INVESTIGATING THE ROLE OF

EPITHELIAL-MESENCHYMAL

PLASTICITY IN PROSTATE

CANCER

Nataly Stylianou

BAppSc(MedSc)(Hons)

School of Biomedical Sciences

Faculty of Health

Queensland University of Technology (QUT)

Submitted in fulfilment of the requirement for the degree of

Doctor of Philosophy

2017

i Keywords Prostate, prostate cancer, cancer progression, metastasis, epithelial, mesenchymal, epithelial to mesenchymal transition, EMT, mesenchymal to epithelial transition, MET, mesenchymal to epithelial reverting transition, MErT, epithelial- mesenchymal plasticity, EMP, cell plasticity, plasticity, inducible expression models, lentivirus, Snail, Slug, Zeb1, E-cadherin, vimentin, invasion, dormancy, cell cycle, temporal transcriptional profiling, microarray, gene signature, survival, biochemical recurrence, predictive, metabolism, animal models

ii Abstract

Prostate cancer is the most diagnosed cancer and the second leading cause of cancer- related deaths amongst Australian men. Unfortunately, the majority of deaths occur due to the metastatic spread of cancer cells to distant organs. The propensity of cancer cells to transition between epithelial and mesenchymal phenotypic states via the epithelial-mesenchymal transition (EMT) program is considered to be critical in metastatic processes, cancer progression and treatment resistance. The activation of EMT imparts epithelial cancer cells with invasive characteristics, aiding their dissemination to distant organs. However, recent research suggests that a reversion back to their proliferative epithelial phenotype via a mesenchymal-epithelial reverting transition (MErT) is imperative for the formation of overt metastases. With increasing evidence that cell plasticity plays an important role in cancer progression, more accurate models are required to study these events more thoroughly.

To investigate a transient EMT in prostate adenocarcinoma, inducible and reversible models of EMT were generated whereby the expression of EMT inducing transcription factors Snail, Slug, or Zeb1 could be experimentally controlled both in vitro and in vivo via treatment with Doxycycline Hyclate (Dox). Exposure of cells to Dox induced a rapid transition of cells to a mesenchymal-like phenotype, which was confirmed via the reduction in epithelial specific markers and an increase in mesenchymal markers at the mRNA and protein level. Upon removal of Dox from cells having undergone an EMT, the EMT markers returned to their basal levels along with the re-establishment of the epithelial phenotype, overall signifying a MErT. Cells induced into the EMT-state became highly invasive in 3D-on-top Matrigel™ assays, and this was paralleled by a loss of proliferative ability and entry into a dormant-like state. Subsequent removal of Dox saw the re-awaking of these dormant cells which was characterised by the reacquisition of cell proliferation.

The reversible EMT models were successfully grown in the prostate of mice and the expression of the EMT transcription factors could be regulated by Dox treatment. While their use for in vivo studies requires further optimisation, this study is the first to generate orthotopic models whereby the induction and reversal of the EMT program can be experimentally controlled.

iii Comprehensive transcriptional profiling of a reversible EMT using a custom made 180k probe Agilent microarray elucidated for the first time the temporal dynamics of the less characterised MErT process. This revealed that the transcriptional profile of MErT is not a simple mirror image of EMT. The MErT not only consisted of dynamically reversible transcripts but also transcripts that remained persistently altered following an EMT or became uniquely activated with MErT. Both reversible and novel transcriptional subprograms were enriched in samples of lethal metastatic castration resistant prostate cancer (mCRPC), supporting the clinical role of a reversible EMT in metastasis. From this enrichment, a metastasis-derived gene signature (MPS) was identified to predict more rapid cancer relapse and reduced survival in not only prostate cancer patients but also across a number of other human carcinoma types.

Interestingly, The MPS was found to have a high enrichment of metabolism related genes. As there is little evidence supporting the involvement of EMT in cancer metabolism, further investigation revealed that the metabolism-related genes were also predictive of poor patient outcome when examined in isolation. Furthermore, functional testing of the effect of EMT on cell metabolism showed the EMT program to regulate the metabolic phenotype of prostate cancer cells. Cells actively undergoing EMT entered a metabolically quiescent state. Surprisingly, this study provides prototypical evidence that MErT not only relieved this effect, but resulted in a metabolically more energetic phenotype than that observed prior to EMT.

Cumulatively, this study provides first-in-field evidence to support the association of epithelial plasticity with poor clinical outcomes in multiple human carcinoma types. Furthermore, the identified novel signatures are promising candidates for discovering 1) biomarkers that can identify these transitions in clinical samples; and 2) drivers of these transitions that could be used to develop better targeted therapies; something that is currently lacking in the field.

iv Table of Contents

Keywords ...... ii

Abstract ...... iii

Table of Contents ...... v

Table of Tables ...... x

Table of Figures ...... xii

List of Abbreviations ...... xvi

QUT Confidentiality Undertaking...... xix

Statement of Original Authorship ...... xx

List of Publications and Awards ...... xxi

Acknowledgments ...... xxv

Chapter 1: Literature Review ...... 1

1.1. Prostate cancer: incidence, detection, and management...... 2

1.2 Epithelial-mesenchymal plasticity (EMP) and prostate gland development...... 4

1.3 EMP in cancer...... 6

1.4 Regulation of the Epithelial to Mesenchymal Transition (EMT)...... 8

1.4.1 EMT markers...... 8

1.4.2 EMT regulators...... 10

1.4.2.1 The Snail family...... 10

1.4.2.2 The Zeb family...... 12

1.4.2.3 The basic/helix-loop-helix (bHLH) family...... 12

1.4.3 EMT inducers...... 13

1.5 The role of EMP in the metastatic cascade...... 15

1.5.1 EMP and tumour heterogeneity...... 16

1.5.2 EMP and metastatic dissemination...... 18

1.5.3 EMP and the metastatic niche...... 20

v 1.5.4 EMP and metastatic colonisation...... 21

1.6 EMT and therapy resistance...... 23

1.7 EMT and cancer stem cells (CSCs)...... 24

1.8 Study aims ...... 25

1.8.1 Rationale ...... 25

1.8.2 Hypotheses ...... 26

1.8.3 Aims ...... 26

Chapter 2: Materials and Experimental Methods ...... 27

2.1 Tissue culture...... 28

2.2 pINDUCER20 vector system...... 28

2.3 HEK293T transfection and lentivirus production...... 33

2.4 Cell transduction and infection...... 33

2.5 Three-dimensional (3D)-on-top Matrigel™ assays...... 33

2.6 Immunofluorescence staining...... 35

2.7 RNA extraction and cDNA preparation...... 36

2.8 Quantitative real-time polymerase chain reaction (qRT-PCR)...... 36

2.9 Protein extraction and Western blotting...... 36

2.10 Cell viability and proliferation assays...... 38

2.11 Flow cytometry...... 38

2.12 Seahorse assay to measure metabolic parameters...... 39

2.13 Microarray gene expression profiling and analysis...... 39

2.14 Statistical analysis...... 41

2.15 Generation of the LNCaP-iSnailRFP/LUC , LNCaP-iSlugRFP/LUC, and LNCaP- iGFPRFP/LUC cells...... 41

2.16 Intraprostatic injection procedure...... 41

2.17 Bioluminescent imaging and tumour progression...... 43

vi 2.18 Mouse termination and organ collection...... 43

2.19 Tissue processing, embedding and sectioning...... 43

2.20 Haematoxylin and Eosin staining (H&E)...... 44

2.21 Immunohistochemistry (IHC)...... 44

Chapter 3: Characterisation of reversible EMT models ...... 46

3.1 Introduction ...... 47

3.2 Results ...... 49

3.2.1 Optimisation of Dox concentration to induce cDNA expression in the reversible EMT models...... 49

3.2.2 Assessing the reversibility of the pINDUCER20 construct following removal of Dox...... 53

3.2.3 Assessing EMT-like changes in the LNCaP-iSnail, LNCaP-iSlug, and LNCaP-iZeb1 models following Dox treatment...... 56

3.2.4 Assessing the invasive characteristics of the LNCaP-iSnail and LNCaP- iSlug models following Dox treatment...... 59

3.2.5 Transcriptional profiling of LNCaP-iSnail and LNCaP-iSlug models during 4 days of Dox treatment in 3D Matrigel™ cultures...... 63

3.2.5.1 Assessment of EMT-related markers ...... 63

3.2.5.2 Identification of enriched public datasets and biological processes ...... 71

3.2.6 Assessing the reversibility of EMT in the LNCaP-iSnail and LNCaP- iSlug models...... 71

3.3 Discussion ...... 79

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression...... 86

4.1 Introduction ...... 87

4.2 Results ...... 89

4.2.1 MErT reawakens EMT-induced dormant-like tumour cells...... 89

vii 4.2.2 Temporal dynamics of gene expression associated with MErT...... 93

4.2.3 MErT is not the antithesis of EMT – revealing a transcriptional footprint of plasticity in mCRPC...... 102

4.2.4 A metastasis-derived plasticity signature expressed in primary tumour samples predicts poor patient prognosis...... 105

4.2.5 The MPS is not reliant on cell cycle gene expression...... 111

4.2.6 Epithelial-mesenchymal plasticity (EMP) fuelled metabolic reprogramming is a predictor of clinical outcome...... 115

4.3 Discussion ...... 128

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study ...... 133

5.1 Introduction ...... 134

5.2 Results ...... 135

5.2.1 Generation and characterisation of RFP/LUC variants of the LNCaP- iSnail, LNCaP-iSlug, and LNCaP-iGFP cells...... 135

5.2.2 Experimental design of in vivo experiment...... 142

5.2.3 Optimising the cell concentration for the intraprostatic injections...... 146

5.2.4 Examination of the primary tumour growth across groups...... 147

5.2.5 Examination of inducible protein expression following transient Dox treatment in the LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP- iGFPRFP/LUC tumours...... 152

5.2.6 Examination of EMT-related marker expression at the primary site. .. 156

5.2.7 Examination of the tumour stroma...... 165

5.2.8 Assessment of distant tumour cells...... 167

5.3 Discussion ...... 173

Chapter 6: Final discussion and future directions ...... 180

6.1 The reversible epithelial-mesenchymal transition (EMT) models are unique tools for investigating epithelial-mesenchymal plasticity (EMP) in LNCaP cells...... 181

viii 6.2 MErT is a kinetically dynamic process and is enriched in mCRPC...... 183

6.3 MErT imprints unique transcriptional features that are enriched in mCRPC and could serve as clinical identifiers/therapeutic targets for cancer cells that have undergone a reversible EMT...... 183

6.4 Evidence of EMP involved in the androgen signalling axis...... 186

6.5 Intra-tumoural cell plasticity via the EMT program at the primary site is predictive for poor patient outcome...... 188

6.6 EMP reprograms the metabolic phenotype of LNCaP cells...... 188

6.7 Suggested roles for EMP in cancer progression...... 189

6.8 The reversible EMT models and their use for in-vivo studies...... 191

6.9 Summary ...... 193

Reference List ...... 194

ix Table of Tables Table 2.1 Table of forward and reverse primer sequences used for qRT-PC of individual genes...... 37

Table 2.2. Table of primary antibodies and the concentrations used for Western blotting...... 38

Table 2.3 Table of primary antibodies and dilutions used for immunohistochemistry...... 45

Table 4.1. The Metastatic Plasticity Signature (MPS) ...... 106

Table 4.2. Enrichment of the Metastatic Plasticity Signature (MPS) in multiple PCa patient cohorts...... 108

Table 4.3. Enrichment of the Metastatic Plasticity Signature (MPS) in multiple cancer patient cohorts...... 112

Table 4.4. The Metastatic Plasticity Signature with cell cycle genes removed (MPSCCR)...... 114

Table 4.5. Enrichment of the Metastatic Plasticity Signature devoid of cell cycle genes (MPSCCR) in multiple cancer patient cohorts...... 118

Table 4.6. The metabolism-related genes from the Metastatic Plasticity Signature (MPSMETAB)...... 120

Table 4.7. Enrichment of the metabolism-related genes from the Metastatic Plasticity Signature (MPSMETAB) in multiple cancer patient cohorts...... 125

Table 5.1. Tumour take of LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP- iGFPRFP/LUC cells following intraprostatic injection...... 146

Table 5.2. Percentage of Ki67 positive tumour cells to Ku70 positive tumour cells within the tumours...... 162

Table 5.3. Number of mice that presented with tumour cells within the tumour stroma or blood vessels...... 165

x Table 5.4. Number of mice that presented with Ku70 positive tumour cells in the lymph node ...... 169

Table 5.5. Number of mice that presented with Ku70 positive tumour cells in the lungs...... 171

xi Table of Figures

Figure 1.1 The localisation of the prostate gland and prostate cancer...... 2

Figure 1.2. Development of the prostate gland in a C57BL/6J mouse embryo...... 5

Figure 1.3. The spectrum of epithelial – mesenchymal plasticity...... 9

Figure 1.4. Domain structure of the EMT-TFs Zeb1/2, Snail1/2/3 and Twist1/2. .... 10

Figure 1.5. (A-F) The metastatic cascade...... 15

Figure 2.1. Plasmid map of the pINDUCER20 construct...... 29

Figure 2.2. Schematic diagram and mechanism of the pINDUCER-20 vector backbone...... 30

Figure 2.3. Schematic of the generated pINDUCER20 constructs...... 31

Figure 2.4: Plasmid map of the pENTR™ 223.1 plasmid...... 32

Figure 2.5. Schematic diagram of a 3D-on-top Matrigel™ assay...... 34

Figure 2.6. Schematic of the pMig-Luc2-DsRed plasmid...... 42

Figure 3.1. Optimisation of the minimum effective dose of Dox to induce maximum expression of GFP in the LNCaP-iGFP cells...... 50

Figure 3.2. Optimisation of the minimum effective dose of Dox to EMT-like marker expression in the LNCaP-iSnail model...... 52

Figure 3.3. Assessing the reversibility of the LNCaP-iGFP cell model...... 54

Figure 3.4. Assessing the reversibility of the LNCaP-iSnail cell model...... 55

Figure 3.5. Assessing the induction of an EMT in the LNCaP-iSnail and LNCaP- iSlug models...... 58

Figure 3.6. Immunofluorescent imaging of EMT proteins in the LNCaP-iSnail and LNCaP-iSlug models...... 60

Figure 3.7 Assessing the induction of an EMT in the LNCaP-iZeb1model...... 61

xii Figure 3.8. Induction of LNCaP-iSnail and LNCaP-iSlug tumour spheroid invasion following Dox treatment...... 65

Figure 3.9. Assessing EMT-like protein expression in the LNCaP-iSnail and LNCaP-iSlug models following 4 days of Dox treatment...... 66

Figure 3.10. A Snail or Slug-induced invasive EMT leads to changes in hallmark EMT genes ...... 70

Figure 3.11. The Snail or Slug-induced EMT gene signature is enriched for genes reported in other EMT-related datasets...... 72

Figure 3.12. Biological processes enriched in a Snail-induced EMT ...... 73

Figure 3.13. Biological processes enriched in a Slug-induced EMT ...... 74

Figure 3.14. Defining the reversibility of the LNCaP-iSnail and LNCaP-iSlug models...... 77

Figure 3.15. Western blot of EMT-related proteins demonstrates the reversibility of the LNCaP-iSnail and LNCaP-iSlug models...... 78

Figure 4.1. MErT breaks EMT-induced cell dormancy ...... 91

Figure 4.2. Relative enrichment score of enriched biological processes from the gene ontology consortium (GO-BPs) (Ashburner et al. 2000; Gene Ontology 2015) at the indicated treatment times...... 92

Figure 4.3. MET rescues EMT-induced decrease in cell proliferation...... 94

Figure 4.4. The MErT is a kinetically dynamic transcriptional process that ...... 96

Figure 4.5. mRNA expression of indicated epithelial and mesenchymal markers. ... 98

Figure 4.6. Enriched GO-BPs in the indicated clusters...... 100

Figure 4.7. MErT is enriched in prostate cancer metastasis...... 101

Figure 4.8. MET is not the antithesis of EMT and is enriched in mCRPC ...... 104

Figure 4.9. The Metastatic Plasticity Signature (MPS) overlaps and is significantly expressed in the metastasis of a number of prostate cancer datasets...... 107

Figure 4.10. The MPS predicts poor patient outcome across multiple cancers...... 110

xiii Figure 4.11. Bubble chart indicating the enrichment of biological processes from the Gene Ontology Consortium in the Metastatic Plasticity Signature (MPS)...... 113

Figure 4.12. The Metastatic Plasticity Signature (MPS) devoid of cell cycle genes (MPSCCR) predicts poor patient outcome in prostate, breast, and lung cancer...... 117

Figure 4.13. Bubble chart showing the enrichment of biological processes from the Gene Ontology Consortium in the Metastatic Plasticity Signature with cell cycle genes removed (MPSCCR)...... 119

Figure 4.14. The metabolism-related genes from the Metastatic Plasticity Signature (MPSMETAB) predict poor patient outcome in prostate, breast, and lung cancer...... 124

Figure 4.15. Examining the effect of a Snail-induced EMT and MErT on metabolic reprogramming...... 126

Figure 4.16. Measure of Basal oxygen consumption rate (OCR), and extracellular acidification rate (ECAR) in LNCaP-iGFP cells treated with Dox for up to 5 days followed by Dox removal for 10 days...... 127

Figure 5.1. The inducible cell models were sorted for the top 27-38% RFP- expressing cells...... 137

Figure 5.2. Luciferase expression as detected by the IVIS of LNCaP-iSlugRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iGFPRFP/LUC cells at indicated cell numbers following incubation with D-luciferin (150 μg/mL) for 5 minutes...... 138

Figure 5.3. Cell morphology of LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox...... 139

Figure 5.4. Cell morphology, RFP and GFP expression of LNCaP-iGFPRFP/LUC cells following transient treatment with Dox...... 140

Figure 5.5. Expression of Slug in LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox...... 141

Figure 5.6. Expression of E-cadherin in LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox...... 143

Figure 5.7. Expression of vimentin in LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox...... 144

xiv Figure 5.8. The experimental design...... 145

Figure 5.9. Luciferase emission over time of the (A) LNCaP-iGFPRFP/LUC, (B) LNCaP-iSnailRFP/LUC, and (C) LNCaP-iSlugRFP/LUC primary tumours in mice that received either 2x105 or 1x106 cells intraprostatic injections...... 149

Figure 5.10. H&E staining of tumour xenographs...... 150

Figure 5.11. Example of the stromal septae (S) and necrosis (N) in the tumours (T) stained with haematoxylin and eosin...... 151

Figure 5.12. The scale utilised for scoring the staining intensity...... 153

Figure 5.13. Expression of GFP, Snail, or Slug in tumour xenographs...... 154

Figure 5.14. Expression of GFP (brown) in LNCaP-iGFPRFP/LUC tumour cross- sections from tumours treated with Dox for 19 weeks (EMT)...... 155

Figure 5.15. Expression of Ku70 in the tumour xenographs...... 157

Figure 5.16. Expressionof GFP in the tumour xenographs...... 158

Figure 5.17. Expression of Snail in the tumour xenographs...... 159

Figure 5.18. Expression of Slug in the tumour xenographs...... 160

Figure 5.19. Expression of vimentin in the tumour xenographs...... 161

Figure 5.20. Expression of E-cadherin in the tumour xenographs...... 163

Figure 5.21. Expression of Ki67 in the tumour xenographs...... 164

Figure 5.22. Expression of Ku70 in the tumour stroma...... 166

Figure 5.23. Expression of Ku70 in the lymph nodes...... 168

Figure 5.24. H&E staining of the lymph nodes...... 170

Figure 5.25. Expression of Ku70 in the lungs...... 172

Figure 6.1. Expression of POU4F1 and TUBB3 in treatment naïve prostate cancer and castrate-resistant prostate cancer (CRPC) primary tumours...... 187

Figure 6.1. Hypothetical roles of EMT and MErT in cancer progression...... 192

xv List of Abbreviations

# number % percentage ~ approximately   less than  more than ≤ less than or equal to ≥ more than or equal to µg/µL micrograms per microliter µg/mL micrograms per millilitre µm micrometre µM micromolar µmol micromole ⁰C degrees Celsius 2D two-dimensional 2-DG 2-deoxyglucose 3D three-dimensional ADT androgen deprivation therapy AIHW Australian Institute of Health and Welfare ANOVA analysis of variance APCRC-Q Australian Prostate Cancer Research Cancer Queensland AR androgen receptor ATCC American Type Culture Collection avg average bHLH basic/helix-loop-helix BSA bovine serum albumin CCP cell cycle and proliferation CCR cell cycle removed cDNA complimentary DNA cm centimetre

CO2 carbon dioxide CRPC castrate resistant prostate cancer CSC cancer stem cell CtBP C-terminal binding protein binding site

xvi CTC circulating tumour cell C-terminal carboxy-terminal  change DMEM Dulbecco's modified Eagles medium DNA deoxyribonucleic acid Dox doxycycline hyclate ECAR extracellular acidification rate EDTA Ethylenediaminetetraacetic acid EMP epithelial to mesenchymal plasticity EMT epithelial to mesenchymal transition FACS fluorescently activated cell sorting FBS foetal calf serum FCCP carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone GFP green fluorescent protein GO-BP biological processes from the Gene Ontology Consortium GSEA gene set enrichment analysis H&E haematoxylin and eosin HEK human kidney embryonic hr hour IF immunofluorescence IHC immunohistochemistry IPA Ingenuity Pathway Analysis L litter LIMMA Linear models for microarray data LUC luciferase M molar mCRPC metastatic castrate resistant prostate cancer MErT mesenchymal to epithelial reverting transition MET mesenchymal to epithelial transition MFI mean fluorescence intensity min minute mL millilitre mm millimetre mM millimolar

xvii MMP matrix metalloprotease mol mole MPS metastatic plasticity signature MPSCCR metastatic plasticity signature with cell cycle genes removed mRNA messenger RNA mSigDB Molecular Signatures Database mths months n sample size nmol nanomoles N-terminal amino-terminal

O2 oxygen OCR oxygen consumption rate OR odds ratio p p-value PBS phosphate buffered saline PCa prostate cancer PCR polymerase chain reaction pmol/min picomoles/minute PSA prostate specific antigen qRT-PCR quantitative real time polymerase chain reaction QUT Queensland University of Technology RFP red fluorescent protein RNA ribonucleic acid RPMI-1640 Rosewell Park Memorial Institute medium 1640 SD standard deviation SEM standard error mean TBS tris-buffered saline TBST tris-buffered saline with Tween-20 TE Trypsin Ethylenediaminetetraacetic acid TF transcription factor UGM urogenital sinus mesenchyme UGE urogenital sinus epithelium β beta

xviii QUT Confidentiality Undertaking The author of the PhD has disclosed confidential information which has a unique value to the Author and/or QUT, and may be the basis of future patent applications. The Authors and/or QUT will be prejudiced by any unauthorised disclosure of the confidential information, may be precluded from being granted patents, and may suffer financial loss as a result of such unauthorised disclosure. As such, the information within this PhD thesis is presented in confidence.

xix Statement of Original Authorship The work contained in this thesis has not been previously submitted to meet the requirements for an award at this or any other higher education institution. To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made.

QUT Verified Signature Signature:

Date: 30.05.2017

xx List of Publications and Awards Research articles in preparation or submitted:

Nataly Stylianou, Melanie L. Lehman, Chenwei Wang, Lidija Jovanovic, Anja Rockstroh, Martin C. Sadowski, Abhishek S. Kashyap, Thomas F. Westbrook, Elizabeth D. Williams, Jennifer H. Gunter, Colleen C. Nelson, Brett G Hollier. Revealing the plastic nature of human carcinoma metastasis. Manuscript under preparation.

Phoebe Sarkar, Wendy Lee, Elizabeth Williams, Amy Lubik, Nataly Stylianou, Ali Shokoohmand, Melanie Lehman, Brett Hollier, Jennifer Gunter, and Colleen Nelson. Insulin accelerates invasion and migration in androgen-deprived prostate cancer cells through activation of FOXC2. Endocrine-Related Cancer (Manuscript submitted May 2017)

Research articles accepted:

Brian W.C. Tse,*, Marianna Volpert,*, Ellca Ratther,*, Nataly Stylianou, Mannan Nouri, Kayla McGowan, Melanie L. Lehman, Stephen J. McPherson, Mani Roshan-Moniri, Miriam S. Butler, Josselin Caradec, Cheryl Y. Gregory-Evans, Jacqui McGovern, Rajdeep Das, Mandeep Takhar, Nicholas Erho, Mohamed Alshalafa, Elai Davicioni, Robert B. Jenkins, R. Jeffrey Karnes, Robert B. Den, Ladan Fazil, Luke A. Selth, Philip A. Gregory, Martin E. Gleave, Elizabeth D. Williams, Paul S. Rennie, Ralph Buttyan, Jennifer H. Gunter, Pamela J. Russell, Colleen C. Nelson & Brett G. Hollier . Neuropilin-1 is up-regulated in castrate resistant prostate cancer, promotes metastasis and is a prognostic biomarker of metastatic relapse and patient mortality. Oncogene, 2016

Research articles published:

L. A. Selth, R. Das, S. L. Townley, I. Coutinho, A. R. Hanson, M. M. Centenera, N. Stylianou, K. Sweeney, C. Soekmadji, L. Jovanovic, C. C. Nelson, A. Zoubeidi, L. M. Butler, G. J. Goodall, B. G. Hollier, P. A. Gregory and W. D. Tilley. A ZEB1- online miR-375-YAP1 pathway regulates epithelial plasticity in prostate cancer. Oncogene advance publication 6 June 2016; doi: 10.1038/onc.2016.185

xxi Mannan Nouri, Ellca Ratther, Nataly Stylianou, Colleen C. Nelson, Brett G. Hollier and Elizabeth D. Williams. Androgen-targeted therapy-induced epithelial mesenchymal plasticity and neuroendocrine transdifferentiation in prostate cancer: an opportunity for intervention. Front Oncol. 2014 Dec 23;4:370. doi:

10.3389/fonc.2014.00370. eCollection 2014.

Lubik A.A., Gunter J.H., Hollier B.G., Ettinger S., Fazli L., Stylianou N., Hendy S.C., Adomat H.H., Gleave M.E., Pollak M., Herington A., Nelson C.C. IGF2 increases de novo steroidogenesis in prostate cancer cells. Endocr Relat Cancer. 2013 Mar 22;20(2):173-86. doi: 10.1530/ERC-12-0250. Print 2013 Apr.

Awards:

Best Poster - Judges Award – Investigating the Role of the Epithelial- Mesenchymal Transition in Prostate Cancer Metastasis. IHBI Inspires Postgraduate Student Conference. 2012, Gold Coast, Australia.

Runner Up – Student Poster Award – Investigating the role of the Epithelial- Mesenchymal Plasticity in Prostate Cancer Progression. TRI Annual Poster Symposium. 2014, Brisbane, Australia.

Runner Up – Student Poster Award – Investigating the role of the Epithelial- Mesenchymal Plasticity in Prostate Cancer Progression. TRI Annual Poster Symposium. 2015, Brisbane, Australia.

Best Early Career Researcher Oral Presentation Award - MErT is enriched in lethal metastatic castrate resistant prostate cancer and correlates with an overall poor prognosis across multiple cancers. The Epithelial-Mesenchymal Transition International Association – VII. 2015, Melbourne, Australia.

Best Poster Presentation – Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer. Novel Concepts in the Biology and Treatment of Metastatic Cancer Conference. 2016, Paphos, Cyprus.

xxii Oral presentations:

Investigating the Role of the Epithelial-Mesenchymal Transition in Prostate Cancer Metastasis. Australian Prostate Cancer Research – Queensland Seminar Series. 2012, Brisbane, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. IHBI Inspires Postgraduate Student Conference. 2015, Gold Coast, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. IHBI Gala Dinner. 2015, Brisbane, Australia.

MErT is enriched in lethal metastatic castrate resistant prostate cancer and correlates with an overall poor prognosis across multiple cancers. TRI Annual Symposium. 2016, Brisbane, Australia

Poster presentations:

Investigating the Role of the Epithelial-Mesenchymal Transition in Prostate Cancer Metastasis. Australian-Canadian Prostate Cancer Collaborative Research Alliance Symposium. 2012, Daydream Island, Australia.

Investigating the Role of the Epithelial-Mesenchymal Transition in Prostate Cancer Metastasis. 14th International Biennial Congress of the Metastasis Research Society. 2012, Brisbane, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. Australian-Canadian Prostate Cancer Collaborative Research Alliance Symposium. 2013, Port Douglas, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. IHBI Inspires Postgraduate Student Conference. 2013, Gold Coast, Australia.

xxiii Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. The Epithelial-Mesenchymal Transition International Association - VI. 2013, Alicante, Spain.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. Prostate Cancer World Congress. 2013, Melbourne, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. IHBI Inspires Postgraduate Student Conference. 2014, Brisbane, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. Australian-Canadian Prostate Cancer Collaborative Research Alliance Symposium. 2015, Brisbane, Australia.

Investigating the Role of the Epithelial-Mesenchymal Plasticity in Prostate Cancer Progression. IHBI Inspires Postgraduate Student Conference. 2016, Gold Coast, Australia

xxiv Acknowledgments What a journey! Brett, you crazy, crazy man… you took me on back in 2011 for Honours, probably from a combined effort of desperation on your part and keenness/naivety on mine. Regardless, thank you for giving me the time of day. It’s needless to say that I wouldn’t be where I am today without your mentoring, guidance, enthusiasm, and friendship. Thank you for sticking with me through thick and thin and shaping me into the researcher that I am today. You were always there for me in your own way, your “tough love” definitely paid off in the end.

Jenni, thanks for calmly receiving my heated ranting sessions, you kept me sane when things were falling apart! Elizabeth, you came into my PhD life at a time of great need, thank you for the mentoring, counselling, and advice. Your enthusiasm and passion for research is contagious and definitely kept me going when things got tough and the tough got going! Colleen, you are such an inspiring and understanding person, thank you for having faith in me despite the numerous extensions this project endured.

There are many APCRC-Q, QUT, and TRI staff members that I would like to acknowledge. Dr. Eliza Whiteside, thanks for steering me towards the right path early on! You pulled me aside and told me to get on with the Honours program and move through to a PhD, and now, here I am! Anja, you definitely sparked my interest in the intense world of bioinformatics and for that I thank you immensely! Also, thank you for always engaging in in-depth conversations with me at the most random times! Martin and Steve, aside from the never ending debates, chatting about Game of Thrones and TV shows was a welcoming break every week. Chenwei and Mel, thanks for all your bioinformatics insights and for putting up with my ridiculous requests! Grega, thanks for the laughs, Lidija thanks for always having a kind word to say. Pam, Sam, Leisl, Rob, Vanessa, aka “the Lenti Room people,” I had a blast tissue culturing by your side! Jodie, thanks for pulling all the strings to make life easier around the lab. Gordon, Jo, Brent, and the rest of the dock team, thanks for cheerfulness, friendliness and helpfulness! Thank you to all the animal facility staff for always keeping things rolling smoothly. Sandrine, Ali, Brian, Dalia, Rob, Felicity, Crystal, Tony, and the rest of the Histology, Microscopy, FACS, and Imaging facility staff, thank you for your friendly banter, advice, and help with my

xxv experiments and analysis. Trish, Kathryn, Mandy, Julie, thanks for your patience when dealing with the admin work and paperwork involved when organising my last minute conference related travelling.

I can’t leave out my peeps. Taylor!! What would I have done without you, thanks for being there during the toughest time of my PhD!!! Tiffany, Kayla, Mannan, (and Phoebe), we had some good times, you were the perfect minions. Special shout out to Esha (and Mama Shah - I miss your hot cuisine!), Jacky, Ali, Rachel, Marianna, Chelsea, Katrina P and Katrina S, Christine, Mei, and Jess, you made every day lab life a little bit more interesting. Ellca G, what can I say, thanks for always making me laugh no matter what was going on. You’re hands down the coolest person I know! You even saved my life that day when “that truck was going to end me!” Pat, mate, you are one of the most hard working peeps that I know, and we had some good times honouring and PhDing together. Ashley Cooper (Ash), I really enjoyed our in-depth conversations and impromptu coffee/rant sessions, you were always there in a split second when I needed some time out. Phoebe, thanks for bringing girl power to the forefront! Abhi, thanks for having the patience when I first started Honours to teach me all your fantastic ways! Lipsa, you’re the most patient person I know, thanks for keeping me sane when we were in Spain!

Outside of the lab, Alyssa, my bestest friend in the whole wide world, I seriously couldn’t have done it without your unconditional support and belief in me. Mum, you can finally tell the world that you have a “Doctor” for a daughter, and Dad, you raised me to have the same passion for excellence like you do, this PhD is as much yours as it is mine. Nan, thanks for your continuous support, you were always there for a chat when I needed one. Roger, you endured this journey alongside me from the start, thanks for keeping things easy going. Nick, you really lifted my spirits when I needed it the most, words cannot describe how thankful I am.

And lastly, I thank my stubborn perseverance for shining through at the best and absolute worst of times.

“Success is the ability to go from failure to failure without losing your enthusiasm” – Winston Churchill.

xxvi “Have faith” – Dr. Brett Hollier.

Nataly.

xxvii

Chapter 1: Literature Review

Chapter 1: Literature Review 1 1.1. Prostate cancer: incidence, detection, and management. The prostate is an androgen -dependent gland that is situated between the bladder and the urethra (Figure 1.1). Its function is to produce seminal fluid that contains nutrients to nourish and protect sperm. Prostate cancer is associated with old age; in fact, the risk of Australian men developing prostate cancer in 2015 by the age of 85 years old was calculated to be 1:7 (AIHW 2014). Other risk factors include ethnicity and a positive family history of prostatic carcinoma. The incidence of prostate cancer varies worldwide, however, in Australia prostate cancer is the most commonly diagnosed male cancer and the second leading cause of cancer-related deaths (AIHW 2015).

Figure 1.1 The localisation of the prostate gland and prostate cancer.

The prostate gland is located between the bladder and the urethra. The left panel depicts a normal prostate gland and the right panel a cancerous prostate gland. (Figure sourced from www.mayoclinic.org)

Prostate cancer is initially diagnosed by the detection of a persistent rise in the serum biomarker prostate-specific antigen (PSA). While PSA is secreted by prostate cells and is specific to the prostate, its elevation does not always indicate the

Chapter 1: Literature Review 2 presence of cancer. PSA can also be elevated by other conditions such as benign prostate hyperplasia (BPH) or prostatitis. Therefore, the definitive diagnosis of prostate cancer is made by a transrectal ultrasound (TRUS)-guided needle biopsy combined with a digital rectal examination (DRE). Depending on the severity of their condition at the time of diagnosis, patients are offered a range of therapeutic options. Patients with organ-confined prostate cancer can opt for active surveillance, watchful waiting, or a more aggressive approach such as radical prostatectomy or radiotherapy with or without hormone therapy (ACNWP 2003; D'Amico et al. 2004; Denham et al. 2011; Roach et al. 2008; Bohmer et al. 2016). Unfortunately, there are no curative options available for patients with detectable metastases at time of diagnosis. For those patients, therapies such as ADT or radiation are administered to reduce tumour burden and extend or better their quality of life. On the other hand, while the available treatment methods are successful for most patients with organ- confined prostate cancer, approximately 25% of patients will relapse with a secondary rise in serum PSA levels (Paller and Antonarakis 2013). At this stage, most patients will be administered androgen deprivation therapy (ADT), which exploits the reliance of prostate cancer cells on androgens to survive (Paller and Antonarakis 2013; Asmane et al. 2011). While ADT reduces tumour burden and prolongs life, invariably most patients will relapse once again at a median time of 18-24 months with castrate-resistant prostate cancer (CRPC) (Asmane et al. 2011). CRPC has an aggressive phenotype that often results in metastases. As prostate cancer in general has prevalence to metastasise to the bones, majority of metastases observed in CRPC patients is also to the bones. (Shah et al. 2004; Bubendorf et al. 2000; Petrylak 2013). Therapies available for CRPC are namely second generation anti-androgens (i.e., enzalutamide, abiraterone acetate (Scher et al. 2010; Armstrong and Gao 2015; Attard, Belldegrun and de Bono 2005; Tran et al. 2009), and taxane based chemotherapies (Kroon et al. 2016), however, they are only effective for a limited amount of time with most patients acquiring resistance and dying within 2 years (Rescigno et al. 2012; Armstrong and Gao 2015; Lorente and De Bono 2014). Therefore, there is an urgent need for new therapies targeting CRPC and more importantly, to prevent or inhibit prostate cancer metastasis and growth.

Chapter 1: Literature Review 3 1.2 Epithelial-mesenchymal plasticity (EMP) and prostate gland development.

The epithelial-mesenchymal transition (EMT) was first discovered for its paramount role during embryonic morphogenesis (Shook and Keller 2003; Thiery et al. 2009). EMT is characterised by the collaboration of biochemical and phenotypic changes as cells transform from an epithelial profile to a mesenchymal one. This transition sees the loss of epithelial markers, cell-cell adhesions and apico-basal polarity. Concomitantly, newly transitioned cells gain mesenchymal markers and features including motility, invasiveness, and an elongated cell shape (Greenburg and Hay 1982; Thiery et al. 2009). Cell plasticity involved in embryonic development is orchestrated by organised waves of EMT and its reverse, mesenchymal to epithelial transition (MET) to determine cell differentiation and fate. The term MET has been routinely used to describe the reversal of EMT; this thesis uses both the terms MET and mesenchymal to epithelial reverting transition (MErT) to refer to EMT reversal. Collectively, the ability of cells to transition between epithelial and mesenchymal states is termed herein as epithelial- mesenchymal plasticity (EMP).

A critical phase in early development is the process of gastrulation, which gives rise to three germ layers: the ectoderm, mesoderm, and endoderm (Grant and Kyprianou 2013). This process relies on a transforming growth factor beta (TGF-β)- induced EMT and the expression of transcription factors (TF) Snail1 and Snail2 (Thiery et al. 2009; Grant and Kyprianou 2013; Arnold et al. 2008). Specifically, Snail1 is considered to be a master regulator of this process as it upregulates invasion-related genes such as fibronectin, vimentin, Zeb1, and matrix metalloproteases (MMPs) 2 and 9 (Baritaki et al. 2009). Indeed, loss of Snail1 leads to failed gastrulation (Acloque et al. 2009; Carver et al. 2001; Nieto et al. 1994). Following gastrulation, a primary EMT takes place to form the neural crest with a secondary EMT to generate mesodermal cells. These cells then undergo a mesenchymal to epithelial reverting transition (MErT) to become the notochord, somites, and the precursors to the urogenital system (Thiery et al. 2009). The development and differentiation of the prostate gland is dependent on androgens stimulating the androgen receptor (AR). This interaction initiates budding of the urogenital sinus epithelium (UGE) into the surrounding urogenital sinus

Chapter 1: Literature Review 4 mesenchyme (UGM) (Figure 1.2) (Marker et al. 2003; Cunha and Lung 1978). This interaction is critical as failure to do so, for instance by a non-functional AR, results in the absence of a prostate gland and testicular feminisation-like syndrome (Hayward and Cunha 2000). The UGE buds then elongate, forming ducts that begin to branch once they reach the mesenchymal pads of the UGM. Meanwhile, progenitor cells differentiate into epithelial basal or luminal cells by the expression of the AR and a coordinated change in the types of cytokeratin expressed (Prins and Putz 2008). Concurrently, contact of the UGE buds with the UGM stimulates smooth muscle differentiation and the arrangement of mesenchymal cells in a distinctive pattern along the basement membrane. Further cell differentiation takes place inside the ducts to form mature fibroblasts that express vimentin. The prostate gland continues to branch out postnatally and reaches maturation during puberty when circulating androgen levels rise rapidly. Taken together, the development of the prostate gland involves the precise coordination of EMP and androgen signalling.

Figure 1.2. Development of the prostate gland in a C57BL/6J mouse embryo.

A. The urogenital sinus (UGS) lies between the bladder and the urethra. B. Formation of the urogenital epithelium (UGE) buds (prostatic buds). The UGE is

Chapter 1: Literature Review 5 stained blue and the urogenital mesenchyme (UGM) white. (Sourced from (Schneider, Branam and Peterson 2014))

1.3 EMP in cancer. In a cancer context, the term EMT generally refers to phenotypic switching of cancer cells from an epithelial to a mesenchymal state characterised by loss of cell- cell adhesion and the acquisition of invasive and aggressive properties. EMT has been proposed and supported by multiple studies, to be re-activated in cancer to facilitate its progression and cancer metastasis (Thiery et al. 2009; Tsai and Yang 2013) For instance, the transcription factor SRY-related high-mobility-group box 9 (SOX9) typically expressed by foetal basal prostate cells during UGE budding, was found to be expressed in a subset of prostate cancer cells with increased expression in relapsed hormone-refractory prostate cancer (Wang et al. 2008). Further investigation revealed that the over-expression of SOX9 in PCa xenografts enhanced their invasive abilities, growth and angiogenesis, overall implying the latent activation of the EMT program in prostate cancer progression (Wang et al. 2008).

The Wnt signalling pathway is another developmental pathway involved with EMP that is thought to become reactivated in cancer. A number of proteins involved in the Wnt signalling pathway such as β-catenin, cyclin D1, and glycogen synthase kinase have been shown to regulate AR-driven transcriptional activity (Robinson, Zylstra and Williams 2008; Terry et al. 2006). Specifically, β-catenin is a protein that interacts with E-cadherin and the cytoskeleton to modulate cell behaviour. Expressed in the membrane of non-invasive cells, β-catenin translocates to the cytosol and the nucleus to activate Wnt signalling and invasion (Polette et al. 2007; Kypta and Waxman 2012). Β-catenin has also been shown to bind to the AR, enhancing its transcriptional activity (Yang, Chen, et al. 2006; Lee, Ha and Logan 2015). This enhances expression of the AR-dependent genes mouse mammary tumour virus (MMTV) and PSA where their increased expression contributes to an invasive and malignant prostate cancer phenotype (Yang et al. 2002). Many studies have associated elevated β-catenin and AR expression with CRPC that may explain the emergence of this phenotype (Whitaker et al. 2008; Jaggi et al. 2005; Patriarca et al. 2003; Assikis et al. 2004; Chen et al. 2004; Yokoyama et al. 2014). Indeed, the Wnt signalling pathway has been shown to be one of the most mutated pathways in

Chapter 1: Literature Review 6 CRPC patient samples, with a specific increase in the expression of the Wnt pathway regulator genes such as adenomatous polyposis coli (APC) (Hieronymus and Sawyers 2012; Grasso et al. 2012; Robinson et al. 2015). Other studies report decreasing β-catenin expression with increasing Gleason score (Whitaker et al. 2008; Horvath et al. 2005) supporting that the downregulation of β-catenin/AR signalling in localised prostate cancer allows for tumorigenesis. While more research is required to clarify the role of β-catenin in prostate cancer progression, evidence supports the interaction of β-catenin with the AR and parallel activation of the Wnt pathway, and thus EMP.

Furthermore, examination of prostate cancer tissue following radical prostatectomy revealed that the elevated expression levels of the EMT inducing transcription factor (EMT-TF) Twist and the mesenchymal marker Vimentin were independent predictors of biochemical recurrence (Behnsawy et al. 2013; Wu et al. 2014; Raatikainen et al. 2015). Similarly, loss of epithelial marker E-cadherin was associated with increased Gleason score and poor patient outcome (Whiteland et al. 2013). These findings support the involvement of EMT in prostate cancer progression as the downregulation of E-cadherin and the upregulation of vimentin are commonly used hallmarks of EMT.

However, the involvement of EMP in cancer progression has been questioned on many fronts with most stemming from the inherent difficulty of tracking such events in pathological samples. This is partly due to the heterogeneity of patient samples, which results in variable expression of EMT markers (Rodriguez-Gonzalez et al. 2013; Alkatout et al. 2013). A key issue is the lack of a standardised definition of what markers constitute an EMT. To date, the assessment and extent of EMT relies on the expression of one or a few genes, which limits the comparability between studies (Chui 2013; Nieto 2011). Furthermore, most markers associated with EMT do not discern between mesenchymal tumour cells and benign mesenchymal cells such as fibroblasts in the stroma (Tarin, Thompson and Newgreen 2005; Garber 2008; Ledford 2011; Chui 2013). Adding to the complexity is also the question as to whether identified mesenchymal tumour cells are mesenchymal by origin, acquired mutations over time, or arose through an EMT.

Chapter 1: Literature Review 7 There is compounding evidence that primary tumours and circulating tumour cells may present with EMT features. However, metastases often reflect the epithelial phenotype of the primary tumour, suggesting that if EMT is indeed involved in metastatic progression a reversal must occur (Chaffer and Weinberg 2011; Gingrich et al. 1996; Kallakury et al. 2001; Rubin et al. 2001; Thiery 2002; Satelli et al. 2015; Tsai et al. 2012; Ocana et al. 2012). The proposition that a MErT may be involved in this process has renewed the interest in studying EMP rather than EMT as a unidirectional program. However, the limited evidence available for either EMT or MErT occurring in pathological samples calls for further research directed at finding unique markers that can identify these transitions.

1.4 Regulation of the Epithelial to Mesenchymal Transition (EMT). EMT is a complex program that involves the cooperation of multiple signalling pathways and regulators (Zheng and Kang 2014; De Craene and Berx 2013; Thiery et al. 2009; Tsai and Yang 2013). This section will provide a general overview of key markers that constitute EMT, how EMT is regulated, and what induces an EMT in a cancer context.

1.4.1 EMT markers. EMT is predominantly identified by the absence (or downregulation) of epithelial and the presence (or upregulation) of mesenchymal-related markers, which ultimately define the identity of a cell. One of the first alterations during an EMT is the loss of cell-cell adhesion and polarity proteins such as E-cadherin, tight junction protein 1 (TJP1) and epithelial cell adhesion molecule (EpCAM) (Figure 1.4) (Thiery et al. 2009). Multiple studies have identified the downregulation of E- cadherin as a molecular hallmark of EMT (Thiery et al. 2009). The downregulation of E-cadherin can be mediated via direct transcriptional repression (Batlle et al. 2000; Cano et al. 2000; Hajra, Chen and Fearon 2002), promoter methylation (Graff et al. 1995; Kanai et al. 1997; Saito et al. 1998) or post-translational protein phosphorylation or degradation (Bachelder et al. 2005; Zhou et al. 2004; Lester et al. 2007). Also, the downregulation of E-cadherin is often mirrored by a switch from type I and II cytokeratin intermediate filaments to type III vimentin intermediate filaments that enhance cell integrity, flexibility and motility (Mendez, Kojima and Goldman 2010). EMT is mostly known for bestowing cells an invasive phenotype

Chapter 1: Literature Review 8 and therefore proteins that facilitate cell migration, and invasion are also considered EMT markers (Figure 1.4). Amongst them are the secreted proteins fibronectin and a number of MMPs such as MMP2, 9, and 13, that degrade the extracellular matrix (ECM) and promote cell migration and invasion (Hynes and Yamada 1982; Radisky and Radisky 2010). Another mesenchymal protein associated with the invasive phenotype is N-cadherin, and its upregulation is concomitant with the downregulation of E-cadherin, often referred to as “cadherin switching” (Gravdal et al. 2007; Zhang, Liu, et al. 2013). Cell surface markers such as CD44 and integrin β6 are also associated with invasive and aggressive features related to EMT (Cho et al. 2012; Kuo et al. 2009; Bates et al. 2005). While these markers denote a classic and full EMT, it has been shown that cells are able to transition into an intermediate or “metastable” phase, whereby both epithelial and mesenchymal associated proteins are expressed (Figure 1.3) (Hazan et al. 1997; Lee et al. 2006; Jordan, Johnson and

Abell 2011), highlighting the full spectrum of epithelial plasticity (Tan et al. 2014).

Figure 1.3. The spectrum of epithelial – mesenchymal plasticity.

Epithelial cells transition into a mesenchymal phenotype by the progressive loss of epithelial markers and acquisition of mesenchymal markers. This process is reversible and constitutes the epithelial-mesenchymal plasticity spectrum. (Adapted from (Kalluri and Weinberg 2009)).

Chapter 1: Literature Review 9 1.4.2 EMT regulators. EMT is regulated by embryonic transcription factors (TFs) that coordinate gene expression of epithelial and mesenchymal genes. There are currently three known superfamilies of transcription factors, the Snail, Zeb1, and basic/helix-loop-helix families (Figure 1.4).

1.4.2.1 The Snail family. The Snail family of TFs is comprised of three members: SNAI1 (Snail), SNAI2 (Slug), and SNAI3 (Smuc) (Nieto 2002) (Figure 1.4). In relation to EMT, Snail and Slug have been the most studied family members with Smuc only recently being shown to be potentially implicated in EMT (Zhao et al. 2013; Kataoka et al. 2000). This superfamily has a highly conserved cluster of zinc fingers at the C-terminal that bind directly to the E-pal sequence “CAGGTG” within the E-box of the E-cadherin promoter to repress its expression (Mauhin et al. 1993; Kataoka et al. 2000; Hajra, Chen and Fearon 2002; Dominguez et al. 2003; Mingot et al. 2009). Between Snail and Slug, Snail has a stronger binding affinity to these sequences than Slug (Bolos et al. 2003; de Herreros et al. 2010).

Figure 1.4. Domain structure of the EMT-TFs Zeb1/2, Snail1/2/3 and Twist1/2.

(Sourced from (Sanchez-Tillo et al. 2012))

Chapter 1: Literature Review 10 The Snail family also shares a highly conserved SNAG domain at the N-terminal end that regulates binding to Sin3A/HDAC1/HDAC2, Ajuna-PRMT5-PRC2, and LSD1- coREST complexes that in turn mediate repression of E-cadherin (Peinado et al. 2004; Hou et al. 2008; Herranz et al. 2008). The composition of Snail and Slug differs in the proline-serine rich region with Snail having a serine-rich domain and a nuclear export sequence (Dominguez et al. 2003; Zhou et al. 2004) whereas Slug has a C-terminal binding protein binding site (CtBP) and a SLUG domain (Nieto 2002; Sanchez-Tillo et al. 2012). Both CtBP and SLUG domains have been shown to be involved in efficient E-cadherin repression (Molina-Ortiz et al. 2012; Tripathi et al. 2005). Snail forms a complex with SMAD3/4 that enhances repression of E- cadherin and can also repress other epithelial gene promoters including tight junction proteins CAR and ZO-1, occludin, claudin-1, and claudin-3 (Vincent et al. 2009; Ohkubo and Ozawa 2004; Martinez-Estrada et al. 2006). The repression of epithelial proteins initiates a cascade affecting the expression or stabilisation of downstream epithelial and mesenchymal markers. For instance, downregulation of E-cadherin by Snail or Slug deregulates the E-cadherin/β-catenin complex, releasing β-catenin to the cytoplasm (Medici, Hay and Olsen 2008; Zheng et al. 2013). Cytoplasmic β-catenin then becomes stabilised by GSK-3β and translocates to the nucleus to forms a complex with T-cell factor 4 (TCF-4) that in turn activates the promoter of TGF-β3 (Eger et al. 2000; Zheng et al. 2013). Synthesised TGF-β3 then cooperates with the SMAD signalling pathway to produce lymphoid enhancer factor 1 (LEF-1) that also binds with β-catenin (Medici, Hay and Olsen 2008; Eger et al. 2000; Bienz and Clevers 2000). This correlates with the upregulation of mesenchymal markers vimentin, fibronectin, uPAR, metalloproteinases, and the downregulation of E-cadherin and cytokeratin (Medici, Hay and Olsen 2008; Eger et al. 2000; Gilles et al. 2003; Mann et al. 1999; Caraci et al. 2008), leading to an EMT. Snail and Slug themselves can also be modulated post-transcriptionally by alteration of their protein stabilisation and localisation. Similar to β-catenin, GSK-3β can also phosphorylate Snail protein by translocating it to the cytoplasm where its function is inactivated (Zhou et al. 2004; Zheng et al. 2013). PKD1 can also phosphorylate Snail and transport it from the nucleus to the cytoplasm, render it inactive (Du et al. 2010; Zheng et al. 2014). Conversely, phosphorylation by PAK1 translocates Snail from the cytoplasm to the nucleus, activating the repressive functions of Snail (Yang et al.

Chapter 1: Literature Review 11 2005). Snail has been shown to destabilise cell polarity by transcriptionally repressing the Crumbs3 and human lethal giant larvae homologue 2 (HUGL2) promoters, which abolishes the expression of Par, Crumbs, and HUGL2 at the tight junctions (Whiteman et al. 2008; Aigner et al. 2007; Lamouille, Xu and Derynck 2014; Harder et al. 2012). Snail also interacts with the MAPK and phosphoinositide 3-kinase (PIK3) signalling pathways to activate the promoter of MMP9, a protein that degrades the ECM to pave the way for cell migration and invasion (Jorda et al. 2005; Smith et al. 2014). Lastly, Snail activates other EMT-TFs including Slug, Zeb1 and Zeb2 that in turn mediate and sustain an EMT (Sanchez-Tillo et al. 2012).

1.4.2.2 The Zeb family. The Zeb family is made up of the two members Zeb1 and Zeb2. The Zeb family are of zing-finger/homeodomain type and are highly conserved across species (Gheldof et al. 2012). They are large transcription factors, with sequences spanning 1124 amino acids for Zeb1 and 1214 amino acids for Zeb2 (Figure 1.4) (Sanchez- Tillo et al. 2012). Within their sequences, they have multiple independent domains that interact with other transcriptional regulators (Sanchez-Tillo et al. 2011; Gheldof et al. 2012). Zeb family members initiate an EMT similar to the Snail family, by binding to the E-box sequence of the E-cadherin promoter to repress its expression (Comijn et al. 2001; Peinado, Olmeda and Cano 2007; Grooteclaes and Frisch 2000; Wong, Gao and Chan 2014). They additionally suppress the expression of other cell- cell adhesion proteins such as plakophilin 2, connexin 26, and ZO-3 (Vandewalle et al. 2005). Zeb1 also destabilises cell polarity by binding to the E-box sequences of the Crumbs3 and HUGL2 to repress their expression (Aigner et al. 2007; Spaderna et al. 2008). Furthermore, the Zeb family can control members of the miR-200 family, forming a double-negative feedback loop where their repression also leads to the downregulation of E-cadherin and EMT (Christoffersen et al. 2007; Bracken et al. 2008; Burk et al. 2008; Gregory et al. 2008; Korpal and Kang 2008; Korpal et al. 2008; Park et al. 2008; Kim et al. 2011; Chen et al. 2014; Brabletz et al. 2011).

1.4.2.3 The basic/helix-loop-helix (bHLH) family. The bHLH family of transcription factors includes numerous transcription factors that contain a bHLH domain. The most extensively characterised members of this family in the context of EMT are Twist1 (Yang et al. 2004) and Twist2 (Fang et al.

Chapter 1: Literature Review 12 2011) (Figure 1.4). Twist1 and Twist2 share a “Twist box” at the C-terminus, which functions both as a gene suppressor and activator (Franco et al. 2011). Twist members regulate gene expression in multiple ways; they can directly bind to other transcriptional regulators, dimerise with other proteins, or mediate post-translational modifications (Sanchez-Tillo et al. 2012). Twist proteins activate expression of N- cadherin, AKT2 or Gli1 by binding to the E-box sequence of their promoters (Alexander et al. 2006; Cheng et al. 2007). Suppression of E-cadherin by Twist1 is achieved indirectly by activating expression of BMI1, which in turn binds to the PRC1 and PRC2 elements of the E-cadherin promoter (Yang et al. 2010). Additionally, Twist1 can also repress E-cadherin by activating Snail proteins (Smit et al. 2009; Casas et al. 2011). Furthermore, Twist1 influences gene repression by inhibiting the activity of other transcription factors and cofactors such as Runx2 and p300/pCAF (Yang et al. 2011; Hamamori et al. 1999). Twist1 is also thought to contribute to the activation of N-cadherin and the repression of E-cadherin by binding to H4K20 methyltransferase SET8 (Yang et al. 2012). The characterisation of the upstream signalling pathways Twist elicits is not as comprehensive as with pathways elicited by the Snail and Zeb families. However, Twist is often upregulated by multiple EMT-related pathways and thus plays an integral role in the mediation of EMT.

1.4.3 EMT inducers. Genetic alterations of EMT-TFs have not been linked with predisposing cancer cells to undergo an EMT, probably due to their necessary role in embryogenesis. Rather, cancer cells are thought to undergo an EMT in response to a number of extracellular signals from the tissue microenvironment. These extracellular signals often modulate critical developmental pathways related to EMT such as the Wnt, Notch, tumour related TGF-β and other growth factor signalling cascades. TGF-β is thought to be one of the primary inducers of EMT in carcinomas as it is secreted by stromal fibroblasts in the tumour extracellular matrix (Hanahan and Weinberg 2011; Katsuno, Lamouille and Derynck 2013). Secreted TGF-β binds to its type II receptor (TGF-βRII) and trans phosphorylates the type I receptor (TGF-βRI) (Shi and Massague 2003). This phosphorylates SMAD2 and SMAD3 proteins, which then form complexes with SMAD4. Accumulation of these complexes in the nucleus collaborate with other TFs to regulate gene expression such as core EMT-TFs

Chapter 1: Literature Review 13 Snail1/2, Zeb1/2 and Twist1, leading to an EMT (Thiery 2009; Eckert et al. 2011; Shi and Massague 2003). However, TGF-β signalling involves the cooperation of multiple other pathways such as the activation of the Ras kinase cascade or the Wnt/β-catenin/LEF-1 pathway. Other growth factors including fibroblast growth factors (FGF and FGF2), epidermal growth factor (EGF), insulin-like growth factors (IGF-I and IGF-II), platelet-derived growth factor (PDGF), and hepatocellular growth factor (HGF) have also been shown to activate EMT in cancer cells (Gonzalez and Medici 2014). For instance, EGF has been demonstrated to promote the endocytosis of E-cadherin, and induce expression of Snail and Twist1 proteins that concurrently repress the expression of E-cadherin and other epithelial targets (Lee, Chou, et al. 2008; Lo et al. 2007; Lu et al. 2003). IGF-I binding to the IGF-I receptor stimulates expression of Zeb1 in prostate cancer cells (Graham et al. 2008). IGF-I has also been found to activate expression of Snail and the NFκB pathway in mammary epithelial cells, inducing them to undergo an EMT (Graham et al. 2008; Kim et al. 2007). Collectively, growth factors activate multiple signalling pathways and TFs that lead to an EMT (Thiery 2002; Gonzalez and Medici 2014).

In addition to growth factors, inflammatory cytokines can also induce an EMT in cancer cells. For instance, tumour necrosis factor α (TNFα) activates the NFκB pathway which stabilises the protein expression of Snail1 to maintain an EMT (Wu et al. 2009). In parallel, TNFα also induces expression of Twist1 via the activation of NFκB p65 and IKK-β (Li et al. 2012). Other cytokines such as interferons (IFNs) can activate Stat3 through JAK kinase, leading to the expression of Twist1 (Darnell, Kerr and Stark 1994). Lastly, hypoxic regions within cancers have been associated with the induction of EMT. Hypoxia induces the expression of hypoxia-inducing factor 1 (HIF1) that in turn activates Snail and Twist1 expression, leading to an EMT (Yang et al. 2008; Mak et al. 2010).

