The Pennsylvania State University. The Graduate School of. Integrative Biosciences

Transcription

1 The Pennsylvania State University The Graduate School of Integrative Biosciences THE REGULATION OF MESENCHYMAL AND CANCER STEM CELL PHENOTYPES IN HEPATOCELLULAR CARCINOMA A Dissertation in Molecular Medicine by Hien T Dang 2012 Hien T Dang Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy December 2012

2 iii The dissertation of Hien T Dang was reviewed and approved* by the following: Kent Vrana Elliot S. Vesell Professor and Chair of Pharmacology Dissertation Advisor Chair of Committee Charles Lang Distinguished Professor of Cellular and Molecular Physiology and Surgery Director, Moelcular Medicine Program Williard Freeman Associate Professor of Pharmacology Harriet Isom Distinguished Professor of Microbiology and Immunology *Signatures are on file in the Graduate School

3 iv ABSTRACT Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide and is the third leading cause of cancer deaths. Unfortunately, patients with HCC present with invasive/metastatic disease that is resistant to traditional cytotoxic treatment and are not eligible for curative therapy such as surgical resection or liver transplantation. Increasing evidence indicates that cancer stem cells (CSCs) are the origin of a subset of HCC. Furthermore, the epithelial-to-mesenchymal transition (EMT) program has been demonstrated to generate cells with CSC characteristics. Resistance to therapy and metastatic disease are two factors that correlate a progenitor-phenotype in HCC with poor survival. The EMT program allows epithelial cells to migrate and intravasate into the blood stream and metastasize to distant organ sites. Accordingly, these disseminated cancer cells require self-renewal capability and other characteristics similar to those exhibited by stem cells in order to survive and promote tumor initiation. These observations lead to the hypothesis that the EMT program is responsible for generating stem-like cells that can spawn macroscopic metastases. Our central hypothesis is that cancer stem cells and the EMT program are linked through central regulators such as the TGF and c-met/hgf signaling axes, and that targeting these mechanisms will improve the design of novel therapeutics for advanced HCC. In murine epithelial HCC cells, TGF induced EMT through Snail1 is indicated by CSC characteristics including increased tumorsphere formation (self-renewal assay) and increased CSC-related genes including CD44, Nanog, and Bmi-1. We demonstrated that Snail1 induced a CSC phenotype through the regulation of the Nanog promoter and that the inhibition of Snail1 leads to the loss of the EMT and CSC phenotypes. In both CSC and EMT-like murine and human HCC cells, CD44 mrna expression was high; thus we hypothesized that CD44 plays a central role in regulating the CSC and EMT phenotypes. CD44 undergoes alternative splicing, generating multiple splice variants, each with a different function.

4 v Recent evidence indicates that CD44s (CD44 standard form) regulates EMT. c-met has been demonstrated to regulate CD44 splicing and c-met + HCC cells demonstrated a mesenchymal phenotype with CSC characteristics including increased CD44s expression and tumorsphere formation capability. To investigate the regulation of CD44s by c-met, we inhibited c-met signaling. The down regulation of c-met decreased CD44s expression and decreased CSC and EMT phenotypes. Further analysis demonstrated that c-met regulates CD44s through p-akt and that subsequent loss of CD44s leads to decreased EMT and CSC characteristics in the human MHCC97-H cell line. To further document that c-met regulates tumor initiation through CD44s, we knocked down CD44s in MHCC97-H cells. The down regulation of CD44s produced loss of both the EMT and CSC phenotype, minimal change in c-met signaling. In vivo data demonstrates that CD44s is required for tumor initiation. Moreover, the overexpression of CD44s in Huh7 (CD44s neg /c-met neg ) increased tumorsphere formation, increased stemness genes including NANOG, BMI-1 and OCT4, and increased mesenchymal phenotypes including loss of E- cadherin and increased fibronectin expression. Our data suggest that the induction of tumor initiation is dependent on CD44s in c-met + HCC. We investigated whether there is a correlation between c-met and CD44s expression in human clinical HCC tissues. In 101 HCC cases, 39% of HCC were c-met + /CD44s +, supporting the hypothesis that there is a relationship between c-met and CD44s. According to our data, more than 39% of HCC are c-met +. c-met is associated with a poor prognosis and has been demonstrated to be resistant to chemotherapy, therefore targeting c- Met may prove beneficial for HCC patients with invasive/metastatic disease. We further investigated whether the inhibition of c-met will inhibit tumor growth in vivo by using a selective c-met inhibitor, PHA Indeed, c-met inhibition led to tumor stasis but was not enough to inhibit tumor growth completely. Further analysis demonstrated that epidermal growth factor receptor (EGFR) /ERBB3 expression increased after c-met inhibition and acted as a compensatory mechanism for tumor survival. To further test our hypothesis, we investigated

5 vi whether combination therapy with PHA and Gefitinib (EGFR inhibitor) could inhibit tumor growth. Interestingly, combination therapy was able to suppress tumor growth compared to c-met or EGFR inhibition alone. In summary, this dissertation work demonstrates the importance of understanding EMT and its relation to CSCs. The link between the EMT and CSC programs play an important role in the resistance, tumor initiation and metastatic disease seen in HCC patients, and through targeting of key regulators of the EMT/CSC phenotype (i.e., c-met), HCC progression may be significantly decreased.

6 vii Table of Contents LIST OF FIGURES... ix LIST OF TABLES... xi ACKNOWLEDGEMENTS... xii ABBREVIATIONS... xiii Chapter 1 Literature Review Introduction Hepatocellular carcinoma Cancer stem cells Epithelial-to-mesenchymal transition Transforming Growth Factor c-met/hgf signaling axis and Epithelial-to-Mesenchymal Transition CD44 Signaling CD44 and cancer stem cells Epithelial-to-mesenchymal transition generates cells with stem-like characteristics Goals and summary of dissertation Chapter 2 Transforming growth factor- induces epithelial-to-transition and cancer stem cell characteristics Abstract Introduction Materials and Methods Results Mesenchymal cells acquire CSC characteristics post-emt Resistance to chemotherapy is linked to cell proliferation TGF -induced EMT results in CSC characteristics Inhibition of Snail1 blocks CSC characteristics TGF regulates Snail and Nanog through Smad signaling Snail1 directly regulates Nanog promoter Inhibition of Snail1 results in decreased tumor growth in vivo Epithelial and mesenchymal differences in human HCC Discussion Chapter 3 Induction of tumor initiation is dependent on CD44s in c-met + hepatocellular carcinoma Abstract Introduction Materials and Methods Results CD44 expression correlates with c-met in human HCC... 61

7 3.4.2 c-met + CD44s + HCC cells have increased mesenchymal and CSC characteristics Down-regulation of c-met in the MHCC97-H cells decreases CD44s expression, mesenchymal and CSC phenotype CD44s regulates mesenchymal and CSC phenotype CD44s modulates cell proliferation c-met activation of mesenchymal and CSC characteristics occurs through CD44s CD44s regulates tumor initiation in vivo Discussion Chapter Abstract Introduction Materials and Methods Results Mesenchymal cells with cancer stem cell characteristics are c-met positive c-met + cells are sensitive to PHA c-met inhibition reduces tumor growth in vivo Upregulation of EGFR signaling occurs after c-met inhibition sirna screen identifies EGFR as potential by-pass survival mechanism after c-met inhibition Inhibition of c-met results in up-regulation of EGFR/ERBB Combination therapy provides additional benefit compared to c-met inhibition alone Combination therapy provides additional benefit to c-met inhibition in vivo Discussion Chapter 5 Overall Discussion and Conclusions Introduction Discussion and Future Directions Clinical Implications Limitations Appendix Supplementary Data References viii

8 ix LIST OF FIGURES Figure 1-1 Hepatocellular carcinoma model... 3 Figure 1-2 Hypothetical models of intratumoral heterogeneity... 7 Figure 1-3 The origins of cancer stem cells... 8 Figure 1-4 Implications for targeting CSCs... 9 Figure 1-5 The process of epithelial-to-mesenchymal transition (EMT) Figure 1-6 Transcriptional regulation of TGF -induced EMT Figure 1-7 CD44 and c-met signaling Figure 1-8 CD44 gene and protein structures Figure 1-9 Outline of Dissertation Figure 2-1 Diagram of how murine P2E and P2M were derived Figure 2-2 Murine epithelial and mesenchymal liver cancer cells Figure 2-3 Mesenchymal cells demonstrate up-regulation of CSCs characteristics Figure 2-4 Resistance to chemotherapy is linked to cell proliferation Figure 2-5 TGF -induced EMT cells with CSC characteristics Figure 2-6 TGF -induced EMT cells acquire stemness gene and protein expression Figure 2-7 Snail1 regulates EMT and CSC characteristics in mesenchymal cells Figure 2-8 TGF regulates Snail1 and Nanog through Smad signaling Figure 2-9 Repression of Snail1 attenuates Nanog promoter activity and tumor proliferation Figure 2-10 Repression of Snail1 attenuates tumor growth Figure 2-11 Human epithelial and mesenchymal liver cancer cells Figure 3-1 CD44s correlates with c-met expression in human HCC samples Figure 3-2 CD44s + HCC cells have mesenchymal and tumor-initiating stem-like characteristics Figure 3-3 c-met regulates mesenchymal and CSC phenotype and CD44s expression... 65

9 x Figure 3-4 c-met regulates CD44s expression through AKT signaling Figure 3-5 CD44s regulates mesenchymal and tumor-initiating stem-like characteristics Figure 3-6 CD44s promotes CSC and mesenchymal characteristics in Huh7 cells Figure 3-7 CD44s modulates p-akt signaling Figure 3-8 CD44s modulates cell proliferation and cell growth Figure 3-9 CD44s is highly expressed in tumorspheres compared with adherent cells Figure 3-10 CD44s regulates tumorsphere formation Figure 3-11 CD44s recovers CSC characteristics after c-met inhibition Figure 3-12 CD44s regulates tumor initiation in vivo Figure 4-1 c-met + MHCC97-L and MHCC97-H cells have mesenchymal and CSC characteristics Figure 4-2 PHA selectively inhibits c-met + HCC cells and downstream PI3K/AKT and MAPK/ERK signaling Figure 4-3 PHA selectively inhibits c-met + HCC tumor growth Figure 4-4 Down regulation of c-met in MHCC97-H cells increases EGFR signaling Figure 4-5 Inhibition of c-met increases EGFR signaling Figure 4-6 Combination of c-met and EGFR inhibitors has an additive effect on the inhibition of MHCC97-H cell growth in vitro Figure 4-7 Combination therapy with c-met and EGFR inhibitors significantly inhibits xenografted tumor growth Figure 5-1 TGF induces EMT and CSC through Snail Figure 5-2 CD44 isoform switching during EMT Figure 5-3 Clincal implications for targeting c-met + HCC cells Figure 5-4 Combination therapy with EGFR and c-met inhibitors inhibits HCC tumor

10 xi LIST OF TABLES Table 4-1 Seven targets validated after sirna screening

11 xii ACKNOWLEDGEMENTS I would like to dedicate this work to my mom, who supported me throughout my life, encouraging only a vision of accomplishments and success. This would have been impossible without your love and your infallible dedication to working hard to make your children s dreams come true. May this work be worthy of your sustained support and make you proud. I would also like to thank my boyfriend Harun Karaman for being supportive, thoughtful and patient. It is you that gave me the strength and motivation to appreciate failure as a means to only do better. Without your support, I would have given up so many times. I would like to thank Dr. C. Bart Rountree, my mentor, who empowered me as a scientist and inspired me as a human being. His patience, support, and advice on the many days of defeat made my graduate school experience a fun and challenging privilege. It is Dr. Rountree s belief and motivation in my potential as an independent scientist that helped shape my optimistic academic career and made me successful today. I am in debt to him. I would like to thank Dr. Kent Vrana for his career advice and moral support during my graduate career. I want to thank him for his help and patience during my last year of graduate school. I could not have completed this dissertation without his mentorship and guidance. Dr.Vrana s devotion to my current and future professional career has made my future in academic science one step closer. I would like to thank Dr. Bill Freeman for his availability, mentorship, and 10 minute (which usually ends up being more than an hour) science and career discussions. These stimulating conversations are what made my long hours at HCAR worthwhile and fun. I would like to acknowledge and thank Dr. Wei Ding for the great technical and intellectual support in my work in the laboratory. I would like to thank him for his continuous scientific guidance, his rigorous science and stimulating conversations. It was a great pleasure and honor to have worked and trained with him. Without Dr. Ding s help and advice, I would not be where I am today. I would like to also thank Dr. Charles Lang for being on my committee, his committed support and belief in me from the beginning. His guidance and moral support during challenging times has helped motivate me to work harder and pursue my dreams. Last but not least, I would like to thank Dr. Harriet Isom for her commitment to my career and for being my committee member. Her scientific and career advice has made my graduate career fun and worthwhile. She has been a great role model and I hope to be successful like her one day. I would also like to thank the people at HCAR and my friends. Without them, my graduate experience would not have been this wonderful.

12 xiii ABBREVIATIONS ABCG AML BMP CD44s CD44v ATP-binding cassette sub-family G Acute myeloid leukemia Bone morphogenic proteins CD44, standard form CD44 variant CD44v3 CD44 variant 3 CD44v6 CD44 variant 6 CSCs CTCs DMEM EMT ERM EGFR HA HBV HCC HCV HGF LCSCs MAPK MDCK Cancer stem cells Circulating tumor cells Dulbecco s Modified Eagle Medium Epithelial-to-mesenchymal transition Ezrin, radixin, moesin Epidermal growth factor receptor Hyalauranon Hepatitis B virus Hepatocellular carcinoma Hepatitis C virus Hepatocyte growth factor Liver cancer stem cells Mitogen-activated protein kinase Madin-Darby canine kidney cells MDR1 Multi-drug resistance 1 MEK Mitogen-activated protein kinase kinase

13 xiv MET P2E P2M PI3K PKC PTEN RLU RTKs SP Mesenchymal-to-epithelial transition Passage 2 epithelial Passage 2 mesenchymal Phosphoinositol-3-kinase Protein kinase C epsilon Phosphatase and Tensin Homolog Renilla luciferase units Tyrosine Kinase Receptors Side populations TGF Transforming growth factor alpha TGF TGF R TRAIL Transforming growth factor beta Transforming growth factor beta receptor TNF-related apoptosis-inducing ligand

14 Chapter 1 Literature Review

15 2 1.1 Introduction Hepatocellular carcinoma (HCC) is one of the most common malignancies worldwide and is the third leading cause of cancer deaths (1). More importantly, the incidence of HCC during the past two decades in the United States has tripled while the 5-year survival rate has remained below 12% (2). Unfortunately, the majority of HCC patients presents with invasive/metastatic disease and are not eligible for curative therapy such as surgical resection or liver transplantation. Increasing evidence indicates that liver cancer stem cells (LCSCs) are the origin of a subset of HCC. Recent research on cancer stem cells (CSCs) implicate that within the whole tumor bulk, only CSCs possess the distinct survival mechanism and stem cell properties crucial for the maintenance and propagation of the tumor. There are two widely accepted functional propertie LCSCs possess that provide an explanation for the failure of cytotoxic chemotherapy; resistance to cytotoxic chemotherapy and radiation and efficient transplantability (3, 4). Current research indicates that the epithelial-to-mesenchymal transition (EMT) program generates cells with CSC characteristics. EMT is the initial process during metastasis that allows single cancer cells to migrate and therefore, invade and subsequently seed tumors in distant organ sites. Thus, it is important to understand the regulation of CSCs and develop novel therapeutic approaches that target specific CSC populations to effectively treat HCC. 1.2 Hepatocellular carcinoma HCC is on the rise in the United States with more than 20,000 new cases per year (5-7). Hepatocarcinogenesis is a complex process that involves chronic hepatocyte injury, inflammation, proliferation and genomic instability (Figure 1-1). Three major etiological agents of HCC are: Hepatitis B (HBV), Hepatitis C (HCV), and chronic alcohol abuse. These risk factors

16 3 disrupt regulatory signals and are a part of a multistep process that leads to alterations in several oncogenic pathways such as inactivation of the tumor suppressor p53 and dysregulation of key signaling pathways including transforming growth factor- (TGF- ), c-met/hepatocyte growth factor (HGF), and Wnt/ -catenin pathways. In addition, genetic and epigenetic alterations can inactivate cell cycle regulation in prolonged chronic inflammatory states, such as cirrhosis and chronic hepatitis, and promote unregulated cellular growth. There are two main cellular origins of HCC; the mature hepatocytes and the multipotent tumor initiating stem-like cells. Hepatocytes are highly differentiated cells that, in response to liver injury, can restore original liver mass by self-replication. During chronic injury caused by HBV or HCV, hepatocytes are constantly undergoing cell death and compensatory cellular proliferation. This excessive cell death requires extensive proliferation to maintain a stable liver mass. Such prolonged self-replication in an inflammatory microenvironment can result in the accumulation of genetic mutations necessary for cancer formation. Figure 1-1 Hepatocellular carcinoma model Hepatocellular carcinoma is a multistage process that takes up to 40 years. Chronic alcohol abuse, HCV or HBV are three major risk factors in HCC. Hepatic stem cells are also ideal targets for carcinogenesis. They are capable of infinite self-renewal and have the ability to differentiate into multiple lineages including bile epithelia, hepatocytes, and cholangiocytes (8). These represent two important characteristic of the liver

17 4 cancer stem cell. HCC typically arises after a prolonged period of chronic hepatitis and subsequent pre-neoplastic and dysplastic changes that are associated with oval cell proliferation. Oval cells are adult liver progenitor cells within the non-parenchymal fraction of the liver that reside near the terminal bile ducts and are capable of differentiating into both hepatocytes and biliary cells. Moreover, when hepatocyte proliferation is blocked or delayed, oval cells proliferate and differentiate as a compensatory response. Thus, mutations in oval cells may be expressed in its differentiated progeny. Further proliferation of these differentiated progenitor cells can be promoted by their microenvironment and allow mutations, all of which are preliminary steps for cancer formation. Recent evidence suggests that the dysregulation of these intricate regulatory pathways in adult oval or progenitor/stem cells could give rise to HCC (8). We and others have defined cancer stem cells as a CD133 expressing murine oval cell population (9). Furthermore, microarray analyses propose that progenitor-based HCC has a more aggressive phenotype with increased expression of CD133, EpCAM and CD44 (10). The treatment for HCC is dependent on the cancer stage by the use of the Barcelona Clinic Liver Cancer staging system (2). Surgical resection is performed for very early-stage HCC with an overall survival rate of 90%. However, early stage HCC is currently difficult to diagnose because it requires a single lesion measuring less than 2cm in diameter with no vascular or distant metastasis. Typically, early stage HCC patients present with a solitary tumor that is less than 5cm in diameter or less than 3 nodule (each nodule has be less than 3cm in diameter) that is not invasive or metastatic is treated according to the severity of the liver dysfunction and the patient s status with respect to coexisting conditions. The most appropriate treatment for early-stage HCC is liver transplantation. However, the waiting process for liver transplantation is fairly long and therefore, the disease may progress while awaiting transplantation (11). The primary cause of transplant failure for HCC is recurrent/metastatic HCC derived from circulating tumor cells (CTCs) (12-14). In addition, over 80% of HCC cases are diagnosed at an advanced stage and are

18 5 no longer operable with surgical treatments and are not transplant candidates (15-17). Late-stage or advanced HCC patients present with mild cancer-related symptoms, vascular invasion or intrahepatic spread. The ideal treatment for late-stage HCC is chemotherapy. However, chemotherapy remains an unsatisfactory approach due to resistance. Currently, the drugs that have been effective in shrinking the tumors are doxorubicin, 5-fluorouracil, and cisplatin. However, these drugs shrink less than 1 of 5 tumors. New developments of systemic agents for the treatment of advanced HCC include two randomized studies with sorafenib and sunitinib as front-line therapy for advanced HCC. Although these studies provide evidence that systemic agents can affect the progression of HCC when cirrhosis is minimized, survival benefits were modest with only an additional 2.5 months (18, 19). While sunitinib, a multi-kinase inhibitor, shows similar effects as sorafenib, it appears to be highly toxic. Brivanib, a multi-kinase inhibitor (shows potent and selective inhibition of vascular endothelial growth factor receptor 2, vascular growth factor receptor, and fibroblast growth factor receptor) manufactured by Bristo-Myers Squibb, is now undergoing clinical trials Phase III to treat patients who are tolerant or resistant to sorafenib treatment. Thus far, brivanib shows promise with an increase in median survival of 10 months in clinical trials Phase II (20). Furthermore, patients acquire resistance after sorafenib treatment and therefore, a second-line therapy is required to treat patients with advanced HCC who progress after sorafenib treatment. Brivanib is currently undergoing clinical trials phase III. Apart from conventional systemic chemotherapy treatment, recent evidence on the presence of CSCs has also provided a novel therapeutic direction for the treatment of HCC. According to the CSC hypothesis, CSCs are resistant to cytotoxic treatments and are responsible for tumor recurrence (8). Furthermore, HCC patients with a CSC gene expression mrna profile have a significantly worse prognosis compared to patients with a more differentiated HCC (hepatocyte-like) (21). This leads to the possibility that CSCs may play an important role in

19 promoting an aggressive HCC phenotype and that by targeting CSCs, one could improve overall survival Cancer stem cells Phenotypic and functional heterogeneity is a defining feature of many solid tumors. Several factors contribute to intratumor heterogeneity including genetics, epigenetics, microenvironment and genomic instability. For years, tumor initiation and development has been regarded as a multistep process where tumor cell phenotypes are determined by all the aforementioned factors. These cellular phenotypes are not stable as they change over time and contribute to drug resistance and tumor progression. This stochastic model, known as the clonal evolution model, suggests that the all cancerous clonal population(s) possess equal tumorigenic capability (22, 23). The discovery of cancer stem cells (CSCs) in solid tumors has changed our view of tumor progression. Unlike the clonal evolution model, the cancer stem cell model assumes that cancers are maintained and propagated by a subpopulation of cells within the tumor that possess or have acquired the self-renewal capability and other properties of stem cells (24). In this hierarchical model, only the cancer stem cell can drive tumor progression and drug resistance (8, 22, 23, 25). The two models do not have to be mutually exclusive, and their combination is also plausible (Figure 1-2). By definition, CSCs possess stem cell features including resistance to apoptosis and selfrenewal (25-27). Even though the very first experiments supporting the cancer stem cell model were only performed in the 1960s, the cancer stem cell hypothesis had been studied much earlier. In the early 1900s, studies of teratocarcinoma provided evidence for the stem cell origin of cancer by demonstrating the ability of transplantable teratocarcinomas in mice to differentiate into mature benign cells (28, 29). The stem cell origin of leukemia was later established with studies

20 7 that tested the ability of various purified hematopoietic stem cell populations to form leukemia in immunodeficient (NOD-SCID) mice. These studies led to the isolation and characterization of cancer stem cells on the basis of their phenotype (30, 31). In solid tumors, CSCs have now been described in many different malignancies including HCC (32, 33). Figure 1-2 Hypothetical models of intratumoral heterogeneity (A)The cancer stem cell (CSC) model. The CSC hypothesis postulates that there is a hierarchical organization of cells such that only a small subset of neoplastic cells are responsible for sustaining tumorigenesis and establishing the cellular heterogeneity within the primary tumor. Therefore, only the cancer stem cells drive tumorigenesis and drug resistance. (B) The clonal evolution model is a stochastic model that depicts that all tumor cells have the capacity to undego self-renewal, thus they all have the potential to contribute to tumor progression and drug resistance. (Figure modified from Polyak et al, 2007 (34)). One of the earliest studies that demonstrates that CSCs exists in solid tumors was performed in primary human breast cancer by Al-Hajj and colleagues, who showed that CD44 + /CD24 - cell fractions from metastatic primary invasive breast tumors had significantly higher tumorigenic potential when injected into athymic nude mice than CD44 - /CD24 + cell fractions (35). This initial study was rapidly followed by others showing similar findings in a wide variety of solid tumors including brain tumors, prostate, pancreatic, melanoma, and hepatocellular carcinomas (36, 37).

