Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
THE ROLE OF THE EPITHELIAL-MESENCHYMAL TRANSITION IN AGGRESSIVE TUMOUR PHENOTYPES by Alexandria Marie Haslehurst A thesis submitted to the Department of Pathology and Molecular Medicine In conformity with the requirements for the degree of Doctor of Philosophy Queen’s University Kingston, Ontario, Canada (October, 2014) Copyright ©Alexandria Marie Haslehurst, 2014 Abstract The epithelial-mesenchymal transition (EMT) is an evolutionarily conserved developmental process characterized by the loss of intercellular contacts, changes in cell polarity and increased migration and invasion. EMT has recently been shown to have a significant impact on chemotherapy response and metastatic progression in cancer cells. We hypothesized that because EMT contributes to aggressive cancer phenotypes, expression profiles of key EMT modulators could be profiled in ovarian and prostate tumours to distinguish aggressive from nonaggressive disease. Our results demonstrated that increased EMT gene expression was correlated with chemotherapy resistance in ovarian cancer, and with higher Gleason pattern in prostate cancer, confirming an association between EMT and increased risk for cancer progression. This underscores the potential for EMT gene signatures to be used as clinically relevant predictive and prognostic biomarkers. Functional studies using in vitro models of Cisplatin-resistant ovarian cancer and invasive prostate cancer demonstrated that inhibition of EMT signaling, both through the dysregulation of key transcription factors (SNAI1, SNAI2, ZEB1), and repressed expression of phenotypically important genes such as VIM, SPARC, CLTC and HIF1A resulted in loss of aggressive attributes in these cell lines. This included re-sensitization of our ovarian cancer cells to the effects of Cisplatin, while significantly reducing cell migration and invasion in our prostate cancer cells, both in 2- and 3-dimensional culture models. Furthermore, using our prostate cancer knockdown cell lines, we were able to assess gene expression changes in response to single gene repression. As a result, those genes that appear to have the highest degree of connectivity within the context of our EMT network were identified, thus contributing to the overall understanding of EMT regulation. ii Ultimately, we have been able to assess the role of the epithelial-mesenchymal transition in aggressive tumour behaviours, such as resistance to chemotherapy and progression towards metastatic events. The results of this work indicate that gene expression signatures associated with EMT may be useful as predictive and prognostic biomarkers in various cancer types. Furthermore, gene knockdown studies have provided insight into EMT signaling network regulation, as well as the functional contribution of a number of genes to alterations in migratory and invasive cancer cell programs. iii Co-Authorship Below is a summary in recognition of the contributions of others to various chapters of this thesis. Chapter 3: The role of EMT in chemotherapy response in ovarian cancer Haslehurst, A et al. EMT transcription factors Snail and Slug directly contribute to Cisplatin resistance in ovarian cancer. BMC Cancer 2012, 12:91. The following authors were listed on this publication: Alexandria M Haslehurst, Dr. Madhuri Koti, Dr. Moyez Dharsee, Dr. Paulo Nuin, Dr. Ken Evans, Dr. Joseph Geraci, Dr. Timothy Childs, Dr. Jian Chen, Jieran Li, Dr. Johanne Weberpals, Dr. Scott Davey, Dr. Jeremy Squire, Dr. Paul C Park and Dr. Harriet Feilotter. I carried out analysis of cell line microarray data, completed in vitro studies including cell migration, invasion and drug sensitivity assays, as well as gene knockdown experiments, and drafted the manuscript. Dr. Madhuri Koti provided microarray data for the primary tumours. Dr. Moyez Dharsee, Dr. Paulo Nuin and Dr. Joseph Geraci assisted with bioinformatic analysis of microarray data. Dr. Moyez Dharsee and Dr. Paulo Nuin also designed peptide sequences for SRM-MS. Dr. Ken Evans, Dr. Jian Chen and Jieran Li carried out targeted protein quantification via SRM-MS and subsequent analysis. Dr. Johanne Weberpals carried out classification of primary tumours as drug sensitive or resistant using clinical data. Dr. Timothy Childs examined the primary tumours and calculated the percent tumour cells. Dr. Scott Davey provided guidance on biological interpretations. Dr. Jeremy Squire and Dr. Harriet Feilotter conceived of the study, and participated in its design. Dr. Scott Davey, Dr. Jeremy Squire, Dr. Paul C Park and Dr. Harriet Feilotter and I all made significant intellectual contributions to this study. All authors contributed to the editing of this manuscript. Chapter 4: Evaluating the role of EMT as an indicator of metastatic potential in prostate cancer The following authors contributed significantly to this work: Alexandria M Haslehurst, Dr. Robert Gooding, Andrew Day and Dr. Paul C Park. Dr. Park completed pathology review of the prostate tissues in our study cohort. Mr. Day and Dr. Gooding completed all normalization and statistical modeling of our nCounter expression data, and identified the multivariate model that is discussed in this chapter. I performed all experiments outlined in this study. Experimental and study design were completed by Dr. Park and I. iv Acknowledgements First and foremost I would like to thank my supervisors, Drs Harriet Feilotter and Paul Park, for their constant support and guidance over the past five years. I am truly grateful for all that you have taught me, and more importantly, all that you have encouraged me to teach myself. As I move forward with my academic training, and eventual career, I will look back on the time that I have spent working with you both with an extreme sense of fondness. Thank you again, for everything. To my Park/Feilotter lab-mates, thank you all for enriching my experience at Queen’s with your friendships and contributions to my academic pursuits. Though I have only had the privilege of working with many of you for a short period of time, I have certainly taken something away from our interactions, both on a personal and professional level. I wish nothing but the best for each of you as you move forward with your training and future career endeavours. A special thank you to our collaborators, Dr. Bob Gooding and Andrew Day – without whom, I would have been lost. To my wonderful friends, both within the Queen’s community, and those of you from outside Kingston – I cannot express how appreciative I am for all your support over the past five years. Your encouraging words and practiced ability to nod along in understanding while I tell you about my research have meant the world to me. I would specifically like to thank Jill Baker for being undoubtedly the most amazing and supportive person I have ever encountered. Your willingness to listen to me rant and brag in equal measure, at any time and for any reason, has provided me a much needed outlet during this process, and I am eternally grateful. Finally, my family – to my grandparents, aunts, uncles and cousins – thank you for all of the encouragement and kind words that have been sent in my direction as I have pursued my degree. I truly recognize how fortunate I am to have such an immense support system, and will be forever thankful to have you all in my life. v To my sister Stephanie, your love and support have meant the world to me. Thank you for reminding me that there is more to life than ‘being a nerd’ – it has been an important lesson in finding balance. Last but certainly not least – Mom and Dad – from the time I was very young you encouraged me to follow my passions and to strive to achieve any goal that I ever set for myself. While I know you have not always understood my academic pursuits, you have continually engaged me in conversations about my research, and your pride in my accomplishments has always been recognized and appreciated. Thank you for teaching me the value of hard work and perseverance, those qualities have served me well over the course of my graduate studies. I love you both. vi Statement of Originality I hereby certify that all of the work described within this thesis is the original work of the author. Any published (or unpublished) ideas and/or techniques from the work of others are fully acknowledged in accordance with the standard referencing practices. Alexandria Marie Haslehurst October, 2014 vii Table of Contents Abstract ............................................................................................................................................ ii Co-Authorship ................................................................................................................................ iv Acknowledgements .......................................................................................................................... v Statement of Originality................................................................................................................. vii List of Figures ................................................................................................................................. xi List of Tables ................................................................................................................................. xii List of Abbreviations .................................................................................................................... xiii Chapter 1 : Introduction ................................................................................................................... 1 1.1 The Epithelial-mesenchymal transition ................................................................................. 1 1.1.1 EMT overview ................................................................................................................ 1 1.1.2 Current model of EMT regulation during cancer progression ........................................ 3 1.1.3 The mesenchymal to epithelial transition (MET) ......................................................... 12 1.1.4 EMT and cancer stem cells ........................................................................................... 13 1.1.5 EMT as a biomarker of cancer metastasis and chemotherapy resistance ..................... 14 1.1.6 Targeting EMT for cancer therapy................................................................................ 16 1.2 Ovarian Cancer .................................................................................................................... 18 1.2.1 Overview ....................................................................................................................... 18 1.2.2 Diagnosis and Prognosis ............................................................................................... 19 1.2.3 Treatment and Chemotherapy Resistance ..................................................................... 21 1.2.4 Molecular mechanisms contributing to platinum resistance in ovarian cancer............. 22 1.3 Prostate Cancer .................................................................................................................... 25 1.3.1 Introduction to prostate cancer ...................................................................................... 25 1.3.2 Molecular events in prostate cancer initiation and progression .................................... 25 1.3.3 Natural history of prostate cancer: What are the true risks? ......................................... 27 1.3.4 Clinical aspects of prostate cancer diagnosis, prognosis and treatment options ........... 29 1.3.5 Challenges associated with current clinical management of prostate cancer ................ 35 1.3.6 EMT as a factor in prostate cancer progression ............................................................ 38 Chapter 2 : Rationale, Hypothesis and Objectives ......................................................................... 40 Chapter 3 : EMT transcription factors, Snail and Slug, contribute to Cisplatin resistance in epithelial ovarian cancer ................................................................................................................ 43 3.1 Abstract ................................................................................................................................ 43 3.2 Introduction .......................................................................................................................... 45 viii 3.3 Materials and Methods ......................................................................................................... 47 3.4 Results .................................................................................................................................. 53 3.4.1 Cisplatin resistance in A2780cis correlates with changes in cellular morphology consistent with EMT signaling .............................................................................................. 53 3.4.2 Whole transcriptome profiling and LC/SRM-MS analysis identify components of EMT signaling networks in A2780cis relative to A2780 ................................................................ 55 3.4.3 Cisplatin resistant cells display increased potential for migration and invasion ........... 58 3.4.4 Knockdown of Snail and Slug reverses the EMT phenotype and reduces cellular resistance to Cisplatin ............................................................................................................ 58 3.4.5 Drug resistant human ovarian tumours can be differentiated from drug sensitive ovarian tumours using a biomarker panel of EMT-related genes .......................................... 62 3.5 Discussion ............................................................................................................................ 65 3.6 Conclusions .......................................................................................................................... 67 Chapter 4 : Evaluating the role of EMT as an indicator of metastatic potential in prostate cancer68 4.1 Abstract ................................................................................................................................ 68 4.2 Introduction .......................................................................................................................... 70 4.3 Materials and Methods ......................................................................................................... 72 4.4 Results .................................................................................................................................. 82 4.4.1 A data mining and systems biology-based approach to the development of a comprehensive EMT biomarker panel ................................................................................... 82 4.4.2 Genes associated with EMT signaling are differentially expressed between Gleason pattern 3 and Gleason pattern 4 prostate cancers ................................................................... 85 4.4.3 Principal component analysis combined with logistic regression modeling identifies a multivariate model capable of differentiating between Gleason pattern 4 and pattern 3 prostate cancers ...................................................................................................................... 88 4.4.4 Functional validation of EMT-associated genes ........................................................... 92 4.4.4.1 Inhibited expression of Snail, ZEB1 and HIF1α, but not SPARC, VIM or CLTC, results in loss of mesenchymal morphological characteristics .......................................... 94 4.4.4.2 Assessment of EMT gene knockdown on cell migration....................................... 96 4.4.4.3 Assessment of EMT gene knockdown on invasive characteristics in 2D and 3D cultures ............................................................................................................................... 98 4.4.5 Evaluation of frequently dysregulated gene expression in response to EMT-associated gene knockdown .................................................................................................................. 103 4.5 Discussion .......................................................................................................................... 105 ix 4.5.1 The utility of EMT-associated gene expression in differentiating between Gleason pattern 3 and Gleason pattern 4 prostate cancers ................................................................. 105 4.5.2 Assessment of gene function in EMT-associated phenotypes .................................... 108 4.5.2.1 SPARC ................................................................................................................. 109 4.5.2.2 Vimentin (VIM) ................................................................................................... 110 4.5.2.3 Clathrin, heavy chain (CLTC) ............................................................................. 112 4.5.3 Evaluating the relationships between EMT-associated genes and their impact on the signaling networks regulating EMT ..................................................................................... 114 4.6 Conclusions ........................................................................................................................ 117 Chapter 5 : General Discussion.................................................................................................... 119 5.1 Revisiting EMT in the regulation of chemotherapy response in ovarian cancer ............... 119 5.2 The tumour microenvironment and EMT .......................................................................... 122 5.3 EMT and cancer stem cells ................................................................................................ 122 5.4 The utility of EMT as predictive or prognostic biomarkers in cancer ............................... 125 5.5 Targeting EMT in the therapeutic management of aggressive cancer ............................... 128 Chapter 6 : Future Directions ....................................................................................................... 133 Chapter 7 : Concluding Remarks ................................................................................................. 136 References .................................................................................................................................... 139 Apendices..................................................................................................................................... 168 Appendix A: Supplementary Materials (Chapter 3) .................................................................... 168 Appendix B: Detailed report on the approach to multivariate model identification .................... 169 Appendix C: Supplementary Materials (Chapter 4) .................................................................... 171 x List of Figures Figure 1.1 Summary of EMT characteristics ................................................................................... 2 Figure 1.2 Summary of the current model of EMT regulation ........................................................ 5 Figure 1.3 Snail-mediated repression of E-cadherin ........................................................................ 8 Figure 1.4 Cellular processes implicated in platinum resistance ................................................... 24 Figure 1.5 Representation of true and false 'low-risk' prostate biopsies ........................................ 37 Figure 3.1 Morphology of A2780 and A2780cis cells................................................................... 54 Figure 3.2 Upregulation of gene associated with EMT in resistant cells....................................... 56 Figure 3.3 Knockdown of Snail and Slug reverses EMT phenotype ............................................. 60 Figure 3.4 Cisplatin sensitivity with Snail and Slug knockdown .................................................. 61 Figure 3.5 EMT gene signatures in drug resistant ovarian tumours .............................................. 63 Figure 4.1 Overlapping gene expressions from invasive prostate cancer cell lines. ...................... 83 Figure 4.2 Overlap between genes dysregulated during prostate cancer cell invasion and genes associated with EMT...................................................................................................................... 84 Figure 4.3 Differential EMT-associated gene expression in Gleason pattern 4 versus pattern 3 prostate cancer ............................................................................................................................... 87 Figure 4.4 ROC-AUC analysis of the first three principal components ........................................ 89 Figure 4.5 Fit plot demonstrating the predictive power of the first and third principal components ..................................................................................................................... 90 Figure 4.6 Validation of gene knockdown ..................................................................................... 93 Figure 4.7 Cell morphology in 2D culture ..................................................................................... 95 Figure 4.8 Impact of gene knockdown on cell migration .............................................................. 97 Figure 4.9 Impact of gene knockdown on cell invasion ................................................................ 99 Figure 4.10 Impact of gene knockdown on colony formation and invasion in 3D culture .......... 102 Figure 4.11 Identification of key nodal points in EMT regulation .............................................. 104 Figure 5.1 Potential mechanisms contributing to EMT-regulated chemotherapy resistance in high grade serous ovarian cancer ................................................................................................. 121 xi List of Tables Table 1.1 Summary of FIGO staging system for ovarian malignancies ........................................ 20 Table 1.2 TNM staging criteria for prostate cancer ....................................................................... 32 Table 1.3 Risk stratification in prostate cancer .............................................................................. 33 Table 3.1 Summary of SRM-MS quantification results for EMT-associated proteins .................. 57 Table 3.2 List of genes used for unsupervised hierarchical clustering of primary tumours .......... 64 Table 4.1 Genes demonstrating the highest degree of correlation between PC three and signal intensity.......................................................................................................................................... 91 Supplementary Table S3.1 Clinical data for ovarian tumour samples…………………………..169 Supplementary Table S4.1 NanoString probe sequences……………………………………….172 Supplementary Table S4.2 Primer sequences for multiplex target enrichment…………………176 Supplementary Table S4.3 Targets for EMT gene expression profiling in prostate cancer…….180 Supplementary Table S4.4 Altered EMT-associated gene expression in response to individual gene knockdown………………………………………………………………………………...183 xii List of Abbreviations # Abbreviation Definition 2D 2-dimensional 3D 3-dimensional A ADT AJ AR AS AUC Androgen deprivation therapy Adherens junction Androgen receptor Active surveillance Area under the curve B BCR BPH Biochemical recurrence Benign prostatic hyperplasia C CA125 CLTC CSC CTC Cancer antigen-125 Clathrin, heavy chain (referred to as Clathrin) Cancer stem cell Circulating tumour cell D DRE Digital rectal exam E EBRT ECM EMT EMT-TF EOC External beam radiation therapy Extracellular matrix Epithelial-mesenchymal transition Epithelial-mesenchymal transition transcription factor Epithelial ovarian cancer F FDR FFPE FIGO FN1 False discovery rate Formalin-fixed paraffin embedded International Federation of Gynecology and Obstetrics Fibronectin G G3 G4 GEO GS Gleason pattern 3 Gleason pattern 4 Gene Expression Omnibus Gleason score H HGPIN HGSC HRE High grade PIN High grade serous cancer Hypoxia response element xiii I IHC Immunohistochemistry K KD KI Knockdown Knock-in L LOH Loss of heterozygosity M MET miRNA miR-200f MMP MOI MTE Mesenchymal-epithelial transition MicroRNA MicroRNA 200 family Matrix metalloproteinase Multiplicity of infection Multiplex target enrichment N NCCN National Comprehensive Cancer Network P PC PIN PSA PTEN Principal component Prostatic intraepithelial neoplasia Prostate specific antigen Phosphatase and tensin homolog R RNA ROC RP Ribonucleic acid Receiver operating characteristic Radical prostatectomy S SCRM shRNA SNAI1 SNAI2 SPARC Scrambled control Short hairpin RNA Snail family zinc finger 1 (Snail) Snail family zinc finger 2 (Slug) Osteonectin T TJ TNM TRUS Tight junction Tumour-node-metastasis Trans-rectal ultrasound V VIM Vimentin xiv Chapter 1: Introduction 1.1 The Epithelial-mesenchymal transition 1.1.1 EMT overview The epithelial-mesenchymal transition (EMT) is a reversible and evolutionarily conserved developmental process, whereby a polarized epithelial cell undergoes physiological and biochemical changes that allow it to adopt a mesenchymal phenotype (Timmerman 2004, Mani 2008, Xu 2014). It is characterized by loss of apical-basal cell polarity, increased production of extracellular matrix (ECM) components, as well as increased capacity for migration and invasion, and resistance to apoptosis (Yu 2014, Savagner 2010). Activation of EMT-driving transcription factors (EMT-TFs) results in disassembly of intercellular junctions, and changes in cytoskeletal organization, thereby altering cellular morphology and facilitating migration (Tsai 2012). Additionally, cells up-regulate production of enzymes involved in ECM degradation, thus facilitating invasive behaviour (Tsai 2012) (Summarized in Figure 1.1). It is important to note that EMT represents a combination of genomic and phenotypic alterations that occur within a cell in response to various stimuli, and as such is an operational definition, rather than being defined as a specific signaling pathway. 1 Figure 1.1 Summary of EMT characteristics Epithelial cells demonstrate apical-basal polarity, mediated by their engagement with a basement membrane and the presence of intercellular contacts such as adherens junctions, tight junctions and desmosomes. At the onset of EMT, intercellular contacts are dissociated, initiated primarily through the loss of E-cadherin. The dismantling of the intracellular contacts results in dissociation of the cells from the basement membrane, eventually leading to a loss of cell polarity. EMT progression is associated with upregulation of matrix metalloproteinases (MMPs) and cytoskeletal rearrangements which work to facilitate cell migration and invasion through the basement membrane. Cells that have undergone EMT also upregulate production of extracellular matrix components, such as Vimentin (VIM) and Fibronectin (FN1), both considered mesenchymal markers. The reverse of this process, mesenchymal-epithelial transition (MET) is also possible, such that cells lose their mesenchymal characteristics and reestablish their epithelial traits. 2 Epithelial-mesenchymal transitions occur throughout a number of biological processes and the context in which EMT occurs impacts the nature of its regulation. Tightly regulated and highly conserved processes, such as embryogenesis have more well-defined mechanisms of EMT initiation and progression, whereas in cancer, EMT is a byproduct of mass signaling dysregulation and therefore more difficult to characterize. As a result of these differences, in 2009 Kalluri and Weinberg suggested the definition of three distinct types of EMT, each with their own functional and genomic variances (Kalluri 2009). Type I EMT functions during implantation, embryogenesis, and organogenesis (Fitchett JE 1989, Hay 1990). It is involved in many stages of human development, most notably in the formation and separation of the germinal layers during gastrulation (Hay 1990). Type II EMT functions in wound-healing, tissue repair and fibrosis (Hudson 2009, Xi 2014). This EMT type appears to be predominantly initiated in response to inflammation and facilitates cell migration into the wound space, and subsequently re-generation of epithelial tissue (Savagner 2005, Malchionna 2012). Type III EMT, the primary focus of our research, occurs during cancer progression, particularly during the initiation of metastatic spread (Liang 2014, Jing 2014, Xu 2014), and is discussed in more detail in the following section. 1.1.2 Current model of EMT regulation during cancer progression From initiation to completion, EMT relies on the engagement of many distinct, and interconnected, molecular events, resulting in a shift towards a mesenchymal phenotype (Summarized in Figure 1.2). In the context of cancer biology, EMT can be initiated in response to inflammation in the tumour microenvironment (Agnihotri 2013), increasing hypoxia (Hwang-Verslues 2013) and changes in the density and composition of the ECM (Vargas 2013). 3 Tumour-associated macrophages in the microenvironment release inflammatory cytokines such as IL6, TNF and IL1B which activate TGFβ signaling and as a result, up-regulate EMT-TFs, SNAI1 (Snail) and ZEB1 (Leibovich-Rivkin 2013, Techasen 2012). Additionally, EMT-TFs such as Snail and TWIST1 can be activated in response to hypoxia, through the upregulation of hypoxia-inducible factor, HIF1A (HIF1α) (Yang 2008). HIF1α regulates the expression of Snail by binding to hypoxia response elements (HREs) in the Snail promoter (Zhang 2013). HIF1α-mediated Snail expression leads to loss of CDH1 (E-cadherin) and promotes cells invasion, through the upregulation of CTNNB1 (β-catenin) transcriptional activity (Zhang 2013, Zhao 2011, Liu 2010). Additional cues from the cell environment can regulate EMT in cancer, such as changes in cell-ECM interactions or mechanical stress. ECM density has been shown to contribute to EMT, such that an increasing collagen concentration results in weakened cell-cell adhesion, through negative regulation of E-cadherin (Kumar 2014). Spatial geometry and mechanical stress appear to also contribute to EMT. Epithelial sheet cultures demonstrate that stimulation with TGFβ results in increased levels of Vimentin and SMA in the cells at the sheet periphery, which are not seen in cells in the interior of the sheet. Additionally, mechanical stress applied across the cell sheets induced patterned EMT in high stress areas (Gomez 2010). 4 Figure 1.2 Summary of the current model of EMT regulation EMT is primarily initiated in response to stress cues from the tumour microenvironment, such as hypoxia, inflammation and cytotoxic stress. This results in the upregulation of signaling molecules such as TGFβ, Notch, Integrins, Wnt and a number of RTK ligands (HGF, HGF, EGF, PDGF), which initiate their respective signaling cascades. In response, EMT-TFs, primarily Snail-, ZEB- and Twist-family members, are upregulated and work to alter intercellular contacts, cell polarity and enzyme production. Feed-back and feed-forward signaling loops have been established in the regulation of EMT, such that EMT-TFs can regulate the expression of themselves and each other, while also facilitating activation of upstream signaling pathways. At the same time, downstream alterations such as loss of E-cadherin and upregulation of MMPs can contribute to differential regulation and function of the EMT-TFs. Ultimately, the combination of these molecular cues translates into the adoption of mesenchymal characteristics such as increased cell migration, invasion and resistance to apoptosis. 5 EMT progression is regulated by extra- and intra-cellular signaling molecules such as TGFβ (Geng 2014, Kim 2014), EGF (Cho 2014), IGF (Vazquez-Martin 2014), HGF (Farrell 2014, Zaritsky 2012), Wnt/β-catenin (Gnemmi 2014), Notch (Capaccione 2014), and estrogens (Sun 2014). EMT in cancer progression has also been linked to oncogenic mutations such as RasV12 (Safina 2009), ERBB2 (Gupta 2014, Nilsson 2014) or mutant TP53 (p53) (Rieber 2014). Despite the fact that regulatory mechanisms which facilitate EMT are broad-ranging, specific phenotypic hallmarks are consistently observed in relation to EMT progression towards the mesenchymal state. One of the earliest phenotypic changes associated with EMT is the dismantling of intercellular junctions, and subsequent loss of cell polarity. Of the intercellular contacts, tight junctions (TJs) are the first to be disassembled following PAR complex phosphorylation by TGFβR2. This results in active PAR6 binding to SMURF1 (an E3-ubiquitin ligase) and leads to ubiquitinylation, and subsequent degradation, of RHOA. This leads to disruptions in the cortical actin cytoskeleton which contributes to TJ loss (Ozdamar 2005). Dismantling of TJs is also associated with decreased expression of members of the Claudin and Zona occludens families, major components of these junctional complexes (Ikenouchi 2003). Loss of E-cadherin, resulting in dissociation of adherens junctions (AJs), is one of the best characterized molecular traits associated with EMT. E-cadherins are single-span transmembrane glycoproteins that interact in a homophilic, calcium-dependent manner with Ecadherins on neighbouring cells to regulate cell-cell contact and mediate intercellular signaling (Yilmaz 2009, Bhatt T 2013). These E-cadherin-based cell adhesion complexes are anchored to the actin cytoskeleton and interact with β-catenin to aid in cytoskeletal organization (Niessen 2011, Kovacs 2002). In this way, E-cadherin function is critical in the development and maintenance of a polar epithelium (Hajra 2002). Loss of E-cadherin, and the resultant loss of adherens junctions and cell polarity, have been demonstrated in gastrointestinal, pancreatic, 6 ovarian, colorectal and prostate cancers, and are correlated with poor overall prognosis and survival (Yun 2014, Lee 2013, Corso 2013, Hong 2011, Quattrocchi 2011, Whiteland H 2013). E-cadherin loss during EMT can be mediated both transcriptionally and posttranscriptionally. Transcriptional repression of E-cadherin results from interaction with transcriptional repressors including Snail, Slug, ZEB1 and 2 and TWIST (as reviewed in Peinado 2007). These factors can be activated in response to molecular cues, such as TGFβ, EGF, HGF, Wnt and Notch signaling, as outlined previously. Once activated, these factors participate in positive and negative feedback loops that regulate their own expression, and the expression of one another (Shin 2010). Snail, for example, is able to regulate its own expression by binding to its promoter through recognition of the E-box, resulting in down-regulation of Snail mRNA (Peiro 2006). Ultimately, association of these transcription factors with the promoter region of the E-cadherin gene results in epigenetic silencing by a variety of histone modifications, and eventually resulting in DNA hypermethylation. Upon Snail expression, the SNAG domain of the protein associates with histone deacetylases (HDAC) 1 and 2, forming a repressor complex with Sin3A, which then targets the E-box regions of the E-cadherin promoter, resulting in histone H3 and H4 deacetylation (Peinado 2004). Promoter deacetylation facilitates recruitment of the polycomb repressor complex 2 (PRC2) which then acts to repress E-cadherin expression (Herranz 2008). After initial repression of E-cadherin transcript levels, Snail induces the expression of ZEB1, which results in further inhibition of E-cadherin, through PRC2-independent mechanisms (Herranz 2008). Recruitment of methyltransferases, such as EHMT2 (G9a), also supports hypermethylation of the E-cadherin promoter, leading to repression (Hou 2008, Dong 2012) (Summarized in Figure 1.3). 7 Figure 1.3 Snail-mediated repression of E-cadherin (1) Activated Snail recruits and binds HDACs 1 and 2 through interactions with its SNAG domain. Together with Sin3A they form a repressive complex which targets the E-box regions of the E-cadherin promoter, resulting in histone deacetylation (2). Deacetylation of the promoter region allows for PRC2 and G9a recruitment (3) and results in E-cadherin repression through promoter hypermethylation (4). Additionally, Snail upregulates ZEB1, which modulates PRC2independent repression of E-cadherin (5), though these mechanisms are less well-defined. 8 In addition to transcriptional regulation, post-translational internalization of E-cadherin through Clathrin-mediated endocytosis, also occurs and leads to lysosomal degradation (Janda 2006, Akhtar 2001). However, these mechanisms are less well-characterized. Loss of E-cadherin, and subsequent dismantling of AJs, contributes to destabilization of desmosomes. This is likely due to increased activation of ZEB-family transcription factors, particularly ZEB2, which is known to repress transcription of Plakophilins and Desmoplakins, key elements of desmosome complexes (Vandewalle 2005). One major consequence of the dissolution of intercellular contacts is loss of cell polarity. The apical polarity Crumbs complex is dismantled via Snail-regulated repression of CRB3 (Whiteman 2008) and basolateral polarity is disrupted through loss of Scribble complexes, where LGL2 is believed to be a direct target of ZEB1 transcriptional repression (Aigner 2007). Ecadherin loss contributes to depolarization by preventing contact between SCRIB and the lateral membrane (Navarro 2005). Ultimately, loss of apical-basal polarity corresponds with cellular adoption of front-rear polarity, facilitated by actin cytoskeleton reorganization and mediated by localized induction of Rho GTPases. In particular, CDC42 and RAC1 are activated in the front of the cell, resulting in Arp2/3 complex-mediated actin assembly (Rotty 2013), and microtubule stabilization by DIAPH1, APC and EB1 complex formation (Wen 2004). In the rear of the cell, RHOA is activated and works to regulate contractile actomyosin filaments which aid in cell detachment and motility (Ridley 2003). In addition to enabling alterations in cell polarity, dissolution of intercellular contacts, specifically through the loss of E-cadherin, can have an extensive impact on the transcriptional and functional progression of EMT. For example, repression of E-cadherin alone can lead to increased cancer cell migration and invasion by activating upstream signaling networks and EMT-TFs, which work to further repress E-cadherin (Onder 2008). The establishment of this 9 feed-forward loop suggests that once EMT is initiated, E-cadherin loss may act to stabilize and sustain the program, and thus contribute to more comprehensive EMT regulation. Loss of intercellular contacts and changes in cell polarity ultimately result in phenotypic progression towards cell migration and invasion, characteristic traits of cancer cells with metastatic potential. In order for cells to move through their environment, they must first upregulate the production of a number of matrix metalloproteinases (MMPs) to facilitate degradation of the basement membrane and various ECM components (Son 2010). Both secreted and membrane-bound MMPs participate in EMT-mediated cell invasion. In particular, the secreted MMPs -9, -3 and -2, and the membrane-bound MT1-MMP (Chen 2013, Lin 2011, Bae 2013, Yang 2013) have significant roles, not only in ECM degradation, but in initiating EMT through disruption of intercellular contacts (Radisky 2005). Activation of EMT-TFs directly regulates the expression of MMPs leading to cell invasion. Activation of Snail results in proinvasive and pro-angiogenic cell behaviour through upregulation of MMP14 (MT1-MMP) and MMP15 (MT2-MMP) (Ota 2009). This behaviour is abrogated upon silencing of either MMP (Ota 2009). Similarly, Snail family member, Slug, regulates pancreatic cancer cell migration and invasion, through induction of MMP9 and reorganization of F-actin filaments (Zhang 2011). In addition to EMT-TF-regulated modulation of proteinase expression, cell migration and invasion is dependent upon the formation of invadopodia, filopodia-like structures which possess proteinase activity, allowing for cell extension through the ECM and past the basement membrane (Bowden 2006). MT1-MMP appears to be the primary proteinase associated with the invasive function of these cell projections. MT1-MMP trafficking, which is dependent upon Clathrin-mediated endocytosis and stabilized via cortactin and F-actin (Poincloux R 2009), allows for their concentrated localization, and high turn-over at membrane protrusions (Watanabe 2013). When frequency or duration of MT1-MMP expression is altered, it results in significant reduction of the invasive behaviour of the invadopodia structure, thus suggesting the importance of these 10 proteins in invasive progression (Watanabe A 2013). Invadopodia-mediated cell invasion has been linked to EMT. Eckert MA et al (2011) demonstrate that TWIST1, through the induction of PDGFRα and SRC signaling, increases the invasive potential of breast cancer cells by regulating the formation invadopodia. Furthermore, structural filaments associated with mesenchymal cells, such as Vimentin, have been shown to be necessary for invadopodia elongation during cell invasion, though they do not appear to initiate formation (Schoumacher 2010). Taken together, these studies provide clear evidence of the importance of EMT-regulated invadopodia formation in cell invasion through the ECM. Posttranscriptional regulation via microRNA (miRNA), contributes significantly to the progression of EMT. For example, the miR-200 family (miR-200f) targets multiple sequences in the 3’ UTR of ZEB2, leading to downstream regulation of E-cadherin expression (Christoffersen 2008). Further studies have demonstrated that ZEB1 is also a target of this miRNA family, and that inhibition of miR-200 results in reduced E-cadherin, and increased Vimentin expression (Park 2008). However, it seems that ZEB1 and miR-200f participate in a feedback loop as evidence indicates that ZEB1 recognizes and binds to E-box domains upstream of miR-200f promoter regions, repressing their expression (Burk 2008, Bracken 2008). Similarly, a feedback loop between Snail and the miR-34 family (miR-34f) appears to be involved in EMT regulation. MiR-34 targets a conserved sequence in the 3’ UTR of not only Snail, but Slug and ZEB1, while Snail and ZEB1 have been shown to repress the miR-34 promoter (Siemens 2011). TP53-mediated miR-34 expression is necessary for the prevention of cancer metastasis, and inhibition its expression through IL-6/STAT signaling results in upregulation of Snail, and is correlated with increased invasiveness of colorectal tumours in vivo (Rokavec 2014). 