tmpD2.tmp.pdf

of warfarin-mediated inhibition of the receptor tyrosine kinase AXL Thesis for the degree of philosophiae doctor (PhD) at the University of Bergen

2 Copyright Gry Sandvik Haaland The material in this publication is protected by copyright law. Year: 2017 Title: Author: Print: Investigations of the cancer therapeutic and protective effects of warfarinmediated inhibition of the receptor tyrosine kinase AXL Gry Sandvik Haaland AiT Bjerch AS / University of Bergen

3 Scientific environment This work was performed at the Department of Biomedicine, Centre for Cancer Biomarkers, Faculty of Medicine, University of Bergen, during the period of 2011-2017. The work has been conducted with Professor James B. Lorens as main supervisor and Professor Oddbjørn Straume as co-supervisor. From August 2012-December 2012, I had a predoctoral fellowship at UT Southwestern, Dallas, Texas, USA, in the research lab of Dr. Rolf Brekken. The Faculty of Medicine provided financial support for the PhD-fellowship. The experimental work was supported by the Faculty of Medicine, Helse-Vest, and the Research Council of Norway through its Centers of excellences funding scheme.

4 Acknowledgements First of all, I would like to thank my supervisor James B. Lorens for giving me the opportunity to do this work, and for letting me be a part of his inspiring research group for all this time. Thank you for your energy, and for an admirable ability to always see things from the positive side. Also, thank you for many interesting scientific discussions, and for believing in and letting me go through with my ideas. I would also like to thank my co-supervisor Oddbjørn Straume, for help and advice during my project, and for the encouragement and support to conduct the register based part of this work. Thanks to Dr. Rolf Brekken for letting my stay in his lab for 4 months during this work. I really appreciate the opportunity to experience a research environment of your caliber, and to be part of your group for this time. I would also like to thank all the members of the Lorens laboratory (previous and current), for making such a positive work environment. A special thanks to Sissel Vik Berge for always knowing the answer to every question, and keeping track of everything in the lab. Also, a special thanks to Kjersti, my office-mate for the last six years. Thank you for all the laughs, coffees and shared frustrations, you have made this work bearable, even on the most challenging days. Thanks to Stefan and Kristina, for lifesaving in Dallas, for lunches and friendship. A special thanks to my friends, Line and Solveig, for all the talks and laughs during the years, and to Ingrid and Helene, for dedicated time, also when schedules are busy. This work has been easier knowing I have friends like you. Thanks to my parents, Inger and Svein for always believing in me, no matter what. Also thanks to my sister Marte, for being constantly supportive and for proofreading the thesis. To my husband, Helge, thank you for the constant encouragement, patience, love and support, this would not have been possible without you. And last but not least, thank you to Sverre and Vilde. You make me keep focus on what is important in life.

5 Abbreviations AC ACC AML ATP BAD Apoptotic cells Acinar cell carcinomas Acute myeloid leukemia Adenosine triphosphate Bcl- 2 associated death promoter Bcl-2 B cell lymphoma 2 Bcl-XL BCSC BRAF B cell lymphoma extra large Breast cancer stem cells B-Raf proto-oncogene serine/threonine kinase BRCA2 Breast cancer 2 C1-TEN Cbl CDKN2A CI CLL CRN CSC C1 domain containing phosphatase and TENsin homologue Casitas B-lineage lymphoma Cyclin dependent kinase inhibitor 2A Confidence Interval Chronic lymphatic leukemia Cancer registry of Norway Cancer stem cell DKK3 Dickkopf related protein 3 DPC4 Deleted in pancreas cancer 4 E-cadherin ECM EGF EGFR EMT ERK Epithelial cadherin Extracellular matrix Epidermal growth factor Epidermal growth factor receptor Epithelial to mesenchymal transition Extracellular signal regulated kinase

6 FGF Fibroblast growth factor GAS6 Growth arrest spesific 6 GEMM GGCX Genetically engineered mouse models Gamma-glutamyl carboxylase GSK3 Glycogen synthase kinase 3 HGF HIF ICD ICD-O-3 IFNAR IG IGF IPMN IRR KO KRAS LMWH MAPK MET mir MK MMP N-cadherin NF-κB NK cells NOAC NorPD Hepatocyte growth factor Hypoxia induced transcription factor International classification of diseases International classification of diseases for oncology Interferon α/β receptor Immune globuline Insulin like growth factor Intraductal papillary mucinous neoplasm Incidence rate ratio Knockout Kirsten rat sarcoma viral oncogene homolog Low molecular weight heparin Mitogen activated protein kinase Mesenchymal to epithelial transition MicroRNA Menaquinone Matrix metallo- proteinases Neural Cadherin Nuclear factor κb Natural Killer cells Non-vitamin k anticoagulant Norwegian Prescription database

7 PAK1 PanIN PI3K P-21 activated kinase Pancreatic intraepithelial neoplasias Phosphoinosytol 3 OH kinase PD1 Programmed Death Protein 1 PDAC PNET PtdSer ROS RTK saxl Pancreatic ductal adenocarcinoma Pancreatic neuroendocrine tumors Phosphatidylserine Reactive oxygen species Receptor tyrosine kinase Soluble AXL SMAD4 SMAD family member no 4. SOCS Sp TAM TERT TLR TGF-β TNFα TNM VEGF-A VHL VKOR VKDP VSMC ZEB Suppressor of cytokine signaling specificity protein Tyro, AXL, Mer Telomerase reverse transcriptase Toll-like receptor Transforming growth factor β Tumor necrosis factor α Tumor, Node, Metastasis Vascular endothelial growth factor A Von Hippen Lindau protein Vitamin K epoxide reductase Vitamin K dependent proteins Vascular smooth muscle cells Zinc finger E-box binding homeobox

8 Abstract Cancer is a major health issue all over the world. Cancer related deaths are one of the major causes of deaths, and are in > 90 % of the cases related to metastatic development, and spread of the cancer outside the primary location. The receptor tyrosine kinase AXL is closely associated with the development of cancer and the receptor is upregulated in many different cancer forms. Upregulation is associated with increased invasiveness, and poor overall survival. Warfarin is a known anticoagulant, which also is suitable as an AXL inhibitor. The warfarin-mediated inhibition of AXL is through the depletion of Vitamin K, with a subsequent inhibition of the γ-carboxylation of the Vitamin K dependent proteins in the body. GAS6, the ligand of AXL, is vitamin K dependent and will be unable to activate the receptor following warfarin treatment. In this thesis, we have worked with the warfarin-mediated inhibition of the receptor tyrosine kinase AXL. In five different mouse model systems, we have evaluated the effect of warfarin-mediated AXL-inhibition in the development and metastasis of pancreatic ductal adenocarcinoma. We also evaluated how warfarin-mediated AXL inhibition impacts on expression of EMT markers, and the ability of the cells to migrate and form colonies, which is a hallmark of cancer with metastatic properties. Further, we performed a register based cohort study using the Norwegian National Registry, the Cancer registry of Norway and the Norwegian prescription database. We investigated the cancer incidence in warfarin users compared to non-users in a broad segment of the Norwegian population, with a cohort comprising 1,2 million persons aged 52-82 years. Our work establishes AXL as an important driver of metastatic formation in pancreatic ductal adenocarcinoma. The level of metastatic disease were significantly reduced in all warfarin treated animals. We also confirmed the close relation between AXL and EMT, as epithelial markers were upregulated when AXL was inhibited. The cells migratory and colony forming capabilities were also impaired after AXL inhibition. In the population-based register study we observed an overall cancer protective association, with lowered incidence rate ratio of cancer in warfarin users compared to non-users. This was observed both for all-site cancer, and for the most prevalent cancer diagnoses in the material.

Altogether, our results emphasizes the importance of the receptor tyrosine kinase AXL in the development and progression of cancer. Warfarin-mediated AXL inhibition are shown to have a cancer protective effect, both in murine model systems and in population level studies. The results from this thesis suggest further investigations, to fully illuminate the potential use of warfarin in an anti-cancer setting. 9

10 List of publications I Kirane, A.*, Ludwig, K.*, Sorelle, N., Haaland, G., Sandal, T., Ranaweera, R., Toombs, J., Wang, M., Dineen, Sean., Micklem, D., Dellinger, M., Lorens, J.B., Brekken. R.A. Warfarin blocks GAS6-mediated AXL activation required for pancreatic cancer epithelial plasticity and metastasis. Cancer Res 2015; 75(18); 3699-705. *Authors contributed equally to this work. II Haaland, G.S., Falk, R.S., Straume, O.*, Lorens, J.B.* Lower overall cancer incidence in patients treated with warfarin: A prospective population-based cohort study. (Manuscript submitted) * Authors contributed equally to this work. Other contributions not included in the thesis: Kjersti T. Davidsen, Gry S. Haaland, Maria K. Lie, James B. Lorens, Agnete S.T. Engelsen The role of AXL receptor tyrosine kinase in tumor cell plasticity and therapy resistance. (Chapter 15 in the book Biomarkers of the tumor environment. In press, Springer International publishing)

11 Contents Scientific environment... 3 Acknowledgements... 4 Abbreviations... 5 Abstract... 8 List of publications... 10 1. Introduction... 13 1.1 Cancer... 13 1.2 Pancreatic cancer... 14 1.2.1 Other pancreatic tumors... 15 1.2.2 Staging of pancreatic cancer... 16 1.2.3 Treatment of pancreatic cancer... 16 1.3 Tumor Biology... 18 1.3.1 Mechanisms for metastases... 21 1.4 Epithelial to Mesenchymal transition... 24 1.4.1 Activation of EMT... 25 1.4.2 EMT control... 27 1.4.3 EMT and Cancer... 27 1.5 Receptor tyrosine kinases... 29 1.5.1 AXL receptor tyrosine kinase... 30 1.5.1.2 AXL structure... 30 1.5.1.3 AXL ligand... 31 1.5.1.4 AXL activation... 32 1.5.1.5 Downstream events of AXL... 34 1.5.1.6 AXL regulation... 37 1.5.1.7 AXL in normal physiology... 38 1.5.1.8 AXL and EMT... 40 1.5.1.9 AXL and cancer... 40 1.5.1.10 AXL and drug resistance... 44 1.5.1.11 AXL and Immunotherapy... 44 1.5.1.12 AXL and cancer stem cells... 45 1.6 Vitamin K... 46 1.7 Warfarin... 48 1.7.1 Cancer protective effects of warfarin in a historical perspective... 48

12 1.7.2 AXL and warfarin... 49 1.8 Health registries... 50 1.8.1 The cancer registry of Norway... 50 1.8.2 The Norwegian prescription database... 51 2. Aims of the study... 52 3. Summary of papers... 53 4. Methodological considerations... 55 4.1 Animal experiments... 55 4.1.1 Cell line xenograft models... 55 4.1.2 Syngeneic models... 55 4.1.3 Genetically engineered mouse models... 56 4.2 Mouse strains in use in our work... 56 4.3 In vivo experiments... 58 4.3.1 Medical treatment of animals... 58 4.3.2 Measurements of primary tumor burden and metastases... 59 4.4 Induction of EMT... 59 4.5 Register study... 59 4.5.1 The coupling process of different registries... 60 4.6 Statistics... 61 5. Discussion... 62 5.1 The role of AXL in the development and metastasis of pancreatic ductal adenocarcinoma.. 62 5.2 The role of EMT in warfarin-mediated AXL-inhibition in pancreatic cancer... 66 5.3 Vitamin-K in cancer... 67 5.4 Warfarin use and cancer incidence... 68 5.6 Warfarin in the era of Non-vitamin K anticoagulants... 71 6. Concluding remarks... 72 7. Future perspectives... 73 8. References... 75

