The identification and characterization of colon cancerinitiating

Transcription

1 The identification and characterization of colon cancerinitiating cells by Catherine Adell O Brien A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Institute of Medical Sciences University of Toronto Copyright by COB 2011

2 The identification and characterization of colon cancer-initiating cells Catherine A. O Brien Degree of Doctor of Philosophy Institute of Medical Sciences University of Toronto 2011 Abstract Colorectal cancer is the second leading cause of death from cancer (men and women combined) in the U.S. and Canada. The mainstay of treatment remains surgical resection and although new agents are constantly emerging to treat colorectal cancer, to date none of the agents have been successful at curing patients with advanced disease. In recent years there has been an increasing interest in the notion that cancers are organized as a hierarchy with the cancer-initiating cell (C- IC or cancer stem cell) existing at the apex. The C-ICs only represent a subset of the total tumour cells; however, research indicates that they are responsible for both the initiation and maintenance of tumour growth. In the studies presented here, we determined that human colon cancers are organized in a hierarchical manner. Furthermore, we prospectively isolated a subset of colon cancer-initiating cells (CC-ICs) based on the expression of the cell surface marker, CD133. The identification of CC-ICs has led to a number of questions concerning the molecular mechanisms driving these cells. Functionally all C-ICs are characterized by their ability to: i) generate a xenograft that histologically resembles the parent tumour from which it was derived, (ii) be serially transplanted in a xenograft assay thereby demonstrating the ability to self-renew ii

3 and, (iii) generate daughter cells that possess some proliferative capacity but are unable to maintain the cancer because they lack intrinsic regenerative potential. It is becoming evident that cancer cells evolve as a result of their ability to hijack normal self-renewal pathways, a process that can drive malignant transformation. Studying self-renewal in the context of cancer and C-IC maintenance will lead to a better understanding of the mechanisms driving tumour growth. In this work we demonstrate that the inhibitors of differentiation genes (Id1 and Id3) play a central role in driving self-renewal in the CC-IC subset. Furthermore, we demonstrate that this effect is partially mediated through the cdk-inhibitor, p21cip1/waf1. iii

4 Acknowledgments I gratefully acknowledge the support of my family and friends. I especially appreciated the support of Rita O Brien and Gail Manning. I would also like to thank Antonija Kreso for her invaluable contribution to this work. Finally I gratefully acknowledge the guidance I received from John Dick and Steven Gallinger. iv

5 Table of Contents Table of Contents Acknowledgments... iv Table of Contents... v List of Figures... ix List of Tables... xi List of Abbreviations... xii Chapter Colon cancer-initiating cells Cancer-initiating cells Overview Historical Perspective The C-IC model and solid tumours C-IC phenotypic marker: CD C-IC phenotypic marker: CD Side population assay Colon cancer-initiating cells CD133 and CC-ICs CD44 and CC-ICs Aldehyde dehydrogenase-1 and CC-ICs C-ICs and self-renewal C-ICs and the Wnt, Pten, and Akt pathways C-ICs and the hedgehog pathway C-ICs and the Notch pathway C-ICs and the inhibitors of differentiation Therapeutically targeting self-renewal C-ICs and the microenvironment C-ICs and metastasis C-ICs and therapeutic resistance Chemoresistance and C-ICs v

6 1.7.2 Radioresistance and C-ICs Resistance to differentiation and C-ICs Clinical relevance of C-ICs C-ICs as biomarkers Integrating C-ICs into clinical trials Conclusions Figures References Chapter Identification of a human colon cancer cell capacble of initiating tumour growth in immunodeficient mice Abstract Introduction, Results, and Discussion Figures Materials and Methods Tumour cell preparation Magnetic cell sorting and flow cytometry Transplantation of human colon cancer cells into NOD/SCID mice Limiting dilution analysis and serial tranplantations Purity following magnetic bead separation Xenograft histopathology Immunohistochemistry Statistical analysis References Chapter Id1 and Id3 regulate the self-renewal capacity of human colon cancer-initiating cells through p Addendum Abstract Introduction Results Enrichment of CC-IC following culture of primary human colon cancer tissue vi

7 3.4.2 Comparison of the frequency of cells capable of in vitro sphere and colony formation with CC-IC activity Knock down of Id1/3 expression reduces tumour growth in vivo Knock down of Id1/3 effects proliferation and xenograft microvessel formation Knock down of Id1/3 impairs the ability of CC-ICs to self-renew Knock down of Id1/3 increases sensitivity to oxaliplatin Knock down Id1/3 decreased self-renewal capacity through down-regulation of p Cellular context impacts the Id1/3-p21 regulatory pathway Discussion Figures Materials and Methods Tissue collection, cell isolation, and culture of colon cancer cells Gene knock down Fluorescence activated cell sorting and flow cytometry In vivo tumour formation assays and xenograft injections In vivo limiting dilution assays (LDAs) In vitro limiting dilution assays (LDAs) Cell Proliferation assays Protein isolation and western blotting Histological analysis Real-time PCR Cell cycle Analysis Statistical analysis References Chapter Summary, conclusions, and future directions Summary Conclusions Future Directions C-IC and cell of origin: C-IC experimental design limitations Immunodeficient mouse models and C-IC research Further purification of CC-ICs vii

8 4.3.5 C-ICs and functional genomics C-ICs and plasticity Concluding Remarks References References or Bibliography... Error! Bookmark not defined. viii

9 List of Figures Figure 1.1: Models to explain tumour heterogeneity Figure 1.2: Multiple facets to C-IC self-renewal Figure 1.3: Hypothetical models to account for the role of CC-ICs in tumour relapse and metstasis Figure 2.1 Xenografts generated from both bulk and CD133+ colon cancer cells resemble the original patient tumour Figure 2.2 Expression of CD133 in tumour and normal colonic tissue Figure Supplementary 2.1 Unfractionated and CD133+ colon cancer cells initiate tumours when transplanted under the renal capsule of NOD/SCID mice Figure Supplementary 2.2: Flow cytometric analysis of CD133 and ESA expression Figure Supplementary 2.3: Histological examination following the injection of CD133- colon cancer cells Figure Supplementary 2.4: Analysis of purity following magnetic bead separation Figure 3.1 Characterization of the CC-IC cultures Figure 3.2 Id1/3KD reduces tumour growth Figure 3.3 Effect of Id1/3KD on xenograft histology Figure 3.4 Id1/3KD impairs CC-IC self-renewal Figure 3.5 Id1/3 KD increases the sensitivity of CC-ICs to oxaliplatin Figure 3.6: Id1/3 KD decreased the self-renewal capacity of CC-IC through p21 down regulation ix

10 Figure 3.7: Comparison of p21 levels in cells cultured with and without serum Figure Supplementary x

11 List of Tables Table 1.1: Phenotypic identification of C-ICs in solid tumours Table 1.2: Prospective isolation of CC-ICs Table 1.3: Self-renewal pathways involved in C-IC maintenance Table 2.1 Patient and tumour characteristics Table 2.2 Limiting dilution analysis of the human colon cancer-initiating cell Table 3.1 Comparison of in vivo xenograft formation with in vitro sphere and colony formation Table Supplementary 3.1: In vivo limiting dilution assays of CC-IC samples Table Supplementary 3.2: In vivo LDAs in NOD/SCID versus NSG mice Table Supplementary 3.3: CD44 + colon tumour cells are enriched in CC-ICs as compared to CD44 - tumour cells Table Supplementary 3.4: CC-ICs are enriched in cultures grown in serum-free versus serumsupplemented medium xi

12 List of Abbreviations ABC ALDH APC APC AML BMI-1 BMP4 BMPR1B CEA C-IC CC-IC CFU ATP binding cassette aldehyde dehydrogenase allophycocyanin adenomatous polyposis coli acute myeloid leukemia polycomb group member insertion B lymphoma M0-MLV insertion region-1 bone morphongenic protein-4 bone morphogenic protein receptor 1B carcinoembryonic antigen cancer-initiating cell colon cancer-initiating cell colony forming unit Chk1/2 checkpoint kinases 1 and 2 CK-7 cytokeratin 7 CK-20 cytokeratin 20 CML CPA chronic myelogenous leukemia cyclophosphamide CXCR4 CXC chemokine ligand receptor 4 CXCR12 CXC chemokine ligand receptor 12 CSC DLL4 DMSO EpCAM cancer stem cell delta like 4-ligand dimethyl sulfoxide epithelial cell adhesion molecule xii

13 ESA FACS FAP FCS FITC 5-FU GSH Hh HIF2α HLH HSC IHC Id-1 Id-3 epithelial specific antigen flowcytometry assisted cell sorting familial adenomatous polyposis fetal calf serum fluorescein isothiocyanate 5-fluorouracil glutathione hedgehog hypoxia inducible factor 2α helix-loop-helix hematopoietic stem cell immunohistochemistry inhibitor of differentiation-1 inhibitor of differentiation-3 IL-4 interleukin 4 LDA LSC MACS MFSE limiting dilution assay leukemic stem cell magnetic bead assisted cell sorting mammosphere formation efficiency Muc2 mucin 2 NK natural killer NOD/SCID nonobese diabetic/severe combined immunodeficiency NSG PE NOD/SCID/IL2 receptor gamma chain null phycoerythrin PTCH1 Patched 1 xiii

14 PTEN ROS SCID SDF-1 SFU shrna SRF SP SMO TGF-beta TIC VEGF protein phosphatase and tensin homolog reactive oxygen species severe combined immunodeficiency stromal derived growth factor-1 sphere forming unit short hairpin RNA sphere replating frequency side population smoothened transforming growth factor -beta tumour initiating cell vascular endothelial growth factor xiv

15 1 CHAPTER 1 INTRODUCTION Colon cancer-initiating cells This chapter is based on review articles and a text book chapter that appeared in: Current protocols in stem cell biology 2008 Nov; Chapter 3:Unit 3.1. Authors: A.Kreso and C.A. O Brien Copyright 2008, John Wiley & Sons. Reprinted with permission from the publisher Seminars in radiation oncology 2009 Apr; 19(2):71-7. Authors: C.A. O Brien, A. Kreso, J.Dick Copyright 2009, Elsevier Inc. Reprinted with permission from the publisher Clinical Cancer Research 2010 Jun15;16(12): Authors: C.A. O Brien*, A. Kreso*, and C.J. Jamieson (* these authors contributed equally to this work) Copyright 2010, AACR publications. Reprinted with permission from the published Chapter in Metastasis of Colorectal Cancer 2010 ed. N.Beauchemin and J.Huot. vol.14, pp Authors: A. Kreso, L.Gibson, C.A.O Brien Copyright 2010, Spinger. Reprinted with permission from the publisher

16 2 Chapter 1 Introduction 1 Colon cancer-initiating cells 1.1 Cancer-initiating cells Overview Cancer is characterized by the excessive and uncontrolled growth of abnormal cells, which can invade and destroy other tissues(hanahan and Weinberg, 2000). From this description one may infer that all cancer cells within a tumour are equally abnormal and therefore possess equal capacity to initiate and sustain the tumour. However, it has long been recognized that malignant cells within the same tumour display heterogeneity(wang and Dick, 2005; Shipitsin and Polyak, 2008). In 1965, prior to the introduction of research ethics, Brunschwig et al. published a set of experiments where they autologously injected tumour cells into patients thighs; in order to observe tumour growth, they required a minimum injection of 10 6 cells(brunschwig et al., 1965). Interestingly, the number of tumour cells required to obtain growth was inversely correlated to the stage and grade of the original neoplasm, with more advanced and poorly differentiated tumours requiring fewer cells to seed a tumour in the thigh. These experiments clearly demonstrated that tumour cells are not only morphologically different, but that they also exhibit functional heterogeneity. Two models were proposed to explain tumour heterogeneity: the models differed with respect to their biological explanation of tumour heterogeneity but both models predicted that only a small number of cells within a tumour have the capacity to initiate and sustain tumour growth, typically measured on the basis of initiating a new tumour upon transplantation in either

17 3 syngeneic or xenogeneic recipients. The stochastic model proposed that all cells within a tumour are biologically homogenous and therefore have equal capacity to regenerate the tumour; a characteristic governed by a combination of intrinsic and extrinsic stochastic events such as the expression of critical transcription factors, a supportive microenvironment, and the ability to evade the immune system. According to this model it would be impossible to prospectively identify and isolate the tumour-initiating fraction. In contrast, the hierarchical model (also referred to as the cancer stem cell model) suggested that only a subset of tumour cells possess the capacity to regenerate the tumour(jordan and Lemischka, 1990; Reya et al., 2001; Dick, 2003). According to the hierarchical model it should be possible to prospectively isolate tumour cells into fractions that are tumour-initiating and non-tumour initiating, where only the former are capable of sustaining tumour growth. The tumour-initiating cells, also referred to as cancerinitiating cells (C-IC), are defined by: capacity for self-renewal, potential to develop into any cells in the tumour, and proliferative capacity to drive continued expansion of the tumour population(jordan et al., 2006) (Figure 1.1). It is important to understand that the notion of clonal evolution can be applied to either of these models and refers to the ability of either stochastically or hierarchically driven clones to evolve over the lifetime of a tumour Historical Perspective In recent years the concept of C-ICs or tumour-initiating cells has ignited a great deal of interest in large part because of the potential clinical implications associated with these cells. Simply stated the C-IC hypothesis suggests that the route to fully eradicating a tumour will require the use of agents that target the root of the cancer or C-ICs. The research on C-ICs in solid tumours remains at a very early stage and the question of clinical relevance remains unanswered at this time(o'brien et al., 2009). Preliminary work has fueled much excitement in the field, however, a great deal of work remains to be done both at the level of understanding the

18 4 basic biology of C-IC and their clinical relevance. To best understand the concept of C-ICs and their role in driving tumour growth it is essential to understand the field from a historical perspective. Although the concepts surrounding tumour heterogeneity, cancer stem cells, and the merits of the stochastic versus hierarchical models have a long history extending for many decades, conclusive proof required two important technological developments: cell sorting and reliable and quantitative in vivo xenograft assays. Cell sorting became widely available in the 1980's and was applied to the purification of normal hematopoetic stem cells (HSC). Reliable human normal and leukemia stem cell xenotransplantation assays were first developed in the late 1980's. Initial studies on acute myelogenous leukemia (AML), carried out by John Dick and colleagues in the 1990 s, demonstrated that primary human AML cells could be fractionated based on the CD34 + CD38 - cell surface phenotype and although these cells represented a small percentage of the total leukemic cells, these were the only cells that could give rise to leukemic growth in serial transplantation studies using the SCID (severe combined immunodeficiency) mouse model and could be serially transplanted(lapidot et al., 1994; Bonnet and Dick, 1997). Moreover, the leukemia that arose was an exact phenocopy of the human leukemia from which it was derived thereby demonstrating that the CD34 + CD38 - fraction was able to both self-renew, as well as, give rise to more differentiated progeny(lapidot et al., 1994). This work provided the first conclusive evidence that a rare LSC existed in human AML (based on a repopulation assay), and that this cell could reestablish the entire hierarchy; thereby ruling out the stochastic model in favor of the hierarchical or C-IC model for AML. The development of quantitative assays to assess the frequency of leukemic stem cells (LSC) represented an invaluable tool for carrying out this work; however, the ultimate test is to

19 5 be able to study the biological activity of individual LSCs. The development of in vivo clonal tracking assays provided a means to accomplish this goal. Hope et al used lentivirus vectormediated clonal tracking and the NOD/SCID xenotransplantation system to assay LSCs in AML at a clonal level(hope et al., 2004). Their study was the first to provide direct evidence for the self-renewal of individual LSC in human AML. Furthermore by applying a clonal assay to their system they were able to demonstrate that there is heterogeneity in the self-renewal properties within the LSC pool, thus functionally there is more than one type of LSC in AML and these can be categorized by differing capacities for self-renewal and clonal longevity(hope et al., 2004). Clonal tracking remains the gold standard assay in C-IC biology; however, AML remains the only human malignancy in which a thorough study of clonal tracking has been carried out. 1.2 The C-IC model and solid tumours The principles of the C-IC model were not tested in a solid tumour until a decade after the initial work in leukemia was published(pardal et al., 2003). This was due in large part to the paucity of solid tumour xenograft models generated from single cell suspensions and the inability to sort the cells based on cell surface markers. The technical hurdles were surmounted in 2003 when Michael Clarke s group tested the C-IC model in the context of breast cancer(al Hajj et al., 2003). They were able to prospectively isolate a subset of breast cancer cells (CD44 + CD24 - ) that were solely responsible for sustaining the disease in an immunecompromised mouse model. The C-IC subset could be serially passaged and the xenografts generated were histologically heterogeneous, resembling the parent tumour from which they were derived. These results demonstrated that the same C-IC principles that had previously been shown to apply in an AML model could also be translated to a solid tumour. Since the initial publication in breast cancer, a plethora of papers have been published identifying C-ICs in

20 6 numerous cancers including: brain(singh et al., 2004), colon(o'brien et al., 2007; Ricci-Vitiani et al., 2007), head and neck(prince et al., 2007), pancreas(hermann et al., 2007; Li et al., 2007), lung(eramo et al., 2008), prostate ({Collins, /id}{patrawala, /id}), and sarcoma (Wu et al., 2007) (Table 1.1) C-IC phenotypic marker: CD133 Singh et al published the initial identification of C-IC in human brain tumours based on the identification of the cell surface antigen CD133 (human prominin-1), a five transmembrane domain glycoprotein(singh et al., 2004). CD133 was originally identified as a cell surface antigen present on CD34 + hematopoietic(bidlingmaier et al., 2008) and neural stem cells(coskun et al., 2008; Corti et al., 2007). Much remains to be learned about the biological role of CD133; however, it has been postulated to play a role in cell polarity. In a NOD/SCID xenograft model, only injection of CD133 + brain tumour cells resulted in xenograft formation, with as few as 100 CD133 + cells giving rise to a xenograft that phenotypically and histologically recapitulated the patient tumour from which it was derived. In contrast, injection of up to 10 5 CD133 - cells did not result in xenograft formation, although viable cells could be detected long after transplantation arguing that the microenvironment was supportive of these cells, but they did not possess the capacity to regenerate the tumour. The xenografts that arose from the CD133 + cells could be serially transplanted thereby demonstrating the ability to self-renew, a central tenet of stem cell biology(singh et al., 2004). Since the initial publication on brain cancer, CD133 has been studied as a marker of the C-IC population in a variety of other cancers including: colon(o'brien et al., 2007; Ricci-Vitiani et al., 2007), pancreas(hermann et al., 2007), and lung(eramo et al., 2008). Utilizing model systems very similar to that employed in the brain cancer study, a number of groups were able to

21 7 demonstrate that the tumour-initiating capacity was enriched in the CD133 + cancer cells, as compared to their CD133 - counterparts. More recent studies have yielded other markers that can be used in combination with CD133 to further purify the C-IC fraction; one such example being in pancreatic cancer where researchers employed an orthotopic metastatic NOD/SCID xenograft model to investigate C-IC phenotypes. Utilizing this model they were able to distinguish C-ICs that could initiate metastatic disease based on the cell surface expression of CD133 and CXCR4. Both the CD133 + CXCR4 - and the CD133 + CXCR4 + fractions could initiate disease at the primary site, however, only the CD133 + CXCR4 + C-ICs were capable of initiating metastatic disease(hermann et al., 2007).These findings illustrate that heterogeneity exists within the pancreatic C-IC subset and further work is required to determine if the same organizational structure exists within other tumours, where different subtypes of C-IC give rise to metastatic versus primary lesions. Furthermore, this may also provide insight into why particular cancers preferentially spread to specific organs(fidler, 2003) C-IC phenotypic marker: CD44 Since the initial publication demonstrating that the CD44 + CD24 - expression profile identified breast tumour-initiating cells a number of studies have been published using CD44, a widely expressed adhesion molecule(al Hajj et al., 2003; Naor et al., 2002; Naor et al., 1997; Naor et al., 2008), either alone or in combination with other markers to identify the C-IC subset in a variety of tumours including: colon (ESA + CD44 + )(Dalerba et al., 2007; Dylla et al., 2008), pancreas (ESA + CD44 + CD24 + )(Li et al., 2007), and head and neck (CD44 + )(Prince et al., 2007). Interestingly, both the pancreatic and head and neck studies looked at expression profiles in their C-IC (CD44 + ) versus non-c-ic (CD44 - ) fractions. Prince et al found a preferential expression of BMI-1(bone marrow insertion-1) in the CD44 + subset as compared to the non-tumourigenic fraction(prince et al., 2007). BMI-1 is a member of the polycomb family and has a well

22 8 established role in self-renewal(iwama et al., 2004; Raaphorst, 2003), thereby providing biological evidence to support the functional difference between the CD44 + and CD44 - cells. Similar results were published by Li et al. in the pancreatic C-IC model where they identified a differential expression of SHH (sonic hedgehog) in the ESA + CD44 + CD24 + C-ICs, as compared to the non-tumourigenic fraction(li et al., 2007). Hedgehog pathway activation occurs in a significant number of primary human pancreatic carcinomas and is believed to be an early mediator of pancreatic cancer tumourigenesis(pasca et al., 2006; Thayer et al., 2003). The finding that SHH was preferentially upregulated in the C-IC fraction lends further support to the idea that the C-IC fraction is fundamentally different from their non-c-ic progeny. These studies were among the first to look at expression differences in C-IC and non-c-ic fractions of solid tumours. In the future this will represent an important area of investigation if we are to develop a deeper understanding of the biology and as a result target the pathways that are crucial for C-IC survival Side population assay Another approach being utilized to identify tumourigenic fractions is the use of the side population (SP) assay(wu et al., 2007). This assay exploits the high levels of expression of cell membrane ATP-binding cassette (ABC) transporter proteins which provides a cell with the ability to efflux chemotherapeutic drugs and certain dyes, such as Hoechst 33342(Wu and Alman, 2008). Wu et al. studied the SP + and SP - cells in 23 primary musculoskeletal tumours ranging from benign to high-grade sarcomas and demonstrated that tumour-initiating capacity existed exclusively within the SP + population. Tumours arising from SP + cells recapitulated the heterogeneity of the parent tumour from which they were derived and were able to be serially passaged demonstrating self-renewal capacity. The authors were also able to correlate the