11 1.1.3 The mesenchymal to epithelial transition (MET) It is widely regarded that EMT is a reversible process. This reversal, known as the mesenchymal to epithelial transition (MET), results in the reestablishment of epithelial characteristics in the cells, leading to a reduction in migratory and invasive potential. It is less well-characterized than EMT, but is believed to play a significant role in the establishment of cancer metastases by allowing cells to adapt more readily to changes in environmental cues (Gunasinghe 2012, Wells 2008). For example, expression of the homeobox transcription factor, PRRX1, is sufficient to induce EMT in breast cancer cells, but must be silenced to promote metastatic lung colonization in vivo (Ocana O 2012). This indicates that metastatic colonies are unable to establish themselves while circulating tumour cells (CTCs) sustain EMT-TF expression, and that reversal of EMT is necessary for completion of cancer metastasis. Similarly, induced expression of TWIST1, results in increased Vimentin expression and supports invasion in an in vivo model of squamous cell carcinoma, and must be repressed in order to establish metastatic colonies (Tsai 2012). In this case, loss of TWIST1 correlates with reduced Vimentin and increased E-cadherin expression in the metastatic tumours, again indicating that reversal of EMT is necessary for metastatic colonization (Tsai 2012). Factors leading to the initiation of MET remain largely uncharacterized. However, a study done by Chao, Y et al in 2010 indicates that changes in cues from the tumour microenvironment are likely to play a significant role. In the mesenchymal-like breast cancer cell line, MDA-MB-231, in which E-cadherin expression is repressed through promoter methylation, exposure to cues from a secondary organ microenvironment, via co-culture with hepatocytes, was sufficient to reestablish epithelial characteristics in the cells (Chao 2010). Gain of E-cadherin expression was the result of loss of methylation at the E-cadherin promoter, and was reversible when cells were removed from co-culture with hepatocytes (Chao 2010). This suggests that environment-specific factors contribute to MET initiation, and may explain preferential metastatic sites in different cancer types. 12 Taken together, this evidence suggests that while EMT is required for invasion and intravasation, resulting in cell dissemination from the primary tumour mass, metastasis cannot be completed without initiation of MET, likely through altered cues from the metastatic microenvironment. 1.1.4 EMT and cancer stem cells Cancer stem cells (CSCs) are self-renewing, tumor-initiating cells that have the capacity for pluripotency (Chen 2013). Fundamental associations have been made between EMT activation and the acquirement of these stem cell traits, both in normal and neoplastic cells (Mani 2008). For example, overexpression of Snail or TWIST in cancerous mammary epithelial cells, results in cells undergoing EMT and simultaneously acquiring an antigenic CD44high/CD24low expression pattern, which is widely considered to be a marker for mammary stem cells (Mani 2008, Liao 2007). Furthermore, cells exposed to sustained EMT-promoting signals are capable of initiating tumour formation in athymic nude mouse models, while control cells, not expressing Snail or Twist, are unable to initiate tumour formation (Mani 2008). While it is the case that EMT-TFs can induce CSC characteristics, it is also true that signaling networks associated with stem cell behavior can regulate the expression of EMTassociated proteins. One of the better examples of this is the Wnt/β-catenin signaling network, which is essential for the maintenance of both somatic and cancer stem cells (Zhang 2014, Zhao 2014, Vermeulen 2010). Wnt signaling has also regulates EMT-TF expression, particularly by inhibiting GSK3β, an active Snail repressor, thereby leading to increased Snail expression (Han 2013, Yook 2006). Snail then reinforces Wnt signaling by repressing E-cadherin which normally acts to sequester β-catenin and prevent its translocation to the nucleus, where it functions as a transcriptional co-factor (Heuberger 2010). Therefore, while evidence exits to functionally link EMT and CSC behaviour, it is currently insufficient to suggest that cells that have undergone EMT are, by definition, cancer 13 stem cells. Instead, it appears that the signaling associated with EMT may work in conjunction with those networks regulating CSCs to activate or enhance some characteristic stem cell traits. Given the similarities in the signaling responsible for CSC maintenance and initiation and progression of EMT, it is not unreasonable to conclude that both programs may work synergistically to facilitate aggressive cancer phenotypes such as metastatic colonization and resistance to chemotherapy. Taken together, these findings suggest that initiation of EMT in cells with CSC potential results in the enhancement of stem cell traits such as tumour-initiation, pluripotency and self-renewal, but that EMT itself is not sufficient to generate CSCs. 1.1.5 EMT as a biomarker of cancer metastasis and chemotherapy resistance EMT was first recognized for its role in breast cancer metastasis in the mid-nineties (Pulyaeva 1997) and has continued to garner attention in this field over the past twenty years. Today, EMT is considered by many to have a key role in facilitating tumour cell dissemination, intravasation and extravasation, thereby regulating the primary stages in cancer metastasis (Samatov 2013). As such, the utility of EMT-associated markers as prognostic indicators in cancer has been widely examined. In breast cancer, the highly metastatic basal-like subtype is strongly associated with EMT gene signatures consisting of high Vimentin, N-cadherin and Osteonectin (SPARC) expression and loss of E-cadherin (Sarrio 2008). Similarly, loss of epithelial keratins and subsequent gain of Vimentin are significantly associated with poor overall prognosis in breast cancer (Fuchs 2002). In colorectal cancer, over expression of EMT-TFs, Snail and TWIST1 are significantly associated with lymph node metastasis and poor prognosis (Kim 2014). E-cadherin loss is also significantly associated with reduced disease-free survival and increased rates of metastasis in colorectal cancers (Yun 2014). Additionally, metastatic spread in prostate cancer appears to be associated with EMT, though there are few studies correlating EMT with clinical outcomes in patient samples. One study found that overexpression of SOX4, an EMT-regulating transcription factor, 14 correlates with increasing Gleason score, presence of distant metastases and poor prognosis in prostate cancer (Wang 2013). Furthermore, in vivo models of prostate cancer have demonstrated that cells with high Snail expression and low expression of E-cadherin have a significantly increased rate of metastasis (Deep 2014). Taken together, these data indicate that EMT is noticeably associated with metastatic spread and poor disease prognosis in multiple cancer types. EMT is also correlated with poor prognosis due to its contribution to chemotherapy resistance in cancer. Compared to EMT in metastasis, the mechanisms regulating drug resistance are less well defined, but may be associated with EMT-regulated repression of apoptosis (Du 2011, Zhao 2012) or upregulation of DNA damage repair mechanisms, in the case of platinumbased agents (Hsu 2010). In esophageal cancer, mesenchymal markers, Snail, ZEB1 and Ncadherin, are overexpressed in post-chemotherapy, residual tumours compared to chemonaïve tumours (Hara 2014). This suggests that either EMT is induced in cancer cells during exposure to chemotherapy, or that a subpopulation of cancer cells exists, prior to treatment, that is intrinsically chemoresistant as a result of upregulated EMT signaling. In non-small cell lung cancer, a mesenchymal phenotype defined by increased expression of Vimentin, Fibronectin and N-cadherin, and loss of E-cadherin, is indicative of poor response to EGFR inhibitors (Gefitinib or Erlotinib), while tumours with epithelial characteristics demonstrate favourable response to treatment with EGFR inhibitors (Ren 2013). In many cases, EMT regulation of chemotherapy response is identified using in vitro models. For example, in A2780 ovarian cancer cells, Paclitaxel resistance is found to be the result of EMT induction, caused by a significant upregulation in PI3K signaling (Du 2013). Additionally, Snail and Slug-mediated EMT is initiated in response to Notch3 signaling and contributes to carboplatin resistance in the OVCA429 ovarian cancer cell line (Gupta 2013). In clinical samples, Cisplatin resistance in ovarian cancer correlates significantly with overexpression of EMT-associated genes and microRNAs such as BAMBI, miR-200c and miR-141 (Marchini 2013), again suggesting that 15 chemotherapy resistance in ovarian cancer is at least partially regulated by EMT activation. Ultimately, sufficient evidence exists to suggest that the presence of EMT-associated signaling in different types of cancer may have utility in predicting response to certain chemotherapeutic agents. Given that the relationship between EMT and poor disease prognosis is not limited to a single type of cancer, gene or protein expression associated with this program may have utility in prognostication in a general clinical setting, and is worth examining in the context of biomarker development. 1.1.6 Targeting EMT for cancer therapy Given the apparent link between EMT and cancer metastasis, there is substantial interest in the prospect of targeting the regulatory networks of EMT for cancer therapy. Therapeutic strategies proposed for this purpose can be roughly divided into four targetable stages, as review in Davis et al. (2014). Blockade of extracellular stimuli, responsible for the induction of early EMT, is the first stage to consider. For example, prevention of ligand binding at the cell membrane by targeting specific cell surface receptors would prevent initiation of intracellular signaling leading to EMT. Therapeutic intervention leading to the inhibition of EGF signaling (Lo 2007) and TGFβ signaling (Ehata 2007, Nam 2008, Reka 2011), results in prevention of cancer cell invasion and progression to metastasis, both in vitro and in vivo. However, due to the multiplicity in the upstream signaling that converges on the EMT phenotype, there is a substantial risk of developing resistance to these types of therapeutics (Thompson 2008). Given this limitation, it may be more feasible to target cell surface receptors in combination with internal signal transduction molecules and EMT-TFs. One common example of this is inhibition of STAT3, a molecule which contributes to EMT in a number of different cancers (Liu 2014, Cho 2014, Hogstrand 2013). STAT3 inhibitors such as Stattic (Schust 2006), AG490 (Liu 2014) and NSC74859 (Zhang 2014) have been developed and preliminary evidence 16 suggests they may represent viable therapeutics as they appear to reduce cancer cell invasion (Liu 2014) and modulate radiosensitivity (Zhang 2014). Given these findings, it may also be desirable to target EMT-specific transcription factors such as Snail or ZEB1 for cancer treatment. While few examples of this currently exist, a Co(III)-Ebox conjugate has been developed that can prevent DNA interactions with Snail, Slug and ZEB1 and results in inhibition of Snail-mediated repression of E-cadherin (Harney 2012). Given the association between E-cadherin loss and increased cancer metastasis (Onder 2008, Chen 2012), there is potential that targeting this phenotypic process could be valuable in cancer treatment. Despite the potential utility of inhibiting EMT induction in cancer treatment, this approach would be of little value in postmetastatic disease where tumour cells have already initiated or undergone EMT. In these instances, benefit may be found in targeting cells that are already in their mesenchymal state. There are a number of well-established mesenchymal cell markers such as Vimentin and N-cadherin, which are associated with cancer cell invasion and are expressed in CTCs (Kallergi 2011, Armstrong 2011). Therefore, targeting these mesenchymal proteins may aid in eliminating cancer cells in patients with more advanced disease. Some success has been found in studies looking at the effect of Withaferin-A (WFA), which works by depolymerizing Vimentin filaments (Thaiparambil 2011, Yang 2013). These studies have found that WFA treatment of breast cancer cells results in inhibition of migration and invasion, and is associated with increased apoptosis (Thaiparambil 2011, Yang 2013). N-cadherin is also a viable therapeutic target, such that its inhibition with monoclonal antibodies leads to prevention of metastasis in in vivo models of prostate cancer (Tanaka 2010). Taken together, these results indicate that therapeutic targeting of mesenchymal makers may provide novel advances in treatment of more aggressive cancers. Finally, targeting the reversal of EMT may provide additional avenues for cancer treatment. As outlined previously, the process of MET is still poorly characterized but has been shown to be necessary for metastatic colonization in vivo (Gunasinghe 2012, Tsai 2012). As such, 17 therapeutics that function by preventing the reversal of EMT may have a significant impact in inhibiting cancer metastasis. Given that little is currently known about how this process is regulated, few MET-associated targets for therapeutic intervention have been identified. However one study by Chaffer et al. (2006), demonstrated that circulating cancer cells with epithelial characteristics are more efficient at forming metastatic colonies, and that this was associated with increased expression of FGFR2IIIc. They were also able to show that silencing of FGFR2IIIc and prevention of MET resulted in increased survival in SCID mice (Chaffer 2006). Ultimately, as the biological impacts of EMT and MET continue to be uncovered, new treatment strategies targeting these processes may prove extremely useful in cancer management, both as agents in first line therapeutics, and in advanced or chemoresistant disease. 1.2 Ovarian Cancer 1.2.1 Overview Of the gynecological malignancies, ovarian cancer has the highest associated mortality rate in the western world. While relatively rare, it is estimated that 1 in 68 Canadian women will be diagnosed with some form of ovarian cancer in their lifetime, resulting in 2600 new cases of ovarian cancer in 2013, and 1700 deaths associated with the disease (Canadian Cancer Society 2013). Ovarian cancers can be broadly classified as epithelial or non-epithelial in origin. Nonepithelial cancers account for only 10% of all ovarian cancer cases, and include sex chord stromal and germ cells tumours (Colombo 2012). The more commonly diagnosed epithelial ovarian cancers (EOC) can again be subdivided into five subtypes that encompass nearly 95% of all epithelial-origin tumours. Of these subtypes, high-grade serous cancer (HGSC) is the most common subtype, accounting for 70% of epithelial ovarian cancers, and has been the primary focus of our research. This subtype is frequently associated with mutations in p53 and BRCA1/2, 18 and upregulation of WT1 (Madore 2009, Kuo 2009). Overall, it has a poor 5-year survival rate of approximately 30-40% (Seidman 2012, Verhaak 2013). The high mortality rate associated with this disease is largely due to late stage diagnosis and poor treatment response. 1.2.2 Diagnosis and Prognosis Roughly 75% of patients with HGSC are diagnosed with late stage disease (III or IV), such that the cancer has spread throughout the peritoneal cavity, or to distant metastatic sites, resulting in poor overall 5-year survival (Levanon 2008, Vang 2009). Delayed diagnosis is primarily due to the fact that the disease remains largely asymptomatic in early stages, though significant correlations have been made between early stage disease and some non-specific symptoms such as bloating, constipation and abdominal pain (Goff 2004, Olson 2001). Diagnostic tools which aid in ovarian cancer detection include CA125 measurements, imaging procedures such as transvaginal or pelvic ultrasounds, and tissue biopsies taken during exploratory laparotomy (Cannistra 2004, Nezhat 2013). When a diagnosis of ovarian cancer is made, surgical staging/primary cytoreduction is completed, which typically involves a total hysterectomy and bilateral salpingo-oophorectomy to ensure optimal cancer debulking (Chene 2013). Surgical staging of ovarian cancer is done in accordance with the International Federation of Gynecology and Obstetrics (FIGO) staging system (summarized in table 1.1) (Benedet 2000). Across all epithelial ovarian cancers, disease prognosis is correlated to FIGO stage, such that stage I tumours have an estimated 5-year survival rate of 75-90%, which decreases dramatically to less than 20% survival after 5 years in patients with stage IV disease. 19 Table 1.1 Summary of FIGO staging system for ovarian malignancies FIGO Stage Characteristics I Disease confined to the ovaries IA unilateral, capsule intact, no ascites IB bilateral, capsule intact, no ascites IC Stage IA or IB plus ascites, ruptured capsule, tumor on ovarian surface II Disease spread beyond ovaries, confined to the pelvis III Disease confined to the abdominal cavity (surface of the liver; abdominal lymph nodes; omentum or bowel) IIIA Negative lymph nodes, plus microscopic seeding of peritoneal surface IIIB Negative lymph nodes, peritoneal implants <2 cm IIIC Positive lymph nodes and/or abdominal implants >2 cm IV Spread to liver parenchyma, lung, pleura, or other extra-abdominal sites 20 1.2.3 Treatment and Chemotherapy Resistance In women who are diagnosed with early stage HGSC, where disease remains confined to one or both ovaries, 5-year survival is nearly 90%, largely due to the optimal conditions for surgical intervention (Raja 2012). However, in patients with advanced stage disease (stage III or IV), cytoreductive surgery will be accompanied by chemotherapy as a first line treatment option (Vergote 2010). Typically, adjuvant chemotherapy will involve the use of taxanes and platinumbased drugs, as this combination consistently produces the best clinical response (Ozols 2003, Bookman 2009). Taxanes, such as Paclitaxel and Docetaxel are microtubule stabilizing drugs which prevent cell division (Xiao 2006). They work by binding the β-tubulin subunits that polymerize to form microtubules, thereby preventing depolymerization (Xiao 2006). Paclitaxel blocks cells in the G2/M phase of the cell cycle, thereby preventing normal cell division and leading to apoptosis (Xiao 2006). Taxanes work best in combination with platinum agents, such as Carboplatin and Cisplatin, which induce cytotoxicity by crosslinking with purine bases in DNA, resulting in the formation of adducts that prevent replication and interfere with DNA repair mechanisms by overwhelming nucleotide excision repair (NER) machinery (Dasari 2014, Siddik 2003). Ultimately, accumulation of DNA damage eventually leads to cellular apoptosis (Dasari S 2014). In addition to late stage diagnosis, chemotherapy resistance is partially responsible for the poor overall survival rate in ovarian cancer patients. While 80% of women will initially respond to first line chemotherapy, 20% of patients present with intrinsically resistant disease, defined as having a progression-free interval of less than six months (Chien 2013). Of those who initially respond to first line chemotherapy, approximately 70-80% will have disease relapse (Ushijima 2010) and upwards of 80% of those women will present with chemoresistant disease that will no longer respond to first line therapy, specifically in the case of platinum-based drugs (MantiaSmaldone 2011). Given the high incidence of platinum resistance in ovarian cancer, 21 understanding the molecular mechanisms contributing to this occurrence could have a significant impact on ovarian cancer management strategies. 1.2.4 Molecular mechanisms contributing to platinum resistance in ovarian cancer A number of mechanisms have been established in the development of chemotherapy resistance in ovarian cancer. These mechanisms can be broadly divided into five functional categories, as summarized in Figure 1.4. Given that platinum-based chemotherapeutics work by inducing apoptosis by forming DNA damaging adducts, enhanced DNA repair mechanisms are believed to contribute to platinum resistance. Clinical studies looking at the association between excision repair crosscomplementation group 1 (ERCC1) expression and chemotherapy resistance in ovarian cancer determined that increased expression of ERCC1 is specially found in those tumours that respond poorly to platinum-based therapy (Li 2013). Other factors such as BRCA1/2 status also correlate with platinum response, such that active homologous repair mechanisms reduce platinummediated cell death, thereby conferring resistance (Vencken 2011). Modulation of apoptotic responses also contributes to platinum resistance. For example, upregulation of anti-apoptotic factor, BCL2, and repression of pro-apoptotic, BAX, are found to be significantly correlated with poor platinum response in ovarian cancer (Kassim 1999, Han 2013). Restoration of BAX expression through inhibition of STAT3 results in the restoration of platinum sensitivity (Han 2013). Additionally, upregulation of BCL2L1 upon Cisplatin treatment results in the development of platinum resistance in SKOV3 ovarian cancer cells, which is reversed upon BCL2L1 inhibition (Villedieu 2007). While less extensively evaluated in ovarian cancer, platinum transport and metabolism may contribute to therapy resistance. In the Cisplatin-resistant ovarian cancer cell line, A2780cis, elevated intracellular ATP levels results in increased Cisplatin efflux and therefore reduced platinum accumulation, contributing to drug resistance (Schneider 2013). Outside of ovarian 22 cancer, Cisplatin resistance in breast cancer cells is partially attributed to aberrant intracellular pH regulation, resulting in the prevention of platinum compound activation (Laurencot 1995). Finally, cells that demonstrate reduced proliferative capacity have a higher tolerance for platinum therapy, as slowly dividing cells are less likely to induce apoptosis in response to DNAplatinum adducts. While evidence supporting this theory is limited, it appears that cytotoxic stress can induce cell dormancy, thereby reducing proliferation and enabling disease recurrence after the completion of platinum treatment (Chien 2013). Clearly the development of platinum resistance in ovarian cancer represents a complex dysregulation of normal cellular processes. Therefore, in order to better understand and treat platinum resistant tumours, extensive research still needs to take place. Further characterization of these processes in the context of drug resistance may ultimately lead to the identification of clinical relevant predictive biomarkers and potentially the development of chemotherapeutics which may aid in re-sensitization of ovarian cancer cells to platinum-based therapies. 23 Figure 1.4 Cellular processes implicated in platinum resistance Resistance to platinum-based chemotherapy has been attributed to alterations in five generalized cellular processes. Enhanced DNA repair enables cells to overcome DNA adduct-induced apoptosis by increasing the efficiency of adduct excision. Additionally, inhibition of proapoptotic genes and upregulation of anti-apoptotic genes has been associated with platinum resistance in ovarian cancer. Inhibition of intracellular Cisplatin activation also contributes to resistance. Similarly, cells that have inactivated drug influx transporters, or upregulated efflux transporters, may prevent drug accumulation within the cancer cells. Finally, cells that demonstrate reduced proliferative capacity may have a higher tolerance for platinum therapy, as non-dividing cells are less likely to induce apoptosis in response to DNA-platinum adducts. 24 1.3 Prostate Cancer 1.3.1 Introduction to prostate cancer Prostate cancer is the most commonly diagnosed non-skin cancer amongst Canadian men, with one in seven being affected in their lifetime (Canadian Cancer Society 2013). In 2013 alone it was estimated that 23600 new cases of prostate cancer were diagnosed in and 3900 deaths were associated with the disease (Canadian Cancer Society 2013). According to the National Cancer Institute’s SEER statistics, overall prostate cancer 5-year survival is 98.8% (Howlader 2013). Approximately 93% of patients are diagnosed with localized or regionally confined disease, in which case 5-year survival is 100% (Howlader 2013). However, for the 4% of patients diagnosed with distantly metastatic disease, 5-year survival is substantially lower, at 28% (Howlader 2013), emphasizing the detrimental impact of metastatic progression in this disease. Those at a higher risk of prostate cancer development are men over the age of 50 (risk increases with age), people of African descent, and men with a family history of prostate cancer or germline mutations in BRCA1/2 (Howlader 2013, Johns 2003, Zeegers 2003, Castro 2013). Additional risk factors include high-fat diet, obesity and environmental factors, such as exposure to potential carcinogens (Farrell 2013, Gann 2002) 1.3.2 Molecular events in prostate cancer initiation and progression Prostate cancer appears to originate from regions of hyperproliferative, pre-invasive cells found within the lining of prostatic ducts and acini, referred to as prostatic intraepithelial neoplasia (PIN) (Merrimen 2013). High-grade PIN (HGPIN) is considered the most likely precursor of prostatic carcinoma due to the observation of shared genetic mutations, gene fusions and chromosomal aberrations (Alcazar 2001, Sakr 2001, Qian 1995). For example, loss of heterozygosity (LOH) of chromosomal region, 8p12-21, was found in 63% of HGPIN foci and 90.6% of matched prostate tumours, but not in normal prostate epithelia (Emmert-Buck 1995), indicating that loss of this region is associated with malignant transformation. Progression from 25 HGPIN to prostate cancer remains poorly defined but has been associated with the acquisition of additional genomic alterations, gene mutations and gene fusion events. Hypermethylation and resultant silencing of GSTP1 is considered an early event in prostate cancer progression. It is well documented that GSTP1 is expressed in normal prostate epithelium, however loss of expression of this gene has been detected in approximately 70% of HGPIN and in >90% of prostate cancers (Nakayama 2004, Henrique 2004). Due to its genomic ‘caretaker’ function, early loss of GSTP1 expression is likely to contribute significantly to genomic instability (Lin 2001), leading to cancer development and progression. Another early event in prostate cancer development is the fusion between the TMPRSS2 and ERG genes (TMPRSS2-ERG), which is found in up to 70% of prostate cancers and results in overexpression of the ERG oncogene (Spencer 2013). In a study following 461 patients with HGPIN over 3 years, it was found that 53% of men who had TMPRSS2-ERG fusions present in their PIN foci developed prostate cancer over the study period, while only 35% of men without the gene fusion developed the disease, therefore suggesting an important role for this fusion gene in prostate cancer initiation (Park 2014). Loss of tumour suppressor, phosphatase and tensin homolog (PTEN), is found in 40-70% of prostate cancers (Yoshimoto 2006, Yoshimoto 2007). Given that LOH of the PTEN locus (10q23) is found in 23% of HGPIN lesions and 68% of prostate cancers, this event may also have a significant role in the transition from pre-neoplastic PIN to prostatic carcinoma (Yoshimoto 2006). In the context of progression towards aggressive prostate cancer development, those tumours with both TMPRSS2-ERG fusions and PTEN loss are found to be over 5-fold more likely to have capsular invasion and shortened progression-free survival (Nagle 2013, Leinonen 2013). Disease progression towards androgen independence, or castration resistance, is one of the most clinically detrimental factors in prostate cancer prognosis, and has shown to be 26 associated with inactivating mutations in tumour suppressor genes such as TP53 and RB1, as well as alterations to the androgen receptor (AR). Up to 40% of prostate tumours have mutations in TP53 (Beltran 2013), resulting in loss of AR regulation (Schlomm 2008). Similarly, RB1 mutations, which are rarely found in locally confined prostate cancers, are present in up to 45% of those tumours that develop resistance to androgen deprivation therapy (ADT) (Barbieri 2013). As with p53, this association is due to the resultant inability of RB1 to regulate AR activity. Unchecked activation of the androgen receptor results in hypersensitivity to low androgen levels, thereby facilitating resistance to ADT (Sharma 2010). Modifications to the androgen receptor itself contribute to androgen independence, and therefore poor disease prognosis (Edwards 2003, Gregory 2001, Linja 2001). Furthermore, aggressive prostate tumours can establish androgen independence by bypassing AR-signaling completely, through the modulation of apoptotic responses, such as aberrant activation of apoptotic factors, Bcl2 and BAD, and enhanced Akt signaling (Debes 2004, Martel 2003). Ultimately, biological progression from pre-malignant HGPIN to aggressive androgen independent prostate cancer is a complex, and still poorly understood, process. The great degree of intertumour genomic heterogeneity contributes significantly to the clinical shortcomings in prostate cancer management, which will be discussed in later sections. 1.3.3 Natural history of prostate cancer: What are the true risks? Development and progression of prostate cancer is characteristically slow and as a result, short-term follow-up studies of patient clinical course are unlikely to provide an accurate assessment of disease progression. For patients who present with early stage, locally confined disease, short term survival is 100% (Howlader 2013). However, long term prognosis is more difficult to assess even in low-risk patients. Long term monitoring of early stage, initially untreated prostate cancers has provided some insight into the natural progression of the disease. A 32-year study out of Sweden, monitoring outcome in early stage prostate cancer patients, 27 concluded that only 17% (38 out of 223) of study participants died from prostate cancer-related illness, despite an overall mortality rate of 99% (Popiolek 2012). From the same cohort, 41.4% of men demonstrated local disease progression, while only 18.4% developed distant metastases over the 32-year study period (Popiolek 2012). Looking specifically at those patients classified as lowrisk (well-differentiated tumours with T0/T1 stage), prostate cancer-specific survival remained high (81.1%) up to 15 years post-diagnosis. However, disease-specific survival dropped significantly to 31.4% at 25 years, post-diagnosis. This suggests that low risk prostate tumours, while having minimal potential for metastatic development in the short term, can result in mortality, if given adequate time for progression (Popiolek 2012). Of course, patient age at the time of diagnosis would need to be taken into account in order to assess the significance of these findings in clinical practice. Ultimately, disease progression to metastasis is one of the worst prognostic indicators. Findings from autopsy studies demonstrate that approximately 35% of men with prostate cancer over the age of 40 have evidence of hematogenous metastasis (Bubendorf 2000). Clinical studies following disease progression in patients who have undergone RP for locally confined disease, and have subsequently developed biochemical relapse (BCR), have found that the average time to metastatic progression from initial PSA rise is 8 years (Freedland 2005, Pound 1999). Subsequently, the average time from development of metastatic disease to prostate-cancer specific death is 5 years, with 5-year survival at approximately 43% (Pound 1999). Therefore, disease progression to metastasis significantly reduces prostate cancer survival and should be strictly accounted for when determining cancer prognosis and selecting treatment strategies. Assignment of risk classifications to prostate cancer patients is discussed in the following section. 28 1.3.4 Clinical aspects of prostate cancer diagnosis, prognosis and treatment options In the vast majority of cases, prostate cancer development is detected early through standard screening practices, including measurements of serum prostate-specific antigen (PSA) levels and digital rectal examination (DRE) (Ismail 2013). Current clinical practice encourages PSA testing, in conjunction with DRE, in men over the age of 50, or 40 if the patient falls into a high-risk group, in order to monitor for changes that may be indicative of cancer development (Borza 2013). PSA is a glycoprotein produced by the epithelial component of the prostate gland and detected in patient serum (Barry 2001). It is found to increase in concentration with advancing disease stage (Barry 2001). However, PSA is not a highly specific biomarker and therefore change in PSA can also be indicative of benign prostatic hyperplasia (BPH) and subclinical prostatic inflammation (Barry 2001). In fact, it is estimated that only 1 in 4 men with abnormal PSA results will be diagnosed with prostate cancer (Mistry 2003). Since the introduction of PSA screening, diagnosis of prostate cancer has dramatically increased (Etzioni 2002). However, given the rise in prostate cancer detection, there has only been a marginal reduction in associated mortality rate during the same time period (Djulbegovic 2010). In fact, PSA screening appears to have little impact on overall outcome of prostate cancer, and may actually contribute to over-treatment of the disease (Etzioni 2002, Djulbegovic 2010), such that patients with abnormal PSA measurements, who are lacking histopathological characteristics indicating a likelihood of aggressive disease progression, are choosing to undergo radical treatment options resulting in co-morbidities such as impotence and incontinence (Loeb 2014). While these studies suggest that PSA testing may be more detrimental than beneficial for prostate cancer screening in the general population, its utility remains intact as a biomarker of disease recurrence following radical prostatectomy. Rising post-operative PSA level, referred to as biochemical recurrence (BCR), occurs in up to 35% of patients who have undergone radical prostatectomy (Boorjian 2011). While it is not definitive of clinical relapse, it is considered an early indicator of cancer progression and often precedes the development of metastatic disease 29 and prostate cancer-specific mortality, and typically indicates the need for secondary therapeutic procedures (Boorjian 2011). Ultimately, abnormal PSA and/or DRE may lead to a prostate biopsy, which is used to detect the presence of abnormal or cancerous cells in the prostate, and assist with cancer diagnosis and grading (Nieto-Morales 2013). Using a trans-rectal ultrasound (TRUS)-guided biopsy, twelve needle cores are taken from the prostate and subject to histopathological examination (Gomella 2011). The assignment of a Gleason score is based on glandular differentiation of the prostate tissue and is represented as numerical values associated with the histological pattern (Pascal 2009). Patterns range from 1 (well differentiated) to 5 (undifferentiated) and are combined to provide a Gleason score for the tumour out of 10. As of 2005, Gleason scores from prostate biopsies are calculated by adding the pattern representing the highest tumour volume in the biopsy, to that of the highest pattern present in the biopsy, even if it is tertiary in volume to a lower pattern (Epstein 2005). The grading system from a prostate biopsy differs from that of a radical prostatectomy. When grading a RP, Gleason score is calculated by adding the values from the two most dominant patterns. Higher pattern, tertiary volume disease (<5% of the tumour mass) is recorded, but not included in the score (Epstein 2010). In conjunction with the Gleason score obtained from biopsy, tumour stage is determined to aid in risk assessment for clinical cancer management. Current practice utilizes the tumournode-metastasis (TNM) staging system, which is outlined in Table 1.2 (NCCN 2013). Additional histopathological features are taken into consideration during prostate cancer prognostication, including lymphovascular and perineural invasion, extracapsular extension and seminal vesicle involvement (Lee 2010). Each of these features provides an estimate of the degree of cancer spreading and can be assessed at the time of biopsy to aid in staging, or after radical prostatectomy to support prognostication. The presence of these features is considered 30 indicative of pathologically aggressive prostate cancer and correlates with disease recurrence (Algarra 2014, Sohayda 2000). Ultimately, a combination of Gleason score, tumour stage, pre-operative PSA, as well as information on patient age, family history and ethnicity, are used to risk-stratify patients for optimal treatment options. As of 2014, the National Comprehensive Cancer Network (NCCN) has characterized six risk groups used to assist in prognostication and to best assign treatment strategies (NCCN 2013). Risk groups, and the parameters defining them, are outlined in Table 1.3. 31 Table 1.2 TNM staging criteria for prostate cancer Stage Characteristics Tumour T1a Cancer found in <5% of resected tissue T1b Cancer found in >5% of resected tissue T1c Cancer identified by needle biopsy completed due to abnormal PSA T2a Cancer confined to half of one lobe, or less T2b Cancer present beyond half of one lobe, but not present in the other lobe T2c Cancer present in both lobes T3a Extracapsular extension, no seminal vesicle involvement T3b Invasion of seminal vesicle(s) T4 Invasion of local structures other than SV Node N0 No regional lymph node metastasis N1 Metastasis in regional lymph node(s) Metastasis M0 No distant metastasis M1a Non-regional lymph node metastasis M1b Bone metastasis M1c Metastasis to other site(s) with or without bone involvement 32 Table 1.2 Risk stratification in prostate cancer Risk Group Very Low Characteristics T1c Gleason score ≤6 PSA<10 ng/mL < 3 positive biopsy cores (≤ 50% cancer/core) PSA density <0.15ng/mL/g Low T1-T2a Gleason score ≤6 PSA <10 ng/mL Intermediate T2b-T2c or Gleason score 7 or PSA 10-20 ng/mL High T3a or Gleason score 8-10 or PSA >20 ng/mL Very High T3b-T4 Metastatic Any T, N1 or Any T, Any N, M1 33 Ultimately, risk stratification enables the identification of optimal treatment strategies for prostate cancer management. Ideally, men classified as having a low risk of cancer progression would engage in active surveillance (AS), such that the disease is monitored via continued PSA screening, DRE and follow-up biopsies, thereby avoiding the co-morbidities associated with radical prostatectomy (Klotz 2013). Active surveillance is believed to be sufficient in the management of Gleason score 6 (3+3) prostate cancer as there is ample evidence to suggest that there is a very low, to zero, chance of metastatic progression in these tumours (Ross 2012, Popiolek 2012). For those patients who opt for more aggressive treatment due to concerns about the AS method, or for those identified as having intermediate or high risk, locally confined disease, radical prostatectomy (RP) is typically the first line treatment option (NCCN 2013). The procedure results in the removal of the entire prostate gland, as well as the seminal vesicles and part of the urethra and bladder neck (Rassweiler 2003). If warranted, the regional lymph nodes will also be removed. Depending on the extent of disease, it is possible for some patients to have bilateral nerve-sparing prostatectomy whereby the neurovascular bundles responsible for erectile function are left intact, in an attempt to prevent impotency (Rassweiler 2003). Prostatectomy is typically indicated for men with at least 10 years of life expectancy, those under the age of 75 and those that are healthy enough to undergo anesthesia (Lavallée 2014, NCCN 2013). Radiation, in the form of external-beam radiation therapy (EBRT) or brachytherapy (internal radiation therapy), and cryoablation, are alternative treatment options to RP. Radiation can be used either as a first line treatment of confined or locally regionalized prostate cancer, or as a follow-up treatment after RP, while cryoablation is typically indicated for those who have failed to respond to radiation therapy (NCCN 2013). ADT is the primary treatment for advanced or recurrent prostate cancer, or is used as an adjuvant therapy for men with locally advanced disease who have undergone RP (NCCN 2013). 34 It can also be used in conjunction with radiation therapy in men who are classified as high-risk for progression (NCCN 2013). ADT works to block the effects or production of testosterone and dihydrotestosterone, the main androgens responsible for cancer cell growth (Hoffman 2008). Although it is not a curative treatment, up to 80% of men will initially respond to ADT, though response will decrease over time resulting in castration-resistant disease (Hoffman 2008). Some of the molecular mechanisms contributing to castration resistance in prostate cancer have been discussed previously. Ultimately, the decisions made regarding treatment selection in prostate cancer management are heavily reliant on accurate risk stratification. Unfortunately, categorization of patients into risk groups remains challenging due to a combination of factors that hinder accurate assignment of Gleason score. 