HETEROGENEITY IN MAMMARY CANCER: USING MOUSE MODELS TO INVESTIGATE TUMOR SUBTYPE ORIGINS AND INTERCLONAL INTERACTIONS

Transcription

1 The Pennsylvania State University The Graduate School College of Medicine HETEROGENEITY IN MAMMARY CANCER: USING MOUSE MODELS TO INVESTIGATE TUMOR SUBTYPE ORIGINS AND INTERCLONAL INTERACTIONS A Dissertation in Cell and Molecular Biology by Allison S. Cleary 2014 Allison S. Cleary Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy August 2014

2 The dissertation of Allison S. Cleary was reviewed and approved* by the following: Edward J. Gunther Associate Professor of Medicine Pennsylvania State University College of Medicine Dissertation Advisor Chair of Committee David J Feith Associate Professor of Medicine University of Virginia Adjunct Assistant Professor of Cellular & Molecular Physiology Pennsylvania State University College of Medicine Leslie J. Parent Professor of Medicine and Microbiology and Immunology Co-Director, MD/PhD Program Pennsylvania State University College of Medicine Todd D. Schell Associate Professor of Microbiology and Immunology Pennsylvania State University College of Medicine Raghu Sinha Associate Professor of Biochemistry and Molecular Biology Pennsylvania State University College of Medicine Sarah K. Bronson Associate Professor of Cellular and Molecular Physiology Director, Cell and Molecular Biology Graduate Program Pennsylvania State University College of Medicine *Signatures are on file in the Graduate School. ii

3 Abstract. Breast cancer is a heterogeneous disease on two levels. First, breast cancers display inter-tumor heterogeneity evidenced by the diversity of clinically and molecularly defined tumor subtypes that differ with respect to disease progression and drug sensitivity. Second, individual breast cancers display remarkable intra-tumor heterogeneity evidenced by the diversity of component tumor cell subtypes co-residing within each cancer that differ with respect to cell morphology, proliferation rate, metastatic potential, drug sensitivity, and capacity for tumor reconstitution. The cellular mechanisms that generate and maintain breast cancer heterogeneity, both at the level of tumor subtype and tumor cell subtype, remain poorly understood. Using mouse models of human breast cancer, this dissertation examines whether and how breast cancer heterogeneity derives from the diverse and highly interactive cell subtypes that comprise the normal mammary gland, breast cancer s tissue-of-origin. In one set of experiments directed at uncovering the origin of tumor subtypes, we examined why transgenic mouse models of breast cancer nearly always yield a hormone receptor (HR)-negative mammary cancer subtype. In the mammary gland, mature ducts consist of basal and luminal mammary epithelial cell (MEC) subtypes. The luminal epithelial compartment can be further subdivided into hormone receptor (HR)-positive and HR-negative subsets. While human breast cancers frequently express HRs and depend on ovarian hormones for growth, transgenic mouse models of breast cancers show an unexplained bias toward HR-negative disease. Since the majority of mouse breast cancer models iii

4 use the mouse mammary tumor virus long terminal repeat (MMTV-LTR) as a mammary-specific promoter element, we examined whether MMTV targets transgene expression to a specific MEC compartment. Using the MMTV-LTR to drive expression of a nuclear H2BGFP reporter transgene, we observed nuclear labeling restricted to HR-negative cells within the luminal compartment. Combining this labeling strategy with MMTV-directed expression of the Neu oncogene, we found Neu transgene expression was similarly enriched within HRnegative luminal MECs. Further, Neu-initiated neoplasias were comprised entirely of HR-negative cells from the carcinoma-in-situ stage onward. Thus, MMTV-driven Neu expression targets HR-negative luminal cells, culminating in HR-negative tumors. We propose that the HR-negative phenotype of many mouse breast cancer models can be explained by MMTV-driven transgene expression in HR-negative MECs. In another set of experiments, we sought to study interactions between tumor cell subtypes. To do this, we developed a novel experimental platform for culturing chimeric mammary organoids which permits analysis of both the cellautonomous and non-autonomous effects of oncogene expression. By combining primary MECs from two different transgenic donors, chimeric mammary organoids were assembled consisting of intermingled populations of genetically distinct donor MECs that could each be tracked over time. We tested our system using transgenic mouse models engineered to inducibly express either an activated HRas allele or oncogenic Wnt1 in specific MECs. As expected, HRasexpressing cells expanded in number, which is consistent with a predominantly iv

5 cell autonomous role for oncogenic HRas. By contrast, Wnt1-expressing cells did not expand in number. Instead, luminal expression of Wnt1 produced a dramatic and selective expansion of the basal epithelial cell compartment, as captured by live cell imaging. Thus, secreted Wnt1 primarily drove MEC overgrowth by acting in a paracrine rather than autocrine manner. Overall, chimeric organoid analysis can be used as a sensitive and effective tool for studying complex cellcell interactions in the context of both normal and transformed mammary epithelium. In a final set of experiments directed at explaining how diverse tumor cell subtypes are maintained within mammary cancers, we examined the functional relationship between distinct tumor cell clones. Recent studies highlight the phenotypic and genetic diversity present locally within individual breast tumors, but whether this heterogeneity is a cause or a consequence of tumor progression remains unclear. Here, we used mouse models of breast cancer to demonstrate for the first time that interclonal cooperation can be essential for tumor maintenance. Aberrant expression of the secreted signaling molecule Wnt1 generates mixedlineage mammary tumors composed of basal and luminal tumor cell subtypes, which purportedly derive from a bipotent malignant progenitor cell residing atop a tumor cell hierarchy. Using somatic HRas mutations as clonal markers, we showed that some Wnt tumors indeed conformed to a hierarchical configuration, but others unexpectedly harbored genetically distinct basal HRas mutant (HRas mut ) and luminal HRas wild-type (HRas wt ) subclones. Both subclones were required for efficient tumor propagation, which strictly depended on luminally- v

6 produced Wnt1. When biclonal tumors were challenged with Wnt withdrawal to simulate targeted therapy, analysis of tumor regression and relapse revealed that basal subclones recruited heterologous Wnt-producing cells to restore tumor growth. Alternatively, in the absence of a substitute Wnt source, the original subclones often evolved to rescue Wnt pathway activation and drive relapse, either by restoring cooperation or by switching to a defector strategy. Uncovering similar modes of interclonal cooperation in human cancers may inform efforts aimed at eradicating tumor cell communities. vi

7 Table of Contents. List of Figures. ix List of Tables xii List of Abbreviations xiii Preface. xvii Acknowledgements xviii Chapter 1: Literature Review Introduction The Normal Mammary Gland: Function, Structure, and Development Mammary Gland Disease Hormonal Regulation of the Mammary Gland Estrogen Progesterone Paracrine Signaling by Hormones in the Mammary Gland Paracrine Mediators of ER Signaling Paracrine Mediators of PR Signaling Hormones and Breast Cancer Lineage Hierarchy of Mammary Epithelial Cells Breast Cancer as a Heterogeneous Disease Intra-tumor Heterogeneity Clonal Evolution Hypothesis Linear Evolution Branched Evolution Mutator Phenotype Hypothesis Self-Seeding Hypothesis Cancer Stem Cell Hypothesis Clinical Implications of the Distinct Tumor Cell Heterogeneity Models Tumor Cell Inter-clonal Interactions Oppositional Relationships: Competition Oppositional Relationships: Antagonism Cooperative Relationships: Commensalism Cooperative Relationships: Mutualism Overview.. 41 Chapter 2: Mammary Tumors Initiated by MMTV-LTR-Driven Neu Transgenes Originate within a Hormone Receptor-Negative Luminal Cell Subcompartment Abstract.. 48 vii

8 2.2 Introduction Methods Results Discussion.. 62 Chapter 3: A Novel Chimeric Mammary Organoid Technique for Assessing the Cell-Intrinsic and Extrinsic Effects of Oncogene Expression Abstract Introduction Methods Results Discussion..113 Chapter 4: Tumor Cell Heterogeneity Maintained by Cooperating Subclones in Wnt-Driven Mammary Cancers Abstract Introduction Methods Results Discussion..158 Chapter 5: Overall Discussion Introduction Implications for Human Breast Cancer Implications for Metastasis Implications for Tumor Origins Conclusion..191 References 197 viii

9 List of Figures. Chapter 1 Figure 1.1 Stages of postnatal mammary gland development 42 Figure 1.2 Cell autonomous and paracrine mechanisms of progesteroneinduced cell proliferation in the mammary gland 43 Figure 1.3 Models of the epithelial lineage hierarchy in the mammary gland.. 44 Figure 1.4 Molecular breast cancer subtypes 45 Figure 1.5 Models of tumor cell heterogeneity..46 Chapter 2 Figure 2.1 Schematic of the MMTV-LTR. 68 Figure 2.2 Schematic of MMTV-directed H2BGFP labeling 69 Figure 2.3 MMTV-directed H2BGFP expression labels a morphologically distinct luminal epithelial subset. 70 Figure 2.4 MMTV-driven H2BGFP expression labels a HR-negative luminal MEC subcompartment Figure 2.5 Flow cytometry gating strategy for resolving basal and luminal subsets from mammary tissue. 72 Figure 2.6 Isolation of MMTV-expressing cells by FACS. 73 Figure 2.7 Dox treatment through embryogenesis labels both basal and luminal mammary epithelial subsets Figure 2.8 A Short Dox pulse in juvenile animals labels a luminal epithelial subset in mammary ducts, while TEBs contain labeled cells in both basal and luminal compartments.. 77 Figure 2.9 Neu transgene expression is enriched in the H2BGFP-expressing luminal subset of MMTV-rtTA/ teto-h2bgfp/ MMTV-Neu tritransgenic animals...79 Figure 2.10 Progesterone Receptor expression in MMTV-Neu precursor lesions. 81 Figure 2.11 Neu-induced hyperplasia in MMTV-rtTA/tetO-Neu animals is rapidly reversible upon Dox withdrawal. 83 Figure 2.12 Dox time course of MMTV-rtTA/tetO-H2BGFP/tetO-Neu tritransgenic mice shows transition to HR-negative disease.. 85 Chapter 3 Figure 3.1 Tet-operator driven transgenes 116 Figure 3.2 Ras signaling pathway.117 Figure 3.3 Canonical Wnt signaling pathway Figure 3.4 Time-lapse imaging of Ras-mediated MEC overgrowth 119 Figure 3.5 Time-lapse imaging of Wnt1-mediated MEC overgrowth. 121 ix

10 Figure 3.6 Generation of Chimeric Mammary Organoids Figure 3.7 Luminal expression of TWNT leads to increased cell divisions in the basal epithelial compartment..124 Figure 3.8 Luminal expression of TWNT does not influence cell fate events in the surrounding luminal epithelial compartment Figure 3.9 Luminal expression of TRAS does not influence cell fate events in the surrounding basal epithelial compartment Figure 3.10 Luminal expression of TRAS does not influence cell fate events in the surrounding luminal epithelial compartment..130 Figure 3.11 Time-lapse imaging of Ras-mediated basal MEC overgrowth 132 Figure 3.12 Basal expression of TRAS does not influence cell fate events in the surrounding luminal epithelial compartment..134 Figure 3.13 Basal expression of TRAS does not influence cell fate events in the surrounding basal epithelial compartment Chapter 4 Figure 4.1 Evidence for distinct basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high subclones within some MMTV-Wnt1 tumors..162 Figure 4.2 FACS gating strategy for resolving basal and luminal subsets from mammary tumors..163 Figure 4.3 Hierarchical and Biclonal MMTV-Wnt1 tumors are histologically indistinguishable Figure 4.4 Rescue of basal HRas mut iwnt tumor cells from Wnt withdrawal by heterologous luminal cwnt cells Figure 4.5 Tumor regression following Dox withdrawal. 166 Figure 4.6 Biclonal configuration of reconstituted iwnt/mrfp + tumors. 167 Figure 4.7 Basal subclones from two additional iwnt/mrfp + tumors rescued from Dox withdrawal by heterologous cwnt host cells Figure 4.8 Lineage-restricted subclones recapitulate mosaiform heterogeneity in chimeric cwnt/iwnt tumors..169 Figure 4.9 Biclonal configuration of tumors reconstituted from sorted iwnt/mrfp + tumor cell subsets 170 Figure 4.10 Both sorted basal and sorted luminal cell populations are required to reconstitute biclonal tumors..171 Figure 4.11 Relapse of biclonal tumors through the evolution of either subclone 172 Figure 4.12 Increased HRas MAFs in ßcat mut DITs is not due to gross copy changes at the HRas locus.173 Figure 4.13 Mixed-lineage character of DITs x

11 Chapter 5 Figure 5.1 Models of HR-positive breast tumors Figure 5.2 Models of metastasis for biclonal tumors 194 Figure 5.3 Models of biclonal tumor origins 196 xi

12 List of Tables. Chapter 2 Table 2.1 MMTV-driven tumor models.. 87 Table 2.2 Genotyping primers..90 Chapter 3 Table 3.1 Genotyping primers 138 Chapter 4 Table 4.1 Genotyping primers.175 Table 4.2 Sequencing primers.176 Table 4.3 Primers for generating probes for northern hybridization Table 4.4 Unsorted tumor cells efficiently reconstitute tumors..178 xii

13 List of Abbreviations. 2D 2 Dimensions 3D 3 Dimensions AB Alveolar Bud ADAM17 A Disintegrin and metalloproteinase 17 AF1 Activation function-1 AF2 Activation function-2 Akt Protein Kinase B APC Adenomatous polyposis coli β-cat β-catenin BRCA1 Breast cancer 1, early onset BrDU Bromodeoxyuridine C Celcius cdna Complementary deoxyribose nucleic acid CDP CCAAT binding protein CsCl Cesium chloride cwnt Constitutive Wnt model d Day DIT Doxycycline independent tumor DMEM Dulbeco s DNA Deoxyribose nucleic acid Dox Doxycycline EDTA Ethylenediaminetetaacetic acid EGFR Epidermal growth factor receptor EpCAM Epithelial cell adhesion molecule ER Estrogen receptor ERα Estrogen receptor α isoform ErbB2 v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 2 ERK Extracellular-signal regulated kinase ERKO Estrogen receptor knock out FACS Fluorescence activated cell sorting FAM 6-carboxyfluorescein FGF Fibroblast growth factor FISH Fluorescence in situ hybridization FITC Fluorescein isothiocyanate FVB Friend virus B mouse strain GAP GTPase activating protein GFP Green fluorescent protein xiii

14 GFR Growth factor reduced GSK3 Glycogen synthase kinase 3 GTP Guanosine triposphate H2BGFP Histone 2B- green fluorescent protein fusion protein H&E Hematoxylin and Eosin HER2 Human epidermal growth factor receptor 2 HRas Harvey rat sarcoma viral oncogene homolog HR Hormone receptor HRE Hormone response element IGF1 Insulin growth factor 1 IgG Immunoglobulin G i.p. Intraperitoneal ITS Insulin transferrin selenium iwnt Inducible Wnt model JAK Janus kinase 2 K5 Keratin-5 K8 Keratin-8 K14 Keratin-14 kg Kilogram KRas Kirsten rat sarcoma viral oncogene homolog Lin Lineage markers (CD45/ CD31/ TER119/ BP-1) Lef Lymphoid enhancer binding factor LN Lymph node LOH Loss of heterozygosity LRP5/6 Low-density lipoprotein receptor-related protein 5/6 LTR Long terminal repeat MAF Mutant allelic fraction MAPK Mitogen-activated protein kinase MAP2K Mitogen-activated protein kinase kinase MaSC Mammary stem cell MCF-7 Michigan cancer foundation-7 MEK Mitogen-activated protein/ extracellular signal-regulated kinase MEC Mammary epithelial cell MFP Mammary fat pad mg milligram MHC Major histocompatibility complex ml Milliliter mm Millimeter MMTV Mouse mammary tumor virus xiv

15 MMTV-LTR Mosue mammary tumor virus- long terminal repeat MNU N-methyl-N-nitrosourea mrfp Monomeric red fluorescent protein mtor Mammalian target of rapamycin mut Mutant Neu Neuro/glioblastoma derived oncogene homolog NRas Neuroblastoma Ras viral oncogene homolog NRE Negative response element PBS Phosphate buffered saline PCR Polymerase chain reaction PDGF Platelet derived growth factor PE Phycoerythrin PH Peak Height PI3K Phosphoinositide 3-kinase PR Progesterone receptor PR-A Progesterone receptor- A isoform PR-B Progesterone receptor- B isoform PRKO Progesterone receptor knock out Pten Phosphatase and tensin homolog PyV Polyoma virus qrt-pcr Quantitative real time polymerase chain reaction Raf Rapidly accelerated fibrosarcoma RANK Receptor activator of nuclear factor kappa-b RANK-L Receptor activator of nuclear factor kappa-b ligand Ras Rat sarcoma Ref Reference RNA Ribonucleic acid rpm Revolutions per minute RSV Rous sarcoma virus RTK Receptor tyrosine kinase RT-PCR Reverse transcription polymerase chain reaction rtta Reverse tetracycline transactivator SATB1 Special AT-rich sequence binding protein 1 SAg Super antigen Sca1 Stem cell antigen 1 SFK Src family kinase SMA Smooth muscle actin SOS Son of sevenless Sox Sry-related HMG box Src Sarcoma xv

16 STAT Tcf TEB teto TGFβ TGFP TNF TNM TRAS Tritc TWNT U μl μm Wnt wt YFP Signal transducer and activator of transcription Transcription factor Terminal end bud Tetracyline operator Transforming growth factor beta Tet-operator controlled H2BGFP transgene Tumor necrosis factor Tumor, node, metastasis tumor staging system Tet-operator controlled H-RAS G12V transgene Tetramethylrhodamine isothiocyanate Tet-operator controlled Wnt1 transgene Unit Microliter Micrometer Int-1 homolog of Wingless Wild type Yellow fluorescent protein xvi

17 Preface. The following individuals contributed work to this dissertation: Chapter 2. Figures 2.7b, 2.8b and 2.9b: Shelley A. Gestl assisted with preparations of mammary epithelial cell suspensions and flow cytometric analysis. Chapter 3. Figure 3.6: Kristin A. Plichta performed preliminary experiments (not shown) instrumental to developing the chimeric mammary organoid technique. Chapter 4. Figure 4.4d: Shelley A. Gestl performed the Northern blot. Figure 4.11a and 4.11b: Travis L. Leonard performed the experiments described. Figure 4.12: Travis L. Leonard performed the quantitative PCR analysis. Figure 4.13: Travis L. Leonard performed the immunohistochemistry. Text was co-written by Allison S. Cleary and Edward J. Gunther. xvii

18 Acknowledgements. I want to take this opportunity to express my gratitude to all of those individuals who helped me throughout my graduate career. First, I d like to thank my advisor, Edward Gunther, whose mentorship has been essential to my continuing development as a scientist and doctor. During my time at Penn State, I ve had the pleasure of working with numerous excellent faculty members, both clinical and academic. However, I ve seldom encountered individuals as dedicated, motivating, and nurturing as Ed. His enthusiasm for research is contagious, and has only strengthened my own love of science. Ed is a very involved advisor who likes to see the primary data hot off the bench; as such, he has been there to share in all of the day-to-day triumphs and struggles throughout the entire course of my project. I ll never forget the day in the flow cytometry lab when we both sat and watched the computer monitor as the first chimeric tumor to demonstrate a luminal swap sorted, and how we knew, right then, that we had stumbled onto something big. Also, as a budding physician scientist, being able to observe how Ed balances his clinical and research duties has been invaluable to me. I couldn t have asked for a better experience, and I feel extremely lucky to have such an outstanding mentor, role model and friend. Next, I d like to thank the members of the Gunther lab, both past and present, for contributing to such an open and collaborative work environment and for making my time in the lab so enjoyable. In particular, my thanks go to Shelley Gestl and Travis Leonard for all of their day-to-day help and for teaching me the bulk of my bench skills. Shelley and Travis are not only excellent scientists, but xviii

19 they re wonderful people, and it s been a pleasure getting to know both of you and your families. Thanks go to the members of my thesis committee for their expert guidance. Thank you for attending all of my seminars, for providing constructive feedback, and for all of your suggestions that helped shape my project. I appreciate all of the time and effort you contributed to my education and growth as a scientist. For technical support, I d like to thank Lynn Budgeon for all of those tissue sections, and David Stanford for beautiful DNA sequences. I could not have completed my project without Nate Sheaffer and Joe Bednarczyk who spent countless hours helping me to sort tumor cells into various populations. Extra thanks go to Joe who would often come in early or stay late to help with my experiments, and who once took parts out of another flow cytometer machine to fix the one I needed so that my experiment wouldn t be ruined. Also thanks to Jeannete Mohl and the rest of the animal husbandry staff for taking such good care of my mice. To all of the great friends I ve made along the way, especially Darrin Bann, Marie Bulathsinghala, and Diana Tacelosky who have been on this journey with me since the very beginning, thank you for your support and camaraderie. We ve worked really hard, but we ve also had a lot of fun. It s been a long road (and we re not done just yet), but I never could have gotten this far without you guys. xix

20 Finally, I d like to extend my deepest gratitude to my wonderful family for their unending love and support. To Billy and Jackie who have always been rooting in my corner: thank you for all of the encouragement, for keeping me grounded, for sharing your own passions and achievements with me, and for making each visit back home feel like I never left. And to my parents, who have given me so much, who fought for my dreams, who believed in me when I didn t believe in myself, who taught me that hard work pays off and to always give my best, who were there to pick me back up when I stumbled and who cheered the loudest at each milestone along the way, who valued education, encouraged my curiosity, and instilled in me a deep love of learning: for all this and more I will be forever grateful. xx

21 Chapter 1 Literature Review 1

22 1.1 Introduction. The mammary gland is a complex secretory organ composed of numerous heterogeneous cell types. Paracrine interactions between these various cell types orchestrate the profound morphologic and functional changes that characterize normal mammary gland physiology. The marked heterogeneity within the normal mammary gland translates into breast cancers that also display a startling degree of diversity. Indeed, breast cancer is a heterogeneous disease, comprising several clinically and molecularly defined subtypes that differ with respect to disease progression and drug sensitivity. However, in addition to this inter-tumor heterogeneity, breast cancers also harbor remarkable intra-tumor heterogeneity. Tumor cells within individual breast cancers typically exhibit profound diversity with respect to cell morphology, proliferation rate, metastatic potential, drug sensitivity, and capacity for tumor reconstitution. Recent cancer genome studies add to this complexity; revealing multiple genetically distinct subclonal tumor cell populations coexisting within tumors 1-5. While subclonal tumor cell populations are generally depicted as selfinterested competitors, emerging evidence suggests that, for some tumors, a cooperative relationship may exist between tumor cell subclones 6,7. As such, the complexity of cellular interactions occurring within tumors may rival or surpass that of healthy tissues; however, the nature of these interactions is unknown. Using mouse models of human breast cancer, this dissertation examines how physiologic heterogeneity within the mammary gland contributes to malignant 2

23 breast disease, as well as how individual tumor growth may be influenced by cooperative interactions between heterogeneous tumor cell subclones. 1.2 The Normal Mammary Gland: Function, Structure, and Development. Mammary glands, thought to have evolved from apocrine glands in the skin 8, are one of the defining characteristics of mammals. The primary function of the mammary gland is to provide nourishment to offspring in the form of milk. Mammary glands generally develop in pairs and vary in number from 1 pair (as in humans) to 9 pairs (as in pigs). Mice have 5 pairs of mammary glands that form along two ventrally located milk lines that extend from the thoracic to the inguinal regions of the animal 9. In the mouse, the mammary gland pairs are numbered from 1-5, cranially to caudally. The mouse is a particularly attractive model system for the study of human mammary gland biology and mammary tumor development: mouse and human mammary glands are similar in structure and function 10 and share common developmental and hormonal response features 11. Versatile mouse genetics and tissue transplantation techniques 12,13 create an exceptionally powerful experimental system. Additionally, mouse mammary gland tissue is abundant and easily accessible, further fostering experimental manipulation. The mammary gland is a complex secretory organ composed of numerous heterogeneous cell types. The branched ductal structure of the gland is made up of two distinct epithelial lineages: the luminal epithelial cells which line the ducts 3

24 and form the secretory alveoli responsible for producing milk, as well as the basal, or myoepithelial, cells which have contractile properties and are responsible for milk let-down. Mature mammary ducts have a bi-layered architecture consisting of an inner layer of luminal epithelial cells surrounded by an outer layer of basal epithelial cells. The non-epithelial elements of the mammary gland collectively form the mammary fat pad which surrounds the ductal network. These include the adipocytes, stromal elements including fibroblasts, vascular endothelium, as well as transitory populations of immune cells including macrophages and lymphocytes 9. The mammary gland is unique in that the majority of its development occurs postnatally (Figure 1.1). Mouse mammary gland development begins midgestation at embryonic day 10 (E10.5) when the milk lines appear as a thickening in the surface ectoderm By E12, five pairs of mammary placodes become visible along the milk lines as symmetrically positioned epithelial discs 14. These lens-shaped placodes develop into distinct epithelial buds by E13.5 which begin to elongate into the mesenchyme of the precursor fat pad by day E16. Ductules with a hollow lumen form, and rapid proliferation of epithelial cells produces a rudimentary ductal tree. At this stage (E18), mammary gland development arrests and does not resume until the onset of puberty 15. When the mouse is about 3 weeks old, mammary development recommences with a period of ductal elongation. With the onset of puberty, an increase in ovarian hormones, particularly estrogen, causes rapid expansion of the mammary rudiment into a branched, tree-like, ductal network that extends to fill 4

25 the entire mammary fat pad. During this period, specialized, highly proliferative structures called Terminal End Buds (TEBs) are located at the tips of the growing ducts (Figure 1.1). TEBs are bulbous, club-shaped structures that have multiple layers of epithelial body cells forming the inner core and a single outer layer of pluripotent stem cells, called cap cells, along the leading edge 9. Extensive proliferation in the TEBs allows for ductal elongation, and bifurcation of TEBs forms the branches of the ductal network 15. Apoptosis occurring within the innermost body cells is thought to be responsible for lumen formation 17. When ducts reach the outer limits of the mammary fat pad, proliferation decreases, TEBs regress and growth ceases. The pubertal growth phase of the mammary gland is complete by weeks of age. Upon pregnancy, the adult mammary gland undergoes another phase of dramatic epithelial expansion as the gland prepares for milk production and lactation. A rise in serum progesterone early in pregnancy causes a surge of epithelial proliferation and extensive ductal side-branching 18, markedly increasing the epithelial to adipocyte ratio within the gland. Prolactin signaling later in pregnancy, together with progesterone, induces the formation of alveolar units 19,20 composed of specialized alveolar epithelial cells that make up the secretory apparatus of the gland (Figure 1.1). Alveolar buds are composed of a single layer of secretory alveolar epithelial cells surrounded by a discontinuous layer of basal epithelial cells. A couple of days prior to parturition, the alveolar cells begin secreting milk proteins and lipids 9,21,22. Following parturition, an increase in oxytocin causes the basal epithelial cells to contract in order to force milk out of 5

26 the alveoli and into the ducts 23,24. Lactation continues as long as the pups continue to suckle. Involution is the final stage of mammary gland development as the ductal morphology returns to a state largely resembling that of the pre-parous gland (Figure 1.1). Involution is marked by considerable cell death and remodeling as alveolar epithelial cells undergo apoptosis and alveoli collapse Remarkably, this sequence of gland restructuring is recapitulated with each successive pregnancy. 1.3 Mammary Gland Disease. Mammary gland disease can generally be classified as either inflammatory or neoplastic. The inflammatory diseases typically involve bacterial infection or abscess formation within the ductal epithelium, and are associated with lactation and cigarette smoking, respectively 28. Neoplastic lesions in the mammary gland can be either benign or malignant, with carcinoma of the mammary gland representing the most common malignancy. In humans, breast carcinoma is the most common non-skin malignancy among women, and represents the second leading cause of cancer deaths. In 2013, there were an estimated 232,340 new breast cancer cases and 39,620 breast cancer deaths in the United States alone 29. Women have an approximately 12% life-time risk of developing breast cancer; that is, one in every eight women will develop a breast cancer at some point in her life. The most common risk factors for breast cancer include: gender, increasing age, genetics and family history, breast density, 6

27 and reproductive history 28,29. Other risk factors include: estrogen exposure, radiation exposure, diet and exercise, duration of breast feeding, alcohol, and exposure to environmental toxins 29. While much less common, men can also develop breast cancers. Men have a life-time risk of 0.11% for developing breast cancer, and there are approximately 1500 new cases and 400 deaths each year 28. Risk factors for male breast cancers include: genetics and family history, exposure to estrogens, exposure to radiation, increasing age, decreased testicular function (i.e. Klinefelter syndrome), infertility, and exposure to environmental toxins 28. Recently, a cluster of more than 80 cases of male breast cancers 30 was identified among marines that were stationed for any length of time between 1953 and 1987 at Camp Lejeune, a military base in North Carolina. This increase in breast cancer incidence has been attributed to exposure to trichloroethylene (TCE)- contaminated drinking water 31, Hormonal Regulation of the Mammary Gland. The ovarian steroid hormones, estrogen and progesterone, play a major role in normal mammary gland physiology, as well as mammary gland pathology. Steroid hormones transduce chemical signals through their cognate intracellular receptors. Nuclear Hormone Receptors (HRs), such as the Estrogen Receptor (ER) and Progesterone Receptor (PR) act as transcription factors, and each contain 3 main functional domains: the DNA-binding domain, the activation function 1 (AF1) domain, and the activation function 2 (AF2) domain 33,34. The 7