13 1. Introduction 1.1 Cancer The term cancer describes a diverse group of diseases. These diseases can present very differently, but share common features of uncontrolled cell division, and the ability of metastatic dissemination. The term malignant is used when cells in a tumor has the ability to invade either nearby, or distant tissues. 1 Cancer is a major health problem throughout the world. In Norway, there were 32,592 new cases of cancer reported in 2015. As shown in figure 1, the three most frequent cancer sites for men were prostate, lung and colon, and for women they were breast, colon and lung. Figure 1: The most frequent cancer types in Norway 2011-2015. Adapted from 2 In 2014, 10971 cancer deaths were reported, and death from cancer was the second most common cause of death after heart-diseases. For men, lung cancer (1198) and prostate cancer (1093) are the more frequent causes of cancer death, and lung (960), breast (663) and colon (595) are the most frequent in women. 2

14 Figure 2: Age-standardized (Norwegian standard) mortality rates per 100 000 person-years for selected cancers in Norway, 2014. Adapted from 2 1.2 Pancreatic cancer Pancreatic cancer is the fourth leading cause of cancer death in Norway, with a 5-year relative survival of 6.4%. 2 Pancreatic cancer is the 12th most frequent cancer worldwide, but the high overall mortality, with 330,000 deaths in 2012 makes it to the seventh most leading cause of cancer death. 3 There are several risk factors for developing pancreatic cancer, with increasing age as the major one. The cancer form is rarely seen before the age of 40 years, and the risk is 40 times increased at the age of 80 years. Family members of patients with pancreatic adenocarcinoma has an approximately threefold risk of developing the disease, suggesting a genetic inheritance. 4 Also increasing body mass index, new onset diabetes mellitus, chronic pancreatitis, and smoking are factors known to increase the risk. 5 The most common form for pancreatic cancer is pancreatic ductal adenocarcinoma (PDAC), with over 85% of all pancreatic neoplasms being of this origin. 4 These tumors are located in the head of pancreas in 65% of the cases, and tumors with this localization is normally presenting earlier than other localizations, mainly with symptoms as acute pancreatitis, jaundice and/or biliary obstruction. 5 Several mutations are linked to the progression of PDAC. The most described ones are activation of the oncogene VI-KI- RAS2 Kirsten rat sarcoma viral oncogene homolog (KRAS), followed by inactivation of several tumor suppressor genes, such as cyclin dependent kinase inhibitor 2A (CDKN2A),SMAD family member no 4( SMAD4)/Deleted in pancreatic cancer -4 (DPC4),

15 Tumor protein p53 (TP53) and Breast cancer 2 (BRCA2). 4,6 In most cases, PDACs evolve through non-invasive precursor lesions, so-called pancreatic intraepithelial neoplasias (PanINs). These are microscopic lesions, (<5mm), and not detectable by non-invasive imaging. The PanINs are graded from 1 to 3 after level of cellular atypia. Low-grade PanINs (PanIN1) are increasingly common with increasing age. High-grade PanINs (PanIN3) are usually found together with invasive cancer. 6 The genetic alterations associated with invasive cancer are also found in PanINs and the prevalence of these alterations will increase corresponding to the cytological and architectural atypia in the PanINs. KRAS gene mutations are normally one of the first alterations to be present in these lesions, and are increasingly frequent with the development of more advanced disease. At the stage of PDAC, nearly 100% of the tumors present with KRAS mutations. 4,7 1.2.1 Other pancreatic tumors Neuroendocrine tumors Pancreatic neuroendocrine tumors (PNET) are rare, representing 1-2% of all pancreatic tumors. They originate from pluripotent cells in the pancreatic ductal/acinar system. They could be secreting biologically active hormones, or they could be non-functional (60-90%). Hormone secreting tumors can give many different clinical syndromes, where hyperinsulinemia (insulin-secreting tumors), and Zollinger-Ellisons syndrome (gastrinsecreting tumors) are the most common forms. 8 The malignant potential of these tumors vary from slow-growing tumors with non-invasive growth, to invasive and metastatic tumors. 9,10 Acinar cell carcinoma Acinar cell carcinomas (ACCs) are rare neoplasms, accounting for 1-2% of all pancreatic tumors. 11 These tumors produce high amounts of digestive enzymes, which can give symptoms as skin rashes and joint pain. 12 The distinction between ACCs, and PNET can be unclear, and it is shown that one third of the ACCs have a neuroendocrine component. 11

16 Cystic neoplasms of pancreas Malignant cystic tumors of pancreas represent 3-4% of pancreatic neoplasms and the most common forms are mucinous cystic neoplasms, serous cystic neoplasms and intraductal papillary mucinous neoplasms (IPMN). Mucinous cystic neoplasms are the most frequent type, representing 40% of the cystic neoplasms. The prognosis is similar to PDAC, except patients with IPMN can present with pre-invasive lesions, which have a more favorable prognosis. 5 1.2.2 Staging of pancreatic cancer Pancreatic cancer is staged from 0 to IV regarding size, borders and affection of surrounding tissue and/or lymph node and distant metastases. 6 The staging is based on the Tumor, Node, Metastasis (TNM) classification of malignant tumors. 13 Pancreatic cancer is staged at the point of diagnosis, and the staging will be determinative for choice of treatment. Stage TNM Category Median survival 0 Tis, N0, M0 Local or resectable 17-23 months IA T1, N0, M0 Local or resectable 17-23 months IB T2, N0, M0 Local or resectable 17-23 months IIA T3, N0, M0 Local or resectable 17-23 months IIB T1, N1, M0; T2, N1, M0; T3, N1, M0 Local or resectable 17-23 months III T4, any N, M0 Locally advanced or unresectable 8-14 months IV Any T, any N, M1 Metastatic 4-6 months Table 1: Staging of pancreatic cancer. Adapted from 6 1.2.3 Treatment of pancreatic cancer The treatment opportunities in advanced stages are few, and are yet not very effective. 6 In the following section, the most common choices of treatment are described.

17 1.2.3.1 Surgery Surgery, with complete surgical resection of the tumor is considered to be the only way to cure the disease. Unfortunately, surgery is possible in only ~10% of the patients. 6,14 Metastases at the time of diagnosis is an absolute contraindication for operation. 15 Even after radical surgery, the majority of the patients have a poor prognosis due to recurrence of tumor, and metastatic development. 5 The methods used for surgical intervention, are pancreaticoduodenectomy (Whipples operation), distal resection of pancreas or total pancreatectomy. Operative mortality is low, and the procedure should be performed at centralized surgery wards, with preferably more than 15-20 pancreaticoduodenectomies per year. This is important to keep the complication rate as low as possible. 6 1.2.3.2 Chemotherapy Adjuvant treatment is given with the purpose to prevent, or delay, any recurrent disease. This treatment is given to patients after radical surgery with curative intention. The most used regimens are 5-fluorouracil and leucovorin, or gemcitabine, both for a period of 6 months. In some cases, neo-adjuvant chemotherapy is indicated, but this is still somewhat controversial, and is not yet recommended as standard treatment in Norway. 16 Also FOLFIRINOX (fluorouracil, leucovorin, irinotecan and oxaliplatin) could have a potential in an adjuvant setting, and this is currently in clinical trials (NCT01526135). 17 In a palliative setting, FOLFIRINOX is first choice, when performance status is good. This gives an acceptable quality of life for most of the selected patients with relatively few side effects as long as the Eastern Cooperative Oncology Group (ECOG)-score is 0-1 before start. In addition, albumin-bound paclitaxel (nab-paclitaxel) and gemcitabine, or gemcitabine and capecitabine could be an option for patients when FOLFIRINOX is considered too toxic. In addition, gemcitabine monotherapy is an alternative for patients when therapy that is more intensive is not feasible due to co-morbidities or other complications. If first line treatment gives stable disease for a period of time, it is

18 possible to continue with second line treatment after progression. Gemcitabine or FLOX (fluorouracil and oxaliplatin) can both be used as second line treatment in these settings. 14,18,19 1.2.3.3 New drugs in development Immunotherapy has shown encouraging results in early clinical trials, and different trials have different treatment approaches. Checkpoint inhibitors are promising, because of their ability to enhance the anti-tumor response of the immune system, and several clinical trials is currently ongoing in pancreatic cancer. Programmed Death protein 1 (PD1)/Programmed Death Ligand 1 (PD-L1) inhibitors or Cytotoxic T-lymphocyte associated protein 4 (CTLA-4) inhibitors are given alone or in combination with each other or already established treatment (NCT02311361, NCT02868632, NCT02866383, NCT02777710, NCT02734160). 20 Therapeutic vaccines against pancreas cancer are also currently being tested, and furthermore, different monoclonal antibodies, cytokines and oncolytic virus therapies. 21 1.3 Tumor Biology Cancer development is a complex, multi-step process, normally developing over many years. The development of cancer requires changes, both regarding single cells and for their surroundings. Changes at cell level normally consists of mutations, deletions or upor downregulation of regulatory proteins. An average tumor normally have 2-8 so-called driver gene mutations, providing the tumor with a growth advantage, in addition to 30-60 less important mutations. 22 Also microenvironmental changes and changes in how the environment responds to an atypical cell, is required in the development of a fulminant cancer. 23 In 2000, Hanahan and Weinberg stated six hallmarks of cancer. (See figure 2) 24 In 2011, two new emerging hallmarks were proposed.(figure 2). 25

19 Figure 2: Hallmarks of cancer. Adapted from 25 Sustained proliferative signaling is one of the most important steps in cancer development, and there are many ways to achieve this. Common for many of the different strategies are their influence on the cell cycle. During a normal cell cycle, several checkpoints are established to control the properties of the cell before the entering of a new cell cycle state. In cancer, a common feature is that the cancer cell is continuously kept in an active proliferating state, without being withheld at the checkpoints and transferred to scenescense. 26,27 Cancer cells can produce self-made growth factors giving rise to an autocrine stimulation of proliferation. The cancer cells can also stimulate neighboring cells in the tumor stroma to produce growth factors, or overexpress receptor proteins on their surface to make them hyper-sensitive in situations when limited access to growth factors otherwise could stop the signaling. 28 In addition, ligand independent firing is possible due to structural changes in the receptors. Further, downstream pathways could be activated without receptor activation. 25 The mentioned cellular changes have the potential of making growth factors constitutively active, which will stimulate the cell to increased proliferation. 29 A cancer cell will also deactivate mechanisms in the cell designed to negatively regulate cell proliferation. One example is the tumor suppressor gene TP53, which under normal conditions controls the internal cell machinery, and stops further cell cycle progression if the cell is under stress, or has developed genomic damage. Stressful cell conditions can be hypoxia or suboptimal glucose access, and this can activate TP53 and trigger

20 apoptosis. 30 Many cancers have different mutations in TP53, mainly missense substitutions, and this will alter the proteins ability to suppress cell growth. 31 An inactivating mutation of TP53 will lead to apoptosis evasion, and is seen in many human cancers. 31 To be able to develop into macroscopic tumors, cancer cells will further need the ability to replicate unlimited. Telomeres are protecting the ends of chromosomes, but they are shortening in each division, until they no longer can protect the coding part of the DNA. Under normal circumstances, this will trigger a cell crisis, and subsequent cell death. 25 A majority of cancer cells expresses the enzyme telomerase reverse transcriptase (TERT), which will add segments to the telomeres located at the chromosome ends. This will prolong or even give the cells unlimited replication ability. Mutations in the promoter of the human TERT gene, is one of the most common noncoding cancer related mutations, although common cancers like breast and prostate will normally not have this mutation. 32 It is also essential for every tumor to have access to a sufficient amount of oxygen and nutrients. To achieve this, it is necessary with an adequate vasculature to supply the needs of the developing tumor. During cancer development, preexisting vasculature continues to develop new blood vessels, despite being quiescent under normal conditions. This is often referred to as an angiogenic switch. 33 Different tumors express different proangiogenic factors. Vascular endothelial Growth factor-a (VEGF-A) is the most widespread, but also fibroblast growth factor (FGF) and other members of the VEGF family are expressed in a cancer setting. Tumor blood vessels are different from normal vessels, being more irregular, dilated and with the occurrence of non-functional dead-ends. 34 The majority of the steps involved when a cell develops into a cancer cell, such as elevated levels of metabolic activity and cell division, is increasing the energy-needs of the cell. To meet the new requirements, the developing tumor cells are dependent of changes in energy metabolism. The Warburg effect, first described by Otto Heinrich