23 9 percentage of SP cells with the aggressiveness of the musculoskeletal tumour, with more aggressive tumours possessing a higher percentage of SP + cells(wu et al., 2007). Further support for the role of ABC proteins in C-IC function was published by Schatton et al in human melanoma(schatton et al., 2008). Instead of studying the SP + cells they looked at the expression of one of the chemoresistance mediating ABC proteins (ABCB5) on the cell surface of melanoma cells. The ABCB5 + melanoma cells were enriched for tumour-initiating cells, as compared to their ABCB5 - counterparts, and were capable of being serially transplanted demonstrating their capacity for self-renewal(schatton et al., 2008). The authors took this work a step further by targeting the ABCB5 + cells with systemic administration of a monoclonal antibody capable of inducing antibody-dependent cell-mediated cytotoxicity. In vivo administration of the antibody commenced 14 days following tumour injection into NOD/SCID mice and treatment was continued for 21 days. At the end of the treatment period the mean tumour volume in the mice that received the ABCB5 antibody was 32.7±9.4mm 3, as compared to the control antibody and untreated groups which had mean tumour volumes of 226.6±53.8mm 3 and 165.4±36.9mm 3, respectively(schatton et al., 2008). Therefore not only does expression of the chemoresistance mediator ABCB5 enrich for melanoma C-ICs, but antibodies targeted against surface markers on these cells can result in decreased tumour growth in a xenograft model. 1.3 Colon cancer-initiating cells One of the most active areas in C-IC research has involved colon cancer (CC-IC) and this section will focus on the C-IC markers that have been published in colorectal cancer including:

24 10 CD133(O'Brien et al., 2007; Ricci-Vitiani et al., 2007), CD44(Dalerba et al., 2007), CD166(Dalerba et al., 2007), and ALDH (Dylla et al., 2008; Huang et al., 2009)(Table 1.2) CD133 and CC-ICs CD133 was the first cell surface marker utilized to isolate the tumour-initiating cells in colon cancer(o'brien et al., 2007; Ricci-Vitiani et al., 2007). Two groups utilizing two different murine xenograft models initially published the identification of CD133 as a marker of CC-ICs. Our group employed a NOD/SCID renal subcapsular xenograft model to demonstrate that CD133 + human colon cancer cells were enriched in tumour-initiating capacity(o'brien et al., 2007). In contrast, the CD133 - colon cancer cells did not give rise to xenografts with one exception at the highest cell concentration (2x10 5 cells). Similar results were published by Ricci-Vitiani et al. using CD133 in a SCID mouse subcutaneous xenograft model. They demonstrated that 3,000 CD133 + human colon cancer cells could initiate tumours, whereas 10 6 unfractionated cells did not initiate tumours(ricci-vitiani et al., 2007). Interestingly, the two groups demonstrated significant variation in the percentage of CD133 + cells in the colon cancers studied: the range being 1.8% to 24.5% in one study (O'Brien et al., 2007) and 0.7% to 6.1% in the other(ricci-vitiani et al., 2007). Each of these studies represented a small sample size and therefore future work will be required to determine the range of CD133 expression in a larger sample size of human colon cancers. It is important to recognize that the CD133 + population of colon cancer cells is heterogeneous and not every CD133 + cell represents a CC-IC; rather CD133 enriches for tumour-initiating capacity. This is best illustrated by carrying out in vivo limiting dilution assays (LDAs), which enables one to calculate the CC-IC frequency in the unfractionated colon cancer cells, as well as in the CD133 + subset. Utilizing LDA experiments, O Brien et al. determined

25 11 that the pooled CC-IC frequency for their cohort of 17 patients was approximately 1 in 56,000 unfractionated colon cancer cells(o'brien et al., 2007). The CC-IC activity was significantly enriched in the CD133 + subset, where one in 262 CD133 + colon cancer cells was calculated by LDA to be a CC-IC. This represented a greater than 200 fold enrichment in CC-IC activity in the CD133 + cells, as compared to the bulk colon cancer cells. It should be noted that this CC-IC frequency was derived from pooled LDAs carried out on the 17 colon cancers in the study. Ideally both unfractionated and fractionated LDAs should be completed for each individual tumour, thereby obtaining CC-IC frequencies and percent enrichment specific for each colon cancer CD44 and CC-ICs Initially identified as a C-IC marker in breast cancer(al Hajj et al., 2003), CD44 has since been utilized to identify the tumourigenic fraction in a variety of solid tumours, including: colon(dalerba et al., 2007), pancreas(li et al., 2007), and head and neck(prince et al., 2007). In colon cancer, Dalerba et al. identified that the combined expression of epithelial cell adhesion molecule (EpCAM) and CD44 in colorectal cancer cells enriched for tumour-initiating potential in a NOD/SCID subcutaneous xenograft model(dalerba et al., 2007). They were able to obtain xenograft formation with the injection of, as few as, 200 to 500 EpCAM high CD44 + cells, whereas 10 4 EpCAM low CD44 - cells did not form xenografts. The tumours derived from the EpCAM high /CD44 + cells reproduced the morphological and histological heterogeneity of the parent tumours from which they were derived. The study consisted of a total of eight colorectal cancer specimens and the frequency of the EpCAM high /CD44 + population ranged from 0.8% to 38%. The authors detected CD133 expression on some of the colon cancers studied; they noted that CD44 + cells typically co-expressed CD133, with the CD44 + cells representing the smaller subset(dalerba et al., 2007). The authors also studied CD44 in the context of another marker,

26 12 CD166, a mesenchymal stem cell marker(bruder et al., 1998). It has previously been published that a subset of colorectal cancers express CD166 and that increased CD166 expression levels have been associated with poor clinical outcome in colorectal cancer(weichert et al., 2004). Utilizing two unpassaged primary colorectal cancer specimens and a xenografted sample, the authors demonstrated that the tumour-initiating capacity existed solely in the EpCAM + CD44 + CD166 + colon cancer cells; a subpopulation that ranged from 3.3% to 35.6% of the total tumour cells(dalerba et al., 2007). As per all studies in this field, the sample size was small with six out of the eight samples tested being from high passage xenografts and only two of the samples being tested directly from the patient tumour. Since the Dalerba et al. study, another group has also identified the combined use of CD133 and CD44 as a useful tool to further purify the tumour-initiating fraction(haraguchi et al., 2008). Interestingly, yet another study found that the CD44 alone provided better purification, as opposed to using it in combination with CD133(Chu et al., 2009). It should be acknowledged that published studies in this area typically test a very small sample size, with the last two aforementioned studies testing less than five colon cancer samples each. This is likely one of the main factors leading to the conflicting results Aldehyde dehydrogenase-1 and CC-ICs Additional markers continue to be identified that can segregate CC-ICs and non-cc-ics. For example, aldehyde dehydrogenase 1 (ALDH1), a detoxifying enzyme that oxidizes intracellular aldehydes, has been used to enrich for CC-ICs(Chu et al., 2009; Huang et al., 2009). Huang et al. showed that 100 ALDH1 + primary human colon cancer cells could initiate xenografts in mice, and as few as 25 ALDH1 + colon cancer xenograft-derived cells could be serially passaged in NOD/SCID mice(huang et al., 2009). They also studied ALDH1 in the

27 13 context of known CC-IC markers, CD44 and CD133. The percentage of cancer cells expressing only ALDH1 was %. In contrast, the cell subsets expressing either CD44 or CD133 were significantly larger being % and %, respectively. The proportion of colorectal cancer cells positive for both ALDH1 and CD44 was 1.3%±0.6%. The authors found that the use of CD44 in combination with ALDH1 did not provide any further enrichment of the tumourinitiating activity, as compared to the use of ALDH1 alone. Both the ALDH1 + and the ALDH1 + CD44 + cells displayed tumor formation frequencies of approximately 60%. The combination of CD133 with ALDH1 revealed a population that comprised % of the total cancer cells. Unlike the use of CD44, the combination of CD133 and ALDH1 did reveal enrichment in the tumour-initiating capacity. The ALDH1 + tumour-initiation rate was 58% when used alone. However, this was increased to 89% in the xenografts derived from the cancer cells expressing both CD133 + ALDH1 +. The authors also noted that the injection of CD44 - or CD133 - colon cancer cells did not give rise to xenograft formation; yet, the injection of ALDH + CD44 - or ALDH + CD133 - demonstrated tumour-initiation rates of 70% and 44%, respectively. The authors attributed this observation to the fact that the cells expressed ALDH1, which conferred the tumour-initiating capacity. The ALDH1 + fraction only constitutes a small proportion of the total CD44 - and CD133 - populations, therefore the authors hypothesized that these fractions typically did not form xenografts because the ALDH + CD44 - and ALDH + CD133 - cells are typically present in such low concentrations(huang et al., 2009). Once again, the study consisted of a small sample size with the majority being highly passaged xenografts; nevertheless, it clearly demonstrates that this work is becoming increasingly complex with the addition of each potential CC-IC marker.

28 14 Future studies will be required to validate the combined use of CD133, CD44, CD166, and ALDH1 to further purify the CC-IC fraction. It is very possible that different subtypes of human colon cancers will have CC-ICs identified by distinct cell surface phenotypes; however, determining this will require larger scale studies looking at a variety of markers and multiple combinations thereof. 1.4 C-ICs and self-renewal The emerging complexity of the C-IC phenotype and function is at times daunting and has led to some confusion in the field. However, at its core, the C-IC model is about identifying and characterizing the cancer cells that possess the greatest capacity to regenerate all aspects of the tumour. It is becoming clear that cancer cells evolve as a result of their ability to hijack normal self-renewal pathways, a process that can drive malignant transformation. Studying self-renewal in the context of cancer and C-IC maintenance will lead to a better understanding of the mechanisms driving tumour growth. Increasing evidence suggests that the identification of C-IC cell surface phenotypes can only take the field so far. Only through achieving a better understanding of the self-renewal pathways fuelling C-IC propagation will we start to grasp the functional nature of these cells (Table 1.3). While a number of mouse transgenic studies have demonstrated the importance of self-renewal pathway activation for C-IC maintenance(lessard and Sauvageau, 2003; Yilmaz et al., 2006), few studies have shown this explicitly using human C-IC xenograft models C-ICs and the Wnt, Pten, and Akt pathways Jamieson et al were one of the first groups to demonstrate the importance of a self-renewal pathway in maintaining LSCs(Jamieson et al., 2004; Abrahamsson et al., 2009). They identified aberrant Wnt/ -catenin self-renewal pathway activation to be the driving force in human blast

29 15 crisis LSC propagation(jamieson et al., 2004; Abrahamsson et al., 2009). More recently, increased Wnt/ -catenin signalling has also been implicated in the maintenance of breast C- ICs(Korkaya et al., 2009). The authors demonstrated that the genetic knockdown of protein phosphatase and tensin homolog (Pten) both enriches for breast C-IC markers and increases tumourigenicity in a xenograft model. The effect of Pten knockdown on C-ICs (ALDH1 + ) was mediated by activation of Akt signalling, which resulted in an increase in Wnt/ -catenin activity(korkaya et al., 2009; Ginestier et al., 2007). This work also exemplifies the potential cooperative effect between distinct self-renewal pathways, such as PTEN and Wnt. It is plausible and likely probable that multiple dysregulated self-renewal pathways are functioning to maintain the CSC subset. Our understanding of Wnt activation in the context of C-ICs remains at an early stage, however, it is evident from preliminary work that the Wnt pathway plays a critical role in the initiation and maintenance of C-ICs(Korkaya et al., 2009; Jamieson et al., 2004; Abrahamsson et al., 2009; Takahashi-Yanaga and Kahn, 2010) C-ICs and the hedgehog pathway Another known regulator of self-renewal in the context of embryogenesis is the sonic hedgehog (Hh) signalling pathway; however, little is known about its role in adult stem cells and C- ICs(Ingham and Placzek, 2006; Ingham, 2008; Mechant and Matsui, 2010). The preferential expression of Hh in C-ICs was first published in a pancreatic cancer xenograft model(li et al., 2007). Thereafter, the Hh pathway has also been implicated as having an essential role in maintaining human LSCs(Dierks et al., 2008; Zhao et al., 2009). Loss of the Hh pathway component, smoothened (Smo), resulted in depletion of the chronic myelogenous leukemia (CML) stem cell subset. Moreover, the constitutive activation of Smo resulted in an increased number of CML stem cells and acceleration of the disease(zhao et al., 2009). There is emerging

30 16 evidence that the Hh pathway is also activated in a number of solid tumour C-IC models including: breast(liu et al., 2006), glioblastoma(clement et al., 2007; Bar et al., 2007), and colon(varnat et al., 2009), providing the impetus for a plethora of early phase clinical trials aimed at expunging C-ICs C-ICs and the Notch pathway The Notch pathway is also known to play a critical role in stem cell growth and differentiation(pannuti et al., 2010). Recent work by Hoey et al. demonstrated that the Notch pathway is also activated in the colon C-IC subset(hoey et al., 2009). Utilizing antibodies targeting Delta-like 4 ligand (DLL4), an important component of the Notch pathway, they were able to inhibit the growth of human colon cancer xenografts. One of the mechanisms by which the DLL4 antibody inhibited tumour growth was by directly modulating Notch signalling in the C-IC-enriched population(hoey et al., 2009). An essential role for Notch pathway activation has also been identified in breast (Hoey et al., 2009; Harrison et al., 2010)and glioblastoma (Fan et al., 2010) C-IC models. A myriad of additional pathways such as transforming growth factor beta (TFGβ)(Ikushima et al., 2009) and bone morphogenic protein (BMP)(Piccirillo and Vescovi, 2006) have been shown to influence C-IC initiation and maintenance. Another example is the polycomb group member B lymphoma Mo-MLV insertion region-1 (BMI-1), which has a well established role in self-renewal(park et al., 2004). BMI-1 is preferentially expressed in head and neck C-ICs(Prince et al., 2007) and the genetic knockdown of BMI-1 has been shown to impair C-IC self-renewal capacity in hematopoietic(lessard and Sauvageau, 2003; Rizo et al., 2009), breast(liu et al., 2006)and brain (Abdouh et al., 2009) xenograft models.

31 C-ICs and the inhibitors of differentiation There are numerous additional genes with recognized roles in self-renewal of normal stem cells that remain to be studied in the context of C-ICs. One such family of genes are the inhibitors of differentiation which have recently been identified as central regulators of neural stem cell self-renewal. The inhibitor of DNA binding proteins (Ids) comprise a family of four homologous helix-loop-helix (HLH) transcriptional regulatory factors (Id1-Id4)(Gray et al., 2008). Unlike other HLH factors Ids lack a basic domain and hence are unable to bind DNA directly. Instead they function by forming inactive heterodimers with other HLH factors including: MyoD, E proteins, Ets transcription factors. They have widely recognized roles in: development, senesence, differentiation, angiogenesis, and migration(fong et al., 2004). The constitutive expression of Id proteins has been shown to inhibit the differentiation of various tissues. Although the Ids are highly expressed at different time points in embryonic development, their expression is undetectable in most mature tissues(fong et al., 2004; Benezra, 2001). The one exception being malignant cells where Id proteins are often detected and thought to play multiple roles in tumourigenesis. Experimental models support a role for the Id proteins in tumourigenesis, one example being over-expression of Id1 in hematopoetic cells which leads to disruption of differentiation and leukemia(birkenkamp et al., 2007). Similarly, the overexpression of Id1 in murine intestine led to adenoma formation, albeit it at long latency and low penetrance(wice and Gordon, 1998). More recently, Gupta et al demonstrated that Id1 and Id3 are selective mediators of lung metastatic colonization in the triple negative subset of breast cancers(gupta et al., 2007). Functional studies revealed that Id1 and Id3 were required both in the context of primary tumour formation and during metastatic colonization of the lung microenvironment(gupta et al., 2007). Swarbrick et al demonstrated that the over-expression of Id1 alone in murine mammary epithelium was not sufficient for tumourigenesis, however, mice

32 18 with expression of both Id1 and activated RAS developed metastatic cancer(swarbrick et al., 2008). These tumours expressed high levels of P19arf, p53, p21 waf1 ; leading the authors to conclude that Id1 acts to make cells refractory to p21 waf1 -dependent cell cycle arrest. Conditional inactivation of Id1 in established tumours led to widespread senescence and tumour regression in 40% of the mice, as well as, a significant decrease in pulmonary metastatic load(swarbrick et al., 2008). Interestingly, many of the published papers on the Ids have acknowledged roles for these proteins that overlap with the postulated roles of C-ICs including: promotion of angiogenesis, chemoresistance, and metastasis. The role of Ids in C-ICs and more specifically in maintaining the ability of these cells to self-renewal remains to be investigated(perk et al., 2005; Benezra, 2001) Therapeutically targeting self-renewal Identifying and understanding the role of individual self-renewal pathways in maintaining C-ICs is the first step. However, the eventual goal is to generate targeted therapeutics that inhibit these essential pathways in the C-IC fraction. The targeting of these pathways will likely be complicated by the fact that the same pathways are also pivotal in normal stem cell function. There is preliminary evidence in leukemia models to suggest that there may be subtle differences between how these pathways are operating in malignant versus normal stem cells. One such example is the compound Parthenolide, an agent that selectively targets LSCs, and has no known deleterious effect on the normal hematopoietic stem cells (HSCs)(Jordan, 2008; Neelakantan et al., 2009). Rapamycin was also found to selectively target LSCs in a Pten deletion murine model of leukemia while at the same time improving normal HSC function(yilmaz et al., 2006). Similarly, targeted small molecule inhibition of the sonic hedgehog pathway combined with BCR-ABL inhibition has been shown to markedly reduce CML stem cell propagation(dierks et al., 2008; Zhao et al., 2009).These studies emphasize the importance of defining the cell type and

33 19 context specific effects of self-renewal pathway inhibition to eliminate C-ICs while limiting the effect on normal stem cells. 1.5 C-ICs and the microenvironment Identifying the relevant self-renewal pathways driving C-ICs will also enable us to better understand the role of the microenvironment in initiating and maintaining C-ICs. Over 100 years ago Paget proposed his seed and soil hypothesis to explain why particular cancers preferentially metastasize to certain organs, such as colon cancer metastasizing to the liver(fidler, 2003; Fidler and Poste, 2008; Nakamura et al., 2007). It is plausible that the receptor-ligand interactions between the tumour and local microenvironment govern the activation of specific self-renewal pathways in a particular C-IC, thereby allowing it to initiate tumour growth at a distant site(labarge, 2010). Hermann et al. published work demonstrating that for a xenograft model of human pancreatic cancer the phenotype for the C-IC subset that could give rise to a metastatic deposit differed from that which gave rise to tumour at the orthotopic site, CD133 + CXCR4 + and CD133 + CXCR4 +, respectively(hermann et al., 2007). This demonstrates that the microenvironment, in part, determines which cancer cell possesses the capacity to self-renew. Furthermore, this work illustrates that much remains to be learned about the signals that a C-IC receives from the microenvironment and the role it plays in driving C-IC self-renewal (Figure 1.2). The importance of the microenvironment and in particular tumour-associated stromal cells is best illustrated in elegant studies carried out by Yauch et al. on the Hh pathway(yauch et al., 2008). They demonstrated that inhibition of Hh signalling in pancreatic cancer-associated stromal cells resulted in the suppression of tumour growth. In contrast, inhibition of Hh signalling in the pancreatic cancer cells themselves did not affect the tumour(yauch et al., 2008).