28 DNA-binding domain contains 2 zinc-finger motifs that interact directly with the DNA at specific Hormone Response Elements (HREs) located within the promoter regions of various downstream target genes 35. The AF1 domain is responsible for ligand-independent transcriptional activation, while transcriptional activation by the more potent AF2 domain depends upon ligand binding 36,37. Expression of ER and PR is limited exclusively to the luminal epithelial cells within mammary ducts and has a heterogeneous pattern: in adult glands, between 30% and 50% of luminal epithelial cells express HRs in a non-uniform distribution throughout the ducts While the degree of HR expression varies from animal to animal depending on estrous cycle stage, ER and PR are often coexpressed within the same cells. Indeed, the PR gene is a direct transcriptional target of ER 42, and as such, many progesterone mediated effects also require estrogen. This phenomena is called estrogen priming 33, Estrogen. Estrogens, most notably estradiol, are the primary female sex hormones produced by the granulosa cells during the follicular phase of ovulation. Estradiol is also produced by the placenta during pregnancy and is required for pregnancy maintenance. In the mammary gland, estrogens are thought to be responsible for the ductal elongation that occurs during puberty. In knockout mice lacking the ERα receptor (ERKO), pre-pubertal mammary gland development appears unaffected as mammary ductal rudiments are indistinguishable from timematched wild-type controls However, the glands of ERKO mice do not develop TEB structures and do not demonstrate the ductal elongation typical of 8

29 pubertal mammary gland development; instead, the glands remain as undeveloped ductal rudiments 44,45. Further, transplantation experiments where mammary epithelium from ERKO mice was embedded into the mammary fat pads of ERintact hosts that had previously been cleared of all endogenous epithelium (termed cleared mammary fat pads ; see section entitled Lineage Hierarchy of Mammary Epithelial Cells below) produced no ductal outgrowths, suggesting that ductal elongation is dependent upon epithelial ERα signaling and cannot be rescued by stromal ER expression 45. Moreover, pregnancy failed to induce ductal side branching or alveolar differentiation in ERKO grafts, suggesting a direct or indirect role for estrogen signaling in the mammary gland during pregnancy. Interestingly, ductal growth was restored in bitransgenic ERKO/MMTV-Wnt1 animals that express ectopic Wnt1 in the mammary gland 44. Indeed, ectopic Wnt1 expression produced hyperplastic ductal outgrowths not only in ERKO/MMTV- Wnt1 females, but also in ovariectomized MMTV-Wnt1 females 46 and MMTV- Wnt1 males 47, suggesting that Wnt1 signaling may substitute for estrogen signaling during mammary gland development Progesterone. Progesterone is another cycling ovarian hormone, produced during the luteal phase of ovulation and maintained at high levels during pregnancy. In the mammary gland, progesterone is thought to promote the formation of ductal side branches during di-estrous and early pregnancy. Examination of mammary glands from adult PR knockout (PRKO) mice revealed normal ductal architecture, suggesting that progesterone does not play a role during pubertal mammary gland 9

30 development 48. However, treatment with ectopic estradiol and progesterone to simulate pregnancy failed to produce the lateral side branching and lobuloalveolar development typical of a pregnant gland. Because PRKO mice cannot become pregnant due to ovulation defects, transplantation assays were performed in order to confirm the role of progesterone on mammary gland development during pregnancy. In these experiments, mammary epithelium from PRKO mice was transplanted into cleared mammary fat pads of wild-type hosts who were then mated. Again, lack of epithelial PR did not inhibit mammary ductal elongation in the resulting outgrowths, but the formation of ductal side branches and alveolar units was completely absent upon pregnancy 18. Experiments with mutant mice lacking either the PR-A or PR-B isoforms determined that the effects of progesterone on the mammary gland are mediated primarily through the PR-B isoform 49,50. This is in contrast to the progesterone mediated effects of other reproductive organs such as the uterus and ovaries which act predominantly through the PR-A isoform Paracrine Signaling by Hormones in the Mammary Gland. Despite clear evidence that cell proliferation in the mammary gland is under hormonal control, 98% of proliferating mammary epithelial cells do not express HRs 38,39,51. Recent work from Catherin Brisken s group showed that progesterone-driven cell proliferation in the mammary gland occurs in two distinct waves: a small early wave of PR-positive cell proliferation followed by a much larger wave comprised of PR-negative cell proliferation, as measured by BrdU incorporation following a progesterone challenge 52. Interestingly, in 10

31 epithelium lacking cyclin D1, the initial wave of proliferating PR-positive cells was absent, while the second wave of cell proliferation was unaffected 52. This result suggests that distinct mechanisms regulate progesterone-induced cell proliferation in the PR-positive and PR-negative cell populations of the mammary gland, and that PR-positive cell proliferation depends upon cyclin D1 (Figure 1.2a). Unquestionably, the majority of progesterone-induced cell divisions occurs in PR-negative epithelial cells during the second wave of cell proliferation, despite their lack of HR expression. One possible explanation for this discrepancy could be that dividing cells down-regulate HR expression. In support of this hypothesis, it was shown that, following transactivation, the ERα protein was rapidly degraded by the proteasome in MCF-7 cells 53. Alternatively, hormonal regulation of cell division could have a paracrine signaling mechanism, such that HR-positive cells signal to neighboring HR-negative cells which subsequently divide (Figure 1.2b). Accordingly, when a mixture of ERKO and wild-type mammary epithelial cells was used to reconstitute mammary ducts in cleared fat pads, ductal elongation was restored. Moreover, when exposed to neighboring wild-type cells, ERKO epithelial cells underwent extensive cell divisions and contributed to all aspects of the resulting chimeric outgrowths 45. A similar result was seen in chimeric ductal outgrowths generated from a PRKO:wild-type epithelial admixture: in the presence of wild-type neighbors, PRKO epithelial cells contributed to side branches and formation of lobuloalveolar units 18. Indeed, paracrine signaling circuits may allow for exquisite coordination of the various 11

32 cell types present within the mammary gland during the dramatic morphogenetic changes that occur throughout development and reproductive cycles Paracrine Mediators of ER Signaling. A number of paracrine mediators for both ER and PR signaling function have been identified in the mammary gland. In the pubertal mammary gland, ERα mediated ductal elongation depends upon the EGFR ligand, amphiregulin 54,55. Amphiregulin is a membrane bound protein that is activated through cleavage by the metalloproteinase protein, ADAM Like ERKO epithelium, amphiregulinnull and ADAM17-null epithelia failed to reconstitute cleared mammary fat pads in transplantation experiments 55,57. However, administration of exogenous amphiregulin in ADAM17-null epithelium rescued ductal elongation 57. Interestingly, amphiregulin does not appear to signal through epithelial EGFR to drive cell proliferation and ductal elongation in the pubertal gland. Instead, transplantation experiments revealed that stromal expression of EGFR is required for amphiregulin-mediated ductal elongation Since epithelial ERα produces a paracrine signal to the stroma, then an additional paracrine signal back to the epithelium must be required to induce ductal cell proliferation. Attractive candidates for this stromal epithelial back-signal are insulin-like growth factor1 (IGF1) and fibroblast growth factor (FGF) family members. IGF1 mrna is highly expressed in the mammary gland stroma during puberty and can induce ductal elongation in whole-gland in vitro cultures 60,61. Specifically, IGF1 appears to be necessary for progression of mammary epithelial cells through cell-cycle checkpoints 62. Meanwhile, FGFs are required for branching morphogenesis for 12

33 several ductal epithelial organs including lung 63 and salivary gland 64, and can induce branching of mammary organoids in 3D culture independently of EGFR signaling Paracrine Mediators of PR Signaling. In the adult mammary gland, PR signaling is also facilitated through paracrine mediators, among which is the secreted signaling protein, Wnt4. The Wnt gene family regulates body patterning, cell differentiation, and cell fate. In particular, Wnt4 is important for female sex development. Wnt4 is expressed endogenously in the mammary epithelium 65, and its expression co-localizes with that of PR expression 66. Further, Wnt4 expression becomes upregulated with pregnancy and following ectopic administration of progesterone, whereas its expression is lost in mammary epithelium of PRKO mice 66. Finally, overexpression of Wnt4 produced extensive side-branching reminiscent of pregnancy, while pregnancy failed to induce side-branches in ductal epithelium from Wnt4 knockout mice 66. Together, these data suggest a role for Wnt4 as a paracrine mediator of progesterone signaling, however the specific cell types that Wnt4 acts upon remain unclear, as well as the degree of Wnt4 signaling that occurs via canonical vs. non-canonical pathways. Perhaps the most important paracrine mediator in the adult mammary gland is the Receptor Activator of NF-κB Ligand (RANK-L). RANK-L is a Tumor Necrosis Family (TNF) family member which signals through its cognate receptor, RANK, to orchestrate several physiologic processes including osteoclast differentiation and T-cell activation. Transgenic animals lacking RANK-L or the 13

34 RANK receptor showed defects in alveologenesis resulting in lactation failure, suggesting a role for RANK-L signaling in the mammary gland as well 67. Further, overexpression of RANK-L in the mammary gland produced extensive side branching in virgin glands 68. In PRKO glands, both ectopic administration of RANK-L and expression of an inducible RANK-L transgene rescued the PRKO mammary gland phenotype 69, confirming RANK-L as a downstream mediator of PR signaling. In the mammary gland, RANK-L is expressed exclusively in PRpositive luminal epithelial cells, and expression increases following a progesterone hormone challenge 50,52. PR-positive cells which produce RANK-L are distinct from, but adjacent to the PR-negative cells which express the RANK receptor and divide in response to progesterone stimulation (Figure 1.2b). Indeed, proliferation of PR-positive cells was unaffected in RANK-L-null epithelium following a progesterone challenge, while the major wave of PR-negative cell proliferation was all but eliminated 52. Thus, the paracrine RANK-L signal plays a critical role specifically in progesterone-driven proliferation of the PR-negative epithelial cells within the mammary gland. Furthermore, paracrine mediators of progesterone signaling are implicated in the regulation of adult Mammary Stem Cell (MaSC) homeostasis. MaSCs that have multi-potent potential and self-renewal capabilities are enriched within the basal epithelial compartment (discussed further in the section below entitled Lineage Hierarchy of Mammary Epithelial Cells ). Despite lacking HRs, MaSCs appear to be extremely responsive to PR signaling. The MaSC pool expands during pregnancy and during the leuteal di-estrous phase of the mouse ovarian 14

35 cycle, accompanying the physiologic increase in serum progesterone that occurs during those reproductive stages 70. Administration of exogenous progesterone similarly amplified MaSC numbers while ablation of ovarian hormones through ovariectomy diminished the MaSC pool 70,71. Upregulation of the RANK receptor in MaSCs along with upregulation of RANK and Wnt target genes suggests that RANK-L and Wnt4 may also be important paracrine activators of the MaSC niche (Figure 1.2b) Hormones and Breast Cancer. Clearly, the steroid hormones, estrogen and progesterone, have an indispensable role in normal mammary gland development and homeostasis; however, they play an equally large role in human breast cancer. Reproductive history and lifetime hormone exposure are among the strongest risk factors for breast cancer 71-73, along with age, genetics, and breast density. Accordingly, breast cancer risk increases with the number of ovarian cycles a woman has completed, and is also increased by the use of oral contraceptives and hormone replacement therapy. Despite providing a long-term protective effect, pregnancy produces a short-term increase in breast cancer risk as well. Approximately twothirds of human breast cancers are positive for HR (ER and/or PR) expression: that is, at least one percent of the cells examined within a tumor expresses HRs. The HR status of a tumor can have profound prognostic power, as many HRpositive breast cancers respond well to anti-hormonal therapies. While, like in the mouse, the majority of proliferating cells in normal human breast epithelium do not express HRs, an increased proportion of 15

36 proliferating cells are HR-positive in pre-neoplastic lesions and overt breast tumors. Because of this, a shift from paracrine to cell-autonomous HR signaling is thought to contribute to neoplastic transformation in HR-positive mammary tumors. Supporting this observation, cyclin D1 overexpression and amplification correlate strongly with HR-positivity in breast tumors However, despite this shift to autonomous HR signaling, 11% of human breast cancers express RANK-L 77, leaving open the possibility for a paracrine signaling role in HRpositive breast cancers as well. Indeed, the anti-rank-l therapy, denosumab, has been suggested as a potential adjuvant breast cancer therapy, in addition to its current role as a treatment for osteoporosis. Unlike humans, the vast majority of spontaneous mammary tumor models in mice are hormone-independent and do not express HRs. Yet, in a progestindependent, carcinogen-induced tumor model 78, deletion of the RANK receptor in the mammary gland markedly decreased tumor incidence and increased tumor latency 77. Similarly, overexpression of RANK in this model accelerated tumor onset, while pharmacologic treatment with a RANK-L inhibitor attenuated tumor formation. Together, these data suggest that the physiologic RANK-L/RANK paracrine signaling axis was maintained through tumorigenesis, and may play a role in progestin-driven mammary cancer. Finally, because hormones, particularly progesterone, have been shown to regulate the adult MaSC niche 70,71, MaSC have been implicated as putative targets for breast cancer initiation. Specifically, cycling of MaSC with each ovarian and reproductive cycle may allow for the accumulation of mutations necessary for 16

37 neoplastic transformation. Distinctively, it is not uncommon for HR-positive breast cancers to recur ten or even fifteen years following the initial diagnosis, suggesting that, in these patients, residual malignant cells persist in a dormant but reactivatable state. In mouse studies of dormant mammary cancer, when residual disease lesions containing dormant malignancy were transplanted into cleared mammary fat pads of host mice, complete reconstitution of the mammary ductal epithelium was observed, indicating the existence of multipotent MaSC within the dormant malignant tissue 79. These studies indicate that so-called breast cancer stem cells (discussed further in the section below entitiled Cancer Stem Cell Hypothesis ) may arise from transformed mammary stem cells. Accordingly, a small population of basal-like ER/PR-negative stem cells was identified in xenografts and 3D clonogenic assays of human ER/PR-dual positive breast tumors 80. What s more, these HR-negative basal stem cells expanded following treatment with progestins, while further in vitro studies indicated that this HRnegative stem cell population was more resistant to chemotherapeutic and endocrine therapies. For these reasons, MaSCs have been identified as a potential target for the development of novel chemotherapeutic as well as chemopreventative agents for breast cancer. 1.5 Lineage Hierarchy of Mammary Epithelial Cells. The mammary gland consists of two main epithelial subtypes: the luminal and basal epithelial cells. The luminal subtype can be further subdivided into the ductal luminal cells and the milk-producing alveolar cells. Each of these mammary epithelial subtypes represents a distinct epithelial lineage, and 17

38 consequently expresses unique epithelial markers. The dramatic expansion of mammary epithelium that occurs during puberty and pregnancy, along with the significant regenerative capacity accompanying each successive reproductive cycle strongly implicates the presence of a long-lived population of adult stem cells within the mammary gland. Adult stem cells represent a small population of undifferentiated cells that participate in postnatal development and tissue homeostasis. Adult stem cells have two defining characteristics. First, they have multipotent potential; adult stem cell progeny are capable of differentiating into a diverse range of specialized cells. Second, adult stem cells are capable of selfrenewal. That is, they are able to complete numerous cell division cycles while maintaining their undifferentiated state, thereby preserving the stem cell population within a tissue 81. While traditionally thought to be a relatively quiescent population, stem cells can be activated during particular life-cycle stages or following injury. Once activated, stem cells are thought to undergo an asymmetric division in which one daughter cell differentiates and the other daughter cell replaces the undifferentiated parent stem cell 82,83. Support for the existence of adult MaSCs comes from early mammary gland transplantation experiments. Indeed, the mammary gland transplantation assay pioneered by De Ome and colleagues 13 in the 1950s has become the gold standard assay for evaluating MaSC activity. The assay involves removing or clearing all ductal epithelium from pre-pubertal host mammary glands, while leaving an intact stromal environment. Mammary gland stromal elements include fibroblasts, endothelial cells, macrophages, and adipocytes, and are collectively 18

39 referred to as the mammary fat pad. Epithelial explants from syngenic donor mice consisting of small mammary ductal fragments 84,85 or mammary epithelial cell suspensions 86 are then transplanted into the cleared mammary fat pads, and are capable of regenerating new complete and functional mammary ductal trees. Additional transplantation experiments demonstrated that stem cell activity is present along the entire ductal tree 84, as well as throughout each developmental stage 87. Transplantation of retrovirally tagged donor MECs into cleared mammary fat pads resulted in clonal ductal outgrowths, suggesting that the reconstituted glands originate from a single cell 88. Furthermore, complete ductal reconstitution was observed upon injection of a single microscopically visualized MEC into a cleared mammary fat pad 89,90. Ductal outgrowths contained mammary epithelial cells from all three mammary epithelial lineages, displayed full developmental potential, and contained cells with equal in vivo repopulating activity upon serial transplantation. These experiments provide direct evidence for a population of adult MaSCs capable of multi-lineage differentiation and self-renewal. Within the last several years, protocols have been developed that prospectively enrich for MaSC populations by fluorescence- activated cell sorting (FACS) according to specific cell surface markers. In these experiments, single cell suspensions of MECs were stained with antibodies to various cell surface markers, and then separated into distinct populations by FACS according to their cell surface marker phenotypes. The different cell populations were then injected at limiting dilutions into cleared mammary fat pads and evaluated for their ability to repopulate the mammary ductal epithelium. The Linˉ, CD24 +, CD29 hi subset of 19

40 MECs was identified as being highly enriched for MaSCs activity 89. Additional studies have further defined the MaSC subset as Lin -, CD24 +, CD29 hi, CD49f hi, EpCAM +, Sca-1 -. This sorted population of MECs is not a pure population of MaSCs; rather it contains relatively few MaSCs (<5%) along with mature basal epithelial cells, and basal progenitor cells. Because MaSCs share their cell surface marker phenotype with basal epithelial cells, they are thought to occupy a basal position within mature mammary ducts 89,90. Because a single, prospectively isolated MaSC can regenerate a complete and functional ductal tree consisting of all three epithelial lineage subtypes 89, the mature cells of the mammary ductal epithelium are thought to derive from an epithelial differentiation hierarchy. Borrowing techniques from the hematopoietic stem cell field, researchers have identified differentiation intermediates along the mammary epithelial hierarchy by isolating different cell subsets based on expression of cell surface markers and then testing their differentiation potential via orthotopic transplantation and in vitro clonogenic assays. Thus, a putative epithelial hierarchy has been constructed with MaSCs at the top followed by several intermediary epithelial cell progenitors, which respectively give rise to the terminally differentiated cells of the 3 mammary epithelial lineages (Figure 1.3). While the differentiation potential for the individual progenitors varies, they differ from MaSCs in that they have a limited capacity for self-renewal. Currently, the precise number and nature of intermediate progenitors is unknown. Several putative luminal progenitor populations have been identified to date 91, although the properties of these luminal progenitors appears to vary slightly depending 20

41 upon the isolation protocol. Similarly, a basal progenitor cell population has been postulated, though it has not yet been isolated by this method. Certainly, the classic transplantation assays have strongly implicated MaSCs as master regulators of the dynamic epithelial expansions and reorganizations that occur during the normal stages of mammary gland development and homeostasis. However, recent lineage tracing experiments call into question the physiologic role of a multipotent MaSC. These experiments utilize mouse models that permit inducible, permanent labeling of specific cells within the mammary ducts, such that affected cells, as well as all of their progeny, retain the label for their entire lifespan. In this way, an individual cell s pedigree and differentiation history can be tracked in vivo. In one set of experiments 92, this strategy revealed that embryonic Keratin 14 (K14)-expressing cells, labeled prior to birth, gave rise to both luminal and basal epithelial lineages in the developing gland, providing evidence for a fetal multipotent MaSC. However, when postnatal K14 + basal cells were labeled, they showed strict unipotent potential and gave rise only to other basal cells. Labeled K14 + cells were long lived, surviving multiple rounds of pregnancy and involution. Similarly, labeled Keratin 8 (K8) expressing luminal stem cells were equally long-lived and produced only other luminal cells. When labeled K14 + basal cells were transplanted along with unlabeled luminal epithelial cells, they demonstrated stringent lineage restriction, as only basal cells were labeled in reconstituted outgrowths. Yet, when these same postnatal K14 + cells were transplanted alone into cleared mammary fat pads, they generated complete ductal outgrowths consisting of labeled cells from both basal and 21

42 luminal lineages 92. These results suggest that while a stem cell may be capable of multipotent differentiation, it may not necessarily exercise that capacity under normal physiologic conditions. A second lineage tracing study 93 which labeled stem cells according to expression of Axin2, a reporter gene for active Wnt/βcatenin signaling, also found evidence for long-lived unipotent stem cell populations in the mammary gland. Nevertheless, the same cells that remained unipotent through development showed bipotent capability upon pregnancy, as they gave rise to both basal and luminal components of the alveolar buds. Taken together, these data indicate that a given stem cell s differentiation potential may be highly situational and can depend largely upon the specific conditions it finds itself in. Still, a third very recent lineage tracing study, using Keratin 5 (K5) expression to label cells, does identify a population of bipotent mammary stem cells 94. These basal K5 labeled stem cells contributed both basal and luminal epithelial lineages during pubertal mammary gland development as well as during homeostatic maintenance of adult mammary ducts. However, unlike the previous study 93, separate clonal populations contributed the basal and luminal components of the alveolar units upon pregnancy 94. Obviously, much more work needs to be done in order to reconcile all of these data into a single model. What is clear, however, is the staggering degree of complexity and heterogeneity present within the mammary gland. In addition to a multipotent stem cell sitting at the apex, the mammary epithelial lineage hierarchy likely contains both long and short-lived progenitors that exhibit a spectrum for 22

43 differentiation capacity, which may further be dependent upon their immediate environment. 1.6 Breast Cancer as a Heterogeneous Disease. Considering the diversity of cell types within the normal mammary gland, perhaps the degree of heterogeneity observed among human breast cancers is not surprising. Indeed, individual breast cancers display extremely variable clinical courses and are heterogeneous with respect to disease progression and drug resistance. Clinically, breast cancers are classified into subtypes according to traditional histopathology, grade, and TNM staging systems, as well as by expression status of HRs (ER and PR) and HER2. Oncologists rely upon these clinical classifications to guide treatment decisions. Molecularly, human breast tumors cluster into several distinct subtypes based on gene expression patterns: basal, normal-like, HER2+, luminal A, luminal B, and claudin-low subtypes 95,96. While there is evidence that the different molecular breast cancer subtypes correlate with clinical outcomes 96, this stratification is rarely used to inform treatment decisions in the clinical setting. That said, there is a certain degree of overlap between the clinical and molecular classification strategies, for example: tumors in the luminal A subgroup tend to express HRs, tumors in the HER2+ subgroup tend to have amplified HER2 expression, and tumors in the basal subgroup tend to be so-called triple negative tumors which lack expression of both HRs and HER2. 23

44 Given the heterogeneity of cell types within the normal mammary gland, it is tempting to presume that the different breast cancer subtypes arise from the different mammary epithelial cell types (Figure 1.4). By way of comparison, many hematologic malignancies are classified according their cell-of-origin along either the lymphoid or myeloid branches of the hematopoietic differentiation hierarchy. Similarly, breast tumors with different properties might arise from any of the spectrum of cells in the mammary epithelial hierarchy either via transformation during a certain stage of differentiation or by way of a maturation arrest. Previous work from our own lab highlights the importance of identifying the cell-of-origin for a given tumor: expression of an identical oncogenic signal in different epithelial cell types resulted in activation of divergent effector pathways which, in turn, responded to different inhibitor drugs 97. Thus, the specific epithelial cell-of-origin may profoundly influence the drug sensitivity profile for a given neoplasia. Unfortunately, neither tumor histology nor molecular profiling can be relied upon to predict the originating cell type, as basal-type BRCA1 mutant breast tumors were found, somewhat counterintuitively, to arise from a luminal progenitor cell population Intra-tumor Heterogeneity. In addition to the diversity observed among the different breast tumor subtypes, individual tumors contain a remarkable degree of heterogeneity, both phenotypically and genotypically. This intra-tumoral heterogeneity has been appreciated since the 19 th century when Rudolf Virchow and other early pathologists noted the morphologic heterogeneity among individual tumor cells. 24

45 In the late 20 th century, Bonnet and Dick observed that only a small minority of leukemic cells was capable of reconstituting a leukemia in xenograft models 99, indicating the existence of functionally distinct cell populations within tumors. Around that same time, different chromosome visualization techniques were developed which helped to highlight the heterogeneity within tumors at a genome level. Giemsa banding in tumor cell karyotypes revealed that subpopulations of cells within the same tumor harbored different chromosomal rearrangements 100,101, indicating the presence of multiple distinct subclones. Similarly, fluorescence in situ hybridization (FISH) identified subclonal tumor cell populations based on alterations in gene copy number 102,103. Since then, advancements in gene sequencing technology have made it possible to sequence entire breast cancer genomes from bulk tumor samples 1-4 as well as from individual tumor cells 5, confirming the presence of genetically distinct subclonal populations of cells within tumors. How these different tumor cell subpopulations arise and what influence they may have on disease progression and therapeutic response are topics hotly debated among researchers, and as such, they remain very active areas of investigation. 1.8 Clonal Evolution Hypothesis. Almost without exception, tumors are thought to originate from a single cell, yet established tumors contain considerable heterogeneity. There are several proposed theories that attempt to explain how individual tumors acquire such heterogeneity over time through the process of tumor progression. Perhaps the most widely accepted model is the clonal evolution hypothesis. First proposed in 25

46 1976 by Peter Nowell in a landmark theoretical paper 104, the clonal evolution hypothesis applies the principles of classic Darwinian evolution theory to tumor progression. He suggested that tumor progression depends upon stochastically acquired genetic variability that permits sequential selective expansions of increasingly fit or aggressive subclones. This model depends upon two main assumptions from Darwin s theory of evolution by natural selection 105. First, within a population, individual variances arise which are heritable. Second, individual variances contribute to differences in survival and reproduction. In tumors, increased proliferation rates coupled with error-prone DNA repair processes and exposure to environmental mutagens (e.g. ionizing radiation, cigarette carcinogens, chemotherapeutic agents), allows for the introduction and accumulation of random genetic and epigenetic alterations within individual tumor cells. The majority of acquired mutations are deleterious to the cells and are consequently eliminated though cell extinction. Alternatively, mutations may be inert and have no known biologic significance. The inert changes are termed passenger mutations because, while they don t actively contribute to tumor progression, they do not decrease cell viability and are therefore perpetuated through successive cell divisions. Figuratively, they come along for the ride. Occasionally, however, a mutation arises that confers a selective growth advantage (for example, a mutation in a proto-oncogene or tumor suppressor gene) that leads to increased cell divisions resulting in the clonal expansion of that fitter cell. These biologically advantageous mutations are termed driver mutations because they actively promote tumor growth and 26

47 progression. As a clonal population expands, each of the cell progeny are themselves capable of accumulating additional mutations that further promote or hinder cell growth. Importantly, however, because genetic mutations occur stochastically, different cells acquire different mutations independently of one another. As such, from an initial clonal population of tumor cells, it is possible for several subclonal populations to emerge and co-evolve simultaneously Linear Evolution. Since it was first proposed, several variations of the clonal evolution hypothesis have emerged. The linear, evolution model (Figure 1.5a) suggests that while numerous heterogeneous subclones can arise, at each stage of tumor progression, a single clone ultimately out-competes the others in a selective sweep to dominate the clonal composition of the resulting tumor. Linear clonal evolution is best described by the multistep progression model for colon cancer 106 which posits that tumor progression occurs in discrete steps, each depending upon the accumulation of specific driver mutations in a particular order. Additional support is found in reports that describe monoclonal composition of tumors based on patterns of X-chromosome inactivation 107, or expression of sex-linked genes from heterozygous individuals 108,109, though these studies are limited in that they only measured a single trait. Using more comprehensive sector-ploidy-profiling and comparative genomic hybridization techniques, some breast tumors were also found to consist of only a single major clonal population

48 1.8.2 Branched Evolution. Alternatively, the branched variation of the clonal evolution hypothesis suggests that as a tumor progresses, multiple divergent clones can arise which cohabit the same tumor (Figure 1.5b). Numerous cancer genome sequencing studies provide empirical evidence in support of this model. For example, whole cancer genome sequencing of human breast cancers uncovered numerous genetic mutations that appeared to be present within only a small fraction of the cells within a given tumor 3. When all of the identified mutations within a tumor were graphed together according to the proportion of cells containing each variant, several distinct clusters were apparent. Each cluster represented a distinct subclonal population of tumor cells with its own complement of genetic mutations. Even so, several identified mutations were shared by all of the cells within the tumor, suggesting that the various subclonal populations diverged from a common ancestral clone 3. A similar result was observed in a genome sequencing study of triple-negative breast cancers: multiple mutations were detected at subclonal frequencies 4. Moreover, when deep sequencing techniques were used to analyze the genomes from matched primary and metastasis pairs from a basal-like breast cancer 2 and an ER + lobular breast cancer 1, metastases were seen to stem from a minority of cells within the primary tumor, based on mutant allele frequencies. Still, perhaps some of the most compelling evidence for the branched evolution model comes from studies in renal 111 and pancreatic cancer 112. These studies performed genome analysis on several regions taken from the same 28