In summary, EMT can be induced or sustained by various cues coming from the tumour microenvironment.

Chapter 1: Literature Review 14 1.5 The role of EMP in the metastatic cascade. Cancer metastasis is the leading cause of death for patients with carcinomas such as that of the breast, colon, lung and prostate (Gwak et al. 2014; Li et al. 2015; Shiota et al. 2013; Yao et al. 2014). Cancer metastasis involves a complex series of events whereby cancer cells acquire the ability to detach from the primary site, invade the surrounding stroma, intravasate into the circulation, survive the transit, and finally extravasate at the secondary location where disseminated cancer cells either lie dormant for an undetermined amount of time and/or go ahead and grow into a secondary tumour (Figure 1.5) (Kalluri and Weinberg 2009; Thiery et al. 2009; Chaffer and Weinberg 2011). The hypothesised steps involved in the metastatic cascade are continuously being elucidated with new mechanisms identified, enriching its complexity. There are three speculated methods by which cancer cells can metastasise. Firstly, the growth of primary cancer with or without physical trauma leads to fragmentation, whereby cancer fragments enter the vasculature and lodge at a secondary site. Secondly, epithelial cells can move as a cohesive sheet as seen during the developmental processes of gastrulation and epiboly (Warga and Kimmel 1990) but also shown in cancer metastasis to the lymph node (Giampieri et al. 2009).

Figure 1.5. (A-F) The metastatic cascade.

(sourced from (Chaffer and Weinberg 2011)).

Chapter 1: Literature Review 15 Lastly, it is hypothesised that cancer cells can dissociate from their neighbouring cells and transiently or permanently undergo an epithelial to mesenchymal transition (EMT) to acquire mobile and invasive properties (van Zijl, Krupitza and Mikulits 2011; Giampieri et al. 2009; Nieto et al. 2016). Over the last 20 years, research supports the involvement and spatiotemporal regulation of EMT in mediating some of the steps involved in the metastatic cascade (Celia-Terrassa et al. 2012; Das et al. 2014; Ocana et al. 2012; Kalluri and Weinberg 2009; Thiery et al. 2009; Tsai et al. 2012; Tsuji et al. 2008; Thiery and Chopin 1999; Nieto et al. 2016; Ye and Weinberg 2015). Overall, EMP appears to play a pivotal role in facilitating the multiple steps involved in carcinoma metastasis. The following sections will give a brief overview of the involvement of EMP in these steps.

1.5.1 EMP and tumour heterogeneity. Solid tumours are heterogeneous in composition and are comprised of cancer cells demonstrating varying degrees of differentiation as well as multiple other types of cells. The interactions between the various cell types amongst themselves as well as with the microenvironment can influence their epithelial plasticity which in turn increases the heterogeneity of the primary tumour (Pereira et al. 2015; Nieto et al. 2016). Studies by Malek et al. (Malek et al. 2011) demonstrate increased tumour heterogeneity when the tumour reaches the metastatic stage. Cell plasticity can be altered either via intrinsic or extrinsic factors (Marjanovic, Weinberg and Chaffer 2013). Intrinsic to the cell, mutations in oncogene or tumour suppressor genes elicit a number of biochemical changes, including motility, changes in metabolism, chemo- resistance, proliferation and cell-cell interactions (Boehm and Hahn 2011; Lamouille, Xu and Derynck 2014; Salk, Fox and Loeb 2010). Alterations in cell-cell interactions and the release of paracrine factors can, in turn, trigger the tumour microenvironment to recruit new cell types, form new blood vessels and remodel the extracellular matrix to accommodate the growing tumour (Alphonso and Alahari 2009).

Alternatively, the tumour microenvironment itself can have profound effects on cellular plasticity (Lu et al. 2008; Kenific, Thorburn and Debnath 2010). An unfavourable microenvironment (i.e lack of nutrients or oxygen) may force cancer cells to adapt accordingly, by either differentiating to a state that is compatible with

Chapter 1: Literature Review 16 the new microenvironment, moving to a more favourable microenvironment or attempt to signal for a change in the microenvironment (Gilkes, Semenza and Wirtz 2014; Hanahan and Weinberg 2011; Peinado, Lavotshkin and Lyden 2011). The study of tumour heterogeneity and epithelial plasticity is inherently challenging as it is difficult to separate the various cell types. One way to observe epithelial plasticity as a response to the tumour microenvironment is by immunohistochemical staining of epithelial and mesenchymal proteins at the primary site. Studies in prostate and colorectal cancer show that generally, the centre of the primary tumour mass expresses epithelial marker E-cadherin while the invasive front lacks E-cadherin and expresses the mesenchymal marker vimentin (Sethi et al. 2010; Brabletz et al. 2001). A similar observation was made in colorectal cancer where Snail expression was present at the periphery of invasive tumours (Hoshino et al. 2009). In agreement, studies by Tsuji et al. show that when EMT and non-EMT cells are co-injected subcutaneously, the formed tumours had layers of EMT cells at the periphery and non-EMT cells in the centre (Tsuji et al. 2008).

Another feature that supports the occurrence of an EMT is tumour budding; an observation defined by the infiltration of a small number of cancer cells into the surrounding tumour stroma. These buds have been identified to have decreased expression of epithelial proteins E-cadherin, β-catenin, and proliferation marker Ki67, and increased expression of mesenchymal proteins vimentin and fibronectin compared to non-budding cells (Taira et al. 2012; Yamaguchi et al. 2010; Liang et al. 2013; Zlobec and Lugli 2010; Lugli, Karamitopoulou and Zlobec 2012; Ohike et al. 2010; Dawson and Lugli 2015). Tumour buds with EMT features have been identified across multiple cancers including colon (Zlobec and Lugli 2010; Dawson and Lugli 2015), ampullary (Ohike et al. 2010), lung (Taira et al. 2012; Yamaguchi et al. 2010), breast (Liang et al. 2013), and pancreas (Karamitopoulou et al. 2013), and are considered to be an independent prognostic factor for these cancer types. Collectively the expression profile of the tumour buds is concomitant to an EMT- like profile.

Overall, increased expression of EMT markers at the primary site is linked with tumour heterogeneity and cancer aggressiveness.

Chapter 1: Literature Review 17 1.5.2 EMP and metastatic dissemination. Early studies identified the phenomenon of cancer cells shedding from solid tumours into the circulation (Paget 1989; Ashworth 1869), and since then the identification and enumeration of circulating tumour cells (CTCs) has been established as prognostic for metastatic disease in multiple carcinomas including breast, prostate and colon cancer (Banys-Paluchowski et al. 2015; Cohen et al. 2009; Cristofanilli et al. 2005; de Bono et al. 2008; Nicolazzo and Gradilone 2015; Scher et al. 2009; Bidard et al. 2014). Initially, CTCs were thought to be solely of an epithelial nature as metastases often reflect an epithelial phenotype. EpCAM was, and is still, used as an identifier to differentiate between blood cells and CTCs (Trzpis et al. 2007). However, non-EpCAM based isolation of CTCs such as size differentiating microfluidics or negative enrichment with removal of CD45 positive leukocytes has revealed their heterogeneous nature (Hiley et al. 2014; Martelotto et al. 2014) with a subpopulation of CTCs lacking EpCAM expression and thus escaping detection using EpCAM based techniques (Gorges et al. 2012; Konigsberg et al. 2011). Indeed CTCs from breast and prostate cancer patients have been shown to co-express EMT-related markers such as E-cadherin, vimentin, N-cadherin and cytokeratin (Aktas et al. 2009; Armstrong et al. 2011; Bednarz et al. 2010; Joosse et al. 2012; Khoo et al. 2015; Yu et al. 2013). Furthermore, interactions between CTCs and platelets provide CTCs with a protective cloak against immune surveillance and thus promote metastasis (Gay and Felding-Habermann 2011; Stegner, Dutting and Nieswandt 2014; Li and King 2012). Furthermore, platelets have been shown to secrete TGF-β and PDGF, both of which promote EMT (Ahmad et al. 2011; Labelle, Begum and Hynes 2011) and a mechanism to how CTCs maintain or acquire EMT- like expression while in the circulation.

Studies by Celia-Terrassa et al. (Celia-Terrassa et al. 2012) and Tsuji et al. (Tsuji et al. 2008) have demonstrated a cooperation between epithelial-like (EpCAM high) and mesenchymal-like (EpCAM low) cancer cells that resulted in enhanced metastatic potential. Tsuji et al. showed that hamster cheek pouch carcinoma (HCPC-1) cells that underwent EMT via p12CDK2-AP1, a downstream effector of TGF- β, could invade locally from the primary site but failed to form distant metastasis even when injected intravenously. In their study, they showed that non-EMT cells alone could not intravasate from the primary injection site, but a mix of EMT and

Chapter 1: Literature Review 18 non-EMT cells enabled intravasation of both EMT and non-EMT cells with only the non-EMT cells forming metastasis. They attributed this observation to the lack of adherent properties of EMT cells whereas non-EMT cells had the ability to adhere to the vessel walls and thus extravasate.

Further supporting these observations, studies by Celià-Terrassa et al., show that only PC-3 cells with an epithelial profile (PC-3/Mc) were capable of forming metastasis even though they lacked invasive properties in vitro. On the other hand, the more mesenchymal PC-3 cells (PC-3/S) displayed increased invasive capabilities but could not form distant metastasis. However, when PC-3/Mc cells were co- cultured with PC-3/S cells or murine fibroblasts, the PC-3/Mc cells temporarily gained invasive properties indicating that PC-3/Mc cells could be influenced by both tumoural and non-tumoural cells. Indeed, shortly after orthotopic transplantation of PC-3/Mc, they showed loss of E-cadherin and were able to produce metastasis with the hypothesis that their temporary gain of invasive properties from murine factors allowed for intravasation into the circulation. Intravenous injection of the PC-3/Mc cells alone resulted in bone and lymph node metastases. However, co-injection with PC-3/S cells showed enhanced metastasis to the bone and lymph nodes, earlier colonisation to the lungs, as well as newfound colonisation in the adrenal glands. It was noted that the bone microenvironment consisted of permeable sinusoidal capillary systems, which allowed cancer cells with relatively little invasive properties to colonise. Therefore, their results indicated that the cooperation between PC-3/S and PC-3/Mc enhanced PC-3/Mc extravasation to organs with a tougher capillary barrier than bone.

As previously mentioned, studies by Labelle et al. show that contact between platelets and tumour cells can induce/maintain EMT by the secretion of TGFβ, PDGF and VEGF from platelets (Labelle, Begum and Hynes 2011). In conjunction with the findings outlined above by Celia-Terrassa et al. and Tsuji e al, the EMT- inducing molecules secreted by platelets may also contribute to CTC extravasation. Generally, metastases have an epithelial phenotype and a transient EMT at the sites of intravasation and extravasation could explain that observation. Further supporting the reversion back to an epithelial phenotype at the secondary site are studies by Gao et al. (Gao et al. 2012), showing that the presence of versican at the metastatic niche

Chapter 1: Literature Review 19 promotes a MErT and the re-acquisition of proliferation. Specific to prostate cancer and liver metastases, studies by Yates et al., suggest that the interaction of prostate cancer cells with liver cells downregulates autocrine EGF receptor signalling, thus promoting the re-expression of E-cadherin and MErT (Yates et al. 2007).

Other studies supporting the transient involvement of EMT in cancer metastasis were carried out by Tsai et al., (Tsai et al. 2012) who demonstrated that the constitutive over-expression of EMT-TF Twist1 at the primary site produced fewer metastases when compared to a reversible Twist1 expression. However, it is unclear as to which location the EMT reversal enhanced metastatic colonisation; whether it was necessary to occur while cells were in the bloodstream or already at the secondary site. Additionally, it is unknown whether there was cooperation of cell types at the primary or intravenous sites as seen previously by Celià-Terrassa et al. (Celia-Terrassa et al. 2012), and Tsuji et al. (Tsuji et al. 2008). Providing some insights on this, studies by Ocañia et al. (Ocana et al. 2012) show that intravenous injection of Prrx1 expressing BT549 breast cancer cells only formed metastases following the loss of Prrx1 expression. However, once again, it was unclear as to which location the EMT-reversal was critical.

Recent publications have reinvigorated the debate over the functional role of EMT in cancer dissemination (Fischer et al. 2015; Zheng et al. 2015), however, the involvement of epithelial plasticity in dissemination and/or CTCs is inherently implied whether cancer cells alter their phenotype to escape the primary tumour, later on in the circulation, or at the secondary site. While more research is required to clarify specific details on the timing and involvement of EMT in cancer, many studies support the notion that the regulation of EMT is closely related to metastatic colonisation.

1.5.3 EMP and the metastatic niche. It is still unclear why some cancer types have a tendency to metastasise to preferential organs. For instance, prostate cancer patients tend to develop bone metastasis (Bubendorf et al. 2000; Shah et al. 2004). One theory is the famous “seed and soil” theory by Stephen Paget (Paget 1989) where circulating cancer cells or “seeds” preferentially metastasise to pre-determined favourable microenvironments

Chapter 1: Literature Review 20 or “soil” and that the dissemination is not random. For instance, progressive ovarian cancer cells can grow in the peritoneal cavity and attach to the visceral lining but not metastasise to other visceral organs. Interestingly, when ascites fluid containing ovarian cancer cells was drained directly back into to the circulation via the jugular vein, patients did not present with a higher risk of developing metastases outside of the peritoneum and lung metastasis was rare (Tarin et al. 1984a, 1984b). Specific to prostate cancer, approximately 90% of patients with CRPC examined post-mortem had bone metastasis (Bubendorf et al. 2000; Higano 2004; Scher et al. 2005; Gartrell and Saad 2014). A humanised in vivo mouse model generated by Holzapfel et al., (Holzapfel et al. 2014) was able to replicate this apparent “homing” behaviour and while the reason why prostate cancer cells preferentially go to the bone is still unclear, studies are examining whether disruption of the bone microenvironment or metastatic niche may prevent metastasis (Wang et al. 2014). Other studies by Gao et al. (Gao et al. 2012) identified that the presence of versican at the metastatic niche could induce mesenchymal cancer cells to reacquire their proliferative abilities via a hypothesised MErT, thus promoting metastatic growth. In a similar fashion, interactions between EMT cells and hepatocytes induced the re-expression of E- cadherin by downregulating the EGF signalling pathway (Yates et al. 2007). Adding to the complexity of metastasis, disseminated cancer cells have been found in the parenchyma of secondary tissues and the bone marrow and have been observed to lay dormant for an undetermined amount of time before regrowth (Hedley and Chambers 2009). This observation parallels a number of studies supporting that disseminated mesenchymal cancer cells require a MErT to re-instate proliferation (Ocana et al. 2012; Tsai et al. 2012). In summary, while it is not entirely understood why cancer cells metastasise to preferred secondary locations, indirect evidence at the metastatic niche suggests that epithelial plasticity is involved in the establishment of metastases.

1.5.4 EMP and metastatic colonisation. The involvement of EMT in metastatic colonisation was initially questioned because the metastases often reflected the epithelial phenotype of the primary tumour. However, studies by Bonnomet et al. showed that while primary MDA- MB-468 breast tumour xenografts and resulting lung metastases shared a heterogeneous expression of vimentin, the circulating tumour cells had uniformly

Chapter 1: Literature Review 21 high levels of EMT markers vimentin, Snail1 and Snail2 (Bonnomet et al. 2012). These results imply that the EMT-like CTCs altered their epithelial/mesenchymal balance at the secondary site in order to form metastases with heterogeneous vimentin expression, sparking interest in the dynamic nature of EMT. Indeed, studies reveal the spatiotemporal regulation of EMT in vivo to promote metastatic colonisation and growth. In particular, elegant studies by Tsai et al. using an inducible Twist1 mouse model bearing skin tumours, show that the continuous expression of Twist1 led to local invasion and circulating tumour cells but not the formation of overt metastases (Tsai et al. 2012). However, reversible Twist1 expression resulted in a significantly higher number and size of metastases than continuous EMT. In agreement, studies by Ocaña et al. show enhanced metastasis when Prrx1-expressing BT549 breast cancer cells lose Prrx1 expression following tail vein injection (Ocana et al. 2012). Outside of using reversible systems, studies by Beerling et al., (Beerling et al. 2016) support that both epithelial and mesenchymal cancer cells escape the primary site, however at the secondary site mesenchymal cancer cells revert to their epithelial phenotype within two cell divisions to resume proliferation. Together, these studies support the involvement of EMP in cancer metastasis.

A number of studies have linked EMT with directly repressing cell division. For instance, EMT induced by Snail, Slug, or Zeb1, inhibits cyclin D1 activity (Mejlvang et al. 2007; Vega et al. 2004; Liu et al. 2010). Additionally, EMT cells from various in vivo studies show decreased cell proliferation and cell cycle marker expression such as Ki-67, CDK18, and CCNA2, when compared to non-EMT cells (Fischer et al. 2015; Tsai et al. 2012; Zheng et al. 2015; Ocana et al. 2012; Celia- Terrassa et al. 2012). These observations could explain the presence of disseminated tumour cells found to lay “dormant” at the secondary sites. These dormant cells are hypothesised to be the seeds of future metastasis (Hedley and Chambers 2009), implying that they would have to undergo a MErT to reinitiate tumour growth. However, it is currently unclear what promotes an EMT reversion, whether it is the absence of an EMT stimulator or an environmental cue that induces a MErT. Overall, studies support the regulation of EMT, and the restoration of epithelial characteristics play a pivotal role in metastatic establishment and growth.

Chapter 1: Literature Review 22 1.6 EMT and therapy resistance. The majority of cancer-related deaths are due to tumours acquiring resistance to the available treatments, leading to the uncontrollable growth of metastases. A number of studies have linked EMT with chemoresistance. For example, colorectal cancer cells that were either rendered oxaliplatin resistant (Yang, Fan, et al. 2006) or irradiated (Kawamoto et al. 2012), displayed with EMT-like characteristics such as EMT-related markers and morphological changes. Additionally, tumour samples were taken from colorectal cancer patients that had received pre-operative radiation and chemotherapy (intravenous 5-fluorouracil and oral uracil and tegafur) for one week, presented with a more mesenchymal signature than those from untreated patients (Kawamoto et al. 2012). Similar results were observed in patients with non- small cell lung cancer (NSCLC) that had pre-operative chemoradiotherapy (Shintani et al. 2011), and in breast cancer patients post chemotherapy with postletrozole (Creighton et al. 2009). Consistent with these observations, overexpression of Snail or TGF-β in colorectal cancer cells induced an EMT and increased their chemoresistance to 5-fluorouracil induced apoptosis (Hoshino et al. 2009; Papageorgis et al. 2011). Additionally, studies using EMT lineage tracing in a mouse model of spontaneous breast-to-lung metastasis, identified EMT cells to survive treatment by a number of chemotherapeutic drugs such as cyclophosphamide, doxorubicin, paclitaxel, and fluorouracil, compared to non-EMT cells (Fischer et al. 2015). Conversely, inhibition of Snail or Twist in a mouse model of pancreatic ductal adenocarcinoma (PDAC) increased sensitivity to gemcitabine (Zheng et al. 2015). EMT is thought to confer chemoresistance by decreasing cell proliferation (Vega et al. 2004; Mejlvang et al. 2007; Fischer et al. 2015) and increasing resistance to apoptosis (Inoue et al. 2002; Vega et al. 2004; Kajita, McClinic and Wade 2004; Fischer et al. 2015) as well as upregulating other chemoresistance- related genes (Fischer et al. 2015). As chemotherapeutic drugs are designed to induce apoptosis in highly proliferative cells (Kaufmann and Earnshaw 2000; Hartwell and Kastan 1994), the traits bestowed by EMT interfere with the drug’s mode of action of these drugs (Fischer et al. 2015). Therefore, the targeting of EMT has been proposed as a method to reverse/inhibit EMT-induced chemoresistance, leading to increased efficacy of the chemotherapeutic drugs. However, as discussed in previous sections of this review, an EMT-reversal approach needs to be carefully

Chapter 1: Literature Review 23 considered as it could have deleterious effects by re-initiating metastatic growth in distant disseminated EMT-induced dormant tumour cells.

Similar to the ability of chemotherapeutic drugs to induce an EMT, the administration of androgen deprivation therapy (ADT) to patients with advanced prostate cancer has been identified to also induce an EMT. Specifically, studies by Sun et al. demonstrate ADT to induce EMT in both human LuCaP35 prostate cancer explants and normal mouse prostates, implying a role for androgens in maintaining the epithelial phenotype (Sun et al. 2012). This EMT was shown to be regulated by a negative feedback loop between the androgen receptor and Zeb1, and was accompanied by the upregulation (i.e Zeb1, Zeb2, Twist1, Snail1, Slug, vimentin, N- cadherin) and downregulation (i.e E-cadherin) of classic EMT markers (Sun et al. 2012; Zhu and Kyprianou 2010). Other studies reveal that prolonged treatment of LNCaP95 cells with enzalutamide conferred resistance by activation of Snail and elevation of AR (Ware et al. 2016). Furthermore, ADT-induced EMT was also associated with stem-cell-like features such as increased surface CD44 and NANOG expression (Shang et al. 2015; Sun et al. 2012; Jeter et al. 2011). ADT is the standard of care for many patients, including patients with metastatic disease regardless whether they have responded to local therapy, because it decreases tumour burden and prolongs patient life. However, patients invariably recur with CRPC. Taken together, the studies above imply that ADT generates a population of resistant prostate cancer cells that initiate tumour recurrence. As discussed previously, the emergence of CRPC marks the terminal stage for patients as therapies are no longer effective. This highlights the need to investigate tumour plasticity further and to identify new therapeutic avenues in combating cancer progression.

1.7 EMT and cancer stem cells (CSCs). Cancer stem cells (CSCs), also known as tumour initiating cells (TICs), have been characterised as tumour cells that can grow tumours following serial transplantation. CSCs are identified by expression of a number of surface markers such as CD44high, CD24low, CD133high, as well as regulatory TFs including Notch1, NANOG, SOX2, and OCT4. However, the expression of these markers is cancer type dependent. Studies by Mani et al., and Morel et al. were the first to demonstrate that EMT induced by TGF-β, Snail, Twist, or Zeb1, inferred CSC-like characteristics

Chapter 1: Literature Review 24 to human mammary epithelial cells (Morel et al. 2008; Mani et al. 2008). The first direct link between CSCs and EMT was made by Vesuna et al., where they reported that Twist1 can directly suppress CD24 in breast cancer cells (Vesuna et al. 2009). They also showed that decreasing the expression of Twist1 in Twist1-overexpressing cells only partially reverted their stem cell-like molecular signature. In agreement, studies by Schmidt et al., demonstrate human mammary epithelial cells to acquire stable stem-cell-like properties following long-term transient expression of Twist1 (Schmidt et al. 2015). Other studies show that basal breast cancer non-CSCs can generate CSCs de novo where their plasticity is regulated by the chromatin state of the Zeb1 promoter (Chaffer et al. 2013). Specific to prostate cancer, PC-3 cells induced to undergo EMT via PDGF expressed the stem cell-related markers Notch1, OCT4, NANOG, SOX2 and Lin28B, had enhanced cologenicity and increased tumourigencity in mice (Kong et al. 2010). Taken together, these studies support that EMT generates cancer cells with stem cell-like traits.

However, there are also a number of studies contradicting that EMT infers tumour initiating features. As discussed previously, EMT decreases cell division with studies supporting that a reversion is required for its re-initiation (Beerling et al. 2016; Celia-Terrassa et al. 2012; Ocana et al. 2012; Tsai et al. 2012). Taken together, it is clear that EMT is context and cell type dependent, whereby EMT generates “classic” CSCs with tumour initiating properties in some cell types and others an EMT reversion is required to initiate tumour growth. Further research is necessary to define the distinction between the regulation of CSC, TIC, and EMT properties.

1.8 Study aims

1.8.1 Rationale Uncontrollable metastatic tumour burden is the leading cause of mortality in patients with carcinomas, including prostate cancer. The regulation of EMT is thought to facilitate the multiple steps involved in cancer metastasis, including therapy resistance. However, there is debate over the clinical relevance of EMT and MErT due to the lack of specific markers that can discern cancer cells that have experienced these transitions from the surrounding tumour stromal cells. This is aided by the shortage of appropriate models that allow for the temporal profiling of

Chapter 1: Literature Review 25 cells as they dynamically transition between epithelial and mesenchymal states. Therefore, to better understand the full spectrum of EMP and its clinical relevance, prostate adenocarcinoma cells were generated to reversibly express master EMT-TFs Snail, Slug, or Zeb1.

1.8.2 Hypotheses In context of the literature surrounding EMP in cancer progression, the following hypotheses were formed:

1. Increased EMP in the primary tumour is associated with poor patient outcome. 2. Epithelial prostate cancer cells that have experienced a reversible EMT acquire a transcriptional footprint of the event.

1.8.3 Aims This study has the following aims:

Aim 1: To generate and characterise reversible EMT models regulated by the EMT-TFs Snail, Slug, or Zeb1, in the prostate adenocarcinoma cell line LNCaP. (Chapter 3)

Aim 2: To globally transcriptionally profile a reversible Snail-induced EMT in LNCaP cells. (Chapter 4)

Aim 2A: To identify the transcriptional footprint of LNCaP cells having experienced a reversible EMT. (Chapter 4)

Aim 2B: To examine the clinical relevance of EMP in clinical carcinoma specimens. (Chapter 4)

Aim 3: To establish an in vivo EMP model for prostate cancer metastasis. (Chapter 5)

Chapter 1: Literature Review 26

Chapter 2: Materials and Experimental Methods

Chapter 2: Materials and Experimental Methods 27 2.1 Tissue culture. LNCaP cells (American Type Culture Collection (ATCC), VA, USA) were cultured using phenol red free Roswell Park Memorial Institute-1640 media (RPMI- 1640) (Life Technologies; CA, USA) supplemented with 5% foetal bovine serum (FBS) (Life Technologies), 1% penicillin (Life Technologies), 1% streptomycin (Life Technologies) and 0.01% gentamicin (Life Technologies). Human embryonic kidney cells (HEK293T) (ATCC) were cultured using Dulbecco’s Modified Eagle Medium (DMEM) (Life Technologies) supplemented with 10% deactivated FBS (dFBS), 1% penicillin, 1% streptomycin and 0.01% gentamicin. Cell cultures were maintained at 37°C with 5% CO2 and media was refreshed every 2-3 days. Cell cultures were passaged by washing with phosphate buffered saline (PBS) (Life Technologies), detaching the cells with 0.05% Trypsin Ethylenediaminetetra-acetic acid (TE) (Life Technologies) and resuspending the cells in fresh growth media. Cell numbers were calculated using a haemocytometer or the TC20 Automated Cell Counter (Bio-rad, CA, USA) and viability estimated with Trypan Blue (Life Technologies) exclusion.

2.2 pINDUCER20 vector system. The pINDUCER20 vector system (kindly provided by Dr Thomas Westbrook, Baylor College of Medicine, Houston, Texas) (Figure 2.1) is a lentiviral construct for the Doxycycline (Dox)-inducible expression of cDNA (Figure 2.2). The cDNA of Snail, Slug, or Zeb1 or green fluorescent protein (GFP) was cloned into the pINDUCER20 vector backbone using Gateway Recombination™ (Life Technologies, ) (Figure 2.3) between the pINDUCER20 entry vector and pENTR223 donor vectors containing the cDNA of either Snail, Slug, Zeb1 (Thermofisher Scientific, VIC, AUS) or green fluorescent protein (GFP; kindly provided by Dr Thomas Westbrook, Baylor College of Medicine, Houston, Texas) (Figure 2.4). The optimal Dox concentration to use for successful induction of the pINDUCER20 TRE2 promoter was 1000ng/ml. This was previously validated by Meerbrey et al. (Meerbrey et al. 2011) as well as confirmed in Chapter 3, section 3.2.1.

Chapter 2: Materials and Experimental Methods 28

Figure 2.1. Plasmid map of the pINDUCER20 construct.

(image sourced from https://www.addgene.org/44012/)

Chapter 2: Materials and Experimental Methods 29

Figure 2.2. Schematic diagram and mechanism of the pINDUCER-20 vector backbone.

Once the pINDUCER-20 vector backbone is stably transfected into the host DNA, the Ubc promoter is constitutively active and rtTA3, IRES and Neo are continuously being expressed. The TRE2 promoter is not active at this stage because the rtTA3 cannot bind to the TRE2 promoter when Dox is absent. Once Dox is added, Dox interacts with rtTA3, modifying its structure. The complex of rtTA3 and Dox can then bind to the TRE2 promoter and induce expression of cDNA. The cDNA of interest can be cloned between the recombination sites attR1 and attR2 using Gateway® Cloning. Removal of Dox results in the cessation of cDNA expression.

Chapter 2: Materials and Experimental Methods 30

Figure 2.3. Schematic of the generated pINDUCER20 constructs.

Chapter 2: Materials and Experimental Methods 31

Figure 2.4: Plasmid map of the pENTR™ 223.1 plasmid.

Chapter 2: Materials and Experimental Methods 32 2.3 HEK293T transfection and lentivirus production. The HEK293T cells were seeded at a density of 1.5 x 106 cells per 10 cm petri dish and incubated in their regular growth media (DMEM, 10% dFBS) overnight at

37°C with 5% CO2. Transfection mixtures containing 12 µL FuGENE-6® (Promega, NSW, AUS), 1.8 µg pCMV-deltaR8.2 lentiviral packaging vector (Addgene, MA, USA), 0.2 µg pCMV-VSV-G viral envelope expression vector (Addgene) and 2.0 µg pINDUCER20 expression vector, containing the cDNA of the gene of interest of either Snail, Slug, Zeb1 or GFP, was added onto HEK293T cell cultures overnight. The culture media was then replaced with RPMI-1640 supplemented with 5% FBS and HEK293T culture supernatants containing viral particles were collected at 48 and 72 hours post transfection. Viral supernatants were then filtered using 0.45 µm filters (Pall Corporation, VIC, AUS), and stored at -80°C until further use.

2.4 Cell transduction and infection. Filtered viral supernatants for each pINDUCER20 expression vector (2 mL), along with protamine sulphate (8 µg/mL; Sigma-Aldrich, AUS), were added to 40-

50% confluent LNCaP cells and incubated overnight at 37°C with 5% CO2. Following incubation, the viral supernatants were removed from the cell cultures and were replaced with regular growth medium to allow the cells to recover. Successfully transduced cells with pINDUCER20 vectors (LNCaP-iSnail, LNCaP- iSlug, LNCaP-iZeb1, LNCaP-iGFP) were then selected using neomycin sulphate (neomycin; G418; 500-1000 µg/mL; Life Technologies) until all non-transduced cells died. This was confirmed by adding neomycin to non-transduced LNCaP cells in parallel as a positive control for neomycin-induced cell death.

2.5 Three-dimensional (3D)-on-top Matrigel™ assays. 3D-on-top Matrigel™ assays were performed to assess invasion (Petersen et al. 1992) (Figure 2.5.2). The even coating was achieved by allowing the Matrigel to polymerise during a 15-minute centrifugation, 270g at room temperature. Once polymerised, five thousand cells were suspended in regular growth media containing 2.5% Matrigel™ and seeded over the 100% gelled Matrigel™.

Chapter 2: Materials and Experimental Methods 33

Figure 2.5. Schematic diagram of a 3D-on-top Matrigel™ assay.

Cells are seeded in a suspension with their growth media supplemented with 2.5% Matrigel™ over a bed of 100% gelled Matrigel™. Cells are allowed to grow for 10+ days to form multicellular spheroids. Adapted from Debnath et al., (Debnath, Muthuswamy and Brugge 2003).

Chapter 2: Materials and Experimental Methods 34 Duplicate wells were used for each treatment. Cells were incubated at 37°C 5%

CO2 for 7-10 days to allow for the formation of multicellular spheroids. Media containing 2.5% Matrigel™ was refreshed every 2 days. Once spheres were formed, Dox (1000ng/mL; Sigma-Aldrich) was added to the test wells to induce the expression of either Snail or GFP. Control wells were maintained without Dox.

2.6 Immunofluorescence staining. Cells were fixed with 200 µL of 4% paraformaldehyde in PBS for 30 minutes and then carefully washed twice with 400 µL PBS (Sigma-Aldrich). For cells seeded in 3D-on-top Matrigel™ assays, the cells were carefully washed with PBS: Glycine instead (PBS, 0.1M Glycine, pH 7.4). Cells were then treated with 200 µL of 0.4% Triton-X in PBS (Sigma-Aldrich) for 10 minutes to permeabilise the cells followed by a 1-hour incubation with 200 µL of 5% bovine serum albumin (BSA) (Merck- Calbiochem, Kilsyth, VIC, AUS) to block non-specific binding, at room temperature. Cells were then incubated with human monoclonal primary antibodies (antibody diluted 1:100 in 5% BSA) overnight at 4°C (or room temperature if staining 3D Matrigel™ assays). The following day, cells were washed three times with 0.1% PBS-Tween-20 (PBS-T) and incubated with fluorescently conjugated secondary antibodies for one hour in the dark. AlexaFluor® 633 phalloidin (1:40 PBS) (Life Technologies) was added to cells for staining actin and left for 40 minutes in the dark. Cells were washed with PBS-T three times and incubated with 4’,6-diamidino-2-phenylindole (DAPI) (1:10,000 in PBS) (Life Technologies) for 10 minutes. After washing cells twice with PBS, chambers were removed from the slides and the slides were mounted using Prolong Gold® (Life Technologies). The slides were kept at room temperature in the dark until imaging with either a Zeiss LSM 510 confocal microscope (Zeiss, NSW, AUS) or a DeltaVision OMX Imaging System (GE Healthcare, VIC, AUS)

Primary antibodies used: rabbit anti-Snail (Cell Signalling Technology, Clone C15D3; Beverly, MA, USA), mouse anti-E-cadherin (BD Biosciences, Clone 36), and mouse anti-vimentin (Sigma-Aldrich, Clone V9). Secondary antibodies used: AlexaFluor® 568 goat anti-rabbit IgG and AlexaFluor® 488 goat anti-mouse IgG antibodies (Life Technologies).

Chapter 2: Materials and Experimental Methods 35 2.7 RNA extraction and cDNA preparation. RNA was extracted from cell pellets using TRIzol reagent (Life Technologies) and purified using the Direct-zol RNA MiniPrep kit (Zymo Research, CA, USA) following the manufacturer’s protocol. First strand cDNA synthesis was performed using Superscript III first strand cDNA synthesis kit (Life Technologies). Briefly, cDNA was synthesised by adding 50 pM of Oligo(dT) and random hexamers to 1000ng of RNA diluted in RNase-free water to 11µL. Following a 5 minute incubation at 65°C, a 7µL mixture of 10x reverse transcriptase (RT) buffer, Dithiothreitol (DTT; 0.1M), RNaseOUT™ and SuperScript III ® RT was added. Using a thermal cycler machine (C1000 Touch™ Thermal Cycler, Bio-rad; NSW, AUS), contents were incubated at 25°C for 10 minutes, 50°C for 30 minutes and 85°C for 5 minutes. The mixture was then incubated at 37°C for 20 minutes and synthesised cDNA stored at -20°C until further use.

2.8 Quantitative real-time polymerase chain reaction (qRT-PCR). To analyse mRNA levels, primer pairs were added to 100ng of cDNA to a final concentration of 0.4µM. All reactions were performed in triplicate in a 384-well plate (Life Technologies) using SYBR® Green (Life Technologies) and a 7500fast Real-time PCR System (Life Technologies). PCR amplification was performed following an initial 10-minute denaturation step at 95°C with 40 cycles at 95°C for 15 seconds and 60°C for 60 seconds. A melt curve was included in each run for quality control. Gene expression was quantified using the 2-ΔΔCt method relative to untreated cells (Livak and Schmittgen 2001). RPL32 levels were used to normalise the cDNA loading amounts. Specific primer sequences are indicated in Table 2.1.

2.9 Protein extraction and Western blotting. Protein was extracted by adding radioimmunoprecipitation (RIPA) protein lysis buffer (Sigma-Aldrich) containing phosphatase and protease inhibitor cocktails (Sigma-Aldrich), to the cell pellets. Protein concentration was then quantified using Coomassie Plus® (Thermo Fisher Scientific; VIC, AUS). Proteins were resolved on pre-made NuPAGE® gradient SDS-PAGE gels (4%-12%) (Life Technologies) and transferred to methanol activated PVDF membrane (Millipore, VIC, AUS) at 200 mA in transfer buffer (25mM Tris, 40 mM glycine and 10% methanol) for 2 hours at

Chapter 2: Materials and Experimental Methods 36 Table 2.1 Table of forward and reverse primer sequences used for qRT-PC of individual genes.

Forward 5’ 3’ Reverse 5’ 3’ CCNB1-F AGAGCCATCCTAATTGACTG CCNB1-R CAACCAGCTGCAGCATCTTC CDKN3-F GGAAGAGCTTACAACCTGCC CDKN3-R ACAGGTATAGTAGGAGACAAGC FOXM1-F CTCCTTCTGGACCATTCACC FOXM1-R CCAAGTGCTCGGGCAATTGT MKI67-F CAAATTACAAGACTCGGTCCCTG MKI67-R GGGAGGTCTTCATGGGCTTC CTGF-F CCTGGTCCAGACCACAGAGT CTGF-R TGGAGATTTTGGGAGTACGG SOX9-F AGCGCCCCCACTTTTGCTCTTT SOX9-R CCGCGGCGAGCACTTAGGAAG ZEB1-F CAACTACGGTCAGCCCT ZEB1-R GCGGTGTAGAATCAGAGTC SLUG-F GGGGAGAAGCCTTTTTCTTG SLUG-R TCCTCATGTTTGTGCAGGAG CDH1-F TGCCCAGAAAATGAAAAAGG CDH1-R GTGTATGTGGCAATGCGTTC SNAI1-F CCTCCCTGTCAGATGAGGAC SNAI1-R CCAGGCTGAGGTATTCCTTG NRP1-F AGGACAGAGACTGCAAGTATGAC NRP1-R AACATTCAGGACCTCTCTTGA VIM-F GAGAACTTTGCCGTTGAAGC VIM-R GCTTCCTGTAGGTGGCAATC ESRP1-F TCCTGCTGTTCTGGAAAGTCG ESRP1-R TCCGGTCTAACTAGCACTTCGTG ESRP2-F GGGTCTGGGAAGTCAAGACAATG ESRP2-R CTTCGAAAACAATTGACTGCTGG EPCAM-F TGCTCAAAGCTGGCTGCCAAATG EPCAM-R GTGCCGTTGCACTGCTTGGC RPL32-F GCACCAGTCAGACCGATATG RPL32-R ACTGGGCAGCATGTGCTTTG

4°C. Membranes were then blocked with 5% skim milk in tris-buffered saline with Tween-20 (TBST) (1M Tris, 1.5M NaCl, 0.01% Tween-20) for 1 hour at room temperature followed by overnight incubation at 4°C with primary antibody in blocking buffer (1:1000 in either 5% skim milk or 5% BSA in TBST Table 2.2). Membranes were then washed 6 times over 30 minutes with TBST and incubated with the relevant horseradish peroxidise-conjugated secondary antibodies (anti-rabbit IgG (Cell Signalling Technology) and anti-mouse IgG (Cell Signalling Technology) diluted 1:5000 in 5% skim milk in TBST for 1 hour at room temperature. After further washing with TBST (6 times), protein bands were detected using enhanced chemiluminescence as per manufacturer’s protocol (GE Healthcare) and the signal detected using the ChemiDoc™ system (Bio-rad). The blots were stripped using Restore Plus® stripping buffer (Thermo Fisher Scientific) at 60°C for 50 mins and re-probed for GAPDH, -tubulin, or COXIV expression to validate and assess sample loading.

Chapter 2: Materials and Experimental Methods 37 Table 2.2. Table of primary antibodies and the concentrations used for Western blotting.

Dilution (of Antigen Clone Species stock) Company Cell Signalling Snail C15D3 Rabbit 1:5000 Technologies Cell Signalling Slug C19G7 Rabbit 1:1000 Technologies Zeb1 D80D3 Rabbit 1:3000 Sigma-Aldrich Cell Signalling GFP D5.1 Rabbit 1:1000 Technologies E-cadherin 36 Mouse 1:2500 BD Biosciences Vimentin V9 Mouse 1:500 Sigma-Aldrich Cell Signalling EpCAM D1B3 Rabbit 1:1000 Technologies β-catenin 15B8 Mouse 1:10000 Sigma-Aldrich Fibronectin 10/Fibronectin Mouse 1:1000 BD Biosciences GAPDH GAPDH-71.1 Mouse 1:10000 Sigma-Aldrich Cell Signalling COXIV 3E11 Rabbit 1:5000 Technologies -tubulin GTU-88 Mouse 1:4000 Sigma-Aldrich

2.10 Cell viability and proliferation assays. Cell viability assays were performed using the PrestoBlue® Cell Viability (Life Technologies), a resazurin-based reagent, according to the manufacturer’s protocol. Cell proliferation was assessed by imaging and enumeration of DAPI (Life Technologies) stained nuclei using the Cytell Cell Imaging System (GE Healthcare) at the indicated time points.

2.11 Flow cytometry. For cell cycle analyses, cells were processed, and DNA content analysed using flow cytometry as described previously (Sadowski et al. 2014). The percentage of

Chapter 2: Materials and Experimental Methods 38 cells in each cell cycle phase was calculated with ModFit LT (Verity Software House) based on DNA histograms of 20,000 cells per treatment. Protein biomarker detection was performed on cells harvested from cultures using Accutase® (Sigma- Aldrich). Cells were then fixed with 4% paraformaldehyde and permeabilized with 100% ice cold methanol. Cells at a density of 1x106 cells were fluorescently stained with Alexa Fluor® 647 anti-human Ki-67 (Biolegend, Clone Ki-67) primary antibody for 30 minutes in the dark. Following 3 washes with phosphate-buffered saline (PBS), cells were resuspended in PBS with 2% FBS. Cells were analysed immediately using an FC500 flow cytometer (Beckman Coulter, QLD, AUS) and data analysed using FACS Express software (DE Novo Software).

2.12 Seahorse assay to measure metabolic parameters. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using Seahorse XF24 Extracellular Flux analyser (Seahorse Bioscience, MA, USA). LNCaP-iGFP and -iSnail cells were maintained in RPMI with 5% FBS or treated for 5 days with Dox or 5 days, followed by 10 days in the Dox-free medium. Each treatment was prepared to be ready on the day of analysis. One day before analysis, cells were seeded into 24-well Seahorse plates at 3x105 cells per well with Dox treatments maintained, and an additional group receiving Dox upon seeding (Dox 1 Day). On the day of analysis, the medium was replaced with basal assay medium comprised of unbuffered DMEM (Sigma-Aldrich, D5030) supplemented with glucose (11.1 mM), glutamax (2 mM), sodium pyruvate (1 mM); pH 7.4 for 1hr in a non-CO2 incubator at 37°C. OCR and ECAR were measured in basal conditions and following the sequential addition of 1.2µM oligomycin (injection 1), 1µM carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP) (injection 2) and 1µM each of antimycin A and rotenone (injection 3). At the end of the protocol, total DNA was analysed using the CyQuant DNA quantification kit (Life Technologies) against a standard curve of DNA from a known number of LNCaP cells. Oxygen consumption (pmol/min) and acidification rate (mpH/min) were normalized to cell number.

2.13 Microarray gene expression profiling and analysis. For sample preparation, LNCaP-iSnail and LNCaP-iGFP cells were treated with 1µg/ml of Dox for 5 days, followed by its removal for 20 days. Triplicate samples

Chapter 2: Materials and Experimental Methods 39 for each condition were collected at the indicated time points (No Dox, EMT at 5 days, and MErT at 3, 5, and 20 days) and prepared for microarray profiling as described previously using a custom 180K Agilent array platform (Agilent-027516 VPC Human 180K v2; GPL14873) (Sieh et al. 2012). Microarray raw data were processed using the Agilent Feature Extraction Software (v10.7) as described elsewhere (Sieh et al. 2012). Differential expression was determined using a Bayesian-adjusted t-statistic from a Linear Models for Microarray Data (LIMMA) linear model. The gene expression data have been submitted to Gene Expression Omnibus (GEO) with the accession number GSE80042. Gene expression was considered significant if fold change was ≥1.5 and p <0.05 (adjusted for a false discovery rate (FDR) of 5%). These parameters were chosen so to allow for the detection of minor gene changes that may contribute to overall pathway changes. Before further analysis all probes significantly (fold ≥1.5; p < 0.05) altered by Dox treatment, as compared to non-treatment groups, in the LNCaP-iGFP model at any time point were removed. The filtered gene lists were examined by Ingenuity Pathway Analysis (IPA, Ingenuity Systems Inc.) for functional annotation and gene network analysis. Gene set enrichment analysis (GSEA; http://www.broad.mit.edu/gsea) was used to identify enrichment of gene signatures contained in the Molecular Signatures Database (mSigDB) and metastatic castrate- resistant prostate cancer (mCRPC) samples from GSE35988 (Grasso et al. 2012). Gene set permutation analysis was performed using the “weighted” enrichment statistic and the signal-to-noise metric for gene ranking. Transcript clusters were identified using the pattern matching function within The National Institute on Aging (NIA) Array analysis tools (http://lgsun.grc.nia.nih.gov/ANOVA/). Additional GO- BP and pathway analysis were performed using MetaCore™ (GeneGo; Thomas Reuter) and Ingenuity Pathway Analysis (Qiagen). The microarray data were uploaded to Oncomine™ v4.5 (www.oncomine.com) and overlaid with published microarray datasets using the Concept Analysis tool. Six datasets from the Oncomine™ concept analysis table were selected for testing the prognostic qualities of the MPS. These datasets were two prostate cancers datasets from the Glinsky et al, and Setlur et al studies (Glinsky et al. 2004; Setlur et al. 2008); two breast cancer datasets from the van de Vijver et al, and van’t Veer studies (van de Vijver et al.

Chapter 2: Materials and Experimental Methods 40 2002; van 't Veer et al. 2002); and two lung cancer datasets from the Okayama et al, and Lee et al studies (Okayama et al. 2012; Lee, Son, et al. 2008).

2.14 Statistical analysis. Significance was determined by one-way ANOVA or unpaired two-tailed Student’s t-test using the GraphPad Prism Software and a p < 0.05 was considered statistically significant. For survival analysis, Kaplan–Meier curves were drawn, and differences between the curves were calculated by the log-rank test using GraphPad Prism Software, whereby p < 0.05 was considered statistically significant. Fisher's Exact Test was used to determine enrichment of gene signatures in metastatic samples from prostate cancer patient cohorts (GSE3325, GSE68928, GSE6752, GSE6919, GSE21034 and GSE35988 downloaded from GEO). It is noted that the data obtained from GEO was re-analysed for differential gene expression applying the same methods used for analysing the microarray data generated in this study (Bayesian-adjusted t-statistic from a LIMMA) linear model). GSEA analysis generated a nominal p-value and FDR, reflecting the significance of gene set enrichment, estimated using a gene set-based permutation test. P <0.05 and FDR <0.05 was considered significant.

2.15 Generation of the LNCaP-iSnailRFP/LUC , LNCaP-iSlugRFP/LUC, and LNCaP- iGFPRFP/LUC cells. To be able to detect the reversible EMT models in vivo, the LNCaP-iSnail, LNCaP-iSlug and LNCaP-iGFP cell lines were transduced with the pMig-Luc2- DsRed plasmid (Figure 2.6) following the methods outlined in sections 2.3 and 2.4 (termed from here on LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP- iGFPRFP/LUC). The generated cell lines were then selected by sorting for top 40% of RFP expressing cells using a MoFlo Astrios (Beckman Coulter) cell sorter.

2.16 Intraprostatic injection procedure. Mice were anaesthetised using Ketamine (25mg/kg; ZebraVet, QLD, AUS) and Xylazine (5mg/kg; ZebraVet) before surgery. Once mice were completely unconscious, anaesthesia was maintained during surgery using isofluorane gas (2.5%; ZebraVet). A small incision of approximately 1 cm was made in the skin of the lower abdomen, and the abdomen cavity accessed following blunt dissection

Chapter 2: Materials and Experimental Methods 41 through the abdomen muscle. The bladder and prostate were exteriorized and 10ul containing either 2.0 x 105 or 1 x 106 cells (LNCaP-iSnailRFP/LUC, LNCaP- iSlugRFP/LUC or LNCaP-iGFPRFP/LUC ) was injected directly into the posterior gland of the prostate using a Hamilton needle. The incision was then closed using dissolvable stitches (ZebraVet) for the muscle layer and surgical staples (ZebraVet) for the skin layer. Surgical staples were removed a week later. The mice were given Temgesic (0.1mg/kg; ZebraVet) during surgery as well as the following day to relieve pain.

Figure 2.6. Schematic of the pMig-Luc2-DsRed plasmid.

Chapter 2: Materials and Experimental Methods 42 2.17 Bioluminescent imaging and tumour progression. Mice were imaged for bioluminescence emission every week using the IVIS Spectrum In Vivo Imaging System (IVIS; Perkin Elmer, VIC, AUS). Mice administered D-luciferin via intra-peritoneal injection with D-luciferin (15mg/mL; Perkin Elmer) at 10uL per gram of body weight. The mice were allowed to metabolise the luciferase for approximately 5 minutes prior to anesthetising with isofluorane gas (2.5%) in a chamber. Anaesthesia was maintained in the IVIS holding platform during imaging. Bioluminescent intensity was measured in photons/second using Living Image V4.1 imagining software (Perkin Elmer). Tumour progression was monitored via bioluminescence imaging weekly, and when the tumours reached a palpable size of > 1cm2, or the bioluminescent intensity surpassed 1 x 1010 photons/second, the mice were terminated on ethical grounds (Ethics approval number: 464/12).

2.18 Mouse termination and organ collection. Mice were initially administered D-luciferin via intra-peritoneal injection with D-luciferin and allowed to metabolise it for ~5 minutes. Mice were then anesthetised using isofluorane gas (2.5%) and while under anaesthesia an intracardiac terminal bleed was performed. Following the terminal bleed, mice were terminated via cervical dislocation. To detect metastatic lesions, mice organs were extracted and imaged individually using the IVIS for bioluminescence signal. Where the tumours were large enough, a portion was snap frozen for protein and RNA analysis, and the rest was processed for histological analysis. Otherwise, the whole organ was processed for histological analysis.

2.19 Tissue processing, embedding and sectioning. Tissues were fixed in 10% paraformaldehyde for ~48hrs followed by storage in ice cold 70% ethanol at least 24 hours. Tissued requiring decalcification, such as bones, were then stored in decalcifying Ethylenediaminetetraacetic acid (EDTA) agent for 4 weeks with frequent EDTA changes (3 times a week). Tissues were then processed in the Excelsior ES (Thermo Scientific) to be dehydrated and cleared through a series of graded ethanol and xylene baths (three changes of 70% ethanol for 1 hour, two changes of 90% ethanol for 1hr, three changes of 100% ethanol for 45 minutes and three changes of xylene at 45 minutes). Samples were then infiltrated

Chapter 2: Materials and Experimental Methods 43 with paraffin wax and embedded using a Histocenter 3 (Thermo Scientific) and sectioned using a rotary microtome (Leica RM 2265; Leica, NSW, AUS) at a thickness of 4µm. Sections were stretched on a 43°C water bath and mounted onto Ultra Superfrost glass slides (Thermo Scientific) where they were then baked at 65°C for at least 30 minutes.

2.20 Haematoxylin and Eosin staining (H&E). H&E staining was used to examine the architecture of the tissue and the cell morphology. Mounted sections were deparaffinised through a series of xylene baths (2 x 8 minutes) and rehydrated in graded ethanol (2 x 100% ethanol, 90% ethanol, 70% ethanol) for a minute each and tap water for 2 minutes. Sections were then stained with Mayers Haematoxylin (ProSciTech, QLD, AUS) for 8 minutes followed by bluing in Scotts Tap water (MgSO4, bicarb soda) for 30 seconds. Sections were then stained with Eosin (ProSciTech) for ~30 seconds before clearing by briefly rinsing in 90% ethanol followed by 2 rinses 100% ethanol and finally 2 x 5 minute xylene baths. Slides were coverslipped (Thermo Scientific) using Entellan mounting medium (ProSciTech). Slides were then allowed to dry overnight prior to imaging with an Olympus VS120 Virtual Slide Microscope and Scanner (Olympus, VIC, AUS)

2.21 Immunohistochemistry (IHC). IHC staining was performed using the Ventana automated staining module (Roche, NSW, AUS). Where possible, pre-made and optimised antibodies were purchased from Ventana and staining was performed following manufacturers protocol. Otherwise, antibodies were optimised using the Ventana manual titration protocol before using them with Ventana re-fillable antibody canisters. The antibodies used are listed in Table 2.3. Slides were then vigorously washed in warm soapy water for 1 minute, followed by thoroughly rinsing with water to remove the liquid coverslip. Slides were then cleared as described above in section 2.19. Staining for Snail protein was conducted by our collaborators Dr Antonio Garcià De Herreros and Dr Raùl Peña using their in-house produced monoclonal anti-mouse Snail antibody.

Chapter 2: Materials and Experimental Methods 44 Table 2.3 Table of primary antibodies and dilutions used for immunohistochemistry.

[Antibody] Antigen Clone Species Isotype Company (µg) Vimentin V9 Mouse IgG Pre-diluted Ventana Ki67 MIB-1 Mouse IgG Pre-diluted Ventana Zeb1 polyclonal Rabbit IgG 0.44 Sigma-Aldrich Cell Signalling Slug C19G7 Rabbit IgG 0.5 Technologies Cell Signalling GFP D5.1 Rabbit IgG 0.4 Technologies E-cadherin 36 Mouse IgG2aκ 0.083 BD Biosciences AR ar-441 Mouse IgG1 0.5 Abcam IgG1 - Mouse - As required Abcam IgG2aκ - Mouse - As required Abcam IgG - Rabbit - As required Abcam

Chapter 2: Materials and Experimental Methods 45

Chapter 3: Characterisation of reversible EMT models

Chapter 3: Characterisation of reversible EMT models 46 3.1 Introduction

The metastatic spread of cancer cells to critical organs marks the final disease stage for the majority of patients with carcinomas including prostate cancer. The largest gap in our understanding of cancer metastasis is how cancer cells gain invasive characteristics enabling their metastasis and survival in their local and distant microenvironments. This poses significant challenges when developing therapies to treat patients with aggressive disease. A mechanism that has been hypothesised to be employed by cancer cells to attain the features required for metastasis is the utilisation of the EMT and MErT programs (Thiery 2009). By altering their phenotype, cancer cells can gain the invasive and survival characteristics they need to dissociate from the primary tumour, survive transit within circulatory vessels, and form metastases in other vital organs. Specifically, it is hypothesised that epithelial cancer cells acquire invasive traits by transforming into mesenchymal cells via an EMT induced by environmental factors such as hypoxia or inflammation in the tumour microenvironment (Li et al. 2012; Yang et al. 2011; Mak et al. 2010). Within the bloodstream, it is hypothesised that the mesenchymal state may be maintained by factors present in the blood, such as platelet derived growth factor secreted by platelets (Ahmad et al. 2011; Labelle, Begum and Hynes 2011). Once cancer cells extravasate into distant organs, they have been observed to either lay dormant for an undetermined amount of time or resume growth, leading to metastases (Tsai and Yang 2013). Studies have associated EMT-induced mesenchymal cells with dormancy, which, coupled with the observation that metastases often reflect the epithelial phenotype of the primary tumour, it is hypothesised that cancer cells break dormancy by reverting to their epithelial phenotype via a MErT (Tsai et al. 2012; Ocana et al. 2012; Celia-Terrassa et al. 2012). Therefore, it is critical to be investigating both transitional programs in the order they are thought to occur when in the context of cancer metastasis, that is an EMT followed by a MErT.

To address this, the study herein generated inducible and reversible EMT models using the epithelial prostate adenocarcinoma cell line LNCaP. This allowed for exploring the full spectrum of epithelial plasticity in the same population of cancer cells via the reversible expression of master EMT transcription factors (EMT-TFs)

Chapter 3: Characterisation of reversible EMT models 47 SNAI1 (Snail) and SNAI2 (Slug), as well as Zeb1. This chapter reports the generation and characterisation of these unique models as tools for investigating epithelial-plasticity in prostate cancer.

Chapter 3: Characterisation of reversible EMT models 48 3.2 Results

3.2.1 Optimisation of Dox concentration to induce cDNA expression in the reversible EMT models. To determine the optimal Dox concentration to use for cDNA induction, LNCaP-iGFP cells were exposed to a range of Dox concentrations (62.5 – 2000ng/mL) for 3 days. It was noted that parental LNCaP cells emit an autofluorescence intensity of 3.42 MFI and untreated LNCaP-iGFP cells an intensity of 7.8 MFI indicating minimal GFP expression in the absence of Dox. Expression of GFP was analysed using fluorescence-activated cell sorting (FACS) where GFP expression reached an intensity plateau at 197 mean fluorescence intensity (MFI) with 1000ng/mL Dox (Figure 3.1A). It was then examined whether 1000ng/mL Dox was adequate to sustain longer term and stable GFP expression in LNCaP-iGFP cells treated for 7 days. GFP expression was confirmed using FACS by 1 day following Dox treatment, reached a plateau at 3 days and remained relatively constant over 7 days of Dox treatment (1000 ng/mL) (Figure 3.1B). Therefore, it was confirmed that Dox treatment induced the stable expression of GFP over 7 days.

Although it was determined that 1000ng/mL Dox was adequate for inducing maximum levels of GFP protein in the LNCaP-iGFP model, the Dox concentration curve was repeated with the LNCaP-iSnail cells to determine whether 1000ng/mL Dox would induce sufficient levels of SNAI1 cDNA to trigger EMT. To examine the extent of EMT activation the mRNA and protein expression levels of E-cadherin and vimentin were assessed. The expression of E-cadherin was investigated as it becomes directly repressed by Snail, and vimentin is generally used as a marker of the mesenchymal state. Untreated parental LNCaP cells were also included to assess the “leakiness” of Snail expression in the LNCaP-iSnail cells in the absence of Dox. It was observed that untreated LNCaP-iSnail cells had ~3 fold expression of Snail mRNA (Figure 3.2A) as compared to parental LNCaP cells. However, this expression was not sufficient to produce detectable Snail protein (Figure 3.3B). Moreover, there were no significant alterations in the expression of E-cadherin and vimentin at either the mRNA or protein level (Figure 3.2A & B) in non-treated LNCaP-iSnail cells as compared to parental LNCaP cells.

Chapter 3: Characterisation of reversible EMT models 49

A B

38

2 5 0 29

1 9 7 1 9 7 t 2 0 0 1 8 8 n 19

1 8 0 u o

y 1 6 6 t

1 5 9 C i

s 10

n 1 5 0

e

t n

i 0 0 1 2 3 4 10 10 10 10 10

P 1 0 0 F

G G F P - F IT C 5 0 Sample Colour Median - FITC Geometric Mean - FITC 3 .4 2 7 .8 Negative Grey 3.46 3.5

0 No Dox Black 18.11 18

x

5

0

0

5

0

0

0

.