21 Warburg in 1956, is a metabolic switch observed in cancer cells. In this process, the normal Adenosine triphosphate (ATP)-production via oxidative phosphorylation changes to ATP-production via glycolysis, also under normal oxygen levels. 35,36 Glycolytic ATP-generation is quicker, but is using more glucose than oxidative phosphorylation, which demands a high level of glucose supply from the surroundings. The increased glucose uptake in the tumor tissue is exploited for diagnostic purposes, with the imaging-technique of [ 18 F] fluorodeoxyglucose positron emission tomography (FDG- PET). 37 Altogether, a series of events is required in the development from a normal cell to a cancer cell. The process could stop in any of these steps, and this will stop the cancer from developing further. Different treatment approaches is also able to target different of these steps, aiming to stop the process. As an example, bevacizumab is a monoclonal antibody targeting VEGF, preventing the development of a sufficient tumor vasculature. 38 Also drugs that prevents cells for entering new cell cycles have been developed, targeting the key regulators of the cell cycle, the cyclin dependent kinases. 39 1.3.1 Mechanisms for metastases Metastases account for >90 % of cancer related deaths. 40 An established metastatic tumor at a different and often distant site of the primary tumor is a result of a series of events, which involves local invasion, intravasation, circulative transportation, extravasation, micro-metastatic formation, and finally colonization and formation of a macroscopic metastasis. 41,42 The metastatic process can stop in either of these steps, and the outcome is depending on properties of the tumor cell, but also on responses from the microenvironment at the new site. 42 It is shown that only approximately 0.02 % of the cancer cells that enter circulation is developing into macroscopic metastases. 43 The potential of a cancer cells to metastasize is dependent on the degree of genomic instability in the cell. Cells with high grade of genomic instability will more easily acquire the alterations necessary to metastasize. 44 It is known that the epithelial mesenchymal

22 transition (EMT) is a driver of the metastatic process, and this will be discussed further in section 1.4.2. Experiments have shown the metastatic potential of a tumor is closely related to increased resistance to apoptosis, which is considered as the initial step of the metastatic process. 40 The cell detaches from surrounding cells and extracellular matrix (ECM), and at the same time take a more rounded shape due to degradation of the actin skeleton. In normal conditions, these processes would lead to apoptotic cell (AC) death, through either anoikis (apoptosis induced by cell detaching) or amporphosis (apoptosis induced by disrupted intracellular architecture). 40 In a metastatic setting however, the abnormal cells will escape these processes, and continue to proliferate. It has been shown that overexpression of the anti-apoptotic protein B cell lymphoma 2 (BCL-2) increases the metastatic capacity of mammary epithelial cell, without affecting other important steps as primary tumor growth, cell motility or invasiveness. 45 Also the metastatic steps of intravasation, circulation, extravasation and establishment in a new micro-environment are promoted by anti-apoptotic mechanisms. 40 The steps of metastatic development are illustrated in Figure 3.

23 Figure 3: The pathogenesis of cancer metastasis. The process of metastasis can stop in any of these steps. Adapted from 42 It is known that certain tumor types will metastasize to different specific organs. This problem was addressed as early as in 1889 with Paget s Seed and soil theory, where certain tumor cells (the seed) are hypothesized to have specific preferences for a certain micro environment in specific organs (the soil), independently of the rate of blood flow in the different organs. 46 A few years later, in the 1920s, another model of explanation were suggested. At this point, James Ewing proposed the theory of a circulatory pattern between primary tumor and metastatic organs. According to this theory, the metastatic sites are passive receptors of tumor cells, and the preferred organs depend on the circulatory network between the primary and metastatic site. Later experiments have confirmed that both theories could be valid, as the number and localization of metastases could depend both on mechanical factors as blood supply and tumor cell delivery, and also microenvironmental factor where the local environment at the metastatic site would favor growth of cancer cells from certain primary localizations. 43,47

24 To form a metastasis, the tumor cell has to survive in the new environment at the metastatic site. It is shown that AKT signaling is important in this matter, both when the tumor still is in circulation, but also in specific organs, as lung or bone marrow, to prevent the cell of undergoing apoptotic processes in the early phase of establishing a metastatic tumor. 48 Once localized at the new site, the metastatic cells have the capability of establishing a metastatic niche, in terms of releasing soluble factors or micro-vesicles to make the new microenvironment more facilitated for tumor development. 49 There is also hypothesized, although still debated, that tumor cells already at the primary site or in the circulation can secrete molecules to prepare the microenvironment at the site of metastasis, making a so-called pre-metastatic niche. This is in line with Paget s seed and soil theory. 48,50 It is proposed that modifications of the stroma includes increased levels of fibronectin and matrix metalloproteinases (MMPs), structural changes is ECM and recruitments of bone marrow derived cells to make the environment more favorable for adhesion of the cancer cells, and subsequently metastatic colonization. 51,52 1.4 Epithelial to Mesenchymal transition Epithelial to mesenchymal transition (EMT) is a process with cellular transformation from an epithelial to a mesenchymal phenotype. It is an important process in embryogenesis, which allows cells to migrate to different localizations during phases of development, both in morphogenesis and organogenesis. 53,54 The embryonic form of EMT can be referred to as type 1 EMT. The EMT process is occurring also in adult tissue, both in normal processes such as wound healing and inflammation (Type 2), but also in pathological processes as such cancer, leading to cell invasion, dissemination, and development of therapeutic resistance (Type 3). 55 Characteristics of the EMT process are loss of cell polarity and cell-cell interactions, modulations of the adhesion between cells and ECM, enhanced proteolytic activity, ECM degradation, increased cell motility and reorganization of the cytoskeleton. 56,57

25 Epithelial cells have several features classifying them as epithelial. They have a welldefined apical-basal polarity with a basal membrane and widespread cell-cell contacts, such as tight junctions, which allows communication between the cells. They have a characteristic cobble stone-like shape, and are non-motile. 58,59 Epithelial cadherin (Ecadherin) is an important protein responsible for the formation of adherence junctions, by making protein clusters connected to actin microfilaments. This provides a strong control of the epithelial architecture. 60 E-cadherin is considered the main marker for the epithelial phenotype, and in vitro, a correlation between the lack of E-cadherin and loss of an epithelial phenotype has been demonstrated. 60,61 Other important cell-cell contacts are the tight junctions. The Claudin protein family is the most important component of the tight junctions, followed by the protein occludin. Both Claudin and Occludin are commonly used as markers for an epithelial phenotype, and they are shown to be downregulated during the process of EMT. 53,57,62 The phenotype of mesenchymal cells are quite different from epithelial cells. The shape is more elongated and spindle-like, and they do not have the strict apical-basal polarity seen in epithelial cells. They also lack the cell-cell contacts, which are critical in the epithelial cell structure. Furthermore, mesenchymal cells have the ability to migrate as single cells, and display another set of proteins than the epithelial cells, such as the mesenchymal markers Vimentin and N-cadherin. 59 In cancer, the levels of proteins that are characteristic of mesenchymal cells, and simultaneously loss of epithelial markers, correlates with evidence of tumor progression and poor prognosis. 63 Typically, the cells expressing mesenchymal markers are seen in the invasive front of primary tumors, and are most likely the cells that first will start disseminating. 55,60,64 1.4.1 Activation of EMT The EMT program is activated by developmental transcriptional regulators. The most important of these are TWIST, Zinc finger E-box binding homeobox 1 and 2 (ZEB1 and ZEB2), and two members of the snail superfamily of transcription factors, SNAIL (SNAI1) and SLUG (SNAI2). 65,66 These transcription factors will change the gene expression

26 profile, by repression of the genes encoding for the epithelial proteins (e.g. E-cadherin and β-catenin) and induction of increased expression of mesenchymal proteins, such as Vimentin and N-cadherin. As an example, it is shown that SNAIL and ZEB bind to, and subsequently repress the activity of the E-cadherin promoter, and by that regulating the expression of E-cadherin. 67 SNAIL and ZEB are also contributing to destabilization of epithelial cellular polarity, which is a key feature of epithelial phenotype. By inducing expression of different metalloproteases that will degrade the basal membrane, they stimulate cellular instability and invasion. 68 Also expression of Claudins, which are important for tight junctions are downregulated by SNAIL. This is thought to be via the lysine specific demethylase 1 (LSD1). 69 There is also evidence for a positive feedback loop in this system. The metalloproteinase MMP3 will increase levels of reactive oxygen species (ROS) in the system, and that will again stimulate the expression of SNAIL. 70 Furthermore, it is shown that especially expression of SNAIL is closely related to signaling of transforming growth factor β (TGF-β). 71 This is relevant, both during normal development, and in cancer. 53,71 TGF-β has a two-sided role in the development of cancer. In many conditions, it is an important suppressor of epithelial cell proliferation and subsequently primary tumorigenesis. It will nevertheless serve as a positive regulator of tumor development in other conditions. 72 During tumor progression, there is evidence that the tumor cells will lose their normal TGF- β- related growth inhibition, due to mutational changes. This will lead to increased growth, followed by more mutations, and subsequently cancer progression. 73 In a different pathway, there is evidence that the signaling protein Ras will be activated, and this will enhance the effects of TGF- β that promotes tumor progression, and metastatic development. 73 TGFβ also have the potential to activate the phosphatidylinositol 3OH kinase (PI3K) pathway with its downstream target AKT, which will lead to EMT induction. 74 WNT signaling also have the potential of stimulating EMT. This large family of proteins are involved in several cancer types. Activation of the WNT pathway will phosphorylate Glycogen synthase kinase 3 beta, (GSK3β), a tumor repressor, and this will via β-catenin activate transcription of SNAIL, and stimulate the EMT process. 74

27 1.4.2 EMT control The process of EMT can be controlled by different factors. MicroRNAs (mir) are important in this respect. MiRs are small pieces of RNA (approximately 22 NT) which can bind to target mrna and influence the translation. Especially, mir-200 is associated with EMT by regulating expression of ZEB. 75,76 There is also evidence of down-regulation of mir-200 family members in several human cancers. 77-79 Also SNAIL-dependent EMT can be regulated by mir, most commonly by the mir-34 family. 80,81 Both the mir-200 family and the mir-34 family are controlled by the tumor suppressor TP53. 81,82 TP53 will bind to the promoter of mirna-200, and stimulate its expression. Loss of TP53 in breast cancer will give less expression of mirna-200, increased activation of the EMT-program and development of cells with stem-cells properties. 82 1.4.3 EMT and Cancer In cancer development, EMT is thought to have an important role as a facilitator for dissemination and metastatic spread. This type of EMT is often referred to as type 3 EMT. 55 The process of EMT in cancer is strongly dependent of the tumor microenvironment, and micro-environmental factors will in many cases decide if a cell has the potential to undergo EMT and then metastasize. 83 The loss of E-cadherin during EMT is inversely correlated to cancer grade and patient survival, and E-cadherin downregulation is associated with increased cell growth. 60 A number of different growth factors will contribute to EMT in cancer, and the growth factor signaling will vary in different cell types. Examples of growth factors that can induce EMT are epidermal growth factors (EGFs), fibroblast growth factors (FGFs), hepatocyte growth factor (HGF), insulin-like growth factors (IGFs) and as before mentioned, TGF-β. 54 Also activated Ras/ mitogen activated protein kinases (MAPK), Src kinase and PI3K-signaling are shown to be inducers of the EMT program. 54 These pathways are associated with important hallmarks of cancer, such as ability to regulate cell cycle, sustained proliferation, and also the properties of evading growth suppression and apoptosis. 25 Other tumor-related factors can also trigger EMT in cancer. Intra-tumoral hypoxia could trigger the