49 primary tumor as well as from several matched metastatic lesions from the same patient. Not only were multiple subclonal populations of tumor cells identified, the primary tumors exhibited regional heterogeneity in clonal composition. In the renal carcinoma study 111, in addition to a set of shared mutations, there were several genetic alterations that were unique to either the primary tumor or the metastatic lesions. This pattern indicates a branched or divergent evolution where one branch evolved to produce the various clones present within the primary tumor, while the other branch gave rise to the distinct clones represented among the various metastatic lesions. In contrast, the pancreatic cancer study identified several geographically and genetically distinct subclones within the primary tumor that each contributed to anatomically distinct metastatic lesions. This result indicates that, for this tumor, the different metastatic clones co-evolved independently of one another Mutator Phenotype Hypothesis. The mutator phenotype model 113 adds a further layer of complexity (Figure 1.5c). It suggests that, due to increased genetic instability from alterations in DNA replication or repair machinery, tumor cells develop genetic mutations at a dramatically higher rate than normal cells. Indeed, the mutation rate within neoplastic tissue was shown to be 200 times greater than in normal tissue 114. This model is related to the branched evolution model, as both predict the existence of co-evolving tumor subclones. However, the mutator phenotype posits that, in addition to the driver mutations promoting clonal expansion of tumor cell subpopulations, there is also a high frequency of private, unselected mutations 29

50 that each exist within just one or a few cells within the tumor. In this way, tumors can be depicted as containing a huge number of tiny sub-subclones. Because each of these subclones is defined by the presence of a random inert mutation, they are not thought to contribute directly to overall tumor growth. However, their power rests in their potential: as environmental milieu and selective pressures change throughout tumor progression, a mutation that originally had little or no biologic relevance may suddenly confer a new selective advantage. For example, a mutation that imparts resistance to a particular drug may remain unselected during initial tumor growth, but would facilitate rapid clonal expansion upon treatment with that drug, thereby providing a ready avenue for treatment resistant relapse Self-Seeding Hypothesis. Still another variant of the clonal evolution hypothesis, the self-seeding model (Figure 1.5d) posits that a certain degree of heterogeneity within tumors can be credited to inappropriate tumor cell mobility. It superimposes Steven Paget s historical seed and soil hypothesis 115 on the process of tumor progression. The seed and soil hypothesis addresses the issue of metastatic tropism, suggesting that particular tumor types routinely metastasize to prescribed organs because those organs provide a uniquely well suited environment to support growth of that tumor type (for example, breast cancers preferentially metastasize to bone, brain, liver and lung). Subsequently, the self-seeding tumor model proposes that cells from a primary tumor can escape, circulate, and even take up residence at a distant site before returning to and rejoining the favorable environment of the original tumor 116. In this way, a tumor mass can be viewed as 30

51 a conglomerate of contiguous small growths made up of increasingly aggressive clones, each having withstood a round of dissemination. Support for this model was seen in clever experiments that used a metastasis-derived cell line to generate xenograft tumors within the mammary glands of an immunodeficient mouse. Fluorescently labeled cells were used to generate a tumor in one gland, while an otherwise identical population of unlabeled cells was used to generate a tumor in the paired contralateral gland. Interestingly, subclonal populations of fluorescently labeled cells were detected in the unlabeled tumor from the contralateral side, suggesting contribution from a circulating tumor cell population Cancer Stem Cell Hypothesis. The different variations of the clonal evolution hypothesis all share the same assumption: that each of the cells within a tumor has equal capacity for proliferation. However, ample evidence suggests that this may not be true. Instead, numerous studies indicate that only a small subfraction of cells within a given tumor are capable of growing in in vivo xenograft models 99,118 or in vitro clonogenic assays This functional heterogeneity was re-emphasized in 1997 when Bonnet and Dick found that less than one in 10,000 cells from a human acute myeloid leukemia was able to recapitulate the disease upon transplantation to an immunodeficient mouse xenograft model 99. They proposed that the functional heterogeneity they observed was the result of an intrinsic cell hierarchy with a tumor initiating cell, or cancer stem cell, sitting at the top. Since then, putative cancer stem cell populations have been prospectively identified for a 31

52 number of different solid malignancies, including breast cancer 118, lending support for the stem cell theory of cancer. The cancer stem cell theory posits that tumor growth is driven by a small population of tumor-initiating cells that possess stem-cell like properties 122 (Figure 1.5e). Specifically, these cancer stem cells are thought to have extensive proliferative capacity, self-renewal properties, and give rise to the various, more differentiated, tumor cells that make up the bulk of the tumor. While it is important to emphasize that cancer stem cells do not all necessarily arise from normal tissue stem cells, the longevity of normal tissue stem cells makes them excellent candidates: their long lifespan provides more opportunity for accumulation of the mutations necessary for tumorigenesis. Moreover, growing evidence suggests that normal tissue stem cells and cancer stem cells may be similarly regulated 123. Cancer stem cells are an experimentally defined tumor cell population with a functional, rather than descriptive, phenotype: they are identified according to their behavior in a number of in vitro and in vivo assays to test for differentiation potential, self-renewal, and tumorigenic potential. There is no unique cancer stem cell marker. Different combinations of cell surface markers have been used to isolate cancer stem cell enriched subpopulations within a variety of tumor types. However, tumor-initiating activity is often detected in the un-enriched population as well, albeit at lower frequency. The CD44 + /CD24 - tumor cell sub-population has been identified as a putative cancer stem cellenriched population for a variety of human breast cancers, including invasive 32

53 ductal, invasive lobular, and inflammatory breast cancers 118. While tumor cells of alternate phenotypes failed to form xenografts in mice when injected at high concentrations, as few as 100 cells from the CD44 + /CD24 - sub-population was sufficient to form tumors that could be serially passaged. Despite the body of the work that attempts to identify cancer stem cellenriched cell populations in human tumors, some of the most convincing evidence for the cancer stem cell model comes from studies in mice. In one study 124, a fluorescent YFP reporter was used to indelibly label prominin1 + intestinal stem cells at the same time expression of an oncogenic β-catenin allele was induced in those same cells. Intraepithelial tumors that arose were composed entirely of YFP + cells that expressed nuclear β-catenin, indicating that all of the tumor cells arose from prominin1 + stem cells. However, only a small fraction of tumor cells retained expression of prominin1 while the vast majority of tumor cells were prominin1 -, consistent with a more differentiated phenotype. Thus, malignant transformation of a tissue stem cell generated hierarchically organized tumors wherein a small population of transformed stem cells gave rise to the more differentiated bulk tumor cells. Further, in the MMTV-Wnt1 mouse model of mammary tumorigenesis, constitutive expression of Wnt1 generates mixed-lineage tumors, containing both basal and luminal epithelial lineages. Mammary gland hyperplasias from this model exhibited expansion of a mammary progenitor cell population 125,126, suggesting that malignant transformation may preferentially arise within a bipotent progenitor cell. Supporting this idea, some tumors arising in a Pten 33

54 heterozygous background showed loss of the wild-type Pten allele from all tumor cells 125. Since it is unlikely for identical mutations to arise in both basal and luminal tumor cell subsets independently, a single loss-of-heterozygosity event likely occurred in a common progenitor to the two epithelial subpopulations. Finally, limiting dilution transplantations of tumor cell subsets revealed a small subpopulation of progenitor cells with enhanced tumor initiating capacity that recapitulated tumors with the same mixed-lineage character as the original tumors Together, these data suggest a hierarchical tumor organization with a bipotent tumor initiating cell at the top, giving rise to the basal and luminal epithelial lineages represented in the bulk tumor Clinical Implications of the Distinct Tumor Cell Heterogeneity Models. The variations of the clonal evolution hypothesis together with the cancer stem cell hypothesis are currently the leading models to explain the source of heterogeneity within tumors. However, it is unlikely that these are the only possible explanations. Similarly, it is unlikely that these models are each mutually exclusive of each other. Instead, given the clinical diversity observed between tumors, it is very possible that every tumor is organized differently, and subsequently exhibits the properties of one or any combination of these different models. This becomes especially relevant when considering the best treatment approach for a given tumor, since each of these models of tumor progression has different clinical implications. Indeed, moving forward, treatment success may 34

55 depend upon being able to correctly classify a tumor s organization based on its unique portrait of heterogeneity. The linear and branched evolution models depend on extensive cell proliferation. As such, use of traditional cytotoxic and antiproliferative chemotherapeutics may be most appropriate for treating tumors organized according to these models, with the ultimate goal of eradicating as many tumor cells as possible. Adding an additional layer of complexity, the mutator phenotype suggests that small drug resistant clones may already be present within tumors prior to treatment. Likewise, considering the inherently mutagenic nature of many chemotherapeutic drugs, the mutator phenotype increases the probability that a given treatment will itself induce drug resistance mutations. Accordingly, effective therapy for tumors of this type may require treatment with a combination of multiple agents, each with a different mechanism of action. Conversely, the self-seeding model identifies aberrant cell motility as the primary driver of tumor growth and disease progression, and it places emphasis on the tumor environment. Thus, instead of antiproliferative agents, a self-seeding tumor may respond better to agents that reduce cell migration or create an inhospitable tumor microenvironment, for example, antiangiogenic therapy. Further, because the self-seeding hypothesis contends that primary tumors act as figurative magnets for circulating tumor cells, it predicts, somewhat counterintuitively, that surgical removal of a primary tumor may increase the potential for metastatic deposition. 35

56 Finally, the cancer stem cell model suggests that tumor growth is driven by a small subpopulation of tumor cells with unique tumor initiating capacity, while the bulk of the tumor cells are relatively non-virulent. Traditional chemotherapeutics may eliminate the bulk of the tumor cells, but may not effectively eliminate the cancer-initiating stem cells, which have been shown to be particularly resistant to existing therapies 130. Failure to eliminate the cancer stem cells would ultimately result in tumor recurrence and disease relapse. Therefore, an ideal treatment would specifically target the cancer stem cell population. However, because cancer stem cells may themselves evolve, a combination of multiple cancer stem cell-targeted agents may be necessary to eradicate a hierarchically organized tumor Tumor Cell Inter-clonal Interactions. Regardless of the source of heterogeneity within tumors, the presence of distinct tumor cell subclones raises an interesting question: do cohabitating subclones interact? Interactions between the tumor cell compartment and the surrounding tumor environment are well established. Indeed, increasing evidence depicts tumors as complicated bionetworks, in which tumor cells interact not only with their surroundings, but also with each other. In this way, intra-tumor clonal heterogeneity may not arise solely as a byproduct of tumor progression, but may instead provide a functional role in tumor biology. Gloria Heppner first introduced the idea of interacting members of a cancer cell society 6 after observing that different subpopulations of tumor cells behaved differently when cultured separately than when cultured together. As such, she was the first to 36

57 apply the principles of population biology to tumor cell subpopulations. Under this paradigm, distinct tumor cell subclones can be considered different species 7 or game players 131 which may interact with each other and the surrounding environment according to the same population dynamics and ecological relationships observed in nature. Like in ecology, potential interactions between heterogeneous subclones can be oppositional or cooperative, and multiple different interactions may occur simultaneously within an individual tumor Exploitation or disruption of these potential relationships may represent novel opportunities for intervention Oppositional Relationships: Competition. Since carcinogenesis involves a breakdown in the cell-cell cooperation that normally maintains epithelial tissue architecture, individual subclones within a malignant microenvironment are most commonly depicted as self-interested competitors. Limited space and availability of oxygen, nutrients, and growth factors increases competition between different subclonal populations within the same tumor niche. Gause s principle of competitive exclusion 132 predicts that direct competition between species produces discrete winners and losers, where one species eventually out-competes the others. In tumors, this phenomenon is observed in instances of selective sweeps by positive selection of a dominant tumor cell clone. Currently, directed treatment strategies are often targeted to the dominant clone within a tumor. However, these treatments frequently result in disease relapse, as depression of the dominant clone may increase the competitiveness of a previously under-represented clone. 37

58 Oppositional Relationships: Antagonism. Examples of antagonistic relationships between tumor cell clones have also been described. The doubling rate for tumor cells is orders of magnitude faster than that of normal cells, yet even the most aggressive tumors demonstrate Gompertzian growth kinetics 133,134 in which growth rates ultimately reach a plateau instead of continuing to increase exponentially. This suggests that a large proportion of tumor cells either die 135 or are prevented from proliferating by the surrounding microenvironment or by competing clones. In one study, inhibition of Src family kinase (SFK) in the surrounding stroma prevented proliferation and re-growth of dormant breast cancer cells 136. In an example of clonal interference between tumor cell populations, tumor cell secretion of the growth inhibitory protein, TGF-β, prevented growth of some surrounding tumor cell clones while selecting for clonal outgrowth of the more aggressive cells that could evade the growth regulatory effects of TGF-β 137, Cooperative Relationships: Commensalism. Commensalism refers to a positive interaction in which one party benefits, while the other remains unaffected. A classic example of commensalism in tumors is the interaction between the tumor cells and the surrounding stroma. Reactive or desmoplastic stroma is typical of wound healing and is characterized by fibroblast proliferation, immune infiltration, angiogenic induction, and altered extracellular matrix. These same changes are often induced within the activated stroma surrounding tumors and can act as promoters of tumor growth. Remarkably, work from Mina Bissell s group suggests that correction of 38

59 abnormal signals from the stromal compartment can revert the malignant phenotype of tumors 139. However, cooperative interactions can also occur within the tumor cell compartment, itself. All cancers are thought to exhibit several hallmark phenotypes: sustained proliferative signals, insensitivity to growth suppressors, resistance to programmed cell death, replicative immortality, ability for invasion and metastasis, induced angiogenesis, switch to glycolytic metabolism, and evasion of immune destruction 140,141. It is generally assumed that a single cell must acquire all of these hallmark traits to undergo malignant transformation. However, acquisition of certain traits such as angiogenic capability 142,143 or immune evasion , which depend upon secreted factors, benefit not only the secreting cell, but all of the surrounding cells as well. In this way, a cancer might collectively achieve all of the hallmark traits through contributions from several different cooperating cells (clones). A specific example of cooperation between tumor cell populations was observed in a model of glioblastoma 148. In these experiments, a human glioma cell line was engineered to either overexpress wildtype epidermal growth factor receptor (wtegfr) or a constitutively active mutant EGFR variant (ΔEGFR). When an admixture of these two cell populations was engrafted intracranially, tumor formation was enhanced compared to either cell population alone. Additional experiments determined that cells expressing the ΔEGFR variant unidirectionally stimulated increased proliferation of the wtegfr overexpressing 39

60 cells through secretion of cytokines, thus actively maintaining subclonal heterogeneity within the resulting tumors. Another example was seen in a mouse model of small cell lung cancer 149. Within individual tumors from this model, two phenotypically distinct cell populations were identified which expressed either neuroendocrine or mesenchymal markers. The neuroendocrine cells showed a proliferative advantage when cultured together with the mesenchymal cells. Further, metastatic lesions were observed only when an admixture of the two populations was engrafted onto host animals, but not when either population was engrafted alone. Interestingly, the metastatic lesions contained only cells of neuroendocrine character. So, enhanced metastatic capability was endowed upon the neuroendocrine population as a result of its interaction with the mesencymal cell clone Cooperative Relationships: Mutualism. Mutualism describes a positive interaction in which both contributors benefit. This type of interaction implies a co-dependency. In a tumor, the existence of co-dependent subclones would result in active maintenance of heterogeneity as a stable property. Conversely, uncoupling such an interaction could simultaneously eliminate both clones. In a fruitfly tumor model 150, cells were engineered to express either an oncogenic Ras allele, or a loss-of function mutation in the scribbled tumor suppressor. Neither mutation alone led to tumor formation. However, separate expression of the two mutations in two adjacent cells resulted in aggressive tumor growth through upregulation of JAK/STAT- 40

61 activating cytokines. Downregulation of the JAK/STAT receptor Domeless interrupted the cooperative interaction and suppressed tumor growth 150. Thus tumor growth depended upon a sustained mutualistic interaction between two genetically distinct subclones. Yet, while interclonal cooperation can drive tumorigeneisis in fruitfly models, definitive evidence for functional cooperation between co-dependent tumor cell subclones in mammals is lacking Overview. Using mouse models of human breast cancer, this dissertation examines whether and how breast cancer heterogeneity derives from the diverse and highly interactive cell subtypes that comprise the normal mammary gland. Chapter 2 describes experiments directed at uncovering the origin of tumor subtypes and explores how transformation of a specific mammary epithelial cell type can influence tumor phenotype. Utilizing a novel chimeric mammary organoid model system, Chapter 3 attempts to understand the mechanisms of interclonal cooperation by detailing the cell autonomous and non-autonomous effects of oncogene expression in mammary epithelial cells. Finally, Chapter 4 explores how diverse tumor cell subtypes are maintained within mammary cancers by investigating functional cooperation between heterogeneous tumor cell subclones. 41

62 Figure 1.1: Stages of postnatal mammary gland development. Schematic of postnatal development within the mammary gland. Corresponding pictures are carmine stained whole mounts of representative glands at each indicated developmental stage. Partial lymph node in each picture serves as a landmark. LN: lymph node; MFP: mammary fat pad; TEB: terminal end bud; AB: alveolar bud; G: number of pregnancies; P: number of full-term births, dpc: days post coitum; inv: involution. All pictures taken at same magnification, size bar=500 μm. 42

63 Figure 1.2: Cell autonomous and paracrine mechanisms of progesterone- induced cell proliferation in the mammary gland. a. Progesterone can act directly on HR-positive cells to induce proliferation via a Cyclin D1-dependent mechanism. b. Progesterone can also induce expression of secreted RANK-L in HR-positive Sensor cells which stimulates proliferation within the adjacent HR-negative Responder cells. Figure modified from Brisken C (2013). Nat Rev Cancer. 13(6): Adapted by permission from Macmillan Publishers Ltd. 43

64 Figure 1.3: Models of the epithelial lineage hierarchy in the mammary gland. A multipotent mammary stem cell at the top of the hierarchy gives rise to committed progenitors of the three mammary epithelial lineages: luminal, alveolar, and myoepithelial (basal). The number and nature of the progenitor cell populations is still unknown. An intermediate bipotent progenitor presumably lies upstream of the lineage-committed progenitors, though differentiation potential may depend upon environmental conditions. Figure modified from Fu N, Lindeman G, and Visvader J (2014). Curr Top Dev Biol. 107: Reprinted with permission from Elsevier. 44

65 Figure 1.4: Molecular breast cancer subtypes. In the center, the different molecular breast cancer subtypes are compared to their proposed cell of origin within the normal mammary epithelial hierarchy (left) based on gene expression signature gradients (right). Figure modified from Prat A and Perou C (2009). Nature Medicine. 15: Adapted by permission from Macmillan Publishers Ltd. 45

66 Figure 1.5: Models of tumor cell heterogeneity. Green root nodes represent normal diploid cells. Colored nodes represent different subclonal populations of tumor cells. a. Linear evolution model: a selective sweep results in a homogeneous tumor cell population. b. Branched evolution model creates tumors with several co-habitating clones c. Mutator phenotype generates tumors with multiple diverse clones. d. Self-seeding model creates tumors with a divergent peripheral subpopulation of circulating tumor cells. e. Cancer stem cell model generates tumors in which a minority population of tumorigenic cancer stem cells (pink) supporting the bulk population of tumor cells. Figure modified from Navin N and Hicks J (2010). Mol Oncol. 5(3):302. Reprinted under Elsevier User License. 46

67 Chapter 2 Mammary Tumors Initiated by MMTV-LTR- Driven Neu Transgenes Originate within a Hormone Receptor-Negative Luminal Cell Subcompartment Allison S. Cleary, Shelley A. Gestl, and Edward J. Gunther 47

68 2.1 Abstract. Human breast cancers frequently express hormone receptors (HRs), and most HRpositive tumors depend on ovarian hormones, such as estrogen and progesterone, for growth. By contrast, transgenic mouse models of breast cancers nearly always yield HR-negative, hormone-independent mammary cancers. The mechanisms underlying this bias toward HR-negative disease in mice are not known. Since the majority of mouse breast cancer models use the mouse mammary tumor virus long terminal repeat (MMTV-LTR) as a mammary-specific promoter element, we examined whether MMTV targets transgene expression to a specific mammary cell compartment. Specifically, MMTV-rtTA transactivator mice were crossed to teto-h2bgfp reporter mice, and doxycycline-dependent labeling of mammary epithelial cell (MEC) subtypes was examined in bi-transgenic females. Despite possessing several well-studied hormone response elements, MMTV-directed H2BGFP labeling was restricted to HR-negative cells in the luminal compartment. When this labeling strategy was combined with MMTV-directed expression of the Neu oncogene, Neu transgene expression was enriched within the H2BGFP labeled subpopulation. Further, Neu-initiated neoplasias were comprised entirely of HR-negative cells from the carcinoma-in-situ stage onward. Thus, MMTV-driven Neu expression targets HR-negative luminal cells, culminating in HR-negative tumors. We propose that the HR-negative phenotype of many mouse breast cancer models can be explained by MMTV-driven transgene expression in HR-negative mammary cells. 48

69 2.2 Introduction. The Mouse Mammary Tumor Virus (MMTV) was first discovered in the early 1900s when mouse breeders noticed that certain strains of mice were more susceptible to developing mammary tumors than others, and the susceptibility of a particular mouse depended upon the mother s risk Pup fostering experiments demonstrated that tumor susceptibility was passed vertically from mother to offspring through breast milk in the form of an infectious extrachromosomal or milk factor. 154 This milk factor was later identified as the MMTV retrovirus. MMTV is transmitted through ingestion of infected breast milk. Once ingested, the virus infects the B-lymphocytes within the Peyer s patches of the gut 155. Presentation of MMTV-encoded super antigen (SAg) by the major histocompatibility (MHC) class II proteins on the B-cells leads to extensive proliferation of cognate T-lymphocytes, creating a reservoir of infectioncompetent cells for MMTV dissemination 155. Infected lymphocytes then deliver MMTV to the mammary epithelial cells (MECs) 156. Upon pregnancy and lactation, infected MECs secrete MMTV virus particles into the milk, and the cycle begins again. In addition to the mammary gland, there is evidence that MMTV infects other epithelial organs such as salivary gland, and prostate 157, although malignant transformation by MMTV is generally limited to mammary tissue. 49

70 Unlike Rous sarcoma virus (RSV) or Polyomavirus (PyV) which are directly transforming viruses, the RNA genome of MMTV does not encode its own oncogene. Instead, MMTV proviral insertion in the mouse genome drives increased expression of cellular proto-oncogenes. In the 1980s, Roel Nusse and Harold Varmus discovered that the MMTV provirus most commonly inserted within a highly conserved locus on chromosome 15, leading to increased transcriptional activation of a nearby putative proto-oncogene, int-1 158,159 (later renamed Wnt1 due to sequence homology with the Drosophila gene Wingless). To test the oncogenicity of Wnt1, Varmus group made a transgenic mouse using the long terminal repeat (LTR) region from the MMTV genome as a promoter to drive transcription of a Wnt1 transgene 47. In the mouse mammary glands, ectopic expression of Wnt1 produced a dramatic lobuloalveolar hyperplasia followed by stochastic formation of mammary adenocarcinomas, confirming the role of Wnt1 as an oncogene. Since then, the MMTV-LTR 160 has been used as a tissue specific promoter element in numerous transgenic mouse tumor models to drive transgene expression in the mammary glands (Table 2.1). Like other retroviral promoters, the MMTV-LTR contains U3, R, and U5 regions (Figure 2.1). The U3 region contains the majority of the cis-acting control elements of the MMTV promoter, including the negative regulatory elements (NREs) 161, the MMTV SAg 162, and hormone response elements (HREs) 163, while the R region contains the transcriptional start site. The NREs contribute to the mammary tissue specificity of the MMTV-LTR: expression of transcriptional repressor proteins in MMTV 50

71 non-permissive and semi-permissive tissues negatively regulate MMTV expression by binding to the NREs within the LTR 164,165. Indeed, mice infected with a mutated MMTV strain which lacked the NREs within the U3 region did not develop mammary adenocarcinomas, but instead developed T-cell lymphomas 166. Additionally, there is evidence that NREs temporally regulate MMTV expression within the mammary gland. Because the MMTV virus is transmitted through the breast milk of adult animals, evolutionary selective pressure ought to be permissive toward host survival though adulthood. Accordingly, NREs mediate repression of MMTV transcription during early mammary gland development 161. Conversely, due to stimulation by ovarian hormones and downregulation of negative repressors, MMTV expression is highest during lactation when virus particles are shed into the milk. This increase in MMTV expression during pregnancy may explain, in part, why pregnancy accelerates tumorigenesis in many transgenic tumor models driven by the MMTV-LTR promoter. Additionally, the MMTV-LTR contains a very well characterized HRE that is thought to regulate expression of both endogenous MMTV viral transcripts and/or transgene-encoded oncogenes. The HRE within the MMTV-LTR has been shown to interact with a variety of steroid HRs including progesterone 167, glucocorticoid , mineralocorticoid 174,175, and androgen 167,176 receptors. However, despite its reputation as a hormonally regulated promoter, tumors arising in MMTV-LTR driven transgenic models consistently do not express HRs 51

72 and grow independently of ovarian hormones (Table 2.1). This bias toward hormone-independent mammary tumors in mice is not understood. Here, we use an MMTV-directed labeling strategy to show that, in adult animals, MMTV expression is limited to a specific sub-population of HRnegative luminal MECs. Further, in the classic MMTV-Neu mammary tumor model, Neu transgene expression was limited to this same HR-negative luminal epithelial subset, culminating in HR-negative tumors from the carcinoma-in-situ stage onward. We suggest that in other MMTV-driven transgenic models, the MMTV promoter may direct transgene expression to this specific HR-negative epithelial subset, explaining the propensity for HR-negative mammary tumors in transgenic mouse models. 2.3 Methods. Transgenic Mice. Mice were housed under pathogen-free conditions in the Pennsylvania State University College of Medicine rodent facility with access to water and standard mouse chow (Harlan Tekland, HT2018) ad libitum. All experimental protocols were approved by the Pennsylvania State University College of Medicine s Institutional Animal Care and Use Committee. The teto- H2BGFP (Tg(tetO-HIST1H2BJ/GFP)47Efu/J; stock #005104), teto-neu (B6;SJL-Tg(tetO-Erbb2*)8-4Jek/J; stock #010577), and MMTV-Neu (FVB/N- Tg(MMTVneu)202Mul/J; stock #002376) transgenic lines were obtained from the Jackson Labs. The MMTV-rtTA transgenic line was a gift from Dr. Lewis Chodosh (University of Pennsylvania, Philadelphia, PA). All mice either were 52

73 generated in an inbred FVB/N background or were back-crossed 10 or more generations with FVB/N breeders before initiating experiments. Dox was administered by replacing standard mouse chow with chow containing 2g/kg doxycycline (Bio-serv). Genotyping was performed by PCR using genomic DNA isolated from tail clips and transgene specific primers (Table 2.2). Flow Cytometry and Cell Sorting. Single cell suspensions were generated from primary mammary epithelial tissue through a series of enzymatic digestions. All ten mammary glands were dissected from mice and digested for 6 hours at 37 C in a 10% solution of Collagenase/Hyaluronidase (StemCell Technologies cat# 07912) in Dulbecco's Modified Eagle's Media/F12 (DMEM/F12; Gibco). The digested tissue was centrifuged at 550xg for 10 minutes at 4 C to pellet mammary epithelium, while the supernatant containing the collagenase/hyaluronidase solution with adipose tissue and stromal elements of the mammary fat pad was discarded. The cell pellet was resuspended in 0.25% trypsin-edta and incubated for 2 minutes at room temperature with continuous shaking. Following another 10 minute spin at 550xg, the trypsin solution was removed, taking care to preserve the stringy material floating in the supernatant. The cell pellet along with the retrieved stringy material was resuspended in 4 ml of a Dispase solution in PBS: 5mg/mL Dispase (Roche), 255 U/mL DNase I (Worthington), sterile filtered (0.2μm filter). The sample was incubated at room temperature for at least 2 minutes with shaking, and then filtered through a 0.35 μm cell strainer. Following an additional 10 minute spin at 500xg, samples were incubated for 1-2 minutes at 53

74 room temperature in a 0.64% ammonium chloride solution in distilled water to lyse any red blood cells. Dissociated tumor mammary epithelial cells were then enriched for Lin - (CD45 - / CD31 - / TER119 - / BP-1 - ) mammary epithelial cells with StemCell Technologies EasySep Mouse Epithelial Cell Enrichment Kits per the manufacturer s instructions. Lin - cells were incubated on ice for 20 min with anti- CD49f (α6 integrin) (BD Biosciences ) together with Alexafluor 647 (Invitrogen A21247) in PBS. Cells were spun down for 5 min at 550x g, then incubated with EpCAM-PE conjugated antibody (Biolegend ) in PBS. Cells were either read on a BD FACS Caliber Flow Cytometer equipped with Cell Quest Pro software or sorted on a BD FACS Aria cell sorter machine equipped with Diva software. Cells were separated into their luminal (Lin - /CD49f low /EpCAM high ) and basal (Lin - /CD49f high /EpCAM low ) subpopulations and sorted on the basis of GFP expression. Luminal GFP+, Luminal GFP-, and Basal sorted populations were collected into 5 ml falcon tubes containing PBS. Total RNA was collected from sorted cell populations using Qiagen RNeasy spin column kit. Quantitative RT-PCR. RNA was reversed transcribed using Invitrogen Superscript II First Strand Synthesis kit. We used Taqman Gene Expression Assay mix containing unlabeled PCR primers and a FAM-labeled Taqman probe to detect expression of several genes: GFP (Applied Biosystems Mr _mr), Keratin 5 (Applied Biosystems Mm _m1), Keratin 8 (Applied Biosystems Mm _m1), Progesterone Receptor (Applied 54