0

0

2

5

0

o

2

0

0

1 2

5 Dox +1 day Cyan 425.51 246.03

d

6

1 2

o Dox +3 days Green 685.39 382.1

n L N C a P -iG F P c e lls in Dox +5 days Purple 637.8 355.61 P d o x y c y c lin e (n g /m L )

a Dox +7 days Red 716.92 400.78

C

N L

Figure 3.1. Optimisation of the minimum effective dose of Dox to induce maximum expression of GFP in the LNCaP-iGFP cells.

A. LNCaP-iGFP cells were treated with 62.5 – 2000 ng/mL Dox for after 3 days. GFP intensity was analysed using fluorescently activated cell sorting (FACS); n = 3 technical replicates. B. LNCaP-iGFP cells were treated with 1000 ng/ml Dox for 7 days, and GFP intensity was measured using FACS.

Chapter 3: Characterisation of reversible EMT models 50

Chapter 3: Characterisation of reversible EMT models 51 Figure 3.2. Optimisation of the minimum effective dose of Dox to EMT-like marker expression in the LNCaP-iSnail model.

LNCaP-iSnail cells were treated with 62.5 – 2000 ng/mL Dox and RNA was collected after 3 days. A. Real-time PCR analysis of Snail, E-cadherin and vimentin genes. Data from each treatment day was normalised to RPL32 levels and expressed as the fold change in gene expression relative to parental LNCaP; n = 3 technical replicates. Error bars indicate SEM. One-way ANOVA, p-value: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001, ns: not significant (p>0.05). B. Protein expression of indicated proteins in parental LNCaP cells, untreated LNCaP-iSnail cells and LNCaP-iSnail cells treated with 62.5 – 2000 ng/mL Dox. GAPDH was used as a loading control.

Chapter 3: Characterisation of reversible EMT models 52 With the addition of Dox to the LNCaP-iSnail cells, it was observed that even at the lowest Dox concentration of 62.5ng/mL, there was a robust expression of Snail mRNA and protein, which led to the downregulation of E-cadherin, and upregulation of vimentin mRNA expression (Figure 3.2A). Both Snail and vimentin expression were consistent across all Dox concentrations. However, E-cadherin mRNA expression appeared to decrease continually with increasing Dox concentration (Figure 3.2A). Further examination of the expression of these markers at the protein level was performed via Western blot (Figure 3.2B). Similarly, robust upregulation of Snail protein expression was observed across all Dox concentrations (Figure 3.2B). Vimentin protein appeared to be variable across the increasing Dox concentrations. However, it was evident that E-cadherin expression reached a minimum at 1000ng/mL Dox. Therefore, it was decided that 1000ng/mL was the optimal concentration of Dox to use in subsequent experiments using the reversible EMT models.

3.2.2 Assessing the reversibility of the pINDUCER20 construct following removal of Dox. Next, the reversibility of the pINDUCER20 construct was tested by treating the LNCaP-iGFP and LNCaP-iSnail models with Dox for 5 days following removal for up to 5 days (Figure 3.4). GFP intensity was visualised using an epifluorescent microscope where the addition of Dox-induced robust expression of GFP after 1 day and was sustained over 5 days. Removal of Dox prompted a marked decrease in GFP expression after 1 day, which returned to pre-Dox treatment levels by day 5 (Figure 3.3A). No changes were observed in the morphology of the LNCaP-iGFP cells during or after Dox treatment. Similarly, the reversibility of GFP expression with the addition and removal of Dox was further confirmed using FACS (Figure 3.3B). The addition of Dox to LNCaP-iGFP cells for 5 days resulted in the GFP intensity increasing from 9 to 67 mean fluorescent intensity (MFI) (Figure 3.3B) with the removal of Dox restoring GFP intensity back to 8 MFI after 5 days. Similarly in the LNCaP-iSnail model, addition of Dox induced high levels of both Snail mRNA and protein (Figure 3.4A). Removal of Dox induced prompt reduction in Snail mRNA by 24 hours post Dox removal. Western blot analysis of Snail protein showed reduction following 24 hours of Dox removal and back to pre-treated levels by 3 days (Figure 3.4B). Taken together, these data demonstrate that the expression of cDNA in the

Chapter 3: Characterisation of reversible EMT models 53

Figure 3.3. Assessing the reversibility of the LNCaP-iGFP cell model.

LNCaP-iGFP cells were treated with Dox for 5 days and then Dox was removed for 5 days. A. Phase contrast images and epifluorescent images were recorded for untreated cells, cells treated with Dox for 1 and 5 days and then removed for 1 and 5 days. Epifluorescence exposure was 1 second. B. GFP intensity was measured using fluorescently activated cell sorting (FACS) for untreated cells, cells treated with Dox for 5 days and then removed for 5 days.

Chapter 3: Characterisation of reversible EMT models 54 A B

x 2 5 6

o N o D o x D

*

1 2 8 D o x + 1 D a y

o * *

D o x + 3 D a y s

N

x

x

6 4 o o

o * D o x + 5 D a y s

d

t

d

o n

D o x + 5 /-1 D a y s o

n

n

o 3 2

i

l P

D o x + 5 /-3 D a y s i

s a

a L N C a P -iS n a il L N C a P -iS n a il

s

n C

e 1 6 in 1 0 0 0 n g /m L d o x d o x r e m o v e d

S

r

N

i

-

L p

x 8 D a y s

e 1 3 5 7 1 3 5 7

A * * S n a il

N 4

R m

2

e G A P D H

v i

t 1

a

l e

0 .5 R S n a il

Figure 3.4. Assessing the reversibility of the LNCaP-iSnail cell model.

A. Real-time PCR analysis of Snail mRNA in LNCaP-iSnail cells treated with Dox for 1, 3, and 5 days, followed by removal for 1 and 3 days. Dotted lines indicate a fold change of ± 1.5; n = 3 technical replicates. Error bars indicate SEM. One-way ANOVA, p-value: * < 0.05, ** < 0.01. B. Protein expression of Snail in parental LNCaP cells, untreated LNCaP-iSnail cells, and LNCaP-iSnail cells treated with Dox for 1, 3, 5, and 7 days followed by removal for 1, 3, 5 and 7 days. GAPDH was used as a loading control.

Chapter 3: Characterisation of reversible EMT models 55 pINDUCER20 vector could be adequately controlled by the addition and removal of Dox from the cell culture media.

3.2.3 Assessing EMT-like changes in the LNCaP-iSnail, LNCaP-iSlug, and LNCaP-iZeb1 models following Dox treatment. Firstly, it was determined whether a classic EMT could be stimulated following the Dox-induced expression of either Snail or Slug following the addition of Dox to the culture media of the LNCaP-iSnail and LNCaP-iSlug models respectively. The cell models were subjected to a 5-day Dox treatment where alterations to cell morphology, as well as the mRNA and protein expression of known epithelial and mesenchymal markers, were assessed. The LNCaP-iGFP cells maintained typical LNCaP cell morphology following Dox treatment ( Figure 3.5Ai) whereas the LNCaP-iSnail (Figure 3.5 Aii) and LNCaP-iSlug (Figure 3.5 Aiii) cells adopted an elongated phenotype with less cell to cell contacts when compared to the LNCaP- iGFP cells.

Examination of mRNA levels confirmed strong expression of Snail (Figure 3.5 Bi) or Slug (Figure 3.5 Bii) following the addition of Dox to the respective models. This expression was paralleled by a decrease in the expression of epithelial markers E-cadherin and EpCAM and a concomitant increase in the expression of the mesenchymal marker vimentin (Figure 3.5 Bii and iii). In contrast, the addition of Dox to the LNCaP-iGFP cells induced robust expression of GFP mRNA with no alterations in the expression of E-cadherin, EpCAM, vimentin, Snail or Slug (Figure 3.5 Biii). It was then assessed whether the changes observed with the mRNA expression translated to the protein level. Indeed, the addition of Dox-induced strong levels of Snail (Figure 3.5 Ci) or Slug (Figure 3.5 Cii) protein in the respective models after 1 day and was reflected by a prompt decrease in the epithelial proteins E-cadherin and an increase in the mesenchymal protein vimentin. The addition of Dox to the LNCaP-iGFP cells resulted in GFP expression after 1 day and no alterations were observed in the protein expression of E-cadherin, EpCAM and vimentin over the 5 days (Figure 3.5 Ciii).

Chapter 3: Characterisation of reversible EMT models 56

Chapter 3: Characterisation of reversible EMT models 57 Figure 3.5. Assessing the induction of an EMT in the LNCaP-iSnail and LNCaP-iSlug models.

A. Phase contrast images showing the morphology of untreated (i) LNCaP-iGFP, (ii) LNCaP-iSnail, and (iii) LNCaP-iSlug cells and cells treated with Dox for 1, 3, and 5 days. Scale bars indicate 50 μm. B. Real-time PCR of indicated genes in (i) LNCaP-iSnail, (ii) LNCaP-iSlug, and (iii) LNCaP-iGFP cells treated with Dox for 1, 3, and 5 days. Data from each treatment day was normalised to RPL32 levels and expressed as the fold change in gene expression relative to untreated cells; n = 3 biological replicates. Dotted lines indicate a fold change of ± 1.5 Error bars indicate SEM. One-way ANOVA, p-value: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001 C. Protein expression of indicated proteins in untreated (i) LNCaP-iSnail, (ii) LNCaP-iSlug, and (iii) LNCaP-iGFP cells, and treated with Dox for 1,3, and 5 days. Arrows indicate the E-cadherin protein band that decreased following treatment with Dox. Gamma-tubulin was used as a loading control.

Chapter 3: Characterisation of reversible EMT models 58 The protein localisation of the EMT-TF’s Snail or Slug, and E-cadherin was also assessed via immunofluorescent staining in the respective EMT models (Figure 3.6A & B). In the absence of Dox, E-cadherin can be seen expressed predominantly in the cell membrane and the nucleus is devoid of either Snail or Slug protein (Figure 3.6A and B). With the addition of Dox for 4 days, Snail or Slug protein is observed in the nucleus of the LNCaP-iSnail and LNCaP-iSlug models, respectively, which was accompanied by a downregulation of E-cadherin protein (Figure 3.6A and B).

The LNCaP-iZeb1 model was generated late in the study and therefore its characterisation was limited. Similar to results observed with the LNCaP-iSnail and LNCaP-iSlug models, LNCaP-iZeb1 cells gained a mesenchymal-like phenotype following the addition of Dox for 1, 3, and 5 days (Figure 3.7 A). Examination of the mRNA levels via qRT-PCR confirmed robust expression of Zeb1, a downregulation in the expression of epithelial markers E-cadherin and EpCAM, and a downregulation in mesenchymal marker vimentin (Figure 3.7 B). The expression of Zeb1, E-cadherin and vimentin was further assessed at the protein level using western blot. In agreement to that observed at the mRNA level, addition of Dox to the LNCaP-iZeb1 cells induced expression of Zeb1 by as early as 1 day following Dox treatment and this expression remained over the 5 days of Dox treatment (Figure 3.7 C). In contrast, E-cadherin protein was downregulated after 1 day of Dox treatment and remained supressed over the 5 days of Dox treatment (Figure 3.7 C). Vimentin protein became upregulated at 5 days of Dox treatment (Figure 3.7 C).

Taken together, the addition of Dox to the culture media of the LNCaP-iSnail, LNCaP-iSlug, and LNCaP-iZeb1 models induced morphological and gene expression alterations associated with a classical EMT response. Moreover, these changes were not observed in the LNCaP-iGFP model, making it an appropriate model to control for the effects of Dox and as a negative control for EMT induction.

3.2.4 Assessing the invasive characteristics of the LNCaP-iSnail and LNCaP- iSlug models following Dox treatment. EMT is well known for its ability to bestow cells with invasive characteristics. To determine the invasive nature of the models following induction of the EMT- TF’s Snail or Slug, a modified 3D-on-top Matrigel™ assay was utilised

Chapter 3: Characterisation of reversible EMT models 59

Figure 3.6. Immunofluorescent imaging of EMT proteins in the LNCaP- iSnail and LNCaP-iSlug models.

A. LNCaP-iSnail and B. LNCaP-iSlug cells were treated with Dox for 3 days, and the indicated proteins were visualised via immunofluorescent staining and imaging using a Zeiss LSM 510 confocal microscope. Scale bars indicate 50 μm.

Chapter 3: Characterisation of reversible EMT models 60

A N o D o x D o x + 1 D a y D o x + 3 D a y s D o x + 5 D a y s

D a y s in D o x B 1 0 0 0 C * * N o D o x 0 + 1 + 3 + 5

x *

o D o x + 1 D a y s D

* Z e b 1

D o x + 3 D a y s o

N D o x + 5 D a y s

1 0 0

o

t

n E - c a d h e r in

o

i

s

s e r V im e n t in

p 1 0 *

x

e

A N

R  - t u b u li n

m

e 1 *

v i t * * * * * a * * * l * * * * *

e * * * R

0 .1

1 in M in b r t e e A n Z h C e d p a E im -c V E

Figure 3.7 Assessing the induction of an EMT in the LNCaP-iZeb1model.

A. Phase contrast images showing the morphology of untreated LNCaP-iZeb1 cells and cells treated with Dox for 1, 3, and 5 days. Scale bars indicate 50 μm. B. Real-time PCR of Zeb1, E-cadherin, EpCAM, and vimentin, in LNCaP-iZeb1 cells treated with Dox for 1, 3, and 5 days. Data from each treatment day was normalised to RPL32 levels and expressed as the fold change in gene expression relative to untreated cells; n = 3 biological replicates. Dotted lines indicate a fold change of ± 1.5 Error bars indicate SEM. One-way ANOVA, p-value: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001 C. Protein expression of Zeb1, E-cadherin, and vimentin, in untreated LNCaP-iZeb1 cells and cells treated with Dox for 1,3, and 5 days. Gamma- tubulin was used as a loading control.

Chapter 3: Characterisation of reversible EMT models 61

(Debnath, Muthuswamy and Brugge 2003). This assay has been previously validated as an appropriate assay for investigating the invasive properties of PCa cell lines (Harma et al. 2010). LNCaP-iSnail and LNCaP-iSlug cells were seeded as a single cell suspension into the assay and were allowed to form multicellular spheroids for 7-10 days before Dox treatment. It is important to note that the multicellular spheroids that developed over this period had a compact and non- invasive phenotype (Figure 3.8A, First column; Supplemental Video 1 accessible here: http://tinyurl.com/h7s35sr). Following spheroid formation, spheroids were treated with Dox for up to 4 days to induce the expression of the corresponding EMT-TF’s or GFP (Figure 3.8A). As early as 2 days following Dox-induced Snail and Slug expression, cells began to invade and dissociate from the spheroid, characterised by an elongated tear drop shape (Figure 3.8A; second column). By day 4 of treatment, multiple cells were observed to dissociate from spheroids and invade the surrounding Matrigel™ matrix as single cells (Figure3.8A; third and fourth columns; Supplemental Video 2 accessible here: http://tinyurl.com/jhshljq).

In contrast, the spheroids resulted in the expression of GFP protein which was visualised by epifluorescent microscopy, and no further changes were observed in the morphology of the spheroids. The levels of invasiveness were quantified by counting the spheroids in 10 random fields and grouping them into 4 categories depending on their invasive status: category 1: absence of invasive cells, category 2: 1-3 cell protrusions per spheroid, category 3: more than 3 protrusions per spheroid, category 4: multiple dissociated cells (MDC). Overall, 4 days of Dox treatment resulted in the LNCaP-iSnail and the LNCaP-iSlug model having 86% and 65% of total spheroids exhibiting invasive characteristics respectively compared to the LNCaP-iGFP model that only had 9% (Figure 3.8B). Non-treated LNCaP-iSnail and LNCaP-iSlug spheroids retained their non-invasive morphology that resembled that of the LNCaP-iGFP spheroids (Figure 3.8A, first column).

To confirm the induction of EMT following Dox treatment of LNCaP-iSnail and LNCaP-iSlug models, the expression of E-cadherin and vimentin protein levels were examined via immunofluorescent staining (Figure 3.9A and B). In the LNCaP-iSnail model following 4 days of Dox treatment, Snail protein was present in the nucleus,

Chapter 3: Characterisation of reversible EMT models 62 E-cadherin levels were decreased, and vimentin levels increased, particularly, in the cells invading the surrounding matrix (Figure 3.9A). In the LNCaP-iSlug model, the induction of Slug protein was evident in the nucleus and was accompanied by a decrease in E-cadherin protein (Figure 3.9B). In summary, Dox treatment of LNCaP- iSnail and LNCaP-iSlug models induced Snail and Slug protein expression, respectively, which induced EMT-like protein changes leading to an increased invasive phenotype in 3D Matrigel™ assays.

3.2.5 Transcriptional profiling of LNCaP-iSnail and LNCaP-iSlug models during 4 days of Dox treatment in 3D Matrigel™ cultures.

3.2.5.1 Assessment of EMT-related markers Next, the broader transcriptional program of EMT was profiled using the custom APCRC-Q 180K microarray. This microarray incorporates human protein-coding and non-coding probes, as well as probes targeting exonic regions, 3’UTRs, 5’UTRs, and probes targeting intronic and intergenic regions. Transcriptional profiling was performed on untreated LNCaP-iSnail and LNCaP-iSlug cells and cells treated with Dox for 48 and 96 hours. Due to the cost of arrays, only two biological replicates were assessed for each treatment. For control purposes, the LNCaP-iGFP model was also included to not only serve as a negative control for EMT but also to evaluate biological processes potentially affected by Dox itself. Initial clean-up of the microarray data consisted of removing probes from subsequent analyses that became differential ( ± 1.5 fold change (FC) p ≤ 0.05) in LNCaP-iGFP cells treated with Dox for 2 and 4 days. These 153 probes were collapsed to the gene level (104 genes) and examined for their enrichment of biological processes listed within the gene ontology consortium (GO-BPs). However, after assessment with gene set enrichment analysis (GSEA), it was found that no GO-BP’s were significantly enriched in the Dox treated LNCaP-iGFP cells, indicating that 4 days of Dox treatment had a minimal biological effect on LNCaP-iGFP cells (Table 3.1). Overall, the LNCaP- iGFP model served as an appropriate negative control for EMT induction experiments.

Chapter 3: Characterisation of reversible EMT models 63

Chapter 3: Characterisation of reversible EMT models 64 Figure 3.8. Induction of LNCaP-iSnail and LNCaP-iSlug tumour spheroid invasion following Dox treatment.

LNCaP-iSnail, LNCaP-iSlug and LNCaP-iGFP models were grown in modified 3D-on-top Matrigel™ assays for 10 days to form multicellular spheroids and then treated with Dox for 4 days. A. Representative phase contrast images are shown over the 4-day Dox treatment. Scale bars indicate 50 μm. B. Bar chart showing the percentage of invasive colonies in each category for the indicated times. Each category described the invasive status of the spheroids: 1 = absence of invasive cells, 2 = 1-3 protrusions, 3 = more than 3 protrusions, 4 = multiple dissociated cells (MDC).

Chapter 3: Characterisation of reversible EMT models 65 A L N C a P - iS n a il B L N C a P -iS lu g

D A P I

D A P I / F - A c t in D A P I / S n a il D A P I / E - c a d h e r in D A P I / V im e n t in E - c a d h e r in / S lu g

x

o

D

o

N

4

+

s

x

y

o

a

D D

Figure 3.9. Assessing EMT-like protein expression in the LNCaP-iSnail and LNCaP-iSlug models following 4 days of Dox treatment.

LNCaP-iSnail and LNCaP-iSlug cells were grown in 3D Matrigel™ assays to form multicellular spheroids prior to treating with Dox for 4 days. A. Immunofluorescent images showing expression of F-actin (red), and the proteins Snail (red), E-cadherin (red), and vimentin (red), in untreated LNCaP-iSnail cells and cells treated with Dox for 4 days. B. Immunofluorescent images showing he protein expression of E-cadherin (red) and Slug (green) in untreated LNCaP-iSlug cells and cells treated with Dox for 4 days. Images were taken using a Zeiss microscope. Nuclei were visualised with DAPI. Scale bars indicate 50 μm.

Chapter 3: Characterisation of reversible EMT models 66 Table 3.1. Enrichment of biological processes from the Gene-Ontology Consortium in LNCaP-iGFP cells treated with Dox over 4 days.

NAME SIZE NES NOM p-val FDR q-val FWER p-val CENTRAL_NERVOUS_SYSTEM_DEVELOPMENT 2 1.6709269 0.024742268 1 0.784 CATABOLIC_PROCESS 2 1.5609385 0.051282052 1 0.961 INTEGRAL_TO_PLASMA_MEMBRANE 6 1.5512767 0.05836576 0.7675162 0.966 CELLULAR_CATABOLIC_PROCESS 2 1.5422475 0.04191617 0.60959333 0.973 MULTICELLULAR_ORGANISMAL_DEVELOPMENT 7 1.5158646 0.05858586 0.5573185 0.985 CELL_SURFACE_RECEPTOR_LINKED_SIGNAL_TRANSDUCTION_GO_00071664 1.3800225 0.11619048 0.8949551 1 EMBRYONIC_DEVELOPMENT 2 1.3598033 0.10557769 0.8308641 1 INTRINSIC_TO_PLASMA_MEMBRANE 9 1.3448213 0.13184585 0.79698616 1 INTEGRAL_TO_MEMBRANE 8 1.3171527 0.13704497 1 1 PLASMA_MEMBRANE 12 1.2798213 0.18323587 1 1 CELL_CELL_SIGNALING 3 1.2652799 0.204 1 1 BEHAVIOR 1 1.2575681 0.12092131 1 1 ACID_AMINO_ACID_LIGASE_ACTIVITY 1 1.2558564 0.12790698 1 1 ORGAN_DEVELOPMENT 4 1.2555468 0.18503119 1 1 SYSTEM_DEVELOPMENT 6 1.2519095 0.2059925 1 1 PROTEIN_CATABOLIC_PROCESS 1 1.2514938 0.14285715 1 1 OXIDOREDUCTASE_ACTIVITY 5 1.248785 0.22332016 1 1 CELLULAR_MACROMOLECULE_CATABOLIC_PROCESS 1 1.2487247 0.11764706 1 1 SMALL_CONJUGATING_PROTEIN_LIGASE_ACTIVITY 1 1.2429338 0.13168724 1 1 LIGASE_ACTIVITY_FORMING_CARBON_NITROGEN_BONDS 1 1.2416396 0.13424124 0.98334324 1 REGULATION_OF_CELL_PROLIFERATION 3 1.2400961 0.20696722 0.94982046 1 FEEDING_BEHAVIOR 1 1.2399762 0.13426854 0.9082137 1 BIOPOLYMER_CATABOLIC_PROCESS 1 1.2394539 0.14653465 0.87108636 1 CELLULAR_PROTEIN_CATABOLIC_PROCESS 1 1.2394139 0.14677104 0.83522236 1 MACROMOLECULE_CATABOLIC_PROCESS 1 1.2392074 0.14107884 0.803 1 LIGASE_ACTIVITY 1 1.2331758 0.1425819 0.8054917 1 MULTI_ORGANISM_PROCESS 1 1.2308009 0.156 0.7871359 1 SMALL_PROTEIN_CONJUGATING_ENZYME_ACTIVITY 1 1.230386 0.13157895 0.7620448 1 SENSORY_PERCEPTION_OF_CHEMICAL_STIMULUS 1 1.2292093 0.12525253 0.742103 1 UBIQUITIN_PROTEIN_LIGASE_ACTIVITY 1 1.2226748 0.1406551 0.74708253 1 ANATOMICAL_STRUCTURE_DEVELOPMENT 6 1.2173947 0.2358871 0.75063604 1 NEGATIVE_REGULATION_OF_CELL_PROLIFERATION 3 1.2166817 0.22334003 0.73006785 1 REGULATION_OF_MULTICELLULAR_ORGANISMAL_PROCESS 1 1.2146341 0.18674698 0.71839935 1 BRAIN_DEVELOPMENT 1 1.2048922 0.18725869 0.73621476 1 INTRINSIC_TO_MEMBRANE 11 1.1999208 0.21471173 0.7379972 1 HYDROLASE_ACTIVITY_ACTING_ON_ACID_ANHYDRIDESCATALYZING_TRANSMEMBRANE_MOVEMENT_OF_SUBSTANCES1 1.1842035 0.23265307 0.78095007 1 MEMBRANE_FRACTION 3 1.1837668 0.25454545 0.76082295 1 ATPASE_ACTIVITY_COUPLED_TO_MOVEMENT_OF_SUBSTANCES 1 1.1702473 0.25443786 0.79627985 1 ADENYL_NUCLEOTIDE_BINDING 1 1.1698723 0.23921569 0.777829 1 ORGANIC_ANION_TRANSMEMBRANE_TRANSPORTER_ACTIVITY 1 1.1670488 0.24557957 0.7705075 1 ADENYL_RIBONUCLEOTIDE_BINDING 1 1.1665214 0.22524272 0.75280935 1 PRIMARY_ACTIVE_TRANSMEMBRANE_TRANSPORTER_ACTIVITY 1 1.1634161 0.22403258 0.74760133 1 ATP_BINDING 1 1.1626704 0.2485323 0.7330882 1 MICROSOME 1 1.1598978 0.26584867 0.7254962 1 VESICULAR_FRACTION 1 1.1598 0.24318658 0.71000963 1 ELECTRON_CARRIER_ACTIVITY 1 1.1533269 0.24793388 0.71661997 1 HORMONE_METABOLIC_PROCESS 1 1.1507103 0.26367188 0.7104436 1 TRANSMISSION_OF_NERVE_IMPULSE 2 1.14599 0.26271185 0.710356 1 SYNAPTIC_TRANSMISSION 2 1.1312381 0.2733051 0.73948884 1 1 NES: Normalised enrichment score; NOM: Nominal; FDR: False Discovery Rate; FWER: Family-wise error rate.

Chapter 3: Characterisation of reversible EMT models 67 The microarray array data was then filtered to identify differentially expressed probes (±1.5 FC; p-value ≤ 0.05) in the LNCaP-iSnail and LNCaP-iSlug cells treated to undergo EMT for 2 and 4 days compared to untreated cells. To assess EMT, the transcriptional activity of known epithelial and mesenchymal markers was examined following the addition of Dox and EMT onset (Figure 3.10A). This identified multiple epithelial genes to be downregulated (such as E-cadherin (CDH1), EpCAM, occludin (OCLN) and various claudins (CLDN3, 4, 7)) following Dox treatment (Figure 3.10A). Conversely, multiple mesenchymal genes including vimentin, Zeb1, Twist1, MMP2 and various collagens (COL5A2, 3A1) became upregulated (Figure 3.10A). Cumulatively this signified the induction of a comprehensive EMT program. This was further elucidated by pattern matching the expression of the respective EMT-TFs Snail or Slug, E-cadherin, and vimentin in each reversible EMT model, to identify clusters of transcripts that followed similar expression patterns (Figure 3.10 Bi and ii). Lastly, the expression of Snail or Slug, E-cadherin, EpCAM and vimentin were validated via qRT-PCR (Figure 3.10 Ci, ii & iii). The addition of Dox was observed to induce high levels of Snail or Slug mRNA expression after 2 days which remained elevated at 4 days in the respective models (Figure 3.10 Ci and ii).

Expression of either Snail or Slug mRNA was accompanied by a decrease in E- cadherin and EpCAM mRNA expression (Figure 3.10 Ci and ii). For both models, induction of Snail or Slug for 2 days induced upregulation of vimentin mRNA expression after 2 days (Figure 3.10 Ci and ii). The addition of Dox to the LNCaP- iGFP cells showed the robust upregulation of GFP gene expression at 2 and 4 days, and no change was observed in the mRNA expression levels of E-cadherin, EpCAM and vimentin (Figure 3.10 Cii). In summary, the transcriptional changes observed were consistent with an EMT. Next, the expression data from the Snail or Slug- induced EMTs was examined using Gene Set Enrichment Analysis (GSEA) to identify enrichment of public datasets from the Molecular Signatures Database (MSigDB, Broad Institute, MIT). This revealed enrichment of a number datasets related to EMT (Figure 3.11).

Chapter 3: Characterisation of reversible EMT models 68 A

B i

l i

a C D H 1

n S

i C lu s t e r a v g

-

P a S N A I1 C V I M

N C lu s t e r a v g C lu s t e r a v g L

B ii D 0 E 2 E 4 D 0 E 2 E 4 D 0 E 2 E 4

g u

l C D H 1 S

i C lu s t e r a v g

-

P a

C S N A I2 V I M N C lu s t e r a v g

L C lu s t e r a v g

D 0 E 2 E 4 D 0 E 2 E 4 D 0 E 2 E 4

C i C ii C iii

1 2 8 2 5 6 1 2 8 * * * * N o D o x N o D o x * N o D o x x 6 4 1 2 8 *

o D o x + 2 D a y s D o x + 2 D a y s 6 4 D o x + 2 D a y s

D

D o x + 4 D a y s D o x + 4 D a y s D o x + 4 D a y s

o 3 2 6 4 N

3 2 o

t 1 6 3 2

n o

i 1 6

s 8 1 6

s

e r

p 4 8 8

x

e

A 2 * 4

4 N

R *

1 2

m

e 2

v * * i

t 0 .5 1

a * * l

e * * 1

R 0 .2 5 0 .5

0 .1 2 5 0 .2 5 0 .5

l i in M in g in M in P in M in a r t u r t F r t e A n l e A n e A n n C S C G C r e la tiv e S h e h e h e d p d p d p a E im a E im a E im V V V -C -C -C E E E r o w m i n r o w m a x

Chapter 3: Characterisation of reversible EMT models 69 Figure 3.10. A Snail or Slug-induced invasive EMT leads to changes in hallmark EMT genes

The LNCaP-iSnail, LNCaP-iSlug and, LNCaP-iGFP models were treated with Dox for 2 and 4 days in a 3D-on-top Matrigel™ assay and then analysed via microarray. A. Heatmap showing the expression of known epithelial and mesenchymal markers. B. Transcriptional pattern matching of indicated markers (red) in i. untreated LNCaP-iSnail and ii. LNCaP-iSlug cells (D0) and cells treated with Dox for 2 (E2) and 4 days (E4). n= number of probes (grey). Blue line shows the average transcriptional pattern of the cluster (Cluster avg).C. Real-time PCR analysis of indicated genes in i. LNCaP-iSnail cells, ii. LNCaP-iSlug cells, and iii. LNCaP-iGFP cells treated with Dox for 2 and 4 days. Data from each treatment day was normalised to RPL32 levels and expressed as the fold change in gene expression relative to untreated cells. n = 2 biological replicates. Dotted lines represent fold change cut-off of 1.5 units. Error bars indicate SD. Students t-test * = p < 0.05.

Chapter 3: Characterisation of reversible EMT models 70 3.2.5.2 Identification of enriched public datasets and biological processes For example, both Snail and Slug-induced EMTs were found to be positively enriched with the EMT interactome signature developed by Taube et al. (Taube et al. 2010) which represents a core EMT interactome derived using multiple known EMT-inducers in immortalised human mammary epithelial (HMLE) cells (Figure 3.11, first column). Transcriptional alterations of a Snail or Slug-induced EMT were also enriched with genes upregulated in invasive ductal carcinoma relative to non- invasive ductal carcinoma (Schuetz et al. 2006) (Figure 3.11, second column), which is in line with studies supporting the activation of EMT and invasion (Sethi et al. 2010; Christiansen and Rajasekaran 2006; Klymkowsky and Savagner 2009) . Furthermore, the data was enriched with genes upregulated in CD31 negative stromal stem cells compared to CD31 positive non-stem cell counterparts (Boquest et al. 2005) (Figure 3.11, third column), supporting the association between EMT and a stem-cell like phenotype (Mani et al. 2008; Morel et al. 2008; Polyak and Weinberg 2009). As expected, GO-BPs analysis revealed a Snail or Slug-induced EMT signature to be positively enriched with multiple developmental processes such as “anatomical structure morphogenesis” and “system development” but also “cell differentiation” and “neurogenesis” (Figure 3.12A and Figure 3.13A). Conversely, processes related to cell cycle and metabolism such as “mitotic cell cycle” and “small-molecule metabolic process” were negatively enriched in a Snail-induced EMT (Figure 3.12B). Similarly, metabolic related processes such as “regulation of catalytic activity” and “positive regulation of metabolic process” were negatively enriched in a Slug-induced EMT (Figure 3.13B). Taken together, the LNCaP-iSnail and LNCaP-iSlug models transcriptionally recapitulate classic EMT-like biological processes following Dox treatment and expression of Snail and Slug respectively.

3.2.6 Assessing the reversibility of EMT in the LNCaP-iSnail and LNCaP-iSlug models. Next, the reversibility of EMT was assessed in the LNCaP-iSnail and LNCaP- iSlug models. Both models were cultured in 2D and treated with Dox for 5 days followed by its removal for 7 days. The addition of Dox to the models resulted in EMT-like alterations, which was assessed by measuring the expression of E- cadherin, EpCAM, vimentin, and Snail or Slug in their respective models.

Chapter 3: Characterisation of reversible EMT models 71

Figure 3.11. The Snail or Slug-induced EMT gene signature is enriched for genes reported in other EMT-related datasets.

GSEA enrichment plots of indicated public datasets at 4 days of Dox treatment versus untreated (no Dox) in LNCaP-iSnail (top) and LNCaP-iSlug cells (bottom) respectively. NES: Normalised enrichment score; FDR: False discovery rate.

Chapter 3: Characterisation of reversible EMT models 72 A B

U p re g u la te d G O -B P s D o w n re g u la te d G O -B P s

8 0 6 0

6 0

4 0

)

)

e

e

u

u

l

l

a

a

v

v -

- 4 0

p

p

(

(

g

g

o

o

l

l

- -

2 0

2 0

0 0

t

t

t

t

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

e

e

s

s

s

s

s

s

s

s

s

s

n

s

s

s

s

n

n

i

l

l

i

i

i

n

n

n

n

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

s

n

o

o

o

s

c

c

s

s

s

i

i

i

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

e

t

e

e

t

o

t

y

y

e

e

e

e

r

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

c

a

a

a

c

c

m

m

m

m

n

n

n

n

i

u

o

o

o

o

o

o

o

o

o

o

o

o

o

o

o

o

o

o

z

o

o

o

o

o

o

o

o

t

z

l

l

p

p

p

p

i

e

r

r

r

r

r

r

r

r

r

r

r

i

r

r

r

r

r

e

e

e

l

l

r

r

r

r

r

r

r

r

r

r

e

n

o

o

o

o

n

g

p

p

p

p

p

p

p

g

p

p

p

p

n

p

p

g

p

p

p

g

e

e

p

p

p

p

p

p

p

p

p

p

l

l

l

l

n

e

a

o

c

c

l

l

l

o

l

a

o

l

l

o

r

r

r

r

r

e

e

e

e

i

f

c

c

c

c

c

c

c

c

c

c

c

c

c

i

c

c

r

i

i

i

i

i

i

i

i

i

i

i

i

i

g

i

i

m

a

g

a

a

a

h

a

a

a

a

v

v

v

v

l

l

l

l

l

l

l

l

l

a

l

a

l

e

l

l

t

t

b

c

o

b

r

l

t

l

t

t

t

l

r

l

u

f

i

s

c

p

e

e

e

e

o

o

o

o

o

o

o

o

o

o

o

t

f

i

o

o

e

e

i

r

o

u

n

u

n

m

r

n

n

u

o

u

e

r

i

n

l

l

l

l

d

d

d

d

b

b

b

b

b

b

b

b

b

b

b

o

l

n

l

b

g

b

h

h

l

l

s

o

t

e

e

o

e

e

t

n

o

d

t

i

t

t

o

l

i

a

a

a

a

a

a

a

a

a

a

a

i

e

a

e

a

o

a

e

e

n

l

n

e

t

t

t

t

t

l

t

t

t

t

t

t

t

t

n

t

n

n

n

l

m

m

n

m

a

m

m

m

m

c

c

r

g

c

c

e

e

m

e

e

e

e

y

y

e

o

e

e

e

e

e

e

a

e

a

e

r

o

e

p

p

o

p

p

i

e

e

u

f

f

i

r

i

n

m

s

s

n

e

t

t

t

g

t

c

o

t

o

o

o

o

b

m

m

m

m

m

m

m

m

m

m

m

r

m

o

o

m

m

e

r

l

l

l

s

l

-

o

s

s

o

o

o

c

a

i

a

i

i

s

f

u

r

f

n

o

r

y

y

e

e

e

e

e

i

p

y

p

u

z

d

e

e

d

d

d

t

d

z

n

n

n

l

b

b

i

i

l

i

i

i

o

r

i

r

m

v

v

s

v

a

s

v

o

e

a

r

n

c

c

l

t

n

l

o

o

a

g

c

c

m

c

c

n

u

s

m

a

n

i

i

e

e

e

e

g

a

a

i

e

s

n

u

n

u

t

s

t

u

u

g

n

n

l

l

a

c

m

a

o

a

a

a

l

o

l

a

r

i

d

d

d

d

r

g

l

l

n

l

a

o

m

a

u

t

a

o

o

u

u

s

c

e

o

c

i

r

g

i

l

t

l

s

g

i

l

i

f

c

r

l

c

o

e

a

a

o

t

e

s

l

r

r

o

p

t

r

i

c

i

x

r

s

o

u

r

u

l

n

o

m

c

a

o

c

g

c

l

r

n

a

v

-

e

o

p

a

o

n

l

i

e

o

i

l

a

b

y

r

g

g

a

l

r

l

s

m

a

l

l

a

a

f

c

i

e

t

t

a

u

n

m

l

u

x

i

u

l

o

e

e

u

e

l

o

o

g

m

u

c

m

t

u

o

l

n

-

n

r

r

g

n

l

l

i

o

l

r

l

s

u

g

o

g

c

g

n

l

l

i

l

o

l

r

a

e

a

e

e

m

l

t

e

t

u

o

e

b

n

n

l

e

e

a

e

e

m

c

o

i

-

c

r

r

r

l

g

n

n

i

e

r

a

n

a

l

o

g

v

c

l

c

r

o

s

m

i

e

i

a

a

c

e

o

m

n

o

n

t

l

l

t

t

a

i

e

n

u

o

f

c

i

t

s

g

p

a

-

p

a

a

g

l

v

u

a

a

g

i

l

l

s

o

m

o

g

l

t

e

n

n

u

g

e

i

l

m

r

i

r

m

u

s

r

e

t

a

n

e

s

i

s

g

o

o

g

o

m

c

o

n

o

c

n

n

c

i

e

i

t

r

p

o

r

r

s

a

n

l

a

a

l

l

a

u

u

u

l

l

g

g

l

l

r

e

e

e

o

r

c c

Figure 3.12. Biological processes enriched in a Snail-induced EMT

Enrichment of biological processes from the gene ontology consortium in the A. upregulated, and B. downregulated genes in the LNCaP-iSnail treated with Dox for 4 days. The top 20 enriched biological processes are shown here.

Chapter 3: Characterisation of reversible EMT models 73 A B

U p re g u la te d G O -B P s D o w n re g u la te d G O -B P s

5 0 3 0

4 0

2 0 )

) 3 0

e

e

u

u

l

l

a

a

v

v

-

-

p

p

(

(

g

g

o

o

l

l - - 2 0

1 0

1 0

0 0

t

t

t

t

t

t

s

s

y

s

s

s

y

s

s

s

s

s

s

s

s

s

s

y

s

s

s

s

n

n

n

n

n

n

n

g

n

n

n

n

r

i

i

i

i

i

i

i

t

t

t

n

n

n

n

n

s

i

s

i

s

s

s

s

s

i

s

s

s

s

n

o

o

o

o

o

o

o

n

o

o

o

o

o

s

s

s

s

s

s

s

i

i

i

i

i

i

i

i

i

i

i

i

e

e

e

e

e

v

v

v

e

e

e

e

l

e

e

e

e

e

e

e

t

t

t

t

t

t

t

o

t

t

t

t

i

i

i

p

e

e

e

e

e

e

e

r

t

t

t

c

c

c

c

c

c

c

c

c

c

c

c

a

c

a

a

a

a

a

a

a

a

a

m

s

m

m

m

m

n

n

n

n

n

n

n

i

l

i

c

c

c

u

o

o

o

o

o

o

o

o

o

o

o

z

z

t

z

z

n

z

c

t

n

n

p

n

p

p

p

p

i

i

y

e

i

i

e

i

i

e

e

e

e

e

r

r

r

r

r

r

r

r

r

r

r

a

a

a

e

l

l

l

l

r

n

g

u

n

u

o

a

o

o

o

o

g

g

g

g

g

i

n

n

g

g

p

p

p

p

p

p

p

p

p

p

p

f

f

l

l

l

l

l

n

a

a

a

a

r

o

e

e

c

c

e

s

t

o

o

l

o

l

o

o

l

l

a

u

o

o

l

l

l

i

i

r

r

r

r

e

r

e

e

e

e

r

c

c

c

c

f

c

t

h

t

r

s

i

f

h

h

a

g

h

a

h

a

a

h

h

a

a

a

v

m

v

v

v

v

e

a

a

e

a

a

o

o

l

o

o

o

y

y

t

t

t

m

t

t

p

r

u

l

l

l

l

l

l

a

l

l

f

f

l

l

c

o

m

p

s

p

p

p

p

p

e

e

e

e

e

f

f

o

i

i

s

n

m

n

n

n

n

r

r

o

r

e

r

u

r

r

u

u

i

u

i

n

f

f

s

a

n

a

l

l

m

i

d

d

d

d

d

i

g

b

t

t

n

l

l

s

n

o

e

e

e

c

e

e

c

o

o

t

o

n

o

o

o

d

d

m

o

o

i

o

k

l

o

l

n

i

o

a

a

a

l

a

e

e

o

h

e

l

e

s

n

e

l

t

t

t

n

i

l

c

l

l

i

f

m

m

m

a

m

m

m

m

m

m

m

m

m

m

n

r

n

a

c

c

c

c

t

g

e

p

i

e

o

n

a

e

a

l

o

e

o

r

n

o

l

p

p

p

p

p

i

t

t

e

e

g

l

u

o

f

f

c

f

a

r

l

f

m

n

n

i

n

e

t

t

r

t

l

e

g

r

c

o

a

o

o

o

o

o

d

b

n

t

e

r

o

o

m

m

o

e

r

e

l

l

l

m

s

l

l

m

n

o

s

s

-

o

o

c

o

a

u

o

i

g

i

i

e

c

e

f

a

s

u

m

c

f

n

o

-

y

y

e

e

e

e

e

f

f

t

o

l

t

r

e

i

p

u

p

g

t

n

n

n

n

i

v

l

n

n

o

f

r

h

v

s

v

v

s

v

v

o

l

e

e

c

t

c

o

o

r

u

o

l

n

c

t

e

o

l

o

o

a

g

l

o

m

o

s

o

-

i

i

i

e

r

e

e

e

e

o

g

e

a

i

e

a

g

i

a

s

u

n

t

t

s

t

g

j

l

j

g

n

l

l

t

n

n

n

e

o

p

e

r

v

d

d

d

d

d

i

r

l

e

l

g

o

n

a

a

u

t

a

n

o

u

o

o

u

c

b

a

o

o

c

i

l

l

r

l

i

r

n

i

s

l

l

i

i

r

r

r

o

a

t

g

s

i

m

o

l

o

t

g

t

t

a

i

s

u

u

r

u

o

u

m

m

p

a

c

p

t

l

r

a

n

t

v

o

e

a

-

l

i

e

a

a

l

i

s

g

g

a

g

l

r

s

s

l

l

l

g

s

r

i

l

a

a

c

a

c

i

i

l

e

u

s

l

n

u

l

i

e

e

e

u

l

e

l

u

u

m

s

e

e

c

t

u

l

r

n

n

r

r

e

r

i

l

e

r

l

u

g

o

u

g

g

n

g

g

l

o

l

e

r

c

e

a

a

a

l

v

t

u

e

g

l

e

n

i

e

e

e

e

e

m

n

c

r

u

i

g

g

t

e

r

a

r

r

e

u

i

v

v

c

r

r

o

e

m

l

s

i

i

r

e

c

n

l

t

s

t

t

i

e

o

o

e

g

i

e

i

f

t

n

-

-

a

e

a

l

v

o

v

s

v

s

i

o

o

i

i

c

t

e

e

n

u

p

t

t

i

l

l

o

o

h

i

i

a

n

s

g

g

p

p

s

p

m

s

r

o

o

n

n

o

i

o

i

i

t

o

p

p

p

s

s

a

l

m

u

l

l

g

e

e

c r

Figure 3.13. Biological processes enriched in a Slug-induced EMT

Enrichment of biological processes from the gene ontology consortium in the A. upregulated, and B. downregulated genes in the LNCaP-iSlug treated with Dox for 4 days. The top 20 enriched biological processes are shown here.

Chapter 3: Characterisation of reversible EMT models 74 Similar to results observed in section 3.2.3, analysis of the mRNA expression of Snail or Slug, E-cadherin, EpCAM, and vimentin show that addition of Dox induced high levels of Snail (Figure 3.14A) or Slug (Figure 3.14B) in the LNCaP-iSnail and LNCaP-iSlug models, respectively, which was accompanied by a significant decrease in E-cadherin and EpCAM, and a significant increase in vimentin expression (Figure 3.14 A and B). Removal of Dox for 7 days saw the restoration of Snail or Slug mRNA back to pre-treatment levels and the concomitant return of E- cadherin and EpCAM to baseline levels (Figure 3.14 A and B). The expression of vimentin was restored to baseline levels in the LNCaP-iSlug model but not the LNCaP-iSnail model, whereby vimentin remained significantly upregulated 7 days post Dox removal (Figure 3.14 A and B).

Next, the expression of the EMT-TFs, E-cadherin, and vimentin was assessed via immunofluorescent staining. As seen previously (section 3.2.3), the addition of Dox induced robust Snail (Figure 3.14 C, second column) or Slug (Figure 3.14 D, second column) protein expression in the nucleus of the respective models and was paralleled by a decrease in E-cadherin (Figure 3.14 C and D, top row, second panel) and an increase in cytoskeletal vimentin protein expressions (Figure 3.14 C and D, bottom row, second panel). Removal of Dox was marked by the cessation of Snail or Slug protein in the nucleus of the respective models (Figure 3.14 C and D, third column) and the restoration of E-cadherin (Figure 3.14 C and D, third column, top panel) and vimentin (Figure 3.14 C and D, third column, bottom panel) to pre-EMT amounts. Furthermore, cells morphologically re-adopted their pre-EMT phenotype indicated by an increase in cell to cell contacts, which was highlighted by the restoration of membranous E-cadherin staining (Figure 3.14 C and D, third column, top panel). Protein expression was further assessed using Western blot where similar results were observed (Figure 3.15 A & B). Collectively, the reversible EMT models were able to recapitulate a reversible EMT by regulating the expression of EMT- TF’s Snail and Slug with the addition and removal of Dox from the culture media.

Chapter 3: Characterisation of reversible EMT models 75 E M T M E T A C D o x + 5 D a y s /

N o D o x D o x + 5 D a y s R e m o v e d 7 D a y s

l

i a

1 0 0 0 n x

N o D o x S

o

/

D

E M T (D o x + 3 ) I

n

P o

* * E M T (D o x + 5 ) i

r A

N * *

* * M E T (D o x + 5 /-7 ) e D

o 1 0 0

t

h

d

n

o

a

i

c

s -

s *

* * e

* E

r * * * * p 1 0 * *

x *

e

A

N *

l

R

i m 1 a

* e * * n

v * i * * * S t * *

* * I /

a *

l * *

* P e

* n

i

A R

0 .1 t

D n

e il in M in a r t e A n m n i h C e S p d V a E im -c V E

E M T M E T

B D D o x + 5 D a y s /

N o D o x D o x + 5 D a y s R e m o v e d 7 D a y s

g

u

l

S /

x * N o D o x

o * I * 1 0 0 n

D E M T (D o x + 3 )

P i

* *

* r A

o * E M T (D o x + 5 )

e

N

D

M E T (D o x + 5 /-7 ) h

o

d

t

a

n * *

* * c

o i * * - *

s 1 0 *

E

s

e r

p *

x

e

A

N * R 1 * g

*

u

m l *

e * S

v *

i I

* /

t

a

P

l

n

e

i

A

t

R D

0 .1 n e g n n i M i m

u r t i l e A n

S h C e V d p a E im -c V E

Chapter 3: Characterisation of reversible EMT models 76 Figure 3.14. Defining the reversibility of the LNCaP-iSnail and LNCaP- iSlug models.

LNCaP-iSnail and LNCaP-iSlug cells were treated with Dox for 5 days to induce EMT and then Dox was removed for 7 days to assess their reversibility and resultant MET. Real-time PCR of indicated genes in the A. LNCaP-iSnail and B. LNCaP- iSlug cells treated with Dox for 1 and 5 days (EMT) followed by subsequent removal for 7 days (MET). Data from each treatment day was normalised to RPL32 levels and expressed as the fold change in gene expression relative to parental LNCaP. n = 3 biological replicates. Dotted lines indicate a fold change of ± 1.5. Error bars indicate SEM. One-way ANOVA, p-value: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001. Indicated proteins were visualised via immunofluorescence in untreated C. LNCaP-iSnail and D. LNCaP-iSlug cells and cells treated with Dox for 5 days (EMT) followed by removal for 7 days (MET). Nuclei ere visualised with DAPI. Scale bars indicate 50 μm.

ChapterA 3: Characterisation of reversible EMT models B 77

Figure 3.15. Western blot of EMT-related proteins demonstrates the reversibility of the LNCaP-iSnail and LNCaP-iSlug models.

A. Western blot showing the expression of Snail, fibronectin (FN), E-cadherin, B-catenin, and EpCAM proteins in untreated LNCaP-iSnail cells, cells treated with Dox for 7 days (EMT) followed by removal for 7 days (MET). Gamma-tubulin was used as a loading control. B. Western blot showing the expression of Slug and E- cadherin in untreated LNCaP-iSlug cell, cells treated with dox for 3 and 5 days (EMT), followed by removal for 7 days (MET). GAPDH was used as a loading control.

Chapter 3: Characterisation of reversible EMT models 78 3.3 Discussion

Early studies in cancer research investigated EMT as a mean by which tumour cells acquire invasive and pro-survival features. Therefore, the majority of investigations to date have examined the constitutive gain or loss of EMT-inducers such as the TFs Snail and Slug in cancer cell lines. For instance, the constitutive overexpression of Snail in the colorectal cancer cell lines HT29 and HCT116 led to a decrease in the epithelial protein E-cadherin and an increase in the mesenchymal protein fibronectin leading to increased migration and invasion (Fan et al. 2012). Conversely, constitutive knockdown of Snail expression in the cisplatin resistant ovarian adenocarcinoma cell line A2780cis decreased their invasive and migratory capacity and increased their cisplatin sensitivity (Haslehurst et al. 2012). Cumulatively, induction of EMT in cancer cell lines correlates with many features that could ultimately lead to metastatic disease. However, metastases often reflect an epithelial phenotype, therefore, for EMT to be involved in metastatic colonisation a reversion back to the epithelial phenotype via a MET is hypothesised to be required. Unfortunately, by their nature, constitutive models cannot recapitulate an EMT with a sequential MET in the same population of cancer cells. Additionally, the option to study the early temporal events of EMT is missed as investigators can only analyse cells following the stable selection process. Therefore, models that allow for an inducible and reversible EMT are better suited to investigate the temporal events of epithelial-mesenchymal plasticity (EMP) in cancer cells.

One method to achieve a transient EMT is by manipulating external stimuli within the cell culture environment such as altering the available oxygen levels or by the addition and subsequent removal of EMT-inducing growth factors or cytokines from the media. Hypoxia was first discovered to induce EMT when hypoxic regions within hepatocellular carcinoma tissues were found to co-express hypoxia-inducing factor-1a (HIF-1a) and Snail (Mak et al. 2010), prompting further investigation. Indeed, hepatocellular carcinoma cells exposed to hypoxia activated Snail leading to the acquisition of EMT-related markers as well as increased invasion and migration (Zhang, Huang, et al. 2013). Subsequently, studies by Jo et al. (Jo et al. 2009) show that hypoxia followed by reoxygenation treatment of MDA-MB 468 breast cancer cells simulated a transient EMT. While oxygen levels can be manipulated to regulate

Chapter 3: Characterisation of reversible EMT models 79 EMT, hypoxia also regulates additional signalling pathways outside of EMT and has other systemic effects which make discerning the direct influence of EMT difficult.

The use of growth factors, including members of the transforming growth factor-β (TGF-β) family, fibroblast growth factors (FGF and FGF2), epidermal growth factor (EGF), insulin-like growth factors (IGF-I and IGF-II), platelet-derived growth factor (PDGF) and hepatocellular growth factor (HGF) can also be used to achieve a transient EMT. Collectively, they have been shown to activate multiple signalling pathways and TFs that can lead to an EMT (Thiery 2002). For example, EGF promotes the endocytosis of E-cadherin as well as the expression of Snail and Twist that subsequently repress the expression of E-cadherin as well as activate and repress numerous other target genes (Lee, Chou, et al. 2008; Lo et al. 2007; Lu et al. 2003; Xu et al. 2015). Furthermore, IGF-I binding to the type I receptor (IGF-IR) stimulates expression of Zeb1 in PCa cells (Graham et al. 2008), in addition to activating both Snail and the NFκB pathway in mammary epithelial cells (Kim et al. 2007). One of the more extensively studied growth factors is TGF-β1 due to its role in a number of cellular contexts such as in heart development, pulmonary fibrosis, and tumour progression (Mercado-Pimentel and Runyan 2007; Kim et al. 2006; Pickup, Novitskiy and Moses 2013). TGF-β1 binds to its type II receptor (TGF- βRII) causing transphosphorylation of the type I receptor (TGF-βRI) (Massague and Gomis 2006; Neuzillet et al. 2015). This leads to the phosphorylation of Smad2 and Smad3 whereby their activation forms complexes with Smad4. Accumulation of these complexes in the nucleus collaborates with other TFs, such as FoxO, C/EBPβ, and c-Myc to regulate gene expression (Massague and Gomis 2006; Ross and Hill 2008; Gomis et al. 2006; Alexandrow and Moses 1995; Neuzillet et al. 2015). TGF- β1 signalling also activates multiple EMT-TFs including Snail, Slug, Twist, Zeb1, Zeb2 (Thuault et al. 2008; Lamouille, Xu and Derynck 2014). In general, TGF-β is considered to be a prototypical growth factor for induction and maintenance of an EMT as it has been observed to activate EMT-related pathways in multiple cellular contexts, whereas other growth factors can be variable and context dependent (Lee and Nelson 2012; Xu, Lamouille and Derynck 2009; Katsuno, Lamouille and Derynck 2013). For instance, HGF induces EMT during the early development of the cardiac cushion but may inhibit TGF-β induced EMT and fibrosis in the kidney (Mizuno et al. 1998; Lamouille, Xu and Derynck 2014). TGF-β treatment has been

Chapter 3: Characterisation of reversible EMT models 80 previously used to recapitulate a transient EMT in pancreatic cancer cells containing a functional TGF-β pathway (Ellenrieder et al. 2001; Pang et al. 2016). In the study herein, the use of TGF-β was not possible as the LNCaP cell line has a mutated and inactive TGF-β receptor (Guo and Kyprianou 1998). Similar to hypoxia, growth factors regulate multiple pathways independent of EMT and have systemic effects on various tissues, making them impractical for investigating the direct role of EMP on cancer metastasis. Environmental cues such as growth factors and hypoxia are inducers of pathophysiological EMT in cancer. However, due to the array of pathways they regulate outside of EMT, their use makes direct investigation of EMT challenging.

As growth factors and other environmental factors (i.e. hypoxia) activate the expression of EMT-TFs that in turn induce an EMT, studies have used EMT-TFs to isolate the process of EMP better. This has been achieved by the generation of inducible and reversible EMT models. These models are genetically engineered cell lines with the spatiotemporal regulation of EMT triggering genes by the addition or removal of activator molecules as seen with the tamoxifen/modified estrogen receptor (ER) and the doxycycline hyclate (Dox) inducible systems. The tamoxifen/modified ER model was developed by Mani et al (Mani et al. 2008) and was used to show that inducing a Snail- or Twist-induced EMT in the ER-negative human mammary epithelial (HMLE) cells invoked stem cell properties that persisted even after removal of the EMT stimulus. The tamoxifen/modified ER model consists of a mutated ER (G525R) fused to the C-terminal end of EMT-TFs, such as Snail or Twist, whereby treatment with the ER ligand 4-hydroxytamoxifen (4-OHT) translocates the EMT-TF into the nucleus to evoke their transcriptional activity. However, this model is only applicable for use in ER-negative cell lines and therefore not appropriate for the study herein as the LNCaP cell line expresses a functional ER. Furthermore, tamoxifen is a mainstream anti-cancer drug due to its cytostatic and cytotoxic effects in numerous cell types (Taylor et al. 1983; Cuevas and Lindeman 2015). Therefore, models using drugs with fewer side-effects in both in vitro and in vivo situations are preferred.

The most suitable inducible and reversible models currently available are the Dox- inducible models. These models come in two versions, Dox-ON and Dox-OFF

Chapter 3: Characterisation of reversible EMT models 81 where the expression of the transgene is respectively induced or suppressed in the presence of Dox. This study utilised a Dox-ON inducible system (pInducer20) (Meerbrey et al. 2011) whereby induction of EMT-TF expression was controlled by the addition of Dox to the culture media. This system was more appropriate for the context of the current study which involved the transient and reversible expression of the EMT-TF. One concern with using a Dox-ON system is the leaky expression of the transgene in the absence of Dox. Fortunately, in this system, the leaky expression of the EMT-TFs in the absence of Dox was minimal and insufficient to induce any EMT-related changes (Figure 3.2). Furthermore, Dox is a generic antibiotic, and while higher concentrations than those used in this study affect the intestinal flora of mice (Riond and Riviere 1988), the systemic side-effects compared to tamoxifen are minimal.