21 8 Some of the controversial issues surrounding the CSC hypothesis relate to the origin of CSCs. To date, the CSC hypothesis describes two distinct types of malignant cells: 1) stem cells that have undergone a malignant transformation to become CSCs that ultimately give rise to the primary tumor and 2) a rare subpopulation of cells within a primary malignancy, which through slower cell cycle and other mechanisms are able to evade traditional chemotherapy, and replenish the larger neoplastic pool due to its ability to self renew (Figure 1-3). These two populations are not mutually exclusive in that a stem cell-derived CSC will likely represent a rare population within the larger tumor. In liver cancer, chemically induced hepatocarcinogenesis studies established the existence of liver stem cells or oval cells and the dysregulation of these cells can promote HCC, supporting the CSC hypothesis (10, 38-43). Figure 1-3 The origins of cancer stem cells Cancer stem cells originate from the stem cell or progenitor cells whose mutations can become turmorigenic or the differentatiated cell that have acquired the self-renewal capacity through mutations. (Figure is modified from Bjerkvig et al 2005 (22)) Current evidence by the Weinberg group suggests that differentiated cells can acquire a CSC-like phenotype, indicating that mature cells can accumulate mutations to acquire CSC-like

22 properties (44). Because CSCs are poorly defined and further research is required to delineate the cellular origins of CSCs, we hypothesize that LCSCs should possess the following six criteria: 1) Maintain bi-potency or multi-potency (biliary and hepatocyte markers) 2) High level of stem-cell associated gene expression 3) Demonstrate resistance to chemotherapy 4) Maintain a lower rate of cell division compared with progeny (bulk tumor) cells 5) Self-renewal 6) Initiate tumor formation in transplant models at relatively low cell numbers. From the clinical perspective, the eradication of the CSC population would provide patients longterm disease-free survival. Resistance to therapy and metastatic disease are two factors that correlate a CSC-phenotype HCC with poor survival (Figure 1-4). The clinical relevance for the treatment of CSCs is that targeted therapy or cytotoxic chemotherapy is only effective in killing the bulk tumor cells and thus, spares the CSC population. Over time, the CSC populations can propagate and reseed new tumors leading to tumor recurrence and burdens (Figure 1-4). 9 Figure 1-4 Implications for targeting CSCs CSCs are quiescent, slow cycling cells that can self-renew and promote tumor heterogeneity. Because of these stemness characteristics, CSCs are resistant to traditional cytotoxic therapy. By targeting CSCs followed by cytotoxic treatment, tumor growth may be inhibited. CSCs are hypothesized to be the source of metastatic lesions, as a tumor-initiating cell and are also called tumor-initiating stem-like cells (45). Although the term tumor-initiating cell is used to describe CSCs and is a characteristic of CSCs, it is not synonymous with CSCs. A tumor

23 10 initiating cell may not have the capability to self renew or exhibit stem cell-like features but can form tumors in low cell dilutions. In order to isolate CSCs, many groups, including ours, have modeled their CSC subpopulation on hematopoietic malignancies and defined their CSC subpopulations through the expression of immuno-phenotypic cell surface markers such as CD133, EpCAM, CD90, CD34, CD45, c-met and CD44 (9, 21, 32, 46-48). Various criteria have been used for defining CSC populations and thus, no single or group of putative CSC markers have been highly expressed among all tumors or cell lines. Furthermore, definitive experiments with serial transplantability of marker positive cells have yet to be demonstrated for some markers. It has also become evident that the CSC phenotype varies between individual patient tumors, thus raising the question of whether this difference in CSC phenotypes will reflect differences in clinical outcomes. It is also difficult to identify cell surface markers that are specifically associated with normal liver stem cells, HCC cells, or CSCs. The challenge lies in defining the markers specific to these cells at varying stages of differentiation, in normal stem cells, in HCC, and in the CSC. Moreover, there are multiple CSC pools within individual tumor types and therefore, because of the complexity and the heterogeneous populations of solid tumors, multiple cell surface markers have been utilized to identify CSCs (49-51). Another issue with utilizing cell surface markers to identify CSCs is the functional roles of these markers in regulating the CSC phenotype, therefore elucidating the functional roles of the cell surface may be important in designing effective therapeutic targets. Alternatively, the use of side populations (SP), which is based on the functional property of CSCs to exclude Hoechst dye via ABCG2-transporters, has been utilized to isolate CSCs (52, 53). The advantage of using SP to identify CSCs is that they overcome the barrier of diverse phenotype markers that can have shortcomings as presented above and replaces it with more direct functional markers (53). The identification and definitive roles played by liver stem/progenitor cells in human hepatocarcinogenesis requires further

24 11 investigation. Once the CSCs are identified, multiple functional assays are utilized to characterize and confirm the CSC population. Cancer stem cells may share the same fundamental properties of self-renewal, ability to differentiate into a diversity of different mature cell types, and actively express stemness gene expression (i.e. NANOG, BMI-1, SOX2, and POU5F1). Therefore, functional assays such as serial transplantation and self-renewal are required to further confirm the CSC subpopulation. It is important to phenotypically characterize the CSC subpopulation s gene expression profile, ability to initiate tumors in low cell dilution (as low as 100 cells) in vivo, self-renewal capability (tumorsphere formation assay), quiescence and resistance to apoptosis. An ideal in vitro assay would be quantitative, highly specific, and capable of measuring only the cells of interest, as well as sufficiently sensitive to measure candidate stem cells when present at low frequency. Several in vitro assays have been used to identify stem cells, including sphere assays, serial colony-forming unit (CFU) assays (replating assays), and label-retention assays (49). Together, these various assays help better understand the CSC function and its role in hepatocarcinogenesis. 1.4 Epithelial-to-mesenchymal transition Epithelial-to-mesenchymal transition (EMT) is a critical developmental process that plays a central role in the formation and differentiation of multiple tissues and organs. Even though EMT was first described by Elisabeth Hay in the 1960s as a required step in the formation of the chick primitive streak, it wasn t until 1982 that EMT was well established as a phenomenon as an important process in embryonic development (54, 55). During EMT, epithelial cells lose cell-cell adhesion and apical-polarity, and acquire mesenchymal features, such as motility, invasiveness, and resistance to apoptosis (Figure 1-5) (56). There are three distinct

25 12 subtypes of EMT: type I - developmental EMT, type II - repair-associated EMT (wound healing, tissue regeneration, and organ fibrosis); and type III-neoplastic EMT (57). Type I - developmental EMT is associated with implantation, organ formation, and embryo development. EMT is highly conserved in metazoans for organ formation, which initially requires the remodeling of a simple epithelium to generate multilayered epithelium. This process involves delamination and invagination of aggregate groups of plastic epithelial cells. Through EMT, epithelial cells can reversibly or irreversibly convert into mesenchymal cells to facilitate the Figure 1-5 The process of epithelial-to-mesenchymal transition (EMT) EMT involves functional transition of polarized epithelial cells into mobile cells that can interact and secrete extracellular matrix components, or mesenchymal cells. The hallmark of EMT includes the loss of E-cadherin, Cytokeratins, and mir200 family. (Figure modified from Kalluri et al 2007 (57)) formation of these layers by gastrulation (56, 58). In type II-repair associated EMT, the program is initiated as a part of a normal repair mechanism to generate fibroblasts or other cell types to reconstruct tissues following inflammation, trauma, or regeneration. However, in contrast to type I EMTs, type II EMTs are associated with inflammation and cease once inflammation is attenuated, as is seen during wound healing and tissue regeneration. In the setting of liver fibrosis, type II EMTs continue to respond due to chronic inflammation, leading eventually to organ

26 13 destruction. Tissue fibrosis is a chronic form of wound healing due to persistent inflammation (57). In cancer biology, type III EMT is one mechanism to explain the invasive and migratory capabilities that epithelial carcinomas acquire during metastasis (34, 58). Metastasis is a multistep process during which cancer cells disseminate from the site of primary tumors and establish secondary tumors in distant organs. There is evidence that EMT gives rise to dissemination of single carcinoma cells from the primary site, thus suggesting that EMT may be the initial step of metastasis. All three types of EMT have parallel molecular mechanisms that either induce or inhibit EMT processes. One of the key hallmarks of EMT is loss of E-cadherin, a cell-adhesion protein that is regulated by multiple transcription factors including Snail, Slug, Zeb1/2, and Twist. These transcription factors act as E-box repressors and block E-cadherin transcription (56). E-cadherin belongs to classical (type I) cadherin family and is characterized by the presence of extracellular cadherin repeats that contribute to the generation and maintenance of adherens junctions (protein complexes that hold cells together) via cell adhesion. The intracellular domain of E-cadherin interacts with -catenin and -catenin to facilitate cell integrity with the actin cytoskeleton. Loss of E-cadherin expression in epithelial cells leads to abrogation of cell-cell contact and increased motility, whilst forced expression of E-cadherin protein in metastatic tumour cell lines is sufficient for reversal of this phenotype (59). Another key hallmark of EMT is the gain of mesenchymal markers such as fibronectin, Vimentin and N-cadherin. These proteins are involved in the interaction of the extracellular matrix to allow cell migration to occur. The mir200 family plays an important role in dysregulating EMT by repressing ZEB1 and ZEB2, both EMT inducers that function as transcription factors that repress E-cadherin expression (60, 61). These noncoding RNAs have been documented to be regulators of EMT and metastasis by repressing mrna translation. Although the loss of E-cadherin is a key hallmark of EMT, other adhesion molecules such as Cytokeratins, and tight junction protein ZO-1 are also

27 14 considered hallmarks of EMT (57). In summary, the hallmarks of EMT are the loss of E-cadherin and gain of mesenchymal markers including Fibronectin, N-cadherin, Vimentin and increased expression of Snail or ZEB family proteins. Multiple extracellular signals can initiate an EMT program including TGF, c-met/hgf, hypoxia, Notch and WNT signaling (34). Members of the TGF have been extensively characterized as inducers of EMT. Furthermore, recent data indicate that TGF -induced EMT generates cells with a stem cell phenotype (44, 62). Notch signaling has been demonstrated to play a role in the induction of EMT through complex signaling that involves multiple receptors and downstream signaling cascades. Notch can induce EMT by activating the nuclear factor- B (NF B) (63). The receptor tyrosine kinase (RTK) c-met can also induce EMT during tumorigenesis and development (58). The EMT is a transient program that allows cancer cells to migrate and intravasate into the blood stream thus is an important initial step for metastasis. However, once the cancer cells extravasate, the EMT program is no longer necessary and thus cancer cells may undergo a mesenchymal-to-epithelial transition (MET) in order to colonize at the new organ site. While MET is still controversial, the MET program is a significant event in regulating embryogenesis (34). The MET process has been demonstrated to occur in pancreatic cells cultured with normal hepatocytes (64). By co-culturing mesenchymal pancreatic cells with normal hepatocytes, E- cadherin expression is up-regulated and the acquisition of epithelial features including cell-cell interactions is observed. These data suggest that cells without a niche of support by stromal cells can induce an EMT, cells surrounded by epithelial cells may revert back to become epithelial-like through the MET program in order to survive.

28 Transforming Growth Factor- Transforming growth factor- TGF signaling regulates a diverse set of cellular processes including cell proliferation, differentiation, apoptosis, and development during embryogenesis (65). The TGF family contains more than 30 structurally related factors that include TGF, nodal, activins, and bone morphogenic proteins (BMPs). These ligands can bind and activate two types of serine-threonine kinase receptors; TGFR I and II. In canonical TGF signaling, upon ligand binding to TGFR I or II, both receptors dimerize to trigger the phosphorylation of Smad proteins and subsequent downstream activation or repression of gene expression to promote many cellular processes. In addition to the Smad-dependent signaling, TGF signaling also activates phosphoinositol-3-kinase (PI3K) and the mitogen-activated protein kinase (MAPK) signaling pathways (66). Depending on multiple factors including the microenvironment, identity or dosage of the ligand, and cell types; TGF can induce or inhibit cell proliferation. In normal, unstressed tissues, TGF levels are sufficient to maintain homeostasis. However, under conditions of tissue injury or cancer, TGF is released in abundance by blood platelets and stromal cells in order to prevent regenerative cell proliferate, inflammation and malignant transformation. TGF is one of the major inducers of EMT (67). TGF induction of EMT was first recognized in vitro. Upon the treatment with TGF, epithelial cells changed from cuboidal to an elongated spindle-like morphology, with loss of epithelial markers such as E-cadherin and enhanced mesenchymal markers including Fibronectin and Vimentin (68). TGF induces EMT by up-regulating Snail1 via the Smad-dependent pathways or by indirectly regulating the ZEB family proteins (Figure 1-6) (69). While repressing epithelial gene expression such as E-cadherin, Snail1 can activate mesenchymal proteins fibronectin and N-cadherin (70). The induction of ZEB proteins is necessary in order to repress E-cadherin expression and promote cell migration.

29 16 Figure 1-6 Transcriptional regulation of TGF -induced EMT Upon TGF stimulation, Smad2 and 3 form a complex with Smad4, which translocates to the nucleus and regulates transcription of target genes with other DNA binding transcription factors. During wound healing, TGF induces EMT to facilitate the recruitment of inflammatory mediators which aid in tissue repair. In cooperation with this process, anti-apoptotic signaling pathways, such as AKT signaling, are triggered while TGF induces EMT and tissue fibrosis allowing mesenchymal cells to survive (71). Because of TGF s role in EMT regulation of antiapoptotic signals, it is possible that the dysregulation of TGF signaling can promote or enhance carcinogenesis. Indeed, the TGF receptor is inactivated in most human gastrointestinal and pancreatic cancers (65). In liver cancer, aberrant TGF signaling leads to HCC (41). In addition, mutations in Smad2 and Smad4 are found in HCC (72). In early stages of carcinogenesis, TGF serves as a tumor suppressor by inhibiting cell growth, and in later stages of disease, tumor cells escape this growth inhibition. As late stage cancer tends to be resistant to TGF -driven

30 17 growth arrest signals and as TGF is a known inducer of EMT, TGF is proposed to be a facilitator of cancer progression during late stage disease (73-75) c-met/hgf signaling axis and Epithelial-to-Mesenchymal Transition The Met tyrosine kinase is a cell surface receptor for hepatocyte growth factor (HGF, also known as scatter factor ). Upon HGF binding, c-met signaling regulates development, cell motility, epithelial-to-mesenchymal transition (EMT), and cell survival (76, 77). c-met/hgf induction of EMT was first described in Madin-Darby canine kidney cells (MDCK), a polarized epithelial cell line, by Michael Stocker and Michael Perryman (78). The authors demonstrated that MDCK cells incubated with cultured fibroblast medium containing HGF signaling components become migratory. Knockout mice for both c-met and HGF fail to complete the developmental process and die in utero (79, 80). In adult livers, EMT is promoted by HGF/c-Met signaling to regulate mitogenesis and fibrosis, and is important in liver regeneration and repair (80, 81). HGF/c-Met activation plays a critical role in tissue homeostasis in several degenerative diseases including liver cirrhosis (82). HGF binding results in c-met autophosphorylation of tyrosines Y1234 and Y1235 within the activation loop of the kinase domain and subsequent phosphorylation of tyrosine Y1349, which acts as a docking site for transducer and adaptor proteins such as Grb2, Gab1 and Stat3. HGF/c-Met activates a number of pathways including mitogen-activated protein kinase (MAPK) and PI3K-AKT cascades. In addition, c-met interacts with membrane proteins, including CD44, which optimizes the Met-triggered RAS-MAPK pathway (81, 83). The interaction of c-met/cd44 can lead to cytoskeletal changes that promote cell migration through interactions with ezrin, radixin, and moesin (ERM) proteins (84). In addition to its important physiological functions, the proto-oncogene Met is a master regulator of metastasis, tumor invasion and angiogenesis (85, 86). In various types of cancers,

31 18 including gastric, and small cell lung cancer, c-met mutation, amplification and over-expression have been associated with a metastatic phenotype (87). Due to the role of c-met in cell migration through the EMT program, cancer cells can hijack c-met/hgf signaling to promote metastasis. Furthermore, aberrantly activated c-met signaling can cause cellular transformation. This inappropriate activation of Met signaling can be induced by genetic modifications, transcriptional regulation, or ligand-dependent autrocrine or paracrine mechanisms (81, 88). Met over-activation leads to a series of signaling cascades including cell growth, proliferation, invasion and evasion of apoptosis. HGF/c-Met signaling protects cells from apoptosis through both PI3-kinase/AKT and pathways (89-91). Further in vivo studies indicate that the over-expression of wild-type Met in hepatocytes is sufficient to cause HCC, which regresses following transgene inactivation (92). In human HCC, over-expression and mutation of the Met gene are associated with intrahepatic metastases and vascular invasion, two of the most important clinical findings that determine disease outcomes (87, 93). Further studies demonstrate that HGF-induced EMT in HCC cell lines are significantly inhibited by the Raf inhibitor Sorafenib and the MEK inhibitor U0126, but not the PI3K inhibitor wortmannin (94, 95). Notably, our cohort of HCC patients demonstrates 52% (31/60) with c-met + disease, which is in line with current published reports (96-99). Notably, c- Met activation or over expression of HGF has been demonstrated to play a key role in drug resistance in non small cell lung carcinoma (100). Besides its functional role in repair, regeneration and tumorigenesis, c-met also plays an important role in cancer stem cells. In response to HGF, normal stem cells in the embryo express c-met and migrate to accomplish their developmental fate (88, 101). Due to these characteristics, it is likely that cancer cells activate the HGF/c-Met signaling axis to switch to an invasive program. c-met has been utilized as a cell surface marker to select for CSCs in multiple solid tumor types (46, 48, ). c-met + cells demonstrate a CSC phenotype and increased tumor initiation capability compared to c-met neg cells. Thus c-met is used as an enrichment marker for

32 19 CSCs and have been demonstrated to functionally regulate stemness in pancreatic, colon, and glioblastoma cancer stem cells (46, 104, 105). However, some studies suggests that c-met alone is not a robust marker for identifying cancer stem cells. Thus, some have utilized c-met and CD44 as co-stem cell surface markers to identify CSCs. CD44 + and c-met + CSCs have been demonstrated in many solid tumors to have an increased tumor initiation, self-renewal and mesenchymal phenotype compared to c- Met+ CD44 neg or c-met neg CD44 neg cells (106). Functionally, c-met and CD44 physically interact and form a complex that is necessary for c-met/ras and c- Met/PI3K signaling cascades (Figure 1-7) (84, 107, 108). Together, the c-met/cd44 complex plays an important role in promoting cell migration and regulating anti-apoptotic signals to promote cell survival. Notably, in this research, we demonstrate that c-met + mesenchymal cells express CD44 and that the inhibition of c-met leads to the down-regulation of CD44 (109). Furthermore, the inhibition CD44 and subsequent loss of c-met leads to decreased tumorsphere formation, suggesting a CSC relationship between CD44 and c-met. 1.5 CD44 Signaling One of the most well studied subpopulations of CSCs is the CD44 + cell. CD44 + cells have been demonstrated in multiple cancer types but have been extensively studied in breast cancer. CD44 is a transmembrane cell adhesion glycoprotein that participates in many cellular processes including the regulation of cellular growth, survival, differentiation, lymphocyte homing, and motility (Figure 1-8) (110). This variety of roles results from multiple CD44 isoforms produced by alternative splicing ( ). CD44s, the standard form of 80kDa because it is does not include any variant exons, is ubiquitously expressed in most epithelial and non-epithelial tissues (114). Most predominantly, CD44 acts as a hyaluronic acid (HA) receptor; where upon HA binding, CD44 undergoes conformational changes that leads to the interaction with the cell

33 20 cytoskeleton via ankyrin and ERM proteins to promote cell migration and invasion (115, 116). Other ligands include fibronectin, osteopontin, and selectin, all of which are involved in cellular trafficking. Figure 1-7 CD44 and c-met signaling c-met, HGF and CD44v6 form a complex that is important for downstream PI3K and MAPK signaling to regulate multiple cell signaling processes. In the present work, we demonstrate that CD44s interacts with c-met in HCC cells. CD44 can undergo ectodomain cleavage which then triggers intramembrane cleavage by the presenilin- -secretase complex. The cleaved CD44 intracellular domain (ICD) can act as a transcription factor that potentiates the transcriptional co-activators CBP and p300 to increase CD44 and other gene transcription, which provide an auto-regulatory mechanism for CD44 (117). CD44 expression elicits highly tumorigenic properties and CD44 + tumors are resistant to cisplatin (118). In colorectal carcinoma, CD44 promotes tumorigenesis through c-met signaling (119). CD44 contributes to drug resistance by associating with multidrug resistance 1 protein (MDR1), allowing MDR1 stabilization. Another CD44 mechanism of drug resistance is CD44 s ability to modulate Hippo signaling to upregulate p53, down regulate of inhibitor of apoptosis

