BNC1 Regulates Human Epicardial Heterogeneity and Function

BNC1 REGULATES HUMAN EPICARDIAL

HETEROGENEITY AND FUNCTION

Sophie McManus

St Catharine’s College

Department of Clinical Medicine/

Cambridge Stem Cell Institute, Jeffrey Cheah Biomedical Centre

Addenbrooke’s Hospital

University of Cambridge

This dissertation is submitted for the degree of Doctor of Philosophy

October 2019

Dedicated to A.

DECLARATION

This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration except where specifically indicated in the text. It has not been previously submitted, in part or whole, to the University of Cambridge or any other university or institution for any degree, diploma, or other qualification.

In accordance with the Department of Clinical Medicine guidelines, this thesis does not exceed 60,000 words.

Sophie McManus (MA, MRes)

Publications

Part of the work presented in this dissertation has either been submitted or published in the following:

Gambardella L., McManus S.A., Moignard V., Sebukhan D., Delaune A., Andrews S., Bernard W.G., Morrison M., Riley P.R., Le Novѐre N., Sinha S. BNC1 is a master regulator of human epicardial cell heterogeneity and function. Development, accepted for publication in August 2019.

(Development 2019 146: dev174441 doi: 10.1242/dev.174441 Published 13th December 2019.)

ABSTRACT

Name: Sophie McManus

Thesis title: BNC1 regulates human epicardial heterogeneity and function

The epicardium is a transcriptionally heterogeneous cell layer covering the heart, crucial to correct cardiovascular development. Following epithelial-to-mesenchymal transition (EMT), epicardial cells migrate into myocardium, form coronary smooth muscle cells and cardiac fibroblasts, and instruct cardiomyocytes to proliferate and mature. Adult mammalian epicardium is quiescent, but reactivates post-injury with limited effect. However, in zebrafish and in neonatal mouse, epicardial signalling enables robust cardiac regeneration after myocardial infarction. We hypothesise that manipulating human epicardial function could facilitate heart regeneration, via reactivation of embryonic processes. However, epicardial regulation remains incompletely understood; although understanding epicardial mechanisms could be key to potentially manipulating epicardium for therapeutic benefit. This PhD investigates a candidate transcription factor, Basonuclin 1 (BNC1), in functional regulation of human epicardial models, and identifies this gene as a potential key human epicardial regulator. Epicardial-like cells derived from human pluripotent stem cells (hPSC-epi) were previously used for single-cell RNA sequencing (scRNA-seq) in order to investigate possible human epicardial heterogeneity. This identified two distinct hPSC-epi subpopulations: one high in WT1 expression, the other high in TCF21. Bioinformatic analyses identified BNC1 as a potential key node in the hPSC-epi signalling network, via network inference modelling. BNC1 is a transcription factor known to regulate migration and proliferation in other epithelia. Given our network inference analyses and the literature evidence, I hypothesised that BNC1 would have functional relevance in human epicardium, so aimed to investigate its function in hPSC-epi differentiation and epicardial cell migration, as well as identify its putative epicardial targets. Firstly, scRNA-seq data describing hPSC-epi heterogeneity were validated in primary human foetal epicardium and BNC1 expression was confirmed in human epicardial models. BNC1 was subsequently investigated, both by siRNA- knockdown in hPSC-epi and foetal epicardial explants and inducible knockdown cell lines (siKD). siKD hPSC-epi had over 90% BNC1 reduction and displayed significantly altered expression of canonical epicardial genes WT1 and TCF21: hPSC-epi heterogeneity was thereby lost. Altered hPSC-epi proliferation and viability were also

iii

observed. siKD hPSC-epi was subsequently used in a simple epicardial EMT model (epi- EMT). siKD epi-EMT displayed impaired migration and pronounced cortical actin localisation. ChIP sequencing and bulk RNA sequencing identified potentially promising BNC1 targets, such as actin-binding protein supervillin, for future investigation. We conclude that BNC1 is a key functional epicardial regulator in vitro, paving the way for in vivo characterisation. The knowledge that manipulating BNC1 regulates epicardial heterogeneity and function may instruct efforts to harness epicardial potential for future therapeutic benefit.

ACKNOWLEDGEMENTS

Firstly, I am extremely grateful to Dr Sanjay Sinha for providing the opportunity to undertake my PhD within his research group. It has been both a rewarding and challenging experience; I have truly appreciated his continual intellectual input, critical evaluation and overarching support throughout the project, which has passed incredibly quickly. Being a member of the Sinha group for the past few years has been a fantastic experience. I would also like to thank Dr Laure Gambardella, an inspirational day-to-day supervisor. I am especially grateful to Laure for her kindness, advice and encouragement (combined with a great sense of humour); I have very much enjoyed the time spent working with her and have learnt a great deal, particularly with regards to experimental design and data analysis.

I also extend my sincere thanks to all the Sinha group members and collaborators past and present, and in particular thank the following people: Dr Will Bernard, for a great deal of advice and support in establishing the BNC1 knockdown cell lines; Dr Vincent Knight-Schrijver, for his invaluable work in analysing ChIP-seq data and several interesting discussions; Maura Morrison, who was a wonderful rotation student; Dr Maria Colzani, Dr Aishwarya Jacobs and Ms Ping Ong, for invaluable experimental insights; Dr Hongorzul Davaapil, both for scientific guidance and the ImageJ cell counter macro; Alex Petchey for the work on the BNC1 mice; and both Deborah Passey and Dr Peter Holt for general logistical life-saving. All members of the group have been happy to discuss my work, frequently offering helpful feedback and advice: overall, the Sinha group has provided a wonderful working environment, both in terms of scientific achievement and social atmosphere. Dr Helena Kim also offered regular food for thought and professional insight at events outside the lab.

I have greatly appreciated the intellectual input from Dr Helle Jørgensen during regular lab meeting presentations and discussions. I am also thankful to Helle for the time I spent in her group during my rotation project, and for her support in pursuing an early PhD side-project on coronary artery proliferation during embryonic development in the Confetti mouse model. Dr Jenny Harman offered useful advice regarding ChIP experiments, and Annabel Taylor has been a great friend throughout the length of my project, providing plenty of experimental discussion over coffee. Cambridge

Cardiovascular Division members have also provided helpful feedback and critical evaluation of my data following regular divisional presentations.

I am thankful to the Vallier lab for the gift of the inducible knockdown vector and quantities of helpful advice (in particular, Dr Alessandro Bertero, Stephanie Brown and Dr Anna Osnato). While both the Sinha and the Vallier groups have recently relocated, I have many good memories of working in the LRM (notwithstanding the occasional building-related hiccup). Thanks also go to the Phenotyping Hub, which offered support for flow cytometry; Gregory Strachan and Peter Humphreys, for their useful advice regarding imaging; and Xiaoling He, who coordinated the supply of human foetal tissues.

I’ve enjoyed much support from the Oxford girls, Sophie, Helen and Natalia, and have appreciated their unwavering kindness, honesty and humorous take on life. Further thanks are owed to the irrepressible Barton Road girls Anni, Emma and Katie, for their wonderfully absurd sense of humour and fun, as well as several former and continuing members of St Catharine’s College, whom I thank for many great times around Cambridge.

I am naturally extremely grateful to my family, and in particular my parents and grandparents, for offering their valuable perspective and encouragement, as ever.

This PhD project was enabled by generous British Heart Foundation funding under grant code FS/14/59/31282, and I have been both grateful and proud to represent such a great charity over the last few years.

Lastly, I would especially like to thank my fiancé Alex, for being a constant source of support, positivity and kindness.

CONTENTS

1. INTRODUCTION ...... 23 1.1 THE HEART ...... 24 1.1.1 Cardiovascular disease ...... 24 1.1.2 Cardiovascular regeneration after injury ...... 25 1.1.3 Cellular therapy for cardiac regeneration ...... 27

1.2 CARDIOVASCULAR ORIGINS: MESODERM DEVELOPMENT ...... 28 1.2.1 Mesoderm specification: genes and signalling ...... 28 1.2.2 Cardiac lineage formation ...... 30

1.3 PROEPICARDIAL FORMATION, SIGNALLING AND MIGRATION ...... 32 1.3.1 Avian models ...... 34 1.3.2 Zebrafish models ...... 35 1.3.3 Mammalian models ...... 36

1.4 THE EPICARDIUM ...... 37 1.4.1 Epicardial differentiation to cardiac fibroblasts and coronary smooth muscle cells ...... 39 1.4.2 Epicardial differentiation to endothelial cells and cardiomyocytes ...... 40 1.4.3 Gene expression in the PE and epicardium ...... 41

1.5 BACKGROUND TO EPICARDIAL EMT ...... 42 1.5.1 Epicardial EMT during cardiac development ...... 43 1.5.2 Signalling during epicardial EMT ...... 43 1.5.3 Transcriptional regulation of epicardial EMT ...... 45 1.5.4 Cell polarity and morphology changes in EMT ...... 47 1.5.5 EMT and EPDC specification ...... 49

1.6 EPICARDIUM IN HEART REGENERATION: SPECIES-SPECIFIC OUTCOMES ...... 50 1.6.1 Cardiac regeneration in zebrafish ...... 51 1.6.2 Cardiac regeneration in neonatal mouse...... 52 1.6.3. Cardiac injury responses in adult mammals...... 53 1.6.4 Harnessing adult epicardium for cardiac repair ...... 54

1.7 EPICARDIAL HETEROGENEITY ...... 56

vii

1.7.1 Epicardial transcriptional heterogeneity ...... 56 1.7.2 Epicardial morphological heterogeneity ...... 57 1.7.3 Human stem cell models of epicardial function, hPSC-epi ...... 57

1.8 HETEROGENEITY REVEALED IN HPSC-EPI BY SCRNA-SEQ ...... 59 1.8.1 BNC1 identified as candidate gene of interest in hPSC-epi subpopulation .. 61 1.8.2 Gene Ontology analysis predicts differing hPSC-epi subpopulation functions ...... 63 1.8.3 BNC1: a transcription factor with epithelial relevance...... 63

1.9 PHD PROJECT RATIONALE ...... 64

1.10 PHD HYPOTHESIS ...... 65 1.10.1 PhD Project Aims ...... 65 2. MATERIALS AND METHODS ...... 66 2.1 HPSC culture ...... 66 2.2 HPSC differentiation to hPSC-epi ...... 69 2.3 HPSC differentiation to hPSC-epi ...... 70 2.4 Primary human epicardial cultures ...... 71 2.5 BNC1 siRNA-mediated knockdown in human primary foetal epicardial cells . 71 2.6 Inducible knockdown: design and annealing shRNA oligonucleotides ...... 73 2.7 Colony PCR of transformants ...... 75 2.8 Vector digestion ...... 76 2.9 Gene targeting by lipofection ...... 76 2.10 Genotyping siKD hPSC clones ...... 76 2.11 Inducible BNC1 knockdown ...... 79 2.12 Immunofluorescence ...... 79 2.13 Cultured cells: immunofluorescence ...... 79 2.14 F-Actin labelling and quantification ...... 79 2.15 Sequential staining with antibodies raised in the same species...... 80 2.16 Antibody dialysis ...... 80 2.17 Primary antibody fluorophore conjugation ...... 80 2.18 Cryostat sections: immunohistochemistry ...... 81 2.19 Flow cytometry ...... 82 2.20 Calcein assay ...... 83 2.21 Annexin V staining ...... 83 2.22 Zombie Red viability assay ...... 83 2.23 Quantitative real-time polymerase chain reaction (qPCR) ...... 83 viii

2.24 Western Blotting ...... 85 2.25 hPSC-epi Proliferation assay...... 86 2.26 siRNA-mediated knockdown in hPSC-epi model of epithelial-to-mesenchymal transition ...... 86 2.27 EMT model scratch assay ...... 87 2.28 2D InCell migration assay ...... 87 2.29 3D invasion hPSC-epi cardiomyocyte co-culture assay ...... 87 2.30 ChIP qPCR primer design ...... 88 2.31 Chromatin shearing test ...... 88 2.32 Chromatin ImmunoPrecipitation (ChIP) ...... 88 2.33 iPure cleanup of gDNA ...... 89 2.34 ChIP library preparation and sequencing ...... 89 2.35 CHIP qPCR ...... 89 2.36 Bulk RNA sequencing ...... 90 2.37 Data clean-up, alignment, and quality control ...... 90 2.38 Principal Component Analysis...... 91 2.39 Differential gene expression analysis and Gene Ontology enrichment ...... 91 2.40 Single-molecule RNA fluorescent in-situ hybridisation (smRNA-FISH) ...... 91 2.41 smRNA-FISH Probe design ...... 91 2.42 smRNA-FISH slide preparation ...... 91 2.41 Autofluorescence treatment for smRNA-FISH sections ...... 92 2.42 smRNA-FISH hybridisation ...... 92 2.43 Data Presentation and Statistics ...... 92 3. RESULTS...... 94

3.1 HPSC-EPI IS TRANSCRIPTIONALLY HETEROGENEOUS ...... 94 3.1.1 Validating hPSC-epi scRNA-seq results by immunocytochemistry ...... 94 3.1.2 Primary human foetal epicardial cultures used to verify hPSC-epi results .. 96 3.1.3 Validation of hPSC-epi heterogeneity in primary human foetal epicardial cultures ...... 99 3.1.4 Heterogeneous BNC1 expression in hPSC-epi and primary human foetal epicardium ...... 103 3.1.5 PODXL and THY1 are membrane markers for BNC1high and TCF21high hPSC- epi subpopulations ...... 105 3.1.6 PODXL and THY1-positive fractions alter during hPSC-epi differentiation ...... 109

3.2 INVESTIGATING BNC1 FUNCTION IN HPSC-EPI ...... 110 3.2.1 An inducible knockdown strategy to investigate BNC1 function ...... 111 3.2.2 Deriving BNC1 inducible knockdown cell lines ...... 111 3.2.3 Robust BNC1 knockdown mediated in different siKD lines ...... 117 3.2.4 Loss of BNC1 alters expression of canonical epicardial genes WT1 and TCF21 ...... 120 3.2.5 Perturbing BNC1 expression in primary human foetal epicardium alters TCF21 expression ...... 123 3.2.6 BNC1 reduction alters PODXL/THY1 ratios in hPSC-epi ...... 124 3.2.7 BNC1 knockdown increases hPSC-epi proliferative index ...... 127 3.2.8 BNC1 knockdown reduces hPSC-epi viability ...... 130

3.3 BNC1 – A POTENTIAL ROLE IN REGULATION OF EPICARDIAL EMT? ...... 132 3.3.1 Modelling epicardial EMT in vitro with hPSC-epi ...... 132 3.3.2 Characterising the in vitro epi-EMT model ...... 134 3.3.3 PODXL and THY1 expression in the epi-EMT model ...... 136 3.3.4 siRNA-mediated BNC1 knockdown reveals a cortical actin localisation phenotype ...... 137 3.3.4 Cortical actin localisation in siKD epi-EMT ...... 141 3.3.5 Cortical actin localisation is observed both in TGFβ and FGF2-mediated epi- EMT ...... 141

3.4 BNC1 – A POSSIBLE ROLE IN REGULATING CELL MIGRATION? ...... 143 3.4.1 Investigating BNC1 in a 2D epi-EMT migration model ...... 143 3.4.2 BNC1 knockdown significantly reduces 2D epi-EMT cell motility...... 146 3.4.3 Investigating BNC1 in a 3D model of epi-EMT invasion ...... 148 3.4.4 BNC1 knockdown causes non-significant impairment in hPSC-epi invasion ...... 149

3.5 BNC1 – IDENTIFYING POTENTIAL EPICARDIAL TARGETS ...... 151 3.5.1 ChIP sequencing for BNC1 targets ...... 151 3.5.2 ChIP-seq analysis reveals potential targets including actin-binding protein supervillin ...... 153 3.5.3 ChIP-seq target validation via qPCR ...... 156

3.6 FOCUSING ON SVIL AS A PUTATIVE BNC1 TARGET ...... 158 3.6.1 Heterogeneous SVIL expression in hPSC-epi subset ...... 158 3.6.2 SVIL expression falls during epi-EMT progression ...... 159 3.6.3 ChIP for BNC1 in epi-EMT: validating SVIL binding ...... 160 x

3.6.4 SVIL knockdown in epi-EMT model yields cortical actin phenotype ...... 161

3.7 INVESTIGATING TRANSCRIPTOMIC CHANGES INDUCED BY BNC1 KNOCKDOWN IN EPI- EMT ...... 164 3.7.1 Different BNC1 knockdown conditions segregate in three groups via Principal Component Analysis ...... 165 3.7.2 Differential expression analysis reveals changes in relative extracellular matrix and actin remodelling genes ...... 166 3.7.3 Gene ontology enrichment shows distinct phenotypic signatures for each epi- EMT population ...... 172 3.7.4 The d7 and d1 TET epi-EMT populations have a reduced cell motility phenotypic signature ...... 172 3.7.5 d1 TET epi-EMT has a more migratory phenotype than d7 TET epi-EMT . 175 3.7.6 Possible gene mechanisms for BNC1 knockdown epi-EMT phenotypes ..... 176 4. DISCUSSION ...... 178

4.1 BNC1: A KEY REGULATOR IN HPSC-EPI HETEROGENEITY AND FUNCTION ...... 178 4.1.1 BNC1 expression and function in hPSC-epi ...... 180 4.1.2 BNC1 in epi-EMT models ...... 182 4.1.3 Identifying putative BNC1 targets...... 185

4.2 ADVANTAGES AND LIMITATIONS OF OUR MODELS AND EXPERIMENTS...... 186 4.2.1 The hPSC-epi model...... 186 4.2.2 The epi-EMT model ...... 188 4.2.3 ChIP sequencing experiments ...... 190

4.3 FUTURE WORK TO INVESTIGATE BNC1 FUNCTION IN CELL-BASED ASSAYS ...... 191 4.3.1 Sequencing BNC1 knockdown hPSC-epi: network refinement? ...... 191 4.3.2 BNC1 overexpression assays ...... 192 4.3.3 Investigating BNC1 knockdown kinetics ...... 192 4.3.4 Manipulating hPSC-epi populations to ‘enhance’ hPSC-epi effects? ...... 193 4.3.5 Quantifying BNC1 function in epi-EMT models ...... 194 4.3.6 Exploring possible BNC1 targets ...... 195

4.4 FUTURE WORK: A FOCUS ON BNC1 IN VIVO ...... 197

4.5 CONCLUSION ...... 200 5. REFERENCES ...... 201 6. APPENDIX ...... 242 6.1 Carnegie stages in human embryos compared to gestational age (days) and mouse embryonic stage (days) ...... 242

6.2 Gene Ontology enrichment plot for each hPSC-epi subpopulation ...... 243 6.3 RNAscope reveals heterogeneous BNC1 expression in human foetal epicardium ...... 244 6.4 Double-staining for WT1 and BNC1 in human foetal epicardial explant shows both WT1/BNC1 double-positive and WT1/BNC1 single-positive cells ...... 245 6.5 PODXL and WT1 are co-expressed in epicardial cells of the human foetal heart ...... 246 6.6 Targeting rates for each sOPTiKD vector in h9 lipofection ...... 247 6.7 Different BNC1 sOPTiKD vectors mediate a similar relationship between BNC1 reduction, WT1 reduction and TCF21 increase, at mRNA and protein level ...... 248 6.8 Incubation of control sOPTiKD hPSC-epi cells with tetracycline does not mediate cell death ...... 250 6.9 BNC1 knockdown and selected EMT gene expression by qPCR ...... 251 APPENDIX II ...... 253

xii

LIST OF FIGURES

FIGURE 1. EARLY STAGES OF CARDIAC DEVELOPMENT ...... 31

FIGURE 2. SCHEMATIC SHOWING PE LOCATION AND MIGRATION RELATIVE TO DEVELOPING

HEART ...... 34

FIGURE 3. THE EMBRYONIC EPICARDIUM IS A HIGHLY ACTIVE AND DYNAMIC LINEAGE

DURING CARDIOVASCULAR DEVELOPMENT...... 38

FIGURE 4. EPITHELIAL CELLS UNDERGO ALTERATIONS IN CELL-CELL CONTACT AND

MORPHOLOGY IN EPITHELIAL-TO-MESENCHYMAL TRANSITION ...... 49

FIGURE 5. GAMBARDELLA ET AL FOUND TWO DISTINCT HPSC-EPI SUBPOPULATIONS ...... 60

FIGURE 6. CORE EPICARDIAL TRANSCRIPTIONAL NETWORK COORDINATED BY BNC1,

TCF21 AND WT1...... 37

FIGURE 7. AN EXAMPLE OF THE SIKD VECTOR WITH A BNC1 SHRNA INSERT UNDER THE

CONTROL OF A TETRACYCLINE-RESPONSIVE H1 PROMOTER ...... 74

FIGURE 8. HPSC-EPI DIFFERENTIATION FROM H9 EMBRYONIC STEM CELLS ...... 95

FIGURE 9. IMMUNOSTAINING FOR WT1 AND TCF21 SHOWS HETEROGENEOUS EXPRESSION

IN WILD-TYPE HPSC-EPI ...... 96

FIGURE 10. HUMAN FOETAL EPICARDIAL SAMPLES ARE PREPARED IN TWO WAYS ...... 98

FIGURE 11. HUMAN FOETAL EPICARDIAL EXPLANT CULTURES SHOW HETEROGENEITY

SIMILAR TO HPSC-EPI ...... 101

FIGURE 12. HUMAN FOETAL HEART EXHIBITS EPICARDIAL HETEROGENEOUS WT1

EXPRESSION ...... 102

FIGURE 13. HUMAN FOETAL EPICARDIAL EXPLANT CULTURES EXHIBIT HETEROGENEITY AT

DIFFERENT EMBRYOLOGICAL CARNEGIE STAGES ...... 103

FIGURE 14. BNC1 MRNA INCREASES DURING HPSC-EPI DIFFERENTIATION ...... 104

FIGURE 15. BNC1 IS EXPRESSED IN HUMAN FOETAL EPICARDIAL EXPLANT SAMPLES .... 105

FIGURE 16. PRINCIPAL COMPONENT ANALYSIS REVEALED PODXL AS A MARKER FOR

BNC1-HIGH CELLS ...... 107

xiii

FIGURE 17. PRINCIPAL COMPONENT ANALYSIS REVEALED THY1 AS A MARKER FOR TCF21-

HIGH CELLS ...... 108

FIGURE 18. HETEROGENEITY IN THY1 AND PODXL PROTEIN EXPRESSION IN HPSC-EPI ...... 109

FIGURE 19. THE PROPORTION OF PODXL-POSITIVE, THY1-NEGATIVE CELLS INCREASES

DURING HPSC-EPI DIFFERENTIATION ...... 110

FIGURE 20. BNC1 INDUCIBLE LINE DERIVATION FROM TRANSFECTION WITH SOPTIKD

VECTOR ...... 113

FIGURE 21. REPRESENTATIVE GELS FOR THE GENOTYPING OF SIKD CLONES ...... 114

FIGURE 22. CLONES ‘C’, ‘D’ AND ‘E’ HAVE SOME REDUCTION OF BNC1 MRNA AFTER

APPLICATION OF TETRACYCLINE ...... 116

FIGURE 23. ‘E’ IS THE MOST EFFECTIVE VECTOR IN MEDIATING BNC1 KNOCKDOWN .... 117

FIGURE 24. HETEROGENEOUS BNC1 EXPRESSION IN DIFFERENT INDUCIBLE KNOCKDOWN

HPSC-EPI LINES ...... 117

FIGURE 25. DIFFERENT SIKD LINES GENERATED VIA H9 LIPOFECTION WITH SIKD VECTOR

‘E’ ALL EXHIBIT SIGNIFICANT BNC1 KNOCKDOWN...... 118

FIGURE 26. BNC1 IS REDUCED AT THE PROTEIN LEVEL IN SIKD HPSC-EPI ...... 120

FIGURE 27. LOSS OF BNC1 DURING HPSC-EPI RESULTS IN ALTERED EXPRESSION OF

CANONICAL EPICARDIAL GENES WT1 AND TCF21...... 92

FIGURE 28. WT1 IS REDUCED AND TCF21 IS INCREASED AT THE PROTEIN LEVEL IN SIKD

EPI ...... 122

FIGURE 29. BNC1 KNOCKDOWN IN PRIMARY HUMAN FOETAL EPICARDIAL CELLS YIELDS AN

INCREASE IN TCF21 ...... 124

FIGURE 30. BNC1 REDUCTION INCREASES THE PROPORTION OF THY1-POSITIVE CELLS IN

HPSC-EPI ...... 126

FIGURE 31. BNC1 KNOCKDOWN HPSC-EPI DISPLAYS A FIBROBLASTIC MORPHOLOGY .. 127

FIGURE 32. BNC1 KNOCKDOWN INCREASES HPSC-EPI PROLIFERATIVE INDEX ...... 129

FIGURE 33. BNC1 KNOCKDOWN REDUCES HPSC-EPI VIABILITY ...... 131

FIGURE 34. TGFΒ TREATMENT INDUCES REARRANGEMENT OF ACTIN FILAMENTS INTO

STRESS FIBRES TO A MORE PRONOUNCED DEGREE THAN FGF2 TREATMENT ...... 133 xiv

FIGURE 35. AN IN VITRO MODEL OF HPSC-EPI EMT ...... 135

FIGURE 36. TREATING HPSC-EPI WITH TGFΒ ALTERS RELATIVE HPSC-EPI SUBPOPULATION

RATIO ...... 136

FIGURE 37. BNC1 KNOCKDOWN IN AN IN VITRO MODEL OF EMT MEDIATES A CELLULAR

CORTICAL ACTIN PHENOTYPE ...... 138

FIGURE 38. SIKD EPI-EMT EXHIBITS LESS PRONOUNCED CORTICAL ACTIN LOCALISATION

PHENOTYPE THAN ACUTE EPI-EMT BNC1 KNOCKDOWN ...... 140

FIGURE 39. SEVERITY OF CORTICAL ACTIN PHENOTYPE APPEARS TO DEPEND ON WHEN

BNC1 IS KNOCKED DOWN DURING HPSC-EPI DIFFERENTIATION...... 110

FIGURE 40. A CORTICAL ACTIN LOCALISATION PHENOTYPE IS SEEN IN SIKD EPI-EMT

INDUCED BOTH BY TGFΒ AND FGF2 ...... 142

FIGURE 41. SCHEMATIC SHOWING EXPERIMENTAL DESIGN FOR TRACKING 2D EPI-EMT

MIGRATION...... 144

FIGURE 42. MEDUSA PLOTS COMPARING 2D MIGRATION OF BNC1 KNOCKDOWN EPI-EMT

CELLS ...... 145

FIGURE 43. BNC1 KNOCKDOWN IN EPI-EMT CELLS ...... 146

FIGURE 44. INCELL-TRACKED EPI-EMT CELLS DISPLAY REDUCED MOTILITY WHEN BNC1

IS KNOCKED DOWN ...... 147

FIGURE 45. HPSC-EPI INVASION ASSAY SCHEMATIC ...... 149

FIGURE 46. INVASION INTO THE CM LAYER BY HPSC-EPI TRENDS TOWARDS IMPAIRMENT

WHEN BNC1 IS KNOCKED DOWN ...... 150

FIGURE 47. THE EXPERIMENTAL WORKFLOW FOR CHIP-SEQ IN IDENTIFYING POTENTIAL

BNC1 TARGETS ...... 152

FIGURE 48. CHIP SHEAR TEST TO OPTIMISE THE NUMBER OF SONICATION CYCLES ...... 153

FIGURE 49. CHIP-SEQ PEAKS FOR SELECTED PUTATIVE TARGETS HACD1 AND SVIL ... 156

FIGURE 50. CHIP-QPCR TO VALIDATE SELECTED BNC1 TARGETS IDENTIFIED BY CHIP-SEQ ...... 157

FIGURE 51. PRINCIPAL COMPONENT ANALYSIS COLOURED FOR SVIL EXPRESSION ...... 159

FIGURE 52. SVIL MRNA LEVEL IS REDUCED IN THE EPI-EMT MODEL ...... 160

FIGURE 53. SVIL ENRICHMENT IN EPI-EMT CHIP ...... 161

FIGURE 54. SVIL KNOCKDOWN IN EPI-EMT MODEL MEDIATES SIMILAR CORTICAL ACTIN

PHENOTYPE TO THAT SEEN IN BNC1 KNOCKDOWN...... 162

FIGURE 55. BNC1 KNOCKDOWN DOES NOT ALTER SVIL EXPRESSION IN THE EPI-EMT

MODEL...... 130

FIGURE 56. PRINCIPAL COMPONENT ANALYSIS SHOWS CLEAR SEPARATION FOR THE THREE

BNC1 KNOCKDOWN CONDITIONS FOR BULK SEQUENCING IN EPI-EMT...... 165

FIGURE 57. VENN DIAGRAM SHOWING THE NUMBER OF DIFFERENTIALLY EXPRESSED GENES

BETWEEN DIFFERENT EPI-EMT CONDITIONS ...... 166

FIGURE 58. VOLCANO PLOT SHOWING GENES THAT ARE DIFFERENTIALLY EXPRESSED

BETWEEN D7 TET EPI-EMT AND NO TET EMT ...... 168

FIGURE 59. VOLCANO PLOT SHOWING DIFFERENTIAL GENE EXPRESSION ANALYSIS

BETWEEN THE D1 TET EPI-EMT AND NO TET EPI-EMT POPULATIONS ...... 169

FIGURE 60. VOLCANO PLOT SHOWING DIFFERENTIAL GENE EXPRESSION ANALYSIS

BETWEEN THE D1 TET AND D7 TET EPI-EMT POPULATIONS ...... 171

FIGURE 61. GENE ONTOLOGY ENRICHMENT PLOT FOR D7 TET EPI-EMT RELATIVE TO NO

TET EPI-EMT ...... 173

FIGURE 62. GENE ONTOLOGY ENRICHMENT PLOT FOR D1 TET EPI-EMT RELATIVE TO NO

TET EPI-EMT ...... 174

FIGURE 63. GENE ONTOLOGY ENRICHMENT PLOT FOR D7 TET EPI-EMT RELATIVE TO D1

TET EPI-EMT ...... 176

FIGURE 64. A POSSIBLE MODEL FOR HPSC-EPI CELL RESPONSES TO TGFΒ ...... 183

LIST OF SUPPLEMENTARY FIGURES

FIGURE S1. PREDICTED TISSUE AND CELLULAR SPECIFICITIES OF BNC1HIGH AND

TCF21HIGH CELLS...... 206

FIGURE S2. RNASCOPE FOR BNC1 IN HUMAN HEART ...... 244

FIGURE S3. WT1 AND BNC1 ARE BOTH DETECTED BY IMMUNOFLUORESCENCE IN

EPICARDIAL EXPLANT CULTURES FROM EMBRYONIC HUMAN HEART AT 8 WEEKS GA ...... 245 xvi

FIGURE S4. THERE IS APPARENT CO-EXPRESSION OF PODXL AND WT1 IN SOME

EPICARDIAL CELLS IN HUMAN FOETAL HEART, SHOWN BY IMMUNOHISTOCHEMISTRY ...... 246

FIGURE S5. BNC1 REDUCTION MEDIATES WT1 REDUCTION AND TCF21 INCREASE IN

DIFFERENT CLONES VIA DIFFERENT SOPTIKD VECTORS ...... 248

FIGURE S6. CONTROL B2M VECTOR SIKD HPSC-EPI CULTURED UNDER THE PRESENCE OF

TETRACYCLINE DOES NOT EXHIBIT CELL DEATH ...... 250

FIGURE S7. BNC1 KNOCKDOWN FROM D5 OF HPSC-EPI DIFFERENTIATION CAUSES NO

SIGNIFICANT DIFFERENCE IN EXPRESSION OF GENES ASSOCIATED WITH EMT BY QPCR ...... 251

LIST OF TABLES

TABLE 1. SIRNA SEQUENCES FOR SIRNA-MEDIATED KNOCKDOWN EXPERIMENTS ...... 73

TABLE 2. SHRNA SEQUENCES FOR PSOPTIKD VECTOR CONSTRUCTION. BGLII OVERHANG

IS IN RED, TERMINATOR SEQUENCE/SALI OVERHANG IN BLUE. HAIRPIN LOOP

SEQUENCE IS IN BOLD. STEM IS UNDERLINED...... 75

TABLE 3. PCR PRIMER SEQUENCES AND CYCLING CONDITIONS FOR GENOTYPING PSOPTIKD

CLONES ...... 79

TABLE 4. ANTIBODIES USED IN IMMUNOFLUORESCENCE EXPERIMENTS ...... 81

TABLE 5. ANTIBODIES USED IN FLOW CYTOMETRY ...... 83

TABLE 6. QRT-PCR PRIMER SEQUENCES ...... 85

TABLE 7. ANTIBODIES FOR WESTERN BLOT ...... 86

TABLE 8. CHIP QPCR PRIMERS ...... 90

TABLE 9. BNC1 CHIP-SEQ PEAKS IDENTIFIED BY DR KNIGHT-SCHRIJVER AFTER DATA

CLEAN-UP IN BNC1 CHIP-SEQ...... 156

xvii

LIST OF SUPPLEMENTARY TABLES

TABLE S1. A TABLE TO SHOW AN APPROXIMATE COMPARISON BETWEEN HUMAN

GESTATIONAL AGE, CARNEGIE EMBRYO STAGING, AND EMBRYONIC DAY IN MOUSE ...... 243

TABLE S2. SUMMARY TABLE OF GENOTYPING RESULTS FOR SIKD CLONES ...... 247

LIST OF ABBREVIATIONS AND ACRONYMS

AAV-prom AAVS1 locus promoter

APLNR Apelin receptor

BAMBI BMP and Activin Membrane Bound Inhibitor

BCAM Basal cell adhesion molecule

BMP Bone morphogenetic protein

BNC1 Basonuclin 1 bp Base pairs

CAG CAG promoter

CDH1 E-cadherin

CDH2 N-cadherin

CF Cardiac fibroblasts

CM Cardiomyocytes

COL11A1 Alpha-1 type I collagen

CoSMC Coronary smooth muscle cells

CS Carnegie Stage

CVD Cardiovascular disease

DAPI 4',6-diamidino-2-phenylindole

DEseq Differential expression analysis

EC Endothelial cells

ECM Extracellular matrix xviii

EDN1 Endothelin 1

EGFR Epidermal growth factor receptor

EMT Epithelial-to-Mesenchymal Transition

EPDC Epicardium-derived cells

Epi-EMT In vitro models of EMT

ERKs Extracellular signal-regulated kinases

F-actin Fibrillar actin

FGF Fibroblast growth factor

FHF First heart field

G-actin Globular actin

GA Gestational Age

GO Gene Ontology

H1 H1 human RNA polymerase III promoter

HA Hyaluronan

HAR Homology arm hf-epi Primary human foetal epicardial culture

HH Hamburger-Hamilton hPSC Human pluripotent stem cells hPSC-epi Human pluripotent stem cell-derived epicardium hPSC-epi d1 Day 1 of hPSC-epi differentiation hPSC-epi d3 Day 3 of hPSC-epi differentiation hPSC-epi d5 Day 5 of hPSC-epi differentiation hPSC-epi d7 Day 7 of hPSC-epi differentiation

ISM1 Isthmin 1

L1CAM L1 cell adhesion molecule protein

LAYN Layilin

LAMC2 Laminin γ-2

xix

LPM Lateral plate mesoderm

LTBP1 Latent transforming growth factor beta binding protein 1

MAPK Mitogen activated protein kinase

MAP3K14 Mitogen activated protein kinase kinase kinase 14

MEF2C Myocyte-specific enhancer factor 2C

MEOX1 Mesenchyme homeobox 1

MGP Matrix gla protein

MI Myocardial infarction

MMP Matrix metalloproteinase

MRTFs Myocardin-related transcription factors

MTCL1 Microtubule crosslinking factor 1

MXRA5 Matrix remodelling associated 5

MYOCD Myocardin

NCDH1 N-cadherin

NPNT Nephronectin

Nfatc1 Nuclear factor of activated T-cells cytoplasmic 1

OPTtetR Optimised TET repressor pA Polyadenylation signal

PCA Principal component analysis

PODXL Podocalyxin

PS Primitive Streak

Puro Puromycin resistance qPCR Quantitative real-time PCR

RA Retinoic Acid

RALDH2 Retinaldehyde-dehydrogenase 2

RISC RNA-induced silencing complex

RNAscope/ smRNA-FISH Single-cell RNA in-situ hybridisation

RNAseq RNA sequencing

RXR Retinoid X receptors

SA Splice acceptor

Sca-1 Stem cell antigen-1 siKD Inducible knockdown

SNAI1 Snail sOPTikd Optimised single-step inducible knockdown vector

SPINT1 Serine Protease Inhibitor, Kunitz Type 1

SPTBN2 Beta-III-spectrin

T2A Self-cleaving T2A peptide

T/Bry Brachyury

Tβ4 Thymosin beta-4

TCF21 Transcription factor 21

TET Tetracycline tetR Tet repressor

TGFβ Transforming growth factor beta

THY1 Thy-1 Cell Surface Antigen

TO Tet operon

VASN Vasorin

VIM Vimentin

WNT Wingless-integrated integration site

WT1 Wilm’s Tumour 1

ZO1 Zona-occludens 1

xxi

LIST OF APPENDICES

APPENDIX I 242

APPENDIX II 253

xxii Chapter 1: Introduction

1. INTRODUCTION

The focus of this thesis is the regulation of epicardial development and response to myocardial injury. The rationale behind focussing on the epicardium pertains to its key role in the cardiac injury response. In our group, we hypothesise that understanding epicardial development and the injury response will facilitate or promote heart regenerative strategies. In this Introduction, I will first describe the existing clinical challenge, and touch on previous avenues in the regenerative medicine field, focusing on studies that suggest a key role for the epicardium. To provide background for understanding both epicardial development and the injury response, I will discuss cardiac and epicardial development, touching on common regulatory modules conserved across species, as well as how these inform our current understanding of heart regeneration. Common themes will be explored that may specifically inform human epicardial development from mesoderm and its subsequent functional regulation. Finally, I will discuss the preliminary data obtained in our group on the cellular and molecular regulation of epicardial development, epicardial heterogeneity and function, and the identification of a putative high-level regulator of epicardium, BNC1, that have formed the bases for my thesis questions on epicardial development and epithelial-to- mesenchymal transition.

Sophie McManus – February 2020 23

BNC1 regulates human epicardial heterogeneity and function

1.1 The heart The heart is the first organ to develop in the vertebrate embryo (Sissman, 1970) (DeRuiter, Poelmann, VanderPlas-de Vries, Mentink, & Gittenberger-de Groot, 1992) (Buckingham, Meilhac, & Zaffran, 2005). Throughout life this organ sustains a formidable workload, constantly adjusting to alterations in demand for oxygen by other organs via finely-tuned feedback mechanisms. Continual and efficient cardiac function is enabled by the heart’s specialised structure: the heart is composed of multiple cell types, each with their own key functions. These include cardiomyocytes (CM), endothelial cells (EC), coronary smooth muscle cells (coSMC), cardiac fibroblasts (CF), heart valve cells, endocardial cells and the epicardium, the epithelial lineage covering the heart (Viragh & Challice, 1981). Correct cardiogenesis involves exquisitely regulated coordination of myriad signalling pathways to enable differentiation and integration of different cardiac cells. Dysregulation of these pathways in development may result in congenital heart defects, the commonest type of birth defect worldwide, which are present in around 1 in 100 live births (van der Linde D, Konings E E M, Slager M A, Witsenburg M, Helbing W A, Takkenberg et al., 2011). Furthermore, adult cardiovascular dysfunction remains the world’s leading cause of death (World Health Organisation).

1.1.1 Cardiovascular disease

According to the British Heart Foundation, there are seven million people living in the UK with some form of heart or circulatory disease. Today, while more people than ever survive myocardial infarction (MI) (Gitsels, Kulinskaya, & Steel, 2017), (Krumholz, Normand, & Wang, 2019) the global cardiovascular disease (CVD) burden continues to grow. Cardiovascular disease commonly arises due to coronary artery disease (CAD), the progressive development of atherosclerotic plaques in major vessels: consequences of these atherosclerotic lesions include increased risk of MI.

During myocardial infarction, up to a billion cardiomyocytes die via necrosis and apoptosis, and a huge degree of cardiac pathological remodelling occurs. While emergence of a fibrotic scar stabilises the injured tissue, it also severely impairs contractile function (González, Schelbert, Díez, & Butler, 2018) (Segura, Frazier, & Buja, 2014), with obvious ramifications for the patient’s quality and length of life. Current therapeutic options for patients with CVD include medication, such as the use of aspirin,

24 Sophie McManus - February 2020

Chapter 1: Introduction

statins and beta-blockers; substantial alterations in lifestyle, such as diet and exercise regimes; surgical procedures, and heart transplantation. The latter still represents the only definitive ‘cure’ for heart failure. However, these options all carry shortcomings; in particular, the shortage of donor hearts poses a huge logistical challenge. Patients who do receive a heart transplant will often see great improvements to both their quality and duration of life, but must commit to a heavy regime of immunosuppressant drugs, and contend with their many undesirable side-effects.

1.1.2 Cardiovascular regeneration after injury

Unlike humans, some organisms can regenerate and revascularise the infarcted heart, effectively restoring healthy contractile function. These include the neonatal mouse (Enzo R Porrello et al., 2011), (E. R. Porrello et al., 2013) (Mahmoud, Porrello, Kimura, Olson, & Sadek, 2014) (Polizzotti, Ganapathy, Haubner, Penninger, & Kühn, 2016) (Gunadasa- Rohling et al., 2018), wherein pronounced cardiac regeneration was observed both in apical resection models and coronary ligation injury, and the adult zebrafish, in surgical resection, cryoinjury models and genetic CM ablation (Poss, Wilson, & Keating, 2002) (Kikuchi et al., 2010) (Gonzalez-Rosa, Martin, Peralta, Torres, & Mercader, 2011) (Jinhu Wang et al., 2011). While the mouse heart appears to lose its regenerative ability by postnatal day 7 (P7) (Enzo R Porrello et al., 2011), this is not the case for the zebrafish, which sustains a remarkable regenerative response through adulthood (Poss et al., 2002). Neonatal pigs injured before P3 also display a cardiac regenerative response (Zhu et al., 2018). Biologists have long-hypothesised that this reparative capacity – which extends across multiple tissues in organisms such as axolotls, newts and teleost fish – is an ancestral capacity, maintained in more ‘primitive’ organisms, but lost from mammals (Lepilina et al., 2006). By that logic, we may hope to re-harness or differentially manipulate such regenerative machinery within humans.

Of note, there is also intriguing – albeit sparse – clinical evidence that the human infant possesses a similar capacity to heal and regenerate the injured heart (Cesna, Eicken, Juenger, & Hess, 2013). In one particularly compelling case, a newborn baby who suffered severe cardiac injury due to coronary occlusion showed complete recovery a few weeks post-MI (Haubner et al., 2016). Serum markers for cardiac damage fell to normal levels by two weeks post-injury and later follow up indicated normal heart function

Sophie McManus – February 2020 25

BNC1 regulates human epicardial heterogeneity and function

thereafter. Furthermore, an interesting study (Drenckhahn et al., 2008) investigated cardiac regeneration during the murine embryonic period, via conditional ablation of a portion of cardiomyocytes. A CM-lethal gene was conditionally expressed in some CM of E12.5 female embryos via random X-inactivation. CM-ablated hearts showed rapid restoration of 50% of the ‘lost’ CM, leading the authors to conclude that environmental conditions in the embryo favoured cell-cycle re-entry and CM proliferation. While unknown mechanisms lead to the marked reduction of this beneficial embryonic and early postnatal response to injury, there is some evidence for a limited degree of CM turnover and proliferation in humans, demonstrated via cellular incorporation of radioactive carbon-14. This established that CM can renew at a rate of up to 1% per year (Bergmann et al., 2009; Yutzey, 2017). Although this is a low rate, and obviously not enough to benefit the devastated post-infarct heart, it still presents the possibility that CM proliferation could be stimulated in a therapeutic manner, to enable some degree of the sophisticated regenerative capacity seen in more ‘primitive’ animals.