75 Biosystems Mm _m1), Estrogen Receptor (Applied Biosystems Mm _m1), RANK (Mm _m1), RANK-L (Mm _m1), and ErbB2 (Neu; Applied Biosystems Rn _g1). Relative quantification PCR (ΔΔCt method) was performed in triplicate using Agilent Technologies Stratagene Mx3005P detection system and analyzed using Stratagene MxPro software. Expression levels of the assayed genes in sorted cell populations were normalized to Gapdh transcript levels (Applied Biosystems E) and compared to the unsorted sample (relative expression=1). Immunofluorescence. Tumor samples were fixed in 4% paraformaldehyde on ice for 2 hrs before being paraffin embedded. Paraffin sections (5μm) were stained with antibodies for Progesterone Receptor (PR), Estrogen Receptor α (ER), smooth muscle actin (SMA) and Keratin 8. Primary antibodies used were: rabbit anti-pr (Thermo Scientific 9102-S0, 1:250), rabbit anti-er (Santa Cruz SC-542, 1:250), rabbit anti-sma (AbCAM 5694, 1:250), and rat anti-keratin 8 (Troma-I) (Developmental Studies Hybridoma Bank, University of Iowa, 1:250). Secondary antibodies were: biotinylated goat anti-rabbit IgG (Vector BA-1000, 1:250) and biotinylated rabbit-anti-rat IgG (Dako Cytomation E0468, 1:250). The fluorophore was a streptavidin Tritc (Vector SA-5001, 1:250). Hoechst dye (Invitrogen H1399) was used for nuclear DNA counterstaining, and slides were visualized using a Zeiss wide-field fluorescent microscope equipped with AxioVision 4.8 software. Cell morphology was scored by measuring the length and width of nuclei on paraffin sections. Cells were considered to have a columnar morphology if nuclear length measured at least twice the width. 55

76 2.4 Results. MMTV-driven H2BGFP expression labels a biologically distinct luminal epithelial subset. The MMTV-LTR is generally accepted as a luminal mammary epithelial promoter. We used a previously characterized MMTV-driven rtta transactivator line 177 combined with a teto-h2bgfp reporter line 178 in order to label MMTVexpressing luminal epithelial cells in bitransgenic animals (Figure 2.2). As we have previously reported 97, expression of the H2BGFP label was dependent upon treatment with doxycycline (Dox). Specifically, fluorescent GFP expression was undetectable in the mammary glands of Dox naïve bitransgenic mice, whereas as little as 24 hours of Dox treatment yielded extensive labeling of the luminal mammary epithelial compartment. While MMTV-directed H2BGFP expression was expressed exclusively in luminal epithelial cells, labeling was incomplete and limited to a subset of luminal MECs (Ref. 97 and Shelley Gestl and Jessica Mathers, unpublished data). Closer examination of mammary gland sections from Dox treated bitransgenic mice revealed a non-random distribution of the H2BGFP labeled luminal epithelial subset. Instead, H2BGFP labeling seemed to correlate closely with the morphology of the luminal epithelial cells (Figure 2.3). In tissue sections, the luminal epithelial compartment of the mammary gland consists of cells with a columnar morphology interspersed among cells with a cuboidal morphology (Figure 2.3a). MMTV-directed H2BGFP expression appeared to preferentially label those luminal cells with a columnar morphology. (Figure 2.3b) On mammary gland sections from three adult bitransgenic mice treated 3-4 days 56

77 with Dox, approximately 1200 luminal epithelial cells were scored for cell morphology. A cell was scored as having a columnar morphology if the length of its nucleus measured at least twice its height. By this measure, 76% of H2BGFP labeled cells had a columnar morphology, though this figure is likely an underestimate since histologic cell morphology depends, to a certain extent, on the plane of the section through the duct. By contrast, only 19% of GFP-negative cells within the luminal compartment scored as having a columnar morphology. To determine whether the H2BGFP-expressing cells represented a biologically distinct epithelial cell subset, mammary glands were collected from adult bitransgenic mice that were treated 3-4 days with Dox. Mammary gland sections were then stained with antibodies against epithelial lineage markers and HRs (Figure 2.4). As expected, the GFP-positive cells were interspersed throughout the Keratin 8 (K8)-positive luminal epithelial layer, though the K8 staining intensity appeared weaker in the GFP-positive subset than in the GFPnegative luminal subset, or was absent altogether in the GFP-positive cells (Figure 2.4a). Further, no GFP-positive cells stained with smooth muscle actin (SMA), a basal epithelial marker (Figure 2.4b). Finally, the staining pattern for both PR and ERα was mutually exclusive of H2BGFP expression. (Figure 2.4c, d). Thus, MMTV-directed H2BGFP expression labeled a distinct subset of HR-negative luminal epithelial cells. 57

78 Isolation of HR-positive MECs by FACS on the basis of MMTV-directed labeling. Next, pooled mammary glands collected from 7-8 bitransgenic mice treated 3-4 days with Dox were dissociated into single cell suspensions for FACS analysis. Using an established flow cytometry strategy 129 (Figure 2.5), mammary epithelial cells were separated into their component basal and luminal subsets, and then sorted on the basis of GFP expression (Figure 2.6a). In agreement with the histology, distinct GFP-positive and GFP-negative subsets of cells were present within the Luminal Gate, while nearly all of the cells within the Basal Gate were negative for GFP expression (Figure 2.6b). Luminal GFP+, Luminal GFP-, and Basal epithelial subpopulations were collected, and total RNA was isolated for qrt- PCR analysis. Again, expression of both ER and PR was restricted to the Luminal GFP- epithelial subpopulation, consistent with the immunofluorescence data above (Figure 2.6c). Similarly, RANK-L expression was limited to the same Luminal GFP- subset, while expression of the RANK receptor was upregulated in the Luminal GFP+ and Basal sorted populations (2.6c). MMTV expression pattern is developmental stage-dependent. A recent study 179 using the same MMTV-driven reporter system described significant H2BGFP labeling within the basal epithelial compartment following a continuous Dox treatment from conception through postnatal day 28. Because expression patterns of several transgene promoters 92,93 have been shown to be dependent upon the developmental stage of the mammary gland, we asked whether embryonic induction of MMTV-directed H2BGFP expression would 58

79 show a different labeling pattern than a postnatal induction. Bitransgenic MMTVrtTA/tet-O-H2BGFP mice were treated continuously with Dox throughout embryogenesis by initiating Dox treatment in breeding pairs prior to conception (Figure 2.7a). After birth, pups continued to receive Dox treatment through their diet until they were 28 days old, at which point, mammary glands were collected for histologic and flow cytometric analysis. In agreement with the previous report 179, mammary gland sections showed extensive labeling of both the basal and luminal epithelial compartments (Figure 2.7c,d). Bright H2BGFP expressing cells were discernable in the basal layer of mammary ducts as well as in the cap cell layer of the terminal end buds (TEBs). Regardless of localization, H2BGFP expressing cells did not co-express PR (Figure 2.7e). Curiously, despite prominent labeling of the basal epithelial compartment in mammary gland sections, flow cytometric analysis displayed few GFP-positive cells in either the Luminal or Basal gates. Approximately 11% of cells within the Luminal gate were positive for GFP expression, while less than 2% of cells within the Basal gate exhibited high GFP intensity (Figure 2.7b). In contrast, when bitransgenic mice were treated with a short Dox pulse from postnatal day 25 to day 28 (Figure 2.8a), H2BGFP expression was observed only in the luminal compartment of mammary ducts (Figure 2.8c,d). Again, H2BGFP expression was mutually exclusive to PR expression. In TEBs, H2BGFP expression continued to be discernable in the cap cell layer, but at a lesser intensity. By flow cytometry, the vast majority of GFP-positive cells fell within the Luminal Gate (Figure 2.8b), in agreement with the histology. Interestingly, in 59

80 histograms depicting GFP intensity vs. event counts, a distinct pulse peak of cells with high GFP expression was present for the mammary glands treated short-term with Dox (Figure 2.8b). This peak was absent for the glands treated continuously through embryogenesis (Figure 2.7b). Further, the ratio of luminal: basal MECs was increased in chronically treated glands (L:B=2.8) compared with the glands treated with only a short Dox pulse (L:B=1.5), raising the possibility that longterm induction of this particular histone-tagged GFP reporter may not be completely benign. Transgenic oncogene expression is limited to H2BGFP labeled MECs in a classic MMTV-driven tumor model. Next, we asked whether this MMTV expression pattern was translatable to other MMTV-driven transgenic models. MMTV-rtTA/tet-O-H2BGFP reporter mice were crossed with MMTV-Neu oncomice to generate MMTV-rtTA/tet-O- H2BGFP/MMTV-Neu tritransgenic animals. Pooled mammary glands from 4 adult tritransgenic mice treated 3 days with Dox were dissociated into a single cell suspension and subjected to FACS analysis as before (Figure 2.9a). Cells were separated into luminal and basal subpopulations. Again, cells within the Luminal Gate belonged to one of two distinct subgroups: the Luminal GFP+ group or the Luminal GFP- group (Figure 2.9b). Whereas, just one distinct group of GFPnegative cells was present within the Basal Gate (Figure 2.9b). Using probes specific to the rat Neu transgene, qrt-pcr analysis revealed Neu transgene expression was markedly enriched within the Luminal GFP+ subpopulation, showing a nearly 9 fold increase in expression as compared to the Luminal GFP- 60

81 subgroup (Figure 2.9c). By contrast, PR and ER were preferentially expressed in the Luminal GFP- group (Figure 2.9c). A third population of CD49f High /EpCAM - /GFP - cells appeared on the 2-parameter flow cytometry plot (Figure 2.9a). However, because these cells failed to express epithelial keratins by qrt-pcr, they were excluded from further analysis as a non-epithelial population (not shown). Targeted transgene expression to a HR-negative epithelial subset yields HRnegative tumors. Tumors arising in the MMTV-Neu transgenic model do not express HRs and grow in a hormone-independent manner. Yet, MMTV-Neu induced mammary hyperplasias still contain interspersed HR-positive cells. We asked at what point during tumor progression MMTV-Neu tumors lost their HR positivity. MMTV-rtTA/tet-O-H2BGFP/MMTV-Neu mice were aged to 200 days and then treated with Dox for 3 days prior to necropsy to induce MMTV-driven H2BGFP labeling. Mammary glands with grossly undetectable tumors were whole mounted and analyzed histologically in an effort to identify early precursor lesions. Mice with grossly detectable lesions were excluded from the analysis. Precursor lesions were relatively rare, discernible in H&E sections of 28% of the glands analyzed. Staining with antibodies to PR revealed that Neu-induced hyperplasias, including small focal hyperplastic nodules, contained PR-expressing cells (Figure 2.10). As in normal adult mammary gland, PR expression was mutually exclusive to expression of the MMTV-driven H2BGFP reporter. Conversely, early carcinoma in situ lesions were completely negative for PR expression (Figure 2.10). 61

82 To get a more dynamic perspective of tumor progression, we utilized a Dox-dependent teto-neu mouse line that inducibly expresses an activated Neu allele. These mice were crossed to MMTV-rtTA/tet-O-H2BGFP reporter mice to generate tritransgenic MMTV-rtTA/tet-O-H2BGFP/tetO-Neu mice in which the MMTV-rtTA directs Dox-dependent expression of both the H2BGFP reporter and the activated Neu allele. Dox-treatment in adult tritransgenic animals rapidly produced robust hyperplastic overgrowth that involved the entire mammary gland and simultaneously affected all 10 glands within the mouse (Figures 2.11 and 2.12). After just 5 days of Dox-treatment, mammary glands were stiff, thickened, and palpable through the skin. Mice could not tolerate more than 7 days of Doxtreatment, though the phenotype was rapidly reversible upon Dox withdrawal (Figure 2.11). PR staining was detected in hyperplasias throughout the entire 7 day Dox-treatment time course, and PR expression was always mutually exclusive of H2BGFP expression (Figure 2.12). Interestingly however, with longer Dox treatments (5 and 7 days), PR-positive cells were largely detectable in only those areas that maintained an organized, bilayered ductal architecture. Contrastingly, hyperplastic patches of H2BGFP-labeled cells were PR-negative, possibly marking a transition to monoclonality (Figure 2.12). 2.5 Discussion. Transgenic mouse tumor models have been instrumental to our understanding of tumor biology. The MMTV-LTR, in particular, has a long legacy in mouse mammary tumor models, as MMTV-driven models were among the very first transgenic tumor models to be developed 47,180,181. Since then, the 62

83 MMTV-LTR has remained a popular promoter element in the field of mouse mammary gland biology because of its tissue specificity. Yet, biologic models are only informative to the extent that we understand what, precisely, they are modeling. Here, we further refine the expression profile of the MMTV-LTR. Using an MMTV-driven rtta transactivator combined with a teto-h2bgfp reporter gene to achieve Dox dependent labeling of MMTV-expressing cells in the mammary glands of bitransgenic mice, we found MMTV-directed H2BGFP expression is restricted to a specific HR-negative luminal epithelial subset in adult animals. Interestingly, in agreement with a recent report 179, in utero induction of MMTV-directed H2BGFP expression produced a broader labeling pattern, such that, when glands were examined at postnatal day 28, a significant proportion of basal epithelial cells displayed H2BGFP expression together with the luminal epithelial cell subset. However, a short pulse from postnatal day 25 through 28, reliably induced H2BGFP expression in only HR-negative luminal epithelial cells of mature ducts, consistent with the labeling pattern observed in adult glands. Recent lineage tracing experiments highlight the influence that mammary gland developmental stage can have on the expression profile of various promoters. For instance, in one set of experiments using Axin2 to drive expression of an inducible label, in utero induction labeled only MECs from the luminal epithelial lineage. Whereas, postnatal induction using the same promoter instead labeled only basal MECs 93. Similarly, in a second lineage tracing experiment, in utero induction of a Keratin 14 (K14)-driven label marked both basal and luminal 63

84 epithelial cells, while postnatal induction labeled only cells from the basal epithelial subset 92. The MMTV-LTR, then, may be an additional example of a gene promoter whose specific expression pattern is developmental stagedependent. While its expression in the embryonic and juvenile mammary gland appears promiscuous, additional work from our lab suggests that the MMTV-LTR promoter has a stable expression profile in sexually mature animals: short Dox treatments throughout all stages of post-pubertal mammary gland development and pregnancy reliably induced H2BGFP expression that was limited to a subset of luminal epithelial cells (Shelley Gestl and Jessica Mathers, data not shown). Chemical carcinogenesis experiments in rats have defined the interval between puberty and the first pregnancy as the susceptibility window for cellular transformation 182,183. Thus, adult MMTV expression is likely most relevant to tumor formation in MMTV-driven tumor models, such as the MMTV-Neu model. The question remains: why is MMTV expression limited to such a specific epithelial subset in adult animals? The MMTV lifecycle is closely linked to the stages of mammary gland development, as viral transmission depends upon shedding of virus particles in the milk of lactating animals. Upon pregnancy, an increase in serum progesterone stimulates dramatic expansion of the mammary ductal epithelium as the mammary gland prepares for milk production and lactation. In particular, progesterone signaling has been shown to induce proliferation of the HR-negative luminal epithelial cell subset, which ultimately contributes to the secretory apparatus of the gland 18,52,184. Accordingly, this secretory subset would be an appropriate place for viral gene expression and 64

85 virion assembly. It stands to reason, then, that MMTV might target this expanding cell population in an effort to maximize viral transmission. Ultimately, experiments that sort MMTV-infected mammary epithelium into its component epithelial cell subpopulations may provide confirmatory evidence for the MEC subtype specific expression of MMTV. Because MMTV provirus integrates randomly into its host genome, differences in integration might contribute to differences in expression pattern. Nonetheless, binding of transcriptional repressor proteins such as CCAAT displacement protein (CDP) 161,185 and special AT-rich sequence binding protein 1 (SATB1) 165 to NREs in the MMTV-LTR have been shown to help regulate spatial and temporal expression of MMTV. It will be interesting to determine whether a similar mechanism impedes MMTV expression within the HR-positive mammary epithelial subset, thereby facilitating selective expression of MMTV in the HRnegative luminal epithelial cell subpopulation. Given the extensive amount of evidence supporting hormonal regulation of the MMTV promoter, it was surprising to us that MMTV-driven transgene expression was absent from the HR-expressing mammary epithelial subpopulation. Yet, for several transgenic models, MMTV-driven phenotypes have been reported in animals that lacked systemic ovarian hormones, such as male animals or ovariectomized females 47,186. Indeed, MMTV-directed H2BGFP labeling of luminal MECs was preserved in both male and ovariectomized female animals following a short Dox pulse (Shelley Gestl, personal communication). Similarly, serum levels of ovarian hormones plummet prior to parturition and 65

86 remain low throughout lactation, coinciding with the phases of highest MMTV expression. Together, these observations suggest that MMTV expression does not depend upon ovarian hormones. Rather, regulation of MMTV expression in vivo may instead depend more heavily on glucocorticoid and/or mineralocorticoids. In our further experiments with the classic MMTV-Neu transgenic model, we found Neu transgene expression was also enriched within this same HRnegative luminal cell subpopulation, culminating in HR-negative disease from the carcinoma in situ stage onward. We propose that MMTV-directed transgene expression may be similarly limited to the specific HR-negative luminal epithelial cell subset in other MMTV-driven mouse models. In particular, this MEC subsetspecific expression pattern may explain the preponderance of HR-negative, hormone independent mouse tumor models. Differences in transgene integration may affect expression patterns, so it will be important to determine if the other MMTV-driven transgenic models exhibit this same subset restricted expression pattern. Notably, this information may influence how experiments with these models are interpreted. While the MMTV-LTR targets transgene expression to the HR-negative luminal epithelial subset, there are currently no promoters that specifically target the HR-positive luminal epithelial population. Here, we successfully used MMTV-directed labeling to isolate HR-positive luminal epithelial cells from the HR-negative luminal cell subset by FACS. In a previous report, expression of Prominin1 together with CD24 was used to similarly enrich for a HR-positive luminal subpopulation 187. It will be interesting to determine whether these two 66

87 methods isolate the same population of cells. Moving forward, comparative gene chip analysis between the HR-positive and HR-negative luminal epithelial subsets may identify genes unique to the HR-positive subset that may ultimately serve as more effective promoters for generating transgenic HR-positive breast cancer models. 67

88 Figure 2.1: Schematic of the MMTV-LTR. The MMTV-LTR contains U3, R, and U5 regions. The U3 region contains several cis-acting regulatory regions, depicted as boxes with different hatch markings: the promoter-proximal negative regulatory element (pnre), the promoter-distal negative regulatory element (dnre), the junction between the pnre and the dnre (jnre). The U3 region also contains the hormone response element (HRE) harboring several HR binding sites. Arrow indicates the transcriptional start site (+1) within the R region. The first base of the LTR is shown as +1 (-1195 from the transcription start site). Figure modified from Zhu Q and Dudley JP (2002). J. Virol. 76(5) Reprinted with permission from the American Society for Microbiology. 68

89 Figure 2.2: Schematic of MMTV-directed H2BGFP labeling. Doxycycline (Dox) dependent TetO-H2BGFP expression directed by an rtta transactivator driven by the MMTV-LTR promoter. 69

90 Figure 2.3: MMTV-directed H2BGFP expression labels a morphologically distinct luminal epithelial subset. a. H&E stained section through a mammary duct from an adult virgin mouse highlighting the different morphology of cells within the luminal MEC layer. Arrows indicate representative columnar luminal MECs. Arrowheads indicate representative cuboidal luminal MECs. b. Hoechst stained section through a mammary duct from an adult virgin MMTVrtTA/tetO-H2BGFP bitransgenic mouse treated 3 days with Dox. H2BGFP expression is seen primarily in columnar-shaped luminal MECs. Arrows indicate representative GFP-positive columnar luminal MECs. Arrowheads indicate representative GFP-negative cuboidal luminal MECs. 70

91 Figure 2.4: MMTV-driven H2BGFP expression labels a HR-negative luminal MEC subcompartment. Sections through mammary ducts from an adult virgin MMTV-rtTA/tetO- H2BGFP bitransgenic mouse treated 3 days with Dox to induce MMTV-driven H2BGFP labeling. Sections stained with antibodies against a) Keratin-8, b) Smooth muscle actin, c) Progesterone Receptor (PR) or d) Estrogen Receptor (ER) and counterstained with Hoechst dye. Expression of H2BGFP is mutually exclusive of both PR and ER expression. Scale bar, 50μm. 71

92 Figure 2.5: Flow cytometry gating strategy for resolving basal and luminal subsets from mammary tissue. Pooled Mammary glands from several adult virgin mice were mechanically and enzymatically dissociated into single cell suspensions. a. Negative selection against Lin + cells using Stem Cell Technologies EasySep Mouse Epithelial Cell Enrichment Kit. Resulting Lin - (CD45 - / CD31 - / TER119 - / BP-1 - ) cells were then immunostained with antibodies for CD49f (α6 integrin) and EpCAM and analyzed by flow cytometry. b. Exclusion of cell debris and dead/ dying cells. Dead/dying cells collect as a band along the bottom of a FSC-A vs. SSC-A twoparameter plot, and these were gated out in P1. c. Cell doublets were discarded in P2. d. Basal and Luminal mammary epithelial cell populations were separated by immunophenotype. Basal epithelial cells are CD49f high / EpCAM low (P3) and luminal epithelial cells are CD49f Low / EpCAM high (P4). e. Gating tree showing gating strategy for flow cytometric analysis as well as parent and total cell percentages within each of the gates for a representative experiment. 72

93 Figure 2.6: Isolation of MMTV-expressing cells by FACS. a. Following a 3 day Doxtreatment, pooled mammary glands from 7-8 adult virgin mice were dissociated into a single cell suspension and separated by immunophenotype into basal (CD49 High /EpCAM Low ) and luminal (CD49 Low /EpCAM High ) subpopulations by flow cytometry. b. Basal and luminal subpopulations were sorted by FACS on the basis of MMTV-driven H2BGFP expression into GFP-positive and GFP-negative subgroups. Upper panels show two-parameter plots depicting GFP intensity vs. forward scatter. Lower panels show histograms depicting GFP intensity vs. event counts. Percentages show +/- S.E.M. c. Relative gene expression levels between sorted MEC populations measured by quantitative RT-PCR. Gene expression levels normalized to Gapdh and set relative to the Unsorted population (relative expression =1). Data points represent two independent sorts, with black lines indicating median values. Un, unsorted; B, basal; L GFP-, luminal GFP-negative; L GFP+, luminal GFP-positive groups. 73

94 74

95 Figure 2.7: Dox treatment through embryogenesis labels both basal and luminal mammary epithelial subsets. a. Schematic depicting period of Dox treatment. b. Mammary glands were collected from MMTV-rtTA/tetO-H2BGFP bitransgenic animals at postnatal day 28 following chronic Dox treatment. Glands were dissociated into a single cell suspension and separated by immunophenotype into basal (CD49 High /EpCAM Low ; blue) and luminal (CD49 Low /EpCAM High ; green) subpopulations by flow cytometry. Percentages show +/- S.E.M. c. GFP expression within Basal and Luminal gates. Upper panels show two-parameter plots depicting GFP intensity vs. forward scatter. Lower panels show histograms depicting GFP intensity vs. event counts. Percentages show +/- S.E.M. d-f. Sections through mammary ducts (upper panels) or terminal end buds (TEB, lower panels) of 28 day old chronically Dox treated bitransgenic animals. Sections stained with antibodies against d) Keratin-8 (K8), e) Smooth muscle actin (SMA), and f) Progesterone receptor (PR). Scale bar, 50 μm. 75

96 76

97 Figure 2.8: A Short Dox pulse in juvenile animals labels a luminal epithelial subset in mammary ducts, while TEBs contain labeled cells in both basal and luminal compartments. Schematic depicting period of Dox treatment. b. Mammary glands were collected from bitransgenic animals at postnatal day 28 following a short Dox pulse. Glands were dissociated into a single cell suspension and separated by immunophenotype into basal (CD49 High /EpCAM Low ; blue) and luminal (CD49 Low /EpCAM High ; green) subpopulations by flow cytometry. Percentages show +/- S.E.M. c. GFP expression within Basal and Luminal gates. Upper panels show two-parameter plots depicting GFP intensity vs. forward scatter. Lower panels show histograms depicting GFP intensity vs. event counts. Percentages show +/- S.E.M. d-f. Sections through mammary ducts (upper panels) or terminal end buds (TEB, lower panels) of 28 day old animals following 3d Dox pulse. Sections stained with antibodies against d) Keratin-8 (K8), e) Smooth muscle actin (SMA), and f) Progesterone receptor (PR). Scale bar, 50 μm. 77

98 78

99 Figure 2.9: Neu transgene expression is enriched in the H2BGFP-expressing luminal subset of MMTV-rtTA/ teto-h2bgfp/ MMTV-Neu tritransgenic animals. a. Following a 3 day Dox-treatment, pooled mammary glands from 4 adult virgin MMTV-rtTA/ teto-h2bgfp/ MMTV-Neu tritransgenic mice were dissociated into a single cell suspension and separated by immunophenotype into basal (CD49 High /EpCAM Low ; blue) and luminal (CD49 Low /EpCAM High ; green) subpopulations by flow cytometry. b. Basal and luminal subpopulations were sorted by FACS on the basis of MMTV-driven H2BGFP expression into GFP-positive and GFP-negative subgroups. Upper panels show two-parameter plots depicting GFP intensity vs. forward scatter. Lower panels show histograms depicting GFP intensity vs. event counts. c. Relative gene expression levels between sorted MEC populations measured by quantitative RT-PCR. Gene expression levels normalized to Gapdh and set relative to the Unsorted population (relative expression =1). Data represents a single sort. Un, unsorted; B, basal; L GFP-, luminal GFPnegative; L GFP+, luminal GFP-positive groups. 79

100 80

101 Figure 2.10: Progesterone Receptor expression in MMTV-Neu precursor lesions. MMTVrtTA/tet-O-H2BGFP reporter mice were crossed with MMTV-Neu mice to generate MMTV- Neu/MMTV-rtTA/tet-O-H2BGFP tritransgenic animals. 200 day old mice were Dox-treated for 3 days to induce MMTV-driven H2BGFP labeling, and mammary glands with grossly undetectable lesions were collected to identify early precursor lesions. Depicted are H&E stained sections showing representative precursor lesions alongside sections of stage-matched lesions stained with antibodies recognizing Progesterone Receptor (PR). MMTV-Neu induced hyperplasias contained scattered PR-positive cells, while Carcinoma-In-Situ lesions were completely PR-negative. 81

102 82

103 Figure 2.11: Neu-induced hyperplasia in MMTV-rtTA/tetO-Neu animals is rapidly reversible upon Dox withdrawal. Dox time course of MMTV-rtTA/tetO-Neu animals. MMTVrtTA/tet-O-H2BGFP mice were crossed to teto-neu mice to generate tritransgenic animals in which the MMTV-rtTA directs Dox-inducible expression of both the H2BGFP reporter transgene and an activated Neu oncogene. Depicted are carmine stained whole mounts of #4 mammary glands treated with Dox as indicated. Left panels show glands from MMTVrtTA/tetO-H2BGFP animals. Right panels show glands from MMTV-rtTA/ teto-h2b/ teto-neu littermates. 83

104 84

105 Figure 2.12: Dox time course of MMTV-rtTA/tetO-H2BGFP/tetO-Neu tritransgenic mice shows transition to HR-negative disease. MMTV-rtTA/tet-O-H2BGFP mice were crossed to tet-o-neu mice to generate tritransgenic animals in which the MMTV-rtTA directs Doxinducible expression of both the H2BGFP reporter transgene and an activated Neu oncogene. Mice were treated with Dox for up to 7 days, as indicated. Left-most panels depict carmine stained whole mounts of #4 mammary glands from Dox-treated animals. Middle panels show H&E stained sections through mammary glands. Right-most panels show mammary gland sections stained with antibodies against Progesterone Receptor (PR) and counterstained with Hoechst. With longer Dox treatments (5 and 7 days), PR-positive cells are present in areas that maintain an organized, bilayered architecture, indicated with brackets. Contrastingly, hyperplastic patches of H2BGFP-labeled cells (outlined with dotted line) are PR-negative. 85

106 86

107 MMTV-Driven Tumor Models Model Transgene HR Status MMTV-Wnt1 Wnt1 + MMTV-RANK RANK + MMTV- CyclinD1 MMTV- D1T286A MMTV-ErbB2 MMTV-c-neu* MMTV-c-neu* Cyclin D1 +/- Mutated cyclin D1 +/- Wild type ErbB2 (c-neu) - Mutated ErbB2 (c-neu) (Muller et al.) Mutated ErbB2 (c-neu) (Bouchard et al.) - - Notes Hormone independent tumors; median latency 6 months; mixed lineage tumors. No tumors in virgin animals. Progesterone dependent tumors arose only in multiparous (median latency 26.5 months) or MPA/DMBA treated animals; mixed lineage tumors. 37% of tumors expressed HR, the remaining 63% were HR-negative and hormone independent; median latency 20 months; mixed lineage tumors. 50% of tumors expressed HR, the remaining 50% were HR-negative and hormone independent; median latency 16 months; mixed lineage tumors. Hormone independent tumors; stochastic tumor formation; latency 7-12 months; luminal epithelial character. Hormone independent tumors; synchronous tumor formation; median latency 3 months; luminal epithelial character. Hormone independent tumors; stochastic tumor formation; latency 5-10 months; luminal character. Reference 158,188 77,78,189, , , MMTV-Myc c-myc - Hormone independent tumors; latency 11 months; mixed lineage tumors. 180,186 87