28 expression of SNAIL, and subsequently the process of EMT. Thus, this is an important factor in tumor development, together with acidic conditions, inflammation and low blood glucose. 54,84,62 The role of EMT in cancer progression is still not fully understood. It is proven that EMT is not required for establishment of metastases, despite being very important for the invasive development. 57 Furthermore, cells with different EMT status could be present within the same tumor. It is hypothesized that there is an EMT gradient in different segments of the tumor, and that the cells goes through different states of intermediate EMT levels, where they gradually lose their epithelial phenotype. 57,85 One theory is that a malignant tumor has an invasive front, where the cells have undergone EMT and have mesenchymal properties, whereas the main part of the tumor still largely is epithelial. 57 Both in embryogenesis, and in cancer development, EMT is a transient phenomenon. After cell dissemination and spread to distant sites, the cells will go through the opposite process, mesenchymal to epithelial transition (MET) in order to establish macrometastases at the new localization. 60,62 In metastases, the cancer cells exhibit histopathological similar traits as the cells in the primary tumor, lacking the mesenchymal phenotype, supporting the theory of the process of MET. 55 It is proposed that this is because the microenvironment in the new localizations will not provide the EMT stimulating signals present at the primary location, leading to a reversal of the process. 86 EMT is important also in other steps in the malignant development, such as resistance to cell death. TGF- β can induce EMT in mammary cells, and at the same time inhibit apoptosis. It is also shown that EMT induction can give rise to a phenotype with resistance of cell senescence induced by oncogenes. 87 1.4.3.1 EMT and Cancer stem cells The EMT program has in several studies also been linked to the development of cancer stem cells (CSCs). A subset of the cells undergoing EMT is exhibiting stem-like properties, or being in a condition, just ready to enter the stem cell state. 88-90 The hypothesis is that

29 these cells will exhibit some of the same traits as normal stem cells, such as the ability to self-renew and serve as progenitors for cell clones with adaptive characteristics. 91 There is not consistency about the origin of cancer cells with stem-like traits, and different theories have been suggested. Dedifferentiation of the a cancer cell together with EMT is one of the proposed mechanisms, together with the hypothesis of malignant transformation of a normal stem cell, and induction of pluripotent cancer cells. In the invasive front of tumors there is evidence for expression of both stemnessassociated genes, and EMT-related genes, supporting the theory of a dedifferentiation of a subpopulation of cancer cells together with EMT. 92 It is shown that both the transcription factors SNAIL and TWIST stimulate the acquisition of stem cell properties in cancer cells. 87 Investigations of cells that have undergone EMT and exhibit mesenchymal markers show that a significant proportion of these cells will express cell surface markers compatible with stem cell traits, with a CD44high/CD24 low ratio. 88 1.4.3.2 EMT and drug resistance It has been shown that the EMT-related transcription factor SNAIL can induce resistance to both chemotherapy and immunotherapy, and also immunosuppression. Furthermore, TWIST is capable of inducing resistance to senescence. 87,93 Investigations of cell lines resistant to chemotherapy revealed that these cells present with a mesenchymal phenotype, and express markers of EMT. 94,95 In acquired resistance, evidence shows that the process of EMT gives rise to more mesenchymal cells with chemo-refractory abilities and stem cell like features. 96 Also de novo resistance is related to EMT. In treatment of lung cancer with EGFR kinase inhibitors, sensitive tumors have elevated E-cadherin levels, while drug resistant cells have properties that are more mesenchymal. 97 1.5 Receptor tyrosine kinases Receptor tyrosine kinases (RTKs) are a group of transmembrane proteins, functioning as cell surface receptors and regulators of many cellular processes. In the human genome there are 58 RTKs, separated in 20 subfamilies. 98 The molecular structure of all

30 the RTKs are similar, containing an extracellular ligand-binding domain, and an intracellular tyrosine kinase region, separated by a transmembrane helix. RTKs are normally activated by ligand binding, followed by receptor dimerization and activation of an intracellular kinase domain. 98 Overexpression of RTKs, commonly because of gene amplification, is closely related to many human cancers. 99 Figure 4 gives an overview of human RTKs. Figure 4: Human receptor tyrosine kinases. Adapted from 99 1.5.1 AXL receptor tyrosine kinase AXL is a RTK and a member of the TAM (TYRO, AXL, MER), family of RTKs which is a group of transmembrane RTKs. AXL is located at chromosome 19, and is encoded by 20 exons. The receptor is approximately 140 kda in a fully glycosylated state. 100,101 The name AXL is from the Greek anexelekto which means uncontrolled. 100 AXL was first discovered in Chronic myelogenous leukemia, as an unidentified transforming gene. 102 In normal tissues, AXL has a ubiquitously distribution. Detectable levels of AXL are found in endothelial cells, heart, kidney, liver, monocytes/macrophages, platelets, skeletal muscle, and testis. Also in the normal brain, there is evidence of AXL, most notably in cerebellum and hippocampus. 103

31 1.5.1.2 AXL structure The structure of AXL is similar to other RTKs. AXL consists of an extracellular domain (Nterminal), containing two fibronectin type III domains, and two immunoglobulin (IG)-like domains. 100,104 The intracellular tyrosine kinase domain (C-terminal) contains an unusual KWIAIE amino acid sequence, which is unique for the TAM family of RTKs. This sequence is similar, but different to the consensus sequence for all the tyrosine kinases. 100 The intracellular domain has an ATP-binding site, which catalyzes receptor autophosphorylation and serves as a docking site for cytoplasmic signaling proteins containing domains for protein tyrosine binding and src-homology-2. 105 This structure is common for all the TAM RTKs. Figure 5 shows the structure of AXL and its ligand GAS6. Figure 5: Structure of the receptor tyrosine kinase AXL and its ligand GAS6. Adapted from 106 1.5.1.3 AXL ligand AXL is activated by its ligand Growth arrest specific 6 (GAS6) which was first identified in 1995. 107 GAS6 is a vitamin-k dependent protein, containing a gamma-glutamic acid residue (Gla) rich domain at the N-terminal. The Gla-domain is under normal

32 conditions carboxylated from glutamate to gamma-carboxyglutamate prior to receptor binding. This vitamin-k dependent process is necessary for GAS6-mediated activation of AXL. 108-110 In addition to the Gla-domain, GAS6 also consists of four EGF-like repeats and a sex hormone binding globulin at the C-terminal, which includes two laminin G-like domains. 111 Structural studies have shown that there are two binding sites between AXL and GAS6, one major contact between LG1 GAS6 and IG1 AXL and one minor contact between LG1 GAS6 and IG2 AXL. Both the points of contact are required for AXL activation. 112 1.5.1.4 AXL activation After binding, AXL and GAS6 are creating a strong 1:1 GAS6:AXL complex, followed by a dimerization of two 1:1 GAS6:AXL complexes. 112 After gamma-carboxylation, the Gladomain of GAS6 will bind to phosphatidylserine (PtdSer) on neighboring cells in a calcium-dependent process. Many cell types throughout the body are expressing PtdSer. In most conditions, PtdSer is located at the inner leaflet of the plasma membrane, but activated platelets and ACs are presenting the PtdSer on the outside. 113 The interaction between the Gla-domain and the PtdSer on neighboring cells is optimizing receptor activation. 110 There are conflicting reports regarding the possibility of AXL activation without simultaneously binding of PtdSer. The evidence is consistent in the findings of that the PtdSer binding is not essential for the binding of GAS6 to the receptor, but the presence of PtdSer is thought to increase the affinity. 114 A study from 2014 argues that the gamma carboxylation of the Gla-domain is essential for receptor activation, but there is no need of following binding to PtdSer for receptor signaling. 115 Another report from the same year argues that AXL signaling depends both on the carboxylation of GAS6 and subsequent binding of PtdSer. 110 On the contrary, a more recent report demonstrates a possibility of receptor activation without gamma carboxylated Gla domain, and subsequently no binding to PtdSer but the activation will then be of weaker character. 116 Figure 6 illustrates the GAS6-mediated activation of AXL.

33 Figure 6: GAS6:AXL activation. Illustrates that AXL is activated of GAS6 and not Pros1, and that AXL is dependent of carboxylated Gla-domain of GAS6 binding to PtdSer on neighboring cells. Adapted from Lew et al 110 There is also evidence that ligand-independent processes can activate AXL. This can be through interaction of two monomers on neighboring cells, causing cell aggregation, ligand independent dimerization and heterotypic receptor-dimerization with a non- TAM receptor. 111,117 Ligand-independent activation of AXL is related to AXL overexpression, and is therefore more likely to occur under pathological conditions, such as cancer. 118,119 Especially the homophilic binding of the extracellular domains on neighboring cells can lead to cell aggregation, and is associated with cancer, when the receptor is overexpressed. 120 Figure 7 illustrates the different mechanisms leading to AXL activation.

34 Figure 7: Mechanisms of AXL receptor activation/inactivation. A: Ligand induced activation. High affinity 1:1 GAS6:AXL complex followed by a dimerization of two GAS6:AXL complexes. B: Hemophilic binding of extracellular domain of AXL expressed on neighboring cells can lead to cell aggregation especially when AXL is overexpressed. C: Ligand independent hemophilic dimerization of AXL and auto-phosphorylation in response to ROS. D: Proteolytic cleavage of AXL to form soluble AXL (saxl). Adapted from 118 After the dimerization of the complexes, there will be a subsequent autophosphorylation of the intracellular kinase domain of AXL. There are several intracellular tyrosine residues which can be phosphorylated, the most described sites are Y779, Y821 and Y866, all located in the C-terminal kinase domain. 118,121 After activation of the GAS6:AXL complex, there is a cleavage of the extracellular domain from the cell surfaces, a process conducted by proteases. 122 There is evidence that elevated soluble AXL (saxl) in blood can be a marker of different conditions of variable character. For example, saxl is suggested as a potential biomarker for inflammation, and early stage hepatocellular carcinoma. 123,124 1.5.1.5 Downstream events of AXL The activation of AXL is linked to different intracellular signaling cascades, where several also are related to tumor development. The downstream signaling is thought to be similar to other RTKs, and different pathways are activated at different time points,

35 determined by tissue type, cell type and extracellular environment. 101 The PI3Kpathway, with the downstream targets AKT and S6K, and also the phosphorylation of nuclear factor-κb (NF-κB) is one of the major downstream pathways of AXL activation. PI3K is an intracellular kinase and the key component of a pathway mediating several cellular responses related to cancer development, such as growth, motility and survival. 125 It is evidenced that the p85 (α and β) subunit of PI3K is interacting with a multi-substrate docking site at the tyrosine 821 on AXL. 121,126 Activation of PI3K and subsequently NF-κB, will lead to increased expression of anti-apoptotic proteins such as B-cell lymphoma 2 (BCL-2) and B-cell lymphoma extra-large (BCL-XL), and inhibition of pro-apoptotic proteins like caspase 3. 103 Furthermore, AKT will phosphorylate the proapoptotic BAD, which subsequently will not be able to interact with BCL-XL and BCL-2. This will lead to decreased apoptosis, and increased cell survival. 127 It has been shown in Gonadotropin releasing hormone -neurons that PI3K can mediate activation of p38, which in turn will lead to phosphorylation of Heat-shock protein 25. This is a regulator of actin modelling, and its activation will regulate remodeling of actin, which will favor increased migration. 128 Also p-21-activated kinases -1 (PAK1) is known to be activated through the PI3K/AKT pathway. PAK1 stimulates cell invasion when activated. 129 Another downstream target for AKT is GSK3. This protein has been shown to mediate several oncogenic traits, such as cell survival, proliferation and cell cycle progression. 130 Overall, AXL-mediated activation of the PI3K-pathway is linked to increased cell survival, proliferation and cell migration. The MAPK/extracellular signal regulated kinases (ERK) signal transduction cascade is also initiated upon AXL phosphorylation. The tyrosine kinase domain of AXL will after auto-phosphorylation bind to the intracellular Grb2 protein, which then will activate the MAPK/ERK pathway. 121 This pathway is often linked to AXL-mediated proliferation. 103,131 This GAS6-mediated induction of ERK is both in strength and duration comparable to what is seen in response to more well-described growth factors, such as EGF and platelet derived growth factor. 132 The growth-stimulating effect of GAS6 is additive to