Heart function is supported by many cell types other than CM, including fibroblasts, which make up approximately 20% of the cells within the heart, and provide essential structural support and aid electrical conduction (M. Fang, Xiang, Braitsch, & Yutzey, 2016; Furtado, Nim, Boyd, & Rosenthal, 2016; Pinto et al., 2016). Following cardiac injury however, there is amplification of resident CF and recruitment of haematopoietic cells to the site of injury and their transformation into CF (Humeres & Frangogiannis, 2019). The post-injury fibrosis may also be modulated by epicardial function; for example, interstitial fibrotic regions in injured hearts express epicardial markers (C. M. Braitsch, Kanisicak, van Berlo, Molkentin, & Yutzey, 2013). Hence a putative regenerative response might entail manipulation of the cardiac fibrosis mechanisms that emerge post-MI.

Today there is a great deal of research interest in elucidating the mechanisms that allow a regenerative response, whether in fish or in rodent, with the aim of manipulating these – whether that be to enhance cardiomyocyte proliferation, or modulate cardiac fibrosis – to eventually induce a similar outcome in adult human patients.

26 Sophie McManus - February 2020

Chapter 1: Introduction

1.1.3 Cellular therapy for cardiac regeneration

Stem cells have been considered as a source of hope for regenerative medicine, in particular for cardiovascular medicine, but this field has been dogged by controversy and fraud. The early 2000s saw the publication of papers showing heart regeneration from bone-marrow derived stem cells, or an endogenous cardiac population of c-kit-positive stem cells (Beltrami et al., 2003; Orlic et al., 2001). It was argued that over half of the injured heart could be repaired by bone marrow-derived or c-kit positive stem cells, despite robust evidence to the contrary, as well as failure to reproduce these results (Murry et al., 2004; van Berlo et al., 2014). Nonetheless, the furore surrounding cardiac stem cells resulted in several clinical trials (Assmus et al., 2006) (K. et al., 2006) (E.J. van den Bos, W.J. van der Giessen, 2008) (Steele, Macarthur, & Woo, 2017). These have yielded negligible benefits to patients. Overall, these cells do not appear to offer the regenerative capacity that was once promised. To date, there have been 18 retractions for work from the Anversa group, and more are recommended (Retraction Watch) (Ozkan, 2019).

Alternative options for improving cardiac function after MI involve purification of cardiac progenitors or cardiomyocytes differentiated from human embryonic or induced pluripotent stem cells, (Burridge, Keller, Gold, & Wu, 2012) (Burridge et al., 2014) for subsequent injection or engrafting. In this vein, use of cardiac patches, or engineered heart tissues (EHTs) has yielded promising results in a variety of animal models, including pigs (Zimmermann et al., 2006) (J. Tang et al., 2018; Xiong et al., 2011; Ye et al., 2014) (J. Zhang, Zhu, Radisic, & Vunjak-Novakovic, 2018) (Curtis & Russell, 2009) (Park & Yoon, 2018). Furthermore, the function of such EHTs has been significantly enhanced by the addition of stem cell-derived epicardial cells (Bargehr et al., 2019). The benefits of this synergistic approach will be discussed in more detail in Section 1.6. While these studies are promising, they are still at the pre-clinical stage. A common theme among research approaches regarding possible cardiac repair is the involvement of the epicardium, which is the epithelial lineage that covers the heart and governs cardiac development. For example, the epicardium can be successfully primed to derive de novo cardiomyocytes post-injury. Epicardium has also been used in transplantation approaches to encourage restoration of cardiac function (Bargehr et al., 2019; Smart et al., 2007; Elizabeth M. Winter et al., 2009). The promise of the epicardium in cardiac therapy to date is discussed in detail in Section 1.6.4.

Sophie McManus – February 2020 27

BNC1 regulates human epicardial heterogeneity and function

To bolster efforts to repair the damaged heart, there is substantial research interest in understanding endogenous developmental mechanisms that ‘build’ a healthy heart, with the aim of manipulating these for therapeutic gain. Development and disease are often regarded as different sides of the same coin. Robust understanding of developmental mechanisms and their collective coordination is key if we are to recapitulate such mechanisms in order to repair or regenerate the injured heart (Freire, Resende, & Pinto- Do-Ó, 2014). Hence honing in on specific signalling events, heterogeneous gene expression patterns and transcriptional networks that are active in embryogenesis has relevance to the wider field of cardiac regeneration, in a form of ‘bottom-up’ approach.

1.2 Cardiovascular origins: mesoderm development The heart is composed of three layers: the endocardium, the myocardium and the epicardium (Santini, Forte, Harvey, & Kovacic, 2016) each with clear functional roles; the cardiovascular layers are predominantly mesoderm-derived, with some additional contribution from the anterior neural crest (Vincent & Buckingham, 2010), although I will mainly focus on the role of the mesoderm. Gastrulation represents the critical point of embryogenesis, during which the three germ layers (endoderm, mesoderm and ectoderm) are defined from the blastula. Formation of the primitive streak (PS) occurs during gastrulation, and this developmental event is conserved across avian and mammalian embryos. PS formation occurs at Hamburger-Hamilton (HH) stage 2 in chick, at embryonic day E6.0 in mouse, and at the third week of gestation in human (Carnegie Stage 6). The primitive streak lies at the centre of the gastrulating embryo (Downs, 2009). During gastrulation, epithelial-to-mesenchymal transition (EMT) occurs at the primitive streak, and endoderm and mesoderm migrate bilaterally; some pre-cardiac mesoderm cells subsequently migrate anteriorly, start expressing myocardial markers, and form the cardiac crescent in mammals (Tam, Parameswaran, Kinder, & Weinberger, 1997) (Buckingham et al., 2005). Of note, Auda-Boucher et al also found that some cardiac precursor cells are present pre-gastrulation (Auda-Boucher et al., 2000).

1.2.1 Mesoderm specification: genes and signalling

Mesoderm induction is accomplished along the embryonic posterior/anterior axis via the integration of complex signalling pathways. Mesoderm can be subdivided into paraxial

28 Sophie McManus - February 2020

Chapter 1: Introduction

mesoderm, which gives rise to somites; intermediate mesoderm, which forms the urinary and reproductive systems; and lateral plate mesoderm (LPM), which has a major role in generating the cardiovascular system (Gilbert, 2000). I will briefly highlight the main signalling pathways and genes involved in induction of LPM and its derivative cardiac cells in different vertebrate species.

A series of elegant experiments in chick embryos and explant cultures revealed that subtypes of chicken endoderm can signal to induce expression of NKX2.5, a cardiac marker gene, in presumptive LPM (Schultheiss, Xydas, & Lassar, 1995). Later, LPM migration out of the PS was impacted by WNT3a and WNT5a signalling (Sweetman, Wagstaff, Cooper, Weijer, & Münsterberg, 2008). Post-gastrulation, controlled levels of BMP proteins are crucial in sustained induction of lateral plate mesoderm (Schultheiss, Burch, & Lassar, 1997) (Andrée, Duprez, Vorbusch, Arnold, & Brand, 1998). Cooperative signalling between BMP2 and FGF8 has been implicated in induction of chick cardiac mesoderm (Alsan & Schultheiss, 2002) and this targets NKX2.5 (Lee, Evans, Ruan, & Lassar, 2004). Canonical WNT signalling inhibition in LPM enables cardiac mesoderm specification (Marvin, 2001). In summary, chick LPM specification relies on coordination of BMPs, WNTs and FGFs.

Alongside avian models, the zebrafish represents a highly manipulable and useful model organism in approaching questions in developmental biology, particularly as the embryo is transparent, hence allowing easy visualisation of the cardiovascular system. There are similarities with other vertebrates during zebrafish cardiac mesoderm specification; for example, nkx2.5 is an early zebrafish cardiac mesoderm marker and fgf8 is essential to zebrafish heart development (Reifers, F., Walsh E., Leger S., Stainier D, 2000) . Moreover, coordinated wnt signalling is important in zebrafish, as in chick (Ueno et al., 2007). Regarding bmp signalling, Marques et al found that loss of bmp receptor alk8 resulted in failure of atrial tissue formation from the LPM (Marques & Yelon, 2009). Hence the signalling molecules key to forming chick cardiac mesoderm are generally conserved in zebrafish, though the temporal control of their influences may differ; for example, the zebrafish displays discrete phases of wnt/β-catenin signalling (Dohn & Waxman, 2012).

Sophie McManus – February 2020 29

BNC1 regulates human epicardial heterogeneity and function

Signalling conservation also exists with regard to cardiac mesoderm specification in chick and in mouse. Conditional deletion of Bmp receptors in cardiac mesoderm Mesp1-positive cells caused inadequate cardiac crescent formation, with reduced Nkx2.5 and Isl1 expression (A. Klaus, Saga, Taketo, Tzahor, & Birchmeier, 2007). However inhibition of canonical Wnt effector β-catenin did not impact cardiac crescent formation from lateral plate mesoderm (although the heart tube did not loop properly, and right ventricular formation was stunted). The same authors showed that Wnt is involved in mediating Nkx2.5 and Isl1 expression in mouse SHF (Alexandra Klaus et al., 2012). Recently, Row et al characterised a conserved mechanism of Bmp- and Fgf-mediated mesodermal patterning in both zebrafish and mouse (Row et al., 2018). There are obvious hurdles associated with understanding early human mesoderm signalling. To overcome such challenges, several groups have developed in vitro models of PS development and germ layer derivative formation from pluripotent stem cells, with the aim of elucidating similarities and differences between the human state and what is known from animal models (Koh et al., 2016; Loh et al., 2016). .

1.2.2 Cardiac lineage formation

Cardiac lineages arise from the lateral plate mesoderm once mesoderm cells have egressed from the PS into the lateral anterior portion of the embryo (Lawson & Pedersen, 1992) (Tam et al., 1997). When mesoderm is newly specified in the PS, markers such as MESP1 and MESP2 are expressed; early markers of lateral plate mesoderm (LPM), which will become cardiac mesoderm (Saga et al., 1999) (Brand, 2003). With regard to heart development, LPM can be further subdivided into two layers; the somatic mesoderm (dorsal) and the splanchnic mesoderm (ventral). Dorsal LPM will form bones, ligaments, blood vessels, and connective tissue of the limbs, whereas cells within the LPM splanchnic mesoderm coalesce into the aforementioned murine cardiac crescent by E7.5 (Kelly, Buckingham, & Moorman, 2014). A schematic of early cardiogenesis is shown in Figure 1.

30 Sophie McManus - February 2020

Chapter 1: Introduction

Figure 1. Early stages of cardiac development, from ventral view of the murine embryo. (A) At around E6.0, cardiac mesoderm progenitors egress from the primitive streak, PS (egression indicated by yellow arrows). (B) At E7.0 a subset of cardiac mesoderm cells form the cardiac crescent (red) first heart field (FHF) – second heart field (SHF) cells are recruited medially to the FHF (green). (C) By E8.0 the heart tube has formed and begun beating. SHF cells are continually recruited to the tube, which elongates at the arterial pole (AP) and venous pole (VP). (D) At E8.5 the heart loops and continues expanding. Presumptive right atrium (PRA) and left atrium (PLA) are shown, as well as the outflow tract (OFT), right ventricle (RV) and left ventricle (LV); the heart has not yet gained its four-chamber structure.

The first heart field (FHF) forms from the cardiac crescent (Brade, Pane, Moretti, Chien, & Laugwitz, 2013). When the FHF is exposed to BMPs and FGFs, there is induction of genes including nascent cardiac markers Nkx2.5, Tbx5 and Gata4 (Hatcher, Goldstein, Mah, Susan Delia, & Basson, 2000; Lints, Parsons, Hartley, Lyons, & Harvey, 1993) (Heikinheimo, Scandrett, & Wilson, 1994). By E8.0 an early heart tube has formed (Harvey, 2002). In humans, the heart tube forms during the third week of gestation (Carnegie Stages 9-10). The mammalian linear heart tube is formed from FHF cells (Buckingham et al., 2005) (Lescroart et al., 2014) (Brade et al., 2013), which egress before counterpart second heart field (SHF) cells and have different markers (Lescroart et al., 2014). The SHF was identified in mouse by fate-mapping experiments and explant manipulation (Mjaatvedt et al., 2001). SHF cells robustly express ISL1, (C.-L. Cai et al., 2003) and contribute to the cardiac outflow tract, a large portion of the inflow region (putative atria) and right ventricle (Zaffran, Kelly, Buckingham, 2004) (Brade et al., 2013), whereas FHF cells generate the left ventricle. Recently, multiphoton confocal

Sophie McManus – February 2020 31

BNC1 regulates human epicardial heterogeneity and function

analysis has provided an elegant tool for imaging the dynamics of FHF cells and SHF cells in the murine heart tube via use of fluorescent reporters (Ivanovitch, Temiño, & Torres, 2017) (Reiter et al., 1999) (DY et al., 1996) (Serluca, 2008) (Hami, Grimes, Tsai, & Kirby, 2011) (Martinsen, 2005) (Mjaatvedt et al., 2001) (Waldo et al., 2001).

The mammalian linear heart tube recruits cells from the second heart field (SHF), elongates, begins beating, and undergoes looping by E8.5 - E9.0 (Buckingham et al., 2005) (Brade et al., 2013). Looping involves involvement of genes such as Lefty and Nodal (Harvey, 2002), and allows chamber expansion and the appearance of the heart’s characteristic morphology; the heart is thereby fully functional at birth (Bruneau, 2013). Up until looping, critically, the heart is composed of just two of the types of cardiac layer, the myocardium and the endocardium. At E9.5, there is migration of cells from the proepicardial organ (PE) onto the looping heart; this PE attaches and expands to cover the heart with an epithelial layer (epicardium), the third cardiac cell type, which is a crucial event to allow the next wave of cardiovascular development. I shall briefly address the developmental events underlying epicardial formation from PE during embryonic development, before outlining what is known regarding epicardial functions across different contexts, alongside current gaps in understanding and how these may be broached. Understanding how the epicardium develops offers scope for its potential manipulation.

1.3 Proepicardial formation, signalling and migration The proepicardium is a transient collection of proliferative cells present in various vertebrate species at the sinus venosus between the heart and liver (J. Liu & Stainier, 2010; Smits, Dronkers, & Goumans, 2018) (Figure 2). This is a signalling node that orchestrates correct cardiac development, migrating to the heart tube and forming a continuous epicardial layer 3-5 days post-fertilisation in zebrafish, approximately HH17- 26 in chick, by E10-11 in mouse, and between embryonic Carnegie stage 11-15 in human (Viragh & Challice, 1981) (Nahirney, Mikawa, & Fischman, 2003) (Jinhu Wang, Cao, Dickson, & Poss, 2015) (Hirakow, 1992) (C. A. Risebro et al., 2015). (A comparison table for mouse and human embryo staging is included in the Appendix, Table S1.) While the chick PE has been described as containing two types of cell population, this is species- specific, and there has been no equivalent morphological heterogeneity described in

32 Sophie McManus - February 2020

Chapter 1: Introduction

mammalian PE (Rodgers, Lalani, Runyan, & Camenisch, 2008). The human proepicardium has not been described.

Epicardial importance has been long-recognised, since a host of experiments wherein PE disturbance impaired heart formation, and quail-chick PE grafts resulted in swathes of epicardium formation at the transplant site in the host organism (Männer, 1993) (José María Pérez-Pomares, Macías, García-Garrido, & Muñoz-Chápuli, 1998). PE disruption has been induced via various surgical methods, particularly in avian models, and consequences of this include dramatic cardiac defects, such as failure to form coronary vasculature and the pronounced ballooning of a thinned ventricular wall (Gittenberger- De Groot, Vrancken Peeters, Bergwerff, Mentink, & Poelmann, 2000). More recently, mouse and zebrafish work has shed further light on mechanisms governing PE development, hence epicardial development. Not all epicardium is PE-derived; some epicardial portions are derived from pericardial epithelium (José M. Pérez-Pomares, Phelps, Sedmerova, & Wessels, 2003). Furthermore, a portion of epicardium forms from a second proepicardial organ at the arterial pole, which covers a region of intra-pericardial aorta and the pulmonary trunk (Adriana C. Gittenberger-de Groot et al., 2012). Overall, however, most of the epicardium is derived from the transient PE that arises in septum transversum once the heart has formed.

Although many cardiac developmental mechanisms are conserved across different species, as outlined in Section 1.2, differences have been described with regards to PE origins, signalling, and migration to the developing heart tube. Such inter-species differences are an example of why studying epicardial biology in a human model is so important. I shall briefly outline the consensus on proepicardial signalling in different animal models.

Sophie McManus – February 2020 33

BNC1 regulates human epicardial heterogeneity and function

Figure 2. Schematic showing PE location and migration relative to developing heart. The developing heart is composed of two layers, the endocardium (endo) and the myocardium (myo). The proepicardium (PE) is forms by the sinus venosus (SV) distal to the developing heart. PE cells migrate towards the heart and attach (some attachment shown, epi cells). By E10.5 the heart is covered by a continuous epicardial layer. Schematic adapted from a design by Servier Medical Art, freely licensed under Creative Commons Attribution Unported Licence 3.0.

1.3.1 Avian models

The origin of the chick PE is a topic of some debate. In the chick, fate-mapping studies have provided some evidence for a somatic mesoderm origin for PE, which develops asymmetrically (as in many vertebrate embryos) with development known to be mediated by FGF8 and SNAI1. TWIST1 appears to be a key signal in this context (Schlueter & Brand, 2013), as its suppression led to disturbances in PE villous projections, as well as dysregulated WT1 and TCF21 expression.

34 Sophie McManus - February 2020

Chapter 1: Introduction

As outlined in Section 1.3, the migration and attachment of the PE to the heart tube is species-dependent, and in recent years different mechanisms in various models have been characterised (Y. Cao & Cao, 2018). In the chick, tissue bridging between villous PE projections and the myocardium was observed via electron microscopy (Nahirney et al., 2003); tissue bridges were positive for heparan sulphate and fibronectin, and heparinase injection disrupted PE movement. The signalling underlying this bridge formation and cell movement is thought to be predominantly BMP-based, as Ishii et al demonstrated that ectopic BMP application in chick caused excessive PE targeting to the heart, whereas Noggin expression obliterated PE-myocardial attachment (Ishii, Garriock, Navetta, Coughlin, & Mikawa, 2010). TBX5 has also been implicated in chick PE migration in a dose-dependent manner (Hatcher et al., 2004): PE cells transfected with an antisense oligonucleotide against TBX5 failed to migrate.

1.3.2 Zebrafish models

In fish, it is also believed that the PE arises from lateral plate mesoderm (Serluca, 2008). Pandora/spt6 mutation in zebrafish abolishes PE formation (Serluca, 2008) as does hand2 ablation (J. Liu & Stainier, 2010). Liu and Stainier investigated whether bmp signalling was also instrumental in zebrafish PE formation, to compare to what had been found in the chick. They derived zebrafish mutants lacking Bmp type I receptor acvr1. This indeed resulted in a failure of PE formation, and a lack of expression in canonical epicardial genes tbx18 and tcf21 (J. Liu & Stainier, 2010).

In the zebrafish, a dual-mechanism PE-epicardium translocation has been argued; Peralta et al used sophisticated optical imaging to show that the fish heartbeat generates pericardial fluid advections, which are required for PE cluster formation, cell release, and transfer to the myocardial surface from dorsal pericardium. In this experiment, blocking heart contractions inhibited PE transfer to the heart (Peralta et al., 2013), as visualised by GFP-labelling of PE cells. Another study lent weight to this, using fluorescent reporters to trace PE and epicardial cells (tcf21-positive cells) alongside a pard3:GFP line, thought to mark the developing epicardium (Hirose, 2006). Reporter expression both in fixed and live embryos in this study indicated that a cellular bridge formed between the pericardium and myocardium (Plavicki et al., 2014). Recently, it was shown that proepicardial

Sophie McManus – February 2020 35

BNC1 regulates human epicardial heterogeneity and function

delamination is facilitated by constriction of dorsal pericardial cells, which allows PE extrusion (Andrés-Delgado et al., 2019).

1.3.3 Mammalian models

In mouse, it is known that Nkx2.5 and Isl1-expressing progenitors contribute to proepicardium (Zhou, Gise, Ma, Rivera-Feliciano, & Pu, 2008); however, most work involving PE origins and signalling pathways has been undertaken in avian and fish models rather than mammals. Gata4-null mice do not form a proepicardium (Watt, Battle, Li, & Duncan, 2004). Moreover, there is some evidence that Wnt signalling plays a part in regulating appropriate epicardial specification, as Dkk1 and Dkk2 double-knockout mice display epicardial hyperplasia and die perinatally (Phillips, Mukhopadhyay, Poscablo, & Westphal, 2011). Events during PE development and signalling in mammals remain incompletely characterised. With regard to the human proepicardium, discovery has naturally been somewhat limited owing to embryonic sample availability and experimental restrictions, although a recent paper by Cui et al used single-cell transcriptomic analysis of human embryos to identify cells with a possible proepicardial signature. Authors thereby found that putative PE cells (termed ‘proEPs’) specifically expressed HEY1 and HEY2, two target genes of the NOTCH signalling pathway, implying that PE fate may be at least in part controlled by NOTCH signalling (Cui, Zheng, Liu, Wen, et al., 2019). This finding complements previous work by del Monte et al, who described an essential role for Notch in murine PE and epicardial development (Del Monte et al., 2011).

In mouse, the picture of PE migration to the heart appears more complicated, as there are multiple proposed mechanisms for PE translocation: to date it seems as though different mechanisms operate in tandem. Initially, it was proposed that mouse PE cells migrate as clusters, or cysts, across the pericardial cavity to enable attachment to the heart tube by E9.0 (Komiyama, Ito, & Shimada, 1987). Par3 knockout mice exhibit lethality due to failure in PE migration to the heart tube (Hirose, 2006), caused by impaired epicardial cyst formation. Latter work suggested that both this mechanism and a direct-contact model both occur in tandem (Rodgers et al., 2008), whereas a more recent paper by Li and colleagues confirmed this multiple-mechanism understanding, alongside reporting some other cell migration along the inflow tract. Authors of this study also went on to

36 Sophie McManus - February 2020

Chapter 1: Introduction

propose the cell division control protein Cdc42 as a key mediator in PE migration (Li et al., 2017).

Once the PE has attached to the heart, the signals and events that allow PE expansion across the heart to form epicardium are not well-characterised, as studying these would likely pose myriad technical challenges. Overall, proepicardial migration to the heart is an example of a species-dependent phenomenon, and much is still incompletely understood regarding regulation of proepicardial cell migration, particularly in mammals. There is overlap in genes expressed in PE and in epicardium (Section 1.4.3) so perhaps there are also species-dependent overlap in cell migratory mechanisms for PE translocation and for epicardial EMT (introduced and expanded in Section 1.5). What is clear is that the epicardial lineage, once established, is utterly critical to correct cardiovascular development (Smits et al., 2018).

1.4 The epicardium The epicardium is the epithelial layer covering the heart, derived from the proepicardial organ. This lineage was first described over a century ago in the dissected chick embryo by the histologist Tadeusz Kurkiewicz in his PhD thesis (‘Zur Kenntniss der Histogenese der Herzmuskels der Wirbeltiere’¸ Jagiellonian University, Poland), before more comprehensive investigations via electron microscopy were made possible a few decades later. Of note, human epicardium is multi-layered in certain regions, whereas in different animal model organisms it is an epithelial monolayer. Human ventricular epicardium has multiple cell layers, while the atrial epicardium is a monolayer (Catherine A Risebro, Vieira, Klotz, Riley, et al., 2015). The ventricular and atrial epicardial cells have also been reported to have different morphologies and spontaneous differentiation propensities.

During embryonic life, the epicardium provides paracrine signals for proliferation, survival, and maturation to cardiomyocytes (Trembley, Velasquez, de Mesy Bentley, Small, et al., 2015) (J. M. Pérez-Pomares et al., 2002) (Masters & Riley, 2014). There is reciprocal signalling from the myocardium, which provides signals inducing proliferation and epithelial-to mesenchymal-transition (EMT) in the epicardium. During EMT, a subset of epicardial cells delaminate from the heart’s surface to yield epicardium-derived cells

Sophie McManus – February 2020 37

BNC1 regulates human epicardial heterogeneity and function

(EPDCs), which invade the myocardium and differentiate to form cardiac derivative cells. A schematic of various embryonic epicardial functions is shown in Figure 3.

Figure 3. The embryonic epicardium is a highly active and dynamic lineage during cardiovascular development. Key functions include epithelial-to-mesenchymal transition (EMT), wherein a subset of epicardial cells (in green and red) delaminate from the heart surface and migrate into the developing myocardium as epicardium-derived cells (EPDCs, in green), differentiating to become coronary smooth muscle cells and cardiac fibroblasts (shown respectively as SMC, CF). There is also cross-talk between the epicardium and the myocardium to facilitate compaction, maturation and proliferation in the latter. Schematic has been adapted from one created by Dr Laure Gambardella.

38 Sophie McManus - February 2020

Chapter 1: Introduction

1.4.1 Epicardial differentiation to cardiac fibroblasts and coronary smooth muscle cells

Epicardial cells in the myocardium differentiate to form cardiac fibroblasts (CF) and coronary smooth muscle cells (coSMC); this has been observed via quail-chick chimeras, dye-labelling, and transgenic mice (Dettman, Denetclaw, Ordahl, & Bristow, 1998) (Acharya et al., 2012) (Q. Liu et al., 2016).

Classical papers by Mikawa et al and Dettman et al in chick traced EPDC migration and differentiation to conclude that epicardial cells form coronary smooth muscle cells and cardiac fibroblasts; endothelial cells were also reported, but this finding has since been challenged, as discussed below (Dettman et al., 1998; Mikawa & Gourdie, 1996). Mikawa’s work entailed use of vital dyes and replication-defective retroviruses encoding beta-galactosidase. Epicardial lineage tracing was achieved by either direct labelling of putative vasculogenic cells in the proepicardium in ovo, or via tagging dissected-out PE cells following by stage-matched transplantation (Mikawa & Gourdie, 1996). This showed PE/epicardial contributions to coronary smooth muscle, connective tissue and coronary endothelium. Dettman used chick-quail chimeras to broadly agree with this finding (Dettman et al., 1998). Mechanistically, Lu et al demonstrated that coSMC differentiation from quail epicardial cells involves serum response factor (SRF)- dependent expression of SMC marker genes such as calponin and SM22α and, when rhoA was inhibited, there was impaired actin filament reorganisation and coSMC formation (J. Lu et al., 2001).

In mice, epicardial contribution to coSMC and CF has also been demonstrated in several mouse lineage tracing studies using different epicardial gene reporters (C.-L. Cai et al., 2008) (Acharya et al., 2012) (Wessels et al., 2012). The latter showed that EPDCs migrate into myocardium by E12.5 and that EPDCs populate specific regions within leaflets of atrioventricular valves.

In zebrafish, Kikuchi et al used tcf21 as an epicardial reporter to examine derivative formation in favour of wt1b or tbx18, as these genes showed cardiomyocyte expression in the fish, while tcf21 did not. They thus observed formation of perivascular cells from

Sophie McManus – February 2020 39

BNC1 regulates human epicardial heterogeneity and function

epicardial progenitors, including smooth muscle cells of the outflow tract (Kikuchi et al, 2011).

The EMT enabling epicardial migration and subsequent differentiation to CF and coSMC shall be described in Section 1.5.

1.4.2 Epicardial differentiation to endothelial cells and cardiomyocytes

There is some controversy regarding embryonic epicardial capacity to form other cell types, for example EC and CM, as mentioned above. In some examples, the experimental strategy may preclude a definitive conclusion. For example, Tbx18-Cre lineage tracing suggested PE/epicardial cells could form CM in mice, as Tbx18-derived cells found in myocardium co-expressed cardiac markers such as troponins and Gata4. However, Tbx18 is commonly known to be extra-epicardially expressed, a pertinent criticism of this study’s conclusion (Christoffels et al., 2009). Another paper that also suggested CM derivation from epicardial cells carries a similar caveat, as authors employed Wt1- CreERT2 and Wt1-GFPCre mice, while Wt1 has been shown to be expressed in cardiac endothelial cells, as well as extra-cardiac tissues (Zhou, Ma, et al., 2008) (Duim, Kurakula, Goumans, & Kruithof, 2015). In any case, the proportion of CM considered to be epicardium-derived in this study was low. Regarding the zebrafish, Kikuchi and colleagues found no epicardial-cardiomyocyte contribution either in development or disease (Kikuchi, Gupta, et al., 2011). Given these points, it seems clear that the epicardium does not typically form CM.

Cardiac endothelial cells (EC) have also been traced to epicardial origins (Dettman et al., 1998; Mikawa T, 1996), however, this is controversial. Uncertainty regarding this conclusion clearly stems in part from experimental methodology; while quail-chick grafts (for example) have proven a powerful technique in the epicardial field, there always exists the possibility of transplanting non-epicardial tissue, which could provide another source of EC. The identification of Scx and Sema3D as PE and epicardial markers enabled a new lineage tracing strategy for Katz and colleagues in 2012, who found that Scx-Cre–marked and Sema3D-Cre–marked cells contributed to coronary endothelial cells to a similar degree as they did to coSMC. However, the proportion of coronary endothelial cells labelled was fairly low, hence a definitive epicardial origin of coronary endothelial cells

40 Sophie McManus - February 2020

Chapter 1: Introduction

remained unclear (Katz et al., 2012). A more recent paper concluded that 20% coronary EC were derived from PE (Cano et al., 2016). Other work has disputed these findings (Acharya et al., 2012). Recently, Lupu et al found no evidence for substantial epicardial contribution to coronary EC. Wt1-positive tdTomato traced cells contributed to just 3.4% of CD31-positive EC. Instead, authors proposed that endothelial precursors may instead be present in the septum transversum mesenchyme, challenging Katz et al’s and Cano et al’s findings (Lupu, Redpath, & Smart, 2019). In summary, while evidence for epicardial formation of coSMC and CF is clear, it appears the epicardium does not contribute extensively to either CM or EC.

1.4.3 Gene expression in the PE and epicardium

The growth factors and signalling cascades that form the PE and epicardium generate gene expression patterns which define and regulate epicardial identity and fate. Across species, many genes are today considered canonical markers of PE; in many instances, expression of these genes is maintained in the epicardium. However, much about the epicardial gene expression network is still ill- or incompletely defined; this is elaborated upon in Section 1.7. Furthermore, it is difficult to distinguish the proepicardium from other cells in same SV region, as gene expression is similar in PE cells and in septum transversum (C.-L. Cai et al., 2008; Lupu et al., 2019). Examples of genes that are highly expressed in PE and epicardium include transcription factors such as Tbx18, Wt1 and Tcf21/epicardin (Robb et al., 1998). Tbx18 is also expressed in some cells of the myocardium (Christoffels et al., 2009). Wt1 is expressed in the podocytes of kidney (Guo, 2002), as well as follicular cells in the ovary (Gao et al., 2014) and cardiac endothelium (Duim et al., 2015), while Tcf21 is expressed in lung and kidney epithelia (Quaggin et al., 1999). The fact that these genes are expressed outside the epicardium has resulted in certain epicardial fate-mapping studies attracting criticism, as mentioned previously (Rudat & Kispert, 2012). However, Wt1 and Tcf21 expression do not tend to overlap or occur in adjacent cells outside the epicardium, so when these genes are both present in the same tissue this strongly implies an epicardial definition. There is evidence for heterogeneous epicardial gene expression in different models; this shall also be addressed in Section 1.7.

Sophie McManus – February 2020 41

BNC1 regulates human epicardial heterogeneity and function

Wt1 KO embryos die due to abnormal epicardial formation, impaired EPDC migration and aberrant development of the coronary vasculature (A. W. Moore, McInnes, Kreidberg, Hastie, & Schedl, 1999) (Martínez-Estrada et al., 2010). Tcf21-null mice also exhibit embryonic lethality, accompanied by increased smooth muscle deposition on the heart surface (rather than in myocardium), as well as clear paucity of interstitial CF formation (C. M. Braitsch, Combs, Quaggin, & Yutzey, 2012; Tandon, Miteva, Kuchenbrod, Cristea, & Conlon, 2013). The roles of Wt1 and Tcf21 in EMT will be discussed further in Section 1.5.3. These genes are both expressed in embryonic epicardium, down-regulated in the adult, and re-expressed following myocardial injury.

Genetic ablation of Gata4 causes a complete lack of PE development (Watt et al., 2004) whereas VCAM-1 nulls also die during development due to failure of epicardial formation (Kwee et al., 1995). Other genes important in epicardial development and function are a host of signalling molecules and growth factors, such as retinoic acid (RA), FGFs, TGFs, PDGFs, and WNTs (Masters & Riley, 2014). Various cell-to-cell adhesion molecules and tight junction proteins such as e-cadherin (CDH1), cytokeratin and zona occludens-1 (ZO-1) are also expressed in the epicardium (von Gise & Pu, 2012).

Scx and Sema3D were reported as markers of distinct compartments of the PE and the epicardium, apparently well-restricted to these lineages (Katz et al., 2012). The selectivity of these genes’ expression has recently been refuted by Lupu et al, who investigated the expression of Wt1, Tcf21, Tbx18, Sema3d and Scx via single-cell RNA in-situ hybridisation (RNAscope) in embryonic mouse epicardium at different time points and reported co-expression of each until E13.5 (Lupu et al., 2019).

Work by Bochmann et al also identified several novel epicardial markers including Bnc1, Gpm6a, and Anxa8 (Bochmann et al., 2010) in the adult mouse. Bnc1 was enriched in adult adult mouse epicardium and significantly downregulated at 3 day post-injury, when the mouse heart undergoes fibrotic remodelling. The particular relevance of this work to my thesis will be described in context in Section 1.8.

1.5 Background to epicardial EMT Epithelial-to-mesenchymal transition is a crucial process in multiple lineages and contexts during embryonic development, as well as adult homeostasis – for example,

42 Sophie McManus - February 2020

Chapter 1: Introduction

embryonic generation of mesoderm (Hay, 1995) and wound healing. EMT also occurs in aberrant contexts during metastasis of various cancers and progression of fibrosis (Brabletz, Kalluri, Nieto, & Weinberg, 2018) (Tomic-Canic et al., 2016).

EMT is a hallmark of epicardium, and much insight into epicardial EMT has come from studies using dye-labelling approaches, quail-chick engraftment, and mouse Cre-Lox genetic lineage tracing studies. In epicardial EMT, a subset of cells delaminate from the sheet covering the heart, and migrate through the sub-epicardial layer and into the myocardium (von Gise & Pu, 2012). This cell delamination marks the onset of various drastic morphology changes, as well as alterations in cytoskeletal organisation, topics which are expanded upon in Section 1.5.4.

1.5.1 Epicardial EMT during cardiac development

The quail-to-chick chimera model has been widely employed and highly informative in visualising epicardial cell movements during EMT. In these experiments, transplantation of quail tissue into chick allows straightforward tracking of quail cells via labelling with an anti-quail antibody. The fate of the transplanted quail epicardium can be inferred via labelling for various markers of cell type. This approach has allowed identification of epicardial cells that have migrated into epicardium and thus differentiated into CF and coSMC (Dettman et al., 1998) (José María Pérez-Pomares et al., 1998) (A C Gittenberger- de Groot, Vrancken Peeters, Mentink, Gourdie, & Poelmann, 1998). Similar studies have been informative in mouse and zebrafish, as mentioned in Section 1.4.1. Importantly however, we lack information pertaining to events during human epicardial EMT.

1.5.2 Signalling during epicardial EMT

Many signalling pathways have been reported to mediate epicardial cell EMT, and this has been well-reviewed, for example by Blom et al, 2013, and von Gise et al, 2012. I will summarise the known major mediators. Overall, many growth factors act in combination to mediate this highly complex physiological phenomenon.

Firstly, members of the TGFβ superfamily are powerful inducers of EMT, both through canonical SMAD signalling and non-SMAD pathways (Xu, Lamouille, & Derynck, 2009) (Lamouille, Xu, & Derynck, 2014). TGFβ dissolves epithelial cells’ basement

Sophie McManus – February 2020 43

BNC1 regulates human epicardial heterogeneity and function

membrane, aids in actin rearrangement and enhancement of cell motility, and enables transition to a mesenchymal cell phenotype (Bax et al., 2011) (von Gise & Pu, 2012) (Dronkers, Moerkamp, van Herwaarden, Goumans, & Smits, 2018). TGFβ1 and TGFβ2 deficient mice die perinatally, exhibiting dramatic cardiac defects, alongside a range of other organ system aberrations; in particular, TGFβ2 deficient mice show impaired EMT (Sanford et al., 1997). Interestingly, after cardiac injury, TGFβ1 and TGFβ2 levels rise (J. Wu, Jackson-Weaver, & Xu, 2018). TGFβ binds to type I and type II receptors, which induces Smad2/Smad3-dependent and -independent (MAPK and Akt) signalling. These events mediate a wide range of transcriptional outcomes. Alk5 is the major TGFβ type I receptor; a lack of Alk5 in mouse and in human epicardial cells hence blocks in vitro EMT. (Sridurongrit, Larsson, Schwartz, Ruiz-Lozano, & Kaartinen, 2008) (Bax et al., 2011). Furthermore, ablation of Alk5 in zebrafish heart diminishes the regenerative response, showing how important TGFβ signalling is for cardiac repair (Chablais & Jazwinska, 2012).

Another factor important in mediating epicardial EMT is PDGF. Smith et al generated mice lacking PDGFRα, PDGFRβ or both PDGF receptor types in the epicardium. When both receptors were knocked out, EMT did not progress, although interestingly this was rescued by Sox9 (which, authors noted, was expressed heterogeneously, in a subset of E13.5 epicardial cells). Of great interest, mutants lacking a single PDGFR subtype exhibited failure of EMT in a lineage-specific fashion; i.e. loss of receptor PDGFRα caused a loss of cardiac fibroblast formation, whereas loss of PDGFRβ led to loss of coronary smooth muscle cells. To further dissect these differences in phenotype, expression of each PDGF receptor was investigated by flow cytometry, which established that by E16.5 expression of each receptor was mutually exclusive. Notably, the EMT phenotypes were observed despite the PDGFR knockout (KO) epicardium maintaining expression of key epicardial markers (Smith, Baek, Sung, & Tallquist, 2011) (Mellgren et al., 2008).

Retinoic acid (RA) exerts effects via RA receptors and retinoid X receptors (RXR). Epicardial RA signalling promotes epicardial EMT; when RA receptor expression is perturbed, aberrant coronary vessel formation is observed. Von Gise et al indicated that epicardial EMT is at least partially regulated by Wt1-dependent RA signalling (von Gise & Pu, 2012).

44 Sophie McManus - February 2020

Chapter 1: Introduction

Lavine and colleagues showed that FGFs are expressed in mouse epicardium, and that Fgf9 expression was induced by RA signalling. Consequences of Fgf signal disruption included premature EPDC differentiation (Kory J. Lavine et al., 2005). Another paper has shown that Fgf10 is essential to EMT with regard to epicardial-to-CF formation and migration: there is reciprocal signalling between the two cardiac cell layers (Vega- Hernández, Kovacs, De Langhe, & Ornitz, 2011). Consistent with this, Lavine et al found that epicardial-myocardial FGF signalling induces Hedgehog (HH) activation, implicated in coronary vascular formation (K. J. Lavine et al., 2006).

WNTs are also involved in epicardial EMT. Zamora et al showed the essential requirement for Wnt effector molecule β-catenin in epicardial EMT via a Gata5 proepicardial knockout model. β-catenin-null mice displayed failed expansion of the subepicardial space, reduced EPDC myocardial invasion, blunted EPDC differentiation to coSMC and impaired coronary plexus modelling (Zamora, Männer, & Ruiz-Lozano, 2007). Wu and colleagues suggested that loss of β-catenin causes randomised spindle orientation when EPDCs enter the myocardium during EMT (M. Wu et al., 2010). .

Another important factor is Tβ4, a G-actin-binding protein known to regulate cell motility and invasion during EMT via cytoskeletal reorganisation (Hong, Lee, Hong, & Hong, 2016; H.-C. Huang et al., 2007). Tβ4 is expressed throughout the myocardium; its effects are therefore exerted via paracrine myocardial-epicardial signalling. Tβ4 knockdown embryos displayed epicardial nodules and impaired EPDC migration through the myocardium; embryos also showed impaired vessel development (Smart et al., 2007). Tβ4 and its epicardial mechanisms of action will be further discussed in Section 1.6.4.

1.5.3 Transcriptional regulation of epicardial EMT

I will briefly summarise some of the transcriptional events thought to mediate EMT. Many studies have outlined roles for the SNAIL superfamily of transcription factors (Barrallo-Gimeno & Nieto, 2005) (Y. Wang, Shi, Chai, Ying, & Zhou, 2013). In different EMT contexts, SNAIL1 binds to the E-cadherin promoter to repress expression, thus promoting the breakdown of cell-cell junctions, enhancing cell motility (Krainock et al., 2016). E-cadherin is known to control transcriptional activity of β-catenin and NF-kappa- B with a downstream block in mesenchymal gene expression; Snail has been shown to

Sophie McManus – February 2020 45

BNC1 regulates human epicardial heterogeneity and function

reverse this (Solanas et al., 2008). In avian epicardium, Snail over-expression induced epicardial EMT in vitro (Tao, Miller, & Lincoln, 2013). However, an in vivo study employing Snail1-null mice implied that a Snail1-independent EMT may occur in mammalian epicardium (Casanova, Travisano, & de la Pompa, 2013). There is also evidence for Twist involvement in EMT, but direct epicardial evidence is lacking (Krainock et al., 2016).

A host of transcription factors control EPDC migration into myocardium during developmental EMT, although our understanding is far from complete. TF shown to be relevant during epicardial EMT include Nuclear factor of activated T-cells cytoplasmic 1 (Nfatc1), which is expressed in a subset of proepicardial cells, epicardial cells and EPDCs; its deletion impairs late embryonic cell migration due to impaired breakdown of myocardial extracellular matrix (Combs, Braitsch, Lange, James, & Yutzey, 2011) (Smits et al., 2018).

Myocardin-related transcription factors (MRTFs) are also involved in mediating epicardial EMT. Trembley and colleagues noted enrichment of MRTFs in the peri-nuclear cytoplasm of epicardial cells, expression of which moved to the nucleus upon TGFβ stimulation. In vivo deletion of either Mrtfa or Mrtfb impacted migration of epicardial cells into myocardium and caused sub-epicardial haemorrhage (Trembley, Velasquez, de Mesy Bentley, & Small, 2015). ETS1 and 2 are further transcription factors shown to mediate migration of EPDCs during EMT, as antisense oligonucleotide silencing in chick embryos impaired cell migration and the formation of coronary vasculature (Lie-Venema et al., 2003, 2007)

Various canonical epicardial TF are also required to effectively regulate epicardial EMT. For example, Wt1 is required for correct mediation of EMT and formation of epicardium- derived cells (EPDCs); this has been shown via mutant mice (A. W. Moore et al., 1999) and in vitro, for example in work by Martinez-Estrada and colleagues using stem cell models. This study showed that Wt1 knockout in vivo and in vitro reduced epicardial cells’ propensity to form mesenchymal cells, and that a possible mechanism in regulating EMT was via binding of Snail1 and Cdh1 (Martínez-Estrada et al., 2010). A similar experimental design indicated a mechanism by which Wt1 and Tbx18 bi-directionally control the epicardial EMT via impacting Snail2 expression in murine primary epicardial

46 Sophie McManus - February 2020

Chapter 1: Introduction

cells. In contrast, to the aforementioned study, in this instance Wt1 reduction induced epicardial EMT, identified via increased cell migration, alongside upregulated expression of N-cadherin and reduced Zo-1, alongside altered Tbx18 (Takeichi, Nimura, Mori, Nakagami, & Kaneda, 2013). This variation in results implies that while WT1 certainly does appear to regulate EMT, the exact mechanism may be model-dependent. In vivo, Tbx18 seems to be redundant for EMT, possible due to compensatory Tbx20 expression (Krainock et al., 2016).