1.3.5 Challenges associated with current clinical management of prostate cancer It is well documented that prostate cancer is a largely heterogeneous disease at the genomic, pathological and clinical levels. As described previously, development and progression of prostate cancer can be associated with an array of chromosomal, genetic and epigenetic abnormalities, which differ between patients categorized within the same clinical risk groups. However, intratumour heterogeneity also exists, such that different tumour foci within a single prostate can have diverse histopathological traits (different Gleason patterns) as well as differences in chromosomal, genetic and epigenetic profiles (Gerlinger 2012). The multifocal nature of the disease, combined with the potential for non-uniformity of the histopathological features between tumour foci introduces uncertainty to the biopsy procedures used for Gleason scoring, and subsequently, risk stratification. Currently, biopsy sampling inadequacy remains a significant challenge in proper risk assessment, such that small volume tumour foci with high Gleason pattern are frequently missed during the biopsy procedure (Fine 2008). As a result, up to 30% of patients diagnosed with Gleason score 6 prostate cancer at the time of biopsy are 35 pathologically upgraded to Gleason score 7 following radical prostatectomy (Fine 2008) (true ‘low-risk’ and false ‘low-risk’ biopsies are demonstrated in Figure 1.5). In these cases, patients are falsely categorized as ‘low-risk’ for disease progression, when in fact they should have been clinically upgraded to an ‘intermediate-risk’ group. The issues with sampling inadequacy are further compounded by a substantial degree of interobserver variability during pathology review, thus contributing to inaccurate risk assignment (Netto 2011). Unfortunately, inaccurate risk assignment detrimentally impacts treatment selection for prostate cancer patients. There is strong evidence suggesting that true `low-risk` prostate cancers have no potential for metastatic progression within five years of diagnosis (Ross 2012), and therefore, these patients would be excellent candidates for active surveillance. However, given the inability to confidently predict Gleason score in nearly 30% of patients, appropriate treatment selection remains challenging. Ultimately, men who fall into the ‘low-risk’ category at the time of biopsy are faced with the decision to undergo more aggressive treatment than is likely necessary, resulting in undesirable side effects which impact quality of life, or to forego treatment in favour of active surveillance, often resulting in a high degree of patient anxiety from having to live with cancer (Nicolaisen M 2014, Venderbos LD 2014). As such, the identification of biomarkers that could assist in risk-stratification by increasing the accuracy and utility of prostate biopsies is of critical importance, and would vastly improve disease management and lead to a reduction in prostate cancer over-treatment. To accomplish this goal, biomarkers need to be identified which can be evaluated in Gleason pattern 3 cancer to predict the presence of higher pattern disease that was not detected in the prostate biopsy. As the clinically significant difference between Gleason pattern 3 and Gleason pattern 4 cancers is the increase in associated risk of disease progression, the first step in developing prognostic biomarkers is to identify the molecular differences which define these two pathologies. 36 Figure 1.5 Representation of true and false 'low-risk' prostate biopsies Sampling inadequacies in prostate cancer biopsies can lead to incorrect assignment of Gleason score. True ‘low-risk’ biopsies are obtained when the cores taken accurately represent the range of Gleason patterns present within the prostate (A). False ‘low-risk’ biopsies are the result of cores not being taken in regions of small volume, higher pattern disease, thereby underrepresenting the Gleason score of the tumour (B). In this case, it may be possible to identify the presence of undetected Gleason pattern 4 through the identification of shared molecular signatures that can be profiled in the pattern 3 biopsies. 37 1.3.6 EMT as a factor in prostate cancer progression It is well established that EMT significantly contributes to metastatic progression in many types of cancer, as outlined previously. Recent evidence in prostate cancer suggests that the expression levels of EMT-associated proteins such as E-cadherin, Snail, Twist and Vimentin, are significantly correlated with biochemical recurrence and advancing Gleason score, suggesting a role for this cellular program in progression to aggressive disease (Behnsawy 2013). Therefore, EMT may represent a clinically relevant option for the development of prognostic indicators for prostate cancer progression. EMT in prostate cancer appears to be predominantly initiated in response to hypoxia (through HIF1α) and growth factor signaling, such as TGFβ and EGF (Bao 2012, Mak 2010, Gan 2010). Hypoxia and TGFΒ signaling are both associated with high grade prostate cancer and are believed to work in conjunction with VEGFA to promote EMT by activating Snail transcription (Mak 2010). Furthermore, the STAT3/HIF1α signaling axis upregulates EMT transcription factor TWIST1 in response to both TGFΒ and EGF signaling (Cho 2014). In the context of aggressive tumour phenotypes, upregulation of STAT3 signaling occurs in response to loss of AR expression, ultimately leading to EMT activation and the upregulation of stem cell markers, leading to the development of castration-resistant, invasive populations (Lee 2013, Izumi 2013). Additionally, upregulation of CCL2 upon AR loss in advanced prostate cancer leads to the recruitment of tumour-associated macrophages, which in conjunction with the onset of EMT can facilitate metastatic progression (Izumi 2013). The functional role of cells from the tumour microenvironment in contributing to EMT and prostate cancer aggressiveness should not be overlooked. Signaling from cancer-associated fibroblasts results in the upregulation of factors such as HIF1α, beta-catenin and NFkB, leading to the initiation of EMT, aggressive tumour behaviour and stem cell-like characteristics (Giannoni 2011). Correlative studies have indicated that high levels of Snail expression are associated with advancing Gleason score, while E-cadherin is found predominantly in low grade tumours (Poblete 2014). Similarly, increased expression levels of the 38 EMT-TF, TWIST (Raatikainen 2014) are strongly correlated with increasing Gleason score. Taken together, this evidence strongly suggests that genes associated with EMT, may have utility in distinguishing between ‘low-risk’ Gleason pattern 3 and ‘intermediate-risk’ Gleason pattern 4 prostate cancers. Should this be the case, it may eventually translate into the ability to distinguish Gleason pattern 3 disease in a Gleason score 3+4 biopsy from that in a Gleason score 3+3 biopsy. In that way, EMT gene signatures may be useful in increasing the accuracy of risk stratification from prostate cancer biopsies, by identifying those that contain undetected higher pattern disease. 39 Chapter 2: Rationale, Hypothesis and Objectives At the onset of this research project, exploratory studies in our lab aimed at characterizing dysregulated signaling networks in chemotherapy resistant ovarian cancer lead to the identification of upregulated EMT-associated genes in Cisplatin resistant ovarian cancer cells, relative to drug sensitive cells. At that time, the importance of the EMT phenotype in the behaviour of malignant cells was just being recognized, and our initial discovery in ovarian cancer prompted us to think more broadly about the potential role of this pathway in cancer. The literature at that time was only beginning to highlight the correlation between EMT and chemoresistance, but already strongly supported the idea that the EMT was involved in other aggressive cancer behaviour, such as progression to metastasis. Therefore, our focus was broadened to allow us to explore several key aspects of EMT, using various cancer models. In order to assess the involvement of EMT in chemotherapy response and cancer metastasis, optimal disease models for these behaviours were selected for further study. Ovarian cancer provides an excellent model in which to evaluate chemotherapy resistance, as it is one of the greatest clinical challenges associated with successfully treating this disease. Up to 80% of initially chemoresponsive patients will relapse with chemotherapy resistant disease within the first five years. Therefore, being able to predict which patients are unlikely to respond to first-line chemotherapeutics would ultimately lend itself to risk stratification and delineation for alternative treatment options. However, ovarian cancer does not provide an optimal model to study EMT in the context of the initiation of metastasis, due to the fact that nearly 75% of patients are diagnosed with disease that is no longer organ-confined. In this case, the contribution of EMT to the onset of metastatic events is better modeled in prostate cancer, where progression to metastasis is typically slow and associated with well-characterized histopathological features that may be correlative 40 with EMT gene signatures. Again, being able to predict which patients have an increased risk for disease progression will lead to improved clinical management of potentially aggressive prostate cancer. Additionally, developing a better understanding of the molecular regulation of EMT in the context of aggressive cancer phenotypes may lend itself to other aspects of cancer management, such as the identification of novel therapeutic strategies. Therefore, given the evidence linking EMT with poor cancer prognosis, we hypothesize that EMT contributes to aggressive cancer phenotypes, specifically chemoresistance and progression to metastasis, and that the associated gene expression can be profiled in ovarian and prostate tumours to distinguish aggressive from non-aggressive disease. To assess the contribution of EMT to chemotherapy resistance in ovarian cancer and the potential for metastatic progression in prostate cancer, the following objectives were identified: 1. Evaluating the contribution of EMT to chemotherapy resistance in ovarian cancer A. Establish that the observed over-expression of EMT-associated genes translates to EMTspecific phenotypes such as changes in cell morphology and increases in rates of cell migration and invasion, and that these phenotypes can be reversed through modulation of EMT transcription factors, Snail (SNAI1) and Slug (SNAI2). B. Assess the impact of Snail and Slug on Cisplatin resistance in the A2780cis ovarian cancer cell line. C. Evaluate the potential contribution of EMT to chemotherapy resistance in primary ovarian tumours. 41 2. Evaluating EMT as an indicator of metastatic potential in prostate cancer A. Identify a comprehensive network of genes which are functionally associated with EMT, to be used in gene expression studies interrogating the presence of EMT in prostate cancer. B. Using a retrospective cohort of 62 FFPE radical prostatectomy specimens (31 Gleason score 3+3 and 31 Gleason score ≥ 3+4), assess EMT-associated gene expression from tumour regions predominant in Gleason pattern 3 and 4 disease in order to identify gene signatures that can be used to distinguish between patterns, and potentially be used in patient risk stratification. C. Employ 2D and 3D in vitro models to functionally validate the identified genes of interest in order to assess their roles in EMT-associated phenotypes. 42 Chapter 3: EMT transcription factors, Snail and Slug, contribute to Cisplatin resistance in epithelial ovarian cancer 3.1 Abstract Background EMT is a molecular process through which an epithelial cell undergoes transdifferentiation into a mesenchymal phenotype. The role of EMT in embryogenesis is wellcharacterized and increasing evidence suggests that elements of the transition may be important in other processes, including metastasis and drug resistance in various different cancers. Methods Agilent 4 × 44 K whole human genome arrays and selected reaction monitoring mass spectrometry were used to investigate mRNA and protein expression in A2780 Cisplatin sensitive and resistant cell lines. Invasion and migration were assessed using Boyden chamber assays. Gene knockdown of Snail and Slug was done using targeted siRNA. Clinical relevance of the EMT pathway was assessed in a cohort of primary ovarian tumours using data from Affymetrix GeneChip Human Genome U133 plus 2.0 arrays. Results Morphological and phenotypic hallmarks of EMT were identified in the chemoresistant cells. Subsequent gene expression profiling revealed upregulation of EMT-related transcription factors including Snail, Slug, TWIST2 and ZEB2. Proteomic analysis demonstrated up regulation of Snail and Slug as well as the mesenchymal marker Vimentin, and down regulation of Ecadherin, an epithelial marker. By reducing expression of Snail and Slug, the mesenchymal phenotype was largely reversed and cells were resensitized to Cisplatin. Finally, gene expression 43 data from primary tumours mirrored the finding that an EMT-like pathway is activated in resistant tumours relative to sensitive tumours, suggesting that the involvement of this transition may not be limited to in vitro drug effects. Conclusions This work strongly suggests that genes associated with EMT may play a significant role in Cisplatin resistance in ovarian cancer, therefore potentially leading to the development of predictive biomarkers of drug response or novel therapeutic strategies for overcoming drug resistance. 44 3.2 Introduction Of the gynecological malignancies, ovarian cancer has the highest associated mortality rate in the western world (Khalil 2010, Kobel 2008). While relatively rare at 1 in 71 women affected in Canada (Canadian Cancer Society 2011), approximately 70-80% of patients with ovarian cancer will succumb to the disease within five years of diagnosis (Holschneider 2000). The high mortality rate is due, in part, to the fact that ovarian cancer is often diagnosed in advanced stage, because of a lack of measurable early symptoms and ineffective screening techniques (Moore 2010, Kulasingam 2010). Of equal importance, 20% of tumours display primary resistance to platinum compounds while the majority of initial responders will relapse, often as a result of acquired drug resistance (Visintin 2008, Dressman 2007). Standard treatment for ovarian cancer involves tumour debulking and platinum-based chemotherapy administered intravenously or intraperitoneally (Yen 2001, Hogberg 2010). Cisplatin, the most common first line chemotherapeutic drug, is a platinum compound that binds to and cross-links DNA (Marsh 2009). During cell division Cisplatin-DNA adducts block replicative machinery, inducing the DNA damage response, and eventually apoptosis (Marsh 2009, Galluzzi 2011). It has been proposed that decreased cellular uptake of drug as well as increased capacity for DNA damage repair and anti-apoptotic signaling may play a role in Cisplatin resistance displayed by many tumours (Galluzzi 2011, Konstantinopolous 2008, Li 2007, Cheng 2006, Ahmad 2010, Burger 2011). Recent evidence has suggested that EMT may play a role in the development of chemoresistance. EMT is a critical process in embryogenesis (Hay 2005) and has been well studied in that context. It is characterized by up-regulation of extracellular matrix components, a loss of intercellular cohesion, increased rate of cellular migration and invasion, as well as increased resistance to apoptosis, and is modulated by a number of transcription factors, namely SNAI1 (Snail) and SNAI2 (Slug) (Kalluri 2009, Micalizzi 2010). In early embryogenesis, these 45 cellular traits enable both the formation of the germinal layers during gastrulation by facilitating formation of the mesoderm and endoderm from cells in the primitive streak, and derivation of migratory neural crest cells from the epithelial neural plate (Acloque 2009). EMT also has a significant role later in embryo development during tissue reorganization and organ modeling (Soo 2002, Ekblom 1996). The same cellular remodeling and signaling networks appear to be active during metastasis, and may also contribute to the development of drug resistance in tumour cells (Geiger 2009, Zlobec 2010). During cancer progression, EMT appears to promote dissemination of cells from the tumour mass (Sabe 2011) and facilitates tissue invasion by regulating the production of matrix metalloproteinase s and altering cytoskeletal organization (Zhang 2011, Zhao 2011). In models of drug resistant breast and ovarian cancers, EMT gene signatures have been found to correlate with the presence of drug resistance (Iseri 2011, Helleman 2010) and manipulation of EMT transcriptional regulators modulates resistance to chemotherapeutic drugs in lung and bladder cancers (Chang 2010, McConkey 2009). Additional evidence suggests that EMT may contribute to the acquisition of drug resistance by altering expression of key genes involved in cell cycle regulation, drug transport and apoptosis (Vega 2004, Saxena 2011, Wu 2009). We hypothesized that the genes that regulate EMT have a role in Cisplatin resistance in ovarian cancer. Using a cell line model of Cisplatin-resistant ovarian cancer, we demonstrate features of a mesenchymal phenotype in the resistant cells relative to their epithelial parent cells. Additionally, we demonstrate increased capacity for migration and invasion in resistant compared to sensitive cells, characteristics that are reverted following reduction of expression of Snail and Slug. Using a cohort of primary human ovarian tumours, we demonstrate that EMT genes are upregulated in chemonaïve drug resistant tumours, suggesting that these genes may act as biomarkers of chemotherapy resistance. Finally, we modulate resistance to Cisplatin in A2780cis 46 cells by manipulating levels of Snail and Slug, suggesting that the key regulators of EMT are directly contributing to the drug resistant phenotype in ovarian cancer cells. 3.3 Materials and Methods Cell lines and culture conditions The Cisplatin-sensitive A2780 ovarian adenocarcinoma cell line and its daughter line, A2780cis, were obtained from the European Collection of Cell Cultures (Salisbury, UK). Cells were cultured in RPMI with 10%FBS, 1% penicillin/streptomycin and 1% L-glutamine at 37°C in 5% CO2. A2780cis cells were maintained in media with 1 μM Cisplatin. For all assays, cells were grown to 70-80% confluence and harvested following trypsinization. Analysis of cell morphology was done at 20× magnification using a Zeiss Axiovert 25 Phase Contrast Inverted Microscope. Digital images were captured using a Canon Power Shot G10 equipped with a Carl Zeiss 426126 lens. Human ovarian tumours Consent for tumour banking was obtained and the study was approved by the Research Ethics Boards at both Kingston General Hospital and The Ottawa Hospital. Tumour samples were obtained from the Division of Gynecologic Oncology Ovarian Tissue Bank and the Ontario Tumour Bank. All tumours were chemonaïve at collection. Seventeen tumours were classified as chemosensitive (progression-free interval of greater than 18 months) and eleven as chemoresistant (progression-free interval of less than 6 months) using available follow-up clinical data. Histological assessment of samples confirmed that each sample contained > 70% tumour cells. Gene expression profiling and analysis Total RNA was extracted from the cell lines and tumours using the Qiagen miRNeasy Mini kit (Toronto, Canada). RNA quality was assessed to have an RNA integrity number of at least eight using an Agilent 2100 Bioanalyzer (Mississauga, Canada). For cell lines, total RNA 47 was labeled with Cyanine-3 dye using a Quick Amp Labeling Kit (Agilent, Mississauga, Canada) and hybridized to Agilent Whole Human Genome (4 × 44 K) Microarrays (Mississauga, Canada) for 17 hours in a rotating SciGene model 700 oven (Sunnyvale, USA). Arrays were scanned using an Agilent Technologies DNA Microarray Scanner and data were feature extracted using Feature Extraction Software 10.5.1.1 (Agilent) and statistically analyzed using default settings on GeneSpring GX 11.0.1 software (Agilent). Expression profiles from the tumor RNA were obtained using Affymetrix GeneChip Human Genome U133 plus 2.0 arrays. Raw data were imported into GeneSpring GX 11.0.1 and analyzed. Unsupervised hierarchical clustering of the tumour samples was completed using the self-organizing maps algorithm in the GeneSpring GX 11.0.1 package. qRT-PCR Taqman™ arrays Snail and Slug expression levels were analyzed using Taqman™ assays (Applied Biosystems, Streetsville, Canada, item Hs00195591_m1 and item Hs00950344_m1) and the SuperScript III First-Strand Synthesis SuperMix kit for qRT-PCR (Invitrogen, Burlington, Canada). PCR conditions were 50°C for 2 minutes, 95°C for 10 minutes, 40 cycles of 95°C for 15 seconds, 60°C for 1 minute. GAPDH was used as an internal control. As a measure of relative change in expression between the parental and resistant samples, ΔΔCt values were calculated and converted to approximate fold change values (2-ΔΔCt) (Schmittgen 2008). Approximate fold change = 2-ΔΔCt 2-ΔΔCt = [(Ct gene of interest – Ct internal control)sample A – (Ct gene of interest – Ct internal control)sample B)] 48 Cell proliferation assay Cells were plated in 24-well plates at 5 × 104 cells/well. Cells were harvested and counted using a haemocytometer after 24, 48 and 72 hours. Average cell counts were used to produce growth curves, from which cell doubling time was calculated. Wound healing assay Cell migration was assessed using wound-healing assays. Cells were grown in a confluent monolayer in a 60 mm plate. A wound was inflicted in the cell layer by scratching the plate with a sterile pipette tip. Plates were rinsed gently with media twice prior to incubation to remove non-adherent cells. Digital images of the wound were obtained at times 0 hours, 12 hours and 36 hours at 10× magnification. Effects of proliferation were controlled for by using a reduced serum medium (3%FBS) and monitored via cell count. Boyden chamber migration and invasion assays Cells were serum starved for 24 hours prior to use. Media with 10% FBS was added to the wells of a 24-well plate. BD Falcon™ Cell Culture Inserts (8µm pores) (BD Biosciences, Mississauga, Canada) were placed in each well. 2 × 103 cells in serum-free media were added to the interior of each insert. Plates were incubated for 24 hours at 37°C in 5% CO2, and media removed from the insert, which was then washed with PBS. Insert membranes were fixed with cold methanol for 10 minutes, stained with 0.5% Crystal Violet in 25% methanol for 10 minutes and rinsed with water to remove excess dye. Membranes were removed from the insert, placed under a microscope and the number of cells that migrated through the porous membrane was counted. Invasion assays were done as described above using BD BioCoat™ Matrigel™ invasion chambers (BD Biosciences, Mississauga, Canada). For each assay, three independent experiments were completed with triplicate technical replicates, per experiment. Data are presented as mean ± standard deviation. 49 MTT assays 5 × 103 cells/well were seeded in 96-well plates in 100 µl medium with 10 μM Cisplatin and without phenol red and left to incubate for 48 hours at 37°C and 5% CO2 . After 48 hours, 10 μl MTT (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) (Sigma-Aldrich, Oakville, Canada) was added to each well and cells were left for 4 hours. After incubation, 150 μl MTT solvent (0.1 N HCl in anhydrous isopropanol) was added to each well and mixed thoroughly by pipetting until all formazan crystals were dissolved. Colourimetric change was measured at 570 nm and background absorbance at 690 nm. Final values were obtained by subtracting OD690 nm from OD570 nm. MTT assays for siRNA optimization were done without adding Cisplatin and the initial seeded cells were only incubated for 24 hours prior to MTT addition. Gene knockdown Pre-designed siRNA oligos for Snail (cat# SASI_Hs01_00039785, duplex sequences: 5'GCCUUCAACUGCAAAUACU and 5'AGUAUUUGCAGUUGAAGGC) and Slug (cat# SASI_Hs01_00159363, duplex sequences: 5'GCAUUUGCAGACAGGUCAA and 5'UUGACCUGUCUGCAAAUGC) were purchased from Sigma-Aldrich (Oakville, Canada). Optimization of gene knockdown was done using AllStars Hs Cell Death Control siRNA (Qiagen, Toronto, Canada), a mix of siRNAs that target genes essential for cell survival. 5 × 104 cells/well were plated in 24-well plates and incubated overnight. After 24 hours cells were transfected with 7.5 ng, 19 ng, 37.5 ng or 75 ng of AllStars Hs Cell Death Control siRNA and 1.5 μl, 3 μl or 7.5 μl of HiPerFect Transfection Reagent (Qiagen, Toronto, Canada). Cells were incubated for 72 hours and cell death measured using MTT assays. Maximum cell death was achieved using 19 ng siRNA with 3 μl of transfection reagent. Optimal time points were then established using siRNA targeted against Snail and Slug. 24, 48 and 72 hours after transfection the level of Snail and Slug transcript was determined by qRT-PCR TaqMan assay. Optimal 50 knockdown of both genes was seen 72 hours post transfection. Efficiency of transfection in the A2780 and A2780cis cells was determined to be 87% and 84%, respectively. For knockdown experiments, 5 × 104 cells/well were plated in 24-well plates and incubated overnight. 19 ng of siRNA and 3 μl of HiPerFect transfection reagent were diluted in 100 μl serum-free RPMI and were incubated for 10 minutes at room temperature. Transfection complexes were then added drop-wise onto the cells. Cells were incubated for 72 hours at 37°C and 5% CO2. Media was changed as necessary. Transfection of A2780cis cells with AllStars Negative Control (Qiagen), an siRNA sequence with no homology to any known mammalian gene, was used as a control for this experiment. Cells with reduced expression were designated as follows: A2780cisSN (Snail knockdown), A2780cisSL (Slug knockdown), A2780cisSN/SL (both Snail and Slug knocked down). Statistical analysis Each assay was performed in triplicate. Data are expressed as the mean ± standard deviation (SD). Statistical significance of all data was evaluated using the Student's t-test, p < 0.05. Sample preparation for proteomic analysis Samples were lysed in RIPA buffer containing Halt Proteinase and Phosphatase Inhibitor Cocktail (PIERCE, IL, USA). Cells were sonicated and lysates incubated at 4°C for 30 minutes with shaking. Supernatants were separated by centrifugation for 15 minutes at 4°C. Protein concentration was measured with a Bio-Rad DC Protein Assay kit and 30 μg of protein were used for tryptic digestion. Each aliquot was dried, dissolved in 0.5 M triethylammonium bicarbonate (TEAB), reduced by adding 2 μl of tris-(2-carboxyethyl)-phosphine (TCEP), incubated at room temperature for 1 hour, alkylated using 1 μl of methyl methane thiosulfonate (MMTS) and incubated at room temperature for 1 hour in the dark. 2 μg trypsin were added for overnight 51 digestion at 37°C. The tryptic digest was desalted by using a C18 spin column (PIERCE, USA), dried under vacuum and resuspended in 0.1% formic acid for SRM analysis. LC/SRM-MS analysis The LC/SRM-MS analytical system consisted of an Eksigent nanoflow HPLC (AB Sciex, USA) coupled to a 5500 QTRAP® hybrid triple quadrupole/linear ion trap mass spectrometer (AB SCIEX, USA). 1 μg of digested protein was loaded onto a trap column (0.3 mm I.D, 5 mm L), packed with 5 μm Zorbax SB-C18, 300 Å pore (Agilent, USA). Peptides were separated on a 75 μm I.D., 15 cm long nanoflow column with 15 μm spray tip (New Objectives, USA). A linear gradient profile was employed starting from 5% solvent B to 40% B in 60 minutes (solvent A was 2% acetonitrile in water with 0.1% formic acid; solvent B was 2% water in acetonitrile with 0.1% formic acid). The ion spray voltage was set at 2300 V, and source temperature at 160°C. The declustering voltage was 100 V, and collision energy value was selected for each transition as generated by MultiQuant® software (AB Sciex, USA). To ensure sensitive and reliable quantification, tryptic peptides and SRM transitions were generated by MRMPilot software (AB SCIEX, USA) based on common chemical rules of peptide fragmentation. The specificity of each peptide was verified using BLAST alignment against the NCBI-NR human protein database. Raw SRM-MS data was preprocessed using MultiQuant 2.0.2 software (AB SCIEX). A 2-point Gaussian smoothing window was applied to all transition peaks. Peak area for each transition was averaged over three replicate experiments per resistant and sensitive cell line sample; fold-change was calculated from the ratio of the average peak area in resistant cells to that in sensitive cells. A transition was discarded if peak area coefficient of variation (CV) across replicates was greater than 0.2 or peak area in any replicate was below the 25th percentile. Peptides with at least 2 transitions satisfying these constraints were conserved. 52 3.3 Results 3.3.1 Cisplatin resistance in A2780cis correlates with changes in cellular morphology consistent with EMT signaling We examined the morphological characteristics of the cell lines during exponential growth. Parental A2780 cells formed cohesive clusters with round cellular morphology in vitro (Figure 3.1a), consistent with an epithelial phenotype. In contrast, A2780cis cells, grown in the presence of 1 μM Cisplatin, display a spindle-like morphology and formed dyscohesive sheets (Figure 3.1b). Additionally, the resistant cells exhibit the formation of pseudopodia (Figure 3.1b, inset), not seen in the parental cells. 53 Figure 3.1 Morphology of A2780 and A2780cis cells Cells visualized under 20× magnification. Drug sensitive cells (A) have round morphology and grow in tight clusters with substantial cellular cohesion. Comparatively, the drug resistant cells (B) have a more fibroblastic appearance and demonstrate reduced intercellular contacts. Additionally, drug resistant cells extend pseudopodia (inset). 54 3.3.2 Whole transcriptome profiling and LC/SRM-MS analysis identify components of EMT signaling networks in A2780cis relative to A2780 To confirm the involvement of EMT elements in the development of Cisplatin resistance, whole transcriptome microarrays (Agilent, Mississauga, Canada) were used to compare gene expression between A2780 and A2780cis cells during exponential growth. In repeated assays, EMT pathway elements, including Snail, Slug, TWIST2 and ZEB2, were over-expressed in the Cisplatin-resistant cell line relative to the parental line (Figure 3.2a). Technical validation by qRT-PCR using Taqman assays confirmed relative overexpression of Snail by 3.7 ± 0.3 (SD) fold and Slug by 6.9 ± 0.4 (SD) fold, in the drug-resistant cells compared to the drug-sensitive cells (Figure 3.2b). Protein analysis by liquid chromatography/selected reaction monitoring mass spectrometry (LC/SRM-MS) confirmed upregulation of Snail, Slug and Vimentin, as well as downregulation of epithelial marker E-cadherin in the drug resistant cells relative to the drug sensitive cells (Table 3.1). 55 Figure 3.2 Upregulation of gene associated with EMT in resistant cells Genes known to 'regulate EMT were shown to be upregulated in the A2780cis cells compared to the A2780 cells. Data from the gene expression microarrays (a) and technical validation of Snail and Slug expression levels using qRT-PCR TaqMan assay (b) demonstrate that upregulation. Approximate fold change calculated as ΔΔCt for Snail is 3.7 ± 0.3 (p-value = 0.0008) and for Slug is 6.9 ± 0.4 (p-value = 0.001). 56 Table 3.1 Summary of SRM-MS quantification results for EMT-associated proteins Protein Peptide GPFPKNLVQJK E-cadherin VFYSITGQADTPPV GVFIIER EYQDLLNVK Vimentin ILLAELEQLK Snail SFLVR HFNASK Slug VSPPPPSDTSSK Fragment b8 b9 b10 y8 y9 y10 y6 y7 y8 y6 y7 y8 y9 b3 b4 y3 b3 b5 b10 b11 y9 Fold Change -1.18 -1.55 -1.63 -4.48 -4.16 -4.01 1.15 1.14 1.11 1.01 1.08 1.10 1.08 1.25 1.21 1.29 1.10 1.27 1.05 1.19 1.24 Mean Peak Area Sensitive Resistant 5.82E+04 3.22E+04 1.57E+04 1.01E+06 3.33E+04 2.05E+04 2.10E+04 5.02E+03 8.25E+04 1.99E+04 8.45E+04 2.11E+04 4.78E+06 2.52E+06 2.19E+06 2.50E+06 1.27E+05 1.14E+05 4.58E+06 4.62E+06 7.34E+06 7.94E+06 1.11E+07 1.22E+07 1.35E+06 1.47E+06 6.00E+04 7.50E+04 7.38E+04 8.89E+04 2.14E+04 2.76E+04 1.92E+04 2.10E+04 2.40E+05 3.06E+05 9.96E+04 1.04E+05 4.47E+04 5.34E+04 4.00E+04 4.96E+04 CV Sensitive 0.01 0.07 0.03 0.09 0.02 0.07 0.01 0.02 0.02 0.02 0.04 0.04 0.04 0.14 0.03 0.16 0.18 0.06 0.13 0.07 0.07 Resistant 0.13 0.08 0.19 0.10 0.12 0.11 0.01 0.01 0.05 0.03 0.05 0.04 0.04 0.03 0.07 0.04 0.14 0.16 0.07 0.05 0.06 Mean peak area and coefficient of variation (CV) for each peptide fragment ion were derived from peak areas measured in three replicate assays of each of the sensitive and resistant cell line samples. Transitions with CV > 0.20 in either sample are not shown. 57 3.3.3 Cisplatin resistant cells display increased potential for migration and invasion Migratory capacity was measured using Boyden chamber assays. In three independent experiments, 23.7 ± 5.1 drug sensitive cells and 194.0 ± 7.0 drug-resistant cells migrated through the membrane after 24 hours (p-value = 0.003) (Figure 3.3c). Resistant cells also showed a fivefold increased invasive capability relative to sensitive cells using a Matrigel-coated Boyden chamber assay. After 24 hours, an average of 36.3 ± 5.03 sensitive cells and 188.3 ± 4.04 resistant cells invaded the matrix (p-value = 0.003) (Figure 3.3c). 3.3.4 Knockdown of Snail and Slug reverses the EMT phenotype and reduces cellular resistance to Cisplatin SiRNA was used to reduce the levels of Snail and Slug transcript in the A2780cis cells to those of the parent A2780 cell line (A2780cisSN - Snail knockdown, A2780cisSL - Slug knockdown, A2780cisSN/SL - both Snail and Slug knocked down). Knockdown of these genes was confirmed by qRT-PCR (results not shown) and resulted in reversion of cell morphology of the resistant cells to that of the parent cell line (Figure 3.3a). Additionally, the knockdowns resulted in a reduction of the doubling time of the resistant cells from 45.7 ± 2.4 hours to 31.1 ± 2.4 hours (p-value = 0.017), similar to the 27.3 ± 1.6 hour doubling time of the drug-sensitive A2780 cells (p-value = 0.15) (Figure 3.3b). Migratory capacity of the transfected cells was reduced from 194.0 ± 7.0 (untransfected controls) to 33.67 ± 3.1 cells migrating through the membrane (p-value = 0.002) (Figure 3.3c). Invasive capability was reduced from 188.3 ± 4.04 (untransfected controls) to 56 ± 4.58 cells invading the matrix (p-value = 0.004) (Figure 3.3c). Finally, we showed increased cellular sensitivity to Cisplatin in transfected cells relative to controls (Figure 3.4). Following 48 hours of growth in 10 μM Cisplatin, transfected cells displayed 62 ± 2.5% cell death compared to controls at 37 ± 2.4% (p-value = 0.005). As a comparator, parental drug-sensitive cells displayed 75 ± 3.3% cell death. 58 59 Figure 3.3 Knockdown of Snail and Slug reverses EMT phenotype Snail and Slug knockdown resulted in an epithelial morphology (a) and reduced the doubling time of the A2780cisSN/SL cells (p-value = 0.017) compared to the A2780cis cells (b). Additionally, knockdown of Snail and Slug resulted in a reduction in migration from 194.0 ± 7.0 cells to 33.67 ± 3.1 cells migrating through the membrane (p-value = 0.002) as well as reducing invasion rate from 188.3 ± 4.04 cells to 56 ± 4.58 invading the matrix (p-value = 0.004), in each case, making the A2780cisSN/SL knockdown cells more similar to the A2780 cell line, having migration and invasion rates of 23.7 ± 5.1 and 36.3 ± 5.03, respectively (c). The control group, A2780cis cells transfected with a scrambled siRNA sequence, shows no statistically significant changes in morphology, doubling time, migration or invasion. 60 Figure 3.4 Cisplatin sensitivity with Snail and Slug knockdown Cells with Snail and Slug knockdowns were grown in 10 μM Cisplatin for 48 hours and cell survival was determined by MTT assay. Approximately 75 ± 3.3% of the drug sensitive cells died after treatment compared to 37 ± 2.4% of the drug resistant cells. With Snail and Slug knockdown, cell death was significantly increased to 62 ± 2.5% (p-value = 0.005). 61 3.3.5 Drug resistant human ovarian tumours can be differentiated from drug sensitive ovarian tumours using a biomarker panel of EMT-related genes Human ovarian tumour samples were classified as chemosensitive (progression-free interval of greater than 18 months) or chemoresistant (progression-free interval of less than 6 months) based on available clinical data (Supplementary Table S1). Gene expression data from 17 sensitive and 11 resistant tumours were mined to investigate the relative expression levels of EMT-related genes. In addition to Snail, Slug, TWIST2 and ZEB2, the resistant tumour samples also showed increased expression of TWIST1 and ZEB1, two other important regulators of EMT (Figure 3.5a). We used published information about EMT to derive a list of 17 additional genes with a role directly upstream or downstream of Snail and Slug (Table 3.2). Unsupervised hierarchical clustering using expression data from those genes allowed reasonable separation between the drug sensitive and drug resistant ovarian tumours (Figure 3.5b). 62 Figure 3.5 EMT gene signatures in drug resistant ovarian tumours EMT gene signatures are present in primary drug resistant ovarian tumours. (a) Gene expression arrays identify increased expression of genes integral to EMT in the resistant tumours relative to the sensitive tumours. (b) Unsupervised hierarchical clustering based on a panel of genes related to EMT supports differentiation between drug sensitive (blue) and drug resistant (red) tumours. 63 Table 3.2 List of genes used for unsupervised hierarchical clustering of primary tumours Gene Symbol VIM EGFR PPARG IGF1 TGFΒ2 FN1 ZEB1 SNAI2 TWIST2 RXRA KRT5 KRT15 KRT17 KRT18 KRT7 KRT16 KRT4 Description Vimentin Epidermal growth factor receptor Peroxisome proliferator-activated receptor gamma Insulin-like growth factor 1 Transforming growth factor beta 2 Fibronectin Zinc finger E-box-binding homeobox 1 Slug (Zinc finger protein) Twist-related protein 2 Retinoid X receptor alpha Keratin5 Keratin15 Keratin17 Keratin18 Keratin7 Keratin16 Keratin4 Genes used to perform unsupervised hierarchical clustering on the primary ovarian tumours, chosen based on their documented involvement in EMT in recent literature. 64 3.5 Discussion Ovarian cancer exhibits a high rate of platinum sensitivity in the first-line setting, but resistance frequently develops in recurrent disease (Solar 2011). As such, understanding the signaling networks that regulate chemoresistance is critical for successful treatment. This can be said of any cancer commonly treated with Cisplatin such as small cell lung cancer, head and neck cancer, colorectal and hepatocellular cancers. While EMT has been widely studied for its role in early development and, more recently, cancer metastasis, it may also contribute to a cellular ability to evade the effects of platinumbased therapies. EMT results in the transformation of a differentiated epithelial cell to a mesenchymal cell with stem-like properties, and is characterized by loss of cell-to-cell adhesion, specifically through the dismantling of adherens, tight and gap junctions, as well as loss of cell polarity and increased motility (Moustakas 2007). In embryogenesis, EMT functions by promoting migration of mesenchymal cells, first, during gastrulation and neural crest cell migration, then later during tissue remodeling and organogenesis, ultimately contributing to the development of differentiated tissues with specific phenotypes (Hay 2005). In cancer progression, EMT appears to be at least partially responsible for the invasive nature of tumour cells and facilitates metastasis by converting a non-motile cancerous epithelial cell into a motile mesenchymal cell capable of disseminating from the tumour mass and entering the circulatory or lymphatic system (Yang 2008). It is widely believed that embryogenesis and cancer metastasis represent two facets of EMT, the former, a self-contained process that functions to generate diverse cell types and tissues, while the latter is affected by oncogenes and tumour suppressor genes resulting in invasion and metastatic spread through the circulatory system (Zeisberg 2009, Kalluri 2009). It is likely that EMT in drug resistance may rely on many of the same transcription factors that function in embryogenesis and metastasis-related EMT, although details of their 65 regulation are largely unknown. Studies from embryology and metastasis provide evidence to suggest that these cellular changes provide a survival advantage to cells under chemotoxic stress, for example, by down-regulating pro-apoptotic factors such as caspases and Dap1, or upregulating anti-apoptotic factors Dad1 and Bcl2 (LaGamba 2005, Sun 2011). Additionally, oxaliplatin-resistant colorectal cancer cells have been shown to display many of the hallmarks of EMT, including increased migration and invasion as well as a spindle-cell shape, loss of polarity and formation of pseudopodia (Yang 1996). Increased expression of ZEB1 and decreased expression of E-cadherin have been associated with drug resistant pancreatic cell lines and reduction of ZEB1 expression has been implicated in increased drug sensitivity (Arumugam 2009). In ovarian cancer cell lines, upregulation of Snail and Slug has been correlated with resistance to radiation and Paclitaxel and shown to directly participate in p53-mediated prosurvival signaling (Kurrey 2009). Therefore the idea that the EMT genes play a significant role in Cisplatin resistance in ovarian cancer is supported by previous evidence, and our own demonstration of a direct impact of these genes on platinum resistance. Through the manipulation of key EMT signaling molecules, we have been able to show that we can re-sensitize drug resistant cells to the effects of Cisplatin, approaching wild-type sensitivity after 72 hours in culture. Our findings suggest that genes known to regulate EMT directly contribute to Cisplatin resistance in this ovarian cancer model. We do recognize that the A2780 cell line may not provide a faithful model of serous ovarian carcinoma, given that lineage fidelity for this line may not be maintained in long term culture. However, we conclude that the translation of our cell line derived results to a series of primary ovarian carcinomas provides evidence for the effectiveness of this model in this instance. We have demonstrated that a panel of EMT-related genes provides a reasonable model of classifying primary ovarian tumours according to their chemoresistance status. While we recognize that optimization of the biomarker panel would be required before suggesting this could provide a clinical benefit, this initial gene 66 list identifies key differences in the underlying molecular circuitry of drug sensitive and resistant ovarian tumours. We have not determined whether this signature derives entirely from tumour cells, given that we did not enrich for tumour cells in our experiments; however, the fact that all of the primary tumours showed at least 70% tumour cells by histology would support the idea that the tumour likely contributes substantially to the signature we noted. To our knowledge, this is the first study in primary tumours demonstrating an EMT gene signature that can be used to differentiate between chemosensitive and chemoresistant human ovarian tumours and suggests that our overall findings may provide important clues about chemotherapy resistance in ovarian cancer. 