34 21 proteins (IAP) and cleavage of caspase 3 (108, 120). Most recently, in vivo knockdown of CD44 in glioblastoma resulted in the inhibition of cell proliferation, an increase in apoptosis and sensitization of cells to chemotherapy drugs (106). CD44 -/- mice with melanoma have impaired angiogenesis and decreased tumor growth (121). We have demonstrated that c-met neg HCC has no response in vitro or in vivo to c-met tyrosine kinase inhibition. c-met + HCC (MHCC97-L and MHCC97-H) demonstrates a significant inhibition of tumor growth during c-met inhibitor treatment, with loss of c-met phosphorylation and significant reduction in CD44 expression (109). Besides binding to HA, CD44 also acts as a co-receptor, which triggers signal transduction via interaction of the cytoplasmic tail with receptor tyrosine kinases (RTKs) such as c-met, transforming growth factor receptor- I (TGF RI), epidermal growth factor receptor (EGFR), ERBB2 (Her2/Neu) and ERBB4 (107, 110, ). As a co-receptor, CD44 cooperates with RTKs in mediating receptor activation by modulating kinase activity. In several cell lines, coimmunoprecipitation and in vivo studies give evidence of a CD44v6 (CD44 variant 6) and c-met interaction and HGF/c-Met/CD44v6 complex. Notably, HGF-induced c-met activation and signaling depends on the presence of CD44v6 for activation of MAPK pathway, indicating that CD44v6 is essential for c-met activation of the Ras/MAPK signaling (84, 107, 125). More specifically, the cytoplasmic tail of CD44v6 is required to amplify c-met/ras/mapk signaling by interacting with ERM proteins upon HGF presentation (84, 107). Furthermore, HGF/c-Met activation of the MAPK pathway leads to increased CD44 expression, suggestive of a positive CD44/c-Met feedback loop (126). While many studies implicate CD44v6 in amplifying the RAS/MAPK pathway, little is known about the role of CD44s in activation of the MAPK and PI3K pathway. In addition, our data indicate that CD44s, not CD44v6, results in the activation of c-met independent of HGF. CD44 also functions as a co-receptor for the ERBB receptor tyrosine kinase (RTK) family including epidermal growth factor receptor (EGFR), ERBB2, and ERBB4. CD44 can be

35 22 co- immunopreciptated with the ERBB RTK family members in several cell lines and primary cells (123, 124, 127). CD44 interacts with the ERBB family by forming an active signaling complex with ERBB2 that anchors matrix metalloprotease-7 (MMP7) to the membrane, leading Figure 1-8 CD44 gene and protein structures (A-C) CD44 undergoes extensive alternative splicing to generate more than 10 different variants whose functions include cell proliferation, cell motility, and cell cycle. (Figure modified from Zoller et al, 2011 (108))

36 23 to the proteolytic cleavage of heparin-binding epidermal growth factor (HBEGF). Activated HBEGF binds and activates ERBB4 leading to the induction of cell survival by suppressing apoptotic signals (110). CD44v3, CD44 variant 3, has been shown to tightly interact with TGFRI. In breast cancer cells, HA or TGF ligand activates the TGFRI/CD44v3 complex leading to phosphorylation of Smad proteins (Smad2 and Smad3) to promote cell migration (122). Furthermore, the HA- or TGF -mediated activation of the TGFRI/CD44v3 complex can lead to the phosphorylation of CD44v3. The phosphorylation of CD44v3 can enhance the binding to the cytoskeletal protein ankyrin, which in turn, promotes cytoskeletal rearrangement to induce cell migration. Accordingly, the CD44v3/TGFR1 interaction plays an important role in oncogenesis signaling. Together, CD44 proteins can trap and concentrate growth factors that are relevant for cellular growth, increase RTK signaling by promoting phosphorylation or ligand presentation and regulate transcription of cell survival genes. Owing to the involvement of CD44 in the crosstalk with the microenvironment, tyrosine kinase receptors and drug resistance; CD44 have been implicated in cancer stem cells. CD44 has been demonstrated to be required for the maintenance of human acute myeloid leukemia (AML), a heterogeneous clonal disorder marked by accumulation of undifferentiated myeloid blasts which are continuously replenished by rare CSCs (128). In solid tumors, CD44 + cells have a CSC phenotype and increased metastatic capability (129). Even though CD44 has been extensively utilized as a CSC marker in multiple solid tumors including liver cancer, CD44 s functional role in CSC remains to be unknown. Furthermore, because of CD44 s migratory function, it has been implicated in the epithelial-to-mesenchymal program. Collectively, CD44 seems to play an important part in both the EMT and CSC program.

37 CD44 and cancer stem cells CD44 has been extensively utilized as a cell surface marker to identify CSCs in breast cancer (130). In 2003, Al-Hajj and colleagues used various human breast tumor samples to isolate CD44 + /CD24 neg CSCs and xenografted into the mammary glands of NOD/SCID mice. The authors demonstrated that as low as 100 CD44 + /CD24 neg breast cancer cells were enriched for tumorigenic potential (129). Soon after, CD44 + CSCs were discovered in multiple cancers including prostate, colon, glioblastoma, colorectal, and liver cancers (47, 120, ). While increasing evidence indicates that CD44 + cells are CSCs, the functional importance of CD44 has yet to be elucidated. CD44 s functional role in leukemia has been elucidated as a regulator of leukemic stem cell function such as differentiation and homing (128, 135). In solid tumors, the inhibition of CD44 by sirna in colorectal primary cells prevented clonal formation and inhibited tumorigenicity in xenograft models (136). Together these data suggest that CD44 contributes to the activation of the stem cell regulatory genes and can be a target of these genes. However, there are no compelling data that CD44 has a central role in selfrenewal. CD44 does not seem to be a direct master regulator of stemness genes such as Nanog, Oct4 (POU5F1) or Bmi-1(108). Nonetheless, CD44 is involved in many signaling pathways that regulate stem cell characteristics such as the Wnt pathway, indirect regulation of Nanog, TGF and the microenvironment. The Wnt pathway regulates CD44 expression by restricting it to the intestinal crypts in non-transformed tissues. However, in human and mice dysplastic crypts and ademomas, CD44 is over-expressed (137). Furthermore, in TCF4 (transcription factor 4) mutant mice, the loss of CD44 are accompanied by the loss of intestinal stem cells (137). These data suggest that Wnt signaling regulates CD44, which is required for the stem cell program in the intestinal crypts. Another mechanism by which CD44 may play an important role in the

38 25 regulation of CSCs is through the indirect regulation of Nanog. HA-CD44 binding promotes the activation of protein kinase C (PKC ) and this increases Nanog phosphorylation and translocation to the nucleus (138). Nanog, a transcription factor involved in stem cell self-renewal can promote transcriptional regulation of many stem cell genes (139). In this dissertation, we show that CD44 standard form regulates tumor initiation, a CSC characteristic. While the evidence for CD44 s functional role on CSCs is lacking, it is clear that CD44 has an important role in CSCs. 1.6 Epithelial-to-mesenchymal transition generates cells with stem-like characteristics Resistance to therapy and metastatic disease are two factors that correlate a progenitorphenotype HCC with poor survival. CSCs are hypothesized to be the source of metastatic lesions, as a tumor-initiating cell (45). The EMT program allows epithelial cells to migrate and intravasate into the blood stream to further metastasize to distant organ sites. However, how the cells extravasate into distant organ sites in order to spawn macroscopic metastases is controversial. Accordingly, these disseminated cancer cells would require self-renewal capability and other characteristics similar to that exhibited of stem cells in order to survive and promote tumor initiation. These observations lead to the hypothesis that the EMT program can generate cells with stem like characteristics in order to spawn macroscopic metastases. Although this hypothesis remains controversial, recent work establishes a connection between epithelialmesenchymal-transition (EMT) and a CSC-phenotype (44, 62, 140). As additional evidence linking EMT to CSCs, TGF has been shown to regulate Nanog expression, a transcription factor that contributes to self-renewal and cell fate determination in embryonic stem cells (141, 142). In human HCC cell lines, TGF regulates CD133 expression, a marker of CSCs, through induction of epigenetic modifications of the CD133 promoter (67, 143).

39 26 Thus, several studies have demonstrated that TGF drives EMT through Snail1 up-regulation, and other studies have correlated EMT to the acquisition of CSC characteristics. Recent evidence demonstrates that TGF regulates stemness through Nanog expression (141). Furthermore, aberrant TGF signaling in liver progenitor cells results in HCC and metastatic breast CD44 + CSCs express a strong TGF signature (41, 144). Thus, it is likely that TGF links the EMT program to the CSC phenotype. Indeed, many studies have demonstrated that TGF induced EMT generates cells with stem cell properties (11, 44, 62). Most interestingly, epithelial cells treated with TGF undergo EMT and have increased CD44 + cells and stem cell gene expression including increased NANOG, BMI-1 and POU5F1 (Oct4) (44). Together, these data link the EMT and CSC phenotype together as an important means for cell survival and metastasis. In liver cancer, TGF induced EMT increases CD44 expression and generates mesenchymal cells with tumorsphere forming capabilities (62). Recent evidence in glioblastoma demonstrates that the inhibition of TGF pathway decreases CD44 + CSC populations (145). Interestingly, TGF -induction of EMT has been implicated in regulating CD44 alternative splicing, switching from CD44v (variant forms) to CD44s (standard form), suggesting an important functional role for the TGF /CD44 relationship in CSCs (11). Together, there is compelling evidence in supporting an EMT/CSC link that may be important in metastasis. 1.7 Goals and summary of dissertation The overall goal of this research was to better understand the regulation of epithelial-tomesenchymal transition (EMT) and cancer stem cells (CSC) in hepatocellular carcinoma (HCC) (Figure 1-9). Our hypothesis is that the cancer stem cell and epithelial-to-mesenchymal program are linked through central regulators such as the TGF or the c-met/hgf signaling axis, and

40 27 that this mechanism will play an important role in the effective design of novel therapeutics for advanced hepatocellular carcinoma. HCC is the third leading cause of cancer deaths and is usually diagnosed in advanced metastatic states for which there are little or no effective treatments. Poor clinical outcome has been correlated with advanced HCC with a mesenchymal and CSC-like phenotype. Invasive HCC are hard to treat and metastasis occurs frequently. The EMT program has been implicated as an initial step for metastasis. Furthermore, cancer cells that undergo the EMT program have been linked to acquire CSC characteristics. CSCs have been identified in several tumor types including HCC and are characterized by their relative quiescence and capacity to self-renew indefinitely. The EMT and CSC characteristics allow cancer cells to be resistant to therapy and further promote tumor heterogeneity, making cancer treatment difficult. Thus, targeting both the EMT and the CSC program may prove to be therapeutically important in the treatment of HCC. For the first part of our study, we aim to link the TGF -induced EMT program with CSC characteristics by using murine cell lines P2E (epithelial) and P2M (mesenchymal) established in our laboratory (Chapter 2). Accordingly, mesenchymal cells are post EMT and have CSC characteristics such as tumorsphere formation capability and increased stemness gene expression. We show that upon TGF stimulation, epithelial cells underwent EMT through the up-regulation of Snail1 in a Smad-dependent manner. Most interestingly, TGF -induced EMT highlighted the importance of Snail1 in regulating CSC marker Nanog. In summary, we discovered that TGF regulates Snail1 to induce the EMT program and that Snail1 regulates Nanog in order to promote the CSC phenotype. We discovered that TGF -induced EMT increased CD44 expression followed by the acquisition of CSC characteristics. When EMT was blocked by using Snail1 sirna, CD44 expression was decreased significantly. While CD44 has been well established as a CSC marker

41 28 in breast cancer, its functional role in the regulation of the CSC phenotype has not been studied. In Chapter 3, we investigated the regulation of CD44, specifically CD44s (standard isoform), by c-met. In human mesenchymal HCC cells, c-met + HCC is resistant to chemotherapy and associated with advanced/metastatic disease. We highlight the important role of CD44s in regulating CSC and mesenchymal phenotypes and its regulation by c-met. We propose that this mechanism is an important, yet undescribed, oncogenic signal within advanced HCC and that targeting this mechanism will ultimately lead to improved survival. Finally, we investigated the role of c-met as a potential target for HCC therapeutics. Current evidence indicates that approximately 30-40% of HCC are c-met + and that c-met + HCC demonstrate poor clinical outcomes. In Chapter 4, we explored whether inhibition of c-met can inhibit tumor growth in vivo. We demonstrate that the inhibition of c-met leads to tumor stasis and that single RTK treatment is not enough for complete tumor regression. We further investigated the survival mechanism after c-met inhibition and observed an activation of EGFR/ERBB3 signaling by increase expression of TGF. Finally, we showed that combination therapy with both the c-met inhibitor and EGFR inhibitor can inhibit tumor growth. This dissertation work is novel because it demonstrates, for the first time, a regulatory relationship between Snail1 and Nanog in HCC and the complex interaction between CD44s and c-met. We discuss the controversial ideas of the EMT and CSC program and clinical implications of our studies (Chapter 5). Together, this dissertation research help illuminate the importance of understanding the underlying mechanism of the EMT program and CSC-EMT link. Importantly, this understanding of the CSC and EMT link can lead to better novel therapeutics that is necessary to address an unmet need within HCC therapy.

42 29 Figure 1-9 Outline of Dissertation The cancer stem cell and epithelial to mesenchymal transition program share a common theme: CD44 expression. In Chapter 2, we link the CSC and EMT program in HCC. In Chapter 3, we show that CD44 is an important CSC and EMT regulator and c-met that regulates CD44 expression through AKT signaling. In Chapter 4, we seek to target c-met in order to inhibit the CSC and EMT program and subsequently, inhibit tumor growth.

43 Chapter 2 Transforming growth factor- induces epithelial-to-transition and cancer stem cell characteristics Hien Dang, Wei Ding, Dow Emerson, and C. Bart Rountree The text represents a modified version of the journal article Snail1 induces epithelial-tomesenchymal transition and tumor initiating stem cell characteristics published in BMC Cancer (2011), 11:396. HD carried out the molecular and in vivo studies and drafted the manuscript and figures. WD assisted in in vivo experiments. DE participated in molecular in vitro studies (Figure 2-4E). CBR conceived of the study, and participated in its design and preparation of the manuscript.

44 Abstract Background: Cancer stem cells (CSCs), also known as tumor initiating stem-like cells, are a subset of neoplastic cells that possess distinct survival mechanisms and self-renewal characteristics crucial for tumor maintenance and propagation. The induction of epithelialmesenchymal-transition (EMT) by TGF has been recently linked to the acquisition of CSC characteristics in breast cancer. In HCC, a CSCs and EMT phenotype correlates with a worse prognosis. In this work, our aim is to elucidate the underlying mechanism by which cells acquire tumor initiating characteristics after EMT. Methods: Gene and protein expression assays and Nanog-promoter luciferase reporter were utilized in epithelial and mesenchymal phenotype liver cancer cell lines. EMT was analyzed with migration/invasion assays. CSCs characteristics were analyzed with tumor-sphere self-renewal and chemotherapy resistance assays. In vivo tumor assay was performed to investigate the role of Snail1 in tumor initiation. Conclusion: TGF induced EMT in epithelial cells through the up-regulation of Snail1 in Smaddependent signaling. Mesenchymal liver cancer post-emt demonstrates CSC characteristics such as tumor-sphere formation but is not resistant to cytotoxic therapy. The inhibition of Snail1 in mesenchymal cells results in decreased Nanog promoter luciferase activity and loss of selfrenewal characteristics in vitro. These changes confirm the direct role of Snail1 in some CSC traits. In vivo, the down-regulation of Snail1 reduced tumor growth but was not sufficient to eliminate tumor initiation. In summary, TGF induces EMT and CSC characteristics through Snail1 and Nanog up-regulation. In mesenchymal cells post-emt, Snail1 directly regulates Nanog expression, and loss of Snail1 regulates tumor growth without affecting tumor initiation.

45 Introduction Cancer stem cells (CSCs), also known as tumors initiating stem-like cells, are a subpopulation of neoplastic cells that possess distinct survival and regeneration mechanisms important for chemotherapy resistance and disease progression (49, 146). By definition, CSCs possess stem cell features including resistance to apoptosis and self-renewal (25-27). After their initial discovery and characterization within hematological malignancies (30, 31), CSCs have now been described in many different malignancies including hepatocellular carcinoma (HCC) (32, 33). Further evidence supports that HCC arises as a direct consequence of dysregulated proliferation of hepatic progenitor cells (40, 41). More recently, transcriptome analysis of human HCC clinical samples demonstrated that a progenitor-based (CSC-phenotype) expression profile is associated with a poor prognosis compared to differentiated tumors (hepatocyte-phenotype) (10, 42, 43). Resistance to therapy and metastatic disease are two factors that correlate a progenitorphenotype HCC with poor survival. CSCs are hypothesized to be the source of metastatic lesions, as a tumor-initiating cell (45). Although this hypothesis remains controversial, recent work establishes a connection between epithelial-mesenchymal-transition (EMT) and a CSC-phenotype (44, 140). EMT is a critical developmental process that plays a central role in the formation and differentiation of multiple tissues and organs. During EMT, epithelial cells lose cell-cell adhesion and apical-polarity, and acquire mesenchymal features, such as motility, invasiveness, and resistance to apoptosis (56). One of the key hallmarks of EMT is loss of E-cadherin, a cell-adhesion protein that is regulated by multiple transcription factors including Snail, Slug, and Twist. These transcription factors act as E-box repressors and block E-cadherin transcription (56). In cancer biology, EMT is one mechanism to explain the invasive and migratory capabilities that epithelial carcinomas

46 33 acquire during metastasis (34, 58). In HCC, increased expression of the E-cadherin repressors Twist and Snail correlates with poor clinical outcomes (147). In breast cancer, EMT is associated with the acquisition of a CSC CD44 + /CD24 low phenotype (44, 148). One of the major inducer of EMT is transforming growth factor- (TGF ), a multifunctional cytokine that regulates cell proliferation, differentiation and apoptosis (67). In early stages of carcinogenesis, TGF serves as a tumor suppressor by inhibiting cell growth, and in later stages of disease, tumor cells escape this growth inhibition. As late stage cancer tends to be resistant to TGF -driven growth arrest signals and as TGF is a known inducer of EMT, TGF is proposed to be a facilitator of cancer progression during late stage disease (73-75). TGF induces EMT by up-regulating Snail1 via the Smad-dependent pathways (69). Mishra and colleagues have reviewed the complexity of TGF signaling during hepatocarcinogenesis, specifically as related to 2-Spectrin loss and stem cell malignant transformation (45, ). As additional evidence linking EMT to CSCs, TGF regulates Nanog expression, a transcription factor that contributes to self-renewal and cell fate determination in embryonic stem cells (141, 142). In prostate cancer, increased Nanog expression is implicated in tumor progression, and the co-expression of Nanog and Oct4 promotes tumor-sphere formation (26, 152, 153). In colon cancer, increased Snail1 expression correlates to increased Nanog expression (154). In human HCC cell lines, TGF regulates CD133 expression, a marker of CSCs, through induction of epigenetic modifications of the CD133 promoter (67, 143). Thus, several studies have demonstrated that TGF drives EMT through Snail1 upregulation, and other studies have correlated EMT to the acquisition of CSC characteristics. What is lacking is an understanding of the mechanism of how liver cancer cells acquire CSC characteristics through EMT. Our hypothesis is that mesenchymal cells acquire CSC traits after EMT through Snail1-dependent mechanisms. In this report, we demonstrate that mesenchymal

47 34 liver cancer cells (post-emt) possess several CSC characteristics compared to epithelial cells. TGF induces EMT and CSC characteristics in epithelial cells through Snail1. In mesenchymal cells, knock-down of Snail1 result in loss of Nanog and CSC traits. In vivo studies demonstrate that Snail1 regulates tumor growth but does not fully control tumor initiation. 2.3 Materials and Methods Cell Culture. Epithelial and mesenchymal murine liver cancer cells were cultured in Dulbecco s modified Eagle s medium (DMEM)/F12 (Sigma) supplemented with 10% fetal bovine serum as described (155). The human HCC cell line Huh7 was provided by Jianming Huh, Penn State College of Medicine and cultured as described (143, 156). The human HCC cell lines MHCC97- L were provided by Xinwei Wang, National Cancer Institute, under agreement with the Liver Cancer Institute, Zhongshan Hospital, Fudan University, Shanghai, China and cultured as described (157). Transfections. For Snail1 transient knockdown, cells were transfected with 100pM of Snail1 Stealth sirna (Invitrogen) using Lipofectamine 2000 (Invitrogen). For Smad signaling inhibition, cells were transfected with 2ug of DNA using Fugene 6 (Roche). To generate Snail1 knockdown stable transfectants, mesenchymal cells were transfected with Snail1 Mission shrna lentivirus (Sigma) and selected with 2ug/ml of puromycin. Luciferase Assay. pcmv5-smad7-ha (Plasmid 11733), prk-smad3dc (Plasmid 12626), and Nanog-Luc (Plasmid 16337) were provided by Addgene. Cells were plated in 12 well plates, incubated overnight, and transfected with the Nanog-Luc plasmid and Renilla for 24 hours (4:1 Nanog-Luc:Renilla ratio). Cells were washed with 1X PBS, serum free starved for 2 hours, and

48 35 treated with 5 ng/ml of TGF for 24 hours. Following cell lysis, luciferase activity was measured using the Dual Luciferase Assay Kit (Promega) and a Sirius Luminometer V3.1 (Zylux). Luciferase reading light units (RLU) was normalized to Renilla RLU and a fold change was calculated. qrt-pcr. Trizol (Invitrogen) was used to isolate total RNA from cells according to manufacturer s protocol. Isolated RNA was quantified using the ND-1000 spectrophotometer (NanoDrop) and complementary single strand DNA was synthesized using the Omniscript RT Kit according to the manufacturers protocol (Qiagen). qpcr was performed using Taqman Gene Expression Assays and ABI-Prism 7700 Thermal Cycler (Applied Biosystems). Normalization was performed using β-actin or Gapdh as an endogenous control and relative gene expression was calculated using the comparative 2 (-ΔΔCt) method with SDS software (143). Cell Viability Assays. Cell viability was performed using the XTT (2,3-bis(2-methoxy-4-nitro-5- sulfophenyl)-2h-tetrazolium-5-carboxanilide) kit (Trevigen) according to the manufacturer s protocol. 5 X 10 3 cells were plated in 96-well plates, incubated for 24 hours at 37 C, and treated with specified agents at defined time points. Western Blot Analysis. Cells were washed twice with ice cold 1XPBS and cell lysates were harvested by the addition of lysis buffer (40nM Tris [ph 7.4], 150mM NaCl, 10mM ethylene diamine tetraccetic acid, 10% glycerol, 1% Triton X-100, 10mM glycerophosphate, 1mM Na3VO4, 1mM phenylmethylsulfonyl fluoride) supplemented with protease inhibitor cocktail tablets (Roche). BCA protein assay (Thermo Fisher Scientific) was used to determine protein concentration as described (158). 30ug of protein lysates were separated on a NuPAGE 4-12% Bis-Tris Gel (Invitrogen) and the separated proteins were transferred onto a polyvinylidene

49 36 difluoride membrane (Invitrogen). After blocking for 60 min with 5% non-fat dry milk, membranes were incubated with the primary antibody overnight at 4 C followed by incubation with corresponding secondary antibody for 60 min at room temperature. The membranes were developed using enhance chemiluminescence solutions (Thermo Fisher Scientific) (159). Cell Migration Assay. The capability of tumor cell migration was assessed using a wound-healing assay. Confluent cell adherents were manually wounded by scraping the cells with a 1,000μL pipette tip down the center of the well. The cell culture medium was replaced and migration was assessed at 24 hours (155). Matrigel Invasion Assay. Cell invasion was assessed using 6-well Transwell permeable inserts with 8-μm pores (Corning) (155). In brief, 1 X 10 5 cells were cultured in a serum-free DMEM/F12 medium in an insert coated with Matrigel (BD). Below the insert, the chamber of 6- well plates contained DMEM/F12 supplemented with 10% FBS. Cells were incubated in a 37 C incubator for 48 hours and the number of cells that invaded across the membranes and fallen onto the bottom of the plate was counted. Transcriptome analysis. Using the cell lines from the liver specific Pten -/- model described (155) P2E (epithelial) and P2M (mesenchymal) messenger RNA were analyzed using an Illumina mouse gene chip according to the manufacturer s protocol and as described (155). Housekeeping genes were used as standards to generate expression levels, and data analysis was conducted using 1.4-fold or greater change in expression with p<0.05 as significant. The full complement of the expression data is available at (Accession number GSE18255).

