Matrigel (Figure 5.8). Several had no effect on the 184-hTERT acini, including Fat4, Gpr115, and Zpld1. One abnormal structure was  134  formed when Fry was silenced in 1 of the 2 duplicate experiments performed. Further investigation is required to determine if this was a real event, as spontaneous branching of the 184-hTERT cells has not been previously observed under any other condition. Silencing of Shisa4 lead to an exact phenocopy of epithelial cell branching seen with RNA interference of Celsr1. This suggests that Celsr1 and Shisa4 may act in conjunction with each other to regulate branching morphogenesis. Further work is required to formalize a physical interaction between these two molecules. Irrespective of a formal interaction, this work has revealed two novel regulators of branching morphogenesis in the mammary gland residing in spheres of research that have not previously been implicated in mammary development. 5.2.5  Preliminary phenotype of Celsr1 loss in mouse mammary glands The current gold standard for assessing branching morphogenesis in the  mammary gland is to investigate in vivo branching within the mouse. Human xenotransplantation assays are not yet sufficiently robust to be used in rigorous investigations of branching and thus a switch in our model organism was required. To study the in vivo effects of silencing Celsr1, we constitutively deleted Celsr1 to examine the resultant epithelial ductal tree in adult mice. Mice with loxP sites flanking the transmembrane region of the Celsr1 gene were generously provided to us by Dr. F. Tissir [174]. We crossed these mice with pCX-NLS-Cre mice, a transgenic line that constitutively expresses nuclear-localized Cre recombinase [175]. Roughly half of the homozygous deleted mice die in utero due to defects in neural tube closure. The remaining mice have kinked tails and whorled fur patterning, as previously described [174]. We have observed that the homozygous deleted male mice will not breed with heterozygous deleted females and thus our breeding strategy necessitates crossing heterozygous deleted male and female mice. Preliminary investigations have revealed that the homozygous deleted males produce 50% less sperm than their wild-type counterparts and that these sperm lack tails rendering them immobile (data not shown). Carmine alum stained mammary explants from homozygous deleted female mice and their wild-type littermates are shown in Figure 5.9 (estrus cycles matched between pairs by observing the gross morphology of the right uterine horn). In two of the Celsr1  135  deleted mice, the ductal tree does not fill the entire fat pad during development, leaving one-third or more of the fat pad devoid of epithelium (Figure 5.9A and Figure 5.9C). Less side branching is also evident in these glands compared to their matched controls. In one of the deleted mice, the epithelial tree does extend to the ends of the fat pad (Figure 5.9E). However, there is considerably less side branching in this mouse. This suggests that variegated effects can occur following Celsr1 deletion across mice. Additionally, in subsequent homozygous deleted female mice we have observed variegated effects of Celsr1 deletion on the reproductive system. While the numbers of mice observed are too few to establish a phenotype with certainty, we have subsequently observed homozygous deleted females with distended uterine horns, 1 or more absent uterine horns and 1 or more absent ovaries. Because of this, we must be cautious with the observations made regarding the mammary glands as the hormonal milieu greatly influences branching morphogenesis. While we established the estrus stage by observing the right uterine horn in the 3 mice from which the mammary gland wholemounts were obtained, we neglected to confirm the presence of the left uterine horn and both ovaries. Further experiments must be performed to separate out the effect of Celsr1 deletion from any potential defects in the hormonal milieu within these mice. Initially, the levels of circulating estrogen and progesterone can be monitored in Celsr1 deleted mice to establish if differences are present prior to resecting the mammary glands for wholemount analysis. Artificial estrogen and progesterone supplementation in the Celsr1 deleted mice can also be used to eliminate the potential confounding factor of a hormone deficiency contributing to the hypobranching observed upon Celsr1 deletion. Alternatively, transplantation of the Celsr1 deleted epithelium obtained prior to puberty into cleared wild-type mammary fat pads could be performed to control for the differences in hormones that may be present within these mice. The latter experiment would also establish if hypobranching resulting from Celsr1 depletion is affected by the Celsr1 status of the stromal cells. The breeding strategy employed to obtain homozygous Celsr1 deleted mice and the subsequent embryonic lethality of these mice limits the number of homozygous deleted females that can be obtained. As such, we sought to confirm the hypobranching seen with Celsr1 deletion using a tamoxifen inducible Cre recombinase system. For this,  136  the Celsr1 floxed mice were crossed with Rosa26-CreER mice [343]. Double homozygotes were treated with 1 mg of tamoxifen for 2 days at 3 weeks of age to induce Cre-mediated deletion of Celsr1. Mammary explants were examined 4 weeks posttreatment. In all 4 tamoxifen treated mice, at least one-third of the fat pad was devoid of epithelium and reduced side branching was evident (data not shown). Deletion of Celsr1 in the mammary gland after treatment was confirmed through PCR. No effect on branching was observed when Rosa26-CreER mice were treated with tamoxifen using the same dosing schedule. Despite this, the use of tamoxifen to establish an effect on mammary gland branching is not ideal as the disruption of the hormonal milieu at puberty may confound the effect of Celsr1 deletion. As outlined above, further experiments are required to ensure that the hypobranching seen is not related to a disruption of estrogen and progesterone levels in these mice. The in vivo mouse data has the opposite effect to what is observed in vitro when human cells are placed in Matrigel. These contradictory responses may reflect physiological differences between species, or may be related to the assay systems. Matrigel is an artificial assay system and thus may not represent true physiology. Alternatively, the effect of Celsr1 may be heavily dependent upon the concentration of Celsr1 present in a cell. With RNA interference, residual expression can still be present and may produce a different response than when no copies of functional Celsr1 are transcribed due to gene deletion. To begin to investigate this, the CommaD-β mouse epithelial cell line was transfected with siRNA prior to seeding in Matrigel. These cells did not provide a uniform response in Matrigel in either the control or test conditions (Figure 5.10). While the acini that formed were mostly spherical, several non-spherical structures formed that ranged from protrusions stemming out of the spherical acini to small branched structures. While the number of non-spherical structures was greater when Celsr1 was silenced, this intermediate response was not particularly informative. Therefore, primary mouse mammary cells will need to be plated within this Matrigel assay in order to assess their response. If branching occurs, then the differential effects are likely to be assay dependent. If not, the level of Celsr1 expression will need to be reduced in wild-type primary mammary cells with RNA interference prior to seeding in Matrigel. Failure to recapitulate the branching effects seen in human cells after these 137  investigations would leave species specific effects as the most likely explanation for the differences seen between the in vivo and in vitro results. Nevertheless, Celsr1 is affecting branching morphogenesis within the mammary gland, despite the inconsistencies within the system. This is the first description of a functional role for Celsr1 in the mammary gland. 5.3  Discussion This work has identified two distinct roles for Celsr1 within the mammary gland.  Deregulation of Celsr1 leads to an increase in the number of bi-potent progenitor cells detected in human tissue. It also regulates branching morphogenesis through a potentially novel signaling axis co-mediated by Shisa4. It is unknown whether these two functional effects are related, or if they occur through independent means. This work has lead to more questions than was feasible to address here. However, the observations made here clearly outline the next steps that are required to fully characterize the role of Celsr1 in mammary development. As cells progress from a bi-potent progenitor cell state to differentiated luminal cells through the process of lineage commitment, the expression of Celsr1 gradually increases. However, it is not clear if these expression changes are due to the loss of the basal phenotype or rather the loss of the primitive characteristics associated with progenitor cells. As the location of the bi-potent progenitor cells within the mammary tissue remains controversial, we cannot answer this question by simply staining for Celsr1 in sectioned mammary tissue. Until methods are developed to prospectively fractionate a pure population of differentiated myoepithelial cells devoid of myoepithelial progenitors and contaminating stroma, we will be unable to address this issue further. When Celsr1 is silenced in unsorted primary cells prior to plating in the colonyforming assay, the number of detectable bi-potent progenitor cells increases. There is no effect on the luminal progenitor cells. If Celsr1 were involved in maintaining cells in a primitive state, the loss of Celsr1 in the luminal cells would seemingly contribute to the increase in the number of bi-potent progenitor cells through de-differentiation. However, when bi-potent progenitor cells are prospectively purified prior to plating within this assay, their numbers are still increased upon silencing of Celsr1 showing that this  138  phenomenon is not related to the de-differentiation of luminal cells. However, the bipotent progenitor cell enriched fraction is not composed entirely of progenitor cells and contains some differentiated myoepithelial cells. Thus, it is possible that dedifferentiation of these cells is occurring. Alternatively, Celsr1 actively prevents bipotent progenitor cells from undergoing self-renewal and silencing its expression allows these cells to expand. As luminal progenitor cells express higher levels of Celsr1, the lack of an effect within this population may be a function of differential expression between these cells. RNA interference can leave residual levels of mRNA transcripts within cells. If endogenous levels of Celsr1 are higher, such as in the luminal progenitor cells, the levels of residual expression may remain high enough to prevent self-renewal from occurring within this population and thus no effect is seen in the colony-forming assay. Further investigations are required to pinpoint the manner in which Celsr1 is regulating bi-potent progenitor cells. However, this will be difficult to accomplish until cell fractionation schemes are further improved upon and purer populations of cells can be isolated. An alternate approach would be to increase the expression of Celsr1 in the bi-potent progenitor cells and look for the disappearance of colony-forming cells. The size of Celsr1 negates the possibility of generating a viral cDNA construct that would successfully produce infectious particles and the limited transfectability of these cells would hinder the direct transfer of this construct into the cells. Careful consideration of the limitations imposed by this particular cell population and working with a protein this large will be required to produce a successful method to evaluate this further. The branching effects seen in human cells plated in Matrigel are robust and dramatic, yet entirely opposite to the hypobranching seen within homozygous deleted mouse mammary glands. Explaining these differences is complex due to the myriad of potential confounding factors. Initially, replication of the exact Matrigel assay conditions in the primary mouse epithelial cells will help to determine is these differences are species specific or assay dependent. Irrespective of these results, the hypobranching within the mouse model system will still require further investigation to delineate the role of Celsr1 in mammary development. In the mouse model, Celsr1 is deleted in both the epithelial cells and the stroma. Celsr1 may be required for normal epithelial cell function, for the deposition of the correct extracellular matrix, or for stromal-epithelial cell  139  interactions that occur during branching. To assess this, wild-type mammary epithelial cells must be transplanted into a cleared mammary fat pad of a Celsr1 knockout mouse, and vice versa, to determine if hypobranching is the result of an altered signaling environment within the stroma or a deficiency within the epithelial cells. Further to this, Celsr1 could be deleted in either the basal or luminal cell compartments alone using a lineage specific promoter to drive Cre expression. If the effect were intrinsic to the epithelial cells, this approach would determine if cells form the two compartments are equally dependent upon Celsr1 for branching to occur. Establishing the fundamental framework for how Celsr1 functions in the mouse mammary gland is an important step in understanding the fundamental role Celsr1 may be playing in branching morphogesis. Celsr1 has traditionally been viewed as controlling planar cell polarity. Recently, an alternate signaling axis has been suggested but not yet defined [342]. We have identified Shisa4 as being coordinately down-regulated upon silencing of Celsr1. Silencing of Shisa4 phenocopied the branching effect seen in 184-hTERT cells in Matrigel. Shisa family members have been show to antagonize Wnt and FGF signaling by preventing the maturation and surface presentation of their receptors [344]. This function has not been reported with Shisa4, potentially due to its predicted location on the plasma membrane as opposed to retention within the endoplasmic reticulum seen with the other Shisa isoforms [345]. Whilst not confirmed, the predicted location of Shisa4 suggests that it has the opportunity to directly interact with Celsr1 on the plasma membrane. Initially, co-localization of these proteins should be confirmed through fluorescence microscopy. Following that, immunoprecipitation of Shisa4 and/or Celsr1 will allow for a direct interaction to be interrogated. The coordinate regulation of these two proteins suggests that they are influenced by a similar feedback mechanism. The down-regulation of Celsr1 within cells may provide a signal to suppress the polarity or an alternate signaling programme through a common transcriptional mechanism. Initially, the transcription factor Stat5A should be explored as a potential regulator of this phenomenon as it is up-regulated when Celsr1 is silenced. Stat5A is required for luminal cell development from mammary stem cells and for secondary branching during ductal morphogenesis [346, 347]. Manipulating Stat5a within cells should lead to changes in the expression of both Shisa4 and Celsr1 if it is responsible for regulating these genes. If  140  this is the case, performing chromatin immunoprecipitation with Stat5a will provide evidence of a direct regulatory effect. If a connection fails to be found, surveying the promoter regions of both Shisa4 and Celsr1 to find common transcriptional regulators would be the next logical approach. Finding a mechanism for co-regulation may be a difficult venture, but one that would ultimately identify a new signaling axis in mammary development.  141  A  BPE  10  i3T3  Relative Expression  8  6  4  2 BPE  0 0  B  Celsr1 2 C-terminal domain  Celsr1 1 N-terminal domain  Patient 1  Patient 2  3 Celsr2  4 Celsr3  Patient 3  5  Patient 4  Relative Expression  10  Bi-potent progenitors  1  0.1  0.01 0  Luminal progenitors  1  Differentiated luminal  C  2  Differentiated myoepithelial/ stroma  3  D  10 +m  142  Figure 5.1 – Celsr1 is differentially expressed in fibroblast-dependent growth and in mammary progenitor cell subsets A) 184-hTERT-L9 cells were grown with either irradiated fibroblasts in assay media, or with MEGM (assay media containing BPE). After 48 hours, RNA was harvested and RT-QPCR performed with primer/probe pairs detecting the region of the mRNA transcript corresponding to the N-terminal and C-terminal domains of Celsr1, in addition to Celsr2 and Celsr3. Expression when cells were grown with fibroblasts is shown relative to when they were grown with BPE (red line). Bars denote 95% confidence intervals (n=3). B) Reduction mammoplasty samples from 4 patients were depleted of endothelial and hematopoietic cells prior to prospective fractionation into bi-potent progenitor, luminal progenitor, differentiated luminal and differentiated myoepithelial/stromal fractions based on the expression of CD49f, Muc-1 and Thy-1. RNA was harvested and RT-QPCR performed to assess Celsr1 expression levels relative to the bi-potent progenitor cell enriched population (red line). Bars denote 95% confidence intervals (n=3). C) 184-hTERT-L9 cells were grown on glass coverslips, fixed with paraformaldehyde, and stained with an antibody for Celsr1. Secondary antibody alone did not produce background staining. D) Comma D-beta cells were transfected with a mouse Celsr1 cDNA construct fused to GFP using Lipofectamine 2000.  143  A  6  Mixed/Myo colonies  Luminal colonies  Fold change  5 4 3 2  Control  1 0 shRNA #1  B  shRNA #2  shRNA #3  shRNA #1  shRNA #2  shRNA #3  80  Colony Count  60  40  20  0 0.5  Control  1.5 shRNA #1 2.5 shRNA #2 3.5 shRNA #3 4.5  Figure 5.2 – Celsr1 negatively regulates the colony-forming ability of bi-potent progenitor cells A) Reduction mammoplasty samples from 3 patients were dissociated into single cell suspensions and infected with lentiviral shRNA constructs targeting GFP (control) and Celsr1 (shRNA #1 to #3). After 18 hours, cells were plated in the Colony-forming Assay for 10 days. Using a dissecting microscope, colony types were scored into mixed/myoepithelial and luminal colonies. Data is shown relative to the control condition (n=3). B) ) Reduction mammoplasty samples from 3 patients were depleted of endothelial and hematopoietic cells prior to prospective fractionation into bi-potent progenitor cell enriched fractions based on the expression of CD49f and EpCam. Cells were infected with lentiviral shRNA constructs targeting GFP (control) and Celsr1 (shRNA #1 to #3). After 18 hours, cells were plated in the Colony-forming Assay and the number of resultant colonies was scored after 10 days (n=3).  144  145  Figure 5.3 – Celsr1 regulates branching in 184-hTERT 3D Matrigel cultures 184-hTERT-L9 cells were transfected with siCONTROL3 or pooled siRNAs targeting Celsr1 and E-cadherin. After 48 hours and 96 hours, cells were harvested and re-plated in 3D Matrigel culture. By Day 8, branched tubules arose in the Celsr1 silenced condition. Images were captured on Day 21.  146  1  Relative Expression  0.8  0.6  0.4  0.2  0 0.5  Day 0  1.5  Day 10  2.5  Day 21  3.5  Figure 5.4 – Celsr1 silenced through siRNA remains repressed in Matrigel 184-hTERT-L9 cells were transfected with pooled siRNAs targeting Celsr1 or Lipofectamine 2000 alone. After 48, cells were harvested and re-plated in 3D Matrigel culture. RNA was extracted from cells just prior to plating, on Day 10 in culture, and on Day 21. RT-QPCR was performed to assess Celsr1 expression levels relative to Lipofectamine 2000 alone (relative expression of control = 1). Bars denote 95% confidence intervals (n=3).  