34 20 This suggests that in some tumours paracrine, as opposed to autocrine or endogenous Hh signalling, is essential for maintaining tumour growth. The exact mechanism by which Hh pathway inhibition in the stromal microenvironment suppresses tumour growth remains to be determined. Adding further complexity to the field, there is some evidence to suggest that the inhibition of the Hh pathway in stromal cells can lead to changes in Wnt pathway components(yauch et al., 2008). Recent publications have established that Hh signalling is activated in leukemic(dierks et al., 2008; Zhao et al., 2009), breast(liu et al., 2006), brain(clement et al., 2007; Bar et al., 2007), and colon (Varnat et al., 2009) C-ICs, however, the role of cancer associated stromal cells in initiating and maintaining these C-ICs remains to be determined. One hypothesis is that a cancer cell s ability to function as a C-IC depends on whether it possesses the ligand/receptor required to respond to the self-renewal signals being emitted by the surrounding stroma. If this is proven to be the case it may explain the predilection of C-ICs for specific metastatic sites because the self-renewal pathways that are utilized by an individual C-IC will depend on the microenvironment of the organ in which it exists. 1.6 C-ICs and metastasis There are striking similarities between the understanding of how cancer cells metastasize and the emerging functional definition of C-ICs. Interestingly, both the C-IC and fully metastatic cell represent rare clones within primary tumours(chiang and Massague, 2008). In animal models, 0.01% or fewer of cancer cells entering the circulation develop into metastases(chambers et al., 2002; Luzzi et al., 1998). The low metastatic efficiency of cancer cells may be related to the extreme environmental stresses that these cells are exposed to upon entering the circulation, including: lack of oxygen or nutrients, a low ph, and reactive oxygen

35 21 species (ROS)(Gupta and Massague, 2006). It has been postulated that only those cancer cells that acquire an aggressive phenotype capable of overcoming these environmental stressors will survive to give rise to a metastatic lesion. According to the C-IC model, only C-ICs are capable of initiating and maintaining the tumour, which often leads to the assumption that C-ICs must be the cells responsible for metastatic dissemination. The relationship between C-ICs and metastasis, although theoretically plausible, requires further experimental proof. The first evidence that supported a potential role for C-ICs in metastasis was published in breast cancer, where it was determined that the majority of early disseminated cancer cells detected in the bone marrow of breast cancer patients have a putative breast cancer stem cell phenotype(balic et al., 2006). This led researchers to question whether C-ICs in breast cancer have biological properties that facilitate their metastatic spread and enable them to colonize distant sites. In support of this hypothesis, Diehn et al. published work in breast cancer, as well as in other solid tumours, which indicated that C-ICs express increased levels of free radical scavenging systems and consequently have lower levels of ROS when compared to their nontumour initiating counterparts (Diehn et al., 2009). To assess whether the lower levels of ROS detected in the C-IC fraction was providing them with a survival advantage the authors pharmacologically depleted glutathione (GSH), this in turn led to an increase in the ROS levels. GSH is a critical cellular antioxidant that has been implicated in both the chemotherapy and radiotherapy resistance of cancer cells. The depletion of GSH and the resulting increase in ROS levels in the C-IC subset significantly decreased their clonogenicity suggesting that the low ROS levels provide the C-ICs with a survival advantage. High ROS levels are one of the recognized barriers to a cancer cell s ability to successfully metastasize(chiang and Massague, 2008). Therefore, it is interesting that C-ICs possess mechanisms to maintain lower ROS levels, which

36 22 may function to protect these cells against environmental stressors and give them an advantage during the metastatic process. More recently, Li and colleagues demonstrated that in brain cancer xenografts, the C-IC subset preferentially expresses hypoxia-inducible factor 2α (HIF2α) and multiple HIF-regulated genes, as compared to the non-tumour-initiating fraction(li et al., 2009). Targeting HIF2α in the C-IC population inhibited self-renewal, proliferation, and attenuated xenograft growth in vivo. They also determined that HIF2α is required for vascular endothelial growth factor (VEGF) expression in the C-IC subset(li et al., 2009). This result supports the notion that C-ICs can preferentially survive in hypoxic environments and furthermore it suggests that C-ICs possess the ability to establish the vascular niche for the tumour. In summary, evidence is emerging that C-ICs share a number of characteristics with the cells that are able to successfully metastasize. It is easy to assume that the C-ICs are therefore responsible for metastatic disease but the experimental evidence is very limited at this time and will require extensive work to conclusively answer this question. A very elegant study was published by Hermann et al. that explored the question of C- ICs in metastasis in the context of pancreatic adenocarcinoma(hermann et al., 2007). First, they determined that CD133 marks only a small subset of pancreatic tumour cells and that only the CD133 + cells are able to initiate tumours when injected into the pancreas of immunocompromised mice(hermann et al., 2007). Injection of 10 6 CD133 - cells did not produce tumours, thereby confirming that in their system the C-IC activity existed only within the CD133 + pancreatic cancer cells. In addition to primary pancreatic cancer tissue, they also studied two pancreatic cancer cell lines, L3.6pl and FG, which have high and low metastatic potential, respectively. Utilizing an orthotopic xenograft model, they demonstrated that both cell

37 23 lines possessed a CD133 + C-IC subset, however, only CD133 + cells from the L3.6pl cell line possessed the ability to metastasize. The CD133 + cells from the FG cell line could form a xenograft at the site of injection but lacked any metastatic capability(hermann et al., 2007). The authors noted that only the L3.6pl cell line expressed the chemokine (C-X-C motif) receptor CXCR4 and questioned whether this might be playing a role in the ability of these CD133 + cells to metastasize. Interestingly, CXCR4 and its ligand, stromal derived growth factor 1 (SDF-1) (also known as CXC chemokine ligand 12 [CXCR12]), play important roles in the survival of both breast and renal cancer cells at sites of metastasis (Muller et al., 2001; Staller et al., 2003; Zagzag et al., 2005). To determine whether CXCR4 is functionally important in pancreatic adenocarcinoma C-ICs and metastasis, the L3.6pl (high metastatic ability) CD133 + cell population was fractionated based on CXCR4 expression. The CD133 + CXCR4 + and CD133 + CXCR4 - cell subsets were both capable of generating xenografts at the site of injection in NOD/SCID mice, demonstrating that the tumour-initiating capacity of the two cell subsets was equivalent. However, when they analyzed the blood from the xenograft-bearing mice they found that only the double positive cells (CD133 + CXCR4 + ) could be detected in the circulation and this correlated with a metastatic phenotype in the mice. Furthermore, they inhibited CXCR4 pharmacologically, which resulted in abrogated metastasis in their murine xenograft model(hermann et al., 2007). In 2010 Pang et al. identified a subpopulation of C-ICs from human colon cancers that are capable of forming metastases in an orthotopic SCID mouse model. Interestingly, all colon cancer patients included in the study that presented with synchronous or metachronous liver metastases harbored a CD133 + subset that expressed CD26(Pang et al., 2010). When injected orthotopically into SCID mice only the CD133 + CD26 + cells led to the formation of liver metastases, in contrast the CD133 + CD26 - cells were enriched for tumour formation at the site of

38 24 injection in the cecum but were unable to establish metastatic deposits in the liver. Similar to the work by Hermann et al. this demonstrates that heterogeneity exists within the CC-IC fraction, with distinct subsets being responsible for the generation of the primary tumour and metastatic deposits(pang et al., 2010). In an attempt to correlate these results with clinical outcomes with CD26 expression the authors analyzed 27 primary colon cancers without metastasis at the time of presentation. A total of 8 out of the 27 cases had an identifiable subpopulation of CD26 + cells, of which 5 of these patients developed liver metastases during follow-up. In contrast, none of the 19 patients without CD26 + cells developed distant metastases during the follow-up(pang et al., 2010). These results suggest that CD26 expression should be further investigated not only as a CC-IC marker but as a possible biomarker of colon cancers with increased metastatic propensity. These publications were the first to describe heterogeneity in the C-IC compartment of solid tumours, where a distinct subset of C-ICs was responsible for metastatic dissemination. One could hypothesize that there may be a subset of C-ICs that preferentially disseminates to a specific organ depending on the cell surface phenotype and micro-environmental niche that best support its survival. It is clear that there are currently more questions than answers as they relate to C-ICs and their role in metastasis. Understanding this relationship will provide insight into the mechanisms that cancer cells employ to survive hostile environments, as well as potentially providing insight into why tumours have well-defined patterns of metastasis (Paget, 1989). 1.7 C-ICs and therapeutic resistance The identification and functional isolation of C-ICs represents the first step of what remains a very young field in the area of solid tumour biology. The interest focused on the field stems in large part from the potential clinical promise of studying the C-IC subset. The identification of these cells has led cancer researchers to question whether these are the cells most responsible for

39 25 disease recurrence and metastases (Figure 1.3). If this is proven to be the case it would change how future adjuvant therapies are developed. Instead of the current standard of practice where adjuvant agents are chosen for their activity against the bulk of the tumour cells, the focus would shift to identifying agents that specifically target the C-IC subset. Preliminary evidence exists to indicate that the C-IC subset are preferentially resistant to the effects of both radiation(bao et al., 2006) and chemotherapy(li et al., 2007; Todaro et al., 2008), when compared to the non-c-ic fraction within the same tumour. Furthermore, there is evidence that the C-IC subset may be responsive to differentiation therapies, which in the future may represent a new way in which to approach the treatment of solid tumours(piccirillo and Vescovi, 2006) Chemoresistance and C-ICs There have been a number of studies published in xenograft models demonstrating chemotherapy resistance of the C-IC fraction. One such study was published by Hermann et al who treated both C-IC (CD133 + ) and non-c-ic (CD133 - ) pancreatic cancer cells with the standard chemotherapy agent gemcitabine. The CD133 + cells showed a preferential drug resistance to gemcitabine treatment, as compared to the CD133 - cells from the same tumour. Following 5 days of in vitro treatment with gemcitabine the percentage of CD133 + cells had increased approximately 50 fold, from 1.5% to 47%. Furthermore the authors carried out an in vivo study where they evaluated xenografts derived from mice which had received either gemcitabine or vehicle treatment biweekly for 21 days. The tumours treated with gemcitabine displayed a profound enrichment of CD133 + cells(hermann et al., 2007). These preliminary results suggest that the pancreatic C-IC fraction is resistant to conventional chemotherapy, however, these results must be verified in the clinical setting with human patients.

40 26 Todaro et al. were the first to publish evidence that the CC-IC and non-cc-ic fractions of human colon cancers responded differently to conventional chemotherapeutic agents, such as 5-Fluorouracil (5-FU) and oxaliplatin(todaro et al., 2008). Utilizing CD133 as the CC-IC marker, they demonstrated that CD133 + cells were relatively resistant to standard chemotherapeutic agents. In contrast, CD133 - cells derived from the same tumours were highly sensitive to both in vivo and in vitro treatment with 5-FU or oxaliplatin. The authors determined that one of the mechanisms by which the CD133 + cells protect themselves from conventional chemotherapies is through the autocrine production of interleukin-4 (IL-4). By treating mice with a neutralizing antibody against IL-4, they were able to render previously resistant CD133 + colon cancer cells sensitive to treatment with oxaliplatin and/or 5-FU, thereby leading to an overall decrease in the number of CD133 + cells(todaro et al., 2008). Interestingly, the increased chemosensitivity in the cancer cells, following treatment with the IL-4 antibody, was associated with a reduction in pro-survival molecules, such as cflip, Bcl-x L, and PED in the CD133 + fraction(todaro et al., 2008). Another study published by Dylla et al. looked at the response of CC-ICs (ESA + CD44 + ) to irinotecan or cyclophosphamide treatment(dylla et al., 2008). Human colon cancer cells were utilized to generate xenografts, which were subsequently treated for two weeks with irinotecan. Following treatment, the xenografts were harvested and it was determined that the percentage of tumour cells displaying the CC-IC phenotype was increased by 61% in the xenografts treated with irinotecan, as compared to untreated control xenografts. Furthermore, the xenografts were tested in a limiting dilution serial transplantation assay and only the cells displaying the CC-IC phenotype could generate tumours in both the irinotecan and control treated groups(dylla et al., 2008). The authors demonstrated that not only was there an increase in CC-IC marker

41 27 expression after exposure to irinotecan, but that this change was also functionally important because it was associated with an increase in the tumour-initiating capacity. To investigate the relationship between cyclophosphamide (CPA) treatment and CC-ICs the authors used a different approach. Existing literature suggests that resistance to CPA results from high ALDH activity. Interestingly, when Dylla et al. studied colon cancer cells for ALDH1 expression, it was determined that the CC-IC fraction (ESA + CD44 + ) possessed the majority of the ALDH1 activity(dylla et al., 2008). Moreover, the tumour-initiating potential was greater in the ESA + CD44 + ALDH1 + versus the ESA + CD44 + ALDH1 - colon cancer cells. Phenotypically the ALDH1 + subpopulation of ESA + CD44 + cells was consistently higher after CPA treatment, as compared to the vehicle treated controls (67±6.3% versus 56.8±6.8% respectively). Interestingly, the inhibition of ALDH1 activity either by short-hairpin RNA interference or chemical-mediated inhibition by diethylaminobenzaldehyde did not significantly inhibit xenograft growth. However, the decreased ALDH1 activity did render the cells sensitive to CPA treatment, resulting in a significant inhibitory effect on xenograft growth(dylla et al., 2008). Interestingly, the inhibition of ALDH1 activity did not have any effect on the resistance of CC- ICs to irinotecan, indicating that the mechanism of resistance is drug-specific. This work was the first to demonstrate the complex mechanisms utilized by CC-ICs to protect themselves from the stressors in their environment and confer a survival advantage. One can assume that there are multiple survival mechanisms utilized by the CC-ICs; deciphering these mechanisms and developing methods to perturb them will require better purification and a deeper understanding of CC-IC biology.

42 Radioresistance and C-ICs The evidence for the radioresistance of the C-IC subset stems mainly from work carried out in brain tumours. Using a xenograft model of glioma Bao et al. found that the CD133 + C-IC fraction was significantly increased following irradiation, resulting in tumours that possessed an increased percentage of CD133 + cells relative to the parent tumour from which they were derived(bao et al., 2006). These findings were consistent whether the cells were irradiated in vitro, prior to injection, or post-implantation into NOD/SCID mice. The enrichment in the C-IC fraction correlated with these cells having a survival advantage following irradiation, as compared to their non-c-ic counterparts (CD133 - cells). It was determined that the increased survival of the CD133 + cells following irradiation stemmed from their ability to activate the DNA damage response more readily than their CD133 - counterparts. The use of a small molecule inhibitor against the Chk1/2 kinases (members of the DNA damage and replication checkpoint) rendered the CD133 + cells radiosensitive and eliminated their survival advantage over their CD133 - counterparts(bao et al., 2006). These findings suggest that the DNA damage response plays an important role in the survival of glioma C-IC, however, future studies are required to confirm these findings in a clonogenic assay(baumann et al., 2008). More recently Zhang et al. demonstrated that C-ICs isolated from p53 null mouse mammary tumours are better poised to repair DNA damage following in vivo ionzing radiation, as compared to bulk tumour cells. The C-IC subpopulation in the p53null mouse model was characterized by the expression of Lin-CD29highCD24high, markers which have been previously utilized to isolate normal mammary stem cells. Lending support to the notion that C- ICs are inherently more radioresistant, microarray analysis revealed increased expression of cell cycle checkpoint and DNA damage repair genes within the C-IC subpopulation(zhang et al.,2010). Utilizing a Wnt reporter virus the authors also demonstrated that the C-IC fraction

43 29 preferrentially expressed increased Wnt signaling, as compared to the bulk cells of the tumour. Furthermore, through inhibition of the Akt pathway the authors were able to both inhibit Wnt signaling and repair of DNA damage selectively in C-ICs, thereby sensitizing them to ionzing radiation. This work not only demonstrates that C-ICs are radioresistant but proceeds to unravel some of the mechanisms driving this resistance (Zhang et al.,2010) Resistance to differentiation and C-ICs By its nature, a hierarchy model implies that cancer cells are capable of maturation, albeit abnormal. Thus, another avenue which is being investigated is the identification of methods that can induce differentiation of the C-IC population. In a xenograft glioblastoma model, researchers demonstrated that transient in vitro exposure or in vivo treatment of established tumours with bone morphogenic protein-4 (BMP4) abolished the ability of transplanted glioblastoma cells to establish intracerebral xenografts(piccirillo and Vescovi, 2006). The exposure of glioblastoma cells to BMP4 resulted in a reduction in proliferation and increased expression of markers of neural differentiation. Interestingly there was no observed effect on cell viability, as determined by rates of cell death and apoptosis. There was an observed decrease in the size of the CD133 + pool by approximately 50% which correlated with the observed decrease in clonogenic ability(piccirillo and Vescovi, 2006). The results from this study demonstrate that the C-IC fraction may still possess the ability to respond to biological cues for maturation which may in the future lead to new non-toxic therapies being developed that utilize these findings to drive differentiation of malignant tissues. More recently another study demonstrated that there is a subset of glioblastoma C-ICs that do not undergo differentiation in response to BMP4 treatment. The inability to respond to BMP4 treatment was the result of EZH2-dependent epigenetic silencing of BMP receptor 1B

44 30 (BMPR1B), a state normally found in early embryonic neural stem cells. Forced expression of BMPR1B, either by transgene expression or demethylation of the promoter, restored the differentiation capacity of the glioblastoma C-IC and induced the loss of their tumourigenicity in response to BMP4 treatment(lee et al., 2008). This elegant work takes the concept of C-IC a step further by attempting to better understand the operative and aberrant differentiation pathways in glioblastoma C-ICs. It also stresses the importance of understanding the biology of the C-IC fraction and developing an appreciation for the biological diversity that exists within the subset that we currently lump together as the C-IC fraction. If novel therapeutic approaches are to be successful there is an enormous amount of understanding that must be achieved so that treatments can be chosen based on a solid understanding of the aberrant biology that underlies the cells that drive tumour formation. 1.8 Clinical relevance of C-ICs C-ICs as biomarkers The identification of C-IC markers has led to interest in being able to utilize these cell surface phenotypes as prognostic indicators. The most logical method in which to approach this question is through the utilization of immunohistochemistry (IHC) techniques to quantitate the expression of the cell surface markers and correlate this information back to stage, grade, and disease-free survival. The number of publications in this field remain limited and although initial results have been very interesting, larger studies are required before clinically relevant biomarkers are validated(choi et al., 2009; Ginestier et al., 2007; Horst et al., 2008; Horst et al., 2009a; Horst et al., 2009c; Horst et al., 2009b). The first study reviewed a series of 77 colorectal cancer patients, of whom 21 (27%) died of colorectal cancer within 5 years of diagnosis(horst et al., 2008). They employed a semi-quantitative approach where tumours were scored as 0, <50%,

45 31 or 50% CD133 + cells. CD133 expression did not correlate with age, gender, or stage of the tumour. However, by multivariate analysis CD133 expression was demonstrated to be an independent marker for decreased patient survival: CD133 high expression represented a relative risk of 2.45 as compared to the CD133 low group(horst et al., 2008). More recently, Choi et al. studied the relationship between the expression of CC-IC phenotypic markers (CD133 and CD44) as they relate to colorectal cancer: stage, differentiation status, and outcomes(choi et al., 2009). The retrospective study consisted of a consecutive series of 523 colorectal cancer cases with complete histopathological data. The authors concluded that CD133 and CD44 expression had no effect on overall survival by univariate or multivariate analysis. However, they did demonstrate that CD133 protein expression was higher in the more advanced stage colon cancers, whereas larger tumours (>5.5cm) showed greater CD44 expression(choi et al., 2009). These studies provide good examples of the ongoing controversy in the field of CC-IC research and demonstrate why no definite conclusions can be drawn based on the published literature. An important issue with the use of known C-IC markers for IHC studies is the understanding that the markers identified to date represent an enrichment of tumour-initiating activity. Further purification of the C-IC subset is required to better delineate the most relevant cells. An interesting approach to this problem is to combine C-IC markers with molecular markers of pathways postulated to have an essential role in the maintenance of the C-IC subset. This approach was employed by Horst et al. in a recent publication where they studied the coexpression of CD133 and nuclear β-catenin and correlated this to outcome data in colorectal cancer patients(horst et al., 2009a). Previous work by this group had demonstrated that most colon cancers display a heterogeneous expression pattern of nuclear β-catenin, and increased

46 32 expression was correlated with decreased survival(horst et al., 2009a). Furthermore, there is data from other cancers that nuclear β-catenin may play an important role in maintaining C- ICs(Korkaya et al., 2009). The authors hypothesized that nuclear β-catenin, like CD133, may also mark the CC-IC subset. To test this, they studied the co-expression of the two markers in 162 stage IIa colon cancers. The combined evaluation proved to be very powerful in identifying patients with stage IIa disease that harbored an increased risk of recurrence. For the subset of patients whose tumours demonstrated the highest co-expression of CD133 and nuclear β-catenin, the 5-year survival rate was only 47±13%, a figure approaching the five-year survival rates in stage IIIc patients(horst et al., 2009a). However, the authors caution that these results remain preliminary and must be confirmed in larger studies. Future research using similar methods of combining CC-IC phenotypes with molecular markers from relevant pathways will hopefully lead to these biologically relevant marker combinations being employed as clinically useful prognostic indices. A recent study by Ginestier et al. demonstrated that another marker, ALDH1, could prospectively identify the breast cancer cell subset capable of xenograft formation and serial transplantation(ginestier et al., 2007). As mentioned previously in this chapter, ALDH1 has also been identified as a C-IC marker in colon cancer(huang et al., 2009). Interestingly, apart from identifying ALDH1 as a C-IC marker in breast cancer, they also studied a series of 577 breast cancer patients and determined that expression of ALDH1 correlated with poor prognosis. The relative risk of death due to cancer was 1.76 for patients with ALDH1-positive tumours as compared to patients with ALDH1-low or -negative tumours(ginestier et al., 2007). The potential value of ALDH1 as a prognostic marker in colon cancer remains to be studied.

47 Integrating C-ICs into clinical trials There is a keen interest to determine the clinical relevance of C-ICs and an essential aspect to answering this question requires incorporating the study of C-ICs into clinical trials. Li et al. were the first to study C-ICs in the context of a neo-adjuvant clinical trial for locally advanced breast cancer patients(li et al., 2008). The authors compared the C-IC subset (CD44 + CD24 - ) following conventional chemotherapy (docetaxol or doxorubicin and cyclophosphamide) alone or with the addition of lapitinib (an epidermal growth factor receptor/her2 inhibitor). Tumour samples were obtained pre- and post-treatment and tested both phenotypically and functionally for C-IC activity. The functional testing included in vitro mammosphere formation efficiency (MFSE) and in vivo xenograft formation. The tumours treated with chemotherapy alone demonstrated an enrichment in cells displaying the C-IC CD44 + CD24 - phenotype. In contrast, the patients that received chemotherapy with lapatinib did not display any enrichment in the C-IC phenotype(li et al., 2008). It is important to appreciate that the phenotypic assessment is insufficient to draw any conclusions when used in isolation without functional correlation. The gold standard is to functionally assess tumourigenicity or self-renewal potential of the cancer cells by in vivo limiting dilution analysis. Li et al. were able to demonstrate that not only was there a phenotypic change in the profile of the tumours, but this also translated to a functional change because the tumours treated with lapatinib possessed decreased self-renewal capacity as measured by MFSE and xenograft formation, as compared to the tumours only treated with conventional chemotherapy(li et al., 2008). The authors hypothesized that the observed phenotypic and functional C-IC changes may in part explain the survival benefit conferred by lapatinib. However, the trial was preliminary and larger trials will have to be carried out to confirm these findings. To date no other clinical trials have been published that have incorporated the study of C-ICs. The principles applied by Li et al. provide

48 34 an excellent framework for colorectal cancer, as well as other solid tumours. By sampling the tumours pre- and post-treatment, they were able to both phenotypically and functionally test the self-renewal capacity of the tumours, both aspects being essential when determining the clinical relevance of C-ICs. The ability to carry out informative clinical trials will be essential to proving that the C-IC hypothesis is an important and relevant clinical entity. 1.9 Conclusions The study of C-ICs in solid tumours has expanded exponentially over the past five years following the initial publication by Michael Clarke s group(al Hajj et al., 2003). It has provided a different way to study tumour initiation and maintenance. As a result, it has also led to numerous questions and controversies(jordan et al., 2006; Morrison, 2005). Preliminary evidence indicates that CC-ICs are both biologically and clinically relevant(dylla et al., 2008; Horst et al., 2008; Todaro et al., 2008). Albeit preliminary, the research relating to their clinical relevance suggests that these are the cells capable of surviving conventional chemotherapy. Furthermore, C-ICs share many characteristics with the rare clones within a cancer that are capable of successfully metastasizing. This has led to the question of whether C-ICs are the cells responsible for establishing metastases. There is evidence in support of this notion, particularly in the case of pancreatic adenocarcinoma(hermann et al., 2007); however, definitive proof is pending. This will require a better understanding of the biology of the CC-ICs and the mechanisms they exploit to preferentially survive under environmentally stressful conditions. Larger scale studies are also required to develop a more complete appreciation of the biological diversity that exists within the C-IC subset. Current research should be focused on improved purification of CC-ICs, as well as understanding the biology that makes them so unique. Studying the functional biology of C-ICs

49 35 and more specifically the self-renewal pathways driving C-IC regeneration is essential because it will provide insight into how these cells initiate and maintain tumour growth. A better understanding of the functional biology of these cells will also require studying the microenvironment in which they exist and the role of this interaction in maintaining and possibly defining which cancer cells can function as C-ICs. The interest in the cancer stem cell field is driven in large part by the potential clinical relevance.the C-IC model suggests that the route to eradicating a tumour will require agents that expunge the root cause of the cancer -C-ICs. This will likely prove to be very challenging because the same self-renewal pathways driving C-ICs are also essential in maintaining normal stem cells. However, there is preliminary evidence in LSCs that indicates there are subtle differences in how C-ICs and normal stem cells utilize the same pathways(yilmaz et al., 2006; Neelakantan et al., 2009). If C-ICs are to be successfully targeted in the clinical setting, it will require a thorough understanding of how these pathways function in both normal and malignant cells and developing targeted agents to exploit these differences.