36 the effect of EGF, which suggests that GAS6 utilizes other pathways than those utilized by EGF. 126,132 AXL signaling is heavily involved in the human immune system. Together with the other TAM receptors, the GAS6:AXL complex will protect innate immune cells such as macrophages, dendritic cells and natural killer cells (NK cells), from apoptosis. Via the type 1 interferon receptor (IFNAR), GAS6:AXL will initiate phosphorylation of the transcription factor STAT 1, a member of the Signal transducers and activators of transcription family of transcription factors, and subsequently the expression of suppressor of cytokine signaling 1 and 3 (SOCS 1 and 3), which are inhibitors of cytoplasmic cytokines. SOCS 1 and 3 will also inhibit Toll-like receptors (TLR) on dendritic cells, and by this inhibit their inflammatory response to pathogens. These pathways are very important for controlling inflammatory responses. 118,133 Also STAT3 is linked to GAS6:AXL. STAT 3, as STAT1, belongs to the STAT family of transcription factors, and is persistently activated in many cancers. STAT 3 is known to be an important mediator of the oncogenic effects of EGF. 134 In head and neck-cancer and colorectal cancer, it is shown that inhibition of AXL will lead to reduced phosphorylation of STAT3. 135,136 Furthermore, there is shown a relationship between AXL and the cytokine interferon-α (IFN-α). 104 Secretion of IFN-α will upregulate AXL in macrophages, which in turn will lead to increased TWIST expression and reduced tumor necrosis factor α (TNFα) production. TNFα is a strong inflammatory cytokine, and reduced production will give a weaker inflammatory response. 137 The downstream events of AXL is illustrated in Figure 8.

37 Figure 8: Downstream events of AXL. This figure show the downstream events after AXL activation. The boxes show the final outcome after the intracellular processes are activated. 1.5.1.6 AXL regulation Some intracellular proteins have the ability to inhibit and regulate the activity of AXL. C1 domain containing phosphatase and TENsin homologue (C1-TEN) is a protein that will bind to the intracellular compartment of AXL. This will negatively regulate AXL signaling through PI3K/AKT. The detailed mechanisms behind this inhibition are still unclear. When overexpressed, C1-TEN will inhibit the cells ability to proliferate and migrate. 138 There is also evidenced that soluble forms of AXL (saxl) are circulating in plasma, and will, by binding to GAS6, inhibit receptor activation. 115,139 As other RTKs, AXL is regulated by a mono-ubiquitination signal, which leads to endosomal internalization and degradation by lysosomes. The regulating ligase for AXL internalization is Casitas B- lineage Lymphoma-c (c-cbl). 101,140. In kidney cancer it is shown that cells with low levels of VHL have increased AXL expression, due to binding of hypoxia-inducible transcription factor 1 and 2 (HIF-1 and HIF-2) to the AXL promoter. 141,142 Myeloid Zinc finger 1 is another protein that is regulating AXL expression by binding to the promoter. This will activate the promoter, and increases the expression of AXL. 143

38 There is also evidence for AXL regulation at the mrna level. Two mirs are identified (mir-34a and mir-199a) which binds to the 3 UTR of AXL, having an inhibitory effect on AXL protein levels. 144,145 A recent study in ovarian cancer shows that expression of mir- 34a will lead to decreased AXL expression followed by significantly inhibited cell migration. 146 Similar findings have been shown in prostate cancer where expression of mir-34a leads to downregulation of AXL, and induced apoptosis, and growth inhibition. 147 In osteosarcoma it is shown that mir-199a-3p will downregulate the expression of the AXL gene, and this will inhibit progression of the disease. Also, low levels of mir-199a-3p significantly correlates with recurrence of lung metastases, and low levels of mir-199a are a predictor of poor prognosis in osteosarcoma. 148 Specificity protein 1 and 3 (Sp1/Sp3) transcription factors are also regulators of AXL expression by binding to Sp motifs upstream for the AXL promoter, and by that driving AXL expression. In low AXL-expressing cells, these motifs are methylated which will restrict AXL transcription. This is in contrast to high AXL-expressing cells, where there is evidence for a hypo-methylation of the Sp motifs. Experiments with demethylation of these areas in low AXL-expressing cells, lead to increased AXL expression. 149 1.5.1.7 AXL in normal physiology In the normal physiology, the expression of AXL is widespread throughout the body, although mainly in the mature immune, nervous, reproductive and vascular systems. 150 Still, none of the TAM receptors is essential for embryonic development, as TAM - / - are viable after birth. 114 Hemostasis AXL-expression on platelets will mediate thrombogenesis and platelet stabilization. 151 The receptor will be activated by PtdSer on aggregating platelets, and simultaneous release of GAS6 from granules in the platelets. This process is contributing to stabilizing the clot formation. GAS6 - / - mice show signs of impaired platelet aggregation, with prolonged bleeding time. 114,152 TAM RTKs are also involved in other parts of the vascular homeostasis, such as reestablishment of the endothelial barrier function after vascular damage, and also by promoting survival of endothelial cells. 153 Furthermore, AXL is

39 known to play a role in neovascularization, and it is shown that AXL is important for VEGF-A induced endothelial cell migration and subsequent formation of new blood vessels. This is effectuated through the downstream PI3K/AKT pathway. 154 Mediation of phagocytosis of apoptotic cells In the process of phagocytosis, the TAM ligands GAS6 and Pros1 binds to PtdSer on ACs, and serves as a bridging molecule between the AC and a TAM-receptor on a neighboring phagocyte. After activation of the TAM-receptor, this linking of the cells will push the apoptotic process forward. 104 This is an important step in normal physiology, and necessary to prevent a state of continuous inflammation. 153 In AXL knockout (KO) animals the consequences is shown to be severe, with an accumulation of dead ACs. Especially regarding sperm production and in retina, this is important, leaving the AXL KO animals blind and sterile. 113 Immunology All the TAM receptors, have an important role in the inhibition of the innate immune system. 114 The receptors functions as a safety-system to prevent prolonged and overintense immune reactions and will promote tissue-repair after inflammatory responses. Dendritic cells have medium AXL expression at steady state, but the expression will be upregulated following pathogen invasion, which gives activation of TLR and further by type 1 IFN. This will contribute to the termination of the immune response after the specific pathogen reaction. 133 Situations with low AXL will always present with chronic inflammation and a prolonged immune reaction. AXL KO mice have severe autoimmune disease in a clinical pattern similar of systemic lupus erythematosus or rheumatoid arthritis. 155 When a virus enters the human organism, it is possible that AXL activation dampens the immune response, and thereby making it easier for the virus to escape the immune reaction. Many viruses have PtdSer on their external surface, and will by that activate AXL, and enter the cell. Infections like Zika virus, and also Ebola and West Nile virus has been coupled to AXL as a receptor for cell entrance. 156

40 1.5.1.8 AXL and EMT AXL is known to be closely related to the EMT process. Over-expression of EMT-related transcription factors, such as SLUG, SNAIL, TWIST and ZEB2 is linked to up-regulation of AXL. 65 AXL up-regulation will additionally have a positive feedback on the transcription factors, leading to sustained expression of SLUG, SNAIL and TWIST. 65,157 There is also a tight connection between expression of EMT-related proteins and AXL. An example of this is the mesenchymal protein Vimentin. This is a protein is an important regulator of mesenchymal cell migration, and a marker of EMT. The level of AXL is closely related to the level of Vimentin, and increased expression of both proteins will enhance the cancer cells migratory capacity. 158 There is also evidence for a relationship between AXL upregulation and acquisition of drug resistance and an EMT phenotype. This is described in several cancers, such as lung cancer, breast cancer and chronic myeloid leukemia (CML). 159-162 Furthermore, AXL expression is shown to be enhanced in breast cancer metastases relative to the primary tumor (investigated for matched samples). These findings strongly indicate that AXL is associated with epithelial plasticity and has a role in malignant progression and metastatic development. 65 1.5.1.9 AXL and cancer AXL is associated with many different cancers. It was first described in Chronic myeloid leukemia (CML) in 1988. In the beginning, it was described as an unidentified transforming gene. 102,163 Activating mutations or amplifications associated with AXL are rare, rather up-regulation and increased ligand-induced activation is associated with cancer. 100,164 In the recent years, there have been many reports of AXL up-regulation in several different cancer types. (Table 1). It is believed that the upregulation of AXL is induced by hypoxia in the tumor environment, which is a common feature of most solid tumors. Hypoxic conditions will stimulate HIF-1 and HIF-2 to express several genes as response to this, amongst these, AXL. 142 Overexpression of AXL has several implications. It is related to poorer prognosis 65,165,166, development of drug resistance 159,161,167 and increased invasiveness. 158 Malignancies related to AXL-upregulation, and the correlation with poor prognosis are summarized in Table 2.

41 Malignancies Up-regulation Human Poor Independent prognostic factor tumor prognosis Astrocytic brain tumors 168-173 169,171,172 169 169 Breast cancer 65,174-185 65,176-183,186 65,182,183,185 65 Gallbladder cancer 187 187 187 GI cancers Colon cancer 136,143,144,188-191 136,190,191 136,191 191 Esophageal cancer 167,192-194 167,193,194 193,194 Gastric cancer 195,196 195,196 Gynecological cancers: Ovarian cancer 197-202 197-202 197,200,202 197 Uterine cancer 203-205 203-205 205 Head and neck cancer 206-212 135,207,210-212 135,207,210,212 210,212 Liver - HCC 213-217 214,217 214,217 214,217 Leukemias: AML 166,218-220 166,218-220 166,218 166,218 CLL 221-223 221-223 CML 100,162,220,224 100,220 Lung cancer: 225 SCLC 144,159,165,188,226-159,165,231-234 230-232,234 NSCLC Malignant melanoma 235-238 238 234 Mesothelioma 239-241 239 Pancreatic cancer 109,242-244 242-244 242-244 242 Sarcomas: Ewing Sarcoma 245 245 Kaposis sarcoma 246 246 Liposarcoma 247,248 247,248 247 247 Osteosarcoma 249-251 249 249 249 Undifferentiated pleomorphic sarcoma 252 252 252 Skin SCC 253,254 253,254 Thyroid cancer 255-258 255,256,258 Urological cancers: Bladder cancer 259-261 259 Prostate cancer 262-265 263,265 RCC 142,266-270 266-270 142,268 268 Abbreviations: HCC: Hepato-cellular carcinoma, AML: Acute myeloid leukemia, CLL: Chronic lymphatic leukemia, CML: Chronic myeloid leukemia, SCLC: Small cell lung carcinoma, NSCLC: Non small cell lung carcinoma, RCC: Renal cell carcinoma. Table 2: AXL upregulation and correlation with prognosis. Adapted from 271