147  Figure 5.5 – Celsr1 regulates branching in 3D culture of primary mammary cells A reduction mammoplasty sample was dissociated into a single cell suspension and infected with lentiviral shRNA constructs targeting GFP (control) and Celsr1 (shRNA #1 to #3). After 18 hours, cells were plated into 3D Matrigel culture and stained for CD49f and GM130 after 21 days. Structures did not arise from shRNA #3. Branching is otherwise observed when Celsr1 is silenced, with polarity markers re-organized within the tubular structures.  148  Figure 5.6 – Celsr1 regulates polarity in 184-hTERT 3D Matrigel culture 184-hTERT-L9 cells were transfected with pooled siRNAs targeting Celsr1 or Lipofectamine 2000 alone. After 48 hours, cells were harvested and re-plated in 3D Matrigel culture. After 21 days, cultures were stained for markers of basal polarity (CD49f) and apical polarity (GM130). While basal polarity is maintained, GM130 is no longer properly aligned within the 3D acinar structures when Celsr1 is silenced.  149  Figure 5.7 – Silencing of Celsr1 in MCF10A cells leads to luminal filling and hyperproliferation in 3D Matrigel culture MCF-10A cells were transfected with pooled siRNAs targeting Celsr1. After 48 hours, cells were harvested and re-plated in 3D Matrigel culture. After 21 days, cultures were stained for markers of basal polarity (CD49f) and apical polarity (GM130). In this system, Celsr1 suppression does not affect apical polarity, but luminal filling and hyperproliferation is observed.  150  Figure 5.8 – Silencing of Shisa4 phenocopies the regulation of branching observed with Celsr1 184-hTERT-L9 cells were transfected with pooled siRNAs targeting surface molecules that are coordinately down-regulated when Celsr1 is silenced. After 48 hours, cells were harvested and re-plated in 3D Matrigel culture. Brightfield images were captured after 21 days. One branched structure was observed when Fry is silenced over duplicate cells and can thus be considered a rare event. Shisa4 produced branched structures similar to what is seen with Celsr1 in duplicate well (n=1). All images were captured using a 10x objective on a brightfield microscope.  151  152  Figure 5.9 – Celsr1 deletion results in hypobranching of the epithelium within mouse mammary gland The fourth inguinal mammary glands were resected from Celsr1 -/- mice generated from crossing pcxNLScre mice with Celsr1 fx/fx to obtain heterozygous Celsr1 deleted lines that were then interbred. Wild-type littermates in the same stage of estrus were selected by observing the gross morphology of the uterus upon autopsy (matched pairs A/B, C/ D, and E/F). Mammary glands were fixed onto glass slides in Carnoy’s fixative and stained with Carmine Alum. Representative images from a dissecting microscope are shown with final magnifications indicated.  153  154  Figure 5.10 – Silencing of Celsr1 affects the morphology of 3D structures in mouse cells A) Comma D Beta cells were transfected with pooled siRNAs targeting Celsr1 or Lipofectamine 2000 alone. After 48 hours, cells were harvested and re-plated in 3D Matrigel culture. After 21 days, cultures were enumerated for the number of structures formed in matrigel that were of typical spherical morphology or non-spherical morphology (branched or bulbous). The overall number of structures decreases with Celsr1 silencing, with a commensurate increase in the number of non-spherical structures. Bars denoted 95% confidence intervals (n=3). B) Representative brightfield images of spherical and non-spherical structures enumerated above.  155  Table 5.1 – Program of coordinate gene regulation upon Celsr1 silencing Changes in gene expression that occur upon siRNA mediated silencing of Celsr1 in 184-hTERT-L9 cells relative to a Lipofectamine 2000 only control. Expression levels were assessed by Affymetrix Human Exon 1.0 ST Expression Arrays and validated by RTQPCR (p-values less than 0.05). Gene ID  DESCRIPTION  IL1RL1 ATP6V0D2 ZPLD1 GEM EPCAM ADAMTS1 FAT4  interleukin 1 receptor-like 1 ATPase, H+ transporting, lysosomal 38kDa, V0 subunit d2 zona pellucida-like domain containing 1 GTP binding protein overexpressed in skeletal muscle epithelial cell adhesion molecule ADAM metallopeptidase with thrombospondin type 1 motif, 1 FAT tumor suppressor homolog 4 (Drosophila)  ODZ1 BPIL2  odz, odd Oz/ten-m homolog 1(Drosophila) bactericidal/permeability-increasing protein-like 2  ELMOD1 SHISA4 FRY GPR115 SERPING1 NCRNA00204 CSF3 SERPINB3 TRIB2 PLAT  ELMO/CED-12 domain containing 1 shisa homolog 4 (Xenopus laevis) furry homolog (Drosophila) G protein-coupled receptor 115 serpin peptidase inhibitor, clade G (C1 inhibitor), member 1 non-protein coding RNA 204 colony stimulating factor 3 (granulocyte) serpin peptidase inhibitor, clade B (ovalbumin), member 3 tribbles homolog 2 (Drosophila) plasminogen activator, tissue  CDT1  chromatin licensing and DNA replication factor 1  Abs Fold Change Affy 9.04 8.51 2.67 7.67 2.59  Down Down Down Down Down  Fold Change RTQPCR 65.31 10.39 9.88 6.22 4.83  5.20 2.47  Down Down  4.79 4.23  Down Down  2.46 5.41  Down Down  4.05 3.82  Down Down  2.42 2.84 2.46 2.80 4.01 9.04 3.07 7.41 2.34 11.31  Down Down Down Down Down Down Up Up Up Up  3.27 3.10 3.02 2.35 2.10 2.09 4.67 3.50 2.65 2.36  Down Down Down Down Down Down Up Up Up Up  Up  2.15  Up  2.80  Fold Change Direction  156  Fold Change Direction Down Down Down Down Down  6  Epigenetic regulation of genome stability in non-transformed  mammary epithelial cells 6.1  Introduction Genome instability is a pathological event that allows cells to acquire the genetic  alterations necessary for tumourigenesis [113]. In its basic form, it is the unfaithful transmission of genetic information from a dividing cell to its progeny. This can include a spectrum of conditions from subtle alterations in the DNA sequence to gene amplification or deletions, chromosomal rearrangements or aneuploidy [348]. DNA sequence alterations, such as point mutations and microsatellite instability, can accumulate in cancers due to defects in nucleotide excision repair, base excision repair, or mismatch excision repair [349, 350]. Microsatellite instability is not a common feature of breast cancer [351]. Chromosomal instability, manifested as gross chromosomal rearrangements or aneuploidy, is a prominent feature of breast cancers and is closely linked with clinical subtype, prognosis and malignant progression [352-354]. In colorectal carcinomas, where a progression model of disease is clearly defined and well accepted, allelic imbalances are found in 90% of early adenomas, with the adjacent normal tissue being cytogenetically normal [355]. In breast cells obtained from fine needle aspirates of patients with atypical hyperplasia, 89% of cases have detectable copy number changes [356]. Gross chromosomal rearrangements can result from defects in double stranded break repair pathways, such as non-homologous end joining and homologous recombination [357]. Additionally, growth of immortalized cells in the absence of active telomerase can lead to the accumulation of chromosomal fusions and translocations [358]. Aneuploidy results primarily from failures in mitotic chromosome transmission or defects in the spindle assembly checkpoint. Within normal tissues, several protective mechanisms are in place to maintain genome stability at the chromosomal level. Mutations within genes that control DNA damage sensing and repair, and subsequent progression through the cell cycle checkpoints, yield cells that are more permissive of chromosomal instability and susceptible to neoplastic change. Germline mutations of genes within these pathways, such as BRCA1, BRCA2, ATM, BUB1B, NBS1, RAD50, and p53, are responsible for 157  hereditary cancer syndromes and highlight the role for genome integrity in maintaining stasis [359-365]. Detecting parallel driver mutations in sporadic cancers has been complicated by both the inter- and intra-tumoural heterogeneity found within breast cancer. Advances in next-generation sequencing have overcome these technical limitations and has made it possible to robustly characterize patient specific mutational landscapes. Through this approach, our lab recently identified HAUS3 as being homozygously mutated early in the evolution of a lobular breast tumour [107]. Haus3 is part of the augmin complex, the disruption of which leads to destabilization of the kinetochore and subsequent formation of multi-polar spindles; this may be essential for maintaining genome stability within the breast in some scenarios [366]. 20% of the familial risk for breast cancer can be attributed to mutations in a handful of high and moderate-penetrance breast cancer susceptibility genes. All of these genes are involved in DNA damage repair and thus act to maintain genome integrity [367, 368]. BRCA1 and BRCA2 are the most common hereditary breast cancer susceptibility genes and together account for 16% of the familial risk for breast cancer [369]. BRCA1 aids in maintaining genome integrity through its role in double-stranded break repair [370]. Patients with BRCA1 mutations display higher rates of chromosomal instability compared to those without this mutation [371]. In many tissues, the DNA damage response is constitutively activated in precursor lesions and pre-invasive tumours, preceding genome instability and malignant conversion [372, 373]. An alternate, perhaps complementary, method to generating chromosomal instability results from telomere dysfunction and the breakage-fusions-bridge cycles that result during divisions [374]. In the culture of normal mammary epithelial cells, the gradual erosion of telomere ends leads to massive chromosomal instability that eventually expunges the culture [375]. There is little consensus on the status of telomerase activity in normal or malignant breast tissue and thus its role in malignant progression is unclear [376]. Recently, arguments for the epigenetic control of genome stability in cancer have been made and stand as an attractive hypotheses when genetic changes cannot solely explain the acquisition of allelic imbalance [377, 378]. Initial work characterizing epigenetic regulation during mammary development focused around the ability of the gland to produce milk in response to lactogenic  158  hormones. Hypomethylation of the β- and γ-casein genes in the lactating mouse mammary gland is observed and correlated with an up-regulation in gene expression compared to non-lactating glands and liver tissue [154]. Similarly, in the rat mammary gland, expression of the κ-casein gene during lactation or following prolactin treatment is associated with hypomethylation of the region surrounding the gene [155]. Continued investigation into the functional differentiation of mammary cells subsequently revealed roles for chromatin conformation and histone modifications in the expression of a complement of milk proteins [156-158, 379]. DNA methylation patterns specific to mammary progenitor cells during lineage commitment have emerged as a potential mechanism regulating the associated global gene expression changes amongst these cells [159]. A selection of these methylation differences remained conserved when assessed in analogous cell types within malignant tissue. The interplay between these changes and known transcription factors that regulate self-renewal and lineage commitment remains uncharacterized, but will likely emerge as the epigenetic programmes of these cells continue to be explored. The mammary gland does not exist as a flat plane of cells, and thus 3D culture models were developed to more accurately reflect the physiology of these cells within their native environment. As with many other processes, DNA methylation and histone acetylation patterns differ dramatically between 2D and 3D culture conditions [380, 381]. When placed within an artificial extracellular matrix (such as Matrigel), mammary cells will undergo a limited number of divisions prior to the outer layer of cells establishing apico-basal polarity and the inner layer of cells apoptosing to form a hollow lumen [338]. Acinar morphogenesis is accompanied by nuclear reorganization and histone deacetylation, which ultimately increases chromatin condensation and causes a global decrease in overall gene expression [380]. Disrupting the nuclear organization or increasing histone acetylation within these acini causes depolarization and allows the cells to bypass growth arrest [382]. Additionally, cells maintained in a constant state of hypomethylation through continuous exposure to the demethylating agent 5-Aza-2'deoxycytidine are unable to establish apico-basal polarity within this assay [381]. Epigenetic regulation of transcription is thus a major player in establishing and  159  maintaining acini in 3D culture and must be considered when contemplating tissue homeostasis within the mammary gland. In breast cancer, DNA is globally hypomethylated in comparison to adjacent normal tissue [160, 161]. However, in a rather contradictory fashion, regional hypermethylation in a host of genes is also seen and is thought to arise from changes in chromatin structure related to histone modifications rather than through general demethylase activity [383]. Promoter hypermethylation explains the loss of tumour suppressor genes, known to be mutated in hereditary cancer syndromes, that are genetically unperturbed in sporadic cancers [384]. Of particular interest in breast cancer is the p16INK4a promoter locus, which is found to be hypermethylated in 31% of primary breast tumours [385]. Inactivation of this gene correlates with progressive shortening of telomeres in tumour tissue, which can ultimately lead to genome instability and malignant progression [386]. As discussed in Chapter 3, sup-populations of mammary epithelial cells in culture emerge from growth-induced senescence with silenced p16INK4a expression [203]. It remains controversial whether p16INK4a silencing is an artifact of culture, or whether it is a variant population of cells from the original tissue containing this aberrant methylation that subsequently dominates the culture [204, 205]. Regardless of the in vitro mechanism, a subpopulation of cells is found to exist within normal breast tissue with p16INK4a promoter hypermethylation [204]. These cells are more susceptible to oncogenic transformation and are proposed to be one mechanism through which progression to malignancy can occur [387]. Epigenetic regulation of the homeobox-containing transcription factors has also been implicated in progression from ductal carcinoma in situ to invasive breast cancer, with increasing promoter hypermethylation seen during progression [388]. The identification here of a model system with an epigenetic mechanism regulating chromosomal instability will aid in the further understanding of how the epigenome regulates malignant progression in breast cancer. In Chapter 3, we generated a series of seemingly syngenic clonal cell lines. However, upon closer investigation we identified a variant line, the 184-hTERT-L2 cells, that differ in comparison to the remaining cell lines in its propensity to gain chromosome 20. In addition to rapid ascendance to aneuploidy, the 184-hTERT-L2 cells display other  160  characteristics commonly attributed to malignant cells. This includes the loss of cell cycle checkpoints and aberrations in the mitotic spindle assembly, which may ultimately be permissive factors in the gain of aneuploidy [389, 390]. They also form disorganized acinar structures when placed into 3D matrigel culture, similar to what is seen when malignant cells are assayed in this manner [220]. The integration site for the 184hTERT-L2 cells differs from the remaining cell lines, which implies that their cell of origin in the primary sample from which they were generated also differs. Genetic and epigenetic heterogeneity exists within normal mammary tissue, and thus the phenotypic differences encountered between these lines either results from intrinsic differences between the originating cells or disruptive integration of the telomerase cDNA viral construct. To identify the telomerase integration site, we employed Vectorette PCR methodology. In the case of the 184-hTERT-L2 cells, integration of the viral telomerase construct occurred in a non-coding genomic region located at 8q24.3 and integration of the virus within this site did not disrupt the expression of neighboring genes. There was also no evidence of viral fusion products and thus it is unlikely that the 184-hTERT-L2 cell phenotype is a result of the integration of the telomerase virus. Several approaches exist for surveying transcribed genetic alterations within cells. These can be broadly categorized into array-based and sequencing-based methods. Array-based methods are limited by our current knowledge as only pre-defined sequences can be detected. Technical considerations of these methods include the optical quality of reagents used in the production of a specific batch of arrays, non-linear amplification of oligonucleotide probes, and the condition of the lasers and detectors employed for scanning. These factors directly influence the signal-to-noise ratio, the resolution and the reproducibility of array-based data. These issues are of particular concern in gene expression studies as they hinder the ability to accurately detect and profile low abundance transcripts [391]. Sequencing-based methods are superior to array-based approaches for absolute quantification of gene expression, detection of low level transcript expression, and discovery of novel transcripts, transcript variants, and mutations [392]. Early attempts at high-throughput analysis of differential gene expression came in the form of sequencing cDNA libraries generated from a sample of interest. A single sequencing reaction of a few hundred basepairs can be performed from  161  either direction of randomly selected clones to create libraries of 5’ and 3’ Expressed Sequence Tags (ESTs) [393]. ESTs have proven valuable for mutation studies as they represent cloned single alleles. However, they remain of limited utility as their sequences are often unverified and display high error rates, especially around ends of the sequencing reads [394, 395]. Limitations imposed by sequencing capacities and cloning biases not withstanding, EST libraries have proven useful in using transcript abundance levels to gauge temporal gene expression changes [396, 397].  