50 Figures Figure 1.1: Models to explain tumour heterogeneity

51 37 Figure 1.1: Models to explain tumour heterogeneity: (A) The stochastic model predicts that the distinct ability of tumour cells to initiate tumours is governed by random variables, such as entry into the cell cycle, a low probability stochastic event. According to this model, it should not be possible to fractionate cells with differential tumour initiation properties. (B) The hierarchical or cancer stem cell model predicts that there is a subset of cancer cells that are responsible for establishing and maintaining tumour growth. According to this model, it should be possible to prospectively isolate these cancer stem cells, or cancer-initiating cells (C-ICs), based on the expression of specific cell surface markers.

52 38 Figure 1.2: Multiple facets to C-IC self-renewal Figure 1.2: Multiple facets to C-IC self-renewal. Increasing evidence is emerging to support the notion that CSC self-renewal decisions can be guided by the activation of several pathways, including Wnt, Notch, Hedgehog, and others. A CSC may autonomously trigger the appropriate signaling cascade to maintain self-renewal with minimal niche support. It is likely that some CSCs need the appropriate microenvironment to provide the stimuli for uncontrolled self-

53 39 renewal. Finally, some cancer cells have lost the capacity to self-renew regardless of stimulating molecules, and hence cannot initiate a tumour. Figure 1.3: Hypothetical models to account for the role of CC-ICs in tumour relapse and metstasis Figure 1.3: Hypothetical models to account for the role of CC-ICs in tumour relapse and metastasis. (A) After chemotherapy, refractory CC-ICs survive which possess the ability to selfrenew and proliferate extensively, allowing them to re-generate a new tumour. (B) CC-ICs are postulated to be responsible for the establishment of metastatic disease. Based on their ability to endure environmental stressors, it is thought that C-ICs may possess the inherent biological mechanisms that give them an advantage during the metastatic process.

54 40 Table 1.1: Phenotypic identification of C-ICs in solid tumours Tumour type Marker(s) used to enrich for CSCs Acute myelogenous leukemia CD34 + CD38 - Breast CD44 + CD24 - Breast ALDH1 + Brain CD133 + Prostate CD44 + high α 2 β 1 CD133 + Head and neck CD44 + Colon CD133 + Colon EpCAM high CD44 + Colon ALDH1 + Pancreas ESA + CD44 + CD24 + Pancreas CD133 + Mesenchymal Side population Lung CD133 + Liver CD90 + Melanoma ABCB5 + Ovarian CD133 +

55 41 Table 1.2: Prospective isolation of CC-ICs Reference CC-IC marker(s) Injection site Mouse strain Minimum cell number for tumour engraftment Range of marker expression in tumours O'Brien et al., 2007 Ricci-Vitiani et al., 2007 Dalerba et al., 2007 Dalerba et al., 2007 Todaro et al., 2007 Haraguchi et al., 2008 Haraguchi et al., 2008 Haraguchi et al., 2008 Chu et al., 2009 Chu et al., 2009 Huang et al., 2009 Huang et al., 2009 Huang et al., 2009 CD133 + Renal capsule NOD/SCID % to 24.5% CD133 + Subcutaneous SCID % to 6.1% EpCAM high CD44 + Subcutaneous NOD/SCID 200 to % to 38% EpCAM high CD44 + CD166 + Subcutaneous NOD/SCID % to 35.6% CD133 + Subcutaneous Nude 2500 to % to 3% CD133 + Subcutaneous NOD/SCID % to 82% CD44 + Subcutaneous NOD/SCID % to 58.4% CD133 + CD44 + Subcutaneous NOD/SCID % to 50.5% CD44 high Subcutaneous SCID/Bg % to 24% CD44 high ALDH + Subcutaneous SCID/Bg % to 9% ALDH + Subcutaneous NOD/SCID 25 to %±1% ALDH + CD44 + Subcutaneous NOD/SCID no data 1.3%±0.6% ALDH + CD133 + Subcutaneous NOD/SCID no data 0.9%±0.2%

56 42 Table 1.3: Self-renewal pathways involved in C-IC maintenance WNT Breast cancer Chronic myeloid leukemia Hedgehog Breast cancer Pancreatic cancer Glioblastoma Chronic myeloid leukemia Colon cancer Notch Colon cancer Breast cancer Glioblastoma BMI1 Murine acute myeloid leukemia Breast cancer Head and neck squamous cell cancer Glioblastoma Acute myeloid leukemia PTEN Murine leukemia Breast cancer BMP Glioblastoma TGF-beta Glioblastoma

57 References Abdouh,M., Facchino,S., Chatoo,W., Balasingam,V., Ferreira,J., and Bernier,G. (2009). BMI1 sustains human glioblastoma multiforme stem cell renewal. J. Neurosci. 29, Abrahamsson,A.E., Geron,I., Gotlib,J., Dao,K.H., Barroga,C.F., Newton,I.G., Giles,F.J., Durocher,J., Creusot,R.S., Karimi,M., Jones,C., Zehnder,J.L., Keating,A., Negrin,R.S., Weissman,I.L., and Jamieson,C.H. (2009). Glycogen synthase kinase 3beta missplicing contributes to leukemia stem cell generation. Proc. Natl. Acad. Sci. U. S. A 106, Al Hajj,M., Wicha,M.S., Benito-Hernandez,A., Morrison,S.J., and Clarke,M.F. (2003). Prospective identification of tumorigenic breast cancer cells. Proc. Natl. Acad. Sci. U. S. A 100, Balic,M., Lin,H., Young,L., Hawes,D., Giuliano,A., McNamara,G., Datar,R.H., and Cote,R.J. (2006). Most early disseminated cancer cells detected in bone marrow of breast cancer patients have a putative breast cancer stem cell phenotype. Clin. Cancer Res. 12, Bao,S., Wu,Q., McLendon,R.E., Hao,Y., Shi,Q., Hjelmeland,A.B., Dewhirst,M.W., Bigner,D.D., and Rich,J.N. (2006). Glioma stem cells promote radioresistance by preferential activation of the DNA damage response. Nature 444, Bar,E.E., Chaudhry,A., Lin,A., Fan,X., Schreck,K., Matsui,W., Piccirillo,S., Vescovi,A.L., Dimeco,F., Olivi,A., and Eberhart,C.G. (2007). Cyclopamine-mediated hedgehog pathway inhibition depletes stem-like cancer cells in glioblastoma. Stem Cells 25, Baumann,M., Krause,M., and Hill,R. (2008). Exploring the role of cancer stem cells in radioresistance. Nat. Rev. Cancer 8, Benezra,R. (2001). Role of Id proteins in embryonic and tumor angiogenesis. Trends Cardiovasc. Med. 11, Bidlingmaier,S., Zhu,X., and Liu,B. (2008). The utility and limitations of glycosylated human CD133 epitopes in defining cancer stem cells. J. Mol. Med.

58 44 Birkenkamp,K.U., Essafi,A., van der Vos,K.E., da Costa,M., Hui,R.C., Holstege,F., Koenderman,L., Lam,E.W., and Coffer,P.J. (2007). FOXO3a induces differentiation of Bcr-Abltransformed cells through transcriptional down-regulation of Id1. J. Biol. Chem. 282, Bonnet,D. and Dick,J.E. (1997). Human acute myeloid leukemia is organized as a hierarchy that originates from a primitive hematopoietic cell. Nat. Med. 3, Bruder,S.P., Ricalton,N.S., Boynton,R.E., Connolly,T.J., Jaiswal,N., Zaia,J., and Barry,F.P. (1998). Mesenchymal stem cell surface antigen SB-10 corresponds to activated leukocyte cell adhesion molecule and is involved in osteogenic differentiation. J. Bone Miner. Res. 13, Brunschwig,A., Southam,C.M., and Levin,A.G. (1965). Host resistance to cancer. Clinical experiments by homotransplants, autotransplants and admixture of autologous leucocytes. Ann. Surg. 162, Chambers,A.F., Groom,A.C., and MacDonald,I.C. (2002). Dissemination and growth of cancer cells in metastatic sites. Nat. Rev. Cancer 2, Chiang,A.C. and Massague,J. (2008). Molecular basis of metastasis. N. Engl. J. Med. 359, Choi,D., Lee,H.W., Hur,K.Y., Kim,J.J., Park,G.S., Jang,S.H., Song,Y.S., Jang,K.S., and Paik,S.S. (2009). Cancer stem cell markers CD133 and CD24 correlate with invasiveness and differentiation in colorectal adenocarcinoma. World J. Gastroenterol. 15, Chu,P., Clanton,D.J., Snipas,T.S., Lee,J., Mitchell,E., Nguyen,M.L., Hare,E., and Peach,R.J. (2009). Characterization of a subpopulation of colon cancer cells with stem cell-like properties. Int. J. Cancer 124, Clement,V., Sanchez,P., de Tribolet,N., Radovanovic,I., and Altaba,A. (2007). HEDGEHOG- GLI1 signaling regulates human glioma growth, cancer stem cell self-renewal, and tumorigenicity. Curr. Biol. 17, Corti,S., Nizzardo,M., Nardini,M., Donadoni,C., Locatelli,F., Papadimitriou,D., Salani,S., Del Bo,R., Ghezzi,S., Strazzer,S., Bresolin,N., and Comi,G.P. (2007). Isolation and characterization

59 45 of murine neural stem/progenitor cells based on Prominin-1 expression. Exp. Neurol. 205, Coskun,V., Wu,H., Blanchi,B., Tsao,S., Kim,K., Zhao,J., Biancotti,J.C., Hutnick,L., Krueger,R.C., Jr., Fan,G., de Vellis,J., and Sun,Y.E. (2008). CD133+ neural stem cells in the ependyma of mammalian postnatal forebrain. Proc. Natl. Acad. Sci. U. S. A 105, Dalerba,P., Dylla,S.J., Park,I.K., Liu,R., Wang,X., Cho,R.W., Hoey,T., Gurney,A., Huang,E.H., Simeone,D.M., Shelton,A.A., Parmiani,G., Castelli,C., and Clarke,M.F. (2007). Phenotypic characterization of human colorectal cancer stem cells. Proc. Natl. Acad. Sci. U. S. A 104, Dick,J.E. (2003). Breast cancer stem cells revealed. Proc. Natl. Acad. Sci. U. S. A 100, Diehn,M., Cho,R.W., Lobo,N.A., Kalisky,T., Dorie,M.J., Kulp,A.N., Qian,D., Lam,J.S., Ailles,L.E., Wong,M., Joshua,B., Kaplan,M.J., Wapnir,I., Dirbas,F.M., Somlo,G., Garberoglio,C., Paz,B., Shen,J., Lau,S.K., Quake,S.R., Brown,J.M., Weissman,I.L., and Clarke,M.F. (2009). Association of reactive oxygen species levels and radioresistance in cancer stem cells. Nature 458, Dierks,C., Beigi,R., Guo,G.R., Zirlik,K., Stegert,M.R., Manley,P., Trussell,C., Schmitt- Graeff,A., Landwerlin,K., Veelken,H., and Warmuth,M. (2008). Expansion of Bcr-Abl-positive leukemic stem cells is dependent on Hedgehog pathway activation. Cancer Cell 14, Dylla,S.J., Beviglia,L., Park,I.K., Chartier,C., Raval,J., Ngan,L., Pickell,K., Aguilar,J., Lazetic,S., Smith-Berdan,S., Clarke,M.F., Hoey,T., Lewicki,J., and Gurney,A.L. (2008). Colorectal cancer stem cells are enriched in xenogeneic tumors following chemotherapy. PLoS. ONE. 3, e2428. Eramo,A., Lotti,F., Sette,G., Pilozzi,E., Biffoni,M., Di Virgilio,A., Conticello,C., Ruco,L., Peschle,C., and De Maria,R. (2008). Identification and expansion of the tumorigenic lung cancer stem cell population. Cell Death. Differ. 15,

60 46 Fan,X., Khaki,L., Zhu,T.S., Soules,M.E., Talsma,C.E., Gul,N., Koh,C., Zhang,J., Li,Y.M., Maciaczyk,J., Nikkhah,G., Dimeco,F., Piccirillo,S., Vescovi,A.L., and Eberhart,C.G. (2010). NOTCH pathway blockade depletes CD133-positive glioblastoma cells and inhibits growth of tumor neurospheres and xenografts. Stem Cells 28, Fidler,I.J. (2003). The pathogenesis of cancer metastasis: the 'seed and soil' hypothesis revisited. Nat. Rev. Cancer 3, Fidler,I.J. and Poste,G. (2008). The "seed and soil" hypothesis revisited. Lancet Oncol. 9, 808. Fong,S., Debs,R.J., and Desprez,P.Y. (2004). Id genes and proteins as promising targets in cancer therapy. Trends Mol. Med. 10, Ginestier,C., Hur,M.H., Charafe-Jauffret,E., Monville,F., Dutcher,J., Brown,M., Jacquemier,J., Viens,P., Kleer,C.G., Liu,S., Schott,A., Hayes,D., Birnbaum,D., Wicha,M.S., and Dontu,G. (2007). ALDH1 is a marker of normal and malignant human mammary stem cells and a predictor of poor clinical outcome. Cell Stem Cell 1, Gray,M.J., Dallas,N.A., Van Buren,G., Xia,L., Yang,A.D., Somcio,R.J., Gaur,P., Mangala,L.S., Vivas-Mejia,P.E., Fan,F., Sanguino,A.M., Gallick,G.E., Lopez-Berestein,G., Sood,A.K., and Ellis,L.M. (2008). Therapeutic targeting of Id2 reduces growth of human colorectal carcinoma in the murine liver. Oncogene 27, Gupta,G.P. and Massague,J. (2006). Cancer metastasis: building a framework. Cell 127, Gupta,G.P., Perk,J., Acharyya,S., de Candia,P., Mittal,V., Todorova-Manova,K., Gerald,W.L., Brogi,E., Benezra,R., and Massague,J. (2007). ID genes mediate tumor reinitiation during breast cancer lung metastasis. Proc. Natl. Acad. Sci. U. S. A 104, Hanahan,D. and Weinberg,R.A. (2000). The hallmarks of cancer. Cell 100, Haraguchi,N., Ohkuma,M., Sakashita,H., Matsuzaki,S., Tanaka,F., Mimori,K., Kamohara,Y., Inoue,H., and Mori,M. (2008). CD133+CD44+ population efficiently enriches colon cancer initiating cells. Ann. Surg. Oncol. 15,

61 47 Harrison,H., Farnie,G., Howell,S.J., Rock,R.E., Stylianou,S., Brennan,K.R., Bundred,N.J., and Clarke,R.B. (2010). Regulation of breast cancer stem cell activity by signaling through the Notch4 receptor. Cancer Res. 70, Hermann,P.C., Huber,S.L., Herrler,T., Aicher,A., Ellwart,J.W., Guba,M., Bruns,C.J., and Heeschen,C. (2007). Distinct populations of cancer stem cells determine tumor growth and metastatic activity in human pancreatic cancer. Cell Stem Cell 1, Hoey,T., Yen,W.C., Axelrod,F., Basi,J., Donigian,L., Dylla,S., Fitch-Bruhns,M., Lazetic,S., Park,I.K., Sato,A., Satyal,S., Wang,X., Clarke,M.F., Lewicki,J., and Gurney,A. (2009). DLL4 blockade inhibits tumor growth and reduces tumor-initiating cell frequency. Cell Stem Cell 5, Hope,K.J., Jin,L., and Dick,J.E. (2004). Acute myeloid leukemia originates from a hierarchy of leukemic stem cell classes that differ in self-renewal capacity. Nat. Immunol. 5, Horst,D., Kriegl,L., Engel,J., Jung,A., and Kirchner,T. (2009a). CD133 and nuclear beta-catenin: the marker combination to detect high risk cases of low stage colorectal cancer. Eur. J. Cancer 45, Horst,D., Kriegl,L., Engel,J., Kirchner,T., and Jung,A. (2008). CD133 expression is an independent prognostic marker for low survival in colorectal cancer. Br. J. Cancer 99, Horst,D., Kriegl,L., Engel,J., Kirchner,T., and Jung,A. (2009b). Prognostic significance of the cancer stem cell markers CD133, CD44, and CD166 in colorectal cancer. Cancer Invest 27, Horst,D., Scheel,S.K., Liebmann,S., Neumann,J., Maatz,S., Kirchner,T., and Jung,A. (2009c). The cancer stem cell marker CD133 has high prognostic impact but unknown functional relevance for the metastasis of human colon cancer. J. Pathol. 219, Huang,E.H., Hynes,M.J., Zhang,T., Ginestier,C., Dontu,G., Appelman,H., Fields,J.Z., Wicha,M.S., and Boman,B.M. (2009). Aldehyde dehydrogenase 1 is a marker for normal and

62 48 malignant human colonic stem cells (SC) and tracks SC overpopulation during colon tumorigenesis. Cancer Res. 69, Ikushima,H., Todo,T., Ino,Y., Takahashi,M., Miyazawa,K., and Miyazono,K. (2009). Autocrine TGF-beta signaling maintains tumorigenicity of glioma-initiating cells through Sry-related HMG-box factors. Cell Stem Cell 5, Ingham,P.W. (2008). Hedgehog signalling. Curr. Biol. 18, R238-R241. Ingham,P.W. and Placzek,M. (2006). Orchestrating ontogenesis: variations on a theme by sonic hedgehog. Nat. Rev. Genet. 7, Iwama,A., Oguro,H., Negishi,M., Kato,Y., Morita,Y., Tsukui,H., Ema,H., Kamijo,T., Katoh- Fukui,Y., Koseki,H., Van Lohuizen,M., and Nakauchi,H. (2004). Enhanced self-renewal of hematopoietic stem cells mediated by the polycomb gene product Bmi-1. Immunity. 21, Jamieson,C.H., Ailles,L.E., Dylla,S.J., Muijtjens,M., Jones,C., Zehnder,J.L., Gotlib,J., Li,K., Manz,M.G., Keating,A., Sawyers,C.L., and Weissman,I.L. (2004). Granulocyte-macrophage progenitors as candidate leukemic stem cells in blast-crisis CML. N. Engl. J. Med. 351, Jordan,C.T. (2008). Can we finally target the leukemic stem cells? Best. Pract. Res. Clin. Haematol. 21, Jordan,C.T., Guzman,M.L., and Noble,M. (2006). Cancer stem cells. N. Engl. J. Med. 355, Jordan,C.T. and Lemischka,I.R. (1990). Clonal and systemic analysis of long-term hematopoiesis in the mouse. Genes Dev. 4, Korkaya,H., Paulson,A., Charafe-Jauffret,E., Ginestier,C., Brown,M., Dutcher,J., Clouthier,S.G., and Wicha,M.S. (2009). Regulation of mammary stem/progenitor cells by PTEN/Akt/betacatenin signaling. PLoS. Biol. 7, e Labarge,M.A. (2010). The difficulty of therapeutically targeting cancer stem cell niches. Clinical Cancer Research 16.

63 49 Lapidot,T., Sirard,C., Vormoor,J., Murdoch,B., Hoang,T., Caceres-Cortes,J., Minden,M., Patterson,B., Caligiuri,M.A., and Dick,J.E. (1994). A cell initiating human acute myeloid leukemia after transplantation into SCID mice. Nature 367, Lee,J., Son,M.J., Woolard,K., Donin,N.M., Li,A., Cheng,C.H., Kotliarova,S., Kotliarov,Y., Walling,J., Ahn,S., Kim,M., Totonchy,M., Cusack,T., Ene,C., Ma,H., Su,Q., Zenklusen,J.C., Zhang,W., Maric,D., and Fine,H.A. (2008). Epigenetic-mediated dysfunction of the bone morphogenetic protein pathway inhibits differentiation of glioblastoma-initiating cells. Cancer Cell 13, Lessard,J. and Sauvageau,G. (2003). Bmi-1 determines the proliferative capacity of normal and leukaemic stem cells. Nature 423, Li,C., Heidt,D.G., Dalerba,P., Burant,C.F., Zhang,L., Adsay,V., Wicha,M., Clarke,M.F., and Simeone,D.M. (2007). Identification of pancreatic cancer stem cells. Cancer Res. 67, Li,X., Lewis,M.T., Huang,J., Gutierrez,C., Osborne,C.K., Wu,M.F., Hilsenbeck,S.G., Pavlick,A., Zhang,X., Chamness,G.C., Wong,H., Rosen,J., and Chang,J.C. (2008). Intrinsic resistance of tumorigenic breast cancer cells to chemotherapy. J. Natl. Cancer Inst. 100, Li,Z., Bao,S., Wu,Q., Wang,H., Eyler,C., Sathornsumetee,S., Shi,Q., Cao,Y., Lathia,J., McLendon,R.E., Hjelmeland,A.B., and Rich,J.N. (2009). Hypoxia-inducible factors regulate tumorigenic capacity of glioma stem cells. Cancer Cell 15, Liu,S., Dontu,G., Mantle,I.D., Patel,S., Ahn,N.S., Jackson,K.W., Suri,P., and Wicha,M.S. (2006). Hedgehog signaling and Bmi-1 regulate self-renewal of normal and malignant human mammary stem cells. Cancer Res. 66, Luzzi,K.J., MacDonald,I.C., Schmidt,E.E., Kerkvliet,N., Morris,V.L., Chambers,A.F., and Groom,A.C. (1998). Multistep nature of metastatic inefficiency: dormancy of solitary cells after successful extravasation and limited survival of early micrometastases. Am. J. Pathol. 153, Mechant,A. and Matsui,W. (2010). Therapeutic targeting of the Hedgehog signaling pathway in cancer. Clinical Cancer Research 16.