108 MMTV-HaRas HRas - MMTV-PyMT PyMT - MMTV- CyclinE MMTV- ΔN89β-cat MMTV- TGFB-DNIIR LMW Cyclin E - Activated β- catenin - Dominant negative TGFβ NR MMTV-c-rel c-rel NR MMTV-TGFA TGFα NR MMTV- Notch4 Notch4 (int3) NR MMTV-cdc37 CDC37 NR MMTV-Cox2 COX2 NR MMTV-NDF NDF/heregulin NR MMTV-Wnt10 Wnt10 NR MMTV- rtta/teto- Wnt1 Tet operator controlled Wnt1 + Hormone independent tumors; latency 6 months Hormone independent tumors; latency 1 month for females, 2-3 months for males. Hormone independent tumors; latency months. Hormone independent tumors; latency 7 months; tumor formation accelerated in multiparous females; mixed lineage tumors. Hormone independent tumors; latency 27.5 months Hormone independent tumors; latency 20 months Hormone independent tumors; latency 16 months Hormone independent tumors; latency 2-7 months No tumors in virgin animals; tumor latency 19 months for multiparous females. No tumors in virgin animals; tumors arose after multiple rounds of pregnancy. Hormone independent tumors; latency 12 months. Hormone independent tumors; latency months for females, 13 months for males. Hormone independent tumors; latency 5 months following Dox induction; mixed lineage tumors , , , ,211 88

109 MMTV- rtta/teto- HRas Tet operator controlled mutant HRas - Hormone independent tumors; 177,212 MMTVrtTA/tetO-Neu Tet operator controlled mutant ErbB2 (c-neu) - Hormone independent tumors; tumors arise synchronously in every gland within days of Dox induction 177 MMTV- rtta/teto-c- Myc Tet operator controlled c-myc - Hormone independent tumors; latency 22 weeks following Dox induction 177,213 MMTV-Cre/ p53 Floxed p53 - Hormone independent tumors; median latency 17.5 months for virgins, tumor formation accelerated in multiparous females; 214 MMTV-Cre/ BRCA1 Floxed BRCA1 - Hormone independent tumors; latency 6-8 months in p53 +/- background 215,216 MMTV- Cre/Pten Floxed Pten - Hormone independent tumors; latency 10 months; tumor formation accelerated in multiparous females. 217 MMTV- Cre/Rb Floxed Rb - Hormone independent tumors; latency 18 months 218 Table 2.1: MMTV-driven tumor models. The Hormone receptor (HR) status of the various tumor models is indicated: +, tumors express HRs (ER and/or PR); -, tumors do not express HRs; NR, HR status not reported. 89

110 Genotyping Primers Gene Primer Sequence MMTV-rtTA F1: 5 - TGCCGCCATTATTACGACAAGC- 3 R1: 5 - ACCGTACTCGTCAATTCCAAGGG- 3 F2: 5 - ATCCGCACCCTTGATGACTCCG- 3 R2: 5 - GGCTATCAACCAACACACTGCCAC- 3 teto-h2bgfp F1: 5 - AAGTTCATCTGCACCACCG- 3 F2: 5 - CCTTGATGCCGTTCTTCTGCTTGTC- 3 R: 5 - TCCTTGAAGAAGATGGTGCG- 3 teto-neu F1: 5 - AGCAGAGCTCGTTTAGTG- 3 R1: 5 - GGAGGCGGCGACATTGTC- 3 F2: 5 - CAAATGTTGCTTGTCTGGTG- 3 R2: 5 - GTCAGTCGAGTGCACAGTTT- 3 MMTV-Neu Maintained as a homozygous colony Table 2.2: Genotyping Primers. Genotyping of transgenic mouse lines was performed by PCR using genomic DNA isolated from tail clips and transgene specific primers. F indicates the forward primer. R indicates the reverse primer. 90

111 Chapter 3 A Novel Chimeric Mammary Organoid Technique for Assessing the Cell-Intrinsic and Extrinsic Effects of Oncogene Expression Allison S. Cleary, Kristin A. Plichta, and Edward J. Gunther 91

112 3.1 Abstract When establishing a malignant microenvironment, an individual oncogene may wield both cell autonomous and non-autonomous effects. Indeed, mounting evidence portrays tumors as complex bionetworks, in which tumor cells interact with one another and with the surrounding environment to direct tumor maintenance and progression. While numerous studies detail the cell-autonomous effects of oncogenes, determining how oncogene-expressing cells impact neighboring cells remains a challenge. To address this, we developed a method for creating chimeric mouse mammary organoids, which permits analysis of both cell autonomous and non-autonomous effects of oncogene expression within the context of the bilayered, polarized architecture of normal mammary ductal epithelium. By combining primary mammary epithelial cells (MECs) from two different transgenic donors, chimeric mammary organoids were assembled consisting of intermingled populations of genetically distinct donor MECs. Using donor epithelia that expressed different fluorescent reporters allowed discrimination between distinct donor populations and tracking of subsequent cell fate changes within each population over time. We tested our system using transgenic mouse models engineered to inducibly express either an activated HRas allele or oncogenic Wnt1 in specific MECs. These oncogene-expressing cells were incorporated into wild-type mammary organoid scaffolds, and imaged over time in 3D Matrigel culture. As expected, HRas-expressing cells expanded in number, which is consistent with a predominantly cell autonomous role for oncogenic HRas. By contrast, Wnt1-expressing cells did not expand in number. 92

113 Instead, luminal expression of Wnt1 produced a dramatic and selective expansion of the basal epithelial cell compartment, as captured by live cell imaging. Thus, secreted Wnt1 primarily drove MEC overgrowth by acting in a paracrine rather than autocrine manner. Altogether, chimeric organoid analysis can be used as a sensitive and efficient tool for studying complex cell-cell interactions in the context of both normal and transformed mammary epithelium. 3.2 Introduction. The multiple hit hypothesis of cancer pathogenesis posits that cancer is the result of an accumulation of genetic alterations in proto-oncogenes and tumor suppressor genes. This suggests that cooperative interactions exist among acquired mutations via crosstalk between their respective signaling pathways. While cooperative oncogenic interactions are traditionally thought to occur autonomously within individual tumor cells, emerging evidence suggests that oncogenic interactions can also occur across tumor cell groups. Indeed, the complexity of cellular interactions occurring within tumors may equal or surpass that of healthy tissues. As such, expression of a given oncogene may produce both cell-intrinsic and extrinsic effects. For example, in addition to promoting proliferation and cell survival in a cell autonomous manner, expression of oncogenic Ras has been shown to promote neovascularization and stromal activation through paracrine signaling mechanisms Whereas numerous studies detail the cell-autonomous effects of oncogene expression, few address how that expression may impact neighboring cells. In order to begin to understand the mechanisms of how tumor cells interact with one another, it is 93

114 necessary to first appreciate both the cell-autonomous and non-autonomous effects for a given oncogene. Technical challenges make the study of these intercellular interactions difficult. Because cell-cell interactions often depend upon an intact tissue architecture, most cell culture models are inadequate for this type of study. On the other hand, individual cellular interactions can be difficult to trace in in vivo models. Alternatively, partially disaggregated mammary ductal epithelium can be grown in 3D Matrigel culture as mammary organoids. Organoids maintain the bilayered architecture of mammary ducts, with distinct basal and luminal cell compartments, making them particularly well-suited for studying interactions between the epithelial cell populations 222. Administration of growth factors induces organoids to undergo morphologic branching, reminiscent of the important developmental changes seen in mammary ducts in vivo 223, while transgenic activation of oncogenes induces aberrant organoid overgrowth 224. As indicated, organoids retain the bi-layered structure and cellular polarity of intact mammary ducts. Thus, similar to mammary epithelium in vivo, transgene expression can be targeted to epithelial subsets within mammary organoids. We have previously described mouse models 224 that inducibly express tet operatorcontrolled transgenes in a compartment-restricted manner (Figure 3.1). Specifically, either the mouse mammary tumor virus- long terminal repeat (MMTV-LTR) or the Keratin-5 (K5) promoter was used to drive expression of the rtta transactivator in luminal or basal MECs respectively 177,225. Crossing each transactivator strain to a reporter strain expressing a tet-operator-controlled 94

115 histone H2BGFP fusion transgene (hereafter, TGFP) 178 demonstrated Doxdependent fluorescent labeling of either luminal (luminal GFP ) or basal (basal GFP ) MECs in bitransgenic organoids. Nuclear TGFP labeling in mammary organoids permitted real time visualization of individual mitosis and cell death events. Chimera analysis is an essential tool for elucidating the role of cell-cell interactions in complex phenotypes. Here, we have developed a novel technique for generating chimeric mammary organoids. By combining primary mammary epithelia from two different transgenic donors, chimeric mammary organoids were assembled consisting of intermingled populations of genetically distinct donor MECs. Use of donor epithelia that expressed different fluorescent reporters allowed discrimination between distinct donor populations and tracking of subsequent cell fate changes within each component population over time. We used this technique to discern the cell autonomous and non-autonomous effects of two fundamentally different oncogenes: the cell-intrinsic signaling molecule, Ras, and the secreted signaling ligand, Wnt1. Ras signaling. The Ras/MAPK pathway (Figure 3.2) plays key roles in cell proliferation, differentiation, survival, and cell death. The Ras (Rat sarcoma) GTPase is typically activated by the SOS (son of sevenless) guanine nucleotide exchange factor, downstream of receptor tyrosine kinase (RTK) signaling Upon activation, Ras initiates the MAPK signaling cascade, consisting of three sequentially activated protein kinases: Raf, MEK, and ERK. Ras activates the first kinase Raf (MAP3K) , which in turn phosphorylates and activates MEK 95

116 (MAP2K) , which ultimately phosphorylates and activates the third and final kinase of the cascade, ERK (extracellular signal-regulated kinase) 234,235. Activated ERK then translocates to the nucleus where it interacts with transcription factors to induce expression of various downstream growth-promoting target genes In addition to the MAPK pathway, activated Ras can signal through the PI3K (phosphatidylinositol 3 kinase) pathway (Figure 3.2). Here, Ras activates PI3K which in turn activates Akt (protein kinase B) through phosphorylation 244,245. Akt can then activate a variety of effectors, including mtor (mammalian target of rapamycin) to regulate cell proliferation, survival, and growth. PTEN is an important negative regulator of Akt, and acts as a tumor suppressor gene. Increased Ras signaling is observed frequently in cancer across numerous tumor types. Activating mutations in one of the highly conserved Ras family member genes, HRAS, KRAS, or NRAS, occur in 20-30% of tumors 246. Mutations most commonly arise in one of 3 specific codons: 12, 13 or 61. Each of these mutations results in constitutive Ras activity by disrupting normal GTPase function. Mutations in codons 12 and 13 occur in the phosphate binding loop of the protein and prevent inactivation of Ras by GTPase activating proteins (GAPs) 247, while mutations at codon 61 occur in the catalytic domain and destabilize the transition state for GTP hydrolysis 248. Activation of the Ras signaling pathways can also be achieved through mutations in any of the downstream signaling kinases. For example, melanomas and colorectal cancers often harbor activating mutations in the downstream B-RAF gene 249,250, while 96

117 breast cancers frequently harbor gain-of-function mutations in PI3KCA Loss of negative regulators like PTEN is seen frequently in prostate cancers and glioblastomas, and loss of heterozygosity (LOH) at the PTEN locus is detected in approximately 40% of breast cancers 253. Finally, upstream activation of Ras is achieved when growth factor RTKs are mutated or overexpressed, such as EGFR (epidermal growth factor receptor) in lung cancer 254 and HER2/neu (human epidermal growth factor receptor 2) in breast cancer 255. Wnt/β-catenin signaling. Canonical Wnt signaling is a highly evolutionarily conserved signaling pathway responsible for regulating cell fate decisions. It performs key roles in both embryonic development and adult tissue homeostasis, and has also been shown to have important roles in tumorigenesis. Wnts comprise a large family of secreted signaling glycoproteins: there are as many as 19 mammalian Wnt family members 256. As a secreted signaling ligand, Wnts can act in a paracrine or autocrine manner. Wnt proteins bind to the Frizzled receptor- LRP5/6 co-receptor complex to stimulate the canonical Wnt signaling pathway (Figure 3.3). The binding of Wnt ligand to its receptor complex activates the cytoplasmic protein, Dishevelled 256,260, which in turn inhibits the β-catenin destruction complex, composed of Axin, adenomatous polyposis coli (APC), and glycogen synthase kinase 3 (GSK3) 261. In the absence of Wnt ligand, the β-catenin destruction complex phosphorylates β-catenin, targeting it for proteosomal degradation 262. Wnt mediated inhibition of the β-catenin destruction complex leads to stabilization and accumulation of β-catenin in the cytoplasm. Stabilized 97

118 β-catenin can then translocate to the nucleus 262 where it activates the Tcf/Lef family of transcription factors In cancer, the Wnt signaling pathway is activated in a number of different ways. For example, 30-40% of hepatocellular cancers 267 sustain an activating mutation in the CTNNB1 gene, which encodes β-catenin, the key downstream effector of the pathway. Alternatively, loss-of-function mutations in the APC tumor suppressor gene result in inappropriate stabilization of β-catenin; colorectal cancers almost invariably activate the Wnt/β-catenin signaling pathway in this manner 106. Finally, the Wnt pathway can be activated via epigenetic silencing of several negative regulators: secreted Frizzled-related protein, Wnt inhibitory factor, Dikkopf-related protein, and the nuclear protein SOX 268. In human breast cancer, mutations are rarely detected in Wnt/β-catenin pathway members, yet 40% of breast cancers 269 display abnormal β-catenin expression or localization, suggesting activation of the pathway. Downregulation of various negative regulators of the Wnt signaling pathway through promoter methylation 270 or chromatin remodeling is thought to contribute to the disregulated β-catenin signaling in breast cancer, and these are thought to be particularly important for the so called triple negative breast cancer subset Methods. Transgenic Mice. Mice were housed under pathogen-free conditions in the Pennsylvania State University College of Medicine rodent facility with access to water and standard mouse chow (Harlan Tekland, HT2018) ad libitum. All 98

119 experimental protocols were approved by the Pennsylvania State University College of Medicine s Institutional Animal Care and Use Committee. The teto- H2B-eGFP (Tg(tetO-HIST1H2BJ/GFP)47Efu/J; stock #005104), and mrfp (B6.Cg-Tg(CAG-mRFP1)1F1Hadj/J; stock #005884) transgenic lines were obtained from the Jackson Labs. The teto-hras (FVB-Tg(tetO-HRAS)65Lc; stock #01XB4) line was obtained from the NCI Mouse Repository. The MMTVrtTA and teto-wnt1 transgenic lines were a gift from Dr. Lewis Chodosh (University of Pennsylvania, Philadelphia, PA). The Keratin-5-rtTA line was a gift from Adam Glick (Pennsylvania State University, University Park, PA). All mice either were generated in an inbred FVB/N background or were back-crossed 10 or more generations with FVB/N breeders before initiating experiments. Dox was administered by replacing standard mouse chow with chow containing 2g/kg doxycycline (Bio-serv). Genotyping was performed by PCR using genomic DNA isolated from tail clips and transgene specific primers (Table 3.1). Mammary Organoid Culture. Mammary organoids were generated from primary mammary epithelium through mechanical and enzymatic digestion and differential centrifugation. Briefly, following mechanical dissociation, mammary tissue was partially digested in a solution of Dulbecco's Modified Eagle's Media/F12 (DMEM/F12; Gibco) containing 300U/mL collagenase (Sigma) and 100U/mL hyaluronidase (Sigma) for 20 minutes in a 37 C incubated shaker set at 100 rpm. The digested tissue was centrifuged at 180x g for 10 minutes at 4 C to pellet mammary epithelium, while the supernatant containing the collagenase/hyaluronidase solution with adipose tissue and stromal elements of 99

120 the mammary fat pad was discarded. The cell pellet was resuspended in 10 ml DMEM/F12 and centrifuged again at 180x g to remove any remaining adipose from the sample. To remove free-floating DNA, the pellet was resuspended in 4 ml DMEM/F12 plus 30 μl Deoxyribonuclease I (DNase, 2.6U/ μl; Worthington) and incubated for 5 minutes at room temperature, with occasional shaking. The DNase solution was removed following a 10 minute spin at 180x g. Finally, mammary organoids were separated from single cells through differential centrifugation: pellets resuspended in 10 ml DMEM were spun at 400x g for 1 minute, and the supernatant was discarded. This was repeated 2 additional times. Following the final centrifugation steps, organoids were filtered through a 500 μm mesh to remove large clumps, and resuspended in Growth Factor Reduced Matrigel 3D basement membrane matrix (GFR Matrigel, BD Biosciences) for plating. Organoids were plated in wells of a 96 well cell culture plate that were pre-coated with a 10 μl GFR Matrigel underlay (40 μl organoid suspension/well). Once the Matrigel surrounding the plated organoids was set, 200 μl of culture media was added to each well (Culture Media: DMEM/F12 containing 2x antibiotic/antimycotic and ITS (10 μg/ml insulin, 5.5 μg/ml transferrin, and 2 ng/ml sodium selenite, all from Sigma). Plated organoids were allowed to equilibrate overnight in a 37 C incubator before initiating experiments. Culture Media was exchanged every 2 days. For Dox treatment, Doxycycline (1000 ng/μl, Sigma) was added to the Culture Media. 100

121 Chimeric Mammary Organoid Culture. Chimeric mammary organoids were generated by combining the primary mammary epithelium from two separate donor mice, ultimately incorporating single MECs from epithelia of one genotype onto scaffold organoids of another genotype. Specifically, mammary epithelium from one donor was partially dissociated into mammary organoids, as described above. Meanwhile, in a separate preparation, mammary epithelium from the second donor was fully dissociated into a single cell suspension. Briefly, mechanically dissociated epithelium was enzymatically digested in a DMEM/F12 solution containing 450U/mL collagenase and 100U/mL hyaluronidase for 1 hour in a 37 C incubated shaker set at 100 rpm. Resulting cells were centrifuged twice at 2000x g for 8 minutes, discarding the supernatant solution each time. Finally, resulting single cells were resuspended in DMEM/F12 and passed through a 35 μm cell strainer. Both preparations were combined in 3 ml DMEM/F12 plus 5x antibiotic/antimycotic and plated together in one well of a 6 well cell culture plate. The plate was then rotated for 3 hours at 150 rpm on a rotary shaker at 37 C. The cell suspension was removed from the plate, and the well was rinsed several times with 1 ml DMEM/F12. Each rinse was added to the sample until a total volume of 10 ml was reached. The sample was then centrifuged at 400x g for 2 minutes. The resulting pellet was resuspended in 10 ml DMEM/F12 and passed through the 500 μm mesh to remove any large clumps. The sample was spun down one last time and resuspended in GFR Matrigel for plating, as above. Chimeric organoid preparations contain varying amounts of 3 different types of 101

122 organoids, which are all cultured together: 1) chimeric organoids that incorporated epithelium from both donors, 2) scaffold organoids which contain only epithelium from the first donor, and 3) assembled organoids which contain only epithelium from the second donor. Because different fluorescent reporters are used to label the two donor epithelia, the three types of organoids can be readily classified according to their fluorescent expression. Imaging. Multi-parameter time lapse images were taken using a Zeiss AxioVision Observer microscope and AxioCam MRm camera. An incubated chamber surrounding the microscope stage together with a Zeiss Temp Module S, CO2 Module S, and Heating Unit XL were used to maintain culture conditions at 37 C with 5% carbon dioxide. A robotic stage with AxioVision Multidimensional Imaging Modules was used to track multiple individual organoids within the same experiment, and images were taken of each selected organoid every 15 minutes for 7 days. Acquired images were saved in daily blocks. Each day, the z-axis parameter was adjusted for better image focus. Image Analysis. Organoid size was determined by tracing the organoid boarders and calculating 2D area measurements using AxioVision 4.8 software. Mitosis and cell death events were quantitated by assembling daily image blocks into movies and tallying the number of events observed during frame-by-frame analysis. 102

123 3.4 Results. Expression of activated Ras and Ectopic Wnt1 produce distinct phenotypes in mammary organoid culture. The teto-hras (hereafter, TRAS) mouse model is a tet-operator regulated model of inducible HRAS G12V expression. HRAS G12V is an activated HRas allele which confers constitutive GTPase activity of the HRas protein. HRAS G12V has a single amino acid substitution (glycine-->valine) at codon 12 in the HRas coding sequence. Notably, an analogous activating mutation is represented among the somatic mutations commonly identified as cooperating mutations in mouse breast cancer models. When paired with the MMTV-rtTA transactivator strain, MMTVrtTA/TRAS (luminal RAS ) bitransgenic animals exhibit luminally-restricted HRAS G12V expression 224. In order to track the cell-autonomous effects of oncogenic HRas expression, organoids were derived from MMTV-rtTA/TRAS/TGFP (luminal RAS/GFP ) tritransgenic mice, plated in Growth Factor Reduced Matrigel and cultured for one week in the presence or absence of Dox. Multiparameter time lapse images were taken of several representative organoids every 15 minutes for the course of the week. Dox naïve tritransgenic organoids underwent very modest morphologic changes and did not grow appreciably in size (Figure 3.4a). Similarly, while Dox treated oncogene naïve MMTV-rtTA/ TGFP (luminal GFP ) organoids displayed nuclear TGFP expression within several hours of Dox induction, they failed to increase appreciably in size or exhibit any morphologic 103

124 phenotype. (Figure 3.4a) However, in agreement with previous work from our lab 97, expression of TRAS in the luminal cell compartment reliably produced dramatic, Dox dependent phenotypes in luminal RAS/GFP organoids. Specifically, luminally-restricted TRAS expression led to significant increases in organoid area, and organoids developed numerous aberrant branched structures which were distinct from the physiologic branches of growth-factor induced organoids (Figure 3.4a, b). Additionally, several hours following Dox induction, nuclear TGFP expression was discernable within the TRAS-expressing cell subset. Time lapse images collected over the week long experiment were assembled into videos, and these were used to count the individual mitotic and cell death events that occurred within the labeled TRAS expressing cell population of Dox treated tritransgenic organoids (Figure 3.4c). In luminal RAS/GFP organoids, the number of mitotic events within the labeled cell population increased as compared with oncogene naive luminal GFP and controls, and this increase in cell divisions was sustained over several days of Dox treatment. In contrast, the number of cell death events remained unchanged (Figure 3.4c). Thus TRAS induced organoid overgrowth can be attributed to increased cell divisions within the luminal TRAS-expressing cell compartment. Analogous to the TRAS model, the tet-o-wnt1 mouse model (hereafter, TWNT) features the tet-operator-controlled expression of Wnt1. When crossed to an MMTV-rtTA transactivator strain, Dox-fed MMTV-rtTA/TWNT (luminal WNT ) bitransgenic mice express Wnt1 transgene in a subset of MMTV-expressing luminal MECs, phenocopying the constitutive MMTV-Wnt1 model 177. As before, 104

125 organoids were derived from MMTV-rtTA/TWNT/TGFP (luminal WNT/GFP ) tritransgenic mice, and cultured for one week in Growth Factor Reduced Matrigel. Multiparameter time lapse images were taken of several representative organoids every 15 minutes for the course of the week. Dox naïve tritransgenic organoids underwent very modest morphologic changes and did not grow appreciably in size (Figure 3.5a). Likewise, MMTV-rtTA/ TGFP (luminal GFP ) oncogene naïve controls exhibited nuclear TGFP expression within several hours of Dox induction but failed to increase in area or produce a morphologic phenotype (Figure 3.5a). In contrast, Dox treated luminal WNT/GFP organoids reliably displayed significant Dox dependent growth (Figure 3.5a, b). Additionally, while they did not form branches, they did develop a laminar, almost onion-skinned appearance, and by the end of the week, most developed a dark central crater region of unknown significance (Figure 3.5a). Cells continued to be present within the dark crater regions, making them distinct from the physiologic lumens which form in organoids in the absence of oncogene expression. Nuclear TGFP labeling of TWNT-expressing cells was detected within several hours following initiation of Dox treatment, and individual mitosis and cell death events were counted from frame-by-frame analysis of time lapse videos (Figure 3.5c). There was no change in the number or mitosis or cell death events in luminal WNT/GFP compared to oncogene naïve luminal GFP controls. In fact, as the luminal WNT/GFP organoids became larger, the TGFP expressing cell population progressively became the minority cell population. Thus growth of luminal WNT/GFP organoids 105

126 cannot be explained by expansion of the TWNT-expressing cell population via autocrine Wnt1 activity. Luminally-derived Wnt1 drives selective expansion of the basal epithelial compartment. Because expression of TWNT did not appear to drive organoid overgrowth in a cell-autonomous manner, we hypothesized that luminally-derived Wnt1 promoted organoid growth through a paracrine signaling mechanism. To test this, we generated chimeric mammary organoids that allowed us to track the effects of luminal TWNT expression within the neighboring wild type cells (Figure 3.6). First, we tested whether luminally-derived Wnt1 affects cell fate events in the surrounding basal epithelial cells. Luminal WNT mice were crossed with an mrfp reporter line, to generate red-labeled luminal WNT/RFP tritransgenic animals. Mammary epithelium collected from a donor luminal WNT/RFP mouse was fully dissociated into a suspension of single cells. Meanwhile, epithelium from a second basal GFP donor animal was partially disaggregated to produce mammary organoids. The cell preparations from both donor animals were combined and incubated together in rotary culture for several hours before being plated in 3D Matrigel culture in the presence of Dox. The basal GFP organoids served as a scaffold for the luminal WNT/RFP cells, and the result was chimeric mammary organoids in which TWNT was expressed in red-labeled luminal epithelial cells while nuclear TGFP expression was restricted to the basal epithelial cell compartment (Figures 3.6 and 3.7a). In this way, it was possible to track the cell 106

127 non-autonomous effects of luminally-derived Wnt1 in the surrounding basal cell population. In addition to chimeric organoids, the preparation contained assembled organoids which self-assembled from the luminal WNT/RFP single cells, as well as basal GFP scaffold organoids which failed to incorporate any additional cells. All three types of organoids were plated together in 3D Matrigel culture in the presence of Dox for one week. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). As before, organoids were cultured for one week with images collected every fifteen minutes. Dox-treated chimeric basal GFP :: luminal WNT/RFP organoids produced the same morphologic phenotype as non-chimeric luminal WNT organoids: they grew in size, developed a laminar onion-skin appearance, and developed a dark central crater (Figure 3.7a). Chimeric basal GFP :: luminal WNT/RFP organoids grew to the same extent as their non-chimeric luminal WNT/RFP counterparts (Figure 3.7b). Interestingly, the growth effects of secreted Wnt1 were restricted to those organoids that harbored Wnt-1 producing cells (basal GFP :: luminal WNT/RFP chimeric organoids and luminal WNT/RFP assembled organoids). By contrast, scaffold basal GFP organoids, even those immediately adjacent to the other Wnt1-producing organoids from the same preparation, remained unaffected by secreted Wnt1. Instead, the scaffold basal GFP organoids behaved similarly to oncogene naïve controls: over the course of the week, scaffold organoids exhibited unremarkable growth with only modest morphologic changes (Figure 3.7a). This is consistent with the prevailing notion that Wnts provide very short- 107

128 range signals. Basal expression of nuclear TGFP permitted tracking of the cell fate events that occurred within this cell population. Frame-by-frame analysis of time lapse videos revealed that significantly more mitotic events occurred within the basal cell population of the chimeric organoids than in the basal GFP scaffold organoids from the same preparation (Figure 3.7c). What s more, this increased mitotic rate was sustained over several days of Dox treatment. No difference was observed in the number of cell death events (Figure 3.7c). Next, we asked whether luminally-derived Wnt1 also acted upon the neighboring luminal cells which do not, themselves, express TWNT. We combined the mammary epithelium from luminal GFP and luminal TWNT/RFP donor animals to generate luminal GFP :: luminal WNT/RFP chimeras such that some of the luminal cells within the organoid expressed TWNT, while a separate population of luminal cells expressed the TGFP label to permit nuclear tracking. The chimeric organoid preparation was cultured in 3D Matrigel culture for one week. Dox treated chimeric luminal GFP :: luminal WNT/RFP and assembled luminal WNT/RFP organoids adopted the characteristic TWNT phenotype and developed a laminar appearance with central crater (Figure 3.8a). However, while the chimeric luminal GFP :: luminal WNT/RFP organoids did grow significantly larger than the luminal GFP scaffold organoids, they did not grow to the same extent as their nonchimeric luminal WNT/RFP assembled counterparts (Figure 3.8a,b). Analysis of time lapse videos showed no difference in the number of mitotic events or cell death events in labeled luminal cells between chimeric luminal GFP :: luminal WNT/RFP and scaffold luminal GFP organoids (Figure 3.8c). All together, these data suggest that 108