Tag-based capillary sequencing  approaches such as serial analysis of gene expression and cap analysis of gene expression provide more robust transcript-frequency information [398]. However, these methods suffer constraints due to their reliance upon restriction enzyme sites during their upstream sample preparation steps. They possess limited efficacy in assessing transcribed mutations due to their short sequencing reads and low sequence coverage. Their ultimate depth of sequencing has previously been constrained by costly, low-throughput capillary sequencing [399]. The development of next generation sequencing-by-synthesis devices has eliminated the limitations imposed by capillary-based sequencing on transcriptome analysis. These devices, currently produced by 454 Life Sciences (a division of Roche), Applied Biosystems and Illumina (formerly Solexa), sequence millions of DNA fragments in parallel by measuring single nucleotide incorporation[400]. This technology is essentially an improved version of early massively parallel signature sequencing [401]. Next generation sequencing eradicates the limitations imposed on early transcriptome profiling methodologies by capillary sequencing and abolishes dependence upon restriction enzyme sites when the correct upstream sample preparation methods are selected [397, 402]. Whole transcriptome shotgun sequencing (RNA-seq) libraries were generated from paired end sequencing reads obtained from the 184-hTERT-L2 and 184-hTERT-L9 cells. Single base pair mutations and small insertions and deletions were identified using the MAQ multiple alignment algorithm. This method aligns the short sequencing reads produced by Illumina sequencing to a reference genome [403]. From the alignment, a consensus sequence is generated that is compared back to the reference genome and mismatches are identified. Mismatches were modeled in silico by replacing the variant nucleotide(s) in the corresponding reference transcript and comparing the resultant amino  162  acid sequence. Non-synonymous mutations and smaller insertions and deletions that were present in the 184-hTERT-L2 cells but not the 184-hTERT-L9 cells were verified through re-sequencing and found to be false events. The absence of single nucleotide polymorphisms (SNP) was verified by Affymetrix SNP 6.0 arrays, as was the absence of any copy number variations between the two cell lines. Thus, no genomic alterations are present in the 184-hTERT-L2 cells that can account for the rapid gainer phenotype. Exploration of epigenetic mechanisms regulating the rapid gainer phenotype began with interventions to globally perturb these processes and search for reversions in phenotype. The most striking feature of the 184-hTERT-L2 cells is the rapid gain of chromosome 20, and thus cells were epigenetically perturbed prior to extended passaging and examination of this phenotype. We evaluated the DNA demethylating agent 5-aza2'-deoxycytidine, the histone deacetylase inhibitor trichostatin A (TSA), the global histone methylation inhibitor 3-deazaneplanocin A (DZNep), and the histone lysine methyltransferase inhibitor BIX-01294 [404, 405]. Figure 6.13 shows minimal effects on the level of aneuploid cells after 184-hTERT-L2 cells were treated with these histone modifiers prior to extended passaging of the cells. Treatment with 5-aza-2'deoxycytidine led to a substantial decline in the percentage of aneuploid cells and was not related to differential toxicity of the treatment on aneuploid cells. To characterize the changes in DNA methylation contributing to this phenotype, the cells were subjected to MRE-seq and MeDIP-seq. Through comparative analysis of the methylomes of these cells, we identified CENPI as being differentially methylated and overexpessed in the 184-hTERT-L2 cells. Although further investigations are required, we currently believe that the deregulation of CENPI within these cells is a strong candidate for inducing the rapid gainer phenotype. 6.2 6.2.1  Results 184-hTERT cells have the propensity to gain chromosome 20 As described in Chapter 3, the 184-hTERT parental cells are diploid at early  passages, and polyploid for chromosome 20 in later passages. The gain of chromosome 20 in the heterogeneous parental cells is quantified in Figure 6.1A. At passage 11 post  163  telomerase infection, 93% of cells are diploid as assessed by interphase FISH. By passage 29, 95% of the cells have 3-4 copies of chromosome 20. This gain of chromosome 20 is not specifically related to the overexpression of the telomerase cDNA as 184-HMEC cells immortalized through exposure to benzo[a]pyrene were reported to predominantly display trisomy 20 (in addition to a host of other genetic aberrations) [190]. As the parental line was not initially clonally derived, chromosome 20 gain could be attributed to either the outgrowth of a variant cell population habouring this genotype, or an intrinsic cellular mechanism allowing all cells from this patient to gain this chromosome. To address this, we analyzed the propensity of the clonal lines derived from the 184-hTERT parental cells to gain chromosome 20 during passaging (see Chapter 3). Figure 6.1B-E shows that four of the clonal lines (184-hTERT-L5, -L8, -L9 and -C2) display low level gain of chromosome 20 after extended passaging. At early passage numbers (passage 8 to 11 post telomerase infection), 1.3 to 4.7% of cells within these cell lines harboured extra copies of chromosome 20. By passage 22 to 24 post infection, 15.6 to 25.2% of cells were aneuploid. The propensity for all of the clonal lines to gain chromosome 20 signifies that it is an intrinsic mechanism within the cells and is not related to selection of an aneuploid variant during culturing. The 184-hTERT-L2 clonal line gained chromosome 20 at an exceedingly rapid rate (Figure 6.1F). Similar to the other clonal lines, 3.3% of these cells were aneuploid at an early passage. However, over passaging the percentage of aneuploid cells and the extent of amplification within each individual cell increased over time. By passage 22, 92.9% of the cells were aneuploid and up to 10 copies of chromosome 20 were detected per cell. Although these cells were confirmed to be diploid at passage 11 by M-FISH (Figure 3.6), the rapid accumulation of aneuploid cells warranted further analysis into the kartyotype of these cells at a later passage. M-FISH was conducted on 184-hTERT-L2 cells at passage 22 (Figure 6.2A). 15 metaphases were analyzed and revealed a host of structural abnormalities represented at low levels (Table 6.1). In order to quantitatively assess these changes, we selected 2 chromosomes shown to be altered in the M-FISH and investigated their numerical status in the 184-hTERT-L2 and 184-hTERT-L9 cells. Chromosome 8 was selected for analysis as numerical abnormalities were present in 3  164  metaphases and trisomy 8 (distinct from aneuploidy) has been reported in 38% of invasive ductal carcinomas (n=40)[406]. Figure 6.8B shows the percentage of cells with 3 or more copies of chromosome 8 enumerated with probes targeting the centromere and region 8q24.12-q24.13 (spanning the MYC gene locus). Eleven percent of the 184hTERT-L2 cells at passage 22 have 3 or more copies of centromere 8, and 19.6% have MYC amplification. This discrepancy between the probes is not unexpected as amplification of MYC above that of chromosome 8 is reported in 9.7% of invasive breast carcinomas (n=245)[407]. Also selected for further investigation was chromosome 17; aneuploid in 13% of cases (n=175) [408]. The centromere and the HER-2 gene locus, spanning 17q11.2-q12, were probed and at passage 22, 3.9% of the 184-hTERT-L2 cells were hyperploid for centromere 17 and 12.9% displayed HER-2 amplification (Figure 6.2C). Taken together, these results confirm that other than the rapid gain of chromosome 20, the 184-hTERT-L2 cells are relatively cytogenetically normal at later passages. 6.2.2  Rapid gainers of chromosome 20 have a gene expression profile dominated  by genes involved in cell cycle regulation and DNA damage repair 184-hTERT-L2 cells differ from the other 184-hTERT clonal lines on the basis of rapid gain of chromosomal 20. Having been derived from the same patient, these lines are all theoretically genetically identical prior to the emergence of chromosome 20 hyperploidy. We sought to determine if any anomalies exist between these clonal lines that may antecede the rapid gainer phenotype. One would anticipate that if the cells were truly synonymous, their expression profiles would be similar up until gene dosage effects related to chromosome 20 gain appeared in the 184-hTERT-L2 cells. Using Affymetrix Exon Arrays and RT-QPCR, expression differences in the rapid gainer cells at passage 11, when the cells are considered diploid, were identified in comparison to the normal lines (Appendix F and Appendix G). In Figure 6.3, the expression changes that validated by RT-QPCR are depicted as a heat map along with the relative fold change for the RTQPCR. 184-hTERT-C6 and 184-hTERT-L5 cells were included as independent comparators for the RT-QPCR as they have the same integration sites as the 184-hTERTL2 and 184-hTERT-L9 cells, respectively. The expression profile has a unique  165  distribution with respect to the direction of change; 94% of genes were up-regulated in the 184-hTERT-L2 cells. The majority of these differentially expressed genes are involved in cell cycle regulation and DNA damage repair, as categorized by their Gene Ontology annotations (Figure 6.4A) [283]. The RT-QPCR validated gene changes were also analyzed using the “Core Analysis” function in Ingenuity Pathways Analysis platform (www.ingenuity.com). This interpreted the expression data in relation to interaction networks and predominant canonical pathways. Figure 6.4B depicts the top regulatory network deregulated in the 184-hTERT-L2 cells (associated with cell cycle, cellular assembly and organization, and DNA replication, recombination and repair functions). Ingenuity Pathways Analysis surveys the top associated canonical pathways within a data set and conducts a right-tailed Fisher’s test to calculate a p-value determining the probability of association between the genes in the data set and the canonical pathway. The top 5 canonical signaling pathways within this dataset (p < 0.00435) include the mitotic role of polo-like kinase, G2/M DNA damage checkpoint regulation, nicotinate metabolism, ATM signaling, and hereditary breast cancer signaling. The expression profile of the 184-hTERT-L2 cells guided us to examine the cell cycle checkpoint regulation of these cells. 6.2.3  Mitotic spindle and cell cycle checkpoint defects are present in rapid gainers  of chromosome 20 The expression profile of the 184-hTERT-L2 cells compelled us to investigate the ability of these cells to form a functional mitotic spindle assembly. Figure 6.5 shows alpha-tubulin staining of the mitotic spindles and pericentrin staining to visualize the centrosomes during metaphase and anaphase. No abnormal mitoses were seen in the normal comparator cell line, the 184-hTERT-L9 cells (n=20). In the mitotic 184-hTERTL2 cells analyzed, multiple centrosomes, abnormal spindle assembly and/or misaligned chromosomes were seen. The acquisition of additional centrosomes through overduplication or splitting leads to the formation of multipolar spindles and ultimately the missegregation of chromosomes. The resulting aneuploid cells would either be nonviable or, by chance, acquire a chromosome complement that allows for survival [409, 410]. Centrosome hypertrophy and multipolar spindles are a characteristic of breast  166  cancer, with centrosome size and number correlating with aneuploidy and chromosomal instability within tumours [411, 412]. Despite what would be anticipated from the gene expression profile of the 184hTERT-L2 cells, the cell cycle profile of this line, as assessed by BrdU incorporation, is identical to the normal comparator line (Figure 6.6A). However, intervention with treatments blocking progression through the cell cycle at various phases reveals deficiencies in the S-phase and G2-M phase checkpoint mechanisms. Serum-starvation, which arrests cells at G1, does not differentially affect the 184-hTERT-L2 cells (Figure 6.6B). However, early S-phase arrest achieved by aphidicolin-induced inhibition of DNA polymerase, or mitomycin C DNA cross-linking, is not as efficient in the 184-hTERT-L2 cells (Figure 6.6C and Figure 6.6D) [413, 414]. The topoisomerase II inhibitor, etoposide, is also not as effective at stalling the 184-hTERT-L2 cells in late S-phase/early G2 (Figure 6.6E) [415]. Treatment of cells with colchicine delays mitotic progression by activating the spindle assembly checkpoint (Figure 6.6F) [416]. As anticipated from the imaging results above, this checkpoint is not fully functional and 184-hTERT-L2 cells can escape this blockade. Cells lacking cell cycle checkpoints have been defined as more sensitive to ionizing radiation as they are not able to halt cell cycle progression to repair DNA damage inflicted by the insult [417]. The surviving fraction of 184-hTERT-L2 cells in response to increasing doses of ionizing radiation is reduced when compared to 184-hTERT-L9 and 184-hTERT-L5 cells (Figure 6.7). Additionally, complete growth arrest in the survival assay is achieved at a lower dose in the rapid gainer cells. The defects in multiple cell cycle checkpoints, combined with the presence of multipolar mitotic spindles, outlines a clear phenotypic difference in the manner through which the rapid gainer cells divide and provides a potential explanation for the accumulation of aneuploidy within these cells. 6.2.4  Rapid gainers of chromosome 20 have disrupted 3D acinar morphogenesis Within 2D culture, the rapid gainer and normal clonal lines have comparable  morphologies. However, growth on tissue culture plastic does not reflect the in vivo organization or environment that mammary cells exist within. We have shown that like mammary progenitor cells, these clonal lines are able to form acinar structures when  167  placed into 3D Matrigel culture and are able to differentiate to express markers of both lineages. However, when the 184-hTERT-L2 cells are placed within these conditions, they form disorganized acinar structures that lack proper apico-basal polarity (Figure 6.8). Apical polarity was assessed through staining with the golgi marker GM130. As seen in the 184-hTERT-L9 cells, when apical polarity is established the Golgi will align above the nuclei in an organized manner parallel to the basement membrane. In the 184hTERT-L2 cells, GM130 is found in no particular order in the outer layer of cells. Disorganization within acinar structures occurs when malignant primary tumours and breast cancer cell lines are placed in 3D culture [220]. Kenny et. al. were able to group a series of breast cancer cell lines based upon the morphology of the structures they were able to form under these conditions, which was ultimately related to their gene expression profiles [418]. These patterns, in part, correlated with the site of origin of the primary tumour with 6/7 lines forming ‘mass-like’ structures originating from a primary tumour and 8/9 lines forming ‘grape-like’ structures derived from a metastatic site. The 184hTERT-L2 cells have a 3D morphology similar to these ‘mass-like’ structures. These cells did not form tumours when embedded in collagen and placed under the kidney capsule of immunocompromised mice (Chapter 3). Together, this suggests that the 184hTERT-L2 cells are a transformed but non-tumourigenic cell line. 6.2.5  The rapid gainer phenotype is not caused by the integration of viral  telomerase The propensity to gain chromosome 20 is not specifically related to the forced expression of telomerase as cells immortalized through benzo[a]pyrene exposure predominantly display trisomy 20 (in addition to a host of other genetic aberrations) [190]. However, the rapid gain of chromosome 20 in the 184-hTERT-L2 cells could result from a non-specific effect related to the location of the telomerase integration site or a gene fusion created through displacement of the viral long-terminal repeats. We decided to test whether the viral integration site, or products of the integration site, were associated with producing the 184-hTERT-L2 phenotype. First, the viral integration sites for the 184-hTERT-L2 and 184-hTERT-L9 cells were determined through Vectorette PCR (Figure 6.9A) [419]. For this procedure,  168  vectorettes generated from synthetic oligonucleotides with a mismatched central region are ligated onto the ends of restriction enzyme digested genomic DNA. A primer specific to a known region of DNA, in this case the viral LTR, is used to PCR amplify the 1st strand of the template of interest and the vectorette. A vectorette specific primer that will only bind to the complement of the bottom strand of the mismatched vectorette is used in conjunction with the aforementioned sequence specific primer for the second and subsequent PCR rounds. Sanger sequencing of the PCR product provides the identity of the unknown region of DNA adjacent to the integrated LTR (Figure 6.9B). Both integration sites mapped to intergenic regions of the genome and were not associated with gene structures or promoter features (Figure 6.9C). More specifically, the telomerase cDNA construct for 184-hTERT-L2 integrated at 8q24.3 between RPL8 and ZNF517. Integration in the 184-hTERT-L9 cells occurred at 7q36.1 between ZNF746 and ZNF767. A survey of the 2 megabase region of DNA surrounding the telomerase integration site of the 184-hTERT-L2 cells by RT-QPCR did not reveal any dramatic expression changes relative to the 184-hTERT-L9 cells (Figure 6.10, Appendix H). The 184-hTERT-C6 and 184-hTERT-L5 cells were included as comparator lines as they have the same integration sites as the 184-hTERT-L2 and 184-hTERT-L9 cells, respectively. The 2 genes on either side of the integration site had a less than 2-fold change in relative expression compared to the 184-hTERT-L9 cells. RECQL4 was the closest up-regulated gene, but its expression is unlikely to be related to the integration site as it is located 287 Kb away. It is unlikely that any expression changes stemming from the integration of the viral telomerase are responsible for the rapid gainer phenotype. The long-terminal repeat (LTR) sequences found at either end of the retrovirus, used to stably integrate telomerase into the genome, possesses all of the requisite signals for gene expression. While the integrated provirus is not able to produce functional viral particles, it does possess the ability to reverse transcribe itself and integrate elsewhere within the genome. However, the presence of a single integration site for these clonal lines was confirmed by Southern Blotting in Chapter 3. As the viral LTR contains promoter and enhancer regions, this could result in the aberrant expression of distant genes. To eliminate this as a possibility, whole transcriptome shotgun sequencing libraries generated from the 184-hTERT-L2 cells were scoured for the presence of any  169  transcripts whose expression may be driven by the viral promoter. Appendix K lists all of the next-generation sequencing reads that are composed of any region of the pLXIN telomerase cDNA construct fused to a transcribed region of the genome. These fusions were represented 1-2 times within the sequencing results indicating a low confidence in their validity. None of these fusions validated as real events by PCR (PCR primers in Appendix L). Furthermore, amplification from sequences located 100 bases on either side of the putative fusions generated PCR products from the cDNA of a size consistent with uninterrupted genomic DNA. This indicates that the sequencing reads did not detect true viral fusion constructs and that expression driven from a viral LTR promoter in an unanticipated region of the genome is not responsible for the rapid gainer phenotype. 6.2.6  The genotype of the cell of origin does not lead to the rapid gainer  phenotype With the viral integration site ruled out as a determinant in establishing the rapid gainer phenotype, we looked at possible mechanisms related to the cell of origin. The 184-hTERT-L9 and 184-hTERT-L2 cells were characterized as different from each other based upon their integration sites. This also signifies that they had different parental cells within the heterogeneous 184-HMEC culture. Taken from this, the originating cell within the patient tissue may have harboured an inherent genetic difference between these otherwise syngenic lines that directly resulted in the rapid gainer phenotype. To assess this, we used Whole Transcriptome Shotgun Sequencing of the 184-hTERT-L2 and 184hTERT-L9 cells to identify single nucleotide variants, micro-insertions, and microdeletions that were present in the rapid gainer line and not in the normal comparator. Sequencing identified 211 single nucleotide variants and 62 micro insertions and deletions in the coding regions of expressed genes that were specific to the 184-hTERTL2 cells (Appendix I, Appendix J). Upon PCR amplification of these regions in both the 184-hTERT-L2 and 184-hTERT-L9 cells and re-sequencing by next-generation sequencing, none of these variations validated as real events. Affymetrix Genome-Wide Human SNP 6.0 Arrays hybridized with genomic DNA from the 184-hTERT-L2 and 184-hTERT-L9 cells confirmed the absence of any single nucleotide variants or copy number variants that differ between these cell lines. Representative plots depicting an  170  absence of copy number variations for chromosome 20 are shown in Figure 6.11. There is no evidence that any genomic changes specific to the 184-hTERT-L2 clonal line exists and thus the rapid gainer phenotype is not related to a genomic change within these cells. 