64 50 Morrison,S.J. (2005). Cancer stem cells. Clin. Adv. Hematol. Oncol. 3, Nakamura,T., Fidler,I.J., and Coombes,K.R. (2007). Gene expression profile of metastatic human pancreatic cancer cells depends on the organ microenvironment. Cancer Res. 67, Naor,D., Nedvetzki,S., Golan,I., Melnik,L., and Faitelson,Y. (2002). CD44 in cancer. Crit Rev. Clin. Lab Sci. 39, Naor,D., Sionov,R.V., and Ish-Shalom,D. (1997). CD44: structure, function, and association with the malignant process. Adv. Cancer Res. 71, Naor,D., Wallach-Dayan,S.B., Zahalka,M.A., and Sionov,R.V. (2008). Involvement of CD44, a molecule with a thousand faces, in cancer dissemination. Semin. Cancer Biol. 18, Neelakantan,S., Nasim,S., Guzman,M.L., Jordan,C.T., and Crooks,P.A. (2009). Aminoparthenolides as novel anti-leukemic agents: Discovery of the NF-kappaB inhibitor, DMAPT (LC-1). Bioorg. Med. Chem. Lett. 19, O'Brien,C.A., Kreso,A., and Dick,J.E. (2009). Cancer stem cells in solid tumors: an overview. Semin. Radiat. Oncol. 19, O'Brien,C.A., Pollett,A., Gallinger,S., and Dick,J.E. (2007). A human colon cancer cell capable of initiating tumour growth in immunodeficient mice. Nature 445, Paget,S. (1989). The distribution of secondary growths in cancer of the breast Cancer Metastasis Rev. 8, Pang,R., Law,W.L., Chu,A.C., Poon,J.T., Lam,C.S., Chow,A.K., Ng,L., Cheung,L.W., Lan,X.R., Lan,H.Y., Tan,V.P., Yau,T.C., Poon,R.T., and Wong,B.C. (2010). A subpopulation of CD26+ cancer stem cells with metastatic capacity in human colorectal cancer. Cell Stem Cell 6, Pannuti,A., Foreman,K., Rizzo,P., Osipo,C., Golde,T., Osborne,B., and et al (2010). Targeting cancer stem cells through the notch signaling. Clinical Cancer Research 16.

65 51 Pardal,R., Clarke,M.F., and Morrison,S.J. (2003). Applying the principles of stem-cell biology to cancer. Nat. Rev. Cancer 3, Park,I.K., Morrison,S.J., and Clarke,M.F. (2004). Bmi1, stem cells, and senescence regulation. J. Clin. Invest 113, Pasca,d.M., Sekine,S., Ermilov,A., Ferris,J., Dlugosz,A.A., and Hebrok,M. (2006). Hedgehog/Ras interactions regulate early stages of pancreatic cancer. Genes Dev. 20, Perk,J., Iavarone,A., and Benezra,R. (2005). Id family of helix-loop-helix proteins in cancer. Nat. Rev. Cancer 5, Piccirillo,S.G. and Vescovi,A.L. (2006). Bone morphogenetic proteins regulate tumorigenicity in human glioblastoma stem cells. Ernst. Schering. Found. Symp. Proc Prince,M.E., Sivanandan,R., Kaczorowski,A., Wolf,G.T., Kaplan,M.J., Dalerba,P., Weissman,I.L., Clarke,M.F., and Ailles,L.E. (2007). Identification of a subpopulation of cells with cancer stem cell properties in head and neck squamous cell carcinoma. Proc. Natl. Acad. Sci. U. S. A 104, Raaphorst,F.M. (2003). Self-renewal of hematopoietic and leukemic stem cells: a central role for the Polycomb-group gene Bmi-1. Trends Immunol. 24, Reya,T., Morrison,S.J., Clarke,M.F., and Weissman,I.L. (2001). Stem cells, cancer, and cancer stem cells. Nature 414, Ricci-Vitiani,L., Lombardi,D.G., Pilozzi,E., Biffoni,M., Todaro,M., Peschle,C., and De Maria,R. (2007). Identification and expansion of human colon-cancer-initiating cells. Nature 445, Rizo,A., Olthof,S., Han,L., Vellenga,E., de Haan,G., and Schuringa,J.J. (2009). Repression of BMI1 in normal and leukemic human CD34(+) cells impairs self-renewal and induces apoptosis. Blood 114,

66 52 Schatton,T., Murphy,G.F., Frank,N.Y., Yamaura,K., Waaga-Gasser,A.M., Gasser,M., Zhan,Q., Jordan,S., Duncan,L.M., Weishaupt,C., Fuhlbrigge,R.C., Kupper,T.S., Sayegh,M.H., and Frank,M.H. (2008). Identification of cells initiating human melanomas. Nature 451, Shipitsin,M. and Polyak,K. (2008). The cancer stem cell hypothesis: in search of definitions, markers, and relevance. Lab Invest 88, Singh,S.K., Hawkins,C., Clarke,I.D., Squire,J.A., Bayani,J., Hide,T., Henkelman,R.M., Cusimano,M.D., and Dirks,P.B. (2004). Identification of human brain tumour initiating cells. Nature 432, Staller,P., Sulitkova,J., Lisztwan,J., Moch,H., Oakeley,E.J., and Krek,W. (2003). Chemokine receptor CXCR4 downregulated by von Hippel-Lindau tumour suppressor pvhl. Nature 425, Swarbrick,A., Roy,E., Allen,T., and Bishop,J.M. (2008). Id1 cooperates with oncogenic Ras to induce metastatic mammary carcinoma by subversion of the cellular senescence response. Proc. Natl. Acad. Sci. U. S. A 105, Takahashi-Yanaga,F. and Kahn,M. (2010). Wnt Signaling: Can we safely eradicate cancer stem cells? Clinical Cancer Research 16. Thayer,S.P., di Magliano,M.P., Heiser,P.W., Nielsen,C.M., Roberts,D.J., Lauwers,G.Y., Qi,Y.P., Gysin,S., Fernandez-del Castillo,C., Yajnik,V., Antoniu,B., McMahon,M., Warshaw,A.L., and Hebrok,M. (2003). Hedgehog is an early and late mediator of pancreatic cancer tumorigenesis. Nature 425, Todaro,M., Perez,A.M., Scopelliti,A., Medema,J.P., and Stassi,G. (2008). IL-4-mediated drug resistance in colon cancer stem cells. Cell Cycle 7, Varnat,F., Duquet,A., Malerba,M., Zbinden,M., Mas,C., Gervaz,P., and Altaba,A. (2009). Human colon cancer epithelial cells harbour active HEDGEHOG-GLI signalling that is essential for tumour growth, recurrence, metastasis and stem cell survival and expansion. EMBO Mol. Med. 1,

67 53 Wang,J.C. and Dick,J.E. (2005). Cancer stem cells: lessons from leukemia. Trends Cell Biol. 15, Weichert,W., Knosel,T., Bellach,J., Dietel,M., and Kristiansen,G. (2004). ALCAM/CD166 is overexpressed in colorectal carcinoma and correlates with shortened patient survival. J. Clin. Pathol. 57, Wice,B.M. and Gordon,J.I. (1998). Forced expression of Id-1 in the adult mouse small intestinal epithelium is associated with development of adenomas. J. Biol. Chem. 273, Wu,C. and Alman,B.A. (2008). Side population cells in human cancers. Cancer Lett. Wu,C., Wei,Q., Utomo,V., Nadesan,P., Whetstone,H., Kandel,R., Wunder,J.S., and Alman,B.A. (2007). Side population cells isolated from mesenchymal neoplasms have tumor initiating potential. Cancer Res. 67, Yauch,R.L., Gould,S.E., Scales,S.J., Tang,T., Tian,H., Ahn,C.P., Marshall,D., Fu,L., Januario,T., Kallop,D., Nannini-Pepe,M., Kotkow,K., Marsters,J.C., Rubin,L.L., and de Sauvage,F.J. (2008). A paracrine requirement for hedgehog signalling in cancer. Nature 455, Yilmaz,O.H., Valdez,R., Theisen,B.K., Guo,W., Ferguson,D.O., Wu,H., and Morrison,S.J. (2006). Pten dependence distinguishes haematopoietic stem cells from leukaemia-initiating cells. Nature 441, Zagzag,D., Krishnamachary,B., Yee,H., Okuyama,H., Chiriboga,L., Ali,M.A., Melamed,J., and Semenza,G.L. (2005). Stromal cell-derived factor-1alpha and CXCR4 expression in hemangioblastoma and clear cell-renal cell carcinoma: von Hippel-Lindau loss-of-function induces expression of a ligand and its receptor. Cancer Res. 65, Zhang, M.,Atkinson,R.L., Rosen,J.M. (2010). Selective targeting of radiation-resistant tumorinitiating cells. PNAS. vol.107(8) Zhao,C., Chen,A., Jamieson,C.H., Fereshteh,M., Abrahamsson,A., Blum,J., Kwon,H.Y., Kim,J., Chute,J.P., Rizzieri,D., Munchhof,M., VanArsdale,T., Beachy,P.A., and Reya,T. (2009). Hedgehog signalling is essential for maintenance of cancer stem cells in myeloid leukaemia. Nature 458,

68 54 CHAPTER 2 Identification of a human colon cancer cell capable of initiating tumour growth in immunodeficient mice This chapter is a version of a research article published in Nature 2007 vol445(4) January. The list of authors is as follows: C.A. O Brien, A.Pollett, S.Gallinger, J.Dick Copyright 2007, Blackwell Publishing. Reprinted with permission from the publisher

69 55 Chapter 2 2 Identification of a human colon cancer cell capacble of initiating tumour growth in immunodeficient mice 2.1 Abstract Colon cancer is one of the best understood neoplasms from a genetic perspective(fearon, 1991; Fearon and Jones, 1992; Radtke and Clevers, 2005), yet it remains the second leading cause of cancer related death indicating that some cancer cells are not eradicated by current therapies(nelson et al., 2001; O'Connell et al., 2004). What has yet to be established is whether every colon cancer cell possesses the potential to initiate and sustain tumor growth (known as the stochastic model), or whether the tumour is hierarchically organized where only a subset of cells, the so-called cancer stem cells (CSCs), possess these abilities (the CSC model)(dick, 2003; Wang and Dick, 2005). Here, we demonstrate the existence of colon CSC based on the identification of a human colon cancer-initiating cell (CC-IC) assayed by renal capsule transplantation in immunodeficient NOD/SCID mice. By limit dilution analysis, there was 1 CC- IC in 5.7x10 4 unfractionated tumour cells. Purification experiments established that all CC-IC were CD133 + enabling enrichment by >200 fold. By contrast, the CD133 - cells that comprised the majority of the tumour, were unable to initiate tumour growth. CC-IC within the CD133 + population were able to maintain themselves as well as differentiate and reestablish tumour heterogeneity upon serial transplantation. The identification of colon CSC that are distinct from the bulk tumour cells provides strong support for the hierarchical organization of human colon cancer according to the CSC model. The existence of CSC predicts that for therapeutic strategies to be effective, they must target the CSC.

70 Introduction, Results, and Discussion There is a long history of studying human tumour biology in experimental xenogeneic colon cancer models, typically generated by injecting cell lines or implanting pieces of primary tumours into immunodeficient mice(pocard et al., 1996; Ravi et al., 2004; Golas et al., 2005). However, cell lines do not recapitulate all aspects of primary tumours and a quantitative assay for single cells is required to determine whether CC-IC exist in colon cancer. Therefore, we developed a reliable xenograft model through subrenal capsule implantation of human colon cancer cell suspensions into pre-irradiated immunodeficient NOD/SCID mice. Tumour formation occurred in 17 out of 17 samples tested including: 6 primary colon cancers, 10 liver metastases, and 1 retroperitoneal metastasis (Table 1 and Supplemental Fig. 1a and b). The histology and degree of differentiation of all xenografts resembled the original tumours from which they were derived (Fig. 1). The tumours were positive for cytokeratin-20 (CK-20) and negative for cytokeratin-7 (CK-7), a pattern seen almost exclusively in colonic adenocarcinoma(sack and Roberts, 1997). Xenografts and parent tumours exhibited similar patterns of expression for multiple mucin antigens, and markers highly associated with colon cancers including carcinoembryonic antigen (CEA)(Ishida et al., 2004) and p53(liang et al., 2004). The degree of tumour cell proliferation, as revealed by MIB-1 staining(ishida et al., 2004), was similar in xenografts and parent tumours (Fig. 1). Thus, the xenografts generated in this model recapitulated the phenotypes of the original tumours. To determine whether this xenotransplant system was quantitative and able to detect single CC-IC, we performed limiting dilution experiments. Groups of NOD/SCID mice were transplanted with replicate doses of human colon cancer cells over a range where some doses were unable to initiate tumour growth while others always initiated tumour formation (Table 2).

71 57 The tumour forming capacity and phenotypic appearance was the same for primary xenografts and tumours passaged into secondary and tertiary recipients. Based on the similar behavior of the primary and passaged tumours, it was possible to combine data to calculate the average frequency of CC-IC in these tumours using the maximum likelihood estimation method of limiting dilution assay (LDA) (Porter and Berry, 1964; Wang et al., 1997). We calculated that on average there was 1 CC-IC per 5.7x10 4 (95%CI: 1 per 3.4x10 4 to 1 per 9.3x10 4 ) unfractionated colon cancer cells, although the limited number of recipients employed for any single sample precluded precise estimation of the patient to patient variation. Thus, as previously shown for breast(al Hajj et al., 2003) and brain(singh et al., 2004) cancer, only a small subset of colon cancer cells are able to initiate tumour growth. By combining the quantitative assay with cell fractionation, we were able to test whether human colon cancer adheres to the stochastic model, where every tumour cell has equal tumour initiation potential(reya et al., 2001; Dick, 2003; Wang and Dick, 2005), or the CSC model, where some cell fractions are enriched for CC-IC activity while others are completely devoid of CC-IC(Reya et al., 2001; Dick, 2003; Wang and Dick, 2005).We focused on fractionation based on CD133 expression. Recently, the phylogenetically conserved protein CD133(Handgretinger et al., 2003; Corbeil et al., 2000) was identified as a potential CSC marker in brain(singh et al., 2004) and prostate 19 cancer. CD133 expression ranged from 1.8 to 24.5% in the colon cancer samples described in Table 1 (Fig. 2a). By immunohistochemistry, CD133 was expressed in clusters amid negative cells (Fig. 2b). Normal colon tissue also expressed CD133 but at much lower levels than primary colonic tumours (0.4%-2.1% normal vs. 8.9%-15.9%) (Table 1; Fig. 2a). To determine if CD133 expression enriches for CC-IC, colon cancer cells were separated into CD133 - and CD133 + fractions and injected into NOD/SCID mice. Of 47 mice (dose range: 2x10 3 to 2.5x10 5 ) injected with CD133 - cells, only 1 mouse transplanted with the highest cell

72 58 dose (Table 2) generated a tumour. Since the CD133 - fraction was contaminated with 5 to 15% of cells that expressed only low levels of CD133, we conclude that neither CD133 - nor CD133 low cells possess CC-IC activity. By contrast, tumours were consistently generated following injection of 1x10 3 colon cancer cells expressing the highest levels of CD133 (CD133 + ), and injection of 100 CD133 + cells resulted in tumour growth in one of four mice. Thus, while significantly enriched, not every CD133 + cell represents a CC-IC. In total, 45 of 49 mice injected with CD133 + cells developed tumours (Table 2). All tumours generated from CD133 + cells were phenotypically similar to original tumours (Fig. 1). Moreover, CD133 expression ranged from 1.7 to 22.4% in the xenografts, similar to that seen in the original tumours. The isolation of tumourigenic and non-tumourigenic fractions, based on CD133 expression, provides strong support for the cellular organization of human colon cancer according to the CSC model. Another prediction of the CSC model is that CC-IC should self-renew to generate new CSC as well as differentiate to generate non-tumourigenic progeny. Serial transplantation experiments from 10 primary xenografts demonstrated that only CD133 + and not CD133 - cells were able to initiate tumour growth in serially transplanted secondary and tertiary mice. Since tumours, either primary or passaged, could be infiltrated with non-malignant cell types the high proportion (>98%) of CD133 - cells that co-expressed the human specific epithelial specific antigen (ESA) from both primary and passaged tumours confirmed they were human colon cancer cells and not infiltrating murine cells or non-epithelial human cells that were somehow co-passaged (Supplemental Fig. 2). Moreover, we showed unequivocally that CD133 - cells remained viable and stained positive for ESA under the renal capsule but were unable to regenerate tumour for as long as 15 to 21 weeks post-injection (Supplemental Fig. 3a,b, and c). Furthermore, ESA + cells were also malignant, staining positive for p53, in cases where the parent tumours were p53 + (Supplemental Fig. 3d). Moreover, since these studies were performed on passaged tumours, it is

73 59 highly unlikely the CD133 - cells are pre-malignant cells, rather than malignant cells. Therefore, the CD133 - cells are generated from the CD133 + cells. Thus we can conclude that only CD133 + CC-IC can be serially passaged, forming xenografts that reestablish tumour heterogeneity, generating both CD133 + and CD133 - progeny in a similar ratio to the patient tumour (Fig. 2c). To determine the frequency of CC-IC within the CD133 + subset we carried out an LDA, using the same principles as described for unfractionated tumour cells 14,15. The passaged xenografts recapitulated the phenotype and tumour forming capacity of the parent tumours, enabling us to combine data from passaged and primary cells. The frequency was calculated to average 1 CC-IC in 262 CD133 + colon cancer cells (95%CI: 1 in 129 to 1 in 534) representing a 216 fold enrichment of CC-IC compared to unfractionated colon cancer cells. Interestingly, the estimate of CC-IC frequency when back calculated to take into account the proportion of CD133 + cells within the unfractionated tumour is ~20 fold higher than when unfractionated cell suspensions were assayed. For example, multiplication of the CC-IC frequency (1/262) by the mean level of CD133 expression for all samples (12%) yields an estimate of ~20 CC-IC per unfractionated tumour cells, instead of the 1/57000 measured in the initial LDA. One possible explanation for this finding is that the CD133 - progeny are negatively regulating the growth of the CD133 + CC-IC fraction, thereby requiring a greater overall number of CD133 + cells to give rise to a tumour, a proposition observed for human hematopoietic stem cells(madlambayan et al., 2005). Here, we identified and characterized CC-IC from human colon tumour samples based on their ability to initiate human colon cancer after transplantation into NOD/SCID mice. CC-IC possessed two key criteria that define stem cells: after transplantation at limit dilution, single CC-IC proliferated extensively and differentiated producing tumours that were phenotypically

74 60 similar to the original patient tumours, and as a population, they self-renewed enabling reestablishment of colon cancer in secondary and tertiary recipient mice. CC-IC were almost exclusively CD133 +, while the CD133 - fraction that comprised between 81 to 98% of the tumour mass had no CC-IC activity. Thus colon cancer, like acute myelogenous leukemia (AML)(Lapidot et al., 1994), breast(al Hajj et al., 2003) and brain(singh et al., 2004) cancer is organized as a hierarchy where a small population of CSC sustain the tumour. Since the calculated frequency of CC-IC was, on average, 1 in 262 CD133 + cells, clearly the majority of CD133 + cells are not CC-IC. As described for CD34 expression on AML leukemic stem cells, this result suggests there may be a hierarchy of CC-IC and progenitors(hope et al., 2004). Thus, future studies employing additional cell surface markers in combination with CD133 are necessary to further purify the CC-IC fraction. Finally, clonal tracking studies need to be carried out to establish self renewal at the single cell level and determine whether different subclasses of CC-IC exist(hope et al., 2004). Although we found CD133 + CC-IC in primary and metastatic tumours, most primary colon cancers tested were derived from right sided tumours and may not be representative of all forms of colon cancer. Nevertheless, our findings should stimulate future studies directed towards increasing the range of colon cancer samples tested and addressing whether qualitative or quantitative CC-IC differences have prognostic value. Analysis of CC-IC using molecular genetic techniques should further our understanding of the genetic abnormalities commonly associated with colon cancer, such as microsatellite status. Furthermore, as our understanding of normal colon stem and progenitor cell biology improves it should be possible to gain insight into the cell of origin of colon cancer and the cellular context within which the well characterized sequence of genetic events occurs(al Hajj and Clarke, 2004; Polyak and Hahn, 2006).

75 61 The existence of tumourigenic and non-tumourigenic cells within colon cancers implies that not all the cells within a tumour are able to initiate and sustain neoplastic growth. This concept has important therapeutic implications, and may explain the observation that small numbers of disseminated cancer cells can be detected in the circulation of patients that never develop metastatic disease(reya et al., 2001). The identification of CC-IC provides a powerful tool to develop a better understanding of tumour progression and the metastatic process as the CSC model predicts that the unit of selection in tumour progression would be the CSC itself. Moreover, since CC-IC are the driving force sustaining tumour growth, developing adjuvant therapies directed at specifically eliminating the CC-IC fraction may prove to be a more effective strategy to decrease both local and distant recurrence(dick, 2003; Al Hajj et al., 2004). The model described here will provide the means to further purify and functionally characterize the biological properties of the CC-IC fraction with the goal of developing new therapeutic strategies directed specifically against CC-ICs.