3.6 Conclusions In summary, we have demonstrated, through the use of an ovarian cancer cell line, A2780, and its Cisplatin-resistant daughter line, A2780cis, that the genes that regulate the epithelial to mesenchymal transition directly contribute to Cisplatin-resistance in ovarian cancer. We have shown that when Snail and Slug, two key regulators of EMT, are knocked down in our Cisplatin resistant cell line, the EMT phenotype is largely reversed and drug sensitivity is restored. Additionally, we demonstrate that this gene signature is present in drug resistant human ovarian tumours and that we can distinguish between drug sensitive and drug resistant ovarian tumours using the differential expression of a panel of EMT-related genes. Therefore, as it appears that EMT is connected to the drug resistant phenotype, this process and corresponding signaling network may be relevant biomarkers of drug resistance in ovarian cancer. Additionally, these molecules may represent targets for novel therapeutic strategies, used to overcome chemotherapy resistance in ovarian cancer, thereby modulating drug response in these patients and reducing the mortality rate associated with this disease. 67 Chapter 4: Evaluating the role of EMT as an indicator of metastatic potential in prostate cancer 4.1 Abstract Since there are currently no reliable diagnostic assays to separate aggressive from indolent prostate cancers, many men opt for treatments with serious morbidities. The problem lies specifically with cases that present as Gleason grade 3+3 on biopsy, of which up to 30% harbor small volume Gleason pattern 4, that is likely to be associated with a higher risk of progression. Therefore, there is an acute need for prognostic biomarkers in the management of this disease. Evidence suggests that EMT is a prerequisite for the onset of metastatic events in prostate cancer, whereby the molecular changes consistent with this process may act by priming tumors for metastatic progression. We hypothesize that EMT is associated with higher pattern prostate cancer, and that profiling of gene expression associated with this program may be useful in distinguishing between Gleason pattern 3 and Gleason pattern 4 disease, thus providing insight into molecular events in disease progression. To test this hypothesis, a cohort of 31 Gleason pattern 3 and 31 Gleason pattern 4 tumours has been analyzed for differential expression of EMT-associated genes. Ultimately, it was possible to identify a multivariate model that could distinguish between Gleason pattern 4 and pattern 3, based solely on the differential expression of EMT-associated genes. Furthermore, functional validation studies were completed to assess the roles of six genes of interest in characteristic EMT phenotypes such as migration and invasion. Migration rates were significantly reduced in response to SNAI1, ZEB1, HIF1A and SPARC knockdowns, while invasion was repressed as a result of SNAI1, ZEB1, HIF1A, SPARC, VIM and CLTC 68 knockdowns, in both 2D and 3D culture. Gene expression profiles were assessed in these cell lines, and resulted in the identification of 5 nodal points (CDH1, FN1, SPARC, NUPR1, IL6) upon which the dysregulated signaling appears to converge, thereby providing some insight into EMT gene regulation. Ultimately, we have been able to assess the role of EMT in the context of increasing metastatic risk in prostate cancer, and consequently identified a multivariate model capable of distinguishing Gleason pattern 3 and 4 prostate cancers. Our functional assessment of a subset of EMT-associated genes provided insight into the phenotypic and gene regulatory changes associated with this process. Given sufficient validation, we expect that the genes identified in this study may eventually provide utility in risk stratification for prostate cancer management, while further elucidating the molecular mechanisms underlying tumor progression in relation to EMT. 69 4.2 Introduction Prostate cancer is the most commonly diagnosed cancer amongst Canadian men, with one in seven being affected in their lifetime (Canadian Cancer Society 2013). According to the National Cancer Institute’s SEER statistics, overall prostate cancer 5-year survival is 98.8%. Approximately 93% of patients are diagnosed with localized or regionally confined disease, in which case 5-year survival is nearly 100% (Howlader 2013). However, long term studies (>15 years after diagnosis) into the natural progression of prostate cancer have demonstrated that up to 35% of patients will develop indications of distant metastases (Pound 1999, Bubendorf 2000). Unfortunately, after the development of metastatic disease, the 5-year survival rate drops into the range of 25-40% (Pound 1999, SEER 2013). Current clinical practice in prostate cancer management relies on patient risk stratification to estimate disease progression potential and to direct treatment decisions. Risk classification is assigned based on a combination of TNM stage, pre-operative PSA level and Gleason score. Independently, the assigned Gleason score (GS) remains the best prognostic indicator of prostate cancer progression, such that higher Gleason scores correlate with more aggressive disease (GS ≥8), while lower Gleason scores are associated with indolent disease (GS ≤6). Evidence suggests that those prostate cancers which contain only Gleason pattern 3 (GS 6) have zero risk for metastatic progression within the first five years after diagnosis (Ross 2012). However, due to biopsy sampling inadequacies, nearly 30% of patients diagnosed with Gleason score 6 (3+3) prostate cancer on biopsy, actually have undetected small volume pattern 4 identified after radical prostatectomy (Fine 2008). In these cases, the presence of pattern 4 disease results in a clinical upgrade from ‘low-risk’ to ‘intermediate-risk’ of cancer progression. Given this relatively high rate of discordance, resulting in a false ‘low-risk’ categorization of nearly 30% of patients, risk-stratification and subsequent treatment assignment remain challenging (Fine 70 2008). In particular, this issue results in substantial overtreatment of prostate cancer as many men opt for more aggressive treatment than is necessary, rather than engage in active surveillance and live with an uncertain prognosis. Therefore, increasing the clinical utility of prostate cancer biopsies, by identifying molecular signatures associated with increased risk of progression, may provide a foundation on which to delineate false ‘low-risk’ from true ‘low-risk’ cancers, even in the absence of detected higher pattern disease. This would ultimately facilitate more confident decision making with respect to treatment strategies employed in men with prostate cancer. Recent evidence suggests that EMT functions in metastatic progression in prostate cancer (Xu 2006, Banyard 2013). Traditionally, EMT was believed only to occur in cells at the invasive edge of tumours in order to facilitate metastatic spread (Matuszak 2011). However, recent immunohistochemical studies reveal that increased staining of mesenchymal markers such as Snail (SNAI1), TWIST1 (TWIST1) and Vimentin (VIM), as well as weak staining of the epithelial marker, E-cadherin (CDH1) are pervasive throughout prostate tumours (Behnsawy 2013). This indicates that EMT may be induced prior to the initiation of metastasis, in order to prime tumours for invasive progression. As such, it may be that the onset of EMT is associated with high pattern disease (≥4), and subsequently increased risk for metastatic progression. Therefore, genes associated with EMT-regulation may be differentially expressed in Gleason pattern 3 and Gleason pattern 4 prostate cancers in accordance with their acknowledged differences in metastatic potential. If this proves to be the case, EMT-associated genes could serve as prognostic biomarkers in distinguishing true ‘low-risk’ from false ‘low risk’ prostate cancers at the time of biopsy. 71 4.3 Materials and Methods Gene Expression Omnibus (GEO) Data Gene expression data from prostate cancer cell lines grown in 3-dimensional (3D) cultures were obtained from GEO (accession number GSE19426). These data were generated for a study by Harma, V. et al. (2010) for the purpose of investigating the genomic and phenotypic differences in prostate cancer cell growth and behaviour in 2-dimensional (2D) versus 3D cell culture, and to comment on the relevance of 3D culturing in drug discovery and disease modeling. Based on the results from their study, data were collected from four cell lines that display invasive characteristics when grown in Matrigel-based 3D culture: PC-3, PC-3M, ALVA-31, RWPE-2-w99. Gene expression data from these cells grown in 2D culture were used as a baseline against which post-invasion (day 11-13), 3D culture cell, gene expression was measured. The assessment of changes in gene expression was completed using the GEO2R program built into and designed specifically for GEO-formatted expression data. The GEO2R program incorporates the GEOquery and limma R packages from Bioconductor, which enable parsing of the data into R data structures, and basic statistical analysis, allowing for the analysis of differentially expressed genes between data sets in the GEO repository. Statistical analysis was completed using the Benjamini and Hochberg false discovery rate method for multiple testing corrections. Tabulated results include raw p-value, adjusted p-value and log2 fold change between selected groups (3D culture, post-invasion vs 2D culture). Gene network analysis Network building and analysis were completed using two different online, open access network building algorithms: String v9.1 and GeneMania. For each program individually, the genes which were differentially expressed between invasive and non-invasive states were parsed 72 into groups based on biological function (eg. Hypoxia, cancer stem cells, EMT-TFs, etc.). Connecting nodes that linked functional gene groups were identified and subject to literature review prior to inclusion in our EMT gene panel. The STRING (http://string-db.org/) database is comprised of known and predicted protein interactions. It covers over five-million proteins from 1133 organisms and works to build interaction networks based on physical and functional protein associations. Information on these interactions is derived from genomic context, high-throughput experiments, conserved coexpression and previous knowledge from the recent literature. Similarly, the GeneMANIA (http://www.genemania.org/) database includes curated information on protein and genetic interactions, pathways, co-expression, co-localization and protein domain similarity. It works to find genes that are related to a set of input genes, using mined functional association data. Networks are created based on assigned weights for each gene in the input list. Weights are calculated by linear regression to increase interaction between input genes, and reduce interactions between those genes falling outside the predicted Gene Ontology (GO)-based annotations, thereby ensuring a high degree of gene relatedness. Cell lines and culture conditions PC-3 cells were purchased from the American Type Culture Collection (ATCC) (#CRL1435) and were grown in F12-K media supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. Cultures were grown at 37ºC in 5% CO2. All cultures were determined to be mycoplasma-free using the Universal Mycoplasma Detection Kit from the ATCC (#301012K). Cell images were captured using an EVOS® FL Cell Imaging System (LifeTech, Burlington ON). 73 Stable lentivirus-mediated gene knock-down Transduction-ready MISSION short hairpin RNA (shRNA) lentiviral particles containing pLK0.1-puro vectors with U6 promoters were purchased from Sigma-Aldrich (Oakville, Ontario). Target Gene SNAI1 shRNA Sequence TRC Number CCGGTGCTCCACAAGCACCAAGAGTCTCGAGACTCTTGGTGCTTGTGGAGCATTTTTTG TRCN0000453110 ZEB1 CCGGCCTCTCTGAAAGAACACATTACTCGAGTAATGTGTTCTTTCAGAGAGGTTTTT TRCN0000017565 HIF1Α CCGGGTGATGAAAGAATTACCGAATCTCGAGATTCGGTAATTCTTTCATCACTTTTT TRCN0000003810 SPARC CCGGCGGTTGTTCTTTCCTCACATTCTCGAGAATGTGAGGAAAGAACAACCGTTTTT TRCN0000008709 VIM CCGGGCAGGATGAGATTCAGAATATCTCGAGATATTCTGAATCTCATCCTGCTTTTT TRCN0000029121 CLTC CCGGCGGTTGCTCTTGTTACGGATACTCGAGTATCCGTAACAAGAGCAACCGTTTTTG TRCN0000342755 Scrambled control CCGGCAACAAGATGAAGAGCACCAACTCGAGTTGGTGCTCTTCATCTTGTTGTTTTT n/a (Catalogue #SHC002V) Transduction Procedure Optimal multiplicity of infection (MOI) was determined using GFP-tagged pLK0.1 control vectors. MOIs of 0.5, 2 and 5 were tested in the PC-3 cell line. The success of transduction was assessed via puromycin selection and GFP expression. PC-3 cells were transduced using a viral MOI of 5. On the first day of the transduction procedure, 5x104 cells per well in a 24-well plate were seeded and allowed to proliferate for 24 hours in complete F12K media. Cultures were maintained at 70-80% confluency prior to proceeding with transduction. For cell transduction, viral particles were added to complete F12K media at an MOI of 5 (based on viral titre of each construct), to a total volume of 500µl. Hexadimethrine bromide was added to a final concentration of 8 μg/ml then mixed gently by inversion. Old culture medium was removed and replaced with 500ul of diluted viral particles. For one well (mock well control), 500µl of complete medium with 8 μg/ml hexadimethrine bromide were added to cells with no virus particles. Cultures were incubated at 37°C with 5% CO2 for 24 hours, and then viruscontaining media was replaced with fresh, virus-free complete media. Cultures were incubated for 48 hours in virus-free, F12K media, and split as needed. 74 Following the 48 hour incubation period, puromycin selection was completed to isolate those cells that had been successfully transduced. The puromycin concentration used for selection was 2μg/ml, in complete F12K media, which was replaced every 2-3 days until drug-resistant colonies began to propagate, and control cells were no longer viable. Migration and Invasion Assays Migration and invasion assays were completed as outlined in the materials and methods section of Chapter 3 (page 50). Embedded 3-Dimensional Cell Cultures Growth Factor Reduced (GFR), Phenol Red-free, Matrigel Basement Membrane Matrix (Corning, #356231), was left to thaw on ice at 4ºC overnight. All culture plates and pipet tips were pre-chilled at -20ºC prior to plating Matrigel. To generate embedded 3D cultures in 8 chamber optical slides, 20ul of Matrigel were added to each chamber carefully spread across the slide surface using a pipet tip, ensuring a uniform layer without the introduction of bubbles. Slides were then incubated at 37ºC for 15-30 minutes allowing for Matrigel polymerization. For each cell line, 5x103 cells were added into a 1:3 media:Matrigel mixture (300ul total volume – 225ul Matrigel, 75ul media/cells), being kept on ice throughout. Matrigel and cells were mixed thoroughly but carefully to evenly distribute cells without introducing bubbles. This mixture was then plated on top of the first Matrigel layer. The culture slides were returned to the incubator for 30-45 minutes to allow for polymerization. When Matrigel was completely set, 500ul of appropriate cell media were added to the cultures. Media was changed every 2-3 days. 75 Human prostate tumour samples Archived FFPE prostate cancer samples were collected from Kingston General Hospital in Kingston, Ontario. The retrospective study cohort consisted of 62 formalin-fixed paraffin embedded (FFPE) radical prostatectomy specimens gathered between 1994 and 2013. Each sample was classified as either Gleason score 3+3 (31 samples) or ≥ Gleason score 4+3 (31 samples). Review of the histological slides from each case was completed by a pathologist and appropriate tissue blocks were selected for further processing. Reference sections from each block were cut and mounted on glass slides then H&E stained. Two sections from each block were mounted on 1.0 PEN membrane slides (Zeiss, # 415190-9041-000) to facilitate dissection. A second pathology review was completed for each prostate cancer section in order to identify those regions rich in Gleason patterns 3 or 4 from the Gleason score 3+3 and ≥ 3+4 samples, respectively (from here on referred to simply as Gleason pattern 3 and Gleason pattern 4). Manual, macrodissection of the identified tumour-dense regions was completed for each sample. Laser-Capture Microdissection For nine samples where the tumour foci were too small to manually dissect, laser-capture microdissection, using a Zeiss PALM CombiSystem microscope, was employed. Auto-cut and LPC laser tools were used at the following settings: laser pressure catapult (LPC) energy: 68, LPC focus: 73, and cut energy: 41, cut focus: 78. Tissue areas were selected and dissected under 20X magnification. For each sample, one to two squared millimeters were dissected in order to ensure an appropriate RNA yield for downstream experimentation. Tissue was collected in microscope-compatible adhesive cap 0.2ml tubes (Zeiss, # 415190-9181-000). After dissection, all collected tissues were stored at -80ºC until ready for RNA extraction. 76 RNA extraction A. FFPE tissue All extractions were completed using components of Qiagen RNeasy® FFPE Kits (# 73504), combined with Roche PCR-grade Proteinase K, and following a modified protocol from Zeiss for working with small tissue quantities. Briefly, tissues were digested with 10ul Proteinase K and 150ul of buffer PKD at 56ºC for 4-5 hours, then heated at 80ºC for 15 minutes to inactive the reaction, followed by a three-minute incubation on ice. To remove DNA contamination, samples were subject to DNase digestion using 16μl DNase Booster Buffer and 10 μl DNase I stock solution, and left to incubate for 15 minutes at room temperature. The lysates were then transferred to 1.5ml tubes and mixed with 320ul of buffer RBC and 720ul of 100% ethanol. After thorough mixing of the samples via pipetting, 700ul of each sample were added to RNeasy MinElute spin columns and centrifuged for 15 seconds at >10000rpm. This step was repeated until the full sample volume had passed through the column. Following this, two 500ul washes with buffer RPE were passed through each column by centrifugation at >10000rpm for 15 seconds. The columns were then transferred to clean collection tubes and centrifuged with their caps open at 12000rpm for 5 minutes to dry any remaining ethanol in the column. Dried columns are placed in 1.5ml centrifuge tubes and 20ul of RNAse-free water are added to the column membrane and left to incubate at room temperature for 5 minutes. RNA was eluted via centrifugation at >10000rpm for 1 minute. RNA was quantified using Agilent BioAnalyzer PicoChips, following manufacturer’s protocol, and stored at -20°C until use. On average, RNA samples had RINs (RNA integrity numbers) of 2-2.5 and demonstrated that 50% of the sample contained fragments of at least 200 nucleotides. 77 B. Cell lines RNA was obtained using a Qiagen RNeasy extraction kit, following the manufacturer’s protocol without deviation. RNA was eluted in 50µl of RNase-free water and stored at -20ºC until use. Concentration was determined via NanoDrop spectrophotometry. Multiplexed Target Enrichment for nCounter Analysis When working with FFPE-extracted RNA, low quantity and poor quality of RNA samples makes multiplexed target enrichment (MTE) a requirement in order to be compatible with the NanoString nCounter platform. MTE Primers were designed by NanoString, using Primer3 with the following parameters: amplicon size 170-230nt, primerTm 60ºC, primer optimal size 18nt. The primers were designed such that the 100nt target region for the NanoString probes is fully contained in the amplicon. Primers were purchased from IDT (Coralville, Iowa) and were supplied in a single pool with final specifications of 0.25nmole/oligo at a final concentration of 0.5uM in 500uL of IDTE buffer pH 7.5 (see supplementary Table S4.1 for MTE primer sequences). The amplification reaction was completed following NanoString protocols for single cell analysis using total RNA. Briefly, cDNA was generated in a low-volume 5ul reaction composed of 4ul RNA (totaling 1ng) and 1ul of SuperScript VILO Master Mix (LifeTechnologies #11755050). Conversion to cDNA began with a 10-minute incubation at 25°C, followed by a 60 minute incubation at 42°C. The SuperScript enzyme was inactived with a 5-minute incubation at 85°C. Using the whole reaction volume from the cDNA conversion, samples were prepared for amplification by adding 1ul of pooled primers and 5ul of TaqMan® PreAmp Master Mix (LifeTechnologies #4391128). Amplification via PCR was carried out beginning with a 10 78 minute denaturation at 94ºC, followed by 10 cycles of 94°C for 15 seconds, then 60°C for 4 minutes. Amplified samples were stored at -20°C until NanoString probe hybridization. Sample Hybridization and nCounter Analysis Prior to probe hybridization, pre-amplified samples (cDNA) were denatured by incubation at 94°C for 2 minutes then snap-cooled on ice. When analyzing unamplified RNA from our knockdown cell lines, hybridization was completed without denaturation, as conversion to cDNA was not required. Probe hybridization reactions were set up following NanoString protocol using 5ul (100ng) total, unamplified RNA or 11ul of amplified cDNA (complete amplification reaction volume). Each RNA/cDNA sample was combined with 10ul hybridization buffer, 10ul reporter probes and finally 5ul capture probes (For capture and reporter probe sequences for each target gene, see Supplementary Table S4.2). Samples were mixed by gently flicking the tubes, followed by brief centrifugation. Vortex and rigorous pipetting were avoided so as to not sheer the probes. Samples were incubated at 56°C overnight (at least 16 hours, and not longer than 24 hours) to allow for probe hybridization to the RNA/cDNA. Following hybridization, samples were loaded into the nCounter cartridge (12 samples/cartridge) using the fully automated nCounter Prep Station. During this procedure, the hybrid probe/RNA molecules are bound to the cartridge surface by the capture probes (biotin/streptavidin interaction), and aligned across the cartridge surface via an electromagnetic gradient. For our purposes, all cartridges were prepared using the ‘high sensitivity’ setting. Once the cartridges were prepared, they were transferred to the digital analyzer which collects images of the immobilized reporter probes. Images are processed internally and exported in data files which contain both the target name and count number, along with outputs for positive and negative internal controls, which were used to guide quantification. 79 nCounter Data Analysis Data pre-processing and normalization were completed using the open-source R package, NanoStringNorm (Waggott 2012), under the most conservative settings as defined by the following calls: CodeCount=’geo.mean’ Background=’mean.2sd’ SampleContent=’housekeeping.geo.mean’ Following data normalization, 67 genes and 56 prostate cancer samples (27 Gleason pattern 3 and 29 Gleason pattern 4) passed all thresholds set for quality control and were included in our downstream analysis. Data pre-processing and normalization were completed by Dr. Robert Gooding from the Department of Physics at Queen’s University. Statistical approaches and multivariate modeling Statistical analysis and multivariate modeling were completed by Andrew Day of the Kingston General Hospital Clinical Research Centre, and Dr. Robert Gooding from the Department of Physics at Queen’s University. For detailed statistical methods in multivariate model generation, as outlined by Mr. Day, see Appendix B. In brief, using the pre-processed and normalized data, significantly differentially expressed genes between Gleason patterns 3 and 4 were determined using the non-parametric Mann-Whitney U test, set at p=0.05. Nineteen genes were identified, though no single gene intensity had the ability to significantly distinguish between Gleason patterns. Therefore, multivariate models were investigated for this purpose. Given the use of a relatively small cohort, compared to the number of genes being targeted in the analysis, standard methods for multivariate model identification were considered to be overly optimistic, such that the models were likely to be fitting noise within the data. Based 80 on the observation that a large number of our genes were found to be highly correlated with one another, it was possible to represent the data in three principal components (PCs). A logistic regression model using the first three principal components as variables was able to predict Gleason pattern with an ROC-AUC of 0.77 (p=0.002). Of the principal components, the third was found to be significantly correlated with Gleason pattern (p=0.0067), while the first was approaching statistical significance (p=0.0593). By identifying those genes whose intensities are most highly correlated with the third PC, a five-gene signature was identified that independently achieves an AUC or 0.7 (p=0.003). 81 4.4 Results 4.4.1 A data mining and systems biology-based approach to the development of a comprehensive EMT biomarker panel In order to comprehensively assess the molecular representation of EMT in our prostate cancer cohort, a biomarker panel of EMT-associated genes was identified to be interrogated during targeted gene expression analysis using the nCounter platform. As a starting point for the biomarker panel, genes associated with cell invasion in 3D Matrigel cultures were identified from four invasive prostate cancer cell lines (PC-3, PC-3m, ALVA31 and RWPE-2-w99). By cross-referencing gene expression profiles associated with invasive progression, 1669 genes were identified as being commonly dysregulated in at least three out of four cell lines (Figure 4.1). To isolate those genes from the list that are known to be functionally associated with EMT, the 1669 genes were cross-referenced with a panel of previously identified EMT-associated genes from a qRT-PCR array developed by SABiosciences. This panel contained 84 genes which spanned a number of signaling pathways, and included many of the known genes which function during EMT in the context of embryogenesis and wound healing. However, this panel does not comprehensively represent EMT in the context of metastatic progression in cancer and therefore was not independently sufficient for our purposes. Given the high degree of commonality in the signaling networks which regulate EMT in embryogenesis and wound healing, and those that regulate EMT in cancer progression, it was a feasible to use this panel to isolate those genes from our invasive prostate cancer cell lines which are also well-established EMT-associated genes. Comparison of these two gene lists resulted in the identification of 33 genes (Figure 4.2). The 33 genes identified formed the ‘seed list’ from which the rest of the EMT panel was built (see Supplementary table S4.1). 82 Figure 4.1 Overlapping gene expressions from invasive prostate cancer cell lines. Gene expression profiles were identified from four invasive prostate cancer cell lines after having been grown in 3D Matrigel cultures. Differential gene expression was determined by comparing post-invasion gene expression to the gene expression profiles of cells grown in 2D culture. Crossreferencing the gene expression profiles associated with invasion in 3D culture identified 1669 genes that were commonly dysregulated in at least three out of 4 cell lines (white dotted line). 83 Figure 4.2 Overlap between genes dysregulated during prostate cancer cell invasion and genes associated with EMT Venn analysis was completed to compare genes which function during EMT in embryogenesis and wound healing, to those that are commonly dysregulated in prostate cancer cell line invasion in 3D culture. 33 genes were identified through this analysis. 84 Expansion of our initial list of EMT/ invasion genes was completed using a combination of network building algorithms and literature mining. Primary transcription factors known to regulate EMT that were not identified through previous analysis (Snail, Slug, ZEB1, TWIST1, FOXC2) were automatically included and were used in conjunction with the initial panel of 33 genes to build networks for further gene identification. Genes from the primary list were grouped based on biological function, or signaling pathway into the following categories: TGFβ signaling, MAPK signaling, hypoxia response, apoptosis, morphogenesis, stress response, cell junction organization and cell migration/invasion. These small groups of genes were input into the network building algorithms, STRING and GeneMANIA, and nodal points that were shown to connect the networks with the highest confidence (based on evidence classifications and FDRs) were noted. Genes identified through network building were subject to literature review prior to inclusion in the final gene panel. Those genes with known functions in EMT were included, as were genes associated with EMT-related phenotypes such as migration and invasion, and genes linked to cancer stem cell behaviour, which has been functionally linked to EMT. In total, 101 genes (plus six housekeeper genes) were identified for our EMT biomarker panel, to be used for gene expression profiling in our prostate cancer cohorts (Supplementary table S4.1). 4.4.2 Genes associated with EMT signaling are differentially expressed between Gleason pattern 3 and Gleason pattern 4 prostate cancers To determine the utility of EMT-associated genes in differentiating between Gleason pattern 3 and 4 prostate cancer, gene expression analysis using the nCounter platform from NanoString was completed in a cohort of 62 FFPE radical prostatectomy samples (31 Gleason score 3+3 and 31 Gleason score ≥4+3). Following data pre-processing and normalization which reduced the sample cohort to 27 Gleason pattern 3 and 29 Gleason pattern 4 prostate cancers, 19 genes were determined to be significantly differentially expressed between the groups, using a Mann-Whitney U test at p=0.05 (Figure 4.3). These 19 genes were selected solely on statistical 85 significance such that no fold change cut-off was applied to reduce the number of genes identified for univariate analysis. Data are represented as average fold change (Gleason pattern 4 vs. pattern 3) ± the standard error. 86 Figure 4.3 Differential EMT-associated gene expression in Gleason pattern 4 versus pattern 3 prostate cancer Following data normalization using the NanoStringNorm package for R, nineteen genes which are functionally associated with EMT were found to be significantly upregulated in Gleason pattern 4 prostate cancer (n=29), relative to Gleason pattern 3 prostate cancer (n=27). Statistical significance was determined using a Mann-Whitney U test at p=0.05. Data are displayed as average fold change (G4 vs G3), ± standard error. 87 4.4.3 Principal component analysis combined with logistic regression modeling identifies a multivariate model capable of differentiating between Gleason pattern 4 and pattern 3 prostate cancers Although 19 genes were identified as having significantly different expression between Gleason pattern 3 and 4 prostate cancers, univariate analysis revealed that no single gene intensity was capable of distinguishing between the groups. Therefore, the potential for multivariate model development was examined. A logistic regression model using the first three principal components (PCs) of the data as variables was able to significantly predict Gleason pattern with an area under the curve (AUC) of 0.77 (p=0.002). Figure 4.4 shows the receiver operating characteristic (ROC) curve for the model built using the first three PCs. Assessing each of the principal components individually for their association with Gleason pattern demonstrated that only principal component 3 was significantly predictive of Gleason pattern (p=0.0067), while principal component 1 approached significance (p=0.0593). Figure 4.5 demonstrates the predictive power of the first and third principal components in distinguishing Gleason pattern 4 from pattern 3 prostate cancers. A five-gene model was selected based on the identification of the highest correlations between individual gene intensities and the third principal component (Table 4.1) This gene signature has correlation with the third PC of 0.87. When used to distinguish Gleason pattern 4 from Gleason pattern 3, this model achieves an AUC of 0.70 (p=0.003). 88 Figure 4.4 ROC-AUC analysis of the first three principal components The ROC curve built for the multivariate model developed from logistic regression of the first 3 principal components, representing the differential gene expression between Gleason patterns, indicates an AUC of 0.77 (p=0.002). 89 Figure 4.5 Fit plot demonstrating the predictive power of the first and third principal components This plot graphically represents the predictive power of the first and third principal components in predicting Gleason pattern 4 (red) over Gleason pattern 3 (blue). It demonstrates that samples with higher principal component three values are more likely to be Gleason pattern 4. 90 Table 4.1 Genes demonstrating the highest degree of correlation between PC three and signal intensity Gene Symbol Correlation between PC 3 and log-2 intensities MST1R 0.5131 TWIST2 -0.5030 AHNAK 0.4648 FN1 0.4428 POU5F1 -0.4059 The intensities of these genes have the highest correlation with principal component 3, which is most significantly associated with Gleason pattern. As such, this five-gene signature represents a model capable of distinguishing between Gleason pattern 4 and Gleason pattern 4 prostate cancers. 91 4.4.4 Functional validation of EMT-associated genes Genes of interest were selected from the target gene panel based upon the results from the gene expression data in our prostate tumour cohort. To functionally validate these in the context of EMT-specific phenotypes, stable knockdown cell lines were generated for six candidates: Snail, ZEB1, HIF1α, CLTC, VIM and SPARC. Each gene was stably knocked down in the invasive PC-3 prostate cancer cell line which demonstrates endogenously high levels of EMTassociated genes. Confirmation of gene knockdown (KD) was completed via NanoString gene expression analysis (Figure 4.6) Impact of each KD on cell morphology, migration, invasion and growth in 3D culture was assessed to confirm the functional contribution of these genes to EMT, and subsequently their potential involvement in facilitating metastatic progression through this cellular program. 92 Figure 4.6 Validation of gene knockdown Gene knockdown was assessed using the NanoString nCounter platform and the EMT gene panel designed for this study. Expression of HIF1α was reduced by 73%, CLTC by 68%, Snail by 15%, SPARC by 70%, Vimentin by 82% and ZEB1 by 46%. 93 4.1.1.1 Inhibited expression of Snail, ZEB1 and HIF1α, but not SPARC, VIM or CLTC, results in loss of mesenchymal morphological characteristics Examination of alterations in cell morphology, post gene knockdown, revealed that reduced expression of Snail, ZEB1 and HIF1α all result in loss of the characteristic ‘mesenchymal’ morphology of the untransfected, and scrambled sequence control (SCRM), PC-3 cells. These knockdown cell lines demonstrate typical epithelioid morphology, such that the cells have lost their elongated, spindle-like shape and instead exhibit a rounded morphology. Additionally, Snail, ZEB1 and HIF1α KD cell lines exhibit higher cell-cell adhesion and grow in cohesive clusters, compared to control cell lines, which have reduced intercellular contacts and grow in dyscohesive sheets (Figure 4.7A). In contrast, SPARC, VIM and CLTC knockdowns appear to have no effect on 2-dimensional cell morphology, as these cells maintain the same mesenchymal morphological characteristics as the control cell lines (Figure 4.7B). 94 Figure 4.7 Effects of EMT gene knockdown on cell morphology in 2D culture Cells visualized under 20X magnification. Reduced expression of Snail, ZEB1and HIF1α (A) result in cell morphological conversion from spindle-like mesenchymal characteristics to the rounded morphology associated with epithelial cells. An increase in intercellular contacts is also seen, resulting in the appearance of cohesive colony growth. In contrast, knockdowns of SPARC, CLTC and VIM (B) show minimal impact on the morphological characteristics of the cells, when compared to the untransduced and SCRM control lines. 95 4.1.1.2 Assessment of EMT gene knockdown on cell migration One of the primary phenotypic changes associated with EMT is the increased propensity for cell migration. To evaluate the impact of our genes of interest on cell migration, Boyden chamber assays were employed. Untransduced and SCRM control PC-3 cells show 286.2 ± 61.2 cells and 318.6 ± 47.1 cells, respectively, migrating through the Boyden chamber membrane. As expected, Snail, ZEB1 and HIF1α knockdowns all resulted in significantly reduced cell migration, when compared to control cell lines, with 209 ± 20.1 cells, 162 ± 52.8 cells and 185.8 ± 52.9 cells migrating through the membrane, respectively (all p-values <0.01) (Figure 4.8). While knockdowns of CLTC and VIM showed no significant difference in rates of migration compared to controls (261.8 ± 36.5 cells [p-value=0.08] and 259.2 ± 66 cells [p-value=0.14], respectively), SPARC knockdown resulted in a significant reduction in migration (178.2 ±81.2 [p-value=0.01]) (Figure 4.8). 96 Figure 4.8 Impact of gene knockdown on cell migration Migration rates were assessed using Boyden chamber assays. All data are represented as average ± standard deviation, across replicates. Compared to control cell lines (PC-3: 286.2 ± 61.2 cells and SCRM control: 318.6 ± 47.1 cells), Snail, ZEB1, HIF1α and SPARC knockdowns all demonstrate significantly reduced rates of migration, with 209 ± 20.1 cells, 162 ± 52.8 cells, 185.8 ± 52.9 cells and 178.2 ±81.2 cells, respectively, migrating through the membrane (all pvalues ≤0.01). In contrast, CLTC and VIM knockdowns have no statistically significant impact on migration, resulting in 261.8 ± 36.5 migratory cells (p-value=0.08) and 259.2 ± 66 migratory cells (p-value=0.14), respectively. 97 4.1.1.3 Assessment of EMT gene knockdown on invasive characteristics in 2D and 3D cultures Augmented invasive behaviour is another phenotypic hallmark associated with EMT, and perhaps the most clinically relevant in the context of cancer progression. Modified Boyden chamber assays were used to assess the contribution of our genes of interests to invasive behaviour in prostate cancer cells. Untransduced and SCRM control cells demonstrated 284±26.3 cells and 313±23.2 cells, respectively, invading through the Matrigel coated membrane. In comparison, all knockdown cell lines demonstrated reduced levels of cell invasion. As expected, Snail, ZEB1 and HIF1α, all well established for their function in cell invasion, demonstrated 201.5 ± 24.2 cells (p-value <0.01), 156 ± 38.3 cells (p-value=0.014) and 191 ± 25.9 cells (p-value < 0.01) invading through the Matrigel layer, respectively. Vimentin, Clathrin and SPARC knockdowns also demonstrated reduced invasive potential at rates of 63 ± 14 cells (p-value < 0.01), 137 ± 18.7 cells (p-value <0.01) and 14.7 ± 7.8 cells (p-value <0.01) invading through the matrix, respectively (Figure 4.9). 98 Figure 4.9 Impact of gene knockdown on cell invasion Assessment of cell invasive potential was made using modified Boyden chamber assays. Untransduced and SCRM control cells demonstrated 284±26.3 cells and 313±23.2 cells, respectively, invading through the Matrigel coated membrane. Snail, ZEB1 and HIF1α knockdowns demonstrated 201.5 ± 24.2 cells (p-value < 0.01), 156 ± 38.3 cells (p-value < 0.01) and 191 ± 25.9 cells (p-value < 0.01) invading through the Matrigel layer, respectively. Finally, Vimentin, Clathrin and SPARC knockdowns also demonstrated reduced invasive potential at rates of 63 ± 14 cells (p-value < 0.01), 137 ± 18.7 cells (p-value <0.01) and 14.7 ± 7.8 cells (pvalue < 0.01) invading through the matrix, respectively. All data are represented as averages ± the standard deviation, across triplicates. 99 Given that cells grown in 2-dimensional culture are not representative of the true biological context in which tumours form and progress to invasion, it is useful explore the impact of our genes of interest on cell behaviour in the context of growth and invasion in a 3D extracellular environment. In embedded Matrigel cultures, PC-3 untransfected and SCRM control cell lines recapitulate the expected progression to invasion (Windus 2012, Harma 2010), showing development of invadopodia, from the cell colonies within 7 days, and progressing towards extensive invasion after 12 days in culture (Figure 4.10A and B). Knockdowns of Snail and ZEB1 demonstrate morphology in 3D culture which is consistent with results seen in non-invasive prostate cancer cell lines such as DU145 (Harma 2010, Windus 2012), such that over the course of 12 days in Matrigel, cells form spheroid colonies with no evidence of invadopodia formation (Figure 4.10C and D). Knockdown of HIF1α follows a similar trend, forming non-invasive colonies. However, morphologically, the colonies appear more lobular then those of the Snail and ZEB1 knockdown cell lines (Figure 4.10E). Cells with reduced expression of Vimentin (Figure 4.10G) show no indication of invasive potential after 12 days in culture. However, compared to other non-invasive cell lines (Snail and ZEB1 knockdowns) Vimentin knockdown cell colonies were slower to form. CLTC and SPARC knockdown cells display delayed onset to invasion when compared to the PC-3 control cell lines (Figure 4.10F and H). These cells exhibit significant colony formation prior to the development of invasive projections. However, clear evidence of invasive potential is indicated within the first week in culture, though not to the extent of control cells. 100 Figure 4.10 (Continued on page 102) 101 Figure 4.10 Impact of gene knockdown on colony formation and invasion in 3D culture Cultures visualized under 20X magnification. PC-3 untransduced and SCRM control lines (A, B) demonstrate typical progression to invasion of the PC-3 cell line, with well-established invasive processes by the end of the first week, and evidence of extensive invasion after 12 days in culture. PC-3 cells with Snail, ZEB1 and HIF1α knocked down show no progression towards invasion after 12 days in culture, although colony formation ability remains intact (C-E). HIF1α knockdown colonies (E) differ from the Snail and ZEB1 knockdowns in that they appear to have Vimentin knockdown cells show similar results, with no evidence on invasion, or formation of invasive structures, after 12 days in culture, though colony formation progresses at a slower rate than in the Snail, ZEB1 and HIF1α cultures (G). CLTC and SPARC knockdown cell lines demonstrate delayed cell invasion, compared to controls, forming sizeable colonies prior to the onset of invasive behaviour. After 12 days in culture, both cell lines have established clear invasive structures (F and H). 