To better study EMP in prostate cancer (PCa), the study herein generated reversible EMT models using the PCa cell line LNCaP. From the broad array of possible models that could have been made for inducing EMT, the inducible expression of embryonic TFs was chosen. This was for three reasons; firstly, their well- documented role in regulating EMP via the EMT and MET programs during embryogenesis, and secondly, the direct mechanism by which they mediate an EMT. Snail and Slug directly bind the promoter region of E-cadherin and other epithelial proteins, including claudins and occludins, to suppress their expression (Peinado et al. 2004; Cano et al. 2000; Batlle et al. 2000; Grooteclaes and Frisch 2000; Eger et al. 2005; Sanchez-Tillo et al. 2012). While this can be cell type specific, the suppression of epithelial genes often initiates a cascade of transcriptional alterations that indirectly upregulate mesenchymal markers to facilitate the phenotypic switch, amplifying this transition (Yokoyama et al. 2003; Jorda et al. 2005; Guaita et al. 2002). Furthermore, Snail, Slug, and Zeb1 are considered master EMT-regulators as they are tightly regulated by multiple signalling pathways capable of stimulating EMT in both developmental and pathological systems (Zheng and Kang 2014). Lastly, the expression of EMT-TFs has been documented in various carcinomas including PCa (Moody et al. 2005; Tuhkanen et al. 2009; Franci et al. 2009; Haider et al. 2015; Jang et al. 2015; Goscinski et al. 2015; Pomp et al. 2015; Heeboll et al. 2009). Snail expression has been found to correlate with high Gleason score PCa when compared to low grade or benign hyperplasia samples (Poblete et al. 2014;

Chapter 3: Characterisation of reversible EMT models 82 Heeboll et al. 2009). Additional studies by Haider et al. (Haider et al. 2015) show that a subset of cancer cells in CRPC bone metastasis expressed nuclear Slug, Zeb1 and Twist with an EMT-like phenotype. Interestingly, they also expressed epithelial cytokeratin hinting at a partial EMT, partial MET, or the emergence of a metastable population (Jordan, Johnson and Abell 2011; Huang et al. 2013). Other studies by Sun et al. linked the commonly administered androgen deprivation therapy (ADT) to inducing a Zeb1 regulated EMT in both normal and cancer cells, thus supporting that the therapy itself may have deleterious effects on patient outcome (Sun et al. 2012). In conclusion, EMT-TFs are expressed in PCa and correlate with disease progression supporting the involvement of EMP in disease progression.

From the limited cell lines available within the PCa research field, the LNCaP cell line is accepted to be a representative model of AR positive and hormone-dependent PCa. The LNCaP cell line also expresses multiple epithelial markers and forms epithelial-like tumour spheroids when grown in 3D Matrigel™ culture (Sieh et al. 2012; Harma et al. 2010). Importantly, the LNCaP cell line is tumorigenic in vivo and can sustain growth when transplanted orthotopically or subcutaneously (Horoszewicz et al. 1983; Horoszewicz et al. 1980). Additionally, the LNCaP cell line is weakly metastatic when implanted orthotopically (Stephenson et al. 1992; Sato et al. 1997) and does not generally form clinically relevant metastases in distant organs. Hence, the LNCaP cell line allows for studies examining the influence of EMP on promoting metastasis in vivo (discussed in Chapter 5).

The aim of this chapter was two-fold; firstly to confirm that a robust and classical EMT could be obtained following the Dox-induced expression of the respective EMT-TFs; and secondly, that an EMT reversal was achieved following cessation of the EMT-TF expression following the withdrawal of Dox. Indeed, addition of Dox and the expression of Snail, Slug, or Zeb1 in the respective models resulted in a representative EMT highlighted by a distinct change in cell morphology, characteristic EMT gene / protein alterations (Figure 3.5, Supplemental Figure 3.2), and the acquisition of invasive abilities (Figure 3.8, Supplemental Figure 3.2). Expression of Snail, Slug, or Zeb1 following the addition of Dox was rapid as their expression was detected at both the mRNA and protein level after 24 hours (Figure 3.5, Supplemental Figure 3.1). Across all models, the EMT induction was prompt

Chapter 3: Characterisation of reversible EMT models 83 with hallmark markers E-cadherin decreasing and vimentin increasing after 24 hours of Dox treatment (Figure 3.5, Supplemental Figure 3.1).

To assess the expression of a broader spectrum of known epithelial and mesenchymal markers, the LNCaP-iSnail and LNCaP-iSlug models were induced to undergo an EMT in 3D Matrigel™ assays and were then transcriptionally profiled using a custom-made Agilent 180K probe microarray developed by the APCRCQ laboratory. This transcriptional profiling strategy allowed for systems biology approaches to be applied, including pathway analysis to identify enriched biological processes, and integration with public gene expression datasets. This was characterised by the downregulation of multiple known epithelial markers including E-cadherin and the upregulation of multiple known mesenchymal markers including vimentin (Thiery et al. 2009). Additionally, the transcriptional profiles of a Snail- or Slug-induced EMT were positively enriched with the “EMT interactome” generated by Taube et al (Taube et al. 2010) where a common core EMT signature was derived using various EMT inducers in the human mammary epithelial (HMLE) cell line, that included Snail (but not Slug). The EMT profiles were also enriched with datasets related to cancer invasion and stem cells, consistent with studies reporting EMT to promote these phenotypes in cancer cells (Mani et al. 2008; Morel et al. 2008; Polyak and Weinberg 2009; Fan et al. 2012; Emadi Baygi et al. 2010; Thiery et al. 2009; Heerboth et al. 2015). Further supporting the achievement of a robust EMT was the positive enrichment of biological processes related to development and cell differentiation (Lim and Thiery 2012; Thiery 2003; Nieto et al. 2016). This is in line with the paramount role of EMT in embryogenesis. Overall, the EMT achieved by the reversible Snail or Slug EMT models was consistent with the current literature on EMT.

The novel aspect of the models generated herein was their reversible nature. Removal of Dox from the culture media swiftly terminated expression of the EMT- TFs and allowed for MET to occur, which was portrayed by the loss of vimentin and re-acquisition of E-cadherin in the LNCaP-iSnail and LNCaP-iSlug models (Figure 3.10). Furthermore, MET restored the cell morphology to a pre-EMT state which was indicated by an increase in cell to cell contacts (Figure 3.10). Unfortunately, time did not permit for the full characterisation of the LNCaP-iZeb1 model in this

Chapter 3: Characterisation of reversible EMT models 84 study. However, this model is currently being characterised by other members of the Hollier laboratory. In summary, the LNCaP-iSnail and LNCaP-iSlug models recapitulate a classic EMT with the addition of Dox and allow for the investigation of the reverse process of MET by removal of Dox.

The models generated within offer high customisability when designing experiments to investigate the role of EMP in cancer progression and metastasis. The tight regulation of the EMT-TFs by Dox allows for the spatiotemporal and in-depth investigation of the early and late events of EMT and MET; both processes are studied and discussed in Chapters 4 and 5 respectively. Furthermore, these models could be used in conjunction with the Matrigel™ assay as a pre-clinical platform for testing therapeutics targeting EMP and invasion.

Chapter 3: Characterisation of reversible EMT models 85

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 86 4.1 Introduction

Metastatic tumour burden is the primary cause of cancer-related mortality in carcinomas arising in the epithelia of organs such as the breast, lung, colon and prostate (Li et al. 2015; Shiota et al. 2013; Yao et al. 2014). Epithelial-mesenchymal transition (EMT) can endow carcinoma cells with the phenotypic mesenchymal traits thought to facilitate seeding of distant metastases (Thiery et al. 2009; De Craene and Berx 2013; Hanahan and Weinberg 2011; Lamouille, Xu and Derynck 2014). The conceptual framework of carcinoma cells undergoing an EMT requires that these cells revert to an epithelial phenotype as observed through characterisation of distant metastases (Imai et al. 2004; Rubin et al. 2001; Tarin, Thompson and Newgreen 2005). Indeed, recent studies in preclinical models show that the induction of EMT and subsequent reversion of tumour cells to their epithelial phenotype (termed herein as the mesenchymal-epithelial reverting transition; MErT) is required for the formation of overt metastases (Celia-Terrassa et al. 2012; Ocana et al. 2012; Roca et al. 2013; Tsai et al. 2012). Despite this recognition, in comparison to studies focused on characterising the EMT program, the investigation of MErT has largely been ignored. In part, this stems from the lack of robust preclinical models able to tightly regulate EMT and MErT in human cancer cell populations, but also from the preconception that MErT is simply the antithesis of EMT. This assumption has remained due to the shortage of comprehensive studies characterising the dynamic transcriptional events that constitute a reversible EMT in a cancer context. As a result, there is limited direct evidence in clinical samples for the relevance of epithelial-mesenchymal plasticity in metastasis and patient prognosis.

As there are no markers that can identify the transitions, it is inherently difficult to show their occurrence in patient samples. Currently, assessment of patient samples relies on the identification of the phenotypic state of the cancer cells. Examination of the distribution of cancer cells within primary tumours, as well as the expression of EMT-TFs supports the occurrences of EMT at the primary site. However, there is relatively little evidence of a MErT occurring at either primary or metastatic sites. To address this, the study herein established prostate cancer models driven by the inducible and reversible expression of the master EMT-TFs Snail, Slug or Zeb1 and their characterisation has been described in the previous chapter. In this chapter, the

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 87 LNCaP-iSnail model was selected for further profiling, with a particular focus on transcriptionally characterising the complete spectrum of epithelial-mesenchymal plasticity (EMP) and its association with cancer progression in clinical samples of primary and metastatic origin.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 88 4.2 Results

4.2.1 MErT reawakens EMT-induced dormant-like tumour cells.

LNCaP-iSnail cells were treated with Dox for 5 days to induce EMT, followed by removal for 20 days to revert to an epithelial phenotype. To gain insights into the biological processes altered during the oscillation between epithelial and mesenchymal states, LNCaP-iSnail cells were transcriptionally profiled using a custom made 180K probe Agilent microarray and Gene Set Enrichment Analysis was used with the Molecular Signatures Database “Hallmark” gene collection (MSigDB; Broad Institute, MIT) (Liberzon et al. 2015) (Figure 4.1A). As anticipated, the transcriptional alterations evoked as cells transitioned from their parental epithelial state to an induced mesenchymal state (Figure 4.1A; x-axis, EMT5 vs. No Dox) were positively enriched in hallmark signatures of EMT, hypoxia and myogenesis (Figure 4.1A, bottom right quadrant). This was mirrored by a negative enrichment for signatures related to cell cycle and metabolism, including “E2F Targets”, “G2/M Checkpoint” and “Oxidative Phosphorylation” (Figure 4.1A, top left quadrant). The reversion of EMT-induced cells (EMT5) back to an epithelial phenotype via a 20 day MErT (y-axis, MErT20 vs. EMT5) reprogrammed cells to have a positive enrichment for these cell cycle and metabolism signatures (Figure 4.1A, top left quadrant). A larger collection of Gene Ontology (GO) gene sets (1454 gene sets; MSigDB) was used to validate these findings and further revealed the temporal dynamics operating during these transitions (Figure 4.2). For example, processes related to development such as “organ development” became positively enriched with EMT and then temporally returned to baseline by 20 days of MErT. On the other hand, processes related to cell cycle such as “mitosis” reflected the opposite trend.

Recent experimental evidence suggests that disseminated EMT-induced tumour cells need to regain their epithelial characteristics via a hypothesised MErT to reacquire the proliferative abilities necessary to grow as metastatic lesions (Ocana et al. 2012; Tsai et al. 2012). In support of this hypothesis, EMT saw the transcriptional decrease of critical mediators of cell cycle progression, including E2F Targets (i.e., CDKN3, MKI67 and PCNA) and G2/M checkpoint regulators (i.e., CCNB1, FOXM1

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 89

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 90 Figure 4.1. MErT breaks EMT-induced cell dormancy

(A to C) LNCaP-iSnail cells were treated with Dox for 5 days to induce an EMT (EMT5), followed by removal for 3, 5 and 20 days to allow for an MErT (MErT3- 20). A. Enrichment of indicated hallmark datasets from the MSigDB (Liberzon et al. 2015) at the EMT5 timepoint as compared to untreated cells (NoDox; x-axis) and at the MErT20 timepoint as compared to the EMT5 timepoint (y-axis). B. Heatmap showing the expression of selected cell cycle related genes. Colour gradient refers to z-score of the gene expression within each row. C. Quantitative real-time PCR showing the gene expression of CDKN1, CCNB1, FOXM1, and MKI67 relative to untreated cells (No Dox). Gene expression was normalised to RPL32. Error bars indicate SEM, n = 3 biological replicates, one-way ANOVA p-value: *< 0.05, **<0.01, *** <0.001, **** < 0.0001, ns: not significant p > 0.05 D. Cell cycle analysis showing percentage of cells in G0/G1, S, and G2/M phase for untreated LNCaP-iSnail and LNCaP-iGFP cells (No Dox), cells treated with Dox for 5 days (EMT5), followed by removal for 5 and 10 days (MErT5-10). E. Expression of Ki- 67 antigen surface expression in untreated LNCaP-iSnail cells (black/NoDox), cells treated with Dox for 5 days (red/EMT5), followed by removal for 5 days (blue/MErT5). Isotype control was IgG (grey). (F) Enrichment score of previously published signatures of Fi. cell cycle progression (CCP) (Cuzick et al. 2011) and Fii. tumour dormancy (Kim et al. 2012) in untreated LNCaP-iSnail cells, cells treated with Dox for 5 days (EMT5) followed by removal for 3, 5, and 20 days (MErT3-20). Error bars indicate SEM. One-way ANOVA p-value: *< 0.05, **<0.01, *** <0.001, **** < 0.0001, ns: not significant p > 0.05. Gi. Cell confluency of LNCaP- iSnail and control LNCaP-iGFP cells treated with or without Dox over 7 days relative to Day 0 was measured using the Incucyte. Asterisks indicate the statistical significance between Dox treated and untreated LNCaP-iGFP (black) or LNCaP-iSnail (red) cells. Two-way ANOVA; p-value: *<0.05, ns: not significant (p>0.05) Gii. Cell number of LNCaP-iSnail and control LNCaP-iGFP cells treated with or without Dox for 5 days. H. Representative images of LNCaP-iSnail cells treated with Dox for 5 days in a monolayer (top left, magnification 10x) before seeding as single cells in 3D Matrigel™ assays (top right, magnification 40x). LNCaP-iSnail cells were then either maintained in Dox (bottom right, magnification 40x) or had Dox removed (bottom left, magnification 40x) for an additional 7 days.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 91

Figure 4.2. Relative enrichment score of enriched biological processes from the gene ontology consortium (GO-BPs) (Ashburner et al. 2000; Gene Ontology 2015) at the indicated treatment times.

LNCaP-iSnail cells were treated with Dox for 5 days to induce an EMT (EMT5), followed by removal for 3, 5, and 20 days (MErT3-20). No Dox: untreated LNCaP- iSnail cells. NES: Normalised enrichment score.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 92 and CDC20), and this was reversed during MErT as tumour cells reacquired their epithelial phenotype (Figure 4.1 B and C).

Downstream functional assessment of these alterations confirmed EMT-induced cells to be in G0/G1-phase arrest (Figure 4.1D) with a substantially reduced population of proliferating antigen-Ki67 expressing cells (Figure 4.1E). Release from the mesenchymal state with a MErT restored both the cell cycle and Ki67 expressing profile of cells to a pattern similar to the parental epithelial state (Figure 4.1 D and E). Comparative examination of previously published tumour-derived signatures of cell cycle progression (Cuzick et al. 2011) and dormancy (Kim et al. 2012), revealed MErT to reprogram EMT-induced cells out of a dormant-like state and to restore the transcriptional program promoting cell cycle progression (Figure 4.1 Fi and ii). Reflective of this, LNCaP-iSnail cells induced into an EMT were observed to have reduced proliferative capacity (Figure 4.1 Gi and ii), which was restored with MErT (Figure 4.3). Confirming this phenotypic plasticity, LNCaP- iSnail cells held in an EMT state remained as individual invasive cells following seeding into 3D Matrigel™ cultures (Figure 4.1 H, bottom right panel). In contrast, cells allowed to undergo MErT had the capacity to re-initiate proliferation and form multi-cellular tumour spheroids (Figure 4.1 H, bottom left panel). Collectively, these results provide direct evidence of the re-awakening of EMT-induced dormant tumour cells by MErT to establish proliferation.

4.2.2 Temporal dynamics of gene expression associated with MErT.

The temporal transcriptional dynamics of a reversible Snail-mediated EMT was profiled. Microarray-based transcriptional profiling combined with Principal Component Analysis (PCA) identified a time-dependent connectivity pattern as cells reverted from a mesenchymal to an epithelial state (Figure 4.4A). Assessment of the temporal expression patterns of established mesenchymal and epithelial markers during MErT revealed that the majority of transcripts returned to pre-EMT baseline levels after 3 days of MErT (Figure 4.4B; MErT3). Other transcripts such as vimentin (VIM), fibronectin (FN1) and Quaking (QKI) were observed to have slower reverting kinetics (up to 20 days; Figure 4.4 B and C).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 93

Figure 4.3. MET rescues EMT-induced decrease in cell proliferation.

LNCaP-iSnail and LNCaP-iGFP cells were treated with Dox for 10 days followed by removal for 10 days. A. Schematic of experimental procedure. Bi Cell count of LNCaP-iSnail cells and Bii. LNCaP-iGFP cells. Error bars indicate SEM, n = 3 biological replicates, two-way ANOVA, p-value: **<0.01, **** < 0.0001, ns: not significant (p < 0.05).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 94

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 95 Figure 4.4. The MErT is a kinetically dynamic transcriptional process that

Microarray analysis of LNCaP-iSnail cells treated with Dox for 5 days to induce an EMT (EMT5) followed by removal over 20 days to allow for an MErT (MErT3- 20). A. Principal Component Analysis (PCA) as cells transition from an epithelial (No Dox) to a mesenchymal (EMT5) and back to an epithelial (MErT3-20) state. B. Heatmap of the expression of known epithelial and mesenchymal markers. Heatmap colour gradient refers to z-score of the gene expression within each row. C. Transcriptional clusters generated through pattern matching of select genes (blue) and classified by temporal reversion pattern. Clusters I and II reverting expression by day 3 of MErT, Clusters III and IV by day 5 and Clusters V and VI by day 20.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 96 In total, six discrete transcriptional patterns were identified to operate during MErT (Figure 4.4C). Clusters I and II contained transcripts reverting to baseline levels early (3 days of MErT; cluster I, n = 1208 transcripts; cluster II, n = 572 transcripts; Figure 4.4Ci); clusters III and IV contained transcripts reverting with intermediate kinetics to baseline (5 days of MErT; cluster III, n = 712 transcripts; cluster IV, n = 33 transcripts; Figure 4.4Cii); and clusters V and VI contained transcripts with a slower reversion to baseline levels (20 days of MErT; cluster V, n = 269 transcripts; cluster VI, n = 36 transcripts; Figure 4.4Ciii). The expression of characteristic epithelial markers, such as E-cadherin (CDH1) and EpCAM, were reactivated early during MErT (cluster II), and this was mirrored by the prompt reduction in mesenchymal markers, such as the established EMT-regulating transcription factor, ZEB1 (cluster I). The hallmark mesenchymal marker vimentin, as well as CTGF and SOX9, were among the slower reverting transcripts (cluster V; Figure 4.4Ciii). Information on the transcripts included in each cluster can be found at the following link: http://tinyurl.com/hdmm9ez. A number of genes were validated via qRT-PCR and were in agreement with the transcriptional pattern observed from the microarray (Figure 4.5). Gene ontology analyses of the biological processes enriched across the clusters revealed that majority of processes were enriched early in the MErT (3 days) compared to the later time points (5 and 20 days MErT) (Figure 4.6). Taken together, these findings provide the first temporal snapshot of the dynamic and reversible transcriptional alterations occurring with the cycling of tumour cells between EMT and MErT-like states.

Having obtained this transcriptional profile, the expression the MErT profile was examined in clinical samples of localised prostate cancer and lethal metastatic castrate-resistant prostate cancer (mCRPC) obtained from the Grasso (Figure 4.7 Ai), and Taylor (Figure 4.7 Aii) datasets (Grasso et al. 2012; Taylor et al. 2010). To generate the MErT profile gene list, all genes that were differentially expressed at 20 days of MErT compared to 5 days of EMT (~4600 genes) were ranked by signal-to- noise ratio using GSEA (descending order). The MErT profile was comprised of the top and bottom 500 genes of the ranked list. Interogation of the MErT profile with the two datasets revealed the metastatic samples to be significantly positively enriched with the MErT profile. To ensure that this trend was not primarily influenced by the proliferative status of the samples, genes associated with cell cycle

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 97 2 .0 1 .5 1 .5 2 .0 1 .5 C D H 1 E S R P 1 E S R P 2 E P C A M T O P 2 A **** 1 .5 1 .5 ** 1 .0 1 .0 1 .0

1 .0 1 .0

0 .5 *** 0 .5 0 .5 *** 0 .5 0 .5 * *** e * **** **** ****

g 0 .0 0 .0 0 .0 0 .0 0 .0

n

a h 1 0 0 4 0 6 0 0 1 5 1 5 C E M IL IN 2 **** S N A I1 Z E B 1 S O X 9 **** V IM **** 8 0 **** **** d ****

l 3 0

4 0 0 1 0 1 0 o 6 0 *** F **** 2 0

A 4 0 * 2 0 0 5 5 N 1 0

2 0

R m

0 0 0 0 0

e

v

i t

4 5 4 2 0 0 5 a l ** C O L 4 A 5 **** N R P 1 IH H E F N B 2 C T G F e * ** 4 * ** 4

R 3 3 1 5 0 ** ** 3 3 * 2 2 1 0 0 2 2

1 1 5 0 1 1

0 0 0 0 0

3

5

3

5

3

5

3

5

3

5

0

0

0

0

0

5

5

5

5

5

x

x

x

x

x

2

2

2

2

2

T

T

T

T

T

T

T

T

T

T

T

T

T

T

T

o

o

o

o

o

r

r

r

r

r

r

r

r

r

r

T

T

T

T

T

M

M

M

M

M

D

D

D

D

D

r

r

r

r

r

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

E

o

o

o

o

o

M

M

M

M

M

M

M

M

M

M

M

M

M

M

M

N

N

N

N N

Figure 4.5. mRNA expression of indicated epithelial and mesenchymal markers.

Quantitative real-time PCR showing the expression of 5 upregulated (CDH1, ESRP1, ESRP2, EPCAM, TOP2A) and 10 downregulated (SNAI1, ZEB1, EMILIN2, SOX9, VIM, NRP1, IHH, EFNB2, CTGF, COL4A5) genes in LNCaP-iSnail cells treated with Dox for 5 days to undergo an EMT (EMT5) followed by removal for 3, 5, and 20 days to allow for a MErT (MErT3-20). Expression was normalised to RPL32 and fold change is relative to untreated cells (No Dox). n = 3 biological replicates. Error bars indicate SEM. One-way ANOVA, p value: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 98

A I

I

I

V

I

r

r

r

e

e

t

e

t

t

s

s

s

u

u

l

u

l

l

C

C C

F O X C 1 QKI MMP2 EFNB2 VIM NRP1 IHH

S O X 9 PDGFRB EMP1 CTGF

Z E B 1 T B X 2

EPCAM N A V 2 C LD N4 CDH1 N C A P G 2 DDC ESRP1 A D D 3 PRUNE2

T O P 2A K LH L1

I

I

V

I

V

I

r

r

r

e

e

e

t

t

t

s

s

s

u

u

l

u

l

l

C

C C

N

E

M

N

M

M

E

M

M

M

N

E

M

M

M

M

M

o

o

M

o

E

E

E

E

E

E

E

E

E

T

r

T

r

r

r

D

r

r

D

T

r

r

r

D

T

T

T

T

T

T

T

T

T

5

5

5

o

o

o

3

5

3

2

5

2

3

5

2

x

x

x

0

0

0

r e la tiv e

r o w m i n r o w m a x A

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 99 Figure 4.6. Enriched GO-BPs in the indicated clusters.

Gene set enrichment analysis (GSEA) was performed to identify enriched GO- BPs across all the clusters (first column), clusters I & II (second column), III & IV (third column), and V & VI (fourth column). Left y-axis represents the normalised enrichment score (NES) for the enriched GO-BPs in untreated LNCaP-iSnail cells (No Dox; first row), cells treated with Dox for 5 days to undergo an EMT (EMT5; second row) and followed by removal for 3, 5, and 20 days to allow for a MErT (MErT3-20; third – fifth row). Blue dotted lines represent a NES of ± 2; the cut-off used to determine enriched GO-BPs. Right y-axis represents the false discovery rate (FDR) for each enriched GO-BP. Red dotted lines represents a FDR of 0.05; the cut- off used to identify significant GO-BPs

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 100 A i ii M E rT M E rT 4 0 0 4 0 0

* * * *

e * * * * r

o 2 0 0 2 0 0

c

S

e

r 0 0

u

t

a

n g

i -2 0 0 -2 0 0 S

-4 0 0 -4 0 0

ry is ry is a s a s a ) a m ) t m 1 t ri 9 s ri 3 s ) 5 a a 9 P t ) P 1 t = e = e 1 n 3 n = ( M 3 ( M n = ( (n

ii B i C C R C C R M E rT M E rT 3 0 0 2 0 0

* * * * * * e

r 2 0 0

o 1 0 0

c

S

1 0 0 e

r 0 u

t 0

a

n g

i -1 0 0

-1 0 0 S

-2 0 0 -2 0 0

ry is ry is a s a s a ) a m ) t m 1 t i 9 s ) i s ) r 5 r 3 P ta 3 P 1 ta 9 = 3 = e 1 n e = = ( M (n M (n (n

Figure 4.7. MErT is enriched in prostate cancer metastasis.

(A) Dot plots showing the enrichment score of the MErT profile in samples of localised and metastatic prostate cancer from the Ai. Grasso dataset (Grasso et al. 2012) and, Aii. the Taylor dataset (Taylor et al. 2010). (B). Dot plots showing the enrichment score of the MErT profile with cell cycle genes removed (MErTCCR) in samples of primary and metastatic prostate cancer from the Bi. Grasso dataset and, Bii. the Taylor dataset. Error bars indicate SEM, unpaired t-test, p-value: **<0.01, **** < 0.0001.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 101 (identified by using Ingenuity Pathway Analysis; IPA) were removed from the MErT profile, and the samples were re-examined to show that a similar trend remained (Figure 4.7 Bi and ii). In summary, the data supports an enrichment of the MErT profile in mCRPC.

4.2.3 MErT is not the antithesis of EMT – revealing a transcriptional footprint of plasticity in mCRPC.

It was examined whether cells having passed through a reversible EMT acquired a transcriptional footprint of this experience. This identified four distinct clusters of transcripts that persisted in their differential expression. Clusters VII and VIII contained transcripts significantly altered during EMT and persisted throughout MErT (cluster VII, n = 65 transcripts; cluster VIII, n = 54 transcripts; Figure 4.8 Ai and ii); and clusters IX and X that contained transcripts not significantly altered during EMT and then changed early during MErT (at 3 days) that persisted throughout the 20 day MErT (cluster IX, n = 56 transcripts; cluster X, n = 22 transcripts; Figure 4.8 Bi and ii) (for cluster transcript identities see Table S1 accesible here: http://tinyurl.com/hdmm9ez). Next, the expression of these transcriptional clusters was assessed in clinical patient samples of localised PCa and mCRPC (Grasso et al. 2012). This revealed mCRPC samples to have increased enrichment of the “EMT persistent” (Figure 4.8Ci; combined clusters VII/VIII) and “MET unique” (Figure 4.8Cii; combined clusters IX/X) signatures, indicating more prominent expression of these transcriptional clusters in mCRPC than in localised PCa samples.

Having potentially identified transcriptional hallmarks of cells having experienced a reversible EMT in samples of mCRPC, the co-expression of these clusters was examined. The “EMT persistent” (clusters VII/VIII) and “MET unique” (clusters IX/X) signature scores had a strong positive correlation in both localised and mCRPC samples (Pearson correlation, r = 0.86; p < 0.0001; Figure 4.8Di). This revealed the co-expression/co-existence of these signatures, or lack thereof in negative scoring samples, at both primary and metastatic sites. This correlation was not observed in benign prostate tissue samples collected from the same patient cohort (Figure 4.8Dii) supporting the involvement of a reversible EMT in cancer

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 102 A i C lu s te r V II B i C lu s te r IX 1 0 1 2

T U B B 3 P O U 4 F 1 y

y L R R T M 4 TP R N

t t

i 1 1

i s s A R H G A P 2 2

F A R P 1 n

n e

e 8 t t 1 0 R B B P 7 S L C 1 6 A 3 n n P O U 4F1

T U B B 3 i

i

l

l a

C lu s te r a v g a C lu s te r a v g

n

n

g

g i

i 7

s

s

6

2

2

g

g

o o

l 6 l

4 5

x x

0 0

3 5 3 5

5 5

o o

2 2

T T T T

T T

r r r r

D D

T T

M M

r r

E E E E

x x

0 0

3 5 3 5

5 5

o o

E E

E E

o o

M M M M

2 2

T T T T

T T

N N

r r r r

M M

D D

T T

M M

r r

E E E E

o o

E E

E E

M M M M

N N

M M

ii C lu s te r V III ii C lu s te r X 1 0 1 2

D O C K 11 M A P 3 K 5 S W A P 7 0

y t

C A D P S 2 y C lu s te r a v g i C lu s te r a v g

t S G T B

i s

s 1 1 n D O C K 1 1 n IQ G A P 1

e 8

t

e

t n

i D S C 3

n

i H L A -D M B

l

l

a 1 0

a n

n S W A P 7 0

g

i

g

i

s

6 s 8

2

2

g

g

o

l

o l 7

4

x x

0 0

3 5 3 5

5 5

o o

2 2

T T T T

T T

r r r r

D D

T T

M M

r r

E E E E

x

0

3 5

5

o o

x

E E

5

0

3

5

E E

o

M M M M

2

T T

T

o

N N

2

T

T

T

r r

M M

r

D r

T

D

T

M

r

E E

M

r

E

E

o

E

o

E

E

E M

M re la tiv e re la tiv e

M

N M

N

M M

ro w m in ro w m a x ro w m in ro w m a x

C i E M T p e rs is te n t ii M E T u n iq u e D i ii 6 0 4 0 6 0 2 0

**** **** r = 0 .8 5 5 1 r = 0 .2 6 7 3

e

e r

4 0 r 4 0 p = 0 .1 6 9 2 o

o p < 0 .0 0 0 1

e

e c

c 2 0 1 0

u

u

S S

2 0 q 2 0 q

i

i

n

n

e

e

u

u

r r

0 0 0 0

u

u

T

T

t

t

a a

-2 0 E -2 0 E

n

n

M

M g

g -2 0 -1 0 i i .P r i m a r y

B e n ig n S S -4 0 -4 0 . .m C R P C -6 0 -4 0 -6 0 -2 0 -4 0 -2 0 0 2 0 4 0 -2 0 -1 5 -1 0 -5 0 5 ry C ry C a P a P im R im R E M T p e rs is te n t E M T p e rs is te n t r C r C P m P m

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 103 Figure 4.8. MET is not the antithesis of EMT and is enriched in mCRPC

LNCaP-iSnail cells were treated with Dox for 5 days to induce an EMT (EMT5), followed by removal for 3, 5, and 20 days to allow for a MErT (MErT3-20). (A) Clusters representing EMT persistent transcripts: Ai. transcripts upregulated during EMT and stay upregulated and Aii. transcripts downregulated during EMT and stay downregulated during MErT (B) Clusters representing uniquely activated MET transcripts: Bi. transcripts upregulated at 3 days of MErT and stay up regulated over the 20 days and Bii. transcripts downregulated at 3 days of MErT and stay downregulated over the 20 days. Heatmap colour gradient refers to z-score. (C) Enrichment of the Ci. EMT persistent and Cii. MET unique signatures in primary PCa and mCRPC. (D) Correlation plot of the EMT persistent and MET unique signatures in Di. primary PCa and mCRPC, and Dii. benign prostate tissue from Grasso et al (Grasso et al. 2012). Error bars indicate SEM, n = 3 biological replicates, unpaired t-test, p-value: **** < 0.0001. Cluster avg: the average transcriptional pattern of the cluster.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 104 progression. Collectively, these data provide evidence for the clinical relevance of these transcript expression clusters in mCRPC and their potential to serve as novel sentinel biomarkers for cells that have experienced a complete cycle of EMT/MErT.

4.2.4 A metastasis-derived plasticity signature expressed in primary tumour samples predicts poor patient prognosis.

Next, GSEA was used to identify a clinically relevant set of genes from the EMT/MErT transcriptional data obtained using the LNCaP-iSnail cells. For this, the same ranked list used previously in section 4.2.2 for the delineation of the MErT profile, was interrogated with a list of genes found to be upregulated in lethal samples of mCRPC compared to treatment naïve PCa samples from the Grasso dataset (Grasso et al., 2012). The analysis identified a core set of 698 plasticity genes, that were enriched in mCRPC, termed herein as the Metastatic Plasticity Signature; MPS) (Table 4.1). A description of the genes within the MPS can be found here: https://tinyurl.com/gvknxd8. To further validate that the MPS was enriched in metastases, the MPS was examined in a number of additional prostate cancer patient cohorts containing metastatic samples (Chandran et al. 2007; LaTulippe et al. 2002; Taylor et al. 2010; Varambally et al. 2005; Yu et al. 2004). This identified the MPS to overlap significantly with gene signatures upregulated in metastatic samples across all patient cohorts examined (Figure 4.9). Concept analysis using the Oncomine database (www.oncomine.org) revealed MPS genes to be significantly associated (p 0.01, odds ratio OR ≥ 1.9) with not only metastatic prostate cancer samples, but also in primary prostate tumour samples from patients with recurrent disease (1-5 years) and poor survival (1-5 years; Table 4.2). The stratification of patients based on a positive or negative MPS expression score revealed patients with MPS positive primary tumours (MPS positive; Figure 4.10) to have a significantly shorter time to biochemical recurrence (Figure 4.10Ai; log rank p < 0.0001; hazard ratio [HR]: 7.08) and shorter overall survival (Figure 4.10Aii; log rank p < 0.0001; HR: 3.27) compared to patients with primary tumours having MPS negative scores (MPS negative; Figure 4.10 Ai and ii), respectively.

As many human carcinoma types share similar mechanisms of tumour progression, including evidence of EMT (Hoshino et al. 2009; Sethi et al. 2010; Malek et al. 2011; Pereira et al. 2015), the prognostic ability of the MPS was assessed in

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 105 Table 4.1. The Metastatic Plasticity Signature (MPS)

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 106

Figure 4.9. The Metastatic Plasticity Signature (MPS) overlaps and is significantly expressed in the metastasis of a number of prostate cancer datasets.

Datasets examined: Varambally et al (Varambally et al. 2005), Taylor et al (Taylor et al. 2010), LaTulippe et al (LaTulippe et al. 2002), Chandran et al (Chandran et al. 2007), and Yu et al (Yu et al. 2004). Dotted line indicates p-value of 0.01.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 107 Table 4.2. Enrichment of the Metastatic Plasticity Signature (MPS) in multiple PCa patient cohorts.

Dataset Property Overlap (# of genes) P-Value Odds Ratio Top % Taylor Prostate 3 Metastatic Status 287 4.38E-109 7.2 10 Varambally Prostate Metastatic Status 238 6.14E-73 5.4 10 LaTulippe Prostate Metastatic Status 107 2.40E-24 3.8 10 Yu Prostate Metastatic Status 103 4.74E-22 3.6 10 Chandran Prostate Metastatic Status 121 8.59E-14 2.3 10 Vanaja Prostate Metastatic Status 128 2.44E-13 2.2 10 Lapointe Prostate Metastatic Status 99 3.38E-13 2.5 10 Holzbeierlein Prostate Metastatic Status 50 7.67E-08 2.5 10 Tamura Prostate Metastatic Status 93 9.64E-08 1.9 10 Ramaswamy Multi-cancer Metastatic Status 74 2.42E-06 1.9 10 Magee Prostate Metastatic Status 21 0.003 2.1 5 Ramaswamy Multi-cancer 2 Metastatic Status 35 0.003 1.7 5 Glinsky Prostate Recurrence Status at 3 Years 93 2.94E-08 2 10 Recurrence Status at 5 Years 90 2.32E-07 1.9 10 Nakagawa Prostate* Recurrence Status at 1 Year 7 3.42E-04 7.5 5 Recurrence Status at 3 Years 12 7.50E-06 7.4 10 Recurrence Status at 5 Years 12 7.50E-06 7.4 10 Survival Status at 3 Years 12 7.50E-06 7.4 10 Survival Status at 5 Years 9 3.52E-06 12 5 Nakagawa Prostate 2* Recurrence Status at 1 Year 7 8.37E-04 6.4 5 Recurrence Status at 3 Years 12 3.94E-05 6 10 Recurrence Status at 5 Years 7 8.37E-04 6.4 5 Survival Status at 3 Years 13 5.95E-06 7 10 Survival Status at 5 Years 10 0.001 4.3 10 Setlur Prostate Survival Status at 1 Year 46 2.28E-04 1.9 10 Survival Status at 3 Years 46 8.94E-14 4.4 5 Survival Status at 5 Years 50 1.84E-16 4.9 5 Taylor Prostate 3 Recurrence Status at 3 Years 84 4.31E-14 2.8 5 Recurrence Status at 5 Years 48 8.95E-27 9.1 1 Barwick Prostate Recurrence Status at 5 Years 7 3.42E-04 7.5 5

* Nakagawa datasets were profiled using the Illumina DASL human cancer panel V1 (502 genes; Nakagawa Prostate) or

Illumina DASL expression microarrays (526 genes; Nakagawa Prostate 2) associated with PCa progression.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 108 Ai ii

l P ro s ta te P ro s ta te

a

l

c

l

a

i

a

v

i m 1 .0 v 1 .0

M P S + v e (n = 2 3 ) i p < 0 .0 0 0 1 M P S + v e (n = 5 9 )

v

e

r

v

r h

u M P S -v e (n = 5 6 ) M P S -v e (n = 3 0 4 )

c

u

s

s

o

i

e

f

b e

1 2 o

r 1 4 5

f

f

-

y

o t

0 .5 m th s 0 .5 e

i m th s

l

y

c

i t

i 7 2

n

b

l

i

e

a b

r m th s

b

r

a

o

u b

p < 0 .0 0 0 1 r

c

o

P e

r 0 .0 0 .0

r P

0 5 0 1 0 0 1 5 0 0 1 0 0 2 0 0 3 0 0

M o n th s M o n th s B i ii

B re a s t B re a s t

l

l

a

a

v i

v M P S + v e (n = 1 1 2 ) i

1 .0 v 1 .0 M P S + v e (n = 3 7 )

r

v f

r M P S -v e (n = 1 8 2 ) u

o M P S -v e (n = 3 0 )

u

s

s

y

t

e

i

f

l

i

e

o

r

b

f

y

-

a t

i 0 .5

s 0 .5

l

i

b

i s

1 3 2 o

b r

a 4 0

a

t

P

s b

m th s

a m th s

o t r p < 0 .0 0 0 1

e p = 0 .0 1 1 3 P

0 .0 0 .0 m 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 0 5 0 1 0 0 1 5 0 2 0 0

M o n th s M o n th s C i ii

Lung Lung

l

l

a

a

v

v i

1 .0 M P S + v e (n = 3 3 ) i 1 .0 M P S + v e (n = 1 3 )

v

v

r

r

f

f

u

u o

M P S -v e (n = 1 9 3 ) o M P S -v e (n = 4 5 )

s

s

y

y

t

t

e

e

i

i

l

l

i

i

e

e r

r 1 0 5

b

b

f

f

-

- a

0 .5 a 0 .5 e

e m th s

b

b c

c 7 .7

o

o n

5 7 n

r

r

e

e P

P m th s

r

r

m th s

r

r

u

u c

p = 0 .0 0 3 8 c p = 0 .0 0 1 6 e

0 .0 e 0 .0

r r 0 5 0 1 0 0 1 5 0 0 5 0 1 0 0 1 5 0 2 0 0

M o n th s M o n th s

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 109 Figure 4.10. The MPS predicts poor patient outcome across multiple cancers.

Kaplan-Meyer curves showing the probability of recurrence and survival of patients based on their positive or negative MPS score. Ai. Biochemical recurrence and Aii. survival of patients with PCa (Glinsky et al. 2004; Setlur et al. 2008). Bi. Patient survival for patients with breast cancer (van de Vijver et al. 2002) and Bii. time to breast cancer metastasis (van 't Veer et al. 2002). (Ci and Cii) Tumour recurrence in patients with lung cancer (Okayama et al. 2012; Lee, Son, et al. 2008). Dotted line indicates the median time. Statistical test used was log-rank (Mantel-Cox). MPS: Metastatic Plasticity Signature.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 110 additional human cancer types. As seen with the prostate cancer patient cohorts, MPS genes were significantly (p 0.01, OR ≥ 1.3) associated with recurrent disease, time to metastasis and poor overall survival across a number of carcinoma types, including breast, lung, kidney and colon cancers (Table 4.3).

For example, breast cancer patients (n=294) (van de Vijver et al. 2002) with MPS positive primary tumour scores had a significantly shorter overall survival (Figure 4.10Bi; log rank p < 0.0001; HR: 4.18) than patients with an MPS negative tumour scores. In a separate cohort of breast cancer patients (n= 97) (van 't Veer et al. 2002), patients with MPS positive tumour scores had a shorter time to metastasis (Figure 4.10Bii, log rank p = 0.0113; HR: 2.22) than patients with MPS negative scoring tumours. Additionally, in two distinct cohorts of lung adenocarcinoma patients (n=226) (Okayama et al. 2012), n=58 (Lee, Son, et al. 2008)), patients with MPS positive primary tumour scores also presented with shorter time to recurrence (Figure 4.10Ci; log rank = 0.0038; HR: 3.07; and Figure 4.10Cii; log rank = 0.0016; HR: 5.40). Taken together, these results not only associate the MPS with poor clinical outcome but also provide evidence that the cycling of epithelial and mesenchymal states occurs within primary tumours and is related to tumour progression and treatment resistance.

4.2.5 The MPS is not reliant on cell cycle gene expression.

To elucidate biological pathways responsible for plasticity-associated poor clinical outcomes, the MPS was examined for enrichment of biological pathways from the Gene Ontology Consortium (GO-BPs). As observed in previous analyses (Figure 4.1A), the MPS was also predominated by cell cycle/proliferation, DNA repair/recombination, cell movement/motility and metabolism-related processes (Figure 4.11). Cell cycle, DNA repair and cell motility have well-established roles in promoting metastatic progression and poor clinical outcomes. A number of established diagnostic signatures currently used in the clinic are enriched with cell cycle and proliferation-related genes. Indeed, patients with primary tumours that have a high proliferation score have poorer prognosis than patients that have a low score (Cooperberg et al. 2013; Cuzick et al. 2011). Therefore, it was examined whether the predictive power of the MPS was also reliant on the cell cycle gene content. For this, a MPS devoid of cell

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 111 Table 4.3. Enrichment of the Metastatic Plasticity Signature (MPS) in multiple cancer patient cohorts.

Dataset Property Overlap (# of genes) P-Value Odds Ratio Top % vandeVijver Breast Metastatic Event Status at 5 Years 141 1.35E-59 7.3 5 Loi Breast 3 Metastatic Event Status at 5 Years 218 6.34E-59 4.7 10 vandeVijver Breast Recurrence Status at 5 Years 140 9.81E-59 7.2 5 Curtis Breast Survival Status at 5 Years 151 5.23E-58 6.3 5 Loi Breast 3 Recurrence Status at 5 Years 213 1.28E-55 4.5 10 Curtis Breast Survival Status at 3 Years 144 1.71E-52 5.9 5 vandeVijver Breast Recurrence Status at 3 Years 132 5.04E-52 6.6 5 vandeVijver Breast Metastatic Event Status at 3 Years 125 2.06E-46 6.1 5 DirectorsChallenge Lung Survival Status at 3 Years 113 3.63E-42 6.1 5 vandeVijver Breast Survival Status at 5 Years 117 2.62E-40 5.5 5 Schmidt Breast Metastatic Event Status at 5 Years 109 4.56E-39 5.8 5 DirectorsChallenge Lung Survival Status at 5 Years 109 4.56E-39 5.8 5 vantVeer Breast Metastatic Event Status at 3 Years 102 1.75E-38 6 5 Pawitan Breast Survival Status at 5 Years 123 8.33E-38 4.8 5 Loi Breast 3 Recurrence Status at 3 Years 183 1.64E-37 3.6 10 Schmidt Breast Metastatic Event Status at 3 Years 105 4.63E-36 5.5 5 Kao Breast Survival Status at 5 Years 179 2.49E-35 3.4 10 Loi Breast 3 Metastatic Event Status at 3 Years 179 2.49E-35 3.4 10 Curtis Breast Survival Status at 5 Years 177 8.13E-35 3.4 10 Wang Breast Recurrence Status at 3 Years 148 3.32E-33 3.8 10 Wang Breast Recurrence Status at 5 Years 146 4.84E-32 3.8 10 DirectorsChallenge Lung Survival Status at 1 Year 99 9.88E-32 5 5 vantVeer Breast Metastatic Event Status at 5 Years 91 2.74E-30 5.1 5 Loi Breast Metastatic Event Status at 5 Years 169 4.02E-30 3.2 10 Loi Breast Recurrence Status at 5 Years 169 4.02E-30 3.2 10 vandeVijver Breast Survival Status at 3 Years 102 8.50E-30 4.5 5 Pawitan Breast Survival Status at 3 Years 164 1.17E-28 3.1 10 Bos Breast Metastatic Event Status at 3 Years 109 6.33E-28 4 5 Desmedt Breast Metastatic Event Status at 5 Years 93 1.26E-27 4.6 5 Laurent Melanoma Metastatic Event Status at 5 Years 163 3.58E-27 3 10 Taylor Prostate 3 Recurrence Status at 5 Years 48 8.95E-27 9.1 1 Zhan Myeloma 2 Survival Status at 3 Years 162 1.08E-26 3 10 Loi Breast Metastatic Event Status at 3 Years 106 3.80E-26 3.8 5 Loi Breast Recurrence Status at 3 Years 106 3.80E-26 3.8 5 Phillips Brain Survival Status at 3 Years 104 1.04E-25 3.8 5 Okayama Lung Recurrence Status at 5 Years 105 1.45E-25 3.8 5 Lee Lung Recurrence Status at 1 Year 104 5.49E-25 3.7 5 Schmidt Breast Metastatic Event Status at 1 Year 133 6.41E-25 3.3 10 Esserman Breast Survival Status at 3 Years 153 4.71E-22 2.7 10 Boersma Breast Survival Status at 1 Year 84 6.64E-22 4 5 Zhan Myeloma 2 Survival Status at 1 Year 151 1.11E-21 2.7 10 TCGA Renal Survival Status at 3 Years 151 9.81E-21 2.6 10 Phillips Brain Survival Status at 1 Year 146 2.42E-20 2.6 10 Kao Breast Survival Status at 3 Years 146 1.48E-19 2.6 10 Nakayama Sarcoma 2 Survival Status at 5 Years 144 9.83E-19 2.5 10 Okayama Lung Recurrence Status at 3 Years 143 2.50E-18 2.5 10 Desmedt Breast Recurrence Status at 3 Years 119 3.80E-18 2.8 10 Okayama Lung Survival Status at 5 Years 91 5.44E-18 3.1 5 Yang Renal Survival Status at 5 Years 91 5.44E-18 3.1 5 Phillips Brain Survival Status at 5 Years 140 7.40E-18 2.5 10 1 Table shows the top 50 significant (p<0.01) datasets (n=262 datasets, accessible here: http://tinyurl.com/gs53j9r).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 112

Figure 4.11. Bubble chart indicating the enrichment of biological processes from the Gene Ontology Consortium in the Metastatic Plasticity Signature (MPS).

GeneGo MetaCore was used to identify the enriched GO-BPs in the MPS. Y-axis represents increasing significance and the x-axis the gene count. The size of the circles indicates the fold enrichment (observed genes/expected genes within each GO-BP). Plotly was used to generate the bubble chart.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 113 Table 4.4. The Metastatic Plasticity Signature with cell cycle genes removed (MPSCCR).

EBLN2 NACC2 PRAME FAM73B MAST2 TRAPPC9 LARP4 STK11IP AK4 ABCC5 ZBTB37 KCNG1 ASH1L FANCI FANCB SS18L1 PRRC2B OPLAH ORAI2 ASB6 AASDH CYFIP2 PIF1 AGBL5 BTN2A3P PDE7A SMG6 BBC3 ARGLU1 VANGL1 SLC2A1 TET2 PLEKHS1 RNF216 ZFAND5 PDS5B ZNF395 TMEM184A RFC5 IP6K2 PROSER1 EFNA1 TIGD1 TNIP1 PRR12 CLPB SFI1 DISP2 TICRR METRNL WIPI2 ZNF493 SYT4 SGK2 SNHG7 KIF24 FAM215A TCF25 SDCBP YEATS2 MUCL1 PACSIN1 WDR4 TUBA1C MGME1 FAM120C NPEPL1 CIC TTC39B POLD3 SREBF2 PUS1 LRRC75B BSDC1 ZNF250 B4GALNT4 RCE1 KHDRBS3 FAM195A TRA2B CCDC174 ZER1 VMO1 CCBL1 ERP44 SLC25A13 MYO19 VPS13B XXYLT1 PDIA2 ZNF236 PGLS SLC7A5 SLC7A6 HPRT1 RBM6 SIMC1 EMD LRRC8A ZNF253 PAQR4 GAPDH STRBP NPW PPP1R35 ZNF783 SYNRG ULK1 RNFT2 CENPP GINS2 NR2C2 ARHGEF39 BGLAP SMARCC2 CHST12 OVOL1 PGP CDCA4 TMSB10 CACFD1 LPGAT1 MORC2 RP9 A2ML1 ATL3 ARRDC1 MAP3K3 CCDC171 DDX6 PQLC1 ZNF197 TRIM10 SPC24 ARHGAP11A HSD17B2 YBX3 ZC3H11A MSX1 KLF11 RIMKLA SH3GLB2 ZNF524 INSIG1 CTSV APEX2 THBS3 CGGBP1 TRIB3 STC2 WDR90 ATAD2 DDX39A SECISBP2L SCAMP5 PDRG1 PRKAG2 ECE2 ZMIZ2 HIST1H3B AGO2 UBE2Z SETD5 KIAA0907 BRWD1 HYAL3 HSD11B2 BBX METTL21A FAM193B CRIP2 EHMT2 SLC5A6 UBE2T NUDT19 CD2BP2 ARHGAP35 RGS12 UNK SOX11 TRPM2 MYO1C FUT3 ATXN2 CCDC183 DFNB31 VGLL4 ITPKA PPFIBP1 GTPBP3 NDUFS8 CRYGS MCM3AP-AS1 CCDC50 GTF2IRD1 RYBP OSBPL3 WDR5 NOXA1 HTT KMT2C PI4K2A NFE2L1 ARFGAP1 IFITM5 FGL1 PLEKHA1 PTTG1IP NAPRT USP42 SRGAP2 CHD8 CCDC88C CDCP1 SRP19 TTLL5 KANSL1 ANKIB1 PIP5K1C PIP4K2B KIAA1257 ABCB6 PPM1F ETF1 SLC52A2 ZNF160 FAM160B2 RNF44 MCCC2 MLPH ZMYND19 HIST1H1D PDXDC2P PPP1R3F MED13 PRR7 OAS3 MRPL38 EPPK1 TPRN PKM TIA1 AP3M2 TMEM39B TMC8 LRSAM1 SCARB1 PTMS RNF216P1 HOXB9 TULP3 CRTC2 DONSON ABCA3 RASSF7 KIAA1549 KRBOX4 SEMA6C TOM1L2 DIRAS1 AGPAT3 DHODH TADA2B TMEM201 SUSD2 KRBA1 MYO9B PXN ALCAM SGK3 SLC46A1 CXXC5 HCN3 ZFYVE27 SMYD4 ZBTB43 PROM2 WHSC1 HN1 UBE2H FAM13A AIFM2 DSTYK GPSM1 UTP23 PUS7 ALG3 ASCC3 EPB41L4B GGA3 DCAF5 MMP11 RHPN1 TIMELESS POLR3E SLC25A19 CPT1C ATXN7L2 CCDC120 SETDB1 GRPEL2 TRMU FBXL6 BRI3BP HSD11B1L ZFYVE19 SHPRH TMEM132A RALGAPA2 RNF213 EVPL FADS1 UBAP2 CRTC1 MPHOSPH8 LRRC61 NPDC1 POC1A GAS2L3 PSD3 CCDC40 CDC42BPB TECPR2 ACAP3 PTK6 SECTM1 MID1IP1 LRP8 GRB14 AGPAT1 SOX12 LPP RNF24 ARTN WDR62 CCDC85C SEC14L1 SLFNL1 SERPINA1 ZNF821 ADAP2 MNX1 FKBP4 ZDHHC21 BCOR POU2F1 EXOC4 RBM12B OIP5 HIP1R UCK2 GRB7 TTC21A SLC27A1 USP4 ZNF3 GLDC GRPEL1 TRMT61A ZNF165 PHF21A HDAC7 ZNF625 ZFP36L1 ABHD8 OXR1 LAD1 USP43 GDI1 ELMOD3 AKAP8L CMTM7 RALGPS1 LARS LPIN1 ZCCHC6 IRX5 PILRB TOP1MT UBOX5 ZBTB7B SCLY POP1 VWA1 CPS1 PHKA2 RAPH1 PFKP REEP3 KIF16B FAM64A RPP25 TEP1 MZF1 CACNB3 ZKSCAN1 FGFR1OP PDXDC1 WDR54 C2CD2L ERC1 PCMTD2 CERK CCDC151 FARP2 NUP210 CCDC15 PGK1 ANK3 LUC7L PIK3R3 GSDMB RAB26 DAZAP1 NMU RNF208 EGLN3 WWC3 TTYH3 KIF3C RDM1 GRAMD1C TMEM105 SDC1 RERE SH3BGRL3 S100PBP BPTF GET4 EEF1A2 RMND5B CLN3 R3HDM2 POGZ ZNF646 ZNF500 TMEM106C SLC26A1 ATP5D MRPS17 SLC16A3 FBXL17 PLXNA1 FOXJ1 SDR16C5 STX2 RTN3 POLQ ALPK1 ARHGAP33 ZNF609 ZNF587 CPT1A PSME3 NOL6 CEP78 ALB MAP4K4 KDM5A CELSR3 KIAA1671 SULT1A1 RNASEH2A JAG2 PORCN SENP5 TYK2 AGRN MCF2L CPNE7 PKP3 DNAH11 TAPBP CCNL2 TSC22D4 FAM69B ARHGAP26 SYNE2 TROAP VDR ITIH4 TEF GON4L ORAOV1 GNPTAB BSPRY LLGL2 DHCR7 FOXD3 CYHR1 TBX1 LHX2 COMTD1 AIM1L ISYNA1 DOK3 SNX29 CLCN7 KCNQ1OT1 RABGAP1 ABHD11 MAST1 REEP4 LRRC45 F8 FBXO11 ZNF785 METRN PPFIA1 SLCO4A1 ITSN2 GOLGA3 HIST1H4B HIST1H3D ITPR3

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 114 cycle and proliferation genes was established via the removal of genes annotated within the GO-BP category of “cell cycle” and cell cycle progression (CCP) signature genes (Cuzick et al. 2011) (termed herein as the MPS Cell Cycle Removed; MPSCCR; 515 genes; Figure 4.15 and Table 4.4). A description of the genes within the MPSCCR can be found here: https://tinyurl.com/gvknxd8. Patients from prostate, breast and lung cancer cohorts that were previously examined using the MPS (Figure 4.10) were stratified into MPSCCR positive and negative primary tumour scoring groups. As seen previously for the MPS (Figure 4.10), the prognostic ability of the MPSCCR in prostate (Figure 4.15 Ai and Aii), breast (Figure 4.12 Bi and Bii) and lung cancer (Figure 4.12 Ci and Cii) was retained, whereby patients with a MPSCCR positive primary tumour scores had a significantly worse outcome than patients with MPSCCR negative primary tumour scores. Additionally, MPSCCR genes were significantly (p 0.01, OR ≥ 1.4) associated with recurrent disease, time to metastasis and poor overall survival across a number of carcinoma types, including breast, lung, kidney and colon cancers (Table 4.5). These results support that the predictive ability of the MPS is not reliant on cell cycle genes.

4.2.6 Epithelial-mesenchymal plasticity (EMP) fuelled metabolic reprogramming is a predictor of clinical outcome.

Previous analysis of the GO-BPs enriched in the MPS performed in section 4.2.5, indicated that after cell cycle the next most prominent category of enriched processes were related to metabolism. Indeed, analysis of the GO-BPs enriched in the MPSCCR show high enrichment of metabolically related processes (Figure 4.13). The involvement of cell cycle and proliferation has established roles in cancer progression as a predictor of poor patient outcome (Cooperberg et al. 2013; Cuzick et al. 2011), however, the investigation of metabolism has been less studied. While the reprogramming of central metabolic pathways, such as oxidative phosphorylation (OXPHOS) and glycolysis, are “hallmarks” of cancer with well-established roles in regulating tumour growth and survival (Hanahan and Weinberg 2011), it is only recently that metabolic reprogramming has been implicated in tumour invasion and metastasis using pre-clinical models (LeBleu et al. 2014; Dupuy et al. 2015).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 115 Ai ii

l P ro s ta te P ro s ta te

a

l

l

c

a

i a

v C C R C C R

i v m M P S + v e (n = 2 3 ) M P S + v e (n = 4 4 )

1 .0 i 1 .0

v

e r

C C R v C C R r h M P S -v e (n = 3 1 9 )

u M P S -v e (n = 5 6 )

u

c

s

s o

p = 0 .0 0 2 1

i

e

f

b e

o 1 3 5

1 6

r

f

f

y

-

o t

m th s m th s

0 .5 i 0 .5

e

l

y

i

c

t

i

b n

l 1 1 6

i

a

e

b

r

b r

a m th s

o

u

r b

c p < 0 .0 0 0 1

o

P

e

r 0 .0 0 .0

r P

0 5 0 1 0 0 1 5 0 0 1 0 0 2 0 0 3 0 0

M o n th s M o n th s

B i ii

B re a s t B re a s t

l

l

a a

C C R v C C R i

v M P S + v e (n = 9 7 ) M P S + v e (n = 3 1 ) i

1 .0 v 1 .0

r v

C C R f C C R r

M P S -v e (n = 1 9 7 ) u M P S -v e (n = 6 6 )

o

u

s

s

y

t

e

i

f

l

i

e

o

r

b

f

y

-

a

t i

0 .5 s 0 .5

l

i

b

i

s

o b

1 4 1 r

a 4 0

a

t

P

s b m th s

a m th s

o t r p < 0 .0 0 0 1

e p = 0 .0 0 1 4 P

0 .0 0 .0 m 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 0 5 0 1 0 0 1 5 0 2 0 0

M o n th s M o n th s C i ii

Lung Lung

l

l

a

a v

C C R v i i C C R

1 .0 M P S + v e (n = 2 4 ) 1 .0 M P S + v e (n = 5 )

v

v

f

r r

C C R f C C R

o u M P S -v e (n = 2 0 2 ) u

o M P S -v e (n = 5 3 )

s

s

y

y

t

t

i

e e

i 4 5

l

l

i

i

e

e

r

r

b

b f

f m th s

-

- a

0 .5 a 0 .5

e

e

b

b

c c

o 4 .6

o

r n

4 1 n

r

e

e P

P m th s

r

r

m th s

r

r

u

u c

p = 0 .0 0 0 3 c p = 0 .0 0 0 8 e

0 .0 e 0 .0

r r 0 5 0 1 0 0 1 5 0 0 5 0 1 0 0 1 5 0 2 0 0

M o n th s M o n th s

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 116 Figure 4.12. The Metastatic Plasticity Signature (MPS) devoid of cell cycle genes (MPSCCR) predicts poor patient outcome in prostate, breast, and lung cancer.