Acharya et al indicated that Tcf21 is important in initiating EMT; primary epicardial cultures exposed to Tcf21 acquired actin stress fibres and displayed EMT gene expression. Tcf21-null mouse hearts had reduced EPDC migration into myocardium (Acharya et al., 2012) and a lack of CF formation, although coSMC appeared to develop normally.

These studies have all implicated various transcription factors key to modulating EMT in animal models. However, the picture is far from complete, and there is a particular dearth of understanding with regards to transcriptional regulation of human epicardial EMT, just as is the case for gene expression networks in human epicardial cells. A deeper appreciation of epicardial transcriptional regulation itself is therefore key to inform understanding of TF involved in EMT. Understanding transcriptional control of EMT is important from the broader context of possible epicardial-mediated cardiac therapy.

1.5.4 Cell polarity and morphology changes in EMT

Epithelial cells are characterised by defined apical-basal cell polarity, which is maintained by cell-cell junctions, including tight junctions, adherens junctions, gap junctions, and desmosomes (Blom & Feng, 2018) (Lamouille et al., 2014). Polarity in epithelia is maintained by PAR3, PAR6, and αPKC (Horikoshi 2009), which are localised in the apical membrane. Regulation of individual cell polarity has been shown to affect epithelial function; for example, Par3 knockouts die due to failed epicardial migration during cardiovascular development, which was linked to failed polarity cue interpretation (Hirose, 2006). Cx43 has also been suggested as an important modulator of epicardial cell polarity, with further implications for correct coronary vascular development (Rhee et al., 2009).

Sophie McManus – February 2020 47

BNC1 regulates human epicardial heterogeneity and function

At the onset of EMT, epicardial cells lose their epithelial morphology and take on a mesenchymal cell fate, involving marked biological changes (Zhou et al., 2012). In epithelia, E-cadherin (CDH1) is key to maintaining adhesion junctions, whereas N- cadherin is a marker of a cell undergoing EMT (CDH2). Cell switching from CDH1 to CDH2 expression is a hallmark of EMT (Blom & Feng, 2018), as upon EMT initiation Cdh1 is cleaved and endocytosed while Cdh2 is upregulated (David & Rajasekaran, 2012). Cells lose their rounded morphology in favour of a spindle shape. There is concomitant loss of expression of adhesion molecules such as ZO-1, and upregulation of transcription factors such as SNAIL1 and SNAIL2 has been reported (Villarejo, Cortés- Cabrera, Molina-Ortíz, Portillo, & Cano, 2014) (Lamouille et al., 2014) (von Gise & Pu, 2012).

Remodelling of the cytoskeleton is pivotal to EMT. Actin is a ubiquitously expressed protein that can bind to many other proteins; its expression is essential to many fundamental cellular processes, including EMT (Izdebska, Zielińska, Grzanka, & Gagat, 2018; Sun, Fang, Li, Chen, & Xiang, 2015). During EMT, actin is repackaged from its globular form (G-actin) in the cortex into its filamentous state (F-actin). Actin filaments can exist in lamellopodia and filopodia, as well as aligned stress fibres (Blanchoin, Boujemaa-Paterski, Sykes, & Plastino, 2014). Lamellopodia consist of dense, organised filaments at the leading edge of a cell, which allow its propulsion forward, whereas ‘finger-like’ filopodia are projections that enable a cell to probe its surrounding environment. Stress fibres are composed of actin and myosin (which allows cell contraction) and connect the cell cytoskeleton to the cortex; the cell cortex shape, and changes therein, are also maintained by actin modelling (Blanchoin et al., 2014). Generally, F-actin remodelling enables coordinated cell migration and invasion (Haynes, Srivastava, Madson, Wittmann, & Barber, 2011a; Izdebska et al., 2018; Rhee et al., 2009).

A schematic summary of epithelial-mesenchymal transition is shown in Figure 4.

48 Sophie McManus - February 2020

Chapter 1: Introduction

Figure 4. Epithelial cells undergo alterations in cell-cell contact and morphology in epithelial-to-mesenchymal transition. (A) Epithelial cells form a continuous sheet via cell-cell gap junctions and tight junctions; apical-basal polarity is maintained by the polarity complex; actin is arranged in cortical bundles. (B) In EMT, cell-cell gap junctions and tight junctions dissolve and actin is reorganised into defined stress fibres, allowing cells to undergo migration and invasion. Figure after Lamouille et al, 2014.

1.5.5 EMT and EPDC specification

Perez-Pomares et al found that avian EPDCs that have delaminated from the heart and invaded the subepicardial space continue to express WT1 and RALDH2, before they populate myocardium and differentiate to coSMCs (whereupon both markers were downregulated) (J. M. Pérez-Pomares et al., 2002). In other work, a Wt1-Cre transgene labelled

Sophie McManus – February 2020 49

BNC1 regulates human epicardial heterogeneity and function

over 90% of cardiac vascular smooth muscle, suggesting that the vast majority of coSMC arise from epicardium via EMT of Wt1-positive cells (Wilm, 2005).

Acharya et al used Tcf21 reporter mice to find that robust Tcf21 expression at E14.5 was linked to the specific acquisition of a cardiac fibroblast identity by E18.5, and that Tcf21 expression persisted in both postnatal and adult heart CF. However, when E10.5 cells were lineage traced, some of these cells could become coSMC. Therefore authors concluded that Tcf21-positive cells are initially multipotent; however, by E14.5, Tcf21 expression indicates epicardial cells are redisposed to form CF rather than SMC (Acharya et al., 2012). Tcf21-null mice in this study formed coSMC but not CF. However other studies have shown that Tcf21-null mice cannot form an epicardium, let alone go on to undergo EMT (Tandon et al., 2013).

Lupu et al performed scRNA-seq on Wt1-tdTomato lineage traced epicardial cells at E15.5, the time point when epicardial cell fates are specified. They found that expression of epicardial markers Wt1 and Sema3d was reduced, whereas Tcf21 increased in EPDCs relative to epicardial cells, though expression did fall in epicardial-derived mural cells (Lupu et al., 2019). Lupu et al found no specificity in epicardial gene expression according to cell fate branching, and so suggest that cell fate may in fact be defined post- EMT, and Tcf21 may not be essential for CF formation. It seems likely that further lineage-tracing studies centring on epicardial EMT and EPDC specification will be informative in defining exactly which epicardial cells can form CF, the degree of epicardial ‘plasticity’ in forming derivative cells, and what additional regulation may be in play.

1.6 Epicardium in heart regeneration: species-specific outcomes In the adult, the epicardium is transcriptionally quiescent. However, upon injury, some organisms such as adult zebrafish and neonatal mouse can fully regenerate the heart, as introduced previously; this is concomitant with robust epicardial reactivation (Poss et al., 2002) (Enzo R Porrello et al., 2011). Over the years, various roles for the epicardium in this context have been suggested, such as enhanced paracrine signalling to enhance cardiomyocyte proliferation and dampening of inflammation (Poss et al., 2002) (Y. Cao & Cao, 2018; Jopling et al., 2010).

50 Sophie McManus - February 2020

Chapter 1: Introduction

1.6.1 Cardiac regeneration in zebrafish

In the zebrafish, cardiac injury is followed by sustained CM proliferation to regenerate the damaged region; the epicardium is indispensable for this repair response (Schnabel, Wu, Kurth, & Weidinger, 2011). When the zebrafish heart is injured, the epicardium reactivates its developmental gene expression profile and rapidly starts to proliferate, becoming a multi-layered lineage and undergoing EMT to yield new coronary vessels (Lepilina et al., 2006). The authors concluded that this pronounced reactivation response is underpinned by interaction between the fgf17b ligand in the myocardium and fgfr2 and fgfr4 in epicardium, and noted that raldh2 and tbx18 were upregulated fewer than 24 hours after injury. Similar work has underscored the importance of raldh2 in this regenerative context (Kikuchi, Holdway, et al., 2011). Cardiac injury thereby stimulates activation of a coordinated epicardial repair programme. In the regenerating zebrafish heart, extracellular component hyaluronic acid has been identified as a key player in reactivation of EMT (Missinato, Tobita, Romano, Carroll, & Tsang, 2015). Proteomics has also revealed the key role of fibronectin secretion by the epicardium in mediating an extracellular environment that promotes impressive regeneration for up to 30 days after injury (Jinhu Wang, Karra, Dickson, & Poss, 2013).

Zebrafish epicardial induction of cardiac regeneration also requires PDGF signalling, as found by Kim and colleagues, who examined explant-derived primary cultured epicardial cells in vitro as well as the regenerating zebrafish heart’s in vivo signalling. An upregulation of pdgfrβ was noted in regenerating hearts; observing explanted epicardial cells in culture revealed PDGF-induced proliferation and that PDGF was needed to alter epicardial cell morphology to a more mesenchymal phenotype, for example via organisation of F-actin stress fibres and loss of cell-cell adhesion (J. Kim et al., 2010). This seems particularly intriguing in light of the findings made by Smith et al regarding PDGF and epicardial EMT, wherein signalling via different PDGF receptors induced varying cell fates (Smith et al., 2011), lending weight to the idea that epicardial heterogeneity may have functional consequences in particular contexts.

Emphasising the vital role of the epicardium during zebrafish cardiac regeneration, the Poss group performed a highly elegant study wherein authors ablated tcf21-positive epicardial cells via targeted use of a bacterial nitroreductase, so that a non-toxic substrate,

Sophie McManus – February 2020 51

BNC1 regulates human epicardial heterogeneity and function

Mtz, is converted to a cytotoxin. Loss of approximately 45% of epicardial cells resulted in a CM proliferative index drop of a third; epicardial-ablated animals also exhibited reduced muscularisation and vascularisation after injury compared to controls. Interestingly however, the ablated epicardium itself regenerated robustly, via vigorous expansion of spared cells: remarkably, this epicardial regeneration was observed even with a loss of 90% of epicardial cells (Jinhu Wang et al., 2015). The dynamics of epicardial regeneration were examined by Cao et al via imaging in vivo and in explants; these revealed that epicardium regenerates via an endoreplication wavefront due to elevated tension at the leading edge (J. Cao et al., 2017). In my view, this paper supports arguments that cardiac regeneration may be a type of concerted, highly efficient wound repair response that has been maintained in certain ‘lower’ animals, and this study underscores just how effective the epicardium can be in the context of heart regeneration. Similar studies in ablating mammalian epicardium would be highly informative, as outlined below.

1.6.2 Cardiac regeneration in neonatal mouse

Porrello and colleagues first characterised remarkable replenishment of cardiomyocytes in the neonatal mouse, finding that removal of up to 15% of the ventricle was adequately repaired by 30 days post-injury (Enzo R Porrello et al., 2011) – as long as the injury was induced before P7. Cardiomyocytes were generated from pre-existing CM in the damaged heart, visualised via Cre-Lox labelling. Other groups have also shown cardiac regeneration in neonatal mouse (Mahmoud et al., 2013; Xin et al., 2013). Nonetheless, there is some disagreement regarding the extent of regeneration, with some groups reporting little to no repair (W. Cai et al., 2019). It seems as though this disparity is due to differing experimental approaches, for example the reporter used in lineage tracing, or the exact amount of tissue resected during injury (Lam & Sadek, 2018).

While the repair response is similar in the zebrafish and the neonatal mouse with regard to the phenomenon of pre-existing CM proliferation, the neonatal mouse heart is a still- developing and immature organ, so it seems unsurprising the myocardial injury repair response is so sustained. In contrast, the zebrafish heart is post-mitotic, so repair in this organism requires coordinated reactivation of embryonic programmes (Aguirre, Sancho- Martinez, & Izpisua Belmonte, 2013). Nevertheless, the fact that the neonatal mouse heart

52 Sophie McManus - February 2020

Chapter 1: Introduction

is amenable to regeneration, unlike the adult, posits this model as a powerful tool in understanding mechanisms for development and repair. Given the indispensable role of epicardium in zebrafish cardiac regeneration, its relevance in the mouse model is of great interest. Interestingly, the regenerative window in the neonatal mouse coincides with the responsive period for cardiomyocytes when they are exposed to epicardial trophic factors (T. H. P. Chen et al., 2002). The same authors also found that epicardial paracrine signalling diminishes in mouse after P4, suggesting that CM proliferation and repair in the neonatal mouse during the regenerative window might be stimulated by epicardial signalling. Although this isn’t direct experimental evidence, it appears highly suggestive.

There is a published model of epicardial ablation (Stevens, von Gise, VanDusen, Zhou, & Pu, 2016) wherein Wt1-positive cells can be ablated via selective expression of diphtheria toxin. This model revealed that epicardial ablation reduces cardiac macrophage numbers in the foetal heart; I believe that it would be of great interest to ablate the epicardium during the neonatal regeneration window and observe how this might affect neonatal mouse heart repair.

1.6.3. Cardiac injury responses in adult mammals

By P7, the murine response to cardiac injury has dramatically altered. While the epicardium is transcriptionally reactivated after injury, re-expressing developmental transcription factors, (Smart et al., 2011) (van Wijk, Gunst, Moorman, & van den Hoff, 2012), it is far less dynamic than in the embryonic setting (C. M. Braitsch et al., 2013). Although there is some EMT, the result of this is an excess of fibroblasts arising during acute cardiac remodelling following MI. While formation of a fibrous scar stabilises the injury site, and to some degree repairs the injured heart, the heart’s overall contractile function is greatly reduced.

Wnt and RA expression rises in the epicardium following injury, which appear to contribute to a post-injury EMT response (Duan et al., 2012) (Bilbija et al., 2012). In a 2013 paper, authors noted that when injury was induced in mouse heart, the subepicardium thickened and showed signs of fibrosis, alongside reactivation of epicardial genes including Wt1 and Tcf21. Areas of heart with interstitial fibrosis in different injury models (pressure-overload and hypertension) displayed Tbx18, Tcf21 and Wt1

Sophie McManus – February 2020 53

BNC1 regulates human epicardial heterogeneity and function

expression. Furthermore, fibrotic human hearts exhibited a similar expression pattern (C. M. Braitsch et al., 2013). Such work suggests that while the epicardium is transcriptionally reactivated after injury, there is a failure to effectively reinitiate embryonic pathways, either due to endogenous epicardial inefficiency or due to a hostile injury environment, or a combination of both.

1.6.4 Harnessing adult epicardium for cardiac repair

Despite this, there is some promising work showing that various epicardial functions are in fact manipulable in the injury setting. Smart et al found that priming the adult mouse heart with Tβ4 enhanced the epicardial transcriptional reactivation that accompanies injury; furthermore, this priming also induced migration of Wt1-positive cells into myocardium, increased proliferative index, and aided in formation of de novo cardiomyocytes from EPDCs. The possibility that the new cardiomyocytes were derived from pre-existing CM or coronary vascular cells rather than de facto EPDCs was excluded via transplant studies and co-staining (Smart et al., 2011). The EPDCs resulting from Tβ4 treatment were found to be a highly heterogeneous population, with associate variation in cardiovascular potential; while some embryonic genes were expressed, these cells were distinct from their embryonic equivalent (Bollini et al., 2014).

Zhou and colleagues found that the adult epicardium produces paracrine factors, including VEGF and FGF2 that have a beneficial conditioning effect on adult injured heart. Impressively, particularly considering the presumable low concentration of these factors, epicardial conditioned medium reduced infarct size and improved cardiac function. When hearts were injured, there was epicardial transcriptional reactivation and thickening, but in this case EPDCs did not migrate into myocardium, instead exerting paracrine effects (Zhou et al., 2011). It seems likely that adult reactivated epicardial EPDCs lack the potency of the embryonic equivalent. Hence understanding factors that might regulate epicardial EMT in vitro may prove informative to ‘boost’ EPDC effects in a therapeutic setting.

Other studies, such as those conducted by Winter et al (E M Winter et al., 2007), have demonstrated that primary adult epicardial cells can improve cardiac function after injury, furthermore that epicardium combined with cardiac cells has a greater benefit than a

54 Sophie McManus - February 2020

Chapter 1: Introduction

single-cell-type approach. A further encouraging 2015 study used ‘enhanced’ epicardial patches to induce some recovery in mammalian adult injured hearts. Epicardial cells were known to express follistatin-like, or FSTN1 during development; however, following MI, expression typically shifted to myocardial cells, concomitant with cardiac fibrosis. Epicardial patches exogenously treated with FSTN1 applied to the damaged heart induced cardiogenic activity via paracrine signalling. The authors concluded that the epicardial- myocardial shift in FSTL1 expression is a maladaptive response to injury, hence restoration of epicardial FSTL1 could potentially reverse deleterious remodelling seen after myocardial infarction in humans (K. Wei et al., 2015).

Most recently, our lab demonstrated that combining stem cell-derived epicardial cells and CM results in enhanced maturation of 3D-EHTs, with resultant improvements in contractility and electrical conduction (Bargehr et al., 2019). These striking improvements were shown to be specifically due to epicardial cells, rather than primary mesenchymal stromal cells or stem cell-derived mesenchymal cells. Furthermore, when cell suspensions containing stem cell-derived epicardial cells in combination with stem cell-derived CM were injected into infarcted rat hearts, there was enhanced microvascular density and cellular proliferation in hearts that were grafted with epicardial- cardiomyocyte suspensions compared to controls. Graft size was also 2.6-fold larger for epicardial-cardiomyocyte injections than for CM-only or epicardial-only, and cardiovascular outcomes were significantly improved. These results illustrate the highly- promising impact of the epicardium in the context of adult mammalian cardiac repair post-injury, in particular for long-term engraftment. Unlike previous studies (E M Winter et al., 2007; Elizabeth M. Winter et al., 2009), this approach employed stem cell-derived epicardium, rather than adult epicardium, perhaps explaining the noticeable enhanced vascularisation and stable cardiac engraftment, as epicardial cells with a more ‘foetal’ phenotype may have greater potency. RNA sequencing for the epicardial secretome revealed fibronectin as a key candidate for the beneficial response observed in epicardial- cardiomyocyte EHT grafts (Bargehr et al., 2019), mirroring observations made by Wang and colleagues in the context of zebrafish cardiac regeneration.

Taken together, this work underscores the promise of the epicardium in prospective strategies to mediate adult cardiac repair in mammals via several possible strategies. As

Sophie McManus – February 2020 55

BNC1 regulates human epicardial heterogeneity and function

the ultimate goal is human therapy, it is important to thoroughly characterise human epicardial function and regulation; however, much remains to be defined in this respect.

1.7 Epicardial heterogeneity As outlined above, the epicardium has clear importance during embryonic development, and, in certain species, during cardiac regeneration. When considering the functional roles of the epicardium, it is important to note that this is a heterogeneous lineage, both in terms of transcription and morphology. Proepicardial and epicardial heterogeneity have been described in the literature in various animal models at different time points; however, both the relevance and regulation of this heterogeneity are incompletely understood, particularly with regard to human physiology.

1.7.1 Epicardial transcriptional heterogeneity

Canonical epicardial genes WT1 and TCF21, discussed in Section 1.4.3, are differentially expressed in the epicardium (C. M. Braitsch et al., 2012) (Vicente-Steijn et al., 2015). This is true in both mouse and zebrafish epicardium. Heterogeneous expression of PDGF receptor subtypes has also been reported (Mellgren et al., 2008). The latter study presented evidence for functional importance of receptor subtype heterogeneity, and represents a rare insight into the biological relevance of epicardial heterogeneity.

Transcriptomic heterogeneity at the single-cell level has been explored within the whole mouse heart (Cui, Zheng, Liu, Wen, et al., 2019; DeLaughter et al., 2016), but the epicardial cell subset was too small for extensive analysis, and there is no equivalent epicardial-focused work in either mouse or human, although there are such studies in the zebrafish. For example, Weinberger and colleagues recently used reporter lines at 5 days post-fertilisation for single-cell RNA sequencing, demonstrating that the zebrafish epicardium is comprised of three sub-populations expressing differing combinations of tbx18, tcf21 and wt1b (Weinberger, Simões, Patient, Sauka-Spengler, & Riley, 2018). Epicardial population 1 (Epi 1) expressed all three markers tbx18, tcf21 and wt1b; population 2 (Epi 2) expressed tbx18 and some smooth muscle cell markers; population 3 (Epi 3) expressed tcf21 only. Authors noted a relative increase in GO terms pertaining to epithelial cell adhesion and migration in Epi 1, vasoconstriction and cell migration during heart development GO-terms in Epi 2, and non-cardiac cell migration terms in Epi

56 Sophie McManus - February 2020

Chapter 1: Introduction

3, implying the zebrafish epicardium may be subdivided into populations that are to some extent functionally distinct. Other zebrafish work has also proposed the presence of three different epicardial cell populations via single-cell transcriptomics (J. Cao et al., 2016).

It is highly important to consider the question of epicardial transcriptional identity; while work across different models has proffered a generally accepted signature for epicardial cells (e.g. expression of canonical epicardial genes such as WT1 and TCF21, characteristic ‘cobblestone’ morphology), the full epicardial transcriptional network identity and regulation thereof is not well-defined. This is a topic that has relevance to harnessing the epicardium for therapeutic benefit. While there are still key regulatory genes ‘missing’ in our understanding of epicardial transcription, the regulatory picture remains incomplete, hence so does our capacity to optimise its function.

1.7.2 Epicardial morphological heterogeneity

Epicardial heterogeneity is not just transcriptional, but also exists with regard to epicardial thickness and cell alignment, at least in human foetal samples. It is possible that human foetal epicardium is composed of distinct cell types, with corresponding disparity in function; in particular, atrial and ventricular epicardium display differing cell morphology, as well as propensity to spontaneously differentiate (Sinha group, unpublished observations) (Catherine A Risebro, Vieira, Klotz, Riley, et al., 2015). This phenomenon is likely relevant to development of future regenerative strategies; for example, ventricular human foetal epicardium has been shown to undergo spontaneous EMT more than atrial epicardial cells do. A thorough comparison of relative characteristics and differentiation capacities of each epicardial region would therefore be useful. Epicardial morphological differences could be a function of the transcriptional heterogeneity described in Section 1.7.1; however, as yet little-to-no work has been performed on possible transcriptional-morphological correlation.

1.7.3 Human stem cell models of epicardial function, hPSC-epi

While model organisms such as mouse and zebrafish are enormously instructive, the goal of biomedical research is typically to extrapolate the situation in animal models to that in humans, to inform current and future clinical strategies. Mice and humans exhibit many similarities in their anatomical organisation, embryonic developmental events,

Sophie McManus – February 2020 57

BNC1 regulates human epicardial heterogeneity and function

physiology and genome; moreover, mice represent a highly manipulable model organism in genetic lineage tracing studies and mutant models. I have alluded to common themes across species during cardiovascular development in Sections 1.2.1 and 1.2.2. Conservation of signalling mechanisms and gene expression allow us to infer a great deal from animal models, but this does not negate our lack of human data. Furthermore, there are still substantial differences between mouse and human gene expression, gestation period and development, so human studies are essential to validate or challenge animal data, and test approaches that will become relevant to eventual patient treatment.

However, studying human epicardial function presents challenges, not least due to contending with a lack of healthy adult samples and an unpredictable supply of staged human embryos. For example, a 2015 paper outlining the WT1 expression pattern in human embryonic epicardium showed clear staining in epicardial cells, as well as invading EPDCs, at different stages, but the time points were confined to weeks 5, 10 and 20 of gestational age (Duim, Smits, Kruithof, & Goumans, 2016). Human foetal epicardial explant cultures are a useful tool (Catherine A Risebro, Vieira, Klotz, & Riley, 2015) (Moerkamp et al., 2016) (Dronkers et al., 2018), but again are derived from embryos at different developmental stages, with differing genetic backgrounds that may influence experimental observations; the number of cells or tissue samples available in this context is also limited. Therefore, a tractable human stem cell model of the epicardium represents a highly versatile tool, offering the opportunity to closely study human developmental events, model diseases in a dish, and generate a large quantity of well-characterised cells for study and translational opportunities.

To this end, our group established a robust in vitro model for deriving epicardial cells from human pluripotent stem cells (hPSC-epi), which can be stimulated in serum- and feeder-free culture to derive coronary-like smooth muscle cells and cardiac fibroblast-like cells (Iyer et al., 2015). Other protocols to generate in vitro epicardial models also exist (Witty et al., 2014) (Bao et al., 2017; Guadix et al., 2017).

Given evidence in the literature that the epicardium is a heterogeneous lineage, combined with highly promising results from employing hPSC-epi in a regenerative context (Bargehr et al., 2019), there is a great deal of interest in establishing any functional relevance of epicardial heterogeneity (Catherine A Risebro, Vieira, Klotz, Riley, et al.,

58 Sophie McManus - February 2020

Chapter 1: Introduction

2015). For example, could we generate large quantities of functionally optimised hPSC- epi for use in the clinic? If epicardial heterogeneity does confer functional implications, could some epicardial populations yield greater benefit in a regenerative context than others? Could we manipulate such populations in the laboratory prior to engraftment, or for use in drug screens? In summary, hPSC-epi models present both the opportunity to study fundamental biological questions and to build tools for therapy.

1.8 Heterogeneity revealed in hPSC-epi by scRNA-seq In order to approach the question of whether functionally relevant human epicardial heterogeneity exists, our hPSC-epi was used for single-cell RNA sequencing (scRNAseq) to investigate human epicardial transcriptional heterogeneity, as well as possible functional relevance of this, via Smart-Seq2. 232 hPSC-epi cells were examined from three different differentiations. This provided evidence for heterogeneity in an in vitro human epicardial model, and has allowed definition of a potential human epicardial transcriptional network, to some extent mirroring data from mouse and chick embryos. Two subpopulations were identified by principal component analysis (PCA) of single- cell RNA sequencing data (Gambardella et al., 2019), shown in Figure 5. The two hPSC- epi subpopulations identified were TCF21highWT1low and TCF21lowWT1high; more information is contained in the Appendix (Figure S2).

Sophie McManus – February 2020 59

BNC1 regulates human epicardial heterogeneity and function

Figure 5. Gambardella et al found two distinct hPSC-epi subpopulations at d9 of hPSC- epi differentiation via scRNA-seq. Principal component analysis shown, coloured for single-cell expression of WT1, BNC1 and TCF21. One of the d9 hPSC-epi subpopulations was high in expression level for epicardial transcription factor WT1; the other population was high in transcription factor TCF21. The expression of each gene is usually mutually exclusive. The WT1-high subpopulation expressed BNC1 (red shows high expression level) to a high degree. BNC1 expression anti-correlated with TCF21 (BNC1 in green, TCF21 in red).

60 Sophie McManus - February 2020

Chapter 1: Introduction

1.8.1 BNC1 identified as candidate gene of interest in hPSC-epi subpopulation

While WT1 and TCF21 are accepted epicardial genes known to have relevance in different animal models, knowledge of the epicardial transcription network is far from complete. Our scRNAseq data identified a potential new facet within the transcription network, the transcription factor Basonuclin 1 (BNC1), as a gene that is differentially expressed to a high level between the two epicardial subpopulations. BNC1 was highly expressed in WT1high hPSC-epi cells (the TCF21lowWT1highBNC1high hPSC-epi subpopulation) but expressed at a low level in TCF21high cells (the TCF21highWT1lowBNC1low hPSC-epi subpopulation), as shown in Figure 5. The strongest loadings of TCF21 and WT1 were on the second component (PC2). Over-representation analyses using the 100 genes with strongest negative and positive PC2 loadings defined two different molecular signatures for the TCF21 and WT1 subpopulations; for the latter, the strongest PC2 loading was for Podocalyxin and the second-strongest was for BNC1. BNC1 expression was better for aligning cells along PC2 than WT1 was.

Importantly, network inference methods applied to our system by Dr Le Novère positioned BNC1 as a putative key regulator with relation to the modelled epicardial regulatory network, Figure 6 (Gambardella et al., 2019). To better understand the implications of BNC1, TCF21 and WT1 in the regulation of the epicardial development and function, a transcriptional regulatory network was generated. Variation in the system was generated by using bulk sequencing transcriptomic data from different stages of cell development, including smooth muscle cells differentiated from hPSC-derived lateral plate mesoderm (Cheung, Bernardo, Trotter, Pedersen, & Sinha, 2012), hPSC-epi, hPSC- epi-CF and hPSC-epi-SMC (Iyer et al., 2015).

The top 100 predicted functional interactions between any TF and each of BNC1, TCF21 and WT1 were retained. BNC1 and TCF21 shared 3 interactors, BNC1 and WT1 shared 17 interactors, and WT1 and TCF21 shared 21 interactors. 11 other transcription factors were shown to interact with the three baits, TCF21, WT1 and BNC1. The top 100 influences involving TCF21 showed a balanced picture with 48 influences originating from TCF21, and 52 targeting TCF21; furthermore, many influences were bidirectional. The conclusion was similar for WT1. However, 68% of influences involving BNC1 originated from this gene, the imbalance being even more striking in the 50 strongest

Sophie McManus – February 2020 61

BNC1 regulates human epicardial heterogeneity and function

interactions, where only 9 influences targeted BNC1. These findings suggested that BNC1 may be a key regulator of epicardial function. A proposed hPSC-epi transcription factor network is outlined below in Figure 6 (Gambardella et al., 2019).

Figure 6. Core epicardial transcriptional network coordinated by BNC1, TCF21 and WT1. The network is built using the 100 strongest inferred influences between any of BNC1, TCF21 and WT1 and other transcription factors. The central nodes interact with all 3 baits, the nodes on the middle circle interact with 2 of our baits while the nodes on the external circle only interact with one bait. Node colours represent the relative expression of the transcription factor in the two populations, turquoise for BNC1high and magenta for TCF21high. The thickness and density of the edges reflect the likelihood of the inferences. Note that since the network is directed, some pairs of nodes are linked by

62 Sophie McManus - February 2020

Chapter 1: Introduction

two edges going in opposite directions, although in most cases only one edge passed the threshold.

1.8.2 Gene Ontology analysis predicts differing hPSC-epi subpopulation functions

Gene Ontology (GO) over-representation analyses suggested a different phenotypic signature for each hPSC-epi subpopulation. This analysis filtered for terms related to cardiac tissues, hence suggested terms related to migration and muscle differentiation for the BNC1high population, compared to terms related to adhesion/angiogenesis for the TCF21high population (Gambardella et al., 2019). The BNC1high population expressed more genes involved in muscle differentiation, migration and cell-cell interaction. In contrast, the TCF21high population was characterised by adhesion with the term ‘Cell substrate-adhesion’ showing high significance and high specificity particular to this population. GO enrichment is shown in the Appendix, in Figure S1.

1.8.3 BNC1: a transcription factor with epithelial relevance

BNC1 is a cell-type specific TF first detected in basal keratinocytes of epithelia (Tseng & Green, 1992); various models have revealed function in stratified epithelia (Iuchi & Green, 1997) (X. Zhang & Tseng, 2007). BNC1 affects RNApol-I and RNApol-II- dependent gene expression and can shuttle between the cytoplasm and the nucleus, depending on its level of serine phosphorylation (Junwen Wang, Zhang, Schultz, & Tseng, 2006).

Bnc1-null mice are known to be born at a reduced rate, and have shown impaired proliferation in a corneal epithelium model of wound repair (X. Zhang & Tseng, 2007) as well as reduced fertility (D. Zhang et al., 2018). BNC1 has also been studied in the context of cancer – it has been indicated that BNC1 expression modulates TGF-β1- induced epithelial dedifferentiation of mammary epithelial cells (Feuerborn, Mathow, Srivastava, Gretz, & Gröne, 2015). It is currently suggested that TGFβ signalling induces an increase in BNC1 mRNA via Smad3, as well as multiple other TGFβ-linked genes, for example c-Jun. Connexin 43 (Cx43) and E-cadherin (CDH1) were also suggested as

Sophie McManus – February 2020 63

BNC1 regulates human epicardial heterogeneity and function

potential downstream targets of BNC1 (Junwen Wang et al., 2006) in silico via DNase fingerprinting, with some validation using a keratinocyte cell line.

Work by the Rosenthal group identified Bnc1 as a novel epicardial marker in adult mouse heart via laser capture microdissection. Transcriptomics in adult mouse heart pre- and post-infarct revealed Bnc1 expression in the epicardium – post-infarct, Bnc1 expression was downregulated (Bochmann et al., 2010). BNC1 had not been investigated in human heart to date. Interestingly, in zebrafish, bnc1 is expressed in adult epicardium, as is the case in the mouse (Weinberger et al, personal communication). However, unlike in mouse, wherein Bnc1 was downregulated after injury, in zebrafish bnc1 levels increased by day 3 post-injury during the regenerative response. Our scRNA-seq data and network inference models raised the question of whether BNC1 is in fact a key epicardial regulator, and if so, what functions it mediates in human epicardial cells. Given this information, BNC1 appeared to be a highly promising candidate of interest in our hPSC-epi model.

1.9 PhD Project Rationale While we know a great deal about epicardial capacity to undergo EMT and differentiate to key cardiac cells during embryogenesis in different animal models, the key mediators of epicardial transcriptional regulation and potential functional implications of epicardial heterogeneity remain largely unexplored. This is particularly true in human epicardium. Characterising and understanding epicardial regulation is key if we are to best-utilise this lineage in a regenerative setting. For example, manipulating heterogeneous epicardium in vitro may provide a more effective cell type for transplant (for example, in EHTs), either by facilitating a magnified EMT response, or via increasing paracrine signalling, hence enhancing vasculogenesis (Bargehr et al., 2019).

In particular, our single-cell RNA sequencing identified marked heterogeneous expression of a transcription factor, BNC1. Network inference models placed this gene on top of an hPSC-epi regulatory network. This gene is typically enriched in mouse epicardium, and downregulated after injury, but had not been investigated in the human heart (Bochmann et al., 2010). In other epithelia, Bnc1 is known to regulate cell proliferation and some aspects of cell migration (Feuerborn et al., 2015; X. Zhang & Tseng, 2007). The intention of this PhD was therefore to establish if the BNC1

64 Sophie McManus - February 2020

Chapter 1: Introduction

transcription factor imparts functional relevance in human epicardium, by addressing three specific project aims outlined below.

1.10 PhD Hypothesis Given our single-cell RNA sequencing data and network inference modelling, I hypothesised that BNC1 may regulate the hPSC-epi transcriptional network, hence affect hPSC-epi heterogeneity, and affect hPSC-epi migration. I therefore planned to address the following aims during my PhD.

1.10.1 PhD Project Aims

1. Investigate BNC1 function in regulation of hPSC-epi differentiation and heterogeneity

2. Characterise BNC1 function in epi-EMT migration models

3. Identify BNC1 targets

Sophie McManus – February 2020 65

BNC1 regulates human epicardial heterogeneity and function

2. MATERIALS AND METHODS

2.1 HPSC culture

Human embryonic stem cells, hPSC, (H9 line, Wicell, Madison, WI) were cultured under chemically defined conditions as described previously (Brons et al., 2007) in an incubator

maintained at 37ºC and 5% CO2. H9 cells were cultured in a chemically defined medium (CDM) containing bovine serum albumin fraction A II (CDM-BSA II) supplemented with Activin A (10 ng/ml, R&D systems) and FGF2 (12 ng/ml, R&D systems) on 6-well tissue culture-treated plates (Corning). Culture plates were pre-treated with porcine gelatine (1mg/ml, Sigma) for 10 minutes at room temperature, before gelatine was removed and replaced with mouse embryonic fibroblast media (MEF media) overnight in the incubator. CDM-BSA media comprised Iscove’s modified Dulbecco’s medium (Gibco) plus Ham’s F12 NUT-MIX (Gibco) medium supplemented with Glutamax-I in a 1:1 ratio, supplemented with BSA (5 mg/ml; Europa Bioproducts), chemically defined lipid concentrate (Life Technologies), transferrin (15 µg/ml, Roche Diagnostics), insulin (7 µg/ml, Roche Diagnostics), penicillin-streptomycin (Sigma, added to a final volume of 1%) and monothioglycerol (450 μM, Sigma).

Cells were passaged once wells were 80% confluent, typically every 5-7 days. For passaging, cells were washed with PBS and treated with Collagenase IV (1mg/ml, Invitrogen) for 5 minutes at room temperature until colony edges began to detach. Colonies were then manually detached from the plate surface by gentle horizontal and vertical scraping with a 2ml pipette. Colonies were collected in a 15ml falcon tube filled with CDM-BSA media; they were left to settle to the bottom of the tube for 5-10 minutes, to allow preferential culture of larger colonies and avoid subsequent culture of single cells. Once colonies were settled, the supernatant media was aspirated, thereby removing smaller colonies and single cells from the culture. The colony pellet was re-suspended in media and triturated with a 5ml pipette 1-3 times. Cells were split at a ratio of 1:6, with average colonies of around 100-150 cells. The culture medium was changed daily.

66 Sophie McManus - February 2020

Chapter 2: Materials and Methods

Media compositions are shown below.

MEF media

Stock concentration or Component Final volume Supplier volume

Advanced DMEM F-12 500 ml 440 ml Gibco

Foetal bovine serum 500 ml 50 ml Gibco

L-Glutamine 100 ml 5 ml Gibco

β-mercaptoethanol 100 ml 3.5 ul Sigma

Penicillin-streptomycin 100 ml 1 ml Sigma

CDM-BSA

Stock concentration or Component Final volume Supplier volume

IMDM 500 ml 250 ml Gibco

Ham’s F12 GlutaMAX-1 500 ml 50 ml Gibco

L-Glutamine 100 ml 5 ml Gibco

β-mercaptoethanol 100 ml 3.5 ul Sigma

Penicillin-streptomycin 100 ml 1 ml Sigma

Sophie McManus – February 2020 67

BNC1 regulates human epicardial heterogeneity and function

CDM-PVA (basal media for FlyB, FB50, WBR)

Stock concentration or Component Final volume Supplier volume

IMDM 500 ml 250 ml Gibco

Ham’s F12 GlutaMAX-1 500 ml 50 ml Gibco

L-Glutamine 100 ml 5 ml Gibco

β-mercaptoethanol 100 ml 3.5 ul Sigma

Penicillin-streptomycin 100 ml 1 ml Sigma

RPMI

Stock concentration or Component Final volume Supplier volume

Thermo RPMI 1640 500 ml 250 ml Fisher

Thermo B27 supplement 10 ml 5 ml Fisher

Penicillin-streptomycin 100 ml 1 ml Sigma

68 Sophie McManus - February 2020

Chapter 2: Materials and Methods

Type IV Collagenase

Stock concentration or Component Final volume Supplier volume

Advanced DMEM/F-12 500 ml 400 ml Gibco

Knockout serum Thermo 500 ml 100 ml replacement Fisher

L-Glutamine 100 ml 5 ml Gibco

β-mercaptoethanol 100 ml 3.5 ul Sigma

1-Thioglycerol 1.25g/ml 20 ul Sigma

Collagenase IV 1 g 0.5 g Life Tech

2.2 HPSC differentiation to hPSC-epi

Differentiation of HPSC was performed in CDM-PVA, which has the same composition as CDM-BSA, apart from substituting polyvinyl alcohol (PVA, 1 mg/ml, Sigma) for BSA II. For early mesoderm differentiation, cells were grown in CDM-PVA with FGF2 (20 ng/ml), LY294002 (10 μM, Sigma) and BMP4 (10 ng/ml, R&D) for 36 h. Cells were then treated with CDM-PVA with FGF2 (20 ng/ml) and BMP4 (50 ng/ml) for 3.5 days to

Sophie McManus – February 2020 69

BNC1 regulates human epicardial heterogeneity and function

generate LM as previously described (Cheung et al., 2012). To generate epicardium after 3.5 days of FB50 media, cells were collected in CDM-PVA and spun for 3 minutes at 1200rpm before resuspension and plating in CDM-PVA with BMP4 (50 ng/ml), recombinant human WNT3A (25 ng/ml, R&D systems) and RA (4 μM, Sigma) at a seeding density of 2.5×103/cm2 for up to 30 days. hPSC-epi cell WBR media was not changed for the first week of differentiation, but wells were topped up with extra WBR in the event of the media changing colour. After the first week of differentiation, WBR was topped up or changed every few days.

2.3 HPSC differentiation to hPSC-epi

Differentiation of cardiomyocytes from hPSC was performed as follows. For the first stage, 24-well Matrigel hPSC-qualified matrix (Fisher Scientific) plates were prepared in the morning, one hour before commencing differentiation; 500 ul of Matrigel was used to coat each well. Confluent H9 HPSC were used to start the cardiomyocyte differentiation. Cells were washed with PBS and detached by incubation with TryplE for 4-5 minutes at 37 Celsius. Cells were collected in CDM-BSA and spun for 3 minutes at 1200rpm before resuspension and plating at either 1.5 or 2x106 cells/ml in CDM-BSA supplemented with FGF2 (12ng/ml), Activin (30 ng/ml) and Y-27632, ROCK inhibitor (MilliPore, 10uM), with each well of the 24-well plate containing 1 ml of cells.

Mesoderm induction was started in the evening, by transferring the cells to FlyAB media: CDM-BSA supplemented with FGF2 (20 ng/ml) Ly (10 uM) Activin (50 ng/ml) and BMP4 (10 ng/ml). Cells were washed with PBS before 1 ml of FlyAB was added to each well. Cells were incubated in FlyAB for 42 hours. After 42 h of FLyAB induction, cells were washed with PBS and incubated with FBRI media1 ml/well: CDM-BSA supplemented with FGF2 (8 ng/ml), BMP4 (10 ng/ml) Retinoic acid (1 µM) and IWR1 (1uM). FBRI media was refreshed after 2 days. On day 4 cells were incubated in FB media: CDM-BSA supplemented with FGF2 (8 ng/ml) and BMP4 (10 ng/ml) for a further 2 days. The final step of cardiomyocyte differentiation required one more wash with PBS before 1 ml of CDM-BSA (no cytokines) was added per well. Beating typically commenced 1-2 days after addition of CDM-BSA, which was changed every other day.

70 Sophie McManus - February 2020

Chapter 2: Materials and Methods

2.4 Primary human epicardial cultures

Human embryonic and foetal tissues were obtained following therapeutic pregnancy interruptions performed at Cambridge University Hospitals NHS Foundation Trust, with ethical approval (East of England Research Ethics Committee, REC) and informed consent in all instances. The REC approval code was 96/085. For human embryonic epicardial explants, one method involved taking 8-week post-conception embryonic hearts were harvested and set up under coverslip on gelatin coated plates (0.1% gelatin for 20 minutes at RT, followed by advanced DMEM F12 + 10% FBS for storage at 37०C) and primary epicardium medium [1:1 mixture of Dulbecco’s modified Eagle’s medium (DMEM, Sigma) and Medium 199 (M199, Sigma) containing 100 U/ml penicillin, 100 μg/ml streptomycin, 10% heat-inactivated foetal bovine serum (FBS, Sigma)]. After a few days, when epicardial cells had started to explant, SB-435142 (Sigma Aldrich), 10µM final concentration, was added to the medium. Cells were maintained in this media until fixation.