6.2.7  Epigenetic alterations contribute to the rapid gainer phenotype The lack of genetic variation between the rapid gainer and comparator 184-  hTERT clonal lines suggests that an epigenetic mechanism could likely be responsible for generating the noted phenotypic differences. If this is the case, erasing the epigenome to allow for re-programming might alter the rapid gainer phenotype. We decided to investigate whether the rapid gainer phenotype was susceptible to changes in DNA methylation or histone de-acetylation. 5-aza-2'-deoxycytidine is a cytidine analogue that incorporates into the DNA of dividing cells and inhibits DNA methylation [420]. Silenced genes are reactivated through the reversible inhibition of DNA methyltransferase activity. This analogue occupies the active site of DNA methyltransferase, forming a high affinity bond that prevents DNA methylation at other sites by making the enzyme unavailable [421, 422]. Re-methylation occurs after the analogue is removed and it is diluted away by DNA divisions. When the clonal lines are treated for 72 hours with 5-aza-2'-deoxycytidine prior to extended passaging, the rapid gainer phenotype is suppressed (Figure 6.12A). After 11 passages, the percentage of aneuploid 184-hTERT-L2 cells decreased from 79% to 36%. This is not simply due to selective toxicity of the treatment on aneuploid cells as cells at passage 23, treated in a similar manner and passaged twice, did not show any decrease in aneuploidy (Figure 6.12B). This phenomenon was specific to methylation, as treatment with trichostatin A, an inhibitor of acetylation, does not alter the percentage of aneuploid cells (Figure 6.13). Treatment of the 184-hTERT-L2 cells with 5-aza-2'-deoxycytidine also serves to correct the acinar disorganization and loss of polarity noted in these cells when they are placed into 3D Matrigel culture. When treated at increasing dosages of 5-aza-2'deoxycytidine for 72 hours prior to plating in Matrigel, the acinar structures begin to hollow out and the outer layer of cells organize more clearly into a multi-layered ring (Figure 6.14A). Both this line and the 184-hTERT-L9 cells increased their overall size and lumen size in response to de-methylation. Figure 6.14B depicts this trend in both cell  171  lines by plotting to scale the spherical dimensions of the acinar structures and their lumens 21 days after drug treatment. The methylome of the 184-hTERT-L2 cells thus appears to regulate the chromosomal instability and the 3D architecture of these cells; properties that must remain stable within normal cells to prevent malignant progression to breast cancer. 6.2.8  MRE-sequencing reveals regions of differential methylation in the rapid  gainer cells The suppression of the rapid gainer phenotype seen after demethylation of the 184-hTERT-L2 cells prompted us to investigate early changes within the methylome of these cells. To accomplish this, we subjected the 184-hTERT-L2 cells, and 184-hTERTL5 and 184-hTERT-L9 comparator lines, to a combination of methylated DNA immunoprecipitation and sequencing (MeDIP-seq) and methyl-sensitive restriction enzyme sequencing (MRE-seq). These are two complementary approaches that use nextgeneration sequencing to detect both methylated and un-methylated DNA [423]. MeDIPseq couples immunoprecipitation of DNA with a 5-methylcytosine-specific antibody to next-generation sequencing to identify sites of DNA methylation [424]. MRE-seq uses digestion with multiple methylation sensitive enzymes to fragment DNA prior to sequencing to detect un-methylated CpGs [425]. Integration of these two methods allows for the identification of sites with mixed, or allele-specific, methylation patterns [426]. Using the MAQ algorithm, MeDIP-seq and MRE-seq reads were mapped to the human genome assembly (hg18). For the MRE-Seq, the read was only included if the 5′ end mapped to a CpG site within a methyl-sensitive restriction enzyme site. Methylation scores were generated for each CpG site and visualized on the UCSC Genome Browser RefSeq Gene track [426]. Overall, the methylation profile of the 184-hTERT-L2 cells was similar to that of the comparator lines. No differences were seen in the methylation pattern of the region surrounding the viral telomerase integration sites for these cell lines. When focusing on the genes that are differentially expressed between these cell lines (Figure 6.3), only 18 of 103 had any differential methylation within the coding or intergenic sequences and/or the regions directly 5’ or 3’ of these genes. Four of these genes have methylation pattern differences within a CpG islands located around the 5’  172  region of the gene sequence, suggestive of interference with their putative promoter regions. These four genes are UBE2C, CHTF18, ATP8B2 and CENPI. Of these, UBE2C displays mixed methylation in the 184-hTERT-L2 cells within its first intron; the DNA is un-methylated in the controls. This does not match the increase in UBE2C expression seen in these cells and is thus not likely to be a defining event resulting in the establishment of the rapid gainer phenotype. A similar discrepancy is seen with CHTF18. Interestingly, CHTF18 is required for sister chromatid cohesion, which plays an integral role in maintaining genome stability by preventing premature sister chromatid separation [427]. However, it is the absence of this gene and not its over-expression that leads to genome instability in other model systems [428, 429]. Therefore, despite the observed differences in methylation, we do not believe that CHTF18 is affecting genome stability within the 184-hTERT-L2 cells. ATP8B2 has CpG island methylation in the 184-hTERT-L2 cells but not the 184-hTERT-L5 and 184-hTERT-L9 cells; this matches the observation that the expression of ATP8B2 is decreased in the 184-hTERT-L2 cells (Figure 6.15). Finally, CENPI is unmethylated in the 184-hTERT-L2 cells and methylated in the comparator lines, which matches the observed increase in expression in the 184-hTERT-L2 cells (Figure 6.15). CENPI, along with CENPA, is a core component of the centromere that remains stable during turnover of the other centromere proteins during cell cycle progression [430]. Interestingly, overexpression of CENPA led to genome instability in HCT116 cells [431]. Additionally, CENPI is required for the proper localization of CENPF during kinetochore assembly [432]. Up-regulation of CENPF in primary breast tumours was associated with poor prognosis and chromosomal instability [433]. Because of this, we believe that the overexpression of CENPI in the 184-hTERT-L2 cells stemming from this aberrant methylation pattern is responsible for generating the rapid gainer phenotype in these cells. Further investigations whereby CENPI is silenced in the 184-hTERT-L2 cells to revert this phenotype, and overexpressed in the 184-hTERT-L9 cells to induce this phenotype, will be required to solidify this hypothesis. Additionally, similar experiments with UBE2C, CHTF18, and ATP8B2 should be performed to confirm our belief that the deregulation of these genes is not affecting the rapid gainer phenotype.  173  6.3  Discussion One of the major benefits of working with models of normal physiology is the  ability to study premalignant changes leading to a diseased state. While the generalized notion stands that prevention is more desirable than a cure, this is hard to accomplish when little is understood about disease onset. Oncogenesis of transformed cells has been well studied in breast cancer, but not necessarily through mechanisms that closely resemble native disease progression. For example, immortalized cells have been rendered tumourigenic through the overexpression of h-Ras, Raf-1, and c-Myc oncogenes [185, 434, 435]. However, activating mutations in Ras rarely occur in breast cancer (less than 5% of cases) and c-myc amplification is only present in 15.5% of cases [436, 437]. Raf-1 activation is better represented with 28.6% of patients eligible for tamoxifen treatment expressing phosphorylated Raf-1 [438]. Overall, generating models of disease progression through the activation of potent oncogenes is only an accurate reflection of a small percentage of tumours. This gap between immortalization and malignant transformation remains poorly understood and requires better model systems in order to delineate the processes regulating this progression. Our model system identifies epigenetic changes that contribute to this process and identifies the need to further explore non-genetic components in premalignant cells. We were able to identify a model system of syngenic lines with an unexplained acquisition of premalignant changes including cell cycle defects, mitotic spindle defects, acquisition of aneuploidy, and disorganized acinar growth. One of the major differences between these cells lines is their propensity to gain chromosome 20 over time. All of the 184-hTERT cell lines have an intrinsic mechanism to gain chromosome 20; only a unique variant is able to do so at a rapid rate. Chromosome 20 is commonly gained in breast cancer and thus the identification of a normal mammary epithelial cell line that selectively acquires this chromosome is immediately relevant to understanding malignant progression. When assessed by array CGH, 58 of 305 invasive breast cancer cases had gained the q-arm of chromosome 20 [439]. Gain of 20q is the fourth most common chromosomal change seen in breast cancer primary tumours and cell lines, with gain of 1q, 8q and 17q being more prevalent [440-442]. Within primary tumours, the level of chromosome 20q amplification is associated with increasing percentages of Ki67 positive  174  cells and cells in S-phase [443, 444]. In analysis of individual breast tumours with adjacent intraductal hyperplasia, atypical ductal hyperplasia, ductal carcinoma in situ, and invasive ductal carcinoma, increasing levels of 20q amplification is associated with progression through these stages of carcinogenesis [445]. Low-level amplification of 20q is the most common chromosomal change seen in ductal hyperplasia without atypia and has been proposed to be an early genetic change that initiates clonal expansion of a precursor lesions into ductal carcinoma [446]. Despite these associations, we cannot definitively explain why chromosome 20 would be preferentially gained in breast tumours or the 184-hTERT cells. We speculate that the specific gain of chromosome 20 within the 184-hTERT cells results from a non-specific generation of aneuploidy. If a mal-adaptation within these cells leads to chromosome 20 providing a growth advantage, the continued accumulation of this karyotype would result. Numerous genes involved in regulating proliferation and preventing apoptosis are found encoded on Chromosome 20 and their deregulation may ultimately be providing the advantage. Chromosomal instability itself thus becomes a selective pressure to acquire and maintain this karyotype. The rapid gain of Chromsomome 20 seen in the 184-hTERT-L2 variant may be a function of an increased rate of missegregations due to the cell cycle checkpoint defects found within these cells. Thus, the accumulation of cells retaining this phenotype occurs at an amplified rate. The mal-functioning cell cycle checkpoints within the rapid gainers is likely to be the crucial event within these cells that allows for the differential rate of chromosomal gain. As these cells are completely genetically identical and the phenotypes can be reversed by de-methylation, an epigenetic mechanism would have to be responsible for the initial accumulation of these changes. In addition to the rapid ascendance to aneuploidy, the 184-hTERT-L2 cells display other characteristics commonly attributed to malignant cells. This includes the loss of cell cycle checkpoints and aberrations in the mitotic spindle assembly, which may ultimately be permissive factors in the gain of aneuploidy [389, 390]. In culture, they exhibit dyskaryosis and failure of abscission (data not shown). They form disorganized acinar structures when placed into 3D matrigel culture, similar to what is seen when malignant cells are assayed in this manner [220]. Despite this in vitro resemblance of  175  malignant cells, the 184-hTERT-L2 cells fail to form tumours when engrafted under the kidney capsule in immunocompromised mice. These cells lie somewhere within the spectrum of immortalization and malignancy. While these cells can be defined as transformed, the true meaning of this in relation to malignant progression cannot be solidified. Therefore, these cells have limited malignant properties and we cannot extrapolate our results beyond this point until they are characterized in an alternate model system. This accumulation of premalignant changes occurred without the activation of a potent oncogene or the accumulation of genetic anomalies. The ability to reverse these effects through de-methylation and epigenetic reprogramming signifies that an initial event leading to these changes occurred within the methylome. Methylation sequencing has revealed CENPI as a likely candidate in inducing this phenotype within the 184hTERT cells. Confirmation of this association will provide the first description of a role for CENPI in controlling genome stability and in malignant progression.  176  A  1 copy  Barrett p11  B L5 p11  1 copy 2 copies  2 copies  3 copies  3 copies  Barrett p15  4 copies  L5 p14  4 copies  5 copies  5 copies Barrett p24  6 copies  6 copies  L5 p18  Barrett p29  L5 p22  gfp - late passage clonal  0% 0%  20%  40%  60%  80%  20%  40%  60%  80%  100%  100%  C  1 copy L8 p11  D  1 copy L9 p11  2 copies  2 copies  3 copies  3 copies  4 copies  L8 p14  4 copies  L9 p14  5 copies  5 copies  6 copies  6 copies  L8 p18  L9 p18  L8 p24  L9 p22  0%  20%  40%  60%  80%  100%  E  0%  1 copy C2 p8  20%  40%  60%  80%  100%  F  1 copy L2 p8  2 copies  2 copies 3 copies  3 copies 4 copies  C2 p11  5 copies  4 copies  L2 p12  5 copies 6 copies  6 copies C2 p17  7 copies  L2 p16  8 copies C2 p22 0%  G  10 copies  L2 p22 20%  40%  60%  80%  100%  0%  20%  40%  60%  80%  100%  H  Figure 6.1 – 184-hTERT cells gain extra copies of chromosome 20 during passaging A-F) Distribution of chromosome 20 levels in 5 clonal lines and the parental 184-hTERT lines over extensive passaging. Levels were enumerated by interphase FISH with a probe targeting 20q13.2 (n=150 to 152). 184-hTERT-L2 cells rapidly gain chromosome 20.  177  G) Representative image of early passage 11 parental Barrett 184-hTERT lines showing 2 copies of both the p and q arm of chromosome 20, labeled green and orange, respectively. DNA was counterstained with Dapi. H) Sample image of late passage 29 parental Barrett 184-hTERT lines showing 4 copies of both the p and q arm of chromosome 20, labeled green and orange, respectively. DNA was counterstained with Dapi.  178  A  B  c-myc  100%  C  80% 60% 40% 20% 0%  CEP17  Her2  100% Percent Aneuploid  Percent Aneuploid  CEP 8  80% 60% 40% 20% 0%  L2 p22  L9 p22  L2 p22  L9 p22  Figure 6.2 – 184-hTERT cells do no gain chromosome 8 or 17 during passaging A) Multiplex FISH of 184-hTERT-L2 cells (passage 22) reveals structural abnormalities in addition to chromosome 20 gain in some of the analyzed metaphases. B) Percentage of cells with 3 or more copies of centromere 8 and c-myc in interphase FISH at passage 22 in 184-hTERT-L9 and 184-hTERT-L2 cells. C) Percentage of cells with 3 or more copies of centromere 17 and Her2 in interphase FISH at passage 22 in 184-hTERT-L9 and 184-hTERT-L2 cells. 179  QPCR C6 vs L5  QPCR L2 vs L9  Affy L2 vs L9 11  0  -6  1711  0  -911  71  0  -207 Not Expressed  180  QPCR C6 vs L5  QPCR L2 vs L9  Affy L2 vs L9  QPCR C6 vs L5  QPCR L2 vs L9  Affy L2 vs L9  QPCR C6 vs L5  QPCR L2 vs L9  Affy L2 vs L9  Figure 6.3 – Global expression differences exist between 184-hTERT-L2 and 184hTERT-L9 cells The Affymetrix GeneChip® Human Exon 1.0 ST Array platform was utilized to initially interrogate expression differences between the 184-hTERT-L2 and 184-hTERT-L9 cell lines. Genes with a greater than 3.5 fold change in expression with a p-value of less than 0.01 were validated by probe-based reverse-transcriptase real-time PCR comparing the L2 and L9 cell lines, and the C6 and L5 cell lines. L2 and C6 lines have the same integration site, as assessed by Southern Blotting. L5 and L9 cells have the same integration site. Gradient legends with relative expression values are indicated for each comparison. Blue represents decreased expression relative to the L2 and C6 lines, red represents increased expression, white indicates no change and yellow signifies that no expression was detected.  181  A  cell cycle 8  6  6 33  cell division mitosis  10  chromosome segregation 11  protein amino acid phosphorylation 30 29  DNA repair DNA replication cell proliferation  B  182  Figure 6.4 – Biological Process Gene Ontology annotations for genes with increased expression in 184-hTERT-L2 cells A) Categorization of the Gene Ontology (as of December 20, 2010) for the genes overexpressed in 184-hTERT-L2 cells in comparison to 184-hTERT-L9 cells identified from the Affymetrix analysis and validated by RT-QPCR. B) Ingenuity Pathway Analysis showing RT-QPCR validated differentially expressed genes from the network containing the greatest percentage of genes and involved in cell cycle regulation. Red indicates genes that are up-regulated in the 184-hTERT-L2 cells, green represents genes that are down-regulated, and white indicates no change in expression relative to the 184-hTERT-L9 cells. Direct relationships are displayed with solid arrows, indirect relationships with dashed arrows.  183  _-tubulin  pericentrin  Draq5  merge  L9  L2  L2  L2  5um  Figure 6.5 – Centrosome amplification and spindle multipolarity are seen in 184hTERT-L2 cells 184-hTERT-L9 and 184-hTERT-L2 cells were grown on glass coverslips for 48 hours, fixed in methanol, and stained for alpha-tubulin and pericentrin. 184-hTERT-L2 cells in mitosis often had several centrosomes, and always displayed an abnormal pattern of alpha-tubulin. All of the 184-hTERT-L9 cells in mitosis displayed normal spindle assembly and centrosome counts (n=20).  184  184-hTERT-L9  A  B  Control  Anti-BrdU FITC  G0-G1 S G2-M  C  Serum Starved  51.4 31.6 12.7  G0-G1 S G2-M  74.1 14.5 11  4  104  3  10  10  4  10  103  10  400  600  3  2  100 0  200  400  600  800 1000  3  200  400  600  800 1000  200  400  600  G0-G1 S G2-M  10  10  2  10  102  101  101  101  800 1000  3  100 0  200  400  600  800 1000  41.4 1.13 56.3  104  3  100 0  4Gy Irrad.  9.72 4.82 83.5  10  2  100 0  G0-G1 S G2-M 104  10  2  G  Colchicine  33.1 26.6 38.9  104  10  101  800 1000  10.3 71.5 15.4  104  3  100  200  G0-G1 S G2-M  101  10  0  Etoposide  G0-G1 S G2-M  1  101 0  Mitomycin C  20.5 71.5 6.97  10  10  F  G0-G1 S G2-M  10  2  E  Aphidicolin  2  10  10  D  100 0  200  400  600  800 1000  0  200  400  600  800 1000  Propidium Iodide  184-hTERT-L2  A  B  C  Control  Serum Starved  Anti-BrdU FITC  G0-G1 S G2-M  G0-G1 S G2-M  57.4 29.2 12.3  77.8 9.97 11.8  D  E  Aphidicolin  Mitomycin C  Etoposide  G0-G1 S G2-M  G0-G1 S G2-M  G0-G1 S G2-M  35.8 50.4 11.4  18.3 58.2 19.4  F Colchicine  48.7 13 36.6  G0-G1 S G2-M  10  4  104  104  10  10  10  103  103  103  103  103  10  10  2  102  102  10  10  10  101  101  101  101  101  10  100  0  0  10  0  200  400  600  800 1000  4  2  200  400  600  800 1000  2  100  10 0  4  0  200  400  600  800 1000  100 0  200  400  600  800 1000  0  200  400  600  800 1000  G 4Gy Irrad.  49.3 5.3 42.8  G0-G1 S G2-M  56.5 1.97 40.5  4  4  10  3  103  2  10  1  101  2  100  100 0  200  400  Propidium Iodide  185  600  800 1000  0  200  400  600  800 1000  Figure 6.6 – 184-hTERT cells have a defective G2-M cell cycle checkpoint 184-hTERT lines were plated 24 hours prior to treatment with agents listed above. After an additional 24 hours post-treatment, 50 µM BrdU was applied for 30 min prior to processing for cell cycle analysis by way of flow cytometry. Representative cell cycle plots and average percentage of cells in each phase of the cycle are shown (n=5).  186  Surviving Fraction (%)  100 L9 L5 L2  80 60 40 20 0 0  1  2  3  4  5  6  X-ray Dose (Gy)  Figure 6.7 – 184-hTERT-L2 cells display increased sensitivity to irradiation Clonal lines were harvested, subjected to increasing doses of ionizing radiation, and replated at clonal densities in MEGM. The surviving fraction of cells was determined by comparing the number of resultant colonies after 7 days to a no irradiation control. Error bars denote standard error (n=3).  187  L9  20um Phallodin  B  GM130  Propidium Iodide  Merge  L2  L5  L9  10um  Figure 6.8 – 184-hTERT-L2 cells form disorganized acinar structures in 3D Matrigel culture Confocal microscopy of 21-day Matrigel cultures stained for F-actin (phallodin) and GM130 shows disorganization and a lack of apical polarity within the 184-hTERT-L2 acinar structures.  188      A  %"
    