42 Invasion and metastasis Overexpression of AXL is thought to be more related to metastatic dissemination and poor overall survival than primary tumor growth. 188,244 These observations correspond to the receptors close relation to EMT. AXL will enhance invasiveness in many cancers, which corresponds to the association to metastatic development. The AXL-related PI3K/AKT activation has in breast, gastric and ovarian cancer been shown to enhance tumor cell invasion, via the NF-κB pathway. 65,195,201. In hepatic carcinoma, AXL downregulation will lead to reduced expression and less activity in the PI3K/AKT pathway. 216 PAK1 is known to be activated through the PI3K/AKT pathway, and PAK1 stimulates cell invasion when activated. 129 Inhibition of the PI3K/AKT pathway and subsequently PAK1 were shown to strongly reduce cell invasion ability. 216 The E3 ubiquitin ligase Casitas B-lineage lymphoma-b (Cbl-b) is) is activating Natural Killer cells (NK-cells). By inhibiting this ligase, the NK-cells will be triggered to reject metastatic tumors. AXL, together with the rest of the TAM family, is shown to be substrate for Cblb, so by AXL inhibition, the activity of the ligase is inhibited, and the NK-cells will have an inhibitory effect of metastasis formation. 272 Cell survival Overexpression of AXL in tumor cells is in many conditions linked to increased cell survival. It is reported in different cancers that increased AXL expression leads to prevention of apoptosis, and subsequently increased survival. A work in breast cancer reports that elevated estrogen levels stimulate increased AXL expression and reduced levels of apoptosis. 176 Also in osteosarcoma there is shown a strong relationship between increased AXL expression and protection from apoptosis. 251 Similar reports is seen in astrocytoma 170, chronic lymphatic leukemia (CLL) 222 and ocular melanoma 237. Some of the anti-apoptotic potential of AXL, goes through the PI3K pathway. Activation of PI3K and subsequently NF-κB, will lead to increased expression of anti-apoptotic proteins such as Bcl-2 and Bcl-XL, and inhibition of pro-apoptotic proteins like caspase 3. In addition, the inhibitory interaction between BAD and Bcl-2 and Bcl-XL will be

43 blocked due to AKT-mediated phosphorylation of BAD. The total effect will be prevention of apoptosis and increased cell survival. 103,127 Angiogenesis Furthermore, there is strong evidence supporting AXL having a role in angiogenesis. 273,274 In the normal cellular environment, AXL is involved with repair of vascular injury. 275 Studies have shown that overexpression of AXL in cancer is present not only in the tumor cells, but also in surrounding vascular cells. 169,233 Vascular smooth muscle cells (VSMC) express GAS6, and exogenous application of GAS6 will stimulate proliferation and mobility of VSMC. 276 There is also evidence that AXL knockdown in endothelial cells (HUVEC) will impair tube formation, and that AXL is an important driver of proliferation of HUVEC cells. 273 Furthermore, it is shown that AXL will influence angiogenesis through modulation of signaling, via angiopoietin/tie2w and Dickkopf related protein 3 (DKK3) pathways. These proteins are known regulators of angiogenesis. In addition, combination of AXL knockdown and anti-vegf therapy resulted in enhanced inhibition of tube formation compared to anti-vegf therapy alone. 188 Furthermore, Ruan et al show that AXL is essential for the VEGF-dependent activation of PI3K/AKT, supporting the evidence of AXL having important implications in angiogenesis. 154 In vivo mouse studies have further shown a strong relationship between AXL and angiogenesis, and suggest that AXL inhibition will suppress formation of new blood vessels. 157,273 Tumor microenvironment AXL can also regulate factors in the tumor microenvironment. Malignant tumors have the ability of invading tumor-surrounding tissues. A crucial trait to achieve this is the capacity of producing matrix-degrading enzymes. MMPs are important enzymes regarding this. MMP-9 is a type IV collagenase, which degrades type IV collagen, an important structure in the basement membrane and ECM. 277 It is shown that AXL enhances the expression of MMP-9, by regulating the promoter activity through the MAP kinase kinase MEK/ERK pathway. 278. In the environment of a tumor, the conditions

44 are often hypoxic compared to the surrounding tissues. Hypoxia will increase the expression of HIF1α which will then promote increased AXL transcription. 142 1.5.1.10 AXL and drug resistance AXL is a facilitator for acquisition of drug resistance in many cancers. The mechanisms for this are not fully elucidated, but it is shown that AXL is significantly overexpressed in therapy resistant cancers compared to tumors in other stages. 279 Especially in breast and lung cancer, this is thoroughly described. 159,161,227,280 It is known that increased expression of RTKs is a compensatory reaction to therapy-induced inhibition of a specific signaling pathway. For AXL, this mechanism has been shown for both antimitotic drugs, and also targeted agents. 281 In lung cancer with a situation of AXL overexpression, AXL will dimerize with EGFR, and by that activate downstream effects leading to limited sensitivity for anti-egfr therapy and acquired resistance for RTK inhibitors. 206 In head and neck cancer there is evidence that AXL over-expression will give inhibition of the signaling of c-abl/p73, and subsequently reduced response to DNA-damage and decreased levels of apoptosis, which in turn will give resistance against DNA-damaging drugs such as cisplatin. 167 Also in malignant melanoma, AXL is important in developing resistance to MAPK inhibitors in B-Raf proto-oncogene serine/threonine kinase (BRAF) -mutant melanomas, and targeted AXL-inhibition increases the effect of treatment with BRAF-inhibitors when given together. 282 1.5.1.11 AXL and Immunotherapy In normal physiology, it is known that AXL is important in dampening the immune system, preventing immune overreactions and auto-immunity. 104 In a cancer setting, with AXL overexpression, this is important, because AXL would facilitate for the tumor cell to survive attacks from the immune system. 283 AXL-inhibition will activate the immune system in several manners. For example, NK-cells will be stimulated to antimetastatic activity and make it more likely for the tumor cells to be killed in an immune reaction. 272 There will also be an increase in activation of dendritic cells, an increase in proinflammatory and anti-tumoral cytokines, and promotion of intra-tumoral infiltration of cytotoxic cells. 284 All these events will make it more likely for a tumor cell

45 to be rejected by the immune system. AXL overexpression is related to development of resistance to PD1-inhibition, and the hypothesis is that AXL inhibition will enhance the effect of checkpoint inhibitors. 285 This is currently in clinical trial (NCT02872259). The potential of AXL inhibition in a cancer immunotherapy setting is addressed in a recent review, and more research to illuminate this further is of great importance. 284 1.5.1.12 AXL and cancer stem cells There is also evidence for a correlation between AXL and CSCs. A study regarding breast cancer stem cells (BCSCs) showed that AXL was capable of inducing EMT, by upregulating expression of Vimentin and N-cadherin and downregulation of E-cadherin in BCSCs. AXL also regulates the tumorigenicity and ability of invasion and migration of BCSCs. 286 Cells with high AXL expression also have high surface-expression of CD44, which is a known marker of stem-cells. CD44 low cells, have less AXL expression, arguing that AXL expression is linked to the development of cancer stem cell traits. 287

46 1.6 Vitamin K Vitamin K was discovered in the 1930s, first described by Henrik Dam, a Danish scientist, who received the Nobel prize for the discovery. The major role of the vitamin is to serve as a co-factor in γ-carboxylation of glutamic acid residues in specific vitamin K dependent proteins (VKDP). The most known VKDPs are related to blood coagulation, but they are also found in bone homeostasis and the vasculature, preventing calcification. 288 The known VKDPs are coagulation factors II, VII, IX and X, protein S, protein Z and protein C, and further GAS6, osteocalcin, matrix Gla protein and Gla rich protein, together with four unknown integral proteins (PRGP1-4). 289 The Gla domain on VKDPs gives the proteins the ability to bind metal-ions. After binding of calcium-ions, the protein will change structural conformation in the matter of gaining the ability to bind to phospholipids. This is necessary for membrane binding, which is a known feature of these proteins. 290 Dietary supplements of vitamin K is essential. Vitamin K is fatsoluble, and has several subtypes, where vitamin K1 (phylloquinone) is the most available in the human diet, through plant-based nutrients, such as green vegetables, grains, fruits and dairy products. 291 Vitamin K2 (menaquinone) is synthesized by bacteria, and can also be available for humans, either from gut flora bacteria, or from bacteria within food, for example cheese, or from liver-products. 291,292 Menaquinone is a group of proteins, where 7 of the subgroups is considered relevant for human intake (menaquinones 4-10). Interestingly, menaquinon-4 can also be produced in mammalians, with a conversion of vitamin K1, or other menaquinones. 293 This reaction is catalyzed by an enzyme called UBIAD1, which was first described in 2010. 294 It is suggested that vitamin K1 mainly is taken up in the liver, and vitamin K2 in arteries and other extra-hepatic locations. 289 As vitamin K is taken up in the body in the quinone form, it is necessary to convert it to the reduced form, hydroquinone, or vitamin KH2 to make it biologically active. This step is catalyzed by the enzyme vitamin K epoxide reductase (VKOR), which was first discovered in 2004. 295 Alternatively, the reaction

47 could be catalyzed by a more specific NAD(P)H-dependent quinone reductase. This enzyme is manly active in the liver. 291 The Vitamin K cycle is illustrated in Figure 9. The enzyme mediating the γ-carboxylation of glutamic residues is called γ-glutamyl carboxylase (GGCX). In the carboxylation reaction, one molecule of Glutamate (Glu) on the VKDPs is converted to γ-carboxyglutamate (Gla), at the same time as one molecule of vitamin KH2 is converted to its inactive form, vitamin K epoxide. (See figure 9). In the conversion of Glu to Gla, one molecule of CO2 is incorporated into the glutamic acid binding to γ-carbon. 290 To be suitable for reuse as carboxylation substrate, vitamin K epoxide has to be converted back to its reduced form, a reaction catalyzed by VKOR. 289 This recycling of vitamin K is preventing clinical vitamin K deficiency in humans, which is a rare condition. 296 The liver is the site of synthesis of coagulation factors, and a substantial part of the vitamin K in the body are stored here. 10 % of the stored vitamin K is vitamin K1, the rest is vitamin K2, mainly MK-7-MK-13. 292 The heart and pancreas also contains comparable levels of vitamin K1 as in the liver, and lower levels are observed in the brain, kidneys and lungs. Vitamin K2 are found at significant levels in the brain, kidney and pancreas. 291 Figure 9: Schematic presentation of hepatic vitamin K metabolism. A: In normal conditions. B: In presence of warfarin. 1) γ-glutamyl carboxylase, 2) VKOR. 3) NAD(P)H-dependent quinone reductase. Figure adapted from 291

48 1.7 Warfarin Warfarin is one of the most used oral anticoagulants worldwide. Statistics from the Norwegian Prescription Database shows that it was approximately 70,000 daily users in Norway in 2015. 297 Several studies show that the drug is used by approx. 2-10% of the adult population, with increasing prevalence with increasing age. 298,299 The clinical available form of warfarin is a racemic mix of equal amounts of (R)- and (S)- enantiomers, where the S-enantiomers is 3 to 5 times more potent. 300 It is almost completely absorbed after oral administration and 98-99% is bound to plasma protein, especially albumin. 301 Warfarin reaches peak plasma concentration after 4 hours. 302 It is almost completely metabolized in the liver, mostly by CYP2C9. 300,301 Warfarin acts as a vitamin-k antagonist. It interferes with the cyclic conversion of vitamin K epoxide to vitamin K. This process is catalyzed by the enzyme vitamin K epoxide reductase (VKOR), which is inhibited by warfarin. Consequently, this will stop the regeneration of vitamin K from its inactive form, and lead to a depletion of active Vitamin K in the body. 301,303 Warfarin shares a common ring structure with vitamin KH2, and can by binding to VKOR inhibit the reductase reaction. 290 There has been a discussion if there is a competitive or non-competitive binding of warfarin to VKOR, but resent research is suggesting a shared binding site for warfarin and vitamin K, and consequently a competitive binding to the enzyme. 304 1.7.1 Cancer protective effects of warfarin in a historical perspective Reports about anti-cancer effects of warfarin have been published occasionally the last 50 years. In 1968, Ryan et al reported a reduced incidence of spontaneous metastases after Coumadin (warfarin) therapy. 305 The group performed a subcutaneous mouse model with anaplastic sarcoma and mammary adenocarcinoma, and evaluated the development of pulmonary metastases. In both models, there was a significant reduction in the metastatic formation in the Coumadin (warfarin) treated group. In the early years, the prevailing hypothesis was that the observed cancer protective effect of