76 Figures Figure 2.1 Xenografts generated from both bulk and CD133+ colon cancer cells resemble the original patient tumour

77 63 Figure 2.1: Xenografts generated from both bulk and CD133 + colon cancer cells resemble the original patient tumour. The parent tumour (Tumour 14) is compared with xenografts generated from both primary and secondary passages of the tumour. The initial passage represents a xenograft generated from the injection of 1x10 5 bulk human colon cancer cells. The secondary xenograft was generated from the injection of 500 CD133 + colon cancer cells. The histology of the three tumours, as expressed by H&E staining, shows well to moderately differentiated mucinous adenocarcinomas with intestinal differentiation including numerous goblet cells and intraluminal mucin (5x magnification). The immunohistochemical markers (including: CK-20, CK7, CEA, Muc2, MIB-1, and p53) reveal comparable staining patterns in both the bulk and CD133 + xenografts, as compared to the parent tumour (20x magnification).

78 Figure 2.2 Expression of CD133 in tumour and normal colonic tissue 64

79 65 Figure 2.2: Expression of CD133 in tumour and normal colonic tissue. a, Flow cytometric contour plots demonstrating the variable expression of CD133 between normal colon tissue (Normal 7) and colon cancer tissue from the same patient (Tumour 7) and a representative third tumour (Tumour 5) showing higher CD133 expression. The upper and lower panels depict istotype controls and CD133 staining, respectively. b, Immunohistochemical staining for CD133: Normal 7, and Tumours 7 and 5 (20x magnification). c, Flow cytometric contour plots demonstrate preservation of CD133 expression through primary, secondary, and tertiary passages as exemplified by tumours 17 and 11. Tumours from each passage were stained with CD133 PE and an isotype specific antibody (IgG1 PE) and the proportion of CD133 + cells is shown in each quadrant. CD133 expression varied between tumour 11 and 17.Each tumour maintained consistent levels of CD133 expression through three separate tumour passages. The first passage represented injection of bulk colon cancer cells, however, subsequent passages involved the isolation and injection of CD133 + colon cancer cells.

80 66 Table 2.1 Patient and tumour characteristics Table 1: Patient and tumor characteristics Tumor Tumor Tumor Xenograft CD133 + in CD133 + in Patient No. Age/Sex site stage differentiation formation tumor(%) normal(%) P1 73/M right colon IIIB moderate yes P2 80/F liver IV moderate yes 5.2 P3 73/M liver IV poor yes 14.7 P4 70/F liver IV moderate yes 1.8 P5 74/F paracolic IV poor yes 24.5 P6 64/M liver IV moderate yes 6.3 P7 75/F right colon I well yes P8 34/F right colon IIIC well yes P9 63/F liver IV well to moderate yes 12.1 P10 66/F right colon IIIC well to moderate yes P11 74/F liver IV moderate yes 19 P12 65/F right colon IIIC poor yes P13 84/F sigmoid I moderate yes P14 58/M liver IV moderate yes 17.6 P15 53/M liver IV moderate yes 18.2 P16 75/F liver IV not stated yes 10.4 P17 56/M liver IV moderate yes 3.2 Table 2. 1 NOD/SCID mice were injected with bulk colon cancer cells from each tumour. All seventeen tumours generated xenografts in NOD/SCID mice. Ten of the tumours were carried through to tertiary passages in mice (Tumours 7-15, and 17). Of the remaining tumours all except three (Tumours 1-2, and 16) were carried through to secondary mice. CD133 expression was determined by flow cytometry for each tumour prior to implantation. For the 6 primary colonic tumours, CD133 expression was also determined for normal colonic tissue.

81 67 Table 2.2 Limiting dilution analysis of the human colon cancer-initiating cell Table 2: Limiting dilution analysis of the human colon cancer initiating cell Colon No.and ID of No. of 1 ary mice No. of 2 ndary & 3 ary Total No. of mice cancer samples with tumor/ total mice with tumor/ mice with tumors/ cell source Cell dose tested (P) No. injected total No. injected total No. injected (%) BULK 1 x , /4 0/4 0/8 (0) 2.5 x ,6,10-14,17 1/6 0/2 1/8 (12.5) 5 x ,15 2/5 2/5 4/10 (40) 7.5 x ,5,7-9, /8 -- 4/8 (50) 1 x ,7-13 6/6 4/4 10/10 (100) 1 x / /17 (100) 2 x , /8 -- 8/8 (100) CD x ,8,14, /4 1/4 (25) 5 x ,6,11,13,14,17 1/1 4/5 5/6 (83.33) 1 x ,10,12,17 1/1 6/6 7/7 (100) 5 x ,15,17 1/1 7/7 8/8 (100) 1 x ,7-14,,17 1/1 9/9 10/10 (100) 2 x ,9-11, /9 9/9 (100) CD133-5 x ,9,11,12, /5 0/5 (0) 1 x ,10,12,14,15,17 0/1 0/5 0/6 (0) 2 x ,9,10,13 0/1 0/6 0/6 (0) 5 x ,7-9,11,12,14,15 0/1 0/7 0/8 (0) 1 x ,8,10,12-15,17 0/1 0/7 0/8 (0) 2.5 x ,7-9,11,13-15, /9 1/9 (11.1) Table 2.2: NOD/SCID mice were transplanted with: BULK (n=61), CD133 + (n=49), and CD133 - (n=47) human colon cancer cells. All doses are displayed with the exception of 2x10 3 for CD133 + (n=5) and CD133 - (n=5), where tumour formation rates were 100% and 0%, respectively. Mice were sacrificed at 621 weeks post-injection. Mice were considered negative if no tumour tissue was identified. The tumour identification numbers (p) are depicted as subscripts

82 68 in the No. of samples tested column. Only doses that resulted in a mix of positive and negative mice were utilized to calculate the limiting dilution experiments.

83 69 Figure Supplementary 2.1 Unfractionated and CD133+ colon cancer cells initiate tumours when transplanted under the renal capsule of NOD/SCID mice Figure Supplementary 2.1: Unfractionated and CD133+ colon cancer cells initiate tumours when transplanted under the renal capsule of NOD/SCID mice. a, A representative tumour 15 weeks post- injection, the tumour was generated from the injection of 5x104 colon cancer cells (Tumour 3). b, A representative tumour at the time of sacrifice 21 weeks post-injection. The tumour was generated from 5x103 CD133+ colon cancer cells (Tumour 15). Arrows point to tumour tissue.

84 CD133/1 PE IgG1 PE 70 Figure Supplementary 2.2: Flow cytometric analysis of CD133 and ESA expression Tumor 9 Tumor IgG1 FITC ESA FITC Figure Supplementary 2.2: Flow cytometric analysis of CD133 and ESA expression. Tumours 9 and 12, at passage 4, were stained with isotype specific antibodies (upper panels), or with antibodies specific for CD133 PE and ESA FITC (lower panels).

85 71 Figure Supplementary 2.3: Histological examination following the injection of CD133- colon cancer cells Figure Supplementary 2.3: Histological examination following injection of CD133- colon cancer cells. a, Analysis of the site of injection at 20 weeks of 2x104 CD133- cells revealed ESA+ cells (2x magnification) (right colon cancer (Tumour 10); second passage). b, Analysis of the site at 17 weeks of injection of 1x105 CD133- cells revealed ESA+ cells (2x magnification) (liver metastasis (Tumour 17)); third passage).c, At higher magnification (20x) the ESA+ cells were consistent in appearance with colon cancer cells (Tumour 17).d, ESA+ cells were stained for p53 in the case where the parent tumour (Tumour 17) was p53+.

86 CD133/2 PE CD133/2 PE CD133/2 PE 72 Figure Supplementary 2.4: Analysis of purity following magnetic bead separation b CD133 + fraction after separation a Tumor sample before separation unstained c CD133 - fraction after separation 1 unstained unstained Figure Supplementary 2.4: Analysis of purity following magnetic bead separation. a, A representative example of CD133 expression of an acutely dissociated tumour, obtained at the time of surgical resection (Tumour 14). b, Representative analysis of the purity of CD133+ cells (93.5%) following magnetic bead separation. c, Representative analysis of purity of CD133- cells (88.5%) following MACS purification.

87 Materials and Methods Tumour cell preparation Colon cancer specimens were obtained from consenting patients, as approved by the Research Ethics Board at The University Health Network (UHN) (Toronto, Ontario). Tumour tissue was mechanically dissociated and incubated with Collagenase Type IV (Sigma) followed by magnetic bead separation to remove dead cells (Miltenyi Biotec). Following the generation of a single cell suspension, tumour cells were resuspended in 100μl of 1x binding buffer (Miltenyi- Biotec) and incubated for 15 minutes at room temperature with 100μl Dead Cell Removal Microbeads (Miltenyi-Biotec). Magnetic separation was carried out with a mini-macs Separator (Miltenyi-Biotec) using a positive selection MS column (Miltenyi-Biotec). The process was repeated to ensure increased purity. At the end of the separation, viable tumour cells were resuspended in HamF12:DMEM media. Viability was confirmed by utilizing trypan blue dye exclusion. Each tumour specimen yielded approximately 2-3 million cells Magnetic cell sorting and flow cytometry Human colon cancer cells were magnetically labeled and separated by double passage using a CD133 Cell Isolation Kit (Miltenyi Biotec). Prior to separation, samples were assessed using a FACSCalibur flow cytometer (BD Biosciences), mouse IgG1 conjugated to phycoerythrin (PE) or fluorescein isothyiocyanate (FITC) were used as isotype controls (BD Biosciences). CD133 expression was assessed using anti-cd133/1 PE (Miltenyi-Biotec). To confirm the cells as human colon cancer, samples were tested using anti-epithelial specific antigen (ESA) FITC (Biomeda)(Supplemental Fig.2). At least 10,000 events were acquired for each sample and all cells positive for propidium iodide were gated out. Following magnetic bead separation samples were assessed by flow cytometry for purity.

88 Transplantation of human colon cancer cells into NOD/SCID mice NOD/LtSz-scid/scid (NOD/SCID) mice were bred and maintained under defined conditions at the Ontario Cancer Institute (OCI) under conditions approved by the Animal Care Committee of the OCI. Colon cancer cells were suspended in a 1:1 mixture of media and matrigel (BD Biosciences) and injected under the renal capsule of mice (8 weeks of age) that were sublethally irradiated (350cGy). Mice were anaesthetized and cells injected under the renal capsule. All mice were sacrificed when the tumour measured 1cm, at the first sign of suffering, or between weeks post-transplantation Limiting dilution analysis and serial tranplantations Unsorted human colon cancer cells were injected at doses ranging from 1x10 4 to 2x10 6 cells. Fourteen out of the seventeen patient samples were injected over a range of at least four cell doses with one mouse per dose resulting in an aggregate of between 8-17 mice per group for each injection dose. For purification experiments, colon cancer cell suspensions were magnetically separated on the basis of CD133 expression prior to injection. Purified cells were injected at doses ranging from 100 to 2x10 4 and 2x10 3 to 2.5x10 5 per mouse, for CD133 + and CD133 - cells, respectively. There were between 4-10 mice per group for each injection dose. Serial transplantation experiments were completed by taking tumours from primary mice injected with bulk primary cells and separating the cells based on CD133 expression. The CD133 + and CD133 - cells were re-implanted into secondary mice. Tumours generated by the injection of CD133 + cells were subsequently injected into tertiary mice Purity following magnetic bead separation After passage through the mini-macs Separator, CD133 + and CD133 - cells were once again evaluated by flow cytometry, using anti-cd133/2 (293C3) PE to determine the efficiency of the

89 75 separation. Purities ranged from 78 to 95% for the CD133 + cells (median 89%) (Supplemental Fig. 3a and b) and from 85.3 to 99% for the CD133 - cells (median 92%) (Supplemental Fig. 3a and c) Xenograft histopathology Tissue sections taken from xenografts were fixed in 4% paraformaldehyde for 48 hours and then transferred to 70% ethanol. Once fixed, the tissue samples were processed using Tissue-Tek VIP and embedded in paraffin. The blocks were sectioned at 4μm thickness on a Microm HM 200 cryotome, and stained with haematoxylin-and-eosin (H&E) as per standard histopathological technique Immunohistochemistry Paraffin-embedded, 4 m formalin fixed sections were dewaxed in xylene and rehydrated with distilled water. The anti-human CD133 antibody was incubated at room temperature overnight (Miltenyi-Biotec). All remaining antibodies were incubated for 1 hour at room temperature; Muc-1, Muc-2, Muc-5AC, Muc-6, CEA, ESA, CEACAM-1 (Vector), and MIB-1, p53, CK7, CK20 (Dako). All sections, with the exception of CEA and ESA, were treated with heat-induced epitope retrieval technique using a 10mM citrate buffer at ph 6.0. Antibody staining was followed by 30 minute incubations with a biotinylated linking reagent (ID labs) and horseradich peroxidase-conjugated ultrastreptavidin labeling reagent (ID labs). Development and counterstaining were done with NovaRed solution (Vector) and Mayer s hematoxylin (ID labs), respectively.

90 Statistical analysis For limiting dilution assays, the frequency of the cancer initiating cell was calculated using the Porter and Berry method, as previously described 14,15,27. The calculation utilized doses at which only a fraction of the injected mice demonstrated tumour growth.

91 References Al Hajj,M., Becker,M.W., Wicha,M., Weissman,I., and Clarke,M.F. (2004). Therapeutic implications of cancer stem cells. Curr. Opin. Genet. Dev. 14, Al Hajj,M. and Clarke,M.F. (2004). Self-renewal and solid tumour stem cells. Oncogene 23, Al Hajj,M., Wicha,M.S., Benito-Hernandez,A., Morrison,S.J., and Clarke,M.F. (2003). Prospective identification of tumorigenic breast cancer cells. Proc. Natl. Acad. Sci. U. S. A 100, Corbeil,D., Roper,K., Hellwig,A., Tavian,M., Miraglia,S., Watt,S.M., Simmons,P.J., Peault,B., Buck,D.W., and Huttner,W.B. (2000). The human AC133 hematopoietic stem cell antigen is also expressed in epithelial cells and targeted to plasma membrane protrusions. J. Biol. Chem. 275, Dick,J.E. (2003). Breast cancer stem cells revealed. Proc. Natl. Acad. Sci. U. S. A 100, Fearon,E.R. (1991). A genetic basis for the multi-step pathway of colorectal tumorigenesis. [Review]. International Symposium of the Princess Takamatsu Cancer Research Fund 22, Fearon,E.R. and Jones,P.A. (1992). Progressing toward a molecular description of colorectal cancer development. [Review]. Faseb J 6, Golas,J.M., Lucas,J., Etienne,C., Golas,J., Discafani,C., Sridharan,L., Boghaert,E., Arndt,K., Ye,F., Boschelli,D.H., Li,F., Titsch,C., Huselton,C., Chaudhary,I., and Boschelli,F. (2005). SKI- 606, a Src/Abl inhibitor with in vivo activity in colon tumor xenograft models. Cancer Res. 65, Handgretinger,R., Gordon,P.R., Leimig,T., Chen,X., Buhring,H.J., Niethammer,D., and Kuci,S. (2003). Biology and plasticity of CD133+ hematopoietic stem cells. Ann. N. Y. Acad. Sci. 996,

92 78 Hope,K.J., Jin,L., and Dick,J.E. (2004). Acute myeloid leukemia originates from a hierarchy of leukemic stem cell classes that differ in self-renewal capacity. Nat. Immunol. 5, Ishida,H., Miwa,H., Tatsuta,M., Masutani,S., Imamura,H., Shimizu,J., Ezumi,K., Kato,H., Kawasaki,T., Furukawa,H., and Kawakami,H. (2004). Ki-67 and CEA expression as prognostic markers in Dukes' C colorectal cancer. Cancer Lett. 207, Lapidot,T., Sirard,C., Vormoor,J., Murdoch,B., Hoang,T., Caceres-Cortes,J., Minden,M., Patterson,B., Caligiuri,M.A., and Dick,J.E. (1994). A cell initiating human acute myeloid leukemia after transplantation into SCID mice. Nature 367, Liang,J.T., Huang,K.C., Jeng,Y.M., Lee,P.H., Lai,H.S., and Hsu,H.C. (2004). Microvessel density, cyclo-oxygenase 2 expression, K-ras mutation and p53 overexpression in colonic cancer. Br. J. Surg. 91, Madlambayan,G.J., Rogers,I., Kirouac,D.C., Yamanaka,N., Mazurier,F., Doedens,M., Casper,R.F., Dick,J.E., and Zandstra,P.W. (2005). Dynamic changes in cellular and microenvironmental composition can be controlled to elicit in vitro human hematopoietic stem cell expansion. Exp. Hematol. 33, Nelson,H., Petrelli,N., Carlin,A., Couture,J., Fleshman,J., Guillem,J., Miedema,B., Ota,D., and Sargent,D. (2001). Guidelines 2000 for colon and rectal cancer surgery. J. Natl. Cancer Inst. 93, O'Connell,J.B., Maggard,M.A., and Ko,C.Y. (2004). Colon cancer survival rates with the new American Joint Committee on Cancer sixth edition staging. J. Natl. Cancer Inst. 96, Pocard,M., Tsukui,H., Salmon,R.J., Dutrillaux,B., and Poupon,M.F. (1996). Efficiency of orthotopic xenograft models for human colon cancers. In Vivo 10, Polyak,K. and Hahn,W.C. (2006). Roots and stems: stem cells in cancer. Nat. Med. 12, Porter,E.H. and Berry,R.J. (1964). The efficient design of transplantable tumour assays. Br J Cancer 17,

93 79 Radtke,F. and Clevers,H. (2005). Self-renewal and cancer of the gut: two sides of a coin. Science 307, Ravi,R., Jain,A.J., Schulick,R.D., Pham,V., Prouser,T.S., Allen,H., Mayer,E.G., Yu,H., Pardoll,D.M., Ashkenazi,A., and Bedi,A. (2004). Elimination of hepatic metastases of colon cancer cells via p53-independent cross-talk between irinotecan and Apo2 ligand/trail. Cancer Res. 64, Reya,T., Morrison,S.J., Clarke,M.F., and Weissman,I.L. (2001). Stem cells, cancer, and cancer stem cells. Nature 414, Sack,M.J. and Roberts,S.A. (1997). Cytokeratins 20 and 7 in the differential diagnosis of metastatic carcinoma in cytologic specimens. Diagn. Cytopathol. 16, Singh,S.K., Hawkins,C., Clarke,I.D., Squire,J.A., Bayani,J., Hide,T., Henkelman,R.M., Cusimano,M.D., and Dirks,P.B. (2004). Identification of human brain tumour initiating cells. Nature 432, Wang,J.C. and Dick,J.E. (2005). Cancer stem cells: lessons from leukemia. Trends Cell Biol. 15, Wang,J.C., Doedens,M., and Dick,J.E. (1997). Primitive human hematopoietic cells are enriched in cord blood compared with adult bone marrow or mobilized peripheral blood as measured by the quantitative in vivo SCID-repopulating cell assay. Blood 89,

94 80 CHAPTER 3 Id1 and Id3 regulate the self-renewal capacity of human colon cancer-initiating cells through p21 This chapter is a version of a research article submitted to Cancer Cell with the following list of authors: C.A.O Brien, A.Kreso, P.Ryan, L.Gibson, S.Gallinger, J.Dick. A.Kreso contributed equally to this research article

95 81 Chapter 3 3 Id1 and Id3 regulate the self-renewal capacity of human colon cancer-initiating cells through p Addendum Following the submission of this manuscript to Cancer Cell it came to our attention that the six cell lines were actually sub-clones of one original cell line. Interestingly, despite the fact that all the lines originated from the same patient the lines were distinct with respect to their mutational status for PI3kinase. Three of the lines possessed activating PI3kinase mutations while the remaining lines were PI3kinase wild type. Furthermore, each line possessed distinct CC-IC frequencies as demonstrated by in vivo limiting dilution assays. These results indicate that both functionally and genetically distinct CC-IC clones can exist within one colon cancer. Recent work in the C-IC field has focused largely on phenotypic characterization of C-ICs; however, these results clearly demonstrate that there is significant diversity even between CC- ICs originating from the same tumour. This is an important finding because it emphasizes that much remains to be understood about the complexity of C-ICs both from the genetic and functional perspective. Since the submission of the initial manuscript we have gone on to knock down the inhibitors of differentiation 1 and 3 (Id1 and Id3) in 3 additional genetically distinct colon cancer samples derived from patient specimens. Furthermore we have overexpressed p21 in combination with knockdown of Id1 and Id3. The serial transplantation results clearly demonstrate that the biological results we observed in the initial colon cancer specimen can be repeated using distinct

96 82 colon cancer samples. We are currently completing the serial transplantation experiments and plan on resubmitting this manuscript within the next 2 months.

97 Abstract There is increasing evidence that many tumours, including colon cancer, are organized as cellular hierarchies sustained by a relatively rare population of cells that possess stem cell properties termed, cancer-initiating cells (C-ICs). Although the capacity for self-renewal is a hallmark of all stem cells, little is known of the genes that control self-renewal in C-ICs. Here we establish that Id1 and Id3 govern the self-renewal capacity of colon C-ICs (CC-ICs). The combined knock down of Id1 and Id3 decreases the expression level of cell cycle inhibitor p21 and increases entry into cell cycle. Genetic rescue experiments found that the regulation of p21 by Id1/3 is a central mechanism by which CC-IC self-renewal is maintained. Additionally, silencing of Id1/3 increases the sensitivity of CC-ICs to the chemotherapeutic agent oxaliplatin, linking self-renewal with chemoresistance and opening a new avenue for future therapeutic strategies to prevent recurrence of drug resistant disease.