##  %"
    	  $"
  $"
  	         B  L2 L9  Amplified Vector  Amplified Vector Integration Site  Integration Site  L2 L9  189  Figure 6.9 – Viral integrated telomerase resides at 8q24.3 in 184-hTERT-L2 cells and 7q36.1 in 184-hTERT-L9 cells A) Schematic depicting the Vectorette PCR methodology. DNA is digested with HaeII prior to the addition of vectorettes generated from synthetic oligonucleotides with a mismatched central region. A primer specific to a known region of DNA, in this case the viral LTR, is used to PCR amplify the 1st strand of the template of interest and the vectorette. A vectorette specific primer that will only bind to the complement of the bottom strand of the mismatched vectorette is used in conjunction with the aforementioned sequence specific primer for the second and subsequent PCR rounds. Sanger sequencing of the PCR product provides the identity of the unknown region of DNA adjacent to the integrated LTR B) Vectorette PCR amplification of the 184-hTERT-L2 and 184-hTERT-L9 integration sites and the alignment of the Sanger sequenced PCR product in Ensembl (assembly NCBI35). Green arrow heads indicate the direction of the integrated construct.  190  191  Figure 6.10 – Retroviral integration does not substantially interfere with normal gene regulation in 184-hTERT-L2 cells RT-QPCR was used to assess the expression of genes flanking the 184-hTERT-L2 viral integration. Expression levels (y axis) were measured as relative mRNA quantity (RQ) after normalization for the level of GAPDH in comparison to 184-hTERT-L9 cells, and plotted on a log(10) scale from the median level (1). Error bars represent the upper and lower limits of the 95% confidence interval (n=3). 184-hTERT-C6 and 184-hTERT-L5 cells were included as independent comparators for the RT-QPCR as they have the same integration sites as the 184-hTERT-L2 and 184-hTERT-L9 cells, respectively. Expression results are aligned along an Ensemble view of genomic region 8q24.3 (NCBI35) in their relative positions with the 184-hTERT-L2 and 184-hTERT-C6 viral integration site marked by the dashed green line.  192  q13.31  5 q13.33  q13.32  5  q13.33  q13.32  q13.31  4  q13.2  q13.13  q13.12  q13.11  4  q13.2  q13.13  q13.12  q13.11  3 q12  q11.23  q11.22  3  q12  q11.23  q11.22  2 q11.21  p11.1 q11.1  p11.21  p11.22  2  q11.21  p11.1 q11.1  p11.21  p11.22  1 p11.23  p12.1  1  p11.23  p12.1  0 p12.2  p12.3  0  p12.2  6  p12.3  p13  Copy number  6  p13  Copy number  184-hTERT-L2 Chromosome 20 (bp) 6  4  2  0  ï  ï  ï x 10  7  184-hTERT-L9 Chromosome 20 (bp)  6  4  2  0  ï  ï  ï  x 10  7  193  Figure 6.11 – Genetic variation is not detected between the 184-hTERT-L2 and 184hTERT-L9 cells DNA from the 184-hTERT-L2 or 184-hTERT-L9 cells grown in MEGM was hybridized to Affymetrix Genome-Wide Human SNP 6.0 Arrays. No differences in single nucleotide polymorphisms or copy number variation existed between the two cell lines. Copy number status from the log ratio intensity data was predicted through HMMK11 modeling (available from http://compbio.bccrc.ca/?page_id=401). Copy number plots for chromosome 20 are depicted here and are representative of all chromosomes.  194  A  0uM 5-AZDC  10uM 5-AZDC  90% 79%  80%  % aneuploid cells  70% 60% 50% 36%  40% 30%  22%  20%  18%  10% 0% L9p23  B  0uM 5-AZDC  L2p23 10uM 5-AZDC 86%  90%  83%  80%  % aneuploid cells  70% 60% 50% 40% 28%  30% 20%  15%  10% 0% L9p25  L2p25  Figure 6.12 – Gain of chromosome 20 in 184-hTERT-L2 cells can be reversed with 5aza-2'-deoxycytidine A) Clonal lines at passage 11 were treated with 5-aza-2'-deoxycytidine prior to extended passaging. Levels of aneuploidy were enumerated by interphase FISH with a probe targeting 20q13.2 (n=150 to 155). When treated, less of the 184-hTERT-L2 cells gain chromosome 20 over time.  195  B) Clonal lines at passage 23 were treated with 5-aza-2'-deoxycytidine and passaged twice. Interphase FISH for chromosome 20 reveals that 5-aza-2'-deoxycytidine treatment in itself is not inherently toxic to aneuploid cells (n=150 to 156).  196  Control  BIX-01294  DZNep  TSA  90%  % aneuploid cells  80% 70% 60% 50%  40%  40%  45%  41% 31%  30% 17%  20% 10%  6%  8%  14%  0% L9p20  L2p19  Figure 6.13 – Gain of chromosome 20 in 184-hTERT-L2 cells is not affected by histone modification Clonal lines at passage 11 were treated with histone modifiers for 3 days prior to extended passaging. Levels of aneuploidy were enumerated by interphase FISH with a probe targeting 20q13.2 (n=150 to 159). Treatments included the histone-lysine methyltransferase inhibitor BIX-01294, the global histone methylation inhibitor trichostatin A (TSA), and the histone deacetylase inhibitor 3-deazaneplanocin A (DZNep). Minimal changes in levels of aneuploidy were observed in the 184-hTERT-L2 cells after treatment.  197  A  0 +m  0.01 +m  0.05 +m  0.1 +m  0.5 +m  1 +m  L9  L2  800  Spheroid size/thickness (+m)  B  600  400  200  0  0  0.01  0.05  0.1  0.5  1  Spheroid size/thickness (+m)  800  600  400  200  0  0  0.01  0.05  0.1  0.5  5-Aza-2'-Deoxycytidine (+m)  198  Figure 6.14 - 5-aza-2'-deoxycytidine corrects the aberrant morphology of 184-hTERTL2 cells in 3D Matrigel culture A) Clonal lines at passage 11 were treated with 5-aza-2'-deoxycytidine prior to plating in Matrigel. Confocal microscopy of 21-day Matrigel cultures stained for F-actin (phallodin) shows that the disorganization and luminal filling within the 184-hTERT-L2 acinar structures regresses with increasing concentrations of 5-aza-2'-deoxycytidine. B) Spheroid thickness and size (plotted to scale) increase along with increasing dosing of 5aza-2'-deoxycytidine.  199  ATP8B2  CENPI  CENPI (magnified view)  200  Figure 6.15 – Differential CpG island methylation with correlated gene expression changes in the 2 regions of the 184-hTERT-L2 Methylation scores from MeDIP-Seq and MRE-Seq were generated for each CpG site and visualized on the UCSC Genome Browser RefSeq Gene track. Methylated CpG sites are depicted in red, unmethylated in blue, and mixed methylation in yellow. Overall, the methylation profile of the 184-hTERT-L2 cells was similar to that of the normal comparator lines (184-hTERT-L9 and 184-hTERT-L5). Genes with differential expression between the 184-hTERT-L2 cells and the 184-hTERT-L9 cells were manually analyzed for methylation changes in their putative promoter regions. ATP8B2 has CpG island methylation in the 184hTERT-L2 cells but not the 184-hTERT-L5 and 184-hTERT-L9 cells; this matches the observation that the expression of ATP8B2 is decreased in the 184-hTERT-L2 cells. CENPI is unmethylated in the 184-hTERT-L2 cells and methylated in the comparator lines, which matches the observed increase in expression in the 184-hTERT-L2 cells.  201  Table 6.1 - Structural and numerical karyotypic abnormalities in 184-hTERT-L2 Multiplex FISH of 184-hTERT-L2 cells (passage 22) revealed a host of low-level structural abnormalities in addition to the expected chromosome 20 gain in metaphase spreads (n=15). Chromosome 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 X Marker  Structural Numerical +del(1) [2] +del(3) [1] +del(4) x 2 [1] +del(5) [3] +del(6) [1]  +4 [1] +5 [1]  +del(8) [2]  -7 [2] +8 [2] -8[1]  +del(9) x2 [1] +del(9) [1] ?del(10) [2]  -10 x2 [1] -10 [1]  +del(11) [1] +12 [2] -13 [1] -14 [2] -16 [1] -17 [1] -18 [1] -19 [1] +20 [11] -20 [1] -21 [1] -22[1] -X [2] +mar [1] +mar [1] +mar [1]  202  7 7.1  Discussion and future directions Major contributions Breast tissue homeostasis and specifically the factors regulating growth and  differentiation of mammary epithelium are incompletely understood. This thesis addresses the regulation of mammary epithelial cell growth and differentiation through the use of a model system developed specifically for this work and through systematic, genome-wide methods for identifying the genes important in these processes. In what was initially an effort to uncover the role of fibroblasts in regulating mammary progenitor cell growth, we identified a system that allowed us to study the regulation of fibroblast-driven epithelial cell growth, the role of polarity genes in mammary development and genome stability in nontransformed cells. A greater understanding of how the mammary gland is regulated has implications for breast cancer research in that is allows us to identify the changes that lead to malignancy. There are several well-established model systems with which the study of normal human mammary epithelial cells can be conducted. The most rigorous method is the use of discarded reduction mammoplasty tissue. While this tissue can be readily available when cooperative surgeons are within close proximity of a laboratory, the inherent nature of this tissue limits its usefulness in certain settings. Some of these limitations include the inability to culture these cells for extended periods of time, stromal contamination of the epithelial cells, limited transfection efficiency for plasmids and siRNA, extreme patient-to-patient variability, and limited material per patient sample. In efforts to avoid these limitations, several cell lines have been generated from primary mammary tissue through either spontaneous or forced immortalization and/or transformation. However, none of the described lines comprise all of the qualities that are desirable when attempting to work within a model of normal physiology. In this work, we generated a collection of clonal 184-hTERT cell lines that are non-transformed, diploid, genomically stable, express markers of normal epithelium, and resemble primary epithelium in 2D and 3D culture. These lines have a closer resemblance to normal bi-potent progenitor cells that any other cell lines, are readily amenable to genetic modifications, and have been extensively profiled throughout this work. These lines are useful for studying normal mammary physiology, provide an excellent 203  comparison for in vitro breast cancer studies, and provide a model to study premalignant progression from a normal state. Whilst generating these lines, we identified a clonal variant that rapidly gains chromosome 20 and displayed mitotic spindle and cell cycle checkpoint defects. There was no evidence for a genetic event within these cells that lead to the onset of this rapid gainer phenotype. Rather, we identified a region of differential methylation in the promoter of CENPI between the rapid gainer and normal comparator lines that likely resulted in this phenotype. This is significant in that it represents a heterogeneous epigenetic change within normal cells that led to the accumulation of premalignant changes. This is a unique model in that this change was not artificially induced and was not selected for within the culture. This change merely subsisted within a population of cells comprising the parental line. Despite the availability of next-generation sequencing technologies, it would have been extremely difficult, if not impossible, to detect this change within the parental cell line or any other normal tissue specimen. This is the first known association of CENPI with genome stability in mammary epithelium and one of the few descriptions of promoter hypomethylation in mammary cells leading to premalignant changes. The stromal compartment of the mammary gland plays a major role in supporting mammary epithelial cell growth during development. Fibroblasts, which comprise the major cellular compartment of the stroma, support both the in vivo and in vitro growth of mammary progenitor cells. This relationship is clearly evident in the well-defined colony-forming assay whereby freshly isolated mammary progenitors can be quantified when they are plated alongside fibroblasts in 2D culture. Little is known about the growth regulation of these progenitor cells within this, or any other, context. This work addressed this by developing a model co-culture system with the 184-hTERT clonal cell lines. Using an RNA interference screen, we were able to identify all of the molecules that transduce the signals required for epithelial cell growth within this context. As such, we were able to assign a novel growth regulatory function to 49 genes involved in regulating fibroblast-driven mammary cell growth. These genes had surprisingly diverse previously described functions, which was not anticipated at the outset of this study. Also, the effects of these genes on growth are more potent than previously identified growth regulators in the mammary gland. Several of these genes had an effect on bi-potent and luminal progenitor cell growth, either coordinately or  204  differentially regulating these populations. This provided a solid understanding of the normal growth physiology of these cells and the most comprehensive profile of functional effects on mammary progenitor cells performed to date. Finally, this work identified a role for Celsr1 within the mammary gland, a first for any canonical planar cell polarity gene. When Celsr1 was silenced in 184-hTERT cells and primary cells prior to growth in a 3D culture system, induction of branching from the normally spherical acini was observed. Celsr1 thus plays a role in modulating branching morphogenesis within mammary epithelial cells, a process that is continually required throughout the female reproductive life cycle. In addition to this, there was a marked increase in the number of detectable bi-potent progenitor cells in the colony-forming assay when Celsr1 is down-regulated in primary mammary tissue and progenitor enriched sorted fractions. Additionally, we identified Shisa4 as coordinately regulated with Celsr1 and capable of producing similar effects on branching in 3D culture. We believe that Celsr1 and Shisa4 may be acting together to regulate branching. This has implications within both the mammary gland and other organ systems where Celsr1 is not suspected of acting through its canonical signaling pathway. This work has attributed new functions to genes with no previously known roles in the mammary gland and has discovered new pathways regulating growth, branching morphogenesis and genome stability. 7.2 7.2.1  Implications and future directions New mediators of mammary progenitor cell regulation Despite being well studied, the growth regulation of mammary progenitor cells  remains somewhat elusive due to the presence of undefined factors within the growth assays. The inability to properly replace fibroblasts with a combination of up to 15 exogenous growth factors and hormones suggests that either the growth signaling is extremely complex, or that the proper combinations of factors were not tested [239]. In Chapter 4 we identified 49 cell surface or secreted factors whose silencing has a greater effect on epithelial cell growth than the silencing of the insulin or epidermal growth factor receptors. The factor with the greatest effect on bi-potent and luminal progenitor cell growth was Gpr39, a zinc activated signaling receptor [447]. Zinc was shown to be required for cell proliferation, but  205  its role in media is largely ignored due to inconsistencies in its concentration across medias and the lack of reporting of its concentration by media manufactures [448]. This may have been an important factor that was overlooked in development of culture conditions for mammary epithelial cells. Beyond in vitro culture, the growth regulators identified here provide new potential paracrine effectors between epithelial cells and the stroma. Recombinant xenografting assays make it possible to alter these putative signal receptors in the epithelial cells alone, and/or their cognate ligands in the stromal cells alone, to directly assess their effect on progenitor cell growth and function in vivo. Systematically assessing the effect of these genes would undoubtedly lead to a better understanding of response of epithelial cells to their surrounding environment. 7.2.2  Improved progenitor cell isolation strategies From the original identification of a mouse mammary stem cell in 2006, efforts have  been directed towards refining isolation strategies for both mouse and human progenitor cells [12, 13, 15, 18, 449, 450]. When working with human cells, prospective identification of progenitor cells is reliant upon cell surface molecules or intracellular enzymes whose activity can alter fluorescence in an introduced molecule. The latter is difficult to accomplish as the molecule must be cell permeable initially, not leach out of the cells upon enzymatic cleavage, and be detectable by flow cytometry. As such, mammary progenitor cell enrichment protocols have mainly relied on antibodies to cell surface proteins. In a circular conundrum, the search for new markers to further purify these cells is hindered by the inability to obtain purer population of cells to profile for differential gene expression. Chapter 4 focused on the functional identification of genes regulating mammary progenitor cell growth. The majority of these genes coordinately regulated bi-potent and luminal progenitor growth. However, we found that SerpinH1, Nkain4 and Kcnj5 had differential effects on the growth of these populations. These surface markers are ideal candidates to assess for enrichment capability as they have a known functional effect in a specific progenitor cell population. 7.2.3  Novel signaling paradigm for planar cell polarity genes Although there is evidence emerging to the contrary, Celsr1 is predominantly thought  to act through its canonical planer cell polarity signaling pathway [342]. In this work, we  206  have identified Shisa4 as potentially mediating the effects of Celsr1 on branching morphogenesis in the mammary gland. Isoforms of Shisa, but not Shisa4, have been implicated in preventing the maturation of Wnt and FGF receptors in the mouse [345]. They bind to immature forms of frizzled and FGF receptor within the endoplasmic reticulum, effectively suppressing their trafficking to the cell membrane [344]. As opposed to its homologues, Shisa4 is predicted to reside on the plasma membrane thus proving it with the opportunity to physically interact with Celsr1. Confirmation of a functional interaction between these two molecules would identify a role for Shisa4 and provide an alternate signaling pathway through which Celsr1 may act. With their shared ability to bind isoforms of frizzled, it is possible that a common domain exists between these proteins that mediates this interaction. This would have implications outside of mammary development, as both molecules are expressed during embryogenesis. Specifically, Celsr1 and Shisa4 are both expressed in the lung, the spinal cord, the dorsal root ganglia, and the sub-ventricular region of the brain when assessed at embryonic day 12.5 and 14.5, respectively [323, 345, 451]. This interaction would be a prime candidate to investigate in the neuroepithelium whereby a role Celsr1 signaling outside of the planar cell polarity pathway has been proposed but not yet identified [342]. 