50 37 Tumorspheres Formation Assay. The capability of self-renewal was assessed using Corning Ultra-Low Attachment Surface (Corning). 5 X10 3 cells were seeded and incubated in a cell culture incubator for 1 week in DMEM/F12 supplemented with 10% FBS or serum free medium and phase-contrast images were obtained. In vivo tumor growth assay. Cells were counted with trypan blue exclusion and suspended in a 1:3 dilution of Matrigel (Matrigel:DMEM/F12 supplemented with 10% FBS) (143). 1X10 4 and 1X10 5 cells/50 L were injected subcutaneously into 10-week-old nude mice. Caliper measurements of tumor volume (length width height) were conducted every 2 days. After 3 weeks, mice were sacrificed for tumor analysis. All procedures were in compliance with our institution's guidelines for the use of laboratory animals and approved by the Penn State College of Medicine Institutional Animal Care and Use Committee. Statistical Analysis. Microarray statistical analysis was performed as describe (155). Student t test was used comparing two groups. One-way ANOVA was used comparing multiple groups followed by Tukeys post-hoc test. All analysis with a p<0.05 was considered significant. 2.4 Results Mesenchymal cells acquire CSC characteristics post-emt In a previous report, we established a model of EMT using liver cancer cell lines derived from Pten -/- mice (155). In this model, we transplanted epithelial liver cancer cells, and from the resulting tumors, harvested epithelial and mesenchymal cells. The epithelial tumor cells were identical to parent cells, labeled P2-Epithelial (P2E), and the mesenchymal, fibroblastoid cells,

51 38 Figure 2-1 Diagram of how murine P2E and P2M were derived P2E represents epithelial and P2M represents mesenchymal. (Figure modified from Ding et al, 2010 (155)) were labeled P2-Mesenchymal (P2M) (Figure 2-1). Both epithelial and mesenchymal cells demonstrated Pten -/- genotype (155). In support of the EMT-metastasis paradigm, mesenchymal cells demonstrated significant metastatic potential (155). To confirm the persistence of epithelial and mesenchymal phenotypes, we analyzed the expression of key EMT genes and migratory/invasion in vitro. The mesenchymal cells demonstrate loss of E-cadherin, gain of E- box transcription repressors Snail1 and Zeb2, significant migration in wound assay, and increased invasion through Matrigel pores compared to epithelial cells (Figure 2-2B-E). In mesenchymal cells, transcriptome profiling demonstrated increased expression of multiple liver CSC markers (Figure 2-3A). Real-time PCR validated up-regulated Nanog, Oct4, CD44, and EpCam (Figure 2-3B). Although CD133 is a strong CSC marker in previous reports, the mesenchymal cells have no detectable CD133 expression, making comparative analysis impossible. In terms of self-renewal assay, the mesenchymal cells were able to form large tumor-spheres in low adherent plates

52 39 (Figure 2-3C). Increased stem cell markers and tumor-sphere formation indicates that the mesenchymal cells have a CSC phenotype. Figure 2-2 Murine epithelial and mesenchymal liver cancer cells (A) Representative phase-contrast images (100X magnification) of epithelial and mesenchymal cells. (B) Heatmap of mesenchymal markers generated from raw microarray values. (C) Relative expression of mrnas endcoding E-cadherin, Snail1, Zeb1 and Zeb2 normalized to endogenous control Gapdh. Bars represent mean±sem of triplicates, *p<0.01. Western blot analysis of E-cadherin, Snail1, and β-actin. Data are representative of two independent experiments. (D) Wound healing assay. Bars represents mean ± SEM of triplicates, *p<0.01. (E) Matrigel invasion assay. Phasecontrast images (4X magnification) of cells that have invaded the membrane and adhered to the bottom of the plate. Data are expressed as the total number of invasive cells at the bottom of the plate, with five fields counted per well, n=3 wells/cell line, and reported as the mean±sem, *p<0.01.

53 40 Figure 2-3 Mesenchymal cells demonstrate up-regulation of CSCs characteristics (A) Microarray heatmap of CSCS markers. (B) Relative expression of mrnas encoding CSCS genes Bmi-1, Nanog, Oct4, Abcg2, CD44, Cxcr4, and EpCam normalized to endogenous control Gapdh. Data represent mean±sem of triplicates, *p<0.01. (C) Tumor-sphere assay was performed for two weeks using non-adherent plates. Phase contrast images are representation of three independent experiments (40X magnification) Resistance to chemotherapy is linked to cell proliferation To test the hypothesis that mesenchymal cells are resistant to chemotherapy, a CSC feature, cells were treated with doxorubicin and 5 Fluorouracil. The mesenchymal cells demonstrate increased sensitivity to genotoxic agents compared to epithelial cells (Figure 2-4A and B). In terms of cell cycle progression, the mesenchymal cells are highly proliferative compared to the epithelial cells (Figure 2-4E). Thus, we conclude that resistance to chemotherapy is linked to the level of cell proliferation, not mesenchymal status, consistent with the mechanism of action of cytotoxic agents. In addition to rate of proliferation, Abcg2 expression correlated with chemotherapy resistance (Figure2-4A and B, 2-3B), indicating that drug resistance may be dependent on the

54 41 ATP-binding cassette expression as a mechanism of drug efflux. ATP-binding cassette efflux has been highly correlated to epithelial phenotype liver CSCs (10, 160). In addition to resistance to genotoxic agents, we assessed whether the mesenchymal cells are resistant to TRAIL- and TGFβinduced apoptosis. Although there was no significant difference in response to TRAIL stimulation (Figure 2-4C), the mesenchymal cells demonstrate resistance to TGFβ-induced apoptosis (Figure 2-4D), a characteristic of CSCs (158). Figure 2-4 Resistance to chemotherapy is linked to cell proliferation (A-D) Cell viability evaluation using XTT assay of cells treated with doxorubicin, 5 Fluorouracil, TRAIL or TGFβ for 48 hours. Data reported as mean±sem, n=8, *p<0.01. (E) Cell proliferation of epithelial and mesenchymal cells. 1X10 4 cells were plated on 60mm 2 culture plates for 48 hours followed by cell count using cytometer at specific time points. Data are reported as mean±sem of triplicates, *p<0.01.

55 TGF -induced EMT results in CSC characteristics During later stages of disease, TGF induces EMT and contributes to disease progression (45, 161) After TGF stimulation; epithelial cells undergo a morphological change from cuboidal to fibroblastic-like cells (Figure 2-5A). In addition to morphology change, TGF treatment resulted in increased cell migration and the formation of larger tumorspheres in low adherent plates (Figure 2-5B and C). This TGF -induced change was associated with typical EMT characteristics, including decreased E-cadherin and increased Snail1 and Nanog (Figure 2-6D and E). Figure 2-5 TGF -induced EMT cells with CSC characteristics (A) Phase-contrast images of treated and untreated epithelial cells after 48 hours of TGF stimulation (200X). (B) Representative images of wound healing assay of TGF treated and untreated epithelial cells. Data represent mean±sem of triplicates, *p<0.01. (C) Tumorsphere formation assay of TGF treated and untreated epithelial cells. Cells were cultured in low adherent plates for two weeks (40X magnification). Data represent mean±sem of triplicates, *p<0.01.

56 43 A B Figure 2-6 TGF -induced EMT cells acquire stemness gene and protein expression (A) Relative expression of mrna encoding EMT and CSCSs genes normalized to endogenous control Gapdh after TGF stimulation. (B) Western blot analysis of CD44, Snail1, E-Cadherin, Nanog, and β-actin Inhibition of Snail1 blocks CSC characteristics In HCC, a CSC phenotype with Snail1 over-expression is associated with poor prognosis (147). To test the specific role of Snail1 in up-regulating CSC characteristics, we utilized sirna to knock down Snail1 in mesenchymal cells. After Snail1 sirna treatment, CSC markers Nanog and CD44 decreased significantly (Figure 2-7A), which was associated with decreased tumorspheres formation (Figure 2-7B) and decreased migration (Figure 2-7C) TGF regulates Snail and Nanog through Smad signaling The primary mechanism of TGF -induced EMT is through Smad-dependent signaling. Following activation of TGF receptors, Smad2 and Smad3 are phosphorylated and form the Smad2/3/4 heterocomplex, which translocates to the nucleus to regulate Snail1 transcription (58, 69, 162). After TGF stimulation in epithelial cells, Snail1 increased (Figure 2-6A and D). In

57 44 order to confirm that TGFβ induces Snail1 through Smad-dependent pathways in our model, we utilized inhibitory Smads, Smad7 and dominant-negative Smad3 (ΔSmad3), which block heterocomplex formation. Epithelial cells were transfected with Smad7 or DSmad3 vectors 24 hours prior to TGF stimulation. qpcr and western blot analysis demonstrated that inhibitory Smads significantly attenuated TGF -induced Snail1 up-regulation (Figure 2-8A and B). TGF regulates Nanog promoter activity through Smad signaling in human embryonic stem cells (141). To confirm that TGF can induce Nanog promoter activity in our model, epithelial cells were Figure 2-7 Snail1 regulates EMT and CSC characteristics in mesenchymal cells (A) Epithelial cells were treated with Snail1 sirna for 48 hours and mrna expression was analyzed for E-cadherin, Zeb1, Zeb2, Bmi-1, Nanog, and CD44 normalized to Gapdh. Bars represent mean±sem of triplicates, *p<0.01. Western blot analysis of Snail1, CD44, Nanog and -actin, with data representative of two independent experiments. (B) Tumor-sphere formation assay of mesenchymal cells transfected with scrambled or Snail1 sirna. Cells were cultured in low adherent plates (40X magnification). (C) Wound assay of mesenchymal cells transfected with either scrambled or Snail1 sirna. The number of cells migrated towards the wound was calculated. Data presented are mean ± SEM of triplicates, *p<0.01.

58 45 Figure 2-8 TGF regulates Snail1 and Nanog through Smad signaling (A) Relative Snail1 mrna expression of epithelial cells. One-way ANOVA with Tukeys posthoc test was performed. Data shown as mean±sem of triplicates, *p<0.01. (B) Western blot of Snail1 and -actin. Data represent two independent experiments. (C) Epithelial cells were transfected with Nanog-Luc plasmid and treated with TGF. Luciferase activity was normalized to Renilla. Data are shown as mean ± SEM of triplicates, *p<0.05. (D) Relative Nanog luciferase activity after 24 hours of TGF stimulation. Data are shown as mean ± SEM of triplicates, *p<0.05. co-transfected with Nanog-Luc and Smad7 or DSmad3 vectors. Following TGF stimulation, Nanog-Luc activity was significantly attenuated by inhibitory Smads (Figure 2-8C and D), indicating that TGF stimulates Nanog promoter activity through Smad-dependent signaling Snail1 directly regulates Nanog promoter After transient knockdown of Snail1, Nanog expression is decreased, indicating that Snail1 directly regulates CSC genes in mesenchymal cells (Figure 2-7A). To further investigate this Snail1-driven CSC expression profile, we established stable Snail1 knockdown in mesenchymal-

59 46 Snail1-shRNA cells (Figure 2-9A). In these mesenchymal-snail1-shrna cells, down regulation of Snail1 corresponded to decreased Nanog promoter activity and decreased Nanog and CD44 expression (Figure 2-9A and B). Figure 2-9 Repression of Snail1 attenuates Nanog promoter activity and tumor proliferation (A) Relative gene expression of Snail1, Nanog, and CD44 of mesenchymal-scrambledshrna compared to mesenchymal-snail1-shrna cells. Bars represent mean±sem of triplicates, *p<0.05. Western blot of Snail1 and actin, with blots representative of two independent experiments. (B) Inhibition of Snail1 reduces Nanog luciferase activity. Data presented are shown as mean±sem of three independent experiments, *p< Inhibition of Snail1 results in decreased tumor growth in vivo. As demonstrated, Snail1 is a key regulator of CSC characteristics in vitro. To investigate the role of Snail1 in tumor initiation, we inoculated 1X10 4 mesenchymal-snail1-shrna cells into nude mice. The mesenchymal-snail1-shrna cells demonstrate reduced in tumor growth compared to control mesenchymal cells. Analysis of tumors demonstrates that Snail1 expression was down regulated in 1X10 4 cell initiated tumors from mesenchymal-snail1-sir cells (Figure 2-10A). However, tumor initiation was not affected by Snail1 suppression, as evidence by all inoculations forming tumors, even in Snail1 inhibited cells.

60 47 Figure 2-10 Repression of Snail1 attenuates tumor growth Tumors of indicated number of cells of mesenchymal-scrambled-shrna or mesenchymal-snail1-shrna knock-down cells. Tumor volume reported as mean±sem, *p<0.05; N=4/group. Relative Snail1 mrna expression of tumor tissues. Data presented are reported as mean±sem of all Scrambled and Snail1 tumor tissues, *p<0.05; N=4/group Epithelial and mesenchymal differences in human HCC In order to investigate SNAIL1 and NANOG expression in human HCC cells, we utilized Huh7 and MHCC97-L cells. Huh7 cells have been described to be epithelial whereas MHCC97-L cells are mesenchymal with metastatic potential (156, 157). Accordingly, MHCC97-L cells demonstrate significant migration and invasion, increased expression of SNAIL1, NANOG and decreased expression of E-Cadherin (Figure 2-11B-D). Mesenchymal MHCC97-L cells also demonstrate CSC characteristics including increased NANOG, BMI-1, CD44 and OCT4 (POU5F1) mrna expression as well as increased tumorsphere formation (Figure 2-10E and F).

61 48 Figure 2-11 Human epithelial and mesenchymal liver cancer cells (A) Representative phase-contrast images (200X magnification) of Huh7 and MHCC97- L cells. (B) Representative images (40X magnification) of Huh7 and MHCC97-L wound healing assay after 24 hours of scratching. Experiments were performed in triplicates. (C) Relative expression of mrnas endcoding SNAIL1, ZEB1 and ZEB2 normalized to endogenous control GAPDH. Bars represent mean±sem of triplicates, *p<0.01. Western blot analysis of E-Cadherin and β-actin. Data representative of two independent experiments. (D) Matrigel invasion assay. Phase-contrast images (40X magnification) of cells that have invaded the membrane and adhered to the bottom of the plate. Data are expressed as the total number of invasive cells at the bottom of the plate, with five fields counted per well, n=3 wells/cell line, and reported as the mean±sem, *p<0.01. (E) Relative expression of mrnas encoding CSC genes BMI-1, NANOG, OCT4, ABCG2, CD44, CXCR4, and EPCAM normalized to endogenous control GAPDH. 2.5 Discussion Although liver transplantation has significantly improved survival in patients with early stage HCC, the prognosis for late stage HCC remains poor (2). Causes of poor prognosis in late stage

62 49 disease include invasive/metastatic disease and tumor recurrence after treatment. In breast cancer, EMT has been linked to CSC characteristics and resistant disease. Although this link between EMT and CSCs has been established in other cancers, including breast, prostate, nasopharyngeal, and colon cancer, this relationship has yet to be defined in HCC (44, 148, 163). One potential link between EMT and CSCs in liver cancer is TGF. TGF has a dual role in HCC either as a tumor suppressor in early stages or tumor promoter in later stages (45, 161). One of the mechanisms of early neoplastic transformation is through the evasion of cytostatic effects of TGF (161). During the late stages of HCC tumorgenesis, TGF stimulates cellular invasion through the EMT program (162). TGF induces EMT through Snail1, which represses E-cadherin by binding to E- box promoter elements (56, 58, 70). In cancer patients, an EMT-phenotype transcriptome profile, with increased Snail1 expression, correlates with invasive tumors (147, 164, 165) profile, with stimulation of epithelial liver cancer cells results in a mesenchymal phenotype with fibroblastoid appearance, loss of E-cadherin, increased invasion and migration, and an up-regulation of Snail1. In addition, TGF treatment induces a CSC phenotype in epithelial cells. Although TGF - induced EMT generates CSC characteristics (44, 148), the underlying mechanism has not yet been elucidated. Based on our results, we hypothesize that these CSC characteristics are Snail1 dependent. Inhibition of Snail1 causes the down-regulation of Nanog, Bmi-1 and CD44, loss of a migration and self-renewal as evidenced by decreased tumor-sphere formation. Another key regulatory signaling pathway known to induce EMT in liver cells is the Hedgehog (Hh) signaling pathway. Hh promotes EMT in response to chronic liver injury (166). Hh promotes EMT in response to chronic liver injury (166). In addition, Hh signaling has been suggested to play an important role in the maintenance of CSCs, and BMI-1, the polycomb group protein, may directly mediate Hh signaling in order to confer a self-renewal capacity in CSCs (40, 163, 167). However, within our system, we were unable to see significant differences of BMI-1

63 50 between epithelial and mesenchymal cells. TGF also directly controls Nanog in human embryonic stem cells (141). Nanog is a key transcription factor that regulates self-renewal in stem cells (26, 168). Recent studies demonstrate that Nanog promotes CSC characteristics, and the down regulation of Nanog inhibits sphere formation and tumor development (26, 153, 154, 169). In this report, Nanog is up-regulated by TGF through Smad signaling. In addition, Snail1 directly regulates Nanog promoter activity. In our murine system, Snail1 inhibition resulted in loss of tumor-sphere formation, decreased expression of CD44 and Nanog, and decreased tumor growth. According to our in vitro results, Snail1 clearly regulates CSC characteristics. However, the loss of Snail1 is not sufficient to inhibit tumor initiation, as evidenced by in vivo results. These findings are not un-expected in that the proposed CSC-driven tumor initiation is an early event in tumorigenesis, and cells that acquire CSC characteristics after EMT are a late event in tumor progression. In addition, Snail1 is one of many regulators of EMT, and thus manipulation of multiple factors may be required to fully inhibit tumor initiation. In summary, we demonstrated that TGF induces EMT and CSC characteristics through the up-regulation of Snail1 and Nanog. In addition, Snail1 directly regulates Nanog promoter activity. Inhibition of Snail1 alone is not sufficient to inhibit tumor initiation, but does result in reduction of tumor growth in vivo.

64 Chapter 3 Induction of tumor initiation is dependent on CD44s in c-met + hepatocellular carcinoma 51 Hien Dang, Steven N Steinway, Wei Ding, and C. Bart Rountree The text represents a modified version of the journal article c-met induction of tumor initiation occurs through CD44s submitted to the journal Carcinogeneis (August 2012). HD conceived the study, carried out all the molecular and in vivo experiments and drafted the manuscript and figures. SS assisted in in vivo work and contributed editorial input to the manuscript. WD helped with in vivo experiments. CBR provided intellectual support and contributed editorial input to the manuscript.