The MPS was regenerated devoid of cell cycle genes (MPSCCR; Table 4.4) Kaplan-Meyer curves showing the probability of recurrence and survival of patients based on their positive or negative MPSCCR score. Ai. Biochemical recurrence and Aii. survival of patients with PCa (Glinsky et al. 2004; Setlur et al. 2008). Bi. The probability of survival and Bii. metastasis-free survival in patients with breast cancer (van de Vijver et al. 2002; van 't Veer et al. 2002). Ci and Cii. The probability of recurrence-free survival in patients with lung cancer (Lee, Son, et al. 2008; Okayama et al. 2012). Dotted line indicates the median time. Statistical test used was log-rank (Mantel-Cox). MPS: Metastatic Plasticity Signature

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 117 Table 4.5. Enrichment of the Metastatic Plasticity Signature devoid of cell cycle genes (MPSCCR) in multiple cancer patient cohorts.

Dataset Property Overlap (# of genes) P-Value Odds Ratio Top % Curtis Breast Survival Status at 5 Years 112 3.02E-17 2.8 10 Loi Breast 3 Metastatic Event Status at 5 Years 112 9.15E-17 2.7 10 Curtis Breast Survival Status at 5 Years 70 1.75E-15 3.3 5 vandeVijver Breast Metastatic Event Status at 5 Years 95 4.09E-15 2.8 10 Loi Breast 3 Recurrence Status at 5 Years 107 1.21E-14 2.5 10 TCGA Renal Survival Status at 3 Years 108 3.14E-14 2.5 10 Curtis Breast Survival Status at 3 Years 103 1.88E-13 2.5 10 vandeVijver Breast Recurrence Status at 5 Years 59 2.36E-13 3.4 5 Laurent Melanoma Metastatic Event Status at 5 Years 103 4.82E-13 2.4 10 vandeVijver Breast Metastatic Event Status at 3 Years 86 2.64E-11 2.4 10 vandeVijver Breast Recurrence Status at 3 Years 85 6.50E-11 2.4 10 Loi Breast 3 Metastatic Event Status at 3 Years 97 8.31E-11 2.2 10 Loi Breast 3 Recurrence Status at 3 Years 96 1.87E-10 2.2 10 Smith Colorectal Survival Status at 3 Years 96 1.87E-10 2.2 10 Laurent Melanoma Metastatic Event Status at 3 Years 94 9.16E-10 2.1 10 Kao Breast Survival Status at 5 Years 93 1.99E-09 2.1 10 Lenz Lymphoma Survival Status at 3 Years 92 4.25E-09 2.1 10 Steidl Lymphoma Recurrence Status at 1 Year 92 4.25E-09 2.1 10 Loi Breast Metastatic Event Status at 5 Years 90 1.87E-08 2 10 Loi Breast Recurrence Status at 5 Years 90 1.87E-08 2 10 Monti Lymphoma Survival Status at 1 Year 90 1.87E-08 2 10 Wang Breast Recurrence Status at 3 Years 71 3.47E-08 2.2 10 Kuner Lung Recurrence Status at 1 Year 89 3.83E-08 2 10 Williams Renal Recurrence Status at 5 Years 64 4.97E-08 2.3 10 vandeVijver Breast Survival Status at 5 Years 77 5.03E-08 2.1 10 DirectorsChallenge Lung Survival Status at 3 Years 44 6.34E-08 2.7 5 Hou Lung Survival Status at 1 Year 88 7.76E-08 2 10 Schmidt Breast Metastatic Event Status at 3 Years 70 7.77E-08 2.2 10 Esserman Breast Recurrence Status at 5 Years 89 8.74E-08 2 10 Kao Breast Survival Status at 3 Years 87 1.55E-07 1.9 10 Kuner Lung Survival Status at 1 Year 87 1.55E-07 1.9 10 TCGA Brain 2 Survival Status at 5 Years 52 1.70E-07 2.4 5 Wang Breast Recurrence Status at 5 Years 69 1.71E-07 2.1 10 Esserman Breast Recurrence Status at 5 Years 88 1.73E-07 1.9 10 Loi Breast Metastatic Event Status at 3 Years 86 3.05E-07 1.9 10 Loi Breast Recurrence Status at 3 Years 86 3.05E-07 1.9 10 Nakayama Sarcoma 2 Survival Status at 3 Years 86 3.05E-07 1.9 10 Esserman Breast Survival Status at 5 Years 87 3.39E-07 1.9 10 Lee Bladder Survival Status at 1 Year 51 3.63E-07 2.3 5 Forster Gastric Recurrence Status at 3 Years 52 4.22E-07 2.3 5 Pawitan Breast Survival Status at 5 Years 84 4.57E-07 1.9 10 Kotliarov Brain Survival Status at 3 Years 52 4.99E-07 2.3 5 Kotliarov Brain Survival Status at 5 Years 52 4.99E-07 2.3 5 Lenz Lymphoma Survival Status at 1 Year 85 5.93E-07 1.9 10 vantVeer Breast Metastatic Event Status at 3 Years 63 5.94E-07 2.1 10 Forster Gastric Metastatic Event Status at 3 Years 19 6.22E-07 4.3 1 TCGA Breast Survival Status at 5 Years 52 9.60E-07 2.2 5 Shaknovich Lymphoma Survival Status at 1 Year 84 1.14E-06 1.9 10 Shaknovich Lymphoma Recurrence Status at 1 Year 84 1.14E-06 1.9 10 Esserman Breast Survival Status at 3 Years 85 1.24E-06 1.9 10 1 Table shows the top 50 significant (p<0.01) datasets (n=201 datasets, accessible here: http://tinyurl.com/gs53j9r).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 118

Figure 4.13. Bubble chart showing the enrichment of biological processes from the Gene Ontology Consortium in the Metastatic Plasticity Signature with cell cycle genes removed (MPSCCR).

GeneGo MetaCore was used to identify the enriched GO-BPs in the MPSCCR. Y- axis represents increasing significance and the x-axis the number of genes. The size of the circles indicates the fold enrichment (observed genes/expected genes within each GO-BP). Plotly was used to generate the bubble chart.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 119 Table 4.6. The metabolism-related genes from the Metastatic Plasticity Signature (MPSMETAB).

AK3L1 PCMTD2 B4GALNT4 COMTD1 SREBF2 USP42 C20orf72 WWC3 ZNF253 ABHD11 MYO19 ANKIB1 C3orf21 SH3BGRL3 ULK1 ZDHHC21 HPRT1 ZNF160 C7orf47 SENP5 CGGBP1 GRB7 STRBP PPP1R3F CSDA CCNL2 EHMT2 ZNF165 GINS2 HOXB9 DDX39 TEF SOX11 USP43 ARHGAP11A GGA3 FAM119A CYHR1 RYBP CYFIP2 ZNF524 CDC42BPB MLL3 FBXO11 ARFGAP1 SGK2 HSD11B2 AGPAT1 NAPRT1 RBM6 CHD8 PACSIN1 NUDT19 SLC27A1 PAR1 NR2C2 PIP4K2B SLC25A13 FUT3 HDAC7 PDXDC2 MAP3K3 TMEM39B SLC7A6 NDUFS8 PILRB PKM2 PRAME CRTC2 GAPDH NOXA1 KCNQ1OT1 ZNF673 ZBTB37 DIRAS1 PGP SRP19 ZNF785 CPT1C SMG6 GPSM1 ATL3 PPM1F BBX HSD11B1L WIPI2 SETDB1 ECE2 SCARB1 CD2BP2 IRX5 NPEPL1 ZNF821 HYAL3 SLC46A1 HTT CPS1 ZNF250 ZNF3 UBE2T ALG3 PTTG1IP TEP1 ZNF236 PFKP MYO1C POLR3E ASB6 EGLN3 SMARCC2 KIF3C GTPBP3 FBXL6 TNIP1 RERE MORC2 ZNF500 WDR5 MID1IP1 ZNF493 SLC16A3 MSX1 FOXJ1 FGL1 FKBP4 YEATS2 ALPK1 VGLL4 ZNF587 ABCB6 UCK2 CIC ALB NFE2L1 ORAOV1 MRPL38 LPIN1 OAS3 ITIH4 SRGAP2 LHX2 LRSAM1 ATP5D AGPAT3 FOXD3 PIP5K1C CXXC5 ABCA3 NOL6 PROM2 AGBL5 TULP3 UBE2H DHODH RNASEH2A UTP23 TET2 SMYD4 ASCC3 SGK3 LLGL2 RNF24 IP6K2 SHPRH AASDH WHSC1 ISYNA1 GLDC ABCC5 SOX12 ARGLU1 PUS7 DNAH11 ABHD8 PDE7A SERPINA1 ZNF395 TRMU VDR FARP2 TCF25 USP4 MUCL1 RNF213 DHCR7 RDM1 PDIA2 ZNF625 SLC7A5 HIP1R HSD17B2 CPT1A ZNF783 TOP1MT OVOL1 OXR1 SLC25A19 GNPTAB LPGAT1 CERK A2ML1 SCLY ZCCHC6 MAST1 DDX6 PIK3R3 PRKAG2 NUP210 INSIG1 PGK1 ZC3H11A ZNF646 SLC5A6 DAZAP1 ATAD2 SDC1 APEX2 TSC22D4 TRPM2 PSME3 FADS1 CLN3 RGS12 GON4L CCDC88C SULT1A1 RPP25 MRPS17 PI4K2A TBX1 MCCC2 CPNE7 PSD3 FANCB PIF1 SLC2A1 RFC5 SFI1 WDR4

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 120 To investigate whether the metabolism-related genes within the MPS are predictive for patient outcome, all genes within the processes related to metabolism where isolated from the MPS using IPA (239 genes; MPSMETAB ; Table 4.6; A description of the genes within the MPSMETAB can be found here: https://tinyurl.com/gvknxd8.) and the same datasets tested in sections 4.2.4 (Figure 4.10) and 4.2.5 (Figure 4.12) were re-examined by stratifying patients into MPSMETAB positive and negative groups (Figure 4.14) . This revealed that the prognostic ability of the MPSMETAB was similar to that of the MPS and the MPSCCR whereby patients from prostate (Figure 4.14 Ai and Aii) and breast cancer cohorts (Figure 4.14 Bi and Bii) with a MPSMETAB positive scoring primary tumour had worse outcome than patients with a negative score. It was noted that from the two lung cancer patient cohorts examined, the MPSMETAB was able to predict faster recurrence in one dataset (Figure 4.14 Ci) but not the other (Figure 4.14 Cii). Furthermore, similar to the MPS and MPSCCR, the MPSMETAB was also significantly associated (p 0.01, OR ≥ 1.6) with recurrent disease, time to metastasis, and poor overall survival in a number of additional carcinoma types such as bladder and colon cancer (Table 4.7). Overall, the data supports that the metabolic genes within the MPS may be driving the predictive power of the signature.

To examine the metabolic reprogramming fuelled by epithelial-mesenchymal plasticity, the XF Analyzer (Seahorse Biosciences) was used to measure the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) to quantify levels of OXPHOS and glycolysis, respectively. Treatment of LNCaP-iSnail cells over 5 days (EMT1 and EMT5) resulted in a dramatic decrease in both basal and maximal rates of OXPHOS and glycolysis (Figure 4.15 A and B) inducing a state of overall metabolic quiescence (Figure 4.15C, lower left quadrant). This state was not observed following treatment of control LNCaP-iGFP cells with Dox over the same period (Figure 4.16). MErT induced by the removal of Dox treatment for 10 days returned metabolic activity to pre-EMT levels and reactivated cell cycle progression (Figure 4.15 A to C). Of note, both basal and maximal levels of OXPHOS following MErT were significantly elevated compared to the levels detected in cells before EMT induction (Figure 4.15A). This revealed tumour cells that have passed through a reversible EMT to acquire an elevated capacity for OXPHOS and have a more energetic metabolic phenotype (Figure 4.15 C).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 121 Taken together, these results suggest that 1) the association of MPS with poor clinical outcome is not solely dependent on intrinsic cell cycle/proliferative- associated gene expression, and 2) provide evidence for the potential utility of metabolism associated gene panels as novel predictors of clinical outcome in human cancer types.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 122 Ai ii

l P ro s ta te P ro s ta te

a

l

l

c

a

i a

v M E T A B M E T A B

i v m M P S + v e (n = 2 1 ) M P S + v e (n = 4 0 )

1 .0 i 1 .0

v

e v

r M E T A B M E T A B r

h M P S -v e (n = 5 8 ) M P S -v e (n = 3 2 3 )

u

u

c

s

s o

p = 0 .0 1 5 2

i

e

f

b e

o 1 3 5

2 1

r

f

f

y

- o t m th s m th s

0 .5 i 0 .5

e

l

y

i

c

t

i

b n

l 7 7

i

a

e

b

r b

r m th s

a

o

u

r b

c p = 0 .0 0 0 1

o

P

e

r 0 .0 0 .0

r P

0 5 0 1 0 0 1 5 0 0 1 0 0 2 0 0 3 0 0

M o n th s M o n th s

B i ii

B re a s t B re a s t

l

l

a a

M E T A B v

i M E T A B

v M P S + v e (n = 1 0 0 )

i M P S + v e (n = 3 1 )

1 .0 v 1 .0

r v M E T A B f M E T A B

r M P S -v e (n = 1 9 4 )

u M P S -v e (n = 6 6 )

o

u

s

s

y

t

e

i

f

l

i

e

o

r

b

f

y

-

a

t i

0 .5 s 0 .5

l

i

b

i s

1 3 2 o

b r

a 3 8

a

t

P

s b

m th s

a m th s

o t r p < 0 .0 0 0 1

e p = 0 .0 0 0 2 P

0 .0 0 .0 m 0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 0 5 0 1 0 0 1 5 0 2 0 0

M o n th s M o n th s C i ii

Lung Lung

l

l

a

a

v

v i M E T A B i M E T A B

1 .0 M P S + v e (n = 2 7 ) 1 .0 M P S + v e (n = 6 )

v

v

r

r f

M E T A B f M E T A B

u

u o

M P S -v e (n = 1 9 9 ) o M P S -v e (n = 5 2 )

s

s

y

y t

t 3 5

e

e

i

i

l

l

i

i

e

e r

r m th s

b

b

f

f

-

- a

0 .5 a 0 .5

e

e

b

b

c

c

o

o n

5 7 n

r

r

e

e

P

P

r

r

m th s 7 .5

r r

p = 0 .0 8

u

u c

p = 0 .0 0 2 7 c m th s ( n s ) e

0 .0 e 0 .0

r r 0 5 0 1 0 0 1 5 0 0 5 0 1 0 0 1 5 0 2 0 0

M o n th s M o n th s

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 123 Figure 4.14. The metabolism-related genes from the Metastatic Plasticity Signature (MPSMETAB) predict poor patient outcome in prostate, breast, and lung cancer.

The MPS was regenerated with only the genes related to metabolism (MPSMETAB; Table 4.6). Kaplan-Meyer curves showing the probability of recurrence and survival of patients based on their positive or negative MPSMETAB score. Ai. Biochemical recurrence and Aii. survival of patients with PCa (Glinsky et al. 2004; Setlur et al. 2008). Bi. The probability of survival and Bii. metastasis-free survival in patients with breast cancer (van de Vijver et al. 2002; van 't Veer et al. 2002). Ci and Cii. The probability of recurrence-free survival in patients with lung cancer (Lee, Son, et al. 2008; Okayama et al. 2012). Dotted line indicates the median time. Statistical test used was log-rank (Mantel-Cox). MPS: Metastatic Plasticity Signature

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 124 Table 4.7. Enrichment of the metabolism-related genes from the Metastatic Plasticity Signature (MPSMETAB) in multiple cancer patient cohorts.

Dataset Property Overlap (# of genes) P-Value Odds Ratio Top % vandeVijver Breast Metastatic Event Status at 5 Years 60 4.18E-15 3.9 10 vandeVijver Breast Recurrence Status at 5 Years 55 2.95E-12 3.4 10 TCGA Renal Survival Status at 3 Years 39 5.15E-11 3.8 5 Curtis Breast Survival Status at 5 Years 37 2.81E-10 3.8 5 vandeVijver Breast Recurrence Status at 3 Years 49 3.36E-09 2.9 10 Loi Breast 3 Metastatic Event Status at 5 Years 53 9.32E-09 2.7 10 Loi Breast 3 Recurrence Status at 5 Years 53 9.32E-09 2.7 10 Curtis Breast Survival Status at 5 Years 52 1.36E-08 2.7 10 vantVeer Breast Metastatic Event Status at 3 Years 41 1.77E-08 3 10 Laurent Melanoma Metastatic Event Status at 5 Years 52 2.52E-08 2.6 10 Curtis Breast Survival Status at 3 Years 50 9.68E-08 2.5 10 Kuner Lung Recurrence Status at 1 Year 50 1.71E-07 2.5 10 vandeVijver Breast Metastatic Event Status at 3 Years 29 3.07E-07 3.3 5 Laurent Melanoma Metastatic Event Status at 1 Year 49 4.29E-07 2.4 10 vandeVijver Breast Survival Status at 5 Years 44 5.70E-07 2.5 10 Loi Breast 3 Recurrence Status at 3 Years 31 7.14E-07 3 5 Esserman Breast Survival Status at 5 Years 31 7.85E-07 3 5 Esserman Breast Recurrence Status at 5 Years 48 1.18E-06 2.3 10 Loi Breast 3 Metastatic Event Status at 3 Years 47 2.50E-06 2.3 10 Lenz Lymphoma Survival Status at 3 Years 47 2.50E-06 2.3 10 Laurent Melanoma Metastatic Event Status at 3 Years 47 2.50E-06 2.3 10 Esserman Breast Recurrence Status at 5 Years 46 6.48E-06 2.2 10 TCGA Breast Survival Status at 3 Years 29 8.46E-06 2.7 5 vantVeer Breast Metastatic Event Status at 5 Years 35 8.79E-06 2.5 10 Wang Breast Recurrence Status at 3 Years 37 9.11E-06 2.4 10 Steidl Lymphoma Recurrence Status at 1 Year 45 1.32E-05 2.2 10 TCGA Breast Survival Status at 5 Years 45 1.99E-05 2.1 10 vantVeer Breast Metastatic Event Status at 1 Year 34 2.21E-05 2.4 10 Lee Bladder Survival Status at 1 Year 27 2.49E-05 2.6 5 Kuner Lung Survival Status at 1 Year 44 2.90E-05 2.1 10 Lee Lung Recurrence Status at 1 Year 27 4.61E-05 2.5 5 Schmidt Breast Metastatic Event Status at 3 Years 35 5.39E-05 2.3 10 DirectorsChallenge Lung Survival Status at 3 Years 35 5.39E-05 2.3 10 Williams Renal Recurrence Status at 5 Years 32 5.69E-05 2.4 10 Loi Breast Metastatic Event Status at 5 Years 43 6.23E-05 2 10 Loi Breast Recurrence Status at 5 Years 43 6.23E-05 2 10 Lenz Lymphoma Survival Status at 1 Year 43 6.23E-05 2 10 Murat Brain Survival Status at 1 Year 43 6.23E-05 2 10 Zhan Myeloma 2 Survival Status at 3 Years 43 6.23E-05 2 10 Pawitan Breast Survival Status at 5 Years 42 6.99E-05 2.1 10 Carrasco Myeloma Survival Status at 1 Year 26 1.18E-04 2.4 5 vanDoorn Lymphoma Survival Status at 1 Year 26 1.18E-04 2.4 5 Wang Breast Recurrence Status at 1 Year 34 1.24E-04 2.2 10 Smith Colorectal Survival Status at 3 Years 42 1.30E-04 2 10

1 Table shows the top 50 significant (p<0.01) datasets (n=281 datasets, accessible here: http://tinyurl.com/gs53j9r ).

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 125

Figure 4.15. Examining the effect of a Snail-induced EMT and MErT on metabolic reprogramming.

Ai. Measure of basal oxygen consumption rate (OCR) in LNCaP-iSnail cells treated with Dox 1 and 5 days (EMT1-5) followed by removal for 10 days (MErT10). Aii. Bar chart representation showing the basal and maximum respiration rate relative to untreated LNCaP-iSnail cells (No Dox). One-way ANOVA; p value: ****<0.0001. Error bars indicate SEM of triplicates. FCCP: carbonyl cyanide 4- (trifluoromethoxy) phenylhydrazone. Bi. Measure of extracellular acidification rate (ECAR) in LNCaP-iSnail cells treated with Dox 1 and 5 days (EMT1-5) followed by removal for 10 days (MErT10). Bii. Bar chart representation showing the basal and maximum respiration rate relative to untreated LNCaP-iSnail cells (No Dox). One- way ANOVA; p value: ****<0.0001. Error bars indicate SEM of triplicates. 2-DG: 2-deoxyglucose. C. Energy map showing the metabolic phenotype of untreated LNCaP-iSnail cells, and cells treated with Dox for 1 ad 5 days followed by removal for 10 days.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 126 Ai B i

R o te n o n e & )

) O lig o m y c in F C C P

A c tim y c in A s

l

s l l G lu c o s e 2 -D G l O lig o m y c in

5 0 0 e 1 0 0 e

c / / N o D o x

c

5 5 8 0 / /

4 0 0 0 E M T 1

0

1 1 / / / / E M T 5

n 6 0

i n

i 3 0 0 / / M E rT 1 0

m

m /

/ 4 0

l H

o 2 0 0

p 2 0

m

m

p

(

(

1 0 0 0

R

R

A C

0 C -2 0 O 0 2 0 4 0 6 0 8 0 1 0 0 E 0 2 0 4 0 6 0 8 0 1 0 0 M in u te s M in u te s ii ii

B a s a l M a xim u m B a s a l M a xim u m

1 5 0 1 5 0

x

x

o

o D

1 0 0 D 1 0 0

o

o

N

N

f

f

o

o

5 0 5 0

% %

0 0

Figure 4.16. Measure of Basal oxygen consumption rate (OCR), and extracellular acidification rate (ECAR) in LNCaP-iGFP cells treated with Dox for up to 5 days followed by Dox removal for 10 days.

Ai. Measure of basal oxygen consumption rate (OCR) in LNCaP-iGFP cells treated with Dox 1 and 5 days (EMT1-5) followed by removal for 10 days (MErT10). Aii. Bar chart representation showing the basal and maximum respiration rate relative to untreated LNCaP-iSnail cells (No Dox). One-way ANOVA; p value: ****<0.0001. Error bars indicate SEM of triplicates. FCCP: carbonyl cyanide 4- (trifluoromethoxy) phenylhydrazone. Bi. Measure of extracellular acidification rate (ECAR) in LNCaP-iGFP cells treated with Dox 1 and 5 days (EMT1-5) followed by removal for 10 days (MErT10). Bii. Bar chart representation showing the basal and maximum respiration rate relative to untreated LNCaP-iGFP cells (No Dox). One- way ANOVA; p value: ****<0.0001. Error bars indicate SEM of triplicates. 2-DG: 2-deoxyglucose.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 127 4.3 Discussion

The complete cycle of EMT and MErT (collectively termed epithelial-mesenchymal plasticity; EMP) represents a dynamic and complex set of phenotypic transitions that contribute and are perhaps paramount, to metastatic dissemination and colonisation. Despite a number of elegant studies in support of this concept (Tsai et al. 2012; Ocana et al. 2012), recent reports using lineage tracing strategies in spontaneous metastasis models of breast and pancreatic ductal carcinoma (Fischer et al. 2015; Zheng et al. 2015) have reinvigorated the debate over the functional role of EMT and MErT in cancer metastasis. This discussion is aided by the lack of evidence for these phenotypic transitions in clinical patient specimens. In part, this stems from technical challenges in distinguishing cancer cells that have undergone EMT and/or MErT from surrounding tumoural and non-tumour cell types as they can share similar marker expression (e.g. vimentin is present in EMT cells as well as in fibroblasts in the tumour stroma). Our limited understanding of the spatiotemporal regulation of the complex gene expression events underpinning these transitions has resulted in the limited prognostic utility of EMT-derived signatures in clinical specimens (Taube et al. 2010; Chikaishi, Uramoto and Tanaka 2011; Tan et al. 2014). The present study aimed to address a number of these issues, with a focus on the identification and clinical evaluation of the transcriptional dynamics involved in an experimentally reversible EMT.

Using a prostate cancer model of reversible Snail-induced EMT, this study revealed the sophisticated and orchestrated temporal transcriptional events that occur during the reversible transition of cancer cells between epithelial and mesenchymal states (Figure 4.1 and Figure 4.4). Clinical analysis of prostate cancer patient cohorts revealed that the transcriptional signature of MErT was enriched in metastatic specimens (Figure 4.4). It has been speculated that for dormant tumour cells held in an EMT state to regain their proliferative capacity, they need to reacquire their epithelial phenotypic characteristics (Thiery 2002; Tsai and Yang 2013). Previous studies have reported EMT-inducers to reduce cell division in a variety of tumour models (Vega et al. 2004; Hugo et al. 2013; Bierie and Moses 2006), whereby removal of these inducers, such as Twist (Tsai et al. 2012) and Prrx1 (Ocana et al. 2012), promoted metastatic growth. The study described herein provides direct

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 128 evidence that EMT-induced invasive cells enter a state of cell cycle arrest and metabolic dormancy, which is broken by MErT (Figure 4.1 and Figure 4.15).

Previously, the gene expression events associated with EMT reversion (i.e., MErT) have been regarded as the mirror image of those activated or repressed during EMT. Only recently has there been a growing realization that the MErT program may have unique features, or impart cells with stable phenotypic traits after passage through a reversible EMT (Schmidt et al. 2015). This study provides evidence that while the vast majority of the transcriptional alterations evoked by EMT are reversed during MErT, cells may retain a “transcriptional footprint” as seen in their persistent expression of transcript changes (Figure 4.8). These persisting transcripts have the potential to serve as a sentinel signature of cancer cells that have experienced a reversible EMT; however, further studies will be necessary to determine the resilience of these alterations over longer durations of MErT and/or during multiple EMT/MErT cycles. Previously the inducible expression of Snail in human mammary epithelial cells was reported to bind transiently to its target promoters, triggering both transient and long-lasting chromatin alterations contributing to EMT (Javaid et al. 2013). Since EMT can evoke an epigenetic switch in normal and malignant cell types (Tam and Weinberg 2013), the integration of studies examining the dynamics of epigenetic alterations as cells cycle through EMT/MErT will likely provide novel insights into the persistent transcriptional alterations identified herein.

The results provide prototypical evidence supporting the role of EMP in prostate cancer. The general MErT-associated epithelial gene expression profile was found to be significantly different in its enrichment between localised prostate cancer and metastatic castrate-resistant prostate cancer (mCRPC) samples (Figure 4.4). Similarly, the expression of the EMT persistent and MET unique genes were highly enriched in mCRPC compared to localised prostate cancer (Figure 4.8). Furthermore, the expression of the EMT persistent and MET unique genes highly correlated more so in the mCRPC samples compared to the localised samples. Importantly, this correlation was not present in benign prostate tissue samples (Figure 4.8). Cumulatively, the results support that the spatiotemporal regulation of EMT is activated in a cancer context and facilitates cancer progression.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 129 Despite the observation for the occurrence of EMT at the invasive front of multiple carcinoma types (Brabletz et al. 2001; Trimboli et al. 2008; Sethi et al. 2010), there are opposing views as to whether the EMT status of localised tumours is able to predict the patients outcome (Chikaishi, Uramoto and Tanaka 2011; Tan et al. 2014; Loboda et al. 2011). For instance, detection of an EMT signature in lung cancer patients was unable to predict post-operative recurrence or disease free survival (Chikaishi, Uramoto and Tanaka 2011). In contrast, presence of an EMT signature in colon cancer patients was associated with recurrence, decreased time to metastasis and survival (Loboda et al. 2011). Studies by Tan et al., (Tan et al. 2014) support that enrichment of an EMT signature in ovarian and colorectal cancer patients is associated with poor prognosis, however, this was not observed in breast cancer patients.

Herein, it was identified that when a plasticity-derived metastasis signature (MPS) was expressed in the primary tumours of the prostate, breast and lung cancer patients, it was associated with poor prognosis (Figure 4.10). Furthermore, it was demonstrated that this association is not solely dependent on intrinsic cell cycle/proliferation-related gene expression, a previously validated predictor of poor clinical outcome (Cuzick et al. 2011; Cooperberg et al. 2013), as a MPS devoid of these genes (MPSCCR) independently retained prognostic value similar to that of the MPS (Figure 4.12). Closer examination of the GO-BPs enriched in the MPSCCR revealed enrichment in metabolism-related gene expression (Figure 4.13). Indeed, when the metabolism genes were examined in isolation for prognostic abilities, they were able to predict poor patient outcome in almost all datasets examined using the MPS Figure 4.14. Further functional testing of the effect of EMT on cell metabolism revealed that LNCaP cells in an EMT state entered a metabolically quiescent phenotype which was rescued with MErT Figure 4.15 and Figure 4.16). Interestingly, cancer cells acquired a more energetic metabolic state following MErT than what was observed prior to the EMT (Figure 4.15). Together, these results imply that biomarker gene sets altered during the metabolic reprogramming of cancer cells may have clinical utility as novel predictors of patient outcome across multiple human cancer types. Moreover, these biomarker panels may aid in the identification of patients that may benefit from intervention with metabolic pathway- targeted drugs (Vander Heiden 2011) to slow disease progression and provide

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 130 insights into new therapeutic strategies. However, it must be noted that metabolic processes are highly complex and reliance on transcriptomic profiles may not be the most effective way to monitor. This is because transcriptomic profiles may not be reflective of the activity of enzymatic processes. Therefore, it is advised that metabolic processes identified via transcriptomic approaches be used as a guide and a functional product of these metabolic pathways (i.e. an enzyme or protein that can be found in the blood stream) should be used as a marker for the activity of these processes.

It is recognised that the MPS is a large signature which a characteristic that is not ideal for use in diagnostics. To address this, collaboration with GenomeDx was initiated to not only examine the MPS in their own exclusive cancer patient cohorts, but to test whether the MPS can be refined into a smaller gene list without loss of its predictive qualities. While the MPS will undoubtedly require further refinement and testing in extended patient cohorts, the data herein suggest that the cycling of epithelial and mesenchymal states via EMT/MErT may also occur at the primary tumour site. This is in contrast to the prevalent view that reversion of EMT only occurs for cancer cells at the primary site to disseminate to distant tissue sites. The enrichment of the MPS was present in a subset of primary tumours within each patient cohort, and these subgroups of patients had a shorter time to recurrence and reduced overall survival (Figure 4.10). Moreover, this was not unique to PCa patient cohorts, but broadly applicable across a number solid tumour types, including breast and lung carcinomas (Figure 4.10). Extensive studies will be required to confirm the presence and prevalence of EMT/MErT cycling within primary tumours. However, it is speculated that as EMT-induced invasive cells move throughout the tumour microenvironment and are exposed to varying spatial and paracrine cues from surrounding cells and tissue types, they may oscillate between epithelial and mesenchymal states more than previously appreciated.

In summary, the results of this study lay the foundation for understanding the transcriptional landscape of epithelial-mesenchymal plasticity that occurs during cancer progression. In addition to providing evidence for a transcriptional signature of this plasticity in human cancer metastasis, this study also identifies the potential of cancer cells cycling between epithelial and mesenchymal states within the primary tumour, whereby this plasticity significantly predicts poor patient outcomes.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 131 Additionally, the reawakening of EMT-induced dormant-like cells via MErT will need to be carefully considered for the application of potential anti-EMT therapeutic strategies. In particular, these treatments may have deleterious effects on metastatic tumour growth in patients with detectable circulating tumour cells or dormant occult metastases. Nevertheless, there is substantial evidence that EMT confers chemoresistance, whereby the inhibition of EMT results in resensitization of cells and suppression of metastasis (Zheng et al. 2015; Fischer et al. 2015). This study proposes that the full spectrum of epithelial-mesenchymal plasticity needs to be considered for the application of new anti-metastatic therapies aimed at inhibiting EMT. A more efficient approach may be to combine these therapies with agents targeting pathways arising from MErT that increases tumour cell survival and treatment resistance.

Chapter 4: Revealing the transcriptional landscape of epithelial-mesenchymal plasticity in cancer progression 132

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 133 5.1 Introduction Cancer cell plasticity via the EMT program has been implicated in facilitating cancer progression to the metastatic stage across multiple types of carcinomas. In prostate cancer, studies have associated the EMT phenotype with the emergence of lethal metastatic disease and development of castrate-resistant prostate cancer (CRPC). However, the investigation of EMT in vivo often examines cells in end- point epithelial or mesenchymal states, failing to capture and investigate the biology of cancer cells dynamically transitioning between these states. Furthermore, in vivo examination of the EMT program, even when using reversible models like the models generated herein, typically only provide a snapshot in time. However, a reversible in vivo model of EMT would allow for the examination of multiple time points, aiding further to detail the role of cancer cell plasticity in cancer biology and metastasis.

There are currently no in vivo models able to control the induction and reversion of the EMT program in human prostate cancer cells. Therefore, the inducible EMT cell models described previously (Chapters 3 and 4) were examined under in vivo conditions to assess whether their established in vitro performance could translate to altered tumour behaviour in an in vivo setting. This chapter outlines the results from this first-in-field attempt in utilising these reversible human prostate cancer EMT cell models to simulate EMT and MErT in cells implanted orthotopically into the prostate to mimic the clinical scenario.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 134 5.2 Results

5.2.1 Generation and characterisation of RFP/LUC variants of the LNCaP- iSnail, LNCaP-iSlug, and LNCaP-iGFP cells. To be able to detect the reversible EMT models in vivo, the LNCaP-iSnail, LNCaP-iSlug and LNCaP-iGFP cell lines were transduced with the pMig-Luc2- DsRed plasmid (termed from here on LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP-iGFPRFP/LUC) as detailed in Chapter 2, section 2.15. To obtain a more homogeneously bright population of RFP/LUC expressing cells, the cell lines were sorted and the top 27-38% of RFP expressing cells collected (Figure 5.1).

Next, the luciferase emission was examined using the IVIS Spectrum in vivo imaging system (IVIS). All three cell lines were plated in a 96 well plate, in a dilution series ranging from 40,000 cells to 19 cells and subsequently treated with 150 μg/mL of D-Luciferin. Following a 5 minute incubation period to allow for the D-Luciferin to be metabolised, the cell lines were measured for emitted bioluminescence using the IVIS. All three cell lines emitted bioluminescence and the IVIS was capable of detecting a minimum of 39-78 cells in this format (Figure 5.2).

Lastly, the induction of EMT following Dox treatment was assessed to ensure the performance of the models prior to using for in vivo experimentation. LNCaP- iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells treated with Dox for 5 days acquired a similar mesenchymal phenotype as previously seen in chapter 3 section 3.2.4 (Figure 5.3). By contrast, LNCaP-iGFPRFP/LUC cells retained their epithelial-like morphology and expressed GFP following treatment with Dox (Figure 5.3 and Figure 5.4). Removal of Dox from the LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells for 7 days saw their return to a pre-EMT cell phenotype (Figure 5.3), and the cessation of GFP expression in the LNCaP-iGFPRFP/LUC cells (Figure 5.4).

The protein expression of Slug, E-cadherin, and vimentin was assessed using IHC. The induction of Slug in LNCaP-iSlugRFP/LUC cells treated with Dox for 5 days was heterogeneous as characterised by the variability in staining across cell nuclei. Slug protein was only present in LNCaP-iSlugRFP/LUC cells treated with Dox and absent in the rest of the experimental groups, including following the removal of Dox (Figure 5.5).

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 135

A i L N C a P -iS n a ilR F P /L U C A ii

t

s

a

r

t

n

o

C

e

s

a

h

P

P

F R

B i L N C a P -iS lu g R F P /L U C B ii

t

s

a

r

t

n

o

C

e

s

a

h

P

P

F R

C ii L N C a P -iG F P R F P /L U C C ii

t

s

a

r

t

n

o

C

e

s

a

h

P

P

F R

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 136 Figure 5.1. The inducible cell models were sorted for the top 27-38% RFP- expressing cells.

Histogram of the expression of RFP from Ai. LNCaP-iSnailRFP/LUC,(Bi. LNCaP- iSlugRFP/LUC, and Ci. LNCaP-iGFPRFP/LUC cells before sorting. Y-axis shows the cell count and the x-axis shows the RFP intensity. The R4 gate measured RFP negative cells and the R5 gate was used to collect high expressing RFP cells. Phase contrast (top) and epifluorescent (bottom) images showing the RFP expression of Aii. LNCaP-iSnailRFP/LUC, Bii. LNCaP-iSlugRFP/LUC, and Cii. LNCaP-iGFPRFP/LUC cells post sorting. Scale bars indicate 50μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 137

N u m b e r o f c e lls

0

0

0

0

0

0

0

0

0

0

0

0

0

0

5

5

2

6

,

,

,

0

5

2

8

9

9

2

1

5

0

0

0

,

,

,

4

2

1

5

2

1

6

3

1

7

3

1

C

U

L

/

P

P

a

F

R

C

g

N

u

l

L

S

i

-

C

U

L

/

P

P

F

a

R

C

l

i

N

a

L

n

S

i

-

C

U

L

/

P

P

a

F

R

C

P

N

F

L

G

i -

Figure 5.2. Luciferase expression as detected by the IVIS of LNCaP- iSlugRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iGFPRFP/LUC cells at indicated cell numbers following incubation with D-luciferin (150 μg/mL) for 5 minutes.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 138

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.3. Cell morphology of LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox.

Representative images showing the cell morphology of untreated (vehicle) LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells, treated with Dox for 5 days (EMT), and followed by removal for 7 days (MET). Cells were fixed in formalin and stained with H&E prior to imaging. Scale bars indicate 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 139

V e h ic le E M T M E T

t

s

a

r

t

n

o

c

e

s

a

h

P

C

U

L

/

P

F

e

R

c

n

P

e

F

)

c

s

P

G

i

e

F

-

r

R

o

P

(

u

l

a

f

i

C

p

N

E

L

e

c

n

e

)

c

s

P

e

F

r

G

o

(

u

l

f

i

p E

Figure 5.4. Cell morphology, RFP and GFP expression of LNCaP- iGFPRFP/LUC cells following transient treatment with Dox.

Phase contrast (top row) and epifluorescence of RFP (middle row) and GFP (bottom row) of untreated LNCaP-iGFPRFP/LUC cells (vehicle) and cells treated with Dox for 5 days (EMT) followed by removal for 7 days (MET). Scale bars indicate 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 140

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.5. Expression of Slug in LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox.

Expression of Slug protein (brown) in untreated (vehicle) LNCaP-iGFPRFP/LUC , LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells, treated with Dox for 5 days (EMT) followed by removal for 7 days (MET). Cells counterstained with haematoxylin. Scale bars indicate 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 141 The expression of E-cadherin protein decreased in the LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells treated with Dox, however, some cells still retained E- cadherin at the cell membrane and this was highlighted by the strong staining found at the cell junctions (Figure 5.6). Untreated and removal of Dox resulted in the reinstatement of E-cadherin at the cell junctions (Figure 5.6). Examination of vimentin expression revealed that the LNCaP cell line contains <1% of cells expressing vimentin (Figure 5.7). Induction of Snail or Slug by the addition of Dox to the LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells increased the number of vimentin positive cells. However, it was clear that removal of Dox for 5 days (i.e. MET) was not enough to restore vimentin protein expression to pre-induction levels (Figure 5.7). This was in line with previous analyses showing that vimentin mRNA required up to 20 days of MET to return to baseline levels (Chapter 4).

In conclusion, all three models expressed sufficient amounts of RFP and luciferase for in vivo use, and the LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells continued to recapitulate a phenotypic EMT and MErT consistent with that described in previous chapters.

5.2.2 Experimental design of in vivo experiment. The objective of the in vivo study was to recapitulate EMT and MErT in a prostate cancer specific setting to assess the effects of cell plasticity on prostate cancer progression to metastasis. To better simulate prostate cancer as it presents in clinical patients, the inducible EMT cell models (LNCaP-iGFPRFP/LUC, LNCaP- iSnailRFP/LUC, LNCaP-iSlugRFP/LUC) were injected orthotopically into the posterior gland of the murine prostate of 4-5 week old male NOD/SCID mice at a cell dose of either 2x105 or 1x106 cells. Following successful tumour establishment, defined as 2- 3 consecutive weeks of positive luciferase emission, mice were sorted into their experimental groups (Figure 5.8). The vehicle group (4 mice per cell dose) received drinking water supplemented with 50 g/L sucrose (as above) for the duration of the experiment (19 weeks). The EMT group consisted of 4 mice per cell dose on water supplemented with sucrose and 2 g/L Dox for the duration of the experiment (19 weeks). The MET group (4 mice per cell dose) received water supplemented with sucrose and Dox for 4 weeks, followed by a return to water supplemented with sucrose for the remainder of the experiment (15 weeks). When the tumours reached

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 142

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.6. Expression of E-cadherin in LNCaP-iGFPRFP/LUC , LNCaP- iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox.

Expression of E-cadherin protein (brown) in untreated (vehicle) LNCaP- iGFPRFP/LUC , LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells, treated with Dox for 5 days (EMT) followed by removal for 7 days (MET). Cells counterstained with haematoxylin. Scale bars indicate 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 143

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.7. Expression of vimentin in LNCaP-iGFPRFP/LUC , LNCaP- iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC and cells following transient treatment with Dox.

Expression of vimentin protein (brown) in untreated (vehicle) LNCaP- iGFPRFP/LUC , LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells, treated with Dox for 5 days (EMT) followed by removal for 7 days (MET). Cells counterstained with haematoxylin. Scale bars indicate 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 144

Figure 5.8. The experimental design.

Four to five week old male NOD/SCID mice received an intraprostatic injection of the EMT models, and tumour establishment monitored via bioluminescent imaging. Following tumour establishment (2-3 weeks), mice were placed in the indicated experimental groups for up to 19 weeks. Once tumours reached approximately 1 cm3, mice were sacrificed and organs collected for ex-vivo examination and subsequent histological analysis.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 145 the ethical limit of 1000 mm3, mice were sacrificed and their tumours and organs assessed for metastasis via ex vivo luciferase imaging and subsequently IHC staining of tissue sections.

5.2.3 Optimising the cell concentration for the intraprostatic injections. One aspect of the in vivo pilot study was to determine the optimal cell concentration for optimal tumour take intraprostatic injection of LNCaP models. All three cell lines (LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP-iGFPRFP/LUC) were examined at either 2x105 or 1x106 cells/10 μL PBS injection. Following injection into the posterior gland of the prostate, mice were imaged weekly for bioluminescent signal to monitor tumour establishment. Tumour establishment was confirmed by a detectable bioluminescent signal of more than 1x106 radiance (photons/second) for at least 2 consecutive weeks. It was observed that across all three cell lines there was higher tumour establishment in the mice that received 1x106 cells (27/30 mice; 90%) compared to mice that received 2x105 cells (23/35; 66%) (Table 5.1).

Table 5.1. Tumour take of LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP-iGFPRFP/LUC cells following intraprostatic injection.

Mice with established Cell line Cell dose tumours / total mice (%) LNCaP-iGFPRFP/LUC 1x106 11/11 (100%) 2x105 7/11 (64%)

LNCaP-iSnailRFP/LUC 1x106 6/9 (67%) 2x105 7/12 (58%)

LNCaP-iSlugRFP/LUC 1x106 10/10 (100%) 2x105 9/12 (75%)

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 146 5.2.4 Examination of the primary tumour growth across groups. Once tumours were established, mice were randomly allocated into the experimental groups outlined above. Tumour growth was monitored weekly by bioluminescent imaging (Figure 5.9). It was determined that a luciferase signal of 1x1010 radiance, on average, corresponded to a tumour size of approximately 1000 mm3, the ethical limit for this study. Although the sample number within each group was too small to test for statistically significant differences, weekly bioluminescent imaging of the tumours across the experimental groups and within each cell dose revealed no distinct differences in tumour growth pattern. Furthermore, unexpected death or early sacrifice of mice within each group resulted in graph anomalies. It was also observed that on 3 occasions, imaged mice would present with an uncharacteristically low luciferase emission, causing severe dips in the graphs. This was attributed to either incorrect intraperitoneal injection of the luciferin or that the mice were imaged before or after the peak luciferase emission timeframe. Tumours were harvested once they grew to a palpable size of approximately 1000 mm3. Across all cell lines, treatments, and cell doses, the tumours were encapsulated and had a dark red colour, indicating the highly vascularised and haemorrhagic nature of LNCaP tumours. Tumours were fixed, processed and embedded in paraffin blocks before sectioning with a microtome. The architecture of the tumour sections, as well as select markers, was examined via haematoxylin and eosin (H&E) staining and IHC.

The primary tumour’s architecture was initially assessed by H&E staining of tumour cross-sections (Figure 5.11). It is noted that the following analyses were performed on a single section per tumour and may not represent the tumours as a whole. Examination of the H&E staining revealed the multifocal nature of the tumours across all cell lines, treatments and cell doses. The tumour foci were characterised by having stromal septae separating each focus (Figure 5.11). The H&E staining also revealed areas of necrosis (marked by pink areas with or without the presence of red blood cells) and the proportion of necrotic area was on average ~25% of the tumour area observed across all cell lines, treatments and cell doses (Figure 5.12).

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 147 1 x 1 0 6 c e lls 2 x 1 0 5 c e lls

1 0 1 1 1 0 1 1 C 1 0 1 0

U 1 0 1 0

L

/

P

e

e

F 9 9

c

c

1 0 c 1 0

R

n

n

a

a

P

i

i

d

d

F 8 8

a

a

1 0 a 1 0

G

R

R

V e h ic le ( 3 ) R V e h ic le ( 3 )

i -

P 1 0 7 E M T ( 4 ) 1 0 7 E M T ( 2 ) a

C M E T ( 3 ) M E T ( 2 )

N 1 0 6 1 0 6 L 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 W e e k s W e e k s

1 0 1 1 1 0 1 1 C

U 1 0 1 0 1 0 1 0

L

/

P

F

e e

e e e

e 9 9

c c

c c c

c 1 0 1 0

R

n n

n n n

n

l

i

a a

a a a

a

i i

i i i

i

a

d d

d d d d

8 8

a a

a a a a

n 1 0 1 0

R R

R R R

R V e h ic le ( 1 ) V e h ic le ( 2 )

S

i - 7 E M T ( 2 ) 7 E M T ( 2 ) P 1 0 1 0

a M E T ( 3 ) M E T ( 3 ) C 6 6

N 1 0 1 0 L 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 W e e k s W e e k s

1 0 1 1 1 0 1 1

V e h ic le ( 2 ) C

U 1 0 1 0 1 0 1 0 E M T ( 3 )

L /

P M E T ( 3 )

e e e

e e e

F 9 9

c c c

c c

1 0 c 1 0

R

n n n

n n n

g

a a a

a a a

i i i

i i i

d d d

d d d u

l 8 8

a a a

a a

1 0 a 1 0

S

R R R

R R

V e h ic le ( 4 ) R

i -

P E M T ( 3 )

1 0 7 1 0 7 a

C M E T ( 3 )

N 1 0 6 1 0 6 L 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 2 W e e k s W e e k s

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 148 Figure 5.9. Luciferase emission over time of the (A) LNCaP-iGFPRFP/LUC, (B) LNCaP-iSnailRFP/LUC, and (C) LNCaP-iSlugRFP/LUC primary tumours in mice that received either 2x105 or 1x106 cells intraprostatic injections.

Y-axis shows luciferase emission in photons/second. X-axis shows weeks from time of intraprostatic injection. Mice in the vehicle group received water supplemented with sucrose for up to 20 weeks. Mice in the EMT and MET groups were treated with water supplemented with Dox and sucrose from week 2 onwards. Mice in the MET group were removed from Dox treatment on week 6 and put back on water supplemented with sucrose for the remaining time. Number in brackets is the number of mice per group. Error bars indicate standard deviation when n ≥ 2 mice.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 149 V e h ic le E M T M E T

P

T S V

C

U

L /

P T F N P R T T

P N

F

G

i -

P N a N

C T

N T T L T

C T T T

U

L /

P N F N

R N

l i

a T

n

S i - T N

P N T a

C T

N T L T P N

S V

C T

U L / P

P T

F R N

g P u

l T N

S

i

-

P

a

C N L B N T P S V

Figure 5.10. H&E staining of tumour xenographs.

Representative H&E stained sections showing the tumour architecture of the LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC at time of termination (~19 weeks). Time of termination was approximately 19 weeks post intraprostatic injections. Tumours are from mice that received intraprostatic injections of 1 x 106 cells. T: tumour, N: necrotic area, SV: seminal vesicles, B: bladder, P: mouse prostate tissue. Scale bar indicates 2 mm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 150 T N

S

N

Figure 5.11. Example of the stromal septae (S) and necrosis (N) in the tumours (T) stained with haematoxylin and eosin.

This tumour is included in Figure 5.01 (2nd row; first panel; LNCaP- iSnailRFP/LUC; MET). Scale bar indicates 2 mm in the left panel and 200 μm in the right panel.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 151 5.2.5 Examination of inducible protein expression following transient Dox treatment in the LNCaP-iSnailRFP/LUC, LNCaP-iSlugRFP/LUC, and LNCaP- iGFPRFP/LUC tumours. The expression of the inducible proteins of each cell line xenograft (i.e. Snail, Slug, or GFP) following transient treatment with Dox was assessed via IHC. The staining intensity was categorised as: negative, weak staining, moderate staining, strong staining, and very strong staining (Figure 5.12).

LNCaP-iGFPRFP/LUC tumours in the vehicle control group were predominantly GFP negative with approximately 10% of cells expressing weak levels of GFP (Figure 13, row one, first panel). The weak GFP staining was anticipated due to the minimal leaking that was previously observed (Chapter 3, section 3.2.2). All LNCaP-iGFPRFP/LUC tumours in the EMT group presented with 100% of tumour cells expressing GFP (5 tumours) (Figure 5.13, row one, second panel, and

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 152

Figure 5.14). The intensity of GFP varied from weak to very strong, and while the distribution of the staining intensity varied between tumours, the average staining intensity was moderate to strong. The variability observed in the staining was in line with previous results showing that Dox treated LNCaP-iGFP cells expressed heterogeneous levels of GFP (Chapter 3, section 3.2.2). Tumours from mice which Dox was removed, the MET group, presented with a similar staining pattern to those from the vehicle control group (Figure 13, row one, third panel). Cumulatively, these results indicate that the Dox system was functioning as expected whereby strong GFP staining was only present in tumours treated with Dox, and the levels of GFP returned to pre-treatment levels following the withdrawal of Dox.

Next, the expression of Snail was examined in the LNCaP-iSnailRFP/LUC tumours. Tumours from the vehicle control group showed no staining for Snail. However, isolated cells were detected that were weakly positive (Figure 5.13, row two, first panel). Tumours from the EMT group presented with Snail-positive cells that were either arranged in tight groups, scattered, or isolated (Figure 5.13, row two, second panel). The staining intensity of the Snail-positive cells ranged from weak to very strong and some areas of the tumour presented with no staining. Tumours from the MET group had a staining pattern comparable to the vehicle control group.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 153 N e g a tiv e W e a k M o d e ra te

S tro n g V e ry S tro n g

Figure 5.12. The scale utilised for scoring the staining intensity.

The representative images for negative and weak staining were acquired from mice with LNCaP-iGFPRFP/LUC tumours from the vehicle group. Images for the moderate, strong, and very strong, were obtained from mice with LNCaP- iGFPRFP/LUC tumours from the EMT group.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 154

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

P

F

F

G

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

l

i

i

a

a

n

n

S

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

g

u

u

l

l

S

S

i

-

P

a

C

N L

Figure 5.13. Expression of GFP, Snail, or Slug in tumour xenographs.

Expression of GFP, Snail, and Slug (brown) in LNCaP-iGFPRFP/LUC , LNCaP- iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours when in the absence of Dox (vehicle), treated with Dox for 19 weeks (EMT) and transient Dox treatment (MET; 4 weeks Dox, Dox removed ~15 weeks). Tissue sections counterstained with haematoxylin. Scale bars indicate 100 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 155

T S V

T T T

S V N T T P P

T T T T T N N N

T T P S V T T

T

T N

T P

Figure 5.14. Expression of GFP (brown) in LNCaP-iGFPRFP/LUC tumour cross-sections from tumours treated with Dox for 19 weeks (EMT).

Tissue sections counterstained with haematoxylin. T: tumour; N: necrotic area; P: mouse prostate tissue; SV: seminal vesicles. Scale bar indicates 1 mm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 156 Examination of the LNCaP-iSlugRFP/LUC tumours revealed similar staining patterns as was observed in the LNCaP-iSnailRFP/LUC tumours (Figure 5.13, row three). Overall, the data supports that the inducible markers of each cell line became activated in the presence of Dox at 19 weeks (EMT group), and their expression returned to pre-treatment levels in the MET group whereby mice were treated with Dox for 4 weeks followed by removal for up to 15 weeks.

5.2.6 Examination of EMT-related marker expression at the primary site. Next the expression of EMT-related markers E-cadherin and vimentin were visualised via IHC staining of serial tumour sections across the three cell lines. Alongside, serial sections were stained for GFP, and Slug, however, staining for Snail protein was performed on earlier sections by collaborators in Spain (Dr Antonio Garcias de Herreros and Dr Raul Pina, Institut Hospital del Mar d’Investigacions Mèdiques (IMIM)). To differentiate between human and murine cells, serial tumour sections were also stained with an antibody specific to human Ku70 (Figure 5.15).

Although expression of GFP, Snail, and Slug was already examined in Figure 5.14, they were examined again as part of staining other EMT-related markers across serial sections (Figures 5.15 to 5.21). This was to address the heterogeneity of the tumours and allowed to more directly compare the expression of the EMT-related markers in similar areas of the tumours assessed. As expected, GFP protein was present only in the LNCaP-iGFPRFP/LUC tumours from the EMT group and absent in the tumours from the vehicle and MET groups. GFP staining was also absent in the LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC tumours across all groups (Figure 5.16). Similarly, the expression of Snail or Slug was most prominent in the respective LNCaP-iSnailRFP/LUC (Figure 5.17) or LNCaP-iSlugRFP/LUC (Figure 5.18) tumours from the EMT group, and was absent from the LNCaP-iGFPRFP/LUC tumours.

Interestingly, examination of vimentin expression using a human specific vimentin antibody revealed the LNCaP cell line to express higher levels of vimentin in vivo when compared to that previously seen in vitro. This was evident in the LNCaP-iGFPRFP/LUC tumours from all experimental groups as well as the LNCaP-

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 157

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.15. Expression of Ku70 in the tumour xenographs.

Expression of human Ku70 (brown) in LNCaP-iGFPRFP/LUC, LNCaP- iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 158

Figure 5.16. Expressionof GFP in the tumour xenographs.

Expression of GFP (brown) in LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 159

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.17. Expression of Snail in the tumour xenographs.

Expression of Snail (brown) in LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC , and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 160

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.18. Expression of Slug in the tumour xenographs.

Expression of Slug (brown) in LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 161

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.19. Expression of vimentin in the tumour xenographs.

Expression of vimentin (brown) in LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 162 iSnailRFP/LUC and LNCaP-iSlugRFP/LUC tumours in the vehicle control groups (

Figure 5.19). Interestingly, the levels of vimentin expression did not increase in LNCaP-iSnailRFP/LUC or LNCaP-iSlugRFP/LUC tumours from the EMT group (

Figure 5.19, second panel of rows 2 and 3) when compared to tumours from the vehicle groups. The images shown in Figures 16 to 22 are serial sections of the representative tumours, and it is acknowledged that the image shown for the untreated LNCaP-iSlugRFP/LUC tumours appears to have less vimentin expression than the ones from the EMT or MET groups, however, the expression of vimentin within the tumours was heterogeneous with that particular area expressing less than other regions of the tumour.

The levels of E-cadherin appeared to remain unaltered across all cell models and experimental groups (Figure 5.20). Lastly, the percentage of Ki67 expressing cells to cells expressing human Ku70 was assessed to examine potential differences in cell proliferation as seen previously with EMT. This analysis was performed on a randomised selection of tumours across the cell lines and treatments (2 tumours per cell line per treatment). The expression of Ku70 and Ki67 was measured using the Visiopharm software to detect positive nuclei across the entire tumour cross section. The ratio of Ki67 to Ku70 nuclei revealed no distinct differences in Ki67 positive cells across the cells lines or treatments groups (Figure 5.21; Table 5.2) These results will be further discussed in section 5.3.

Table 5.2. Percentage of Ki67 positive tumour cells to Ku70 positive tumour cells within the tumours.

Cell line Vehicle* EMT* MET* LNCaP-iGFPRFP/LUC 38% 44% 47% LNCaP-iSnailRFP/LUC 29% 50% 34% LNCaP-iSlugRFP/LUC 63% 54% 42%

* Untreated tumours: Vehicle; Tumours treated with Dox for 19 weeks: EMT; Tumours treated with Dox for 4 weeks followed by removal for up to 15 weeks: MET.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 163

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.20. Expression of E-cadherin in the tumour xenographs.

Expression of E-cadherin (brown) in LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 164

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.21. Expression of Ki67 in the tumour xenographs.

Expression of Ki67 (brown) in LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from untreated mice (vehicle), treated with Dox for 19 weeks (EMT) and mice treated with Dox for 4 weeks followed by removal for up to 15 weeks (MET). Tissue sections were counterstained with haematoxylin. Scale bars indicate 200 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 165 Overall, while the expression of vimentin and E-cadherin remained unaltered, the inducible Dox system functioned as expected on the induction of the GFP, Snail, and Slug proteins in the LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP- iSlugRFP/LUC tumours.

5.2.7 Examination of the tumour stroma. Study of the tumour stroma revealed human cells within circulatory vessels. The human cells were identified by their positive Ku70 expression and the vessels were recognised from their organised endothelial cells and circular shape (Figure 5.22). It was observed that the occurrence of human cells within murine blood vessels in the stroma was more frequent in the mice that received the 1 x 106 cell dose than mice that had received the 2 x 105 cell dose (Table 5.3). However, the sample number across groups was too small, and the analysis of more tumour sections is required to determine whether the differences observed were statistically significant.

Table 5.3. Number of mice that presented with tumour cells within the tumour stroma or blood vessels.

5 * * * 2 x10 cell dose Vehicle EMT MET LNCaP-iGFPRFP/LUC 0/2 1/2 0/2 LNCaP-iSnailRFP/LUC 0/0 1/2 3/3 LNCaP-iSlugRFP/LUC 0/2 0/3 0/4

1 x106 cell dose Vehicle* EMT* MET* LNCaP-iGFPRFP/LUC 1/2 2/3 3/3 LNCaP-iSnailRFP/LUC 1/1 2/3 0/3

LNCaP-iSlugRFP/LUC 2/4 2/3 1/3

* Untreated tumours: Vehicle; Tumours treated with Dox for 19 weeks: EMT; Tumours treated with Dox for 4 weeks followed by removal for up to 15 weeks: MET.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 166

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.22. Expression of Ku70 in the tumour stroma.