98 Introduction There is increasing experimental evidence from cell fractionation experiments that many, but perhaps not all, tumours are organized as a cellular hierarchy sustained by a so called cancerinitiating cell (C-IC) or cancer stem cell (CSC)(Al Hajj and Clarke, 2004; Dick, 2008; O'Brien et al., 2009). Several attributes distinguish C-ICs from the remaining cells of a tumour, including: initiation of cancer growth in xenotransplantation assays, restoration of the tumour hierarchy by generating non-c-ics, and the capacity for long-term self-renewal(dick, 2003). Self-renewal is the cardinal property of all stem cells, whether normal or malignant, and has been best characterized in the context of normal development. Normal stem cells require mechanisms that not only preserve their undifferentiated state, but also enable the production of more differentiated daughter cells that have lost the capacity to self-renew. It is becoming evident that the acquisition of dysregulated self-renewal mechanisms represents a key step in the generation of C-ICs(Morrison and Kimble, 2006; Shenghui et al., 2009). The strongest evidence for selfrenewal in cancer has come from serial xenotransplantation assays and clonal tracking studies carried out in hematological malignancies(bonnet and Dick, 1997; Hope et al., 2004). Experimental data is now accumulating that several solid tumours, including breast(al Hajj et al., 2003), brain(singh et al., 2004), and colon(o'brien et al., 2007; Ricci-Vitiani et al., 2007), adhere to the hierarchical model and contain self-renewing C-ICs. The focus of our study is on colon C-ICs (CC-ICs), which typically represent a small subset of the total colon cancer cell population and can be prospectively isolated based on the expression of specific cell surface (i.e. CD133(O'Brien et al., 2007; Ricci-Vitiani et al., 2007), CD44(Dalerba et al., 2007), CD166(Dalerba et al., 2007)) or functional (aldehyde dehydrogenase-1(dylla et al., 2008; Huang et al., 2009)) markers. Evidence is also emerging that the C-ICs from several cancers possess

99 85 properties that make them resistant to chemotherapy and radiation, including expression of drug efflux pumps, altered DNA damage response, or cellular quiescence(bao et al., 2006; Todaro et al., 2008; Hermann et al., 2007; Schatton et al., 2008; Viale et al., 2009). Collectively, little is known of the molecular regulation of self-renewal in solid tumour C-ICs and whether intrinsic stem cell properties such as self-renewal are directly linked to survival mechanisms and therapeutic resistance. As a first step to uncover specific factors that govern self-renewal in CC-ICs, we identified genes that have been implicated in the self-renewal properties in somatic or embryonic stem cells, narrowed the list to those that are dysregulated in cancer and finally to those implicated in human colon cancer specifically. One family of proteins that satisfies these criteria are the inhibitor of DNA binding proteins (Ids); a family of homologous helix-loop-helix (HLH) transcriptional regulatory factors(gray et al., 2008a) (Id1 to Id4) with recognized roles in development, senescence, differentiation, angiogenesis, and migration(fong et al., 2004). The ability of Id proteins to drive self-renewal is well established in embryonic stem cells, where the up-regulation of Ids by bone morphogenic protein 4 (BMP4) is required to maintain self-renewal and pluripotency(hollnagel et al., 1999; Ruzinova and Benezra, 2003). Recent studies found that over expression of Id1 in murine cortical neural stem cells increased their self-renewal capacity, illustrating the important effect of Ids in governing normal neural stem biology(nam and Benezra, 2009). Evidence suggesting the importance of Ids in cancer comes from studies that have shown increased expression in a variety of solid tumours, including pancreatic(kleeff et al., 1998), cervical(schindl et al., 2001), ovarian(schindl et al., 2003) prostate(ouyang et al., 2002), breast(lin et al., 2000; Fong et al., 2003), and colon(meteoglu et al., 2008; Gray et al., 2008b); Id up-regulation correlates with both poor prognosis and chemoresistance(cheung et al., 2004; Hu et al., 2009; Li et al., 2007). Furthermore, studies from Gupta et al using murine models of

100 86 breast cancer demonstrated a role for Id1/3 in the initiation of metastases, a process that may be closely related to the concept of C-ICs(Gupta et al., 2007). Considering the overall importance of Ids in cancer biology and their role in embryonic and somatic stem cell self-renewal, they represent prime candidates to evaluate in CC-IC self-renewal. In some systems one of the Id proteins, Id1, functions to maintain self-renewal through effects on the cell cycle inhibitor, p21/cip1/waf1 (p21)(ciarrocchi et al., 2007; Jankovic et al., 2007). Additionally, several reports have linked p21 to self-renewal of leukemic and normal hematopoietic stem cells(cheng et al., 2000; Viale et al., 2009). While there have been no reports on a role for p21 in CC-IC self-renewal, p21 has a well established role in protecting colon cancer cells against a variety of stress stimuli, including exposure to radiation and chemotherapy(mahyar-roemer and Roemer, 2001; Bene and Chambers, 2009; Gorospe et al., 1996; Sharma et al., 2005; Tian et al., 2000). These studies point to the plausibility that therapy resistance and self-renewal mechanisms may be linked and that the pathways that drive selfrenewal may also function to protect CC-ICs when exposed to environmental stress. Mechanistic studies on self-renewal require large numbers of CC-ICs to carry out functional genomic experiments aimed at identifying the key players in CC-IC self-renewal. However, this has proved difficult because CC-ICs are typically rare in primary human cancers, and culture systems that permit genetic studies as well as the production of large numbers of CC-ICs are not well established. Here we developed a robust culture system that enabled the expansion and genetic manipulation of CC-ICs. The silencing of Id1 and Id3 together significantly impaired tumour initiation and CC-IC self-renewal capacity. This impairment was partially rescued through over expression of p21, a molecule that had previously not been identified as playing a role in CC-IC self-renewal. Our studies also linked the expression of Id1/3 to a decreased

101 87 response to oxaliplatin, a commonly used chemotherapeutic agent in colon cancer(alberts and Wagman, 2008). Thus we have uncovered components of the self-renewal machinery of human CC-ICs and linked them to chemotherapy resistance at the stem cell level, opening the way for therapeutic strategies to manipulate this process for more effective colon cancer treatments. 3.4 Results Enrichment of CC-IC following culture of primary human colon cancer tissue In colon cancer the C-IC fraction is typically small rendering these cells difficult to identify and making molecular studies aimed at manipulating self-renewal programs very challenging(o'brien et al., 2007). As a first step we focused on methods to enrich for tumourinitiating activity, which would in turn facilitate our ability to genetically study self-renewal. To develop such a system, six colon cancer samples were obtained at the time of surgical resection (Fig 1A) and established as sphere cultures in serum free, growth factor enriched media(kreso and O'Brien, 2008). The expression of known CC-IC surface markers, CD133 and CD44, changed significantly following in vitro culturing for all six tumours. CD133 expression was completely lost within three days of culturing, consistent with findings described by other groups(patru et al., 2010). By contrast, CD44 was rapidly expressed on all cultured cells (Fig 1B). Since CD44 expression is often associated with CC-IC activity(dalerba et al., 2007; Huang et al., 2009), in vivo limiting dilution assays (LDA) were used to determine the CC-IC frequency in each sample. The frequencies ranged from 1 in 49 to 1 in 217 colon cancer cells, representing a significant enrichment when compared to samples tested directly from patients, where the CC- ICs are on average 1 in 60,000(O'Brien et al., 2007). The frequencies were maintained upon serial passage indicating that the proportion of CC-IC to bulk cancer cells is stably maintained (Table 1; Suppl. Table 1). The results also clearly show a dissociation of function and phenotype

102 88 with respect to CD44 since the entire population expressed CD44, rendering this an unreliable marker of CC-IC activity in the in vitro setting. Recently a more immune-deficient mouse strain has become available (NSG) that results in a dramatic enhancement in the detection of C-ICs in some tumours, like melanoma(quintana et al., 2008). To determine if the CC-IC frequencies differed between the two strains, parallel in vivo LDAs were carried out to compare NOD/SCID with NSG recipients. The CC-IC frequencies were similar in both recipients indicating that the wide variation observed in melanoma does not apply to colon cancer (Suppl. Table 2). To determine whether CC-ICs derived from the highly enriched primary cultures retained their ability to reestablish a neoplastic cellular hierarchy upon xenotransplantation, they were transplanted into mice and the differentiation properties of the resulting xenografts were assessed by flow cytometry and histology. By contrast to the in vitro setting where all cells were CD44 +, a significant proportion of cells (47 to 72%) comprising the xenograft were CD44 - (Fig 1C). Serial transplantation of CD44 + and CD44 - xenograft derived cells demonstrated that CD44 + cells possessed increased CC-IC capacity as compared to the CD44 - fraction. There was a gradient of CC-IC frequency that ranged from 1 in 142 CD44 + cells to 1 in 128,135 CD44 - cells (Suppl. Table 3). The histological analysis of the xenografts derived from the CC-IC cultures demonstrated strong similarity to the parent tumours from which they were derived (Fig 1D). For example, parent tumours that possessed signet ring cells or mucinous phenotype gave rise to xenografts that continued to exhibit these histological features. Thus, by robust functional and phenotypic criteria we established that the CC-ICs enriched from primary samples retain their capacity to generate a tumour hierarchy.

103 Comparison of the frequency of cells capable of in vitro sphere and colony formation with CC-IC activity The gold standard test for enumerating the frequency of C-ICs and measuring their selfrenewal capacity is in vivo serial transplantation of clonal C-ICs, where clonality is established by LDAs. However, recently many groups have commenced utilizing sphere or colony forming assays as a time-saving surrogate(todaro et al., 2008; Korkaya et al., 2009). To determine whether these in vitro assays constitute valid surrogates for the measurement of C-IC capacity we tested each culture in parallel for colony and sphere replating ability and in vivo C-IC frequency. In all cases, we determined the actual frequency of sphere and colony forming capacity using an LDA approach to ensure that our measurements were derived from single clonogenic cancer cells. When compared to the vivo LDA results, the sphere and colony replating assays estimated 2.89 and 57.8 fold greater initiating cell frequencies, respectively (Table 1). The sphere assay more closely reflected the in vivo C-IC frequencies and therefore we chose to use it as an in vitro surrogate assay; although findings were always confirmed by in vivo LDAs Knock down of Id1/3 expression reduces tumour growth in vivo Given the established role of Id1 and Id3 in maintaining self-renewal in stem cells and driving cancer cell growth we were interested in determining whether they played any role in maintaining the CC-IC fraction. As an initial step we determined the level of expression in CD44 + and CD44 - xenograft derived cell fractions by qpcr. The Id1 mrna expression levels were similar in both the CD44 + and CD44 - fractions (Fig 2A); by contrast, Id3 mrna was consistently expressed at higher levels in the CD44 + fraction (Fig 2B). To determine the functional significance of Id1 and Id3 expression shrna interference technology was used to transduce CC-IC cultures with control, Id1, Id3, or Id1 and 3 combined knock down (KD)

104 90 lentivectors. Western analysis of the lysates from shrna transduced cells demonstrated reductions in Id1 and Id3 of at least 50%, compared with parental and control transduced cells (Fig 2C). Following transduction and selection, the colon cancer cells were injected subcutaneously into NOD/SCID mice. The knock down of Id1 or Id3 expression individually resulted in partial inhibition of colon cancer xenograft growth in the majority of the samples. However, the combined knock down of Id1 and Id3 (Id1/3KD) resulted in a profound decrease in the ability of these cells to form colon cancer xenografts, in all six CC-IC cultures (Fig 2D-H). Since the role of Id1 in vasculogenesis is well established(lyden et al., 1999), we investigated whether the drastic difference in tumour size was in part related to poor vascularization at the site of injection. To test this Id1/3KD and control transduced colon cancer cells were injected into intra-splenic, rather than subcutaneous sites, since the spleen is heavily vascularized. Similar to the subcutaneous model, the injection of Id1/3KD cells intra-splenically resulted in a significant reduction in tumour growth, as well as a decrease in metastatic burden in the liver (Suppl. Fig.1). This result was reproduced in two distinct xenograft models establishing that Id1/3KD was significantly impairing the capacity of CC-IC to either initiate or sustain tumour growth Knock down of Id1/3 effects proliferation and xenograft microvessel formation To elucidate the mechanism of action of Id1/3KD on colon cancer cells, the xenografts generated from each experimental group were examined for microvessel density (MVD), apoptosis, necrosis, and proliferation (MIB-1). There was a significant difference in MVD that was decreased by 50% in the Id1/3KD group as compared to control group (Fig.3A). This result was anticipated because previous studies have recognized the importance of Id1 expression in vasculogenesis both in endothelial and cancer cells(lyden et al., 1999).There was no significant

105 91 difference in the apoptotic rates in the control, Id1KD, Id3KD, and Id1/3KD groups, as determined by immunohistochemical staining with an M30 antibody (Fig 3B). Similarly, the MIB-1 immunostaining and histological quantification of necrosis revealed no significant difference between the experimental and control xenografts with respect to proliferative index or percentage necrosis (Fig 3C-E). In contrast to xenografts, the in vitro proliferative capacity was significantly less in the experimental knock down groups in 4 of 6 CC-IC cultures, as compared to controls (Fig 3F). The effect on in vitro proliferation, albeit statistically significant, was typically less than 2 fold and occurred to the same degree in all three knock down groups. The decreased proliferation observed in vitro could explain the significantly smaller xenografts in the Id1KD and Id3KD groups as compared to control xenografts. However, by itself the effect on proliferation could not account for the profound decrease in xenograft growth observed in the Id1/3KD group compared to Id1KD and Id3KD alone Knock down of Id1/3 impairs the ability of CC-ICs to self-renew Since tumour formation is dependent on the stem cell properties of CC-IC, it was important to determine whether knock down of Id1 and Id3 was affecting the self-renewal capacity of CC-ICs. Initially self-renewal was assessed using surrogate sphere replating experiments, which showed that Id1/3KD reduced the sphere forming capacity of all six samples. Serial replating assays revealed that the mean sphere forming capacity decreased from an average of 1 in 39 (1 in 22 to 1 in 58) in the control cells to 1 in 3139 in the Id1/3KD cells (1 in 769 to 1 in 5555) (Fig 4A,B). The decrease in sphere initiating capacity was maintained upon serial replating experiments. These results suggested that Id1 and Id3 play an essential role in the ability of CC-ICs to self-renew.

106 92 To conclusively establish that Id1 and Id3 were affecting CC-ICs at the clonal level, in vivo serial transplantation studies were carried out at limiting dilution to assess self-renewal capacity of single CC-ICs (Fig 4C). In one sample (1066), the CC-IC frequency was 1 in 49 for the control group versus 1 in for the Id1/3KD cells (Fig. 4D), representing a 200 fold decrease in self-renewal capacity. The in vivo results confirmed our in vitro observations with the self-renewal capacity of the Id1/3KD cells being 50 to 100 fold lower than their control transduced counterparts (Fig 4D). The decreased capacity for self-renewal was only observed in the Id1/3KD group and not when Id1 or Id3 were targeted individually (Fig 4E). Since the in vitro proliferative capacity was equally reduced in all three experimental groups (Id1, Id3, and Id1/3) (Fig 3E), our results suggest that the major effect of combined Id1/3KD is not through impaired proliferation but rather through inhibition of the self-renewal capacity of CC-ICs Knock down of Id1/3 increases sensitivity to oxaliplatin The chemoresistance of C-ICs has emerged as an important cellular property that enables tumours to regrow following initial cytoreductive therapy. Prior literature has identified a potential role for Id1 in maintaining the chemoresistance of cancer cells in a variety of solid tumours, however, no such data exists for colon cancer(cheung et al., 2004). Oxaliplatin is a chemotherapeutic agent commonly used in colon cancer therapy. To study whether Id1/3 KD changes the resistance of colon cancer cells to oxaliplatin, IC50 measurements of oxaliplatin sensitivity were carried out for each CC-IC enriched culture at baseline and following Id1/3KD. In one CC-IC sample (1066), the IC50 for the control cells was 200 ug/ul oxaliplatin versus 50 ug/ul in the Id1/3KD group. Similar results were obtained with the remaining 5 CC-IC enriched cultures (Fig 5A,B). The enhanced chemosensitivity was also reflected in the increased proportion of apoptotic cells that were generated following oxaliplatin exposure at the IC50 dose (61%±12% vs. 27%±11% for the Id1/3KD and control groups, respectively) (Fig 5C,D). Thus,

107 93 Id1/3KD resulted in increased sensitivity to oxaliplatin as demonstrated by decreased proliferative capacity and increased apoptosis. To determine whether oxaliplatin had any effect on self-renewal capacity, sphere replating assays were carried out on the control transduced and Id1/3KD cells in the presence of oxaliplatin. In vitro treatment of the control transduced cells with oxaliplatin (IC50 dose) did not significantly change their sphere replating capacity. By contrast, the Id1/3KD cells displayed a further decrease in sphere replating of 5 to 10 fold following treatment (Fig 5E,F). To determine if this result could be translated to an in vivo model, xenograft studies were carried out with ongoing oxaliplatin treatment to transplanted mice. Despite a highly significant decrease in tumour growth in the Id1/3KD versus control groups, the addition of oxaliplatin led to a further significant decrease in mean tumour weight (Fig 5G). Interestingly, there was no significant difference in the tumour weights in the oxaliplatin-treated versus the vehicle treated control groups. This result is most likely related to the dose of oxaliplatin used and based on our in vitro results, it is plausible that a higher dose would have generated a measurable difference in tumour weights in vivo between the control xenograft groups. Nevertheless, even in the absence of an objective tumour response in the control group, our results demonstrated a further decrease in xenograft formation in the Id1/3KD oxaliplatin treated group, as compared to the vehicle treated Id1/3KD group, indicating that Id1/3KD was potentiating tumour cell killing Knock down Id1/3 decreased self-renewal capacity through downregulation of p21 To further delineate how Id1/3 may affect the self-renewal mechanism, we investigated a known target, p21/cip1/waf1 (p21). Id1 mediated repression of p21 has been identified as a mechanism to preserve self-renewal capacity in endothelial progenitor cells(ciarrocchi et al., 2007). Our interest was to determine whether p21, a cell cycle inhibitor, was playing a similar

108 94 role in CC-ICs. Interestingly, western blot analysis in the CC-IC lines revealed that p21 expression levels in vitro were high in the parental and control transduced cells, but almost undetectable in the Id1/3KD cells (Figure 6a); the exact opposite of the pattern reported in endothelial progenitor cells(ciarrocchi et al., 2007). The silencing of p21 has previously been shown to impair leukemic and hematopoietic stem cell self-renewal(cheng et al., 2000; Viale et al., 2009); however, decreased p21 expression has never been identified in the context of Id1/3KD. To determine whether the decreased expression of p21 observed in the Id1/3KD group was playing a functional role in maintaining self-renewal capacity, a genetic rescue experiment was designed to reintroduce p21 into the Id1/3KD colon cancer cells. Two CC-IC cultures were tested by in vivo xenograft formation; Id1/3KD/p21 over expressing xenografts were significantly larger than those generated by injection of Id1/3KD cells alone (Figure 6B). However, the xenografts generated from the Id1/3KD/p21 over expressing cells remained significantly smaller than their control transduced counterparts. In the two CC-IC samples tested, the reintroduction of p21 resulted in a partial rescue of tumour formation indicating that the decreased expression of p21 in the Id1/3KD cells was functionally important for tumour initiation and maintenance. Furthermore, examination of p21 expression by qpcr in the CD44 + versus CD44 - xenograft derived colon cancer cells demonstrated that p21 expression was significantly greater in the CD44 + cell subset, which is enriched for tumour-initiating capacity (Fig 6C). The lower levels of p21 expression detected in the CD44 - fraction is concordant with our findings that CD44 - cells also possess tumourinitiating capacity but at a lower frequency. These results lend additional support to the notion that p21 is playing a critical role in the maintenance of the self-renewal capacity in the CC-IC subset and that the Id/p21 regulatory axis is important for the maintenance of stem cell function in human CC-ICs.