65 Abstract Background: In Chapter 2, we demonstrated that the TGF -induced epithelial-to-mesenchymal transition (EMT) increased CD44 expression with acquired cancer stem cell (CSC) characteristics. CD44 is a cell surface receptor that identifies mesenchymal CSCs in several cancers. However, the functional role of CD44 has yet to be elucidated. c-met, a high affinity receptor for hepatocyte growth factor (HGF), plays a critical role in cancer growth, invasion and metastasis. Hepatocellular carcinoma (HCC) patients with active HGF/c-Met signaling have a significantly worse prognosis. Furthermore, CD44 + /c-met + cells demonstrate a CSC phenotype in pancreatic cancer. Accordingly, the functional role of CD44 as a regulator of the CSC phenotype has yet to be elucidated. In this work, we examine the complex interaction between c-met and CD44s (standard form) and investigate the specific role of CD44s as a tumor initiator and stemness marker in HCC. Methods: Gene and protein expression assays and sirna and shrna assays were utilized in epithelial and mesenchymal phenotype liver human HCC cancer cell lines. Microarray analysis was performed for mrna characterization of cell lines. CSC characteristics were analyzed with tumorsphere self-renewal and chemotherapy resistance assays. In vivo tumor assay was performed to investigate the role of CD44s in tumor initiation. Immunohistochemical analysis was performed for human HCC samples. Conclusion: In a transcriptome profile analysis of human HCC samples, we identified a positive correlation (R 2 =0.717) between up-regulated CD44 and c-met expression. Immunohistochemical and immunoblot analysis of human HCC samples confirmed that 39% of human HCC samples express c-met and CD44s. MHCC97-H cells, which are CD44 + /c-met +, were utilized to

66 53 investigate the relationship between c-met and CD44s. The knockdown of c-met in MHCC97-H cells results in decreased CD44s and subsequent reduction in CSC characteristics and loss of tumorsphere formation. Furthermore, we demonstrate that the inhibition of PI3K/AKT signaling decreased CD44s expression and this subsequent loss of CD44s decreased tumorsphere formation. The down-regulation of CD44s leads to a significant loss of a CSC and mesenchymal phenotype, no change in c-met, increased AKT phosphorylation, and increased cell proliferation; suggestive of a negative feedback loop. A direct interaction between CD44s and c-met was confirmed with immunoprecipitation and immunoblot analysis. Finally, the down-regulation of CD44s in MHCC97-H cells decreased tumor initiation in vivo compared with the scrambled control. In summary, our data indicate that CD44s modulates c-met-pi3k-akt signaling to dysregulate cell proliferation and support a mesenchymal and CSCS phenotype, and highlights the therapeutic potential of targeting the c-met/cd44 axis in HCC.

67 Introduction Hepatocellular carcinoma (HCC) is the third leading cause of cancer related deaths worldwide (7). Evidence suggests that HCC arises as a direct consequence of dysregulated proliferation of hepatic progenitor cells (40, 41). Such progenitors, called cancer stem cells (CSCs) or tumor-initiating stem-like cells, have been described in many different malignancies, including HCC, and may account for poor survival and chemotherapy resistance within specific tumors (32, 33). Transcriptome analysis of HCC has demonstrated that a progenitor-based (CSCphenotype) expression profile is associated with a poor prognosis compared with differentiated tumors (hepatocyte-phenotype) (10, 42, 43). CSCs exhibit the capacity for rapid tumorsphere formation, enriched stem cell gene expression profile, and efficient tumor initiation in vivo. Furthermore, CSCs share multiple gene networks involved in self-renewal, drug efflux, survival, and pluripotency with embryonic stem cells (144). c-met is a receptor tyrosine kinase that, upon activation by its ligand hepatocyte growth factor (HGF), promotes malignant progression and metastasis in multiple cancers, including HCC ( ). Interestingly, 40% of HCC cases are c-met +, and c-met expression is associated with a poor prognosis (93, 97, 170). Aberrant c-met activation can occur through multiple mechanisms, including autocrine or paracrine ligand-dependent stimulation, mutational activation or gene amplification ( ). During development, homozygous deletion of HGF or c-met is embryonic lethal (174, 175). HGF/c-Met signaling has been demonstrated to be important in liver regeneration, hepatocyte survival, and tissue remodeling after acute injury (80, 176, 177). Following c-met phosphorylation and activation, multiple signaling pathways are involved as downstream targets, such as the PI3K/AKT and MAPK/ERK1/2 pathways (178, 179). CD44 is a transmembrane cell adhesion glycoprotein that participates in many cellular processes, including the regulation of cellular growth, survival, differentiation, lymphocyte

68 55 homing, and motility (108, 110). The variety of cellular processes affected by CD44 is likely the result of multiple CD44 isoforms produced by alternative splicing ( ). CD44s, the smallest standard form of CD44 of about 80kDa, is devoid of all CD44 variable exons and is the predominant variant that is ubiquitously expressed in epithelial tissues and has recently been proposed to be essential for epithelial-to-mesenchymal transition (EMT) (114). Splice variant 6 of CD44 (CD44v6) has been described as a co-receptor for receptor tyrosine kinases, which trigger signal transduction via interaction of the cytoplasmic tail (110). CD44v6 has been described as an adaptor protein that facilitates HGF binding to c-met and enhances downstream activation (84, 107, 125, 126). CD44 expression has also been described within CSC populations (108). In breast cancer, cells undergoing EMT exhibit increased CD44 expression and CSC characteristics (44, 62). Recently, CD44 + /c-met + cells have been demonstrated to be tumorigenic with stemness characteristics in pancreatic cancer, which suggests a dual role of c-met and CD44 as regulators of tumor initiation (26). Using HCC cell lines, we have previously demonstrated that pharmacologic inhibition of c-met results in the decreased expression of CD44s, which indicates a potential link between CD44s and c-met activation (170). In the current study, we investigate the co-regulation of c-met and CD44s, independent of CD44v6. Here, we define a specific functional role of CD44s as a tumor-initiating regulator in HCC. Our results demonstrate that c- Met regulates tumor initiation and stemness through phospho (p)-akt activation of CD44s. Consequently, our study establishes a novel regulatory relationship between CD44s and c-met within HCC and demonstrates the critical aspect of CD44s in maintaining a CSC and mesenchymal phenotype in c-met + cells.

69 Materials and Methods Cell viability assays/tumorspheres formation assay/ qrt-pcr: See chapter 2 Materials and Methods section (62). Cell culture: The human HCC cell line MHCC97-H was provided by Dr. Xinwei Wang, from the National Cancer Institute (NCI), under agreement with the Liver Cancer Institute, Zhongshan Hospital, Fudan University, Shanghai, China, and cultured as previously described (157, 170). The human HCC cell line Huh7 was maintained as previously described (180). The human SK- Hep1 cells were provided by Dr. Brian Barth, Penn State College of Medicine, and maintained in Dulbecco s modified Eagles Medium 1X supplemented with 10% defined FBS (Hyclone Laboratories, Logan, UT), 1 mm GlutaMAX-1 (Life Technologies), 100 U/ml penicillin and 100 µg/ml streptomycin. The cells were cultured in a humidified incubator with 5% CO 2 at 37⁰C. sirna and shrna plasmid constructs and generation of stable cell lines: c-met sirna was acquired from Thermo Scientific (Dharmacon, Chicago, IL). Stable shrna:tg HuSH 29mer shrna constructs against c-met in the pgfp-v-rs vector were purchase from OriGene (Rockville, MD). The following constructs have been validated using real-time PCR assays and have been used for developing stable c-met knock-down cell line. The c-met shrna targeting sequence of construct 1: 5 -TACTGCTGACATACAGTCGGAGGTTCACT-3 and construct 2: 5 -ACACTCCTCATTTGGATAGGCTTGTAAGT-3. The scrambled shrna construct with the pgfp-v-rs backbone was purchased from OriGene (Cat# TR30013). Short-hairpin construct oligonucleotide inserts of CD44s were generated for the psirna-h7sk G1 (clone sites: Bbsl/Bbsl) expression vector. Sequencing was performed to verify the presence of the sirna. The CD44s shrna targeting construct was 5 -CAAGTGGACTCAACGGAGA-3. MHCC97-H

70 57 cells were transfected with either scrambled shrna, c-met shrna, or CD44s shrna using Fugene 6 transfection reagents (Promega, Sunnyvale, CA ). Twenty-four hours after transfection, puromycin (2 g/ml) was added to select stable c-met shrna clones, and 100 ug/ml of zeocin was added to select stable CD44s clones. Multiple pooled clones of stable MHHCC97-H cells containing scrambled shrna and CD44s shrna and single clones containing c-met shrna were isolated and expanded. Knock-down of c-met and CD44s expression was validated using both real-time PCR and western blot assays as previously described (98, 158). Western blot analysis: See chapter 2-materials and methods section. c-met, p-c-met (1349), p-c- Met (1234/1235), AKT, p-akt, ERK1/ERK2, p-erk1/erk2, CD44, E-cadherin, vimetin, and moesin antibodies were purchased from Cell Signaling Technology (Danvers, MA). -actin antibody was obtained from Sigma (Allentown, PA). CD44v6 was obtained from ebioscience (San Diego, CA) and fibronectin was obtained from BD Sciences (San Jose, CA). Cell lysates were collected, and western blot was performed as previously described (180). Immunoprecipitation: Lysed MHCC97-H proteins were immunoprecipitated with Dynabeads Protein G (Invitrogen, Grand Island, NY) for 24 h after coating the beads with IgG, anti-c-met or CD44 and eluted per the manufacturer s protocol (Invitrotgen, Grand Island, MA). Eluted samples were run on a 2D gel and stained with Sypro Ruby (Molecular Probes) overnight at room temperature and washed with 10% methanol/6% acetic acid for 30 min followed by two washes with ddh 2 O. Gel plugs were sectioned and sent to Nextgensciences (Ann Arbor, MI) for mass spectrometry analysis. Parallel immunoblot analysis was performed to confirm the pull-downs. Immunohistochemistry:Paraffin embedded slides were labeled with anti-cd44 (Cell Signaling, Danvers, MA) and anti-c-met antibodies (Cell Signaling) and stained as previously described

71 58 (170). Microarray analysis: Using the MHCC97-H CD44s shrna, MHCC97-H c-met shrna or MHCC97-H scrambled shrna cells, mrna was analyzed using an Illumina human gene chip as previously described (155). Housekeeping genes were used as standards to generate expression levels, and data analysis was conducted using 1.4-fold or greater change in expression with P < 0.05 as significant. The full complement of the expression data is available at (accession number GSE38343). Animal care and xenograft transplantation experiment:nude Mice (Jackson Laboratory, Bar Harbor, ME) were fed and housed as previously described (170). All of the procedures were in compliance with our institution s guidelines for the use of laboratory animals and approved by the Institutional Animal Care and Use Committee. The cells were counted with trypan blue exclusion and suspended in a 1:3 dilution of Matrigel (Matrigel: DMEM/F12 supplemented with 10% FBS). Three different cell dilutions were used for bilateral subcutaneous injection: 1X10 4 cells/100 ml, 1X10 3 cells/100 L and 1X10 2 cells/100 L. Serial diluted cells were inoculated into 10-week-old nude mice (Jackson Laboratory, Bar Harbor, ME). Tumor initiation was checked every 3-4 days after injection. Caliper measurements of tumor volume (length width height) were conducted at the end of the study. The mice were sacrificed, and tumor tissues were fixed for histology studies or frozen for protein extraction. Immunofluorescence: 2.5X10 3 MHCC97-H cells were cultured on 8-well chamber slides overnight. The cells were fixed with 10% formalin for 15 min at room temperature and washed with 1X PBS three times for 5 min each. The cells were blocked with 5% normal goat serum, and IgG, CD44 or c-met (Cell Signaling) was incubated overnight at 4 C and washed with 1X PBS.

72 59 Cells were incubated with goat anti-mouse IgG H&L (Cy2 ) or goat anti-rabbit IgG H&L (Cy3 ) secondary antibody (Abcam, Cambridge, MA) for 2 h, washed with 1X PBS three times for 5 min each, and placed on slides with Vectashield Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA). CD44v6 sirna. The CD44v6 sirna targeting sequence construct (5 - GCAACTCCTAGTAGTACAA-3 ) was generated by Thermo Scientific (LaFayette, CO). MHCC97-H cells were transfected with sirna for 48 h using RNAiMax per the manufacturer s protocol (Invitrogen, Grand Island, NY). Flow Cytometry (FACS) Analysis. FACS experiments were performed using one million cells, incubated with mouse anti-human CD44-PE (BD Biosciences, Falcon Lakes, NJ) or anti-human c-met/2-apc (ebiosciences, San Diego, CA ). Analysis was performed using a FACS Calibur (BD Biosciences, Falcon Lakes, NJ). Post-FACS analysis was performed using the Flow-Jo program (Tree Star, Ashland, OR). Positive and negative gates were determined using immunoglobulin G (IgG)-stained and unstained controls. Gene Amplification Analysis. Genomic DNA was isolated from MHCC97-H and Huh7 cells using Puregene Core Kit according to the manufacturers protocol (Qiagen, Maryland, USA). MET and Methylenetetrahydrofolate reductase (MTHFR) (endogenous control) levels were evaluated using the following primers: MET-sense: 5'-CCATCCAGTGTCTCCAGAAGTG-3'; MET-anti-sense: 5'- TTCCCAGTGATAACCAGTGTGTAG-3'; MTHFR-sense: 5'- CCATCTTCCTGCTGCTGTAACTG-3'; MTHFR-anti-sense: 5'- GCCTTCTCTGCCAACTGTCC-3'. For c-met, 20 ng of genomic DNA was amplified for 35

73 60 cycles (95C for 20sec, 58C for 20 sec and 72C for 20 sec). For MTHFR, 20 ng of genomic DNA was amplified for 35 cycles (95C for 20sec, 60C for 20 sec and 72C for 20 sec). PCR products were run on a 3% agarose gel for 40 minutes at 120 v. Statistical analysis: Microarray statistical analysis was performed as describe (155). Students t test was used comparing two groups. One-way ANOVA was used when comparing multiple groups followed by Tukeys post-hoc test to look for differences amongst groups. All analysis with a p<0.05 was considered statistically significant.

74 Results CD44 expression correlates with c-met in human HCC To investigate the correlation between CD44 and c-met expression in HCC, we utilized publicly available HCC gene expression datasets from NIH GEO DataSets database (10, 181). We observed that CD44 and c-met expression has a high correlation in human HCC (R 2 =0.717) (Figure 3-1A). To further confirm this correlation between CD44 and c-met in HCC, we performed tissue immunohistochemical staining and immunoblot staining analysis on 101 HCC samples (Figure 3-1B and C and Appendix: Supplementary Figure S1A and S1B). Immunohistochemical analysis demonstrated that 39% (27/68) of the human HCC samples are both c-met + CD44 +. Immunoblot analysis of an additional 33 HCC samples demonstrated a similar correlation between c-met and CD44s in 45% (15/33) of the samples (Figure 3-1B and Appendix: Supplementary Figure S1A and S1B). Interestingly, in human HCC tissues, CD44 and c-met are co-localized (Appendix: Supplementary Figure S2) c-met + CD44s + HCC cells have increased mesenchymal and CSC characteristics To study the potential relationship between CD44s and c-met in HCC, we characterized four human HCC cell lines: Huh7, Hep3B, Sk-Hep1 and MHCC97-H. Microarray analysis demonstrated that SK-Hep1 and MHCC97-H cells have increased mesenchymal (i.e. fibronectin) and CSC markers (NANOG, SOX9, and BMI-1) and low expression of epithelial markers (i.e. E- cadherin) compared with the Huh7 and Hep3B cells (Figure 3-2A and B). Fluorescence activated cell sorting (FACS) analysis demonstrate that both the SK-Hep1 and MHCC97-H cell lines are 99% CD44 + compared with the Huh7 and Hep3B cells, which are less than 1.5% CD44 + (Appendix: Supplementary Figure S3A). Further characterization of the four cell lines

75 62 Figure 3-1 CD44s correlates with c-met expression in human HCC samples (A) Relative fold change of CD44 and c-met mrna expression of 149 HCC samples from microarray datasets from the Gene Expression Omnibus Database ( (B) Representative images of CD44 and c-met immunohistochemistry performed on 68 human HCC tissues using anti-cd44 and c-met antibodies (400X). (C) Representative western blot in which 7 out of 33 human clinical HCC samples demonstrating c-met, p-c-met Y1234/Y1235, p-c-met Y1349, and CD44s co-expression. See Appendix: Supplementary Figure S1 for all 33 samples.

76 63 demonstrates that CD44 + cell lines can readily form tumorspheres, have a mesenchymal phenotype with decreased E-cadherin, and have resistance to cytotoxic chemotherapy (Figure 3-2C-E and Appendix: Supplementary Figures S3B and C). The MHCC97-H cells demonstrated increased expression of both CD44s and c-met; thus, the MHCC97-H cells provide the best model for the c-met + /CD44s + HCC cells that have been observed in human samples. Although MHCC97-H cells express CD44v6, and the c-met/cd44v6 relationship has been established, human HCC samples demonstrated that CD44s (15/33) is highly expressed and correlated with c- Met expression, not CD44v6 (2/33) (Appendix: Supplementary Figure S1A) Down-regulation of c-met in the MHCC97-H cells decreases CD44s expression, mesenchymal and CSC phenotype Previously c-met has been demonstrated to regulate CD44 alternative splicing and CD44v6 expression (107). Thus we seek to investigate whether the inhibition of c-met affects CD44s or CD44v6 expression. The inhibition of c-met by PHA66572, a selective inhibitor of c- Met, had a greater effect on CD44s than CD44v6 (Appendix: Supplementary Fig. S4) (170). Based on this result, we hypothesized that c-met may regulate CD44s. As previously described, MHCC97-H cells have an active c-met, which occurs through c-met gene amplification rather than a mutational or HGF secretion (Appendix: Supplementary Figure S4A) (170). Because CD44v6 has been described to interact with c-met as a co-receptor, we wanted to investigate whether there is a direct interaction between CD44s and c-met. We performed an immunoprecipitation of CD44s and c-met and performed mass spectrometry analysis and immunoblot analysis in parallel (Appendix: Supplementary Figure S6 and Table S1). In both assays, we demonstrated that anti-cd44s precipitated c-met and vice versa (Appendix: Supplementary Figure S6A). The co-localization of CD44s and c-met in MHCC97-H cells was

77 64 Figure 3-2 CD44s + HCC cells have mesenchymal and tumor-initiating stemlike characteristics (A) Microarray heatmap of epithelial, mesenchymal and tumor-initiating stem-like markers in four HCC cell lines. (B) Relative mrna expression of mrnas encoding CD44, E-cadherin and c-met normalized to the endogenous control GAPDH. Data represent mean±sem of triplicates, *p<0.01. (C) HCC cells were treated with sorafenib for 24 h. Endogenous protein expression of apoptotic markers PARP, cleaved PARP, and β-actin. (D) Protein expression of endogenous CD44s, c-met, mesenchymal markers, and downstream signaling. The data are representative of three independent experiments. (E) Tumorsphere assay was performed for two weeks in non-adherent culture plates, and the number of tumorspheres was counted. The data represent the mean±sem of triplicates, *p<0.01. Phase-contrast images are representative of triplicates (40X magnification).

78 65 confirmed using immunofluorescence and confocal microscopy (Appendix: Supplementary Figure S6B). Taken together, our data indicate that CD44s and c-met interacts. Figure 3-3 c-met regulates mesenchymal and CSC phenotype and CD44s expression (A) Stable down-regulation of c-met decreased the mesenchymal markers CD44s, moesin, and fibronectin and increased E-cadherin expression as determined by immunoblotting. (B) c-met regulates tumorsphere formation. The data represent the mean±sem of triplicates, *p<0.05. To investigate the relationship between c-met and CD44s, we developed a stable MHCC97-H c- Met shrna cell line (Figure 3-3A). In the MHCC97-H c-met shrna cells, p-akt, p-erk1/2 signaling, and CD44s expression are down-regulated compared with the control MHCC97-H scrambled shrna cells. Down-regulation of c-met leads to increased E-cadherin and decreased tumorsphere formation (Figure 3-3A and B). To investigate which c-met downstream signaling regulates CD44s expression, we targeted both the p-akt and p-erk1/2 using the small molecule inhibitors LY and PD98059, respectively and blotted for p-akt and p-erk1/2. CD44s expression was significantly decreased after LY treatment compared with DMSO or PD98059 treatment (Figure 3-4A), which indicated that CD44s expression is activated by c-metp-akt signaling. To test whether the inhibition of p-akt activity and subsequent loss of CD44s

79 66 inhibits tumorsphere formation, we treated MHCC97-H cells with a low dose of LY or PD98059 for two weeks in low adherent culture conditions. LY was able to significantly inhibit tumorsphere formation compared with DMSO or PD98059 treatment (Figure 3-4B). Figure 3-4 c-met regulates CD44s expression through AKT signaling (A) MHCC97-H cells were treated with DMSO, LY (PI3K inhibitor) or PD98059 (MEK inhibitor) for 24 h, and immunoblot analysis was performed. (B) MHCC97-H cells were cultured in low-adherent culture dishes with 25 M DMSO, LY or PD98059 for two weeks, and the number of tumorspheres was counted. The data represent the mean±sem of triplicates, *p< CD44s regulates mesenchymal and CSC phenotype The down-regulation of CD44s and the significant decrease in tumorsphere formation after p-akt inhibition indicate that CD44s may be a critical CSC marker. To test whether CD44s regulates tumor-initiating characteristics, we generated a stable MHCC97-H CD44s shrna cell line (Figure 3-5A and Appendix: Supplementary Figure S7A). Compared with the MHCC97-H scrambled shrna cells, the down-regulation of CD44s slightly decreased CD44v6, decreased expression of mesenchymal, and CSC markers and increased epithelial markers (Figure 3-5B-D and Appendix: Supplementary Figure S7B). Notably, CD44s down regulation had little or minimal affect on c-met and c-met phosphorylation (Figure 3-5C).

80 67 Figure 3-5 CD44s regulates mesenchymal and tumor-initiating stem-like characteristics (A) Endogenous protein levels of CD44s and the mesenchymal markers E-cadherin and fibronectin in two pooled stable CD44s shrna cell lines. (B-D) Knockdown of CD44s in MHCC97-H cells. Relative mrna expression of stem cell genes in CD44s shrna #1 compared with the scrambled control normalized to GAPDH. The data represent the mean±sem of triplicates, * p< To investigate whether the over-expression of CD44s can induce CSC and mesenchymal characteristics, we generated a stable cell line which over expressed CD44s in Huh7 cells (Huh7 CD44s), which are CD44s neg /c-met neg. Over expression of CD44s in Huh7 leads to an increase in mesenchymal markers including Twist1 and Zeb2, increased fibronectin and decreased in E-

81 68 cadherin protein expression, increased tumorsphere, formation and increased in both stemness genes BMI-1, Nanog and OCT4 (Figure 3-6A-D). Figure 3-6 CD44s promotes CSC and mesenchymal characteristics in Huh7 cells (A) Immunoblot analysis demonstrates increased CD44s expression after CD44s overexpression in Huh7 HCC cells. The data are representative of two independent experiments. (B) Tumosphere assay of Huh7 and Huh7-CD44s after two weeks. Phase contrast images are taken (40X magnification). (C-D) Relative mrna expression of mesenchymal and CSC genes. Data are representative of three independent experiments and shown as mean±sem of triplicates, * p<0.001.