Representative sections showing Ku70 positive human cancer cells (brown) found in the tumour stroma of LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, and LNCaP-iSlugRFP/LUC tumours from mice that received intraprostatic injections of 1 x 106 cells. Tissue sections were counterstained with haematoxylin. Scale bar indicates 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 167 5.2.8 Assessment of distant tumour cells. At the time of termination, select mouse organs were examined ex-vivo for bioluminescent signal to detect metastatic deposits. Consistent with previous literature on the metastatic capacity of the LNCaP cell line, almost all mice had luciferase emission from the lymph nodes situated behind the primary tumour (44/47 mice). The lymph nodes were processed and stained with H&E to histologically identify lymph node and cancer cells (Figure 5.24). Unfortunately, the lymph nodes of three mice were lost during processing due to their small size. From the 44 mice, lymph nodes were identified in the tissue blocks for 37 mice. The remaining 7 tissue blocks contained muscle and stroma tissue but were devoid of lymphoid tissue and cancer cells. This was attributed to either inexperienced sectioning or that the lymph nodes were lost during processing. From the 37 mice whose lymph nodes were identified via H&E stain, 21 contained cancer cells that were confirmed by staining with Ku70 (Figure 5.23). The remaining 16 mouse lymph nodes were devoid of cancer cells even though 14 of them emitted a positive luciferase emission signal. This was attributed to either inexperienced sectioning or a false positive luciferase signal. Overall, it was observed that the confirmed presence of cancer cells within the lymph nodes was more frequent in the mice that received the 1 x 106 cell dose than mice that had been given the 2 x 105 cell dose (Table 5.4). However, the sample number across groups was too small to calculate whether the differences observed were statistically significant.

The following organs were also removed and imaged ex-vivo for luciferase signal: testes, spleen, kidneys, liver, lungs, and brain. None of the organs displayed macroscopic metastases, however, some organs were found to emit low levels of luciferase. It was observed that after the lymph nodes, the most probable organ to emit luciferase were the lungs. Coupled with the detection of cancer cells within murine circulatory vessels in the stroma surrounding the primary tumours, the lungs from all mice were serial sectioned and stained with Ku70 to detect human cancer cells (Figure 5.25). Indeed, disseminated cancer cells were identified in all treatment groups across all three cell lines and both cell doses. The identified cells were either single, in doublets, or small clusters. No discernible pattern was observed across the treatment groups, cell line or cell doses (Table 5.5).

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 168

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.23. Expression of Ku70 in the lymph nodes.

Representative sections showing Ku70 positive human cancer cells (brown) found in the lymph nodes of mice that received intraprostatic injections of 1 x 106 LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, or LNCaP-iSlugRFP/LUC cells. Tissue sections were counterstained with haematoxylin. Scale bar indicates 100 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 169 Table 5.4. Number of mice that presented with Ku70 positive tumour cells in the lymph node

5 * * * 2 x10 cell dose Vehicle EMT MET LNCaP-iGFPRFP/LUC 0/2 1/1 2/2 LNCaP-iSnailRFP/LUC 0/1 1/1 2/2 LNCaP-iSlugRFP/LUC 0/0 0/3 0/3

1 x106 cell dose Vehicle* EMT* MET* LNCaP-iGFPRFP/LUC 1/2 3/3 2/2

LNCaP-iSnailRFP/LUC 1/1 2/2 1/3

RFP/LUC LNCaP-iSlug 1/3 2/3 1/3

* Untreated tumours: Vehicle; Tumours treated with Dox for 19 weeks: EMT; Tumours treated with Dox for 4 weeks followed by removal for up to 15 weeks: MET.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 170

Figure 5.24. H&E staining of the lymph nodes.

Representative H&E sections showing tumour cells (T) within the lymph nodes of mice that received intraprostatic injections of 1 x 106 LNCaP-iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, or LNCaP-iSlugRFP/LUC cells. Scale bar indicates 100 μm. L = Lymphoid tissue, N = necrosis.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 171

Table 5.5. Number of mice that presented with Ku70 positive tumour cells in the lungs.

2 x 105 cells Vehicle* EMT* MET* LNCaP-iGFPRFP/LUC 1/1 1/2 2/2

LNCaP-iSnailRFP/LUC 2/2 1/2 3/3 LNCaP-iSlugRFP/LUC 0/2 2/2 2/3

1 x 106 cells Vehicle* EMT* MET* LNCaP-iGFPRFP/LUC 2/2 2/3 3/3 LNCaP-iSnailRFP/LUC 1/1 1/2 1/3

LNCaP-iSlugRFP/LUC 3/4 3/3 1/3

* Untreated tumours: Vehicle; Tumours treated with Dox for 19 weeks: EMT; Tumours treated with Dox for 4 weeks followed by removal for up to 15 weeks: MET..

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 172

V e h ic le E M T M E T

C

U

L

/

P

F

R

P

F

G

i

-

P

a

C

N

L

C

U

L

/

P

F

R

l

i

a

n

S

i

-

P

a

C

N

L

C

U

L

/

P

F

R

g

u

l

S

i

-

P

a

C

N L

Figure 5.25. Expression of Ku70 in the lungs.

Representative sections showing Ku70 positive human cancer cells (brown) found in the lungs of mice that received intraprostatic injections of 1 x 106 LNCaP- iGFPRFP/LUC, LNCaP-iSnailRFP/LUC, or LNCaP-iSlugRFP/LUC cells. Tissue sections were counterstained with haematoxylin. Scale bar indicates 50 μm.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 173 5.3 Discussion A number of in vivo studies have demonstrated that the inherent ability of cancer cells to shift phenotypic states facilitates metastasis. For instance, skin cancer cells induced into a mesenchymal state via induced expression of EMT-TF Twist were not capable of forming overt metastasis unless allowed to revert to their epithelial phenotype following the loss of Twist expression (Tsai et al. 2012). Similarly, metastasis by intravenous injection required the loss of the EMT-TF Prrx1 from Prrx1 expressing breast cancer cells (Ocana et al. 2012). In prostate cancer, studies by Celia-Terrassa further demonstrate that while mesenchymal cancer cells are capable of escaping the primary tumour, metastatic outgrowth requires cancer cells to be in an epithelial state (Celia-Terrassa et al. 2012). Also, mesenchymal cancer cells, as well as murine stromal tumour fibroblasts, have been shown to influence epithelial cancer cells to undergo EMT when in close proximity, bringing forward the hypothesis of possible collaboration between cancer cells in epithelial and mesenchymal states. Indeed, co-injection of hamster cheek pouch carcinoma cells that have undergone EMT and ones that didn’t (non-EMT cells) into the bloodstream resulted in metastasis to more vascularly complex organs such as the liver or kidneys compared to non-EMT cells which metastasised to lymph nodes and bone (Tsuji et al. 2008). Other studies highlight that an EMT induced mesenchymal state evokes chemoresistant properties and promotes invasion from the primary tumour and into the vasculature (circulating tumour cells; CTCs). However, metastases were commonly of epithelial nature with only a small population of cancer cells identified to have experienced phenotypic plasticity sometime throughout the metastatic process (Fischer et al. 2015; Zheng et al. 2015). Taken together, there is strong evidence that cancer cell plasticity plays a role in facilitating metastasis either directly or by collaborating with neighbouring tumour and stromal cells.

While these in vivo models provide proof-of-principle evidence for the role of cancer plasticity in promoting metastasis, the temporal characterisation of cancer cell plasticity in vivo is still lacking. Furthermore, there are no available in vivo orthotopic prostate cancer models able control the induction and reversal of EMT. To address this, the reversible EMT models herein were examined in vivo to assess whether they could simulate prostate cancer cell plasticity in a controlled manner. However, before attempting to investigate the temporal effects of cell plasticity via

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 174 the EMT program, this initial first in-field pilot study focused on determining not only whether the inducible models perform in vivo with the regulation of Dox, but also other experimental variables such as optimising the injected cell dose for maximum tumour take rate. This study also provided insights into the orthotopic growth of the LNCaP sublines and allowed for determining the time of growth to the ethical limit. Additionally, the metastatic capacity of the LNCaP cell line, outside of an experimentally induced EMT was also examined. Overall, the pilot study herein provided crucial information required for future experiments.

It was determined that a cell dose of 1 x 106 resulted in higher tumour take compared to 2 x 105 cells (Table 5.1), and this cell dose produced orthotopic tumours that grew to the ethical limit of 1000 mm3 by a maximum of 19 weeks post implantation (Figure 5.10). This provided a timeframe that can be used to design future experiments aiming to temporally assess cancer plasticity via the EMT program in prostate tumours. It was observed that almost all mice acquired metastasis in the lumbar lymph nodes situated behind the prostate, regardless of experimental conditions (Table 5.4, Figures 5.24 and 5.25). This is in line with a previous study demonstrating that LNCaP cells reliably metastasise to the lymph nodes when implanted orthotopically (Zeng et al. 2006; Sato et al. 1997). Furthermore, disseminated cancer cells were detected in the lungs of almost all mice (Table 5.5, Figure 5.26), indicating that LNCaP cells have the capacity to enter the bloodstream but do not form overt metastasis in distant organs in the timeframe of up to 19 weeks.

Before implantation of the EMT cell models in vivo, the expression of the inducible EMT-TFs as well as the EMT markers E-cadherin and vimentin was assessed following transient treatment with Dox by IHC. This revealed the heterogeneous nature of the EMT cell models, characterised by the wide range of staining intensities observed across all of the proteins assessed, including areas of negative staining where protein levels were below the detection limit (Figures 5.6, 5.7, and 5.8). Nevertheless, the reversible EMT models were used in vivo as is, without sorting for a homogeneous population of cells that represented a classic EMT, which is high levels of vimentin and low levels of E-cadherin, when the EMT-TF was expressed. This was based on the clinical observation that epithelial carcinomas,

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 175 including prostate cancer, are heterogeneous in nature and often express epithelial and mesenchymal markers in a heterogeneous pattern. Thus, maintaining the inducible EMT models as a heterogeneous population of cells, they more closely represented clinical prostate cancer. Additionally, it was hypothesised that the heterogeneous nature of the EMT-TF and markers represent a spectrum of EMT intensity, which once again may be more representative of a clinical scenario. To further recapitulate how metastasis is thought to occur in the patient, the cell models were allowed to form primary prostate tumours, prior to inducing a transient EMT.

The regulation of the transgene within each cell model by administration of Dox was successful as highlighted by the control LNCaP-iGFPRFP/LUC cell line (Figure 5.14). Additionally, the level of transgene expression was heterogeneous as seen previously during in vitro experimentation outlined in section 6.2.1 (Figures 5.14 and 5.15). While there was an increase in the number of cells staining positive for vimentin in LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells treated with Dox in vitro, it was noted that a small number of untreated LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC cells, as well as the LNCaP-iGFPRFP/LUC cells treated with or without dox stained positive for vimentin (Figure 5.7). This indicates that the LNCaP cell line inherently has a small population of vimentin positive cells. Following in vivo implantation, all three LNCaP models showed expression of vimentin with no notable increase in expression in the LNCaP-iSnailRFP/LUC and LNCaP-iSlugRFP/LUC tumours treated with dox (Figure 5.20). Simialrly, the expression of E-cadherin remained unaltered across all experimental groups and did not follow the pattern observed when they were assessed in vitro (Figure 5.21). However, considering the EMT timeframe was over the long period of 19 weeks, combined with the fact that the cell models were heterogeneous, it was hypothesised that perhaps the subset of cancer cells that underwent a “robust” EMT invaded away from the primary tumour, leaving behind their more epithelial counterparts. Initially, the experimental design included collecting primary tumours from mice following 1 week of treatment to assess their EMT status. Unfortunately due to unexpected mouse deaths, that collection time point was omitted to maintain group numbers for the longer collection time points. Therefore, the EMT status of the primary tumour was only assessed at the endpoint and not at 1-week post Dox administration. This is a crucial control that must be

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 176 included in follow-up studies, particularly for ensuring that an EMT has occurred before inducing a MErT.

The timeframe chosen to induce EMT with a subsequent MErT was 4 weeks and 15 weeks respectively. The EMT timeframe was selected to ensure that the LNCaP cells underwent a complete and robust EMT. In hindsight, the 4 week EMT timeframe was not tested in vitro and therefore it was unknown whether this may have caused epigenetic modifications to the LNCaP cells, potentially hindering reversal to the epithelial phenotype by becoming fixed in an intermediate state, which would inhibit potential metastatic outgrowth. A number of studies have identified the ability of EMT-TFs Snail and Slug to recruit various histone-modifying complexes such as HDAC1/2/3 and SIN3A, leading to the epigenetic silencing of epithelial genes including E-cadherin (Cano et al. 2000; Hajra, Chen and Fearon 2002; Batlle et al. 2000; Bolos et al. 2003). The study herein assessed the expression of E-cadherin during a 5 day Snail or Slug-induced EMT followed by a 20 day MErT under in vitro conditions, whereby aberrant E-cadherin expression decreased with EMT and reversed with MErT. However, a 4 week EMT induction and the effect on E- cadherin expression following MErT had not yet been assessed in vitro and will need to be examined prior to future animal experimentation.

The lack of overt metastases in the MET groups may also be due to the relatively long-term induction of EMT. LNCaP cells in an EMT state may have been unable to extravasate to the secondary location increasing their chance of becoming sheared by the exerted circulatory fluid forces (Mitchell and King 2013; Michor et al. 2011; Swartz and Lund 2012). Cumulatively, the EMT induction timeframe requires optimisation and shorter timeframes will be assessed in future experiments as they are thought to be more physiologically relevant compared to long-term induction such as the 4-week timeframe used in the study herein (Tsuji et al. 2008; Celia- Terrassa et al. 2012).

It was considered whether metastases had established in distant organs, but were undetectable because the MErT timeframe was limited by the primary tumour size. Unlike studies investigating breast cancer in vivo where the primary tumour is resected, it was not possible to remove the prostate tumour to allow additional time for the metastases to grow. One option to overcome this limitation would be to opt

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 177 for subcutaneous injection instead of intraprostatic injection. Alternatively, another option would be to induce EMT in vitro, followed by intracardiac injection of the cells into mice treated with or without Dox to either simulate a continuous EMT or allow for a MErT. While this study design misses the early stages of the metastatic cascade such as local invasion and intravasation (Saxena and Christofori 2013), the effect of EMT or MErT on organ colonisation once the cancer cells have already entered the bloodstream would be examined. This could provide a longer timeframe for potential metastases to grow to a macroscopic and detectable size. However, this approach would require careful optimisation as injection of a large bolus of cancer cells is not clinically representative (Ottewell, Coleman and Holen 2006), and can inadvertently lead to fast metastasis in critical vital organs due to the physical entrapment of cancer cells in capillaries. Another factor to consider would be the heterogeneity of the cell lines, and they may need to be sorted in an EMThigh and EMTlow population to assess for any collaborative effects between the cell types. Prior to approaching a sorting based experimental design, the LNCaP sublines should be sorted in vitro and the resulting populations characterised and examined for stability.

From a technical standpoint, it was observed that the use of luciferase to detect cancer cells was equivocal when attempting to detect low numbers of cancer cells. Serial sectioning of mouse lungs from all treatment groups revealed the presence of cancer cells, however, not all lungs emitted luciferase signal at the time of sacrifice. This was attributed to the luciferase emission being below detectable limits of the charged-coupled device (CCD) due to the small number of cells present in the lungs. Similarly, some organs emitted distinct levels of luciferase. However, no metastasis was confirmed following serial sectioning and staining with human specific Ku70. Furthermore, the location and depth of the cancer cells within the organ impacts the luciferase detection, therefore imagining all organ planes is advised (Zinn et al. 2008). The use of luciferase is considered an effective method of monitoring tumour growth in small animals, however, its limitations in detection sensitivity must be taken into consideration for future projects and additional detection methods should be employed for identifying micrometastases in distant organs. Such methods include serial sectioning of organs and staining for human specific markers such as Ku70 to identify the cancer cells or to genetically modify the LNCaP sublines with

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 178 the Lac-Z reporter gene for β-galactosidase expression where reaction with X-gal results in a dark blue colour, identifying the micrometastases (Buller et al. 2003). Alternatively, nanoparticle targeting systems with magnetic resonance imaging (MRI) could be utilised (Yang et al. 2009). This system requires the generation of cancer cells with cell surface receptors that can bind and internalise magnetic iron oxide (IO) nanoparticles introduced into the bloodstream. The internalisation of the IO nanoparticles provides a contrast when imaging with MRI which aids in the detection of tumours.

Another imaging technique that could be incorporated to increase tumour resolution down to individual cancer cells is intravital microscopy (IVM) and multiphoton imaging (Ellenbroek and van Rheenen 2014). IVM could be used to track individual cancer cells in the circulation, to aid in determining experimental treatment timeframes. For instance, in the study herein, mice with orthotopic LNCaP- iSnailRFP/LUC or LNCaP-iSlugRFP/LUC tumours could be treated with Dox to induce an EMT until cancer cells appeared in the circulation prior to removal from Dox to allow for a MErT. The method that is used to visualise circulatory vessels is relatively non-invasive and can be repeated daily as vessels can be directly optically accessed through the superficial layers of the mouse skin. IVM also allows for assessing the expression of certain markers if the cancer cells have pre-integrated fluorescently active reporters for those markers. For instance, the activity of E- cadherin could be evaluated by generating the reversible EMT models with a fluorescently labelled E-cadherin reporter. This would not only allow for determining the phenotype of the disseminating cancer cells but could also be used to monitor the behaviour of cancer cells in the primary tumour. To be able to visualise the implanted prostate tumours, a more invasive technique would be required that include the instalment of a “viewing window” in the lower abdomen, exposing the prostate tumour (Baron et al. 2011). This viewing window would allow for day-to-day visualisation of the prostate cancer cells and further elucidate the intra-tumoural behaviour of the reversible EMT models when grown orthotopically in vivo. Once the in vivo use of these cell models is optimised to a level whereby experimental metastasis is achieved, viewing windows could be installed in secondary organs to assess the effect of EMT-regulated plasticity on disseminated cancer cells and metastatic growth.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 179 In summary, this initial pilot study has provided valuable information on how the reversible EMT cell models grow and respond to Dox treatment when implanted orthotopically in vivo, and also provided insight on how the LNCaP cell line itself behaves under in vivo conditions. Further optimisation of the reversible EMT models is required before embarking on full-scale animal experiments assessing the temporal effects of cancer cell plasticity via the EMT program.

Chapter 5: Developing an in vivo model of EMP in PCa – A pilot study 180

Chapter 6: Final discussion and future directions

Chapter 6: Final discussion and future directions 181 6.1 The reversible epithelial-mesenchymal transition (EMT) models are unique tools for investigating epithelial-mesenchymal plasticity (EMP) in LNCaP cells. This thesis focused on investigating prostate cancer cell plasticity as regulated by the epithelial-mesenchymal transition (EMT) and mesenchymal to epithelial reverting transition (MErT) programs. This was addressed by developing inducible EMT prostate adenocarcinoma models that allowed for controlling cell plasticity by the reversible expression of major EMT transcription factors Snail, Slug, or Zeb1. To my knowledge, these reversible models are the first in the prostate cancer field to control the induction and reversal of EMT in a temporal manner, enabling interrogation of cancer cell plasticity in the same isogenic population of human prostate adenocarcinoma cells (LNCaP cells). This is of particular importance as there are limited models that are able to capture these transitions as they occur, instead of assessing end-point transitions, where cancer cells are at either end of the plasticity spectrum. Additionally, the ability to determine these transitions sequentially in a controlled manner allowed for further elucidating the temporal biology of cancer cells as they oscillate between phenotypic states.

Characterisation of the reversible EMT models in Chapter 3 showcased the tight control they offer over the expression of the EMT-TF by addition and removal of Dox. The induction of the EMT-TF resulted in a robust and “classical” EMT as shown by the swift upregulation of known mesenchymal markers and the downregulation of known epithelial markers (Figure 3.6). This was accompanied by a classic shift in the LNCaP cell morphology and behaviour, as highlighted when the inducible EMT models were cultured in 3D Matrigel™. When induced to undergo EMT in 3D Matrigel™, LNCaP cells that typically form multicellular non-invasive tumour spheroids, exhibited strong invasive properties that were characterised by single cells disassociating from the tumour spheroid and invading into the surrounding matrix (Figure 3.8). The reversible EMT models also adhere to the current concept that EMT induces epithelial cancer cells to enter a “dormant-like” non-proliferative state when in an induced mesenchymal state (Tsai et al. 2012; Celia-Terrassa et al. 2012; Ocana et al. 2012; Fischer et al. 2015; Zheng et al. 2015). Reduction of proliferation and associated proliferation markers was observed in EMT induced cells, and this was relieved when the expression of the EMT-TF was terminated following the removal of Dox (Figure 4.1). Removal of Dox induced a

Chapter 6: Final discussion and future directions 182 reversal to the epithelial phenotype (i.e. a MErT) that was reflected by the re- establishment of epithelial markers and an epithelial phenotype (Figure 3.14). Collectively, the use of these models in both 2D and 3D culturing conditions, allows for high customisability when designing experiments regarding epithelial plasticity, both in temporal terms or long-term end-point experiments. For instance, these models are excellent candidates for high throughput screening assays examining the effect of various drug or gene-targeted therapies on cancer cell plasticity. Their ability to recreate EMT in a 3D space also provides a platform for assessing the phenotypic attributes of EMT such as invasion or cell dormancy, which are particularly relevant prior to in vivo testing.

The reversible EMT models generated in this study recapitulated a reversible EMT in the LNCaP cell line, which was sufficient for investigating epithelial- mesenchymal plasticity for the purposes of this project. However, it would be of value to develop additional EMT models using other epithelial-like prostate cancer cell lines. Additionally, the EMT models could be generated using other types of cancer cell lines such as from breast, colon, or lung cancer, to allow for cross- comparing the effects of EMT across cancer types. This would not only serve for validation purposes but to also address heterogeneity between cell lines. Furthermore, this approach would also examine whether other cell types are capable of a reversible EMT. For instance, some cell types may be unable to revert with a MErT following an EMT and thus become “locked” in an EMT state. It is important to note that these “locked” states may also be dependent on the duration of the EMT induction as well as the type of EMT-TF induced. Specific to the experimental procedures used in this thesis, the induction of EMT was examined over a relatively short period of 5-7 days via inducible expression of Snail, Slug, or Zeb1. It would be of interest to assess whether the LNCaP cell line can still experience a reversible EMT if the EMT induction timeframe was extended to longer periods (>7 days) and whether there are any differences between the EMT-TFs used.

Ultimately, the successful generation of additional models of EMT in a number of cell lines using various EMT-TFs could be collated into an “EMT model library” in which users could test the effect of their favourite drug or gene target on the EMT program across a broader spectrum.

Chapter 6: Final discussion and future directions 183 6.2 MErT is a kinetically dynamic process and is enriched in mCRPC. The reversible EMT models were predominantly used to transcriptionally characterise a cycle of Snail-induced EMT and MErT, with a selective focus on the less studied process of MErT (Chapter 4). The assessment of a cycle of epithelial- mesenchymal plasticity (EMP) in the same isogenic population of human prostate cancer cells provided insights into the temporal re-establishment of the epithelial phenotype at a transcriptional level. The reversion of mesenchymal cells to an epithelial state revealed for the first time the dynamic nature of the MErT program, characterised by the identification of distinct clusters of transcripts transiently reverting to pre-EMT expression levels. The majority of transcriptional reversions occurred early in the MErT (within 3 days), however, some transcriptional clusters returned to pre-EMT expression levels in the later stages of the MErT timeframe tested (up to 20 days) (Figure 4.4). Examination of the MErT signature in clinical samples revealed MErT to be enriched in the metastases of patients with prostate cancer as well as castrate-resistant prostate cancer (mCRPC), providing prototypical evidence supporting the involvement of MErT in clinical metastasis.

6.3 MErT imprints unique transcriptional features that are enriched in mCRPC and could serve as clinical identifiers/therapeutic targets for cancer cells that have undergone a reversible EMT. Interestingly, while the vast majority of transcripts followed a reverting pattern with MErT, it was observed that a select number of probes either failed to revert or became newly altered with the onset of MErT (Figure 4.8). These novel genes were found to be positively enriched in the metastases of patients with castrate-resistant prostate cancer (mCRPC), supporting the hypothesis of a MErT occurrence at the secondary site (Figure 4.8). It is acknowledged that this is an indirect indication of an EMT/MErT event. However, the positive enrichment of those specific genes in the metastases of patients with mCRPC, along with their negative enrichment at the primary site of treatment naïve patients, supports the further examination of their potential function in the metastatic process.

To date, assessment of EMT and MErT relies on the expression of known epithelial and mesenchymal markers to identify the cell phenotype in clinical samples. However, this approach only identifies the cell phenotype and cannot

Chapter 6: Final discussion and future directions 184 distinguish cells that have undergone an EMT or MErT. Furthermore, the currently used markers cannot differentiate between tumour cells and non-tumour cell types found within the stroma. The novel EMT and MErT markers identified in this study are promising candidates to serve as hallmarks for these transitions, with the MErT activated genes serving as markers for a cycle of EMT/MErT. Of particular interest are the MErT activated genes, as they may act as gatekeepers of the MErT program and reacquisition of the epithelial phenotype. Similarly, the EMT persistent genes could be linked to epigenetic alterations and may serve as a permanent marker of cells that have experienced an EMT event. The identified novel genes, both EMT persistent and MErT activated, need to be further characterised and their function in cell plasticity elucidated as they may be attractive therapeutic targets for promoting or inhibiting the EMT and MErT programs. For instance, promoting EMT/ inhibiting MErT would maintain cells in an EMT induced mesenchymal state that may slow or prevent outgrowth of metastases in distant organs. On the other hand, promoting a MErT/inhibiting an EMT would result in having cancer cells in an epithelial state. As EMT promotes chemoresistance (Zheng et al. 2015), inducing/maintaining cancer cells in an epithelial state may increase the effectiveness of the administered therapies. However, relatively little is known about the chemoresistance of cancer cells following MErT or even multiple cycles of EMT/MErT. Therefore, it is important to investigate the chemoresistant abilities of cancer cells once they have experienced these transitions. As EMT and MErT are mostly reverse transitions of each other, often the inhibition of one transition promotes the other; it would be of particular interest to investigate whether any of the identified novel genes can influence one transition without affecting the other. For example, the silencing of a MErT activated gene could perhaps prevent a mesenchymal cancer cell from reverting to its epithelial phenotype without inducing or promoting an EMT; thus, acting to block rather than promote a reversion.

Therefore, it is important to determine the permanency of these novel gene alterations following multiple cycles of EMT and MErT. To best isolate those genetic changes, the reversible EMT models would need to be sorted for cells that have undergone EMT following each cycle. This is to ensure that only cells that have undergone an EMT are assessed in subsequent cycles. The sorting is proposed because of the heterogeneous nature of the reversible EMT models as was observed

Chapter 6: Final discussion and future directions 185 by the heterogeneous expression of the EMT-TF following Dox induction (Chapter 5, Figure 6). This may have been limited by the detection sensitivity of the antibody and technique used. A more robust approach in identifying the cancer cells that are actively undergoing EMT would be to utilise an E-cadherin reporter construct. For instance, an RFP-tagged reporter that targets downstream the E-cadherin E-box sequence would identify cells actively undergoing EMT, allowing for precise sorting by flow cytometry at every EMT cycle. The assessment of cancer cells following cycles of EMT and MErT may reveal additional markers for these transitions. These could serve as biomarkers of these transitions in patient samples, something that is currently lacking in the field. Additionally, as discussed above, these permanent markers may play important roles for these transitions, and may hold prognostic and therapeutic value for the patients.

An alternative to identifying genes that either promote or inhibit these transitions is the identification of lethality inducing genes that are specific to cancer cells that have experienced these transitions. A proposed method for identifying such genes would be by performing a siRNA library screen on cancer cells maintained in specific phenotypic states induced by these transitions (i.e., untreated cells, EMT, or MErT-induced cells). This approach would identify genes that are imperative to the survival of cancer cells in those states. For example, selective lethality towards cancer cells that have transitioned from an EMT-induced mesenchymal state to an epithelial state (i.e., MErT), but not in untreated cancer cells, could perhaps inhibit/limit the outgrowth of metastases in distant organs. On the other hand, targeting cancer cells in an induced EMT state could eliminate cancer cells that are likely to invade the primary tumour and disseminate in the circulation. This would have to be performed on multiple prostate cancer cell lines, and perhaps even in cell liens from different cancer types to allow for heterogeneity. Ultimately, identifying lethality inducing genes for cancer cells that have experienced these transitions would provide an alternative therapeutic approach for patients with metastatic prostate cancer.

Chapter 6: Final discussion and future directions 186 6.4 Evidence of EMP involved in the androgen signalling axis. The expression of two candidates from the identified novel transcriptional clusters, TUBB3 (from the unregulated EMT persistent gene cluster) and POU4F1 (from the upregulated MET unique gene cluster) was briefly examined in clinical samples of primary PCa (data obtained by Dr Hollier in collaborative studies with Prof Ralph Buttyan, Dr Ladan Fazli and D/Prof Martin Gleave, Vancouver Prostate Centre, Vancouver, Canada). Interestingly, approximately half of the treatment naïve patient samples (49%) had co-expression of TUBB3/POU4F1 proteins (Figure 6.1). However, this increased to over 90% of patient tumours that progressed to CRPC. Assuming that the expression of TUBB3/POU4F1 would come about following a cycle of EMT/MErT, these results provide evidence supporting the occurrence of intra-tumoural EMT/MErT at the primary site. While the expression of additional markers from the novel clusters should be examined, these results support that cancer cell plasticity is influenced by the androgen signalling axis, and perhaps, increased cancer cell plasticity associates with resistance to androgen targeted therapies and progression to a castration-resistant phenotype. Indeed, androgen deprivation therapy has been linked with inducing an EMT in prostate cancer cells (Sun et al. 2012). Future studies to confirm these findings would be to assess cell plasticity by regulating the expression of the androgen receptor (AR), either by utilising inducible knockdown shAR models (already established in Dr. Hollier’s laboratory) or by manipulation of the available androgens present in the cell media. This would be performed by either removal (i.e., use of growth media supplemented with charcoal-stripped foetal bovine serum (FBS) as opposed to non-charcoal stripped FBS) or AR antagonists like enzalutamide to model conditions of androgen deprivation. Information from these experiments would further elucidate the role of cancer plasticity and the androgen axis, which is a driving factor for prostate cancer progression.

Chapter 6: Final discussion and future directions 187 P O U 4 F 1 p o s /T U B B 3 p o s

P O U 4 F 1 p o s /T U B B 3 n e g

n e g p o s P O U 4 F 1 /T U B B 3

n e g n e g P O U 4 F 1 /T U B B 3 P O U 4 F 1 T U B B 3

4 9 %

9 %

3 0 %

e v

i 2 0 %

a N

C 9 1 %

P

R C 0 % 4 % 5 %

Figure 6.1. Expression of POU4F1 and TUBB3 in treatment naïve prostate cancer and castrate-resistant prostate cancer (CRPC) primary tumours.

Tissues counterstained with haematoxylin. Scale bar indicates 100 μm.

Chapter 6: Final discussion and future directions 188 6.5 Intra-tumoural cell plasticity via the EMT program at the primary site is predictive for poor patient outcome. Further supporting a role for cell plasticity in driving aggressive tumour behaviour was the discovery of the metastatic plasticity signature (MPS). From the identified spectrum of genes that were altered with a cycle of Snail-induced EMT and MErT, a core set was found to be upregulated in patients with mCRPC as compared to treatment naïve PCa. Further investigation of this signature revealed that its positive expression in primary prostate tumours was predictive of shorter biochemical recurrence and survival in multiple patient cohorts (Figure 4.10, Table 4.2). The predictive ability of the MPS was not only effective in prostate cancer patient groups, but also in patient cohorts of other epithelial cancers such as that of the breast, lung, and colon (Figure 4.10, Table 4.3). The MPS has led to the submission of an Australian Provisional Patent Application (#2015905357) and has attracted interest from industry partners in developing/refining the signature for use as a prognostic test of disease progression. Overall, the results highlight that while the models used herein focused on prostate cancer, the findings apply to other epithelial cancers, indicating that cancer cell plasticity is potentially a general trait for cancer aggressiveness and progression.

6.6 EMP reprograms the metabolic phenotype of LNCaP cells. A number of previously established and validated gene panels used for patient disease prediction, such as the Decipher signature (Erho et al. 2013), or the Cell Cycle Progression (CCP) signature (Cuzick et al. 2011), are mostly reliant on the expression of genes related to cell cycle processes. Indeed, the enrichment of cell cycle related genes portrays a highly proliferating primary tumour, which ultimately leads to a worse patient outcome. Due to this, many chemotherapeutic approaches target cell cycle and highly proliferating cells to reduce tumour burden. While the MPS also included a substantial number of cell cycle related genes, its predictive power was not solely reliant on those genes. Examination of an MPS devoid of genes related to cell cycle (MPSCCR) revealed that the MPS was still predictive of patient outcome (Figure 4.12 and 4.15, Table 4.5). Investigation of the remaining non-cell cycle related genes revealed enrichment in mostly metabolism-related processes. Indeed, examination of a MPS signature consisting of only genes found within the metabolism related processes (MPSMETAB) revealed that the MPSMETAB was still

Chapter 6: Final discussion and future directions 189 predictive of patient outcome and was associated with poor patient outcome in a number of cancer types.

As briefly shown in Chapter 5, EMT suppressed the OXPHOS and glycolytic rate of LNCaP cells, and this was re-established with MErT (Figure 4.12). Interestingly, MErT saw an increase in both basal and maximal OXPHOS rates when compared to untreated cells. This elevation perhaps supports that cancer cells that have experienced EMP enter a more active metabolic state, favouring growth. There is currently mounting evidence supporting that EMT shifts cancer cell metabolism from biosynthetic pathways such as lipogenesis and glycolysis to bioenergetics pathways such as the tricarboxylic acid cycle (TCA cycle) and oxidative phosphorylation (Jiang, Deberardinis and Boothman 2015; Zadran et al. 2014; Zhou et al. 2014; LeBleu et al. 2014). This metabolic switch was found to be bi-directional as shown following experimental regulation of fatty acid synthase (FASN) via transient treatment with the EMT-inducing growth factor TGF-β1 (Jiang, Deberardinis and Boothman 2015). Further studies showed that direct knockdown of FASN resulted in a decrease in primary tumour growth, but enhanced metastasis (Jiang, Deberardinis and Boothman 2015). Collectively, the data supports that the induction of EMT inhibits proliferation (Zhou et al. 2014) and reprograms cancer cell metabolism towards bioenergetic - ATP-producing oxidative phosphorylation to promote invasion and metastasis. This is hypothesised to revert with MErT at the metastatic site, and the data herein supports that this reversion leads to enhanced OXPHOS and increased growth. The regulation of metabolism via the EMT and MErT programs requires further investigation as targeting metabolic pathways to inhibit metabolism is becoming increasingly studied as a new therapeutic approach for cancer patients.

6.7 Suggested roles for EMP in cancer progression. The hypothesis that the EMT and MErT programs facilitate cancer progression and metastasis originated from the observed roles that these transitions have in embryonic development. During development, embryonic cells undergo multiple cycles of EMT and MErT as they differentiate to form various tissues and organs (Thiery et al. 2009). When epithelial cells undergo EMT, they gain invasive properties allowing them to migrate, a feature that could explain how cancer cells

Chapter 6: Final discussion and future directions 190 initially invade away from the primary site and into the circulation. In cancer patients, it is often observed that the centre of large tumours is typically hypoxic and expresses mesenchymal markers, including EMT-TFs such as Snail, Slug, or Zeb1 (Mak et al. 2010). Experimentally, the exposure of cancer cells to hypoxic conditions in vitro induces an EMT and this observation indirectly supports that cancer cells can alter their plasticity in response to the tumour microenvironment. Considering that EMT provides cancer cells with invasive abilities, the hypothesis that EMT allows cancer cells to invade into the circulation is an attractive concept. However, since metastases often reflect an epithelial phenotype, coupled with the in vitro observation that cancer cells in an induced mesenchymal state have reduced proliferative capabilities, the proposition that a reversion occurs at the secondary site is currently being investigated. The recapitulation of these processes in animal models by controlling the expression of EMT-inducing TFs has yielded promising results in supporting that cancer cells can accomplish the multiple steps required in the metastatic cascade by altering their plasticity (Celia-Terrassa et al. 2012; Tsai et al. 2012; Ocana et al. 2012). Currently, the majority of information pathologists can extract from patient samples is the identity of cancer cells by their expression/ co- expression of epithelial or mesenchymal markers. A shortcoming of this method is that it can underestimate the identification of cancer cells from surrounding non- cancer cells and the current markers used in the clinic cannot determine whether these transitions have occurred. Therefore, there is an urgent need to identify markers that are unique to these transitions to determine their occurrence directly in patient samples.

The novel EMT persistent and MErT activated genes identified herein are promising candidates for identifying cancer cells that have undergone these transitions. Their positive enrichment in metastasis further supports the occurrence of an EMT and MErT, and while these findings indirectly suggest the involvement of cell plasticity in the process of metastasis, it is acknowledged that these markers may only indicate intra-tumoural plasticity and not a means for metastatic dissemination. In combination with the finding that cancer patients with an MPS positive enriched primary tumour face a worse outcome than patients that don’t, it is hypothesised that intra-tumoural plasticity increases the aggressiveness and progression to the metastatic stage (Figure 6.2) Experimental modelling of metastatic

Chapter 6: Final discussion and future directions 191 dissemination via the EMT program in murine models has yielded mixed results. On one hand, the experimental control of EMT-TF expression in cancer cells increases metastatic tumour burden (Celia-Terrassa et al. 2012; Tsai et al. 2012; Ocana et al. 2012). On the other hand, spontaneous models of metastasis whereby the plasticity of cancer cells could be monitored by the use of reporter vectors targeting mesenchymal gene promoters, revealed that while circulating cancer cells undergo EMT, the majority of cancer cells in the metastases did not (Fischer et al. 2015; Zheng et al. 2015). It is important to note that the use of these reporters marked cancer cells that had undergone an EMT, but could not identify whether those cancer cells reverted to their epithelial phenotype via MErT, as the reporter activation was an irreversible process. However, a small percentage of cancer cells in the metastases were found to be EMT positive, which supports that cancer cells can undergo EMT (and possible MErT) without experimentally inducing an EMT, and their presence at the metastatic site warrants further investigation in their role while in transit/extravasation or at the metastatic site. Alongside, other studies support that epithelial and mesenchymal cancer cells influence each other, as was shown by an increase in metastatic load when co-injected in the bloodstream, compared to either cell phenotype alone (Celia-Terrassa et al. 2012; Tsuji et al. 2008). Taken together, the data supports that cancer cell plasticity via the EMT and MErT programs may be involved in more than just cancer cell dissemination and is an area that warrants further investigation.

6.8 The reversible EMT models and their use for in-vivo studies. An important feature of the reversible EMT models generated herein is their potential applications for studying EMP in vivo. The first attempt to use these models in vivo established their growth pattern and composition when implanted orthotopically. While the in vivo study conducted herein requires more optimisation in defining the timeframes for induction of EMT and MErT, it was clear that the expression of the EMT-TF was regulated by Dox as seen for in vitro experiments. A limiting factor that could explain the lack of gross metastases in distant organs was the early termination of the mice due to the primary tumour size reaching ethical limits. Hence, not allowing enough time for detectable metastases to establish. One way to overcome this would be to investigate the effect of cancer cells in an EMT or

Chapter 6: Final discussion and future directions 192

Figure 6.2. Hypothetical roles of EMT and MErT in cancer progression.

A. Cancer cells undergo EMT at the primary site to gain metastatic properties followed by a MErT at the secondary site to allow for metastatic growth. B. Cancer cells undergo both EMT and MErT at the primary site, whereby increased intra- tumoural plasticity results in a more aggressive and metastatic cancer.

Chapter 6: Final discussion and future directions 193 MErT state on the metastatic establishment in vital organs following delivery into the circulation via intracardiac injection. While this technique also bypasses a number of the initial steps in the metastatic cascade, such as local invasion and intravasation, it would allow for the direct assessment of cell plasticity on extravasation and metastatic growth following entry into the circulation. This approach would serve as a more appropriate model for assessing the effectiveness of therapies in inhibiting metastasis in prostate cancer as often patients already have disseminated disease at the time of diagnosis.

6.9 Summary In summary, the reversible EMT models generated herein allowed for the detailed investigation of EMP in prostate cancer due to the tight regulation they offer over the activity of the EMT program. Temporal investigation of the transcriptional events partaking during EMP mapped for the first time the transcriptional dynamics of a reversible EMT and revealed that MErT imparts cancer cells with new transcriptional signatures. The enrichment of MErT in primary tumours supported the occurrence of intra-tumoural plasticity via the EMT/MErT programs. From this enrichment, a plasticity-related gene signature (the MPS) was found to be predictive of faster recurrence and shorter survival across multiple types of epithelial cancers. Importantly, the prognostic ability of the MPS was not reliant on cell-cycle-related genes as removal maintained its performance. Furthermore, the regulation of EMT reprogrammed the metabolic phenotype of LNCaP cells and this was found to be transcriptionally predictive for poor patient outcome. The novel subprograms imparted by a reversible EMT were found to be enriched in mCRPC, providing prototypical evidence of the involvement of epithelial plasticity in human carcinoma progression. These novel signatures are promising candidates for discovering biomarkers that can identify the EMT or MErT transitions in clinical samples; something that is currently lacking in the field.

Chapter 6: Final discussion and future directions 194

Reference List

Reference List 195

Acloque, H., M. S. Adams, K. Fishwick, M. Bronner-Fraser and M. A. Nieto. 2009. "Epithelial- mesenchymal transitions: the importance of changing cell state in development and disease." J Clin Invest 119 (6): 1438-49. http://www.ncbi.nlm.nih.gov/pubmed/19487820. doi: 10.1172/JCI38019.

ACNWP. 2003. "Australian Cancer Network Working Party on Management of Localised Prostate Cancer. Clinical Practice Guidelines: Evidence-based information and recommendations for the management of localised prostate cancer. Canberra: Commonwealth of Australia.

Ahmad, A., Z. Wang, D. Kong, R. Ali, S. Ali, S. Banerjee and F. H. Sarkar. 2011. "Platelet- derived growth factor-D contributes to aggressiveness of breast cancer cells by up- regulating Notch and NF-kappaB signaling pathways." Breast Cancer Res Treat 126 (1): 15-25. http://www.ncbi.nlm.nih.gov/pubmed/20379844. doi: 10.1007/s10549- 010-0883-2.

Aigner, K., B. Dampier, L. Descovich, M. Mikula, A. Sultan, M. Schreiber, W. Mikulits, et al. 2007. "The transcription factor ZEB1 (deltaEF1) promotes tumour cell dedifferentiation by repressing master regulators of epithelial polarity." Oncogene 26 (49): 6979-88. http://www.ncbi.nlm.nih.gov/pubmed/17486063. doi: 10.1038/sj.onc.1210508.

AIHW. 2014. "Australian Institute of Health and Welfare. Cancer in Australia: an overview 2014." Cancer series no. 90. Cat. no. CAN 88.

AIHW. 2015. "Australian Institute of Health and Welfare. Cancer in Australia 2014: Actual incidence data from 1982 to 2011 and mortality data from 1982 to 2012 with projections to 2014." Asia Pac J Clin Oncol 11 (3): 208-20. http://www.ncbi.nlm.nih.gov/pubmed/26264473. doi: 10.1111/ajco.12407.

Aktas, B., M. Tewes, T. Fehm, S. Hauch, R. Kimmig and S. Kasimir-Bauer. 2009. "Stem cell and epithelial-mesenchymal transition markers are frequently overexpressed in circulating tumor cells of metastatic breast cancer patients." Breast Cancer Res 11 (4): R46. http://www.ncbi.nlm.nih.gov/pubmed/19589136. doi: 10.1186/bcr2333.

Alexander, N. R., N. L. Tran, H. Rekapally, C. E. Summers, C. Glackin and R. L. Heimark. 2006. "N-cadherin gene expression in prostate carcinoma is modulated by integrin- dependent nuclear translocation of Twist1." Cancer Res 66 (7): 3365-9. http://www.ncbi.nlm.nih.gov/pubmed/16585154. doi: 10.1158/0008-5472.CAN- 05-3401.

Reference List 196 Alexandrow, M. G. and H. L. Moses. 1995. "Transforming growth factor beta and cell cycle regulation." Cancer Res 55 (7): 1452-7. http://www.ncbi.nlm.nih.gov/pubmed/7882352.

Alkatout, I., M. Wiedermann, M. Bauer, A. Wenners, W. Jonat and W. Klapper. 2013. "Transcription factors associated with epithelial-mesenchymal transition and cancer stem cells in the tumor centre and margin of invasive breast cancer." Exp Mol Pathol 94 (1): 168-73. http://www.ncbi.nlm.nih.gov/pubmed/22985790. doi: 10.1016/j.yexmp.2012.09.003.

Alphonso, A. and S. K. Alahari. 2009. "Stromal cells and integrins: conforming to the needs of the tumor microenvironment." Neoplasia 11 (12): 1264-71. http://www.ncbi.nlm.nih.gov/pubmed/20019834.

Armstrong, A. J., M. S. Marengo, S. Oltean, G. Kemeny, R. L. Bitting, J. D. Turnbull, C. I. Herold, P. K. Marcom, D. J. George and M. A. Garcia-Blanco. 2011. "Circulating tumor cells from patients with advanced prostate and breast cancer display both epithelial and mesenchymal markers." Mol Cancer Res 9 (8): 997-1007. http://www.ncbi.nlm.nih.gov/pubmed/21665936. doi: 10.1158/1541-7786.MCR- 10-0490.

Armstrong, C. M. and A. C. Gao. 2015. "Drug resistance in castration resistant prostate cancer: resistance mechanisms and emerging treatment strategies." Am J Clin Exp Urol 3 (2): 64-76. http://www.ncbi.nlm.nih.gov/pubmed/26309896.

Arnold, S. J., U. K. Hofmann, E. K. Bikoff and E. J. Robertson. 2008. "Pivotal roles for eomesodermin during axis formation, epithelium-to-mesenchyme transition and endoderm specification in the mouse." Development 135 (3): 501-11. http://www.ncbi.nlm.nih.gov/pubmed/18171685. doi: 10.1242/dev.014357.

Ashburner, M., C. A. Ball, J. A. Blake, D. Botstein, H. Butler, J. M. Cherry, A. P. Davis, et al. 2000. "Gene ontology: tool for the unification of biology. The Gene Ontology Consortium." Nat Genet 25 (1): 25-9. http://www.ncbi.nlm.nih.gov/pubmed/10802651. doi: 10.1038/75556.

Ashworth, T. R. 1869. "A case of cancer in which cells similar to those in tumors were seen in the blood after death." Aust. Med. J (14): 146-149.

Asmane, I., J. Ceraline, B. Duclos, L. Rob, V. Litique, P. Barthelemy, J. P. Bergerat, P. Dufour and J. E. Kurtz. 2011. "New strategies for medical management of castration- resistant prostate cancer." Oncology 80 (1-2): 1-11. http://www.ncbi.nlm.nih.gov/pubmed/21577012. doi: 10.1159/000323495.

Reference List 197 Assikis, V. J., K. A. Do, S. Wen, X. Wang, J. H. Cho-Vega, S. Brisbay, R. Lopez, et al. 2004. "Clinical and biomarker correlates of androgen-independent, locally aggressive prostate cancer with limited metastatic potential." Clin Cancer Res 10 (20): 6770-8. http://www.ncbi.nlm.nih.gov/pubmed/15501953. doi: 10.1158/1078-0432.CCR-04- 0275.

Attard, G., A. S. Belldegrun and J. S. de Bono. 2005. "Selective blockade of androgenic steroid synthesis by novel lyase inhibitors as a therapeutic strategy for treating metastatic prostate cancer." BJU Int 96 (9): 1241-6. http://www.ncbi.nlm.nih.gov/pubmed/16287438. doi: 10.1111/j.1464- 410X.2005.05821.x.

Bachelder, R. E., S. O. Yoon, C. Franci, A. G. de Herreros and A. M. Mercurio. 2005. "Glycogen synthase kinase-3 is an endogenous inhibitor of Snail transcription: implications for the epithelial-mesenchymal transition." J Cell Biol 168 (1): 29-33. http://www.ncbi.nlm.nih.gov/pubmed/15631989. doi: 10.1083/jcb.200409067.

Banys-Paluchowski, M., N. Krawczyk, F. Meier-Stiegen and T. Fehm. 2015. "Circulating tumor cells in breast cancer-current status and perspectives." Crit Rev Oncol Hematol. http://www.ncbi.nlm.nih.gov/pubmed/26563820. doi: 10.1016/j.critrevonc.2015.10.010.

Baritaki, S., A. Chapman, K. Yeung, D. A. Spandidos, M. Palladino and B. Bonavida. 2009. "Inhibition of epithelial to mesenchymal transition in metastatic prostate cancer cells by the novel proteasome inhibitor, NPI-0052: pivotal roles of Snail repression and RKIP induction." Oncogene 28 (40): 3573-85. http://www.ncbi.nlm.nih.gov/pubmed/19633685. doi: 10.1038/onc.2009.214.

Baron, V. T., J. Welsh, P. Abedinpour and P. Borgstrom. 2011. "Intravital microscopy in the mouse dorsal chamber model for the study of solid tumors." Am J Cancer Res 1 (5): 674-86. http://www.ncbi.nlm.nih.gov/pubmed/21994905.

Bates, R. C., D. I. Bellovin, C. Brown, E. Maynard, B. Wu, H. Kawakatsu, D. Sheppard, P. Oettgen and A. M. Mercurio. 2005. "Transcriptional activation of integrin beta6 during the epithelial-mesenchymal transition defines a novel prognostic indicator of aggressive colon carcinoma." J Clin Invest 115 (2): 339-47. http://www.ncbi.nlm.nih.gov/pubmed/15668738. doi: 10.1172/JCI23183.

Batlle, E., E. Sancho, C. Franci, D. Dominguez, M. Monfar, J. Baulida and A. Garcia De Herreros. 2000. "The transcription factor snail is a repressor of E-cadherin gene expression in epithelial tumour cells." Nat Cell Biol 2 (2): 84-9. http://www.ncbi.nlm.nih.gov/pubmed/10655587. doi: 10.1038/35000034.

Reference List 198 Bednarz, N., E. Eltze, A. Semjonow, M. Rink, A. Andreas, L. Mulder, J. Hannemann, et al. 2010. "BRCA1 loss preexisting in small subpopulations of prostate cancer is associated with advanced disease and metastatic spread to lymph nodes and peripheral blood." Clin Cancer Res 16 (13): 3340-8. http://www.ncbi.nlm.nih.gov/pubmed/20592016. doi: 10.1158/1078-0432.CCR-10- 0150.

Beerling, E., D. Seinstra, E. de Wit, L. Kester, D. van der Velden, C. Maynard, R. Schafer, et al. 2016. "Plasticity between Epithelial and Mesenchymal States Unlinks EMT from Metastasis-Enhancing Stem Cell Capacity." Cell Rep 14 (10): 2281-8. http://www.ncbi.nlm.nih.gov/pubmed/26947068. doi: 10.1016/j.celrep.2016.02.034.

Behnsawy, H. M., H. Miyake, K. Harada and M. Fujisawa. 2013. "Expression patterns of epithelial-mesenchymal transition markers in localized prostate cancer: significance in clinicopathological outcomes following radical prostatectomy." BJU Int 111 (1): 30-7. http://www.ncbi.nlm.nih.gov/pubmed/23107154. doi: 10.1111/j.1464- 410X.2012.11551.x.

Bidard, F. C., D. J. Peeters, T. Fehm, F. Nole, R. Gisbert-Criado, D. Mavroudis, S. Grisanti, et al. 2014. "Clinical validity of circulating tumour cells in patients with metastatic breast cancer: a pooled analysis of individual patient data." Lancet Oncol 15 (4): 406-14. http://www.ncbi.nlm.nih.gov/pubmed/24636208. doi: 10.1016/S1470- 2045(14)70069-5.

Bienz, M. and H. Clevers. 2000. "Linking colorectal cancer to Wnt signaling." Cell 103 (2): 311-20. http://www.ncbi.nlm.nih.gov/pubmed/11057903.

Bierie, B. and H. L. Moses. 2006. "Tumour microenvironment: TGFbeta: the molecular Jekyll and Hyde of cancer." Nat Rev Cancer 6 (7): 506-20. http://www.ncbi.nlm.nih.gov/pubmed/16794634. doi: 10.1038/nrc1926.

Boehm, J. S. and W. C. Hahn. 2011. "Towards systematic functional characterization of cancer genomes." Nat Rev Genet 12 (7): 487-98. http://www.ncbi.nlm.nih.gov/pubmed/21681210. doi: 10.1038/nrg3013.

Bohmer, D., M. Wirth, K. Miller, V. Budach, A. Heidenreich and T. Wiegel. 2016. "Radiotherapy and Hormone Treatment in Prostate Cancer." Dtsch Arztebl Int 113 (14): 235-41. http://www.ncbi.nlm.nih.gov/pubmed/27146591. doi: 10.3238/arztebl.2016.0235.

Bolos, V., H. Peinado, M. A. Perez-Moreno, M. F. Fraga, M. Esteller and A. Cano. 2003. "The transcription factor Slug represses E-cadherin expression and induces epithelial to

Reference List 199 mesenchymal transitions: a comparison with Snail and E47 repressors." J Cell Sci 116 (Pt 3): 499-511. http://www.ncbi.nlm.nih.gov/pubmed/12508111.

Bonnomet, A., L. Syne, A. Brysse, E. Feyereisen, E. W. Thompson, A. Noel, J. M. Foidart, P. Birembaut, M. Polette and C. Gilles. 2012. "A dynamic in vivo model of epithelial- to-mesenchymal transitions in circulating tumor cells and metastases of breast cancer." Oncogene 31 (33): 3741-53. http://www.ncbi.nlm.nih.gov/pubmed/22120722. doi: 10.1038/onc.2011.540.

Boquest, A. C., A. Shahdadfar, K. Fronsdal, O. Sigurjonsson, S. H. Tunheim, P. Collas and J. E. Brinchmann. 2005. "Isolation and transcription profiling of purified uncultured human stromal stem cells: alteration of gene expression after in vitro cell culture." Mol Biol Cell 16 (3): 1131-41. http://www.ncbi.nlm.nih.gov/pubmed/15635089. doi: 10.1091/mbc.E04-10-0949.

Brabletz, S., K. Bajdak, S. Meidhof, U. Burk, G. Niedermann, E. Firat, U. Wellner, et al. 2011. "The ZEB1/miR-200 feedback loop controls Notch signalling in cancer cells." EMBO J 30 (4): 770-82. http://www.ncbi.nlm.nih.gov/pubmed/21224848. doi: 10.1038/emboj.2010.349.

Brabletz, T., A. Jung, S. Reu, M. Porzner, F. Hlubek, L. A. Kunz-Schughart, R. Knuechel and T. Kirchner. 2001. "Variable beta-catenin expression in colorectal cancers indicates tumor progression driven by the tumor environment." Proc Natl Acad Sci U S A 98 (18): 10356-61. http://www.ncbi.nlm.nih.gov/pubmed/11526241. doi: 10.1073/pnas.171610498.

Bracken, C. P., P. A. Gregory, N. Kolesnikoff, A. G. Bert, J. Wang, M. F. Shannon and G. J. Goodall. 2008. "A double-negative feedback loop between ZEB1-SIP1 and the microRNA-200 family regulates epithelial-mesenchymal transition." Cancer Res 68 (19): 7846-54. http://www.ncbi.nlm.nih.gov/pubmed/18829540. doi: 10.1158/0008-5472.CAN-08-1942.

Bubendorf, L., A. Schopfer, U. Wagner, G. Sauter, H. Moch, N. Willi, T. C. Gasser and M. J. Mihatsch. 2000. "Metastatic patterns of prostate cancer: an autopsy study of 1,589 patients." Hum Pathol 31 (5): 578-83. http://www.ncbi.nlm.nih.gov/pubmed/10836297.

Buller, C. J., X. P. Zang, E. W. Howard and J. T. Pento. 2003. "Measurement of beta- galactosidase tissue levels in a tumor cell xenograft model." Methods Find Exp Clin Pharmacol 25 (9): 713-6. http://www.ncbi.nlm.nih.gov/pubmed/14685298.

Burk, U., J. Schubert, U. Wellner, O. Schmalhofer, E. Vincan, S. Spaderna and T. Brabletz. 2008. "A reciprocal repression between ZEB1 and members of the miR-200 family

Reference List 200 promotes EMT and invasion in cancer cells." EMBO Rep 9 (6): 582-9. http://www.ncbi.nlm.nih.gov/pubmed/18483486. doi: 10.1038/embor.2008.74.

Cano, A., M. A. Perez-Moreno, I. Rodrigo, A. Locascio, M. J. Blanco, M. G. del Barrio, F. Portillo and M. A. Nieto. 2000. "The transcription factor snail controls epithelial- mesenchymal transitions by repressing E-cadherin expression." Nat Cell Biol 2 (2): 76-83. http://www.ncbi.nlm.nih.gov/pubmed/10655586. doi: 10.1038/35000025.

Caraci, F., E. Gili, M. Calafiore, M. Failla, C. La Rosa, N. Crimi, M. A. Sortino, F. Nicoletti, A. Copani and C. Vancheri. 2008. "TGF-beta1 targets the GSK-3beta/beta-catenin pathway via ERK activation in the transition of human lung fibroblasts into myofibroblasts." Pharmacol Res 57 (4): 274-82. http://www.ncbi.nlm.nih.gov/pubmed/18346908. doi: 10.1016/j.phrs.2008.02.001.

Carver, E. A., R. Jiang, Y. Lan, K. F. Oram and T. Gridley. 2001. "The mouse snail gene encodes a key regulator of the epithelial-mesenchymal transition." Mol Cell Biol 21 (23): 8184-8. http://www.ncbi.nlm.nih.gov/pubmed/11689706. doi: 10.1128/MCB.21.23.8184-8188.2001.

Casas, E., J. Kim, A. Bendesky, L. Ohno-Machado, C. J. Wolfe and J. Yang. 2011. "Snail2 is an essential mediator of Twist1-induced epithelial mesenchymal transition and metastasis." Cancer Res 71 (1): 245-54. http://www.ncbi.nlm.nih.gov/pubmed/21199805. doi: 10.1158/0008-5472.CAN- 10-2330.

Celia-Terrassa, T., O. Meca-Cortes, F. Mateo, A. M. de Paz, N. Rubio, A. Arnal-Estape, B. J. Ell, et al. 2012. "Epithelial-mesenchymal transition can suppress major attributes of human epithelial tumor-initiating cells." J Clin Invest 122 (5): 1849-68. http://www.ncbi.nlm.nih.gov/pubmed/22505459. doi: 10.1172/JCI59218.