Alternatively, the heart was removed from foetuses that were past 10-weeks post- conception (Crown-Rump-Length, CRL, of at least 30mm). Several patches of the epicardial layer were then peeled off with fine dissecting tweezers and set up to grow in a gelatin-coated 12-well tissue culture plate in primary epicardial medium. After 5 days in a humidified incubator at 37°C and 5% CO2, the cells were dissociated with TrypLE Express Enzyme and replated in primary epicardial medium supplemented with SB- 435142, 10μM final. Cells were maintained in the same conditions and passaged 1:2 or 1:3 when confluent.

2.5 BNC1 siRNA-mediated knockdown in human primary foetal epicardial cells siRNA-mediated knockdown was performed by transfection via Optimem (Gibco) and Dharmafect (Dharmacon) and siRNA assays s2011 and s57438 against BNC1 (from Thermo Fisher) as per manufacturers’ directions, with a scrambled siRNA sequence (AllStar) and no siRNA (just transfection mix) as controls. Final siRNA concentration was 40nm. siRNA-containing media was removed the following day after transfection, and cells were rinsed and fed with human foetal medium. Cells were imaged and typically

Sophie McManus – February 2020 71

BNC1 regulates human epicardial heterogeneity and function

fixed at day 5 post-knockdown for qPCR analysis and ICC. The siRNAs used for these experiments are outlined below in Table 1.

Assay Target Supplier Sense (5' to 3') Antisense (5' to 3')

GUACCAGGCUGCAACACCA UGGUGUUGCAGCCUGGUACT s57438 BNC1 Thermo TT T

GGAAGUCCAGUAUGCCUAU AUAGGCAUACUGGACUUCCT s2011 BNC1 Thermo TT G

CCACGGUUGCUCCAAUGUA UACAUUGGAGCAACCGUGGA s13663 SVIL Thermo TT T

GGAGAAUGCAAUCCGCUUA UAAGCGGAUUGCAUUCUCCA s13662 SVIL Thermo TT G

72 Sophie McManus - February 2020

Chapter 2: Materials and Methods

10587 GACAUUUGCCUUGCUUGA HACD1 Thermo CUCAAGCAAGGCAAAUGUCTG 1 GTT

10587 GAAUCUUUAUGGUGUGGC HACD1 Thermo AGCCACACCAUAAAGAUUCTT 2 UTT

Table 1. siRNA sequences for siRNA-mediated knockdown experiments

2.6 Inducible knockdown: design and annealing shRNA oligonucleotides

Oligonucleotides were designed using TRC sequences from Sigma, and are shown in Table 2. Hairpin A was selected as a validating hairpin as it was demonstrated to work previously in downregulating B2M expression. The oligonucleotides were annealed according to the protocol supplied by Bertero and colleagues (Bertero et al., 2016), and then ligated into the cut optimised inducible knockdown sOPTiKD vector (a kind gift from the Vallier laboratory) using T4 ligase for two hours at room temperature. A Snapgene vector map showing the sOPTiKD vector can be found in Figure 7. The ligation mix was transformed into alpha-select competent cells (BioLine) according to manufacturers’ directions. The transformations were plated onto LB agar plates containing ampicillin before colony PCR screening of transformants.

Sophie McManus – February 2020 73

BNC1 regulates human epicardial heterogeneity and function

Figure 7. An example of the siKD vector with a BNC1 shRNA insert under the control of a tetracycline-responsive H1 promoter. The vector has BamHI cutting sites displayed. Successfully targeted cells are subsequently selected via ampicillin selection.

74 Sophie McManus - February 2020

Chapter 2: Materials and Methods

shRNA ID Top oligonucleotide (5' to 3') Bottom oligonucleotide (5' to 3')

GATCCCGGACTGGTCTTTCTATCTCTTCAAGA TCGACAAAAAAGGACTGGTCTTTCTATCTCTC B2M (A) GAGAGATAGAAAGACCAGTCCTTTTTTG TCTTGAAGAGATAGAAAGACCAGTCCGG

BNC1.1 GATCCCGCCACACCATTTCAGGTTGAAACTC TCGACAAAAAACCACACCATTTCAGGTTGAAA (B) GAGTTTCAACCTGAAATGGTGTGGTTTTTTG CTCGAGTTTCAACCTGAAATGGTGTGGCGG

BNC1.2 GATCCCGCCTTGTCTTGAAGATTCTAAACTCG TCGACAAAAAACCTTGTCTTGAAGATTCTAAA (C) AGTTTAGAATCTTCAAGACAAGGTTTTTTG CTCGAGTTTAGAATCTTCAAGACAAGGCGG

BNC1.3 GATCCCGCCTCATAAGCAACATGACTTTCTC TCGACAAAAAACCTCATAAGCAACATGACTTT (D) GAGAAAGTCATGTTGCTTATGAGGTTTTTTG CTCGAGAAAGTCATGTTGCTTATGAGGCGG

BNC1.4 GATCCCGCCTGCACTGATAGGGTCATTGCTC TCGACAAAAAACCTGCACTGATAGGGTCATT (N/A) GAGCAATGACCCTATCAGTGCAGGTTTTTTG GCTCGAGCAATGACCCTATCAGTGCAGGCGG

BNC1.5 GATCCCGGAGAGTAGTGAAGATCATTTCTCG TCGACAAAAAAGGAGAGTAGTGAAGATCATT (E) AGAAATGATCTTCACTACTCTCCTTTTTTG TCTCGAGAAATGATCTTCACTACTCTCCGG

Table 2. shRNA sequences for psOPTIkd vector construction. BglII overhang is in red, terminator sequence/SalI overhang in blue. Hairpin loop sequence is in bold. Stem is underlined.

2.7 Colony PCR of transformants

Transformants grown on LB agar plates were picked in the morning after plating for colony PCR. AAVsingiKD forward (CGAACGCTGACGTCATCAACC) and reverse (GGGCTATGAACTAATGACCCCG) primers were used; thermocycling conditions were as follows: 95°C for five minutes, then 35 cycles of: 95°C for 30 seconds, 63°C for 30s and 72°C for 1 minute. These PCR reactions were run on a 1.5% agarose gel, with positive colonies running at 520bp. Positive colonies were mini-prepped (Qiagen, mini prep kit, used according to manufacturers’ directions) before Sanger sequencing via Source BioScience, using the protocol specified for strong hairpin structures. Miniprepped vectors which showed correct insertion of our shRNA sequence were selected for midiprep (Qiagen) and restriction digest with BamHI enzyme (ThermoFisher) to check vector fragment size on a 1% agarose gel.

Sophie McManus – February 2020 75

BNC1 regulates human epicardial heterogeneity and function

2.8 Vector digestion

Briefly, the inducible knockdown (siKD) vector (kindly supplied by Ludovic Vallier’s laboratory) was digested using restriction enzymes Bgl II and Sal I (ThermoFisher) in FastDigest buffer (ThermoFisher) for 30 minutes at 37°C to allow insertion of different shRNA sequences against BNC1. The digested vector product was purified with the QIAquick PCR purification kit (QIAGEN) and run on a 0.8% agarose gel before extraction using the QIAEX II Gel Extraction Kit (QIAGEN).

2.9 Gene targeting by lipofection

For generation of inducible knockdown hPSC, AAVS1 targeting was performed by lipofection. hPSCs were transfected 24-48 h following cell passaging with 4 μg of DNA and 10 μl per well of Lipofectamine 2000 in Opti-MEM media (Gibco), according to manufacturer’s instructions. Briefly, cells were washed twice in PBS before incubation at room temperature for up to 45 minutes in 1 ml OptiMEM (Gibco). While cells were incubated in OptiMEM, DNA-OptiMEM mixtures were prepared. Mix 1 comprised 4μg DNA (equally divided between the two AAVS1 ZFN plasmids, also received from Ludovic Vallier’s laboratory, and our shRNA targeting vector) in 250μl OptiMEM per well of a six-well plate. Mixture 2 comprised 10μl lipofectamine in 250μl OptiMEM per well. Mixtures 1 and 2 were prepared and mixed gently before incubation at room temperature for five minutes. 250μl of Mixture 2 was then added to 250μl Mixture 1 before incubation at room temperature for 20 minutes. 500μl transfection mix of 1:1 Mixture 1:Mixture 2 was added in a drop-wise spiral manner around the well of hPSC. Cells were incubated in transfection mix at 37°C overnight before washing in CDM-BSA II media the next day approximately 18 hours post-transfection. After 2 days, 1 μg ml-1 of puromycin was added to the CDM-BSA II culture media. Individual hPSC clones were picked and expanded in culture in CDM-BSA II following 7-10 days of puromycin selection.

2.10 Genotyping siKD hPSC clones

Clones from gene targeting were screened by genomic PCR to verify site-specific targeting, determine whether allele targeting was heterozygous or homozygous, and check for off-target integrations of the targeting plasmid. (See Table 3 for PCR primers

76 Sophie McManus - February 2020

Chapter 2: Materials and Methods

and thermocycling conditions and Table S2 in the Appendix for targeting success rates with different vectors).

Locus PCR (‘PCR 1’) for wild-type AAVS1 locus indicates a non-targeted allele (‘PCR 1’). ‘PCR 2’ is a locus PCR/loss-of-allele PCR). ‘PCR 2’ and ‘3’ refer to 5’INT/3’INT PCR respectively, PCR reactions for vector backbone 5’-end and 3’-end genomic integration region, indicative of specific transgene targeting. PCRs ‘4’ and ‘5’ refer to PCRs for vector backbone 5’-end and 3’-end genomic integration region, which were run to indicate non-specific off-target plasmid integration.

All PCRs were performed using 100ng of genomic DNA as template in a 25μl reaction volume using LongAmp Taq DNA Polymerase (NEB) according to manufacturers’ instructions, including 2.5% volume dimethyl sulphoxide (DMSO). DNA was extracted using the genomic DNA extraction kit from Sigma Aldrich, according to manufacturers’ instructions.

Sophie McManus – February 2020 77

BNC1 regulates human epicardial heterogeneity and function

78 Sophie McManus - February 2020

Chapter 2: Materials and Methods

Table 3. PCR primer sequences and cycling conditions for genotyping psOPTIkd clones

2.11 Inducible BNC1 knockdown

One homozygous-targeted clone for each vector transfection was initially selected for subsequent differentiation into hPSC-epi with or without the addition of 1μg/ml tetracycline (Sigma) to culture media with the aim of mediating BNC1 knockdown. hPSC-epi was differentiated from each clone in the presence and absence of tetracycline. Subsequently, further clones were generated with vector (E) (BNC1.5).

2.12 Immunofluorescence

Antibodies used were as follows:

Unconjugated or Alexa Fluor-488 conjugated Rabbit Anti-WT1 [CAN-R9(IHC)-56-2] from Abcam; Rabbit anti-BNC1 from Atlas Antibodies; Rabbit anti-TCF21 from Atlas Antibodies; Mouse Anti-THY1 clone 5E10 (ThermoFisher); Mouse Anti-CNN1 (Sigma); Rabbit Anti-periostin (Abcam), Mouse anti-ZO1 (ZO1-1A12). All other antibody catalogue numbers and dilutions used are outlined in Table 4.

2.13 Cultured cells: immunofluorescence

For immunocytochemistry of transcription factors, cells were fixed with 4% PFA for 15 minutes at room temperature, then permeabilised and blocked with 0.5% Triton- X100 / 5% BSA/PBS for 30 min at room temperature. Unless otherwise stated, primary antibody incubations were performed at 4°C overnight and Alexa Fluor-tagged secondary antibodies (Invitrogen) were applied for 1 hour at room temperature. Nuclei were counterstained with DAPI (10 μg/ml, Sigma). Images were acquired on a Zeiss LSM 700 confocal microscope and analysed with ImageJ software.

2.14 F-Actin labelling and quantification

F-actin filaments were labelled in hPSC-epi cells or epi-EMT cells using conjugated Phalloidin-TRITC, suppled by Sigma (P1951) at a dilution of 1 in 500. Cells were counted in ImageJ (via the ‘CellCounter’ plugin), and assigned a phenotype: either ‘cortical’, ‘aligned’ or ‘indeterminable’ before the relative proportion of ‘cortical’ cells from three

Sophie McManus – February 2020 79

BNC1 regulates human epicardial heterogeneity and function

independent experiments was averaged and calculated in GraphPad Prism. The FibrilTool plugin (Burian et al., 2013) was used to assign cells an anisotropy score, as a proxy for actin filament organisation. 125 cells per condition from each experiment from at least three fields of view were scored with FibrilTool; there were three independent experiments.

2.15 Sequential staining with antibodies raised in the same species

For double staining of TCF21/WT1, TCF21 and WT1 were detected sequentially. Rabbit anti-TCF21 was first applied overnight and detected with a Rhodamin-FAB fragment goat anti-Rabbit IgG antibody (H+L chains, Abcam) for 1 hour at room temperature. The next day, wells were incubated with rabbit anti-WT1 antibody conjugated to Alexa Fluor- 488 for 2 hours at RT. Dilutions are outlined in Table 4.

2.16 Antibody dialysis

Slide-a-Lyzer MINI dialysis units (Thermo Scientific) were used to remove glycerol from primary antibody solution as per manufacturers’ directions. Briefly, the dialysis unit was soaked for 15 minutes in dH2O. Antibody was added to the unit and dialysis was carried out with PBS for 1 hour.

2.17 Primary antibody fluorophore conjugation

BNC1 antibody was labelled with either Alexa Fluor-568 or Alexa Fluor-647 via use of a kit supplied by Life Technologies as per manufacturers’ directions. Briefly, a 1M solution of sodium bicarbonate was added up to 1/10 the final volume of dialysed antibody. This mix was transferred to a vial of reactive dye, then inverted to fully dissolve the dye. The solution was incubated for one hour at room temperature, with inversion every 15 minutes. The antibody-dye solution was subsequently purified by passing through a spin column filled with proprietary purification resin.

80 Sophie McManus - February 2020

Chapter 2: Materials and Methods

Protein Species Supplier (cat. no) Dilution detected

BNC1 Rabbit Atlas (HPA063183) 1/100

WT1 Rabbit Abcam (ab89901) 1/500

MYH11 Rabbit Abcam (ab53219) 1/500

Calponin Mouse Sigma (2687) 1/10,000

TCF21 Rabbit Atlas (HPA013189) 1/250

Phospho- Histone H3 Mouse Santa Cruz (sc-9790S) 1/400

Ki67 Rabbit CST (9129) 1/400

MYH11 Rabbit Abcam (ab125884) 1/200

SMA22alpha Rabbit Abcam (ab14106) 1/200

Endomucin Rat Santa Cruz (sc-65496) 1/200

Periostin Rabbit Abcam (ab14041) 1/200

ZO-1 Mouse Thermo (33-9100) 1/300

Table 4. Antibodies used in immunofluorescence experiments

2.18 Cryostat sections: immunohistochemistry

Human foetal hearts were harvested as described for primary human culture of epicardium. Whole hearts were snap-frozen in liquid nitrogen and stored at -80C before

Sophie McManus – February 2020 81

BNC1 regulates human epicardial heterogeneity and function

cryostat sectioning after embedding in OCT. 10µm sections were collected onto SuperfrostPlus slides and stored at -80C until immunostaining. Immunostaining was performed as described above.

2.19 Flow cytometry

In flow cytometry experiments, hPSC-epi cells in culture were washed with PBS and detached with Tryple-E. Cells were resuspended in flow labelling buffer (PBS with 0.5M EDTA) and spun at 1200rpm for 3 minutes. The cell pellet was then resuspended in 100ul flow labelling buffer containing antibody. Antibody details are shown in Table 5. Incubations were carried out for 30 minutes on ice before the addition of 1.5 ml labelling buffer to the samples and centrifugation for 3 minutes at 1200rpm. Cells were resuspended in 700ul labelling buffer and filtered through 50um Celltrics (Sysmex) filters before analysis at a BD Biosciences FACS Calibur instrument.

Antibody Species Supplier Dilution

Mouse IgG2A APC- conjugated Antibody Mouse IgG2A R&D 1/200 (control for PODXL stain)

Mouse IgG1 kappa Isotype Control, PE- conjugated, Mouse IgG1 Thermo 1/100 eBioscience-50 µg (control for THY1 stain)

Human Podocalyxin Allophycocyanin MAb Mouse IgG2A R&D 1/200 (Clone 222328)

82 Sophie McManus - February 2020

Chapter 2: Materials and Methods

CD90 (Thy-1) Monoclonal Antibody Mouse IgG1 Thermo 1/100 (eBio5E10 (5E10)), PE- conjugated

Table 5. Antibodies used in flow cytometry

2.20 Calcein assay hPSC-epi cells were incubated with calcein-AM dye to assess cell viability. Briefly, a 1mM stock of calcein-AM (BioTechne) was prepared by adding 50ul DMSO to one vial of 50ug calcein-AM. This was diluted 1 in 1000 in CDM-PVA media and hPSC-epi cells were incubated in this mix for 15 minutes in the incubator. Plates were imaged with the EVOS microscope (Thermo Fisher).

2.21 Annexin V staining

Cells were washed with PBS and detached with TrypLE treatment for 1-2 minutes in the incubator. Cells were then collected in a falcon tube and spun at 1200rpm for 3 minutes, then resuspended in 100ul of Annexin-V 1x buffer (Thermo Fisher). 1ul of anti-Annexin- V-488 was added to cells for 15 minutes at RT, and they were protected from light. Subsequently 300ul of PBS was added and cells were kept on ice. 10,000 events were measured by flow on the Cyan flow cytometer (Ann McLaren laboratory). An unstained control was used in each experiment.

2.22 Zombie Red viability assay

100ul of DMSO was added to one vial of lyophilised Zombie Red dye, (BioLegend) as per manufacturers’ directions, and mixed until fully dissolved. Cells were washed with PBS and incubated with PBS containing Zombie Red dye (1:1000 dilution) for 20 minutes in the incubator. Cells were imaged with the EVOS microscope (Thermo Fisher).

2.23 Quantitative real-time polymerase chain reaction (qPCR)

Total RNA was extracted using the RNeasy mini kit (Qiagen) as per manufacturers’ directions, and 260/280 and 260/230 ratios were checked on the Nanodrop (Thermo

Sophie McManus – February 2020 83

BNC1 regulates human epicardial heterogeneity and function

Fisher). cDNA was synthesised from a minimum of 250 ng RNA using the Maxima First Strand cDNA Synthesis kit (Fermentas). Quantitative real-time polymerase chain reaction (qPCR) reaction mixtures were prepared with SYBR green PCR master mix, (Applied Biosystems) and run on the 7500 Fast Real-time PCR system, either by the quantitative relative standard curve protocol against standards prepared from pooled cDNA from each experiment, or by the ΔΔ-CT method. Melt curves were checked for each experimental run. CT values were normalised to housekeeper genes porphobilinogen deaminase (PBGD) or GAPDH. Primers used are listed in Table 6.

Gene Forward sequence Reverse sequence

GAPDH AACAGCCTCAAGATCATCAGC GGATGATGTTCTGGAGAGCC

WT1 CACAGCACAGGGTACGAGAG CAAGAGTCGGGGCTACTCCA

TCF21 TCCTGGCTAACGACAAATACGA TTTCCCGGCCACCATAAAGG

BNC1 GGCCGAGGCTATCAGCTGTACT GCCTGGGTCCCATAGAGCAT

CX43 GGTGACTGGAGCGCCTTAG GCGCACATGAGAGATTGGGA

PAR3 CAGGTGCATCGCTTGGAAC GCTGAGACATTGTTGGTGCC

PAR6 AGCATCGTCGAGGTGAAGAG GTATAGCCAAGTAGCACGTCC

SMURF1 AGATCCGTCTGACAGTGTTATGT CCCATCCACGACAATCTTTGC

ZO1 CTGGTGAAATCCCGGAAAAATGA TTGCTGCCAAACTATCTTGTGA

CDH1 GGCTGGACCGAGAGAGTTTC CGACGTTAGCCTCGTTCTCA

CDH2 AGGGATCAAAGCCTGGAACA TTGGAGCCTGAGACACGATT

SNAIL1 AAGCCTAACTACAGCGAGCT GAGTCCCAGATGAGCATTGG

84 Sophie McManus - February 2020

Chapter 2: Materials and Methods

SNAIL2 AGCATTTCAACGCCTCCAAA TGGTTGTGGTATGACAGGCA

TGFB1 CAGCAGGGATAACACACTGC CATGAGAAGCAGGAAAGGCC

TWIST1 AGTCTTACGAGGAGCTGCAG ATCTTGCTCAGCTTGTCCGA

TWIST2 AGAGCGACGAGATGGACAAT CTAGTGGGAGGCGGACATG

CASP3 AAAATACCAGTGGAGGCCGA GCACAAAGCGACTGGATGAA

TUBAL3 GCGAGGCCGTTACTCTGTG AGAGAGACGTAAACCCTGAACC

SMAD3 GCGTGCGGCTCTACTACATC GCACATTCGGGTCAACTGGTA

PODXL AAGGCCAGGGGTTCACAT AGCCTCGCATCCCTCTAACT

POSTN CCCGTGACTGTCTATAAGCCAA GTGTGTCTCCCTGAAGCAGT

MOESIN ATGCCCAAAACGATCAGTGTG ACTTGGCACGGAACTTAAAGAG

VIM GAAGGCGAGGAGAGCAGGATT CAAGGTCATCGTGATGCTGAG

ZEB CATATTGAGCTGTTGCCGCTG TCTTGCCCTTCCTTTCCTGTGT

Table 6. qRT-PCR primer sequences

2.24 Western Blotting

To assess BNC1 protein level by immunoblotting, the lysate from one confluent well of hPSC-epi cells in a six-well plate was separated by SDS PAGE on an 8% acrylamide gel. 10ul of Precision Plus Protein All Blue standard was loaded. Protein was then transferred overnight onto a polyvinylidine difluoride membrane (Merck Millipore, IPVH00010). Membranes were blocked with 5% fat-free milk in Tris-buffered saline and 0.05% Tween 20 (TBS-T) for 1 hour at room temperature, followed by incubation overnight at 4 degrees Celsius. Membranes were washed with TBS-T three times for 5 minutes at RT before incubation with secondary anti-rabbit or anti-mouse HRP-conjugated antibody (Sigma) for 1 hour at RT. Membranes were washed with TBS-T and protein was detected with ECL Western blotting detection reagents (PierceTM). If necessary, the blot was stripped

Sophie McManus – February 2020 85

BNC1 regulates human epicardial heterogeneity and function

with mild stripping buffer for 1 hour and probed again. Antibodies used in Western blot are outlined in Table 7.

Antibody type Protein Species Supplier Dilution

Atlas Primary BNC1 Rabbit 1/100 (HPA063183)

Primary β tubulin Mouse CST (2148S) 1/1000

Secondary - Anti-rabbit HRP NEB (7074S) 1/10,000

Secondary - Anti-mouse HRP NEB (7076S) 1/10,000

Table 7. Antibodies for Western Blot

2.25 hPSC-epi Proliferation assay

In order to assay the proliferative index of hPSC-epi, cells were grown with or without tetracycline in the media until day 7 of differentiation, then fixed in 4% PFA and stained for mitotic marker Phospho-histone H3. H3-positive nuclei were quantified with an Image J cell counting macro developed in the Sinha lab by Dr Hongorzul Davaapil. Five fields of view per condition were quantified per experiment. Three independent experiments were quantified.

2.26 siRNA-mediated knockdown in hPSC-epi model of epithelial-to-mesenchymal transition

H9 ES cells were differentiated to form hPSC-epi as described above. At day 8 of epicardial culture, cells were split into CDM-PVA media (without pen/strep) containing 2 ng/ml TGFβ or 10ng/ml FGF2, to induce epithelial-to-mesenchymal transition (EMT). siRNA-mediated knockdown was performed by transfection via Optimem (Gibco) and Dharmafect (Dharmacon) and siRNA sequences against BNC1 (Thermo Fisher) as per manufacturers’ directions, with a scrambled siRNA sequence (AllStar) and no siRNA as controls. Final siRNA concentration was 40nm. siRNA sequences used are listed in Table 1. The following morning, cells were rinsed in CDM-PVA with TGFβ and penicillin-

86 Sophie McManus - February 2020

Chapter 2: Materials and Methods

streptomycin. Cells were fed daily and monitored. Cells were pelleted for RNA extraction and subsequent q-RT PCR at least 48 hours post-transfection. Cells were also fixed with 4% PFA for immunocytochemistry.

2.27 EMT model scratch assay

When cells were confluent enough to perform scratch assay (usually two-three days post- transfection) a scratch was made using a P1000 tip. The scratch border was monitored frequently by imaging on the Evos FL Cell Imaging System. Cells were fixed before scratch borders met, usually 2 days after scratch, by 4% PFA for 15 minutes at RT. Fixing was followed by immunocytochemistry and phalloidin staining to visualise actin filaments.

2.28 2D InCell migration assay

At d8 of hPSC-epi differentiation, cells were washed with PBS, detached with TrypLE treatment for 2-3 minutes, and resuspended in CDM-PVA before centrifugation at 1200rpm for 3 minutes. 200,000 cells were then re-plated in EMT media (CDM-PVA with 2ng/ml TGFβ) per well of a 6-well plate. The cells were left to attach for 4-5 hours in the incubator. Following attachment, the plate lid was removed and the plate was sealed with Breathe-Easy sealing membrane (Merck). The plate was then placed in the InCell microscope, where 10 fields of view were imaged per well every 30 minutes over a period of 48 hours in order to allow subsequent cell tracking.

Wells were blinded and cells’ migration was tracked using the ‘Pointing Cell Tracking’ plugin in ImageJ. Data were plotted in Medusa plots and cells’ accumulated distance and velocity were quantified using the Chemotaxis software, supplied by Ibidi.

2.29 3D invasion hPSC-epi cardiomyocyte co-culture assay mStrawberry-positive H9 and GFP-positive H9 cells were received from Laure Gambardella. mStrawberry-positive cardiomyocytes were differentiated from H9 stem cells in Matrigel, as described previously. Once the cardiomyocytes were beating, they were dissociated and re-plated at a density of 100,000 CM per well of a u-angiogenesis slide (Ibidi). Cells were resuspended before plating in a 50:50 mix of RPMI media mixed

Sophie McManus – February 2020 87

BNC1 regulates human epicardial heterogeneity and function

with Matrigel. 10ul of RPMI:Matrigel mix containing cells was then pipetted into each well and allowed to set in the incubator for 2-3 hours. Subsequently, GFP-positive d7 hPSC-epi was detached from 6-well plates and resuspended in WBR (with or without tetracycline) at a concentration of 25,000 cells per 50ul of media. This cell suspension was pipetted on top of the CM/Matrigel matrix. Wells were fixed with 4% PFA at day 1, 2 and 3 post-seeding to allow confocal imaging of relative hPSC-epi invasion into the cardiomyocyte culture. Images were merged and 3D-rendered in ImageJ. To quantify the relative Z-position for the green hPSC-epi cells and the red CM, Imaris quantification was performed 1 day after cell seeding, using the Imaris 3D viewer ‘surfaces’ tool. Surfaces were defined for the green and red cells, before software calculation of surface statistics including average Z position in 3D. The average distance value between the hPSC-epi layer and CM layer in each condition (um) was calculated for three independent experiments and values were plotted in ImageJ.

2.30 ChIP qPCR primer design qPCR primers against putative ChIP targets identified by Dr Vincent Knight-Schrijver’s analysis of ChIPseq data were designed against the FASTA sequence comprising 1kb bases around the identified peak using NCBI PrimerBlast. At least two primer pairs were tested to assess melt curves before use with ChIPseq material.

2.31 Chromatin shearing test

To test the degree of chromatin shearing, a shear test was carried out using the Chromatin Shearing Optimisation Kit (Low SDS) for transcription factors (Diagenode) according to manufacturers’ directions. Shearing for 4, 5, 6, 7, 8 and 9 cycles (30s on, 45s off) was performed. Sheared chromatin was run on a 1.5% agarose gel to compare the relative fragment size yielded by differing shear regimes.

2.32 Chromatin ImmunoPrecipitation (ChIP)

ChIP was performed using either the Low-Cell ChIP kit according to manufacturers’ directions, or the iDeal ChIP-seq kit for Transcription Factors (both supplied by Diagenode). Briefly, 30 million hPSC-epi cells were harvested and fixed for 10 minutes at room temperature with 1% fresh formaldehyde in PBS. Subsequently, either 10 million

88 Sophie McManus - February 2020

Chapter 2: Materials and Methods

cells or 4 million cells were used in ChIP, depending on the type of kit used. Sonication was performed using a BioRuptor according to the following regime: six sonication cycles, with 30 seconds on and 45 seconds off per cycle. 2.5 ug of antibody was used per condition: IgG negative control, CTCF positive control (supplied with the kit), BNC1 (HPA063183) and TCF21 (HPA013189). Some gDNA was reserved as the input. Immunoprecipitation was carried out overnight at 4 degrees Celsius.

2.33 iPure cleanup of gDNA

The Diagenode iPure kit was used to clean up CHIP DNA according to manufacturers’ directions.

2.34 ChIP library preparation and sequencing

Material for ChIP was measured by Qubit assay as per manufacturers’ directions and ChIP-seq libraries were prepared by the Stem Cell Institute under the direction of Maike Paramor. Sequencing was carried out at Cambridge Cancer Research UK. All samples were sequenced on one lane of HiSeq4000 (SE50). Analysis was conducted by Dr Vincent Knight-Schrijver using the R software package. Data were cleaned via heuristic analysis (comparison between peaks from antibodies of interest and Input and IgG controls), before peak-calling algorithms were run using BayesPeak.

2.35 CHIP qPCR

For ChIP qPCR, each IP was diluted 1 in 5. qPCR was carried out using the quantitative relative standard curve method, against 6 standards of diluted gDNA, or via the ddCT method. Samples were run in triplicate and analysed in Excel using the fold enrichment method, comparing each ChIP condition with IgG control. ChIP qPCR primers are shown in Table 8.

Gene Forward sequence Reverse sequence

HACD1 GCATTCCTGCCCGGTTCTAT GGCATTGCCTAATACGGTGC

RCOR1 CTAGGAAGAAAGCCCGAGCG ACCCCAACCAAGCGCTAAT

Sophie McManus – February 2020 89

BNC1 regulates human epicardial heterogeneity and function

MPC1 CACTGTCACCGGCTCTCTAC GAGAGGGTGCGCTTGTCAG

SVIL CCCAAGGAAACCGCGATGAA GCTAGCCCAAGCATCTTCCC

AHR ATAGTGCTGAGAAGCGGGTG TCCAGTCCCTGTACCTGACC

CDK12 CATCGCGAGACTCAGGTGAA GCGAGATCAGGTCCCAAACT

COX11 TGGCCTAAGAAACGGCTCTAC CGAGACGGAGCACGCC

Table 8. ChIP qPCR primers

2.36 Bulk RNA sequencing

RNA for bulk sequencing was extracted from three differentiations of hPSC-epi cells (1Ei hPSC-epi line) using the Qiagen RNeasy kit, as per manufacturers’ directions. RNA libraries were prepared by Dr Maike Paramor (Cambridge Stem Cell Institute) after quality control, via use of the Takara Pico V2 kit (Takara). Libraries were sequenced via single-end reads using the HiSeq500 sequencing system. Bulk RNA sequencing differential expression analysis was conducted by Dr Nicolas Le Novère in the DESeq2 software package. Workflow steps were as follows in Sections 2.37, 2.38 and 2.39. The ‘ReadMe’ file of the work undertaken is contained in Appendix II.

2.37 Data clean-up, alignment, and quality control

Briefly, FASTQ sequencing files were cleaned using FASTQC and fastp, a software package that removes bad reads and trims remaining adapter sequences. The directory containing the reads was aligned to the human genome GRCh38, using HISAT2. Quality control of the RNAseq results (analysing where the reads were located) was performed with the software SeqMonk. All mRNA isoforms were merged, and matched to the Gene names. Based on visual inspection, the libraries were found opposite-strand specific, and counted accordingly. The counts were loaded in R, and counts for probes matched to the same feature (i.e. gene) were aggregated. Counts were corrected for library size. Genes with extremely low expression and genes for which expression did not vary by at least twofold across samples were removed.

90 Sophie McManus - February 2020

Chapter 2: Materials and Methods

2.38 Principal Component Analysis

PCA plots were generated according to the following method: removal of genes with fewer than 10 counts per million in at least one sample, followed by removal of genes that do not vary by at least twofold across all samples, prior to normalisation of gene expression using DESeq2's rlog procedure, and PCA creation using the prcomp R package. Values were subsequently normalised using the DESeq2 rlog procedure.

2.39 Differential gene expression analysis and Gene Ontology enrichment

Differential expression analyses were performed with the DESeq2 package. Significance was set at a p-value adjusted for multiple testing of 0.01. The background for DESeq2- related enrichment was the list of all genes expressed in at least one cell. Gene Over- representation analyses for Gene Ontology enrichment plots were performed using Webgestalt. We retained all the terms with a p-value adjusted for multiple testing with a false discovery rate (FDR) under 0.05.

2.40 Single-molecule RNA fluorescent in-situ hybridisation (smRNA-FISH)

All steps and solutions were kept RNAase free to maximise RNA recovery and signal.

2.41 smRNA-FISH Probe design

Probes for single-molecule RNA FISH were either ShipReady off-the-shelf (GAPDH) or Custom probes (BNC1) supplied by 2B Scientific. A custom probe for BNC1 were designed with the online BioSearch probe designer using sequences from the Ensembl database, variants removed as specified by the probe designer.

2.42 smRNA-FISH slide preparation

H9 or hPSC-epi single-cell suspensions were washed and seeded in CDM-BSA or PVA onto 18x18 mm glass coverslips (VWR). These coverslips were prepared as follows: washing in 2 ml 70% EtOH for 15 mins, then a wash in dH2O and drying before incubation in 0.01% (w/v) Poly-(L)-lysine in dH2O for 15 minutes. Coverslips were dried before use.

Sophie McManus – February 2020 91

BNC1 regulates human epicardial heterogeneity and function

2.41 Autofluorescence treatment for smRNA-FISH sections

To quench section autofluorescence prior to smRNA-FISH hybridisation, human foetal cryosections were incubated with 2% NaBH4 and 200mM VRC in RNAase-free PBS for 15 minutes at RT.

2.42 smRNA-FISH hybridisation

Stellaris probes were resuspended at 12.5 μM in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0). Cell and section hybridisation was performed according to manufacturer’s instructions, using commercial buffers (Wash A, Wash B and hybridisation buffer); available from 2b Scientific. Other home-made wash buffer (recipes supplied by Magda Bienko, Karolinska Institutet) and hybridisation buffer were also used, as these had been successfully trialled previously. Briefly, cells on coverslips were fixed in 2 ml 4% PFA in 1xPBS for 15 minutes at RT, before two washes in 2 ml 1xPBS and immersion in 2 ml 70% RNAase-free ethanol overnight at 4 °C. Next, cells were washed with 2 ml Wash Buffer A before hybridisation overnight in 100 μL of probe solution dissolved in hybridisation buffer inside a humified chamber at 30°C. Different probe concentrations were trialled: 250nM, 500nM and 1000nM were used on different occasions. After hybridisation, samples were then washed in Wash A for 1hr at 30°C and stained with 5 ng/ml 4',6-diamidino-2-phenylindole (DAPI, Sigma) in Wash A for 30 minutes at 30°C. Coverslips were mounted with ProLong Diamond antifade reagent (Thermo Scientific), cured for 24 hours at RT before imaging. Imaging was performed on a Leica SP8 confocal microscope with 63x, 1.4 numerical aperture objective (Institute of Metabolic Science, Cambridge) and resolution improved through pinhole reduction (1- 0.2 au).

2.43 Data Presentation and Statistics

Data were plotted in GraphPad Prism. Data are shown from at least three independent experiments, with error bars representing standard deviation, unless stated otherwise in the figure legend. Student’s t-test was performed for comparison of two groups; ANOVA with post-hoc significance testing was performed for multiple groups.

92 Sophie McManus - February 2020

Chapter 2: Materials and Methods

Sophie McManus – February 2020 93

BNC1 regulates human epicardial heterogeneity and function

3. RESULTS

3.1 hPSC-epi is transcriptionally heterogeneous Although the epicardium has been long-considered a homogenous tissue, evidence for epicardial heterogeneity has been documented in several different animal models of the epicardium, in particular in mouse and zebrafish (C. M. Braitsch et al., 2013) (Weinberger et al., 2018). However, there is little understanding of the precise extent of epicardial heterogeneity, in particular in the human setting rather than in model organisms. Moreover, functional regulation of epicardial heterogeneity and the physiological relevance of this has not been well-characterised. Via single-cell RNA sequencing, we found transcriptional heterogeneity in a human stem cell model of epicardium, hPSC-epi; in particular, we documented the existence of two distinct hPSC-epi subpopulations, TCF21highWT1low and TCF21lowWT1high (Gambardella et al., 2019), as depicted in Figure 5 (Chapter 1). During my PhD I initially aimed to validate this transcriptional heterogeneity, via immunostaining for epicardial genes that defined the two hPSC-epi subpopulations.

3.1.1 Validating hPSC-epi scRNA-seq results by immunocytochemistry

Our hPSC-epi was differentiated as described in Chapter 2 (Iyer et al., 2015); a schematic of the differentiation protocol is shown in Figure 8A. During differentiation, hPSC cultured in discrete colonies are dissociated, before differentiation to early mesoderm and lateral plate mesoderm-like cells that generate a confluent epithelial sheet, exhibiting characteristic epicardial cobblestone morphology (C. A. Risebro et al., 2015) (Zhao et al., 2016). The morphologies of pluripotent hPSC colonies and hPSC-epi are shown by brightfield imaging in Figure 8B. During the hPSC-epi differentiation, we observe upregulation of epicardial transcription factors WT1 and TCF21 by qPCR (Figure 8C), with expression peaking at day 10 of culture. The antibodies against WT1 and TCF21 protein that we found to be effective in human cells are both raised in rabbit; hence, previous studies into relative proportions of single-positive and double-positive hPSC-

94 Sophie McManus - February 2020

Chapter 3: Results

epi cells had been carried out via indirect estimate using a flow cytometry strategy outlined in previous work in the Sinha group (Iyer et al., 2015). A strategy developed by Dr Laure Gambardella enabled double-staining of WT1- and TCF21-positive cells, and is outlined in Chapter 2. Hence we see heterogeneous expression of WT1 and TCF21, in either single-positive or double-positive hPSC-epi cells (Figure 8D). Further immunocytochemistry for WT1 and TCF21 in hPSC-epi is shown in Figure 9.

Figure 8. hPSC-epi differentiation from H9 embryonic stem cells. (A) The schematic indicates how hPSC-epi is differentiated from hPSC via lateral plate mesoderm induction and subsequent treatment with epicardial medium containing Wnt3a, BMP4 and RA. (B) Brightfield imaging shows a typical H9 colony, the morphology of lateral plate mesoderm, and the characteristic ‘cobblestone’ morphology of hSPC-epi. Scale bars are

Sophie McManus – February 2020 95

BNC1 regulates human epicardial heterogeneity and function

1000um. (C) Increasing expression of epicardial genes TCF21 and WT1 during hPSC-epi differentiation, shown by qPCR. Gene expression is normalised to housekeeper gene GAPDH and then normalised to that of d10 hPSC-epi. Error bars show standard deviation from the mean (n=4). (D) WT1 (green) and TCF21 (red) are heterogeneously expressed within hPSC-epi, as shown by immunocytochemistry (n=3). Scale bar is 100um.

Figure 9. Representative immunostaining for WT1 and TCF21 shows heterogeneous expression in wild-type hPSC-epi. hPSC-epi from d7 of culture. WT1 and TCF21 expression are shown in green. Scale bars are 100um. (n=3)

3.1.2 Primary human foetal epicardial cultures used to verify hPSC-epi results

To complement our hPSC-epi model, I established primary human foetal epicardial cultures (hf-epi), as outlined in Chapter 2. The use of this culture system is well- described in the epicardial research field (Catherine A Risebro, Vieira, Klotz, Riley, et al., 2015). Example cultures, both from culturing the whole foetal heart, and from peeling the epicardial layer, are shown in Figure 10A and Figure 10B respectively. Our hPSC-

96 Sophie McManus - February 2020

Chapter 3: Results

epi has been extensively compared to human foetal epicardial cultures, with similarities found both in terms of marker expression and cell morphology (Iyer et al., 2015). In establishing my foetal epicardial cultures, I noted expected cobblestone morphology as reported in the literature; cultures were viable for up to a week, and cultures established from epicardial peeling could be split 2-3 times before losing their starting morphology. We noted that some epicardial cells underwent EMT, showing an elongated morphology and differentiating to fibroblast-like or smooth muscle cell-like cells. Studies have shown that this in vitro epicardial EMT occurs via TGFβ type I receptor Alk5 signalling (Sridurongrit et al., 2008). Indeed, we found that use of the Alk5 inhibitor SB-435142 minimised this spontaneous differentiation, hence the hf-epi cultures maintained their epithelial morphology. In the case of peeling epicardial pieces for culture in Figure 10B, we noted a tendency towards spontaneous EMT, despite use of SB-435142, if the embryo had a crown-rump length (CRL) lower than 30mm; hence we elected to peel epicardium from embryos with a CRL over 30, which corresponds to Carnegie Stage (CS) 20, or an approximate gestational age of 8 weeks (Table S1, Appendix). We did not have access to karyotypes for these samples.

Sophie McManus – February 2020 97

BNC1 regulates human epicardial heterogeneity and function

Figure 10. Human foetal epicardial samples are prepared in two ways: (A) culturing the whole heart and allowing epicardial outgrowth in the presence of TGFβ Alk5 receptor inhibitor SB-435142. The majority of the cells in this brightfield image display cobblestone morphology, although some at the edge of the field of view seem to be undergoing some spontaneous differentiation. (B) The epicardium can also be peeled and individual pieces then cultured in the same conditions as (A). Epicardial cobblestone morphology is apparent in both experimental set-ups. Scale bars 1000um. Whole heart cultures were prepared for n=10 cultures; epicardial peeling was performed for n=12 cultures.

98 Sophie McManus - February 2020

Chapter 3: Results

3.1.3 Validation of hPSC-epi heterogeneity in primary human foetal epicardial cultures

In order to assess whether primary human foetal epicardial cultures were heterogeneous, I performed immunocytochemistry on hPSC-epi and cultured human foetal epicardium. An example of this staining is shown in Figure 11, comparing a representative hf-epi explant sample with hPSC-epi at day 10 of culture. TCF21 is shown to be heterogeneously expressed in each culture. Both hf-epi and hPSC-epi display robust ZO-1 expression, a key epithelial membrane protein that is known to be expressed in epicardium (Figure 11) (Smits et al., 2018) (Rhee et al., 2009). I performed immunohistochemistry on primary human foetal cryosections, which revealed clear heterogeneity in epicardial WT1 expression (Figure 12). Immunocytochemistry performed at differing Carnegie stages showed that heterogeneity was present both at CS 20 and CS23, for WT1 and ZO-1 in the former (Figure 13A), and TCF21 in the latter (Figure 13B). We consider that the heterogeneity reported here in cryosection and explant samples is representative of epicardial cells that have not yet undergone EMT, as ZO-1 is a marker of epithelial rather than mesenchymal cells, and epicardial cells that have undergone EMT typically downregulate epicardial genes such as WT1 and TCF21. Furthermore, all explants shown were cultured with TGFβ Alk5 receptor inhibitor SB-435142, to prevent spontaneous EMT.