109 Cellular context impacts the Id1/3-p21 regulatory pathway There have been a number of publications in recent years suggesting serum free culture conditions are superior for maintaining C-ICs(Pollard et al., 2009). There is also a body of literature demonstrating that p21 expression is induced under stress conditions, including serum free culture, raising the question of whether the increased p21 expression levels we observed were a result of the serum free culture conditions(abbas and Dutta, 2009). Interestingly, following the addition of serum to two CC-IC cultures, the expression of p21 decreased to levels observed in the Id1/3KD cells (Fig 7A). To determine if the decreased p21 levels induced by serum exposure would result in a concomitant decline in CC-IC activity and self-renewal capacity, a quantitative analysis of the CC-IC frequency was carried out. The tumour-initiating capacity was significantly less in the cells cultured in serum versus the same cells cultured in serum free media. For example, in sample 328 the frequencies were 1 in 111 in serum free media versus 1 in 15,166 in the presence of serum (Suppl. Table 4). This experiment was repeated with a tumour sample taken directly from the patient and similar results were obtained: serum-free culture increased tumour-initiating capacity compared to cells cultured in the presence of serum (Suppl. Table 4). The xenografts derived from cells grown in the presence or absence of serum displayed similar histology and proportion of CD44 + cells (Fig 7B). Immunohistochemical staining for p21 unexpectedly showed similar p21 expression levels in all the xenografts irrespective of whether the cells were initially grown with or without serum. Thus, p21 was reexpressed in tumours that originally were derived from cells grown in serum where p21 levels were not detectable. Collectively, these data together with the in vivo genetic p21 rescue experiment indicate that the Id1/3-p21 pathway is important in regulating human CC-ICs. The results also indicate that serum-free culture conditions are superior at enriching for C-IC activity. Our data also indicate that CC-ICs grown in the presence of serum are able to maintain a

110 96 hierarchy once reinjected into mice, indicating that these biological properties are not irreversibly lost upon exposure to culturing. 3.5 Discussion In this paper we establish that Id1 and Id3 govern the self-renewal capacity of CC-ICs from primary colon cancer samples and cultures derived from primary samples. Silencing of both genes together resulted in a dramatic loss of tumour-initiating potential that was due to a loss of self-renewal capacity in the CC-IC fraction. Two independent measures of self-renewal were made on the basis of serial replating of tumourspheres and the gold standard assay of selfrenewal based on secondary tumour initiation of single CC-ICs derived from LDA experiments. Second, we found that Id1/3 orchestrate their regulation of CC-IC self-renewal via p21, providing the first linkage of the Id1/3-p21 regulatory axis with maintenance of self-renewal in any solid tumour C-IC. Finally, we show that in the presence of oxaliplatin, a commonly used chemotherapeutic agent, Id1 and Id3 function to protect the self-renewal capacity of CC-ICs. Following the knock down of Id1/3, exposure to oxaliplatin resulted in a further decrease in selfrenewal capacity. Thus, our study connects the self-renewal machinery with chemoresistance of CC-ICs and opens a new avenue for future therapeutic strategies to prevent recurrence of drug resistant disease. As expected from prior studies of Id family proteins in other model systems, we observed effects on proliferation and angiogenesis upon silencing of Id1 and Id3 individually or together. However, these effects could not fully explain the drastic decrease in xenograft growth that occurred exclusively in the Id1/3KD group compared to individual Id knock downs. Instead, clonal quantitative analysis of serially passaged CC-IC established that the major impact of Id1/3 silencing on tumour formation was impaired self-renewal capacity. The fact that Id genes play an

111 97 important role in embryonic and neural stem cell self-renewal lends further support to our conclusion(hollnagel et al., 1999; Nam and Benezra, 2009). A prior study from Gupta et al. reported that silencing Id1/3 in MMTV-Wnt1 mammary cancer cells impaired both mammosphere formation and tumour initiation, as well as formation of lung metastases in a syngeneic transplant model. They did not fully assess self-renewal, attributing the major effect on tumour growth to decreased proliferative capacity following Id1/3 KD(Gupta et al., 2007). However, it seems likely that our findings from the human colon cancer model would extend to their model, as well as other solid tumour models and especially epithelial cancer models, making it critical that assessment of C-IC self-renewal becomes more widely investigated. We initially hypothesized that the mechanism by which Id1/3KD was decreasing selfrenewal capacity in the CC-IC cultures may be similar to the effect observed in endothelial progenitor cells where Id1KD resulted in increased p21 levels and a subsequent decrease in selfrenewal capacity(ciarrocchi et al., 2007). Unexpectedly, we observed the exact opposite; p21 was highly expressed in the parental and control transduced cells, whereas the Id1/3KD cells displayed an almost complete loss of p21 expression in all 6 CC-IC samples tested. Moreover, in xenografts CD44 + cells expressed significantly higher p21 levels as compared to CD44 - cells, closely correlating with the much higher frequency of CC-ICs in CD44 + cells. This result suggested that in the context of CC-ICs, p21 may function to maintain Id1/3 dependent selfrenewal potential. Interestingly, an association between Id1 over expression and high levels of p21 was initially identified by Swarbrick et al.. Studying a mouse mammary carcinoma model they showed that following Id1 over expression tumours continued to proliferate despite high levels of p21(swarbrick et al., 2008). They hypothesized that Id1 must act downstream of p21, and therefore render cells refractory to p21-dependent cell cycle arrest. However, this was not the case in CC-IC cultures because the reintroduction of p21 resulted in a partial rescue of the

112 98 Id1/3KD induced self-renewal defect, providing strong genetic evidence that p21 plays a functional role in maintaining CC-ICs. The rescue was only partial because the p21 expressing xenografts remained significantly smaller than those in the control group. One possible explanation for this observation is that the effect of Id1 and Id3 on proliferation is independent of their combined effects on p21 and self-renewal. Each of the Id1KD, Id3KD, and Id1/3KD groups exhibited an approximate 2 fold decrease in proliferative capacity, yet the self-renewal effect was only seen in the Id1/3KD cells. Alternatively, Id1/3KD may have influenced other downstream pathways in addition to p21 that are involved in the maintenance of self-renewal in the CC-ICs. Although the pathways governing stem cell self-renewal are only starting to be characterized, studies of a variety of normal and neoplastic stem cell systems are pointing to the integrated functioning of multiple genetic and epigenetic components working together to maintain selfrenewal potential(morrison and Spradling, 2008; Shackleton et al., 2009; Shenghui et al., 2009). Along this line, although we see a massive reduction in the frequency of cells that are capable of self-renewal in the Id1/3KD cells, a small proportion still possess this capacity. This may be due to incomplete knock down of Id1 and Id3, but it could also signify the existence of other selfrenewal pathways still operating in these cells. Further work is required to elucidate these other pathways and our system is highly amenable to taking a genome wide screening approach to identify self-renewal genes that remain active in the face of Id1/3 silencing. The notion that a cell cycle inhibitor such as p21 is maintaining self-renewal seems counterintuitive. However, there is strong evidence from the study of normal and leukemic hematopoietic stem cells that p21 is an important regulator of self-renewal. p21 is required for maintaining the self-renewal capacity of normal HSCs and leukemic stem cells (LSCs) in a murine model of acute promyelocytic leukemia. In the absence of p21, the proliferation of HSCs or LSCs increased and they underwent exhaustion, resulting in an inability to maintain the

113 99 malignant clone(cheng et al., 2000; Viale et al., 2009). Similar to the results obtained in leukemia, our results support a role for p21 in the prevention of CC-IC exhaustion through cell cycle restriction. Furthermore, our results are consistent with a number of publications over the past twenty years that have recognized a role for p21 in the protection of cancer cells from stress (Mahyar-Roemer and Roemer, 2001; Bene and Chambers, 2009; Gorospe et al., 1996; Sharma et al., 2005; Tian et al., 2000). Bunz et al were the first to demonstrate that p21-/- colon cancer cells undergo aberrant progression through S and M phases of the cell cycle triggering apoptosis, after treatment with Adriamycin, a DNA damaging agent(bunz et al., 1998). Furthermore, there are numerous reports linking p21 expression with protection of colon cancer cells from apoptosis induced by a wide range of insults including exposure to radiation(tian et al., 2000), chemotherapeutic agents(mahyar-roemer and Roemer, 2001; Bene and Chambers, 2009), and cryotherapy(sharma et al., 2005). Finally, clinical trial data from patients with rectal tumours undergoing neoadjuvant chemoradiation show association between increased p21 expression and the development of resistance resulting in decreased disease specific survival(kuremsky et al., 2009; Rau et al., 2003). Taken together, these observations support the clinical relevance of our findings and extend the functional roles of p21 to include preservation of CC-IC self-renewal. Our studies provide a direct link between the capacity for self-renewal and the chemoresistance of CC-ICs. While existing literature supports a role for Id1 and p21 in maintaining chemoresistance in some solid tumours(cheung et al., 2004; Hu et al., 2009; Li et al., 2007), the effect of Id1/3 KD on colon cancer cells and their response to treatment with oxaliplatin had not been investigated. We found that the dose of oxaliplatin (IC50) that reduced overall cell proliferation did not inhibit the sphere replating capacity of control transduced cells. This finding may in part explain clinical observations related to oxaliplatin treatment. When oxaliplatin is used in the adjuvant setting with 5-Fluorouracil (5-FU), the tumour response rates

114 100 are in the range of 40-50%, whereas the actual survival advantage conferred is on average less than 10%(Alberts and Wagman, 2008; Chau and Cunningham, 2009). This suggests that the response rates may actually be monitoring decreased proliferative capacity of the bulk cancer cells; however, if there are CC-ICs remaining despite oxaliplatin treatment, the drug does not appear to have an effect on their ability to self-renew. In contrast, oxaliplatin treatment of Id1/3KD cells resulted in reduced self-renewal capacity, which warrants further investigation of the linkage between self-renewal and chemoresistance in colon cancer. If this linkage is universally important in other cancers, then understanding the mechanisms that drive C-IC selfrenewal will lead to the development of therapeutic agents that target this essential aspect of tumour maintenance and may also potentiate the efficacy of existing chemotherapeutic agents. Initial experiments indicate that the effect of Id1/3 and p21 on CC-IC self-renewal may inpart be related to the DNA damage response, however, further work is required to confirm this. In conclusion, our study demonstrates the feasibility of utilizing primary human cancer cells to enrich for CC-IC activity thereby providing a powerful tool to carry out genetic approaches to unravel the molecular pathways driving self-renewal. The findings point to the central role that Id1/3 and p21 play in regulating the self-renewal program of human CC-IC and they link self-renewal with the response of CC-ICs to chemotherapy. Collectively, our findings put forth self-renewal pathways as potential targets for the development of effective therapies to eradicate CC-ICs. If these molecules represent potential targets then future studies will be required to determine whether Id1 and Id3 play a central role in maintaining normal intestinal stem cells self-renewal.

115 Figures Figure 3.1 Characterization of the CC-IC cultures

116 102 Figure 3.1: Characterization of the CC-IC cultures. (A) Origin and differentiation status of the parent tumours from which the CC-IC cultures were derived. The corresponding CD133 and CD44 expression is shown as determined by flow cytometry status at the time of surgical resection.(b) Representative flow cytometric profiles of CD44 and CD133 expression for each of the CC-IC cultures. All samples lost CD133 expression and CD44 expression increased to 89-95%. (C) Flow cytometric profiles of xenografts generated by injecting cells from the CC-IC cultures. Expression of CD44 ranged from 28% to 53%. (D) Comparison of the histology of parent tumours and xenografts from each of the CC-IC enriched cultures (original magnification 200x). Table 3.1 Comparison of in vivo xenograft formation with in vitro sphere and colony formation Colon Cancer CC-IC Frequency SFU Frequency CFU Frequency sample in 124 (229-67) 1 in 34 (68-29) 1 in 1.9 ( ) in 111 (76-156) 1 in 57 (94-35) 1 in 2.3 ( ) in 133 (234-75) 1 in 61 (98-42) 1 in 2.6 ( ) in 49 (83-29) 1 in 25 (40-16) 1 in 1.7 ( ) in 52 (141-19) 1 in 19 (63-10) 1 in 1.8 ( ) in 217 ( ) 1 in 44 (70-27) 1 in 1.8 ( ) Table 3.1: Comparison of in vivo xenograft formation with in vitro sphere and colony formation. The in vivo column enumerates the frequency of CC-ICs, as determined by injection of tumour cells into NOD/SCID mice at limiting dilution. The frequency of non-adherent sphere forming units (SFUs) within each sample was determined by in vitro LDAs. The final column depicts the ability of each sample to establish adherent colony forming units (CFUs),.All data is represented as the frequency of xenograft, SFU, or CFU cells; in parentheses is the 95%

117 103 confidence interval (CI). All in vitro experiments for each sample were repeated a minimum of 3 times. Figure 3.2 Id1/3KD reduces tumour growth Figure 3.2: Id1/3KD reduces tumour growth in vivo. (A) qpcr results revealed that there was no significant difference in the level of Id1 expressed in the CD44 + versus CD44 - fraction. (B) qpcr results revealed that Id3 expression was significantly higher in the CD44 + versus CD44 - fraction; error bar represents standard deviation (SD)(C) Western blot analysis of Id1 and Id3 levels in control transduced (PRS) and Id1 and/or Id3 transduced cells (D) Mean tumour weights for sample 1066 following subcutaneous (SQ) injection of control transduced, Id1KD, Id3 KD,

118 104 or Id1/3KD cells. Bars indicate standard deviation (SD) for 32 tumours per group. The mean tumour weight in the Id1/3KD group (76±10mg) was significantly less than the control (1729±176mg), Id1KD (1534±276mg), and Id3KD (958±188mg) groups (p<0.0001). The Id3KD group did exhibit significantly smaller tumours than both the control and Id1KD group (p<0.001). (E) Mean tumour weights for sample 328 after SQ injection of control transduced, Id1KD, Id3KD, and Id1/3KD. Bars indicate SD for 32 tumours per group. The mean tumour weight in the Id1/3KD group (99 ± 12mg) was significantly less than that observed in the control group (1593 ± 209mg) (p<0.0001). The Id1KD (1129±111mg) and Id3KD (1114±146mg) mean tumour weights did not differ significantly from each other, however, they were significantly smaller than the control tumours and significantly larger than the combined Id1/3KD group (p<0.0001). (D and E) Error bars indicate SD for 20 tumours per group. *** p<0.0001, ** p< (F) Sample 257 demonstrated a significantly smaller mean tumour weight in the Id1/3KD group as compared to the control transduced group (p<0.001). (G) Similar results were obtained for sample 240 (p<0.0001).(h) Photographs of excised tumours from mice injected with either the Id1/3KD cells and control transduced cells.

119 PRS Id1KD Id3KD Id1/3KD PRS Id1KD Id3KD Id1/3KD PRS Id1KD Id3KD Id1/3KD PRS Id1KD Id3KD Id1/3KD Normalized activity %MIB1 % necrosis microvessel density # apoptotic cells per HPF 105 Figure 3.3 Effect of Id1/3KD on xenograft histology Figure 3 A PRS microvessel density *** Id1KD Id3KD Id1/3KD B PRS percent apoptosis Id1KD Id3KD Id1/3KD C 100 MIB1 D 100 percent necrosis PRS Id1KD Id3KD Id1/3KD 0 PRS Id1KD Id3KD Id1/3KD E PRS Id1KD Id3KD Id1/3KD MVD MIB-1 M30 F ** *** NS ** ** *** *** * *** NS

120 106 Figure 3.3: Effect of Id1/3KD on xenograft histology. (A) Microvessel density (MVD) was measured by quantifying CD31 staining on paraffin fixed sections from the xenografts in each group. A total of 20 slides were examined for each group, this included 4 slides from 5 different experiments using 5 separate CC-IC samples. There were significantly fewer CD31 + cells in the Id1/3KD tumours, as compared to the control transduced, Id1KD, and Id3KD groups (p<0.0001). The number of CD31 + cells in the Id1/3KD tumours was approximately 50% less than that observed in the control group (11±0.67 versus 5±0.84). (B) No significant difference was observed in the number of apoptotic cells between the control, Id1KD, Id3KD, and Id1/3KD cells, as determined by M30 antibody staining. (C) There was no significant difference between the proliferative index (MIB1 staining) in the 4 groups(n=20 per group).(d)the percent necrosis was similar in the control transduced and Id1/3KD groups(n=20 per group). For Fig 2(A-D) the data is displayed as the mean per high power field (hpf), with 10 hpfs counted per slide. The bars represent the standard error of the mean (SEM).(E) Histological depiction of the staining (magnification 200x). (F) Metabolic activity was used as an indicator of cell proliferation, and measured utilizing an in vitro alamar blue assay. The graph displays the metabolic activity of the control transduced and Id1/3KD cells for 4 samples (368, 1066, 337, 328). There was a decrease in proliferation in the Id1/3KD cells as compared to control transduced cells, however, this decrease was on average <2 fold. In most samples a similar decrease in proliferation was also observed in the Id1KD and Id3KD cells.

121 Figure 3.4 Id1/3KD impairs CC-IC self-renewal 107

122 108 Figure 3.4: Id1/3KD impairs CC-IC self-renewal. (A) Comparison of the sphere replating frequency (SRF) for the control transduced (PRS) versus Id1/3KD cells, as determined by in vitro LDAs. The SRF for the PRS cells was 1 in 22 to 1 in 58, as compared to 1 in 769 to 1 in 5555 in the Id1/3KD cells. All data is represented as the frequency of SFUs and the 95% CI is indicated in paraenthesis (B) The frequency of sphere forming cells per 100 colon cancer cells as determined by in vitro serial passage LDAs for the PRS and Id1/3KD cells for samples: 328, 337, and There was a statistically significant decrease in sphere forming capacity in all samples tested (p<0.0001).(c) Diagramatic representation of a serial transplantation in vivo LDA assay.(d) In vivo serial transplantation assays of cells derived from PRS or Id1/3KD xenografts. Upon serial passage there was a statistically significant difference in the CC-IC frequency in the control transduced versus Id1/3KD colon cancer cells, as determined by in vivo LDA assays. The CC-IC frequency in sample 328 went from 1 in 100 in control cells to 1 in 5000 in the Id1/3KD cells, a 50-fold decrease. For sample 1066 the CC-IC frequency decreased 200-fold following Id1/3KD going from 1in 49 to 1 in (E) Id1KD or Id3KD alone did not result in a decrease in sphere forming capacity, the frequency being similar to PRS cells. However, combined Id1/3KD together resulted in a highly significant decrease in sphere forming capacity (p<0.0001).

123 PRS Id1/3KD PRS Id1/3KD PRS Id1/3KD # of SFU /100 cells (log) tumor weight (mg) percent apoptosis PerCP-Cy5-5-A PerCP-Cy5-5-A Normalized activity Normalized activity 109 Figure 3.5 Id1/3 KD increases the sensitivity of CC-ICs to oxaliplatin Figure 5 A B 337 PRS PRS 337 Id1/3KD 1066 Id1/3KD *** *** oxaliplatin dose (ug/ul) oxaliplatin dose (ug/ul) 200 C *** *** no oxaliplatin with oxaliplatin D 10 5 control- PRS w/ oxaliplatin Id1/3KD w/ oxaliplatin *** PI PRS PRS Id1/3KD Id1/3KD GFP-A GFP-A Annexin V E CC-IC Enriched Cultures Control - PRS SRF in vitro LDA 1 in 45 (31-60) 1 in 58 (36-93) 1 in 25 (16-70) 1 in 22 (10-53) 1 in 44 (27-74) Control - PRS with oxaliplatin SRF in vitro LDA 1 in 31 (24-34) 1 in 33 (20-54) 1 in 19 (8-24) 1 in 25 (10-49) 1 in 41 (26-66) Id1/3KD SRF in vitro LDA 1 in 5555 ( ) 1 in 769 ( ) 1 in 3571 ( ) 1 in 2469 ( ) 1 in 3333 ( ) Id1/3KD with oxaliplatin SRF in vitro LDA 1 in ( ) 1 in 8333 ( ) 1 in ( ) 1 in 5000 ( ) 1 in ( ) F G *** no oxaliplatin with oxaliplatin ** ** ** *** oxaliplatin PRS PRS Id1/3KD Id1/3KD

124 110 Figure 3.5: Id1/3KD increases the sensitivity of CC-ICs to oxaliplatin. (A and B) Using metabolic activity as an indicator of cell proliferation it was determined that the percent of apoptotic cells was significantly higher in the Id1/3KD cells versus control (PRS), following a 48 hour oxaliplatin exposure.(c) Percent apoptotic cells as measured by flow cytometric annexin V analysis was significantly greater in the oxaliplatin treated Id1/3KD (61±12%) versus control (27±10%) (D) Representative flow cytometric examples of percent apoptosis in Id1/3KD vs control cells. (E) Comparison of the SRF for the control versus Id1/3KD cells in the presence and absence of oxaliplatin, as determined by in vitro LDAs. All data is represented as the number of SFUs for each group; in paraenthesis is the 95% CI. Oxaliplatin exposure did not have a significant effect on SRF in control cells. Following Id1/3KD, exposure to oxaliplatin resulted in a further decrease in SRF, 1 in 769 to 5555 and 1 in 8333 to 35714, respectively. (F) Bar graph demonstrating a 5-10 fold decrease in SFUs after exposing Id1/3KD cells to oxaliplatin (IC 50); no significant difference was detected in the PRS cells treated with oxaliplatin. (G) Control or Id1/3KD cells were injected subcutaneously and the mice treated with oxaliplatin. There was no significant difference observed in tumour size between the untreated and treated control groups. The tumours in the untreated Id1/3KD group were significantly smaller than the control groups (treated or untreated). The oxaliplatin treated Id1/3KD group generated tumours that were significantly smaller than their untreated counterparts (p<0.01).

125 # Cells # Cells Gene expression relative to GAPDH % of cells tumour weight (mg) tumour weight (mg) 111 Figure 3.6: Id1/3 KD decreased the self-renewal capacity of CC-IC through p21 down regulation Figure 6 A IB p21 IB GAPDH PRS ID1/3KD PRS ID1/3KD PRS ID1/3KD B *** 328 ** ** *** 368 ** *** 500 ns PRS Id1/3KD Id1/3KD with p21oe 0 PRS Id1/3KD Id1/3KD with p21oe C *** p21 *** *** CD44+ CD44- D 100 Id1/3KD * PRS *** ns G0/G1 G0/G1 S S G2/M G2/M E Id1/3KD PRS control K 100K 150K 200K 250K PE-A K 100K 150K 200K 250K PE-A

126 112 Figure 6: Id1/3KD decreased the self-renewal capacity of CC-IC through p21 down regulation.(a)western blot analysis demonstrating a consistent decrease in p21 expression in the PRS versus Id1/3KD colon cancer cells. GAPDH was utilized as the housekeeping gene (B) Two graphs displaying mean tumour weights for sample 328 and 368. Three groups were included for each cell line: i) control transduced (prs), ii) Id1/3KD, and iii) Id1/3KD with p21 overexpression. The results demonstrated that over expression of p21 could partially rescue the effect of Id1/3KD on xenograft growth. In both the CC-IC samples the Id1/3KD and p21overexpressing tumours were significantly larger than their Id1/3KD counterparts but significantly smaller than the control transduced cells. (C) qpcr results revealed that p21 expression was significantly higher in the colon cancer cell fraction enriched in tumourigenic capacity, CD44 +, versus the CD44 - fraction. (D) Cell cycle distribution was determined by flow cytometry. This graph depicts the percentage of cells in G0/G1, S, and G2/M for both the Id1/3KD and PRS cells. There were a significantly greater number of PRS cells in G0/G1 (73±3.6%) versus Id1/3KD cells (50±4.3%). Whereas 17±1.6% of the Id1/3KD cells were in S phase and only 8±1.03% of the PRS cells were in S phase. There was no significant difference observed in G2/M, however, it did trend towards being greater in the Id1/3KD group. Each value is expressed as the mean ±SD (n=12 per group; four CC-IC samples and 3 repeats per sample) (E) Flow cytometry diagrams of cell cycle distribution in Id1/3KD and control cells.

127 HCT116p21+/+ HCT116p21-/- 328 w/ serum 328 w/o serum 328 w/ serum PRS 328 w/o serum Id1/3KD 113 Figure 3.7: Comparison of p21 levels in cells cultured with and without serum Supplemental Figure 2 A B serum free with serum IB: p21 IB: GAPDH H&E p21 CD44 Figure 3.7: Comparison of p21 levels in cells cultured with and without serum. (A) Western blot analysis demonstrating the absence of p21 expression in the presence of serum, as compared to the higher expression level of p21 observed in the serum-free cultured parental and control transduced cells. The HCT116p21 +/+ and p21 -/- cell lines act as positive and negative controls. (B) Histological depiction of sample 49367G1P2 xenografts derived from serum-free and serum supplemented media. The sections were stained for H and E, p21, and CD44; representative images of each stain are shown (original magnification 50x). No significant difference was detected between the two groups.

128 weight (gm) weight (gm) wieght (gm) 114 Figure Supplementary 3.1 Figure S1 A PRS Id1/3KD Spleen Liver Spleen Liver B C D Liver weight ** PRS Id1/3KD Spleen weight ** PRS Id1/3KD Liver metastasis weight *** PRS Id1/3KD