Chaffer, C. L., N. D. Marjanovic, T. Lee, G. Bell, C. G. Kleer, F. Reinhardt, A. C. D'Alessio, R. A. Young and R. A. Weinberg. 2013. "Poised chromatin at the ZEB1 promoter enables breast cancer cell plasticity and enhances tumorigenicity." Cell 154 (1): 61-74. http://www.ncbi.nlm.nih.gov/pubmed/23827675. doi: 10.1016/j.cell.2013.06.005.

Chaffer, C. L. and R. A. Weinberg. 2011. "A perspective on cancer cell metastasis." Science 331 (6024): 1559-64. http://www.ncbi.nlm.nih.gov/pubmed/21436443. doi: 10.1126/science.1203543.

Chandran, U. R., C. Ma, R. Dhir, M. Bisceglia, M. Lyons-Weiler, W. Liang, G. Michalopoulos, M. Becich and F. A. Monzon. 2007. "Gene expression profiles of prostate cancer reveal involvement of multiple molecular pathways in the metastatic process." BMC Cancer 7: 64. http://www.ncbi.nlm.nih.gov/pubmed/17430594. doi: 10.1186/1471-2407-7-64.

Reference List 201

Chen, G., N. Shukeir, A. Potti, K. Sircar, A. Aprikian, D. Goltzman and S. A. Rabbani. 2004. "Up-regulation of Wnt-1 and beta-catenin production in patients with advanced metastatic prostate carcinoma: potential pathogenetic and prognostic implications." Cancer 101 (6): 1345-56. http://www.ncbi.nlm.nih.gov/pubmed/15316903. doi: 10.1002/cncr.20518.

Chen, L., D. L. Gibbons, S. Goswami, M. A. Cortez, Y. H. Ahn, L. A. Byers, X. Zhang, et al. 2014. "Metastasis is regulated via microRNA-200/ZEB1 axis control of tumour cell PD-L1 expression and intratumoral immunosuppression." Nat Commun 5: 5241. http://www.ncbi.nlm.nih.gov/pubmed/25348003. doi: 10.1038/ncomms6241.

Cheng, G. Z., J. Chan, Q. Wang, W. Zhang, C. D. Sun and L. H. Wang. 2007. "Twist transcriptionally up-regulates AKT2 in breast cancer cells leading to increased migration, invasion, and resistance to paclitaxel." Cancer Res 67 (5): 1979-87. http://www.ncbi.nlm.nih.gov/pubmed/17332325. doi: 10.1158/0008-5472.CAN- 06-1479.

Chikaishi, Y., H. Uramoto and F. Tanaka. 2011. "The EMT status in the primary tumor does not predict postoperative recurrence or disease-free survival in lung adenocarcinoma." Anticancer Res 31 (12): 4451-6. http://www.ncbi.nlm.nih.gov/pubmed/22199314.

Cho, S. H., Y. S. Park, H. J. Kim, C. H. Kim, S. W. Lim, J. W. Huh, J. H. Lee and H. R. Kim. 2012. "CD44 enhances the epithelial-mesenchymal transition in association with colon cancer invasion." Int J Oncol 41 (1): 211-8. http://www.ncbi.nlm.nih.gov/pubmed/22552741. doi: 10.3892/ijo.2012.1453.

Christiansen, J. J. and A. K. Rajasekaran. 2006. "Reassessing epithelial to mesenchymal transition as a prerequisite for carcinoma invasion and metastasis." Cancer Res 66 (17): 8319-26. http://www.ncbi.nlm.nih.gov/pubmed/16951136. doi: 10.1158/0008-5472.CAN-06-0410.

Christoffersen, N. R., A. Silahtaroglu, U. A. Orom, S. Kauppinen and A. H. Lund. 2007. "miR- 200b mediates post-transcriptional repression of ZFHX1B." RNA 13 (8): 1172-8. http://www.ncbi.nlm.nih.gov/pubmed/17585049. doi: 10.1261/rna.586807.

Chui, M. H. 2013. "Insights into cancer metastasis from a clinicopathologic perspective: Epithelial-Mesenchymal Transition is not a necessary step." Int J Cancer 132 (7): 1487-95. http://onlinelibrary.wiley.com/store/10.1002/ijc.27745/asset/27745_ftp.pdf?v=1& t=iifdc9vg&s=6e8349940ea54d333abbeda51033f7a6d062a102. doi: 10.1002/ijc.27745.

Reference List 202 Cohen, S. J., C. J. Punt, N. Iannotti, B. H. Saidman, K. D. Sabbath, N. Y. Gabrail, J. Picus, et al. 2009. "Prognostic significance of circulating tumor cells in patients with metastatic colorectal cancer." Ann Oncol 20 (7): 1223-9. http://www.ncbi.nlm.nih.gov/pubmed/19282466. doi: 10.1093/annonc/mdn786.

Comijn, J., G. Berx, P. Vermassen, K. Verschueren, L. van Grunsven, E. Bruyneel, M. Mareel, D. Huylebroeck and F. van Roy. 2001. "The two-handed E box binding zinc finger protein SIP1 downregulates E-cadherin and induces invasion." Mol Cell 7 (6): 1267- 78. http://www.ncbi.nlm.nih.gov/pubmed/11430829.

Cooperberg, M. R., J. P. Simko, J. E. Cowan, J. E. Reid, A. Djalilvand, S. Bhatnagar, A. Gutin, et al. 2013. "Validation of a cell-cycle progression gene panel to improve risk stratification in a contemporary prostatectomy cohort." J Clin Oncol 31 (11): 1428- 34. http://www.ncbi.nlm.nih.gov/pubmed/23460710. doi: 10.1200/JCO.2012.46.4396.

Creighton, C. J., X. Li, M. Landis, J. M. Dixon, V. M. Neumeister, A. Sjolund, D. L. Rimm, et al. 2009. "Residual breast cancers after conventional therapy display mesenchymal as well as tumor-initiating features." Proc Natl Acad Sci U S A 106 (33): 13820-5. http://www.ncbi.nlm.nih.gov/pubmed/19666588. doi: 10.1073/pnas.0905718106.

Cristofanilli, M., D. F. Hayes, G. T. Budd, M. J. Ellis, A. Stopeck, J. M. Reuben, G. V. Doyle, et al. 2005. "Circulating tumor cells: a novel prognostic factor for newly diagnosed metastatic breast cancer." J Clin Oncol 23 (7): 1420-30. http://www.ncbi.nlm.nih.gov/pubmed/15735118. doi: 10.1200/JCO.2005.08.140.

Cuevas, M. E. and T. E. Lindeman. 2015. "In vitro cytotoxicity of 4'-OH-tamoxifen and estradiol in human endometrial adenocarcinoma cells HEC-1A and HEC-1B." Oncol Rep 33 (1): 464-70. http://www.ncbi.nlm.nih.gov/pubmed/25351552. doi: 10.3892/or.2014.3565.

Cunha, G. R. and B. Lung. 1978. "The possible influence of temporal factors in androgenic responsiveness of urogenital tissue recombinants from wild-type and androgen- insensitive (Tfm) mice." J Exp Zool 205 (2): 181-93. http://www.ncbi.nlm.nih.gov/pubmed/681909. doi: 10.1002/jez.1402050203.

Cuzick, J., G. P. Swanson, G. Fisher, A. R. Brothman, D. M. Berney, J. E. Reid, D. Mesher, et al. 2011. "Prognostic value of an RNA expression signature derived from cell cycle proliferation genes in patients with prostate cancer: a retrospective study." Lancet Oncol 12 (3): 245-55. http://www.ncbi.nlm.nih.gov/pubmed/21310658. doi: 10.1016/S1470-2045(10)70295-3.

D'Amico, A. V., J. Manola, M. Loffredo, A. A. Renshaw, A. DellaCroce and P. W. Kantoff. 2004. "6-month androgen suppression plus radiation therapy vs radiation therapy

Reference List 203 alone for patients with clinically localized prostate cancer: a randomized controlled trial." JAMA 292 (7): 821-7. http://www.ncbi.nlm.nih.gov/pubmed/15315996. doi: 10.1001/jama.292.7.821.

Darnell, J. E., Jr., I. M. Kerr and G. R. Stark. 1994. "Jak-STAT pathways and transcriptional activation in response to IFNs and other extracellular signaling proteins." Science 264 (5164): 1415-21. http://www.ncbi.nlm.nih.gov/pubmed/8197455.

Das, R., P. A. Gregory, B. G. Hollier, W. D. Tilley and L. A. Selth. 2014. "Epithelial plasticity in prostate cancer: principles and clinical perspectives." Trends Mol Med. http://www.ncbi.nlm.nih.gov/pubmed/25262538. doi: 10.1016/j.molmed.2014.09.004.

Dawson, H. and A. Lugli. 2015. "Molecular and pathogenetic aspects of tumor budding in colorectal cancer." Front Med (Lausanne) 2: 11. http://www.ncbi.nlm.nih.gov/pubmed/25806371. doi: 10.3389/fmed.2015.00011.

de Bono, J. S., H. I. Scher, R. B. Montgomery, C. Parker, M. C. Miller, H. Tissing, G. V. Doyle, L. W. Terstappen, K. J. Pienta and D. Raghavan. 2008. "Circulating tumor cells predict survival benefit from treatment in metastatic castration-resistant prostate cancer." Clin Cancer Res 14 (19): 6302-9. http://www.ncbi.nlm.nih.gov/pubmed/18829513. doi: 10.1158/1078-0432.CCR-08- 0872.

De Craene, B. and G. Berx. 2013. "Regulatory networks defining EMT during cancer initiation and progression." Nat Rev Cancer 13 (2): 97-110. http://www.ncbi.nlm.nih.gov/pubmed/23344542. doi: 10.1038/nrc3447.

de Herreros, A. G., S. Peiro, M. Nassour and P. Savagner. 2010. "Snail family regulation and epithelial mesenchymal transitions in breast cancer progression." J Mammary Gland Biol Neoplasia 15 (2): 135-47. http://www.ncbi.nlm.nih.gov/pubmed/20455012. doi: 10.1007/s10911-010-9179- 8.

Debnath, J., S. K. Muthuswamy and J. S. Brugge. 2003. "Morphogenesis and oncogenesis of MCF-10A mammary epithelial acini grown in three-dimensional basement membrane cultures." Methods 30 (3): 256-68. http://www.ncbi.nlm.nih.gov/pubmed/12798140.

Denham, J. W., A. Steigler, D. S. Lamb, D. Joseph, S. Turner, J. Matthews, C. Atkinson, et al. 2011. "Short-term neoadjuvant androgen deprivation and radiotherapy for locally advanced prostate cancer: 10-year data from the TROG 96.01 randomised trial." Lancet Oncol 12 (5): 451-9. http://www.ncbi.nlm.nih.gov/pubmed/21440505. doi: 10.1016/S1470-2045(11)70063-8.

Reference List 204

Dominguez, D., B. Montserrat-Sentis, A. Virgos-Soler, S. Guaita, J. Grueso, M. Porta, I. Puig, J. Baulida, C. Franci and A. Garcia de Herreros. 2003. "Phosphorylation regulates the subcellular location and activity of the snail transcriptional repressor." Mol Cell Biol 23 (14): 5078-89. http://www.ncbi.nlm.nih.gov/pubmed/12832491.

Du, C., C. Zhang, S. Hassan, M. H. Biswas and K. C. Balaji. 2010. "Protein kinase D1 suppresses epithelial-to-mesenchymal transition through phosphorylation of snail." Cancer Res 70 (20): 7810-9. http://www.ncbi.nlm.nih.gov/pubmed/20940406. doi: 10.1158/0008-5472.CAN-09-4481.

Dupuy, F., S. Tabaries, S. Andrzejewski, Z. Dong, J. Blagih, M. G. Annis, A. Omeroglu, et al. 2015. "PDK1-Dependent Metabolic Reprogramming Dictates Metastatic Potential in Breast Cancer." Cell Metab 22 (4): 577-89. http://www.ncbi.nlm.nih.gov/pubmed/26365179. doi: 10.1016/j.cmet.2015.08.007.

Eckert, M. A., T. M. Lwin, A. T. Chang, J. Kim, E. Danis, L. Ohno-Machado and J. Yang. 2011. "Twist1-induced invadopodia formation promotes tumor metastasis." Cancer Cell 19 (3): 372-86. http://www.ncbi.nlm.nih.gov/pubmed/21397860. doi: 10.1016/j.ccr.2011.01.036.

Eger, A., K. Aigner, S. Sonderegger, B. Dampier, S. Oehler, M. Schreiber, G. Berx, A. Cano, H. Beug and R. Foisner. 2005. "DeltaEF1 is a transcriptional repressor of E-cadherin and regulates epithelial plasticity in breast cancer cells." Oncogene 24 (14): 2375- 85. http://www.ncbi.nlm.nih.gov/pubmed/15674322. doi: 10.1038/sj.onc.1208429.

Eger, A., A. Stockinger, B. Schaffhauser, H. Beug and R. Foisner. 2000. "Epithelial mesenchymal transition by c-Fos estrogen receptor activation involves nuclear translocation of beta-catenin and upregulation of beta-catenin/lymphoid enhancer binding factor-1 transcriptional activity." J Cell Biol 148 (1): 173-88. http://www.ncbi.nlm.nih.gov/pubmed/10629227.

Ellenbroek, S. I. and J. van Rheenen. 2014. "Imaging hallmarks of cancer in living mice." Nat Rev Cancer 14 (6): 406-18. http://www.ncbi.nlm.nih.gov/pubmed/24854083. doi: 10.1038/nrc3742.

Ellenrieder, V., S. F. Hendler, W. Boeck, T. Seufferlein, A. Menke, C. Ruhland, G. Adler and T. M. Gress. 2001. "Transforming growth factor beta1 treatment leads to an epithelial-mesenchymal transdifferentiation of pancreatic cancer cells requiring extracellular signal-regulated kinase 2 activation." Cancer Res 61 (10): 4222-8. http://www.ncbi.nlm.nih.gov/pubmed/11358848.

Reference List 205 Emadi Baygi, M., Z. S. Soheili, F. Essmann, A. Deezagi, R. Engers, W. Goering and W. A. Schulz. 2010. "Slug/SNAI2 regulates cell proliferation and invasiveness of metastatic prostate cancer cell lines." Tumour Biol 31 (4): 297-307. http://www.ncbi.nlm.nih.gov/pubmed/20506051. doi: 10.1007/s13277-010-0037- 5.

Erho, N., A. Crisan, I. A. Vergara, A. P. Mitra, M. Ghadessi, C. Buerki, E. J. Bergstralh, et al. 2013. "Discovery and validation of a prostate cancer genomic classifier that predicts early metastasis following radical prostatectomy." PLoS One 8 (6): e66855. http://www.ncbi.nlm.nih.gov/pubmed/23826159. doi: 10.1371/journal.pone.0066855.

Fan, F., S. Samuel, K. W. Evans, J. Lu, L. Xia, Y. Zhou, E. Sceusi, et al. 2012. "Overexpression of snail induces epithelial-mesenchymal transition and a cancer stem cell-like phenotype in human colorectal cancer cells." Cancer Med 1 (1): 5-16. http://www.ncbi.nlm.nih.gov/pubmed/23342249. doi: 10.1002/cam4.4.

Fang, X., Y. Cai, J. Liu, Z. Wang, Q. Wu, Z. Zhang, C. J. Yang, L. Yuan and G. Ouyang. 2011. "Twist2 contributes to breast cancer progression by promoting an epithelial- mesenchymal transition and cancer stem-like cell self-renewal." Oncogene 30 (47): 4707-20. http://www.ncbi.nlm.nih.gov/pubmed/21602879. doi: 10.1038/onc.2011.181.

Fischer, K. R., A. Durrans, S. Lee, J. Sheng, F. Li, S. T. Wong, H. Choi, et al. 2015. "Epithelial- to-mesenchymal transition is not required for lung metastasis but contributes to chemoresistance." Nature 527 (7579): 472-6. http://www.nature.com/nature/journal/v527/n7579/pdf/nature15748.pdf. doi: 10.1038/nature15748.

Franci, C., M. Gallen, F. Alameda, T. Baro, M. Iglesias, I. Virtanen and A. Garcia de Herreros. 2009. "Snail1 protein in the stroma as a new putative prognosis marker for colon tumours." PLoS One 4 (5): e5595. http://www.ncbi.nlm.nih.gov/pubmed/19440385. doi: 10.1371/journal.pone.0005595.

Franco, H. L., J. Casasnovas, J. R. Rodriguez-Medina and C. L. Cadilla. 2011. "Redundant or separate entities?--roles of Twist1 and Twist2 as molecular switches during gene transcription." Nucleic Acids Res 39 (4): 1177-86. http://www.ncbi.nlm.nih.gov/pubmed/20935057. doi: 10.1093/nar/gkq890.

Gao, D., N. Joshi, H. Choi, S. Ryu, M. Hahn, R. Catena, H. Sadik, et al. 2012. "Myeloid progenitor cells in the premetastatic lung promote metastases by inducing mesenchymal to epithelial transition." Cancer Res 72 (6): 1384-94. http://www.ncbi.nlm.nih.gov/pubmed/22282653. doi: 10.1158/0008-5472.CAN- 11-2905.

Reference List 206

Garber, K. 2008. "Epithelial-to-mesenchymal transition is important to metastasis, but questions remain." J Natl Cancer Inst 100 (4): 232-3, 239. http://jnci.oxfordjournals.org/content/100/4/232.full.pdf. doi: 10.1093/jnci/djn032.

Gartrell, B. A. and F. Saad. 2014. "Managing bone metastases and reducing skeletal related events in prostate cancer." Nat Rev Clin Oncol 11 (6): 335-45. http://www.ncbi.nlm.nih.gov/pubmed/24821212. doi: 10.1038/nrclinonc.2014.70.

Gay, L. J. and B. Felding-Habermann. 2011. "Contribution of platelets to tumour metastasis." Nat Rev Cancer 11 (2): 123-34. http://www.ncbi.nlm.nih.gov/pubmed/21258396. doi: 10.1038/nrc3004.

Gene Ontology, Consortium. 2015. "Gene Ontology Consortium: going forward." Nucleic Acids Res 43 (Database issue): D1049-56. http://www.ncbi.nlm.nih.gov/pubmed/25428369. doi: 10.1093/nar/gku1179.

Gheldof, A., P. Hulpiau, F. van Roy, B. De Craene and G. Berx. 2012. "Evolutionary functional analysis and molecular regulation of the ZEB transcription factors." Cell Mol Life Sci 69 (15): 2527-41. http://www.ncbi.nlm.nih.gov/pubmed/22349261. doi: 10.1007/s00018-012-0935-3.

Giampieri, S., C. Manning, S. Hooper, L. Jones, C. S. Hill and E. Sahai. 2009. "Localized and reversible TGFbeta signalling switches breast cancer cells from cohesive to single cell motility." Nat Cell Biol 11 (11): 1287-96. https://www.ncbi.nlm.nih.gov/pubmed/19838175. doi: 10.1038/ncb1973.

Gilkes, D. M., G. L. Semenza and D. Wirtz. 2014. "Hypoxia and the extracellular matrix: drivers of tumour metastasis." Nat Rev Cancer 14 (6): 430-9. http://www.ncbi.nlm.nih.gov/pubmed/24827502. doi: 10.1038/nrc3726.

Gilles, C., M. Polette, M. Mestdagt, B. Nawrocki-Raby, P. Ruggeri, P. Birembaut and J. M. Foidart. 2003. "Transactivation of vimentin by beta-catenin in human breast cancer cells." Cancer Res 63 (10): 2658-64. http://www.ncbi.nlm.nih.gov/pubmed/12750294.

Gingrich, J. R., R. J. Barrios, R. A. Morton, B. F. Boyce, F. J. DeMayo, M. J. Finegold, R. Angelopoulou, J. M. Rosen and N. M. Greenberg. 1996. "Metastatic prostate cancer in a transgenic mouse." Cancer Res 56 (18): 4096-102. http://www.ncbi.nlm.nih.gov/pubmed/8797572.

Reference List 207 Glinsky, G. V., A. B. Glinskii, A. J. Stephenson, R. M. Hoffman and W. L. Gerald. 2004. "Gene expression profiling predicts clinical outcome of prostate cancer." J Clin Invest 113 (6): 913-23. http://www.ncbi.nlm.nih.gov/pubmed/15067324. doi: 10.1172/JCI20032.

Gomis, R. R., C. Alarcon, C. Nadal, C. Van Poznak and J. Massague. 2006. "C/EBPbeta at the core of the TGFbeta cytostatic response and its evasion in metastatic breast cancer cells." Cancer Cell 10 (3): 203-14. http://www.ncbi.nlm.nih.gov/pubmed/16959612. doi: 10.1016/j.ccr.2006.07.019.

Gonzalez, D. M. and D. Medici. 2014. "Signaling mechanisms of the epithelial-mesenchymal transition." Sci Signal 7 (344): re8. http://www.ncbi.nlm.nih.gov/pubmed/25249658. doi: 10.1126/scisignal.2005189.

Gorges, T. M., I. Tinhofer, M. Drosch, L. Rose, T. M. Zollner, T. Krahn and O. von Ahsen. 2012. "Circulating tumour cells escape from EpCAM-based detection due to epithelial-to-mesenchymal transition." BMC Cancer 12: 178. http://www.ncbi.nlm.nih.gov/pubmed/22591372. doi: 10.1186/1471-2407-12-178.

Goscinski, M. A., R. Xu, F. Zhou, J. Wang, H. Yang, R. Huang, Y. Li, et al. 2015. "Nuclear, cytoplasmic, and stromal expression of ZEB1 in squamous and small cell carcinoma of the esophagus." APMIS 123 (12): 1040-7. http://www.ncbi.nlm.nih.gov/pubmed/26547370. doi: 10.1111/apm.12473.

Graff, J. R., J. G. Herman, R. G. Lapidus, H. Chopra, R. Xu, D. F. Jarrard, W. B. Isaacs, P. M. Pitha, N. E. Davidson and S. B. Baylin. 1995. "E-cadherin expression is silenced by DNA hypermethylation in human breast and prostate carcinomas." Cancer Res 55 (22): 5195-9. http://www.ncbi.nlm.nih.gov/pubmed/7585573.

Graham, T. R., H. E. Zhau, V. A. Odero-Marah, A. O. Osunkoya, K. S. Kimbro, M. Tighiouart, T. Liu, J. W. Simons and R. M. O'Regan. 2008. "Insulin-like growth factor-I- dependent up-regulation of ZEB1 drives epithelial-to-mesenchymal transition in human prostate cancer cells." Cancer Res 68 (7): 2479-88. http://www.ncbi.nlm.nih.gov/pubmed/18381457. doi: 10.1158/0008-5472.CAN- 07-2559.

Grant, C. M. and N. Kyprianou. 2013. "Epithelial mesenchymal transition (EMT) in prostate growth and tumor progression." Transl Androl Urol 2 (3): 202-211. http://www.ncbi.nlm.nih.gov/pubmed/25346895. doi: 10.3978/j.issn.2223- 4683.2013.09.04.

Grasso, C. S., Y. M. Wu, D. R. Robinson, X. Cao, S. M. Dhanasekaran, A. P. Khan, M. J. Quist, et al. 2012. "The mutational landscape of lethal castration-resistant prostate

Reference List 208 cancer." Nature 487 (7406): 239-43. http://www.ncbi.nlm.nih.gov/pubmed/22722839. doi: 10.1038/nature11125.

Gravdal, K., O. J. Halvorsen, S. A. Haukaas and L. A. Akslen. 2007. "A switch from E-cadherin to N-cadherin expression indicates epithelial to mesenchymal transition and is of strong and independent importance for the progress of prostate cancer." Clin Cancer Res 13 (23): 7003-11. http://www.ncbi.nlm.nih.gov/pubmed/18056176. doi: 10.1158/1078-0432.CCR-07-1263.

Greenburg, G. and E. D. Hay. 1982. "Epithelia suspended in collagen gels can lose polarity and express characteristics of migrating mesenchymal cells." J Cell Biol 95 (1): 333- 9. http://www.ncbi.nlm.nih.gov/pubmed/7142291.

Gregory, P. A., A. G. Bert, E. L. Paterson, S. C. Barry, A. Tsykin, G. Farshid, M. A. Vadas, Y. Khew-Goodall and G. J. Goodall. 2008. "The miR-200 family and miR-205 regulate epithelial to mesenchymal transition by targeting ZEB1 and SIP1." Nat Cell Biol 10 (5): 593-601. http://www.ncbi.nlm.nih.gov/pubmed/18376396. doi: 10.1038/ncb1722.

Grooteclaes, M. L. and S. M. Frisch. 2000. "Evidence for a function of CtBP in epithelial gene regulation and anoikis." Oncogene 19 (33): 3823-8. http://www.ncbi.nlm.nih.gov/pubmed/10949939. doi: 10.1038/sj.onc.1203721.

Guaita, S., I. Puig, C. Franci, M. Garrido, D. Dominguez, E. Batlle, E. Sancho, S. Dedhar, A. G. De Herreros and J. Baulida. 2002. "Snail induction of epithelial to mesenchymal transition in tumor cells is accompanied by MUC1 repression and ZEB1 expression." J Biol Chem 277 (42): 39209-16. http://www.ncbi.nlm.nih.gov/pubmed/12161443. doi: 10.1074/jbc.M206400200.

Guo, Y. and N. Kyprianou. 1998. "Overexpression of transforming growth factor (TGF) beta1 type II receptor restores TGF-beta1 sensitivity and signaling in human prostate cancer cells." Cell Growth Differ 9 (2): 185-93. http://www.ncbi.nlm.nih.gov/pubmed/9486855.

Gwak, J. M., H. J. Kim, E. J. Kim, Y. R. Chung, S. Yun, A. N. Seo, H. J. Lee and S. Y. Park. 2014. "MicroRNA-9 is associated with epithelial-mesenchymal transition, breast cancer stem cell phenotype, and tumor progression in breast cancer." Breast Cancer Res Treat 147 (1): 39-49. http://www.ncbi.nlm.nih.gov/pubmed/25086633. doi: 10.1007/s10549-014-3069-5.

Haider, M., X. Zhang, I. Coleman, N. Ericson, L. D. True, H. M. Lam, L. G. Brown, et al. 2015. "Epithelial mesenchymal-like transition occurs in a subset of cells in castration resistant prostate cancer bone metastases." Clin Exp Metastasis.

Reference List 209 http://www.ncbi.nlm.nih.gov/pubmed/26667932. doi: 10.1007/s10585-015-9773- 7.

Hajra, K. M., D. Y. Chen and E. R. Fearon. 2002. "The SLUG zinc-finger protein represses E- cadherin in breast cancer." Cancer Res 62 (6): 1613-8. http://www.ncbi.nlm.nih.gov/pubmed/11912130.

Hamamori, Y., V. Sartorelli, V. Ogryzko, P. L. Puri, H. Y. Wu, J. Y. Wang, Y. Nakatani and L. Kedes. 1999. "Regulation of histone acetyltransferases p300 and PCAF by the bHLH protein twist and adenoviral oncoprotein E1A." Cell 96 (3): 405-13. http://www.ncbi.nlm.nih.gov/pubmed/10025406.

Hanahan, D. and R. A. Weinberg. 2011. "Hallmarks of cancer: the next generation." Cell 144 (5): 646-74. http://www.ncbi.nlm.nih.gov/pubmed/21376230. doi: 10.1016/j.cell.2011.02.013.

Harder, J. L., E. L. Whiteman, J. N. Pieczynski, C. J. Liu and B. Margolis. 2012. "Snail destabilizes cell surface Crumbs3a." Traffic 13 (8): 1170-85. http://www.ncbi.nlm.nih.gov/pubmed/22554228. doi: 10.1111/j.1600- 0854.2012.01376.x.

Harma, V., J. Virtanen, R. Makela, A. Happonen, J. P. Mpindi, M. Knuuttila, P. Kohonen, J. Lotjonen, O. Kallioniemi and M. Nees. 2010. "A comprehensive panel of three- dimensional models for studies of prostate cancer growth, invasion and drug responses." PLoS One 5 (5): e10431. http://www.ncbi.nlm.nih.gov/pubmed/20454659. doi: 10.1371/journal.pone.0010431.

Hartwell, L. H. and M. B. Kastan. 1994. "Cell cycle control and cancer." Science 266 (5192): 1821-8. http://www.ncbi.nlm.nih.gov/pubmed/7997877.

Haslehurst, A. M., M. Koti, M. Dharsee, P. Nuin, K. Evans, J. Geraci, T. Childs, et al. 2012. "EMT transcription factors snail and slug directly contribute to cisplatin resistance in ovarian cancer." BMC Cancer 12: 91. http://www.ncbi.nlm.nih.gov/pubmed/22429801. doi: 10.1186/1471-2407-12-91.

Hayward, S. W. and G. R. Cunha. 2000. "The prostate: development and physiology." Radiol Clin North Am 38 (1): 1-14. http://www.ncbi.nlm.nih.gov/pubmed/10664663.

Hazan, R. B., L. Kang, B. P. Whooley and P. I. Borgen. 1997. "N-cadherin promotes adhesion between invasive breast cancer cells and the stroma." Cell Adhes Commun 4 (6): 399-411. http://www.ncbi.nlm.nih.gov/pubmed/9177902.

Reference List 210

Hedley, B. D. and A. F. Chambers. 2009. "Tumor dormancy and metastasis." Adv Cancer Res 102: 67-101. http://www.ncbi.nlm.nih.gov/pubmed/19595307. doi: 10.1016/S0065-230X(09)02003-X.

Heeboll, S., M. Borre, P. D. Ottosen, L. Dyrskjot, T. F. Orntoft and N. Torring. 2009. "Snail1 is over-expressed in prostate cancer." APMIS 117 (3): 196-204. http://www.ncbi.nlm.nih.gov/pubmed/19245592. doi: 10.1111/j.1600- 0463.2008.00007.x.

Heerboth, S., G. Housman, M. Leary, M. Longacre, S. Byler, K. Lapinska, A. Willbanks and S. Sarkar. 2015. "EMT and tumor metastasis." Clin Transl Med 4: 6. http://www.ncbi.nlm.nih.gov/pubmed/25852822. doi: 10.1186/s40169-015-0048- 3.

Herranz, N., D. Pasini, V. M. Diaz, C. Franci, A. Gutierrez, N. Dave, M. Escriva, et al. 2008. "Polycomb complex 2 is required for E-cadherin repression by the Snail1 transcription factor." Mol Cell Biol 28 (15): 4772-81. http://www.ncbi.nlm.nih.gov/pubmed/18519590. doi: 10.1128/MCB.00323-08.

Hieronymus, H. and C. L. Sawyers. 2012. "Traversing the genomic landscape of prostate cancer from diagnosis to death." Nat Genet 44 (6): 613-4. https://www.ncbi.nlm.nih.gov/pubmed/22641202. doi: 10.1038/ng.2301.

Higano, C. S. 2004. "Understanding treatments for bone loss and bone metastases in patients with prostate cancer: a practical review and guide for the clinician." Urol Clin North Am 31 (2): 331-52. http://www.ncbi.nlm.nih.gov/pubmed/15123412. doi: 10.1016/j.ucl.2004.01.001.

Hiley, C., E. C. de Bruin, N. McGranahan and C. Swanton. 2014. "Deciphering intratumor heterogeneity and temporal acquisition of driver events to refine precision medicine." Genome Biol 15 (8): 453. http://www.ncbi.nlm.nih.gov/pubmed/25222836. doi: 10.1186/s13059-014-0453- 8.

Holzapfel, B. M., F. Wagner, D. Loessner, N. P. Holzapfel, L. Thibaudeau, R. Crawford, M. T. Ling, J. A. Clements, P. J. Russell and D. W. Hutmacher. 2014. "Species-specific homing mechanisms of human prostate cancer metastasis in tissue engineered bone." Biomaterials 35 (13): 4108-15. http://www.ncbi.nlm.nih.gov/pubmed/24534484. doi: 10.1016/j.biomaterials.2014.01.062.

Horoszewicz, J. S., S. S. Leong, T. M. Chu, Z. L. Wajsman, M. Friedman, L. Papsidero, U. Kim, et al. 1980. "The LNCaP cell line--a new model for studies on human prostatic

Reference List 211 carcinoma." Prog Clin Biol Res 37: 115-32. http://www.ncbi.nlm.nih.gov/pubmed/7384082.

Horoszewicz, J. S., S. S. Leong, E. Kawinski, J. P. Karr, H. Rosenthal, T. M. Chu, E. A. Mirand and G. P. Murphy. 1983. "LNCaP model of human prostatic carcinoma." Cancer Res 43 (4): 1809-18. http://www.ncbi.nlm.nih.gov/pubmed/6831420.

Horvath, L. G., S. M. Henshall, C. S. Lee, J. G. Kench, D. Golovsky, P. C. Brenner, G. F. O'Neill, et al. 2005. "Lower levels of nuclear beta-catenin predict for a poorer prognosis in localized prostate cancer." Int J Cancer 113 (3): 415-22. http://www.ncbi.nlm.nih.gov/pubmed/15455387. doi: 10.1002/ijc.20599.

Hoshino, H., N. Miyoshi, K. Nagai, Y. Tomimaru, H. Nagano, M. Sekimoto, Y. Doki, M. Mori and H. Ishii. 2009. "Epithelial-mesenchymal transition with expression of SNAI1- induced chemoresistance in colorectal cancer." Biochem Biophys Res Commun 390 (3): 1061-5. http://www.ncbi.nlm.nih.gov/pubmed/19861116. doi: 10.1016/j.bbrc.2009.10.117.

Hou, Z., H. Peng, K. Ayyanathan, K. P. Yan, E. M. Langer, G. D. Longmore and F. J. Rauscher, 3rd. 2008. "The LIM protein AJUBA recruits protein arginine methyltransferase 5 to mediate SNAIL-dependent transcriptional repression." Mol Cell Biol 28 (10): 3198- 207. http://www.ncbi.nlm.nih.gov/pubmed/18347060. doi: 10.1128/MCB.01435- 07.

Huang, R. Y., M. K. Wong, T. Z. Tan, K. T. Kuay, A. H. Ng, V. Y. Chung, Y. S. Chu, et al. 2013. "An EMT spectrum defines an anoikis-resistant and spheroidogenic intermediate mesenchymal state that is sensitive to e-cadherin restoration by a src-kinase inhibitor, saracatinib (AZD0530)." Cell Death Dis 4: e915. http://www.ncbi.nlm.nih.gov/pubmed/24201814. doi: 10.1038/cddis.2013.442.

Hugo, H. J., L. Pereira, R. Suryadinata, Y. Drabsch, T. J. Gonda, N. P. Gunasinghe, C. Pinto, et al. 2013. "Direct repression of MYB by ZEB1 suppresses proliferation and epithelial gene expression during epithelial-to-mesenchymal transition of breast cancer cells." Breast Cancer Res 15 (6): R113. http://www.ncbi.nlm.nih.gov/pubmed/24283570. doi: 10.1186/bcr3580.

Hynes, R. O. and K. M. Yamada. 1982. "Fibronectins: multifunctional modular glycoproteins." J Cell Biol 95 (2 Pt 1): 369-77. http://www.ncbi.nlm.nih.gov/pubmed/6128348.

Imai, T., A. Horiuchi, T. Shiozawa, R. Osada, N. Kikuchi, S. Ohira, K. Oka and I. Konishi. 2004. "Elevated expression of E-cadherin and alpha-, beta-, and gamma-catenins in metastatic lesions compared with primary epithelial ovarian carcinomas." Hum Pathol 35 (12): 1469-76. http://www.ncbi.nlm.nih.gov/pubmed/15619205.

Reference List 212

Inoue, A., M. G. Seidel, W. Wu, S. Kamizono, A. A. Ferrando, R. T. Bronson, H. Iwasaki, et al. 2002. "Slug, a highly conserved zinc finger transcriptional repressor, protects hematopoietic progenitor cells from radiation-induced apoptosis in vivo." Cancer Cell 2 (4): 279-88. http://www.ncbi.nlm.nih.gov/pubmed/12398892.

Jaggi, M., S. L. Johansson, J. J. Baker, L. M. Smith, A. Galich and K. C. Balaji. 2005. "Aberrant expression of E-cadherin and beta-catenin in human prostate cancer." Urol Oncol 23 (6): 402-6. http://www.ncbi.nlm.nih.gov/pubmed/16301117. doi: 10.1016/j.urolonc.2005.03.024.

Jang, M. H., H. J. Kim, E. J. Kim, Y. R. Chung and S. Y. Park. 2015. "Expression of epithelial- mesenchymal transition-related markers in triple-negative breast cancer: ZEB1 as a potential biomarker for poor clinical outcome." Hum Pathol 46 (9): 1267-74. http://www.ncbi.nlm.nih.gov/pubmed/26170011. doi: 10.1016/j.humpath.2015.05.010.

Javaid, S., J. Zhang, E. Anderssen, J. C. Black, B. S. Wittner, K. Tajima, D. T. Ting, et al. 2013. "Dynamic chromatin modification sustains epithelial-mesenchymal transition following inducible expression of Snail-1." Cell Rep 5 (6): 1679-89. http://www.ncbi.nlm.nih.gov/pubmed/24360956. doi: 10.1016/j.celrep.2013.11.034.

Jeter, C. R., B. Liu, X. Liu, X. Chen, C. Liu, T. Calhoun-Davis, J. Repass, H. Zaehres, J. J. Shen and D. G. Tang. 2011. "NANOG promotes cancer stem cell characteristics and prostate cancer resistance to androgen deprivation." Oncogene 30 (36): 3833-45. http://www.ncbi.nlm.nih.gov/pubmed/21499299. doi: 10.1038/onc.2011.114.

Jiang, L., R. Deberardinis and D. A. Boothman. 2015. "The cancer cell 'energy grid': TGF- beta1 signaling coordinates metabolism for migration." Mol Cell Oncol 2 (3): e981994. http://www.ncbi.nlm.nih.gov/pubmed/27308459. doi: 10.4161/23723556.2014.981994.

Jo, M., R. D. Lester, V. Montel, B. Eastman, S. Takimoto and S. L. Gonias. 2009. "Reversibility of epithelial-mesenchymal transition (EMT) induced in breast cancer cells by activation of urokinase receptor-dependent cell signaling." J Biol Chem 284 (34): 22825-33. http://www.ncbi.nlm.nih.gov/pubmed/19546228. doi: 10.1074/jbc.M109.023960.

Joosse, S. A., J. Hannemann, J. Spotter, A. Bauche, A. Andreas, V. Muller and K. Pantel. 2012. "Changes in keratin expression during metastatic progression of breast cancer: impact on the detection of circulating tumor cells." Clin Cancer Res 18 (4): 993-1003. http://www.ncbi.nlm.nih.gov/pubmed/22228641. doi: 10.1158/1078- 0432.CCR-11-2100.

Reference List 213

Jorda, M., D. Olmeda, A. Vinyals, E. Valero, E. Cubillo, A. Llorens, A. Cano and A. Fabra. 2005. "Upregulation of MMP-9 in MDCK epithelial cell line in response to expression of the Snail transcription factor." J Cell Sci 118 (Pt 15): 3371-85. http://www.ncbi.nlm.nih.gov/pubmed/16079281. doi: 10.1242/jcs.02465.

Jordan, N. V., G. L. Johnson and A. N. Abell. 2011. "Tracking the intermediate stages of epithelial-mesenchymal transition in epithelial stem cells and cancer." Cell Cycle 10 (17): 2865-73. http://www.ncbi.nlm.nih.gov/pubmed/21862874.

Kajita, M., K. N. McClinic and P. A. Wade. 2004. "Aberrant expression of the transcription factors snail and slug alters the response to genotoxic stress." Mol Cell Biol 24 (17): 7559-66. http://www.ncbi.nlm.nih.gov/pubmed/15314165. doi: 10.1128/MCB.24.17.7559-7566.2004.

Kallakury, B. V., C. E. Sheehan, E. Winn-Deen, J. Oliver, H. A. Fisher, R. P. Kaufman, Jr. and J. S. Ross. 2001. "Decreased expression of catenins (alpha and beta), p120 CTN, and E-cadherin cell adhesion proteins and E-cadherin gene promoter methylation in prostatic adenocarcinomas." Cancer 92 (11): 2786-95. http://www.ncbi.nlm.nih.gov/pubmed/11753952.

Kalluri, R. and R. A. Weinberg. 2009. "The basics of epithelial-mesenchymal transition." J Clin Invest 119 (6): 1420-8. http://www.ncbi.nlm.nih.gov/pubmed/19487818. doi: 10.1172/JCI39104.

Kanai, Y., S. Ushijima, A. M. Hui, A. Ochiai, H. Tsuda, M. Sakamoto and S. Hirohashi. 1997. "The E-cadherin gene is silenced by CpG methylation in human hepatocellular carcinomas." Int J Cancer 71 (3): 355-9. http://www.ncbi.nlm.nih.gov/pubmed/9139867.

Karamitopoulou, E., I. Zlobec, D. Born, A. Kondi-Pafiti, P. Lykoudis, A. Mellou, K. Gennatas, B. Gloor and A. Lugli. 2013. "Tumour budding is a strong and independent prognostic factor in pancreatic cancer." Eur J Cancer 49 (5): 1032-9. http://www.ncbi.nlm.nih.gov/pubmed/23177090. doi: 10.1016/j.ejca.2012.10.022.

Kataoka, H., T. Murayama, M. Yokode, S. Mori, H. Sano, H. Ozaki, Y. Yokota, S. Nishikawa and T. Kita. 2000. "A novel snail-related transcription factor Smuc regulates basic helix-loop-helix transcription factor activities via specific E-box motifs." Nucleic Acids Res 28 (2): 626-33. http://www.ncbi.nlm.nih.gov/pubmed/10606664.

Katsuno, Y., S. Lamouille and R. Derynck. 2013. "TGF-beta signaling and epithelial- mesenchymal transition in cancer progression." Curr Opin Oncol 25 (1): 76-84. http://www.ncbi.nlm.nih.gov/pubmed/23197193. doi: 10.1097/CCO.0b013e32835b6371.

Reference List 214

Kaufmann, S. H. and W. C. Earnshaw. 2000. "Induction of apoptosis by cancer chemotherapy." Exp Cell Res 256 (1): 42-9. http://www.ncbi.nlm.nih.gov/pubmed/10739650. doi: 10.1006/excr.2000.4838.

Kawamoto, A., T. Yokoe, K. Tanaka, S. Saigusa, Y. Toiyama, H. Yasuda, Y. Inoue, C. Miki and M. Kusunoki. 2012. "Radiation induces epithelial-mesenchymal transition in colorectal cancer cells." Oncol Rep 27 (1): 51-7. http://www.ncbi.nlm.nih.gov/pubmed/21971767. doi: 10.3892/or.2011.1485.

Kenific, C. M., A. Thorburn and J. Debnath. 2010. "Autophagy and metastasis: another double-edged sword." Curr Opin Cell Biol 22 (2): 241-5. http://www.ncbi.nlm.nih.gov/pubmed/19945838. doi: 10.1016/j.ceb.2009.10.008.

Khoo, B. L., S. C. Lee, P. Kumar, T. Z. Tan, M. E. Warkiani, S. G. Ow, S. Nandi, C. T. Lim and J. P. Thiery. 2015. "Short-term expansion of breast circulating cancer cells predicts response to anti-cancer therapy." Oncotarget 6 (17): 15578-93. http://www.ncbi.nlm.nih.gov/pubmed/26008969. doi: 10.18632/oncotarget.3903.

Kim, H. J., B. C. Litzenburger, X. Cui, D. A. Delgado, B. C. Grabiner, X. Lin, M. T. Lewis, et al. 2007. "Constitutively active type I insulin-like growth factor receptor causes transformation and xenograft growth of immortalized mammary epithelial cells and is accompanied by an epithelial-to-mesenchymal transition mediated by NF- kappaB and snail." Mol Cell Biol 27 (8): 3165-75. http://www.ncbi.nlm.nih.gov/pubmed/17296734. doi: 10.1128/MCB.01315-06.

Kim, K. K., M. C. Kugler, P. J. Wolters, L. Robillard, M. G. Galvez, A. N. Brumwell, D. Sheppard and H. A. Chapman. 2006. "Alveolar epithelial cell mesenchymal transition develops in vivo during pulmonary fibrosis and is regulated by the extracellular matrix." Proc Natl Acad Sci U S A 103 (35): 13180-5. http://www.ncbi.nlm.nih.gov/pubmed/16924102. doi: 10.1073/pnas.0605669103.

Kim, R. S., A. Avivar-Valderas, Y. Estrada, P. Bragado, M. S. Sosa, J. A. Aguirre-Ghiso and J. E. Segall. 2012. "Dormancy signatures and metastasis in estrogen receptor positive and negative breast cancer." PLoS One 7 (4): e35569. http://www.ncbi.nlm.nih.gov/pubmed/22530051. doi: 10.1371/journal.pone.0035569.

Kim, T., A. Veronese, F. Pichiorri, T. J. Lee, Y. J. Jeon, S. Volinia, P. Pineau, et al. 2011. "p53 regulates epithelial-mesenchymal transition through microRNAs targeting ZEB1 and ZEB2." J Exp Med 208 (5): 875-83. http://www.ncbi.nlm.nih.gov/pubmed/21518799. doi: 10.1084/jem.20110235.

Reference List 215 Klymkowsky, M. W. and P. Savagner. 2009. "Epithelial-mesenchymal transition: a cancer researcher's conceptual friend and foe." Am J Pathol 174 (5): 1588-93. http://www.ncbi.nlm.nih.gov/pubmed/19342369. doi: 10.2353/ajpath.2009.080545.

Kong, D., S. Banerjee, A. Ahmad, Y. Li, Z. Wang, S. Sethi and F. H. Sarkar. 2010. "Epithelial to mesenchymal transition is mechanistically linked with stem cell signatures in prostate cancer cells." PLoS One 5 (8): e12445. http://www.ncbi.nlm.nih.gov/pubmed/20805998. doi: 10.1371/journal.pone.0012445.

Konigsberg, R., E. Obermayr, G. Bises, G. Pfeiler, M. Gneist, F. Wrba, M. de Santis, R. Zeillinger, M. Hudec and C. Dittrich. 2011. "Detection of EpCAM positive and negative circulating tumor cells in metastatic breast cancer patients." Acta Oncol 50 (5): 700-10. http://www.ncbi.nlm.nih.gov/pubmed/21261508. doi: 10.3109/0284186X.2010.549151.

Korpal, M. and Y. Kang. 2008. "The emerging role of miR-200 family of microRNAs in epithelial-mesenchymal transition and cancer metastasis." RNA Biol 5 (3): 115-9. http://www.ncbi.nlm.nih.gov/pubmed/19182522.

Korpal, M., E. S. Lee, G. Hu and Y. Kang. 2008. "The miR-200 family inhibits epithelial- mesenchymal transition and cancer cell migration by direct targeting of E-cadherin transcriptional repressors ZEB1 and ZEB2." J Biol Chem 283 (22): 14910-4. http://www.ncbi.nlm.nih.gov/pubmed/18411277. doi: 10.1074/jbc.C800074200.

Kroon, J., S. Kooijman, N. J. Cho, G. Storm and G. van der Pluijm. 2016. "Improving Taxane- Based Chemotherapy in Castration-Resistant Prostate Cancer." Trends Pharmacol Sci 37 (6): 451-62. https://www.ncbi.nlm.nih.gov/pubmed/27068431. doi: 10.1016/j.tips.2016.03.003.

Kuo, Y. C., C. H. Su, C. Y. Liu, T. H. Chen, C. P. Chen and H. S. Wang. 2009. "Transforming growth factor-beta induces CD44 cleavage that promotes migration of MDA-MB- 435s cells through the up-regulation of membrane type 1-matrix metalloproteinase." Int J Cancer 124 (11): 2568-76. http://www.ncbi.nlm.nih.gov/pubmed/19243022. doi: 10.1002/ijc.24263.

Kypta, R. M. and J. Waxman. 2012. "Wnt/beta-catenin signalling in prostate cancer." Nat Rev Urol 9 (8): 418-28. http://www.ncbi.nlm.nih.gov/pubmed/22710668. doi: 10.1038/nrurol.2012.116.

Labelle, M., S. Begum and R. O. Hynes. 2011. "Direct signaling between platelets and cancer cells induces an epithelial-mesenchymal-like transition and promotes metastasis."

Reference List 216 Cancer Cell 20 (5): 576-90. http://www.ncbi.nlm.nih.gov/pubmed/22094253. doi: 10.1016/j.ccr.2011.09.009.

Lamouille, S., J. Xu and R. Derynck. 2014. "Molecular mechanisms of epithelial- mesenchymal transition." Nat Rev Mol Cell Biol 15 (3): 178-96. http://www.ncbi.nlm.nih.gov/pubmed/24556840. doi: 10.1038/nrm3758.

LaTulippe, E., J. Satagopan, A. Smith, H. Scher, P. Scardino, V. Reuter and W. L. Gerald. 2002. "Comprehensive gene expression analysis of prostate cancer reveals distinct transcriptional programs associated with metastatic disease." Cancer Res 62 (15): 4499-506. http://www.ncbi.nlm.nih.gov/pubmed/12154061.

LeBleu, V. S., J. T. O'Connell, K. N. Gonzalez Herrera, H. Wikman, K. Pantel, M. C. Haigis, F. M. de Carvalho, et al. 2014. "PGC-1alpha mediates mitochondrial biogenesis and oxidative phosphorylation in cancer cells to promote metastasis." Nat Cell Biol 16 (10): 992-1003, 1-15. http://www.ncbi.nlm.nih.gov/pubmed/25241037. doi: 10.1038/ncb3039.

Ledford, H. 2011. "Cancer theory faces doubts." Nature 472 (7343): 273. http://www.ncbi.nlm.nih.gov/pubmed/21512545. doi: 10.1038/472273a.

Lee, E., S. Ha and S. K. Logan. 2015. "Divergent Androgen Receptor and Beta-Catenin Signaling in Prostate Cancer Cells." PLoS One 10 (10): e0141589. http://www.ncbi.nlm.nih.gov/pubmed/26509262. doi: 10.1371/journal.pone.0141589.

Lee, E. S., D. S. Son, S. H. Kim, J. Lee, J. Jo, J. Han, H. Kim, et al. 2008. "Prediction of recurrence-free survival in postoperative non-small cell lung cancer patients by using an integrated model of clinical information and gene expression." Clin Cancer Res 14 (22): 7397-404. http://www.ncbi.nlm.nih.gov/pubmed/19010856. doi: 10.1158/1078-0432.CCR-07-4937.

Lee, J. M., S. Dedhar, R. Kalluri and E. W. Thompson. 2006. "The epithelial-mesenchymal transition: new insights in signaling, development, and disease." J Cell Biol 172 (7): 973-81. http://www.ncbi.nlm.nih.gov/pubmed/16567498. doi: 10.1083/jcb.200601018.

Lee, K. and C. M. Nelson. 2012. "New insights into the regulation of epithelial-mesenchymal transition and tissue fibrosis." Int Rev Cell Mol Biol 294: 171-221. http://www.ncbi.nlm.nih.gov/pubmed/22364874. doi: 10.1016/B978-0-12-394305- 7.00004-5.

Reference List 217 Lee, M. Y., C. Y. Chou, M. J. Tang and M. R. Shen. 2008. "Epithelial-mesenchymal transition in cervical cancer: correlation with tumor progression, epidermal growth factor receptor overexpression, and snail up-regulation." Clin Cancer Res 14 (15): 4743-50. http://www.ncbi.nlm.nih.gov/pubmed/18676743. doi: 10.1158/1078-0432.CCR-08- 0234.

Lester, R. D., M. Jo, V. Montel, S. Takimoto and S. L. Gonias. 2007. "uPAR induces epithelial- mesenchymal transition in hypoxic breast cancer cells." J Cell Biol 178 (3): 425-36. http://www.ncbi.nlm.nih.gov/pubmed/17664334. doi: 10.1083/jcb.200701092.

Li, C. W., W. Xia, L. Huo, S. O. Lim, Y. Wu, J. L. Hsu, C. H. Chao, et al. 2012. "Epithelial- mesenchymal transition induced by TNF-alpha requires NF-kappaB-mediated transcriptional upregulation of Twist1." Cancer Res 72 (5): 1290-300. http://www.ncbi.nlm.nih.gov/pubmed/22253230. doi: 10.1158/0008-5472.CAN- 11-3123.

Li, H., X. Chen, Y. Gao, J. Wu, F. Zeng and F. Song. 2015. "XBP1 induces snail expression to promote epithelial- to-mesenchymal transition and invasion of breast cancer cells." Cell Signal 27 (1): 82-9. http://www.ncbi.nlm.nih.gov/pubmed/25280941. doi: 10.1016/j.cellsig.2014.09.018.

Li, J. and M. R. King. 2012. "Adhesion receptors as therapeutic targets for circulating tumor cells." Front Oncol 2: 79. http://www.ncbi.nlm.nih.gov/pubmed/22837985. doi: 10.3389/fonc.2012.00079.

Liang, F., W. Cao, Y. Wang, L. Li, G. Zhang and Z. Wang. 2013. "The prognostic value of tumor budding in invasive breast cancer." Pathol Res Pract 209 (5): 269-75. http://www.ncbi.nlm.nih.gov/pubmed/23561623. doi: 10.1016/j.prp.2013.01.009.

Liberzon, A., C. Birger, H. Thorvaldsdottir, M. Ghandi, J. P. Mesirov and P. Tamayo. 2015. "The Molecular Signatures Database (MSigDB) hallmark gene set collection." Cell Syst 1 (6): 417-425. http://www.ncbi.nlm.nih.gov/pubmed/26771021. doi: 10.1016/j.cels.2015.12.004.

Lim, J. and J. P. Thiery. 2012. "Epithelial-mesenchymal transitions: insights from development." Development 139 (19): 3471-86. http://www.ncbi.nlm.nih.gov/pubmed/22949611. doi: 10.1242/dev.071209.

Liu, J., B. Uygur, Z. Zhang, L. Shao, D. Romero, C. Vary, Q. Ding and W. S. Wu. 2010. "Slug inhibits proliferation of human prostate cancer cells via downregulation of cyclin D1 expression." Prostate 70 (16): 1768-77. http://www.ncbi.nlm.nih.gov/pubmed/20564361. doi: 10.1002/pros.21213.

Reference List 218 Livak, K. J. and T. D. Schmittgen. 2001. "Analysis of relative gene expression data using real- time quantitative PCR and the 2(-Delta Delta C(T)) Method." Methods 25 (4): 402-8. http://www.ncbi.nlm.nih.gov/pubmed/11846609. doi: 10.1006/meth.2001.1262.

Lo, H. W., S. C. Hsu, W. Xia, X. Cao, J. Y. Shih, Y. Wei, J. L. Abbruzzese, G. N. Hortobagyi and M. C. Hung. 2007. "Epidermal growth factor receptor cooperates with signal transducer and activator of transcription 3 to induce epithelial-mesenchymal transition in cancer cells via up-regulation of TWIST gene expression." Cancer Res 67 (19): 9066-76. http://www.ncbi.nlm.nih.gov/pubmed/17909010. doi: 10.1158/0008-5472.CAN-07-0575.

Loboda, A., M. V. Nebozhyn, J. W. Watters, C. A. Buser, P. M. Shaw, P. S. Huang, L. Van't Veer, et al. 2011. "EMT is the dominant program in human colon cancer." BMC Med Genomics 4: 9. https://www.ncbi.nlm.nih.gov/pubmed/21251323. doi: 10.1186/1755-8794-4-9.

Lorente, D. and J. S. De Bono. 2014. "Molecular alterations and emerging targets in castration resistant prostate cancer." Eur J Cancer 50 (4): 753-64. http://www.ncbi.nlm.nih.gov/pubmed/24418724. doi: 10.1016/j.ejca.2013.12.004.

Lu, Z., S. Ghosh, Z. Wang and T. Hunter. 2003. "Downregulation of caveolin-1 function by EGF leads to the loss of E-cadherin, increased transcriptional activity of beta- catenin, and enhanced tumor cell invasion." Cancer Cell 4 (6): 499-515. http://www.ncbi.nlm.nih.gov/pubmed/14706341.

Lu, Z., R. Z. Luo, Y. Lu, X. Zhang, Q. Yu, S. Khare, S. Kondo, et al. 2008. "The tumor suppressor gene ARHI regulates autophagy and tumor dormancy in human ovarian cancer cells." J Clin Invest 118 (12): 3917-29. http://www.ncbi.nlm.nih.gov/pubmed/19033662. doi: 10.1172/JCI35512.

Lugli, A., E. Karamitopoulou and I. Zlobec. 2012. "Tumour budding: a promising parameter in colorectal cancer." Br J Cancer 106 (11): 1713-7. http://www.ncbi.nlm.nih.gov/pubmed/22531633. doi: 10.1038/bjc.2012.127.

Mak, P., I. Leav, B. Pursell, D. Bae, X. Yang, C. A. Taglienti, L. M. Gouvin, V. M. Sharma and A. M. Mercurio. 2010. "ERbeta impedes prostate cancer EMT by destabilizing HIF- 1alpha and inhibiting VEGF-mediated snail nuclear localization: implications for Gleason grading." Cancer Cell 17 (4): 319-32. http://www.ncbi.nlm.nih.gov/pubmed/20385358. doi: 10.1016/j.ccr.2010.02.030.

Malek, J. A., E. Mery, Y. A. Mahmoud, E. K. Al-Azwani, L. Roger, R. Huang, E. Jouve, et al. 2011. "Copy number variation analysis of matched ovarian primary tumors and peritoneal metastasis." PLoS One 6 (12): e28561.

Reference List 219 http://www.ncbi.nlm.nih.gov/pubmed/22194851. doi: 10.1371/journal.pone.0028561.

Mani, S. A., W. Guo, M. J. Liao, E. N. Eaton, A. Ayyanan, A. Y. Zhou, M. Brooks, et al. 2008. "The epithelial-mesenchymal transition generates cells with properties of stem cells." Cell 133 (4): 704-15. http://www.ncbi.nlm.nih.gov/pubmed/18485877. doi: 10.1016/j.cell.2008.03.027.

Mann, B., M. Gelos, A. Siedow, M. L. Hanski, A. Gratchev, M. Ilyas, W. F. Bodmer, et al. 1999. "Target genes of beta-catenin-T cell-factor/lymphoid-enhancer-factor signaling in human colorectal carcinomas." Proc Natl Acad Sci U S A 96 (4): 1603-8. http://www.ncbi.nlm.nih.gov/pubmed/9990071.

Marjanovic, N. D., R. A. Weinberg and C. L. Chaffer. 2013. "Cell plasticity and heterogeneity in cancer." Clin Chem 59 (1): 168-79. http://www.ncbi.nlm.nih.gov/pubmed/23220226. doi: 10.1373/clinchem.2012.184655.