7.2.4  Potential strategy for maintaining genome stability Through a process of elimination, we identified an epigenetic change in the promoter  of CENPI as the aberration that is likely responsible for the mitotic spindle defects, cell cycle checkpoint defects, and rapid ascendance to aneuploidy that is seen in the 184-hTERT-L2 cells. This is an early change within these cells that precedes the genome instability evident in the later passages. Overexpression of the centromere protein CENPF is associated with chromosomal instability and poor patient survival in breast cancer [433]. Furthermore, an inhibitor to CENPE administered to mice with xenografts of colon cancer cells displayed marked tumour regression due to apoptosis within the tumour cells [452]. This inhibitor is currently in Phase I clinical trials in patients with refractory cancer (GlaxoSmithKline). CENPI expression has not been investigated in breast cancer. These experiments suggest that promoter hypomethylation and subsequent overexpression of CENPI is an early event that would be found within premalignant tissue. Thus, a survey of CENPI expression in  207  hyperplasic tissue is warranted to confirm its applicability across patient samples. Following this, the association between CENPI overexpression and genome instability within breast tumours should be investigated. If CENPI is an early change that leads to genome instability within breast cancer, as suggested by the experiments in Chapter 6, a small molecule inhibitor to CENPI should be developed. If CENPI parallels what is seen with CENPA, inhibiting its expression will restore the proper assembly of bi-polar spindles and inhibit aberrant mitotic progression [431]. This would prevent further genomic changes from occurring within the cancer cells that may contribute to advancing malignant progression through the acquisition of de novo characteristics mediated by genome instability. Aside from providing potential treatment options, the presence of CENPI promoter hypomethylation in normal tissue may act as a prognostic factor for breast cancer risk. If this epigenetic change is able to induce tumour formation, it will provide evidence that genetic insults are not always responsible for tumour initiation and will expand our perceptions of what constitutes a tumour driver. 7.3  Concluding comments In this thesis, we have identified new regulators of mammary development and have  uncovered a novel mechanism controlling genome stability in normal cells. This work utilized new technologies to address old problems. As a function of this, the large majority of the growth regulators we identified in Chapter 4 have not been described to play a role within the mammary gland and their cognate ligands are often still to be discovered. Approaching this problem using a functional screen thus uncovered a host of factors previously unidentified to be important in fibroblast-driven epithelial cell growth when assessed by expression arrays, proteomic analysis or directed screening. The application of functional screening to other biological problems may advance our understanding of the functions of lesser-described genes in the regulation of seemingly simple physiological processes. The discovery of promoter hypomethylation as the aberration responsible for genome stability in the 184-hTERT cells would not have been possible without nextgeneration sequencing technologies. Demonstrating that a genetic aberration was not responsible for producing the rapid gainer phenotype required that we carefully profiled the genome of these cells along with a comparator line. This required high resolution and high  208  confidence that no region was left un-profiled. This was also true for the methylation sequencing, which uncovered CENPI as differentially methylated and expressed. Without these technologies, it would have been extremely difficult to identify the epigenetic change associated with the rapid gainer phenotype and impossible to rule out a contributing genetic change. This work has uncovered a series of novel genes that regulate mammary progenitor cell growth, branching morphogenesis and genome stability. Additional follow-up will be required to extend these discoveries into breast cancer. However, understanding normal physiology is the first step in understanding disease. De-regulation of growth signaling, disrupted tissue architecture and genome instability are all hallmarks of breast cancer and the advances we made in the normal regulation of these processes will undoubtedly be implicated in breast cancer progression.  209  Bibliography 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.  Hughes, E.S., The Development of the Mammary Gland: Arris and Gale Lecture, delivered at the Royal College of Surgeons of England on 25th October, 1949. Ann R Coll Surg Engl, 1950. 6(2): p. 99-119. Howard, B.A. and B.A. Gusterson, Human breast development. J Mammary Gland Biol Neoplasia, 2000. 5(2): p. 119-37. Fritsch…, H., Color atlas of human anatomy: Internal organs. 2008. Russo, J. and I.H. Russo, Development of the human breast. Maturitas, 2004. 49(1): p. 2-15. Russo, J., R. Rivera, and I.H. Russo, Influence of age and parity on the development of the human breast. Breast Cancer Res Treat, 1992. 23(3): p. 211-8. Eyden, B.P., et al., Intralobular stromal fibroblasts in the resting human mammary gland: ultrastructural properties and intercellular relationships. J Submicrosc Cytol, 1986. 18(2): p. 397-408. Emerman, J.T. and A.W. Vogl, Cell size and shape changes in the myoepithelium of the mammary gland during differentiation. Anat Rec, 1986. 216(3): p. 405-15. DEOME, K.B., et al., Development of mammary tumors from hyperplastic alveolar nodules transplanted into gland-free mammary fat pads of female C3H mice. Cancer research, 1959. 19(5): p. 515-20. Daniel, C.W. and K.B. DEOME, GROWTH OF MOUSE MAMMARY GLANDS IN VIVO AFTER MONOLAYER CULTURE. Science, 1965. 149: p. 634-6. Smith, G.H. and D. Medina, A morphologically distinct candidate for an epithelial stem cell in mouse mammary gland. J Cell Sci, 1988. 90 ( Pt 1): p. 173-83. Young, L.J., et al., The influence of host and tissue age on life span and growth rate of serially transplanted mouse mammary gland. Exp Gerontol, 1971. 6(1): p. 49-56. Shackleton, M., et al., Generation of a functional mammary gland from a single stem cell. Nature, 2006. 439(7072): p. 84-88. Stingl, J., et al., Purification and unique properties of mammary epithelial stem cells. Nature, 2006: p. 1-5. Loeffler, M. and I. Roeder, Tissue stem cells: definition, plasticity, heterogeneity, self-organization and models--a conceptual approach. Cells Tissues Organs (Print), 2002. 171(1): p. 8-26. Asselin-Labat, M.-L., et al., Gata-3 is an essential regulator of mammary-gland morphogenesis and luminal-cell differentiation. Nat Cell Biol, 2007. 9(2): p. 201-209. Sheffield, L.G. and C.W. Welsch, Transplantation of human breast epithelia to mammary-gland-free fat-pads of athymic nude mice: influence of mammotrophic hormones on growth of breast epithelia. Int J Cancer, 1988. 41(5): p. 713-9. Kuperwasser, C., et al., Reconstruction of functionally normal and malignant human breast tissues in mice. Proc Natl Acad Sci USA, 2004. 101(14): p. 4966-71. Eirew, P., et al., A method for quantifying normal human mammary epithelial stem cells with in vivo regenerative ability. Nat Med, 2008. 14(12): p. 1384-9. Loeffler, M. and I. Roeder, Tissue stem cells: definition, plasticity, heterogeneity, self-organization and models--a conceptual approach. Cells Tissues Organs, 2002. 171(1): p. 8-26.  210  20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37.  Pluznik, D.H. and L. Sachs, The cloning of normal "mast" cells in tissue culture. J Cell Physiol, 1965. 66(3): p. 319-24. Bradley, T.R. and D. Metcalf, The growth of mouse bone marrow cells in vitro. Aust J Exp Biol Med Sci, 1966. 44(3): p. 287-99. Senn, J.S., E.A. McCulloch, and J.E. Till, Comparison of colony-forming ability of normal and leukaemic human marrow in cell culture. Lancet, 1967. 2(7516): p. 5978. Robinson, W.A., J.E. Kurnick, and B.L. Pike, Colony growth of human leukemic peripheral blood cells in vitro. Blood, 1971. 38(4): p. 500-8. Taylor-Papadimitriou, J., P. Purkis, and I.S. Fentiman, Cholera toxin and analogues of cyclic AMP stimulate the growth of cultured human mammary epithelial cells. J Cell Physiol, 1980. 102(3): p. 317-21. Smith, H.S., et al., Clonal proliferation of cultured nonmalignant and malignant human breast epithelia. Cancer Res, 1981. 41(11 Pt 1): p. 4637-43. Stoker, M., M. Perryman, and R. Eeles, Clonal analysis of morphological phenotype in cultured mammary epithelial cells from human milk. Proc R Soc Lond, B, Biol Sci, 1982. 215(1199): p. 231-40. Edwards, P.A., I.M. Brooks, and P. Monaghan, Expression of epithelial antigens in primary cultures of normal human breast analysed with monoclonal antibodies. Differentiation, 1984. 25(3): p. 247-58. Edwards, P.A., et al., Clonal analysis of expression of epithelial antigens in cultures of normal human breast. J Cell Sci, 1986. 80: p. 91-101. O'Hare, M.J., et al., Characterization in vitro of luminal and myoepithelial cells isolated from the human mammary gland by cell sorting. Differentiation, 1991. 46(3): p. 209-21. Stingl, J., et al., Phenotypic and functional characterization in vitro of a multipotent epithelial cell present in the normal adult human breast. Differentiation, 1998. 63(4): p. 201-13. Stingl, J., et al., Characterization of bipotent mammary epithelial progenitor cells in normal adult human breast tissue. Breast Cancer Res Treat, 2001. 67(2): p. 93-109. Lim, E., et al., Aberrant luminal progenitors as the candidate target population for basal tumor development in BRCA1 mutation carriers. Nat Med, 2009. 15(8): p. 907913. Wildt, L., G. MARSHALL, and E. Knobil, Experimental induction of puberty in the infantile female rhesus monkey. Science, 1980. 207(4437): p. 1373-5. Fechner, A., S. Fong, and P. McGovern, A review of Kallmann syndrome: genetics, pathophysiology, and clinical management. Obstet Gynecol Surv, 2008. 63(3): p. 189-94. Gajdos, Z.K., J.N. Hirschhorn, and M.R. Palmert, What controls the timing of puberty? An update on progress from genetic investigation. Current Opinion in Endocrinology, Diabetes and Obesity, 2009. 16(1): p. 16-24. Seminara, S.B., et al., The GPR54 gene as a regulator of puberty. N Engl J Med, 2003. 349(17): p. 1614-27. Kallman, F.J. and W.A. Schonfeld, Psychiatric Problems in the Treatment of Eunuchoidism. American Journal of Mental Deficiency, 1943. 47(4): p. 386-395.  211  38. 39. 40. 41. 42. 43. 44.  45. 46.  47. 48. 49. 50. 51. 52. 53. 54. 55.  McCann, S.M., et al., Hypothalamic control of gonadotrophin and prolactin secretion, implications for fertility control. J Reprod Fertil Suppl, 1973. 20(0): p. 4359. Evans, W.S., et al., Contemporary aspects of discrete peak-detection algorithms. II. The paradigm of the luteinizing hormone pulse signal in women. Endocr Rev, 1992. 13(1): p. 81-104. Zhang, K., et al., Onset of Ovulation after Menarche in Girls: A Longitudinal Study. Journal of Clinical Endocrinology & Metabolism, 2008. 93(4): p. 1186-1194. Cunha, G.R., et al., Elucidation of a role for stromal steroid hormone receptors in mammary gland growth and development using tissue recombinants. Journal of mammary gland biology and neoplasia, 1997. 2(4): p. 393-402. Mueller, S.O., et al., Mammary gland development in adult mice requires epithelial and stromal estrogen receptor alpha. Endocrinology, 2002. 143(6): p. 2357-65. Clarke, R.B., et al., Dissociation between steroid receptor expression and cell proliferation in the human breast. Cancer research, 1997. 57(22): p. 4987-91. Clarke, R.B., A. Howell, and E. Anderson, Estrogen sensitivity of normal human breast tissue in vivo and implanted into athymic nude mice: analysis of the relationship between estrogen-induced proliferation and progesterone receptor expression. Breast Cancer Res Treat, 1997. 45(2): p. 121-33. Lydon, J.P., et al., Mice lacking progesterone receptor exhibit pleiotropic reproductive abnormalities. Genes & Development, 1995. 9(18): p. 2266-78. Hofseth, L.J., et al., Hormone replacement therapy with estrogen or estrogen plus medroxyprogesterone acetate is associated with increased epithelial proliferation in the normal postmenopausal breast. Journal of Clinical Endocrinology & Metabolism, 1999. 84(12): p. 4559-65. Chlebowski, R.T., et al., Estrogen plus progestin and breast cancer incidence and mortality in postmenopausal women. JAMA, 2010. 304(15): p. 1684-92. Asselin-Labat, M.L., et al., Control of mammary stem cell function by steroid hormone signalling. Nature, 2010. Joshi, P.A., et al., Progesterone induces adult mammary stem cell expansion. Nature, 2010. 465(7299): p. 803-807. Badders, N.M., et al., The Wnt Receptor, Lrp5, Is Expressed by Mouse Mammary Stem Cells and Is Required to Maintain the Basal Lineage. PLoS ONE, 2009. 4(8): p. e6594. Gusterson, B.A., et al., Distribution of myoepithelial cells and basement membrane proteins in the normal breast and in benign and malignant breast diseases. Cancer Res, 1982. 42(11): p. 4763-70. Pitelka, D.R., et al., Cell contacts in the mouse mammary gland. I. Normal gland in postnatal development and the secretory cycle. J Cell Biol, 1973. 56(3): p. 797-818. Shin, K., V.C. Fogg, and B. Margolis, Tight junctions and cell polarity. Annu Rev Cell Dev Biol, 2006. 22: p. 207-35. Gumbiner, B.M., Regulation of cadherin-mediated adhesion in morphogenesis. Nat Rev Mol Cell Biol, 2005. 6(8): p. 622-34. Lakhani, S.R. and M.J. O'Hare, The mammary myoepithelial cell--Cinderella or ugly sister? Breast Cancer Res, 2001. 3(1): p. 1-4.  212  56. 57. 58. 59. 60.  61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74.  Williams, J.M. and C.W. Daniel, Mammary ductal elongation: differentiation of myoepithelium and basal lamina during branching morphogenesis. Dev Biol, 1983. 97(2): p. 274-90. Mailleux, A.A., et al., BIM regulates apoptosis during mammary ductal morphogenesis, and its absence reveals alternative cell death mechanisms. Dev Cell, 2007. 12(2): p. 221-34. Gudjonsson, T., et al., Normal and tumor-derived myoepithelial cells differ in their ability to interact with luminal breast epithelial cells for polarity and basement membrane deposition. J Cell Sci, 2002. 115(Pt 1): p. 39-50. Slade, M.J., et al., The human mammary gland basement membrane is integral to the polarity of luminal epithelial cells. Exp Cell Res, 1999. 247(1): p. 267-78. Streuli, C.H., N. Bailey, and M.J. Bissell, Control of mammary epithelial differentiation: basement membrane induces tissue-specific gene expression in the absence of cell-cell interaction and morphological polarity. J Cell Biol, 1991. 115(5): p. 1383-95. Streuli, C.H., et al., Laminin mediates tissue-specific gene expression in mammary epithelia. J Cell Biol, 1995. 129(3): p. 591-603. Harris, T.J. and M. Peifer, The positioning and segregation of apical cues during epithelial polarity establishment in Drosophila. J Cell Biol, 2005. 170(5): p. 813-23. Harris, T.J. and M. Peifer, Adherens junction-dependent and -independent steps in the establishment of epithelial cell polarity in Drosophila. J Cell Biol, 2004. 167(1): p. 135-47. McGill, M.A., R.F. McKinley, and T.J. Harris, Independent cadherin-catenin and Bazooka clusters interact to assemble adherens junctions. J Cell Biol, 2009. 185(5): p. 787-96. Morais-de-Sa, E., V. Mirouse, and D. St Johnston, aPKC phosphorylation of Bazooka defines the apical/lateral border in Drosophila epithelial cells. Cell, 2010. 141(3): p. 509-23. Bilder, D., M. Schober, and N. Perrimon, Integrated activity of PDZ protein complexes regulates epithelial polarity. Nat Cell Biol, 2003. 5(1): p. 53-8. Tanentzapf, G. and U. Tepass, Interactions between the crumbs, lethal giant larvae and bazooka pathways in epithelial polarization. Nat Cell Biol, 2003. 5(1): p. 46-52. McCaffrey, L.M. and I.G. Macara, The Par3/aPKC interaction is essential for end bud remodeling and progenitor differentiation during mammary gland morphogenesis. Genes Dev, 2009. 23(12): p. 1450-60. Zhan, L., et al., Deregulation of scribble promotes mammary tumorigenesis and reveals a role for cell polarity in carcinoma. Cell, 2008. 135(5): p. 865-78. Luff, A.P., Collective Investigation into the Incidence of Cancer of the Breast, and Its History After Treatment. Br Med J, 1932. 1(3723): p. 897-903. Hernaman-Johnson, F., Radiation treatment of breast cancer. British Medical Journal, 1927. 1927: p. 957-958. Brookes, P., The early history of the biological alkylating agents, 1918-1968. Mutat Res, 1990. 233(1-2): p. 3-14. Adair, F.E. and H.J. Bagg, Experimental and Clinical Studies on the Treatment of Cancer by Dichlorethylsulphide (Mustard Gas. Ann Surg, 1931. 93(1): p. 190-9. Papac, R.J., Origins of cancer therapy. Yale J Biol Med, 2001. 74(6): p. 391-8.  213  75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89. 90. 91. 92. 93. 94.  Frei, E., 3rd, et al., A comparative study of two regimens of combination chemotherapy in acute leukemia. Blood, 1958. 13(12): p. 1126-48. Love, R.R. and J. Philips, Oophorectomy for breast cancer: history revisited. J Natl Cancer Inst, 2002. 94(19): p. 1433-4. Boyd, S., On Oophorectomy in the Treatment of Cancer. Br Med J, 1897. 2(1918): p. 890-6. Boyd, S., Remarks on Oophorectomy in the Treatment of Cancer of the Breast. Br Med J, 1899. 1(1988): p. 257-62. Butlin, H.T., A Clinical Lecture on Oophorectomy in yhe Treatment of Cancer of the Breast: Delivered at St. Bartholomew's Hospital, Wednesday, November 20th. 1901. Br Med J, 1902. 1(2140): p. 10-3. Luft, R. and H. Olivecrona, Experiences with hypophysectomy in man. J Neurosurg, 1953. 10(3): p. 301-16. Luft, R. and H. Olivecrona, [Hypophysectomy in man with carcinoma of the breast and prostate and malignant diabetes mellitus]. Wien Z Inn Med, 1955. 36(2): p. 4958. Jessiman, A.G. and F.D. Moore, Advances in the treatment of cancer. Practitioner, 1958. 181(1084): p. 461-7. Bulbrook, R.D., et al., Hypophysectomy in breast cancer; an attempt to correlate clinical results with oestrogen production. Br Med J, 1958. 2(5087): p. 15-8. Ward, H.W., Anti-oestrogen therapy for breast cancer: a trial of tamoxifen at two dose levels. Br Med J, 1973. 1(5844): p. 13-4. Lippman, M.E. and G. Bolan, Oestrogen-responsive human breast cancer in long term tissue culture. Nature, 1975. 256(5518): p. 592-3. McGuire, W.L., Estrogen receptors in human breast cancer. J Clin Invest, 1973. 52(1): p. 73-7. Spaeren, U., et al., Content of estrogen receptors in human breast cancer. Eur J Cancer, 1973. 9(5): p. 353-7. Freifeld, M.L., P.D. Feil, and C.W. Bardin, The in vivo regulation of the progesterone "receptor" in guinea pig uterus: dependence on estrogen and progesterone. Steroids, 1974. 23(1): p. 93-103. Horwitz, K.B. and W.L. McGuire, Predicting response to endocrine therapy in human breast cancer: a hypothesis. Science, 1975. 189(4204): p. 726-7. Jensen, E.V., et al., Receptors reconsidered: a 20-year perspective. Recent Prog Horm Res, 1982. 38: p. 1-40. Chumsri, S., et al., Aromatase, aromatase inhibitors, and breast cancer. J Steroid Biochem Mol Biol, 2011. Slamon, D.J., et al., Human breast cancer: correlation of relapse and survival with amplification of the HER-2/neu oncogene. Science, 1987. 235(4785): p. 177-82. Hudziak, R.M., et al., p185HER2 monoclonal antibody has antiproliferative effects in vitro and sensitizes human breast tumor cells to tumor necrosis factor. Mol Cell Biol, 1989. 9(3): p. 1165-72. Carter, P., et al., Humanization of an anti-p185HER2 antibody for human cancer therapy. Proc Natl Acad Sci U S A, 1992. 89(10): p. 4285-9.  214  95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115.  Lewis, G.D., et al., Differential responses of human tumor cell lines to antip185HER2 monoclonal antibodies. Cancer Immunol Immunother, 1993. 37(4): p. 255-63. Hudis, C.A., Trastuzumab--mechanism of action and use in clinical practice. N Engl J Med, 2007. 357(1): p. 39-51. Perou, C.M., et al., Molecular portraits of human breast tumours. Nature, 2000. 406(6797): p. 747-52. Sorlie, T., et al., Gene expression patterns of breast carcinomas distinguish tumor subclasses with clinical implications. Proc Natl Acad Sci U S A, 2001. 98(19): p. 10869-74. Sorlie, T., et al., Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc Natl Acad Sci U S A, 2003. 100(14): p. 8418-23. Cheang, M.C.U., et al., Basal-Like Breast Cancer Defined by Five Biomarkers Has Superior Prognostic Value than Triple-Negative Phenotype. Clinical Cancer Research, 2008. 14(5): p. 1368-1376. Cheang, M.C., et al., Ki67 index, HER2 status, and prognosis of patients with luminal B breast cancer. J Natl Cancer Inst, 2009. 