129 expression of Wnt1 within a luminal epithelial subset leads to a dramatic and selective expansion of the basal epithelial compartment. Luminally restricted TRAS expression drives cell proliferation cell autonomously. In luminal RAS/GFP organoids, an increased frequency of cell divisions was observed within the TRAS-expressing cell subset (Figure 3.4c). Next, we asked whether luminal TRAS expression wielded any additional effects on the neighboring wild type cells within the organoids. First, we generated chimeric organoids to determine whether expression of TRAS in a luminal cell population exerted any cell non-autonomous effects upon the surrounding basal epithelial cells. Mammary epithelium was collected from two donor animals: a basal GFP bitransgenic mouse and a luminal RAS/RFP tritransgenic mouse. As before, chimeric mammary organoids were generated such that red-labeled luminal RAS/RFP cells were incorporated into basal GFP scaffold organoids. With nuclear TGFP expression restricted to the basal epithelial compartment, we were able to track cell fate events that occurred in the basal epithelial cells as a result of luminal TRAS expression. Upon administration of Dox, chimeric basal GFP :: luminal RAS/RFP organoids grew significantly and developed numerous aberrant branch structures, reproducing the characteristic TRAS phenotype (Figure 3.9a). Chimeric organoids grew to the same size as their non-chimeric "assembled" counterparts (Figure 3.9b). Basal GFP scaffold organoids which did not incorporate any TRASexpressing cells failed to exhibit any morphologic phenotype (Figure 3.9a). When time lapse videos were analyzed to score cell fate events, no differences were 109

130 observed in the number of cell divisions or cell death events in the chimeric organoids compared with the basal GFP scaffolds (Figure 3.9c). To determine whether luminal TRAS expression affects other surrounding wild type luminal cells, we generated luminal GFP :: luminal RAS/RFP chimeric organoids. In these organoids, red-labeled luminal RAS/RFP cells were incorporated into luminal GFP scaffolds, thereby permitting tracking of cell fate events in wild type luminal cells in response to luminal TRAS expression. Upon Dox administration, the chimeric organoids produced the characteristic TRAS morphologic phenotype (Figure 3.10a). They grew significantly, though not as large as the non-chimeric luminal RAS/RFP assembled organoids (Figure 3.10b). Again, there were no differences in the number of cell divisions or cell death events in the chimeric organoids compared with the luminal GFP scaffold organoids (Figure 3.10c). Basally restricted TRAS expression acts cell autonomously to drive organoid overgrowth. Recently, we reported the presence of a basally restricted somatic HRas mutation in a subset of tumors from a mouse model of mammary tumorigenesis (Chapter 3). Notably, this basal HRas mut clone cooperated with a Wnt1-producing luminal tumor cell clone to maintain tumor growth. Accordingly, we asked whether basally-restricted TRAS expression produced any non-autonomous effects that might facilitate such an interaction across tumor cell subsets. 110

131 First, we generated K5-rtTA/TRAS/TGFP (basal RAS/GFP ) tritransgenic organoids to document the cell autonomous effects of basally-restricted TRAS expression. As before, basal RAS/GFP organoids were grown in 3D Matrigel culture for one week in the presence of Dox. Consistent with previous work 224, expression of TRAS in the basal cell compartment reproducibly produced dramatic Dox dependent phenotypes (Figure 3.11a). Organoid area increased significantly, and organoids developed characteristic aberrant branched structures (Figure 3.11a,b). In contrast, both Dox naïve basal RAS/GFP and oncogene naïve K5- rtta/tgfp (basal GFP ) organoids failed to grow appreciably and exhibited only modest morphologic changes. Several hours following Dox induction, nuclear TGFP expression was detectable within the basal TRAS expressing cell population of basal RAS/GFP organoids. Individual mitotic and cell death events scored via frame-by-frame analysis of time lapse videos revealed an initial increase in the number of mitotic events within basal RAS/GFP organoids compared with their oncogene naïve basal GFP counterparts (Figure 3.11c). Additionally, fewer cell death events were detected in the basal RAS/GFP organoids at early time points (Figure 3.11c). Thus, basally restricted TRAS expression triggered cellautonomous cell proliferation combined with a decreased incidence of cell death, resulting in malignant organoid overgrowth. Next, we tested whether basally-restricted TRAS expression induced any cell non-autonomous effects in the surrounding luminal or basal MECs. We generated chimeric mammary organoids in which TRAS expression was restricted to the basal epithelial compartment, while the luminal compartment expressed the 111

132 nuclear TGFP reporter transgene to permit tracking of cell fate events. To do this, we combined single cells from red-labeled basal RAS/RFP transgenic mammary epithelium with scaffold organoids generated from luminal GFP bitransgenic animals. Upon administration of Dox, chimeric luminal GFP :: basal RAS/RFP organoids grew in size and produced numerous aberrant branched structures, reminiscent of the morphologic phenotype of non-chimeric basal RAS organoids (Figure 3.12a). However, the chimeric organoids did not grow as large as the nonchimeric basal RAS/RFP assembled organoids (Figure 3.12b). Luminal GFP scaffold organoids which did not incorporate any TRAS-expressing cells did not increase in area or exhibit a morphologic phenotype (Figure 3.12a). Analysis of time lapse videos showed no differences in the number of cell divisions or cell death events in the chimeric organoids compared with the luminal GFP scaffold organoids (Figure 3.12c). Finally, we generated basal GFP :: basal RAS/RFP chimeric organoids in which red-labeled basal RAS/RFP cells were incorporated into basal GFP scaffolds, thereby permitting tracking of cell fate events within the surrounding basal cells in response to basal TRAS expression. Again, upon Dox administration, the chimeric organoids produced the characteristic TRAS morphologic phenotype. They grew significantly, though not as large as the non-chimeric basal RAS/RFP assembled organoids, and produced numerous aberrant branched structures (Figure 3.13a,b). However, there were no differences observed in the number of cell divisions or cell death events in the chimeric organoids compared with the basal GFP scaffold organoids (Figure 3.13c). 112

133 Together, these data suggest that TRAS expression has a strictly cellautonomous effect on cell fate events. Unlike in the case of TWNT expression, where luminally-produced Wnt1 lead to a selective and robust expansion of the basal epithelial compartment, TRAS expression in either the luminal or basal MEC compartment produced a dramatic morphologic phenotype driven by an increase in cell-autonomous proliferation, but had no effect on the surrounding luminal or basal cells. 3.5 Discussion. Here, we present 3D chimeric mammary organoid culture as a sensitive and efficient method for discerning both the cell-autonomous and nonautonomous effects of oncogene expression in mammary epithelial cells, as well as for investigating the complex cell-cell interactions that facilitate malignant overgrowth. With this method, we determined a strictly cell-intrinsic role of oncogenic Ras expression in both basal and luminal epithelial subsets. We also successfully detected a paracrine role for oncogenic Wnt1 signaling. Specifically, luminally-derived Wnt1 acted in a paracrine manner to drive selective expansion of the basal epithelial population. This is in agreement with previous studies that report an increased proportion of basal cells by flow cytometry in hyperplastic mammary glands from MMTV-Wnt1 animals 272. However, an increase in basal cell proportion can be accomplished by any of three different mechanisms: 1) increased cell divisions within the basal cell compartment, 2) increased cell death in the luminal cell compartment, or 3) trans- 113

134 differentiation of cells across epithelial compartments. Here, through real time visualization of mitotic events, we confirmed that Wnt1 induced expansion of the basal epithelial population is indeed due to increased cell divisions within that cell compartment. While chimeric organoids faithfully recapitulated the morphologic phenotypes of both the Wnt and Ras oncogenic transgene expression, increases in chimeric organoid area did not always match that of their non-chimeric assembled counterparts. A possibility for this discrepancy could be a difference in the cumulative oncogenic dose. In general, because chimeric organoids represent a mixture of donor epithelia, they tend to contain a variable number of the oncogene expressing cells. If fewer total oncogene expressing cells incorporate into the chimeric organoids, overall oncogene mediated overgrowth might be reduced proportionately, thus resulting in decreased organoid growth. The canonical Wnt/β-catenin and Ras/Mitogen-Activated Protein Kinase (MAPK) pathways are two of the most commonly activated signaling pathways in cancer. Co-activation of these two pathways has been observed in a number of different tumor types including colorectal 250,273, lung 254, skin 249, prostate 274, bladder 275, and pancreatic 276, suggesting a synergistic relationship. However, the mechanism by which these two pathways interact is unclear. There is strong evidence for cooperation between the Wnt and Ras signaling pathways in the MMTV-Wnt1 mouse mammary tumor model. Recently, we reported that for some MMTV-Wnt1 tumors, activating somatic mutations in HRas were restricted to the basal tumor cell subset (see Chapter 4). Given the selective mitogenic effect 114

135 of MMTV-driven Wnt1 expression on the basal epithelial compartment of mammary organoids, it is possible that a cooperative interaction between Wnt and Ras may occur solely within the basal tumor cell subset of MMTV-Wnt1 tumors. While more work is needed to determine the specific mechanism of Wnt/Ras cooperation in MMTV-Wnt1 tumors, Plichta et al. showed that basal (but not luminal) activation of Ras induced organoid overgrowth via the PI3K-Akt-mTOR pathway 224. Similarly, MMTV-Wnt1 tumors arising in a Pten heterozygous background displayed a decreased tumor latency, and these tumors frequently exhibited LOH of Pten 125,277. Thus, for some MMTV-Wnt1 tumors, cooperation between Wnt and Ras may potentially be facilitated through the Ras-PI3K-Akt signaling pathway. 115

136 Figure 3.1: Tet-operator driven transgenes. Transgenes under the control of tet operator sequences were paired with either a luminal MEC transactivator (MMTV-rtTA) or a basal MEC transactivator (K5-rtTA) to direct compartment restricted expression. This system permits Doxinducible expression of oncogenic transgenes: tet-operator-controlled H-RAS G12V (TRAS) and tet-operator-controlled Wnt1 (TWNT), as well as a tet- operator-controlled H2B-GFP nuclear reporter transgene (TGFP). 116

137 Figure 3.2: Ras signaling pathway. Activated Ras signals through the MAP Kinase signaling cascade, or the Pi3K-Akt-mTOR pathway to promote cell growth, proliferation, and survival. Figure modified from Downward J (2008). Nature Medicine 14: Adapted by permission from Macmillan Publishers Ltd. 117

138 Figure 3.3: Canonical Wnt Signaling Pathway. The canonical Wnt signaling pathway results in the stabilization of β-catenin. Stabilized β- catenin can then act as a transcriptional activator. Figure modified from Chien AJ, Conrad WH, and Moon RT (2009). J Invest Dermatol. 129(7): Adapted by permission from Macmillan Publishers Ltd. 118

139 Figure 3.4: Time-lapse imaging of Ras-mediated MEC overgrowth. Organoids derived from Dox-naïve tri-transgenic Luminal RAS/GFP and bitransgenic Luminal GFP mice were plated separately in 3D Matrigel culture on the stage of a live cell imaging station in the presence or absence of Dox, as indicated. A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield and GFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a) In tri-transgenic Luminal RAS/GFP organoids, Dox treatment triggered a robust, highly reproducible TRAS transgene-mediated morphologic transition characterized by significant organoid growth and aberrant organoid branching. Oncogene naïve Luminal GFP and Dox naïve Luminal RAS/GFP displayed only modest morphologic changes. b) Axiovision software was used to calculate organoid area at the indicated time points. n= 7-24 organoids/ time point for each condition. Error bars represent SEM. c) Time-lapse images were assembled into videos and used to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= 4-22 organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=4-22 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Dox-treated Luminal GFP group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with Dox naïve Luminal RAS/GFP group, p<0.05; p<0.01; p<0.001; p<

140 120

141 Figure 3.5: Time-lapse imaging of Wnt1-mediated MEC overgrowth. Organoids derived from Dox-naïve tri-transgenic Luminal WNT/GFP and bitransgenic Luminal GFP mice were plated separately in 3D Matrigel culture on the stage of a live cell imaging station in the presence or absence of Dox, as indicated. A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield and GFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a. In tri-transgenic Luminal WNT/GFP organoids, Dox treatment triggered a robust, highly reproducible TWNT transgene-mediated morphologic transition characterized by significant organoid growth. Oncogene naïve Luminal GFP and Dox naïve Luminal WNT/GFP displayed only modest morphologic changes. b. Axiovision software was used to calculate organoid area at the indicated time points. n= 7-28 organoids/ time point for each condition. Error bars represent SEM. c. Time-lapse images were assembled into videos and used to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n=7-28 organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=7-28 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Dox-treated Luminal GFP group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with Dox naïve Luminal WNT/GFP group, p<0.05; p<0.01; p<0.001; p<

142 122

143 Figure 3.6: Generation of Chimeric Mammary Organoids. a. Organoids derived from the mammary epithelium of Donor 1 were combined with a single cell suspension prepared from the mammary epithelium of Donor 2 and placed in rotary culture for several hours before being plated in 3D Matrigel culture. The resulting chimeric organoids consisted of the scaffold organoid from Donor 1 with incorporated single cells from Donor 2. Using donor epithelia that expressed different fluorescent reporters permitted discrimination between distinct donor populations. 123

144 Figure 3.7: Luminal expression of TWNT leads to increased cell divisions in the basal epithelial compartment. Scaffold organoids derived from Dox-naïve Basal GFP mammary glands were combined with single cells prepared from Dox-naïve mammary epithelia of a Luminal WNT/RFP mouse to generate Basal GFP :: Luminal WNT/RFP chimeric organoids. In addition to chimeric organoids, the prep contained assembled organoids which self-assembled from the Luminal WNT/RFP single cells, as well as Basal GFP scaffold organoids which did not incorporate any additional cells. All three types of organoids were plated together in 3D Matrigel culture in the presence of Dox. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield, GFP and RFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a. Chimeric and assembled organoids reproduced the TWNT morphologic phenotype of MEC overgrowth, while scaffold organoids displayed only modest morphologic changes. b. Axiovision software was used to calculate organoid area at the indicated time points. Chimeric and assembled organoids grew significantly larger than scaffold organoids. n= organoids/ time point for each condition. Error bars represent SEM. c. Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= 4-16 organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=4-16 organoids analyzed for each condition over each time interval. Error bars represent SEM. Significantly more mitotic events occurred in chimeric Basal GFP :: Luminal WNT/RFP organoids than in scaffold Basal GFP organoids, while cell death event numbers were unaffected. Statistics: compared with Doxtreated Luminal GFP group, * p<0.05; ** p<0.01; *** p<0.001; **** p<

145 125

146 Figure 3.8: Luminal expression of TWNT does not influence cell fate events in the surrounding luminal epithelial compartment. Scaffold organoids derived from Dox-treated Luminal GFP mammary glands were combined with single cells prepared from Dox-naïve mammary epithelia of a Luminal WNT/RFP mouse to generate Luminal GFP :: Luminal WNT/RFP chimeric organoids. In addition to chimeric organoids, the prep contained assembled organoids which self-assembled from the Luminal WNT/RFP single cells, as well as Luminal GFP scaffold organoids which did not incorporate any additional cells. All three types of organoids were plated together in 3D Matrigel culture in the presence of Dox. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield, GFP and RFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a. Chimeric and assembled organoids reproduced the TWNT morphologic phenotype of MEC overgrowth, while scaffold organoids displayed only modest morphologic changes. b. Axiovision software was used to calculate organoid area at the indicated time points. n= organoids/ time point for each condition. Error bars represent SEM. c. Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=11-20 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Luminal GFP scaffold group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with chimeric Luminal GFP :: Luminal WNT/RFP group, p<0.05; p<0.01; p<0.001; p<

147 127

148 Figure 3.9: Luminal expression of TRAS does not influence cell fate events in the surrounding basal epithelial compartment. Scaffold organoids derived from a Dox-treated Basal GFP mouse were combined with single cells prepared from mammary epithelia of a Luminal RAS/RFP mouse to generate Basal GFP :: Luminal RAS/RFP chimeric organoids. Chimeric, assembled, and scaffold organoids were plated together in 3DMatrigel culture in the presence of Dox. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield, GFP and RFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a) Chimeric and assembled organoids reproduced the TRAS morphologic phenotype of MEC overgrowth and aberrant branches, while scaffold organoids displayed only modest morphologic changes. b) Axiovision software was used to calculate organoid area at the indicated time points. n= organoids/ time point for each condition. Error bars represent SEM. c) Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=7-29 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Basal GFP scaffold group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with chimeric Basal GFP :: Luminal RAS/RFP group, p<0.05; p<0.01; p<0.001; p<

149 129

150 Figure 3.10: Luminal expression of TRAS does not influence cell fate events in the surrounding luminal epithelial compartment. Scaffold organoids derived from a Dox-treated Luminal GFP mouse were combined with single cells prepared from mammary epithelia of a Luminal RAS/RFP mouse to generate Luminal GFP :: Luminal RAS/RFP chimeric organoids. Chimeric, assembled, and scaffold organoids were plated together in 3DMatrigel culture in the presence of Dox. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield, GFP and RFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a) Chimeric and assembled organoids reproduced the TRAS morphologic phenotype of MEC overgrowth and aberrant branches, while scaffold organoids displayed only modest morphologic changes. b) Axiovision software was used to calculate organoid area at the indicated time points. n= organoids/ time point for each condition. Error bars represent SEM. c) Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n=11-30 organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=11-30 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Basal GFP scaffold group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with chimeric Basal GFP :: Luminal RAS/RFP group, p<0.05; p<0.01; p<0.001; p<

151 131

152 Figure 3.11: Time-lapse imaging of Ras-mediated basal MEC overgrowth. Organoids derived from Dox-naïve tri-transgenic Basal RAS/GFP and bitransgenic Basal GFP mice were plated separately in 3D Matrigel culture on the stage of a live cell imaging station in the presence or absence of Dox, as indicated. A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield and GFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a) In tri-transgenic Basal RAS/GFP organoids, Dox treatment triggered a robust, highly reproducible TRAS transgenemediated morphologic transition characterized by significant organoid growth and aberrant organoid branching. Oncogene naïve Basal GFP and Dox naïve Basal RAS/GFP displayed only modest morphologic changes. b) Axiovision software was used to calculate organoid area at the indicated time points. n= organoids/ time point for each condition. Error bars represent SEM. c) Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= 6-24 organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=8-20 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Dox-treated Basal GFP group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with Dox naïve Luminal RAS/GFP group, p<0.05; p<0.01; p<0.001; p<

153 133

154 Figure 3.12: Basal expression of TRAS does not influence cell fate events in the surrounding luminal epithelial compartment. Scaffold organoids derived from a Dox-treated Luminal GFP mouse were combined with single cells prepared from mammary epithelia of a Basal RAS/RFP mouse to generate Luminal GFP :: Basal RAS/RFP chimeric organoids. Chimeric, assembled, and scaffold organoids were plated together in 3DMatrigel culture in the presence of Dox. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield, GFP and RFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a) Chimeric and assembled organoids reproduced the TRAS morphologic phenotype of MEC overgrowth and aberrant branches, while scaffold organoids displayed only modest morphologic changes. b) Axiovision software was used to calculate organoid area at the indicated time points. n= 6-21 organoids/ time point for each condition. Error bars represent SEM. c) Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=9-18 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Luminal GFP scaffold group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with chimeric Luminal GFP :: Basal RAS/RFP group, p<0.05; p<0.01; p<0.001; p<

155 135

156 Figure 3.13: Basal expression of TRAS does not influence cell fate events in the surrounding basal epithelial compartment. Scaffold organoids derived from a Dox-treated Basal GFP mouse were combined with single cells prepared from mammary epithelia of a Basal RAS/RFP mouse to generate Basal GFP :: Basal RAS/RFP chimeric organoids. Chimeric, assembled, and scaffold organoids were plated together in 3DMatrigel culture in the presence of Dox. The organoids were classified by their fluorescent expression: chimeric (red + green), assembled (solid red), and scaffold (solid green). A robotic stage and multiparameter imaging software was used to capture time-lapse images in the brightfield, GFP and RFP channels at 15 minute intervals. Each row depicts serial images of a representative organoid. Size bar =100μm. a) Chimeric and assembled organoids reproduced the TRAS morphologic phenotype of MEC overgrowth and aberrant branches, while scaffold organoids displayed only modest morphologic changes. b) Axiovision software was used to calculate organoid area at the indicated time points. n= 6-19 organoids/ time point for each condition. Error bars represent SEM. c) Time-lapse images were assembled into videos to detect individual cell fate events occurring within the indicated intervals. For mitotic events, n= organoids analyzed for each condition over each time interval. Error bars represent SEM. For cell death events, n=8-20 organoids analyzed for each condition over each time interval. Error bars represent SEM. Statistics: compared with Basal GFP scaffold group, * p<0.05; ** p<0.01; *** p<0.001; **** p< Compared with chimeric Basal GFP :: Basal RAS/RFP group, p<0.05; p<0.01; p<0.001; p<

157 137

158 Genotyping Primers Gene Primer Sequence MMTV-rtTA F1: 5 - TGCCGCCATTATTACGACAAGC- 3 R1: 5 - ACCGTACTCGTCAATTCCAAGGG- 3 F2: 5 - ATCCGCACCCTTGATGACTCCG- 3 R2: 5 - GGCTATCAACCAACACACTGCCAC- 3 K5-rtTA F1: 5 - TGCCGCCATTATTACGACAAGC- 3 R1: 5 - ACCGTACTCGTCAATTCCAAGGG- 3 F2: 5 - AGGCTGAATTAGGAGGCGTTG- 3 R2: 5 - GACACTGCCTGTGAGGCCTG- 3 teto-h2bgfp F1: 5 - AAGTTCATCTGCACCACCG- 3 F2: 5 - CCTTGATGCCGTTCTTCTGCTTGTC- 3 R: 5 - TCCTTGAAGAAGATGGTGCG- 3 teto-wnt1 F1: 5 - TGCGGTTCCTGATGTATTTTGC- 3 (TWNT) R1: 5 - TGCATTCCTTTGGCGAGAGG- 3 F2: 5 - CAGGAAATTGCTTCTGGTGGC- 3 teto-hras (TRAS) mrfp R2: 5 - TCGAAGATGTTGGGGTGTTGG- 3 F1: 5 - TGAAAGTCGAGCTCGGTA- 3 R1: 5 - TCTGGATCAGCTGGATGGTC- 3 F2: 5 - GGTGTGGGCAAGAGTGCG- 3 R2: 5 - CCTGAGCCTGCCGAGATTCC- 3 Genotyped by assessing for red fluorescence in tail clips under fluorescent microscope. Table 3.1: Genotyping Primers. Genotyping of transgenic mouse lines was performed by PCR using genomic DNA isolated from tail clips and transgene specific primers. F indicates the forward primer. R indicates the reverse primer. 138

159 Chapter 4 Tumor Cell Heterogeneity Maintained by Cooperating Subclones in Wnt-Driven Mammary Cancers Allison S. Cleary, Travis L. Leonard, Shelley A. Gestl, and Edward J. Gunther 139

160 4.1 Abstract. Cancer genome sequencing studies indicate that a single breast cancer typically harbors multiple genetically distinct subclones 1-5. Since carcinogenesis involves a breakdown in the cell-cell cooperation that normally maintains epithelial tissue architecture, individual subclones within a malignant microenvironment are commonly depicted as self-interested competitors 104,133. Alternatively, breast cancer subclones might interact cooperatively to gain a selective growth advantage in some cases. Although interclonal cooperation has been shown to drive tumorigenesis in fruitfly models 150,278, definitive evidence for functional cooperation between epithelial tumor cell subclones in mammals is lacking. Here, we use mouse models of breast cancer to show that interclonal cooperation can be essential for tumor maintenance. Aberrant expression of the secreted signaling molecule Wnt1 generates mixed-lineage mammary tumors composed of basal and luminal tumor cell subtypes, which purportedly derive from a bipotent malignant progenitor cell residing atop a tumor cell hierarchy 125. Using somatic HRas mutations as clonal markers, we show that some Wnt tumors indeed conform to a hierarchical configuration, but others unexpectedly harbor genetically distinct basal HRas mutant (HRas mut ) and luminal HRas wild-type (HRas wt ) subclones. Both subclones are required for efficient tumor propagation, which strictly depends on luminally-produced Wnt1. When biclonal tumors were challenged with Wnt withdrawal to simulate targeted therapy, analysis of tumor regression and relapse revealed that basal subclones recruit heterologous Wnt-producing cells to restore tumor growth. Alternatively, in the absence of a substitute Wnt 140

161 source, the original subclones often evolve to rescue Wnt pathway activation and drive relapse, either by restoring cooperation or by switching to a defector strategy. Uncovering similar modes of interclonal cooperation in human cancers may inform efforts aimed at eradicating tumor cell communities. 4.2 Introduction. Tumor cells within individual cancers typically display remarkable heterogeneity with respect to cell morphology, proliferation rate, metastatic potential, drug sensitivity, and capacity for tumor reconstitution. Recent cancer genome studies add to this complexity; revealing multiple genetically distinct subclonal tumor cell populations coexisting within tumors 1-5. Infiltration of normal stromal, immune, and vascular cells into the tumor microenvironment further adds to the intra-tumor diversity. While cancer progression is known to depend on cooperation between tumor cells and neighboring host cells in the microenvironment, very little is known about how the heterogeneity within the tumor cell compartment influences tumor maintenance and progression. Tumor cell subclones are generally thought to arise as the result of a branched clonal evolution process 104,133. As such, they are commonly depicted as self-interested competitors in a Darwinian system permitting survival of only the fittest clones. Alternatively, some have suggested that cooperation between distinct tumor cell subsets may contribute to the malignant phenotype 6,7,279. In this way, cooperation between subclones may impart a unique fitness advantage favoring co-expansion of both clones, thereby actively maintaining the subclonal 141

162 diversity within the tumor. Supporting this possibility, genetically distinct subclones cooperatively enhanced tumor growth in models engineered to recapitulate a form of tumor cell heterogeneity identified in brain cancers 148. Similarly, phenotypically distinct tumor cell subsets cooperatively enhanced tumor invasion in a murine lung cancer model 149. In the case of human breast cancer, recent studies highlight the phenotypic and genetic diversity present locally within individual tumors 110,280, but whether this heterogeneity is a cause or a consequence of tumor progression remains unclear. Accordingly, we sought definitive evidence for functional cooperation between tumor cell subsets in a classic mouse models of human breast cancer. Tumors arising in the classical MMTV-Wnt1 mouse mammary tumor model 47 contain apparent cellular heterogeneity: ectopic expression of the Wnt1 transgene in the mammary gland stochastically generates mixed-lineage tumors consisting of both basal and luminal epithelial cells. The mixed-lineage character of these tumors is generally thought to originate from transformation of a bipotent progenitor cell whose progeny can then differentiate into component basal and luminal tumor cell subpopulations Interestingly, these tumors maintain basal and luminal cells in a relatively fixed proportion, which is consistent with origination from a common progenitor, but is also suggestive of a potential interaction between the two populations 129. Although mutations in Wnt pathway components are rare in human breast cancers, the transcriptional profile of Wnt1- initiated tumors resembles that of other mammary cancers that commonly show 142

163 mixed-lineage histopathology, including chemical carcinogen-induced rodent mammary cancers and basal subtype human breast cancers 190,281,282. Ectopic Wnt1 expression produces a lobuloalveolar hyperplasia affecting all 10 mammary glands. However tumors in this model arise stochastically following a relatively long latency. Together, this suggests Wnt1 expression alone is not sufficient for tumor initiation, but instead serves as the initiating event in a multi-step process of carcinogenesis 283. Indeed, somatic mutations in HRas encoding an activated oncoprotein have been identified in just over half of the tumors arising in this model 284,285. Using these spontaneously occurring mutations in HRas as genetic tags to track the clonogenicty of tumor cell populations, we identified unexpected genetic heterogeneity within a subset of Wnt-driven mammary tumors. Indeed, for some Wnt-driven mammary tumors we detected distinct subclonal tumor cell populations that demonstrated a functional codependence. 4.3 Methods. Transgenic Mice. Mice were housed under pathogen-free conditions in the Pennsylvania State University College of Medicine rodent facility with access to water and chow ad libitum. All experimental protocols were approved by the Pennsylvania State University College of Medicine s Institutional Animal Care and Use Committee. The MMTV-Wnt1 (FVB.Cg-Tg(Wnt1)1Hev/J; stock #002934) and mrfp (B6.Cg-Tg(CAG-mRFP1)1F1Hadj/J; stock #005884) transgenic lines were obtained from the Jackson Labs. The MMTV-rtTA and 143

164 teto-wnt1 transgenic lines were a gift from Dr. Lewis Chodosh. All mice either were generated in an inbred FVB/N background or were back-crossed 10 or more generations with FVB/N breeders before initiating experiments. Dox was administered by replacing standard mouse chow with chow containing 2g/kg Dox (Bio-serv). Genotyping was performed by PCR using genomic DNA isolated from tail clips and transgene specific primers (Table 4.1). Cell Sorting. Mammary tumors were dissociated into single cell suspensions through mechanical separation and enzymatic digestion as described 129. Dissociated tumor cells were enriched for Lin - (CD45 - / CD31 - / TER119 - / BP-1 - ) mammary epithelial cells with StemCell Technologies EasySep Mouse Epithelial Cell Enrichment Kits per the manufacturer s instructions. Lin - cells were then incubated on ice for 20 min with anti-cd49f (α6 integrin) (BD Biosciences ) together with Alexafluor 647 (Invitrogen A21247) in PBS. Cells were spun down for 5 min at 550x g, then incubated with EpCAM-FITC conjugated antibody (Biolegend ) in PBS. Tumor cells were sorted on a BD FACS Aria cell sorter machine equipped with Diva software into their luminal (Lin - / CD49f low /EpCAM high ) and basal (Lin - /CD49f high /EpCAM low ) subpopulations. Sorted cells were collected into 15ml conical tubes containing PBS. Genomic DNA was collected from sorted cell populations using Qiagen Blood and Tissue DNeasy spin column kit. Total RNA was collected from sorted cell populations using Qiagen RNeasy spin column kit. RNA was reversed transcribed using Invitrogen Superscript II First Strand Synthesis kit. 144