102 4.4.5 Evaluation of frequently dysregulated gene expression in response to EMT-associated gene knockdown Given its vast impact on many areas of organism development and disease progression, EMT has become an increasingly studied area of biomedical research. Despite the growing interest in this process, there is still a lot unknown about the signaling networks that regulate the phenotypic shifts associated with EMT. In an attempt to identify key nodal points in EMT regulation, RNA was collected from each of the generated KD cell lines, and subject to gene expression analysis using the EMT panel designed for the NanoString nCounter system. Gene expression changes in response to individual gene knockdown (SNAI1, ZEB1, HIF1Α, SPARC, CLTC and VIM) are listed in supplementary table S4.4. The assumption was made that the key nodal points in EMT regulation would be those commonly dysregulated in response to multiple gene knockdowns. Of the 101 EMT-associated genes assessed in each cell line, five were found to be dysregulated by two-fold or higher in response to reduced expression of three or more of our six genes of interest (Figure 4.11). Expression of fibronectin (FN1) was increased as a result of gene knockdown in all cell lines except PC3-VIM. E-cadherin (CDH1) expression increased in response to HIF1α, Snail and ZEB1 KD, but was reduced in response to VIM and CLTC knockdown. SPARC expression increased in each knockdown cell line, except PC3-SPARC and PC3-VIM, while NUPR1 expression went up in HIF1α, ZEB1, SPARC and CLTC knockdown cells. IL6 expression also increased in response to knockdown of HIF1α, Snail and ZEB1, but was not affected by reduced expression of SPARC, VIM or CLTC. 103 Figure 4.11 Identification of key nodal points in EMT regulation In an attempt to identify nodal genes in EMT regulation, gene expression analysis was completed for each of the generated knockdown cell lines. Those genes that were altered by greater than 2fold in at least three out of the six knockdown cell lines were identified as being those that may represent EMT signaling nodal points. 104 4.5 Discussion 4.5.1 The utility of EMT-associated gene expression in differentiating between Gleason pattern 3 and Gleason pattern 4 prostate cancers The role of the epithelial-mesenchymal transition has become increasingly noteworthy in the context of cancer progression, and as such, suggests the utility of genes associated with EMT as predictive and prognostic biomarkers in many different cancer types. While EMT is currently profiled in cells and tissues using the expression of a limited number of key transcriptional regulators such as Snail, Slug, ZEB1/2 and TWIST1/2, and a small number of epithelial or mesenchymal markers (E-cadherin, Vimentin), there is a pressing need to identify a comprehensive list of biomarkers that span the many upstream signaling networks and the downstream transcriptional targets that regulate the process as a whole. EMT in the context of cancer progression is induced as a result of aberrant gene expression, and can therefore be associated with a plethora of molecular cues, including growth factors and cytokines, often in the form of signals from the extracellular environment. Since the specific initiation factors of EMT in varying contexts have yet to be delineated, combined with the increasingly complex network of interconnected gene regulation both up- and down-stream of the primary EMT-TFs, there is inherent value in approaching the characterization of this process in a model system using a more inclusive set of biomarkers. In doing so, insight can be provided into the complex signaling networks functioning around the primary transcription factors which regulate this process, and may help delineate the specific biological factors contributing to cancer progression, making them targetable for therapeutic intervention. In the context of prostate cancer, associations between EMT and aggressive behaviour are well-established in the literature. Studies investigating the role of Snail and other EMT transcription factors regulating this process in vitro have reproducibly demonstrated that there is a causal link between EMT and prostate cancer cell migration and invasion (Thakur 2014, Cho 105 2014, Wang 2013). However, few have attempted to assess the utility of EMT-associated biomarkers in a clinical setting. One study, looking at the expression levels of 17 EMT-associated proteins across locally confined prostate cancers, demonstrated a significant association between biochemical recurrence after RP and expression of TWIST and Vimentin (Behnsawy 2013). Ecadherin, Snail, TWIST and Vimentin expression were also significantly correlated with Gleason score, such that E-cadherin expression is lost, while Snail, TWIST and Vimentin are overexpressed in Gleason score 8 and above, relative to Gleason score 7 and below tumours (Behnsawy 2013, Poblete 2014). The association between Vimentin expression and poor prognosis had been identified previously (Zhang 2009), and appears to be associated with TGFβ expression. Independently, TGFβ1 upregulation is significantly associated with Gleason scores ≥7 and poor overall prognosis in prostate cancer (Reis 2011, Wikström 1998), while high ZEB1 expression correlates directly with increasing Gleason score (Graham 2008). High Gleason score has also been correlated with increased expression of ECM component, SPARC, which regulates cell migration and invasion, again suggesting a link between EMT and high-risk prostate cancer (Derosa CA 2012). Reassuringly, our data are largely consistent with the findings of others, demonstrating significantly increased expression of key EMT genes Gleason pattern 4 relative to pattern 3 tumours. This suggests that the increased risk of metastatic progression associated with Gleason pattern 4 disease, may be attributable to the onset of EMT. Despite the fact that these correlations have been identified, no study has yet assessed the utility of these genes and/or proteins as prognostic biomarkers with the purpose of addressing undersampling in prostate cancer biopsies. Our study has resulted in the successful identification of EMT-associated genes that are differentially expressed between Gleason pattern 3 and 4 prostate cancers, as well as a multivariate model that can significantly predict Gleason pattern 4 with an AUC of 0.70. Therefore, we have been able to identify gene expression signatures that may partially define the pre-metastatic state in advancing prostate cancer. If these findings can be 106 reproduced in subsequent validation cohorts, and are found to be capable of distinguishing between Gleason pattern 3 cancers from Gleason score 6 vs Gleason score 7 tumours, then these EMT-associated models may have clinical value in risk stratification and facilitating treatment decisions. Specifically, their utility would be in delineating false ‘low-risk’ biopsies from true ‘low-risk’ biopsies, making decisions about undergoing active surveillance more confident, and thus reducing prostate cancer overtreatment. The five-gene multivariate signature identified in our analysis suggests that differential expression of a subset of EMT-related genes can successfully distinguish between Gleason pattern 3 and 4 prostate cancers with moderate accuracy. Of the five genes identified in the model, TWIST2 is one of the core EMT-TFs and FN1 is a well-characterized ECM protein associated with mesenchymal cells (Amatangelo MD 2012), both of which are correlated with metastatic progression in cancer (Chen 2014, Wang 2014). POU5F1, also known as OCT4, regulates stem cell pluripotency and is expressed by prostate cancer cells with tumour initiating potential (Patrawala L 2006). AHNAK is a desmosome-associated protein that functions in pseudopod formation and cell migration (Shankar J 2010), and MST1R (macrophage stimulating 1 receptor) is a member of the MET receptor tyrosine kinase family and is associated with poor prognosis in gastroesophageal and breast cancers (Catenacci DV 2011, Welm AL 2007). While each of these genes has been linked to aggressive cancer behaviour individually, the evidence connecting them to prostate cancer is minimal at best. Therefore, should this model validate successfully in an independent cohort, it may represent a novel prognostic tool for risk stratification in prostate cancer management. 107 4.5.2 Assessment of gene function in EMT-associated phenotypes While the presence of EMT signaling in cancer cells has been consistently associated with aggressive tumour behaviour, the molecular underpinnings of this process are still largely uncharacterized. Therefore, functionally analyzing the roles of individual EMT-associated genes is an important step towards mapping out the critical regulatory networks guiding this program. Ultimately this will provide a better understanding of EMTs contribution to cancer progression, and may present opportunities for targeted therapeutic intervention. Of the genes for which knockdown cell lines were created, HIF1α, Snail and ZEB1 served as built in models of EMT alteration, as they are already established for their regulatory function in this process. It is well-documented that HIF1α functions in EMT initiation, and leads to downstream activation of migration and invasion programs, which are tightly regulated by EMT-TFs, Snail and ZEB1 (Zhang 2013, Hwang-Verslues 2013). As a result, it is not surprising that inhibition of these three molecules in vitro results in significantly reduced rates of migration and invasion, both in standard 2D culture and in 3D culture. Even in the case of the Snail knockdown where only a 15% reduction in gene expression was achieved, significant impacts on migration and invasion were observed. Given the tightly controlled nature of Snail regulation during EMT, and the fact that it is the first EMT-TF to be upregulated resulting in transcriptional activation of other downstream EMT regulators (Zheng 2014), it is possible that even slight disturbances in its regulation may result in dramatic phenotypic shifts. One unexpected observation in the HIF1α knockdown cell line was the difference in colony morphology in 3D culture, compared to Snail and ZEB1 knockdowns. The lobular, rather than uniform, nature of the colony surface might suggest an inability of HIF1α-deficient cells to spatially organize themselves. While there is no current evidence linking inhibited hypoxia signaling with cell organization or spatial distribution in cancer, it could be the result of dysregulated, EMT-independent, cell-cell contact and communication mechanisms. Alternatively, a cellular inability to regulate oxygen responses may result in the cells attempting to increase 108 their environmentally-exposed surface area, though no evidence exists to support this theory as yet. Ultimately, the novelty of these functional studies comes from examining the roles of SPARC, VIM and CLTC in EMT-associated phenotypes, which remain poorly characterized. 4.1.1.4 SPARC SPARC, or osteonectin, is a non-structural ECM glycoprotein that demonstrates collagen and calcium-binding capabilities and functions in the mediation of cell-matrix interactions (Brekken 2001). Currently, the role of SPARC in cancer behaviour is controversial due to variations in observed function between and within cancer types. For example, high SPARC expression in gastric cancer is significantly correlated with metastatic progression and poor prognosis (Zhao 2010), but in ovarian cancer, low SPARC expression is considered a poor prognostic indicator (Yiu 2001). In prostate cancer, SPARC expression has been correlated with aggressive behaviour and poor prognosis (Darosa 2012, Thomas 2000). However, others suggest that despite its prognostic correlation, SPARC is unlikely to functionally contribute to metastasis (Wong 2008). Our experimental results indicate that reduced expression of SPARC in the highly invasive PC-3 cell line results in significantly inhibited cell migration and invasion. While this is consistent with in vitro SPARC knockdowns in models of cervical cancer and melanoma (Chen 2012, Ledda 1997), there is little supporting in vitro evidence in prostate cancer cell lines to confirm our findings. In fact, contrasting evidence from Wong, S. et al. (2008) indicates that SPARC is not functionally associated with prostate cancer metastasis in murine models. While we consistently observed inhibited cell invasion, in both 2D and 3D cell cultures, it is possible that environmental factors, which were not accounted for in our model system, may significantly contribute to SPARC function in vivo. In fact, variations in the proteinase profile of the tumour microenvironment 109 appear to have an impact on SPARC function due to differential proteolytic processing (IruelaArispe 1995, Koblinski 2005). Functionally, overexpression of SPARC in prostate cancers with a higher risk for metastatic progression is reasonable, given that SPARC is actively produced by osteoblasts during bone formation (Ribeiro 2014), and that the bones are the preferential site for prostate cancer metastasis (Bubendorf 2000). In fact, bone-derived SPARC has been demonstrated to act as an attractant for SPARC-expressing prostate cancer cells (Jacob 1999), suggesting that the molecular environment of the bone is selective for those disseminated cells with high SPARC expression. Therefore, our PC-3 cell line model, which was derived from a prostate cancer metastatic bone lesion (Kaighn 1979) and which expresses high endogenous levels of SPARC, may be selectively dependent upon SPARC function for migration and invasion. This could explain our observation that SPARC loss reduces aggressive prostate cancer cell behaviour. Therefore, SPARC expression may not only correlate with an increased risk of cancer progression, it may prime disseminated cancer cells for bone colonization. 4.1.1.5 Vimentin (VIM) Vimentin is a member of the intermediate filament family of proteins and a wellestablished structural marker associated with mesenchymal cells. It has been demonstrated to be significantly overexpressed in many cancer types and is associated with tumour invasion and overall poor disease prognosis. In prostate cancer specifically, elevated Vimentin expression is prognostic of early BCR, and is also significantly associated with high Gleason score (>8) (Behnsawy 2013). Recently, studies have focused on understanding the role that it plays in regulating EMT signal transduction. Vimentin is capable of upregulating the expression of promigratory kinase, Axl which, when activated by its ligand Gas6, can contribute to breast cancer cell migration (Lee 2013, Vuoriluoto 2011). In fact, induction of EMT via Slug activation has been shown to upregulate Axl in a Vimentin-dependent manner (Lee 2013, Vuoriluoto 2011). 110 Additionally, Vimentin is known to be associated with stabilization and localization of Scribbled, a member of the scribbled polarity complex, which partially regulates front-rear polarity in migrating cells (Phua 2009, Navarro 2005). Given the seemingly well documented link between Vimentin expression and cell migration, it is interesting to note that our study did not demonstrate a significant reduction in cell migration in response to Vimentin KD, when assessing the phenotype using Boyden chamber assays. This could potentially be the result of differences in experimental parameters. Boyden chambers measure single cell migration in response to a chemical stimulus, in this case, FBS. Given the simplicity of the model, it is possible that chemotaxis induces signals that are able to bypass loss of Vimentin, and compensate by upregulating redundant structural components and signaling mechanisms. For example, our gene expression data from the knockdown cell lines have shown that loss of Vimentin results in the upregulation of FBN1, a gene which is involved in the migratory capability of corneal fibroblasts, perhaps by acting as a cofactor with ADAMTS-1 (Ducros 2007). We have also shown that Vimentin knockdown results in reduced prostate cancer cell invasion, using both Boyden chamber assays and 3D culture models. Previous research by Zhao Y et al. (2008) also demonstrated that knockdown of Vimentin in prostate cancer cell lines significantly reduced invasive potential, and that high Vimentin expression was significantly correlated with metastatic spread in prostate cancer. While a number of correlative studies have confirmed this association, little is known about how Vimentin contributes to cancer cell invasion. Wei J et al. (2008) demonstrated that Vimentin promotes tumour invasion by regulating β-catenin/E-cadherin complexes, through activation of c-Src. From a functional perspective, it seems that Vimentin intermediate filaments are necessary for the elongation of invadopodia, such that Vimentin loss results in truncated extension of these projections, thereby inhibiting cell invasion (Schoumacher 2010). 111 Ultimately, our results indicate a potential decoupling of cell migration and invasion in EMT regulation, and suggest that these phenotypes are not moderated by the same cellular components. Certainly, others have indicated that changes in cell migration do not effect EMT progression (Schaeffer 2014). Additionally, hallmark mesenchymal cell morphology is not demonstrated by these Vimentin knockdown cells, again suggesting that the hallmark los s of intercellular contacts and polarity may not be regulated by the same signaling networks which contribute to phenotypic shifts towards invasion. Therefore, it may be the case that Vimentin has no bearing on cell junction assembly and is not strictly necessary for cell movement due to functional redundancy with other filaments, but may have a critical role in regulating progression to invasive behaviour, perhaps through modulation of matrix-degrading enzymes such at MT1MMP (Dave 2014). 4.1.1.6 Clathrin, heavy chain (CLTC) Clathrin heavy chain is a structural protein which associates with Clathrin light chains, to form the triskelion structures responsible for the formation of Clathrin-coated vesicles which function in membrane trafficking and endocytosis (Ma 2013). While the difference in expression of CLTC between Gleason pattern 3 and 4 prostate cancers in our cohort only trended toward significance, biologically it was an interesting target due to its proposed, but largely uncharacterized, role in EMT. Some studies have suggested that Clathrin-mediated endocytosis contributes to E-cadherin sequestration, which may help facilitate EMT (Kon S 2008, Janda E 2006). While research in this area remains limited, there have been indications that loss of Clathrin results in a reduction in cell migration, either through membrane accumulation of Ecadherin which stabilizes adherens junctions, or by inhibiting integrin trafficking, preventing disassembly of focal adhesions (Janda 2006, Ezratty 2009). However, our data indicate that when CLTC is repressed, cell migration is not significantly impacted. This could be due to the function of alternative endocytotic mechanisms. For example, caveolin-mediated endocytosis has also 112 been shown to regulate membrane-associated expression of E-cadherin, as well as facilitate the endocytosis of occludens, leading to dismantling of tight junctions (Izumi 2004, Marchiando 2010). In this way, it may be possible for cells to disrupt intercellular contacts, thereby enabling cell migration, in a Clathrin-independent manner. This conclusion is consistent with our observation that Clathrin knockdown does not affect 2D cell morphology, suggesting that the EMT-associated characteristics of cell-junction dismantling and increased migration may represent separate distinct phenotypes within EMT that are regulated in response to alternative signaling. Conversely, our observation that CLTC knockdown significantly reduces prostate cancer cell invasion is consistent with its role in Clathrin-mediated intracellular trafficking of MT1MMP (Remacle 2003), a membrane-tethered matrix metalloproteinase that is commonly associated with invadopodia (Watanabe 2013). In pancreatic cell lines, CLTC knockdown leads to reduced invasion (Tung 2013), thereby supporting our observations, which provide the first direct correlation between loss of CLTC and inhibition of invasion in prostate cancer cells. Taken together, our results indicate that EMT-initiating factors (HIF1Α) and EMT-TFs (SNAI1 and ZEB1) contribute more globally to EMT signaling and thereby regulate the program in a more general, supervisory manner. In contrast, VIM and CLTC appear to function exclusively in cell invasion, at least within the context of our model system, and therefore may only feedback on direct functionally related signaling molecules. SPARC appears to function in both cell migration and invasion, however this could be an artifact of the cell line model we worked with, considering its potential for being primed for SPARC dependence, given its biological origin. To confirm the effects of these genes in vitro, further work must be completed. First, additional shRNA constructs should be utilized to generate knockdowns in the PC-3 cell line. Doing so will address concerns about any off-target effects that may have be documented using 113 our primary shRNA constructs. Additionally, making use of alternative prostate cancer cell line models would strengthen the association between the genes that were studied and their phenotypic associations with EMT. This would allow for a more general assessment of the molecular regulation of EMT in prostate cancer, rather than limiting the study to model-specific observations. Ultimately, it appears as though the classically defined EMT-associated phenotypes (mesenchymal morphology, migration and invasion) which are typically thought of as being regulated together as part of an all-encompassing signaling network, may actually be individually defined at the regulatory level and therefore dependent on different signaling molecules, and ultimately independent of one another. This would indicate that EMT could be dissected into separate molecular and phenotypic programs. 4.5.3 Evaluating the relationships between EMT-associated genes and their impact on the signaling networks regulating EMT Given the apparent association between EMT and prostate cancer progression, developing a more thorough understanding of the underlying signaling networks regulating this process may facilitate its utility as a biomarker, or therapeutic target in cancer management. To better understand these signaling networks, EMT gene expression was profiled in each of our knockdown cell lines. The results of these expression studies identified five genes (SPARC, FN1, CDH1, NUPR1 and IL6) that were dysregulated by more than 2-fold in response to the repression of three or more of our target genes of interest. Given the degree of overlap between signaling alterations in response to various gene knockdowns, it is possible that these genes represent integral nodal points in EMT regulation. Our data indicate that E-cadherin expression is altered in 5 of the 6 knockdown cell lines, demonstrating high levels of expression in response to HIF1α, Snail and ZEB1 knockdowns, as would be expected given the established role for all three genes in E-cadherin repression. Interestingly, Vimentin and Clathrin knockdowns result in reduced E114 cadherin expression. This is consistent with our observations that Vimentin and Clathrin knockdowns fail to revert 2D cell morphology from ‘mesenchymal’ to ‘epithelial’, such that Vimentin and Clathrin knockdowns appear to inhibit the re-establishment of inter-cellular contacts. However, since each of these knockdown cell lines demonstrates inhibited cell invasion without restoration of adherens junctions, this may provide evidence that EMT-regulated cell invasion is independent of E-cadherin status. While there is currently no evidence in the literature supporting repression of E-cadherin in response to Vimentin downregulation, it is possible that loss of Vimentin feeds back to increase EMT-TF activity with the intention of re-expressing VIM and consequently results in the indirect repression E-cadherin. CLTC knockdown is also found to reduce E-cadherin expression. Again, this association has not been previously addressed in the literature, but may be the result of impaired Clathrin-mediated, E-cadherin trafficking. If Clathrin loss prevents endosomal trafficking and turnover of E-cadherin, its transcriptional activation may be repressed as a result. Mesenchymally associated ECM components SPARC and FN1 appear to be consistently upregulated in response to target gene knockdown, despite expectations to the contrary. Given that our functional data demonstrate that SPARC knockdown results in inhibited cell migration and invasion, theoretically one would expect inhibition of EMT through repressed HIF1α, Snail or ZEB1 to result in SPARC repression. Given that this is not the case, it is possible to theorize that SPARC-mediated inhibition of cell migration/invasion occurs independently of EMTassociated signaling. This theory is further supported by the observation that SPARC knockdown has minimal effect on the expression of genes within our EMT panel, suggesting that it may function in response to signals outside the scope of our targeted gene list. Despite the expectation that fibronectin expression would decrease in response to EMT-gene knockdown, there is evidence to suggest that fibronectin can upregulate its own expression in breast cancer cells, 115 thereby assisting in EMT regulation (Park 2014). As such, FN1 overexpression in response to EMT-gene inhibition may be a compensatory response in the PC-3 cell line. Overexpression of IL6 in response to HIF1α, Snail and ZEB1 knockdown is consistent with the established feedback loop between IL6/Stat3/miR34a signaling and Snail regulation (Rokavec 2014), such that loss of Snail expression (consistent with HIF1α and ZEB1 loss) induces IL6 expression which acts to re-establish Snail expression through repression of miR34a. Overall, IL6 signaling is known to be significantly associated with prostate cancer progression (Rojas 2011). Given its consistent upregulation in response to EMT regulatory gene knockdown, IL6 may represent an important modulatory factor in EMT progression, thereby potentially explaining its association with more aggressive prostate cancers. Finally, NUPR1 expression is increased in response to HIF1α, ZEB1, SPARC and CLTC knockdowns. There is currently no literature-based evidence linking NUPR1 directly to any of the above-listed genes. However, NUPR1 appears to regulate cell stress responses, such as resistance to chemotherapy (Vincent 2012). Given that EMT is initiated in cancer in response to cellular stresses such as hypoxia, inflammation and cytotoxicity, it could be hypothesized that loss of EMT-regulating genes induces its own cellular stress, resulting in upregulation of secondary stress-response mechanisms mediated through NUPR1. Our functional studies indicate that cells overexpressing NUPR1 in response to HIF1α, ZEB1, SPARC or CLTC knockdown, have a reduced capacity for cell invasion. The reverse of this has been previously established in prostate cancer cell lines, such that knockdown of NUPR1 results in invasive cell behaviour. This suggests that in prostate cancer, NUPR1 may have a suppressive effect on aggressive characteristics which is overridden at the time of EMT onset. Further studies will need to be completed to address the relationship between NUPR1 and EMT regulation. Ultimately, examination of these potential nodal points in EMT signaling provides insight into potential signal regulation governing the phenotypes associated with aggressive cancer 116 behaviour. The functional studies, in combination with the observed gene expression alterations, provide a clear indication that EMT can be broken down into distinct sub-phenotypes that be regulated independently of one another. With further examination, there is potential this could lead to the identification of more accurate phenotype-specific biomarkers for cancer prognostication, or potentially provide insight into the best possible molecular targets for inhibiting the undesirable EMT side effects. 4.6 Conclusions In summary, we have demonstrated, through the employment of gene expression-based profiling in a retrospective cohort of FFPE radical prostatectomy samples, that EMT-associated gene signatures can be used to differentiate between ‘low-risk’ Gleason pattern 3 cancers and ‘intermediate-risk’ Gleason pattern 4 cancers. This suggests that the increased risk of metastatic progression associated with higher pattern prostate cancer may be partially attributed to the onset of EMT. Should our identified multivariate model remain successful in differentiating between Gleason patterns in an independent validation cohort, it could represent a novel prognostication tool in prostate cancer management. Eventually this study will need to address the applicability of this model in prostate cancer biopsies, particularly with respect to those Gleason pattern 3 cores that come from both Gleason score 3+3 and Gleason score 3+4 prostate tumours. If our model is capable of predicting occult Gleason pattern 4 from a prostate cancer biopsy, it could aid significantly in clinical risk stratification and in guiding treatment decisions. In addition, we have completed functional validation of a subset of biologically and statistically interesting genes in the context of EMT-associated phenotypes. We have reinforced previous work in the field, confirming the roles of Snail, ZEB1 and HIF1α in the regulation of cell migration and invasion, as indicated in 2D and 3D culture models. Furthermore, we have delineated functional roles for VIM, CLTC and SPARC in EMT using prostate cancer cell line models. SPARC has been shown to contribute significantly to both cell migration and invasion, 117 while VIM and CLTC appear to function primarily during cell invasion in 2D and 3D culture models, while having little impact on cell migration. These results suggest that EMT may not represent a single entity regulating all components simultaneously, but that the various phenotypic aspects of the program can be decoupled and are individually depend on differential factors. Finally, we used our knockdown cell line models to interrogate signaling alterations in response to individual gene dysregulation during EMT. We have identified five nodal points upon which EMT signaling in prostate cancer cells appears to converge (E-cadherin, FN1, SPARC, NUPR1 and IL6). Ultimately, identification of those genes which have the most critical or compensatory roles in EMT regulation may eventually lead to the discovery of optimal therapeutic targets for treating aggressive cancers. 118 Chapter 5: General Discussion 5.1 Revisiting EMT in the regulation of chemotherapy response in ovarian cancer Since the publication of our work additional research has been conducted that confirms our findings regarding a role for EMT in Cisplatin resistance in ovarian cancer. Repeated exposure to Cisplatin in the SKOV-3 ovarian cancer cell line was found to result in the acquisition of a mesenchymal phenotype, concomitant with the development of Cisplatin resistance (Wintzell 2012). These results identify the trademark loss of E-cadherin, upregulation of Vimentin, Twist and Snail, as well as characteristic increased cell motility in the derived Cisplatin-resistant cell line, indicating that a shift towards a mesenchymal state is associated with Cisplatin tolerance. This is consistent with our findings that acquisition of Cisplatin resistance in the A2780cis cell line is associated with upregulation of EMT-TFs and increased cell migration and invasion. EMT also appears to be critical in the development of resistance to other chemotherapeutic agents, such as carboplatin and Paclitaxel (Gupta 2013, Du 2013). The fact that EMT seems to function in generalized resistance to chemotherapy suggests that this program may work to regulate cell survival and apoptotic responses, rather than by interfering with drugspecific modes of action. The mechanisms by which this may be occurring are still largely uncharacterized, although recent evidence that E-cadherin is required for cell-extrinsic apoptosis (Lu 2014). E-cadherin appears to facilitate DR4/5 clustering and DISC assembly through its interaction with the actin cytoskeleton. This suggests that when E-cadherin activity is lost, proapoptotic signaling through receptors DR4 and DR5 is repressed, leading to attenuation of extrinsically initiated apoptosis (Lu 2014). In order to facilitate chemotherapy resistance, mechanisms to block both extrinsic and intrinsic apoptosis programs must be active. While there 119 is no direct evidence linking EMT with inhibition of intrinsic apoptosis in chemotherapy resistant ovarian cancer, associations can be made that suggest the potential for EMT contribution in this area. For example, the EMT-TF, ZEB1, inhibits the expression of pro-apoptotic factors such as Bax, while inducing the anti-apoptotic factors, Bcl-2 and BIRC5, leading to chemotherapy resistance in mantle cell lymphoma (Sanchez-Tillo 2014). In ovarian cancer, ZEB1 is induced by Snail, which can be upregulated in a number of ways, including transcriptional activation through loss of p53 function, a well-established trait of high grade serous ovarian cancer. Therefore, characteristic loss of p53 may prime ovarian tumours for chemotherapy resistance by facilitating the upregulation of Snail, which leads to repression of E-cadherin, resulting in inhibition of death receptor activity, while simultaneously inducing the expression of secondary transcription factors such as ZEB1, thereby modulating intrinsic apoptosis through regulation of pro- and antiapoptotic Bcl-2 family members (Figure 5.1). 120 Figure 5.1 Potential mechanisms contributing to EMT-regulated chemotherapy resistance in high grade serous ovarian cancer The characteristic loss of p53function in high grade serous cancer may prime tumours for chemoresistance by facilitating upregulation of Snail. This results in repression of E-cadherin, leading to inhibition of death receptor activity. Additionally, Snail upregulates secondary transcription factors such as ZEB1 which works to modulate intrinsic apoptosis. 121 5.2 The tumour microenvironment and EMT As the function of EMT in cancer behaviour continues to be unraveled, there is an increasing understanding of the importance of the tumour microenvironment in regulating this process. Gene expression profiling in our prostate cancer cohort revealed upregulation of HIF1α and EPAS1 in Gleason pattern 4 prostate cancer, compared to Gleason pattern 3. The association of hypoxia signaling and increased risk for disease progression suggests a role for the tumour microenvironment in mediating prostate cancer aggressiveness. Prognostic biomarker studies have revealed that upregulation of HIF1α is indicative of poor overall prognosis in rectal cancer (Novell 2014), NSCLC (Wang 2014), prostate cancer (Weber 2012), pancreatic cancer (Hoffmann 2008) and breast cancer (Dong 2013). In response to hypoxic conditions, HIF1α is stabilized and binds to HREs in the Snail promoter regions, resulting in upregulation of Snail expression and ultimately leading to the initiation of EMT signaling cascades (Zhu 2013). Hypoxic conditions also result in increased production and secretion of TGFβ1 in cancerassociated fibroblasts which signals in a paracrine manner to induce EMT in cancer cells (Yu 2014). While increased TGFβ1 signaling in response to environmental stress is known to contribute to metastatic progression, it is also significantly associated with resistance to chemotherapy. In fact, inflammation-mediated production of IL6, alongside fibroblast-secreted TGFβ, can induce EMT in lung cancer cells and modulate response to Erlotinib (Yao 2010). Taken together these data suggest that therapeutic strategies targeting components of the tumour microenvironment may provide benefit in modulating the adverse effects of EMT initiation in tumour cells. These therapeutic strategies will be discussed later on. 5.3 EMT and cancer stem cells While our data, in conjunction with the findings of others, indicates the importance of EMT in the development of aggressive cancer traits, such as metastatic progression and chemotherapy resistance, additional research has indicated the same phenotypes can be attributed 122 to the presence of cancer stem cells (CSCs). It remains largely unknown whether CSCs represent an independent population within a tumour, or if they exist in the context of EMT initiation and progression, such that the EMT program confers stem-like characteristics to a subpopulation of epithelial cells within the tumour. Currently, evidence in this field is conflicting. Arguments have been made suggesting that aggressive traits such as metastatic spread and chemotherapy resistance can be uncoupled from tumour initiation capability. Xie G et al. (2014) demonstrated that the induction of EMT in epithelial breast cells, both cancerous and non-cancerous, did not enhance tumour initiation, but did increase cell invasion and resistance to doxorubicin and radiation. Furthermore, they established that in mesenchymal-like breast cancer cells, reversion to an epithelial phenotype through MET was insufficient to prevent tumour initiation, but did inhibit invasion and therapy resistance. In this case, evidence suggests that tumour initiating cell populations may not rely upon EMT programs for tumour formation, and that aggressive tumour traits may not be attributable to cancer stem cells, but rather to cells undergoing EMT. Contradictory research in this field indicates that EMT programs are active in cells with tumour initiating capability. This is particularly true when examining CTCs where EMT makers are frequently found in conjunction with markers associated with cancer stem cells (Giordano 2012, Raimondi 2011). Of particular importance is the fact that these EMT/stem-like CTCs have been shown to initiate tumour formation. For example, disseminated tumour cells from mouse models of human prostate cancer display increased expression of TWIST1, Vimentin and MMP2, well established EMT markers (Pavese 2014). When these cells are transplanted back into mice, they demonstrate significantly increased metastatic potential and increased resistance to chemotherapy (Pavese 2014). The fact that these cells are capable of forming metastatic tumour masses indicates that they have tumour-initiating capabilities, and suggests that EMT may regulate stem-like behaviour in tumour cells. Similarly, CTCs obtained from men with metastatic, castration-resistant prostate cancer show increased expression of Vimentin alongside an increase 123 in detection of stemness markers, such as CD44 and Separase (Friedlander 2014), again suggesting the relationship between EMT, cancer stem cells and aggressive cancer traits. While our studies lack the in vivo experiments necessary to comment on the relationship between EMT signaling and tumour initiation, some genomic evidence has been obtained that would suggest a connection between cells undergoing EMT and the upregulation of stem cell associated genes. In our cell line model of Cisplatin resistant ovarian cancer, drug-resistant cells, which demonstrate EMT-associated gene signatures, also reveal a 2.2-fold increase in OCT4 expression, and a 15.8-fold increase in ALDH1A1, compared to the drug-sensitive cell line. Both of these genes have been identified in relation to ovarian cancer stem cells (Peng 2010, Landen 2010) and have also been linked to Cisplatin resistance (Tsai 2011, Landen 2010). In our Cisplatin-resistant cells with Snail and Slug knocked down, ALDH1A1 and OCT4 were subsequently downregulated by 5.2- and 7-fold, respectively, compared to control cells. This suggests a gene regulatory link between transcription factors associated with EMT and genes which regulate stem cell characteristics, and their potential contribution to Cisplatin response. This is supported by evidence that the development of Cisplatin resistant lung cancer cells is associated with the upregulation of EMT-related proteins, Snail and Vimentin, as well as the upregulation of stem cell factors, OCT4 and NANOG (Wang 2014). Taken together, these observations support a biological link between EMT and CSC-associated gene expression in the acquisition of chemotherapy resistance in cancer. In our prostate cancer cell line models, expression of genes associated with stem cell behaviour was not found to be associated with an invasive phenotype, such that the highly invasive PC-3 cell line demonstrated low expression levels of stem cell regulating genes, NANOG, OCT4 and SOX2, which were comparable to that of the non-invasive DU145 cell line (data not shown), despite the PC-3s increased expression of EMT-associated genes. Additional observations made during the course of this study indicated that expression of stem cell regulating 124 genes substantially increased in both PC-3 and DU145 cell lines in response to growth in Matrigel-based 3D cultures. However, the invasive potential of either cell line was not affected, with PC-3s maintaining their invasive characteristics, and DU145 colonies remaining noninvasive. This suggests that cancer cell invasion in response to EMT signaling is not associated with the expression of genes which regulate stem cell pluripotency, and therefore, EMT-mediated invasive behaviour may be decoupled from CSC status in cancer progression. However, it is important to note that our study does not address the contribution of signaling from other components of the tumour microenvironment in the regulation of CSCs and cancer invasion. For example, in hepatocellular cancer, signaling from tumour-associated macrophages results in increased expression of CSC markers, BMI1 and EpCAM cancer cells, while simultaneously inducing Snail expression, resulting in E-cadherin loss and Vimentin overexpression (Fan 2014). Together, this shift towards CSC status and the induction of EMT results in increased cancer cell invasion (Fan 2014). This suggests that in more complex environments, CSC characteristics may be regulated by factors from the tumour microenvironment that simultaneously induce EMT, and facilitate enhanced invasive behaviour, in a manner that cannot be sufficiently modelled in vitro. Ultimately, the results of our study indicate that while CSC signaling and EMT may overlap to a certain degree, it appears that it may be possible to dissociate and independently define the regulatory networks guiding these phenotypes, in the context of drug response and tumour initiation. 