Sophie McManus – February 2020 99

BNC1 regulates human epicardial heterogeneity and function

100 Sophie McManus - February 2020

Chapter 3: Results

Figure 11. Human foetal epicardial explant cultures show heterogeneity similar to hPSC- epi. Representative immunocytochemistry for TCF21 (green) and ZO1 (magenta) in human foetal epicardial culture (left) from an embryo at Carnegie Stage 23, compared to similar expression for d10 hPSC-epi (right). TCF21 expression is heterogeneous for each epicardial model. Filled arrowheads indicate TCF21-positive cells, whereas empty arrowheads indicate a lack of TCF21 expression. Scale bars 50um. For hPSC-epi, TCF21 and ZO1 stain was performed in n=1 hPSC-epi culture; for primary human epicardial explants, n=2.

Sophie McManus – February 2020 101

BNC1 regulates human epicardial heterogeneity and function

Figure 12. Human foetal heart exhibits epicardial heterogeneous WT1 expression, as shown by representative immunohistochemistry (WT1 in green, and some WT1-positive cells are indicated by filled arrowheads). Epi, epicardium. Myo, myocardium. Scale bars 50um. CS indicates the embryonic Carnegie Stage. N=3 human foetal hearts.

102 Sophie McManus - February 2020

Chapter 3: Results

Figure 13. Human foetal epicardial explant cultures exhibit heterogeneity at different embryological Carnegie stages. (A) Heterogeneous WT1 (green) at Carnegie stage 20 (n=1). (B) Heterogeneity in TCF21 expression (green) at Carnegie stage 23 (n=2). Scale bars 50um.

Our scRNA-seq data had shown heterogeneity in epicardial WT1 and TCF21 expression at the single-cell RNA expression level. Immunocytochemistry and immunohistochemistry were therefore used to validate heterogeneous expression in both hPSC-epi and hf-epi samples. Having seen that TCF21 and WT1 were heterogeneously expressed at the protein level, I went on to address BNC1 expression, which our scRNA- seq data had identified as a candidate gene of interest in the WT1high hPSC-epi subpopulation.

3.1.4 Heterogeneous BNC1 expression in hPSC-epi and primary human foetal epicardium

Our scRNA-seq data had demonstrated differential BNC1 expression between hPSC-epi subpopulations. Furthermore, network inference analysis had placed BNC1 high up in the transcriptional hierarchy and identified it as a potential key regulator of other epicardial

Sophie McManus – February 2020 103

BNC1 regulates human epicardial heterogeneity and function

genes in the hPSC-epi signalling network (Section 1.8.1, Chapter 1). I therefore verified that BNC1 is expressed during hPSC-epi differentiation by qPCR (Figure 14A) and immunostaining (Figure 14B). By day 3 of epicardial culture, some BNC1 protein was apparent in a few hPSC-epi cells (Figure 14B, indicated by closed arrowheads). BNC1 expression increased during the hPSC-epi differentiation time course. Robust protein expression was observed by d7 of hPSC-epi culture (Figure 14B).

Figure 14. BNC1 mRNA increases during hPSC-epi differentiation, with (A) a notable increase in expression between d3 and d5. mRNA level is normalised to housekeeper GAPDH and plotted relative to d10 hPSC-epi level. Error bars show standard deviation, n=3. (B) Few cells at day 3 of hPSC-epi differentiation express BNC1 (cells with some apparent BNC1 expression are indicated by arrowheads). In contrast, by day 10, most hPSC-epi cells express varying levels of BNC1 protein. BNC1 expression is shown in green. Scale bars are 100um; n=3 for each stage.

BNC1 was previously identified as a marker in healthy adult mouse epicardium (Bochmann et al., 2010), although with no functional data; to date BNC1 had not been investigated in human epicardium. A comparison between immunostaining in hPSC-epi and hf-epi (Figure 15) revealed a comparable heterogeneous expression pattern between hPSC-epi and primary hf-epi samples. RNAscope in human foetal heart cryosections also showed heterogeneous BNC1 expression in epicardium (Dr Laure Gambardella, shown in Figure S2 in Appendix). This is, to our knowledge, the first time that BNC1 expression has been demonstrated in human epicardium at mRNA and protein level.

104 Sophie McManus - February 2020

Chapter 3: Results

In order to enable co-staining for WT1 and BNC1, or TCF21 and BNC1, hence allowing in situ visualisation of the heterogeneous sub-populations, various antibodies were tested; however, the effective antibodies were all raised in rabbit, complicating their combination in immunostaining. Use of an extended blocking step in rabbit serum proved successful for WT1 and TCF21 co-staining in hPSC-epi (Figure 8D), as well as in hf-epi explants; co-detection of BNC1 and WT1 was also successful in hf-epi explant (Gambardella et al., 2019) ((shown in Appendix, Figure S3).

Figure 15. BNC1 is expressed in human foetal epicardial explant samples (top and bottom-right four panels), and its expression appears to be heterogeneous, as in hPSC-epi (bottom and top-left two panels). BNC1 immunostaining is shown in magenta. Some BNC1-positive cells are indicated by filled arrowheads; BNC1-negative cells are indicated by empty arrowheads. Scale bars 100um. n=3 for BNC1 stains in hPSC-epi, and n=3 for BNC1 stains in hf-epi explants.

3.1.5 PODXL and THY1 are membrane markers for BNC1high and TCF21high hPSC-epi subpopulations

Bioinformatic analysis for our scRNA-seq data by Dr Nicolas Le Novère revealed putative membrane markers for each subpopulation. For the WT1 subpopulation, the relevant marker identified was Podocalyxin (PODXL), as this was expressed at a level 54

Sophie McManus – February 2020 105

BNC1 regulates human epicardial heterogeneity and function

times higher in the WT1 population than the TCF21 subpopulation. The suggested marker was THY1 for the TCF21 subpopulation, as this gene was expressed at a level 13 times higher in the TCF21 subpopulation than the BNC1/WT1 subpopulation. A principal component analysis (PCA) plot coloured for relative PODXL and BNC1 expression is shown in Figure 16, while PCA coloured for the levels of TCF21 and THY1 expression is shown in Figure 17. Immunocytochemistry for PODXL and THY1 in hPSC-epi also shows heterogeneous expression of these markers at the protein level in individual cells (Figure 18). Unfortunately, attempts to co-stain BNC1 and PODXL in hPSC-epi were unsuccessful, as membrane permeabilisation destroyed the PODXL signal, whereas an insufficient degree of permeabilisation precluded effective BNC1 staining. In human foetal heart cryosections, we observed co-localisation of WT1 and PODXL protein (Figure S4 in Appendix), but efforts to co-stain for PODXL and BNC1 in human heart were unsuccessful, as BNC1 antibodies tested were incompatible both with frozen and paraffin-embedded tissue sections (not shown).

106 Sophie McManus - February 2020

Chapter 3: Results

Figure 16. Principal component analysis revealed PODXL as a marker for BNC1-high cells. (A) The BNC1-high subpopulation is shown coloured in red, while in (B) PODXL- high cells are also coloured red, showing the respective relative expressions’ overlap. The legend on each panel indicates relative expression level. scRNA-seq data from Gambardella et al.

Sophie McManus – February 2020 107

BNC1 regulates human epicardial heterogeneity and function

Figure 17. Principal component analysis revealed THY1 as a marker for TCF21-high cells. (A) The TCF21 high subpopulation is shown coloured in red, while in (B) THY1-high cells are also coloured in red, showing the respective expressions’ overlap. The legend on each panel indicates relative expression level. scRNA-seq data from Gambardella et al.

108 Sophie McManus - February 2020

Chapter 3: Results

Figure 18. Heterogeneity in THY1 and PODXL protein expression in hPSC-epi, shown via immunocytochemistry and compared to secondary-only control. THY1 expression is shown in magenta, while PODXL expression is shown in green. Scale bars are 100um. n=2 for THY1 stain; n=2 for PODXL stain.

3.1.6 PODXL and THY1-positive fractions alter during hPSC-epi differentiation

I performed flow cytometry for both PODXL and THY1 throughout the hPSC-epi differentiation in order to infer how the WT1high and TCF21high subpopulation ratio may shift throughout differentiation. The PODXL and THY1 antibodies were directly conjugated to APC and PE respectively, thereby allowing double staining and gating for single-positive, double-negative and double-positive cells. I found that the start of the hPSC differentiation was marked by a high proportion of double-positive cells for THY1 and PODXL (61.16 ± 16.45%). However, by day 3 of differentiation there was a marked increase in cells single-positive for PODXL, from 7.75 ± 5.53% at day 1 to 37.16 ± 7.58%. The proportion of cells single-positive for PODXL continued to increase throughout the hPSC-epi differentiation, rising to 65.73 ± 3.98% at day 9 of culture. Conversely, the proportion of single-positive THY1 cells fell during the differentiation time course, from a peak at day 2 of culture (44.65 ± 11.06%) to 12.93 ± 2.34% at day 9. Hence, our flow data implies that subpopulation ratio is dynamic throughout hPSC-epi differentiation. Flow cytometry was also performed on hPSC-epi cultured for one month (n=1); this experiment revealed the same proportions of PODXL/THY1 positive cells as we saw in hPSC-epi at day 9 of culture (not shown). Data are shown from representative flow plots during hPSC-epi differentiation in Figure 19A, while the mean values are in Figure 19B.

Sophie McManus – February 2020 109

BNC1 regulates human epicardial heterogeneity and function

Figure 19. The proportion of PODXL-positive, THY1-negative cells increases during hPSC-epi differentiation. (A) Representative flow cytometry plots at time points between day 1 and day 9 of hPSC-epi differentiation, showing cells gated for PODXL expression (APC-labelled) and THY1 expression (PE-labelled). (B) Graph summarising the flow cytometry data from three independent experiments, error bars show s.d. By d9 of hPSC- epi differentiation, over 60% of cells are single-positive for PODXL expression, and fewer than 20% are single-positive for THY1. n=3.

3.2 Investigating BNC1 function in hPSC-epi Having observed BNC1 heterogeneous expression both in hPSC-epi and in primary human foetal epicardial samples, I aimed to characterise the effect of disrupting BNC1 expression in these epicardial models. Our network inference analysis modelling of the hPSC-epi signalling network had suggested this transcription factor may be a master regulator of other epicardial genes; hence I hypothesised that perturbing its expression would affect epicardial marker expression. Both siRNA-mediated knockdown (outlined

110 Sophie McManus - February 2020

Chapter 3: Results

in Section 3.2.3 and 3.2.5) and an inducible shRNA knockdown cell line-based strategy were used to disrupt BNC1 expression.

3.2.1 An inducible knockdown strategy to investigate BNC1 function

Knockdown by siRNA presents a quick, simple method by which to investigate gene function (Dana et al., 2017), allowing rapid investigation of cell phenotypes via simple genetic manipulation. However, the various caveats of this approach include variability in transfection efficacy between different experiments, potential non-specific knockdown due to high cytoplasmic siRNA concentration, and concerns regarding reagent cytotoxicity. A further disadvantage to siRNA-mediated knockdown is that siRNA concentration dilutes with cell division, preventing long-term knockdown (C. B. Moore, Guthrie, Huang, & Taxman, 2010). In order to increase reproducibility and facilitate stable, reversible knockdown in our hPSC-epi system, allowing further investigation of BNC1 function in epicardium, I aimed to develop shRNA-based vectors to allow inducible, reversible knockdown (siKD) of BNC1 by lipofection of H9 iPSC, after the protocol developed by the Vallier lab (Bertero et al., 2016).

3.2.2 Deriving BNC1 inducible knockdown cell lines

The optimised single-step inducible knockdown (sOPTiKD) vector (Figure 7, Chapter 2) allows the mammalian AAVS1 ‘safe harbour' locus to be targeted with a short hairpin RNA (shRNA) construct (Bertero et al., 2016) (Bertero et al., 2018), hence allowing generation of siKD cell lines via stable transgene expression. Zinc finger nucleases (ZFNs) cause AAVS1 site-specific double-strand breaks to allow homologous recombination-mediated insertion of our sOPTiKD vector containing the shRNA against our gene of interest. (Tiyaboonchai et al., 2014). shRNA sequences consist of two complementary 19–22 bp RNA sequences, linked by a short ‘loop’ structure of 4–11 nucleotides. Following transcription by RNA polymerase, the shRNA sequences are processed by the Drosha enzyme to generate pre-shRNAs, which are then exported to the cytosol and recognised by Dicer enzyme, which processes shRNAs into siRNA duplexes to allow specific knockdown of the gene of interest, via an enzymatic pathway involving the RNA-induced silencing complex (RISC) (C. B. Moore et al., 2010).

Sophie McManus – February 2020 111

BNC1 regulates human epicardial heterogeneity and function

The sOPTiKD-shRNA targeting method is shown in Figure 20C, while the sOPTiKD vector-targeted allele is shown in Figure 20D. This approach facilitates tetracycline- inducible knockdown: in the absence of tetracycline, a modified tet repressor (tetR) binds to the tet operon (TO) and thereby blocks shRNA expression (Gomez-Martinez, Schmitz, & Hergovich, 2013). This tetR has been codon-optimised (OPTtetR) to prevent leaky shRNA expression (Bertero et al., 2016). When tetracycline is added to the cell culture media, this tetR-mediated mechanism is inhibited and shRNA is transcribed by H1 RNA polymerase III for subsequent processing and gene knockdown. I selected five different commercially available BNC1 shRNA sequences (outlined in Chapter 2, Table 2) to test their respective knockdown efficiency in our hPSC-epi system and investigate effects of BNC1 knockdown. sOPTiKD vectors for each of these shRNAs were constructed and miniprepped before sequences were verified by Sanger sequencing (not shown), demonstrating successful cloning of each BNC1 shRNA into the sOPTiKD backbone. Miniprepped vectors with expected sequences were selected for midi-prep and these midi-prepped vectors were verified by BamHI restriction digest (Figure 20A, B), before subsequent hPSC transfection.

112 Sophie McManus - February 2020

Chapter 3: Results

Figure 20. BNC1 inducible line derivation from transfection with sOPTiKD vector. (A) Gel simulation of expected fragment sizes after sOPTiKD-shRNA vector digestions by the BamHI restriction enzyme. (B) sOPTiKD vectors that had correct sequences were midi-prepped and digested with BamHI, showing the correctly sized band fragments for constructed vectors. 1: sOPTiKD_B2M; 2: sOPTiKD_BNC1.1; 3: sOPTiKD_BNC1.2; 4: sOPTiKD_BNC1.3; 5: sOPTiKD_BNC1.4; 6: sOPTiKD_BNC1.5. (C) Lipofection strategy to insert sOPTiKD vector into AAVS1 locus of ES cells. (D) Schematic showing the transgenic allele that is generated via sOPTiKD transfection of hPSC. AAV-prom, AAVS1 locus promoter; 5′-HAR/3′-HAR, 5’ upstream and 3’ downstream homology arms, respectively; SA, splice acceptor; T2A, self-cleaving T2A peptide; Puro, puromycin resistance; H1, H1 human RNA polymerase III promoter; TO, tet operon; pA, polyadenylation signal; CAG, CAG promoter; OPTtetR, optimised TET repressor.

Sophie McManus – February 2020 113

BNC1 regulates human epicardial heterogeneity and function

Following successful vector construction, H9 cells were transfected with the following vectors: sOPTiKD_B2M, sOPTiKD_BNC1.1, sOPTiKD_BNC1.2, sOPTiKD_BNC1.4, and sOPTiKD_BNC1.5. (The sOPTiKD_BNC1.3 vector had a low yield following midiprep and was discarded for the purposes of these experiments.) The B2M hairpin was used as a positive transfection control, as this hairpin’s efficacy has been validated previously (Bertero, 2016). siKD clones that were positively genotyped (Figure 21), showing 5’ and 3’ integration of the sOPTIkd vector, were selected for expansion and differentiation into hPSC-epi. Appendix Table S2 shows the success rate of homozygous, random integration-free targeting hPSC with each of the different sOPTiKD vectors.

Figure 21. Representative gels for the genotyping of siKD clones. (A) Representative gel for homozygous targeted clones; WT H9 give a band at 1692 bp, whereas no band indicates homozygous transgene integration. In this gel, all clones except ‘number 3’ (labelled ‘het’) appear to be homozygous-targeted. (B) 5’ and 3’ vector integration representative gel. 5’ integration is shown by a gel band at 1103bp, whereas 3’ integration

114 Sophie McManus - February 2020

Chapter 3: Results

of insert is represented by a band at 1447bp. In this example, clones 3, 4, 6, 7, 8, 9, 10 and 12 appear to have both 5’ and 3’ integration (circled in yellow). Clone 11 lacks vector integration. (C) Representative blot for off-target vector backbone integration into genomic DNA. In this gel, bands indicate off-target plasmid integration, whereas a lack of band indicates an absence of random vector integration. In this example, clones 4, 6 and 8 appear to be free from random vector integration (indicated by asterisks), whereas 5, 7, 9, 10, 11 and 12 appear to have some random vector integration.

One homozygous-targeted clone for each vector transfection was selected for subsequent differentiation into hPSC-epi with or without the addition of tetracycline to culture media, with the aim of mediating BNC1 knockdown. Tetracycline was applied at the lateral plate mesoderm stage to test each hairpin’s efficacy. hPSC-epi was successfully differentiated from each clone in the presence and absence of tetracycline. hPSC-epi derived from each clone was harvested at d8, by which differentiation time point BNC1 is expressed (Figure 14). qPCR analysis indicated that three of the four BNC1 vectors mediated some degree of BNC1 knockdown; clones from each of these vectors are termed ‘1Ci’, ‘1Di’ and ‘1Ei’ in Figure 22. Of the three clones initially tested, clone Ei had the most pronounced reduction in BNC1 (Figure 22, Figure 23); clones C and D were not much more effective in mediating BNC1 knockdown than siRNA. Furthermore, the viable clone for vector D exhibited off-target vector integration. Hence, after initial characterisation experiments, clone E was selected for experiments, and two further lines using this particular sOPTiKD vector were subsequently generated, to validate results observed with the original ‘1Ei’ cell line.

Sophie McManus – February 2020 115

BNC1 regulates human epicardial heterogeneity and function

Figure 22. Clones ‘C’, ‘D’ and ‘E’ have some reduction of BNC1 mRNA after application of tetracycline. Early mesoderm serves as a negative control for BNC1 expression. hPSC- epi at d10 serves as a positive control for BNC1 expression. The vectors mediating some degree of BNC1 knockdown are outlined in green; these are vectors C, D and E.

116 Sophie McManus - February 2020

Chapter 3: Results

Figure 23. ‘E’ is the most effective vector in mediating BNC1 knockdown. BNC1 normalised to no-tetracycline level for each cell line, termed ‘Ci’, ‘Di’ and ‘Ei’, according to the vector ID. Cell line ‘E’ showed the greatest reduction in BNC1 expression.

3.2.3 Robust BNC1 knockdown mediated in different siKD lines

Lines generated via hPSC-transfection with the sOPTiKD_BNC1.5 vector all heterogeneously expressed BNC1 in the absence of tetracycline, shown by immunocytochemistry (Figure 24). When treated with tetracycline, each line showed at least a 90% reduction in BNC1 mRNA compared to no-tetracycline control cultures (Figure 25). The siKD cell lines were used across different cell assays, to ensure that any phenotype observed was not an artefact. 1Ei under tetracycline had a reduction of BNC1 to 8.22 ± 4.86% (n=3 compared to no-TET control); 1E-8 BNC1 was reduced to 3.21% (n=2) and 1E-17 to 5.93 ± 8.13% (n=3).

Figure 24. Heterogeneous BNC1 expression in different inducible knockdown hPSC-epi lines, shown by immunocytochemistry. D7 hPSC-epi differentiated from each line is shown. Examples of BNC1-positive cells (in green) are indicated by filled arrowheads, and examples of BNC1-negative cells are indicated by empty arrowheads. Scale bars are 100um. n=3 for 1Ei; n=1 for 1E-8, 1E-17.

Sophie McManus – February 2020 117

BNC1 regulates human epicardial heterogeneity and function

Figure 25. Different siKD lines generated via H9 lipofection with siKD vector ‘E’ all exhibit significant BNC1 knockdown by d7 of hPSC-epi differentiation under tetracycline. For Ei and E-17, n=3. *** p-value, p <0.005. for 1E-8, n=2 biological replicates. hPSC-epi has been cultured in the absence or presence of tetracycline since d1 of hPSC-epi differentiation.

Western blot and immunocytochemistry of siKD hPSC-epi also showed a robust loss of BNC1 under tetracycline. When hPSC-epi is cultured under tetracycline since the LPM stage of differentiation, nearly all BNC1 protein appeared to be lost (Figure 26A and 26B), indicating that this knockdown model is extremely effective.

118 Sophie McManus - February 2020

Chapter 3: Results

Sophie McManus – February 2020 119

BNC1 regulates human epicardial heterogeneity and function

Figure 26. BNC1 is reduced at the protein level in siKD hPSC-epi in the presence of tetracycline. (A) Western blot shows that BNC1 level is reduced relative to housekeeper beta-tubulin in d8 hPSC-epi, as is also the case in immunocytochemistry. (B) vs (C) at the same hPSC-epi stage reveal loss of BNC1 (green) in the presence of tetracycline treatment from the LPM stage. n=3.

3.2.4 Loss of BNC1 alters expression of canonical epicardial genes WT1 and TCF21

Network inference analysis had identified BNC1 as a potential signalling node in the hPSC-epi network. Differential gene expression analysis by Dr Le Novère and colleagues revealed that 2494 genes were differentially expressed between the largest clusters: 1454 genes were more highly expressed in the BNC1high population cells, whereas 1040 genes were at a higher level in the TCF21high ones (Gambardella et al., 2019). In addition to a 13.6-fold enrichment in BNC1 expression, the BNC1high cells exhibited 3.6 times more WT1 than TCF21high cells. Genes encoding Podocalyxin, E-Cadherin and P-Cadherin were also strongly enriched.

Given the strong association between WT1 and BNC1 expression in our data, I hypothesised that perturbing BNC1 function may affect epicardial gene expression. I investigated WT1 and TCF21 expression in BNC1 siKD lines at d5, d7 and d8 of differentiation (Figure 27). Upon BNC1 knockdown, qPCR revealed significantly altered expression of WT1 and TCF21 mRNA. WT1 was down-regulated four-fold, while TCF21 was up-regulated six-fold (Figure 27) when hPSC-epi was differentiated without BNC1 expression, with TET application from the lateral plate mesoderm differentiation stage. Of note, the same relationship between BNC1 reduction, WT1 reduction and TCF21 increase was observed in initial characterisation experiments using clone ‘D’, both at the mRNA and protein level (Figure S5 in Appendix).

120 Sophie McManus - February 2020

Chapter 3: Results

Figure 27. Loss of BNC1 during hPSC-epi results in altered expression of canonical epicardial genes WT1 and TCF21. When BNC1 is reduced in hPSC-epi at days 5, 7 and 8 of differentiation, there is a concomitant reduction in WT1: by d8 the reduction in WT1 is 75% relative to no-tetracycline control. Conversely, TCF21 expression is increased to a significant degree at d7 and d8 of hPSC-epi differentiation; by d8 the increase is five- fold. Error bars show s.e.m. n=5. *** p < 0.005; **, p < 0.01.

I also looked at WT1 and TCF21 expression at the protein level by immunocytochemistry, and saw concomitant WT1 reduction and TCF21 increase when BNC1 was reduced (Figure 28).

Sophie McManus – February 2020 121

BNC1 regulates human epicardial heterogeneity and function

Figure 28. WT1 is reduced and TCF21 is increased at the protein level in siKD epi, shown by immunocytochemistry. (A) WT1 (green) is heterogeneously expressed in no- tetracycline d8 hPSC-epi while in (B) WT1 expression is reduced in tetracycline-treated d8 siKD hPSC-epi. (C) TCF21 (green) is heterogeneously expressed in no-tetracycline d8 siKD hPSC-epi whereas (D) TCF21 expression is increased. Scale bars 100um. n=3.

122 Sophie McManus - February 2020

Chapter 3: Results

3.2.5 Perturbing BNC1 expression in primary human foetal epicardium alters TCF21 expression

I also tested the effect of BNC1 knockdown in human foetal samples, in this instance by transfecting primary epicardial cultures with siRNA against BNC1. In line with the results obtained in the hPSC-epi, reduction in BNC1 by over two-thirds (Figure 29A, 29B) induced a significant five-fold increase in TCF21 expression (Figure 29C). An increase in TCF21-positive cells was also seen via immunocytochemistry (Figure 29D). I did not see the same alteration in WT1 expression as I had seen in hPSC-epi; there was great variation in WT1 in each experiment (not shown). This could be due to the experimental methodology used in each case. siKD hPSC-epi had been differentiated in the presence of tetracycline since the LPM stage, hence suppressing BNC1 expression throughout differentiation, whereas hf-epi that did initially express BNC1 was transfected with siRNA to rapidly reduce BNC1 expression at a later stage in epicardial development.

Sophie McManus – February 2020 123

BNC1 regulates human epicardial heterogeneity and function

Figure 29. BNC1 knockdown in primary human foetal epicardial cells yields an increase in TCF21. (A) BNC1 mRNA is reduced by over 60% by siRNA in human foetal epicardial explants, n=3. BNC1 was normalised to level of the housekeeper gene GAPDH before siRNA-treated samples were compared to scrambled siRNA-treated controls. ** p-value, p < 0.01. Error bars are s.e.m. (B) BNC1 (green) is reduced at the protein level in siRNA- treated cells, shown by immunocytochemistry. (C) TCF21 mRNA is increased five-fold in BNC1 siRNA-treated foetal epicardial cells. * p-value, p < 0.05. Error bars are s.e.m. (D) TCF21 protein (green) is also more highly expressed in BNC1 knockdown cultures.

3.2.6 BNC1 reduction alters PODXL/THY1 ratios in hPSC-epi

We had noted that the proportion of PODXL-positive cells by flow cytometry increased continually during the hPSC-epi differentiation to 65.73 ± 3.98%, whereas in hPSC-epi

124 Sophie McManus - February 2020

Chapter 3: Results

at d9 THY1 single-positive cells comprised just 12.93 ± 2.34%. When hPSC-epi was differentiated in the presence of tetracycline, this subpopulation ratio noted by flow was inverted (Figure 30). The single-positive THY1 fraction altered from 11.79 ± 2.99% to 39.43 ± 6.60%, an almost four-fold change, whereas the PODXL single-positive fraction fell by over two-thirds, from 66.60 ± 3.68% to 20.56 ± 5.52% under tetracycline (day 8 hPSC-epi). hPSC-epi differentiated under tetracycline also showed different morphology to controls; cells were elongated and spindle-shaped (Figure 31).

Hence, simply by reducing BNC1 expression in hPSC-epi, we observe that transcriptional heterogeneity is lost; instead of mixed WT1high and TCF21high cells, BNC1 siKD hPSC- epi appears to comprise a homogeneous TCF21high population. Hence perturbing expression of a single epicardial transcription factor appears to ablate cell heterogeneity in a human model of the epicardium.

Sophie McManus – February 2020 125

BNC1 regulates human epicardial heterogeneity and function

Figure 30. BNC1 reduction increases the proportion of THY1-positive cells in hPSC-epi. (A) Immunocytochemistry for membrane marker THY1 (magenta) indicates there is increased THY1 expression in BNC1 siKD hpsc-epi. Scale bars are 100um. (B) Flow cytometry for PODXL and THY1 in no-tetracycline vs tetracycline-treated hPSC-epi reveals an increased proportion of single-positive THY1 cells, decreased proportion of single-positive PODXL cells, and increased double-negative hPSC-epi cells.

126 Sophie McManus - February 2020

Chapter 3: Results

Representative flow cytometry plot shown. (C) Quantification of flow cytometry experiments staining for THY1 and PODXL d9 hPSC-epi cells reveals a significant increase in THY1 single-positive cells and a significant decrease in PODXL single- positive cells. n=3, error bars show standard deviation. *** p-value <0.005.

Figure 31. BNC1 knockdown hPSC-epi displays a fibroblastic morphology. (Ai) At 4x and (Aii) 10x magnification, no-TET hPSC-epi displays cobblestone morphology, while in (Bi) and (Bii) cells under TET display a more elongated morphology. Scale bars in (Ai) and (Bi) 1000um; in (Aii) and (Bii) 400um. n=3.

3.2.7 BNC1 knockdown increases hPSC-epi proliferative index

BNC1 in the literature has been linked to regulation of corneal epithelial cell proliferation (X. Zhang & Tseng, 2007). I hence hypothesised that BNC1 knockdown would reduce

Sophie McManus – February 2020 127

BNC1 regulates human epicardial heterogeneity and function

proliferation in hPSC-epi. This was assessed by phospho-histone H3 (H3) immunostaining in no-tetracycline and tetracycline-treated hPSC-epi. H3 index is a known readout for cell proliferation, and has been indicated as a more accurate cell proliferative index marker than KI67 staining, another known marker of mitosis (J.-Y. Kim et al., 2017). Tetracycline was applied at d3 and I assessed proliferation at day 7 of hPSC culture, as test stains at earlier stages of differentiation revealed that relatively few cells were proliferating prior to day 7. BNC1 reduction appeared to significantly increase the hPSC-epi proliferative index, from 0.83 ± 0.05% to 1.48 ± 0.08% (Figure 32). We considered that this alteration in cell proliferation could be due to increased proliferation in the TCF21/THY1 positive cell compartment. Immunostaining for TCF21 and H3 proved inconclusive, however, and would need to be explored further.

128 Sophie McManus - February 2020

Chapter 3: Results

Figure 32. BNC1 knockdown increases hPSC-epi proliferative index. (A) Immunocytochemistry for proliferative marker phospho-histone H3 (magenta) indicated that more cells were proliferative in the presence of tetracycline (Ai), compared to no- TET controls (Aii) and BNC1 (green) reduction. Left hand side panels show secondary- only control. Scale bars 100um. (B) Quantification in ImageJ by calculating the H3-

Sophie McManus – February 2020 129

BNC1 regulates human epicardial heterogeneity and function

positive nuclei index indicated that hPSC-epi proliferative index was doubled in the absence of BNC1. *** p-value < 0.005. n=3, error bars s.d.

3.2.8 BNC1 knockdown reduces hPSC-epi viability

We had noted that siKD hPSC-epi under tetracycline appeared to exhibit cell death, typically from d7 onwards in culture, as viewed via brightfield imaging. We would not expect tetracycline application itself to be cytotoxic (Gomez-Martinez et al., 2013) (Bertero et al., 2016).To verify this observed death was not simply an effect of tetracycline, I differentiated siKD hPSC-epi containing the B2M vector in the place of BNC1 vector with and without tetracycline; no cell death was observed (Appendix Figure S6).

I therefore went on to investigate cell viability in BNC1 KD hPSC-epi. Use of a Zombie Red assay indicated that the BNC1 KD hPSC-epi had a lower viability index than no- TET controls. Furthermore, the number of cells still present in culture at d8 was significantly reduced in TET vs no-TET controls (Figure 33). We concluded that BNC1 expression is necessary for the survival of a subpopulation of hPSC-epi cells. We can hypothesise that the precursors of the BNC1highWT1high population cannot survive if BNC1 expression is insufficiently up-regulated. The mechanism for this death will need to be explored further, however. Conceivably, the remaining TCF21high compartment may then increase its proliferative index in the event of BNC1high cell death, and this is discussed further in Chapter 4.

130 Sophie McManus - February 2020

Chapter 3: Results

Figure 33. BNC1 knockdown reduces hPSC-epi viability via Zombie Red assay and hPSC-epi cell counts. (A) Increased zombie red dye permeability in BNC1 d8 knockdown hPSC-epi, indicating lesser viability in these cells. Representative images, n=3. Dying cells are visible in the images as white spots and patches. (B) There is a reduced cell number at d8 hPSC-epi differentiation. n=5. Error bars, s.e.m ** p-value < 0.01.

In summary, the first aim of my PhD project was to investigate possible roles for BNC1 in the context of hPSC-epi differentiation, and maintenance of hPSC-epi heterogeneity. Via use of inducible knockdown cell lines I observed that perturbing BNC1 mediates loss of heterogeneous WT1 and TCF21 expression, alters relative hPSC-epi PODXL and THY1 fractions, and alters hPSC-epi proliferative index and viability, suggesting that BNC1 is indeed a key regulator in the context of hPSC-epi differentiation.

Sophie McManus – February 2020 131

BNC1 regulates human epicardial heterogeneity and function

3.3 BNC1 – a potential role in regulation of epicardial EMT?

Epithelial-to-mesenchymal transition is a hallmark function of epicardium, and a lack or a failure of EMT leads to aberrant heart development. As outlined in Chapter 1, while EMT remains a vital function throughout mammalian life in many contexts (such as wound repair), the damaged adult mammalian heart only partially reactivates embryonic EMT pathways (Zhou & Pu, 2011); this partial epicardial reactivation is associated with poor adult cardiac repair, although causation has not been proven. A detailed understanding of epicardial EMT mechanisms is of particular importance if we aim to manipulate these for therapeutic benefit in the post-injury heart.

Bnc1 has been associated with regulation of an EMT response in an in vitro model of breast cancer metastasis (Feuerborn et al., 2015), as well as in corneal wound repair (X. Zhang & Tseng, 2007). Furthermore, Bnc1 expression level has been indirectly linked to the metastatic index of certain invasive epithelial cancers (Pangeni et al., 2015) (Morris et al., 2010). I therefore hypothesised that BNC1 might mediate pathways involved in effecting epicardial EMT, such as changes in epicardial gene expression, actin reorganisation and cell migration. Hence the second specific aim of my PhD project was to investigate BNC1 function in epicardial EMT.

3.3.1 Modelling epicardial EMT in vitro with hPSC-epi

Epithelial-to-mesenchymal transition is a highly conserved process across different epithelia (Y. Wang et al., 2013). In vivo, EMT signalling is mediated by various signalling molecules, for example TGFβ, FGFs, PDGF and RA, as outlined in Chapter 1, Section 1.5.2. TGFβ is known to mediate EMT in chick, mouse and human epicardial cells (Austin, Compton, Love, Brown, & Barnett, 2008) (Bax et al., 2011) (Witty et al., 2014) (Sridurongrit et al., 2008) (Takeichi et al., 2013). Expression of FGFs in epicardium are also fairly well-characterised (Kory J Lavine et al., 2005); moreover, FGF is known to mediate coronary vessel development, via Hedgehog signalling (Tomanek, Hansen, & Christensen, 2008) (K. J. Lavine et al., 2006).

As both these signalling molecules had been described as EMT mediators in the literature, I tested the effect of each on our hPSC-epi. I noted that splitting hPSC-epi at day 8 of differentiation into media containing either 10ng/ml FGF2 or 2ng/ml TGFβ effectively

132 Sophie McManus - February 2020

Chapter 3: Results

induced a change in characteristic epicardial ‘cobblestone’ morphology to yield more elongated cells, as visualised by brightfield (Figure 35B) and via phalloidin staining (Figure 34) (Witty et al., 2014). Actin remodelling is key during EMT, as cells reorganise their cortical bundles to form aligned stress fibres. Hence we considered imaging F-actin to serve as an important indicator that cells were indeed changing their epithelial phenotype (Smith et al., 2011) (Haynes, Srivastava, Madson, Wittmann, & Barber, 2011b). Moreover, I noted that TGFβ treatment appeared to be better-tolerated by the cells, causing a lesser degree of cell death than FGF2 treatment (not shown), and the actin stress fibres’ alignment seemed more pronounced in these cells than those exposed to FGF2 (Figure 34). Given these observations, I chose to use TGFβ treatment for hPSC- epi in a simple in vitro model of EMT (epi-EMT), as it appeared that our hPSC-epi cells responded better to this treatment than they did to FGF, a finding similarly reported previously (Takeichi et al., 2013). Cells were seeded at a density of 300,000 cells per well of a six-well plate; seeding at a density higher than 400,000 cells per well led to a recapitulation of the epicardial phenotype, with respect to expression of epicardial genes such as ZO-1 (not shown).

Figure 34. TGFβ treatment induces rearrangement of actin filaments into stress fibres to a more pronounced degree than FGF2 treatment. Actin filaments visualised via phalloidin staining (in red). The left hand-side panel shows hPSC-epi at d8 of culture, which does not exhibit aligned actin filaments. TGFβ treatment (middle panel) resulted in the presence of aligned actin filaments (indicated by arrowheads), whereas FGF2 (right panel) yielded less striking actin reorganisation. Scale bars are 50um. n=1.

Sophie McManus – February 2020 133

BNC1 regulates human epicardial heterogeneity and function

3.3.2 Characterising the in vitro epi-EMT model

Inducing EMT in various cell models results in down-regulation of epithelial markers such as CDH1 and ZO-1, with concomitant upregulation of genes associated with EMT processes (e.g. invasion), such as CDH2 (Shankar & Nabi, 2015) (Smith et al., 2011) (Derycke & Bracke, 2004). During epicardial EMT, there is a reduction in expression of canonical epicardial transcription factors such as WT1 and TCF21, as cells lose their epithelial fate and take on a mesenchymal phenotype (C. Braitsch & Yutzey, 2013).

I characterised our in vitro model of epi-EMT via qPCR and immunostaining. A schematic for the epi-EMT model is shown in Figure 35A, and cell morphology in each media type is shown in Figure 35B. In our epi-EMT model, I saw that epicardial genes WT1 and TCF21 were downregulated upon EMT induction (Figure 35C), while an expected ‘cadherin-switch’ (Scarpa et al., 2015; Wheelock, Shintani, Maeda, Fukumoto, & Johnson, 2008) was also clearly observed (Figure 35D): CDH1 mRNA expression fell by over 95% relative to its hPSC-epi level, whereas CDH2 expression increased three- fold upon TGFβ treatment. In contrast to other reports (Feuerborn et al., 2015), we found that TGFβ induction reduced the level of BNC1 mRNA (Figure 35C). However, immunocytochemistry revealed that TGFβ treatment did not entirely ablate BNC1 expression at the protein level (Figure 35E). hPSC-epi TGFβ treatment increased expression of SMC marker calponin (CNN) as well as expression of the fibroblast marker periostin (POSTN) (Figure 35E). Hence, we were confident that this model encompassed certain key features of a classical EMT.

134 Sophie McManus - February 2020

Chapter 3: Results

Figure 35. An in vitro model of hPSC-epi EMT. (A) Schematic for epi-EMT. hPSC-epi is split into media containing TGFβ at day 8 of hPSC-epi culture. This induces (B) a change in cell arrangement and morphology, from a continuous hPSC-epi epithelial sheet to more elongated single cells. (C) Epicardial gene mRNA is downregulated upon hPSC- epi treatment with TGFβ, n=3. (D) Upon treatment with TGFβ there is reduction in CDH1 and an increase in CDH2 expression, normalised to GAPDH and compared to hPSC-epi level, n=3. Error bars show standard deviation, s.d. (E) TGFβ treatment causes upregulation of EMT markers including calponin (in red) and periostin (in green), n=2. BNC1 protein is expressed in fewer cells upon TGFβ treatment. Scale bars 50um.

Sophie McManus – February 2020 135

BNC1 regulates human epicardial heterogeneity and function

3.3.3 PODXL and THY1 expression in the epi-EMT model

I also briefly investigated the effect of TGFβ treatment on hPSC-epi on PODXL and THY1 expression. I had previously observed that the proportion of single-positive PODXL cells increased steadily during hPSC-epi differentiation. Interestingly, TGFβ treatment dramatically altered this; after 24 hours of TGFβ treatment, there was an increase in the proportion of cells double-positive for PODXL and THY1 expression, from 10.89 ± 2.63% in hPSC-epi to 31. ± 5.68% in epi-EMT. The proportion of cells single-positive for PODXL expression fell from 66.66 ± 3.68% in hPSC-epi to 30.2 ± 13.8% in epi-EMT, while the proportion of cells single-positive for THY1 increased from 11.79 ± 2.99% to 19.03 ± 15.49% in epi-EMT (Figure 36). This change was concomitant with the previously noted fall in BNC1 and TCF21 expression at the mRNA level upon epi-EMT induction.

Figure 36. Treating hPSC-epi with TGFβ alters relative hPSC-epi subpopulation ratio. (A) Representative flow cytometry plot showing that hPSC-epi is mainly single-positive for PODXL, whereas TGFβ treatment increases the proportion of double-positive PODXL/THY1 cells. (B) Flow cytometry for PODXL/THY1 in TGFβ-treated samples at d1 and d2 of epi-EMT. n=3 experiments, error bars show s.d. (C) While the percentage

136 Sophie McManus - February 2020

Chapter 3: Results

of single-positive PODXL and single-positive THY1 cells drops when hPSC-epi is treated with TGFβ in epi-EMT, BNC1 and TCF21 expression also falls, to a more marked degree. n=3, error bars show s.d.

This raises interesting questions about the possible relationship between respective PODXL and WT1/BNC1 expression and THY1 and TCF21 expression, which was outlined in Section 3.1.5. It seems conceivable that downregulation of epicardial transcription factors would be swiftly followed by a decrease in expression of respective membrane markers. Alternatively, perhaps expression of epicardial transcription factors and these membrane markers becomes uncoupled during EMT, and cells must instead become double-positive for PODXL and THY1 to enable acquisition of a mesenchymal, migratory phenotype.

3.3.4 siRNA-mediated BNC1 knockdown reveals a cortical actin localisation phenotype

To investigate BNC1 function in our epi-EMT system, hPSC-epi cells were split into the epi-EMT model with BNC1 siRNA at day 8 of hPSC-differentiation. BNC1 knockdown via siRNA was confirmed by qPCR (Figure 37A); two siRNAs were used in different experiments, and BNC1 level was reduced to 30.8 ± 8.04% compared to scrambled control. A hallmark of EMT is rearrangement and polymerisation of the actin cytoskeleton, to enable cell motility (Smith et al., 2011; Sun et al., 2015). F-actin filaments in the epi- EMT model were therefore imaged via phalloidin staining. Scrambled siRNA controls exhibited aligned actin stress fibres within the cytoskeleton, as we would expect in cells undergoing a phenotypic switch from a cohesive epithelial layer to individual, migratory, mesenchymal cells. In stark contrast, cells in BNC1 knock-down wells in the epi-EMT model showed pronounced localisation of actin to the membrane (Figure 37C vs 37B) rather than a clear alignment of actin stress fibres. Some cells in the scrambled condition also exhibited this cortical actin localisation, so cell phenotypes for each condition were quantified in ImageJ. The percentage of cells with a cortical actin phenotype was almost

Sophie McManus – February 2020 137

BNC1 regulates human epicardial heterogeneity and function

four-fold higher when BNC1 had been knocked down (36.99 ± 2.15%), when compared to scrambled control (14.00 ± 5.52%) (Figure 37D).

In addition to cortical actin localisation, we can also consider actin filament arrangement. Use of the ImageJ plugin FibrilTool allows a quantitative measure of fibril array anisotropy, a mathematical index for how well-ordered fibres are. Using this anisotropic index, a value of 0 would define a cell containing fibres that have absolutely no order (i.e. the fibres are isotropic) whereas a value of 1 would define a cell with perfectly parallel (anisotropic) fibres (Burian et al., 2013). This approach revealed that SCR siRNA-treated control cells had an average anisotropy index of 0.43 ± 0.008, whereas BNC1 siRNA- mediated knockdown cells’ average index was significantly lower at 0.27 ± 0.008 (Figure 37E). Therefore BNC1 knockdown via siRNA resulted in significantly ‘less organised’ actin filaments in our epi-EMT model.