101(10): p. 736-50. Zhou, W., et al., Long-term survival of women with basal-like ductal carcinoma in situ of the breast: a population-based cohort study. BMC Cancer, 2010. 10: p. 653. Hudis, C.A. and L. Gianni, Triple-negative breast cancer: an unmet medical need. Oncologist, 2011. 16 Suppl 1: p. 1-11. Holland, D.G., et al., ZNF703 is a common Luminal B breast cancer oncogene that differentially regulates luminal and basal progenitors in human mammary epithelium. EMBO Mol Med, 2011. Sircoulomb, F., et al., ZNF703 gene amplification at 8p12 specifies luminal B breast cancer. EMBO Mol Med, 2011. 3(3): p. 153-66. Shipitsin, M., et al., Molecular definition of breast tumor heterogeneity. Cancer Cell, 2007. 11(3): p. 259-73. Shah, S.P., et al., Mutational evolution in a lobular breast tumour profiled at single nucleotide resolution. Nature, 2009. 461(7265): p. 809-813. Noguchi, S., et al., Clonal analysis of human breast cancer by means of the polymerase chain reaction. Cancer research, 1992. 52(23): p. 6594-7. Going, J.J., H.M. Abd El-Monem, and J.A. Craft, Clonal origins of human breast cancer. J. Pathol., 2001. 194(4): p. 406-412. Campbell, L.L. and K. Polyak, Breast tumor heterogeneity: cancer stem cells or clonal evolution? Cell Cycle, 2007. 6(19): p. 2332-8. Stingl, J. and C. Caldas, Molecular heterogeneity of breast carcinomas and the cancer stem cell hypothesis. Nat Rev Cancer, 2007. 7(10): p. 791-9. Nowell, P.C., The clonal evolution of tumor cell populations. Science, 1976. 194(4260): p. 23-8. Hanahan, D. and R.A. Weinberg, The hallmarks of cancer. Cell, 2000. 100(1): p. 5770. Adams, J.M. and A. Strasser, Is Tumor Growth Sustained by Rare Cancer Stem Cells or Dominant Clones? Cancer research, 2008. 68(11): p. 4018-4021. Visvader, J.E., Cells of origin in cancer. Nature, 2011. 469(7330): p. 314-322.  215  116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136.  Al-Hajj, M., et al., Prospective identification of tumorigenic breast cancer cells. Proc Natl Acad Sci USA, 2003. 100(7): p. 3983-8. Prince, M.E., et al., Identification of a subpopulation of cells with cancer stem cell properties in head and neck squamous cell carcinoma. Proc Natl Acad Sci U S A, 2007. 104(3): p. 973-8. O'Brien, C.A., et al., A human colon cancer cell capable of initiating tumour growth in immunodeficient mice. Nature, 2007. 445(7123): p. 106-10. Li, C., et al., Identification of pancreatic cancer stem cells. Cancer research, 2007. 67(3): p. 1030-7. Singh, S.K., et al., Identification of human brain tumour initiating cells. Nature, 2004. 432(7015): p. 396-401. Kelly, P.N., et al., Tumor growth need not be driven by rare cancer stem cells. Science, 2007. 317(5836): p. 337. Quintana, E., et al., Efficient tumour formation by single human melanoma cells. Nature, 2008. 456(7222): p. 593-8. To, K., et al., Y-Box Binding Protein-1 Induces the Expression of CD44 and CD49f Leading to Enhanced Self-Renewal, Mammosphere Growth, and Drug Resistance. Cancer research, 2010. 70(7): p. 2840-2851. Zhang, M., et al., Identification of tumor-initiating cells in a p53-null mouse model of breast cancer. Cancer research, 2008. 68(12): p. 4674-82. Tlsty, T.D. and L.M. Coussens, Tumor stroma and regulation of cancer development. Annu. Rev. Pathol. Mech. Dis., 2006. 1(1): p. 119-150. Molyneux, G., et al., BRCA1 basal-like breast cancers originate from luminal epithelial progenitors and not from basal stem cells. Cell Stem Cell. 7(3): p. 403-17. Bouras, T., et al., Notch signaling regulates mammary stem cell function and luminal cell-fate commitment. Cell Stem Cell, 2008. 3(4): p. 429-41. Pike, M.C., et al., Estrogens, progestogens, normal breast cell proliferation, and breast cancer risk. Epidemiol Rev, 1993. 15(1): p. 17-35. Lambe, M., et al., Transient increase in the risk of breast cancer after giving birth. N Engl J Med, 1994. 331(1): p. 5-9. Berger, S.L., et al., An operational definition of epigenetics. Genes & Development, 2009. 23(7): p. 781-783. Sharma, G., et al., CpG hypomethylation of MDR1 gene in tumor and serum of invasive ductal breast carcinoma patients. Clinical Biochemistry, 2010. 43(4-5): p. 373-379. Holliday, R., The inheritance of epigenetic defects. Science, 1987. 238(4824): p. 16370. Ahuja, N., et al., Association between CpG island methylation and microsatellite instability in colorectal cancer. Cancer research, 1997. 57(16): p. 3370-4. Ji, Y., et al., The In Vivo Pattern of Binding of RAG1 and RAG2 to Antigen Receptor Loci. Cell, 2010. 141(3): p. 419-431. Wossidlo, M., et al., Dynamic link of DNA demethylation, DNA strand breaks and repair in mouse zygotes. The EMBO Journal, 2010. 29(11): p. 1877-1888. Okano, M., S. Xie, and E. Li, Cloning and characterization of a family of novel mammalian DNA (cytosine-5) methyltransferases. Nat Genet, 1998. 19(3): p. 219-20.  216  137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. 151. 152. 153. 154. 155.  Lei, H., et al., De novo DNA cytosine methyltransferase activities in mouse embryonic stem cells. Development, 1996. 122(10): p. 3195-205. Bird, A., et al., A fraction of the mouse genome that is derived from islands of nonmethylated, CpG-rich DNA. Cell, 1985. 40(1): p. 91-9. Saxonov, S., P. Berg, and D.L. Brutlag, A genome-wide analysis of CpG dinucleotides in the human genome distinguishes two distinct classes of promoters. Proc Natl Acad Sci USA, 2006. 103(5): p. 1412-7. Laurent, L., et al., Dynamic changes in the human methylome during differentiation. Genome Research, 2010. 20(3): p. 320-331. Greger, V., et al., Epigenetic changes may contribute to the formation and spontaneous regression of retinoblastoma. Hum Genet, 1989. 83(2): p. 155-8. Sjoblom, T., et al., The Consensus Coding Sequences of Human Breast and Colorectal Cancers. Science, 2006. 314(5797): p. 268-274. Esteller, M., Cancer epigenomics: DNA methylomes and histone-modification maps. Nat Rev Genet, 2007. 8(4): p. 286-298. Christman, J.K., 5-Azacytidine and 5-aza-2'-deoxycytidine as inhibitors of DNA methylation: mechanistic studies and their implications for cancer therapy. Oncogene, 2002. 21(35): p. 5483-95. Stathis, A., et al., Phase I study of decitabine in combination with vorinostat in patients with advanced solid tumors and non-Hodgkin's lymphomas. Clinical Cancer Research, 2011: p. 1-27. Allfrey, V.G., R. Faulkner, and A.E. Mirsky, Acetylation And Methylation Of Histones And Their Possible Role In The Regulation Of Rna Synthesis. Proc Natl Acad Sci USA, 1964. 51: p. 786-94. Anderson, J.D., P.T. Lowary, and J. Widom, Effects of histone acetylation on the equilibrium accessibility of nucleosomal DNA target sites1. Journal of Molecular Biology, 2001. 307(4): p. 977-985. Agalioti, T., G. Chen, and D. Thanos, Deciphering the transcriptional histone acetylation code for a human gene. Cell, 2002. 111(3): p. 381-92. Bannister, A.J. and T. Kouzarides, Reversing histone methylation. Nature, 2005. 436(7054): p. 1103-1106. Kouzarides, T., Chromatin modifications and their function. Cell, 2007. 128(4): p. 693-705. Strahl, B.D. and C.D. Allis, The language of covalent histone modifications. Nature, 2000. 403(6765): p. 41-5. Sanders, S.L., et al., Methylation of histone H4 lysine 20 controls recruitment of Crb2 to sites of DNA damage. Cell, 2004. 119(5): p. 603-14. Dunleavy, E., A. Pidoux, and R. Allshire, Centromeric chromatin makes its mark. Trends Biochem Sci, 2005. 30(4): p. 172-5. Johnson, M.L., et al., Tissue- and cell-specific casein gene expression. II. Relationship to site-specific DNA methylation. J Biol Chem, 1983. 258(17): p. 1080511. Thompson, M.D. and H.L. Nakhasi, Methylation and expression of rat kappa-casein gene in normal and neoplastic rat mammary gland. Cancer research, 1985. 45(3): p. 1291-5.  217  156. 157. 158.  159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174.  Li, S. and J.M. Rosen, Glucocorticoid regulation of rat whey acidic protein gene expression involves hormone-induced alterations of chromatin structure in the distal promoter region. Mol Endocrinol, 1994. 8(10): p. 1328-35. Whitelaw, C.B., et al., Position-independent expression of the ovine betalactoglobulin gene in transgenic mice. Biochem J, 1992. 286 ( Pt 1): p. 31-9. Kabotyanski, E.B., Integration of Prolactin and Glucocorticoid Signaling at the Casein Promoter and Enhancer by Ordered Recruitment of Specific Transcription Factors and Chromatin Modifiers. Molecular Endocrinology, 2006. 20(10): p. 23552368. Bloushtain-Qimron, N., et al., Cell type-specific DNA methylation patterns in the human breast. Proc Natl Acad Sci USA, 2008. 105(37): p. 14076-81. Bernardino, J., et al., DNA hypomethylation in breast cancer: an independent parameter of tumor progression? Cancer Genet Cytogenet, 1997. 97(2): p. 83-9. Soares, J., et al., Global DNA hypomethylation in breast carcinoma: correlation with prognostic factors and tumor progression. Cancer, 1999. 85(1): p. 112-8. Alves, G., A. Tatro, and T. Fanning, Differential methylation of human LINE-1 retrotransposons in malignant cells. Gene, 1996. 176(1-2): p. 39-44. Brooks, J., P. Cairns, and A. Zeleniuch-Jacquotte, Promoter methylation and the detection of breast cancer. Cancer Causes Control, 2009. 20(9): p. 1539-1550. Elsheikh, S.E., et al., Global Histone Modifications in Breast Cancer Correlate with Tumor Phenotypes, Prognostic Factors, and Patient Outcome. Cancer research, 2009. 69(9): p. 3802-3809. Parmar, H. and G.R. Cunha, Epithelial-stromal interactions in the mouse and human mammary gland in vivo. Endocr Relat Cancer, 2004. 11(3): p. 437-58. Visvader, J.E., Keeping abreast of the mammary epithelial hierarchy and breast tumorigenesis. Genes & Development, 2009. 23(22): p. 2563-77. Horikawa, I., et al., Downstream E-box-mediated regulation of the human telomerase reverse transcriptase (hTERT) gene transcription: evidence for an endogenous mechanism of transcriptional repression. Mol Biol Cell, 2002. 13(8): p. 2585-97. Stampfer, M.R., Isolation and growth of human mammary epithelial cells. Methods in Cell Science, 1985. Annab, L.A., et al., Establishment and characterization of a breast cell strain containing a BRCA1 185delAG mutation. Gynecol Oncol, 2000. 77(1): p. 121-8. Debnath, J., S.K. Muthuswamy, and J.S. Brugge, Morphogenesis and oncogenesis of MCF-10A mammary epithelial acini grown in three-dimensional basement membrane cultures. Methods, 2003. 30(3): p. 256-68. Stingl, J., J.T. Emerman, and C.J. Eaves, Enzymatic dissociation and culture of normal human mammary tissue to detect progenitor activity. Methods Mol Biol, 2005. 290: p. 249-63. Richards, J., et al., Method for culturing mammary epithelial cells in a rat tail collagen gel matrix. Methods in Cell Science, 1983. 8(1): p. 31-36. Ishikawa, F., et al., Development of functional human blood and immune systems in NOD/SCID/IL2 receptor {gamma} chainnull mice. Blood, 2005. 106(5): p. 1565. Ravni, A., et al., Planar cell polarity cadherin Celsr1 regulates skin hair patterning in the mouse. J Invest Dermatol, 2009. 129(10): p. 2507-9.  218  175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186.  187. 188. 189. 190.  Nagy, A., Cre recombinase: the universal reagent for genome tailoring. Genesis, 2000. 26(2): p. 99-109. Hammond, S.L., R.G. Ham, and M.R. Stampfer, Serum-free growth of human mammary epithelial cells: rapid clonal growth in defined medium and extended serial passage with pituitary extract. Proc Natl Acad Sci USA, 1984. 81(17): p. 5435-9. Stampfer, M., Isolation and growth of human mammary epithelial cells. Methods in Cell Science, 1985. 9(2): p. 107-115. Foster, S.A. and D.A. Galloway, Human papillomavirus type 16 E7 alleviates a proliferation block in early passage human mammary epithelial cells. Oncogene, 1996. 12(8): p. 1773-9. Yaswen, P. and M.R. Stampfer, Molecular changes accompanying senescence and immortalization of cultured human mammary epithelial cells. Int J Biochem Cell Biol, 2002. 34(11): p. 1382-94. Foster, S.A., et al., Inactivation of p16 in human mammary epithelial cells by CpG island methylation. Mol Cell Biol, 1998. 18(4): p. 1793-801. Ramirez, R.D., Putative telomere-independent mechanisms of replicative aging reflect inadequate growth conditions. Genes & Development, 2001. 15(4): p. 398403. Herbert, B.-S., W.E. Wright, and J.W. Shay, p16(INK4a) inactivation is not required to immortalize human mammary epithelial cells. Oncogene, 2002. 21(51): p. 7897900. Romanov, S.R., et al., Normal human mammary epithelial cells spontaneously escape senescence and acquire genomic changes. Nature, 2001. 409(6820): p. 633-7. Wazer, D.E., et al., Immortalization of distinct human mammary epithelial cell types by human papilloma virus 16 E6 or E7. Proc Natl Acad Sci U S A, 1995. 92(9): p. 3687-91. Thibodeaux, C.A., et al., Immortalization and transformation of human mammary epithelial cells by a tumor-derived Myc mutant. Breast Cancer Res Treat, 2009. 116(2): p. 281-294. Stampfer, M.R., et al., Expression of the telomerase catalytic subunit, hTERT, induces resistance to transforming growth factor beta growth inhibition in p16INK4A(-) human mammary epithelial cells. Proc Natl Acad Sci USA, 2001. 98(8): p. 4498-503. Stampfer, M.R. and J.C. Bartley, Induction of transformation and continuous cell lines from normal human mammary epithelial cells after exposure to benzo[a]pyrene. Proc Natl Acad Sci USA, 1985. 82(8): p. 2394-8. Shay, J.W., et al., The frequency of immortalization of human fibroblasts and mammary epithelial cells transfected with SV40 large T-antigen. Exp Cell Res, 1993. 209(1): p. 45-52. Nonet, G.H., et al., The ZNF217 gene amplified in breast cancers promotes immortalization of human mammary epithelial cells. Cancer Res, 2001. 61(4): p. 1250-4. Walen, K.H. and M.R. Stampfer, Chromosome analyses of human mammary epithelial cells at stages of chemical-induced transformation progression to immortality. Cancer Genet Cytogenet, 1989. 37(2): p. 249-61.  219  191. 192. 193. 194. 195. 196. 197. 198. 199. 200. 201. 202. 203. 204. 205. 206. 207.  Toouli, C.D., et al., Comparison of human mammary epithelial cells immortalized by simian virus 40 T-Antigen or by the telomerase catalytic subunit. Oncogene, 2002. 21(1): p. 128-39. Stampfer, M.R. and P. Yaswen, Culture models of human mammary epithelial cell transformation. J Mammary Gland Biol Neoplasia, 2000. 5(4): p. 365-78. Clark, R., et al., Transformation of human mammary epithelial cells by oncogenic retroviruses. Cancer Res, 1988. 48(16): p. 4689-94. Soule, H.D., et al., Isolation and characterization of a spontaneously immortalized human breast epithelial cell line, MCF-10. Cancer Res, 1990. 50(18): p. 6075-86. Soule, H.D., et al., Isolation and characterization of a spontaneously immortalized human breast epithelial cell line, MCF-10. Cancer research, 1990. 50(18): p. 607586. Kim, Y.M., et al., Continuous in vitro exposure to low-dose genistein induces genomic instability in breast epithelial cells. Cancer Genet Cytogenet, 2008. 186(2): p. 78-84. Briand, P., O.W. Petersen, and B. Van Deurs, A new diploid nontumorigenic human breast epithelial cell line isolated and propagated in chemically defined medium. In Vitro Cell Dev Biol, 1987. 23(3): p. 181-8. Hackett, A., H. Smith, and E. Springer…, Two syngeneic cell lines from human breast tissue: the aneuploid mammary epithelial (Hs578T) and the diploid myoepithelial (Hs578Bst) cell lines. Journal of the …, 1977. Hayflick, L. and P.S. Moorhead, The serial cultivation of human diploid cell strains. Exp Cell Res, 1961. 25: p. 585-621. Raouf, A., et al., Genomic Instability of Human Mammary Epithelial Cells Overexpressing a Truncated Form of EMSY. JNCI Journal of the National Cancer Institute, 2005. 97(17): p. 1302-1306. Bartley, J.C. and M.R. Stampfer, Factors influencing benzo[a]pyrene metabolism in human mammary epithelial cells in culture. Carcinogenesis, 1985. 6(7): p. 1017-22. Stampfer, M., R.C. Hallowes, and A.J. Hackett, Growth of normal human mammary cells in culture. In Vitro, 1980. 16(5): p. 415-25. Brenner, A.J., M.R. Stampfer, and C.M. Aldaz, Increased p16 expression with first senescence arrest in human mammary epithelial cells and extended growth capacity with p16 inactivation. Oncogene, 1998. 17(2): p. 199-205. Holst, C.R., et al., Methylation of p16(INK4a) promoters occurs in vivo in histologically normal human mammary epithelia. Cancer Res, 2003. 63(7): p. 1596601. Novak, P., et al., Stepwise DNA Methylation Changes Are Linked to Escape from Defined Proliferation Barriers and Mammary Epithelial Cell Immortalization. Cancer Res, 2009. 69(12): p. 5251-5258. Moll, R., et al., The catalog of human cytokeratins: patterns of expression in normal epithelia, tumors and cultured cells. Cell, 1982. 31(1): p. 11-24. Nagle, R.B., et al., Characterization of breast carcinomas by two monoclonal antibodies distinguishing myoepithelial from luminal epithelial cells. J Histochem Cytochem, 1986. 34(7): p. 869-81.  220  208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 222. 223. 224. 225. 226.  Jarasch, E.D., et al., Differential diagnosis of benign epithelial proliferations and carcinomas of the breast using antibodies to cytokeratins. Hum Pathol, 1988. 19(3): p. 276-89. Lane, E.B. and C.M. Alexander, Use of keratin antibodies in tumor diagnosis. Semin Cancer Biol, 1990. 1(3): p. 165-79. Barbareschi, M., et al., p63, a p53 homologue, is a selective nuclear marker of myoepithelial cells of the human breast. Am J Surg Pathol, 2001. 25(8): p. 1054-60. Gusterson, B.A., et al., Basal cytokeratins and their relationship to the cellular origin and functional classification of breast cancer. Breast Cancer Res, 2005. 7(4): p. 1438. Miles, C.P., Prolonged Culture Of Diploid Human Cells. Cancer Res, 1964. 24: p. 1070-82. Rehen, S.K., et al., Constitutional aneuploidy in the normal human brain. Journal of Neuroscience, 2005. 25(9): p. 2176-80. Yurov, Y.B., et al., The variation of aneuploidy frequency in the developing and adult human brain revealed by an interphase FISH study. J Histochem Cytochem, 2005. 53(3): p. 385-90. Yurov, Y.B., et al., Aneuploidy and Confined Chromosomal Mosaicism in the Developing Human Brain. PLoS ONE, 2007. 2(6): p. e558. Cimino, M.C., R.R. Tice, and J.C. Liang, Aneuploidy in mammalian somatic cells in vivo. Mutat Res, 1986. 167(1-2): p. 107-22. Gupta, P. and C. Kuperwasser, Disease models of breast cancer. Drug Discovery Today: Disease Models, 2004. 1(1): p. 9-16. Parmar, H., et al., A novel method for growing human breast epithelium in vivo using mouse and human mammary fibroblasts. Endocrinology, 2002. 143(12): p. 4886-96. Ito, M., NOD/SCID/gamma cnull mouse: an excellent recipient mouse model for engraftment of human cells. Blood, 2002. 100(9): p. 3175-3182. Petersen, O.W., et al., Interaction with basement membrane serves to rapidly distinguish growth and differentiation pattern of normal and malignant human breast epithelial cells. Proc Natl Acad Sci USA, 1992. 89(19): p. 9064-8. Taylor-Papadimitriou, J., M. Shearer, and M. Stoker, Growth Requirements Of Human Mammary Epithelial-Cells In Culture. Int J Cancer, 1977. 20(6): p. 903-908. Bhowmick, N.A., TGF- Signaling in Fibroblasts Modulates the Oncogenic Potential of Adjacent Epithelia. Science, 2004. 303(5659): p. 848-851. Sakakura, T., Y. Sakagami, and Y. Nishizuka, Dual origin of mesenchymal tissues participating in mouse mammary gland embryogenesis. Dev Biol, 1982. 91(1): p. 202-7. Kratochwil, K., Organ specificity in mesenchymal induction demonstrated in the embryonic development of the mammary gland of the mouse. Dev Biol, 1969. 20(1): p. 46-71. Cunha, G.R. and Y.K. Hom, Role of mesenchymal-epithelial interactions in mammary gland development. J Mammary Gland Biol Neoplasia, 1996. 1(1): p. 2135. Haagensen, C.D., The physiology of the breast as it concerns the clinician. Am J Obstet Gynecol, 1971. 109(2): p. 206-9.  221  227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239.  240. 241. 242.  Atherton, A.J., et al., Differential expression of type XIV collagen/undulin by human mammary gland intralobular and interlobular fibroblasts. Cell and tissue research, 1998. 