49 warfarin was due to anticoagulation effects. This was also supported by observed cancer protective effects of other anticoagulants, such as unfractionated, and later, low molecular weight heparins. 306,307 Brown et al further strengthened the hypothesis of a cancer protective effect of warfarin in a publication from 1973. The group showed that warfarin effectively reduced the number of lung metastases in a model of KHT sarcoma in mice, and that this effect was abolished when Vitamin K was administered together with warfarin. The authors attributes the effect to the coagulation system, and not as a direct effect of warfarin on the tumor cells. 308 Also McCulloch et al concluded that the observed cancer protective effect of warfarin was related to anticoagulation. 309 There was little knowledge about the molecular effect behind the anti-coagulative properties of warfarin during these first years. The connection with warfarin and Vitamin K was not properly established until a publication from O Reilly et al in 1976, where they acknowledge the inhibitory effect of warfarin on the conversion of Vitamin K epoxide, to Vitamin K. 310 In parallel, theories about a direct warfarin effect on the tumor cells gradually started to emerge. In 1985, Goeting et al. observed an effect on colorectal cancer incidence in rats, both when warfarin was administered in doses affecting the coagulation system, but also in a lower, subclinical dose. The authors suggested that the observed effect was not related to the coagulation system, rather a molecular effect on early stage neoplastic changes. 311 The coupling between warfarin and AXL inhibition was made in 1999, first from warfarin mediated AXL inhibition of mesangial cell proliferation in the kidneys. 312 This connection was further established when Tsou et al showed that γ-carboxylation was necessary for AXL activation and that this process could be inhibited by warfarin 115 1.7.2 AXL and warfarin As described before, the RTK AXL, is dependent on the vitamin K dependent carboxylation of its ligand GAS6 for biological activation. By its inhibition of VKOR, warfarin will give a depletion of vitamin K followed by prevented carboxylation of GAS6 and subsequent inhibited receptor activation. Non-carboxylated GAS6 will have full

50 capability of binding to the receptor, but will function as a selective AXL-antagonist, as it in this state will not be able to induce receptor activation. 313 Warfarin has also been shown to enhance activity of NK-cells, at doses not affecting the coagulation system. 272 1.8 Health registries A health register is defined as a collection of health information, systematically stored and organized so it will be possible to retrieve health information about individuals. 314 In total there are 15 centrally administered health registries in Norway, covering topics as birth, drug prescriptions, cancer occurrences and causes of death. 315 The registries are regulated by the Norwegian law on Health registries, Lov om helseregistre og behandling av helseopplysninger. 316 The free, universal health care system in Norway gives the Norwegian health registries a broad, almost complete, coverage. All inhabitants of Norway are assigned their own unique identification number, registered in the Norwegian National registry, a system in use since 1968. 317 The use of this number is allowing an individual-level linkage of information from different health registries. This gives a wide variety of opportunities to conduct experiments and extract information from these sources. 1.8.1 The cancer registry of Norway The Cancer registry of Norway (CRN) was established in 1951 and is after the cause of death registry the oldest health registry in Norway. The registry is collecting information on all cancer cases in Norway, and all medical doctors are instructed by law to notify new cancer cases to the registry Also cases with cancer suspicion without verified diagnosis, and diagnoses revealed at autopsy should be reported to the registry. 318 The overall coverage is estimated be over 98%. It records detailed information including demographic information, diagnosis, death by cancer, morphology, stage and topography. The registry uses International classification of diseases (ICD), ICD-7, and ICD-10 for diagnosis classification and the international classification of diseases for

51 oncology, 3rd edition (ICD-O-3), for morphological classification of the different lesions. 319 1.8.2 The Norwegian prescription database The Norwegian prescription database (NorPD) was established in 2004, and collects data from all prescribed drugs in Norwegian pharmacies. 320 The registry does not include information on drugs used by hospitalized or otherwise institutionalized patients, nor over the counter drugs, bought without prescription. The registry is using the Anatomical Therapeutic Chemical classification system for classification of the different drugs. 320 When using NorPD for research, the database has several strengths compared to other sources of drug use, with no recall bias, and no primary noncompliance. This will improve the validity of the data of interest. 321

52 2. Aims of the study The central hypothesis of this thesis is that AXL mediated signaling is important for the development and progression of cancer, and that warfarin-mediated AXL inhibition effectively blocks tumor initiation and malignant progression in different cancers. Main aim: Characterize the role of warfarin-mediated AXL-inhibition in the development and progression of cancer. Specific aims: Aim I: Conduct AXL inhibition in pancreatic cancer through warfarin mediated inhibition of GAS6-carboxylation in different mouse models. (Paper I) Aim II: Characterize AXL expression and expression of EMT markers in in vitro pancreatic models, with and without AXL inhibition, to investigate the relationship between AXL expression and EMT. (Paper I) Aim III: Characterize cancer incidence in warfarin users compared to non-users to illuminate the potential effect of warfarin-based AXL inhibition in cancer development. Conduct a prospective population based cohort study, using the Norwegian Cancer registry and the Norwegian prescription database. (Paper II)

53 3. Summary of papers Paper I In this paper, we report that the vitamin-k antagonist warfarin blocks GAS6-mediated activation of the receptor tyrosine kinase AXL in different models of pancreatic cancer. This inhibition reduces both progression and spread of pancreatic cancer. In vivo experiments were performed with human cancer cells and immunocompromised mice, murine cells and immunocompetent mice, and with genetically engineered mouse models (GEMMs), that spontaneously developed PDAC. In all models, we saw a major reduction in the formation of metastases. We also demonstrated an increased effect of gemcitabine when given in combination with warfarin, both in primary tumor growth and metastatic development. The effect of warfarin mediated AXL-inhibition was verified also with other AXL-blocking agents. Through in vitro-experiments, we demonstrated that levels of phosphorylated AXL went down after treatment of warfarin. In addition, phosphorylation of AKT, downstream of AXL, was increased when stimulated with GAS6, and reduced after addition of warfarin. This confirmed that warfarin treatment reduced AXL signaling. Further, the cells ability to form colonies were significantly reduced after warfarin treatment, and these results were confirmed by AXL-knock down cell lines. We also established the close link between AXL and EMT in line with previously published material. Levels of EMT markers were influenced after warfarin-mediated AXL-inhibition, with vimentin being downregulated and E-cadherin being upregulated in warfarin-treated samples. Paper II We report in this paper a clear association between warfarin use and cancer. We defined a cohort with patients from the Norwegian national registry coupled with the Cancer Registry of Norway and the Norwegian Prescription Database to look at the incidence of cancer in warfarin users compared to non-users. We observed a

54 significantly lower cancer incidence in the warfarin user group, with an incidence rate ratio (IRR) for overall cancer of 0.84 (95% Confidence interval (CI) 0.82-0.87) adjusted for sex and age. We further observed a lowered IRR in several of the organ-specific cancer sites, such as prostate, lung and bladder cancer. We also performed a subgroup analysis on patients prescribed warfarin for atrial fibrillation/flutter (AF-group), to eliminate the possible confounding effect of occult malignancy after venous thromboembolism and pulmonary embolism. In these analyses the IRR in overall cancer were even lowered with an IRR of 0.62 (95% CI 0.59-0.65). Also for specific cancer sites, IRR were lowered in the user group compared to the non-users. The findings in this study supports the findings from Paper I, and emphasize the potential of warfarin use in an anti-cancer setting.

55 4. Methodological considerations 4.1 Animal experiments The animal experiments in Paper I has been performed with different approaches. In the experiments, we have used immunocompetent animals, immunocompromised animals and also genetically engineered mouse models (GEMMs) 4.1.1 Cell line xenograft models These models depend on the use of human cancer cell lines in mouse models, which is a common way of modeling cancer development. 322 A prerequisite for this model is the use of immunodeficient mice, to prevent rejection of the injected human cells. This way of modeling cancer has several challenges. The use of immortalized cell lines will not reflect the diversity in a normal tumor, as these cells are preselected cells grown in a favorable environment and often with a different gene expression profile than primary tumor cells. 323 Furthermore, the immune system is important in normal cancer development, and in these models, this factor is eliminated. This makes it impossible to investigate immune-targeted therapies in this setting. Advantages of this model system are the access to numerous and well established cell lines, in many different tumor types. These cells can be injected both subcutaneously and orthotopically. 4.1.2 Syngeneic models In this model, murine cancer cells are transplanted into mice, using a immunocompetent host. With an intact immune system, this will mimic a more realistic tumor environment, including stromal cells and tumor vasculature. Disadvantages of the model is that it is less available, as fewer cell lines suitable for the purpose exists.

56 The clinical translation, with both the tumor and the host being of another species than human is another challenge. 324 4.1.3 Genetically engineered mouse models In these cancer models, the mice are genetically altered, so that they spontaneously will develop the tumor of interest. This has several advantages with the tumor developing in the tissue of origin, and preservation of an intact immune system. The tumor microenvironment will also be intact, with all components such as immune cells, vascular and stromal cells. 325 The major disadvantage of this model system is the complexity of developing the mouse model. 324 4.2 Mouse strains in use in our work NOD/SCID: Non-obese diabetics/ Severe combined immunodeficient mice. This mouse strain is immunocompromised, due to impaired development of T and B cell lymphocytes. In addition, these animals have reduced NK-cells function. 326 These mice are widely used, both for tumor biology and xenograft research. 327 In our work, the strain was used for orthotopic implantation of human cancer cell lines. C57/Bl6: This is an immunocompetent inbred mouse-strain. The strain was first bred in the Jackson laboratory in 1948. With a normal immune system, these mice are widely used in research in the fields of both immunology and cancer. They are robust and longlived compared to other cancer models. 328 We used this mouse model, in order to implant murine cancer cells, which gave us the ability to study the progression of PDAC in a model with a functioning immune system. LSL-Kras G12D. ; Cdkn2a lox/lox. ; p48 Cre (KIC): Genetically engineered mouse strain. This model takes advantage of the pancreas selective transcription factor p48 (Ptf1a). This transcription factor drives the expression of Cre in pancreatic cells. The LSL-Kras G12D

57 gives a mutant Kras, which due to the p48 Cre mutation is specifically expressed in the pancreas, promoting the development of PDAC. Furthermore, the p16/p19 (Ink4a/Cdkn2a) locus is deleted, a common mutational loss in human PDAC. This gives the tumor an aggressive phenotype, with poorly differentiated tumor cells. 329 The development of pancreatic carcinoma in this model is 100% at 4 weeks of age. The model has histopathological features that is consistent with the development and progression of human PDAC, and therefore works well as a model system. 330,331 The mice were bred in the animal facility of UT Southwestern, Dallas, and genotyped shortly after birth. We used the following pancreatic cell lines for our experiments: Human cell lines: AsPc-1: This cell line is derived from cells from the ascites of a patient with pancreatic cancer. 332 Panc-1: This cell line was derived from a primary tumor. The cells do not express significant amounts of carcinoembryonic antigen. 332 C5LM2: A variant of Panc1 cells. The cell line was developed by two passages of in vivo growth and culture of liver metastases from a primary pancreatic cancer. The cell line was developed in the Brekken laboratory. Mia PaCa2: Cells from primary tumor of pancreatic cancer. This tumor did not express carcinoembryonic antigen. 332 Capan1: Derived from a PDAC liver metastasis. 332 Capan 1 does not express AXL. Included as a control cell line. Murine cell line: Pan02: Murine cell line, established from pancreatic tumor in a C57/Bl6 strain. Widely used for research on pancreatic cancer. 333