Figure 37. BNC1 knockdown in an in vitro model of EMT mediates a cellular cortical actin phenotype, as opposed to aligned actin filaments in scrambled controls. (A) BNC1 siRNA treatment mediates over 60% reduction in BNC1 as compared to scrambled control (SCR), p < 0.05. (B) Aligned actin filaments (phalloidin-TRITC, red) observed in scrambled siRNA-treated epi-EMT. (C) Cortical actin localisation in epi-EMT treated with BNC1 siRNA. (B) and (C) error bars are 100um. (D) Manual counting of cells with

138 Sophie McManus - February 2020

Chapter 3: Results

cortical actin localisation in ImageJ shows that the BNC1-knockdown cells are four-fold more likely to exhibit this phenotype than controls. p < 0.01, error bars show s.d. (E) Anisotropic index, analysed using the FibrilTool plugin, is significantly lower in siRNA- treated and BNC1 siRNA-treated cells compared to scrambled controls. p < 0.001, error bars show s.d. n=3.

These initial BNC1 knockdown experiments were performed using siRNA in epi-EMT; the phenotype was noted when BNC1 knockdown was induced at the point when cells were split into TGFβ to induce EMT. I subsequently used our BNC1 siKD cell lines, to validate our siRNA phenotype, as well as to examine the effect of perturbing BNC1 expression at different time points during hPSC-epi differentiation. When siKD hPSC- epi was differentiated in the presence of tetracycline from the LPM stage, I did not observe the striking cortical actin phenotype noted via acute BNC1 siRNA knockdown (Figure 38). Cells with BNC1 knockdown throughout hPSC-epi differentiation had a phenotype similar to that seen in no-TET controls, despite the reduction in BNC1 protein. We therefore hypothesised that BNC1 knockdown should be acutely mediated in order to yield the cortical actin phenotype, as we had previously observed with siRNA applied as cells were split into TGFβ.

Sophie McManus – February 2020 139

BNC1 regulates human epicardial heterogeneity and function

Figure 38. siKD epi-EMT exhibits less pronounced cortical actin localisation phenotype than acute epi-EMT BNC1 knockdown. In the right-hand side panels, BNC1 (green) has been knocked down since hPSC-epi induction; the left-hand side panels show no TET control cells. Actin filaments are shown in red. Scale bars 100um. n=2.

It appears conceivable that BNC1 expression is required for hPSC-epi cells to respond to EMT induction with regards to cellular actin reorganisation. However, if the hPSC-epi has developed without BNC1 since the LPM stage, there could potentially be some degree of compensatory response shown by the TCF21-high subpopulation.

140 Sophie McManus - February 2020

Chapter 3: Results

3.3.4 Cortical actin localisation in siKD epi-EMT

When tetracycline was applied acutely to cells (at d7 of hPSC-epi culture), before they were split into the epi-EMT model, we observed a similar cortical actin localisation phenotype to that observed in Figure 37 (see Figure 39D), validating the result seen with siRNA. Conversely, when BNC1 was knocked down via TET from d3 of hPSC-epi culture, the cortical actin phenotype was less pronounced (Figure 39B).

Figure 39. Severity of cortical actin phenotype appears to depend on when BNC1 is knocked down during hPSC-epi differentiation. LHS panel, secondary only epi-EMT. In (A), there is no BNC1 reduction (BNC1-positive cells in green); in (B) BNC1 has been knocked down with tetracycline since d3 of hPSC-epi differentiation; in (C) since d6 of hPSC-epi differentiation and in (D) since d7 hPSC-epi. EMT is induced at d8 of hPSC- epi differentiation. Cells in panel (D) appear to have a more pronounced cortical actin localisation (actin in red) than in (A), (B), or (C). Scale bars are 100um. n=3.

3.3.5 Cortical actin localisation is observed both in TGFβ and FGF2-mediated epi- EMT

Feuerborn and colleagues had previously presented evidence that Bnc1 modulates TGFβ- induced epithelial dedifferentiation, as siRNA-mediated Bnc1 reduction increased cell- cell adhesion and impacted TGFβ-induced sheet dissolution in mammary epithelial cells. Both the TGFβ canonical activator Smad3 and Bnc1 were also found to associate via chromatin-IP (Feuerborn et al., 2015). I hypothesised that the cortical actin phenotype

Sophie McManus – February 2020 141

BNC1 regulates human epicardial heterogeneity and function

was mediated specifically via disrupted BNC1-TGFβ signalling. I therefore induced siKD epi-EMT with FGF2, and imaged actin filaments to compare FGF2-mediated EMT with the TGFβ-mediated phenotype in siKD epi-EMT. If the cortical actin phenotype seen with BNC1 knockdown was indeed due to a deficiency in BNC1-induced TGFβ signalling, we would expect the FGF2-induced epi-EMT cells to exhibit a similar actin phenotype both in no-TET and TET conditions. However, preliminary results (n=1) showed the same actin localisation phenotype in TGFβ and FGF2-mediated epi-EMT alongside BNC1 knockdown (Figure 40). This experiment suggests that BNC1 is acting downstream of both TGFβ and FGF2 signalling pathways in the regulation of EMT.

Figure 40. A cortical actin localisation phenotype is seen in siKD epi-EMT induced both by TGFβ and FGF2. Top panels show actin filaments at d1 of epi-EMT, visualised by

142 Sophie McManus - February 2020

Chapter 3: Results

phalloidin (red) in no TET conditions for both TGFβ treatment (left-hand side) and FGF treatment (right-hand side); some BNC1-positive cells are visible (green). The bottom panels show both TGFβ (left-hand side) and FGF2 treatment (right-hand side) after acute tetracycline administration to induce BNC1 knockdown; there is cortical actin localisation (red) in both conditions. Fewer cells are BNC1-positive (green). n=1. Scale bars 100um.

3.4 BNC1 – a possible role in regulating cell migration? Gene Ontology over-representation analyses in our hPSC-epi scRNA-seq data revealed that the BNC1high population in our hPSC-epi expressed more genes involved in actin filament regulation, actin filament-based movement, muscle cell migration and positive regulation of locomotion and ameboidal-type migration than the TCF21high population, as outlined in Chapter 1, Section 1.8.2, and shown in Figure S1, Appendix).

In a model of corneal wound repair, Bnc1 knockout mice displayed a slower healing response than controls (X. Zhang & Tseng, 2007). Conversely, a breast cancer cell line transfected with BNC1 siRNA displayed enhanced migration via scratch assay (Pangeni et al., 2015). Given that our BNC1 knockdown epi-EMT cells displayed disorganised actin filaments, and actin remodelling is a key feature of cell migration (Haynes et al., 2011a), I hypothesised that BNC1 disruption would have a negative effect on migration in our hPSC-epi cells. I therefore used in vitro assays in order to investigate a putative role for BNC1 in hPSC-epi migration.

3.4.1 Investigating BNC1 in a 2D epi-EMT migration model

I initially employed scratch assays to assess epi-EMT cell migration, comparing scrambled controls with BNC1 knockdown via siRNA (not shown). However, I found this method to be highly variable, and a sub-optimal approach for quantification of relative cell migration. In particular, scratching tissue culture plates seemed to prevent the majority of cells from migrating across the scratch border, perhaps because scratching removed the plate’s gelatin coating, preventing cell attachment. In the place of scratch assays, I used the InCell microscope to track individual cell movements in the epi-EMT

Sophie McManus – February 2020 143

BNC1 regulates human epicardial heterogeneity and function

model. The experimental setup is shown in Figure 41. I elected to compare no-TET controls with cells that had been cultured with TET from day 5 of hPSC-epi culture and d7 of hPSC-epi culture, to compare the effect of knocking down BNC1 at different stages.

Figure 41. Schematic showing experimental design for tracking 2D epi-EMT migration by use of the InCell microscope. Epi-EMT cells are imaged over 48 hours in the InCell microscope and the migration is subsequently tracked in ImageJ.

Some cells died during the InCell incubation, but overall survival was sufficient to enable individual cell tracking. 96 images were taken per field of view over a period of 48 hours, before I later selected fields at random for tracking. For each experiment, 30 cells were tracked, across at least three different fields of view for each well, and results were compiled from three independent experiments. Representative Medusa plots for tracked cells in each condition are shown in Figure 42, demonstrating that most cells were motile during the incubation period. BNC1 knockdown was verified by immunocytochemistry (Figure 43).

144 Sophie McManus - February 2020

Chapter 3: Results

Figure 42. Medusa plots comparing 2D migration of BNC1 knockdown epi-EMT cells to controls using the InCell microscope. Cells were treated with tetracycline from d5 or d7 of hPSC-epi differentiation to knock down BNC1, and compared to no-TET controls. Plots generated via Point-and-Track plugin in ImageJ, followed by individual coordinate plotting in Ibidi Chemotaxis software to generate Medusa plots. Plot shows cell tracks generated from one representative experiment. Plots were generated using the same XY scale. 30 cells were tracked per condition per experiment, n=3.

Sophie McManus – February 2020 145

BNC1 regulates human epicardial heterogeneity and function

Figure 43. BNC1 knockdown in epi-EMT cells, post-tracking in the InCell. Arrowheads indicate BNC1-positive nuclei (green) in the no TET epi-EMT condition, compared to a loss of BNC1 in the bottom right hand panel. Scale bars 100um. n=1.

3.4.2 BNC1 knockdown significantly reduces 2D epi-EMT cell motility

In agreement with my hypothesis, I found that BNC1 knockdown cells exhibited significantly reduced epi-EMT cells’ distance and velocity when compared to no-TET control cells. Distance was calculated via use of Ibidi’s chemotaxis software and is represented as an index of accumulated distance for each cell. In the no-TET control condition, the mean accumulated cell distance was 963.00 ± 481.10, compared to d5 TET- treated 703.23 ± 473.47, and d7 TET-treated 575.87 ± 366.23 (arbitrary units). Velocity was calculated by the chemotaxis software by dividing accumulated distance by the timeframe of the experimental run. Mean velocities were 6.41 ± 2.80 for no-TET cells, 4.47 ± 2.46 for d5 TET-treated cells, and 3.67 ± 2.01 for d7 TET-treated cells. Interestingly, the more pronounced relative defect in migration was noted in cells treated with tetracycline at d7 of hPSC-epi culture, compared to d5 TET treatment. This is similar to the observation we had made regarding cortical actin phenotype severity corresponding to the more acute BNC1 knockdown, in Section 3.3.4. Data for each tracked cell, as well as mean values from each condition, are shown in Figure 44.

These combined observations indicate that, in our epi-EMT model, while cells may compensate with regards to their actin organisation if they develop without BNC1 expression, an acute BNC1 perturbation results in a pronounced migratory-deficient phenotype, as well as cortical actin localisation. It appears possible that there might be crosstalk between the two hPSC-epi subpopulations, which may be necessary for some of the phenotypes investigated here. For example, the reduction in global cell migration seen when BNC1 is knocked down might be due to a lack of cell-cell signalling between the BNC1high and the TCF21high compartments. I wanted to investigate relative motility (and the effect of BNC1 knockdown) in each hPSC-epi subpopulation compared to cell movements in the mixed hPSC-epi, so performed cell sorting for each population prior to

146 Sophie McManus - February 2020

Chapter 3: Results

seeding cells for InCell tracking. However, so few cells survived each sort that these experiments failed.

Figure 44. InCell-tracked epi-EMT cells display reduced motility when BNC1 is knocked down. (A) d5 and d7 TET-treated epi-EMT cells display significantly reduced accumulated distance compared to no-TET controls. Each dot represents a single tracked cell. ***, p-value < 0.001. (B) Mean distance values. ***, p-value <0.005. Error bars are s.d. (C) TET-treated epi-EMT cells display significantly reduced velocity compared to controls. Each dot represents a single tracked cell. ***, p-value <0.001. (D) Mean velocity values. ***, p-value <0.001. Error bars are s.d. Values were compared via one-way ANOVA. n=3.

Sophie McManus – February 2020 147

BNC1 regulates human epicardial heterogeneity and function

3.4.3 Investigating BNC1 in a 3D model of epi-EMT invasion

In order to investigate epi-EMT cell migration under more physiological conditions, I developed a 3D co-culture Matrigel invasion assay, seeding GFP-positive hPSC-epi cells on top of mStrawberry-positive hPSC-derived cardiomyocytes in angiogenesis slides. The experimental set-up is pictured in Figure 45: the mStrawberry-positive CM are seeded within a Matrigel:media mix at the bottom of the well, and the GFP-positive hPSC-epi cells are added on top in the well, once the Matrigel has set. hPSC- cardiomyocyte generation is described in Chapter 2. The aim of this assay was to emulate, to some extent, the in vivo situation of an epicardial layer covering the myocardium, hence observe the hPSC-epi cells’ invasion into a cardiomyocyte layer. hPSC-epi cells were either no-TET controls or treated with TET either from the start of hPSC-epi differentiation, to emulate the experiments outlined in Section 3.1, wherein alterations in hPSC-epi marker expression were noted in cells that developed from LPM without BNC1. I also seeded some hPSC-epi that had received TET at d4 of differentiation, to try and compare the effect of knocking down BNC1 at different time points. Wells were fixed for imaging at day 1 and day 2 after seeding (n=3 experiments; n=2 for d1, d2 and d3 post-seeding).

148 Sophie McManus - February 2020

Chapter 3: Results

Figure 45. hPSC-epi invasion assay schematic. mStrawberry-positive cardiomyocytes are co-cultured with GFP-positive hPSC-epi cells to enable imaging for relative invasion. mStrawberry-positive cardiomyocytes are contained within 10ul of matrigel:RMPI mix (0.8 mm thick); hPSC-epi were seeded on top within 50ul of WBR media. hPSC-epi cells subsequently invade the cardiomyocyte layer. Cell invasion can then be imaged via Z- stacking confocal microscopy.

3.4.4 BNC1 knockdown causes non-significant impairment in hPSC-epi invasion

No-TET and TET-treated GFP-positive hPSC-epi invasion into mStrawberry-positive CM was imaged before generation of 3D renders in ImageJ. D4-treated hPSC-epi died in two different experiments (not shown), so comparison was only possible between no-

Sophie McManus – February 2020 149

BNC1 regulates human epicardial heterogeneity and function

TET hPSC-epi and hPSC-epi that had been treated with TET from the start of hPSC-epi differentiation. Upon comparison between these two conditions, we noted that TET- treated cells appeared to exhibit reduced migration into the CM layer; this was most striking at d1 post-seeding (shown in Figure 46A vs 46B). Imaris quantification for the average Z-position for each cell surface (green hPSC-epi surface Z-position compared to red CM surface Z-position) revealed a non-significant trend towards a greater mean distance between hpsc-epi and CM in the TET-treated condition (p = 0.29). The mean Z- position difference between hPSC-epi and CM for the no-TET hPSC-epi was 52.59 ± 29.95um, whereas for the TET-treated hPSC-epi condition the mean difference was 65.81 ± 40.35um (Figure 46C). Invading GFP-positive hPSC-epi cells did not appear to vary in their morphology between conditions. Therefore, unlike the 2D migration epi-EMT assay, I did not see a significant difference in invasion for BNC1 knockdown hPSC-epi cells into CM.

Figure 46. Invasion into the CM layer by hPSC-epi trends towards impairment when BNC1 is knocked down. (A) Control hPSC-epi GFP-positive cells appear to move towards the mStrawberry CM layer. (i), (ii) and (iii) indicate independent experiments imaged. (B) In the presence of tetracycline, invasion of GFP-positive hPSC-epi into mStrawberry-positive CM appears to be impaired. (C) There is a non-significant trend

150 Sophie McManus - February 2020

Chapter 3: Results

towards a reduced distance between no-TET hPSC-epi and CM when compared to TET- treated hPSC-epi and CM. n=3 experiments, error bars show s.d.

Overall, it appears that BNC1 may regulate various aspects of hPSC-epi cell migration in vitro, in particular cell motility. I saw a potential role for BNC1 in mediating a cortical actin phenotype and significant alteration in cell F-actin anisotropic index in a simple EMT model via phalloidin staining. We also noted a significant reduction in relative cell migration in an epi-EMT model, particularly via an acute reduction in BNC1. We observed a non-significant trend towards reduced TET-treated hPSC-epi invasion towards CM in a simple invasion assay.

3.5 BNC1 – identifying potential epicardial targets BNC1 is a zinc finger protein and consequently a putative transcription factor, but its potential epicardial target genes are currently uncharacterised. Bioinformatic analyses for our hPSC-epi scRNA-seq had suggested BNC1 as a key regulator in the hPSC-epi signalling network. Furthermore, perturbing BNC1 expression resulted in altered expression of canonical epicardial genes WT1 and TCF21, combined with cellular phenotypes such as increased proliferation and apoptotic indices. These results, combined with network inference models, implied that BNC1 serves as an upstream regulator of a transcriptional hierarchy that regulates and maintains epicardial cell identity in a human stem cell model. We therefore hypothesised that BNC1 would directly target WT1 and TCF21 in hPSC-epi.

3.5.1 ChIP sequencing for BNC1 targets

Protein-DNA interactions underlie many different cellular processes, hence characterising these interactions can give some insight into a gene’s specific mechanisms. I therefore performed chromatin immunoprecipitation (ChIP) combined with high- throughput sequencing (ChIP-seq) for BNC1 in hPSC-epi, to provide an unbiased snapshot of which specific DNA sequences were occupied by BNC1 in our hPSC-epi culture. I used d8 hPSC-epi cells for ChIP-sequencing, as BNC1 is well-expressed at this stage, and this approach is consistent with the time point when hPSC-epi was previously split into TGFβ for the different epi-EMT models.

Sophie McManus – February 2020 151

BNC1 regulates human epicardial heterogeneity and function

The experimental workflow for ChIP-seq is outlined in Figure 47. Prior to ChIP- sequencing, optimisation experiments were carried out; for example, shear testing (Figure 48), which allowed us to check DNA fragment size after sonication, ensuring fragments were in the 200-400bp range. I also performed test ChIPs with positive and negative control antibodies, CTCF and IgG, and verified positive and negative control binding regions for CTCF antibody by qPCR (not shown), to validate the efficacy and specificity of the protocol used.

Figure 47. The experimental workflow for ChIP-seq in identifying potential BNC1 targets. HPSC-epi cells are harvested at d8 of differentiation for DNA-protein crosslinking before lysis and DNA sonication. Sheared DNA is then incubated for immunoprecipitation with BNC1 or control antibody-coated magnetic beads. After washing to remove residual antibody, DNA is purified for qPCR or subsequent library preparation and ChIP-sequencing of potential BNC1 targets.

Input, IgG, BNC1 and TCF21 hPSC-epi ChIPPed-samples were used for ChIP- sequencing. We considered TCF21 to be a reasonable choice for a positive control antibody as there is published ChIP-seq data for TCF21 targets, albeit in coronary SMC rather than epicardial cells (Sazonova et al., 2015).

152 Sophie McManus - February 2020

Chapter 3: Results

Figure 48. CHIP shear test to optimise the number of sonication cycles. Desired fragment length is between 100-400bp. Numbers at the top indicate the number of cycles of 30s sonication. Ladders are 1kb and 100bp respectively. n=1.

3.5.2 ChIP-seq analysis reveals potential targets including actin-binding protein supervillin

ChIP-seq analysis was conducted by Dr Vincent Knight-Schrijver. We expected no protein-binding peaks in the Input sample, as there is no immunoprecipitation; IgG control peaks are also considered to be noise, due to IgG antibodies harbouring some non- specific DNA-protein binding sites. Hence the BNC1 ChIP data was cleaned primarily by simple discounting of peaks that also arose in Input or IgG samples; this was done via a heuristic approach, cutting off samples in controls that were above a normal distribution. This ChIP-seq analysis identified 16 peaks in the first BNC1 ChIP-seq experiment; the genomic regions for these peaks are shown in Table 9. The TCF21 ChIP-seq returned 71 peaks. The majority of peaks for each ChIP-seq sample were in promoter regions of genes: this was the case for 13/16 for BNC1 peaks, and 60/71 for the TCF21 peaks.

The top gene target suggested for BNC1 binding in ChIP-seq was 3-Hydroxyacyl-CoA Dehydratase 1, HACD1 (also known as PTPLA). The peak is shown compared to Input and IgG control in Figure 49. Of interest, this gene is known to be preferentially expressed in both the foetal and the adult heart, compared with lower expression levels seen in skeletal and smooth muscle, leading to the suggestion that this gene may be involved in cardiac development [RefSeq, 2008, and the Human Protein Atlas]. Canine loss of HACD1 causes centronuclear myopathy (Maurer et al., 2012). Loss-of-function

Sophie McManus – February 2020 153

BNC1 regulates human epicardial heterogeneity and function

studies for HACD1 have also revealed myopathy phenotypes in mice and humans; human patients also exhibited mild cardiac defects (Blondelle et al., 2015) (Muhammad et al., 2013) (Sawai et al., 2017). A PCA using our scRNA-seq data coloured for the level of HACD1 and BNC1 indicated that cells which are relatively high in HACD1 expression are also relatively high in BNC1 expression (not shown).

As we saw significant changes in epicardial gene expression upon BNC1 knockdown, I had hypothesised that BNC1 might bind WT1 and TCF21 promoter regions to directly effect these expression changes. However there was no significant peak detected in the WT1 or the TCF21 promoter regions for BNC1 ChIP-seq. It seems conceivable that regulation of WT1 and TCF21 may therefore occur via indirect influence from BNC1. Indeed, some possible targets identified by ChIP-seq are linked to transcriptional regulation via chromatin modifications and RNA binding.

For example, High mobility group protein B1 (HMGB1) is known to mediate changes in chromatin modelling, via binding linker regions of nucleosomes to reduce stability, and bends promoter DNA to enhance binding of transcription factors (Lange & Vasquez, 2009). Other putative targets are also associated with DNA-binding transcription factor activity, transcriptional regulation, and RNA binding: these are Distal-Less Homeobox 2 (DLX2), Zinc finger protein 837 (ZNF837), and Ubiquitin carboxyl-terminal hydrolase isozyme L5 (UCHL5) respectively [GO Central]. Rest Co-Repressor 3 (RCOR3) was also identified, a gene which has been associated with chromatin binding, hence transcriptional repression, alongside other members of the REST family [GO Central] (Yu, Johnson, Kunarso, & Stanton, 2011). A further potential target was Forkhead box protein J3 (FOXJ3); this gene is associated with DNA-binding transcription activator activity, as well as positive regulation of transcription via RNA polymerase II [GO Central] (Alexander et al., 2010). Furthermore, Anthrax Toxin Receptor 1 (ANTXR1) and Protein Phosphatase 1 Regulatory Subunit 12C (PPP1R12C) are both genes associated with cell migration, actin binding and assembly of the actin cytoskeleton [RefSeq 2012]. Another potential target that is particularly strongly associated with actin filament- binding is Supervillin, shown as a peak in BNC1 ChIP-seq in Figure 49, and this gene is discussed in further detail, both in Section 3.6 and in Chapter 4.

154 Sophie McManus - February 2020

Chapter 3: Results

A consensus motif was found for TCF21 binding in 53 of 71 peaks; however, no consensus motif was identified for BNC1. The motif found for TCF21 was TCTCGCGAGA. In contrast, Sazonova and colleagues had previously identified CAGCTG and CATCTG as possible motifs for TCF21 binding in coSMCs. Moreover, the putative targets we found via our TCF21 ChIP-seq were different to those already published (Sazonova et al., 2015). This may be due to our using different cell types.

Gene Peak Score Counts.per.100.peak.bases Chromosome symbol Feature

1 0.999898 21.46523575 10 HACD1 Promoter

2 1 20.26767403 1 FOXJ3 Promoter

3 1 20.00083003 2 DLX2 Promoter

UCHL5 | 4 0.999803 19.46846963 1 TROVE2 Promoter

5 0.9999 18.47008658 14 DNAAF2 Promoter

6 0.999899 17.66739986 1 RCOR3 Promoter

7 0.954291 16.97251199 2 ANTXR1 Promoter

8 0.971409 16.47332046 2 NHEJ1 Promoter

9 0.999998 15.53677531 6 MPC1 Promoter

10 0.955222 14.97574587 14 N/A

HMGB1 | 11 0.999925 14.01221908 13 USPL1 Promoter

12 0.9999 14.01221908 18 C18orf25 Promoter

13 1 13.03572732 19 PPP1R12C

14 0.9999 13.00053952 2 DLX2

15 0.999827 11.00960071 19 ZNF837 Promoter

Sophie McManus – February 2020 155

BNC1 regulates human epicardial heterogeneity and function

16 0.999996 10.19248516 10 SVIL Promoter

Table 9. BNC1 ChIP-seq peaks identified by Dr Knight-Schrijver after data clean-up in BNC1 ChIP-seq.

Figure 49. ChIP-seq peaks for selected putative targets HACD1 and SVIL in BNC1 ChIP compared both to IgG and input controls. ChIP-seq analysis and this figure generation were performed by Dr Vincent Knight-Schrijver.

3.5.3 ChIP-seq target validation via qPCR

In order to validate the peaks found via ChIP-seq I designed qPCR primers for targets suggested by ChIP analysis and ran further BNC1 ChIPs (n=2 ChIP experiments run independently from ChIP-seq) for different hPSC-epi differentiations at d8 of culture, selecting some top targets for the BNC1 ChIP. Targets identified by our TCF21 ChIP-seq were also validated by qPCR (not shown). qPCR analysis for BNC1 putative targets HACD1, MPC1, RCOR3 and SVIL showed enrichment for each target compared to IgG control. Enrichment for HACD1 relative to IgG was 21.55 ± 3.9-fold; for MPC1 it was

156 Sophie McManus - February 2020

Chapter 3: Results

39 ± 3.76-fold; for RCOR3, 40.94 ±14.24-fold; and for SVIL, 36.25 ± 6.66-fold (all shown in Figure 50).

Figure 50. ChIP-qPCR to validate selected BNC1 targets identified by ChIP-seq. Graphs show fold enrichment relative to IgG control for HACD1, RCOR1, MCP1 and SVIL. For HACD1, the mean fold enrichment is 21.55; for MPC1 this is 38.995; for RCOR3, 51.015; for SVIL, 36.25. Values are from two independent hPSC-epi ChIP experiments. Error bars show standard deviation between the two experiments.

Sophie McManus – February 2020 157

BNC1 regulates human epicardial heterogeneity and function

3.6 Focusing on SVIL as a putative BNC1 target A potential candidate of interest reported by our BNC1 ChIP-seq analysis was Supervillin (SVIL) (Figures 49, 50). This gene is a member of the villin and gelsolin family (Pestonjamasp, Pope, Wulfkuhle, & Luna, 1997). Interestingly, supervillin is known to interact with many cytoskeletal proteins, including F-actin and myosin II, as well as to localise at focal adhesions (Bhuwania et al., 2012; Z. Fang et al., 2010; Takizawa, Ikebe, Ikebe, & Luna, 2007). SVIL has also been implicated in the formation of invadosomes (Bhuwania et al., 2012) and mediation of effective cell migration in scratch assays (X. Chen et al., 2018). In a study by Wulfkuhle et al, SVIL overexpression studies yielded an overall increased cellular actin content, as well as disrupted integrity of focal adhesions (Wulfkuhle et al., 1999). Moreover, upregulated SVIL expression has been associated with increased invasiveness of tumours (X. Chen et al., 2018). Given extensive literature evidence for SVIL’s roles in EMT, I hypothesised that BNC1 may be targeting SVIL to effect actin remodelling, hence selected this gene as a particular target of interest.

3.6.1 Heterogeneous SVIL expression in hPSC-epi subset

I first looked for the expression of SVIL in our hPSC-epi single cell dataset. SVIL was expressed in some hPSC-epi cells, but expression was not confined to BNC1high cells (as was the case for HACD1), shown in Figure 51.

158 Sophie McManus - February 2020

Chapter 3: Results

Figure 51. Principal component analysis coloured for SVIL expression to visualise the scRNA-seq dataset from Gambardella et al. PCA shows BNC1 and SVIL expression in hPSC-epi at d9 of differentiation. Whereas BNC1 (bottom panel) is differentially expressed between two hPSC-epi subpopulations, SVIL expression is not segregated (top panel) – a few epicardial cells express SVIL at varying levels.

3.6.2 SVIL expression falls during epi-EMT progression

As mentioned in Section 3.6, SVIL expression is associated with regulation of the actin cytoskeleton during EMT (Crowley, Smith, Fang, Takizawa, & Luna, 2009; Khurana & George, 2008) (X. Chen et al., 2018). I therefore expected that TGFβ treatment would mediate an increase in SVIL expression in our epi-EMT model, as has been previously reported, which may thereby mediate F-actin remodelling (X. Chen et al., 2018; Crowley

Sophie McManus – February 2020 159

BNC1 regulates human epicardial heterogeneity and function

et al., 2009). However, instead SVIL expression seemed to mirror that of BNC1: upon TGFβ treatment, expression of both genes in epi-EMT relative to hPSC-epi fell (Figure 52). Attempts to perform immunocytochemistry and Western Blot for SVIL proved unsuccessful, due to a lack of effective antibody.

Figure 52. SVIL mRNA level is reduced in the epi-EMT model, shown by qPCR. SVIL and BNC1 are reduced concomitantly by TGFβ treatment. mRNA normalised to d7 hPSC- epi pre-TGFβ treatment. n=3 experiments. Error bars show s.d.

3.6.3 ChIP for BNC1 in epi-EMT: validating SVIL binding

I performed further BNC1 ChIP on epi-EMT cells after the first 24 hours of TGFβ treatment, to compare SVIL enrichment in the epi-EMT context to that previously seen in hPSC-epi, thereby aiming to infer whether BNC1 was binding SVIL when cells are exposed to TGFβ. While there was enrichment in the BNC1 ChIP relative to IgG control, the degree was not as high as seen in the hPSC-epi ChIP experiments (15-fold compared to 36.25 ± 6.66-fold) as shown in Figure 53. As BNC1 expression is reduced upon TGFβ addition (Section 3.3.2) perhaps this causes the reduction in fold enrichment for SVIL in BNC1 epi-EMT ChIP.

160 Sophie McManus - February 2020

Chapter 3: Results

Figure 53. SVIL enrichment in epi-EMT ChIP. SVIL is enriched 15-fold relative to IgG control in BNC1 Ab-ChIPPed hPSC-epi-EMT cells. n=1 epi-EMT ChIP experiment.

3.6.4 SVIL knockdown in epi-EMT model yields cortical actin phenotype

I selected SVIL as a candidate to investigate further in our epi-EMT model, hypothesising that its well-characterised effect in mediating actin binding would be relevant to our observed cortical actin phenotype in BNC1 knockdown epi-EMT (Bhuwania et al., 2012; X. Chen et al., 2018; Crowley et al., 2009; Khurana & George, 2008). As a preliminary investigation into potential SVIL function in our epi-EMT model, I performed siRNA- mediated SVIL knockdown. SVIL mRNA level was reduced to 28.4 ± 4.96% of scrambled control (Figure 54A). Intriguingly, imaging F-actin filaments in SVIL epi-EMT knockdown revealed a similar phenotype to that previously seen with BNC1 knockdown (as shown in Figure 54B): actin appeared to be more localised to the cell membrane than aligned in stress fibres. I also performed qPCR for BNC1 in SVIL knockdown samples and found a non-significant trend (p=0.11) (Figure 54C) towards BNC1 increase when SVIL was reduced, implying there may be some degree of a compensatory BNC1 upregulation when SVIL expression is perturbed.

Sophie McManus – February 2020 161

BNC1 regulates human epicardial heterogeneity and function

Figure 54. SVIL knockdown in epi-EMT model mediates similar cortical actin phenotype to that seen in BNC1 knockdown. (A) SVIL mRNA is reduced by over 60% by SVIL siRNA relative to scrambled control, n=3. ** p<0.01. (B) SCR siRNA-treated epi-EMT displays aligned actin filaments (green), compared to SVIL-treated epi-EMT, which exhibits some cortical actin deposition. Scale bars 100um. (C) BNC1 mRNA is increased in SVIL KD, but this is not significant, p=0.11. n=3.

It therefore seemed possible that the cortical actin phenotype and decreased actin filament

162 Sophie McManus - February 2020

Chapter 3: Results

anisotropy seen in BNC1 knockdown studies may be caused by an impaired BNC1-SVIL binding at the point of TGFβ exposure. Therefore I checked SVIL levels in BNC1 knockdown epi-EMT samples, to verify whether reduction of BNC1 concomitantly reduced SVIL, both when BNC1 knockdown had been induced since the LPM stage (n=3) and acutely at d7 hPSC-epi culture (n=2); however, the relative change in SVIL expression that was observed between conditions was highly variable between independent experiments (Figure 55).

Figure 55. BNC1 knockdown does not alter SVIL expression in the epi-EMT model. SVIL is normalised to GAPDH in three independent experiments before normalisation of relative tetracycline-treated epi-EMT SVIL level to no-tetracycline control. Error bars show s.d.

Overall, this preliminary work in identifying possible BNC1 targets has suggested several possible candidates of interest; BNC1 ChIP enrichment for these targets was validated in further ChIP-qPCR experiments. In particular, Supervillin was selected as a candidate of interest, given its well-documented roles in actin binding during EMT. Preliminary SVIL knockdown studies effectively reproduced our BNC1 knockdown phenotype in epi-EMT, lending support to a putative mechanism by which the BNC1 protein binds the SVIL promoter, hence allows actin remodelling in hPSC-epi cells. The low number of peaks

Sophie McManus – February 2020 163

BNC1 regulates human epicardial heterogeneity and function

returned by my ChIP-seq experiment may indicate issues with antibody suitability, as well as a requirement for further protocol troubleshooting, and these caveats will be detailed extensively in Chapter 4. Given these technical challenges, I chose to perform bulk RNA sequencing as an alternative technique by which to identify possible both direct and indirect targets of BNC1. I chose to perform RNA sequencing in the context of epicardial EMT, in order to identify possible mechanisms behind the cortical actin and migratory impairment phenotypes reported in Sections 3.3.5 and throughout Section 3.4.

3.7 Investigating transcriptomic changes induced by BNC1 knockdown in epi-EMT I hypothesised that BNC1 knockdown was blocking the activation of genes involved in EMT processes, such as response to growth factor stimulation, cytoskeletal remodelling and cell migration, hence mediating the observed actin localisation and reduced cell migratory phenotypes. Bulk RNA sequencing was performed for epi-EMT samples to identify potential mechanisms and new hypotheses to test.

Epi-EMT was induced at d8 of hPSC-epi culture, as in previous experiments, and cells were collected for RNA extraction and sequencing after 24 hours of TGFβ treatment. There were four 1Ei epi-EMT technical replicate cultures submitted for RNA library preparation per condition. The RNA sequencing sample conditions were as follows: no- tetracycline epi-EMT induction (no TET), tetracycline from LPM in hPSC-epi culture followed by epi-EMT induction (d1 TET), and tetracycline from d7 of hPSC-epi culture followed by epi-EMT induction (d7 TET).

The aim of this sequencing experiment was to perform differential expression analysis between no TET epi-EMT and epi-EMT with BNC1 knockdown at two different stages of hPSC-epi differentiation. I was particularly interested in investigating the degree of differential expression between no TET and d7 TET, as the latter had resulted in pronounced phenotypes in epi-EMT assays compared to no TET controls. I aimed to identify what differences in gene expression may underlie these. I also hypothesised that the more pronounced d7 TET epi-EMT phenotype seen compared to d1 TET epi-EMT would be mediated by differential expression of genes involved with cytoskeletal remodelling between the two BNC1 knockdown time points.

164 Sophie McManus - February 2020

Chapter 3: Results

3.7.1 Different BNC1 knockdown conditions segregate in three groups via Principal Component Analysis

RNA bulk sequencing analysis was conducted by Dr Nicolas Le Novère, and the methods are described from Section 2.36 onwards, Chapter 2. Principal component analysis showed that samples within each condition (no TET, TET d7 and TET d1) segregated together, with the exception of replicate 3 for TET d7 condition, which grouped within the no-TET group, implying that tetracycline knockdown had been ineffective in this sample (Figure 56); this was therefore treated as an outlier and removed from the analysis. Hence the third replicate for each condition was also removed, before differential gene expression analysis was performed between the three different conditions.

Figure 56. Principal component analysis shows clear separation for the three BNC1 knockdown conditions for bulk sequencing in epi-EMT.

Sophie McManus – February 2020 165

BNC1 regulates human epicardial heterogeneity and function

3.7.2 Differential expression analysis reveals changes in relative extracellular matrix and actin remodelling genes

Differential expression analysis (DESeq) demonstrated that 991 genes were differentially expressed between the d7 TET and no TET conditions, 321 between d1 TET and no TET, and 811 between d1 TET and d7 TET. A Venn diagram summarising the number of differentially expressed genes between the different epi-EMT sequencing conditions is shown in Figure 57.

Figure 57. Venn diagram showing the number of differentially expressed genes between different epi-EMT conditions, and respective overlaps between the differential expression comparisons.

BNC1 knockdown was consistently pronounced across TET-treated replicates relative to no TET controls (d1 TET BNC1 was reduced 5.95-fold relative to no-TET, while d7 TET

166 Sophie McManus - February 2020

Chapter 3: Results

BNC1 expression was reduced 4-fold). Among the genes that were increased in the d7 TET epi-EMT population compared to no TET epi-EMT were those coding for cardiac troponin 2 (TNNT2), which can bind actin (B. Wei & Jin, 2016); a cell adhesion protein, matrix remodelling-associated protein 5 (MXRA5) (Poveda et al., 2017; Walker & Volkmuth, 2002); the apelin receptor (APLNR), the function of which is still under investigation (Chapman, Dupré, & Rainey, 2014; Zhi Wang et al., 2015); ECM component latent TGFβ binding protein 1 (LTBP1) (Robertson et al., 2015), myocardin (MYOCD), which may alter actin expression and therefore affect collagens (Z. Shi & Rockey, 2017); and matrix metalloproteinases (MMPs) 15 and 28. MMPs are involved in ECM remodelling and cellular invasion during migration (Q. K. Chen, Lee, Radisky, & Nelson, 2013; M. Huang et al., 2016).

Conversely, genes including MMP9, which is associated with EMT progression in cancer cells (Bai et al., 2017; Gilles, Newgreen, Sato, & Thompson, 2013), and MMP10 were expressed at a lower level in d7 TET epi-EMT compared to no TET, as were ECM- associated proteins layilin (LAYN) and laminin γ2 (LAMC2). There was reduced expression for mitogen-activated protein kinase (MAPK) signalling genes (MAPK4K and MAP3K14) in d7 TET epi-EMT compared to no TET epi-EMT. These differentially expressed genes are shown in Figure 58 and their possible relevance is discussed in Section 3.7.6.

Sophie McManus – February 2020 167

BNC1 regulates human epicardial heterogeneity and function

Figure 58. Volcano plot showing genes that are differentially expressed between d7 TET epi-EMT and no TET EMT. The plot shows the amplitude of differential gene expression changes (x-axis) and their significance (y-axis). If the highlighted gene is more expressed in d7 TET epi-EMT, it is highlighted in brown, and the x-axis value is positive. If the gene is relatively more expressed in the no TET epi-EMT condition, it is highlighted in blue, and the x-axis value is negative. n=3 technical replicates (three different 1Ei epi- EMT cultures).

Comparing gene expression in d1 TET epi-EMT and no TET epi-EMT revealed that there was increased expression of genes including MYOCD, TNNT2, MEF2C, and members of the actin family, cardiac muscle alpha actin (ACTC1) and smooth muscle actin (ACTA2) in d1 TET epi-EMT relative to no TET. Conversely, differential gene expression analysis revealed reduced expression in genes including L1 cell adhesion molecule protein (L1CAM), kunitz type 1 serine protease inhibitor (SPINT1), basal cell adhesion molecule (BCAM), MMP9, LAYN and ADAM15 in d1 TET epi-EMT relative to no TET epi-EMT,

168 Sophie McManus - February 2020

Chapter 3: Results

which are associated with cell adhesion and ECM organisation (Bono, Rubin, Higgins, & Hynes, 2001; Chang et al., 2017; Hooper, Clements, Quigley, & Antalis, 2001; Kelwick et al., 2015; Samatov, Wicklein, & Tonevitsky, 2016; Yabluchanskiy, Ma, Iyer, Hall, & Lindsey, 2013). The differential expression between d1 TET epi-EMT and no TET populations is shown in Figure 59.

Figure 59. Volcano plot showing differential gene expression analysis between the d1 TET epi-EMT and no TET epi-EMT populations. The plot shows the amplitude of differential gene expression changes (x-axis) and their significance (y-axis). If the highlighted gene is more expressed in d1 TET epi-EMT, it is highlighted in yellow, and

Sophie McManus – February 2020 169

BNC1 regulates human epicardial heterogeneity and function

the x-axis value is positive. If the gene is more highly expressed in the no TET epi-EMT condition, it is highlighted in blue, and the x-axis value for that gene is negative.

The differential gene expression analysis between d1 TET epi-EMT and d7 TET epi- EMT revealed differential expression of genes including those encoding beta-III-spectrin (SPTBN2), MXRA5, ECM component alpha-1 collagen (COL11A1), secreted angiogenesis inhibitor isthmin 1 (ISM1) (Venugopal et al., 2015; Y. Zhang et al., 2011), ECM protein matrix gla (MGP) and developmental transcription factor mesenchyme homeobox 1 (MEOX1), which were relatively higher in the d7 TET epi-EMT population. Beta-III-spectrin is linked to actin binding (Avery, Fealey, et al., 2017; Avery, Thomas, & Hays, 2017), while MGP has been associated with ECM binding (Nishimoto & Nishimoto, 2014).

Conversely, genes such as endothelin 1 (EDN1), vimentin (VIM), vasorin (VASN), microtubule crosslinking factor 1 (MTCL1) and microtubule associated protein RP/EB family member 3 (MAPRE) were more highly expressed in the d1 TET epi-EMT population than d7 TET. Endothelin 1 has many functions, including mediation of migration (Planas-Rigol et al., 2017; L. U. Shi, Zhou, Chen, & Xu, 2017), as do vimentin and vasorin (Battaglia, Delic, Herrmann, & Snider, 2018; Bonnet et al., 2018; Ikeda et al., 2004). There was also differential expression of ADAM and ADAMTS (‘a disintegrin and metalloprotease with thrombospondin repeats’) genes in d1 TET epi-EMT samples relative to d7 TET, which are known to mediate migration via ECM interaction, and are further discussed in Chapter 4 (Edwards, Handsley, & Pennington, 2009). Differentially expressed genes related to ECM organisation and cell migration are highlighted in Figure 60.

170 Sophie McManus - February 2020

Chapter 3: Results

Figure 60. Volcano plot showing differential gene expression analysis between the d1 TET and d7 TET epi-EMT populations. The plot shows the amplitude of differential gene expression changes and their significance. If the highlighted gene is more expressed in d1 TET epi-EMT than in d7 TET epi-EMT, it is highlighted in brown, and the x-axis value is negative. If the gene is more expressed in the d1 TET epi-EMT condition, it is highlighted in yellow, and the x-axis value for that gene is positive.

I did not observe differential expression of SVIL, or other putative BNC1 targets identified via ChIP-seq, between different epi-EMT populations used for bulk RNA sequencing. Possible reasons for differences between the two sequencing data-sets are discussed in detail in Chapter 4.

Sophie McManus – February 2020 171

BNC1 regulates human epicardial heterogeneity and function

3.7.3 Gene ontology enrichment shows distinct phenotypic signatures for each epi- EMT population

Dr Nicolas Le Novère ran Gene Ontology (GO) over-representation analyses via WebGestalt, using genes that were differentially expressed between the three populations. GO over-representation analyses suggested a different phenotypic signature for each epi- EMT population.