5.4 The utility of EMT as predictive or prognostic biomarkers in cancer The results of our studies in chemotherapy resistance and metastatic progression indicate that expression of genes associated with EMT may have utility as both predictive and prognostic biomarkers. While in vitro evidence linking EMT to platinum resistance in ovarian cancer has been further validated since the publication of our data, little evidence exists to support this correlation in primary human tumours. Marchini, S. et al. (2013) identified and validated a 125 limited subset of EMT-associated genes and miRNAs (BAMBI, miR-200c, miR-141) that were significantly correlated with platinum resistance in two independent ovarian cancer cohorts. The association of miR-200c with chemotherapy resistance was also observed in our primary tumour cohort, where it was found to be upregulated 1.5-fold in the resistant tumours, relative to the sensitive tumours (data not shown). Up to 70% of patients with acquired platinum resistance showed dysregulated gene expression related to ECM remodeling and TGF-beta signaling pathways (Marchini 2007), both of which are associated with EMT, and are consistent with our findings. Given the limited number of studies conducted in this field to date, additional research with larger tumour cohorts would need to be completed to confirm this association and identify the most consistent predictive markers for this phenotype, prior to incorporation into clinical practice. The use of EMT genes as prognostic biomarkers in prostate cancer has been addressed by Behnsawy HM. et al. (2013) who found that independently, Vimentin and Twist expression is significantly correlated with shortened time to BCR and that Vimentin, Twist, E-cadherin and Snail expression is correlated with Gleason scores 8 and above. Our work aimed to identify differentially expressed genes that could distinguish between Gleason pattern 3 and 4 prostate cancers, with the eventual goal of enhancing risk-stratification in prognostication. While no single gene from our analysis was capable of significantly distinguishing between Gleason patterns 3 and 4, a multivariate model was identified using the signal intensities of five EMT-associated genes (MST1R, TWIST2, AHNAK, FN1 and POU5F). This model could reasonably distinguish between Gleason pattern 3 and Gleason pattern 4 with an ROC-AUC of 0.70 (p=0.003), suggesting that in fact, EMT-associated gene signatures may have utility as prognostic indicators in prostate cancer, subject to further validation. Despite the evidence submitted by ourselves and others suggesting that EMT may offer utility as a biomarker of cancer prognosis and/or chemotherapy response, there are complications 126 with this strategy that would first need to be addressed. One of the primary difficulties in using EMT-associated gene expression as a biomarker of cancer prognosis is the inherent variability associated with this process across cancer types, as well as between subtypes of the same cancer. EMT in cancer progression, unlike in embryogenesis and wound-healing, is associated with aberrant gene expression and is not clearly defined by strict regulatory mechanisms. As a result, tumours that exhibit different genetic backgrounds, or exposure to distinctive environmental stimuli, appear to initiate and regulate EMT in diverse ways. For example, poor prognosis in ER+ breast cancer is associated with high levels of Snail and Twist expression (van Nes 2012), while in lung cancer, poor prognosis is correlated with increased expression of Snail and Vimentin, as well as loss of E-cadherin (Shi 2013). There are two factors to consider when interpreting findings such as these. The first is that differences in experimental design may be contributing significantly to variation in interstudy results. For example, choosing to profile mRNA versus protein expression may lead to different results, given the array of post-transcriptional and –translational modifications that alter EMT progression. Additionally, chosen quantification methods may contribute to variation in results, such as the decision to profile gene expression using qRT-PCR versus microarray technology, both of which introduce individual study limitations. Furthermore, bioinformatics approaches can have considerable impact on data interpretation, such that the methods used in data pre-processing and normalization can have a substantial impact on the results output. At the same time, the transient nature EMT regulation introduces complications of its own, such that expression of specific genes is dependent on the temporal window in which a particular tumour is profiled. EMT-TFs and the signaling networks they regulate are known to be highly dependent on interconnected, regulatory feedback loops, resulting in oscillating expression of EMT-associated genes that may only present themselves for short periods of time during EMT initiation or progression. 127 To compensate for this inherent variability, it may be more useful to reject the approach of attempting to identify small disease-specific EMT panels, in favour of profiling EMT using larger, more broad-scoped gene panels that can be employed across cancer types. This approach would require the establishment of a weighted scoring system within the panel that could provide insight in the degree of EMT involvement, and be translated into clinically meaningful prognostic information. 5.5 Targeting EMT in the therapeutic management of aggressive cancer Given that EMT is well-established for its contribution to aggressive tumour behavior, specifically in cancer metastasis, and more recently in the development of treatment resistance, it is not surprising that attempts are being made to target this program for new therapy options in cancer management. Strategies proposed for targeting EMT in cancer therapy include: inhibition of initiation of EMT, impeding EMT-TF function, targeting mesenchymal factors and blocking MET. Taking into account the results of our gene expression data comparing Gleason pattern 3 and 4 prostate cancer, which suggest that increased hypoxia signaling may represent an important marker of augmented risk of cancer progression, targeting HIF1α may be an attractive therapeutic option. Indeed, targeting the hypoxia signaling cascade is not a novel idea in cancer therapy. However it has more recently been considered a promising option due to its ability to initiate EMT. Research in breast cancer has demonstrated that prevention of HIF1α signaling, through the use of the hypoxia-activated compound, 2-benzoyl-3-phenyl-6,7-dichloroquinoxaline 1,4-dioxide (DCQ), results in reduced expression of TWIST1, and subsequently leads to inhibition of cell invasion and cancer metastasis in vivo (Ghattass 2014). Given our experimental results demonstrating that Gleason pattern 4 prostate cancer exhibits significantly increased expression of HIF1α, compared to Gleason pattern 3, options such as this may have utility in the treatment of Gleason score 7 prostate cancers by preventing EMT initiation and therefore obstructing metastatic progression. 128 However, 20-40% of patients with Gleason score 7 prostate cancer who undergo RP eventually develop recurrent disease, indicating that cancer cell dissemination may already be occurring at this stage and therefore targeting hypoxia responses may not provide benefit to these patients. Therefore, if Gleason score 7 prostate cancers can be further stratified based on HIF1α expression, targeting hypoxia response factors may provide therapeutic benefit to a subset of patients in which cancer cell dissemination has not yet occurred. Inhibition of the signaling cascades leading to EMT onset is another viable option, both for primary cancer treatment and as an adjuvant therapy to re-sensitize cells to certain drugs. One such example of the latter situation shows that Resveratrol may work to inhibit EMT-dependent Cisplatin resistance in ovarian cancer, thereby renewing the utility of first line chemotherapeutics in disease management (Baribeau 2014). Resveratrol functions by blocking Erk2 phosphorylation, leading to downregulation of the MAPK signaling pathway, one of the common dysregulated pathways leading to EMT. Erk2 inactivation results in decreased expression of Snail, and therefore prevents EMT progression, resulting in increased cell death when exposed to Cisplatin (Baribeau 2014). One of the primary complications with targeting EMT-inducing stimuli and the downstream signaling cascades is that, in the context of cancer progression, unlike in embryogenesis and wound healing, EMT is a highly dysregulated process which can be initiated and sustained through activation of multiple different signaling pathways, both independently and in conjunction with one another. Given this redundancy, targeting a single pathway may prove ineffective in the long term due to the potential for development of therapy resistance. Downstream of the initiating signal cascades, emphasis has been placed on targeting the transcription factors that regulate EMT. While arguments could be made for targeting any of the primary EMT-TFs, there is particular interest in targeting Snail family members, likely due to the fact that they are widely considered the ‘master regulators’ of the program. Given the findings of 129 our studies in both chemotherapy resistant ovarian cancer, and invasive prostate cancer cell lines, which demonstrated an important role for Snail in abrogating both phenotypes, targeting Snail family members may have the broadest impact on inhibiting aggressive cancer traits. To date, only a few successful Snail inhibitors have been developed. One of those compounds, GN-25, works by inhibiting the repressive action of Snail on p53, by disrupting their binding interaction, and thereby reactivating p53 signaling (Lee 2010). In vivo studies with this compound have demonstrated that it is capable of preventing tumour progression and eventually leads to tumour regression (Lee 2010). Given the high frequency of p53 alterations in HGSC, and their contribution to chemotherapy resistance, possibly through the induction of Snail-regulated EMT, this compound could provide benefit for patients who are resistant to standard first line therapy options. However, experimental evidence suggests that this drug is not efficacious in tumours with biallelic p53 mutations, indicating that advanced stage ovarian tumours may see no benefit (Lee 2010). Additional attempts at modulating Snail activity have been made by targeting its Ebox binding domain using a Co(III)-Ebox conjugate (Hearney A 2012). While this drug did not affect viability in breast cancer cells, it did significantly increase the expression of E-cadherin. Given the association between high levels of E-cadherin and reduced cancer metastasis (Onder 2008, Chen 2012), this compound could theoretically be used in conjunction with other therapies to inhibit cancer progression. While targeting EMT-TFs may provide benefit in tumours that have not metastasized, inhibiting their function in patients with signs of metastatic disease is unlikely to aid in cancer treatment. In fact, given that reversal of EMT appears to be necessary for the establishment of metastases (Tsai 2012), inhibition of EMT-TFs after cancer cell dissemination could result in MET and thereby assist in metastatic colonization. For treatment of tumours in which there is evidence that EMT has already facilitated metastatic progression, targeting the mesenchymal phenotype of disseminated cells may provide therapeutic benefit. Results from the functional studies in our prostate cancer cell lines suggest 130 that targeting the mesenchymal marker, Vimentin, may have an impact on aggressive behaviour, such that Vimentin knockdown significantly repressed cell invasion. Withaferin-A, a drug which results in degradation of Vimentin filaments, prevents metastatic colonization in vivo and results in cellular apoptosis at higher concentrations in breast cancer cells (Thaiparambil 2011). Given that overexpression of Vimentin is significantly correlated with poor prognosis in prostate cancer (Behnsawy 2013), drugs that target Vimentin-expressing cells and initiate apoptosis may provide therapeutic benefit for advanced, metastatic disease. In addition to Vimentin, we also investigated the functional roles of CLTC and SPARC in EMT and cancer cell behaviour. While CLTC is unlikely to be an appropriate target for cancer therapy, given the ubiquitous nature of its expression in eukaryotic cells (Kirchhausen 1987), SPARC may represent an appropriate candidate, though its role in cancer remains controversial. Our data indicate the SPARC is upregulated in invasive prostate cancer cells and that reducing its expression results in inhibition of cell migration and invasion. Additionally, upon reexamination of the gene expression data from our Cisplatin resistant ovarian cancer cells and primary tumours, SPARC is found to be overexpressed by 4.8-fold in the resistant A2780cis cells when compared to our drug-sensitive A2780s, and by 2-fold in our drug resistant primary tumours, relative to the drug sensitive tumours. Therefore, SPARC overexpression is found in cells which demonstrate EMT-mediated invasion and chemotherapy resistance. As such, targeting SPARC for therapeutic management of advanced disease may prove successful. Abraxane®, which is currently FDA approved for management of metastatic breast cancer and NSCLC, is an albumin-bound formulation of Paclitaxel (Ma 2013). It has also been used in clinical trials assessing its efficacy in head and neck (Adkins 2013) and pancreatic (Von Hoff 2013) cancers with varying degrees of success. Conjugation to albumin facilitates drug targeting to those cells with high expression of SPARC, an albumin-binding extracellular glycoprotein (Yardley 2013). As such, this drug may 131 prove particularly useful in treating advanced stage, or chemotherapy resistant disease where SPARC has been upregulated in response to EMT progression. Ultimately, EMT provides a number of junctures allowing for therapeutic intervention. Despite this, the contribution of EMT to cancer progression is still largely uncharacterized, and chemotherapeutics interfering with this process should be treated with caution. Given the high degree of plasticity in EMT-MET regulation over the course of cancer progression, inhibition of EMT in the later stages of metastasis may work to facilitate colonization of distant metastases, and therefore worsen overall prognosis. Nonetheless, preliminary evidence suggests a role for EMT-targeting in the therapeutic management of cancers displaying aggressive behaviour, and as such will likely continue to be investigated in the context of drug development. 132 Chapter 6: Future Directions Regarding the next steps in the development of this project, there are number of immediate and long term goals that might be valuable in augmenting the results we have gathered thus far. Firstly, validation of our gene expression findings in independent cohorts must be completed. In the study on chemotherapy resistance in ovarian cancer, new chemosensitive and chemonaïve tumours must be identified and profiled at the mRNA level to confirm our finding that EMT gene signatures can, in fact, differentiate between the two phenotypes. There may be some benefit in applying the method used in the prostate cancer study, such that a comprehensive panel of EMT-associated genes is identified from the microarray data and the best model is selected in an unbiased, statistical manner, as opposed to our initial approach, where genes were chosen based on the observed differences in expression levels between cohorts. If the same genes are identified as being capable of distinguishing resistant from sensitive tumours using a different analytical approach, that could provide an interesting validation of our preliminary conclusions, as well as generate a multivariate model that could be applied to meta-analyses for further validation. Another viable option would be to compile tumour cohorts representing chemotherapy resistance in multiple cancer types and attempt to determine if there are consistently dysregulated EMT genes associated with generalized drug resistance, rather than being tissue specific. In this case, selecting tumours with similar treatment regimens may help eliminate noise from the data, and would provide information about potential mechanisms of EMT-regulated therapeutic resistance in response to certain classifications of drugs. Additionally, the use of alternative ovarian cancer cell line models might be employed to further interrogate the role of EMT in the development of chemotherapy resistance, both to platinum agents and to taxanes. If, for example, cisplatin resistance can be modelled in multiple 133 different cell lines and demonstrate EMT engagement, gene expression profiles can be compared and the most relevant signaling nodes could be identified. Ultimately, this might provide further insight into the bigger picture of EMT regulation in the context of drug response. This work could also be expanded upon to interrogate the presence of CSC-associated gene expression in the development of chemoresistance and its correlation with EMT, which was an unvalidated observation from our data. Ultimately, moving this work towards in vivo models would provide a better representation of the effects that EMT regulation has on drug response in a more complex tumour model. This may also provide insight into possible alternative therapeutic intervention strategies for managing EMT-induced drug resistance. Moving forward with our examination of the role of EMT in metastatic potential in prostate cancer, validation must also be completed using the multivariate model that was identified from our gene expression analysis. If this model can be validated in an independent cohort of radical prostatectomy samples, there would be reason to test its utility in a cohort of prostate cancer biopsies, particularly for the sake of commenting on its function in distinguishing between Gleason pattern 3 biopsy cores that originate from either Gleason score 3+3 or 3+4 tumours. In this way, paired biopsies and radical prostatectomy samples could also be used in validation in an attempt to predict which Gleason 3+3 biopsy specimens actually harbor small volume pattern 4 identified at the time of RP. This step towards validation would provide insight into the clinical applicability of our model, and the possible molecular events that contribute to increased risk of disease progression in prostate cancer. Furthermore, follow-up work in our prostate cancer cell line models could provide additional insight into the differential regulation of EMT-associated phenotypes. As of now, these studies have been completed in a single cell line. Additional prostate cancer cell lines that demonstrate EMT-mediated aggressive behaviour could be employed to evaluate the same 134 genomic manipulations and determine their effects on cell migration and invasion. For example, should we observe the same results in gene-specific phenotypic alteration, this would strongly suggest the necessity of that particular gene in the context of a specific phenotypic trait. If, on the other hand, the same gene knockdown in alternative cell lines results in different phenotypic alterations it could potentially speak to the importance of different genetic and mutational backgrounds in regulating the EMT progression. We could also induce EMT in non-invasive cell lines and document the shift in cell behaviour, as well as resultant gene expression alterations. Comparing these results with those from the previously completed knockdown experiments may provide insight into functionally specific signaling nodes in EMT regulation. Finally, this too could be moved into in vivo models for the sake of examining the role of individual gene knockdowns on metastatic progression. Should any single gene significantly inhibit metastasis, it could represent a viable therapeutic target and could be more closely examined in that capacity. Ultimately, the foundation set by these studies provides ample opportunity for further exploration into EMT-associated aggressive tumour phenotypes. This is particularly true in the context of biomarker development, or in the potential contributions that could be made to the understanding of the complex regulation of EMT in cancer progression. 135 Chapter 7: Concluding Remarks In summation, these studies have provided valuable insight into the significance of EMTassociated signaling in the progression of aggressive cancer traits. We have demonstrated an ability to profile this program at the mRNA level in both ovarian and prostate cancers, such that those tumours with poorer overall prognosis were considerably more likely to express genes associated with a mesenchymal phenotype. This suggests that at the molecular level, EMT initiation and advancement is an important hallmark in the development of unfavourable cancer behaviours, such as resistance to chemotherapy and progression towards metastasis. When our work in chemotherapy resistant ovarian cancer was initiated, the role of EMT in drug response was only just being discussed in the literature. Our findings, in the context of this cancer model, contributed to the general acknowledgement that EMT was at least partially regulating drug resistance in cancer. Since our discovery, others have come to similar conclusions, both in models of ovarian cancer and in other tumour types, as discussed previously. Still, there appears to be minimal available information on the clinical applicability of EMTassociated expression profiles in chemoresistant human ovarian tumours. This could be due to a number of factors, including differences in the parameters of the cohorts being studied, or complexities associated with genotypic alterations in the combination therapies used in ovarian cancer management. It could also be the case that the observations from our primary tumours were strictly artifacts based on a small sample size. None-the-less, modulation of EMT signaling in an in vitro model was capable of demonstrating an impact on Cisplatin-induced apoptosis. At the very least, this indicates a potential biological connection between drug response and EMT. Therefore, the correlations between EMT and clinically relevant chemoresistance in human 136 ovarian cancer are worth further investigation, and could potentially lead to useful predictive tools in disease management. Our work in prostate cancer has demonstrated that increased expression of EMTassociated genes is significantly correlated with higher Gleason pattern, and therefore increasing potential risk for metastatic progression. While the correlation between EMT and metastasis is no longer novel, there is still a great deal of uncertainty regarding the temporal window in which EMT is activated and subsequently results in initiation of metastatic dissemination. Here we have demonstrated that EMT-associated gene signatures are found to be overexpressed in Gleason pattern 4 cancers, which have an increased risk of disease progression relative to ‘low-risk’ pattern 3 cancers. This suggests that upregulation of EMT-associated genes may occur earlier in tumour progression to prime tumours for metastatic events, rather than being restricted to the onset of invasion. Furthermore, we have provided evidence, through a series of gene knockdown experiments, that the phenotypes which are typically associated with EMT, in the context of being a globally regulated entity, may actually be represented by specific regulatory pathways that can be decoupled from one another and individually defined at the molecular level. If this is the case, it could explain some of the variability seen in biomarker identification studies between different cancer types. It may also explain why the data we have generated in tumours demonstrating different EMT- related phenotypes (chemotherapy resistance and progression to metastasis) indicates very little overlap in gene expression profiles, regardless of the indication that EMT signatures are present in both cohorts. Due to this observation, it may be concluded that the best way to use EMT in a predictive or prognostic manner is to target it more comprehensively in tumour tissues. As such, the assignment of a scoring system to the gene/protein panel could potentially work to rank EMT involvement and thereby assign risk levels associated with general prognostic groupings. However, this approach would require a much more thorough understanding of EMT gene regulation than we currently possess. 137 Therefore, while our work, and the work of others, strongly indicates a role for EMT involvement in aggressive cancer behaviour, there is still little that can be said about its dysregulation in cancer, or its potential use as a clinically relevant biomarker or therapeutic target, though advances continue to be made in all areas. Ultimately, given the complexity of the interconnected signaling involved in EMT regulation, as well as its observed transient nature in cells and tissues, a great deal more work will need to be done in order to fully understand how we can exploit our knowledge of this cellular program in order to improve clinical care in cancer patients. 138 References 1. Acloque H, Adams MS, Fishwick K, Bronner-Fraser M, Nieto MA: Epithelialmesenchymal transitions: the importance of changing cell state in development and disease. J Clin Invest 2009, 119(6): 1438–1449. 2. Adkins D, Ley J, Trinkaus K, Thorstad W, Lewis J,Jr, Wildes T, Siegel BA, Dehdashti F, Gay H, Mehan P, et al. A phase 2 trial of induction nab-paclitaxel and cetuximab given with cisplatin and 5-fluorouracil followed by concurrent cisplatin and radiation for locally advanced squamous cell carcinoma of the head and neck. Cancer 2013 Feb 15;119(4):766-73. 3. Agnihotri N, Kumar S, Mehta K. Tissue transglutaminase as a central mediator in inflammation-induced progression of breast cancer. Breast Cancer Res 2013 Feb 25;15(1):202. 4. Ahmad S: Platinum–DNA interactions and subsequent cellular processes controlling sensitivity to anticancer platinum complexes. Chem Biodivers 2010, 7(3):543-66. 5. Aigner K, Dampier B, Descovich L, Mikula M, Sultan A, Schreiber M, Mikulits W, Brabletz T, Strand D, Obrist P, et al. The transcription factor ZEB1 (deltaEF1) promotes tumour cell dedifferentiation by repressing master regulators of epithelial polarity. Oncogene 2007 Oct 25;26(49):6979-88. 6. Akhtar N, Hotchin NA. RAC1 regulates adherens junctions through endocytosis of Ecadherin. Mol Biol Cell 2001 Apr;12(4):847-62. 7. Alcaraz A, Barranco MA, Corral JM, Ribal MJ, Carrio A, Mallofre C, Llopis J, Cetina A, Alvarez-Vijande R. High-grade prostate intraepithelial neoplasia shares cytogenetic alterations with invasive prostate cancer. Prostate 2001 Apr;47(1):29-35. 8. Algarra R, Barba J, Merino I, Tienza A, Tolosa E, Robles JE, Zudaire J. Prognostic value of seminal vesicle involvement due to prostate cancer in radical prostatectomy specimens. Actas Urol Esp 2014 Jul 1. 9. Amatangelo MD, Goodyear S, Varma D, Stearns ME. c-myc expression and MEK1induced Erk2 nuclear localization are required for TGF-beta induced epithelialmesenchymal transition and invasion in prostate cancer. Carcinogenesis 2012 Oct;33(10):1965-75. 10. Armstrong AJ, Marengo MS, Oltean S, Kemeny G, Bitting RL, Turnbull JD, Herold CI, Marcom PK, George DJ, Garcia-Blanco MA. Circulating tumor cells from patients with advanced prostate and breast cancer display both epithelial and mesenchymal markers. Mol Cancer Res 2011 Aug;9(8):997-1007. 11. Arumugam T, Ramachandran V, Fournier KF, Wang H, Marquis L, Abbruzzese JL, Gallick GE, Logsdon CD, McConkey DJ, Choi W: Epithelial to mesencymal transition contributes to drug resistance in pancreatic cancer. Canc Res 2009, 139 69(14):5820-5828. 12. Bae GY, Choi SJ, Lee JS, Jo J, Lee J, Kim J, Cha HJ. Loss of E-cadherin activates EGFR-MEK/ERK signaling, which promotes invasion via the ZEB1/MMP2 axis in nonsmall cell lung cancer. Oncotarget 2013 Dec;4(12):2512-22. 13. Banyard J, Chung I, Wilson AM, Vetter G, Le Bechec A, Bielenberg DR, Zetter BR. Regulation of epithelial plasticity by miR-424 and miR-200 in a new prostate cancer metastasis model. Sci Rep 2013 Nov 6;3:3151. 14. Bao B, Azmi AS, Ali S, Ahmad A, Li Y, Banerjee S, Kong D, Sarkar FH. The biological kinship of hypoxia with CSC and EMT and their relationship with deregulated expression of miRNAs and tumor aggressiveness. Biochim Biophys Acta 2012 Dec;1826(2):272-96. 15. Barbieri CE, Bangma CH, Bjartell A, Catto JW, Culig Z, Gronberg H, Luo J, Visakorpi T, Rubin MA. The mutational landscape of prostate cancer. Eur Urol 2013 Oct;64(4):567-76. 16. Baribeau S, Chaudhry P, Parent S, Asselin E. Resveratrol inhibits cisplatin-induced epithelial-to-mesenchymal transition in ovarian cancer cell lines. PLoS One 2014 Jan 22;9(1):e86987. 17. Barry MJ. Clinical practice. prostate-specific-antigen testing for early diagnosis of prostate cancer. N Engl J Med 2001 May 3;344(18):1373-7. 18. Behnsawy HM, Miyake H, Harada K, Fujisawa M. Expression patterns of epithelialmesenchymal transition markers in localized prostate cancer: Significance in clinicopathological outcomes following radical prostatectomy. BJU Int 2013 Jan;111(1):30-7. 19. Beltran H, Yelensky R, Frampton GM, Park K, Downing SR, MacDonald TY, Jarosz M, Lipson D, Tagawa ST, Nanus DM, et al. Targeted next-generation sequencing of advanced prostate cancer identifies potential therapeutic targets and disease heterogeneity. Eur Urol 2013 May;63(5):920-6. 20. Benedet JL, Bender H, Jones H,3rd, Ngan HY, Pecorelli S. FIGO staging classifications and clinical practice guidelines in the management of gynecologic cancers. FIGO committee on gynecologic oncology. Int J Gynaecol Obstet 2000 Aug;70(2):209-62. 21. Bhatt T, Rizvi A, Batta SP, Kataria S, Jamora C. Signaling and mechanical roles of Ecadherin. Cell Commun Adhes 2013 Dec;20(6):189-99. 22. Bookman MA, Brady MF, McGuire WP, Harper PG, Alberts DS, Friedlander M, Colombo N, Fowler JM, Argenta PA, De Geest K, et al. Evaluation of new platinumbased treatment regimens in advanced-stage ovarian cancer: A phase III trial of the gynecologic cancer intergroup. J Clin Oncol 2009 Mar 20;27(9):1419-25. 23. Boorjian SA, Thompson RH, Tollefson MK, Rangel LJ, Bergstralh EJ, Blute ML, Karnes RJ. Long-term risk of clinical progression after biochemical recurrence following radical prostatectomy: The impact of time from surgery to recurrence. Eur Urol 2011 140 Jun;59(6):893-9. 24. Borza T, Konijeti R, Kibel AS. Early detection, PSA screening, and management of overdiagnosis. Hematol Oncol Clin North Am 2013 Dec;27(6):1091-110, vii. 25. Bowden ET, Onikoyi E, Slack R, Myoui A, Yoneda T, Yamada KM, Mueller SC. Colocalization of cortactin and phosphotyrosine identifies active invadopodia in human breast cancer cells. Exp Cell Res 2006 May 1;312(8):1240-53. 26. Bracken CP, Gregory PA, Kolesnikoff N, Bert AG, Wang J, Shannon MF, Goodall GJ. A double-negative feedback loop between ZEB1-SIP1 and the microRNA-200 family regulates epithelial-mesenchymal transition. Cancer Res 2008 Oct 1;68(19):7846-54. 27. Brekken RA, Sage EH. SPARC, a matricellular protein: At the crossroads of cell-matrix communication. Matrix Biol 2001 Jan;19(8):816-27. 28. Bubendorf L, Schopfer A, Wagner U, Sauter G, Moch H, Willi N, Gasser TC, Mihatsch MJ. Metastatic patterns of prostate cancer: An autopsy study of 1,589 patients. Hum Pathol 2000 May;31(5):578-83. 29. Burger H, Loos WJ, Eechoute K, Verweij J, Mathijssen RH, Wiemer EA: Drug transporters of platinum-based anticancer agents and their clinical significance. Drug Resist Update 2011, 14(1):22-34. 30. Burk U, Schubert J, Wellner U, Schmalhofer O, Vincan E, Spaderna S, Brabletz T. A reciprocal repression between ZEB1 and members of the miR-200 family promotes EMT and invasion in cancer cells. EMBO Rep 2008 Jun;9(6):582-9. 31. Canadian Cancer Society’s Advisory Committee on Cancer Statistics. Canadian Cancer Statistics 2013.Toronto, ON: Canadian Cancer Society; 2013. 32. Canadian Cancer Society (2011) Canadian Cancer Statistics. Statistics Canada/Public Health Agency of Canada www.cancer.ca 33. Cannistra SA. Cancer of the ovary. N Engl J Med 2004 Dec 9;351(24):2519-29. 34. Capaccione KM, Hong X, Morgan KM, Liu W, Bishop JM, Liu L, Markert E, Deen M, Minerowicz C, Bertino JR, et al. Sox9 mediates Notch1-induced mesenchymal features in lung adenocarcinoma. Oncotarget 2014 Jun 15;5(11):3636-50. 35. Castro E, Goh C, Olmos D, Saunders E, Leongamornlert D, Tymrakiewicz M, Mahmud N, Dadaev T, Govindasami K, Guy M, et al. Germline BRCA mutations are associated with higher risk of nodal involvement, distant metastasis, and poor survival outcomes in prostate cancer. J Clin Oncol 2013 May 10;31(14):1748-57. 36. Catenacci DV, Cervantes G, Yala S, Nelson EA, El-Hashani E, Kanteti R, El Dinali M, Hasina R, Bragelmann J, Seiwert T, et al. RON (MST1R) is a novel prognostic marker and therapeutic target for gastroesophageal adenocarcinoma. Cancer Biol Ther 2011 Jul 1;12(1):9-46. 141 37. Chaffer CL, Brennan JP, Slavin JL, Blick T, Thompson EW, Williams ED. Mesenchymal-to-epithelial transition facilitates bladder cancer metastasis: Role of fibroblast growth factor receptor-2. Cancer Res 2006 Dec 1;66(23):11271-8. 38. Chang TH, Tsai MF, Su KY, Wu SG, Huang CP, Yu SL, Yu YL, Lan CC, Yang CH, Lin SB, Wu CP, Shih JY, Yang PC: Slug confers resistance to the epidermal growth factor receptor tyrosine kinase inhibitor. Am J Respir Crit Care Med 2010, 183(8):1071-9. 39. Chao YL, Shepard CR, Wells A. Breast carcinoma cells re-express E-cadherin during mesenchymal to epithelial reverting transition. Mol Cancer 2010 Jul 7;9:179,4598-9-179. 40. Chen K, Huang YH, Chen JL. Understanding and targeting cancer stem cells: Therapeutic implications and challenges. Acta Pharmacol Sin 2013 Jun;34(6):732-40. 41. Chen J, Shi D, Liu X, Fang S, Zhang J, Zhao Y. Targeting SPARC by lentivirusmediated RNA interference inhibits cervical cancer cell growth and metastasis. BMC Cancer 2012 Oct 10;12:464,2407-12-464. 42. Chen QK, Lee K, Radisky DC, Nelson CM. Extracellular matrix proteins regulate epithelial-mesenchymal transition in mammary epithelial cells. Differentiation 2013 Oct;86(3):126-32. 43. Chen R, Dong Y, Xie X, Chen J, Gao D, Liu Y, Ren Z, Cui J. Screening candidate metastasis-associated genes in three-dimensional HCC spheroids with different metastasis potential. Int J Clin Exp Pathol 2014 Apr 15;7(5):2527-35. 44. Chen X, Wang Y, Xia H, Wang Q, Jiang X, Lin Z, Ma Y, Yang Y, Hu M. Loss of Ecadherin promotes the growth, invasion and drug resistance of colorectal cancer cells and is associated with liver metastasis. Mol Biol Rep 2012 Jun;39(6):6707-14. 45. Chene G, Rahimi K, Mes-Masson AM, Provencher D. Surgical implications of the potential new tubal pathway for ovarian carcinogenesis. J Minim Invasive Gynecol 2013 Mar-Apr;20(2):153-9. 46. Cheng TC, Manorek G, Samimi G, Lin X, Berry CC, Howell SB: Identification of genes whose expression is associated with Cisplatin resistance in human ovarian carcinoma cells. Cancer Chemother Pharmacol 2006, 58: 384–395. 47. Chien J, Kuang R, Landen C, Shridhar V. Platinum-sensitive recurrence in ovarian cancer: The role of tumor microenvironment. Front Oncol 2013 Sep 24;3:251. 48. Cho KH, Choi MJ, Jeong KJ, Kim JJ, Hwang MH, Shin SC, Park CG, Lee HY. A ROS/STAT3/HIF-1alpha signaling cascade mediates EGF-induced TWIST1 expression and prostate cancer cell invasion. Prostate 2014 May;74(5):528-36. 49. Christoffersen NR, Silahtaroglu A, Orom UA, Kauppinen S, Lund AH. miR-200b mediates post-transcriptional repression of ZFHX1B. Rna 2007 Aug;13(8):1172-8. 142 50. Colombo N, Peiretti M, Garbi A, Carinelli S, Marini C, Sessa C, ESMO Guidelines Working Group. Non-epithelial ovarian cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann Oncol 2012 Oct;23 Suppl 7:vii20-6. 51. Dasari S, Bernard Tchounwou P. Cisplatin in cancer therapy: Molecular mechanisms of action. Eur J Pharmacol 2014 Oct 5;740C:364-78. 52. Dave JM, Bayless KJ. Vimentin as an integral regulator of cell adhesion and endothelial sprouting. Microcirculation 2014 May;21(4):333-44. 53. Davis FM, Stewart TA, Thompson EW, Monteith GR. Targeting EMT in cancer: Opportunities for pharmacological intervention. Trends Pharmacol Sci 2014 Sep;35(9):479-88. 54. Debes JD, Tindall DJ. Mechanisms of androgen-refractory prostate cancer. N Engl J Med 2004 Oct 7;351(15):1488-90. 55. Deep G, Jain AK, Ramteke A, Ting H, Vijendra KC, Gangar SC, Agarwal C, Agarwal R. SNAI1 is critical for the aggressiveness of prostate cancer cells with low E-cadherin. Mol Cancer 2014 Feb 24;13:37,4598-13-37. 56. Derosa CA, Furusato B, Shaheduzzaman S, Srikantan V, Wang Z, Chen Y, Seifert M, Ravindranath L, Young D, Nau M, et al. Elevated osteonectin/SPARC expression in primary prostate cancer predicts metastatic progression. Prostate Cancer Prostatic Dis 2012 Jun;15(2):150-6. 57. Djulbegovic M, Beyth RJ, Neuberger MM, Stoffs TL, Vieweg J, Djulbegovic B, Dahm P. Screening for prostate cancer: Systematic review and meta-analysis of randomised controlled trials. Bmj 2010 Sep 14;341:c4543. 58. Dong C, Wu Y, Yao J, Wang Y, Yu Y, Rychahou PG, Evers BM, Zhou BP. G9a interacts with Snail and is critical for Snail-mediated E-cadherin repression in human breast cancer. J Clin Invest 2012 Apr 2;122(4):1469-86. 59. Dong M, Wan XB, Yuan ZY, Wei L, Fan XJ, Wang TT, Lv YC, Li X, Chen ZH, Chen J, et al. Low expression of beclin 1 and elevated expression of HIF-1alpha refine distant metastasis risk and predict poor prognosis of ER-positive, HER2-negative breast cancer. Med Oncol 2013 Mar;30(1):355,012-0355-0. Epub 2013 Feb 14. 60. Dressman HK, Berchuck A, Chan G, Zhai J, Bild A, Sayer R, Cragun J, Clarke J, Whitaker RS, Li L, Gray J, Marks J, Ginsburg GS, Potti A, West M, Nevins JR, Lancaster JM: An integrated genomic-based approach to individualized treatment of patients with advanced stage ovarian cancer. J Clin Oncol 2007, 25(5):517-525. 61. Du F, Wu X, Liu Y, Wang T, Qi X, Mao Y, Jiang L, Zhu Y, Chen Y, Zhu R, et al. Acquisition of paclitaxel resistance via PI3Kdependent epithelialmesenchymal transition in A2780 human ovarian cancer cells. Oncol Rep 2013 Sep;30(3):1113-8. 62. Du Z, Qin R, Wei C, Wang M, Shi C, Tian R, Peng C. Pancreatic cancer cells resistant to chemoradiotherapy rich in "stem-cell-like" tumor cells. Dig Dis Sci 2011 Mar;56(3):741143 50. 63. Ducros E, Berthaut A, Mirshahi P, Lemarchand S, Soria J, Legeais JM, Mirshahi M. Expression of extracellular matrix proteins fibulin-1 and fibulin-2 by human corneal fibroblasts. Curr Eye Res 2007 Jun;32(6):481-90. 64. Eckert MA, Lwin TM, Chang AT, Kim J, Danis E, Ohno-Machado L, Yang J. Twist1induced invadopodia formation promotes tumor metastasis. Cancer Cell 2011 Mar 8;19(3):372-86. 65. Edwards J, Krishna NS, Grigor KM, Bartlett JM. Androgen receptor gene amplification and protein expression in hormone refractory prostate cancer. Br J Cancer 2003 Aug 4;89(3):552-6. 66. Ehata S, Hanyu A, Fujime M, Katsuno Y, Fukunaga E, Goto K, Ishikawa Y, Nomura K, Yokoo H, Shimizu T, et al. Ki26894, a novel transforming growth factor-beta type I receptor kinase inhibitor, inhibits in vitro invasion and in vivo bone metastasis of a human breast cancer cell line. Cancer Sci 2007 Jan;98(1):127-33. 67. Ekblom, P: Genetics of kidney development. Curr Opin Nephrol Hypertens 1996, 5(3):282-7. 68. Emmert-Buck MR, Vocke CD, Pozzatti RO, Duray PH, Jennings SB, Florence CD, Zhuang Z, Bostwick DG, Liotta LA, Linehan WM. Allelic loss on chromosome 8p12-21 in microdissected prostatic intraepithelial neoplasia. Cancer Res 1995 Jul 15;55(14):2959-62. 69. Epstein JI. An update of the gleason grading system. J Urol 2010 Feb;183(2):433-40. 70. Epstein JI. Precursor lesions to prostatic adenocarcinoma. Virchows Arch 2009 Jan;454(1):1-16. 71. Epstein JI, Allsbrook WC,Jr, Amin MB, Egevad LL, ISUP Grading Committee. The 2005 international society of urological pathology (ISUP) consensus conference on gleason grading of prostatic carcinoma. Am J Surg Pathol 2005 Sep;29(9):1228-42. 72. Etzioni R, Penson DF, Legler JM, di Tommaso D, Boer R, Gann PH, Feuer EJ. Overdiagnosis due to prostate-specific antigen screening: Lessons from U.S. prostate cancer incidence trends. J Natl Cancer Inst 2002 Jul 3;94(13):981-90. 73. Ezratty EJ, Bertaux C, Marcantonio EE, Gundersen GG. Clathrin mediates integrin endocytosis for focal adhesion disassembly in migrating cells. J Cell Biol 2009 Nov 30;187(5):733-47. 74. Fan QM, Jing YY, Yu GF, Kou XR, Ye F, Gao L, Li R, Zhao QD, Yang Y, Lu ZH, et al. Tumor-associated macrophages promote cancer stem cell-like properties via transforming growth factor-beta1-induced epithelial-mesenchymal transition in hepatocellular carcinoma. Cancer Lett 2014 Oct 1;352(2):160-8. 144 75. Farrell J, Kelly C, Rauch J, Kida K, Garcia-Munoz A, Monsefi N, Turriziani B, Doherty C, Mehta JP, Matallanas D, et al. HGF induces epithelial-to-mesenchymal transition by modulating the mammalian hippo/MST2 and ISG15 pathways. J Proteome Res 2014 Jun 6;13(6):2874-86. 76. Farrell J, Petrovics G, McLeod DG, Srivastava S. Genetic and molecular differences in prostate carcinogenesis between african american and caucasian american men. Int J Mol Sci 2013 Jul 25;14(8):15510-31. 77. Fine SW, Epstein JI. A contemporary study correlating prostate needle biopsy and radical prostatectomy gleason score. J Urol 2008 Apr;179(4):1335-9. 78. Fitchett JE, Hay ED. Medial edge epithelium transforms to mesenchyme after embryonic palatal shelves fuse. Dev Biol 1989 Feb;131(2):455-74. 79. Freedland SJ, Humphreys EB, Mangold LA, Eisenberger M, Dorey FJ, Walsh PC, Partin AW. Risk of prostate cancer-specific mortality following biochemical recurrence after radical prostatectomy. Jama 2005 Jul 27;294(4):433-9. 80. Fuchs IB, Lichtenegger W, Buehler H, Henrich W, Stein H, Kleine-Tebbe A, Schaller G. The prognostic significance of epithelial-mesenchymal transition in breast cancer. Anticancer Res 2002 Nov-Dec;22(6A):3415-9. 81. Galluzzi L, Senovilla L, Vitale I, Michels J, Martins I, Kepp O, Castedo M, Kroemer G: Molecular mechanisms of Cisplatin resistance. Oncogene 2011, DOI: 10.1038/onc.2011.384. 82. Gan Y, Shi C, Inge L, Hibner M, Balducci J, Huang Y. Differential roles of ERK and akt pathways in regulation of EGFR-mediated signaling and motility in prostate cancer cells. Oncogene 2010 Sep 2;29(35):4947-58. 83. Gann PH. Risk factors for prostate cancer. Rev Urol 2002;4 Suppl 5:S3-S10. 84. Geiger TR and Peeper DS: Metastasis mechanisms. Biochimica et Biophysica Acta 2009, 1796:293–308. 