58 All cell lines for animal experiments were grown in a humidified atmosphere with 5% CO2, at 37 C. AsPC-1, Panc-1, Pan02 and MiaPaCa-2 lines were grown in Dulbecco s modified eagle medium (DMEM). Capan-1 cells were grown in Iscove s modified Dulbecco s medium (IMDM). Before implantation, all cell lines were confirmed to be mycoplasma free using e-myco kit (Boca scientific). 4.3 In vivo experiments Animal experiments were performed at University of Texas Southwestern, Dallas, Texas. All animals were housed in a pathogen free facility. The animals had 24-hour access to food and water. All cells were injected orthotopically. For AsPc-1, Panc-1, Mia PaCa2, Capan1 1x10 6 cells were injected, and for Pan02 cells 1x10 5 cells were injected. 4.3.1 Medical treatment of animals The animals were randomized to receive normal drinking water, or water containing warfarin. For immunocompromised mice the warfarin concentration was 1 mg/l (3,0μM). Immunocompetent mice received 0,5 mg/l (1,5μM). Warfarin containing water was in all cases renewed every 3 days. The warfarin treatment were administered with or without gemcitabine 25 mg/kg twice weekly. For Mia Paca2 tumor bearing mice, also 10C9 (250μg ip. twice/week) were given in addition to gemcitabine. The GEMMs started warfarin treatment at three weeks of age. The warfarin treatment continued for 4 weeks until sacrifice. Mice implanted with Panc-1, Capan-1, C5LM2 and Mia Paca2 tumors received warfarin therapy for 6 weeks until sacrifice. AsPc1-bearing mice received 4 weeks of therapy. Pan02 bearing mice received 3 weeks of therapy. All animals were sacrificed when control animals started to be moribund. Differences in aggressiveness and growth rate between the models was the reasons for the varying treatment lengths in the different models.

59 Dosing of warfarin Yanagita et al have shown that warfarin can inhibit GAS6-mediated inhibition of AXL at concentrations below those necessary for affecting the coagulation cascade. 334 In this paper the researchers administered warfarin in drinking water to rats at 0,25-0,5 mg/ml. this gave corresponding serum concentrations of 0,28-1,23 μmol/l. No corresponding anemia, increased bleeding tendency or prolongation of prothrombin time were observed. In Paper I, we administered warfarin in the drinking water with 1mg/L for immunocompromised mice and 0,5 mg/ml in immunocompetent mice. The rationale for different dosing was an observed toxicity in the immunocompetent mice during pilot experiments. We aimed for a dosing with no anticoagulative effect, and no bleeding complications were observed during the experiments. 4.3.2 Measurements of primary tumor burden and metastases The primary tumor burden was measured by weighing pancreas and tumor en block. Metastases were macroscopically counted by visual inspection of liver, diaphragm peritoneal surfaces and the abdominal cavity. Metastatic burden was further confirmed with H&E staining of liver sections. 4.4 Induction of EMT In Paper I, we evaluated the relationship between EMT and AXL, and how AXL expression and inhibition would influence on this process. To establish conditions mimicking EMT, cells were grown on chamber slides coated with collagen, and with addition of TGF-β to the media. This is an established method to induce EMT in artificial environments, and is confirmed by an upregulation of vimentin and a downregulation of E-cadherin. 335 4.5 Register study In Paper II we performed a register study taking advantage of two of the major health registries established in Norway. By using the Norwegian identification number, it was possible to couple a cohort from the Norwegian National registry with the Cancer

60 Registry of Norway and the Norwegian prescription database at an individual level. To be able to perform the coupling we obtained approval from the following instances, in addition to the registries in question: Regional Committees for Medical and Health Research Ethics, The Data protection official for research at University of Bergen and the Norwegian Data Protection Authority. 4.5.1 The coupling process of different registries The coupling of information is possible because of the national identification number. The process requires coordination between different instances, in respect of making the process as quick and smooth as possible. NorPD is a so-called pseudonymous registry. Pseudonymization is the situation where the normal person identifier such as name or personal identification number, is replaced with a pseudonym. This pseudonym is unique for each individual, but will not have any relation to the original identifier of the person. 336 Pseudonyms can be used as personal identification in the coupling process. Because the process included the NorPD, the coupling process had to end and be administered by this registry. Figure 10 illustrates the data collection, and coupling of data between the different instances. Figure 10. Illustration of the coupling process.

61 4.6 Statistics For Paper I, statistical analyses were performed using the software GraphPad Prism (GraphPad Prism version 4.00 for windows). Results were expressed as mean +- s.e.m. of s.d. Data were analyzed by t-test or ANOVA and results were considered significant at p<0.05. For Paper II, statistical analyses were performed using the software STATA IC 13.1 and STATA IC 14. IRRs were calculated by the method of Mantel-Haenzsel, and adjusted for sex and age. Observed IRRs was considered statistically significant if CI did not include 1.

62 5. Discussion In Paper I, we evaluated the role of the RTK AXL in PDAC development and metastasis. Our findings in this paper supports the hypothesis of AXL being an important driver of metastatic formation and cancer progression. Inhibition of AXL leads to decreased expression of EMT markers, which further supports the theory of AXL being important for the metastatic processes in cancer. Our findings in Paper II supports the preclinical findings of the role of AXL in cancer, and shows that AXL also is important in the cancer initiation process, as well as in more advanced stages of the disease. 5.1 The role of AXL in the development and metastasis of pancreatic ductal adenocarcinoma PDAC is one of the most lethal of all cancer forms, with no effective treatment regimen. Any research that can contribute to further developments in the treatment of pancreatic cancer is highly appreciated. The receptor tyrosine kinase AXL is associated with many different cancers, including pancreatic cancer, where AXL overexpression is correlated with reduced overall survival and worsened prognosis. 244 In the literature, AXL has been strongly associated with increased invasiveness, and metastatic formation. 118,188,337 This is reported also in pancreatic cancer, both from tumor samples, and in cell culture experiments. 243 AXL inhibition has been suggested as a potential treatment option, and different strategies to achieve AXL inhibition includes development of small molecule AXL inhibitors, antibody-neutralization of GAS6, the ligand of AXL, and as we performed in Paper I, inhibition of γ-carboxylation of GAS6, with a subsequent AXL inhibition. 157,338 Previous in vitro experiments have demonstrated that warfarin strongly inhibits γ-carboxylation of the Gla-domain of GAS6, with a subsequent diminished receptor activation. 115 Interestingly, the warfarin doses needed for inhibition of γ-carboxylation of the Gla-domain on GAS6, are shown to be at levels beneath those needed for anticoagulation. 272

63 In Paper I, we demonstrated in vitro that warfarin, as expected, inhibited AXL signaling, and that this inhibition could be repealed when vitamin K was added to the media. This demonstrates that the AXL-inhibitory effect of warfarin are through the depletion of Vitamin K. We also observed a decreased expression of GAS6 and γ-carboxyglutamyl when the cells where treated with warfarin, which further confirmed the mechanisms of action (Paper I, Figure 2 A-B). The levels of active phosphorylated AXL were significantly reduced after warfarin treatment, and similarly with treatment with the AXL inhibitor BGB324. Nevertheless, warfarin treatment did not influence on the levels of total AXL in the cells (Paper I, Figure 2 C). We also observed a decreased signaling in phosphorylated AKT downstream of AXL after warfarin treatment, confirming a reduced AXL-mediated signal transduction (Paper I, Figure 2 D-E). We further observed a rescue effect of pakt after addition of GAS6. Knowing that AXL expression is correlated with increased migration 158,286, we conducted a migration assay, showing reduced levels of migration in the warfarin treated cells, but only in the AXL expressing cell lines. In Capan- 1 cells, which do not express AXL, we observed no difference in migratory capacity (Paper I, Figure 2 F). Altogether these findings confirm that AXL expression is important in pancreatic cancer, and that warfarin mediated AXL inhibition will reduce expression of both phosphorylated AXL and its ligand GAS6, which will diminish the migratory and invasive capacity of the cells. Furthermore, we demonstrated in Paper I that warfarin-mediated AXL inhibition reduced metastatic formation in vivo, in 5 different murine PDAC models. (Paper I, Figure 1B) In a syngeneic and a genetically engineered mouse model we also observed a statistically significant effect in primary tumor growth (Paper I, Figure 1A). The findings were confirmed, observing similar results using an AXL-knockdown cell line, and by selective AXL inhibition by 10C9, an AXL antibody (Paper I, Figure 1 F-G). We also observed a reduction in the colony-formation ability of Mia PaCa-2 cells grown in the presence of warfarin (Paper I, Figure 3 A-C). The abilities of a cell to grow surfaceindependent and form colonies are important features of a cancer cell. 339 Our findings was in line with previously published material regarding the ability of anchorage

64 independent growth after AXL-inhibition. 243 Summarized, our findings establishes AXL as a facilitator for the metastatic potential of pancreatic cancer. Pancreatic cancer has a high level of metastatic formation, and it has been proposed that inhibition of the metastatic process has a great potential in extending life expectancy. 340 The results from Paper I suggest warfarin-mediated AXL inhibition as a possible treatment option in pancreatic cancer. We also observed an increased reduction in metastatic formation when warfarin was administered together with the established treatment-option, gemcitabine (Paper I, Figure 3 F-I). The enhanced effect of established treatment corresponds to what has earlier been described in other cancers, like synergistic effects of AXL-inhibition and cisplatin in the suppression of liver metastases in breast cancer, and increased cell death in AXL-knockdown cells after treatment with cisplatin, carboplatin or doxorubicin in lung carcinoma. 157,233 The observed additive effect of warfarin and gemcitabine in Paper I increases the potential of warfarin-treatment in a clinical setting of pancreatic cancer. In Paper I, but also from other groups, it was demonstrated that AXL-inhibition and subsequent anti-cancer effects of warfarin could be achieved when administered in doses not affecting the coagulation system. 272,334 VKDPs located outside the liver are shown to be under-carboxylated in the adult population, leaving this population with a sub-clinical Vitamin-K deficiency. 293 Additionally, it is shown in the liver that another enzyme, NAD(P)H-dependent quinone reductase, can function as an alternative facilitator for the recovering of vitamin K from the epoxide metabolite, when VKOR is inhibited by warfarin. 291 Thus, low doses of warfarin will be able to inhibit VKDPs in peripheral tissues where the steady state concentration of vitamin K is low, while the coagulation-related VKDPs in the liver still will be carboxylated, retaining their biological activity. This finding also supports the theory that the cancer protective effect of warfarin is not due to anticoagulation effects. The possibility of warfarin administration in a low-dose manner is a great advantage for the potential use of the drug in an anticancer setting. The ability of using the drug without unwanted side effects from the coagulations system makes it more suitable as a treatment option for cancer patients.

65 In Paper I, we demonstrated a reduction in metastatic growth in all the investigated mouse models. However, we also observed an inhibition of primary tumor growth in the syngeneic and genetic models. (Paper I, Figure 1A). These models have, in contrast to the orthotopic models, an intact immune system. In normal physiology, AXL is important in moderating effects of the immune system and preventing auto-immune reactions. 156 When AXL is inhibited, the anti-cancer effects are effectuated via two systems, both direct tumor effects and indirect immunological effects because of repealed AXL-mediated inhibition of the immune system. (Figure 11). Figure 11: TAM signaling in cancer therapy. Adapted from 284 Park et al, and later Paolino et al, demonstrated an increased cytotoxic ability of NKcells, after AXL inhibition. 272,341 However in a recent paper, the researchers argue that AXL signaling will increase effector function of NK cells, so this remains still debated. 342 In the model systems in Paper I, the orthotopic models will only benefit from the direct tumor effects seen in figure 11, as these mice are not immunocompetent. The syngeneic and genetically engineered models will benefit both from the direct tumor effects, but