3.7.4 The d7 and d1 TET epi-EMT populations have a reduced cell motility phenotypic signature

The genes down-regulated in the d7 TET epi-EMT population compared to no TET had an enrichment in GO terms in relation to migration, including ‘locomotion’, ‘cell migration’, ‘cell motility’, and ‘epithelial cell migration’, ‘negative regulation of cell–cell adhesion’ as well as terms relating to ‘cellular response to cytokine stimulus’, and ‘MAPK/p38 cascade’. There was an increase in genes that are linked to the GO term ‘actin crosslinking’ and ‘actin cytoskeleton reorganisation’ in d7 TET relative to no TET samples (for example, TNNT2).

Overall, the d7 epi-EMT population’s phenotypic signature corresponds to reduced cell motility, locomotion, and MAPK signalling. p38MAPK signalling is required for TGFβ- mediated EMT progression (Bakin, Rinehart, Tomlinson, & Arteaga, 2002; Gui, Sun, Shimokado, & Muragaki, 2012; Lamouille et al., 2014), and MAPK signalling has been shown to activate MMPs in different models (Cho, Graves, & Reidy, 2000; E.-S. Kim, Kim, & Moon, 2004; Reddy, Krueger, Kondapaka, & Diglio, 1999), so the reduced gene expression associated with these GO terms may be relevant to the impaired EMT we see in BNC1 knockdown samples. There was also GO enrichment in d7 TET relative to no TET for various metabolic processes; however, this could be due to non-specific effects of the tetracycline treatment and will be discussed in Chapter 4. The GO enrichment for d7 TET epi-EMT relative to no TET, highlighting terms that are relevant to EMT, is shown in Figure 61.

172 Sophie McManus - February 2020

Chapter 3: Results

Figure 61. Gene Ontology enrichment plot for d7 TET epi-EMT relative to no TET epi- EMT showing enriched terms such as ‘cell migration’ and ‘motility’ in no TET, and terms such as ‘actin crosslinking’ in d7 TET epi-EMT. Each bubble represents an over- represented GO term, and the disc size is proportional to the degree of GO enrichment. The y-axis presents the significance of the enrichment, while the x-axis indicates if the term enrichment is mostly due to genes over-expressed in no TET epi-EMT cells (negative z-scores) or in d7 TET epi-EMT cells (positive z-scores). Bubble colours show the mean difference of expression, for all the genes annotated by the GO term, between d7 TET epi-EMT cells (brown) and no TET epi-EMT cells (blue).

The d1 TET epi-EMT population had a decrease in gene expression relating to enriched GO terms linked to ‘response to external stimulus’, ‘cell-cell adhesion’, and ‘extracellular matrix organisation’, compared to no TET. There was also an enrichment in terms including ‘locomotion’, ‘chemotaxis’, and ‘cell motility’, indicating that the genes

Sophie McManus – February 2020 173

BNC1 regulates human epicardial heterogeneity and function

regulating these processes were downregulated when BNC1 was reduced from d1 of hPSC-epi differentiation onwards. These terms were enriched due to decreased expression of ADAM genes, which are thought to possess both cell adhesion and protease abilities (Edwards et al., 2009). The GO enrichment for d1 TET epi-EMT compared to no TET is shown in Figure 62.

Figure 62. Gene Ontology enrichment plot for d1 TET epi-EMT relative to no TET epi- EMT showing GO enrichment for terms such as ‘cell migration’, ‘extracellular matrix disassembly’ and ‘regulation of motility’ in no TET, and terms such as ‘cardiac muscle tissue development’ and ‘heart development’ in d1 TET epi-EMT. Each bubble represents an over-represented GO term, and the disc size is proportional to the degree of GO enrichment. The y-axis presents the significance of the enrichment, while the x-axis indicates if the term enrichment is mostly due to genes over-expressed in no TET epi- EMT cells (negative z-scores) or in d7 TET epi-EMT cells (positive z-scores). Bubble

174 Sophie McManus - February 2020

Chapter 3: Results

colours show the mean difference of expression, for all the genes annotated by the GO term, between d1 TET epi-EMT cells (yellow) and no TET epi-EMT cells (blue).

3.7.5 d1 TET epi-EMT has a more migratory phenotype than d7 TET epi-EMT

Interestingly, the d1 TET population had increased gene expression relating to GO enrichment for terms including ‘mesenchyme development’, ‘migration’, ‘motility’ and ‘chemotaxis’ and decreased gene expression relating to GO enrichment for ‘ECM organisation and structure’ and ‘localisation of cell’ compared to d7 TET, reflecting the less pronounced epi-EMT phenotypes seen in samples which had received TET throughout the hPSC-epi differentiation. This is shown in Figure 63.

Sophie McManus – February 2020 175

BNC1 regulates human epicardial heterogeneity and function

Figure 63. Gene Ontology enrichment plot for d7 TET epi-EMT relative to d1 TET epi- EMT showing GO enrichment for terms such as ‘extracellular matrix organisation’ and ‘extracellular matrix disassembly’ in d7 TET epi-EMT and ‘positive regulation of cell migration’ and ‘mesenchyme morphogenesis’ in d1 TET epi-EMT. Each bubble represents an over-represented GO term, and the disc size is proportional to the degree of GO enrichment. The y-axis presents the significance of the enrichment, while the x-axis indicates if the term enrichment is mostly due to genes over-expressed in no TET epi- EMT cells (negative z-scores) or in d7 TET epi-EMT cells (positive z-scores). Bubble colours show the mean difference of expression, for all the genes annotated by the GO term, between d1 TET epi-EMT cells (yellow) and d7 TET epi-EMT cells (brown).

3.7.6 Possible gene mechanisms for BNC1 knockdown epi-EMT phenotypes

I had previously seen that BNC1 reduction in epi-EMT models impacts actin remodelling and cell migration, therefore aimed to investigate transcriptional changes in epi-EMT knockdown samples via bulk RNA sequencing, in order to identify possible mechanisms for the observed phenotypes. Although both the d1 TET and d7 TET epi-EMT populations had phenotypic signatures corresponding to reduced cell migration compared to no TET epi-EMT, the differential gene expression in d1 TET appears insufficient to yield the severe phenotype seen in d7 TET epi-EMT. Comparing the GO term enrichment between d1 and d7 TET confirmed this, and some of the genes differentially expressed between the two conditions are linked to actin binding and ECM remodelling (as shown in Figure 60).

Differential expression analysis combined with GO enrichment has provided possible BNC1 downstream mechanisms to investigate in the EMT context. In particular, differential gene expression analysis indicates that potential BNC1 targets of interest in the epi-EMT model include ECM remodelling genes, such as MMPs and ADAMTs. These could be explored via matrix degradation assays with BNC1 knockdown epi-EMT samples, to compare relative matrix degradation and invasion when BNC1 is acutely reduced (Granata et al., 2017). Other possible genes of interest (e.g. layilin, laminin) that

176 Sophie McManus - February 2020

Chapter 3: Results

were identified by the DEseq could be investigated via knockdown or overexpression assays.

In particular, layilin emerged as an exciting potential candidate for future focus. Layilin was down-regulated with BNC1 knockdown relative to no TET epi-EMT. Layilin is a transmembrane hyaluronan receptor which may interact with F-actin via talin, conveying signals between ECM and the cytoskeleton (Bono et al., 2001). Layilin knockdown in lung basal epithelial cells reduced cell migration (Z. Chen, Zhuo, Wang, Ao, & An, 2008). Layilin was downregulated in our BNC1 knockdown epi-EMT samples, indicating this gene may have a role in regulating migration in our model. Layilin is discussed further in Chapter 4, in particular in Section 4.3.6.

There was also a relative reduction in gene expression pertaining to GO terms ‘MAPK cascade’ and ‘p38 signalling’ in d7 TET epi-EMT samples. TGFβ signalling is known to regulate members of the MAPK family, although MAPK signalling regulation during EMT is not well-characterised (Gui et al., 2012). This signalling mechanism could therefore also be explored via MAPK activation and inhibition assays in control and BNC1 knockdown epi-EMT cells.

Overall, ChIP-sequencing identified few possible BNC1 targets, including SVIL, an actin-binding protein that has been shown to mediate EMT in different models. Technical challenges associated with ChIP-seq led me to perform bulk RNA sequencing as an alternative method to identify putative BNC1 targets. Bulk RNA sequencing has indicated that acutely TET-treated epi-EMT samples display a phenotypic signature related to reduced migration and altered ECM remodelling, and suggest possible gene targets for further investigation into BNC1-mediated TGFβ-induced epi-EMT mechanisms.

Sophie McManus – February 2020 177

BNC1 regulates human epicardial heterogeneity and function

4. DISCUSSION

4.1 BNC1: a key regulator in hPSC-epi heterogeneity and function This thesis has presented BNC1 as a novel human epicardial gene with functional relevance across different contexts. Bioinformatic analyses suggested that BNC1 would act as a key regulator in the hPSC-epi signalling network. Moreover, Bnc1 had been linked to cell migration and proliferation in other epithelial models (Feuerborn et al., 2015; X. Zhang & Tseng, 2007). Hence this gene was selected as a key candidate of interest in epicardial cells. I hypothesised that BNC1 may act as a regulator of epicardial gene expression and epicardial cell migration. I specifically aimed to investigate BNC1 function in hPSC-epi differentiation, develop epi-EMT models to study BNC1 function in epicardial migration, and identify its putative gene targets in epicardial cells.

Firstly, I confirmed that BNC1 is expressed heterogeneously in hPSC-epi and primary human foetal epicardial cultures. In order to investigate BNC1 function, I established tetracycline-inducible knockdown hPSC cell lines. These illustrated that BNC1 expression is indeed pivotal to establishing and maintaining hPSC-epi heterogeneity, as when BNC1 was reduced during hPSC-epi differentiation I saw concomitant alterations in hallmark epicardial genes WT1 and TCF21. A relative increase in TCF21 expression was also noted in primary human foetal epicardial cultures, providing some validation for our stem cell epicardial model. THY1, a membrane marker for the TCF21high subpopulation, was significantly enriched in BNC1 knockdown hPSC-epi compared to controls, while PODXL, a marker for the BNC1high population, was reduced. In addition, BNC1 knockdown hPSC-epi cells displayed a pronounced ‘fibroblastic’ morphology, and cell proliferative index and viability were also altered. Overall, I saw that BNC1 knockdown abolishes hPSC-epi heterogeneity, and results in formation of a predominantly TCF21high THY1high hPSC-epi cell population. These results support the hypothesis that BNC1 is a key regulator in the hPSC-epi signalling network.

Secondly, perturbing BNC1 expression yielded various distinctive phenotypes in a simple in vitro model of epicardial EMT. Both a cortical actin localisation phenotype and

178 Sophie McManus - February 2020

Chapter 4: Discussion

reduced cell F-actin anisotropic index were observed, as well as reduced cell motility in BNC1 knockdown cells. As actin remodelling is a hallmark of effective EMT, this finding implied that BNC1 is a key mediator of actin reorganisation in epi-EMT, depending on the timing of knockdown. Preliminary work in an invasion model revealed a trend towards reduced invasion by BNC1 knockdown cells into a CM layer. Via bulk RNA sequencing, I aimed to investigate the global transcriptional changes in BNC1 knockdown epi-EMT models, in order to identify potential mechanisms by which BNC1 regulates actin organisation and cell migration. Differential expression analysis revealed significant differential expression in genes related to extracellular matrix degradation, cytoskeletal coordination and migration, including MMPs, ADAMTS1, and heparan sulphate sulphotransferases, (Bollini, Riley, & Smart, 2015; Q. K. Chen et al., 2013; Gilles et al., 2013; Ouderkirk & Krendel, 2014; Song, Li, Jiang, Guo, & Li, 2011) providing potential BNC1 downstream epi-EMT gene candidates and mechanisms for future study (Section 4.3) (Tsubakihara & Moustakas, 2018).

There was reduced gene expression associated with GO term enrichments relating to cell migration, cell-cell communication, response to stimuli, cell motility, protein phosphorylation and p38/MAPK signal regulation in d7 TET treated epi-EMT samples relative to no TET. Interestingly, comparing GO enrichment for d1 TET-treated epi-EMT to d7 TET-treated epi-EMT (the latter exhibiting the more pronounced phenotype) revealed that d1 TET samples, compared to d7 TET samples, had increased expression of genes corresponding to GO enrichment for locomotion, cell motility and migration, and decreased enrichment for extracellular matrix organisation and disassembly relative to d7 TET; these observations correlate with the apparently reduced phenotype severity seen in d1 TET-treated cells relative to d7 TET samples. This work could help to identify the particular molecular pathways downstream of BNC1 that are crucial for epi-EMT migration in vitro, thereby enhancing current understanding of mechanisms relevant to epicardial EMT mediation.

With regard to BNC1 targets and possible direct mechanisms of action, preliminary ChIP- seq results identified the actin-binding protein supervillin as a promising potential BNC1 target. I observed SVIL binding in my initial ChIP-seq experiment, in two further validatory ChIP-qPCR experiments, and in a further BNC1 ChIP in epi-EMT cells after 24 hours of TGFβ treatment. Given the literature evidence for SVIL’s roles in mediating

Sophie McManus – February 2020 179

BNC1 regulates human epicardial heterogeneity and function

EMT (X. Chen et al., 2018; Z. Fang et al., 2010) I chose to perform knockdown experiments in epi-EMT cells, knocking down SVIL at the TGFβ induction time point, to verify if perturbing SVIL yielded a phenotype that was similar to BNC1 knockdown, as this could imply functionality for the apparent BNC1-SVIL binding found in ChIPs. Initial SVIL knockdowns in the epi-EMT model revealed a somewhat similar cortical actin phenotype to that seen via BNC1 knockdown, hence SVIL may indeed play a functional role in our system. However, I did not see differential expression of SVIL in epi-EMT bulk sequencing data, which is discussed further in Section 4.3.

Sections 4.1.1, 4.1.2 and 4.1.3 shall discuss the broader context and implications of my thesis work.

4.1.1 BNC1 expression and function in hPSC-epi

The developing epicardium is highly active and dynamic; epicardial EMT yields the coSMC and CF essential to establishing cardiac structure and function, while epicardial trophic factors are crucial to stimulating myocardial proliferation, maturation and compaction. While neonatal mice and zebrafish demonstrate robust cardiac repair post- injury, the degree of epicardial activity and cardiac regeneration is typically limited in adult mammals. Modulating epicardial function back to a more plastic neonatal state could improve cardiac repair post-MI, by reactivation of a more effective EMT, or enhancing paracrine signalling to cardiomyocytes to induce proliferation, thereby aiding cardiac repair. Indeed, there has been progress in stimulating endogenous cardiac repair via epicardial manipulation, indicating this research avenue’s promise for the regenerative medicine field (Bargehr et al., 2019; Riley & Smart, 2009; Smart et al., 2007). A comprehensive understanding of epicardial development, hence its manipulable functions, is essential for the continuation and refinement of such regenerative strategies. While we can hypothesise that human epicardial gene expression and regulation is similar to that in animal models, inferences should be validated in a human system if developmental insights are to yield therapeutic potential. Characterising BNC1 as a functional human epicardial regulator therefore has important implications for the field.

The Rosenthal group originally identified Bnc1 as a novel mouse epicardial marker, noting its enrichment in adult epicardium and significant downregulation in post-MI

180 Sophie McManus - February 2020

Chapter 4: Discussion

hearts (Bochmann et al., 2010). Bnc1 had been previously linked to regulation of proliferation and migration in other epithelia (Feuerborn et al., 2015; D. Zhang et al., 2018; X. Zhang et al., 2012; X. Zhang & Tseng, 2007). To our knowledge, this is the first time that BNC1 expression and function have been investigated in the human heart. I observed heterogeneous expression in hPSC-epi and primary human foetal epicardium, and established several functional consequences accompanying BNC1 knockdown. Indeed, perturbing BNC1 expression at the LPM stage during hPSC-epi differentiation appears to prevent differentiation of epicardial cells, as we see that both hPSC-epi marker expression and cell morphology are drastically altered in our siKD cells. This work indicates that BNC1 should be considered an important facet within the human epicardial transcriptional network.

I saw that proliferative index was increased in my siKD hPSC-epi, while cell viability was reduced. However, proliferation assays were not carried out on WT cells in the presence and absence of tetracycline, so we cannot exclude the possibility that the alteration in proliferation was tetracycline-mediated, rather than an effect of BNC1 reduction. The finding that BNC1 knockdown appeared to significantly increase hPSC- epi cells’ proliferative index is in contrast to findings reported by Zhang and colleagues, wherein Bnc1 knockout mice displayed reduced keratinocyte proliferation in debrided corneal epithelium as a model of wound repair. I speculate that the altered proliferative and viability indices I observed in my siKD hPSC-epi were due to death in the (previously) BNC1high subpopulation, and enhanced proliferation in the remaining TCF21high subpopulation. Preliminary attempts to establish if this is the case were inconclusive. However, my results may simply differ from Zhang and colleagues’ due to model differences; furthermore, my hPSC-epi was not subject to challenge or injury as the corneal epithelium was in this work (X. Zhang & Tseng, 2007).

My work on BNC1 and its influence on maintaining epicardial heterogeneity adds to a growing body of evidence that epicardial heterogeneity has functional consequences (Acharya et al., 2012; C. M. Braitsch et al., 2012; J. Cao et al., 2016; Y. Cao & Cao, 2018; Vicente-Steijn et al., 2015; Weinberger et al., 2018). Other studies have examined single- cell heterogeneity in the human heart, but the epicardium was not the focus (Cui, Zheng, Liu, Yan, et al., 2019). There is not a great deal of work published concerning heterogeneity in human epicardial cells; in particular, functional relevance is not well-

Sophie McManus – February 2020 181

BNC1 regulates human epicardial heterogeneity and function

characterised. Dr Laure Gambardella has recently shown that while both our TCF21highWT1lowBNC1low and TCF21lowWT1highBNC1high subpopulations can effectively differentiate to form coSMC-like cells in culture, only the TCF21highWT1lowBNC1low subpopulation effectively formed CF as well as coSMC. Additional work has implied that loss of BNC1 may decrease the number of SMC progenitors surviving in culture (Gambardella et al., 2019), providing further evidence that BNC1 exerts important functional influences in the epicardial context. Moreover, GO term enrichment analysis for each hPSC-epi subpopulation analysed by scRNA-seq indicated that the BNC1high population possesses a signature pertaining to cell-cell interaction, actin remodelling and migration, while the TCF21high population mapped to terms more related to cell-cell adhesion and vasculogenesis. Taken together, this work could prove highly relevant in potential efforts to ‘purify’ stem cell models of epicardial cells to assist cell therapy efforts in the clinic (Bargehr et al., 2019) (see Section 4.3).

4.1.2 BNC1 in epi-EMT models

While we know that epicardial EMT is a crucial function during cardiovascular development, and several transcription factors pivotal to EMT coordination have been identified in animal models (Combs et al., 2011; Martínez-Estrada et al., 2010; A. W. Moore et al., 1999; Trembley, Velasquez, de Mesy Bentley, & Small, 2015), little is known about transcriptional regulation of human epicardial EMT. Feuerborn and colleagues found that Bnc1 reduction in murine mammary epithelium dysregulated TGFβ-induced cell scattering, while overexpression was found to rapidly dissolve epithelial sheets (Feuerborn et al., 2015). Epithelial features and functions are highly conserved. Given our bioinformatics analyses and literature evidence, I hypothesised that BNC1 may regulate cell migration, and implemented a simple model of EMT. BNC1 had not been investigated in the context of epicardial EMT previously; my findings therefore add to some existing evidence that BNC1 mediates TGFβ-dependent EMT processes, and present the first indication that human epicardial EMT may be at least in part BNC1- regulated.

In my inducible knockdown epi-EMT model, I saw that epi-EMT cells which had developed without BNC1 appeared similar to no-tetracycline controls, apparently exhibiting aligned, well-organised actin filaments. In contrast, acute BNC1 knockdown

182 Sophie McManus - February 2020

Chapter 4: Discussion

by siRNA or tetracycline treatment just prior to epi-EMT culture caused dramatic cortical actin localisation and reduced F-actin anisotropy. This could imply differential mechanisms of action in each hPSC-epi subpopulation in the epi-EMT context. In particular, perhaps the BNC1high subpopulation respond to TGFβ induction via separate mechanisms to the TCF21high subpopulation, thereby typically BNC1-positive cells respond to TGFβ, remodel their actin, and downregulate BNC1, hence migrate. In contrast, TCF21high cells may respond to TGFβ and migrate without the need for BNC1, as when the hPSC-epi develops without BNC1 (i.e. receives tetracycline from the LPM stage) the remaining TCF21high hPSC-epi responds to TGFβ and remodels actin in an apparently organised way. However, if BNC1 is perturbed acutely, that subset of BNC1high cells can no longer respond to TGFβ to allow correct actin remodelling. If this were the case, we might expect to see the pronounced actin phenotype in a distinct subset of epi- EMT cells, the cells that were previously BNC1high and therefore have had BNC1 expression acutely perturbed prior to TGFβ treatment. Instead, we observed the actin phenotype in most of the epi-EMT cells in the event of acute BNC1 knockdown, suggesting the picture may be more complex, potentially involving cell-cell interactions, possibly hPSC-epi subpopulation crosstalk at the point of TGFβ induction. A model for these possible scenarios is shown in Figure 64.

Figure 64. A possible model for hPSC-epi cell responses to TGFβ, wherein (A) shows mixed WT hPSC-epi, (B) shows hPSC-epi that has developed entirely without BNC1 due to TET treatment from the LPM stage, and (C) shows the possible effect of an acute perturbation of BNC1 expression.

Transcription factors such as Nfatc1 and Mrtfs are associated with epicardial cell migration into the myocardium during EMT (Combs et al., 2011; Trembley, Velasquez,

Sophie McManus – February 2020 183

BNC1 regulates human epicardial heterogeneity and function

de Mesy Bentley, & Small, 2015), as was shown by in vivo deletion studies. As actin imaging revealed significant disruption in cytoskeletal remodelling during TGFβ induction, and actin remodelling is key to mediating cell migration, I moved towards examining a potential BNC1 effect on cell migration (D. D. Tang & Gerlach, 2017). This involved imaging epi-EMT cells across a 48-hour time course, before tracking individual cells’ motility. Individual cells’ migratory distance and velocity were tracked in these assays; there was an overall significant difference in mean distance and velocity between conditions, and all cells in the culture were affected. Again, the global reduction in cell distance and velocity when BNC1 was knocked down (rather than apparent subsets of affected cells) may suggest BNC1high TCF21high subpopulation crosstalk that is disrupted when the BNC1high population is perturbed (Figure 64). The reduced migratory phenotype was more apparent when BNC1 expression was most acutely perturbed, a similar observation to that made for actin filament organisation following TET application at different points during hPSC-epi differentiation. However, hPSC-epi cells that had developed entirely without BNC1 displayed a non-significant trend towards reduced invasion in a 3D co-culture assay.

Differential gene expression analysis revealed that BNC1 knockdown in TET-treated epi- EMT samples relative to no-TET controls reduced expression of genes such as layilin, for which reduced expression has been associated with reduced invasive EMT in renal tubule epithelia (Adachi et al., 2015). Of note, layilin is a transmembrane protein which binds the actin cytoskeletal-membrane linked protein talin, as well as the ubiquitous ECM component hyaluronan (HA), allowing mediation of signals from the ECM to the cytoskeleton. A link between F-actin and layilin has been suggested, as has a role in mediating cell motility (Bono et al., 2001), implying that a reduction in expression of this gene could underlie the reduced motility seen in BNC1 knockdown epi-EMT cells. We saw reduced layilin both in d7 and d1 TET epi-EMT samples relative to controls. Layilin is discussed further in Section 4.3.

I also observed reductions in heparin-binding EGF-like growth factor and heparan sulphate glucosamine 3-O- sulfotransferase 3B1, which have also been associated with reduced invasion and migration in keratinocyte EMT models (Song et al., 2011; Stoll, Rittié, Johnson, & Elder, 2012; Yagi, Yotsumoto, & Miyamoto, 2008; Z. Zhang, Jiang, Wang, & Shi, 2018). There was also a relative reduction in matrix metalloproteinase 9

184 Sophie McManus - February 2020

Chapter 4: Discussion

(MMP9), another gene which is associated with enhanced EMT progression in cancer cells (Bai et al., 2017; Gilles et al., 2013). Taken together, these genes present examples of possible mechanisms for BNC1 mediated epi-EMT phenotypes which could be investigated further (see Section 4.3).

Comparing differential gene expression between d1 TET vs d7 TET populations to try to understand why an acute BNC1 knockdown mediates a more pronounced phenotype, I saw differential ADAM and ADAMTS gene expression; members of this family have been linked to regulation of cell migration via ECM remodelling, primarily in models of metastasis (Kelwick et al., 2015). There was reduced expression of ADAMTS1, and ADAMTS15 in the d1 TET condition relative to d7 TET; ADAMTS15 has been linked to negative regulation of cell migration (Kelwick et al., 2015; Wagstaff et al., 2010), while ADAMTS1 has been implicated both in negative and positive regulation of cell migration and invasion (Krampert et al., 2005). Freitas and colleagues showed that ADAMTS1 knockdown stimulated migration, invasion and invadopodia formation in breast cancer cells in vitro (Freitas et al., 2013).

Overall, this differential gene expression analysis indicates that potential targets of interest in the epi-EMT context could include ECM remodelling genes, such as MMPs and ADAMTs. These could be explored via matrix degradation assays with BNC1 knockdown epi-EMT samples, to compare relative matrix degradation and invasion when BNC1 is reduced. There was also a reduction in gene expression pertaining to MAPK cascade and p38 signalling GO term enrichment in d7 TET epi-EMT samples, and these signalling mechanisms could be explored via MAPK activation assays. Caveats to the bulk RNA sequencing approach are addressed in Section 4.2 and potential future work is discussed in Section 4.3.

4.1.3 Identifying putative BNC1 targets

BNC1 is a transcription factor that can interact with RNA polymerase I (Pol I) and RNA polymerase II (Pol II) via three pairs of highly conserved zinc fingers (Iuchi & Green, 1997; Tseng & Green, 1992; Junwen Wang et al., 2006). Using DNase footprinting, Wang and colleagues previously built a computational model for a basonuclin DNA-binding

Sophie McManus – February 2020 185

BNC1 regulates human epicardial heterogeneity and function

module, which was used to identify possible in silico RNA polymerase II target genes in human and mouse promoter databases. This work took previously Bnc1 DNAse I footprinted regions, comparing these protected sequences to mouse and human genomes to identify possible Bnc1 target genes. 21 promoter sequences were suggested by this model, and 11 were validated by ChIPs in a human keratinocyte cell line (HaCaT). Authors suggested that the 52% validation rate for their ChIP could be due to BNC1 partial promoter occupancy, from a large dissociation coefficient, or cell heterogeneity in the HaCaT cell line. I aimed to investigate BNC1 targets via ChIP-seq to identify DNA- protein interactions in human epicardial cells, as a high-throughput method to identify potential downstream mechanisms for the phenotypes observed in hPSC-epi and epi- EMT.

Although ChIP-seq identified just 16 peaks for BNC1 in our hPSC-epi, a promising target was the actin-binding protein Supervillin (SVIL). SVIL binds to F-actin, myosin II and other signalling proteins in lipid rafts (Z. Fang et al., 2010; Khurana & George, 2008; Pestonjamasp et al., 1997; Takizawa et al., 2007). It increases signalling from epidermal growth factor receptor (EGFR) to extracellular signal-regulated kinases (ERKs). Furthermore, this protein is known to mediate EMT (X. Chen et al., 2018). Initial knockdown experimental results were promising, hence this possible BNC1-supervillin interaction could be explored further in epi-EMT models (see Section 4.3).

4.2 Advantages and limitations of our models and experiments

4.2.1 The hPSC-epi model

While animal models have taught us a great deal about epicardial function, there are several clear advantages associated with using a stem-cell derived model of human epicardium; in particular, working to mirror human biology as closely as is currently possible. Although many inferences can be made from animal models with respect to human biology, testing these in the human system is optimal. Furthermore, the option of rapidly generating large numbers of quality-controlled cells without relying on an external source overrides caveats associated with generating primary human foetal epicardial cultures. Moreover, our stem cell-derived hPSC-epi model is highly versatile,

186 Sophie McManus - February 2020

Chapter 4: Discussion

as shown by the successful generation of inducible knockdown cell lines, allowing investigation of my project aims via manipulation of BNC1 expression.

Aside from its practical advantages, our hPSC-epi model has been well-characterised and is highly comparable to primary human foetal epicardium. hPSC-epi expresses epicardial genes, exhibits cobblestone morphology, undergoes EMT to form coSMC and CF, homes in ovo to the subepicardial space, and enhances EHT function and hPSC-epi/CM engraftment in post-MI hearts (Bargehr et al., 2019; Iyer et al., 2015). Nonetheless, validation of hPSC-epi results is key to ensure our observations are not confined to a stem cell line, but are also valid in primary human tissue. To this end, I validated observations in primary human epicardial cells where possible, for example, in performing immunocytochemistry demonstrating heterogeneous BNC1 expression in different human foetal epicardial explants, and siRNA-mediated BNC1 knockdown to examine relative TCF21 increase. It is worth noting that in encouraging recent work, human foetal epicardial cells were also used for single-cell RNA sequencing (Ross, Knight-Schrijver et al, unpublished data). This scRNA-seq in primary human foetal epicardium has also indicated heterogeneous BNC1 expression, complementing previous hf-epi immunocytochemistry, and further demonstrating similarities between our observations in hPSC epi and primary human foetal epicardium via immunostaining (Section 3.1.4, Figure 14) and RNAscope (Appendix, Figure S2).

Our initial scRNA-seq data were a result of sequencing one hPSC-epi time point, d9 of hPSC-epi culture, and were informative in identifying discrete clustered hPSC-epi subpopulations. However, cells and their behaviour are highly dynamic, so clustering analyses have their limitations. For example, cells harvested from the same culture at the same time may exist in a range of different ‘cell states’. In that case, apparent high variation in expression of a particular gene between different cells may be due to differences in those cells’ position along a pathway to transition state. Therefore, established methods for single-cell analysis can be complemented with techniques including single-cell pseudotime analysis (Haghverdi, Büttner, Wolf, Buettner, & Theis, 2016). Cells are thereby ordered along a probabilistic trajectory, and a value is assigned that represents that cell’s progression along the virtual timeline, also indicating the genes

Sophie McManus – February 2020 187

BNC1 regulates human epicardial heterogeneity and function

that are regulated at that particular progression point (Reid & Wernisch, 2016; Trapnell, 2015) (H. Chen et al., 2019).

Our group has now performed further scRNA-seq for hPSC-epi during differentiation, at d1, d2, d3, d4, d8 and d9 in culture. Preliminary pseudotime trajectory analysis has shown that the two hPSC-epi subpopulations previously characterised at d9 are well-defined by d3 of differentiation; prior to d3 there is relatively homogeneous gene expression. This analysis has revealed high relative BNC1 and PODXL expression in the same cells along the pseudotime trajectory, indicating that the BNC1high population can be accurately tracked by PODXL expression throughout hPSC-epi differentiation, complementing our original finding in d9 hPSC-epi. Furthermore, in the absence of BNC1, the PODXL subpopulation was present but not maintained after d5 of differentiation (Dr Laure Gambardella, Dr Vincent Knight-Schrijver, unpublished observations).

4.2.2 The epi-EMT model

Considering BNC1’s role in models of corneal wound repair, I was interested in examining its role in epicardial EMT. I observed reduction in migration distance and velocity in a 2D model of cell motility in epi-EMT. These data, combined with our observations regarding actin organisation in BNC1 knockdown epi-EMT, imply that BNC1 modulates aspects of plasma membrane and cytoskeletal dynamics. I also looked to hf-epi cells, as a more physiological system to generate epi-EMT models to investigate BNC1 function, but observed a great degree of variation between samples, particularly pertaining to actin filament organisation, and so these experiments were stopped due to lack of reproducibility. This may be in part due to the degree of variability in embryonic stage for the foetuses we receive; although crown-rump length and neck-rump length are helpful proxies for Carnegie stage, hence human embryo gestational age, Carnegie staging is not entirely accurate. Therefore samples received may represent several different gestational ages, whereas I consistently performed epi-EMT experiments with hPSC-epi at d8 of differentiation. Moreover, the embryos received all possess very different genetic backgrounds. hf-epi cells were not used in cell migration experiments due to the previous variability in phenotypes observed.

188 Sophie McManus - February 2020

Chapter 4: Discussion

When I set up my 2D model of epi-EMT, I also tested the effect of FGF2 and found this had a less pronounced effect on cell morphology and F-actin remodelling than TGFβ, so selected the latter. As expected during EMT, my model demonstrated a pronounced ‘cadherin-switch’, downregulation of epicardial gene expression, cell morphology changes, actin fibre reorganisation, and some upregulation of EMT markers such as periostin and calponin. One caveat of my model was the lack of pronounced SNAI1 and SNAI2 induction upon TGFβ treatment (Appendix, Figure S7): the SNAIL superfamily is known to mediate EMT in multiple models. However, SNAIL-independent epicardial EMT has been reported (Casanova et al., 2013). To clarify exactly how TGFβ affects hPSC-epi gene expression when cells are split into the epi-EMT model, one would have to perform differential expression analyses between hPSC-epi prior to and post-TGFβ treatment.

While 2D migration assays were a useful tool in investigating my hypothesis, in vivo epicardial EMT is a highly complex biological phenomenon, involving cell morphology changes, migration, and invasion into developing myocardium. To address the limited parallels to this in my 2D cultures, I implemented a 3D invasion assay, wherein fluorescently labelled hPSC-epi cells were seeded on top of fluorescently labelled CM to monitor hPSC-epi cell invasion into CM, providing a more physiological model. While we did not observe a statistically significant reduction in hPSC-epi migration into CM, there was a promising trend towards a greater average distance between each cell layer, suggesting a possible role for BNC1 in this context. This trend, coupled with the noise between experiments, indicates that invasion experiments should be pursued further (see Section 4.3).

The bulk RNA sequencing of epi-EMT samples revealed GO enrichment for terms related to metabolism in the d7 TET epi-EMT samples relative to no TET. However, this may be due to tetracycline-mediated effects on metabolic gene expression (Ahler et al., 2013). In future, a control tetracycline inducible cell line, such as my B2M inducible knockdown line, should be used as a further control for these experiments, as well as in other cell assays; this approach would ensure that the phenotypes observed are entirely due to BNC1 knockdown rather than any non-specific antibiotic effects.

Sophie McManus – February 2020 189

BNC1 regulates human epicardial heterogeneity and function

4.2.3 ChIP sequencing experiments

During my PhD, I aimed to identify putative epicardial BNC1 targets, hence selected ChIP-seq as a powerful and unbiased method to detect DNA-protein interactions in the cellular context, with subsequent corroboration of potential targets in ChIP-qPCR experiments. I verified the ChIP conditions and protocol via shear tests and by using a CTCF control antibody and assessing enrichment for selected positive and negative control sequences, which proved successful. ChIP-seq with our BNC1 antibody identified 16 possible peaks in hPSC-epi cells after data clean-up steps; these were genes related to processes such as transcription factor binding and cell migration. I had hypothesised that BNC1 may directly bind WT1 and TCF21; however, I saw no peaks arising in any epicardial transcription factor genomic regions.

Although ChIP represents a thorough technique for identifying potential gene targets, this approach can involve a high degree of variability (Gade & Kalvakolanu, 2012), for example, due to dependence on a highly specific antibody, as well as sufficient expression of the protein of interest in the selected cell type. The low number of targets identified for our ChIP may therefore be due to relatively low BNC1 level (as this is a transcription factor), or simply poor binding of our BNC1 antibody. Although this antibody is polyclonal, as is recommended in ChIP experiments, and has been successfully used in many other applications, such as immunocytochemistry and Western Blot, it may have failed to adequately bind in my ChIPs.

As a positive control for my ChIP, I also ran ChIP-seq with an antibody against TCF21; ChIP-seq with this antibody has previously yielded over 4000 peaks (Sazonova et al., 2015). However, in our hPSC-epi ChIP-seq experiment, the TCF21 ChIP-seq only resulted in pull-down of 71 peaks after data clean-up. Furthermore, the peaks identified in the publication record did not correspond to those I found in hPSC-epi. This latter point may be because Sazonova and colleagues assessed a different cell type, and at a different differentiation time point.

SVIL was a potential BNC1 target of interest identified by my ChIP experiments. I observed a similar cortical actin localisation in preliminary SVIL knockdown experiments to that seen via BNC1 knockdown. Of note, I did not observe significantly altered SVIL expression by qPCR in BNC1 knockdown compared to wild-type (WT) epi-EMT cells;

190 Sophie McManus - February 2020

Chapter 4: Discussion

nor was SVIL identified as a significantly differentially expressed gene in bulk RNA sequencing for epi-EMT samples at d1 of TGFβ treatment. It seems possible, however, that BNC1-supervillin interactions and functional roles may occur at the onset of TGFb treatment, over a shorter time frame. For example, Bhuwania and colleagues imaged GFP-tagged supervillin via video microscopy and saw acute SVIL recruitment to podosomes, which had a lifetime of just 2-12 minutes (Bhuwania et al., 2012). The time point selected for my bulk RNA sequencing for epi-EMT may have been inadequate to capture any differential expression or binding of supervillin. Overall, future experiments to investigate potential BNC1-supervillin roles in epi-EMT shall likely prove informative (see Section 4.3).

4.3 Future work to investigate BNC1 function in cell-based assays

4.3.1 Sequencing BNC1 knockdown hPSC-epi: network refinement?

When BNC1 is knocked down from the LPM stage, we observe formation of a population of TCF21high THY1high cells that appear fibroblastic, lacking characteristic epicardial ‘cobblestone’ morphology that we see in mixed WT hPSC-epi. Our hPSC-epi network inference analysis predicted 100 transcription factors that were likely to interact with BNC1, WT1 and TCF21, providing further targets that could be investigated in the context of hPSC-epi regulation and function. Sequencing BNC1 siKD cells could refine our network, by defining the transcriptomic effects of BNC1 removal. Single-cell RNA sequencing would also allow us to define the transcriptional basis for the phenotypes observed in BNC1 siKD hPSC-epi; for example, are BNC1 KD cells similar to the TCF21high population previously characterised by scRNA-seq, or do they have a gene signature that significantly differ from hPSC-epi cells altogether? It is interesting to note that although sorted TCF21high subpopulations from WT hPSC-epi could form both coSMC and CF, our BNC1 knockdown hPSC-epi displayed a significantly reduced propensity to form surviving coSMC in culture, despite its increased TCF21 expression. This finding, combined with the more fibroblastic morphology ascribed to siKD hPSC- epi, implies that the BNC1 knockdown cells differ from TCF21high hPSC-epi cells, at least in this context. It would be interesting to compare siKD hPSC-epi expression profiles to our existing data for hPSC-epi derived CF, as well as CF derived from human foetal

Sophie McManus – February 2020 191

BNC1 regulates human epicardial heterogeneity and function

epicardial explants, to explore if there is a particular BNC1 KD ‘fibroblastic’ gene signature. We could also knock down BNC1 after the two subpopulations have been established (rather than at the LPM stage), to investigate whether continued BNC1 expression is essential to the maintenance of two populations. In such an experiment, the TCF21high cells may have a differential capacity to form coSMC than the LPM-stage BNC1 knockdown cells, as they had previously expressed BNC1 during the differentiation. Such questions may be best answered via use of Cre-based lineage tracing cell lines.

4.3.2 BNC1 overexpression assays

Conversely, as we see that BNC1high cells from our mixed hPSC-epi have a higher propensity to form coSMC than CF, would increasing BNC1 expression via an overexpression vector further enrich this relative coSMC formation, in comparison to coSMC formation from mixed hPSC-epi cells? Bnc1 overexpression studies were carried out previously in mammary epithelial cells (Feuerborn et al., 2015), and caused enhanced TGFβ-induced epithelial sheet dissolution; however, such studies are inevitably artificial and lacking physiological control. Hence another potentially promising avenue could be generation of a TCF21 inducible knockdown line, to see if perturbing TCF21 in fact enriches for BNC1 expression, given its own prominent position in our network inference model. If TCF21 knockdown did cause enrichment of BNC1, the inducible TCF21 knockdown cells’ differentiation propensities and migratory properties could also be tested, the former in our coSMC and CF differentiation protocols, the latter in epi-EMT models.

4.3.3 Investigating BNC1 knockdown kinetics

The kinetics of BNC1 knockdown and expression could also be explored to understand how plastic the hPSC-epi cell subpopulations are: if tetracycline is removed from the epicardial media, would BNC1 switch back on? From a practical standpoint, identifying PODXL as a faithful membrane marker for the BNC1high subpopulation is a great advantage to further work in our system, as it allows rapid ‘snapshots’ of relative hPSC- epi PODXL-positive cells (likely to also be BNC1-positive) by flow cytometry. It also raises questions such as whether BNC1 expression induces upregulation of PODXL;

192 Sophie McManus - February 2020

Chapter 4: Discussion

reducing BNC1 does appear to decrease the number of PODXL-positive cells, but perhaps BNC1 is required for maintenance of PODXL expression, rather than its induction.

4.3.4 Manipulating hPSC-epi populations to ‘enhance’ hPSC-epi effects?

Bargehr et al recently reported that addition of hPSC-epi to CM grafts enhanced engraftment and neovascularisation in rat infarct models (Bargehr et al., 2019). This work used heterogeneous hPSC-epi, with both subpopulations present. Our studies into hPSC- epi subpopulation function have implied differential functional properties for each subpopulation (Gambardella et al., 2019). Now that we also see that BNC1 may mediate events during EMT, these data could inform strategies to ‘enhance’ our hPSC-epi for hPSC-epi/CM cell suspension injection or epi/CM EHT grafts. For example, PODXL and THY1 are membrane markers for the BNC1high and the TCF21high populations respectively, so these could be used to sort the subpopulations and allow investigation of their relative effects on EHTs. The infarct environment has been shown to inhibit the hPSC-epi differentiation path to coSMC formation: hPSC-epi delivered to the chorionic vasculature of chick embryos resulted in cell integration into vessel walls and smooth muscle cell marker expression, whereas in the adult infarct zone the hPSC-epi cells formed fibroblast-like cells in place of any coSMC (Bargehr et al., 2019). Nonetheless, a study by Ramjee and colleagues revealed that epicardial Hippo signalling can suppress post-infarct inflammatory response in and around the heart, hence beneficially condition the infarct zone (Ramjee et al., 2017). Overall, ‘enhancing’ hPSC-epi in culture with the aim of further improving engraftment and EHT function may be more readily achieved than conditioning the hostile in vivo injury environment. If one hPSC-epi subpopulation is more effective in mediating compaction and maturation of EHTs, this could be enriched for in future assays, via cell sorting. Foetal epicardial cells have a higher probability of undergoing EMT than adult epicardial cells (Moerkamp et al., 2016); it remains to be seen whether our hPSC-epi subpopulations also exhibit differing propensities for EMT.