85. Geng J, Fan J, Ouyang Q, Zhang X, Zhang X, Yu J, Xu Z, Li Q, Yao X, Liu X, et al. Loss of PPM1A expression enhances invasion and the epithelial-to-mesenchymal transition in bladder cancer by activating the TGF-beta/Smad signaling pathway. Oncotarget 2014 Jul 30;5(14):5700-11. 86. Gerlinger M, Rowan AJ, Horswell S, Larkin J, Endesfelder D, Gronroos E, Martinez P, Matthews N, Stewart A, Tarpey P, et al. Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med 2012 Mar 8;366(10):883-92. 87. Ghattass K, El-Sitt S, Zibara K, Rayes S, Haddadin MJ, El-Sabban M, Gali-Muhtasib H. The quinoxaline di-N-oxide DCQ blocks breast cancer metastasis in vitro and in vivo by targeting the hypoxia inducible factor-1 pathway. Mol Cancer 2014 Jan 24;13:12,459813-12. 145 88. Giannoni E, Bianchini F, Masieri L, Serni S, Torre E, Calorini L, Chiarugi P. Reciprocal activation of prostate cancer cells and cancer-associated fibroblasts stimulates epithelialmesenchymal transition and cancer stemness. Cancer Res 2010 Sep 1;70(17):6945-56. 89. Giordano A, Gao H, Anfossi S, Cohen E, Mego M, Lee BN, Tin S, De Laurentiis M, Parker CA, Alvarez RH, et al. Epithelial-mesenchymal transition and stem cell markers in patients with HER2-positive metastatic breast cancer. Mol Cancer Ther 2012 Nov;11(11):2526-34. 90. Gnemmi V, Bouillez A, Gaudelot K, Hemon B, Ringot B, Pottier N, Glowacki F, Villers A, Vindrieux D, Cauffiez C, et al. MUC1 drives epithelial-mesenchymal transition in renal carcinoma through Wnt/beta-catenin pathway and interaction with SNAIL promoter. Cancer Lett 2014 May 1;346(2):225-36. 91. Goff BA, Mandel LS, Melancon CH, Muntz HG. Frequency of symptoms of ovarian cancer in women presenting to primary care clinics. Jama 2004 Jun 9;291(22):2705-12. 92. Gomella LG, Liu XS, Trabulsi EJ, Kelly WK, Myers R, Showalter T, Dicker A, Wender R. Screening for prostate cancer: The current evidence and guidelines controversy. Can J Urol 2011 Oct;18(5):5875-83. 93. Gomez EW, Chen QK, Gjorevski N, Nelson CM. Tissue geometry patterns epithelialmesenchymal transition via intercellular mechanotransduction. J Cell Biochem 2010 May;110(1):44-51. 94. Graham TR, Zhau HE, Odero-Marah VA, Osunkoya AO, Kimbro KS, Tighiouart M, Liu T, Simons JW, O'Regan RM. Insulin-like growth factor-I-dependent up-regulation of ZEB1 drives epithelial-to-mesenchymal transition in human prostate cancer cells. Cancer Res 2008 Apr 1;68(7):2479-88. 95. Gregory CW, Johnson RT,Jr, Mohler JL, French FS, Wilson EM. Androgen receptor stabilization in recurrent prostate cancer is associated with hypersensitivity to low androgen. Cancer Res 2001 Apr 1;61(7):2892-8. 96. Gunasinghe NP, Wells A, Thompson EW, Hugo HJ. Mesenchymal-epithelial transition (MET) as a mechanism for metastatic colonisation in breast cancer. Cancer Metastasis Rev 2012 Dec;31(3-4):469-78. 97. Gupta N, Xu Z, El-Sehemy A, Steed H, Fu Y. Notch3 induces epithelial-mesenchymal transition and attenuates carboplatin-induced apoptosis in ovarian cancer cells. Gynecol Oncol 2013 Jul;130(1):200-6. 98. Gupta P, Srivastava SK. HER2 mediated de novo production of TGFbeta leads to SNAIL driven epithelial-to-mesenchymal transition and metastasis of breast cancer. Mol Oncol 2014 Jun 18. 99. Hajra KM, Chen DY, Fearon ER. The SLUG zinc-finger protein represses E-cadherin in breast cancer. Cancer Res 2002 Mar 15;62(6):1613-8. 146 100. Han XY, Wei B, Fang JF, Zhang S, Zhang FC, Zhang HB, Lan TY, Lu HQ, Wei HB. Epithelial-mesenchymal transition associates with maintenance of stemness in spheroidderived stem-like colon cancer cells. PLoS One 2013 Sep 9;8(9):e73341. 101. Han Z, Feng J, Hong Z, Chen L, Li W, Liao S, Wang X, Ji T, Wang S, Ma D, et al. Silencing of the STAT3 signaling pathway reverses the inherent and induced chemoresistance of human ovarian cancer cells. Biochem Biophys Res Commun 2013 May 31;435(2):188-94. 102. Hara J, Miyata H, Yamasaki M, Sugimura K, Takahashi T, Kurokawa Y, Nakajima K, Takiguchi S, Mori M, Doki Y. Mesenchymal phenotype after chemotherapy is associated with chemoresistance and poor clinical outcome in esophageal cancer. Oncol Rep 2014 Feb;31(2):589-96. 103. Harma V, Virtanen J, Makela R, Happonen A, Mpindi JP, Knuuttila M, Kohonen P, Lotjonen J, Kallioniemi O, Nees M. A comprehensive panel of three-dimensional models for studies of prostate cancer growth, invasion and drug responses. PLoS One 2010 May 3;5(5):e10431. 104. Harney AS, Meade TJ, LaBonne C. Targeted inactivation of Snail family EMT regulatory factors by a co(III)-ebox conjugate. PLoS One 2012;7(2):e32318. 105. Hay ED. Role of cell-matrix contacts in cell migration and epithelial-mesenchymal transformation. Cell Differ Dev 1990 Dec 2;32(3):367-75. 106. Hay E: The mesenchymal cell, its role in the embryo, and the remarkable signaling mechanisms that create it. Dev Dyn 2005, 233(3):706-20. 107. Helleman J, Smid M, Jansen MP, van der Berg ME, Berns EM: Pathway analysis of gene lists associated with platinum-based chemotherapy resistance in ovarian cancer: the big picture. Gynecol Oncol 2010, 117(2):170-6. 108. Henrique R, Jeronimo C. Molecular detection of prostate cancer: A role for GSTP1 hypermethylation. Eur Urol 2004 Nov;46(5):660,9; discussion 669. 109. Herranz N, Pasini D, Diaz VM, Franci C, Gutierrez A, Dave N, Escriva M, HernandezMunoz I, Di Croce L, Helin K, et al. Polycomb complex 2 is required for E-cadherin repression by the Snail1 transcription factor. Mol Cell Biol 2008 Aug;28(15):4772-81. 110. Heuberger J, Birchmeier W. Interplay of cadherin-mediated cell adhesion and canonical wnt signaling. Cold Spring Harb Perspect Biol 2010 Feb;2(2):a002915. 111. Hoffman P, Djavan B. Androgen deprivation therapy. Rev Urol 2008 Fall;10(4):305-6. 112. Hoffmann AC, Mori R, Vallbohmer D, Brabender J, Klein E, Drebber U, Baldus SE, Cooc J, Azuma M, Metzger R, et al. High expression of HIF1a is a predictor of clinical outcome in patients with pancreatic ductal adenocarcinomas and correlated to PDGFA, VEGF, and bFGF. Neoplasia 2008 Jul;10(7):674-9. 147 113. Hogberg T: Chemotherapy: current drugs still have potential in advanced ovarian cancer. Nat Rev Clin Oncol 2010, 7(4):191-193. 114. Hogstrand C, Kille P, Ackland ML, Hiscox S, Taylor KM. A mechanism for epithelialmesenchymal transition and anoikis resistance in breast cancer triggered by zinc channel ZIP6 and STAT3 (signal transducer and activator of transcription 3). Biochem J 2013 Oct 15;455(2):229-37. 115. Holschneider C and Berek J: Ovarian cancer: Epidemiology, biology, and prognostic factors. Seminars in Surgical Oncology 2000, 19:3–10. 116. Hong SM, Li A, Olino K, Wolfgang CL, Herman JM, Schulick RD, Iacobuzio-Donahue C, Hruban RH, Goggins M. Loss of E-cadherin expression and outcome among patients with resectable pancreatic adenocarcinomas. Mod Pathol 2011 Sep;24(9):1237-47. 117. Hou Z, Peng H, Ayyanathan K, Yan KP, Langer EM, Longmore GD, Rauscher FJ,3rd. The LIM protein AJUBA recruits protein arginine methyltransferase 5 to mediate SNAIL-dependent transcriptional repression. Mol Cell Biol 2008 May;28(10):3198-207. 118. Howlader N, Noone AM, Krapcho M, Garshell J, Miller D, Altekruse SF, Kosary CL, Yu M, Ruhl J, Tatalovich Z,Mariotto A, Lewis DR, Chen HS, Feuer EJ, Cronin KA (eds). SEER Cancer Statistics Review, 1975-2011, National Cancer Institute. Bethesda, MD, http://seer.cancer.gov/csr/1975_2011/, based on November 2013 SEER data submission, posted to the SEER web site, April 2014. 119. Hsu DS, Lan HY, Huang CH, Tai SK, Chang SY, Tsai TL, Chang CC, Tzeng CH, Wu KJ, Kao JY, et al. Regulation of excision repair cross-complementation group 1 by Snail contributes to cisplatin resistance in head and neck cancer. Clin Cancer Res 2010 Sep 15;16(18):4561-71. 120. Hudson LG, Newkirk KM, Chandler HL, Choi C, Fossey SL, Parent AE, Kusewitt DF. Cutaneous wound reepithelialization is compromised in mice lacking functional Slug (Snai2). J Dermatol Sci 2009 Oct;56(1):19-26. 121. Hwang-Verslues WW, Chang PH, Jeng YM, Kuo WH, Chiang PH, Chang YC, Hsieh TH, Su FY, Lin LC, Abbondante S, et al. Loss of corepressor PER2 under hypoxia upregulates OCT1-mediated EMT gene expression and enhances tumor malignancy. Proc Natl Acad Sci U S A 2013 Jul 23;110(30):12331-6. 122. Ikenouchi J, Matsuda M, Furuse M, Tsukita S. Regulation of tight junctions during the epithelium-mesenchyme transition: Direct repression of the gene expression of claudins/occludin by Snail. J Cell Sci 2003 May 15;116(Pt 10):1959-67. 123. Iruela-Arispe ML, Lane TF, Redmond D, Reilly M, Bolender RP, Kavanagh TJ, Sage EH. Expression of SPARC during development of the chicken chorioallantoic membrane: Evidence for regulated proteolysis in vivo. Mol Biol Cell 1995 Mar;6(3):327-43. 124. Işeri OD, Kars MD, Arpaci F, Atalay C, Pak I, Gündüz U: Drug resistant MCF-7 cells exhibit epithelial-mesenchymal transition gene expression pattern. Biomed 148 and Pharmacotherapy 2011, 65(1):40-5. 125. Ismail MT, Gomella LG. Transrectal prostate biopsy. Urol Clin North Am 2013 Nov;40(4):457-72. 126. Iwatsuki M, Mimori K, Yokobori T, Ishi H, Beppu T, Nakamori S, Baba H, Mori M: Epithelial–mesenchymal transition in cancer development and its clinical significance. Canc Science 2009, 101(2):293-299. 127. Izumi G, Sakisaka T, Baba T, Tanaka S, Morimoto K, Takai Y. Endocytosis of Ecadherin regulated by rac and Cdc42 small G proteins through IQGAP1 and actin filaments. J Cell Biol 2004 Jul 19;166(2):237-48. 128. Izumi K, Fang LY, Mizokami A, Namiki M, Li L, Lin WJ, Chang C. Targeting the androgen receptor with siRNA promotes prostate cancer metastasis through enhanced macrophage recruitment via CCL2/CCR2-induced STAT3 activation. EMBO Mol Med 2013 Sep;5(9):1383-401. 129. Jacob K, Webber M, Benayahu D, Kleinman HK. Osteonectin promotes prostate cancer cell migration and invasion: A possible mechanism for metastasis to bone. Cancer Res 1999 Sep 1;59(17):4453-7. 130. Janda E, Nevolo M, Lehmann K, Downward J, Beug H, Grieco M. Raf plus TGFbetadependent EMT is initiated by endocytosis and lysosomal degradation of E-cadherin. Oncogene 2006 Nov 16;25(54):7117-30. 131. Jing Y, Cui D, Guo W, Jiang J, Jiang B, Lu Y, Zhao W, Wang X, Jiang Q, Han B, et al. Activated androgen receptor promotes bladder cancer metastasis via Slug mediated epithelial-mesenchymal transition. Cancer Lett 2014 Jun 28;348(1-2):135-45. 132. Johns LE, Houlston RS. A systematic review and meta-analysis of familial prostate cancer risk. BJU Int 2003 Jun;91(9):789-94. 133. Kaighn ME, Narayan KS, Ohnuki Y, Lechner JF, Jones LW. Establishment and characterization of a human prostatic carcinoma cell line (PC-3). Invest Urol 1979 Jul;17(1):16-23. 134. Kallergi G, Papadaki MA, Politaki E, Mavroudis D, Georgoulias V, Agelaki S. Epithelial to mesenchymal transition markers expressed in circulating tumour cells of early and metastatic breast cancer patients. Breast Cancer Res 2011 Jun 10;13(3):R59. 135. Kalluri R: EMT: When epithelial cells decide to become mesenchymal-like cells. J Clin Invest 2009, 119:1417–1419. 136. Kalluri R, Weinberg RA. The basics of epithelial-mesenchymal transition. J Clin Invest 2009 Jun;119(6):1420-8. 137. Kassim SK, Ali HS, Sallam MM, Fayed ST, Seada LS, abd-Elkawy E, Seada MA, Khalifa A. Increased bcl-2 expression is associated with primary resistance to 149 chemotherapy in human epithelial ovarian cancer. Clin Biochem 1999 Jul;32(5):333-8. 138. Khalil I, Brewer MA, Neyarapally T, Runowicz CD: The potential of biologic network models in understanding the etiopathogenesis of ovarian cancer. Gynecol Oncol 2010, 116(2):282-285. 139. Kim YH, Kim G, Kwon CI, Kim JW, Park PW, Hahm KB. TWIST1 and SNAI1 as markers of poor prognosis in human colorectal cancer are associated with the expression of ALDH1 and TGF-beta1. Oncol Rep 2014 Mar;31(3):1380-8. 140. Kirchhausen T, Harrison SC, Chow EP, Mattaliano RJ, Ramachandran KL, Smart J, Brosius J. Clathrin heavy chain: Molecular cloning and complete primary structure. Proc Natl Acad Sci U S A 1987 Dec;84(24):8805-9. 141. Klotz L. Active surveillance: Patient selection. Curr Opin Urol 2013 May;23(3):239-44. 142. Köbel M, Kalloger SE, Boyd N, McKinney S, Mehl E, Palmer C, Leung S, Bowen NJ, Ionescu DN, Rajput A, Prentice LM, Miller D, Santos J, Swenerton K, Gilks CB, Huntsman D: Ovarian carcinoma subtypes are different diseases: Implications for biomarker studies. PLOS Medicine 2008, 5(12):1749-1760. 143. Koblinski JE, Kaplan-Singer BR, VanOsdol SJ, Wu M, Engbring JA, Wang S, Goldsmith CM, Piper JT, Vostal JG, Harms JF, et al. Endogenous osteonectin/SPARC/BM-40 expression inhibits MDA-MB-231 breast cancer cell metastasis. Cancer Res 2005 Aug 15;65(16):7370-7. 144. Kon S, Tanabe K, Watanabe T, Sabe H, Satake M. Clathrin dependent endocytosis of Ecadherin is regulated by the Arf6GAP isoform SMAP1. Exp Cell Res 2008 Apr 15;314(7):1415-28. 145. Konstantinopolous PA, Spentzos D and Cannistra S: Gene-expression profiling in epithelial ovarian cancer. Nat Clin Pract Oncology 2008, 5:577-587. 146. Kovacs EM, Goodwin M, Ali RG, Paterson AD, Yap AS. Cadherin-directed actin assembly: E-cadherin physically associates with the Arp2/3 complex to direct actin assembly in nascent adhesive contacts. Curr Biol 2002 Mar 5;12(5):379-82. 147. Kulasingam V, Pavlou MP, Diamandis EP: Integrating high-throughput technologies in the quest for effective biomarkers for ovarian cancer. Nature Reviews 2010, 10(5):371378. 148. Kumar S, Das A, Sen S. Extracellular matrix density promotes EMT by weakening cellcell adhesions. Mol Biosyst 2014 Apr;10(4):838-50. 149. Kuo KT, Guan B, Feng Y, Mao TL, Chen X, Jinawath N, Wang Y, Kurman RJ, Shih I, Wang TL. Analysis of DNA copy number alterations in ovarian serous tumors identifies new molecular genetic changes in low-grade and high-grade carcinomas. Cancer Res 2009 May 1;69(9):4036-42. 150 150. Kurrey NK, Jalgaonkar SP, Joglekar AV, Ghanate AD, Chaskar PD, Doiphode RY, Bapat SA: Snail and Slug mediate radioresistance and chemoresistance by antagonizing p53-mediated apoptosis and acquiring a stem-like phenotype in ovarian cancer cells. Stem Cells 2009, 27(9):2059-2068. 151. LaGamba D, Nawshad A and Hay E: Microarray analysis of gene expression during epithelial-mesenchymal transformation. Dev Dynamics 2005, 234:132-142. 152. Landen CN,Jr, Goodman B, Katre AA, Steg AD, Nick AM, Stone RL, Miller LD, Mejia PV, Jennings NB, Gershenson DM, et al. Targeting aldehyde dehydrogenase cancer stem cells in ovarian cancer. Mol Cancer Ther 2010 Dec;9(12):3186-99. 153. Laurencot CM, Andrews PA, Kennedy KA. Inhibitors of intracellular pH regulation induce cisplatin resistance in EMT6 mouse mammary tumor cells. Oncol Res 1995;7(78):363-9. 154. Lavallee LT, Manuel DC, van Walraven C. Survival of men with prostate cancer undergoing radical prostatectomy in ontario. J Urol 2014 Apr 21. 155. Ledda MF, Adris S, Bravo AI, Kairiyama C, Bover L, Chernajovsky Y, Mordoh J, Podhajcer OL. Suppression of SPARC expression by antisense RNA abrogates the tumorigenicity of human melanoma cells. Nat Med 1997 Feb;3(2):171-6. 156. Lee HJ, Lee OJ, Jang KT, Bae YK, Chung JY, Eom DW, Kim JM, Yu E, Hong SM. Combined loss of E-cadherin and aberrant beta-catenin protein expression correlates with a poor prognosis for small intestinal adenocarcinomas. Am J Clin Pathol 2013 Feb;139(2):167-76. 157. Lee JT, Lee S, Yun CJ, Jeon BJ, Kim JM, Ha HK, Lee W, Chung MK. Prediction of perineural invasion and its prognostic value in patients with prostate cancer. Korean J Urol 2010 Nov;51(11):745-51. 158. Lee SH, Shen GN, Jung YS, Lee SJ, Chung JY, Kim HS, Xu Y, Choi Y, Lee JW, Ha NC, et al. Antitumor effect of novel small chemical inhibitors of Snail-p53 binding in K-rasmutated cancer cells. Oncogene 2010 Aug 12;29(32):4576-87. 159. Lee Y, Lee M, Kim S. Gas6 induces cancer cell migration and epithelial-mesenchymal transition through upregulation of MAPK and Slug. Biochem Biophys Res Commun 2013 Apr 26;434(1):8-14. 160. Leibovich-Rivkin T, Liubomirski Y, Bernstein B, Meshel T, Ben-Baruch A. Inflammatory factors of the tumor microenvironment induce plasticity in nontransformed breast epithelial cells: EMT, invasion, and collapse of normally organized breast textures. Neoplasia 2013 Dec;15(12):1330-46. 161. Leinonen KA, Saramaki OR, Furusato B, Kimura T, Takahashi H, Egawa S, Suzuki H, Keiger K, Ho Hahm S, Isaacs WB, et al. Loss of PTEN is associated with aggressive behavior in ERG-positive prostate cancer. Cancer Epidemiol Biomarkers Prev 2013 Dec;22(12):2333-44. 151 162. Levanon K, Crum C, Drapkin R. New insights into the pathogenesis of serous ovarian cancer and its clinical impact. J Clin Oncol 2008 Nov 10;26(32):5284-93. 163. Li FY, Ren XB, Xie XY, Zhang J. Meta-analysis of excision repair crosscomplementation group 1 (ERCC1) association with response to platinum- based chemotherapy in ovarian cancer. Asian Pac J Cancer Prev 2013;14(12):7203-6. 164. Li J, Wood WH, Becker KG, Weeraratna AT, Morin PJ: Gene expression response to Cisplatin treatment in drug-sensitive and drug-resistant ovarian cancer cells. Oncogene 2007, 26: 2860–2872. 165. Liang W, Hao Z, Han JL, Zhu DJ, Jin ZF, Xie WL. CAV-1 contributes to bladder cancer progression by inducing epithelial-to-mesenchymal transition. Urol Oncol 2014 Aug;32(6):855-63. 166. Liao MJ, Zhang CC, Zhou B, Zimonjic DB, Mani SA, Kaba M, Gifford A, Reinhardt F, Popescu NC, Guo W, et al. Enrichment of a population of mammary gland cells that form mammospheres and have in vivo repopulating activity. Cancer Res 2007 Sep 1;67(17):8131-8. 167. Lin CY, Tsai PH, Kandaswami CC, Lee PP, Huang CJ, Hwang JJ, Lee MT. Matrix metalloproteinase-9 cooperates with transcription factor Snail to induce epithelialmesenchymal transition. Cancer Sci 2011 Apr;102(4):815-27. 168. Lin X, Asgari K, Putzi MJ, Gage WR, Yu X, Cornblatt BS, Kumar A, Piantadosi S, DeWeese TL, De Marzo AM, et al. Reversal of GSTP1 CpG island hypermethylation and reactivation of pi-class glutathione S-transferase (GSTP1) expression in human prostate cancer cells by treatment with procainamide. Cancer Res 2001 Dec 15;61(24):8611-6. 169. Linja MJ, Savinainen KJ, Saramaki OR, Tammela TL, Vessella RL, Visakorpi T. Amplification and overexpression of androgen receptor gene in hormone-refractory prostate cancer. Cancer Res 2001 May 1;61(9):3550-5. 170. Liu L, Zhu XD, Wang WQ, Shen Y, Qin Y, Ren ZG, Sun HC, Tang ZY. Activation of beta-catenin by hypoxia in hepatocellular carcinoma contributes to enhanced metastatic potential and poor prognosis. Clin Cancer Res 2010 May 15;16(10):2740-50. 171. Liu RY, Zeng Y, Lei Z, Wang L, Yang H, Liu Z, Zhao J, Zhang HT. JAK/STAT3 signaling is required for TGF-beta-induced epithelial-mesenchymal transition in lung cancer cells. Int J Oncol 2014 May;44(5):1643-51. 172. Lo HW, Hsu SC, Xia W, Cao X, Shih JY, Wei Y, Abbruzzese JL, Hortobagyi GN, Hung MC. Epidermal growth factor receptor cooperates with signal transducer and activator of transcription 3 to induce epithelial-mesenchymal transition in cancer cells via upregulation of TWIST gene expression. Cancer Res 2007 Oct 1;67(19):9066-76. 173. Loeb S, Folkvaljon Y, Makarov DV, Bratt O, Bill-Axelson A, Stattin P. Five-year nationwide follow-up study of active surveillance for prostate cancer. Eur Urol 2014 Jun 30. 152 174. Lu M, Marsters S, Ye X, Luis E, Gonzalez L, Ashkenazi A. E-cadherin couples death receptors to the cytoskeleton to regulate apoptosis. Mol Cell 2014 Jun 19;54(6):987-98. 175. Ma MP, Robinson PJ, Chircop M. Sorting nexin 9 recruits Clathrin heavy chain to the mitotic spindle for chromosome alignment and segregation. PLoS One 2013 Jul 5;8(7):e68387. 176. Ma P, Mumper RJ. Paclitaxel nano-delivery systems: A comprehensive review. J Nanomed Nanotechnol 2013 Feb 18;4(2):1000164. 177. Madore J, Ren F, Filali-Mouhim A, Sanchez L, Kobel M, Tonin PN, Huntsman D, Provencher DM, Mes-Masson AM. Characterization of the molecular differences between ovarian endometrioid carcinoma and ovarian serous carcinoma. J Pathol 2010 Feb;220(3):392-400. 178. Mak P, Leav I, Pursell B, Bae D, Yang X, Taglienti CA, Gouvin LM, Sharma VM, Mercurio AM. ERbeta impedes prostate cancer EMT by destabilizing HIF-1alpha and inhibiting VEGF-mediated Snail nuclear localization: Implications for gleason grading. Cancer Cell 2010 Apr 13;17(4):319-32. 179. Mani SA, Guo W, Liao MJ, Eaton EN, Ayyanan A, Zhou AY, Brooks M, Reinhard F, Zhang CC, Shipitsin M, et al. The epithelial-mesenchymal transition generates cells with properties of stem cells. Cell 2008 May 16;133(4):704-15. 180. Mantia-Smaldone GM, Edwards RP, Vlad AM. Targeted treatment of recurrent platinumresistant ovarian cancer: Current and emerging therapies. Cancer Manag Res 2011;3:2538. 181. Marchiando AM, Shen L, Graham WV, Weber CR, Schwarz BT, Austin JR,2nd, Raleigh DR, Guan Y, Watson AJ, Montrose MH, et al. Caveolin-1-dependent occludin endocytosis is required for TNF-induced tight junction regulation in vivo. J Cell Biol 2010 Apr 5;189(1):111-26. 182. Marchini S, Fruscio R, Clivio L, Beltrame L, Porcu L, Fuso Nerini I, Cavalieri D, Chiorino G, Cattoretti G, Mangioni C, et al. Resistance to platinum-based chemotherapy is associated with epithelial to mesenchymal transition in epithelial ovarian cancer. Eur J Cancer 2013 Jan;49(2):520-30. 183. Marsh S: Pharmacogenomics of Taxane/Platinum therapy in ovarian cancer. Int J Gynecol Cancer 2009, 2:30-34. 184. Martel CL, Gumerlock PH, Meyers FJ, Lara PN. Current strategies in the management of hormone refractory prostate cancer. Cancer Treat Rev 2003 Jun;29(3):171-87. 185. Matuszak EA, Kyprianou N. Androgen regulation of epithelial-mesenchymal transition in prostate tumorigenesis. Expert Rev Endocrinol Metab 2011 May;6(3):469-82. 186. McConkey DJ, Choi W, Marquis L, Martin F, Williams MB, Shah J, Svatek R, Das A, Adam L, Kamat A, Siefker-Radtke A, Dinney C: Role of epithelial-tomesenchymal transition (EMT) in drug sensitivity and metastasis in bladder 153 cancer. Canc Met Rev 2009, 28(3-4):335-344. 187. Melchionna R, Bellavia G, Romani M, Straino S, Germani A, Di Carlo A, Capogrossi MC, Napolitano M. C/EBPgamma regulates wound repair and EGF receptor signaling. J Invest Dermatol 2012 Jul;132(7):1908-17. 188. Merrimen JL, Evans AJ, Srigley JR. Preneoplasia in the prostate gland with emphasis on high grade prostatic intraepithelial neoplasia. Pathology 2013 Apr;45(3):251-63. 189. Micalizzi D, Farabough FM and Ford H: Epithelial-Mesenchymal transition in cancer: Parallels between normal development and tumour progression. J Mam Gland Biol Neoplasia 2010, 15:117-134. 190. Mistry K, Cable G. Meta-analysis of prostate-specific antigen and digital rectal examination as screening tests for prostate carcinoma. J Am Board Fam Pract 2003 MarApr;16(2):95-101. 191. Moore R and MacLaughlin S: Current clinical use of biomarkers for epithelial ovarian cancer. Current Opinion in Oncology 2010, 22(5):492–497. 192. Moustakas A and Heldin CH: Signaling networks guiding epithelial-mesenchymal transitions during embryogenesis and cancer progression. Cancer Sci 2007, 98(10):151220. 193. Nagle RB, Algotar AM, Cortez CC, Smith K, Jones C, Sathyanarayana UG, Yun S, Riley J, Nagy D, Dittamore R, et al. ERG overexpression and PTEN status predict capsular penetration in prostate carcinoma. Prostate 2013 Aug;73(11):1233-40. 194. Nakayama M, Gonzalgo ML, Yegnasubramanian S, Lin X, De Marzo AM, Nelson WG. GSTP1 CpG island hypermethylation as a molecular biomarker for prostate cancer. J Cell Biochem 2004 Feb 15;91(3):540-52. 195. Nam JS, Terabe M, Mamura M, Kang MJ, Chae H, Stuelten C, Kohn E, Tang B, Sabzevari H, Anver MR, et al. An anti-transforming growth factor beta antibody suppresses metastasis via cooperative effects on multiple cell compartments. Cancer Res 2008 May 15;68(10):3835-43. 196. Navarro C, Nola S, Audebert S, Santoni MJ, Arsanto JP, Ginestier C, Marchetto S, Jacquemier J, Isnardon D, Le Bivic A, et al. Junctional recruitment of mammalian scribble relies on E-cadherin engagement. Oncogene 2005 Jun 23;24(27):4330-9. 197. Netto GJ, Eisenberger M, Epstein JI, TAX 3501 Trial Investigators. Interobserver variability in histologic evaluation of radical prostatectomy between central and local pathologists: Findings of TAX 3501 multinational clinical trial. Urology 2011 May;77(5):1155-60. 198. Nezhat FR, Pejovic T, Finger TN, Khalil SS. Role of minimally invasive surgery in ovarian cancer. J Minim Invasive Gynecol 2013 Nov-Dec;20(6):754-65. 154 199. Nicolaisen M, Muller S, Patel HR, Hanssen TA. Quality of life and satisfaction with information after radical prostatectomy, radical external beam radiotherapy and postoperative radiotherapy: A long-term follow-up study. J Clin Nurs 2014 Jun 3. 200. Niessen CM, Leckband D, Yap AS. Tissue organization by cadherin adhesion molecules: Dynamic molecular and cellular mechanisms of morphogenetic regulation. Physiol Rev 2011 Apr;91(2):691-731. 201. Nieto-Morales ML, Fernandez-Ramos J, Perez-Mendez L, Alventosa-Fernandez E, Pastor-Santovena MS, Arias-Rodriguez A, Aguirre-Jaime A. Improving the gleason grading accuracy of transrectal ultrasound-guided biopsy. Acta Radiol 2013 Dec;54(10):1218-23. 202. Nilsson GM, Akhtar N, Kannius-Janson M, Baeckstrom D. Loss of E-cadherin expression is not a prerequisite for c-erbB2-induced epithelial-mesenchymal transition. Int J Oncol 2014 Jul;45(1):82-94. 203. Novell A, Martinez-Alonso M, Mira M, Tarragona J, Salud A, Matias-Guiu X. Prognostic value of c-FLIPL/s, HIF-1alpha, and NF-kappabeta in stage II and III rectal cancer. Virchows Arch 2014 Jun;464(6):645-54. 204. Ocana OH, Corcoles R, Fabra A, Moreno-Bueno G, Acloque H, Vega S, BarralloGimeno A, Cano A, Nieto MA. Metastatic colonization requires the repression of the epithelial-mesenchymal transition inducer Prrx1. Cancer Cell 2012 Dec 11;22(6):709-24. 205. Olson SH, Mignone L, Nakraseive C, Caputo TA, Barakat RR, Harlap S. Symptoms of ovarian cancer. Obstet Gynecol 2001 Aug;98(2):212-7. 206. Onder TT, Gupta PB, Mani SA, Yang J, Lander ES, Weinberg RA. Loss of E-cadherin promotes metastasis via multiple downstream transcriptional pathways. Cancer Res 2008 May 15;68(10):3645-54. 207. Ota I, Li XY, Hu Y, Weiss SJ. Induction of a MT1-MMP and MT2-MMP-dependent basement membrane transmigration program in cancer cells by Snail1. Proc Natl Acad Sci U S A 2009 Dec 1;106(48):20318-23. 208. Ozdamar B, Bose R, Barrios-Rodiles M, Wang HR, Zhang Y, Wrana JL. Regulation of the polarity protein Par6 by TGFbeta receptors controls epithelial cell plasticity. Science 2005 Mar 11;307(5715):1603-9. 209. Ozols RF, Bundy BN, Greer BE, Fowler JM, Clarke-Pearson D, Burger RA, Mannel RS, DeGeest K, Hartenbach EM, Baergen R, et al. Phase III trial of carboplatin and paclitaxel compared with cisplatin and paclitaxel in patients with optimally resected stage III ovarian cancer: A gynecologic oncology group study. J Clin Oncol 2003 Sep 1;21(17):3194-200. 210. Park J, Schwarzbauer JE. Mammary epithelial cell interactions with fibronectin stimulate epithelial-mesenchymal transition. Oncogene 2014 Mar 27;33(13):1649-57. 155 211. Park K, Dalton JT, Narayanan R, Barbieri CE, Hancock ML, Bostwick DG, Steiner MS, Rubin MA. TMPRSS2:ERG gene fusion predicts subsequent detection of prostate cancer in patients with high-grade prostatic intraepithelial neoplasia. J Clin Oncol 2014 Jan 20;32(3):206-11. 212. Park SM, Gaur AB, Lengyel E, Peter ME. The miR-200 family determines the epithelial phenotype of cancer cells by targeting the E-cadherin repressors ZEB1 and ZEB2. Genes Dev 2008 Apr 1;22(7):894-907. 213. Pascal LE, Vencio RZ, Page LS, Liebeskind ES, Shadle CP, Troisch P, Marzolf B, True LD, Hood LE, Liu AY. Gene expression relationship between prostate cancer cells of gleason 3, 4 and normal epithelial cells as revealed by cell type-specific transcriptomes. BMC Cancer 2009 Dec 18;9:452,2407-9-452. 214. Patrawala L, Calhoun T, Schneider-Broussard R, Li H, Bhatia B, Tang S, Reilly JG, Chandra D, Zhou J, Claypool K, et al. Highly purified CD44+ prostate cancer cells from xenograft human tumors are enriched in tumorigenic and metastatic progenitor cells. Oncogene 2006 Mar 16;25(12):1696-708. 215. Pavese JM, Bergan RC. Circulating tumor cells exhibit a biologically aggressive cancer phenotype accompanied by selective resistance to chemotherapy. Cancer Lett 2014 Oct 1;352(2):179-86. 216. Peinado H, Ballestar E, Esteller M, Cano A. Snail mediates E-cadherin repression by the recruitment of the Sin3A/histone deacetylase 1 (HDAC1)/HDAC2 complex. Mol Cell Biol 2004 Jan;24(1):306-19. 217. Peinado H, Olmeda D, Cano A. Snail, zeb and bHLH factors in tumour progression: An alliance against the epithelial phenotype? Nat Rev Cancer 2007 Jun;7(6):415-28. 218. Peiro S, Escriva M, Puig I, Barbera MJ, Dave N, Herranz N, Larriba MJ, Takkunen M, Franci C, Munoz A, et al. Snail1 transcriptional repressor binds to its own promoter and controls its expression. Nucleic Acids Res 2006 Apr 14;34(7):2077-84. 219. Peng S, Maihle NJ, Huang Y. Pluripotency factors Lin28 and Oct4 identify a subpopulation of stem cell-like cells in ovarian cancer. Oncogene 2010 Apr 8;29(14):2153-9. 220. Phua DC, Humbert PO, Hunziker W. Vimentin regulates scribble activity by protecting it from proteasomal degradation. Mol Biol Cell 2009 Jun;20(12):2841-55. 221. Poblete CE, Fulla J, Gallardo M, Munoz V, Castellon EA, Gallegos I, Contreras HR. Increased SNAIL expression and low syndecan levels are associated with high gleason grade in prostate cancer. Int J Oncol 2014 Mar;44(3):647-54. 222. Poincloux R, Lizarraga F, Chavrier P. Matrix invasion by tumour cells: A focus on MT1MMP trafficking to invadopodia. J Cell Sci 2009 Sep 1;122(Pt 17):3015-24. 223. Popiolek M, Rider JR, Andren O, Andersson SO, Holmberg L, Adami HO, Johansson JE. Natural history of early, localized prostate cancer: A final report from three decades of 156 follow-up. Eur Urol 2013 Mar;63(3):428-35. 224. Pound CR, Partin AW, Eisenberger MA, Chan DW, Pearson JD, Walsh PC. Natural history of progression after PSA elevation following radical prostatectomy. Jama 1999 May 5;281(17):1591-7. 225. Pulyaeva H, Bueno J, Polette M, Birembaut P, Sato H, Seiki M, Thompson EW. MT1MMP correlates with MMP-2 activation potential seen after epithelial to mesenchymal transition in human breast carcinoma cells. Clin Exp Metastasis 1997 Mar;15(2):111-20. 226. Qian J, Bostwick DG, Takahashi S, Borell TJ, Herath JF, Lieber MM, Jenkins RB. Chromosomal anomalies in prostatic intraepithelial neoplasia and carcinoma detected by fluorescence in situ hybridization. Cancer Res 1995 Nov 15;55(22):5408-14. 227. Quattrocchi L, Green AR, Martin S, Durrant L, Deen S. The cadherin switch in ovarian high-grade serous carcinoma is associated with disease progression. Virchows Arch 2011 Jul;459(1):21-9. 228. Raatikainen S, Aaltomaa S, Palvimo JJ, Karja V, Soini Y. TWIST overexpression predicts biochemical recurrence-free survival in prostate cancer patients treated with radical prostatectomy. Scand J Urol 2014 Apr 30. 229. Radisky DC, Levy DD, Littlepage LE, Liu H, Nelson CM, Fata JE, Leake D, Godden EL, Albertson DG, Nieto MA, et al. Rac1b and reactive oxygen species mediate MMP-3induced EMT and genomic instability. Nature 2005 Jul 7;436(7047):123-7. 230. Raimondi C, Gradilone A, Naso G, Vincenzi B, Petracca A, Nicolazzo C, Palazzo A, Saltarelli R, Spremberg F, Cortesi E, et al. Epithelial-mesenchymal transition and stemness features in circulating tumor cells from breast cancer patients. Breast Cancer Res Treat 2011 Nov;130(2):449-55. 231. Raja FA, Chopra N, Ledermann JA. Optimal first-line treatment in ovarian cancer. Ann Oncol 2012 Sep;23 Suppl 10:x118-27. 232. Rassweiler J, Seemann O, Hatzinger M, Schulze M, Frede T. Technical evolution of laparoscopic radical prostatectomy after 450 cases. J Endourol 2003 Apr;17(3):143-54. 233. Reis ST, Pontes-Junior J, Antunes AA, Sousa-Canavez JM, Abe DK, Cruz JA, Dall'oglio MF, Crippa A, Passerotti CC, Ribeiro-Filho LA, et al. Tgf-beta1 expression as a biomarker of poor prognosis in prostate cancer. Clinics (Sao Paulo) 2011;66(7):1143-7. 234. Reka AK, Kuick R, Kurapati H, Standiford TJ, Omenn GS, Keshamouni VG. Identifying inhibitors of epithelial-mesenchymal transition by connectivity map-based systems approach. J Thorac Oncol 2011 Nov;6(11):1784-92. 235. Remacle A, Murphy G, Roghi C. Membrane type I-matrix metalloproteinase (MT1MMP) is internalised by two different pathways and is recycled to the cell surface. J Cell Sci 2003 Oct 1;116(Pt 19):3905-16. 157 236. Ren S, Su C, Wang Z, Li J, Fan L, Li B, Li X, Zhao C, Wu C, Hou L, et al. Epithelial phenotype as a predictive marker for response to EGFR-TKIs in non-small cell lung cancer patients with wild-type EGFR. Int J Cancer 2014 Apr 28. 237. Ribeiro N, Sousa SR, Brekken RA, Monteiro FJ. Role of SPARC in bone remodeling and cancer-related bone metastasis. J Cell Biochem 2014 Jan;115(1):17-26. 238. Ridley AJ, Schwartz MA, Burridge K, Firtel RA, Ginsberg MH, Borisy G, Parsons JT, Horwitz AR. Cell migration: Integrating signals from front to back. Science 2003 Dec 5;302(5651):1704-9. 239. Rieber M, Strasberg-Rieber M. p53 inactivation decreases dependence on estrogen/ERK signalling for proliferation but promotes EMT and susceptility to 3-bromopyruvate in ERalpha+ breast cancer MCF-7 cells. Biochem Pharmacol 2014 Mar 15;88(2):169-77. 240. Rojas A, Liu G, Coleman I, Nelson PS, Zhang M, Dash R, Fisher PB, Plymate SR, Wu JD. IL-6 promotes prostate tumorigenesis and progression through autocrine crossactivation of IGF-IR. Oncogene 2011 May 19;30(20):2345-55. 241. Rokavec M, Oner MG, Li H, Jackstadt R, Jiang L, Lodygin D, Kaller M, Horst D, Ziegler PK, Schwitalla S, et al. IL-6R/STAT3/miR-34a feedback loop promotes EMT-mediated colorectal cancer invasion and metastasis. J Clin Invest 2014 Apr 1;124(4):1853-67. 242. Ross HM, Kryvenko ON, Cowan JE, Simko JP, Wheeler TM, Epstein JI. Do adenocarcinomas of the prostate with gleason score (GS) </=6 have the potential to metastasize to lymph nodes? Am J Surg Pathol 2012 Sep;36(9):1346-52. 243. Rotty JD, Wu C, Bear JE. New insights into the regulation and cellular functions of the ARP2/3 complex. Nat Rev Mol Cell Biol 2013 Jan;14(1):7-12. 244. Sabe H: Cancer early dissemination: cancerous epithelial–mesenchymal transdifferentiation and transforming growth factor β signaling. J Biochem 2011, 149(6):633-9. 245. Safina AF, Varga AE, Bianchi A, Zheng Q, Kunnev D, Liang P, Bakin AV. Ras alters epithelial-mesenchymal transition in response to TGFbeta by reducing actin fibers and cell-matrix adhesion. Cell Cycle 2009 Jan 15;8(2):284-98. 246. Sakr WA, Partin AW. Histological markers of risk and the role of high-grade prostatic intraepithelial neoplasia. Urology 2001 Apr;57(4 Suppl 1):115-20. 247. Samatov TR, Tonevitsky AG, Schumacher U. Epithelial-mesenchymal transition: Focus on metastatic cascade, alternative splicing, non-coding RNAs and modulating compounds. Mol Cancer 2013 Sep 23;12(1):107,4598-12-107. 248. Sanchez-Tillo E, Fanlo L, Siles L, Montes-Moreno S, Moros A, Chiva-Blanch G, Estruch R, Martinez A, Colomer D, Gyorffy B, et al. The EMT activator ZEB1 promotes tumor growth and determines differential response to chemotherapy in mantle cell lymphoma. Cell Death Differ 2014 Feb;21(2):247-57. 158 249. Sarrio D, Rodriguez-Pinilla SM, Hardisson D, Cano A, Moreno-Bueno G, Palacios J. Epithelial-mesenchymal transition in breast cancer relates to the basal-like phenotype. Cancer Res 2008 Feb 15;68(4):989-97. 250. Savagner P. The epithelial-mesenchymal transition (EMT) phenomenon. Ann Oncol 2010 Oct;21 Suppl 7:vii89-92. 251. Saxena M, Stephens MA, Pathak H, Rangarajen A: Transcription factors that mediate epithelial–mesenchymal transition lead to multidrug resistance by upregulating ABC transporters. Cell Death and Disease 2011, 2(6):e179. 252. Schaeffer D, Somarelli JA, Hanna G, Palmer GM, Garcia-Blanco MA. Cellular migration and invasion uncoupled: Increased migration is not an inexorable consequence of epithelial-to-mesenchymal transition. Mol Cell Biol 2014 Sep 15;34(18):3486-99. 253. Schlomm T, Iwers L, Kirstein P, Jessen B, Kollermann J, Minner S, Passow-Drolet A, Mirlacher M, Milde-Langosch K, Graefen M, et al. Clinical significance of p53 alterations in surgically treated prostate cancers. Mod Pathol 2008 Nov;21(11):1371-8. 254. Schmittgen T and Livak K: Analyzing real-time PCR data by the comparative CT method. Nature Protocols 2008, 3(6):1101-1108. 255. Schneider V, Krieger ML, Bendas G, Jaehde U, Kalayda GV. Contribution of intracellular ATP to cisplatin resistance of tumor cells. J Biol Inorg Chem 2013 Feb;18(2):165-74. 256. Schoumacher M, Goldman RD, Louvard D, Vignjevic DM. Actin, microtubules, and vimentin intermediate filaments cooperate for elongation of invadopodia. J Cell Biol 2010 May 3;189(3):541-56. 257. Schust J, Sperl B, Hollis A, Mayer TU, Berg T. Stattic: A small-molecule inhibitor of STAT3 activation and dimerization. Chem Biol 2006 Nov;13(11):1235-42. 258. Seidman JD, Yemelyanova A, Cosin JA, Smith A, Kurman RJ. Survival rates for international federation of gynecology and obstetrics stage III ovarian carcinoma by cell type: A study of 262 unselected patients with uniform pathologic review. Int J Gynecol Cancer 2012 Mar;22(3):367-71. 259. Shankar J, Messenberg A, Chan J, Underhill TM, Foster LJ, Nabi IR. Pseudopodial actin dynamics control epithelial-mesenchymal transition in metastatic cancer cells. Cancer Res 2010 May 1;70(9):3780-90. 260. Sharma A, Yeow WS, Ertel A, Coleman I, Clegg N, Thangavel C, Morrissey C, Zhang X, Comstock CE, Witkiewicz AK, et al. The retinoblastoma tumor suppressor controls androgen signaling and human prostate cancer progression. J Clin Invest 2010 Dec;120(12):4478-92. 261. Shi Y, Wu H, Zhang M, Ding L, Meng F, Fan X. Expression of the epithelialmesenchymal transition-related proteins and their clinical significance in lung 159 adenocarcinoma. Diagn Pathol 2013 May 24;8:89,1596-8-89. 262. Shin SY, Rath O, Zebisch A, Choo SM, Kolch W, Cho KH. Functional roles of multiple feedback loops in extracellular signal-regulated kinase and wnt signaling pathways that regulate epithelial-mesenchymal transition. Cancer Res 2010 Sep 1;70(17):6715-24. 263. Siddik ZH. Cisplatin: Mode of cytotoxic action and molecular basis of resistance. Oncogene 2003 Oct 20;22(47):7265-79. 264. Siemens H, Jackstadt R, Hunten S, Kaller M, Menssen A, Gotz U, Hermeking H. miR-34 and SNAIL form a double-negative feedback loop to regulate epithelial-mesenchymal transitions. Cell Cycle 2011 Dec 15;10(24):4256-71. 265. Sohayda C, Kupelian PA, Levin HS, Klein EA. Extent of extracapsular extension in localized prostate cancer. Urology 2000 Mar;55(3):382-6. 266. Solar P and Sytkowski AJ: Differentially expressed genes associated with Cisplatin resistance in human ovarian adenocarcinoma cell line A2780. Cancer Lett 2011, 309(1):11-8. 267. Son H, Moon A. Epithelial-mesenchymal transition and cell invasion. Toxicol Res 2010 Dec;26(4):245-52. 268. Soo K, O'Rourke MP, Khoo PL, Steiner KA, Wong N, Behringer RR, Tam PP: Twist Function Is Required for the Morphogenesis of the Cephalic Neural Tube and the Differentiation of the Cranial Neural Crest Cells in the Mouse Embryo. Dev Biol 2002, 247(2): 251-270. 269. Spencer ES, Johnston RB, Gordon RR, Lucas JM, Ussakli CH, Hurtado-Coll A, Srivastava S, Nelson PS, Porter CR. Prognostic value of ERG oncoprotein in prostate cancer recurrence and cause-specific mortality. Prostate 2013 Jun;73(9):905-12. 270. Sun T, Sun BC, Zhao XL, Zhao N, Dong XY, Che N, Yao Z, Ma YM, Gu Q, Zong WK, Liu ZY: Promotion of tumor cell metastasis and vasculogenic mimicry by way of transcription coactivation by Bcl-2 and TWIST1: A study of hepatocellular carcinoma. Hepatology 2011, DOI: 10.1002/hep.24543 271. Sun Y, Wang Y, Fan C, Gao P, Wang X, Wei G, Wei J. Estrogen promotes stemness and invasiveness of ER-positive breast cancer cells through Gli1 activation. Mol Cancer 2014 Jun 3;13:137,4598-13-137. 272. Tanaka H, Kono E, Tran CP, Miyazaki H, Yamashiro J, Shimomura T, Fazli L, Wada R, Huang J, Vessella RL, et al. Monoclonal antibody targeting of N-cadherin inhibits prostate cancer growth, metastasis and castration resistance. Nat Med 2010 Dec;16(12):1414-20. 273. Techasen A, Namwat N, Loilome W, Bungkanjana P, Khuntikeo N, Puapairoj A, Jearanaikoon P, Saya H, Yongvanit P. Tumor necrosis factor-alpha (TNF-alpha) stimulates the epithelial-mesenchymal transition regulator Snail in cholangiocarcinoma. 160 Med Oncol 2012 Dec;29(5):3083-91. 274. Thaiparambil JT, Bender L, Ganesh T, Kline E, Patel P, Liu Y, Tighiouart M, Vertino PM, Harvey RD, Garcia A, et al. Withaferin A inhibits breast cancer invasion and metastasis at sub-cytotoxic doses by inducing vimentin disassembly and serine 56 phosphorylation. Int J Cancer 2011 Dec 1;129(11):2744-55. 275. Thakur N, Gudey SK, Marcusson A, Fu JY, Bergh A, Heldin CH, Landstrom M. TGFbeta-induced invasion of prostate cancer cells is promoted by c-jun-dependent transcriptional activation of Snail1. Cell Cycle 2014 Jun 19;13(15). 276. Thomas R, True LD, Bassuk JA, Lange PH, Vessella RL. Differential expression of osteonectin/SPARC during human prostate cancer progression. Clin Cancer Res 2000 Mar;6(3):1140-9. 277. Thomson S, Petti F, Sujka-Kwok I, Epstein D, Haley JD. Kinase switching in mesenchymal-like non-small cell lung cancer lines contributes to EGFR inhibitor resistance through pathway redundancy. Clin Exp Metastasis 2008;25(8):843-54. 278. Timmerman LA, Grego-Bessa J, Raya A, Bertran E, Perez-Pomares JM, Diez J, Aranda S, Palomo S, McCormick F, Izpisua-Belmonte JC, et al. Notch promotes epithelialmesenchymal transition during cardiac development and oncogenic transformation. Genes Dev 2004 Jan 1;18(1):99-115. 279. Tsai JH, Donaher JL, Murphy DA, Chau S, Yang J. Spatiotemporal regulation of epithelial-mesenchymal transition is essential for squamous cell carcinoma metastasis. Cancer Cell 2012 Dec 11;22(6):725-36. 280. Tsai LL, Yu CC, Chang YC, Yu CH, Chou MY. Markedly increased Oct4 and nanog expression correlates with cisplatin resistance in oral squamous cell carcinoma. J Oral Pathol Med 2011 Sep;40(8):621-8. 281. Tung KH, Lin CW, Kuo CC, Li LT, Kuo YH, Lin CW, Wu HC. CHC promotes tumor growth and angiogenesis through regulation of HIF-1alpha and VEGF signaling. Cancer Lett 2013 Apr 30;331(1):58-67. 282. Ushijima K. Treatment for recurrent ovarian cancer-at first relapse. J Oncol 2010;2010:497429. 283. van Nes JG, de Kruijf EM, Putter H, Faratian D, Munro A, Campbell F, Smit VT, Liefers GJ, Kuppen PJ, van de Velde CJ, et al. Co-expression of SNAIL and TWIST determines prognosis in estrogen receptor-positive early breast cancer patients. Breast Cancer Res Treat 2012 May;133(1):49-59. 284. Vandewalle C, Comijn J, De Craene B, Vermassen P, Bruyneel E, Andersen H, Tulchinsky E, Van Roy F, Berx G. SIP1/ZEB2 induces EMT by repressing genes of different epithelial cell-cell junctions. Nucleic Acids Res 2005 Nov 24;33(20):6566-78. 285. Vang R, Shih I, Kurman RJ. Ovarian low-grade and high-grade serous carcinoma: Pathogenesis, clinicopathologic and molecular biologic features, and diagnostic 161 problems. Adv Anat Pathol 2009 Sep;16(5):267-82. 