291(3): p. 507-511. Outzen, H.C. and R.P. Custer, Growth of human normal and neoplastic mammary tissues in the cleared mammary fat pad of the nude mouse. J Natl Cancer Inst, 1975. 55(6): p. 1461-6. Yang, J., et al., Phenotypic characterization of collagen gel embedded primary human breast epithelial cells in athymic nude mice. Cancer Lett, 1994. 81(2): p. 11727. Bodgen, A.E. and D.D. Von Hoff, Comparison of the human tumor cloning and subrenal capsule assays. Cancer research, 1984. 44(3): p. 1087. Camps, J.L., et al., Fibroblast-mediated acceleration of human epithelial tumor growth in vivo. Proc Natl Acad Sci USA, 1990. 87(1): p. 75-9. Lasfargues, E., Cultivation and behavior in vitro of the normal mammary epithelium of the adult mouse* 1:: II. Observations on the secretory activity. Experimental Cell Research, 1957. 13(3): p. 553-562. Lasfargues, E.Y., Cultivation and behavior in vitro of the normal mammary epithelium of the adult mouse. Anat Rec, 1957. 127(1): p. 117-29. Emerman, J.T. and D.R. Pitelka, Maintenance and induction of morphological differentiation in dissociated mammary epithelium on floating collagen membranes. In Vitro, 1977. 13(5): p. 316-28. Berdichevsky, F., et al., Branching morphogenesis of human mammary epithelial cells in collagen gels. J Cell Sci, 1994. 107: p. 3557. Li, M.L., et al., Influence of a reconstituted basement membrane and its components on casein gene expression and secretion in mouse mammary epithelial cells. Proc Natl Acad Sci USA, 1987. 84(1): p. 136-40. Lasfargues, E.Y. and D.H. Moore, A method for the continuous cultivation of mammary epithelium. In Vitro, 1971. 7(1): p. 21-5. White, M.T., et al., In vitro analysis of proliferating epithelial cell populations from the mouse mammary gland: fibroblast-free growth and serial passage. In Vitro Cellular & Developmental Biology-Plant, 1978. 14(3): p. 271-281. Bergstraesser, L.M. and S.A. Weitzman, Culture of normal and malignant primary human mammary epithelial cells in a physiological manner simulates in vivo growth patterns and allows discrimination of cell type. Cancer research, 1993. 53(11): p. 2644-54. Shamay, A., et al., Effect of insulin-like growth factor I on deoxyribonucleic acid synthesis and galactopoiesis in bovine undifferentiated and lactating mammary tissue in vitro. Endocrinology, 1988. 123(2): p. 804-9. Smith, J.A., D.P. Winslow, and P.S. Rudland, Different growth factors stimulate cell division of rat mammary epithelial, myoepithelial, and stromal cell lines in culture. J Cell Physiol, 1984. 119(3): p. 320-6. Hovey, R.C., D.D. MacKenzie, and T.B. McFadden, The proliferation of mouse mammary epithelial cells in response to specific mitogens is modulated by the mammary fat pad in vitro. In Vitro Cell Dev Biol Anim, 1998. 34(5): p. 385-92.  222  243. 244. 245. 246. 247. 248. 249. 250. 251. 252. 253. 254. 255. 256. 257. 258. 259. 260. 261.  Plath, A., et al., Expression of transforming growth factors alpha and beta-1 messenger RNA in the bovine mammary gland during different stages of development and lactation. J Endocrinol, 1997. 155(3): p. 501-11. Wilson, S.E., et al., Hepatocyte growth factor (HGF), keratinocyte growth factor (KGF), and their receptors in human breast cells and tissues: alternative receptors. Cell Mol Biol Res, 1994. 40(4): p. 337-50. Silberstein, G.B. and C.W. Daniel, Reversible inhibition of mammary gland growth by transforming growth factor-beta. Science, 1987. 237(4812): p. 291-3. Stampfer, M.R., Cholera toxin stimulation of human mammary epithelial cells in culture. In Vitro, 1982. 18(6): p. 531-7. Silberstein, G.B., et al., In vivo, cAMP stimulates growth and morphogenesis of mouse mammary ducts. Proc Natl Acad Sci USA, 1984. 81(15): p. 4950-4. McGrath, C.M., Augmentation of the response of normal mammary epithelial cells to estradiol by mammary stroma. Cancer research, 1983. 43(3): p. 1355-60. Shyamala, G., Roles of estrogen and progesterone in normal mammary gland development insights from progesterone receptor null mutant mice and in situ localization of receptor. Trends Endocrinol Metab, 1997. 8(1): p. 34-9. Bishop, C., et al., Primary Cilium-Dependent and-Independent Hedgehog Signaling Inhibits p16INK4A. Molecular Cell, 2010. Borgdorff, V., et al., Multiple microRNAs rescue from Ras-induced senescence by inhibiting p21Waf1&sol;Cip1. Oncogene, 2010. 29(15): p. 2262-2271. Smolen, G.A., et al., A genome-wide RNAi screen identifies multiple RSK-dependent regulators of cell migration. Genes & Development, 2010. 24(23): p. 2654-2665. Silva, J.M., et al., Profiling Essential Genes in Human Mammary Cells by Multiplex RNAi Screening. Science, 2008. 319(5863): p. 617-620. Falschlehner, C., et al., High-throughput RNAi screening to dissect cellular pathways: A how-to guide. Biotechnol. J., 2010. 5(4): p. 368-376. Li, S. and G.D. Shipley, Expression of multiple species of basic fibroblast growth factor mRNA and protein in normal and tumor-derived mammary epithelial cells in culture. Cell Growth Differ, 1991. 2(4): p. 195-202. Ulivi, P., et al., c-kit and SCF expression in normal and tumor breast tissue. Breast Cancer Res Treat, 2004. 83(1): p. 33-42. Thike, A.A., et al., Triple negative breast cancer: outcome correlation with immunohistochemical detection of basal markers. Am J Surg Pathol, 2010. 34(7): p. 956-64. Lammie, A., et al., Expression of c-kit and kit ligand proteins in normal human tissues. J Histochem Cytochem, 1994. 42(11): p. 1417-25. Lim, E., et al., Transcriptome analyses of mouse and human mammary cell subpopulations reveal multiple conserved genes and pathways. Breast Cancer Res, 2010. 12(2): p. R21. Wiseman, B.S., Site-specific inductive and inhibitory activities of MMP-2 and MMP3 orchestrate mammary gland branching morphogenesis. The Journal of Cell Biology, 2003. 162(6): p. 1123-1133. Dunne, J., et al., The apparent uptake of fluorescently labeled siRNAs by electroporated cells depends on the fluorochrome. Oligonucleotides, 2003. 13(5): p. 375-80.  223  262. 263. 264. 265. 266. 267. 268. 269. 270. 271. 272. 273. 274. 275. 276. 277. 278. 279. 280.  Poon, S.S., et al., Intensity calibration and automated cell cycle gating for highthroughput image-based siRNA screens of mammalian cells. Cytometry A, 2008. 73(10): p. 904-17. Reynolds, A., et al., Rational siRNA design for RNA interference. Nat Biotechnol, 2004. 22(3): p. 326-30. Boese, Q., et al., Mechanistic insights aid computational short interfering RNA design. Methods Enzymol, 2005. 392: p. 73-96. Birmingham, A., et al., 3' UTR seed matches, but not overall identity, are associated with RNAi off-targets. Nat Methods, 2006. 3(3): p. 199-204. Birmingham, A., et al., A protocol for designing siRNAs with high functionality and specificity. Nat Protoc, 2007. 2(9): p. 2068-78. Tschuch, C., et al., Off-target effects of siRNA specific for GFP. BMC Mol Biol, 2008. 9(1): p. 60. Tang, W., et al., A genome-wide RNAi screen for Wnt/beta-catenin pathway components identifies unexpected roles for TCF transcription factors in cancer. Proc Natl Acad Sci U S A, 2008. 105(28): p. 9697-702. Hitomi, J., et al., Identification of a molecular signaling network that regulates a cellular necrotic cell death pathway. Cell, 2008. 135(7): p. 1311-23. Iorns, E., et al., Integrated functional, gene expression and genomic analysis for the identification of cancer targets. PLoS ONE, 2009. 4(4): p. e5120. Hu, G., et al., A genome-wide RNAi screen identifies a new transcriptional module required for self-renewal. Genes Dev, 2009. 23(7): p. 837-48. Kumar, D., et al., Genome-wide analysis of the host intracellular network that regulates survival of Mycobacterium tuberculosis. Cell, 2010. 140(5): p. 731-43. Parker, J.S., et al., Supervised Risk Predictor of Breast Cancer Based on Intrinsic Subtypes. Journal of Clinical Oncology, 2009. 27(8): p. 1160-1167. Inohara, N., et al., An induced proximity model for NF-kappa B activation in the Nod1/RICK and RIP signaling pathways. J Biol Chem, 2000. 275(36): p. 27823-31. Biswas, D.K., et al., NF-kappa B activation in human breast cancer specimens and its role in cell proliferation and apoptosis. Proc Natl Acad Sci U S A, 2004. 101(27): p. 10137-42. Bertucci, F., et al., How different are luminal A and basal breast cancers? Int J Cancer, 2009. 124(6): p. 1338-48. Donoghue, M., et al., A novel angiotensin-converting enzyme-related carboxypeptidase (ACE2) converts angiotensin I to angiotensin 1-9. Circ Res, 2000. 87(5): p. E1-9. Puddefoot, J.R., et al., The role of angiotensin II in the regulation of breast cancer cell adhesion and invasion. Endocr Relat Cancer, 2006. 13(3): p. 895-903. del Pilar Carrera, M., et al., Mammary renin-angiotensin system-regulating aminopeptidase activities are modified in rats with breast cancer. Tumour Biol, 2010. 31(6): p. 583-8. Maione, A., et al., Angiotensin-converting enzyme inhibitors, angiotensin receptor blockers and combined therapy in patients with micro- and macroalbuminuria and other cardiovascular risk factors: a systematic review of randomized controlled trials. Nephrol Dial Transplant, 2011  .  224  281. 282. 283. 284. 285. 286. 287. 288. 289. 290. 291. 292. 293. 294. 295. 296. 297. 298. 299.  Byrnes, J.J., et al., Effects of the ACE2 inhibitor GL1001 on acute dextran sodium sulfate-induced colitis in mice. Inflamm Res, 2009. 58(11): p. 819-27. Liu, S., et al., Expression of Autotaxin and Lysophosphatidic Acid Receptors Increases Mammary Tumorigenesis, Invasion, and Metastases. Cancer Cell, 2009. 15(6): p. 539-550. Ashburner, M., et al., Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat Genet, 2000. 25(1): p. 25-9. Jackson, A.L., et al., Expression profiling reveals off-target gene regulation by RNAi. Nat Biotechnol, 2003. 21(6): p. 635-7. Birmingham, A., et al., 3ʹ′ UTR seed matches, but not overall identity, are associated with RNAi off-targets. Nat Meth, 2006. 3(3): p. 199-204. Jackson, A.L., et al., Position-specific chemical modification of siRNAs reduces "offtarget" transcript silencing. RNA, 2006. 12(7): p. 1197-205. Fedorov, Y., et al., Off-target effects by siRNA can induce toxic phenotype. RNA, 2006. 12(7): p. 1188-96. Hornung, V., et al., Sequence-specific potent induction of IFN-α by short interfering RNA in plasmacytoid dendritic cells through TLR7. Nat Med, 2005. 11(3): p. 263270. Judge, A.D., et al., Sequence-dependent stimulation of the mammalian innate immune response by synthetic siRNA. Nat Biotechnol, 2005. 23(4): p. 457-462. Pramfalk, C., et al., Insulin receptor activation and down-regulation by cationic lipid transfection reagents. BMC Cell Biol, 2004. 5: p. 7. Sudbery, I., et al., Systematic analysis of off-target effects in an RNAi screen reveals microRNAs affecting sensitivity to TRAIL-induced apoptosis. BMC Genomics, 2010. 11: p. 175. Strickland, P., et al., Slit2 and netrin 1 act synergistically as adhesive cues to generate tubular bi-layers during ductal morphogenesis. Development, 2006. 133(5): p. 823-32. Vaught, D., J. Chen, and D.M. Brantley-Sieders, Regulation of mammary gland branching morphogenesis by EphA2 receptor tyrosine kinase. Mol Biol Cell, 2009. 20(10): p. 2572-81. Tremblay, F., et al., Disruption of G Protein-Coupled Receptor 39 Impairs Insulin Secretion in Vivo. Endocrinology, 2009. 150(6): p. 2586-2595. Sharir, H., et al., Zinc Released from Injured Cells Is Acting via the Zn2+-sensing Receptor, ZnR, to Trigger Signaling Leading to Epithelial Repair. Journal of Biological Chemistry, 2010. 285(34): p. 26097-26106. Rimm, D.L., J.H. Sinard, and J.S. Morrow, Reduced alpha-catenin and E-cadherin expression in breast cancer. Lab Invest, 1995. 72(5): p. 506-12. Knirsh, R., et al., Loss of E-cadherin-mediated cell-cell contacts activates a novel mechanism for up-regulation of the proto-oncogene c-Jun. Mol Biol Cell, 2009. 20(7): p. 2121-9. Ehlers, S., et al., RIP2 regulates growth and differentiation of normal myoblasts and of rhabdomyosarcoma cells. Eur J Cell Biol, 2008. 87(3): p. 163-72. Liu, J., et al., Sex differences in renal angiotensin converting enzyme 2 (ACE2) activity are 17β-oestradiol-dependent and sex chromosome-independent. Biology of Sex Differences, 2010. 1(1): p. 6.  225  300. 301. 302. 303. 304. 305. 306. 307. 308. 309. 310. 311. 312. 313. 314. 315. 316. 317. 318.  Lawrence, P. and P. Shelton, Determination Of Polarity In Developing Insect Retina. J Embryol Exp Morph, 1975. 33(APR): p. 471-486. Usui, T., et al., Flamingo, a seven-pass transmembrane cadherin, regulates planar cell polarity under the control of Frizzled. Cell, 1999. 98(5): p. 585-95. Shimada, Y., et al., Asymmetric colocalization of Flamingo, a seven-pass transmembrane cadherin, and Dishevelled in planar cell polarization. Curr Biol, 2001. 11(11): p. 859-63. Rawls, A.S., Strabismus requires Flamingo and Prickle function to regulate tissue polarity in the Drosophila eye. Development, 2003. 130(9): p. 1877-1887. Wu, J. and M. Mlodzik, The frizzled extracellular domain is a ligand for Van Gogh/Stbm during nonautonomous planar cell polarity signaling. Dev Cell, 2008. 15(3): p. 462-9. Chen, W.-S., et al., Asymmetric homotypic interactions of the atypical cadherin flamingo mediate intercellular polarity signaling. Cell, 2008. 133(6): p. 1093-105. Strutt, D.I., Asymmetric localization of frizzled and the establishment of cell polarity in the Drosophila wing. Molecular Cell, 2001. 7(2): p. 367-75. Feiguin, F., et al., The ankyrin repeat protein Diego mediates Frizzled-dependent planar polarization. Dev Cell, 2001. 1(1): p. 93-101. Axelrod, J.D., Unipolar membrane association of Dishevelled mediates Frizzled planar cell polarity signaling. Genes & Development, 2001. 15(10): p. 1182-7. Curtin, J.A., et al., Mutation of Celsr1 disrupts planar polarity of inner ear hair cells and causes severe neural tube defects in the mouse. Curr Biol, 2003. 13(13): p. 112933. Ross, A.J., et al., Disruption of Bardet-Biedl syndrome ciliary proteins perturbs planar cell polarity in vertebrates. Nat Genet, 2005. 37(10): p. 1135-1140. Song, H., et al., Planar cell polarity breaks bilateral symmetry by controlling ciliary positioning. Nature, 2010. 466(7304): p. 378-382. Sugiyama, Y., et al., Secreted frizzled-related protein disrupts PCP in eye lens fiber cells that have polarised primary cilia. Developmental Biology, 2010. 338(2): p. 193201. Gao, J., et al., Laminin-511 is an epithelial message promoting dermal papilla development and function during early hair morphogenesis. Genes & Development, 2008. 22(15): p. 2111-2124. Lehman, J.M., et al., An Essential Role for Dermal Primary Cilia in Hair Follicle Morphogenesis. J Investig Dermatol, 2009. 129(2): p. 438-448. Devenport, D. and E. Fuchs, Planar polarization in embryonic epidermis orchestrates global asymmetric morphogenesis of hair follicles. Nat Cell Biol, 2008. 10(11): p. 1257-1268. Tissir, F., et al., Lack of cadherins Celsr2 and Celsr3 impairs ependymal ciliogenesis, leading to fatal hydrocephalus. Nature Neuroscience, 2010. 13(6): p. 700-707. Guirao, B., et al., Coupling between hydrodynamic forces and planar cell polarity orients mammalian motile cilia. Nature Publishing Group, 2010. 12(4): p. 341-350. Wallingford, J.B., Planar cell polarity, ciliogenesis and neural tube defects. Human Molecular Genetics, 2006. 15(Review Issue 2): p. R227-R234.  226  319. 320. 321. 322. 323. 324. 325. 326. 327. 328.  329. 330. 331. 332. 333. 334. 335.  Wallingford, J.B., S.E. Fraser, and R.M. Harland, Convergent extension: the molecular control of polarized cell movement during embryonic development. Developmental Cell, 2002. 2(6): p. 695-706. Caddy, J., et al., Epidermal Wound Repair Is Regulated by the Planar Cell Polarity Signaling Pathway. Developmental Cell, 2010. 19(1): p. 138-147. Qu, Y., et al., Atypical cadherins Celsr1-3 differentially regulate migration of facial branchiomotor neurons in mice. J Neurosci, 2010. 30(28): p. 9392-401. Wada, H., et al., Frizzled3a and Celsr2 function in the neuroepithelium to regulate migration of facial motor neurons in the developing zebrafish hindbrain. Development, 2006. 133(23): p. 4749-4759. Yates, L.L., et al., The PCP genes Celsr1 and Vangl2 are required for normal lung branching morphogenesis. Human Molecular Genetics, 2010. 19(11): p. 2251-67. Yates, L.L., et al., The planar cell polarity gene Vangl2 is required for mammalian kidney-branching morphogenesis and glomerular maturation. Human Molecular Genetics, 2010. 19(23): p. 4663-4676. Welsh, J., et al., Age-Related Changes in the Epithelial and Stromal Compartments of the Mammary Gland in Normocalcemic Mice Lacking the Vitamin D3 Receptor. PLoS ONE, 2011. 6(1): p. e16479. Wysolmerski, J.J., et al., Rescue of the parathyroid hormone-related protein knockout mouse demonstrates that parathyroid hormone-related protein is essential for mammary gland development. Development, 1998. 125(7): p. 1285-94. Brisken, C., et al., A paracrine role for the epithelial progesterone receptor in mammary gland development. Proc Natl Acad Sci USA, 1998. 95(9): p. 5076-81. Gallego, M., Prolactin, Growth Hormone, and Epidermal Growth Factor Activate Stat5 in Different Compartments of Mammary Tissue and Exert Different and Overlapping Developmental Effects. Developmental Biology, 2001. 229(1): p. 163175. Lewis, M., The Gli2 Transcription Factor Is Required for Normal Mouse Mammary Gland Development. Developmental Biology, 2001. 238(1): p. 133-144. Robinson, G.W. and L. Hennighausen, Inhibins and activins regulate mammary epithelial cell differentiation through mesenchymal-epithelial interactions. Development, 1997. 124(14): p. 2701-8. van 't Veer, L.J., et al., Gene expression profiling predicts clinical outcome of breast cancer. Nature, 2002. 415(6871): p. 530-6. Carroll, J.S., et al., Chromosome-wide mapping of estrogen receptor binding reveals long-range regulation requiring the forkhead protein FoxA1. Cell, 2005. 122(1): p. 33-43. Simian, M., et al., The interplay of matrix metalloproteinases, morphogens and growth factors is necessary for branching of mammary epithelial cells. Development, 2001. 128(16): p. 3117-31. Wu, J., T.J. Klein, and M. Mlodzik, Subcellular Localization of Frizzled Receptors, Mediated by Their Cytoplasmic Tails, Regulates Signaling Pathway Specificity. Plos Biol, 2004. 2(7): p. e158. Courbard, J.-R., et al., The apical/basal-polarity determinant Scribble cooperates with the PCP core factor Stbm/Vang and functions as one of its effectors. Developmental Biology, 2009. 333(1): p. 67-77.  227  336. 337. 338. 339. 340. 341. 342. 343. 344. 345. 346. 347. 348. 349. 350. 351. 352. 353. 354.  Dow, L.E., et al., The tumour-suppressor Scribble dictates cell polarity during directed epithelial migration: regulation of Rho GTPase recruitment to the leading edge. Oncogene, 2007. 26(16): p. 2272-2282. Hebner, C., V.M. Weaver, and J. Debnath, Modeling Morphogenesis and Oncogenesis in Three-Dimensional Breast Epithelial Cultures. Annu. Rev. Pathol. Mech. Dis., 2008. 3(1): p. 313-339. Debnath, J., et al., The role of apoptosis in creating and maintaining luminal space within normal and oncogene-expressing mammary acini. Cell, 2002. 111(1): p. 2940. Muthuswamy, S.K., et al., ErbB2, but not ErbB1, reinitiates proliferation and induces luminal repopulation in epithelial acini. Nat Cell Biol, 2001. 3(9): p. 785-792. Bilder, D., Cooperative Regulation of Cell Polarity and Growth by Drosophila Tumor Suppressors. Science, 2000. 289(5476): p. 113-116. Vingron, M., et al., Detecting Disease-Specific Dysregulated Pathways Via Analysis of Clinical Expression Profiles. 2008. 4955: p. 347-359. Formstone, C.J., et al., Basal enrichment within neuroepithelia suggests novel function (s) for Celsr1 protein. Molecular and Cellular Neuroscience, 2010. Ventura, A., et al., Restoration of p53 function leads to tumour regression in vivo. Nature, 2007. 445(7128): p. 661-665. Yamamoto, A., et al., Shisa promotes head formation through the inhibition of receptor protein maturation for the caudalizing factors, Wnt and FGF. Cell, 2005. 120(2): p. 223-35. Furushima, K., et al., Mouse homologues of Shisa antagonistic to Wnt and Fgf signalings. Developmental Biology, 2007. 306(2): p. 480-92. Yamaji, D., et al., Devel