165 Tumor reconstitution and propagation. In tumor reconstitution experiments, tumor cells for injection were counted using a hemocytometer and suspended at a concentration of 1000 cells/μl in a 50% Matrigel solution in PBS (BD Biosciences Growth Factor Reduced Matrigel Matrix) cells in 100 μl of Matrigel solution were injected directly into intact #3 or #4 mammary fat pads of anesthetized adult female hosts, after using a small skin incision to expose the injection site. Incisions were then closed with surgical clips. Mice were monitored at least twice weekly for tumor growth and tumors were measured in two dimensions with calipers. To generate a cohort of clonally related tumors for generating Dox Independent Tumors (DITs), tumor fragments were explanted onto the flanks of wild-type Dox-treated host mice. Tumors were permitted to grow to a diameter of 8-10 mm, at which point mice received a single i.p. injection of N-methyl-Nnitrosourea (MNU) (Sigma 50 mg/kg) to accelerate tumor relapse one week prior to Dox withdrawal. Without MNU treatment, only about a third of iwnt tumor explants relapsed during 12 months of continuous Dox withdrawal, and relapses arose after an average latency of 6 months. With MNU treatment, more than 90% of iwnt explants relapsed within 3 months of Dox withdrawal. Tumor regression and relapse was monitored at least twice weekly, and tumors were measured in two dimensions with calipers. DNA sequencing. Genomic DNA or copy DNA isolated from tumor specimens or sorted tumor cell populations was amplified by PCR using gene specific primers (Table 4.2). PCR product was run out on an agarose gel, cut out and isolated using Qiagen QiaQuick Gel Isolation spin column kit. Samples were 145

166 subjected to Sanger sequencing using gene specific primers on an ABI 3130XL Capillary sequencer machine. Sequence traces were analyzed using AB DNA Sequencing Analysis Software v5.2 and AB Sequence Scanner v1.0. Peak height (PH) on sequencing chromatograms was measured using ImageJ 1.46 software and HRas MAF was calculated using the following formula: MAF = PH Mutant / (PH Mutant + PH Wild-type ). Immunofluorescence. Tumor samples were fixed in 4% paraformaldehyde on ice for 2 hrs before being paraffin embedded. Paraffin sections (5μm) were stained with antibodies for smooth muscle actin (SMA) and Keratin 8, which label basal and luminal epithelial cells, respectively. Primary antibodies used were: rabbit anti-sma (AbCAM 5694, 1:250), and rat anti-keratin 8 (Troma-I) (Developmental Studies Hybridoma Bank, University of Iowa, 1:250). Secondary antibodies were: biotinylated rabbit-anti-rat IgG (Dako Cytomation E0468) and biotinylated goat anti-rabbit IgG (Vector BA-1000). The fluorophore was a streptavidin fluorescein (Vector SA-5001). Hoechst dye (Invitrogen H1399) was used for nuclear DNA counterstaining, and slides were visualized using a Zeiss wide-field fluorescent microscope equipped with AxioVision 4.8 software. Quantitative RT-PCR. RNA was reversed transcribed using Invitrogen Superscript II First Strand Synthesis kit. We used Taqman Gene Expression Assay mix containing unlabeled PCR primers and a FAM-labeled Taqman probe to detect expression of the following genes: Wnt1 transgene (Applied Biosystems Mm _g1), keratin 8 (K8) (Applied Biosystems Mm _m1), 146

167 Gata3 (Applied Biosystems Mm _m1), Muc1 (Applied Biosystems Mm _m1), keratin 5 (K5) (Applied Biosystems Mm _m1), keratin 14 (K14) (Applied Biosystems Mm _m1), P-cadherin (Cdh3) (Applied Biosystems Mm _m1) and vimentin (Vim) (Applied Biosystems Mm _m1). Relative quantification PCR (ΔΔCt method) was performed in triplicate using Agilent Technologies Stratagene Mx3005P detection system and analyzed using Stratagene MxPro software. Expression levels of Wnt1 in sorted cell populations were normalized to Gapdh transcript levels (Applied Biosystems E ) and compared to the unsorted sample (relative expression=1). Northern Hybridization. Total RNA was isolated from snap-frozen bulk tumor pieces by CsCl Density Gradient Centrifugation. Northern hybridization was performed as previously described 30 using cdna probes generated by RT-PCR (Table 4.3). After subcloning, the identity of each probe was confirmed by DNA sequence analysis. 4.4 Results. Evidence for distinct basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high subclones within some MMTV-Wnt1 tumors. Mammary cancers arising in the classic MMTV-Wnt1 transgenic mouse model 47 display tumor cell heterogeneity that is widely attributed to malignant transformation of a bipotent mammary progenitor cell Concordantly, MMTV-Wnt1 tumor cells partition into basal and luminal subsets reminiscent of 147

168 the corresponding basal and luminal lineages found in the normal mammary gland. Immunohistochemistry performed on Wnt tumor sections confirmed mosaiform heterogeneity with widespread intermingling of basal and luminal tumor cells (Figure 4.1a). Resolving tumor cells into basal and luminal subsets using an established flow cytometry strategy 129 showed that Wnt tumors reproducibly harbor both cell subtypes in abundance (Figure 4.1b). While studying cooperating oncogenic mutations in the MMTV-Wnt1 model, we found evidence suggesting some Wnt tumors harbored unexpected genetic heterogeneity. About half of all Wnt-initiated mammary tumors spontaneously acquire somatic HRas mutations that encode an activated oncoprotein 284,285. Since HRas mutations act dominantly, HRas mutant allele fractions (MAFs) of approximately 0.5 are expected, barring copy number changes at the HRas locus. Instead, when tumor-derived HRas alleles were amplified by PCR and subjected to DNA sequencing, chromatogram peak heights often indicated smaller HRas MAFs with fractions < 0.3 detected in 4 of 10 tumors. Notably, tumors maintained their small HRas MAFs as a stable property when explanted onto the flanks of syngeneic host mice (Figure 4.1c). This discrepancy could not be explained by contamination of samples with normal (non-tumor) cells since tumor cell content assessed by histopathology consistently exceeded 80%. Moreover, copy number variations leading to either HRas wt allele gain or HRas mut allele loss seemed unlikely driver events. Instead, we considered whether some Wnt tumors might harbor distinct HRas mut and HRas wt subclones, 148

169 noting that biclonal tumors would adopt a mixed-lineage phenotype provided each subclone were committed to a distinct lineage. To search for lineage-restricted HRas mut and HRas wt subclones, dissociated cells prepared from Hras mut Wnt tumors were sorted into basal and luminal subsets (Figure 4.2), then HRas MAFs were determined for each subset and for corresponding samples of unsorted cells. We set a threshold whereby detecting a 5-fold subset-specific enrichment in the HRas MAF would constitute preliminary evidence for the presence of a lineage-restricted subclone. By this measure, half of the HRas mut Wnt tumors analyzed (5 of 10) showed negligible subset-specific enrichment in HRas mut alleles, a pattern consistent with a hierarchical configuration (Figure 4.1d,e). In these cases, basal and luminal cells from the same tumor always harbored identical HRas mut alleles (Figure 4.1e), suggesting they descended from a common bipotent HRas mut progenitor. In contrast, for the remaining half of tumors analyzed, HRas mut alleles were highly enriched within the basal tumor cell subset, a pattern consistent with a biclonal configuration (Figure 4.1e). Basal HRas mut allele enrichment correlated with a lower overall HRas MAF (i.e., as measured in unsorted cells), further suggesting the presence of a private, subclone-restricted mutation. Regardless of whether the distribution of HRas mut alleles fit a hierarchical or biclonal pattern, tumors showed classic mixed-lineage histopathology (Figure 4.3), and luminal tumor cells were invariably the main source of Wnt1 expression as previously reported 129 (Figure 4.1f). Therefore, some Wnt tumors appeared to harbor distinct basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high subclones, implicating interclonal 149

170 cooperation in tumor maintenance. These findings recall early reports in which MMTV-associated mammary tumors initiated by activation of endogenous Wnt genes sometimes were noted to be oligoclonal 286,287. Basal HRas mut iwnt tumor cells rescued from Wnt withdrawal by incorporation of heterologous luminal cells. Seeking stringent proof that some Wnt tumors are biclonal and require interclonal cooperation for maintenance, we attempted to rescue growth of basal HRas mut /Wnt1 low subclones from Wnt1 deprivation by providing access to replacement Wnt1-producing cells. For these experiments, the original MMTV- Wnt1 model (hereafter cwnt denoting constitutive Wnt1 expression) was used in combination with a closely related model engineered for doxycycline (Dox)- dependent transgene expression (MMTV-rtTA/tetO-Wnt1; hereafter iwnt denoting inducible Wnt1 expression). During chronic Dox treatment, iwnt mice and mammary tumors phenocopy their cwnt counterparts, but iwnt tumors regress following Dox withdrawal due to abrogation of Wnt1 transgene expression 288. To enable tracking of cell lineages in tumor reconstitution experiments, iwnt mice were crossed with an mrfp reporter line, generating iwnt/mrfp + mice. As expected, a subset of Dox-dependent iwnt/mrfp + mammary tumors appeared biclonal, since they harbored a basally-restricted HRas mut subclone. After dissociating these tumors into cell suspensions, 10 5 unsorted cells were injected orthotopically into the mammary fat pads of two sets of Dox-treated, mrfp reporter-negative female host mice (Figure 4.4a). Control hosts lacked a transgene capable of rescuing tumor cells from Wnt withdrawal 150

171 (wt/mrfp - ), whereas rescue hosts expressed the constitutive Wnt1 transgene (cwnt/mrfp - ). During chronic Dox treatment, both control and rescue hosts developed mammary tumors in most glands injected with iwnt/mrfp + tumor cells (Figure 4.4b and Table 4.4). As expected, these reconstituted tumors usually regressed when iwnt transgene expression was switched off via Dox withdrawal. On control host mice, tumor regression always was complete and durable in that all sites of regressed tumor remained relapse-free during 6 weeks of monitoring. Interestingly, subclinical disease often persisted during these remissions, since most control hosts subsequently relapsed after Dox re-treatment (Figure 4.4b). By contrast, on cwnt rescue hosts, most reconstituted tumors only partially regressed, then relapsed spontaneously within several weeks of Dox withdrawal (Figures 4.4b and 4.5a). On control hosts, primary tumors were reconstituted almost exclusively from donor mrfp + cells, and relapses triggered by Dox re-treatment remained mrfp + as expected (Figure 4.4c). In contrast, on rescue hosts primary tumors showed varying degrees of chimerism due to incorporation of mrfp - (host-derived) luminal cells (Figure 4.5b), and relapses arising during Dox withdrawal always showed pronounced lineage-restricted chimerism, resulting in mrfp + /basal and mrfp - /luminal subpopulations (Figure 4.4c). To confirm that donor basal subclones recruit host luminal epithelium to serve as a replacement Wnt1 source at relapse, we turned to Northern hybridization analysis of tumor RNAs. Strikingly, tumors reconstituted on rescue hosts expressed the larger iwnt transgene prior to Dox withdrawal (with pertinent exceptions as discussed below), 151

172 then switched to expressing the smaller cwnt transgene at relapse, indicating heterologous rescue (Figure 4.4d). On rescue hosts, primary tumors that arose during Dox treatment incorporated a variable number of cwnt luminal cells, indicating that the crosstalk between heterologous cells required to seed relapse may occur early in tumor reconstitution. Interestingly, one of the three primary tumors analyzed appeared to contain a significant contribution from both donor-derived iwnt/mrfp + /luminal cells and host-derived cwnt/mrfp - /luminal cells. Northern blot analysis confirmed expression of both iwnt and cwnt transgenes (Figure 4.5b). Still, in a different primary tumor from this set, the conversion to lineage-restricted chimerism and cwnt transgene expression was essentially complete, meaning that cwnt-producing cells had replaced iwnt-producing cells despite ongoing Dox treatment (Figure 4.5b). Analysis of this tumor required necropsy of the host, precluding determination of its clinical response to Dox withdrawal, which we propose would have been negligible. Concordantly, in rare cases the growth of sibling primary tumors propagated on rescue hosts continued unimpeded by Dox withdrawal (Figure 4.5a), and these tumors always showed pronounced, lineagerestricted chimerism at necropsy. Elucidating mechanisms whereby host cwnt cells compete with luminal iwnt tumor cells to become the predominant Wntproducing subclone may offer new insights into evolutionary forces shaping tumor microenvironments. Lineage-restricted subclones recapitulate mosaiform heterogeneity in chimeric cwnt/iwnt tumors. 152

173 To determine whether reconstituted tumors recapitulate the biclonal configuration of parental tumors, primary and relapsed tumors from our rescue experiments were separated into basal and luminal tumor cell subsets and subjected to molecular analysis as before. In all cases, the biclonal configuration evident in parental tumors was maintained in that basal cells were HRas mut /Wnt1 low whereas luminal cells were HRas wt /Wnt1 high (Figure 4.6a, b). Expression levels of several epithelial lineage-specific genes were determined to verify the identity of the sorted cell populations and further validate the sorting strategy (Figure 4.c). We repeated these rescue experiments twice, beginning each time with an independent, iwnt/mrfp tumor harboring a distinct, basallyrestricted HRas mutation. In all cases, we observed rescue of donor-derived basal HRas mut /Wnt1 low tumor cells by cwnt host-derived luminal HRas wt /Wnt1 high cells (Figures 4.7 and 3.8a, b). Furthermore, the HRas mut allele detected in relapsed tumors was always identical to that detected in parental tumors, confirming that basal subclones found at relapse were descended from donor-derived tumor cells and were not novel clones. Examination of tumor sections by fluorescence microscopy revealed pervasive intermingling of mrfp + /basal and mrfp - /luminal tumor cells within chimeric relapses (Figure 4.8c). Notably, mrfp - /luminal cells were not clustered together as a clonal outgrowth or concentrated around the perimeter of the tumor. Instead, scattered mrfp - /luminal cells were incorporated throughout the body of the tumor, often immediately juxtaposed to their mrfp + / basal cell counterparts. This pattern is consistent with the prevailing notion that secreted Wnts provide a short-range signal to neighboring cells, and further 153

174 supports a cooperative interaction between the two subpopulations. Whereas socalled collision tumors are very rare, biclonal Wnt tumors apparently show mosaiform genetic heterogeneity, which recent studies suggest may be commonplace in human breast cancers 110,280. Both sorted-basal and sorted-luminal cell populations are required to reconstitute biclonal tumors. To test whether either subclone on its own could reconstitute tumors, biclonal iwnt/mrfp + tumor cells were sorted into basal and luminal subsets, then 10 5 cells of each subtype were injected separately into the mammary glands of Dox-treated wild-type hosts. Unlike unsorted (FACS-naïve) cells, sorted basal and luminal cells each reconstituted tumors inefficiently (Figure 4.9a), perhaps owing in part to loss of cell viability during FACS. When rare tumors arose from one sorted subtype, they always were mixed-lineage and biclonal, comprised of both basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high subclones (7 of 7 tumors analyzed, Figure 4.9b). Given the large number of injected cells and the imperfect separation achieved by FACS (95-98% purity), rare cells cross-contaminating each subset presumably sufficed to permit interclonal cooperation during tumor reconstitution. Consistent with this notion, the relative size of the basal and luminal cell populations within these tumors approximated that found in parental tumors and did not reflect the lineage enrichment achieved by sorting. We confirmed this inference in an experimental context where the putative cooperating subclones were differentially labeled by the mrfp transgene. Here, tumor cells derived from chimeric relapses generated in our rescue experiment 154

175 were studied prospectively. After injecting 10 5 cells as before, neither the basal (mrfp + /HRas mut /Wnt1 low ) nor the luminal (mrfp - /HRas wt /Wnt1 high ) subsets reconstituted tumors efficiently, whereas a 1:1 admixture of both sorted populations reliably reconstituted biclonal tumors (Figure 4.10). Notably, every tumor reconstituted in this experiment, including the lone outlier that arose after injecting sorted luminal cells, faithfully restored the subclonal composition of the source tumor, indicating strong selection favoring interclonal cooperation during tumor reconstitution. Biclonal tumors relapse through the evolution of either subclone. iwnt tumors that regress upon Dox withdrawal frequently relapse weeks later as Dox-independent tumors (DITs), mirroring the clinical scenario of acquired resistance to effective targeted therapy. Next, we asked which subclone(s) contribute when biclonal tumors beget relapse. We reasoned that HRas MAF measurements performed on bulk tumor ought to reflect the relative population sizes of HRas mut and HRas wt subclones. A putative biclonal iwnt tumor with an HRas MAF <0.3 was identified and propagated on host mice, generating a set of Dox-dependent tumor explants. Explants maintained the HRas MAF observed in the parental tumor, suggesting a stable biclonal configuration (Figure 4.11a,b). Next, host mice were subjected to Dox withdrawal and monitored until relapse, generating a set of 20 DITs. In accord with our previous work 289, 18 of the 20 relapses (90%) occurred through one of two mutually exclusive modes of Wnt pathway reactivation. Seven DITs (35%), re-expressed the Wnt1 transgene, and all 7 had acquired one of two rtta mutations (G138R or H100Y) previously 155

176 shown to rescue mammary tumors from oncogene withdrawal by enabling aberrant, Dox-independent expression of teto-controlled transgenes 283. All 7 of these tumors had an HRas MAF comparable to parental tumor (Figure 4.11b), strongly suggesting that rtta mutations originating within the HRas wt sublone restored both Wnt1 expression and cooperation with HRas mut cells, culminating in biclonal relapse. Eleven DITs (55%) instead rescued oncogenic signaling by acquiring an activating mutation in ß-catenin (Ctnnb1, hereafter ßcat), a key downstream Wnt effector (Figure 4.11a,b). Each of these tumors acquired the same G to C transversion mutation at the splice acceptor site between intron 2 and exon 3 of βcat, resulting in a splice variant lacking exon 3. Similar exon 3 deletions have been identified in human colorectal cancers 290. Compared with parental tumor, these relapses showed markedly increased HRas MAFs that were highly reproducible across the tumor set (Figure 4.11b). Therefore, ßcat mutations likely originated within HRas mut cells that later emerged as predominant relapse clones. By activating the Wnt pathway in a cell-autonomous manner, ßcat mutations presumably obviated the need to maintain cooperation with Wnt-producing HRas wt subclones. As such, ßcat mut relapse clones act like defectors in evolutionary game theory terms. HRas MAFs in ßcat mut relapses consistently exceeded 0.5, indicating that ßcat mut clones must harbor additional HRas locus aberrations. These relapses lacked gross changes in HRas gene copy number (Figure 4.12), implicating copy number neutral loss-of-heterozygosity events. Since, HRas MAFs reproducibly exceeded ßcat MAFs by approximately 2-fold 156

177 across the ßcat mut relapse set, elevated HRas MAFs may reflect duplication of the HRas mut allele (e.g., via a gene conversion event) sometime in the life history of ßcat mut subclones. Since restoring Wnt1 expression potentially reactivates Wnt signaling in a cell non-autonomous manner, we examined whether an rtta rescue mutation could be restricted to a relapse subclone. An iwnt/mrfp + tumor previously confirmed as biclonal in our rescue experiments (Figure 4.4) was propagated as above to derive DITs, then relapse-derived tumor cells were separated into basal and luminal subsets and analyzed. We identified one DIT that relapsed via Wnt1 transgene re-expression, and this tumor was biclonal with a luminally-restricted rtta mutation (Figure 4.11c). Trophic support from this luminal rtta mut subclone likely rescued growth of its basal rtta wt counterpart, perhaps obviating the need for the latter to acquire its own rescue mutation. Thus, subclone-restricted mutations appear capable of driving relapse by restoring interclonal cooperation, providing a plausible cellular mechanism that might explain how some types of rescue mutations are maintained at a low MAF. In contrast, our prior analysis indicated that ßcat rescue mutations originated within basal tumor cells, which then became the predominant relapse clones. Despite their basal origin, ßcat mut relapses nonetheless consistently harbored luminal tumor cells in abundance (Figure 4.11c and Figure 4.13) To account for this discrepancy, we hypothesized that acquired ßcat mutations endow basally-restricted subclones with novel bipotent differentiation potential. In this way, a unipotent subclone from a biclonal primary tumor might convert to a 157

178 bipotent, hierarchically-configured clone at relapse, and thereby help to preserve mixed-lineage character. To test this model, we analyzed sorted tumor cells from two ßcat mut DITs generated as above. Both tumors showed negligible subsetspecific enrichment in ßcat mut alleles (Figure 4.11c), consistent with a scenario in which ßcat mut relapse clones acquired bipotency. (An alternative scenario in which each subclone independently aquired matching ßcat mutations cannot be formally excluded, but seems less likely). In our prior experiments (Figures 4.4 and 4.6) and in the rtta mut relapse profiled above, this same subclone invariably behaved in a unipotent manner, remaining basally-restricted when partnered with Wnt1-expressing luminal subclones in the context of primary and relapsed tumors. When both rescue modes are considered together, relapse of iwnt tumors can be seen to stem from adaptations arising within either of the original parental subclones. 4.5 Discussion. Efforts to explain how some cancers stably maintain intratumoral lineage diversity typically invoke tumor cell hierarchies. Here, we show that cooperation between lineage-restricted subclones provides an alternative mechanism for maintaining tumor cell heterogeneity. In our Wnt models, we found evidence for both hierarchically and biclonally configured tumors, yet differently configured tumors were indistinguishable by histopathology, acquired equivalent cooperating HRas mut alleles (albeit with differences in tumor cell compartmentalization), and were comparably Wnt1-dependent. Thus, although distinct clonal configurations evolved, they converged toward analogous malignant phenotypes. These findings 158

179 highlight the difficulties associated with inferring the clonal architecture of cancers from histopathology, even in the simplified context of mouse models. Nonetheless, the Wnt models described here provide an experimentally tractable system for exploring whether and how a tumor s clonal configuration determines its clinical behavior and curability. While we have detected a functional cooperative interaction between tumor cell subclones in biclonal Wnt mammary tumors, a specific mechanism for this interaction remains unknown. Presumably secreted Wnt1 produced by the luminal cell subset acts in a paracrine manner to promote growth within the HRas mut basal cell subset, implying a cooperative interaction between the two oncogenic signals. Though, with both oncogenic signals acting on the basal cell subset, one might expect to see an overabundance of basal cells within biclonal Wnt tumors. Instead, the luminal cell subset consistently outnumbered the basal subset by approximately 2:1. This suggests an additional signal may exist which promotes expansion of the luminal tumor cell compartment. In theory, secreted Wnt1 may also act upon the luminal cells in an autocrine manner to maintain high luminal cell numbers in tumors. However, previous reports contend that luminal cells lack the LRP5/6 co-receptor necessary for Wnt signal transduction 272,291, making this possibility less likely. An alternative possibility includes basal cell- or stromal cell- production of a currently unidentified reciprocal signal which transmits back to the luminal cell subset, and thereby preserves the relatively stable proportions of the two tumor cell populations. Moving forward, understanding the precise nature of interclonal cooperation between tumor cell 159

180 subsets may become crucial for informing novel treatment strategies. Instances of interclonal cooperation may be particularly prevalent in tumors initiated by aberrant expression of secreted signaling molecules, such as Wnt1 and PDGF 292. As such, it will be important to determine how often cooperating subclones arise when initiating events act primarily in a cell-intrinsic manner. For biclonal Wnt tumors, our study does not define when in the course of tumor progression the composite subclones diverge. However, interestingly, in our heterologous rescue experiments, at least one of the primary tumors arising on Dox treated cwnt hosts exhibited additional heterogeneity within the luminal cell compartment, containing luminal cells from both iwnt and cwnt sources (Figure 4.5). This suggests that some Wnt tumors may have an even more complicated clonal architecture than originally thought. In particular, this heterogeneity within the luminal cell population raises the possibility that heterologous incorporation may be polyclonal (as opposed to a single recruited cwnt luminal cell repopulating the luminal tumor cell niche). Additional studies with lineage tracing models may shed light on the nature of heterologous cell recruitment in tumor growth. Furthermore, this same population of recruited luminal cells was incapable of independently initiating tumor growth (Figure 4.10), suggesting that incorporation of these cells into tumors was not sufficient for their malignant transformation. Indeed, both the luminal and basal tumor cell subsets consistently demonstrated co-dependence upon their reciprocal clone, indicating that malignant transformation of the pair might be contingent upon their cooperative interaction. 160

181 Finally, this work has important implications for the tumor heterogeneity research field. Currently, assays for identifying tumor initiating cells, or cancer stem cells, rely upon transplantation of limiting dilutions of an isolated population of tumor cells. However, the possibility for interclonal cooperation calls into question the strength of these assays. By design, such limiting dilution assays would fail to detect the tumorigenic potential of those clones whose growth depends upon cooperative interactions with other cells. Similarly, if interclonal cooperation emerges as a common mechanism for maintaining subclone diversity in malignancies, this scenario would counter a key assumption made when interpreting cancer genome sequences. Specifically, certain mutations detected at low allelic fractions and commonly assumed to be late events in tumor progression, instead may be early events that enable interclonal cooperation. 161

182 Figure 4.1: Evidence for distinct basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high subclones within some MMTV-Wnt1 tumors. a. Immunostaining for smooth muscle actin (SMA) and Keratin-8 performed on serial sections of a representative MMTV-Wnt1 mammary tumor. Scale bar, 50μm. b. Separation of MMTV-Wnt1 tumor cells into basal (CD49f high /EpCAM low ) and luminal (CD49f low /EpCAM high ) cell subpopulations by flow cytometry. Percentages depict mean +/- SEM. c. DNA sequencing chromatograms depicting an HRas CAA61CGA mutation appearing at a fixed MAF during serial propagation of an MMTV-Wnt1 tumor. d. Graphic depiction of the ratio of luminal to basal HRas MAF plotted against the MAF for unsorted cells. Dotted line depicts the threshold at which tumors show 5-fold HRas mut allele enrichment in basal versus luminal cells. Black circles denote tumor values with >5-fold basal enrichment. e. DNA sequencing chromatograms depicting an HRas CAA61CGA mutation (upper panels) and an HRas CAA61CTA mutation (lower panels) detected in representative Wnt tumors whose basal HRas mut allele enrichment fits a hierarchical pattern or biclonal pattern, respectively. f. Tumor cell populations analyzed by DNA sequencing and by qrt-pcr for Wnt1 expression relative to Gapdh. Histograms at left show HRas MAFs determined from chromatogram peak heights. Histograms at right show relative Wnt1 expression with values from unsorted tumor cells set at 1. Un, unsorted; B, basal; L, luminal. Data represent mean +/- SEM. 162

183 Figure 4.2: FACS gating strategy for resolving basal and luminal subsets from mammary tumors. Mammary tumors were mechanically and enzymatically dissociated into single cell suspensions. a. Negative selection against Lin + cells using Stem Cell Technologies EasySep Mouse Epithelial Cell Enrichment Kit. Resulting Lin - (CD45 - / CD31 - / TER119 - / BP-1 - ) cells were then immunostained with antibodies for CD49f (α6 integrin) and EpCAM and analyzed by FACS. b. Exclusion of cell debris and dead/ dying cells. Dead/dying cells collect as a band along the bottom of a FSC-A vs. SSC-A two-parameter plot, and these were gated out in P1. c. Cell doublets were discarded in P2. d. Basal and Luminal mammary epithelial cell populations were separated by immunophenotype. Basal epithelial cells are CD49f high / EpCAM low (P3) and luminal epithelial cells are CD49f Low / EpCAM high (P4). e. Gating tree showing gating strategy for FACS analysis as well as parent and total cell percentages within each of the gates for a representative MMTV-Wnt1 tumor 163

184 Figure 4.3: Hierarchical and Biclonal MMTV-Wnt1 tumors are histologically indistinguishable. a, H&E stained sections from a series of MMTV-Wnt1 mammary tumors whose HRas mut allele distribution pattern suggests hierarchical or biclonal configuration, as indicated. Scale bar, 50 μm. b, Both hierarchical and bi-clonal MMTV-Wnt1 tumors display mixed-lineage character. Serial sections from a hierarchical and bi-clonal MMTV-Wnt1 mammary tumors immunostained for Smooth Muscle Actin (SMA) or Keratin 8 (K8), which recognize basal and luminal epithelial cells respectively. For both, brown pigment is positive staining. Sections were counterstained with hematoxylin. Scale bar, 50 μm. 164

185 Figure 4.4: Rescue of basal HRas mut iwnt tumor cells from Wnt withdrawal by heterologous luminal cwnt cells. a. Schematic of experimental design. b. Growth curves of tumors reconstituted on wild-type or cwnt hosts following injection of iwnt/mrfp + tumor cells. c. Representative FACS plots showing contributions by donor-derived mrfp + cells and hostderived mrfp - cells to reconstituted tumors. Percentages depict mean +/- SEM. Colors indicate events within the basal (blue; CD49f High /EpCAM low ) and luminal (green; CD49f Low /EpCAM high ) gates. d. Northern hybridization analysis of tumor RNA with Wnt1 probe. The larger bicistronic iwnt transcript encodes both Wnt1 and firefly luciferase. 165

186 Figure 4.5: Tumor regression following Dox withdrawal. a. Tumors reconstituted on wildtype or cwnt hosts following injection of iwnt/mrfp + tumor cells were subjected to Dox withdrawal and monitored for regression. *Shown as number of tumor regressions per number of tumors subjected to Dox withdrawal. b. Northern hybridization analysis of tumor RNA with Wnt1 probe. Tumors were reconstituted on Dox-treated cwnt hosts following injection of iwnt/mrfp + tumor cells. Depicted below are the corresponding FACS plots showing the range of contributions by donor derived mrfp + and host-derived mrfp - cells to reconstituted tumors prior to Dox withdrawal. Colors indicate events within the basal (blue; CD49f high /EpCAM low ) and luminal (green; CD49f low /EpCAM high ) gates. 166