Marker, P. C., A. A. Donjacour, R. Dahiya and G. R. Cunha. 2003. "Hormonal, cellular, and molecular control of prostatic development." Dev Biol 253 (2): 165-74. http://www.ncbi.nlm.nih.gov/pubmed/12645922.

Martelotto, L. G., C. K. Ng, S. Piscuoglio, B. Weigelt and J. S. Reis-Filho. 2014. "Breast cancer intra-tumor heterogeneity." Breast Cancer Res 16 (3): 210. http://www.ncbi.nlm.nih.gov/pubmed/25928070. doi: 10.1186/bcr3658.

Martinez-Estrada, O. M., A. Culleres, F. X. Soriano, H. Peinado, V. Bolos, F. O. Martinez, M. Reina, A. Cano, M. Fabre and S. Vilaro. 2006. "The transcription factors Slug and Snail act as repressors of Claudin-1 expression in epithelial cells." Biochem J 394 (Pt 2): 449-57. http://www.ncbi.nlm.nih.gov/pubmed/16232121. doi: 10.1042/BJ20050591.

Massague, J. and R. R. Gomis. 2006. "The logic of TGFbeta signaling." FEBS Lett 580 (12): 2811-20. http://www.ncbi.nlm.nih.gov/pubmed/16678165. doi: 10.1016/j.febslet.2006.04.033.

Mauhin, V., Y. Lutz, C. Dennefeld and A. Alberga. 1993. "Definition of the DNA-binding site repertoire for the Drosophila transcription factor SNAIL." Nucleic Acids Res 21 (17): 3951-7. http://www.ncbi.nlm.nih.gov/pubmed/8371971.

Medici, D., E. D. Hay and B. R. Olsen. 2008. "Snail and Slug promote epithelial-mesenchymal transition through beta-catenin-T-cell factor-4-dependent expression of

Reference List 220 transforming growth factor-beta3." Mol Biol Cell 19 (11): 4875-87. http://www.ncbi.nlm.nih.gov/pubmed/18799618. doi: 10.1091/mbc.E08-05-0506.

Meerbrey, K. L., G. Hu, J. D. Kessler, K. Roarty, M. Z. Li, J. E. Fang, J. I. Herschkowitz, et al. 2011. "The pINDUCER lentiviral toolkit for inducible RNA interference in vitro and in vivo." Proc Natl Acad Sci U S A 108 (9): 3665-70. http://www.ncbi.nlm.nih.gov/pubmed/21307310. doi: 10.1073/pnas.1019736108.

Mejlvang, J., M. Kriajevska, C. Vandewalle, T. Chernova, A. E. Sayan, G. Berx, J. K. Mellon and E. Tulchinsky. 2007. "Direct repression of cyclin D1 by SIP1 attenuates cell cycle progression in cells undergoing an epithelial mesenchymal transition." Mol Biol Cell 18 (11): 4615-24. http://www.ncbi.nlm.nih.gov/pubmed/17855508. doi: 10.1091/mbc.E07-05-0406.

Mendez, M. G., S. Kojima and R. D. Goldman. 2010. "Vimentin induces changes in cell shape, motility, and adhesion during the epithelial to mesenchymal transition." FASEB J 24 (6): 1838-51. http://www.ncbi.nlm.nih.gov/pubmed/20097873. doi: 10.1096/fj.09-151639.

Mercado-Pimentel, M. E. and R. B. Runyan. 2007. "Multiple transforming growth factor- beta isoforms and receptors function during epithelial-mesenchymal cell transformation in the embryonic heart." Cells Tissues Organs 185 (1-3): 146-56. http://www.ncbi.nlm.nih.gov/pubmed/17587820. doi: 10.1159/000101315.

Michor, F., J. Liphardt, M. Ferrari and J. Widom. 2011. "What does physics have to do with cancer?" Nat Rev Cancer 11 (9): 657-70. http://www.ncbi.nlm.nih.gov/pubmed/21850037. doi: 10.1038/nrc3092.

Mingot, J. M., S. Vega, B. Maestro, J. M. Sanz and M. A. Nieto. 2009. "Characterization of Snail nuclear import pathways as representatives of C2H2 zinc finger transcription factors." J Cell Sci 122 (Pt 9): 1452-60. http://www.ncbi.nlm.nih.gov/pubmed/19386897. doi: 10.1242/jcs.041749.

Mitchell, M. J. and M. R. King. 2013. "Computational and experimental models of cancer cell response to fluid shear stress." Front Oncol 3: 44. http://www.ncbi.nlm.nih.gov/pubmed/23467856. doi: 10.3389/fonc.2013.00044.

Mizuno, S., T. Kurosawa, K. Matsumoto, Y. Mizuno-Horikawa, M. Okamoto and T. Nakamura. 1998. "Hepatocyte growth factor prevents renal fibrosis and dysfunction in a mouse model of chronic renal disease." J Clin Invest 101 (9): 1827- 34. http://www.ncbi.nlm.nih.gov/pubmed/9576745. doi: 10.1172/JCI1709.

Reference List 221 Molina-Ortiz, P., A. Villarejo, M. MacPherson, V. Santos, A. Montes, S. Souchelnytskyi, F. Portillo and A. Cano. 2012. "Characterization of the SNAG and SLUG domains of Snail2 in the repression of E-cadherin and EMT induction: modulation by serine 4 phosphorylation." PLoS One 7 (5): e36132. http://www.ncbi.nlm.nih.gov/pubmed/22567133. doi: 10.1371/journal.pone.0036132.

Moody, S. E., D. Perez, T. C. Pan, C. J. Sarkisian, C. P. Portocarrero, C. J. Sterner, K. L. Notorfrancesco, R. D. Cardiff and L. A. Chodosh. 2005. "The transcriptional repressor Snail promotes mammary tumor recurrence." Cancer Cell 8 (3): 197-209. http://www.ncbi.nlm.nih.gov/pubmed/16169465. doi: 10.1016/j.ccr.2005.07.009.

Morel, A. P., M. Lievre, C. Thomas, G. Hinkal, S. Ansieau and A. Puisieux. 2008. "Generation of breast cancer stem cells through epithelial-mesenchymal transition." PLoS One 3 (8): e2888. http://www.ncbi.nlm.nih.gov/pubmed/18682804. doi: 10.1371/journal.pone.0002888.

Neuzillet, C., A. Tijeras-Raballand, R. Cohen, J. Cros, S. Faivre, E. Raymond and A. de Gramont. 2015. "Targeting the TGFbeta pathway for cancer therapy." Pharmacol Ther 147: 22-31. http://www.ncbi.nlm.nih.gov/pubmed/25444759. doi: 10.1016/j.pharmthera.2014.11.001.

Nicolazzo, C. and A. Gradilone. 2015. "Significance of circulating tumor cells in soft tissue sarcoma." Anal Cell Pathol (Amst) 2015: 697395. http://www.ncbi.nlm.nih.gov/pubmed/26167450. doi: 10.1155/2015/697395.

Nieto, M. A. 2002. "The snail superfamily of zinc-finger transcription factors." Nat Rev Mol Cell Biol 3 (3): 155-66. http://www.ncbi.nlm.nih.gov/pubmed/11994736. doi: 10.1038/nrm757.

Nieto, M. A. 2011. "The ins and outs of the epithelial to mesenchymal transition in health and disease." Annu Rev Cell Dev Biol 27: 347-76. http://www.ncbi.nlm.nih.gov/pubmed/21740232. doi: 10.1146/annurev-cellbio- 092910-154036.

Nieto, M. A., R. Y. Huang, R. A. Jackson and J. P. Thiery. 2016. "Emt: 2016." Cell 166 (1): 21- 45. http://www.ncbi.nlm.nih.gov/pubmed/27368099. doi: 10.1016/j.cell.2016.06.028.

Nieto, M. A., M. G. Sargent, D. G. Wilkinson and J. Cooke. 1994. "Control of cell behavior during vertebrate development by Slug, a zinc finger gene." Science 264 (5160): 835-9. http://www.ncbi.nlm.nih.gov/pubmed/7513443.

Reference List 222 Ocana, O. H., R. Corcoles, A. Fabra, G. Moreno-Bueno, H. Acloque, S. Vega, A. Barrallo- Gimeno, A. Cano and M. A. Nieto. 2012. "Metastatic colonization requires the repression of the epithelial-mesenchymal transition inducer Prrx1." Cancer Cell 22 (6): 709-24. http://www.ncbi.nlm.nih.gov/pubmed/23201163. doi: 10.1016/j.ccr.2012.10.012.

Ohike, N., I. Coban, G. E. Kim, O. Basturk, T. Tajiri, A. Krasinskas, S. Bandyopadhyay, et al. 2010. "Tumor budding as a strong prognostic indicator in invasive ampullary adenocarcinomas." Am J Surg Pathol 34 (10): 1417-24. http://www.ncbi.nlm.nih.gov/pubmed/20871215. doi: 10.1097/PAS.0b013e3181f0b05a.

Ohkubo, T. and M. Ozawa. 2004. "The transcription factor Snail downregulates the tight junction components independently of E-cadherin downregulation." J Cell Sci 117 (Pt 9): 1675-85. http://www.ncbi.nlm.nih.gov/pubmed/15075229. doi: 10.1242/jcs.01004.

Okayama, H., T. Kohno, Y. Ishii, Y. Shimada, K. Shiraishi, R. Iwakawa, K. Furuta, et al. 2012. "Identification of genes upregulated in ALK-positive and EGFR/KRAS/ALK-negative lung adenocarcinomas." Cancer Res 72 (1): 100-11. http://www.ncbi.nlm.nih.gov/pubmed/22080568. doi: 10.1158/0008-5472.CAN- 11-1403.

Ottewell, P. D., R. E. Coleman and I. Holen. 2006. "From genetic abnormality to metastases: murine models of breast cancer and their use in the development of anticancer therapies." Breast Cancer Res Treat 96 (2): 101-13. http://www.ncbi.nlm.nih.gov/pubmed/16319986. doi: 10.1007/s10549-005-9067- x.

Paget, S. 1989. "The distribution of secondary growths in cancer of the breast. 1889." Cancer Metastasis Rev 8 (2): 98-101. http://www.ncbi.nlm.nih.gov/pubmed/2673568.

Paller, C. J. and E. S. Antonarakis. 2013. "Management of biochemically recurrent prostate cancer after local therapy: evolving standards of care and new directions." Clin Adv Hematol Oncol 11 (1): 14-23. http://www.ncbi.nlm.nih.gov/pubmed/23416859.

Pang, M. F., A. M. Georgoudaki, L. Lambut, J. Johansson, V. Tabor, K. Hagikura, Y. Jin, et al. 2016. "TGF-beta1-induced EMT promotes targeted migration of breast cancer cells through the lymphatic system by the activation of CCR7/CCL21-mediated chemotaxis." Oncogene 35 (6): 748-60. http://www.ncbi.nlm.nih.gov/pubmed/25961925. doi: 10.1038/onc.2015.133.

Reference List 223 Papageorgis, P., K. Cheng, S. Ozturk, Y. Gong, A. W. Lambert, H. M. Abdolmaleky, J. R. Zhou and S. Thiagalingam. 2011. "Smad4 inactivation promotes malignancy and drug resistance of colon cancer." Cancer Res 71 (3): 998-1008. http://www.ncbi.nlm.nih.gov/pubmed/21245094. doi: 10.1158/0008-5472.CAN- 09-3269.

Park, S. M., A. B. Gaur, E. Lengyel and M. E. Peter. 2008. "The miR-200 family determines the epithelial phenotype of cancer cells by targeting the E-cadherin repressors ZEB1 and ZEB2." Genes Dev 22 (7): 894-907. http://www.ncbi.nlm.nih.gov/pubmed/18381893. doi: 10.1101/gad.1640608.

Patriarca, C., D. Petrella, B. Campo, P. Colombo, P. Giunta, M. Parente, N. Zucchini, R. Mazzucchelli and R. Montironi. 2003. "Elevated E-cadherin and alpha/beta-catenin expression after androgen deprivation therapy in prostate adenocarcinoma." Pathol Res Pract 199 (10): 659-65. http://www.ncbi.nlm.nih.gov/pubmed/14666968.

Peinado, H., E. Ballestar, M. Esteller and A. Cano. 2004. "Snail mediates E-cadherin repression by the recruitment of the Sin3A/histone deacetylase 1 (HDAC1)/HDAC2 complex." Mol Cell Biol 24 (1): 306-19. http://www.ncbi.nlm.nih.gov/pubmed/14673164.

Peinado, H., S. Lavotshkin and D. Lyden. 2011. "The secreted factors responsible for pre- metastatic niche formation: old sayings and new thoughts." Semin Cancer Biol 21 (2): 139-46. http://www.ncbi.nlm.nih.gov/pubmed/21251983. doi: 10.1016/j.semcancer.2011.01.002.

Peinado, H., D. Olmeda and A. Cano. 2007. "Snail, Zeb and bHLH factors in tumour progression: an alliance against the epithelial phenotype?" Nat Rev Cancer 7 (6): 415-28. http://www.ncbi.nlm.nih.gov/pubmed/17508028. doi: 10.1038/nrc2131.

Pereira, L., J. M. Mariadason, R. D. Hannan and A. S. Dhillon. 2015. "Implications of epithelial-mesenchymal plasticity for heterogeneity in colorectal cancer." Front Oncol 5: 13. http://www.ncbi.nlm.nih.gov/pubmed/25699236. doi: 10.3389/fonc.2015.00013.

Petersen, O. W., L. Ronnov-Jessen, A. R. Howlett and M. J. Bissell. 1992. "Interaction with basement membrane serves to rapidly distinguish growth and differentiation pattern of normal and malignant human breast epithelial cells." Proc Natl Acad Sci U S A 89 (19): 9064-8. https://www.ncbi.nlm.nih.gov/pubmed/1384042.

Petrylak, D. P. 2013. "Current state of castration-resistant prostate cancer." Am J Manag Care 19 (18 Suppl): s358-65. http://www.ncbi.nlm.nih.gov/pubmed/24494690.

Reference List 224

Pickup, M., S. Novitskiy and H. L. Moses. 2013. "The roles of TGFbeta in the tumour microenvironment." Nat Rev Cancer 13 (11): 788-99. http://www.ncbi.nlm.nih.gov/pubmed/24132110. doi: 10.1038/nrc3603.

Poblete, C. E., J. Fulla, M. Gallardo, V. Munoz, E. A. Castellon, I. Gallegos and H. R. Contreras. 2014. "Increased SNAIL expression and low syndecan levels are associated with high Gleason grade in prostate cancer." Int J Oncol 44 (3): 647-54. http://www.ncbi.nlm.nih.gov/pubmed/24424718. doi: 10.3892/ijo.2014.2254.

Polette, M., M. Mestdagt, S. Bindels, B. Nawrocki-Raby, W. Hunziker, J. M. Foidart, P. Birembaut and C. Gilles. 2007. "Beta-catenin and ZO-1: shuttle molecules involved in tumor invasion-associated epithelial-mesenchymal transition processes." Cells Tissues Organs 185 (1-3): 61-5. http://www.ncbi.nlm.nih.gov/pubmed/17587809. doi: 10.1159/000101304.

Polyak, K. and R. A. Weinberg. 2009. "Transitions between epithelial and mesenchymal states: acquisition of malignant and stem cell traits." Nat Rev Cancer 9 (4): 265-73. http://www.ncbi.nlm.nih.gov/pubmed/19262571. doi: 10.1038/nrc2620.

Pomp, V., C. Leo, A. Mauracher, D. Korol, W. Guo and Z. Varga. 2015. "Differential expression of epithelial-mesenchymal transition and stem cell markers in intrinsic subtypes of breast cancer." Breast Cancer Res Treat. http://www.ncbi.nlm.nih.gov/pubmed/26467042. doi: 10.1007/s10549-015-3598- 6.

Prins, G. S. and O. Putz. 2008. "Molecular signaling pathways that regulate prostate gland development." Differentiation 76 (6): 641-59. http://www.ncbi.nlm.nih.gov/pubmed/18462433. doi: 10.1111/j.1432- 0436.2008.00277.x.

Raatikainen, S., S. Aaltomaa, J. J. Palvimo, V. Karja and Y. Soini. 2015. "TWIST overexpression predicts biochemical recurrence-free survival in prostate cancer patients treated with radical prostatectomy." Scand J Urol 49 (1): 51-7. http://www.ncbi.nlm.nih.gov/pubmed/24779451. doi: 10.3109/21681805.2014.909529.

Radisky, E. S. and D. C. Radisky. 2010. "Matrix metalloproteinase-induced epithelial- mesenchymal transition in breast cancer." J Mammary Gland Biol Neoplasia 15 (2): 201-12. http://www.ncbi.nlm.nih.gov/pubmed/20440544. doi: 10.1007/s10911- 010-9177-x.

Reference List 225 Rescigno, P., C. Buonerba, J. Bellmunt, G. Sonpavde, S. De Placido and G. Di Lorenzo. 2012. "New perspectives in the therapy of castration resistant prostate cancer." Curr Drug Targets 13 (13): 1676-86. http://www.ncbi.nlm.nih.gov/pubmed/23043326.

Riond, J. L. and J. E. Riviere. 1988. "Pharmacology and toxicology of doxycycline." Vet Hum Toxicol 30 (5): 431-43. http://www.ncbi.nlm.nih.gov/pubmed/3055652.

Roach, M., 3rd, K. Bae, J. Speight, H. B. Wolkov, P. Rubin, R. J. Lee, C. Lawton, R. Valicenti, D. Grignon and M. V. Pilepich. 2008. "Short-term neoadjuvant androgen deprivation therapy and external-beam radiotherapy for locally advanced prostate cancer: long-term results of RTOG 8610." J Clin Oncol 26 (4): 585-91. http://www.ncbi.nlm.nih.gov/pubmed/18172188. doi: 10.1200/JCO.2007.13.9881.

Robinson, D. R., C. R. Zylstra and B. O. Williams. 2008. "Wnt signaling and prostate cancer." Curr Drug Targets 9 (7): 571-80. http://www.ncbi.nlm.nih.gov/pubmed/18673243.

Robinson, D., E. M. Van Allen, Y. M. Wu, N. Schultz, R. J. Lonigro, J. M. Mosquera, B. Montgomery, et al. 2015. "Integrative clinical genomics of advanced prostate cancer." Cell 161 (5): 1215-28. https://www.ncbi.nlm.nih.gov/pubmed/26000489. doi: 10.1016/j.cell.2015.05.001.

Roca, H., J. Hernandez, S. Weidner, R. C. McEachin, D. Fuller, S. Sud, T. Schumann, et al. 2013. "Transcription factors OVOL1 and OVOL2 induce the mesenchymal to epithelial transition in human cancer." PLoS One 8 (10): e76773. http://www.ncbi.nlm.nih.gov/pubmed/24124593. doi: 10.1371/journal.pone.0076773.

Rodriguez-Gonzalez, F. G., D. A. Mustafa, B. Mostert and A. M. Sieuwerts. 2013. "The challenge of gene expression profiling in heterogeneous clinical samples." Methods 59 (1): 47-58. http://www.ncbi.nlm.nih.gov/pubmed/22652627. doi: 10.1016/j.ymeth.2012.05.005.

Ross, S. and C. S. Hill. 2008. "How the Smads regulate transcription." Int J Biochem Cell Biol 40 (3): 383-408. http://www.ncbi.nlm.nih.gov/pubmed/18061509. doi: 10.1016/j.biocel.2007.09.006.

Rubin, M. A., N. R. Mucci, J. Figurski, A. Fecko, K. J. Pienta and M. L. Day. 2001. "E-cadherin expression in prostate cancer: a broad survey using high-density tissue microarray technology." Hum Pathol 32 (7): 690-7. http://www.ncbi.nlm.nih.gov/pubmed/11486167. doi: 10.1053/hupa.2001.25902.

Sadowski, M. C., R. H. Pouwer, J. H. Gunter, A. A. Lubik, R. J. Quinn and C. C. Nelson. 2014. "The fatty acid synthase inhibitor triclosan: repurposing an anti-microbial agent for

Reference List 226 targeting prostate cancer." Oncotarget 5 (19): 9362-81. http://www.ncbi.nlm.nih.gov/pubmed/25313139. doi: 10.18632/oncotarget.2433.

Saito, Y., H. Takazawa, K. Uzawa, H. Tanzawa and K. Sato. 1998. "Reduced expression of E- cadherin in oral squamous cell carcinoma: relationship with DNA methylation of 5' CpG island." Int J Oncol 12 (2): 293-8. http://www.ncbi.nlm.nih.gov/pubmed/9458351.

Salk, J. J., E. J. Fox and L. A. Loeb. 2010. "Mutational heterogeneity in human cancers: origin and consequences." Annu Rev Pathol 5: 51-75. http://www.ncbi.nlm.nih.gov/pubmed/19743960. doi: 10.1146/annurev-pathol- 121808-102113.

Sanchez-Tillo, E., Y. Liu, O. de Barrios, L. Siles, L. Fanlo, M. Cuatrecasas, D. S. Darling, D. C. Dean, A. Castells and A. Postigo. 2012. "EMT-activating transcription factors in cancer: beyond EMT and tumor invasiveness." Cell Mol Life Sci 69 (20): 3429-56. http://www.ncbi.nlm.nih.gov/pubmed/22945800. doi: 10.1007/s00018-012-1122- 2.

Sanchez-Tillo, E., L. Siles, O. de Barrios, M. Cuatrecasas, E. C. Vaquero, A. Castells and A. Postigo. 2011. "Expanding roles of ZEB factors in tumorigenesis and tumor progression." Am J Cancer Res 1 (7): 897-912. http://www.ncbi.nlm.nih.gov/pubmed/22016835.

Satelli, A., A. Mitra, Z. Brownlee, X. Xia, S. Bellister, M. J. Overman, S. Kopetz, L. M. Ellis, Q. H. Meng and S. Li. 2015. "Epithelial-mesenchymal transitioned circulating tumor cells capture for detecting tumor progression." Clin Cancer Res 21 (4): 899-906. http://www.ncbi.nlm.nih.gov/pubmed/25516888. doi: 10.1158/1078-0432.CCR-14- 0894.

Sato, N., M. E. Gleave, N. Bruchovsky, P. S. Rennie, E. Beraldi and L. D. Sullivan. 1997. "A metastatic and androgen-sensitive human prostate cancer model using intraprostatic inoculation of LNCaP cells in SCID mice." Cancer Res 57 (8): 1584-9. http://www.ncbi.nlm.nih.gov/pubmed/9108464.

Saxena, M. and G. Christofori. 2013. "Rebuilding cancer metastasis in the mouse." Mol Oncol 7 (2): 283-96. http://www.ncbi.nlm.nih.gov/pubmed/23474222. doi: 10.1016/j.molonc.2013.02.009.

Scher, H. I., T. M. Beer, C. S. Higano, A. Anand, M. E. Taplin, E. Efstathiou, D. Rathkopf, et al. 2010. "Antitumour activity of MDV3100 in castration-resistant prostate cancer: a phase 1-2 study." Lancet 375 (9724): 1437-46. http://www.ncbi.nlm.nih.gov/pubmed/20398925. doi: 10.1016/S0140- 6736(10)60172-9.

Reference List 227

Scher, H. I., X. Jia, J. S. de Bono, M. Fleisher, K. J. Pienta, D. Raghavan and G. Heller. 2009. "Circulating tumour cells as prognostic markers in progressive, castration-resistant prostate cancer: a reanalysis of IMMC38 trial data." Lancet Oncol 10 (3): 233-9. http://www.ncbi.nlm.nih.gov/pubmed/19213602. doi: 10.1016/S1470- 2045(08)70340-1.

Scher, H. I., M. J. Morris, W. K. Kelly, L. H. Schwartz and G. Heller. 2005. "Prostate cancer clinical trial end points: "RECIST"ing a step backwards." Clin Cancer Res 11 (14): 5223-32. http://www.ncbi.nlm.nih.gov/pubmed/16033840. doi: 10.1158/1078- 0432.CCR-05-0109.

Schmidt, J. M., E. Panzilius, H. S. Bartsch, M. Irmler, J. Beckers, V. Kari, J. R. Linnemann, et al. 2015. "Stem-Cell-like Properties and Epithelial Plasticity Arise as Stable Traits after Transient Twist1 Activation." Cell Rep. http://www.ncbi.nlm.nih.gov/pubmed/25578726. doi: 10.1016/j.celrep.2014.12.032.

Schneider, A. J., A. M. Branam and R. E. Peterson. 2014. "Intersection of AHR and Wnt signaling in development, health, and disease." Int J Mol Sci 15 (10): 17852-85. http://www.ncbi.nlm.nih.gov/pubmed/25286307. doi: 10.3390/ijms151017852.

Schuetz, C. S., M. Bonin, S. E. Clare, K. Nieselt, K. Sotlar, M. Walter, T. Fehm, et al. 2006. "Progression-specific genes identified by expression profiling of matched ductal carcinomas in situ and invasive breast tumors, combining laser capture microdissection and oligonucleotide microarray analysis." Cancer Res 66 (10): 5278- 86. http://www.ncbi.nlm.nih.gov/pubmed/16707453. doi: 10.1158/0008- 5472.CAN-05-4610.

Sethi, S., J. Macoska, W. Chen and F. H. Sarkar. 2010. "Molecular signature of epithelial- mesenchymal transition (EMT) in human prostate cancer bone metastasis." Am J Transl Res 3 (1): 90-9. http://www.ncbi.nlm.nih.gov/pubmed/21139809.

Setlur, S. R., K. D. Mertz, Y. Hoshida, F. Demichelis, M. Lupien, S. Perner, A. Sboner, et al. 2008. "Estrogen-dependent signaling in a molecularly distinct subclass of aggressive prostate cancer." J Natl Cancer Inst 100 (11): 815-25. http://www.ncbi.nlm.nih.gov/pubmed/18505969. doi: 10.1093/jnci/djn150.

Shah, R. B., R. Mehra, A. M. Chinnaiyan, R. Shen, D. Ghosh, M. Zhou, G. R. Macvicar, et al. 2004. "Androgen-independent prostate cancer is a heterogeneous group of diseases: lessons from a rapid autopsy program." Cancer Res 64 (24): 9209-16. http://www.ncbi.nlm.nih.gov/pubmed/15604294. doi: 10.1158/0008-5472.CAN- 04-2442.

Reference List 228 Shang, Z., Q. Cai, M. Zhang, S. Zhu, Y. Ma, L. Sun, N. Jiang, et al. 2015. "A switch from CD44(+) cell to EMT cell drives the metastasis of prostate cancer." Oncotarget 6 (2): 1202-16. http://www.ncbi.nlm.nih.gov/pubmed/25483103. doi: 10.18632/oncotarget.2841.

Shi, Y. and J. Massague. 2003. "Mechanisms of TGF-beta signaling from cell membrane to the nucleus." Cell 113 (6): 685-700. http://www.ncbi.nlm.nih.gov/pubmed/12809600.

Shintani, Y., A. Okimura, K. Sato, T. Nakagiri, Y. Kadota, M. Inoue, N. Sawabata, et al. 2011. "Epithelial to mesenchymal transition is a determinant of sensitivity to chemoradiotherapy in non-small cell lung cancer." Ann Thorac Surg 92 (5): 1794- 804; discussion 1804. http://www.ncbi.nlm.nih.gov/pubmed/22051275. doi: 10.1016/j.athoracsur.2011.07.032.

Shiota, M., J. L. Bishop, K. M. Nip, A. Zardan, A. Takeuchi, T. Cordonnier, E. Beraldi, et al. 2013. "Hsp27 regulates epithelial mesenchymal transition, metastasis, and circulating tumor cells in prostate cancer." Cancer Res 73 (10): 3109-19. http://www.ncbi.nlm.nih.gov/pubmed/23492367. doi: 10.1158/0008-5472.CAN- 12-3979.

Shook, D. and R. Keller. 2003. "Mechanisms, mechanics and function of epithelial- mesenchymal transitions in early development." Mech Dev 120 (11): 1351-83. http://www.ncbi.nlm.nih.gov/pubmed/14623443.

Sieh, S., A. V. Taubenberger, S. C. Rizzi, M. Sadowski, M. L. Lehman, A. Rockstroh, J. An, J. A. Clements, C. C. Nelson and D. W. Hutmacher. 2012. "Phenotypic characterization of prostate cancer LNCaP cells cultured within a bioengineered microenvironment." PLoS One 7 (9): e40217. http://www.ncbi.nlm.nih.gov/pubmed/22957009. doi: 10.1371/journal.pone.0040217.

Smit, M. A., T. R. Geiger, J. Y. Song, I. Gitelman and D. S. Peeper. 2009. "A Twist-Snail axis critical for TrkB-induced epithelial-mesenchymal transition-like transformation, anoikis resistance, and metastasis." Mol Cell Biol 29 (13): 3722-37. http://www.ncbi.nlm.nih.gov/pubmed/19414595. doi: 10.1128/MCB.01164-08.

Smith, B. N., L. J. Burton, V. Henderson, D. D. Randle, D. J. Morton, B. A. Smith, L. Taliaferro- Smith, et al. 2014. "Snail promotes epithelial mesenchymal transition in breast cancer cells in part via activation of nuclear ERK2." PLoS One 9 (8): e104987. http://www.ncbi.nlm.nih.gov/pubmed/25122124. doi: 10.1371/journal.pone.0104987.

Spaderna, S., O. Schmalhofer, M. Wahlbuhl, A. Dimmler, K. Bauer, A. Sultan, F. Hlubek, et al. 2008. "The transcriptional repressor ZEB1 promotes metastasis and loss of cell

Reference List 229 polarity in cancer." Cancer Res 68 (2): 537-44. http://www.ncbi.nlm.nih.gov/pubmed/18199550. doi: 10.1158/0008-5472.CAN- 07-5682.

Stegner, D., S. Dutting and B. Nieswandt. 2014. "Mechanistic explanation for platelet contribution to cancer metastasis." Thromb Res 133 Suppl 2: S149-57. http://www.ncbi.nlm.nih.gov/pubmed/24862136. doi: 10.1016/S0049- 3848(14)50025-4.

Stephenson, R. A., C. P. Dinney, K. Gohji, N. G. Ordonez, J. J. Killion and I. J. Fidler. 1992. "Metastatic model for human prostate cancer using orthotopic implantation in nude mice." J Natl Cancer Inst 84 (12): 951-7. http://www.ncbi.nlm.nih.gov/pubmed/1378502.

Sun, Y., B. E. Wang, K. G. Leong, P. Yue, L. Li, S. Jhunjhunwala, D. Chen, et al. 2012. "Androgen deprivation causes epithelial-mesenchymal transition in the prostate: implications for androgen-deprivation therapy." Cancer Res 72 (2): 527-36. http://www.ncbi.nlm.nih.gov/pubmed/22108827. doi: 10.1158/0008-5472.CAN- 11-3004.

Swartz, M. A. and A. W. Lund. 2012. "Lymphatic and interstitial flow in the tumour microenvironment: linking mechanobiology with immunity." Nat Rev Cancer 12 (3): 210-9. http://www.ncbi.nlm.nih.gov/pubmed/22362216. doi: 10.1038/nrc3186.

Taira, T., G. Ishii, K. Nagai, K. Yoh, Y. Takahashi, Y. Matsumura, M. Kojima, et al. 2012. "Characterization of the immunophenotype of the tumor budding and its prognostic implications in squamous cell carcinoma of the lung." Lung Cancer 76 (3): 423-30. http://www.ncbi.nlm.nih.gov/pubmed/22153829. doi: 10.1016/j.lungcan.2011.11.010.

Tam, W. L. and R. A. Weinberg. 2013. "The epigenetics of epithelial-mesenchymal plasticity in cancer." Nat Med 19 (11): 1438-49. http://www.ncbi.nlm.nih.gov/pubmed/24202396. doi: 10.1038/nm.3336.

Tan, T. Z., Q. H. Miow, Y. Miki, T. Noda, S. Mori, R. Y. Huang and J. P. Thiery. 2014. "Epithelial-mesenchymal transition spectrum quantification and its efficacy in deciphering survival and drug responses of cancer patients." EMBO Mol Med 6 (10): 1279-93. http://www.ncbi.nlm.nih.gov/pubmed/25214461. doi: 10.15252/emmm.201404208.

Tarin, D., J. E. Price, M. G. Kettlewell, R. G. Souter, A. C. Vass and B. Crossley. 1984a. "Clinicopathological observations on metastasis in man studied in patients treated with peritoneovenous shunts." Br Med J (Clin Res Ed) 288 (6419): 749-51. http://www.ncbi.nlm.nih.gov/pubmed/6423061.

Reference List 230

Tarin, D., J. E. Price, M. G. Kettlewell, R. G. Souter, A. C. Vass and B. Crossley. 1984b. "Mechanisms of human tumor metastasis studied in patients with peritoneovenous shunts." Cancer Res 44 (8): 3584-92. http://www.ncbi.nlm.nih.gov/pubmed/6744281.

Tarin, D., E. W. Thompson and D. F. Newgreen. 2005. "The fallacy of epithelial mesenchymal transition in neoplasia." Cancer Res 65 (14): 5996-6000; discussion 6000-1. http://www.ncbi.nlm.nih.gov/pubmed/16024596. doi: 10.1158/0008- 5472.CAN-05-0699.

Taube, J. H., J. I. Herschkowitz, K. Komurov, A. Y. Zhou, S. Gupta, J. Yang, K. Hartwell, et al. 2010. "Core epithelial-to-mesenchymal transition interactome gene-expression signature is associated with claudin-low and metaplastic breast cancer subtypes." Proc Natl Acad Sci U S A 107 (35): 15449-54. http://www.ncbi.nlm.nih.gov/pubmed/20713713. doi: 10.1073/pnas.1004900107.

Taylor, B. S., N. Schultz, H. Hieronymus, A. Gopalan, Y. Xiao, B. S. Carver, V. K. Arora, et al. 2010. "Integrative genomic profiling of human prostate cancer." Cancer Cell 18 (1): 11-22. http://www.ncbi.nlm.nih.gov/pubmed/20579941. doi: 10.1016/j.ccr.2010.05.026.

Taylor, I. W., P. J. Hodson, M. D. Green and R. L. Sutherland. 1983. "Effects of tamoxifen on cell cycle progression of synchronous MCF-7 human mammary carcinoma cells." Cancer Res 43 (9): 4007-10. http://www.ncbi.nlm.nih.gov/pubmed/6871842.

Terry, S., X. Yang, M. W. Chen, F. Vacherot and R. Buttyan. 2006. "Multifaceted interaction between the androgen and Wnt signaling pathways and the implication for prostate cancer." J Cell Biochem 99 (2): 402-10. http://www.ncbi.nlm.nih.gov/pubmed/16741972. doi: 10.1002/jcb.20983.

Thiery, J. P. 2002. "Epithelial-mesenchymal transitions in tumour progression." Nat Rev Cancer 2 (6): 442-54. http://www.ncbi.nlm.nih.gov/pubmed/12189386. doi: 10.1038/nrc822.

Thiery, J. P. 2003. "Epithelial-mesenchymal transitions in development and pathologies." Curr Opin Cell Biol 15 (6): 740-6. http://www.ncbi.nlm.nih.gov/pubmed/14644200.

Thiery, J. P. 2009. "[Epithelial-mesenchymal transitions in cancer onset and progression]." Bull Acad Natl Med 193 (9): 1969-78; discussion 1978-9. http://www.ncbi.nlm.nih.gov/pubmed/20666011.

Reference List 231 Thiery, J. P., H. Acloque, R. Y. Huang and M. A. Nieto. 2009. "Epithelial-mesenchymal transitions in development and disease." Cell 139 (5): 871-90. http://www.ncbi.nlm.nih.gov/pubmed/19945376. doi: 10.1016/j.cell.2009.11.007.

Thiery, J. P. and D. Chopin. 1999. "Epithelial cell plasticity in development and tumor progression." Cancer Metastasis Rev 18 (1): 31-42. http://www.ncbi.nlm.nih.gov/pubmed/10505544.

Thuault, S., E. J. Tan, H. Peinado, A. Cano, C. H. Heldin and A. Moustakas. 2008. "HMGA2 and Smads co-regulate SNAIL1 expression during induction of epithelial-to- mesenchymal transition." J Biol Chem 283 (48): 33437-46. http://www.ncbi.nlm.nih.gov/pubmed/18832382. doi: 10.1074/jbc.M802016200.

Tran, C., S. Ouk, N. J. Clegg, Y. Chen, P. A. Watson, V. Arora, J. Wongvipat, et al. 2009. "Development of a second-generation antiandrogen for treatment of advanced prostate cancer." Science 324 (5928): 787-90. http://www.ncbi.nlm.nih.gov/pubmed/19359544. doi: 10.1126/science.1168175.

Trimboli, A. J., K. Fukino, A. de Bruin, G. Wei, L. Shen, S. M. Tanner, N. Creasap, et al. 2008. "Direct evidence for epithelial-mesenchymal transitions in breast cancer." Cancer Res 68 (3): 937-45. http://www.ncbi.nlm.nih.gov/pubmed/18245497. doi: 10.1158/0008-5472.CAN-07-2148.

Tripathi, M. K., S. Misra, S. V. Khedkar, N. Hamilton, C. Irvin-Wilson, C. Sharan, L. Sealy and G. Chaudhuri. 2005. "Regulation of BRCA2 gene expression by the SLUG repressor protein in human breast cells." J Biol Chem 280 (17): 17163-71. http://www.ncbi.nlm.nih.gov/pubmed/15734731. doi: 10.1074/jbc.M501375200.

Trzpis, M., P. M. McLaughlin, L. M. de Leij and M. C. Harmsen. 2007. "Epithelial cell adhesion molecule: more than a carcinoma marker and adhesion molecule." Am J Pathol 171 (2): 386-95. http://www.ncbi.nlm.nih.gov/pubmed/17600130. doi: 10.2353/ajpath.2007.070152.

Tsai, J. H., J. L. Donaher, D. A. Murphy, S. Chau and J. Yang. 2012. "Spatiotemporal regulation of epithelial-mesenchymal transition is essential for squamous cell carcinoma metastasis." Cancer Cell 22 (6): 725-36. http://www.ncbi.nlm.nih.gov/pubmed/23201165. doi: 10.1016/j.ccr.2012.09.022.

Tsai, J. H. and J. Yang. 2013. "Epithelial-mesenchymal plasticity in carcinoma metastasis." Genes Dev 27 (20): 2192-206. http://www.ncbi.nlm.nih.gov/pubmed/24142872. doi: 10.1101/gad.225334.113.

Reference List 232 Tsuji, T., S. Ibaragi, K. Shima, M. G. Hu, M. Katsurano, A. Sasaki and G. F. Hu. 2008. "Epithelial-mesenchymal transition induced by growth suppressor p12CDK2-AP1 promotes tumor cell local invasion but suppresses distant colony growth." Cancer Res 68 (24): 10377-86. http://www.ncbi.nlm.nih.gov/pubmed/19074907. doi: 10.1158/0008-5472.CAN-08-1444.

Tuhkanen, H., Y. Soini, V. M. Kosma, M. Anttila, R. Sironen, K. Hamalainen, L. Kukkonen, I. Virtanen and A. Mannermaa. 2009. "Nuclear expression of Snail1 in borderline and malignant epithelial ovarian tumours is associated with tumour progression." BMC Cancer 9: 289. http://www.ncbi.nlm.nih.gov/pubmed/19695091. doi: 10.1186/1471-2407-9-289.

van 't Veer, L. J., H. Dai, M. J. van de Vijver, Y. D. He, A. A. Hart, M. Mao, H. L. Peterse, et al. 2002. "Gene expression profiling predicts clinical outcome of breast cancer." Nature 415 (6871): 530-6. http://www.ncbi.nlm.nih.gov/pubmed/11823860. doi: 10.1038/415530a.

van de Vijver, M. J., Y. D. He, L. J. van't Veer, H. Dai, A. A. Hart, D. W. Voskuil, G. J. Schreiber, et al. 2002. "A gene-expression signature as a predictor of survival in breast cancer." N Engl J Med 347 (25): 1999-2009. doi: 10.1056/NEJMoa021967.

van Zijl, F., G. Krupitza and W. Mikulits. 2011. "Initial steps of metastasis: cell invasion and endothelial transmigration." Mutat Res 728 (1-2): 23-34. https://www.ncbi.nlm.nih.gov/pubmed/21605699. doi: 10.1016/j.mrrev.2011.05.002.

Vander Heiden, M. G. 2011. "Targeting cancer metabolism: a therapeutic window opens." Nat Rev Drug Discov 10 (9): 671-84. http://www.ncbi.nlm.nih.gov/pubmed/21878982. doi: 10.1038/nrd3504.

Vandewalle, C., J. Comijn, B. De Craene, P. Vermassen, E. Bruyneel, H. Andersen, E. Tulchinsky, F. Van Roy and G. Berx. 2005. "SIP1/ZEB2 induces EMT by repressing genes of different epithelial cell-cell junctions." Nucleic Acids Res 33 (20): 6566-78. http://www.ncbi.nlm.nih.gov/pubmed/16314317. doi: 10.1093/nar/gki965.

Varambally, S., J. Yu, B. Laxman, D. R. Rhodes, R. Mehra, S. A. Tomlins, R. B. Shah, et al. 2005. "Integrative genomic and proteomic analysis of prostate cancer reveals signatures of metastatic progression." Cancer Cell 8 (5): 393-406. http://www.ncbi.nlm.nih.gov/pubmed/16286247. doi: 10.1016/j.ccr.2005.10.001.

Vega, S., A. V. Morales, O. H. Ocana, F. Valdes, I. Fabregat and M. A. Nieto. 2004. "Snail blocks the cell cycle and confers resistance to cell death." Genes Dev 18 (10): 1131- 43. http://www.ncbi.nlm.nih.gov/pubmed/15155580. doi: 10.1101/gad.294104.

Reference List 233

Vesuna, F., A. Lisok, B. Kimble and V. Raman. 2009. "Twist modulates breast cancer stem cells by transcriptional regulation of CD24 expression." Neoplasia 11 (12): 1318-28. http://www.ncbi.nlm.nih.gov/pubmed/20019840.

Vincent, T., E. P. Neve, J. R. Johnson, A. Kukalev, F. Rojo, J. Albanell, K. Pietras, et al. 2009. "A SNAIL1-SMAD3/4 transcriptional repressor complex promotes TGF-beta mediated epithelial-mesenchymal transition." Nat Cell Biol 11 (8): 943-50. http://www.ncbi.nlm.nih.gov/pubmed/19597490. doi: 10.1038/ncb1905.

Wang, H., I. Leav, S. Ibaragi, M. Wegner, G. F. Hu, M. L. Lu, S. P. Balk and X. Yuan. 2008. "SOX9 is expressed in human fetal prostate epithelium and enhances prostate cancer invasion." Cancer Res 68 (6): 1625-30. http://www.ncbi.nlm.nih.gov/pubmed/18339840. doi: 10.1158/0008-5472.CAN- 07-5915.

Wang, N., F. E. Docherty, H. K. Brown, K. J. Reeves, A. C. Fowles, P. D. Ottewell, T. N. Dear, I. Holen, P. I. Croucher and C. L. Eaton. 2014. "Prostate cancer cells preferentially home to osteoblast-rich areas in the early stages of bone metastasis: evidence from in vivo models." J Bone Miner Res 29 (12): 2688-96. http://www.ncbi.nlm.nih.gov/pubmed/24956445. doi: 10.1002/jbmr.2300.

Ware, K. E., J. A. Somarelli, D. Schaeffer, J. Li, T. Zhang, S. Park, S. R. Patierno, et al. 2016. "Snail promotes resistance to enzalutamide through regulation of androgen receptor activity in prostate cancer." Oncotarget 7 (31): 50507-50521. https://www.ncbi.nlm.nih.gov/pubmed/27409172. doi: 10.18632/oncotarget.10476.

Warga, R. M. and C. B. Kimmel. 1990. "Cell movements during epiboly and gastrulation in zebrafish." Development 108 (4): 569-80. http://www.ncbi.nlm.nih.gov/pubmed/2387236.

Whitaker, H. C., J. Girling, A. Y. Warren, H. Leung, I. G. Mills and D. E. Neal. 2008. "Alterations in beta-catenin expression and localization in prostate cancer." Prostate 68 (11): 1196-205. http://www.ncbi.nlm.nih.gov/pubmed/18459111. doi: 10.1002/pros.20780.

Whiteland, H., S. Spencer-Harty, D. H. Thomas, C. Davies, C. Morgan, H. Kynaston, P. Bose, et al. 2013. "Putative prognostic epithelial-to-mesenchymal transition biomarkers for aggressive prostate cancer." Exp Mol Pathol 95 (2): 220-6. http://www.ncbi.nlm.nih.gov/pubmed/23933194. doi: 10.1016/j.yexmp.2013.07.010.

Reference List 234 Whiteman, E. L., C. J. Liu, E. R. Fearon and B. Margolis. 2008. "The transcription factor snail represses Crumbs3 expression and disrupts apico-basal polarity complexes." Oncogene 27 (27): 3875-9. http://www.ncbi.nlm.nih.gov/pubmed/18246119. doi: 10.1038/onc.2008.9.

Wong, T. S., W. Gao and J. Y. Chan. 2014. "Transcription regulation of E-cadherin by zinc finger E-box binding homeobox proteins in solid tumors." Biomed Res Int 2014: 921564. http://www.ncbi.nlm.nih.gov/pubmed/25197668. doi: 10.1155/2014/921564.

Wu, J. P., W. B. Huang, H. Zhou, L. W. Xu, J. H. Zhao, J. G. Zhu, J. H. Su and H. B. Sun. 2014. "Intensity of stromal changes is associated with tumor relapse in clinically advanced prostate cancer after castration therapy." Asian J Androl 16 (5): 710-4. http://www.ncbi.nlm.nih.gov/pubmed/24875819. doi: 10.4103/1008-682X.129131.

Wu, Y., J. Deng, P. G. Rychahou, S. Qiu, B. M. Evers and B. P. Zhou. 2009. "Stabilization of snail by NF-kappaB is required for inflammation-induced cell migration and invasion." Cancer Cell 15 (5): 416-28. http://www.ncbi.nlm.nih.gov/pubmed/19411070. doi: 10.1016/j.ccr.2009.03.016.

Xu, J., S. Lamouille and R. Derynck. 2009. "TGF-beta-induced epithelial to mesenchymal transition." Cell Res 19 (2): 156-72. http://www.ncbi.nlm.nih.gov/pubmed/19153598. doi: 10.1038/cr.2009.5.

Xu, R., Y. Zhang, L. Gu, J. Zheng, J. Cui, J. Dong and J. Du. 2015. "Arf6 regulates EGF-induced internalization of E-cadherin in breast cancer cells." Cancer Cell Int 15 (1): 11. http://www.ncbi.nlm.nih.gov/pubmed/25678857. doi: 10.1186/s12935-015-0159- 3.

Yamaguchi, Y., G. Ishii, M. Kojima, K. Yoh, H. Otsuka, Y. Otaki, K. Aokage, et al. 2010. "Histopathologic features of the tumor budding in adenocarcinoma of the lung: tumor budding as an index to predict the potential aggressiveness." J Thorac Oncol 5 (9): 1361-8. http://www.ncbi.nlm.nih.gov/pubmed/20631633. doi: 10.1097/JTO.0b013e3181eaf2f3.

Yang, A. D., F. Fan, E. R. Camp, G. van Buren, W. Liu, R. Somcio, M. J. Gray, H. Cheng, P. M. Hoff and L. M. Ellis. 2006. "Chronic oxaliplatin resistance induces epithelial-to- mesenchymal transition in colorectal cancer cell lines." Clin Cancer Res 12 (14 Pt 1): 4147-53. http://www.ncbi.nlm.nih.gov/pubmed/16857785. doi: 10.1158/1078- 0432.CCR-06-0038.

Yang, D. C., M. H. Yang, C. C. Tsai, T. F. Huang, Y. H. Chen and S. C. Hung. 2011. "Hypoxia inhibits osteogenesis in human mesenchymal stem cells through direct regulation of RUNX2 by TWIST." PLoS One 6 (9): e23965.

Reference List 235 http://www.ncbi.nlm.nih.gov/pubmed/21931630. doi: 10.1371/journal.pone.0023965.

Yang, F., X. Li, M. Sharma, C. Y. Sasaki, D. L. Longo, B. Lim and Z. Sun. 2002. "Linking beta- catenin to androgen-signaling pathway." J Biol Chem 277 (13): 11336-44. http://www.ncbi.nlm.nih.gov/pubmed/11792709. doi: 10.1074/jbc.M111962200.

Yang, F., L. Sun, Q. Li, X. Han, L. Lei, H. Zhang and Y. Shang. 2012. "SET8 promotes epithelial- mesenchymal transition and confers TWIST dual transcriptional activities." EMBO J 31 (1): 110-23. http://www.ncbi.nlm.nih.gov/pubmed/21983900. doi: 10.1038/emboj.2011.364.

Yang, J., S. A. Mani, J. L. Donaher, S. Ramaswamy, R. A. Itzykson, C. Come, P. Savagner, I. Gitelman, A. Richardson and R. A. Weinberg. 2004. "Twist, a master regulator of morphogenesis, plays an essential role in tumor metastasis." Cell 117 (7): 927-39. http://www.ncbi.nlm.nih.gov/pubmed/15210113. doi: 10.1016/j.cell.2004.06.006.

Yang, L., X. H. Peng, Y. A. Wang, X. Wang, Z. Cao, C. Ni, P. Karna, et al. 2009. "Receptor- targeted nanoparticles for in vivo imaging of breast cancer." Clin Cancer Res 15 (14): 4722-32. http://www.ncbi.nlm.nih.gov/pubmed/19584158. doi: 10.1158/1078-0432.CCR-08-3289.

Yang, M. H., D. S. Hsu, H. W. Wang, H. J. Wang, H. Y. Lan, W. H. Yang, C. H. Huang, et al. 2010. "Bmi1 is essential in Twist1-induced epithelial-mesenchymal transition." Nat Cell Biol 12 (10): 982-92. http://www.ncbi.nlm.nih.gov/pubmed/20818389. doi: 10.1038/ncb2099.

Yang, M. H., M. Z. Wu, S. H. Chiou, P. M. Chen, S. Y. Chang, C. J. Liu, S. C. Teng and K. J. Wu. 2008. "Direct regulation of TWIST by HIF-1alpha promotes metastasis." Nat Cell Biol 10 (3): 295-305. http://www.ncbi.nlm.nih.gov/pubmed/18297062. doi: 10.1038/ncb1691.

Yang, X., M. W. Chen, S. Terry, F. Vacherot, D. L. Bemis, J. Capodice, J. Kitajewski, et al. 2006. "Complex regulation of human androgen receptor expression by Wnt signaling in prostate cancer cells." Oncogene 25 (24): 3436-44. http://www.ncbi.nlm.nih.gov/pubmed/16474850. doi: 10.1038/sj.onc.1209366.

Yang, Z., S. Rayala, D. Nguyen, R. K. Vadlamudi, S. Chen and R. Kumar. 2005. "Pak1 phosphorylation of snail, a master regulator of epithelial-to-mesenchyme transition, modulates snail's subcellular localization and functions." Cancer Res 65 (8): 3179-84. http://www.ncbi.nlm.nih.gov/pubmed/15833848. doi: 10.1158/0008- 5472.CAN-04-3480.

Reference List 236 Yao, R., H. Jiang, Y. Ma, L. Wang, L. Wang, J. Du, P. Hou, et al. 2014. "PRMT7 induces epithelial-to-mesenchymal transition and promotes metastasis in breast cancer." Cancer Res 74 (19): 5656-67. http://www.ncbi.nlm.nih.gov/pubmed/25136067. doi: 10.1158/0008-5472.CAN-14-0800.

Yates, C. C., C. R. Shepard, D. B. Stolz and A. Wells. 2007. "Co-culturing human prostate carcinoma cells with hepatocytes leads to increased expression of E-cadherin." Br J Cancer 96 (8): 1246-52. http://www.ncbi.nlm.nih.gov/pubmed/17406365. doi: 10.1038/sj.bjc.6603700.

Ye, X. and R. A. Weinberg. 2015. "Epithelial-Mesenchymal Plasticity: A Central Regulator of Cancer Progression." Trends Cell Biol 25 (11): 675-86. http://www.ncbi.nlm.nih.gov/pubmed/26437589. doi: 10.1016/j.tcb.2015.07.012.

Yokoyama, K., N. Kamata, R. Fujimoto, S. Tsutsumi, M. Tomonari, M. Taki, H. Hosokawa and M. Nagayama. 2003. "Increased invasion and matrix metalloproteinase-2 expression by Snail-induced mesenchymal transition in squamous cell carcinomas." Int J Oncol 22 (4): 891-8. http://www.ncbi.nlm.nih.gov/pubmed/12632084.

Yokoyama, N. N., S. Shao, B. H. Hoang, D. Mercola and X. Zi. 2014. "Wnt signaling in castration-resistant prostate cancer: implications for therapy." Am J Clin Exp Urol 2 (1): 27-44. http://www.ncbi.nlm.nih.gov/pubmed/25143959.

Yu, M., A. Bardia, B. S. Wittner, S. L. Stott, M. E. Smas, D. T. Ting, S. J. Isakoff, et al. 2013. "Circulating breast tumor cells exhibit dynamic changes in epithelial and mesenchymal composition." Science 339 (6119): 580-4. http://www.ncbi.nlm.nih.gov/pubmed/23372014. doi: 10.1126/science.1228522.

Yu, Y. P., D. Landsittel, L. Jing, J. Nelson, B. Ren, L. Liu, C. McDonald, et al. 2004. "Gene expression alterations in prostate cancer predicting tumor aggression and preceding development of malignancy." J Clin Oncol 22 (14): 2790-9. http://www.ncbi.nlm.nih.gov/pubmed/15254046. doi: 10.1200/JCO.2004.05.158.

Zadran, S., R. Arumugam, H. Herschman, M. E. Phelps and R. D. Levine. 2014. "Surprisal analysis characterizes the free energy time course of cancer cells undergoing epithelial-to-mesenchymal transition." Proc Natl Acad Sci U S A 111 (36): 13235-40. http://www.ncbi.nlm.nih.gov/pubmed/25157127. doi: 10.1073/pnas.1414714111.

Zeng, Y., K. Opeskin, J. Goad and E. D. Williams. 2006. "Tumor-induced activation of lymphatic endothelial cells via vascular endothelial growth factor receptor-2 is critical for prostate cancer lymphatic metastasis." Cancer Res 66 (19): 9566-75. http://www.ncbi.nlm.nih.gov/pubmed/17018613. doi: 10.1158/0008-5472.CAN- 06-1488.

Reference List 237

Zhang, L., G. Huang, X. Li, Y. Zhang, Y. Jiang, J. Shen, J. Liu, et al. 2013. "Hypoxia induces epithelial-mesenchymal transition via activation of SNAI1 by hypoxia-inducible factor -1alpha in hepatocellular carcinoma." BMC Cancer 13: 108. http://www.ncbi.nlm.nih.gov/pubmed/23496980. doi: 10.1186/1471-2407-13-108.

Zhang, X., G. Liu, Y. Kang, Z. Dong, Q. Qian and X. Ma. 2013. "N-cadherin expression is associated with acquisition of EMT phenotype and with enhanced invasion in erlotinib-resistant lung cancer cell lines." PLoS One 8 (3): e57692. http://www.ncbi.nlm.nih.gov/pubmed/23520479. doi: 10.1371/journal.pone.0057692.

Zhao, W., Y. Zhou, H. Xu, Y. Cheng and B. Kong. 2013. "Snail family proteins in cervical squamous carcinoma: expression and significance." Clin Invest Med 36 (4): E223-33. http://www.ncbi.nlm.nih.gov/pubmed/23906494.

Zheng, H. and Y. Kang. 2014. "Multilayer control of the EMT master regulators." Oncogene 33 (14): 1755-63. http://www.ncbi.nlm.nih.gov/pubmed/23604123. doi: 10.1038/onc.2013.128.

Zheng, H., W. Li, Y. Wang, Z. Liu, Y. Cai, T. Xie, M. Shi, Z. Wang and B. Jiang. 2013. "Glycogen synthase kinase-3 beta regulates Snail and beta-catenin expression during Fas- induced epithelial-mesenchymal transition in gastrointestinal cancer." Eur J Cancer 49 (12): 2734-46. http://www.ncbi.nlm.nih.gov/pubmed/23582741. doi: 10.1016/j.ejca.2013.03.014.

Zheng, H., M. Shen, Y. L. Zha, W. Li, Y. Wei, M. A. Blanco, G. Ren, et al. 2014. "PKD1 phosphorylation-dependent degradation of SNAIL by SCF-FBXO11 regulates epithelial-mesenchymal transition and metastasis." Cancer Cell 26 (3): 358-73. http://www.ncbi.nlm.nih.gov/pubmed/25203322. doi: 10.1016/j.ccr.2014.07.022.

Zheng, X., J. L. Carstens, J. Kim, M. Scheible, J. Kaye, H. Sugimoto, C. C. Wu, V. S. LeBleu and R. Kalluri. 2015. "Epithelial-to-mesenchymal transition is dispensable for metastasis but induces chemoresistance in pancreatic cancer." Nature 527 (7579): 525-30. http://www.nature.com/nature/journal/v527/n7579/pdf/nature16064.pdf. doi: 10.1038/nature16064.

Zhou, B. P., J. Deng, W. Xia, J. Xu, Y. M. Li, M. Gunduz and M. C. Hung. 2004. "Dual regulation of Snail by GSK-3beta-mediated phosphorylation in control of epithelial- mesenchymal transition." Nat Cell Biol 6 (10): 931-40. http://www.ncbi.nlm.nih.gov/pubmed/15448698. doi: 10.1038/ncb1173.

Zhou, H., B. Zhang, J. Zheng, M. Yu, T. Zhou, K. Zhao, Y. Jia, X. Gao, C. Chen and T. Wei. 2014. "The inhibition of migration and invasion of cancer cells by graphene via the

Reference List 238 impairment of mitochondrial respiration." Biomaterials 35 (5): 1597-607. http://www.ncbi.nlm.nih.gov/pubmed/24290441. doi: 10.1016/j.biomaterials.2013.11.020.

Zhu, M. L. and N. Kyprianou. 2010. "Role of androgens and the androgen receptor in epithelial-mesenchymal transition and invasion of prostate cancer cells." FASEB J 24 (3): 769-77. http://www.ncbi.nlm.nih.gov/pubmed/19901020. doi: 10.1096/fj.09-136994.

Zinn, K. R., T. R. Chaudhuri, A. A. Szafran, D. O'Quinn, C. Weaver, K. Dugger, D. Lamar, R. A. Kesterson, X. Wang and S. J. Frank. 2008. "Noninvasive bioluminescence imaging in small animals." ILAR J 49 (1): 103-15. http://www.ncbi.nlm.nih.gov/pubmed/18172337.

Zlobec, I. and A. Lugli. 2010. "Epithelial mesenchymal transition and tumor budding in aggressive colorectal cancer: tumor budding as oncotarget." Oncotarget 1 (7): 651- 61. http://www.ncbi.nlm.nih.gov/pubmed/21317460. doi: 10.18632/oncotarget.101109.

Reference List 239