82 CD44s modulates cell proliferation To test whether the down-regulation of CD44s has an effect on cell proliferation we investigated downstream c-met signaling. Interestingly, CD44s down-regulation resulted in the increased activation of p-akt and slight change of ERK1/2 signaling (Figure 3-7). Figure 3-7 CD44s modulates p-akt signaling Immunoblot analysis demonstrates increased AKT expression after CD44s knockdown. The data are representative of two independent experiments. c-met and CD44s down-regulation leads to decreased mesenchymal phenotypes (Figure 3-3 and 3-5), and the down-regulation of CD44s had little effect on c-met signaling (Figure 3-7). Because p-akt was significantly changed after CD44s knock-down, we tested whether this increased in p-akt signaling leads to increased cell proliferation. As expected, the down-regulation of c-met, which results in a loss of AKT activation, leads to significant decreased cell growth and proliferation (Figure 3-8 A and B) c-met activation of mesenchymal and CSC characteristics occurs through CD44s Our observations thus far suggest that c-met regulates CD44s expression, and CD44s modulates c-met activation of cell proliferation and CSC characteristics. To further confirm the

83 70 Figure 3-8 CD44s modulates cell proliferation and cell growth (A) Propidium Iodide cell cycle analysis of MHCC97-H Scrambled, c-met or CD44s shrna. (B) Cell proliferation analysis of MHCC97-H Scrambled, c-met or CD44s shrna cells. Data represent two independent experiments and are shown as mean ± SEM of 8 replicates, *p<0.05. critical role of CD44s in CSC characteristics, we compared MHCC97-H adherent cultured cells with tumorspheres. Immunoblot analysis demonstrated that there was no difference in c-met expression, but there was a significant increase in CD44s expression in tumorspheres compared with adherent cells (Figure 3-9). Notably, CD44v6 expression is lower in tumorspheres than in adherent cells, which further confirms the importance of CD44s, not CD44v6, as a CSC marker.

84 71 Figure 3-9 CD44s is highly expressed in tumorspheres compared with adherent cells MHCC97-H cells were plated in lowadherent cell or adherent cell culture dishes for two weeks and immunoblotting analysis was confirmed on tumorsphere lysates. The data are representative of three independent experiments. To confirm that CD44s is required for tumorsphere formation downstream of c-met, we performed a tumorsphere assay with MHCC97-H scrambled, c-met and CD44s shrna stable cell lines. The down-regulation of CD44s significantly decreased tumorsphere formation compared with c-met or scrambled shrna (Figure 3-10). To further demonstrate the importance of CD44s, not CD44v6, as the key CSC inducer, we transfected MHCC97-H cells with CD44v6 sirna to test tumorsphere formation capability. The knock-down of CD44v6 had no significant affect on c-met signaling or tumorsphere formation (Appendix: Supplementary Figure S8 A and B). Figure 3-10 CD44s regulates tumorsphere formation MHCC97-H scrambled, c-met and CD44s shrna #1 were grown in low-adherent culture plates for two weeks, and the number of tumorspheres was counted. The data are representative of two independent experiments and are shown as the mean ± SEM of triplicate plates.

85 72 Figure 3-11 CD44s recovers CSC characteristics after c-met inhibition (A-B) CD44s recovers tumorsphere formation. MHCC97-H cells were transfected with c- Met sirna for 24 h followed by overexpression of CD44s-pBabe vector or pbabe empty vector retrovirus for an additional 48 h (immunoblot) or two weeks (tumorsphere assay). Immunoblot data are representative of two independent experiments. The data for the tumorsphere assay data are representative of two independent experiments and are shown as the mean ± SEM of triplicate wells. We also tested the hypothesis that CD44s can recover tumorsphere formation after c-met inhibition by transfecting MHCC97-H cells with c-met sirna or scrambled sirna followed by over-expression of CD44s for 48 h. As previously demonstrated, c-met down-regulation inhibited both CD44s and c-met expression (Figure 3-11A). Subsequent recovery of CD44s expression partially rescued tumorsphere formation (Figure 3-11B).

86 CD44s regulates tumor initiation in vivo To test whether CD44s regulates tumor initiation in vivo, we subcutaneously injected athymic nude mice with 1 x 10 2 (N=10), 1 x 10 3 (N=8), or 1 x 10 4 (N=6) MHCC97-H CD44s shrna or MHCC97-H scrambled shrna cells (Figure 3-12A). In the 1 x 10 2 group, the downregulation of CD44s delayed tumor initiation by 20 days and decreased tumor incidence by 80%, whereas in the MHCC97-H scrambled shrna group, there was a 60% tumor incidence (Figure 3-12B). In the 1 x 10 3 group the down regulation of CD44s also delayed tumor initiation by 20% when compared to Scrambled shrna control. The tumor incidence for MHCC97-H scrambled control shrna group was 100% and 40% for the MHCC97-H CD44s shrna group. Together, these data demonstrate that the down-regulation of CD44s results in inhibition of tumor initiation and incidence in lower cell dilutions compared with the scrambled shrna controls (Figure 3-12B and C). However, when 1x10 4 MHCC97-H CD44s shrna cells were inoculated, there was no significant difference in tumor initiation or incidence. 3.5 Discussion HCC patients with an up-regulation of c-met signaling or CSC transcriptome profile have a poor prognosis (10, 17, 97, 181, 182). In solid tumors, c-met + /CD44 + cells demonstrate increased CSC gene expression profile, increased tumor-sphere formation, and efficient tumor initiation in limited dilution assays (33, 46-48, 102, 183). Although the importance of CD44 in tumor progression and CSC populations has been demonstrated, most reports that define CSC populations with CD44 utilize antibodies that recognizes all CD44 isoforms (108). Therefore, it has yet been identified which CD44 variants are responsible for the CSC phenotypes. Our findings establish for the first time the functional relationship between the CD44 standard variant

87 74 Figure 3-12 CD44s regulates tumor initiation in vivo (A) Tumor initiation graph of MHCC97-H CD44s shrna compared with the scrambled shrna control. (B-D)Bilateral subcutaneous injections of 1x10 2, 1x10 3, or 1x10 4 cells were inoculated into nude mice, and the number of tumors formed and the percent tumor initiation were calculated (1x10 2, N=10; 1x10 3, N=8; or 1x10 4, N=6). Confirmation of down-regulation of CD44s and c-met signaling was performed by immunoblotting.

88 75 (CD44s) and c-met in regulating a CSC phenotype. We confirm that CD44s and c-met coexpress in human HCC by using our own data set and others (10, 181). We discovered a novel complex regulatory relationship between CD44s and c-met that controls cell proliferation, mesenchymal, and CSC phenotype. In this work, we demonstrated the importance of the c-met/akt/cd44s cascade in promoting a CSC phenotype. The down regulation of CD44s significantly decreased tumorsphere formation compared to the down-regulation of c-met. However, CD44s was not able to fully rescue tumorsphere formation after c-met inhibition, suggesting that c-met regulates tumorsphere formation independent of CD44s. Notably, when we inhibited c-met in MHCC97-H CD44s shrna cells, E-cadherin expression was further increased. Together, these data suggest that c- Met alone plays an important role in regulating both CSCs and mesenchymal phenotypes independent of CD44s. The c-met/hgf signaling cascade is important for morphogenesis during embryonic development and organ regeneration by inducing epithelial-to-mesenchymal transition (EMT) and can be highjacked by cancer cells to promote metastasis (171, 176). Furthermore, c- Met has been implicated in regulating the stem/progenitor phenotype by transcriptional regulation of stemness factors including NANOG, POU5F1, and Sox2 (46). Therefore, it is likely that c- Met, through other mechanisms independent of CD44s, can regulate the CSC and mesenchymal phenotype; however, it is predominantly through CD44s that c-met can promote the CSC and mesenchymal phenotype that is important for cell survival. The relationship between c-met and CD44v6 is well established (84, 107, 126). In the MHCC97-H cells both CD44v6 and CD44s isoforms are expressed. The down-regulation of c- Met leads to the loss of CD44v6, suggesting that c-met also regulates CD44v6. The question arose as to whether CD44v6 may also play a role in the CSC and mesenchymal phenotype. This leads to the possibility that CD44v6 may contribute to the regulation of the CSC and mesenchymal characteristics. However, when we compared the expression of CD44v6 and

89 76 CD44s in tumorspheres compared to adherent cultured cells, CD44s was highly expressed in the tumorsphere, whereas CD44v6 expression was slightly lower in the tumorsphere compared to adherent cells. In SK-Hep1 cells, CD44s is highly expressed in tumorspheres compared to adherent cultured cells (Appendix: Supplementary Figure S10). Furthermore, the over expression of CD44s in Huh7 cells increased tumorsphere formation capability and CSC related genes significantly. We further demonstrate that the inhibition of CD44v6 did not significantly decrease tumorsphere formation and that CD44s was able to partially rescue tumorsphere formation after c-met inhibition. We also demonstrate that CD44s, not CD44v6 (data not shown), interacts and co-localizes with c-met and this interaction is independent of HGF. While the CD44s/c-Met interaction is novel, the functional implication in HCC progression is unknown. Future studies will be performed to understand the functional relevance of the CD44s/c-Met interaction in HCC. According to our data, CD44s appears to promote anti-proliferation and enhance a mesenchymal and CSCS phenotype. Thus, we propose that CD44s plays an important role in attenuating cell growth and proliferation to induce a more quiescent state, which is consistent with CSC populations. The MHCC97-H cells have been established as a valuable model of HCC metastasis and have been demonstrated to have high metastatic potential in vivo (157). While CD44v6 does not play a role in the regulation of a CSC phenotype, CD44v6 may play an important in cell migration and metastasis by promoting c-met signaling through ERM (ezrin, radaxin, and moesin) proteins (84, 111, 184). However, this process seems to be HGF dependent. While the role of CD44v6 in cell migration has been well studied in other solid tumors, its functional role in HCC will need to be further investigated. In this study, we demonstrate a positive correlation between CD44s and c-met in clinical HCC samples and show, for the first time, an interaction and functional relationship between CD44s and c-met within HCC. We present evidence that c-met regulates CD44s to drive a mesenchymal and CSC phenotype that is important for cell survival and that the down regulation

90 77 of CD44s decreases tumor initiation both in vitro and in vivo. Our data provides insight into how cell surface markers, c-met and CD44, are functionally responsible for the CSCs and mesenchymal phenotype and thus represents potential targets for the development of new therapeutic agents for targeting CSCs.

91 Chapter 4 78 c-met, a potential therapeutic target for hepatocellular carcinoma Hien Dang, C. Bart Rountree, Steven Steinway, Hanning You, and Wei Ding This chapter is a composite of new experiments as well as studies performed by Hien Dang and included in a manuscript by You, Ding, Dang, Jiang, and Rountree ( c-met represents a potential therapeutic target for personalized treatment in hepatocellular carcinoma ). For each figure, the relative contributions of HD have been explicitly noted. However, it should also be noted that Figure 4-2 was conducted exclusively by HY without input from HD. All new experiments were jointly conducted by HD and WD and will be submitted in a manuscript entitled, TGF stimulates EGFR as a bypass mechanism for the acquired resistance of c-met inhibitors in

92 hepatocellular carcinoma which is to be submitted to BMC Cancer (September 2012) with HD and WD as co-authors and equal contributors Abstract Background: c-met, a high-affinity receptor for Hepatocyte Growth Factor (HGF) plays a critical role in cancer growth, invasion and metastasis. Hepatocellular carcinoma (HCC) patients with an over-active HGF/c-Met signaling pathway have a significantly worse prognosis. As presented in Chapter 3, c-met regulates the epithelial-to-mesenchymal transition and cancer stem cell characteristics through CD44s. These data suggest that c-met inhibition will reduce tumor burden. Although targeting the HGF/c-Met pathway has been proposed for the treatment of multiple cancers, the effect of c-met inhibition in HCC remains unclear. Strategies using c-met tyrosine kinase inhibitors are currently in clinical trials for HCC; however, monotherapy using growth factor receptor tyrosine kinase inhibition in other cancers has demonstrated early success that is not durable. For example, in non-small cell lung, by-pass mechanisms for resistance to epidermal growth factor receptor (EGFR) inhibitors include c-met amplification and activation. Methods: Human HCC cell lines, Huh7, Hep3B, MHCC97-L and MHCC97-H, were utilized in this study to investigate the effect of c-met inhibition using the small molecule, selective c-met tyrosine kinase inhibitors PHA and PF We utilized the human MHCC97-H c- Met positive cell line to explore a mechanism of EGFR activation after c-met inhibition. Gene and protein expression assays were utilized to characterize MHCC97-H liver cancer cells. MHCC97-H c-met or Scrambled shrna expressing stable cell lines were generated for both in vitro and in vivo downstream analysis. Cell proliferation assays were performed to measure cell viability in sirna screenings and following drug treatment in vitro. Tumor growth behavior of MHCC97-H c-met shrna and MHCC97-H scrambled shrna were evaluated after drug

93 80 treatment in vivo. Results: MHCC97-L and MHCC97-H cells demonstrate a mesenchymal phenotype with decreased expression of E-cadherin and increased expression of c-met, fibronectin and Zeb2 compared to Huh7 and Hep3B cells, which have an epithelial phenotype. In addition, c-met positive MHCC97-L and MHCC97-H cells demonstrated cancer stem cell-like characteristics, such as resistance to chemotherapy, tumorsphere formation, and increased expression of CD44s and ABCG2. In the mesenchymal MHCC97-L and MHCC97-H cells, PHA treatment blocked phosphorylation of c-met and downstream PI3K/Akt and MAPK/Erk pathways. PHA treatment suppressed tumor sphere-formation and inhibited CD44s expression, cell proliferation and it induced apoptosis in c-met positive cells. In xenograft models, administration of PHA significantly inhibited c-met positive MHCC97-L and MHCC97-H tumor growth, and PHA treated tumors demonstrated marked reduction of both c-met phosphorylation and cell proliferation. c-met negative Huh7 and Hep3B cells were not affected by c-met inhibitor treatment in vitro or in vivo. Although, c-met inhibition reduced tumor growth, in vitro analysis demonstrates a bypass survival mechanism. sirna screening against 873 kinases and phosphatases identified EGFR as a potential mechanism for survival in MHCC97-H c-met shrna cells and was validated using transcriptome profile analysis. Combination therapy in vivo of MHCC97-H tumors with c-met and EGFR inhibitors result in significant reduction and stabilization of tumor volume compared to c-met inhibition alone. Conclusion: c-met represents a potential target of personalized treatment for HCC with an overactive HGF/c-Met pathway. However, c-met inhibition in HCC results in a compensatory survival mechanism through the up-regulation of the EGFR pathway. This survival mechanism can be partially overcome with combination EGFR and c-met inhibitor therapy.

94 Introduction Hepatocellular carcinoma (HCC) represents the third leading cause of cancer-related death worldwide, and is the only carcinoma with increasing mortality in the United States during the last decade (7). Although surgical resection and transplantation have significantly improved survival in patients with small tumors with no evidence of invasion or metastasis, the prognosis of HCC for late stage disease remains very poor (2). In addition, within HCC transplant patients, recurrent and metastatic disease remain the most important factors in survival (4). In addition to tumor number, size, and vascular invasion observed in imaging studies, one molecular characteristic that appears to predict poor survival in HCC is c-met expression (96-99). Hepatocyte Growth Factor (HGF) is an autocrine and paracrine factor that is produced by stromal cells. HGF acts on c-met, a high affinity tyrosine kinase receptor (87). During development, homozygous deletion of HGF or c-met has an embryonic lethal outcome (174, 175). Although HGF/c-Met signaling does not play a role in liver homeostasis during normal physiologic conditions, many studies have demonstrated the important role of HGF in liver regeneration, hepatocyte survival, and tissue remodeling after acute injury (80, 185). Following c- Met phosphorylation and activation, multiple signaling pathways are activated downstream including the PI3K/Akt and MAPK/Erk pathways (178, 179, 186). Through these intermediary pathways, HGF-induced c-met activation triggers a variety of cellular responses, including cell proliferation, survival, cytoskeletal rearrangements, cell-cell dissociation, and motility (87, 187). In cancer, c-met activation can occur through multiple mechanisms, including autocrine or paracrine ligand-dependent stimulation, mutational activation or gene amplification (171). Within cancer, the HGF/c-Met axis mediates a proliferative advantage and promotes tumor invasion and metastasis (87, ). As a result of the strong clinical correlation between c-met expression and metastatic disease, c-met is considered a therapeutic target against tumor

95 82 growth and metastasis in lymphoma, gastric cancer, melanoma, and lung cancers. In liver cancer, c-met expression is correlated with aggressive, metastatic disease (155). In the last decade, targeted tyrosine kinase therapy has been proposed and tested clinically. In non-small cell lung carcinoma, more than 70% of patients with Epidermal Growth Factor Receptor (EGFR) activating mutations have a favorable initial response to the EGFR inhibitors gefitinib or erlotinib (191). However, the overwhelming majority of EGFR inhibitor responders will develop acquired resistance, through various bypass mechanisms, such as MET amplification and activation (100, 192). These clinical data support a strong link between the c-met and EGFR pathways. We propose that a similar mechanism exists in HCC, through the up-regulation of EGFR signaling. In Chapter 3, we demonstrated that CD44s promotes tumor initiation and a mesenchymal phenotype. Furthermore, we showed that c-met is the upstream regulator of the mesenchymal and CSC phenotype through the regulation of CD44s. In the current study, we demonstrate that by targeting c-met, we inhibit CD44s and subsequently CSC and mesenchymal characteristics. More importantly, we show that by targeting c-met, we can reduce tumor growth; thus demonstrating the importance of c-met as a key regulator of the CSC and mesenchymal phenotype. Although, c- Met inhibition led to reduced tumor growth, single RTK treatment does not completely eliminate the tumor. Finally, we demonstrate that combination c-met and EGFR therapy gives an additive affect in the inhibition of tumor cell growth in vitro, and demonstrated maximal tumor suppression compared with single drug treatment.

96 Materials and Methods Cell viability assays/tumorsphere formation assay/qrt-pcr/shrna plasmid constructs/ Development of c-met shrna HCC cells/western blot/animal Care/Transcriptome Analysis: See chapter 3 Materials and Methods section (62). Cell culture: The human HCC cell line Hep3B [acquired from AddexBio (San Diego, CA)] was maintained as previously described (180). See Chapter 2 and 3 Materials and Methods sections for Huh7, SK-Hep1, MHCC97-L and MHCC97-H. sirna library screening: Invitrogen s sirna screening library covering 873 kinases and phosphatases was utilized to screen for bypass mechanisms following c-met inhibition of MHCC97-H c-met shrna cells were plated in 96-well plates and reverse transfected with individual sirna using the lipid-mediated transfection reagent, Lipofectamine RNAiMAX (Life Technologies Corporation, Grand Island, NY). 48 h following transfection, cell viability was assessed using XTT assays, and sirna with a cell viability Z-score of -2 or less were further validated. Xenograft transplantation experiments: Cells were counted with trypan blue exclusion and suspended in 1 PBS, with 3 X10 5 cells/100 L mixed at a ratio of 1:1 with Matrigel. This cell preparation was tthem inoculated into 6-week-old nude mice subcutaneously. Caliper measurements of tumor volume (length width height) were conducted every 3 days. Daily treatment with PHA (25mg/kg intravenously) or PF suspended in ph 4.0 water and orally administrated by a gavage (50 mg/kg/d) and/or Gefitinib in 1% Tween-80 solution (75

97 84 mg/kg/d by gavage) was initiated when tumors reached 5 or 200 mm 3 in size, with vehicle solution as a control (193, 194). Mice were euthanized after 28 days of treatment and tumor tissues were snap frozen using dry ice and stored at -80 C. Statistical analysis: Student t-test was used to compare data from two groups and one-way ANOVA with Tukey s posthoc testing was used to evaluate the differences amongst multiple groups with p<0.05 established as statistically significant.