187 Figure 4.6: Biclonal configuration of reconstituted iwnt/mrfp + tumors. a. DNA sequencing chromatograms depicting a basally-enriched HRas GGA12GAA mutation detected in the parental tumor. b. Evidence for distinct basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high tumor subclones. Sorted tumor cell subsets were analyzed by DNA sequencing and by qrt-pcr for Wnt1 expression relative to Gapdh. Histograms at left show HRas MAFs determined from chromatogram peak heights. Histograms at right show relative Wnt1 expression with values from unsorted tumor cells set at 1. Un, unsorted; B, basal; L, luminal. Data represent mean +/- SEM. c. For each condition, sorted tumor cell subsets were analyzed by qrt-pcr for expression of several epithelial lineage-specific genes relative to Gapdh, with values for unsorted tumor cells set at 1. Gray bars, unsorted; Blue bars, basal; Green bars, luminal. Data represent mean +/- SEM. 167

188 Figure 4.7: Basal subclones from two additional iwnt/mrfp + tumors rescued from Dox withdrawal by heterologous cwnt host cells. a. Growth curves of tumor outgrowths derived from an iwnt/mrfp + tumor harboring a basally-restricted HRas GGA12AGA mutation. Curves depict regression and relapse of tumors reconstituted on cwnt rescue hosts following Dox withdrawal. b,c. Upper panels. Representative FACS plots showing contributions from donorderived mrfp + cells and host-derived mrfp- cells during tumor reconstitution. Colors indicate events within the basal (blue; CD49f High /EpCAM low ) and luminal (green; CD49f Low /EpCAM high ) gates. Lower panels. DNA sequencing chromatograms showing matching, basally-restricted HRas mutations present in both primary Dox-dependent tumors and chimeric Dox-independent relapses. d-f. Data panels presented as in a-c, showing similar results for an independent iwnt/mrfp + tumor harboring a distinct, basally- restricted HRas CAA61CGA mutation. For both tumors shown here, Northern hybridization analysis confirmed expression of donor-derived iwnt transgene prior to Dox withdrawal, followed by a switch to expression of host-derived cwnt transgene at relapse (data not shown). 168

189 Figure 4.8: Lineage-restricted subclones recapitulate mosaiform heterogeneity in chimeric cwnt/iwnt tumors. a. Schematic. b. DNA sequence chromatograms depicting matching HRas GGA12AGA mutations detected in unsorted and sorted populations from primary and relapsed tumors as indicated. c. Immunostaining of basal (SMA, top panels) and luminal (Keratin-8, lower panels) tumor cells within a Dox-independent relapse arising on a cwnt host. Red fluorescence marks donor-derived iwnt/mrfp + cells intermingled with mrfp - host-derived cells. Scale bar, 50μm. 169

190 Figure 4.9: Biclonal configuration of tumors reconstituted from sorted iwnt/mrfp + tumor cell subsets. a, Sorted tumor cell subsets inefficiently reconstitute tumors. Three independent iwnt/mrfp + bi-clonal tumors were resolved into component basal and luminal tumor cell subsets by FACS. Each tumor harbored a different basally-restricted HRas mutation, as indicated sorted tumor cells were injected orthotopically into intact, post-pubertal mammary glands of wild-type host mice maintained on chronic Dox treatment. *Shown as number of reconstituted tumor outgrowths per injected gland. b, Tumor cells from a parental iwnt/mrfp + tumor harboring a basally-restricted HRas GGA12GAA mutation were resolved into basal and luminal cell subsets by FACS. When these isolated tumor cell subsets were injected orthotopically into the mammary glands of Dox-treated wild-type hosts, few tumors were reconstituted. However, tumors that did arise always were comprised of basal HRas mut /Wnt1 low and luminal HRas wt /Wnt1 high subsets, implicating interclonal cooperation in tumor reconstitution. 170

191 Figure 4.10: Both sorted basal and sorted luminal cell populations are required to reconstitute biclonal tumors. Chimeric tumor relapses generated by injecting iwnt/mrfp + tumor cells onto cwnt rescue hosts were resolved into their component basal (mrfp + /HRas mut /Wnt1 low ) and luminal (mrfp - /HRas wt /Wnt1 high ) cell subsets by FACS. Each sorted population was injected separately (10 5 basal cells/injection or 10 5 luminal cells/injection) or as a 1:1 admixture (5x10 4 basal cells + 5x10 4 luminal cells/injection) onto wildtype, Doxnaïve hosts. All reconstituted tumors faithfully recapitulated the biclonal configuration of the source tumor. Dipicted are FACS plots from parental and reconstituted tumors showing both mrfp + and mrfp - subclonal populations. Colors indicate events within the basal (blue; CD49f high /EpCAM low ) and luminal (green; CD49f low /EpCAM high ) gates. 171

192 Figure 4.11: Relapse of biclonal tumors through the evolution of either subclone. a. DNA sequencing chromatograms depicting matching HRas CAA61CGA mutations detected in primary and relapsed tumors, with an increased MAF detected in the setting of ßcat mutations. b. Histogram depicting MAFs for a series of primary and relapsed tumors derived from a parental biclonal tumor. Upper panel depicts corresponding gene expression patterns for each tumor by Northern hybridization analysis. c. DNA sequencing chromatograms depicting matching HRas GGA12GAA mutations detected in primary and relapsed tumors, with an increased MAF detected in the setting of ßcat mutations. Panels at right depict analysis of unsorted and sorted cells at relapse showing biclonal and monoclonal relapses, depending on the mode of Wnt pathway reactivation. d. Histogram shows Wnt1 expression in unsorted and sorted tumor cells relative to Gapdh with the value measured in unsorted cells from tumor lacking a ßcat rescue mutation set at 1. Un, unsorted; B, basal; L, luminal. 172

193 Figure 4.12: Increased HRas MAFs in ßcat mut DITs is not due to gross copy changes at the HRas locus. Histogram depicts HRas allele copy number relative to ß-actin determined for a cohort of clonally-related Wnt tumor outgrowths. Independent relapse samples are presented in the same order depicted in Fig. 4b. Copy number values were obtained by performing qpcr on genomic DNA from tumor samples and from normal tail, with tail values set at 1. As a positive control, we included a p19 Arf -deficient Wnt tumor sample (~10x Amp) previously found to have approximately 10-fold HRas copy number gain as determined by Southern hybridization. 173

194 Figure 4.13: Mixed-lineage character of DITs. Serial sections of representative Wnt1 transgene re-expressing and β-cat mut relapsed tumors immunostained for Smooth Muscle Actin (SMA) or Keratin-8, which recognize basal and luminal epithelial cells respectively. For both, brown pigment indicates positive staining. Sections were counterstained with hematoxylin. Scale bar, 50 μm. 174

195 Genotyping Primers Gene Primer Sequence MMTV-rtTA F: 5 - TGCCGCCATTATTACGACAAGC- 3 R: 5 - ACCGTACTCGTCAATTCCAAGGG- 3 F: 5 - ATCCGCACCCTTGATGACTCCG- 3 R: 5 - GGCTATCAACCAACACACTGCCAC- 3 teto-wnt1 F: 5 - TGCGGTTCCTGATGTATTTTGC- 3 (iwnt) R: 5 - TGCATTCCTTTGGCGAGAGG- 3 F: 5 - CAGGAAATTGCTTCTGGTGGC- 3 MMTV-Wnt1 (cwnt) mrfp R: 5 - TCGAAGATGTTGGGGTGTTGG- 3 F: 5 - ATCCCACCCTTGATGACTCCG- 3 R: 5 - GGCTATCAACCAACACACTGCCAC- 3 F: 5 - ATCCGCACCCTTGATGACTCCG- 3 R: 5 - GGCTATCAACCAACACACTGCCAC- 3 Genotyped by assessing for red fluorescence in tail clips under fluorescent microscope. Table 4.1: Genotyping Primers. Genotyping of transgenic mouse lines was performed by PCR using genomic DNA isolated from tail clips and transgene specific primers. F indicates the forward primer. R indicates the reverse primer. 175

196 Sequencing Primers Gene Primer Sequence HRas F1: 5 - GGGTCAGGCATCTATTAGCCGTC- 3 R1: 5 - GCCGAGACTCAACAGTGCGAG- 3 βcat F1: 5 - GCGTGGACAATGGCTACTCAAG- 3 R1: 5 - GCGTCAAACTGCGTGGATGG- 3 R2: 5 - GCTACTTGCTCTTGCGTGAAGG- 3 rtta F1: 5 - TGCCGCCATTATTACGACAAGC- 3 R1: 5 - ACCGTACTCGTCAATTCCAAGGG- 3 Table 4.2: Sequencing Primers. Genes to be sequenced were amplified by PCR and then subjected to Sanger sequencing using gene specific primers. F indicates forward primer. R indicates reverse primer. 176

197 Primers for Generating Probes for Northern Hybridization Gene Wnt1 Primer Sequence F: 5ʹ- TGCGGTTCCTGTATTTTGC- 3ʹ R: 5ʹ- TGCATTCCTTTGGCGAGAGG- 3ʹ Axin2 F: 5 - CCGAGCTCATCTCCAGGC- 3 R: 5 - GGACAGAGGCAGCGGACTC- 3 β-actin F: 5 - TGAGACCTTCAACACCCCAG- 3 R: 5 - TGAGACCTTCAACACCCCAG- 3 Table 4.3: Primers for Generating Probes for Northern Hybridization. cdna probes for Northern hybridization were generated by RT-PCR using gene specific primer pairs. F indicates forward primer. R indicates reverse primer. 177

198 Table 4.4: Unsorted tumor cells efficiently reconstitute tumors. Unsorted (FACS naïve) tumor cells from 3 independent iwnt/mrfp + tumors were injected orthotopically into intact, post-pubertal mammary glands of wild-type control hosts or cwnt rescue hosts. Host mice were maintained on chronic Dox treatment. Each tumor harbored a different basally-restricted HRas mutation, as indicated tumor cells were injected into each gland. *Shown as number of reconstituted tumor outgrowths per injected gland. 178

199 Chapter 5 Overall Discussion 179

200 5.1 Introduction. The mammary gland is a heterogeneous organ, composed of numerous different cell types that interact in order to orchestrate the profound morphologic and functional changes that characterize normal mammary gland physiology. The marked heterogeneity within the normal mammary gland translates into breast cancers that also display a startling degree of heterogeneity. The data presented in this dissertation have attempted to elucidate how physiologic heterogeneity within the mammary gland can influence the manifestations and character of malignant breast disease. Most notably, the studies presented here are the first to demonstrate functional cooperation between subclonal tumor cell populations in a spontaneous mammalian tumor model. The bulk of these studies make use of a Wnt-driven mouse mammary tumor model. However, because mutations in Wnt/β-catenin pathway members are rare in human breast cancers, Wnt is not commonly thought to be a driver of human breast tumor growth. Thus, while provocative, our results are not directly translatable to a specific subset of human breast cancer. Still, these studies reveal principles of tumor biology that have broad implications for human disease. Using HR-positive breast tumors as an illustrative example, this chapter speculates as to how interclonal cooperation may manifest in human breast cancers, what the potential clinical implications may be for such an interaction, and highlights avenues for further research. 180

201 5.2 Implications for Human Breast Cancer. The data presented in Chapter 2 underscore the impact that the mammary epithelial cell of origin can have on tumor phenotype. MMTV-directed transformation of a specific population of HR-negative luminal epithelial cells produces, with very few exceptions, mammary tumors that are also HR-negative and hormone independent. Among those few exceptions include models that either produce secreted oncogenic factors or otherwise enhance signaling through a physiologic paracrine signaling axis (see Table 2.1). Interestingly, these HRpositive exceptions are also more likely to exhibit prominent cellular heterogeneity, as those tumors tend to display mixed lineage character. Could malignant exploitation of physiologic paracrine signaling axes contribute to some of the heterogeneity found within tumors? In particular, the MMTV-RANK mouse model generates progesterone dependent mammary tumors following an initial carcinogen treatment and stimulation with exogenous hormone 77. Tumors from this model contain heterogeneous HR expression, containing both HRpositive and HR-negative epithelial cells. Presuming other MMTV-driven tumor models adhere to the pattern of MMTV expression observed in the experiments from Chapter 2, we might assume that MMTV-directed RANK expression is limited to the HR-negative luminal epithelial subset. This, coincidentally, is the same epithelial subset that normally expresses RANK receptor and proliferates in response to PR mediated RANK-L. Thus, overexpression of RANK receptor in this cell population may exaggerate the physiologic RANK/RANK-L paracrine signaling axis and overstimulate RANK-L mediated cell divisions. Consequently, 181

202 one might expect growth of the resulting tumors to be dependent upon the paracrine RANK-L signal produced by PR-positive tumor cells. Indeed, pharmacologic treatment with a selective RANK-L inhibitor attenuated tumor growth 77, confirming that tumors from this model may be only indirectly hormone dependent. That is, instead of having a directly mitogenic effect on tumor cells, hormones may provide growth signals through a downstream paracrine interaction between phenotypically distinct mammary epithelial cells. Specifically, growth of these tumors may conceivably depend upon a cooperative paracrine interaction between a PR-positive subclone which produces RANK-L and a PR-negative subclone which transgenically overexpresses the RANK receptor, though this hypothesis has yet to be experimentally validated. In human HR-positive breast tumors, a shift from paracrine to cellautonomous HR signaling is thought to be responsible neoplastic transformation and tumor growth (Figure 5.1a). However, could it be possible for some human HR-positive breast tumors to depend upon a cooperative paracrine interaction between tumor cell subtypes? A breast tumor is considered positive for HR expression and is eligible for anti-hormonal therapy if greater than 1% of the tumor cells analyzed express HRs (ER and/or PR). Therefore, for some HRpositive tumors, the HR-expressing subpopulation comprises a significant minority of tumor cells. It seems plausible that, for this subset of HR-low or - moderate expressing tumors, ovarian hormones may not have a directly mitogenic role in tumor growth. Rather, hormones may instead stimulate growth through a downstream paracrine signaling mechanism that links hormone-sensing cells to a 182

203 proliferative cell compartment (Figure 5.1b). In one study of human ER-positive tumors 38, ER- expression was detectable in greater than 50% of proliferating tumor cells for just 7 of 19 tumors analyzed. Whereas, for 6 of the tumors, fewer than 10% of cells expressing proliferation markers also expressed ER 38. A second study 293 classified tumors into 3 groups based on the HR status of proliferating cells: those tumors with a positive correlation, those with no correlation, and those with a negative correlation between expression of HRs and cell proliferation markers. Accordingly, it appears that for a significant proportion of HR-positive tumors, the majority of proliferating tumor cells do not express HRs, supporting a potential indirect role for hormones in tumor growth through a downstream paracrine interaction. Recently, Cathrin Brisken s group submitted evidence that paracrine progesterone control mechanisms are conserved across species 11. They showed that RANK-L was necessary and sufficient for progesterone-induced cell division in cultured human breast microstructures. Further, expression levels of breast tissue RANK-L correlated with serum progesterone levels, and PR expression colocalized with RANK-L protein expression in a subset of normal human breast epithelial cells 11. However, whether this paracrine signaling axis is maintained in some human breast cancers is currently unknown. Could some HR-positive human breast cancers have a biclonal tumor organization? Several breast cancer cell lines as well as human primary breast cancers have been shown to express the RANK receptor 294. Additionally, a significant subset of human breast cancers express RANK-L 77, leaving open the possibility for paracrine-driven tumor growth. Such a tumor might contain co- 183

204 dependent subclonal tumor populations: one HR-positive subclone that produces RANK-L together with a second HR-negative subclone that proliferates in response to paracrine RANK/RANK-L signaling (Figure 5.1b). In this scenario, tumor growth would depend upon indirect hormone stimulation through this downstream paracrine interaction, which, consequently, would actively maintain subclonal tumor cell heterogeneity. Further, dependence of biclonal tumors on paracrine RANK-L signaling may implicate a role for anti-rank-l therapy in breast cancer treatment. Indeed, the anti-rank-l therapy, Denosumab, has been suggested as a potential adjuvant breast cancer therapy, in addition to its present role as a treatment for osteoporosis. Currently, an international phase II trial is being conducted to determine the effect of short term neoadjuvant Denosumab treatment on tumor cell proliferation in women with early stage breast cancer 295. Finally, a biclonal tumor organization would have implications for acquired resistance to anti-hormonal therapies and tumor relapse. In our relapse experiments with biclonal tumors (see Figure 4.11c), expression of Wnt1 was abrogated via Dox withdrawal, simulating treatment with targeted therapy. Tumor relapses, then, acquired genetic alterations permitting either Dox independent reexpression of Wnt1 (thereby re-establishing a cooperative paracrine interaction between subclones), or cell-autonomous Wnt/β-catenin signaling (thereby obviating the need for a paracrine interaction). In the case of a hypothetical biclonal human breast tumor dependent upon paracrine RANK-L signaling, antihormonal therapy would similarly interrupt tumor growth, albeit indirectly. It follows that drug resistant relapse may emerge by way of hormone independent 184

205 upregulation of secreted RANK-L. Alternatively, a mutation in a downstream component of RANK/RANK-L signaling might trigger constitutive pathway activations, thereby precluding the need for a cooperative paracrine interaction between RANK-L producing and RANK-L responding subclones. Downstream components of this signaling pathway include several proto-oncogenes, including PI3K, and MAP kinases JNK and ERK. Further, a switch to cell-autonomous signaling within a previously biclonal tumor may produce a monoclonal HRnegative relapse, thus providing a potential mechanism for discordant HR status in primary-relapse pairs. In one study, nearly one third of HR-positive primary tumors produced a HR-negative relapse 296. While loss of HR-expression in disease relapse is a poor prognostic indicator, the mechanisms underlying HR loss are currently unknown. Extrapolating from our mouse studies, we speculate that loss of HR at relapse might reflect conversion of a biclonal tumor to a monoclonal tumor organization for some cases. 5.3 Implications for Metastasis. The potential for co-dependent subclones within tumors also has implications for metastasis. Metastasis is the process whereby a tumor cell escapes from the primary tumor into the circulation and seeds tumor growth at a distant site. The majority of cancer deaths are not caused by primary tumors; rather, patients succumb to the metastatic burden of their disease. However, despite its bearing on cancer related mortality, much remains unknown about the process of metastasis in tumor progression. 185

206 Traditional metastasis models assume that a single cell is responsible for seeding metastatic outgrowths, following accumulation of the necessary mutations that permit cell migration and re-growth at a distant site. For a biclonal tumor with co-dependent subclones, metastasis by this model would require an additional mutation that obviates the need for a cooperative interaction, thus resulting in a self-sufficient subclone (Figure 5.2a). In our tumor relapse experiments (see Figure 4.11), we found evidence for such a mutation which allowed for cell autonomous tumor relapse. Perhaps a similar mutation would be necessary to permit metastasis of this tumor type. This information, in and of itself, would be helpful in guiding treatment decisions for metastatic disease. In such a situation, metastatic lesions composed of cell-autonomous clones would be fundamentally different from their biclonal primary tumor and, consequently, would be unlikely to respond to the same treatments. In support of this idea, matched pairs of primary breast cancers and brain metastases were found to be discordant for HR expression, with metastatic lesions showing loss of ER expression in 7 of 9 cases and loss of PR expression in 9 of 9 cases 297. Therefore, when treating metastatic disease, it may become prudent to consider the metastatic lesions separately from the primary tumor in some instances. Alternatively, could it be possible for tumors with a biclonal organization to produce biclonal metastases? This scenario would require dissemination of cells from both co-dependent subclones. Apart from circulation of a tumor embolus containing both clones, it seems unlikely that cells from two distinct clones would independently acquire metastatic capabilities simultaneously, such 186

207 that both escape from the primary tumor and circulate together. However, sequential achievement of metastatic capability by both clones does seem plausible. In 2008, Podsypanina and colleagues showed that when untransformed mammary epithelial cells were introduced into the circulation of a mouse, they could seed within the lungs and survive undetected for as long as 4 months 298. Later, when oncogenic signaling was transgenically induced in those cells, lung metastases grew. This study counters one of the main assumptions regarding the process of metastatic dissemination in tumor progression: namely, that metastasis is a late event in the progression of tumors. While this study does not address the process of intravasation in metastatic dissemination because mammary epithelial cells were introduced directly into the circulation, it raises the possibility that metastasis may be an early event in the progression of some tumors. Further it suggests that metastatic dissemination of cells does not depend upon prior malignant transformation, rather untransformed cells may disseminate and seed distant sites before they express oncogenic growth signals. Indeed, metastatic dissemination was shown to be an early event in some transgenic mouse models of breast cancer as well as in women with pre-invasive ductal carcinoma in situ (DCIS) lesions 299. Additional support for this idea was seen in a recent renal cancer genome sequencing study that compared genetic alterations in the primary tumor with the matched metastatic lesions from the same patient. This study showed a branched clonal phylogeny whereby the metastatic clone diverged from the main primary tumor clone early in tumor progression and evolved independently of the primary tumor 111. It has been proposed that early 187

208 dissemination of tumor cells may account for dormancy and late relapse of human breast cancers In the case of a biclonal primary tumor with co-dependent subclones, a conceivable scenario for metastasis could include metachronous dissemination of cells where one clone disseminates early, seeds a distant site, and, lacking its partner clone, lays dormant. Later dissemination of the second clone to that same site would then re-establish the cooperative paracrine interaction between subclones and promote growth at that site (Figure 5.2b). Metastasis by this mechanism would have important clinical implications, since such a multi-step process may proffer multiple opportunities for intervention along each stage. Finally, a third possibility is that a single cell from a biclonal tumor disseminates, but only takes up residence at those sites which produce the specific paracrine signals that satisfy the cooperative interaction of the original subclonal pair (Figure 5.2c). This hypothesis builds upon Paget s historic seed and soil theory of metastatic organotropism 115. While Paget surmised that a favorable environment was responsible for preferential metastasis to recurrent sites, he offered no explanation as to what made certain sites more favorable than others for seeding metastatic growth of a particular cancer. In the case of a biclonal tumor, preferred sites of metastasis may be those that express the paired signal from the paracrine circuit driving primary tumor growth. Our heterologous tumor rescue experiments (see Figures 4.4, 4.5, and 4.8) demonstrated that otherwise untransformed epithelial cells from the immediate environment can participate in tumor growth, provided they contribute an appropriate replacement growth signal 188

209 (Figure 5.3b). Cells at distant tissue sites that coincidentally express that same signal might similarly participate in growth of metastatic lesions upon encountering disseminated cells from the complementing subclone. Human breast cancers, especially HR-positive tumors, commonly metastasize to the bone, suggesting a favorable trophic relationship between breast cancer cells and the skeletal microenvironment 115. Indeed, as many as 85% of women dying of breast cancer display bony metastases 301,302, and bone is the most common site of first distant relapse 303. Several breast cancer cell lines as well as human primary breast cancers have been shown to express the RANK receptor 294, while many breast cancer metastases also express high levels of RANK-L 304, implicating a potential role for the RANK/RANK-L signaling axis in bony metastasis. RANK-L is an important osteolytic signal produced by osteoblasts, and has long been thought to participate in metastasis by resorbing bone, thereby creating a physical space for tumor growth 301,305. Indeed, breast tumor cells are thought to produce cytokines that upregulate RANK-L and promote bone resorption. TGF-β released from the resorbing bone further stimulates tumor cell production of osteolytic cytokines, and thus creates the vicious cycle of osteolytic bone metastasis However, given the mitogenic role of RANK/RANK-L signaling in normal mammary gland physiology, it is not unreasonable to suspect that, in addition to stimulating osteoclastic bone resorption, osteoblastic RANK-L may also act on tumor cells to promote growth. In this way, osteoblasts may provide the missing half of a paracrine growth circuit. Osteoblastic RANK-L has even been shown to act as a chemoattractant 189

210 for circulating breast cancer cells 309, perhaps lending support to the idea that circulating tumor cells may seek out those sites that provide complementing trophic support. Indeed, women with RANK-expressing primary tumors are significantly more likely to develop bone metastases than women whose tumors do not express RANK 310. While the majority of breast cancer bone metastases are osteolytic lesions, osteoblastic lesions as well as mixed osteolytic/osteoblastic lesions are also observed. In lesions with an osteoblastic component, osteoblasts (as well as the excess bone deposition) become incorporated into the tumor lesion, further suggesting a potential cooperative interaction between the RANK-L producing osteoblasts and the tumor cells. Finally, treatment with denosumab, a RANK-L inhibitor used to treat osteoporosis, has been shown to significantly delay the incidence of skeletal related events in breast cancer patients with bone metastases 311,312. Currently, a multicenter phase III trial is underway to determine whether denosumab treatment in women with early-stage breast cancer can prevent the formation of bone metastases Implications for Tumor Origins. Almost without exception, tumors are thought to originate from a single cell. By this model, different subclonal populations of tumor cells would derive from evolution of an initial clone (Figure 5.3a). Our studies in biclonal Wnt tumors (Chapter 4) do not define when in the course of tumor progression these composite subclones diverge. However, a biclonal tumor organization raises the 190

211 interesting possibility of a polyclonal tumor origin. While non-traditional, the idea for polyclonal tumors is not unprecedented. In 1996, examination of colonic adenomas from a rare familial adenomatous polyposis (FAP) patient with an XO/XY genotype revealed that majority were polyclonal in origin 314. Since then, the monoclonal origins of several cancers including prolactinomas 315, clear cell renal carcinomas 316, and several mouse tumors 317 have been questioned. In the case of a biclonal tumor with co-dependent subclones, the cooperative interaction between subclones may, in and of itself, enable transformation, setting up a requirement for two clones de novo (Figure 5.3c). While we cannot comment on the clonal origins of the spontaneous primary tumors from our study, in our tumor reconstitution experiments (see Figures 4.9, 4.10, and Table 4.1), neither subclone alone was capable of efficiently reconstituting tumors. Rather an admixture of both clones was required for tumor reconstitution, suggesting that cooperation between subclones was necessary for tumor growth. Moreover, reconstituted relapses on rescue hosts (see Figures 4.4, 4.5, 4.7, and 4.8) were absolutely composed of unrelated subclones, as each subclone was contributed by a different mouse. New lineage tracing models 318,319 may help determine the clonal origins of tumors that depend upon cooperating subclones. 5.5 Conclusion. It will be important to determine whether interclonal cooperation in tumors is limited to instances of aberrant secreted signaling factors. Certainly for the biclonal tumors discussed in Chapter 4, subclonal co-dependence hinged upon secreted oncogenic Wnt signaling. Similarly, the hypothetical breast tumors in 191

212 this discussion were dependent upon paracrine RANK-L signaling. Whether cell intrinsic oncogenes and tumor suppressors can facilitate cooperation between tumor cell subclones is unknown. Using the novel chimeric organoid culture technique described in Chapter 3, we have demonstrated that it is possible to track both the cell autonomous and non-autonomous effects of a given oncogene within the context of a polarized mammary epithelium. Moving forward, this technique may be helpful in dissecting the various interactions that occur between tumor cell subclones. 192

213 Figure 5.1: Models of HR-positive breast tumors. a. Schematic of tumor growth showing cellautonomous response to progesterone signaling. A progesterone signal is detected by the PRpositive cell (depicted with a blue nucleus), stimulating cell division within that cell (cellautonomous cell division depicted by the red arrow). The PR + cell acts as both the sensor cell and the responder cell. The resulting tumor consists of PR + cells that depend upon progesterone signaling for continued growth. b. Schematic of tumor growth showing a paracrine relationship between distinct PR + sensor cell and PR - responder cell populations. The PR + sensor cell detects a progesterone signal which stimulates production of secreted RANK-L. RANK-L then acts on neighboring PR - responder cells to induce cell division (paracrine mitogenic signal depicted by the green arrow). The resulting tumor consists of both PR + and PR - cells and depends upon secreted RANK-L signaling for growth. 193

214 Figure 5.2: Models of metastasis for biclonal tumors. a. Scenario 1: dissemination of a selfsufficient clone. One cell acquires an additional mutation (depicted by a yellow dot) obviating the need for a cooperative interaction between subclonal populations of a biclonally organized tumor (the two subclonal populations of tumor cells depicted in blue or green). This selfsufficient cell disseminates and takes up residence at a distant site. The resulting metastatic lesion has a monoclonal organization. b. Scenario 2: metachronous dissemination of cooperating clones. One clone (blue) from a biclonally organized tumor disseminates and seeds at a distant site, but remains dormant because it lacks the growth advantage afforded from the cooperative interaction between the two clones. At a later time, the second clone (green) disseminates to that same site. The cooperative interaction is restored permitting growth of the metastatic lesion. c. Scenario 3: cells at the distant site provide complementary trophic support. One clone (blue) disseminates, but only takes up residence at sites that produce the specific paracrine signals (depicted by green field) that satisfy the cooperative interaction of the original subclonal pair. 194

215 195

216 Figure 5.3: Models of biclonal tumor origins. a. A divergent clone arises through the clonal evolution process. A positive cooperative interactions between the two clones (blue and red) provides a competitive growth advantage resulting in a biclonal tumor. b. A cooperative interaction between a tumor cell clone and untransformed cells from the surrounding environment provides a competitive growth advantage. The resulting tumor consists of both transformed (blue) and normal cells (red). c. Two tumor cell clones arise and evolve independently of one another. A cooperative interaction between the two clones provides a competitive growth advantage. N indicates normal cell. Darkening shade of color represents evolution within a clone. 196