286. Vargas DA, Bates O, Zaman MH. Computational model to probe cellular mechanics during epithelial-mesenchymal transition. Cells Tissues Organs 2013;197(6):435-44. 287. Vazquez-Martin A, Cufi S, Oliveras-Ferraros C, Torres-Garcia VZ, Corominas-Faja B, Cuyas E, Bonavia R, Visa J, Martin-Castillo B, Barrajon-Catalan E, et al. IGF1R/epithelial-to-mesenchymal transition (EMT) crosstalk suppresses the erlotinibsensitizing effect of EGFR exon 19 deletion mutations. Sci Rep 2013;3:2560. 288. Vega S, Morales AV, Ocaña OH, Valdés F, Fabregat I, Nieto MA: Snail blocks the cell cycle and confers resistance to cell death. Genes Dev 2004, 18(10): 1131– 1143. 289. Vencken PM, Kriege M, Hoogwerf D, Beugelink S, van der Burg ME, Hooning MJ, Berns EM, Jager A, Collee M, Burger CW, et al. Chemosensitivity and outcome of BRCA1- and BRCA2-associated ovarian cancer patients after first-line chemotherapy compared with sporadic ovarian cancer patients. Ann Oncol 2011 Jun;22(6):1346-52. 290. Venderbos LD, van den Bergh RC, Roobol MJ, Schroder FH, Essink-Bot ML, Bangma CH, Steyerberg EW, Korfage IJ. A longitudinal study on the impact of active surveillance for prostate cancer on anxiety and distress levels. Psychooncology 2014 Aug 20. 291. Vergote I, Trope CG, Amant F, Kristensen GB, Ehlen T, Johnson N, Verheijen RH, van der Burg ME, Lacave AJ, Panici PB, et al. Neoadjuvant chemotherapy or primary surgery in stage IIIC or IV ovarian cancer. N Engl J Med 2010 Sep 2;363(10):943-53. 292. Verhaak RG, Tamayo P, Yang JY, Hubbard D, Zhang H, Creighton CJ, Fereday S, Lawrence M, Carter SL, Mermel CH, et al. Prognostically relevant gene signatures of high-grade serous ovarian carcinoma. J Clin Invest 2013 Jan 2;123(1):517-25. 293. Vermeulen L, De Sousa E Melo F, van der Heijden M, Cameron K, de Jong JH, Borovski T, Tuynman JB, Todaro M, Merz C, Rodermond H, et al. Wnt activity defines colon cancer stem cells and is regulated by the microenvironment. Nat Cell Biol 2010 May;12(5):468-76. 294. Villedieu M, Louis MH, Dutoit S, Brotin E, Lincet H, Duigou F, Staedel C, Gauduchon P, Poulain L. Absence of bcl-xL down-regulation in response to cisplatin is associated with chemoresistance in ovarian carcinoma cells. Gynecol Oncol 2007 Apr;105(1):31-44. 295. Vincent AJ, Ren S, Harris LG, Devine DJ, Samant RS, Fodstad O, Shevde LA. Cytoplasmic translocation of p21 mediates NUPR1-induced chemoresistance: NUPR1 and p21 in chemoresistance. FEBS Lett 2012 Sep 21;586(19):3429-34. 296. Visintin I, Feng Z, Longton G, Ward DC, Alvero AB, Lai Y, Tenthorey J, Leiser A, Flores-Saaib R, Yu H, Azori M, Rutherford T, Schwartz PE, Mor G: Diagnostic markers for early detection of ovarian cancer. Clin Cancer Res 2008, 14(4):1065-1072. 297. Von Hoff DD, Ervin T, Arena FP, Chiorean EG, Infante J, Moore M, Seay T, Tjulandin SA, Ma WW, Saleh MN, et al. Increased survival in pancreatic cancer with nab162 paclitaxel plus gemcitabine. N Engl J Med 2013 Oct 31;369(18):1691-703. 298. Vuoriluoto K, Haugen H, Kiviluoto S, Mpindi JP, Nevo J, Gjerdrum C, Tiron C, Lorens JB, Ivaska J. Vimentin regulates EMT induction by Slug and oncogenic H-ras and migration by governing axl expression in breast cancer. Oncogene 2011 Mar 24;30(12):1436-48. 299. Waggott D, Chu K, Yin S, Wouters BG, Liu FF, Boutros PC. NanoStringNorm: An extensible R package for the pre-processing of NanoString mRNA and miRNA data. Bioinformatics 2012 Jun 1;28(11):1546-8. 300. Wang H, Zhang G, Zhang H, Zhang F, Zhou B, Ning F, Wang HS, Cai SH, Du J. Acquisition of epithelial-mesenchymal transition phenotype and cancer stem cell-like properties in cisplatin-resistant lung cancer cells through AKT/beta-catenin/Snail signaling pathway. Eur J Pharmacol 2014 Jan 15;723:156-66. 301. Wang Q, Hu DF, Rui Y, Jiang AB, Liu ZL, Huang LN. Prognosis value of HIF-1alpha expression in patients with non-small cell lung cancer. Gene 2014 May 15;541(2):69-74. 302. Wang T, Li Y, Wang W, Tuerhanjiang A, Wu Z, Yang R, Yuan M, Ma D, Wang W, Wang S. Twist2, the key twist isoform related to prognosis, promotes invasion of cervical cancer by inducing epithelial-mesenchymal transition and blocking senescence. Hum Pathol 2014 Sep;45(9):1839-46. 303. Wang Y, Zhang YX, Kong CZ, Zhang Z, Zhu YY. Loss of P53 facilitates invasion and metastasis of prostate cancer cells. Mol Cell Biochem 2013 Dec;384(1-2):121-7. 304. Watanabe A, Hoshino D, Koshikawa N, Seiki M, Suzuki T, Ichikawa K. Critical role of transient activity of MT1-MMP for ECM degradation in invadopodia. PLoS Comput Biol 2013;9(5):e1003086. 305. Weber DC, Tille JC, Combescure C, Egger JF, Laouiti M, Hammad K, Granger P, Rubbia-Brandt L, Miralbell R. The prognostic value of expression of HIF1alpha, EGFR and VEGF-A, in localized prostate cancer for intermediate- and high-risk patients treated with radiation therapy with or without androgen deprivation therapy. Radiat Oncol 2012 Apr 30;7:66,717X-7-66. 306. Wei J, Xu G, Wu M, Zhang Y, Li Q, Liu P, Zhu T, Song A, Zhao L, Han Z, et al. Overexpression of vimentin contributes to prostate cancer invasion and metastasis via src regulation. Anticancer Res 2008 Jan-Feb;28(1A):327-34. 307. Wells A, Yates C, Shepard CR. E-cadherin as an indicator of mesenchymal to epithelial reverting transitions during the metastatic seeding of disseminated carcinomas. Clin Exp Metastasis 2008;25(6):621-8. 308. Welm AL, Sneddon JB, Taylor C, Nuyten DS, van de Vijver MJ, Hasegawa BH, Bishop JM. The macrophage-stimulating protein pathway promotes metastasis in a mouse model for breast cancer and predicts poor prognosis in humans. Proc Natl Acad Sci U S A 2007 May 1;104(18):7570-5. 163 309. Wen Y, Eng CH, Schmoranzer J, Cabrera-Poch N, Morris EJ, Chen M, Wallar BJ, Alberts AS, Gundersen GG. EB1 and APC bind to mDia to stabilize microtubules downstream of rho and promote cell migration. Nat Cell Biol 2004 Sep;6(9):820-30. 310. Whiteland H, Spencer-Harty S, Thomas DH, Davies C, Morgan C, Kynaston H, Bose P, Fenn N, Lewis PD, Bodger O, et al. Putative prognostic epithelial-to-mesenchymal transition biomarkers for aggressive prostate cancer. Exp Mol Pathol 2013 Oct;95(2):220-6. 311. Whiteman EL, Liu CJ, Fearon ER, Margolis B. The transcription factor Snail represses Crumbs3 expression and disrupts apico-basal polarity complexes. Oncogene 2008 Jun 19;27(27):3875-9. 312. Wikstrom P, Stattin P, Franck-Lissbrant I, Damber JE, Bergh A. Transforming growth factor beta1 is associated with angiogenesis, metastasis, and poor clinical outcome in prostate cancer. Prostate 1998 Sep 15;37(1):19-29. 313. Windus LC, Kiss DL, Glover T, Avery VM. In vivo biomarker expression patterns are preserved in 3D cultures of prostate cancer. Exp Cell Res 2012 Nov 15;318(19):2507-19. 314. Wintzell M, Lofstedt L, Johansson J, Pedersen AB, Fuxe J, Shoshan M. Repeated cisplatin treatment can lead to a multiresistant tumor cell population with stem cell features and sensitivity to 3-bromopyruvate. Cancer Biol Ther 2012 Dec;13(14):1454-62. 315. Wong SY, Crowley D, Bronson RT, Hynes RO. Analyses of the role of endogenous SPARC in mouse models of prostate and breast cancer. Clin Exp Metastasis 2008;25(2):109-18. 316. Wu K and Bonavida B: The activated NF-kappaB-Snail-RKIP circuitry in cancer regulates both the metastatic cascade and resistance to apoptosis by cytotoxic drugs. Crit Rev Immunol 2009, 29(3):241-54. 317. Yang D and Wang AH: Structural studies of interactions between anticancer platinum drugs and DNA. Prog Biophys Mol Biol 1996, 66(1):81-111. 318. Yang MH, Wu MZ, Chiou SH, Chen PM, Chang SY, Liu CJ, Teng SC, Wu KJ. Direct regulation of TWIST by HIF-1alpha promotes metastasis. Nat Cell Biol 2008 Mar;10(3):295-305. 319. Yen MS, Juang CM, Lai CR, Chao GC, Ng HT, Yuan CC: Intraperitoneal Cisplatinbased chemotherapy vs. intravenous Cisplatin-based chemotherapy for stage III optimally cytoreduced epithelial ovarian cancer. Int J Gynaecol Obstet 2001, 72(1):55-60. 320. Xi Y, Tan K, Brumwell AN, Chen SC, Kim YH, Kim TJ, Wei Y, Chapman HA. Inhibition of epithelial-to-mesenchymal transition and pulmonary fibrosis by methacycline. Am J Respir Cell Mol Biol 2014 Jan;50(1):51-60. 321. Xiao D, He J. Epithelial mesenchymal transition and lung cancer. J Thorac Dis 2010 Sep;2(3):154-9. 164 322. Xiao H, Verdier-Pinard P, Fernandez-Fuentes N, Burd B, Angeletti R, Fiser A, Horwitz SB, Orr GA. Insights into the mechanism of microtubule stabilization by taxol. Proc Natl Acad Sci U S A 2006 Jul 5;103(27):10166-73. 323. Xie G, Ji A, Yuan Q, Jin Z, Yuan Y, Ren C, Guo Z, Yao Q, Yang K, Lin X, et al. Tumour-initiating capacity is independent of epithelial-mesenchymal transition status in breast cancer cell lines. Br J Cancer 2014 May 13;110(10):2514-23. 324. Xu J, Wang R, Xie ZH, Odero-Marah V, Pathak S, Multani A, Chung LW, Zhau HE. Prostate cancer metastasis: Role of the host microenvironment in promoting epithelial to mesenchymal transition and increased bone and adrenal gland metastasis. Prostate 2006 Nov 1;66(15):1664-73. 325. Xu MH, Gao X, Luo D, Zhou XD, Xiong W, Liu GX. EMT and acquisition of stem celllike properties are involved in spontaneous formation of tumorigenic hybrids between lung cancer and bone marrow-derived mesenchymal stem cells. PLoS One 2014 Feb 6;9(2):e87893. 326. Yang CC, Zhu LF, Xu XH, Ning TY, Ye JH, Liu LK. Membrane type 1 matrix metalloproteinase induces an epithelial to mesenchymal transition and cancer stem celllike properties in SCC9 cells. BMC Cancer 2013 Apr 1;13:171,2407-13-171. 327. Yang J and Weinberg R: Epithelial-Mesenchymal Transition: At the Crossroads of Development and Tumor Metastasis. Dev Cell 2008, 14(6):818-829. 328. Yang Z, Garcia A, Xu S, Powell DR, Vertino PM, Singh S, Marcus AI. Withania somnifera root extract inhibits mammary cancer metastasis and epithelial to mesenchymal transition. PLoS One 2013 Sep 12;8(9):e75069. 329. Yao Z, Fenoglio S, Gao DC, Camiolo M, Stiles B, Lindsted T, Schlederer M, Johns C, Altorki N, Mittal V, et al. TGF-beta IL-6 axis mediates selective and adaptive mechanisms of resistance to molecular targeted therapy in lung cancer. Proc Natl Acad Sci U S A 2010 Aug 31;107(35):15535-40. 330. Yardley DA. Nab-paclitaxel mechanisms of action and delivery. J Control Release 2013 Sep 28;170(3):365-72. 331. Yilmaz M, Christofori G. EMT, the cytoskeleton, and cancer cell invasion. Cancer Metastasis Rev 2009 Jun;28(1-2):15-33. 332. Yiu GK, Chan WY, Ng SW, Chan PS, Cheung KK, Berkowitz RS, Mok SC. SPARC (secreted protein acidic and rich in cysteine) induces apoptosis in ovarian cancer cells. Am J Pathol 2001 Aug;159(2):609-22. 333. Yook JI, Li XY, Ota I, Hu C, Kim HS, Kim NH, Cha SY, Ryu JK, Choi YJ, Kim J, et al. A wnt-Axin2-GSK3beta cascade regulates Snail1 activity in breast cancer cells. Nat Cell Biol 2006 Dec;8(12):1398-406. 334. Yoshimoto M, Cutz JC, Nuin PA, Joshua AM, Bayani J, Evans AJ, Zielenska M, Squire JA. Interphase FISH analysis of PTEN in histologic sections shows genomic deletions in 165 68% of primary prostate cancer and 23% of high-grade prostatic intra-epithelial neoplasias. Cancer Genet Cytogenet 2006 Sep;169(2):128-37. 335. Yoshimoto M, Cunha IW, Coudry RA, Fonseca FP, Torres CH, Soares FA, Squire JA. FISH analysis of 107 prostate cancers shows that PTEN genomic deletion is associated with poor clinical outcome. Br J Cancer 2007 Sep 3;97(5):678-85. 336. Yu L, Mu Y, Sa N, Wang H, Xu W. Tumor necrosis factor alpha induces epithelialmesenchymal transition and promotes metastasis via NF-kappaB signaling pathwaymediated TWIST expression in hypopharyngeal cancer. Oncol Rep 2014 Jan;31(1):3217. 337. Yu Y, Xiao CH, Tan LD, Wang QS, Li XQ, Feng YM. Cancer-associated fibroblasts induce epithelial-mesenchymal transition of breast cancer cells through paracrine TGFbeta signalling. Br J Cancer 2014 Feb 4;110(3):724-32. 338. Yun JA, Kim SH, Hong HK, Yun SH, Kim HC, Chun HK, Cho YB, Lee WY. Loss of Ecadherin expression is associated with a poor prognosis in stage III colorectal cancer. Oncology 2014;86(5-6):318-28. 339. Zaritsky A, Natan S, Ben-Jacob E, Tsarfaty I. Emergence of HGF/SF-induced coordinated cellular motility. PLoS One 2012;7(9):e44671. 340. Zeegers MP, Jellema A, Ostrer H. Empiric risk of prostate carcinoma for relatives of patients with prostate carcinoma: A meta-analysis. Cancer 2003 Apr 15;97(8):1894-903. 341. Zeisberg M and Neilson E: Biomarkers for epithelial-mesenchymal transitions. J Clin Invest 2009, 119(6):1429–1437. 342. Zhang C, Yang X, Zhang Q, Guo Q, He J, Qin Q, Zhu H, Liu J, Zhan L, Lu J, et al. STAT3 inhibitor NSC74859 radiosensitizes esophageal cancer via the downregulation of HIF-1alpha. Tumour Biol 2014 Jul 1. 343. Zhang K, Chen D, Jiao X, Zhang S, Liu X, Cao J, Wu L, Wang D. Slug enhances invasion ability of pancreatic cancer cells through upregulation of matrix metalloproteinase-9 and actin cytoskeleton remodeling. Lab Invest 2011 Mar;91(3):42638. 344. Zhang P, Chang WH, Fong B, Gao F, Liu C, Al Alam D, Bellusci S, Lu W. Regulation of induced pluripotent stem (iPS) cell induction by Wnt/beta-catenin signaling. J Biol Chem 2014 Mar 28;289(13):9221-32. 345. Zhang Q, Helfand BT, Jang TL, Zhu LJ, Chen L, Yang XJ, Kozlowski J, Smith N, Kundu SD, Yang G, et al. Nuclear factor-kappaB-mediated transforming growth factor-betainduced expression of vimentin is an independent predictor of biochemical recurrence after radical prostatectomy. Clin Cancer Res 2009 May 15;15(10):3557-67. 346. Zhang Q, Bai X, Chen W, Ma T, Hu Q, Liang C, Xie S, Chen C, Hu L, Xu S, et al. Wnt/beta-catenin signaling enhances hypoxia-induced epithelial-mesenchymal transition in hepatocellular carcinoma via crosstalk with hif-1alpha signaling. Carcinogenesis 2013 166 May;34(5):962-73. 347. Zhao N, Sun BC, Zhao XL, Liu ZY, Sun T, Qiu ZQ, Gu Q, Che N, Dong XY. Coexpression of bcl-2 with epithelial-mesenchymal transition regulators is a prognostic indicator in hepatocellular carcinoma. Med Oncol 2012 Dec;29(4):2780-92. 348. Zhao XL, Sun T, Che N, Sun D, Zhao N, Dong XY, Gu Q, Yao Z, Sun BC: Promotion of hepatocellular carcinoma metastasis through matrix metalloproteinase activation by epithelial-mesenchymal transition regulator TWIST1. J Cell Mol Med 2011, 15(3):691-700. 349. Zhao Y, Yan Q, Long X, Chen X, Wang Y. Vimentin affects the mobility and invasiveness of prostate cancer cells. Cell Biochem Funct 2008 Sep-Oct;26(5):571-7. 350. Zhao Z, Lu P, Zhang H, Xu H, Gao N, Li M, Liu C. Nestin positively regulates the Wnt/ss-catenin pathway and the proliferation, survival, and invasiveness of breast cancer stem cells. Breast Cancer Res 2014 Jul 24;16(4):408. 351. Zhao ZS, Wang YY, Chu YQ, Ye ZY, Tao HQ. SPARC is associated with gastric cancer progression and poor survival of patients. Clin Cancer Res 2010 Jan 1;16(1):260-8. 352. Zheng H, Kang Y. Multilayer control of the EMT master regulators. Oncogene 2014 33, 1755–1763. 353. Zhu GH, Huang C, Feng ZZ, Lv XH, Qiu ZJ. Hypoxia-induced Snail expression through transcriptional regulation by HIF-1alpha in pancreatic cancer cells. Dig Dis Sci 2013 Dec;58(12):3503-15. 354. Zlobec I and Lugli A: Epithelial mesenchymal transition and tumor budding in aggressive colorectal cancer: tumor budding as oncotarget. Oncotarget 2010, 1(7):651-61. 167 Apendices Appendix A: Supplementary Materials (Chapter 3) Supplementary Table S3.1 Clinical data for ovarian tumour samples Sample ID Age Stage PFI (months) Classification 1100 D00443 1240 1299 1359 1413 1587 1605 1680 1703 1776 1157 1224 1296 1304 1308 1351 1355 1381 1561 1625 1627 1706 B01440 B01360 54 51 64 53 62 47 54 51 67 48 58 61 61 51 49 55 68 51 63 54 52 61 53 54 IIIc III III IIIb IIIc IV IIIc IIb IIc IIIc IIIc IIIc II IIIa IIIc IV IIa IIIa IIb IIIb IIIa IIIa IIIc IIIc 5 5 <3 3 8 <3 0 3 7 6 6 36 22 39 26 25 No recurrence 23 25 25 26 No recurrence No recurrence 20 19 Resistant Resistant Resistant Resistant Resistant Resistant Resistant Resistant Resistant Resistant Resistant Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Sensitive Available clinical data for each tissue sample including; patient's age at time of diagnosis (years), tumour stage, progression-free interval (PFI, months) and drug response classification. 168 Appendix B: Detailed report on the approach to multivariate model identification Analysis was completed using the normalized data generated by Dr. R Gooding. The data consisted of 27 prostrate samples with Gleason pattern 3 and 29 samples with Gleason pattern 4. Intensities from 68 genes were captured for each of the 56 samples. All intensities were analyzed and reported on the log base 2 scale. Intensities below the detectable threshold have a value of 0 (1 prior to log transformation). CDH2 was found to have no non-zero values and was thus excluded from the remainder of the analysis. Following the generation of a correlation matrix between all possible gene pairs within the target panel, it was found that many of the genes are strongly correlated. Univariate p-values were calculated by the non-parametric Mann-Whitney U test (a.k.a. Wilcoxon Rank Sum test) and identified 19 statistically significant genes distinguishing between patterns 3 and 4. ROC-AUC values for individual genes, and False discovery rates accounting for the number of tests (p=67), were also generated. Thirteen genes were found to have false discovery rates below 10%. However, given the sample size this provides suggestive, but not quite statistically significant, evidence that some of the gene intensities differed between Gleason 3 and 4. Even if the differences in intensities are true, the modest AUC and substantial overlap of intensities between Gleason patterns indicates that no single gene is a strong discriminator between Gleason patterns 3 and 4. The Lasso approach was first used in an attempt to select a multivariable model for predicting Gleason pattern 4 vs. 3. Using this method, a logistic model using the following five gene intensities was selected to predict Gleason pattern: MST1R, MUC1, SORL1, TWIST2 and WNT5A. Applying regular logistic regression to these five predictors, an ROC AUC of 0.89 was obtained. Application of Lasso shrinkage resulted in an ROC AUC of 0.82. While the Lasso method was applied to help reduce model optimism caused by a large number of predictors (p=67) relative to a small sample size (n=56), it can be expected that this AUC is highly 169 optimistic and that this model is likely fitting random noise unique to the population and will not validate in an independent cohort. Following the generation of a correlation matrix, it was observed that there was a high degree of correlation in signal intensities amongst the gene targets. Therefore, it was considered to be possible to represent most of the information from the 67 gene intensities in a small number of principal components (PCs). In fact, the first 3 PCs explain 63% of the total variance among the 67 genes. Thus, with some loss of information, we can reduce this model selection problem from 67 dimensions to 3. The PCs are merely linearly transformations of the gene intensities such that the first PC (linear transformation) has as much variance as possible, the second has as much variance as possible while being orthogonal to the first, and the third has the most variance possible while being orthogonal to the 1st and 2nd, and so forth. The PCs do not consider the Gleason pattern in their creation but are merely an alternative way to express the intensities. Three PCs were selected because the sample size supports approximately 3 predictors and there was little additional variance retained by adding one or two more PCs. A logistic regression model including the first 3 PCs as variables significantly predicted Gleason pattern with a model likelihood ratio test p-value of 0.002 and an ROC AUC of 0.77. Although this ROC is not as high as that obtained by the Lasso regression, it should be appreciated that the PCs were not selected to optimize the AUC in this sample, unlike the Lasso logistic regression. Thus, while the Lasso logistic regression AUC is likely highly optimistic there is no reason to suspect the same of the ROC AUC from the PCs. Independently, the 3rd PC is most predictive of Gleason pattern, with a p-value of 0.0067, followed by the 1st with a p-value of p=0.0593. The 2nd PC does not appear to predict Gleason pattern at all (p-value= p=0.6839). Figure 4.5 displays the probability of being Gleason pattern 4, as estimated by a non-parametric loess smother over the range (technically the domain) of PC1 and PC3. This clearly demonstrates 170 that PC3 is the strongest predictor, such that samples with high PC3 values are more likely to be Gleason pattern 4. Unfortunately, each PC, including PC3, is a linear combination of intensities from all 67 genes. By identifying those genes whose intensities correlate most substantially with PC3, five were found to have a correlation of 0.4 or higher (Table 4.1). By simply adding the three genes with positive correlations and subtracting those with negative correlations (MST1RTWIST2+AHNAK+FN1-POU5F1) a new variable was created that has a correlation of 0.87 with PC3. When used to predict Gleason pattern 4, this variable achieves an AUC of 0.70 (p=0.003). Appendix C: Supplementary Materials (Chapter 4) Supplementary Table S4.1 NanoString probe sequences Gene Symbol NanoString Probe Sequences Capture Probe Sequence Reporter Probe Sequence ACTA2 CCAGCAAAGCCGGCCTTACAGAGCCCAGAGCCATTG ACAATGGATGGGAAAACAGCCCTGGGAGCATCGTCC AHNAK GCAGGTTGTCAAAGTAGATGGTGGCACCCACAATCT CATGGTGTTCAGCAGCTGGGTCACCTCACCCGACT AKT1 AATCGGATTGTTCTGAGGGCTGAGGCCACACCCGGA GCATAGCTGCAGAAGTCCTTAACATTTCCCTACGTG AKT2 CATCGAAGTACCTTGTGTCGACCTCGGACGTGACCT GGGGTGTGATTGTGATGGACTGGGCGGTAAATTCAT AKT3 ATTCTGGAGTGCCACAGAATGTCTTCATGGTGGCTG CATAGTCATTATCTTCTAACACCTCTGGTGCCAGAT AURKA ACTGGAGCTGTAGCCTTAACAGGTCCTGAAATGCAG AATTGCTGAGTCACGAGAACACGTTTTGGACCTCCA BIRC3 TTTTTCTCTGAACTGTCTGTTTTACCAGGCTTCTAC CATTGACTAGTCTATAATTCTCTCCAGTTGCTAGGA BMP7 ACCGGAACTCTCGATGGTGGTAGCGTGGGTGGAAGA TGACAGCTTCCCCTTCTGGGATCTTGGAAAGATCAA CAMK2N1 AGGTTGTTGATTTCATCGTGGGTAGCAAGCTAGTAA TCGCCTCATCTGTCTCCCGGCCTGATACCAGATAC CD82 CGATTCATGAGCTCAGCGTTGTCTGTCCAGTTGTAG TTGACTTCGCAGGAACAGGGGTAGGTGACCTCAGGG CD83 GGAGTCACTAGCCCTAAATGCTTATTTGGGGAGGTA TGCAGAAATCCTGCTCATACCAGTTCTGTCTTGTGA CDH1 CTCGGACACTTCCACTCTCTTTTCAGGAGGCACAAA AGTGTAGGATGTGATTTCCTGGCCCACGCCAAAGTC CDH2 TATGATGTCTACCCTGTTCTCAGGAACTTCACCATA GGGTTGATCCTTATCGGTCACAGTTAGATTAGCTAC CFC1 GGGAAGCGACCTGAGAAGCGGTGCTCGGAAGACAGA CTGGACTGGTTGTGGACACGACAGCCTCCTCAGATT CLTC ATATCAGTTCTGTCTCCACATAGGACTCTCGAGCCT CTAACTCTGCAAGGCGGTTTGTTTTAGCCAGTGCGA COL1A1 AGAGGACGCAGGACAGACTAGGAGGGAGCCGGGAGG TTGTGCTTTGGGAAGTTGTCTCTGAAACCCGGGGAC 171 COL3A1 CATTCCTTGCAGACCAGGAGTACCAGCAGCACCAG TGGACCAGGACTTCCAAGACCTCCTCTTTCTCCAGG CREBBP CGAACCTCTCCGTTTGCTTGCTCTCGTCTCTGACAC TTTTTCATGGTTCGACAATGCGGGAGCGAGCAGGCC CTNNB1 CCACCAGCTAAACGCACTGCCATTTTAGCTCCTTCT TTTGTTTTGTTGAGCAAGGCAACCATTTTCTGCAGC DVL1 GATGTCTTTCCATGTTGAGCGTGACAGTGACGATGT CGTTGCTCTGCCCCACGATGCTGATGCCCAGAAAGT DVL2 CAGATTCTCATGGCTGCTGGACACATTAGGGTGGAA CAGTGACACTACTGACTCGGTTTCTGTCTCAGGCTC EGF TATTTTCTCTGATGTCTCCAACAGAGCCTTCACTCC CAGCCGCTTATCAAGCACATCCAATGACACAGCTGT EGFR TGAAGGAGTCACCCCTAAATGCCACCGGCAGGATGT TATCCAGTTCCTGTGGATCCAGAGGAGGAGTATGTG EPAS1 GTTCGCAGGAAGCTGATTGCCAGTCGCATGATGGAG TCGTTTTCAGAGCAAACTGAGGAGAGGAGCTTGTGT ERBB3 CACCACGCGGAGGTTGGGCAATGGTAGAGTAGAGAA GATGGCAAACTTCCCATCGTAGACCTGGGTCCCTCG FBN1 AATCAGTAACGTTTACTGGCAGCACCCTTGGTGGCT GTCCATTTTGACAGAGATAGCGGACCAACTGGCAGT FGFBP1 TCATCCTTGAGCTTTAGGCATGAGGTTGGATTGCCA CGCAGATTCCGGGCAACTTGTTTCCAATAGACTCTC FN1 ATTCGAAGTTGTGCTGCACCAAAGATGTCCGTCCTG GTCTGTGCAGAAAGAGTATTTCTGGTCCTGCTCATA FOXC2 CGCACGTTGGGGAAAGTTTGCTGCTGGGCCGCGAAC TTCTCAATCCCCAGCCGGTGGGAGTTGAACATCTCC FOXD3 AGAATGGGCGCGATGGTGGCAGTCGTCCGGCTCAG GCGGGCTGCAGAAACTGTCCGGAGAGTGGCACGCTA FZD7 ATCTGCTGATCGCCTCCCAACCCAACCACACAGAGA CCCATGGTTCAAACCTTCCTCTTCGTTCACTATGGT GAPDH GCCCCACTTGATTTTGGAGGGATCTCGCTCCTGGAA CCAGTGGACTCCACGACGTACTCAGCGCCAGCATC GSC AGCTGTCCGAGTCCAAATCGCTTTTACCTTCCTCTT TAGGTAAGTAATACGGGCAAGTGTCCCGCGGCCGTC GSK3B AGCCAACACACAGCCAGCAGACCATACATCTATACT ATCCCCTGGAAATATTGGTTGTCCTAGTAACAGCTC GUSB TAGTAGCCAGCAGATTCTAGGTGGGACGCAGGCTCG TCCAAGGATTTGGTGTGAGCGATCACCATCTTCAAG HIF1Α ACTGTTGGTATCATATACGTGAATGTGGCCTGTGCA CATAGGTGGTTTCTTATACCCACACTGAGGTTGGTT HPRT1 CATCACATCTCGAGCAAGACGTTCAGTCCTGTCCAT ACAGAGGGCTACAATGTGATGGCCTCCCATCTCCTT HPSE CATCTTAGCCGTCTTTCTTCGAGGCTGACCAACATC AATCACTTCTCCACCAGCCTTCAGGAAGCTCTTCAG IGF1 AAGCCCCTGTCTCCACACACGAACTGAAGAGCATCC CTGCTGGAGCCATACCCTGTGGGCTTGTTGAAATAA IGF1R GGCAGTTCTGAAGATCCACTGAGGTACAGGAGGCTT TTTACTGCAGAGAAGCTGTCTCCCGCGGGCAGCAAG IGF2 AACGCCTCGAGCTCCTTGGCGAGCACGTGACCCCG AGAGCAATCAGGGGACGGTGACGTTTGGCCTCCCTG IGFBP3 AGCACATTGAGGAACTTCAGGTGATTCAGTGTGTCT TTGTCACAGTTGGGAATGTGTACACCCCTGGGACTC IGFBP4 GCTACAGGGAGCGAGCCACTGGAAGGATAGGTTCAT AGGTGTAGGGGAAGGAGATATGGAGAGGGAGGCAGA IL17RC CCCTAGGGCAGCCTCGAAGCAGTCATATACCACAGA CCTGGGCTGAGTATAGGACCAGATTCGTACCTCACT IL6 CAGTGCCTCTTTGCTGCTTTCACACATGTTACTCTT TTCAGCCATCTTTGGAAGGTTCAGGTTGTTTTCTGC IL6ST CTGCCATTCGTACCATGTACAATGTGTCACTAGTCA ATTCTGGACCATCCTTCCCACCTTCATCTGTGTATG ILK CATTGTAGAGGGATCCATACGGCATCCAGTGTGTGA TCTGGTCCACGACGAAATTGGTGCCTTCATGTAGTA ITGA5 CCGTAACTCTGGTCACATATAGGAGCTGCTGACCTT TTGGGTTAATGGGGTGATTGGTGGTGCAGTTGAGTC 172 ITGB1 TTCACAATGTCTACCAACACGCCCTTCATTGCACCT ATCCATGTCTTCACTGTTAACTTCATCTGTGCTGCA JAG1 CTTACAGGATTTGGCGTTTACACAAGGTTTGGCCTC GGGAAGACAGTCGCAGTAGTAGCTGGCAATGAGATT KRT7 CCAGCAATCTGGGCCTCAAAGATGTCTGGGAGGCGG CATCCACCTGCAGTGCCTCAAGCTGACCCCGAAGG MIOS AAGGGTCCACTCAGGTCCATTCACTTTCAAGTTTAA CAGGTTTGGTACCGCTCATGTTTACTGATGTGATCA MITF TATAGTCCACGGATGCTTTTAAGATGGTTCCCTTGT CTTTTGCGCGTTGCTGTTCTCGTTGCAACTTTCGGA MLPH CAGGAGTATCAGCCTCCCAGCAACACTTTGAGATTC AGCTGCTAATGTGTGCTCACAGGTCTCCTGAAGTTG MMP3 GGGGCATAGGCATGGGCCAAAACATTTCCAGGTCCA TCATCAAAGTGGGCATCTCCATTAATCCCTGGCCCA MMP9 GGCACTGAGGAATGATCTAAGCCCAGCGCGTGGCCG TCAGTGAAGCGGTACATAGGGTACATGAGCGCCTCC MST1R AAAGCCCGGTGCGAATCCCGAGGCGTCAGCCTTGAG GGAGACAAGATGCTTGGGCAGCACTGACAACGCCAC MUC1 ACCTTCTCATAGGGGCTACGATCGGTACTGCTAGGG GTGTAAGAGAGGCTGCTGCCACCATTACCTGCAGAA NANOG TCAGGGCTGTCCTGAATAAGCAGATCCATGGAGGAA GAAGTGGGTTGTTTGCCTTTGGGACTGGTGGAAGAA NDRG1 TTTCCCCCTTCCTGTTTTGCCTGTTTAAAAGAGGTG TTCCAAATGCCTCTAGCCTCGACATGAATCCCACCT NOTCH1 CCGGGTGGACACAGGCAGGTGAACGAGTTGATGCCG TCATTGACATCGTGCTGGCAGTAGCTGCCCGTGAAG NUPR1 CCCCTCAGAGACTCAGTCAGCGGGAATAAGTCCTAG CACTTTTGCTAGGGTTGGAGGCGCTTTCCTGGTAG PAK1 GATGTAGCCACGTCCCGAGTTGGAGTGACAGGAAGT GGTGGAGTGGTGTTATTTTCAGTAGGTGAAATGGGA PGK1 TGACAGCCTCAGCATACTTCTTGCTGCTTTCAGGAC CCACAGGACCATTCCACACAATCTGCTTAGCCCGAG PIK3CA GCTAAATCCTGCTTCTCGGGATACAGACCAATTGGC TCTAGCTAGTCTGTTACTCAGTCCTGCGTGGGAATA PLEK2 GCTAATTTCTTCCTTGGGGCTTATCTTCTTTTTGTA TTTCACCACCGTGCCACTTAACTCCACAGTGCTCAG POU5F1 GCATGGGAGAGCCCAGAGTGGTGACGGAGACAGGGG CCCCATTCCTAGAAGGGCAGGCACCTCAGTTTGAAT PPARG CCTTTGCTTTGGTCAGCGGGAAGGACTTTATGTATG ATTTGTCTGTTGTCTTTCCTGTCAAGATCGCCCTCG PPL TTCTGCTTCTAACTCATAGTCCTTTACAGCTTGCTG CCTTCCATTCTCCAAGTCGAGAAGAGACCTTAGTTT PTK2 CCAGGTGAGTCTTAGTACTCGAATTTGGTGTGTGAT GCTCCATTGCACCAGGAGAACGTTCCATACCAGTAC RAB5B CCTAAGAGGCTGCGTGAGGGGAAACTGGGCCTTTGA TCTGGGGTCCAACCCCCACGATTAGGGGAAACGCTA RAC1 CAATGGCAACGCTTCATTCGGGAGGCTGTTCTGCTT TGTCGGGAACACGTGCTGCTAACTCACTGGTGAGTT RGS2 CAGAGTGAGACACCACGTTCAGACCACCTATTCCCT ACATTGTCTCTTTCATCATCTTACATTCCTGCTTTT RHOA ACTCCATGTACCCAAAAGCGCCAATCCTGTTTGCCA AAACCTCTCTCACTCCATCTTTGGTCTTTGCTGAAC RUNX2 TGACTGGCGGGGTGTAAGTAAAGGTGGCTGGATAGT GAGTGGTGGCGGACATACCGAGGGACATGCCTGAGG SCNN1A TCCACGTTCTGGGGCCGCGGATAGAAGATGTAGGCA TACCCCCAGGAACTGTGCTTTCTGTAGTCACAGTAC SERPINB5 TCAGATGTGTCTTCACTGAAGATATGTTTCAGCCCT AGGGCCACTCCCTTGGTCTCTGACATTCCAGAGAAA SMAD3 TGTGTGCAGGTTCTGCTGAACACGCACCTCCCAATC TAAAGAGGTGGCGCTCATCGATTTTTCCCGCTGTCC SMAD4 CCATCATACTTGATGGAGCATTACTCTGCAGTGTTA GCTGTCCCTCAAAGTCATGCACATATTCATCCTTCA SNAI1 TGGGGCACTCAGGAGGGAATTCCATGGCAGTGAGAA TTCATCAAAGTCCTGTGGGGCTGATGTGGCCAGAAG 173 SNAI2 GACCTGTCTGCAAATGCTCTGTTGCAGTGAGGGCAA ACATCAGAATGGGTCTGCAGATGAGCCCTCAGATTT SORL1 ACGATTGTGAGTCGGAAGTCGCCATCTGGATTAGCT ACCAGAGCCCTGGGACGATCAAGCACAGAGGAATTG SOX2 TTTCCTTTTTCTTTTTGAGCGTACCGGGTTTTCTCC CAGCTGTCATTTGCTGTGGGTGATGGGATTTTTTTT SOX9 CTGTAGTGTGGGAGGTTGAAGGGGCTGTAGGCGATC CGTACTGTGAGCGGGTGATGGGCGGGTAGGAGGGG SPARC AAAGCGGGTGGTGCAATGCTCCATGGGGATGAGGGG GATGTACTTGTCATTGTCCAGGTCACAGGTCTCGAA STAT3 ACAAAAGCCCCGCCAGCTCACTCACGATGCTTCTCC CGTCCGTGAGAGTTTTCTGCACGTACTCCATCGCTG TFPI2 CTGGAAACCTGTTCTCAATCCGGTTCCGGTGACACC TCTTTGGTGCGCAGAAGCCCATACAAGTAGCTTCAT TGFA AGCAGACGGAGTTCTTGACAGAGTTTTGAAGGCCCA GCTGATTTCTTCTCTAGGTCACACTGAATAACCCCA TGFΒ1 GCTCAACCACTGCCGCACAACTCCGGTGACATCAAA GGCGCTAAGGCGAAAGCCCTCAATTTCCCCTCCACG TGFΒ2 ATGCTCCAGCACAGAAGTTGGCATTGTACCCTTTGG TGCTGTGCTGAGTGTCTGAACTCCATAAATACGGGC TGFΒ3 GAGGCTTGGCAAGAAGGTGCATGAACTCACTGCACT AGGAAAACCAGGCGGCCTCCCCAGATCCCAAAGACT TGFΒR1 GGAAGTGGGTCCCACAACTTCCATCAGCCGTTTTAC GACCCCTCCCTTCCACCTCTAATGACTGAAGGAAAT TGFΒR2 CTGGCGATACGCGTCCACAGGACGATGTGCAGCGGC TCCACATCCGACTTCTGAACGTGCGGTGGGATCGTG TIMP1 TCCCCACGAACTTGGCCCTGATGACGAGGTCGGAAT AACGCTGGTATAAGGTGGTCTGGTTGACTTCTGGTG TMEM132A CCATAGGAACTCAGACAGTTCAAGGGGACTGCTGGA CCCCCACCAGTGCTATTCTCCACCACAAAGTCCAC TNFAIP3 TTCCGGACTTCTCGACACCAGTTGAGTTTCTTCTGG CAATTGCCGTCACCGTTCGTTTTCAGCGCCACAAGC TUBB AGCTCGATACTGCTGGCTTCCACGGCTGGTGAGAG ATCGAAGACCTGCTGGGTGAGTTCCGGCACTGTGAG TWIST1 TCCTTCTCTGGAAACAATGACATCTAGGTCTCCGGC TTATCCAGCTCCAGAGTCTCTAGACTGTCCATTTTC TWIST2 CAACGTTTCGTGGGCTGTCTTGACCAAACACTGCCC CATGTTCTTAGCCATTGCTAATGTGCTCACTCCCGC VDR TCCAGGACATGTCGTCCATGGTGAAGGACTCATTGG CACTGACGCGGTACTTGTAGTCTTGGTTGCCACAGG VEGFA CAGGGTCTCGATTGGATGGCAGTAGCTGCGCTGATA CTCGATCTCATCAGGGTACTCCTGGAAGATGTCCAC VIM AAGATTGCAGGGTGTTTTCGGCTTCCTCTCTCTGAA GACGTGCCAGAGACGCATTGTCAACATCCTGTCTGA VPS13A AAGTCCTAGTGCCATTCCTTCCACAAACTCTTCAGG AGCCAATCCACCAACAGCTCCACCAACTAGTGCCTT WNT5A ATCCACAGTGCTGCAGTTCCACCTTCGATGTCGGAA GCCTATCTGCATCACCCTGCCAAAAACAGAGGTGTT WNT5B CGGGTTCAAAGCTAATGACCACCAGGAGTTGGCGTC CTGGGCACCGATGATAAACATCTCGGGTCTCTGCAC ZEB1 TCATTGTTCTTGGCAGGGATATTTACTTTGCCTGGT TCCTGGGGTTCATTTGCATTTGCAGATTGAGGCTGA ZEB2 CTTCGGCAGCACGCAGGCTCGATCTGCGAAGTCTTG GATCAGATGGCAGTTCGCATGGACTCGGCGCCCTG 174 Supplementary Table S4.2 Primer sequences for multiplex target enrichment Gene Symbol Multiplex Target Enrichment - Primer Sequences Forward Primer Sequence Reverse Primer Sequence ACTA2 CAGCACTGCCTTGGTGTG TCTTTTTGTCCCATTCCCA AHNAK GTGGTCAAGGAGGGGGAC GCCAGGCTCGGGAGAG AKT1 CGGGTTTTAATTTATTTCATCCA AGGGGACACATGGGCAG AKT2 GCCACCCTTCAAACCTCA AACTGGGGGAAGTGGGTC AKT3 CAAAGAAGGGATCACAGATGC CTCCCACACATCATTTCATACA AURKA GGCATCATGGACCGATCT AAGACCCGCTGAGCCTG BIRC3 CCGTGGAAATGGGCTTTA TCTCTCTCCTCTTCCCTTATTTCA BMP7 CGTCAACCTCGTGGAACAT GAAGCGTTCCCGGATGTA CAMK2N1 CATGAAAAGAGAAAGCACTTTGAA TTTATTTTTGCAGAGGAGCC CD82 CAGGTGAAGTGCTGCGG GCCTCGCAGAAGCCC CD83 GGCATGGAACGAGCTTTT CCAGTGTAACAGACAGGCACA CDH1 TGAATGAAGCCCCCATCT CAGTGTCTCTCCAAATCCGA CDH2 CCAGAGTTTACTGCCATGACG CCAGTAGGATCTCCGCCA CFC1 AAGAGCGCCAGGCCA AAGCCTTTGTGGTGTCGC CLTC ACTTGCAGATGGCCCGTA AACAACGGTCACCAACTTGT COL1A1 CACAGCTGCAGCCCATC ACAAAATAGGTACAGAGTCTTTTGC COL3A1 GTGCTGCTGGTCCTCCTG TTTCCCTGGGACACCATC CREBBP GCTGGTTCTACTGCTTCATGC AACTTGGCAGGCTTTCCC CTNNB1 TCTCCACAACCTTTTATTACATCA TGCCATAAGCTAAAATTTGAAGG DVL1 CGACTCCACCATGTCCCT AGCCCCGCCCTTCAT DVL2 GGGACTCAAGGCCTCCAT CCAGCGCCATGCTCA EGF GCAGATCTCGATGGTGTGG TCACAGGAGCAAATAAGAGAATTG EGFR ACCTCCATCAGTGGCGAT CCAAGCCTGAATCAGCAAA EPAS1 CTCCCATCTGGACAAGGC AAACCCTCCAAGGCTTTCA ERBB3 CTCGTGGCCATGAATGAA AGTCAAGCGGAGCTGGC FBN1 TGGAATATCTGTATCCATCTCGG GCTGGAACCCTTTGTTGC FGFBP1 TGAATTTTCCTGTGTCTTTGCT TGCACACTCTGGTTTTCACAG FN1 TCCTGCACCACAGAAGGG AGTGGCACAAGGCACCAT 175 FOXC2 GACCTGAACCACCTCCCC CTGGCAGCTGGCATTG FOXD3 ACCCACCAACCGCTGTC CGGCCATTTGGCTTGA FZD7 TCATCCATATGCTGACCCTG TCCAGAATCACTTTGAAGTTTACC GAPDH TGGAAATCCCATCACCATC ATGACCCTTTTGGCTCCC GSC GTCACCGGAGAAGAGGGA GCGCGTGTGCAAGAAAGT GSK3B TTTGGAGCCACTGATTATACCTC CCTTGTTGGAGTTCCCAGG GUSB CGTGATGTGGTCTGTGGC CCCCTTGTCTGCTGCATAG HIF1Α TGCAACATGGAAGGTATTGC TTCAATATTTGATGGGTGAGGA HPRT1 AAGGGTGTTTATTCCTCATGGA TTGATGTAATCCAGCAGGTCA HPSE AAATGCAAAACTCTATGGTCCTG CTTGGTAGCAGTCCGTCCA IGF1 AGACGCTCTGCGGGG TCCGGAAGCAGCACTCAT IGF1R GAGAGTTTTAACAATCCATTCACA GGGCAGGGAATAAAAAGC IGF2 CCTGCCTGCCCTCCTG ATTGCTGGCCATCTCTGG IGFBP3 GGTCCCTGCCGTAGAGAAA GCTTCCTGCCTTTGGAAG IGFBP4 ACAGCCATGAAGTCACCG CCATCCAGTCAAGCCAGAA IL17RC GCCCTTGTGCAGTTTGGT CCCCTGCAGTCAGGCA IL6 CTCAGCCCTGAGAAAGGAG CACCAGGCAAGTCTCCTCA IL6ST CCCACACAGAATATACATTGTCC CGACTATGGCTTCAATTTCTCC ILK TCCACCTGCTCCTCATCC AAGGCCATGCCCCTTG ITGA5 TGGAACTCAGCTGTCCCC TTTTTGCTGGTGGTGCAG ITGB1 TTTGAGTGTGGCGCGT GCAGACGCACTCTCCATTG JAG1 CGTGCCAGTTAGATGCAAA CTGGCCAAGGCAGTCATT KRT7 AGAAGTCGGCCAAGAGCA TCCTCCACCACATCCTGC MIOS TCACTGATGGGAAGTGAGACT TTCTGAGTCACACACAACAAATCT MITF CCAGACATGCGCTGGAA GAGCAACAAATGCCGGTT MLPH TCCAATCTCTCACTTAAAACTAAGG TGGAAAAACTCCCACGTTT MMP3 CATGGAGACTTTTACCCTTTTGA GCAACGAGAAATAAATTGGTCC MMP9 AGTTTGTTCCTCGTGGCG ATAGAGGTGCCGGATGCC MST1R CTTCAGCCCACGCTCAGT TATACGAAGGCTCCCGTG MUC1 ACCCATGGGCGCTATGT CGACGTGCCCCTACAAGT NANOG TGCCTCACACGGAGACTG TCTTGACCGGGACCTTGT NDRG1 GGGTAATAGTGACATGCAGGC TCTAGATGCACCAACTTTAAAAATC 176 NOTCH1 CTCCTGCTTCAACGGTGG GCCGTCCTGACAGGTGC NUPR1 GCTTGCCACACCCCC ATCTTCCTGGCTCTGCCC PAK1 CACGGTCTGTGATTGAACC CCTCATCAGACATTTTAGGCTTC PGK1 GATGGGCTTGGACTGTGG TCCATGAGAGCTTTGGTTCC PIK3CA GATATGTCAGTGATTGAAGAGCAT TCGTGTAGAAATTGCTTTGAGC PLEK2 GCCCTGTACACTTTTGCTGAG CGACGCACCTTCCAGTTT POU5F1 CCTGAGGGGGAAGCCTTT ACCCTGGCACAAACTCCA PPARG CCTGGCAAAACATTTGTATGA TTGATTTTATCTTCTCCCATCATT PPL GTGCCAATTCCCAGCAGT TTGGTGGCAGGAGATTGG PTK2 GCTGCTTACCTTGACCCC CCAGGTGGTTGGCTCACT RAB5B TGTCAACTGTGGCTCTGTAACTC CAGATCACAGCAGAGCCAA RAC1 CCTATGTAGTTCTCAGATGCGTAA AAGACTGATCCAAAGAGCTGAG RGS2 TCACTGTGTACAGAACGCAAGA AAGGCCACATAATCCCAGA RHOA GCCGGTGAAACCTGAAGA TTTTTCTTCCCACGTCTAGC RUNX2 GCTTCTCCAACCCACGAA GTCCACTCTGGCTTTGGG SCNN1A TGATCAAGGAGTGTGGCTG TGAAACAGCCCAGGTGGT SERPINB5 CCCAAGGCTTGTCTGGAA GGAATCCCCACCATCTTCA SMAD3 TCTCCCTCCCCTCTCAGAA GCAGATCCTTCCAACTTTTCA SMAD4 CCTGGAATTGATCTCTCAGGA GGTGGATGCTGGATGGTT SNAI1 GGAAGGACCCCACATCCT GGGAAGGCCTCTTCCAGA SNAI2 GGGGGAGAAGCCTTTTTCT TGCAGGAGAGACATTCTGGA SORL1 TGCAGGCTTCAAAAAGATTG GGCTTCAGGTCTCCCCAG SOX2 GGGGTGCAAAAGAGGAGAG ATCGCGGTTTTTGCGTG SOX9 TACAGCGAGCAGCAGCAG GCCGCGTGGCTGTAGT SPARC AGCTGGCTCCACTGCGT TCCTTGTCGATATCCTTCTGC STAT3 ACTGCGCTGGACCAGATG CCTCCAATGCAGGCAATC TFPI2 GAAAAATTCTTTTCCGGTGG GGCAGAGCACAGTCCCTC TGFA CACCTGTGCAGCCTTTTG AGGAGTCCGTCTCTTTGCAG TGFΒ1 CGACTCGCCAGAGTGGTT TGAACCCGTTGATGTCCA TGFΒ2 GGGTGGAAATGGATACACG GCAGCAAGGAGAAGCAGA TGFΒ3 AAGTCCTGAAGCTCTCGCA GCAAGGCCTGGAGAGGA TGFΒR1 TGAAGGCTGCTCTGGAGAC GAAGTCAAAGCAGGGGCA 177 TGFΒR2 GTCGGGGGCTGCTCA AGTCCTATTACAGCTGGGGC TIMP1 CCACCCCACCCACAGAC CCCCTAAGGCTTGGAACC TMEM132A GGGCTCACAGAGCCAGAT GGCCTGGCCTGGGTACT TNFAIP3 AAACATCCAGGCCACCCT TCTGTGTCCTGAACGCCC TUBB TTTATGCCTGGCTTTGCC GCAGCCACGGTGAGGTAT TWIST1 CCCACCCCCTCAGCA GGATTTTGCTCTTCTAATTTCCAA TWIST2 CCAAACAGAGCGAAGAGGA CCAAAGTCCTCTTGTGGCA VDR AGGTCATCATGTTGCGCTC TGGAACTTGATGAGGGGC VEGFA TGGTGAAGTTCATGGATGTCT GTCATTGCAGCAGCCCC VIM GGGAGAAATTGCAGGAGGA TCTTCAAAAAGGCAATCTCTTC VPS13A TTACCAGGGAGCCATCCA GCTGCTACCCCCTTAGCC WNT5A CAGGCATCAAAGAATGCCA CGTTCACCACCCCTGCT WNT5B TGGGCTCAGCTTCTGACA TGGCACAGCTTCCTCTGG ZEB1 TTCTGAACCATCTTCTCCTGAA TGGTAAAACTGGGGAGTTAGTCA ZEB2 CCATTTCTGGCCAAAACTACA GCCTCTTGCACCGGG 178 Supplementary Table S4.3 Targets for EMT gene expression profiling in prostate cancer Gene Symbol Description ACTA2 actin, alpha 2, smooth muscle, aorta AHNAK AHNAK Nucleoprotein AKT1 AURKA v-akt murine thymoma viral oncogene homolog 1 v-akt murine thymoma viral oncogene homolog 2 v-akt murine thymoma viral oncogene homolog 3 aurora kinase A BIRC3 baculoviral IAP repeat containing 3 BMP7 bone morphogenetic protein 7 CAMK2N1 calcium/calmodulin-dependent protein kinase II inhibitor 1 CD82 AKT2 AKT3 Included in 'seed' list? Genes passing QC for statistical modeling? References Alcaraz, A et al. (2013) Y Y Shankar, J et al. (2010) Y Xue, X et al. (2014), Iliopoulos, D (2013) Xue, X et al. (2014), Iliopoulos, D (2013) Iliopoulos, D (2013) Chou, CH et al. (2013) Y Y Goode, G et al (2014) McCormack, N et al (2013) Y Y Su, B et al. (2014) CD82 molecule Y Tang Y et al (2014) CD83 CD83 molecule Y Kimbrel EA et al (2014) CDH1 cadherin 1, type 1, E-cadherin (epithelial) Y Gheldof A and Berx G (2013) CDH2 cadherin 2, type 1, N-cadherin (neuronal) Y Gheldof A and Berx G (2013) CFC1 cripto, FRL-1, cryptic family 1 Y Kruithof-de Julio, M et al. (2011) COL1A1 collagen, type I, alpha 1 Y Hospner, NA et al. (2013) COL3A1 collagen, type III, alpha 1 CREBBP CREB binding protein Y Melchionna R et al. (2012) CTNNB1 Y Oh SJ et al. (2013) DVL1 catenin (cadherin-associated protein), beta 1 dishevelled segment polarity protein 1 Y Li, X et al. (2013) DVL2 dishevelled segment polarity protein 2 Y Geng, Y et al. (2014) EGF epidermal growth factor Y Hardy, KM et al (2010) EGFR epidermal growth factor receptor EPAS1 endothelial PAS domain protein 1 Y Y Bangoura, G et al. (2007) ERBB3 v-erb-b2 avian erythroblastic leukemia viral oncogene homolog 3 Y Y Kim, J et al. (2013) FBN1 fibrillin 1 Y Boyer, S et al. (1999) FGFBP1 fibroblast growth factor binding protein 1 Y Wu, S et al. (2012) FN1 fibronectin 1 Y Mikheeva, SA et al. (2010) FOXC2 Y Liu, B et al. (2014) FOXD3 forkhead box C2 (MFH-1, mesenchyme forkhead 1) forkhead box D3 FZD7 frizzled class receptor 7 Y Vincan, E et al. (2007) GSC goosecoid homeobox Y Taube JH et al. (2010) GSK3B glycogen synthase kinase 3 beta Y Katoh M (2006) HIF1Α hypoxia inducible factor 1, alpha subunit Y Philip, B et al. (2013) HPSE heparanase Mao, Q et al (2013) Al Moustafa, AE et al. (2012) Y Chu, TL et al. (2014) Y Y 179 Masola, V et al. (2012) IGF1 Y Chen, CL et al. (2013) IGF1R insulin-like growth factor 1 (somatomedin C) insulin-like growth factor 1 receptor Y Kim, IG et al. (2014) IGF2 insulin-like growth factor 2 Y Chen, CL et al. (2013) IGFBP3 insulin-like growth factor binding protein 3 Natsuizaka M et al (2010) IGFBP4 insulin-like growth factor binding protein 4 Praveen Kumar VR et al (2014) IL17RC interleukin 17 receptor C IL6 interleukin 6 Y Colomiere M et al (2009) IL6ST interleukin 6 signal transducer Y Cheng Y et al. (2013) ILK integrin-linked kinase Mao, Q et al (2013) ITGA5 integrin, alpha 5 Zucchini-Pascal, N et al. (2012) ITGB1 integrin, beta 1 Y JAG1 jagged 1 Y Y Zavadil, J. (2004) KRT7 keratin 7 Y Y Telugu, BP et al. (2013) MIOS missing oocyte, meiosis regulator, homolog Y Senger, S et al. (2011) MITF Y Caramel, J et al. (2013) MLPH microphthalmia-associated transcription factor melanophilin MMP3 Y Wang XM et al. (2013) Y Strom, M et al. (2002) matrix metalloproteinase 3 Y Chen, QK et al. (2013) MMP9 matrix metalloproteinase 9 Y Lin F et al. (2014) MST1R macrophage stimulating 1 receptor Y Welm AL et al. (2007) MUC1 mucin 1, cell surface associated Y Liao G et al. (2014) NANOG Nanog homeobox NDRG1 N-myc downstream regulated 1 Y Santaliz-Ruiz, LE et al. (2013), Lu, Y et al. (2013) Jin, R et al. (2014) NOTCH1 notch 1 Y Capaccione KM et al. (2014) NUPR1 nuclear protein, transcriptional regulator, 1 Y Kim, KS et al. (2012) PAK1 p21 protein (Cdc42/Rac)-activated kinase 1 Y Lee, SH et al. (2011) PIK3CA phosphatidylinositol-4,5-bisphosphate 3kinase, catalytic subunit alpha PLEK2 pleckstrin 2 POU5F1 POU class 5 homeobox 1 PPARG PPL peroxisome proliferator-activated receptor gamma periplakin PTK2 protein tyrosine kinase 2 RAB5B RAB5B, Member RAS Oncogene Family RAC1 ras-related C3 botulinum toxin substrate 1 RGS2 regulator of G-protein signaling 2 RHOA ras homolog family member A RUNX2 runt-related transcription factor 2 SCNN1A sodium channel, non-voltage-gated 1 alpha subunit serpin peptidase inhibitor, clade B (ovalbumin), member 5 SERPINB5 SMAD3 Y Liu, L et al. (2012) Y Y Y Wallin, JJ et al. (2012) Y Hamaguchi N et al. (2007) Y Lu, Y et al. (2013) Sabatino, L et al. (2012) Y Tonoike, Y et al. (2011) Y Hyder CL et al. (2014) Chen, PI et al. (2014) Y Y Chen, X et al (2014) Wu S et al. (2012) Chen, X et al (2014) Y Y Y SMAD family member 3 180 Chimge, NO et al. (2011) Wang S et al. (2013) Y Chou, RH et al (2012) Y Yamazaki K et al. (2014) SMAD4 SMAD family member 4 Y SNAI1 Snail family zinc finger 1 Y Yamazaki K et al. (2014) Thakur N et al. (2014) Sánchez-Tilló E et al. (2012) SNAI2 Snail family zinc finger 2 Y Y Sánchez-Tilló E et al. (2012) SORL1 sortilin-related receptor Y Y Hirayama S et al. (2000) SOX2 SRY (sex determining region Y)-box 2 Y Yang N, et al. (2014) SOX9 SRY (sex determining region Y)-box 9 Y Capaccione KM et al. (2014) SPARC Y Y Fenouille N et al. (2012) TFPI2 secreted protein, acidic, cysteine-rich (osteonectin) signal transducer and activator of transcription 3 tissue factor pathway inhibitor 2 TGFA transforming growth factor, alpha Y TGFΒ1 STAT3 Wendt MK et al. (2014) Y Hibi, K et al. (2010) Y Nakano, T et al. (2003) transforming growth factor, beta 1 Y Lamouille S et al (2014) TGFΒ2 transforming growth factor, beta 2 Y Lamouille S et al (2014) TGFΒ3 transforming growth factor, beta 3 Y Lamouille S et al (2014) TGFΒR1 Y Lamouille S et al (2014) Y Lamouille S et al (2014) TIMP1 transforming growth factor, beta receptor 1 transforming growth factor, beta receptor 2 TIMP metallopeptidase inhibitor 1 Y Y D'Angelo RC et al (2014) TMEM132A transmembrane protein 132A Y Oh-hashi K et al. (2010) TNFAIP3 Y Wang, CM et al. (2011) TWIST1 tumor necrosis factor, alpha-induced protein 3 twist family bHLH transcription factor 1 TWIST2 twist family bHLH transcription factor 2 VDR VEGFA vitamin D (1,25- dihydroxyvitamin D3) receptor vascular endothelial growth factor A Y Xu, M et al. (2014) VIM vimentin Y Satelli A and Li S (2011) VPS13A vacuolar protein sorting 13 homolog A WNT5A ZEB1 wingless-type MMTV integration site family, 5A wingless-type MMTV integration site family, 5B zinc finger E-box binding homeobox 1 ZEB2 zinc finger E-box binding homeobox 2 CLTC GAPDH TGFΒR2 WNT5B Y Sánchez-Tilló E et al. (2012) Y Sánchez-Tilló E et al. (2012) Larriba, MJ et al. (2010) Y An CH et al. (2012) Y Ford, CE et al. (2014) Kato, S et al. (2014) Y Sánchez-Tilló E et al. (2012) Y Sánchez-Tilló E et al. (2012) Clathrin, heavy chain Y N/A housekeeper Y N/A housekeeper GUSB glyceraldehyde-3-phosphate dehydrogenase glucuronidase, beta Y N/A housekeeper HPRT1 hypoxanthine phosphoribosyltransferase 1 Y N/A housekeeper PGK1 phosphoglycerate kinase 1 Y N/A housekeeper TUBB tubulin, beta class I Y N/A housekeeper Y 181 Supplementary Table S4.4 Altered EMT-associated gene expression in response to individual gene knockdown Knockdown Gene Fold Change (KD vs. Control) HIF1Α MMP3 -5.73 GSC -3.58 SOX2 -3.34 SNAI1 -2.62 PTK2 2.02 CAMK2N1 2.03 ZEB1 2.05 MST1R 2.06 TGFA 2.15 CD82 2.15 FGFBP1 2.18 SNAI2 2.20 GSK3B 2.36 SOX9 2.46 CFC1 2.51 TGFΒ1 2.54 ERBB3 2.61 TWIST1 2.63 KRT7 2.64 MUC1 2.65 IL6ST 2.66 SMAD4 2.66 FN1 2.74 TIMP1 2.76 SORL1 2.82 JAG1 2.91 FZD7 3.16 182 SNAI1 KD ZEB1 KD SPARC KD TGFΒR2 3.18 VEGFA 3.34 CDH1 3.46 BIRC3 3.62 RUNX2 4.14 IL6 4.35 SPARC 6.68 NUPR1 15.90 SPARC 2.63 FN1 2.47 COL1A1 2.34 CDH1 2.09 IL6 2.08 FZD7 2.06 IGF2 2.06 TWIST2 -2.02 FBN1 -2.01 IL6ST -1.98 EGF 2.01 FN1 2.29 IL6 2.47 VEGFA 2.48 SORL1 2.83 FGFBP1 2.84 KRT7 2.94 SPARC 3.10 CDH1 3.60 NUPR1 4.50 TGFΒ2 -2.39 183 VIM KD CLTC KD NDRG1 1.96 NUPR1 2.37 FN1 2.68 CDH1 -1.98 FBN1 1.89 FN1 2.74 SPARC 2.69 NUPR1 2.38 WNT5A 1.93 CDH1 -1.87 KRT7 -2.27 TGFΒ2 -2.37 184