98 Results Mesenchymal cells with cancer stem cell characteristics are c-met positive The MHCC97-L and MHCC97-H cell lines have low and high metastatic potential, respectively (157). We investigated whether metastatic HCC cells, MHCC97-L and MHCC97-H, have mesenchymal features in comparison with nonmetastatic Huh7 and Hep3B cells. MHCC97-L and MHCC97-H cells demonstrate low E-cadherin expression, consistent with a mesenchymal phenotype, and high expression of Zeb2 compared with Huh7 and Hep3B cells (data not shown). Protein expression confirmed a mesenchymal phenotype in MHCC97-L and MHCC97-H cells, with low levels of E-cadherin and high fibronectin expression (Figure 4-1B). The mesenchymal phenotype of MHCC97-L and MHCC97-H cells correlates with strong expression and constitutive phosphorylation of c-met (Figure 4-1B). c-met + cells demonstrate also display CSC characteristics with high levels of CD44s protein expression and tumorsphere formation capability compared to c-met neg Huh7 and Hep3B cells (Figure 4-1A). Figure 4-1 c-met + MHCC97-L and MHCC97-H cells have mesenchymal and CSC characteristics (A) Tumorsphere formation assays of all four cell lines. Data are representative of three independent experiments. - HD (B) Representative immunoblot of three independent experiments. A-HD; B-HY and HD (156)

99 c-met + cells are sensitive to PHA Constitutive activation or HGF stimulated tyrosine phosphorylation of c-met is blocked by PHA665752, a small molecule inhibitor that functions as a selective inhibitor of c-met phosphorylation in gastric, lung, and pancreatic cancer cells (194). Using PHA665752, we investigated the effect of tyrosine kinase inhibition on the activation of c-met and downstream signaling pathways in human HCC. As shown in Figure 4-2, PHA treatment inhibited c- Met phosphorylation at multiple tyrosine residues (Y1234/Y1234 and Y1349) and reduced downstream phosphorylation of AKT and ERK (P44/42) in c-met positive MHCC97-L and MHCC97-H cells. PHA had no effect on c-met negative Huh7 and Hep3B cells. Figure 4-2 PHA selectively inhibits c-met + downstream PI3K/AKT and MAPK/ERK signaling HCC cells and Western blot analysis of all four HCC cell lines treated with 1 M of PHA for 24 hours. HY (156)

100 c-met inhibition reduces tumor growth in vivo Because our in vitro data demonstrated that PHA effectively targets c-met and downstream pathways, we investigated whether c-met inhibition was capable of reducing tumor growth in vivo. As depicted in Figure 4-3, PHA administration significantly inhibited growth of c-met positive MHCC97-L and MHCC97-H xenograft tumors. PHA had no significant effect on Huh7- and Hep3B-derived tumors. Immunohistochemical analysis verified that PHA administration inhibited c-met phosphorylation in tumor tissues of MHCC97-L and MHCC97-H (Data not shown). Figure 4-3 PHA selectively inhibits c-met + HCC tumor growth. Nude mice bearing xenograft tumors after subcutaneously inoculation of indicated cells were administered 25mg/kg PHA (~) or vehicle (n) daily intravenously for 12days. Representative pictures demonstrate the tumors at the end point (arrows), with tumor volume expressed as the mean 6 SEM.*p < 0.05 (n ¼ 5-8 per group). HY and HD (156)

101 Upregulation of EGFR signaling occurs after c-met inhibition Previous studies showed that c-met inhibition significantly suppresses CSC and mesenchymal characteristics in c-met activated HCC cells in vitro (Chapter 3). Accordingly, our in vivo studies indicate that c-met inhibition can reduce tumor growth; however, tumor xenografts lose responsiveness over time, suggesting that an alternative pathway may be activated for tumor survival (Figure 4-3). We used the stable MHCC97-H c-met shrna cell line to mimic HCC cells treated with c-met tyrosine kinase inhibitor. These cells demonstrated a profile similar to c-met inhibition using a small molecule inhibitor, PHA (Figure 4-2). To investigate potential signaling pathways that may play an important role in cell survival after c-met down regulation, we performed gene expression analysis. Microarray pathway analysis identified seven significant signaling pathways that are up-regulated after c-met knockdown with the EGFR signaling pathway being the highest followed by Wnt and TGF R signaling. Further microarray analysis demonstrates that upon c-met knockdown, EGFR and ERBB3 expression increased significantly (Figure 4-7A and B). Figure 4-4 Down regulation of c-met in MHCC97-H cells increases EGFR signaling (A) Heatmap of EGFR signaling after c-met knockdown. (B) Microarray gene expression profiling demonstrates an increase in EGFR signaling after c-met knockdown. HD

102 sirna screen identifies EGFR as potential by-pass survival mechanism after c-met inhibition To further investigate potential survival compensation mechanisms following c-met knockdown, we conducted an sirna screen targeting 873 kinases and phosphatases using MHCC97-H c-met shrna cells. From the sirna screening, we identified 17 potential targets including EGFR, upon which validation assays were conducted. To validate the sirna screening results, we used a more stringent assay to exclude false-positives. MHCC97-H c-met shrna cells were Table 4-1 Seven targets validated after sirna screening. - HD generated the c-met shrna cell line, the PSU D4 Core Facility performed the sirna screening, WD performed the validation individually transfected with the 17 potential target sirnas in 8 replicates. Successful validation required a significant suppressed cell viability of targeted (p<0.05) compared to scrambled control sirna. As shown in Table 4-1, seven targets from the sirna screen were validated. Based on our microarray and sirna screening analysis, EGFR was chosen for further investigation.

103 Inhibition of c-met results in up-regulation of EGFR/ERBB3 Gene expression profiling and real-time PCR analyses demonstrated that both EGFR and ERBB3 are up-regulated after c-met knockdown in MHCC97-H cells. Previous evidence in non-smallcell lung cancer demonstrates that combined EGFR/ERBB3 and c-met inhibition reverses resistance to EGFR tyrosine kinase inhibition (100). Thus, we wanted to test whether the EGFR/ERBB3 plays an important role in cell survival after c-met inhibition. As shown in Figure 4-5, real-time PCR analysis confirms a significant increase of EGFR and ERBB3 after c-met knockdown. Furthermore, real-time PCR analysis demonstrates a significant increased expression of ERBB3 after PHA treatment compared to vehicle control compared to vehicle control, whereas gefitinib only slightly up-regulates ERBB3 when it is used alone. Combination therapy with Gefitinib and PHA resulted in the suppression of ERBB3 expression whereas EGFR did not demonstrate any significant changes after treatment. Together, our data demonstrate that inhibition of c-met increases an alternative ERBB3 survival pathway that might be inhibited by the EGFR inhibitor, Gefitinib. A B Figure 4-5 Inhibition of c-met increases EGFR signaling (A) Relative mrna expression of EGFR and ERBB3 in MHCC97-H c-met shrna cells normalized to GAPD endongenous control. Data representative of two independent experiments and error bars represent mean ± SEM of triplicates.(b) ERBB3 expression after PHA or Gefitinib treatment. A-HD, B-WD

104 4.4.7 Combination therapy provides additional benefit compared to c-met inhibition alone 91 In order to verify whether combination therapy of c-met and EGFR tyrosine kinase inhibitors have an additive effect, we performed cell viability assays with different dose combinations of c-met inhibitor PHA (0, 1, 5, 10, and 20 μm) and EGFR inhibitor Gefitinib (0, 5, 10, and 20 μm). As shown in Figure 4-6, compared to cells treated with c-met inhibitor alone, the addition of an EGFR inhibitor further suppressed MHHC97-H cell viability.. Figure 4-6 Combination of c-met and EGFR inhibitors has an additive effect on the inhibition of MHCC97-H cell growth in vitro The c-met inhibitor, PHA665752, combined with Gefitinib, an EGFR inhibitor, was used to treat MHCC97-H cells at different doses for 48 hours (The doses of PHA were 0, 1, 5, and 10μM; and the doses of Gefitinib were 0, 5, 10, and 20 μm). Data reported as mean of 8 replicates. WD performed experiment, HD generated figure Combination therapy provides additional benefit to c-met inhibition in vivo Since the combination of EGFR and c-met inhibitors worked well to suppress proliferation in vitro, we aimed to investigate the response of combination therapy using the xenograft tumor model. Athymic nude mice were inoculated subcutaneously with 3x10 5

105 92 MHCC97-H cells and tumors were allowed to grow until approximately 200 mm 3 volume, at which time the treatment was started. As shown in Figure 4-6A, tumor growth was significantly suppressed in both groups after 10 days of treatment. However, c-met inhibition alone was not enough to inhibit tumor growth after ten days of treatment. Combination therapy of Gefitinib and PF (c-met inhibitor) significantly inhibited tumor growth during long term treatment compared to c-met inhibition alone (p<0.05). In addition, as shown in Figure 4-6B, compared to vehicle treatment, Gefitinib significantly inhibited c-met shrna tumor growth; further supporting the hypothesis that EGFR inhibition provides an additional benefit to c-met inhibition. 4.5 Discussion The HGF/c-Met oncogenic pathway is activated in 20%~50% of HCC, and expression levels of both HGF and c-met are highly correlated with poor clinical outcomes of HCC (96-99). Currently, there are several c-met inhibitors in clinical trial for many cancers, including HCC. As described here and in Chapter 3, c-met positive cells respond well to c-met inhibition in vitro. However, in vivo c-met monotherapy using a small molecule inhibitor is not able to eradicate xenograft tumors, indicating that a bypass tumor survival mechanism is involved in the maintenance of tumor growth. To address this question, we established a stable c-met knockdown cell line that mimics MHCC97-H cells treated with c-met inhibitors. sirna screening indicated that EGFR could be a potential target of this compensatory survival mechanism. Both in vitro and in vivo studies further demonstrate that EGFR inhibition provides additional benefit when combined with c-met inhibitor treatment. It has been reported that c-met and EGFR pathways are activated in lung cancer, as circulating c-met and EGFR ligands are highly expressed in these patients and abrogation of both pathways is required for the maximal tumor inhibition (195). In our system, both c-met and EGFR are abundantly expressed in the

106 93 A B Figure 4-7 Combination therapy with c-met and EGFR inhibitors significantly inhibits xenografted tumor growth (A) Primary MHCC97-H cells were subcutaneously transplanted in nude mice. When average tumor volume reached ~200 mm 3, mice were randomly separated into two groups. In group 1, mice (n=4 with 8 tumors) were treated with PHA alone (25mg/kg), and mice in group 2 (n=4 with 8 tumors) were treated with PHA (25mg/kg) and Gefitinib (50mg/kg). (B) MHCC97-H cells stably transfected with c-met shrna were subcutaneously inoculated in the nude mice. When the average tumor volume reached ~200mm 3, mice were separated into 2 groups. In group 1, mice (n=3, number of tumors=6) were treated with vehicle, and mice in group 2 (n=3, number of tumors=6) were treated with Gefitinib (50mg/kg), *p<0.05. A-WD, B-WD and HD human MHCC97-H cell line. However, monotherapy using an EGFR inhibitor has no significant effect on cell survival and tumor only c-met is active independent of HGF; whereas, EGFR

107 94 requires ligand activation. We show that growth. These results suggest that activated c-met signals are predominantly critical for tumor growth. However, c-met inhibitor monotherapy is not enough to inhibit tumor growth; suggesting that there is a compensatory survival mechanism occurring after c-met inhibition. According to our data, c-met inhibition triggered several survival mechanisms that bypass cell death expression including the EGFR signaling pathway. Although EGFR expression was not up-regulated after c-met inhibition by PHA665752, ERBB2 and ERBB3 was increased; leading to the hypothesis that ERBB2 and/or ERBB3 may be required to activate survival mechanism. While EGFR is not highly increased after c-met inhibition, its abundant expression in the MHCC97-H cells may play a role in stimulating downstream signaling by interacting with ERBB2 or ERBB3. Indeed, the EGFR/ERBB3 and ERBB2/ERBB3 dimerization after ligand stimulation has been shown to activate the PI3K/AKT pathway to promote cell survival in breast and lung cancers (100, 196). The compensatory effect after c-met inhibition does not seem to be mediated by EGFR alone, but through another mechanism. One link between the EGFR/ERBB3 and c-met signaling is TGF Consistent with recent findings; we demonstrate that TGF can activate c-met signaling (197, 198). TGF is up-regulated in xenograft tumors after c-met inhibition, suggesting that an autocrine mechanism may be involved in the tumor survival (Appendix: Supplementary Figure S9). However, the paracrine effect from the tumor micro-environment can t be excluded. The mechanism underlying TGF up-regulation will be further investigated in our future work. In this study, we show that after c-met inhibition we show that combination treatment decreased tumor growth significantly. These results suggest that the MHCC97-H cells acquired multiple EGFR mechanisms for tumor survival and, that the inhibition of both c-met and EGFR will be necessary to completely inhibit tumor growth. In conclusion, we show that a compensatory survival mechanism occurs after c-met inhibition through the activation of the EGFR signaling pathway. This report demonstrates the importance of using combination therapy

108 95 and that both c-met and EGFR inhibitors will be required for the complete suppression of tumor growth in c-met + HCC tumors. Future studies elucidating EGFR/ERRB3 compensatory mechanism will be performed.

109 Chapter 5 Overall Discussion and Conclusions

110 Introduction From 1975 to 2006, hepatocellular carcinoma was the only cancer with increasing mortality in the United States (199). Many HCC patients present with locally advanced or metastatic disease, and therefore are not eligible for liver transplantation and have limited therapeutic options resulting in a five year survival rate of less than 12% (199). Thus, it is important to better understand the underlying mechanism of HCC progression in order to design novel therapeutics to improve clinical patient outcomes. The cancer stem cell (CSC) hypothesis suggests that cancers originate from and/or are maintained by a rare neoplastic subpopulation of cells within the bulk tumor. CSCs possess unique survival mechanisms and harbor distinctive stem/progenitor cell properties including the ability to self-renew, differentiate and evade programmed cell death. These characteristics allow for tumor heterogeneity, tumor propagation and metastasize, and, thus, explain why current cytotoxic therapies, which are designed to target only the bulk tumor mass, are not efficacious in the treatment of HCC. For example, sorafenib, the only FDA approved drug for HCC, adds only 2.5 months of survival to HCC patients with advanced/metastatic disease. Ideally, targeting both the bulk tumor mass and the CSC subpopulation would inhibit tumor growth and recurrence. Recent evidence indicates that the epithelial-to-mesenchymal (EMT) program generates cells with CSC characteristics. Increasing evidence suggests that tumor progression is critically involved with the acquisition of an EMT phenotype, which allows polarized epithelial tumor cells to become mobile and interact with the extracellular matrix components to infiltrate surrounding tissues in order to colonize at distant organ sites. Moreover, the induction of EMT by multiple factors (i.e., TGF and HGF) in epithelial cells promotes a CSC phenotype, drug resistance, and invasion and metastasis. The EMT and CSC link suggests that understanding its mechanism can lead to the development of targets that could block metastasis and tumor recurrence, which make

111 up more than 90% of cancer mortality. This, in turn, will improve the overall survival of patients diagnosed with HCC Discussion and Future Directions In Chapter 2 of this dissertation research, we sought to investigate the EMT and CSC link in murine cell lines. One potential link between EMT and CSCs in liver cancer is Transforming Growth Factor beta (TGF. The TGF signaling pathway is known to play an important role in cellular development, cell differentiation, proliferation, and migration (161). Depending on the differentiated state of the target cell, and the dosage of the ligand, TGF signaling can promote or inhibit cell proliferation, apoptosis or differentiation (73, 74). In early stage HCC, TGF induces apoptosis acting as a tumor suppressor; while in late stage HCC, TGF promotes cell survival and proliferation (73). In fact, the dysregulation of the TGF signaling pathway in hepatic progenitor cells leads to the development of HCC, suggesting that liver tumors can be derived from hepatic stem cells; thus supporting a CSC-progenitor origin in HCC. Moreover, in mammary adenocarcinoma cells, TGF can promote tumor growth as well as increase metastatic potential (200). The current CSC paradigm does not define the origin of CSCs. It has been generally accepted that CSCs are derived from adult stem cells. However, little evidence indicates that CSCs can be derived from somatic cells that have accumulated mutations to acquire the selfrenewal capacity. Accordingly, our studies in Chapter 2 indicate that TGF plays an important role in inducing a mesenchymal phenotype with CSC characteristics in P2E cells. Our data support the notion that dysregulated TGF signaling can negatively impact hepatocytes to acquire CSC characteristics, which is an important step for the development of HCC. These data support

112 99 the hypothesis that CSCs can originate from a differentiated cell (i.e., hepatocytes) that acquires self-renewal capability and stemness characteristics through TGF signaling. As we investigated how TGF induces the EMT program in epithelial cells, we discovered that Snail1 played an important role in the regulation of a mesenchymal phenotype. We were able to confirm that TGF induces EMT in epithelial cells through the upregulation of Snail1 and subsequent loss of E-cadherin, as seen in multiple cancers including breast and lung cancers. Moreover, this Snail1-dependent TGF induced EMT increased the expression of CD44 and Nanog, promoted tumorsphere formation, and cell migration and invasion. When we inhibited Snail1 expression in P2M cells, there was a decrease in CD44 and Nanog expression, decreased tumorsphere formation, decreased cell migration and invasion, and increased E- cadherin expression. While Snail1 has been established to regulate EMT through the repression of E-cadherin, not much is known about how Snail1 regulates CSCs. In contrast to the EMT and CSC hypothesis that cells that undergo EMT are quiescent and resistant to cytotoxic treatments, the P2M mesenchymal cells were more proliferative and more sensitive to chemotherapy drugs such as doxorubicin and 5 fluorouracil compared to P2E cells. One possible mechanism for the increased proliferation and sensitivity to cytotoxic treatments in the P2M cells is the high expression of Snail1 and low expression of Abcg2 (201). Acbg2 is an ATP-binding cassette gene that allows drug efflux and, therefore, cells with increased expression of this gene may be more efficacious in effluxing cytotoxic drugs. While Snail1 is largely known to regulate EMT by repressing E-cadherin expression, Snail1 also promotes cell proliferation by inhibiting Cyclin D1 (202). Previous reports have also suggested that, in skin cancer, epidermal keratinocytes with high expression of Snail1 were found to be highly proliferative. This may contribute to the increased proliferation rate of the mesenchymal

113 100 cells compared to the epithelial cells (203). This might be confirmed by performing a cell proliferation assay on stably transfected Snail1-shRNA in the P2M cells. The induction of EMT and acquired CSC phenotype followed a trend where both CD44 and Nanog expression were increased. This led us to ask the question: does Snail1 regulate the CSC phenotype through CD44 or Nanog expression? Nanog is a transcription factor that regulates stem cell differentiation and self-renewal in embryonic stem cells and CSCs (141). TGF signaling has been demonstrated to regulate Nanog and, in Chapter 2, we show that inhibition of Snail1 down-regulates Nanog expression and promoter activity. For the first time, we show that Snail1 regulates Nanog in order to promote a CSC phenotype in vitro. Together, our data indicate that Snail1 plays an important role in inducing both the EMT and CSC phenotype. Figure 5-1 TGF induces EMT and CSC through Snail1. TGF -induced increased Nanog and CD44 expression. Nanog, independent of Snail1 can promote the CSC phenotype. However, when we investigated Snail1 s role in vivo, Snail1 is not enough to inhibit tumor initiation, an important function of the CSC program. Indeed, when we investigated E- cadherin s role in EMT in the P2M cells, we observed that the E-cadherin gene was highly methylated (data not shown). This epigenetic regulation of E-cadherin proved to be Snail1- independent and thus TGF induction of EMT through Snail1 may be transient. This provides insight into why the inhibition of Snail1 was not able to inhibit tumor initiation. Moreover, TGF has been demonstrated to regulate Nanog independent of Snail1 through Smad signaling cascades

114 101 (141). It is therefore possible that Nanog plays an important role in regulating CSCs independent of Snail1. In Chapter 2, the induction of EMT by TGF through Snail1 increased CD44 expression in murine epithelial cells, suggesting that CD44 plays an important role in both the mesenchymal and CSC phenotypes. CD44, a cell surface glycoprotein that is highly involved in cell-cell adhesion and cellular motility, has been utilized as a CSC marker in human breast cancer extensively. CD44 is also constitutively produced in hepatic stellate cells and highly expressed in biliary cells but not hepatocytes. CD44 is the principal mediator of hyaluronic acid (HA)-induced cell migration. Because of its conserved binding site to HA and subsequent interaction with the extracellular matrix, it is likely that CD44 plays an important role in EMT. However, to date, only one report links the functional role of CD44s to EMT (204). In this study, the authors show that CD44 undergoes isoform switching from CD44v (CD44 variants) to CD44s (CD44 standard form) as epithelial cells undergo EMT. Figure 5-2 CD44 isoform switching during EMT The diverse roles of CD44 in different cellular processes are due to its extensive alternative splicing, however, the function of many CD44v isoforms are still unknown. According to our studies and the literature, CD44s regulates the EMT and CSC phenotypes. Consistent with this finding, in Chapter 3 we documented the importance of CD44s in regulating the mesenchymal phenotype (Figure 5-2). It is important to assess the role of CD44s

115 102 in cell migration and invasion, characteristics of EMT. Therefore, it is important to perform scratch assays or the boyden chamber invasion assay on the stable MHCC97-H CD44s shrna cells to investigate whether CD44s is important in cell migration/invasion. Indeed, because the MHCC97-H cells are CD44 +, do not express high levels of E-cadherin, and express abundant levels of mesenchymal markers (i.e., fibronectin), one can speculate that loss of CD44s may inhibit cell migration or invasion. Furthermore, in vivo studies of orthotopic injection of MHCC97-H CD44s shrna cells into the livers of nude mice may demonstrate that CD44s is required for metastasis. As CD44 is widely utilized as a CSC marker, we speculated that CD44 plays an important role in regulating the CSC phenotype. Indeed, when CD44s was knocked-down in MHCC97-H cells, there was a decreased in tumorsphere formation and a decrease in CSC gene expression including Nanog and BMI-1. Moreover, the over expression of CD44s in CD44 neg Huh7 cells promoted a CSC- and mesenchymal-like phenotype. Together, our data suggest that CD44s plays an important role in the CSC phenotype. The importance of CD44 as marker for identifying CSC populations has been demonstrated; however, most reports that define CSC populations with CD44 utilize antibodies that recognizes all CD44 isoforms (108). Therefore, it has not yet been identified which CD44 variants are responsible for the CSC phenotypes. In accordance with the literature, we utilized a CD44 antibody that recognizes all CD44 isoforms to perform flow cytometry analysis on the human HCC cell lines. In this experiment, we observed that more than 95% of the MHCC97-H cells are CD44 +. CD44s does not include variant exons; its molecular weight is ~80kDa whereas CD44v6 has been estimated to be ~160kDa. According to our immunoblot analysis, the MHCC97-H cells express abundant levels of CD44v6 and CD44s proteins, leading to question as to which CD44 isoform is important for the CSC phenotype. According to previous reports, c- Met regulates CD44 alternative splicing through the RAS/MAPK signaling pathway. Indeed, it is

116 103 specifically CD44v6 that interacts with c-met to promote CD44 alternative splicing. Furthermore, the CD44v6/c-Met interaction enhances the RAS/MAPK signaling cascade to promote cell migration (84, 107). Thus, it may be that CD44v6 interacts with c-met to promote the EMT and CSC phenotype in the MHCC97-H cells. However, when we inhibited CD44v6, there was no significant change in tumorsphere formation or mesenchymal phenotypes observed compared to control scrambled sirna. This further confirms our hypothesis that CD44s is an important regulator of the EMT and CSC phenotype. Although CD44v6 does not seem to play an important role in the CSC or EMT phenotype in the MHCC97-H cells, one cannot rule out this possible role in other cells. Notably, CD44v6 presents HGF to c-met by forming a HGF/c-Met/CD44v6 complex that amplifies downstream c-met signaling. Most interestingly, the MHCC97-H cells are activated due to c-met gene amplification and do not express high levels of HGF, which is suggestive that the c-met/cd44v6 may require HGF to signal. Future studies investigating the functional role of CD44v6 will be necessary in order to answer these questions. In the MHCC97-H cells, more than 95% are CD44 + yet only a small number (~2%) are capable of forming tumorspheres. As the antibody utilized for flow cytometry analysis is pancd44, flow cytometry cannot stratify which CD44 isoforms is responsible. This observation, however, suggests that perhaps CD44 is not directly responsible for the self-renewal capacity as seen in the MHCC97-H cells. This explanation is possible because CD44 species have been shown to regulate Nanog expression through the activation of protein kinase C (PKC Hyaluronic acid (HA) CD binding promotes PKC activation, which can lead to increased phosphorylation of NANOG and translocation to the nucleus (138). In the nucleus, the transcription factor Nanog can activate the transcription of multiple self-renewal genes to promote pluripotency. As the liver undergoes fibrosis, hepatic stellate cells synthesize HA, increasing the HA-CD44 interaction and subsequently, increasing the translocation of NANOG into the nucleus (138). Interestingly, in Chapter 2, down regulation of CD44 was followed by the

117 104 loss of Nanog and in human HCC cells, the down regulation of CD44s decreased Nanog gene expression significantly, suggesting a relationship. This observation further helped confirm the preliminary idea that CD44 may be regulating NANOG; however, future studies investigating the CD44 and Nanog relationship needs to be performed in order to confirm these speculations. In chapter 3, we demonstrated the importance of CD44s as a tumor initiator and a regulator of the mesenchymal phenotype in human HCC cell lines. Moreover, we demonstrated that c-met regulates CD44s through the phosphorylation of AKT. In HCC, more than 40% of HCC are c-met positive, and c-met + HCC are associated with a worse prognosis, thus targeting c- Met + HCC would improve survival. While increasing evidence indicates that CD44 + cells are CSCs, the functional importance of CD44 has yet to be elucidated. CD44 s functional role in leukemia has been hypothesized to be as a regulator of leukemic stem cell function such as differentiation and homing (128, 135). In solid tumors, the inhibition of CD44 by sirna in colorectal primary cells prevented clonal formation and inhibited tumorigenicity in xenograft models (136). Together, these data suggest that CD44 contributes to the activation of the stem cell regulatory genes and can itself be a target of these genes and therefore, may functionally be important for the maintenance of CSCs. CD44 is involved in many signaling pathways that regulate stem cell characteristics such as the Wnt pathway, indirect regulation of Nanog, TGF and the microenvironment (108, 138). Nanog, a transcription factor involved in stem cell self-renewal can promote transcriptional regulation of many stem cell genes (139). In this work, we show that CD44s regulates tumor initiation and CSC-related genes including Nanog. While the evidence for the CD44 variant s functional role in CSCs is lacking, it is clear that CD44s has an important role in CSCs. In this report, we provide evidence that CD44s signaling is important in maintaining a CSC phenotype. Our data suggest that cell surface markers such as CD44 may be robust markers for identifying CSCs due to their capability to signal and regulate CSC-related genes. Other

118 105 markers such as CD133 and ABCG2 have also been implicated to be highly regulated and functionally responsible for CSC characteristics (143, 205). These reports suggest that cell surface markers such as CD133, CD44 and ABCG2 are robust cell surface markers because of their function which may be necessary for the survival of the CSCs. Therefore, by elucidating the CSC signaling mechanism and the function of CSC markers, we would be able to better design pharmacotherapeutics to inhibit CSCs. In addition to its important physiological functions, the proto-oncogene Met is a master regulator of metastasis, tumor invasion and angiogenesis (85, 86). In various types of cancers, including gastric, and small cell lung cancer, c-met mutation, amplification and over-expression have been associated with a metastatic phenotype (87). In adult livers, epithelial-to-mesenchymal transition (EMT) is promoted by HGF/c-Met signaling to regulate mitogenesis and fibrosis, and is important in liver regeneration and repair (80, 81). Besides its functional role in repair, regeneration and tumorigenesis, c-met also plays an important role in cancer stem cells. In response to HGF, normal stem cells in the embryo express c-met and migrate to accomplish their developmental fate (88, 101). Due to these characteristics, it is likely that cancer cells activate the HGF/c-Met signaling axis to switch to an invasive program. Indeed, the EMT program has been proposed by Weinberg and Theiry to be a critical step towards metastasis (56, 58, 206). More recently, studies indicate that the EMT program generates cells with acquired CSC characteristics and that CD44s is critical in EMT (44, 62, 204). Furthermore, the induction of EMT increased CD44 + CSCs, suggesting that CD44 may be the link between the EMT and CSC phenotypes. In this report, we demonstrate that c-met regulates CD44s and that the down regulation of CD44s leads to the loss of fibronectin and increased expression of E-Cadherin, hallmarks of EMT. Notably, the down-regulation of CD44s decreased expression of CSC-related genes and tumor initiation. Furthermore, the over expression of CD44s in Huh7 cells increased CSC-related genes, increased fibronectin protein expression and decreased E-cadherin. Our study demonstrates

119 106 CD44s is a key regulator of the EMT and CSC phenotype in human HCC cells. Furthermore, we demonstrate that the importance of c-met as a regulator of CD44s and that this regulatory mechanism may be important for cell survival and metastasis in c-met + HCC. Future studies will be performed to test the role of CD44s in metastasis. 5.3 Clinical Implications In Chapter 4, we investigated c-met as a potential therapeutic marker in HCC. By targeting c-met, we would subsequently target the CSC and EMT phenotype induced through its regulation of CD44s (Figure 5-3). The tight regulation of HGF/c-Met signaling that is observed in development and regeneration is lost in cancer at multiple levels. These changes in c-met often involve transcriptional deregulation, genetic abnormalities such as gene amplification, crosstalk with other signaling pathways such as EGFR (where HGF can activate EGFR signaling), and activation caused by constitutive co-receptor binding (i.e., CD44s or CD44v6). This wide variety of HGF/c-Met dysregulation occurs in multiple cancers including breast, liver, lung, ovary, kidney, and thyroid (207). In tumor cells, the HGF/c-Met axis activates multiple signaling cascades including migration/invasion, anti-apoptosis, and cell proliferation. In addition to these diverse signaling cascades, we demonstrate that c-met promotes tumor initiation through CD44s. We thus propose that c-met represents a potential therapeutic target for personalized HCC treatment. Indeed, when we treated c-met + MHCC97-L and MHCC97-H cells with PH665752, a selective c-met inhibitor, there was significant tumor inhibition compared to c-met neg Huh7 and Hep3B cells. As described in this dissertation, c-met positive cells are responsive to c-met inhibition in vitro and the inhibition of c-met leads to subsequent inhibition of CSC and EMT (Chapter 3 and Figure 5-3).

120 107 Figure 5-3 Clincal implications for targeting c-met + HCC cells By targeting c-met+ CSCs which are also CD44+, the tumor CSC and EMT program is inhibited and tumor is eliminated. However, c-met+/cd44+ cells are resistant to standard cytotoxic chemotherapy, thus tumor recurrence occurs. Inhibition of c-met leads to down-regulation of both the MAPK/ERK1/2 and PI3K/AKT signaling pathways, both important for tumor cell survival and the EMT/CSC phenotype. However, in vivo c-met monotherapy using a small molecule inhibitor was not able to eradicate xenograft tumors completely. We therefore suspected that a bypass tumor survival mechanism may be involved in the maintenance of tumor growth after c-met inhibition. To investigate the survival mechanism, we performed high throughput sirna screening, which indicated that EGFR was significantly increased after c-met inhibition. Previous reports indicate that EGFR/ERBB3 (Her3) signaling is activated in non small cell lung cancer and that selective inhibition using Gefitinib, an EGFR inhibitor, can reduce tumor growth significantly (195). However, the vast majority of these tumors become resistant due to secondary EGFR (T730M) mutations. Additionally, c-met gene amplification can also promote resistance by interacting with ERBB3, resulting in marked phosphorylation of ERBB3 (100). One link between the c-met and EGFR crosstalk in HCC is transforming growth factor (TGF ) (197). We showed that TGF increased significantly after c-met inhibition, suggesting that the activation of the

121 108 EGFR/ERBB3 signaling may be through the increased expression of TGF. This link will need to be further investigated in order to confirm the regulatory mechanism of tumor survival after c- Met inhibition. In the MHCC97-H cells, c-met is constitutively activated and EGFR levels are high but is not active (no baseline phosphorylation signal). Another question arose as to why there are an abundance of EGFR levels but no activity. We speculate that c-met inhibition leads to the activation of ERBB3, which have been documented to interact with EGFR to activate MAPK/ERK1/2 to promote cell survival. We also show that combination treatment decreased tumor growth significantly. These results suggest that c-met activation is critical for tumor progression and that EGFR/ERBB3 can act as a bypass mechanism for tumor survival and that the inhibition of both c-met and EGFR will therefore be necessary to completely inhibit tumor growth. Figure 5-4 Combination therapy with EGFR and c-met inhibitors inhibits HCC tumor c-met inhibition alone is not enough to completely inhibit tumor growth due to EGFR activation following c-met inhibition. Currently, there are several c-met inhibitors in clinical trials for various cancers, including HCC (171, 207). To date, there are twelve c-met inhibitors that are being used in clinical trials, however, there are only four HCC clinical trials that are in Phase I/II (207). Of the four c-met inhibitors in clinical trials being used to treat HCC, only one is a selective c-met kinase inhibitor. Notably, PHA is a selective c-met kinase inhibitor that is not being used in clinical trials. The HGF/c-Met axis offers a potentially high-value target for HCC cancer drug

122 109 development. In HCC, the HGF/c-Met axis has been demonstrated to be over-activated and c-met mutation has been shown to drive hepatocarcinogenesis; therefore patients who are c-met + can benefit from c-met targeted therapies. Furthermore, EGFR and c-met inhibitors in combination can have synergistic effects, as demonstrated here, to be of benefit to patients with resistance to tyrosine kinase inhibitors (either after c-met or EGFR inhibition) and offer benefits that exceed the inhibition of either target alone. 5.4 Limitations In chapter 2, we utilized a mouse PTEN knock out HCC cell line to study the EMT and CSC link. In humans, PTEN loss is highly associated with liver cancer and the deletion of PTEN is found in more than 40-50% of HCCs; therefore, this is a relevant mouse model for studying hepatocarcinogenesis ( ). While PTEN is an important gene in HCC development, the mouse system is not a good preclinical model for human cancers. For example, multiple studies have demonstrated that a specific drug therapy for selected cancers worked well in animals but were found to be ineffective in clinical trials in humans. Another important limitation in this system is that the P2E and P2M cells are derived from an artificial PTEN knockout cell line generated from multiple generations of xenografted tumors. The P2E and P2M cells were generated by PTEN -/- liver tumors which were minced, cultured in vitro, and subcutaneously inoculated into athymic nude mice for two generations. This process can cause stress signals that can induce genetic or epigenetic mutations that allow the cells to survival under in vitro conditions. A good example is when P2E cells are passaged; the gene expression levels of HGF are lost (data not shown). Therefore, these cells may have multiple confounding factors that are artificial and we should be cautious in interpreting the results. Another important factor is that more than 85% of HCC patients are HBV or HCV positive. In this system, the PTEN loss induces

123 110 inflammation followed by liver carcinogenesis. One may speculate that HBV or HVC infection induced-chronic injury may play an important role in PTEN mutations observed in 40-50% of HCC. As this hypothesis needs further investigation, we must appreciate that the PTEN -/- mouse model has multiple limitations. Due to the limitation of using mouse models to mimic the preclinical HCC seen in humans, we used human five different HCC primary cell lines to study the CSC and EMT link. It is important to note that the primary HCC cell lines, themselves, have limitations. These cell lines 1) are immortalized and have been passaged for many decades and therefore may not represent the true phenotype as presented in the patient, 2) have the same limitations as the murine cell lines in that there are multiple confounding factors that may attribute to the phenotype seen in vitro and 3) as they represent a human HCC tumor population, they are only a very small representative population of the heterogeneous HCC population seen in patients. In order to surpass these barriers, one would work with primary HCC tumors which is very challenging technically. However, given the ethical and technical challenges with in vivo studies, it is best to confirm in vitro phenomenon in human clinical samples. To confirm the c- Met/CD44s relationship, we performed immunoblot and immunohistochemistry analysis on 101 different HCC patient samples. Although clinical information such as survival data to demonstrate the importance of c-met in HCC progression are not available, we were able to show that 40% of HCC samples were c-met and CD44s positive Another important limitation in this dissertation is the in vivo experiments. These experiments are performed subcutaneously and do not take into consideration the liver vs. subcutaneous microenvironment. It is very likely that the microenvironment plays a major role in the promotion of carcinogenesis as well as has a large impact on the survival of CSCs and the induction of EMT. Although these limitations exist, our in vivo studies only measure tumor initiation. Our limited dilution xenograft studies demonstrate that the loss of CD44s reduces

124 111 tumor incidence in low cell dilutions (as low as 100 cells) compared to scrambled control. These data clearly demonstrate the importance of CD44s as a tumor initiator. The microenvironment may also be an important factor when studying the EMT and CSC phenotypes. The microenvironment supports the CSC by providing a CSC niche that is rich in growth factors and supporting proteins such as fibronectin, collagen, and HGF (51). The CSC niche controls their self-renewal and differentiation (211). Furthermore, CSCs can be generated by the microenvironment through induction of CSC features in more differentiated tumor cells. In addition to a role in CSC maintenance, the microenvironment is hypothesized to be involved in metastasis by induction of the EMT, leading to dissemination and invasion of tumor cells (108, 211). The localization of secondary tumors also seems to be orchestrated by the microenvironment, which is suggested to form a pre-metastatic niche fitting the seed and soil hypothesis (206, 212). Thus, the microenvironment seems to be of crucial importance for primary tumor growth as well as metastasis formation; therefore, is an important factor to consider when performing in vivo studies. Importantly, CD44 and c-met both interact with the extracellular matrix in order to activate downstream signaling (108). For example, in colorectal carcinoma, HGF production by the microenvironment activates c-met signaling and HA, which is synthesized during liver fibrosis or injury and can interact with CD44 (213). Moreover, in the liver, CD44 is abundantly expressed in hepatic stellate cells and biliary epithelial cells. Although the CSC hypothesis remains controversial, recent work establishes a connection between epithelial-mesenchymal-transition (EMT) and a CSC-phenotype (44, 140). EMT is a critical developmental process that plays a central role in the formation and differentiation of multiple tissues and organs. During EMT, epithelial cells lose cell-cell adhesion and apicalpolarity, and acquire mesenchymal features, such as motility, invasiveness, and resistance to apoptosis (56). The EMT/CSC link plays a critical role in tumor initiation, tumor propagation, and metastasis and tumor recurrence. By understanding the mechanism as to how cancer cells

125 112 utilize the CSC/EMT program to survive, newly targeted therapeutics can improve patient survival. While EMT has been studied extensively in vitro and some in vivo studies demonstrate the significance of EMT as an important early step in metastasis, the EMT concept is controversial. One of the leading criticisms of the EMT hypothesis is that pathologists have combed through millions of human tissue sections from all tumor types without seeing cells in transition. This has led many skeptics to be concerned that the hypothesis provides false hopes of some day blocking metastasis, which makes up more than 90% of cancers deaths worldwide. Skeptics suggest that cells metastasize due to mutations in the cell-cell adhesion program or break off from the tumor in clumps and travel in packs. However, champions of EMT suggest that tissue sections are snapshots of a moment of the cancer progression and thus you cannot see the transient EMT process. One key experiment to resolve some of these would be to track individual cancer cells from the primary tumor until colonization. However, this is technically challenging but would answer how cancer cells dispatch from the primary site and metastasize to distant organ systems. Another criticism is that markers that are used to identify EMT also are expressed under other important cellular processes such as cell cycle regulation and immune regulation (214). For example, Snail1 has been described by multiple groups to play an important role in EMT by the dysregulation of E-cadherin. However, during acute inflammation, Snail1 recruits macrophages to promote wound healing (214). Also, Snail1 plays an important role in regulating cell cycle by repressing transcription factors to regulate the early to late G1 transition and the G1/S checkpoint (214). This also leads to question whether Snail1 is a good marker for targeting neoplastic EMT. Thus, it is important to design drug targets that are specific to tumor cells, which is challenging but will decrease side off target side effects. Another marker that is involved in EMT but is controversial is fibronectin. During fibrosis, fibronectin is highly expressed in order to activate stellate cells. It is then possible that fibronectin is not a good marker of EMT in that it

126 113 may be a residual effect of fibrosis (215). Together, these data suggests a need for more than one marker to be used in order to identify EMT and careful design of drug therapeutics that can inhibit the neoplastic type I EMT and spare the physiological type I and II EMT.

127 114 Appendix Supplementary Data The supporting material is not essential for the inclusion of chapter 4 but is presented here to further enhance and support the understanding of the conclusions presented in this thesis dissertation.

128 115 A B Cirrhosis HCC Supplementary Figure 1 CD44s and c-met co-expression in Human HCC samples. (a) Immunblot of HCC samples S8-S33 of CD44s, CD44v6, c-met, p-c- Met Y1234/Y1235 and p-c-met Y1349. (b) H&E images of human HCC samples (400X). Arrow points to hepatocytes with abundance CD44 expression.

129 116 Supplementary Figure S2 CD44 and c-met co-localized in human HCC samples. Confocoal images of immunoflourescence stained for CD44 and c-met in human HCC S34 sample. Images are at 430X magnification.

130 117 A B C Supplementary Figure S3 Characterization of human HCC cells. (a) Flow activated cytometry of human HCC cells for CD44 and c-met. Data represent triplicates and experiment was performed two independent times. (b) Cell viability assay of HCC cells after 24 hours of sorafenib or doxorubicin treatment at indicated doses. Data represent two independent experiments and are shown as mean ± SEM of 8 replicates, *p<0.05. (c) Immunoblot analysis for apoptosis after 24 hours of doxorubicin treatment at 2.5ng/ml.

131 118 Supplementary Figure S4 CD44s, not CD44v6, is significantly decreased after c-met inhibition. PHA665752, a c-met inhibitor, inhibits CD44s expression. Supplementary Figure S5 c-met amplification in MHCC97-H cells. Confirmation of c-met status by quantitative PCR c-met is amplified compared to Huh7 HCC cells. PCR was performed for c-met status in MHCC97-H and Huh7 cells normalized to methylenetetrahydrofolate reductase endogenous control. Gels were analyzed using Image J (Bethesda, MD).

132 Supplementary Figure S6 CD44s and c-met interact in MHCC97-H cells. (A) CD44s IP and c-met IP immunoblot. (B) Confocal microscopy of MHCC97-H cells at 430X magnification. Arrow points to co-localization of c-met and CD44 at the cell membrane. 119

133 120 Supplementary Table S8 Iummunoprecipitation analysis of proteins identified in both CD44 and c-met IPs. Identified Proteins in BOTH c-met and CD44 IP Hepatocyte growth factor receptor CD44 antigen Isoform 6 of Myoferlin Transferrin receptor protein 1 Endoplasmic reticulum metallopeptidase 1 Stress-70 protein, mitochondrial Homo sapiens Isoform Alpha-6X1A of Integrin alpha-6 Isoform C of Prelamin-A/C Canalicular multispecific organic anion transporter 1 NAD(P) transhydrogenase, mitochondrial Isoform 2 of ATP-binding cassette sub-family G member 2 Isoform 2 of Transmembrane and TPR repeat-containing protein 3 V-type proton ATPase 116 kda subunit a isoform 2 Aspartyl/asparaginyl beta-hydroxylase Integrin beta-1 Nucleolin Moesin Trifunctional enzyme subunit alpha, mitochondrial Neutral alpha-glucosidase AB Annexin A2 H(+)/Cl(-) exchange transporter 7 Niemann-Pick C1 protein Nodal modulator 3 Isoform 2 of 4F2 cell-surface antigen heavy chain Protein sel-1 homolog 1 Succinate dehydrogenase [ubiquinone] flavoprotein subunit, mitochondrial Isoform 3 of E3 ubiquitin-protein ligase synoviolin Isoform 2 of Golgi apparatus protein 1 Importin-7 Syntaxin-binding protein 3 Large subunit GTPase 1 homolog Isoform 2 of Long-chain-fatty-acid--CoA ligase 5 Symbol MET CD44 MYOF TFR1 ERMP1 GRP75 ITA6 LMNA MRP2 NNTM ABCG2 TMTC3 VPP2 ASPH ITB1 NUCL MOES ECHA GANAB ANXA2 CLCN7 NPC1 NOMO3 4F2 SE1L1 DHSA SYVN1 GSLG1 IPO7 STXB3 LSG1 ACSL5

134 A 121 B Supplementary Figure S7 CD44s shrna cells demonstrate a slight decrease of CD44v6 expression. (A) Immunoflourescence microscopy of MHCC97-H CD44s shrna cells compared to scrambled control at 430X magnification. (B) CD44s shrna cells demonstrate a slight decrease of CD44v6 expression.

135 122 A B Supplementary Figure S8. Cd44v6 does not play an important role in tumor initiation. (a) Immunoblot of c-met, CD44v6 and c-met downstream signaling in CD44v6 sirna compared to Scrambled sirna control in MHCC97-H cells. MHCC97-H cell lysates were collected after 48 hours of transfection. (b) CD44v6 does not regulate tumorsphere formation. After 24 hours of transfection, 1X10 3 cells were plated in a 6-well low adherent culture plate for two weeks. Data represent two independent experiments and are shown as mean ± SEM of triplicates. Supplementary Figure S9. CD44v6 does not play an important role in tumor initiation. Relative TGF mrna expression of MHCC97-H c- Met shrna compared to Scrambled control. Supplementary Figure S10. CD44s expression is highly expressed in tumorspheres in SK-Hep1 cells. Protein expression of CD44s in SK- Hep1 cells. Adh. = adherent cultured cells.