The regulatory network developed by our network inference analysis has also yielded a list of promising molecular pathways, which could potentially be manipulated by small molecules to refine biological properties of hPSC-epi, or potentially even endogenous

Sophie McManus – February 2020 193

BNC1 regulates human epicardial heterogeneity and function

epicardium in future. Work to examine potential hPSC-epi subpopulation paracrine crosstalk could effectively complement ongoing engraftment studies. For example, it is thought that epicardial cells introduced into the mammalian adult injured heart exert beneficial effects via their secretome (Bargehr et al., 2019; E M Winter et al., 2007; Elizabeth M. Winter et al., 2009), for example, via fibronectin secretion; the epicardial subpopulations identified in our hPSC-epi could exhibit heterogeneity in their secretome. If so, potentially this could be modulated. hPSC-epi paracrine effects have not been extensively investigated to date. DEseq between hPSC-epi, hPSCepi derived CF and SMC, and primary smooth muscle cell sources identified nephronectin (NPNT) as the most overexpressed diffusible factor in hPSC-epi (Bargehr et al, 2019). NPNT was also found to be expressed in BNC1high cells via our scRNA-seq. NPNT is the functional ligand of Integrin alpha-8 beta-1, which promotes cell spreading, and this was relatively highly expressed in the TCF21high population, one suggestion of possible cross-talk between the two hPSC-epi populations (Gambardella et al., 2019; Müller, Bossy, Venstrom, & Reichardt, 1995; Schnapp et al., 1995). It seems conceivable that BNC1 knockdown hPSC-epi may exhibit altered paracrine signalling; I speculate that this is a mechanism behind the global reduction in cell migration observed in BNC1 knockdown epi-EMT. Our group is working to characterise hPSC-epi-cardiomyocyte crosstalk (Dr Vincent Knight-Schrijver, ongoing work); comparing hPSC-epi subpopulation paracrine effects may prove highly relevant in this context.

4.3.5 Quantifying BNC1 function in epi-EMT models

While the epi-EMT models I developed during my PhD have provided insights into BNC1 functions in epicardial migration, there is considerable scope to refine these assays. Dr Angela Russell (Oxford University) has developed software for high-throughput scoring of the morphologies of epicardial cells undergoing EMT, which could be used to assess and quantify the relative propensity of siKD epicardial cells compared to controls in losing their epithelial phenotype, potentially via time-lapse imaging. This would provide an additional quantitative measure for cell phenotypes in epi-EMT, and could be combined with my existing 2D migration tracking assay using the InCell microscope. Instead of a binary ‘switch’ from a purely epithelial to mesenchymal cell fate, EMT can be considered as a continuum of cell states: a partial EMT state has been noted in

194 Sophie McManus - February 2020

Chapter 4: Discussion

association with many developmental events (Nieto, Huang, Jackson, & Thiery, 2016). Some of the epi-EMT cells could be partially undergoing EMT; this could potentially be established via Dr Russell’s methodology, and related to relative BNC1 knockdown. My hPSC-epi-CM invasion assay could also be further explored and improved. For example, time-lapse imaging of hPSC-epi cells as they invade the CM layer would provide finer resolution of each cell’s invasive path. Comparisons could also be made between the relative invasion of acute BNC1 knockdown hPSC-epi cells and hPSC-epi that had differentiated entirely without BNC1, to verify if there is the same difference in phenotypic severity as seen in 2D epi-EMT assays.

Acute BNC1 knockdown mediates a distinctive actin phenotype in TGFβ-induced epi- EMT. Previous observations of actin filaments in hPSC-cells over-expressing TCF21 (Dr Laure Gambardella, unpublished observations) also show some degree of cortical actin localisation. Furthermore, Sazonova and colleagues showed GO-term enrichment in TCF21-expressing cells for terms related to actin cytoskeleton and actin filament organisation (Sazonova et al., 2015). It seems possible that the actin phenotype in our BNC1 knockdown cells could be (at least in part) mediated by differential gene expression caused by acute loss of BNC1 in the cells prior to TGFβ treatment, combined with an effect of an increase in TCF21 also prior to TGFβ treatment. As our network suggested such strong potential transcriptional influences from both BNC1 and TCF21, it seems possible that they might act in parallel.

4.3.6 Exploring possible BNC1 targets

Regarding the observed effects of BNC1 knockdown in epi-EMT models, our bulk RNA sequencing identified differential expressed genes pertaining to cell migration, ECM remodelling and cytoskeletal organisation. These might indicate mechanisms by which BNC1 knockdown mediates actin and migratory phenotypes, and present targets of interest for functional validation experiments, as were undertaken in a preliminary fashion for supervillin with siRNA knockdowns in epi-EMT cells. As ECM modelling genes such as MMPs and ADAMTS were identified as potential mechanisms of interest, future experiments could utilise ECM degradation assays, for example, fluorescent gelatin degradation assays, to compare relative effects on ECM from control and siKD cells (P. Lu, Takai, Weaver, & Werb, 2011) (Díaz, 2013).

Sophie McManus – February 2020 195

BNC1 regulates human epicardial heterogeneity and function

Furthermore, genes that were found to be differentially expressed in the BNC1 knockdown epi-EMT models could typically mediate migration of WT hPSC-epi (for example layilin, or MMPs). Such genes could be investigated in primary epicardium, in particular, after cardiac injury, to establish whether there is any upregulation alongside epicardial reactivation. Candidates could also be explored via gain- or loss-of-function studies. Of particular interest, we saw that layilin was downregulated in BNC1 knockdown epi-EMT samples: layilin is a hyaluranon receptor, and HA is important in the context of tissue repair, and cardiac regeneration (Missinato et al., 2015). Indeed, HA hydrogels have proven to increase angiogenesis and improve post-injury cardiac grafting (Bonafè et al., 2014). I saw differential reduction of layilin expression depending on when BNC1 knockdown was performed in epi-EMT samples. Interestingly, our scRNA-seq data revealed that layilin is typically expressed in the BNC1high subpopulation rather than the TCF21high cells, providing further suggestion of a possible link between the two genes. I hypothesise that BNC1 knockdown-mediated reduction of layilin, in combination with altered expression of other extracellular matrix proteins, is key to reduced epi-EMT migration. It seems possible that layilin expression might aid HA-mediated repair post- injury; this could be investigated in vivo, in models of cardiac injury and regeneration.

While my ChIP returned a low number of peaks, some of these appeared to have potential relevance in our system, considering the published evidence. I decided to perform preliminary studies into SVIL function due to its well-characterised roles in cytoskeletal component binding and organisation. SVIL binds F-actin and myosin II domains (Y. Chen et al., 2003) and localises to podosomes, cellular structures comprising an F-actin core encompassed by a ring of plaque proteins (Bhuwania et al., 2012). Podosomes are highly dynamic, with lifetimes of just a few minutes, and aid cell adhesion and migration via contact with ECM. Multiple reports have indicated that SVIL contributes to a rapid cell motility response, and both cell migratory path length and velocity were reduced in SVIL knockdown in previous studies (Crowley et al., 2009; Z. Fang et al., 2010). SVIL expression was also directly associated with extracellular matrix degradation in migration studies via MMP secretion (Crowley et al., 2009). In future work in an epi-EMT context, relative SVIL expression in different BNC1 knockdown conditions should also be characterised in the first 24 hours of EMT induction (X. Chen et al., 2018). Bulk RNA sequencing did not indicate any differential SVIL expression between controls and BNC1

196 Sophie McManus - February 2020

Chapter 4: Discussion

knockdown epi-EMT 24 hours after TGFβ treatment, but SVIL expression was not investigated prior to this time point.

An alternative to ChIP, given the challenges encountered with the experiments to date, could involve a luciferase reporter assay to establish if BNC1 can activate or repress the SVIL promoter. Furthermore, fluorescently tagged SVIL and F-actin bundles could be imaged at a high spatiotemporal resolution to compare remodelling events in BNC1 knockdown compared to control hPSC-epi cells. This, combined with a higher-resolution invasion assay, might allow visualisation of how acutely perturbing BNC1 expression mediates such a dramatic actin phenotype. The mechanism for SVIL involvement in EMT mediation is thought to involve RhoA/ROCK pathway activation of ERK/p38 (X. Chen et al., 2018). Similarly, RhoA signalling has been linked to coSMC differentiation and actin stress fibre reorganisation in EMT (J. Lu et al., 2001). RhoA inhibition and activation could be explored in the epi-EMT model as TGFβ induction occurs, to investigate whether differential regulation of this pathway mediates BNC1 knockdown phenotypes. Overall, ChIP experiments should be optimised and potentially carried out both in epicardial cells and in epi-EMT cells at the point of TGFβ induction.

4.4 Future work: a focus on BNC1 in vivo While cell assays are highly informative, in vivo analysis will be key to verify if the observed in vitro phenotypes have physiological significance. I had aimed to investigate Bnc1 function in vivo, but there was a delay in receiving Bnc1 knockout mice (MRC Harwell). The knockout animals have now arrived and will be the focus of future work. We expected Bnc1 null mice to have reduced viability, given evidence from the publication record (X. Zhang & Tseng, 2007). Given my observations in hPSC-epi differentiation and epi-EMT models, we hypothesise that Bnc1 knockout mice may exhibit altered epicardial marker expression and display signs of impaired EMT, hence potentially an impairment in formation of the coronary vasculature and interstitial CF. Moreover, following our differential expression analysis in BNC1 knockdown epi-EMT samples, expression of candidate genes such as layilin could also be characterised in these mice.

The work on Bnc1 in vivo will involve harvesting WT and KO embryos from E11.5 to term and characterising Bnc1 expression via RNAscope, as this technique has recently

Sophie McManus – February 2020 197

BNC1 regulates human epicardial heterogeneity and function

proven successful for BNC1 detection in human foetal heart sections (Appendix, Figure S2), and we do not have a BNC1 antibody that is effective in immunohistochemistry. Of note so far, some knockout and heterozygote embryos have been observed to have abnormal vasculature at E13.5; furthermore, observed Mendelian ratios in litters born are slightly abnormal, with fewer Bnc1 KO pups born to date relative to WT and heterozygotes (Dr Laure Gambardella, personal communication). Future work will entail characterisation of hearts from control and Bnc1 KO mice at different stages of embryonic development, comparing their morphology, and epicardial gene and protein expression, as we might expect to see altered Wt1 and Tcf21 levels in Bnc1 KO hearts. Coronary vasculature formation can also be examined in whole-mount immunohistochemistry.

We have considered the possibility of possible genetic compensation by Basonuclin 2 (Bnc2) in our Bnc1 knockout mice, and Bnc2 expression can be verified in the knockout animals, however, literature evidences strongly suggests disparate functions for these two genes. Bnc2 shares Bnc1’s three zinc fingers and nuclear localisation signal; amino acid sequences for Bnc1 and Bnc2 are approximately 40% similar, indicating a common origin. However, Bnc2 exhibits different nuclear localisation to Bnc1, localising to nuclear speckles, and Bnc2’s functions are thought to be unrelated to those of Bnc1 (Vanhoutteghem & Djian, 2006). Bnc2 KO exhibit neonatal lethality due to cleft palates and craniofacial abnormalities, which were not rescued by Bnc1 knock-in (Vanhoutteghem et al., 2016). Moreover, we saw no heterogeneity in expression for BNC2 expression in our hPSC-epi, nor differential expression for BNC2 in our d7 vs no TET epi-EMT bulk RNA or d7 TET vs d1 TET epi-EMT RNA sequencing data. There was a 1.49-fold increase for BNC2 in the d1 TET epi-EMT population relative to no TET, indicating a possible slight tendency towards compensation.

Access to Bnc1 KO mice also offers the prospect of exploring Bnc1 function in a disease context, to complement developmental studies. Bochmann and colleagues showed that Bnc1 is enriched in adult mouse epicardium, but that its expression significantly drops by d3 post-infarct. In contrast, we know that in zebrafish larvae, both bnc1 and bnc2 are expressed, but after cardiac injury, bnc1 expression (and not bnc2) increases significantly by 3 days post-injury (dpi) in the wt1b-positive cluster of epicardial cells of the

198 Sophie McManus - February 2020

Chapter 4: Discussion

regenerating fish heart (unpublished communication from Weinberger et al), suggesting that upregulation of bnc genes could be involved in the acute repair and regenerative response in fish. Furthermore, recent data collected by Wang and colleagues reveal that Bnc1 expression increases post-MI in the neonatal mouse heart compared to sham. Bnc1 upregulation at 1.5 dpi occurred if the injury was performed at P1, but not if cardiac injury was performed outside the neonatal regenerative window (P8) (Zhaoning Wang et al., 2019) (Wang et al’s sequencing data are deposited in Gene Expression Omnibus; Bnc1 data were investigated by Dr Laure Gambardella).

In the adult mouse, cardiac injury results in epicardial reactivation, for example, WT1 re- expression, however the benefits of this are constrained (Groot, Winter, & Poelmann, 2010; Zhou et al., 2011; Zhou & Pu, 2011). Inadequate epicardial reactivation is associated with the poor repair seen in adult mammalian hearts (Zhou et al., 2011); although, in this study, some epicardial cells underwent EMT, these EPDCs appeared to be localised to the subepicardium. We have seen that BNC1 knockdown impairs certain EMT hallmarks such as actin remodelling and cell migration. Combining observations made in fish and mouse regarding both their opposing cardiac injury responses and post- infarct epicardial BNC1 expression leads us to speculate that BNC1 may be a positive regulator of epicardial function in the cardiac post-injury setting, particularly given Wang et al’s exciting data in the neonatal mouse MI model.

A powerful investigation could therefore entail use of the neonatal mouse cardiac injury model as follows: typically, we would expect sustained cardiac regeneration in neonatal mice, while Bnc1 KO mice may exhibit impaired epicardial activation and an altered regenerative response. The corneal epithelium developed normally in Bnc1-null mice, but the healing response was impaired after wound induction. Hence I speculate that the heart may require a similar injury challenge to expose any potential role of Bnc1 in epicardial functionality. Conditional Bnc1 knockouts could be generated to allow spatiotemporal control of BNC1 deletion. Furthermore, an epicardial Bnc1 reporter mouse could be a useful tool in visualising migration of Bnc1-positive epicardial cells into developing myocardium. Bnc1 could also be overexpressed after injury in the adult mouse epicardium, to see if this might enhance any reactivation response.

Sophie McManus – February 2020 199

BNC1 regulates human epicardial heterogeneity and function

4.5 Conclusion We know that the epicardium is a crucial lineage to cardiac development, as well as a key player in the cardiac regenerative response in selected organisms. Dissecting epicardial developmental mechanisms with the aim of recapitulating these in the injury setting therefore offers a great deal of promise in the context of regenerative medicine. While the epicardium was considered a homogeneous tissue, there is now a growing appreciation for potential functional relevance of epicardial heterogeneity (Gambardella et al., 2019; Weinberger et al., 2018). The main goal of this thesis was to investigate BNC1 expression and function in human epicardial cells.

Given the data presented here, I conclude that BNC1 appears to have diverse context- dependent roles in various epicardial models, in agreement with reports in other epithelia. BNC1 regulates human epicardial heterogeneity and function, including maintenance of other epicardial genes’ expression, actin filament remodelling, and cell migration. These data lead us to conclude that BNC1 can be classed as a novel epicardial functional regulator in human models of developing epicardium.

Our hPSC-epi has been shown to yield significant benefits during cardiac tissue engraftment post-injury (Bargehr et al., 2019). Characterising functional implications of epicardial heterogeneity could enable enhancement of stem cell-derived therapeutic strategies, via enrichment for optimal hPSC-epi subpopulations and their downstream effects in the injury setting. Moreover, investigating Bnc1 function in vivo will establish whether this gene plays any part in the cardiac regenerative response seen in the neonatal mouse. Insights from developmental mechanisms are highly relevant to work in a disease context. Overall, continual sustained progress in the epicardial research field could provide substantial benefit to patients in future.

200 Sophie McManus - February 2020

Chapter 5: References

5. REFERENCES

Acharya, A., Baek, S. T., Huang, G., Eskiocak, B., Goetsch, S., Sung, C. Y., … Tallquist, M. D. (2012). The bHLH transcription factor Tcf21 is required for lineage-specific EMT of cardiac fibroblast progenitors. Development (Cambridge, England), 139(12), 2139–2149. https://doi.org/10.1242/dev.079970

Adachi, T., Arito, M., Suematsu, N., Kamijo-Ikemori, A., Omoteyama, K., Sato, T., … Kato, T. (2015). Roles of layilin in TNF-α-induced epithelial-mesenchymal transformation of renal tubular epithelial cells. Biochemical and Biophysical Research Communications, 467(1), 63–69. https://doi.org/10.1016/j.bbrc.2015.09.121

Aguirre, A., Sancho-Martinez, I., & Izpisua Belmonte, J. C. (2013, March 7). Reprogramming toward heart regeneration: Stem cells and beyond. Cell Stem Cell, Vol. 12, pp. 275–284. https://doi.org/10.1016/j.stem.2013.02.008

Ahler, E., Sullivan, W. J., Cass, A., Braas, D., York, A. G., Bensinger, S. J., … Christofk, H. R. (2013). Doxycycline Alters Metabolism and Proliferation of Human Cell Lines. PLoS ONE, 8(5). https://doi.org/10.1371/journal.pone.0064561

Alexander, M. S., Shi, X., Voelker, K. A., Grange, R. W., Garcia, J. A., Hammer, R. E., & Garry, D. J. (2010). Foxj3 transcriptionally activates Mef2c and regulates adult skeletal muscle fiber type identity. Developmental Biology, 337(2), 396–404. https://doi.org/10.1016/j.ydbio.2009.11.015

Alsan, B. H., & Schultheiss, T. M. (2002). Regulation of avian cardiogenesis by Fgf8 signaling. Development (Cambridge, England).

Andrée, B., Duprez, D., Vorbusch, B., Arnold, H. H., & Brand, T. (1998). BMP-2 induces ectopic expression of cardiac lineage markers and interferes with somite formation in chicken embryos. Mechanisms of Development, 70(1–2), 119–131. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/9510029

Sophie McManus – February 2020 201

BNC1 regulates human epicardial heterogeneity and function

Andrés-Delgado, L., Ernst, A., Galardi-Castilla, M., Bazaga, D., Peralta, M., Münch, J., … Mercader, N. (2019). Actin dynamics and the Bmp pathway drive apical extrusion of proepicardial cells. Development. https://doi.org/10.1242/dev.174961

Assmus, B., Honold, J., Schächinger, V., Britten, M. B., Fischer-Rasokat, U., Lehmann, R., … Zeiher, A. M. (2006). Transcoronary transplantation of progenitor cells after myocardial infarction. The New England Journal of Medicine, 355(12), 1222–1232. https://doi.org/10.1056/NEJMoa051779

Auda-Boucher, G., Bernard, B., Fontaine-Pérus, J., Rouaud, T., Mericksay, M., & Gardahaut, M.-F. (2000). Staging of the Commitment of Murine Cardiac Cell Progenitors. Developmental Biology, 225(1), 214–225. https://doi.org/10.1006/DBIO.2000.9817

Austin, A. F., Compton, L. A., Love, J. D., Brown, C. B., & Barnett, J. V. (2008). Primary and immortalized mouse epicardial cells undergo differentiation in response to TGFβ. Developmental Dynamics, 237(2), 366–376. https://doi.org/10.1002/dvdy.21421

Avery, A. W., Fealey, M. E., Wang, F., Orlova, A., Thompson, A. R., Thomas, D. D., … Egelman, E. H. (2017). Structural basis for high-affinity actin binding revealed by a β-III-spectrin SCA5 missense mutation. Nature Communications, 8(1). https://doi.org/10.1038/s41467-017-01367-w

Avery, A. W., Thomas, D. D., & Hays, T. S. (2017). β-III-spectrin spinocerebellar ataxia type 5 mutation reveals a dominant cytoskeletal mechanism that underlies dendritic arborization. Proceedings of the National Academy of Sciences of the United States of America, 114(44), E9376–E9385. https://doi.org/10.1073/pnas.1707108114

Bai, X., Li, Y. Y., Zhang, H. Y., Wang, F., He, H. L., Yao, J. C., … Li, S. S. (2017). Role of matrix metalloproteinase-9 in transforming growth factor-β1-induced epithelial— mesenchymal transition in esophageal squamous cell carcinoma. OncoTargets and Therapy, 10, 2837–2847. https://doi.org/10.2147/OTT.S134813

Bakin, A. V, Rinehart, C., Tomlinson, A. K., & Arteaga, C. L. (2002). p38 mitogen- activated protein kinase is required for TGFbeta-mediated fibroblastic transdifferentiation and cell migration. Journal of Cell Science, 115(Pt 15), 3193–

202 Sophie McManus - February 2020

Chapter 5: References

3206. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/12118074

Bao, X., Lian, X., Qian, T., Bhute, V. J., Han, T., & Palecek, S. P. (2017). Directed differentiation and long-term maintenance of epicardial cells derived from human pluripotent stem cells under fully defined conditions. Nature Protocols, 12(9), 1890– 1900. https://doi.org/10.1038/nprot.2017.080

Bargehr, J., Ong, L. P., Colzani, M., Davaapil, H., Hofsteen, P., Bhandari, S., … Sinha, S. (2019). Epicardial cells derived from human embryonic stem cells augment cardiomyocyte-driven heart regeneration. Nature Biotechnology, 37, 895–906. https://doi.org/10.1038/s41587-019-0197-9

Barrallo-Gimeno, A., & Nieto, M. A. (2005, July). The Snail genes as inducers of cell movement and survival: Implications in development and cancer. Development, Vol. 132, pp. 3151–3161. https://doi.org/10.1242/dev.01907

Battaglia, R. A., Delic, S., Herrmann, H., & Snider, N. T. (2018). Vimentin on the move: New developments in cell migration [version 1; referees: 2 approved]. F1000Research, Vol. 7. https://doi.org/10.12688/f1000research.15967.1

Bax, N. A. M., van Oorschot, A. A. M., Maas, S., Braun, J., van Tuyn, J., de Vries, A. A. F., … Goumans, M.-J. (2011). In vitro epithelial-to-mesenchymal transformation in human adult epicardial cells is regulated by TGFβ-signaling and WT1. Basic Research in Cardiology, 106(5), 829–847. https://doi.org/10.1007/s00395-011- 0181-0

Beltrami, A. P., Barlucchi, L., Torella, D., Baker, M., Limana, F., Chimenti, S., … Anversa, P. (2003). Adult cardiac stem cells are multipotent and support myocardial regeneration. Cell, 114(6), 763–776. https://doi.org/10.1016/S0092-8674(03)00687- 1

Bergmann, O., Bhardwaj, R. D., Bernard, S., Zdunek, S., Barnabé-Heide, F., Walsh, S., … Frisén, J. (2009). Evidence for cardiomyocyte renewal in humans. Science, 324(5923), 98–102. https://doi.org/10.1126/science.1164680

Bertero, A., Pawlowski, M., Ortmann, D., Snijders, K., Yiangou, L., Cardoso De Brito, M., … Vallier, L. (2016). Optimized inducible shRNA and CRISPR/Cas9 platforms for in vitro studies of human development using hPSCs. Development.

Sophie McManus – February 2020 203

BNC1 regulates human epicardial heterogeneity and function

https://doi.org/10.1242/dev.138081

Bertero, A., Yiangou, L., Brown, S., Ortmann, D., Pawlowski, M., & Vallier, L. (2018). Conditional Manipulation of Gene Function in Human Cells with Optimized Inducible shRNA. Current Protocols in Stem Cell Biology, 44, 5C.4.1-5C.4.48. https://doi.org/10.1002/cpsc.45

Bhuwania, R., Cornfine, S., Fang, Z., Krüger, M., Luna, E. J., & Linder, S. (2012). Supervillin couples myosin-dependent contractility to podosomes and enables their turnover. Journal of Cell Science, 125(9), 2300–2314. https://doi.org/10.1242/jcs.100032

Bilbija, D., Haugen, F., Sagave, J., Baysa, A., Bastani, N., Levy, F. O., … Valen, G. (2012). Retinoic acid signalling is activated in the postischemic heart and may influence remodelling. PloS One, 7(9), e44740. https://doi.org/10.1371/journal.pone.0044740

Blanchoin, L., Boujemaa-Paterski, R., Sykes, C., & Plastino, J. (2014). Actin dynamics, architecture, and mechanics in cell motility. Physiological Reviews, 94(1), 235–263. https://doi.org/10.1152/physrev.00018.2013

Blom, J. N., & Feng, Q. (2018). Cardiac repair by epicardial EMT: Current targets and a potential role for the primary cilium. Pharmacology & Therapeutics, 186, 114–129. https://doi.org/10.1016/j.pharmthera.2018.01.002

Blondelle, J., Ohno, Y., Gache, V., Guyot, S., Storck, S., Blanchard-Gutton, N., … Pilot- Storck, F. (2015). HACD1, a regulator of membrane composition and fluidity, promotes myoblast fusion and skeletal muscle growth. Journal of Molecular Cell Biology. https://doi.org/10.1093/jmcb/mjv049

Bochmann, L., Sarathchandra, P., Mori, F., Lara-Pezzi, E., Lazzaro, D., & Rosenthal, N. (2010). Revealing New Mouse Epicardial Cell Markers through Transcriptomics. PLoS ONE, 5(6), e11429. https://doi.org/10.1371/journal.pone.0011429

Bollini, S., Riley, P. R., & Smart, N. (2015). Thymosin β4: multiple functions in protection, repair and regeneration of the mammalian heart. Expert Opinion on Biological Therapy, 15 Suppl 1, 163–174. https://doi.org/10.1517/14712598.2015.1022526

204 Sophie McManus - February 2020

Chapter 5: References

Bollini, S., Vieira, J. M. N., Howard, S., Dubè, K. N., Balmer, G. M., Smart, N., & Riley, P. R. (2014). Re-activated adult epicardial progenitor cells are a heterogeneous population molecularly distinct from their embryonic counterparts. Stem Cells and Development, 23(15), 1719–1730. https://doi.org/10.1089/scd.2014.0019

Bonafè, F., Govoni, M., Giordano, E., Caldarera, C. M., Guarnieri, C., & Muscari, C. (2014). Hyaluronan and cardiac regeneration. Journal of Biomedical Science, Vol. 21. https://doi.org/10.1186/s12929-014-0100-4

Bonnet, A. L., Chaussain, C., Broutin, I., Rochefort, G. Y., Schrewe, H., & Gaucher, C. (2018). From vascular smooth muscle cells to folliculogenesis: What about vasorin? Frontiers in Medicine, Vol. 5. https://doi.org/10.3389/fmed.2018.00335

Bono, P., Rubin, K., Higgins, J. M. G., & Hynes, R. O. (2001). Layilin, a novel integral membrane protein, is a hyaluronan receptor. Molecular Biology of the Cell, 12(4), 891–900. https://doi.org/10.1091/mbc.12.4.891

Brabletz, T., Kalluri, R., Nieto, M. A., & Weinberg, R. A. (2018). EMT in cancer. Nature Reviews Cancer. https://doi.org/10.1038/nrc.2017.118

Brade, T., Pane, L. S., Moretti, A., Chien, K. R., & Laugwitz, K.-L. (2013). Embryonic Heart Progenitors and Cardiogenesis. Cold Spring Harbor Perspectives in Medicine, 3(10), a013847–a013847. https://doi.org/10.1101/cshperspect.a013847

Braitsch, C. M., Combs, M. D., Quaggin, S. E., & Yutzey, K. E. (2012). Pod1/Tcf21 is regulated by retinoic acid signaling and inhibits differentiation of epicardium- derived cells into smooth muscle in the developing heart. Developmental Biology, 368(2), 345–357. https://doi.org/10.1016/J.YDBIO.2012.06.002

Braitsch, C. M., Kanisicak, O., van Berlo, J. H., Molkentin, J. D., & Yutzey, K. E. (2013). Differential expression of embryonic epicardial progenitor markers and localization of cardiac fibrosis in adult ischemic injury and hypertensive heart disease. Journal of Molecular and Cellular Cardiology. https://doi.org/10.1016/j.yjmcc.2013.10.005

Braitsch, C., & Yutzey, K. (2013). Transcriptional Control of Cell Lineage Development in Epicardium-Derived Cells. Journal of Developmental Biology. https://doi.org/10.3390/jdb1020092

Brand, T. (2003). Heart development: molecular insights into cardiac specification and

Sophie McManus – February 2020 205

BNC1 regulates human epicardial heterogeneity and function

early morphogenesis. Developmental Biology, 258(1), 1–19. https://doi.org/10.1016/s0012-1606(03)00112-x

Bruneau, B. G. (2013). Signaling and transcriptional networks in heart development and regeneration. Cold Spring Harbor Perspectives in Biology. https://doi.org/10.1101/cshperspect.a008292

Buckingham, M., Meilhac, S., & Zaffran, S. (2005). Building the mammalian heart from two sources of myocardial cells. Nature Reviews. Genetics, 6(11), 826–835. https://doi.org/10.1038/nrg1710

Burian, A., Borowska-Wykre, D., Uyttewaal, M., Wrzalik, R., Kwiatkowska, D., Hamant, O., & Boudaoud, A. (2013). FibrilTool, an ImageJ plug-in to quantify fibrillar structures in raw microscopy images. https://doi.org/10.1038/nprot.2014.024

Burridge, P. W., Keller, G., Gold, J. D., & Wu, J. C. (2012, January 6). Production of de novo cardiomyocytes: Human pluripotent stem cell differentiation and direct reprogramming. Cell Stem Cell, Vol. 10, pp. 16–28. https://doi.org/10.1016/j.stem.2011.12.013

Burridge, P. W., Matsa, E., Shukla, P., Lin, Z. C., Churko, J. M., Ebert, A. D., … Wu, J. C. (2014). Chemically defined generation of human cardiomyocytes. Nature Methods, 11(8), 855–860. https://doi.org/10.1038/nmeth.2999

Cai, C.-L., Liang, X., Shi, Y., Chu, P.-H., Pfaff, S. L., Chen, J., & Evans, S. (2003). Isl1 Identifies a Cardiac Progenitor Population that Proliferates Prior to Differentiation and Contributes a Majority of Cells to the Heart. Developmental Cell, 5(6), 877– 889. https://doi.org/10.1016/S1534-5807(03)00363-0

Cai, C.-L., Martin, J. C., Sun, Y., Cui, L., Wang, L., Ouyang, K., … Evans, S. M. (2008). A myocardial lineage derives from Tbx18 epicardial cells. Nature, 454(7200), 104– 108. https://doi.org/10.1038/nature06969

Cai, W., Tan, J., Yan, J., Zhang, L., Cai, X., Wang, H., … Cai, C. L. (2019). Limited Regeneration Potential with Minimal Epicardial Progenitor Conversions in the Neonatal Mouse Heart after Injury. Cell Reports, 28(1), 190-201.e3. https://doi.org/10.1016/j.celrep.2019.06.003

Cano, E., Carmona, R., Ruiz-Villalba, A., Rojas, A., Chau, Y. Y., Wagner, K. D., …

206 Sophie McManus - February 2020

Chapter 5: References

Pérez-Pomares, J. M. (2016). Extracardiac septum transversum/proepicardial endothelial cells pattern embryonic coronary arterio-venous connections. Proceedings of the National Academy of Sciences of the United States of America, 113(3), 656–661. https://doi.org/10.1073/pnas.1509834113

Cao, J., Navis, A., Cox, B. D., Dickson, A. L., Gemberling, M., Karra, R., … Poss, K. D. (2016). Single epicardial cell transcriptome sequencing identifies Caveolin 1 as an essential factor in zebrafish heart regeneration. Development (Cambridge, England), 143(2), 232–243. https://doi.org/10.1242/dev.130534

Cao, J., Wang, J., Jackman, C. P., Cox, A. H., Trembley, M. A., Balowski, J. J., … Poss, K. D. (2017). Tension Creates an Endoreplication Wavefront that Leads Regeneration of Epicardial Tissue. Developmental Cell. https://doi.org/10.1016/j.devcel.2017.08.024

Cao, Y., & Cao, J. (2018). Covering and Re-Covering the Heart: Development and Regeneration of the Epicardium. Journal of Cardiovascular Development and Disease. https://doi.org/10.3390/jcdd6010003

Casanova, J. C., Travisano, S., & de la Pompa, J. L. (2013). Epithelial-to-mesenchymal transition in epicardium is independent of snail1. Genesis, 51(1), 32–40. https://doi.org/10.1002/dvg.22353

Cesna, S., Eicken, A., Juenger, H., & Hess, J. (2013). Successful treatment of a newborn with acute myocardial infarction on the first day of life. Pediatric Cardiology, 34(8), 1868–1870. https://doi.org/10.1007/s00246-012-0417-2

Chablais, F., & Jazwinska, A. (2012). The regenerative capacity of the zebrafish heart is dependent on TGFβ signaling. Development (Cambridge, England), 139(11), 1921– 1930. https://doi.org/10.1242/dev.078543

Chang, H. Y., Chang, H. M., Wu, T. J., Chaing, C. Y., Tzai, T. S., Cheng, H. L., … Liu, H. S. (2017). The role of Lutheran/basal cell adhesion molecule in human bladder carcinogenesis. Journal of Biomedical Science, 24(1). https://doi.org/10.1186/s12929-017-0360-x

Chapman, N. A., Dupré, D. J., & Rainey, J. K. (2014, August 17). The apelin receptor: Physiology, pathology, cell signalling, and ligand modulation of a peptide-activated

Sophie McManus – February 2020 207

BNC1 regulates human epicardial heterogeneity and function

class A GPCR. Biochemistry and Cell Biology, Vol. 92, pp. 431–440. https://doi.org/10.1139/bcb-2014-0072

Chen, H., Albergante, L., Hsu, J. Y., Lareau, C. A., Lo Bosco, G., Guan, J., … Pinello, L. (2019). Single-cell trajectories reconstruction, exploration and mapping of omics data with STREAM. Nature Communications, 10(1). https://doi.org/10.1038/s41467-019-09670-4

Chen, Q. K., Lee, K., Radisky, D. C., & Nelson, C. M. (2013). Extracellular matrix proteins regulate epithelial-mesenchymal transition in mammary epithelial cells. Differentiation, 86(3), 126–132. https://doi.org/10.1016/j.diff.2013.03.003

Chen, T. H. P., Chang, T. C., Kang, J. O., Choudhary, B., Makita, T., Tran, C. M., … Sucov, H. M. (2002). Epicardial induction of fetal cardiomyocyte proliferation via a retinoic acid-inducible trophic factor. Developmental Biology, 250(1), 198–207. https://doi.org/10.1006/dbio.2002.0796

Chen, X., Zhang, S., Wang, Z., Wang, F., Cao, X., Wu, Q., … Fang, Z. (2018). Supervillin promotes epithelial-mesenchymal transition and metastasis of hepatocellular carcinoma in hypoxia via activation of the RhoA/ROCK-ERK/p38 pathway. Journal of Experimental & Clinical Cancer Research, 37(1), 128. https://doi.org/10.1186/s13046-018-0787-2

Chen, Y., Takizawa, N., Crowley, J. L., Oh, S. W., Gatto, C. L., Kambara, T., … Luna, E. J. (2003). F-actin and Myosin II Binding Domains in Supervillin. Journal of Biological Chemistry, 278(46), 46094–46106. https://doi.org/10.1074/jbc.M305311200

Chen, Z., Zhuo, W., Wang, Y., Ao, X., & An, J. (2008). Down-regulation of layilin, a novel hyaluronan receptor, via RNA interference, inhibits invasion and lymphatic metastasis of human lung A549 cells. Biotechnology and Applied Biochemistry, 50(2), 89. https://doi.org/10.1042/ba20070138

Cheung, C., Bernardo, A. S., Trotter, M. W. B., Pedersen, R. A., & Sinha, S. (2012). Generation of human vascular smooth muscle subtypes provides insight into embryological origin-dependent disease susceptibility. Nature Biotechnology, 30(2), 165–173. https://doi.org/10.1038/nbt.2107

208 Sophie McManus - February 2020

Chapter 5: References

Cho, A., Graves, J., & Reidy, M. A. (2000). Mitogen-activated protein kinases mediate matrix metalloproteinase-9 expression in vascular smooth muscle cells. Arteriosclerosis, Thrombosis, and Vascular Biology, 20(12), 2527–2532. https://doi.org/10.1161/01.ATV.20.12.2527

Christoffels, V. M., Grieskamp, T., Norden, J., Mommersteeg, M. T. M., Rudat, C., & Kispert, A. (2009). Tbx18 and the fate of epicardial progenitors. Nature, 458(7240), E8-9; discussion E9-10. https://doi.org/10.1038/nature07916

Combs, M. D., Braitsch, C. M., Lange, A. W., James, J. F., & Yutzey, K. E. (2011). NFATC1 promotes epicardium-derived cell invasion into myocardium. Development, 138(9), 1747–1757. https://doi.org/10.1242/dev.060996

Crowley, J. L., Smith, T. C., Fang, Z., Takizawa, N., & Luna, E. J. (2009). Supervillin reorganizes the actin cytoskeleton and increases invadopodial efficiency. Molecular Biology of the Cell, 20(3), 948–962. https://doi.org/10.1091/mbc.e08-08-0867

Cui, Y., Zheng, Y., Liu, X., Wen, L., Qiao, J., & Tang, F. (2019). Single-Cell Transcriptome Analysis Maps the Developmental Track of the Human Heart. CellReports, 26, 1934-1950.e5. https://doi.org/10.1016/j.celrep.2019.01.079

Cui, Y., Zheng, Y., Liu, X., Yan, L., Fan, X., Yong, J., … Tang, F. (2019). Single-Cell Transcriptome Analysis Maps the Developmental Track of the Human Heart. Cell Reports, 26(7), 1934-1950.e5. https://doi.org/10.1016/j.celrep.2019.01.079

Curtis, M. W., & Russell, B. (2009). Cardiac Tissue Engineering. The Journal of Cardiovascular Nursing, 24(2), 87–92. https://doi.org/10.1097/01.JCN.0000343562.06614.49

Dana, H., Chalbatani, G. M., Mahmoodzadeh, H., Karimloo, R., Rezaiean, O., Moradzadeh, A., … Gharagouzlo, E. (2017). Molecular Mechanisms and Biological Functions of siRNA. International Journal of Biomedical Science : IJBS.

David, J. M., & Rajasekaran, A. K. (2012). Dishonorable discharge: The oncogenic roles of cleaved e-cadherin fragments. Cancer Research. https://doi.org/10.1158/0008- 5472.CAN-11-3498

Del Monte, G., Casanova, J. C., Guadix, J. A., MacGrogan, D., Burch, J. B. E., Pérez- Pomares, J. M., & De La Pompa, J. L. (2011). Differential notch signaling in the

Sophie McManus – February 2020 209

BNC1 regulates human epicardial heterogeneity and function

epicardium is required for cardiac inflow development and coronary vessel morphogenesis. Circulation Research. https://doi.org/10.1161/CIRCRESAHA.110.229062

DeLaughter, D. M., Bick, A. G., Wakimoto, H., McKean, D., Gorham, J. M., Kathiriya, I. S., … Seidman, C. E. (2016). Single-Cell Resolution of Temporal Gene Expression during Heart Development. Developmental Cell, 39(4), 480–490. https://doi.org/10.1016/j.devcel.2016.10.001

DeRuiter, M. C., Poelmann, R. E., VanderPlas-de Vries, I., Mentink, M. M. T., & Gittenberger-de Groot, A. C. (1992). The development of the myocardium and endocardium in mouse embryos - Fusion of two heart tubes? Anatomy and Embryology. https://doi.org/10.1007/BF00174084

Derycke, L. D. M., & Bracke, M. E. (2004). N-cadherin in the spotlight of cell-cell adhesion, differentiation, embryogenesis, invasion and signalling. The International Journal of Developmental Biology, 48(5–6), 463–476. https://doi.org/10.1387/ijdb.041793ld

Dettman, R. W., Denetclaw, W., Ordahl, C. P., & Bristow, J. (1998). Common Epicardial Origin of Coronary Vascular Smooth Muscle, Perivascular Fibroblasts, and Intermyocardial Fibroblasts in the Avian Heart. Developmental Biology, 193(2), 169–181. https://doi.org/10.1006/dbio.1997.8801

Díaz, B. (2013). Invadopodia Detection and Gelatin Degradation Assay. BIO- PROTOCOL, 3(24). https://doi.org/10.21769/bioprotoc.997

Dohn, T. E., & Waxman, J. S. (2012). Distinct phases of Wnt/β-catenin signaling direct cardiomyocyte formation in zebrafish. Developmental Biology, 361(2), 364–376. https://doi.org/10.1016/j.ydbio.2011.10.032

Downs, K. M. (2009). The enigmatic primitive streak: Prevailing notions and challenges concerning the body axis of mammals. BioEssays. https://doi.org/10.1002/bies.200900038

Drenckhahn, J.-D., Schwarz, Q. P., Gray, S., Laskowski, A., Kiriazis, H., Ming, Z., … Cox, T. C. (2008). Compensatory growth of healthy cardiac cells in the presence of diseased cells restores tissue homeostasis during heart development. Developmental

210 Sophie McManus - February 2020

Chapter 5: References

Cell, 15(4), 521–533. https://doi.org/10.1016/j.devcel.2008.09.005

Dronkers, E., Moerkamp, A. T., van Herwaarden, T., Goumans, M. J., & Smits, A. M. (2018). The isolation and culture of primary epicardial cells derived from human adult and fetal heart specimens. Journal of Visualized Experiments, 2018(134). https://doi.org/10.3791/57370

Duan, J., Gherghe, C., Liu, D., Hamlett, E., Srikantha, L., Rodgers, L., … Deb, A. (2012). Wnt1/βcatenin injury response activates the epicardium and cardiac fibroblasts to promote cardiac repair. EMBO Journal. https://doi.org/10.1038/emboj.2011.418

Duim, S. N., Kurakula, K., Goumans, M.-J., & Kruithof, B. P. T. (2015). Cardiac endothelial cells express Wilms’ tumor-1. Journal of Molecular and Cellular Cardiology, 81, 127–135. https://doi.org/10.1016/j.yjmcc.2015.02.007

Duim, S. N., Smits, A. M., Kruithof, B. P. T., & Goumans, M.-J. (2016). The roadmap of WT1 protein expression in the human fetal heart. Journal of Molecular and Cellular Cardiology, 90, 139–145. https://doi.org/10.1016/j.yjmcc.2015.12.008

DY, S., B, F., JN, C., KS, W., BM, W., SE, M., … MC, F. (1996). Mutations affecting the formation and function of the cardiovascular system in the zebrafish embryo. Development, 123, 285–292.

E.J. van den Bos, W.J. van der Giessen, D. J. D. (2008). Cell transplantation for cardiac regeneration: where do we stand? Netherlands Heart Journal, 16(3).

Edwards, D. R., Handsley, M. M., & Pennington, C. J. (2009, October). The ADAM metalloproteinases. Molecular Aspects of Medicine, Vol. 29, pp. 258–289. https://doi.org/10.1016/j.mam.2008.08.001

Fang, M., Xiang, F.-L., Braitsch, C. M., & Yutzey, K. E. (2016). Epicardium-derived fibroblasts in heart development and disease. Journal of Molecular and Cellular Cardiology, 91, 23–27. https://doi.org/10.1016/j.yjmcc.2015.12.019

Fang, Z., Takizawa, N., Wilson, K. A., Smith, T. C., Delprato, A., Davidson, M. W., … Luna, E. J. (2010). The Membrane-Associated Protein, Supervillin, Accelerates F- Actin-Dependent Rapid Integrin Recycling and Cell Motility. Traffic, 11(6), 782– 799. https://doi.org/10.1111/j.1600-0854.2010.01062.x

Sophie McManus – February 2020 211

BNC1 regulates human epicardial heterogeneity and function

Feuerborn, A., Mathow, D., Srivastava, P. K., Gretz, N., & Gröne, H.-J. (2015). Basonuclin-1 modulates epithelial plasticity and TGF-β1-induced loss of epithelial cell integrity. Oncogene, 34(9), 1185–1195. https://doi.org/10.1038/onc.2014.54

Freire, A. G., Resende, T. P., & Pinto-Do-Ó, P. (2014). Building and repairing the heart: What can we learn from embryonic development? BioMed Research International. https://doi.org/10.1155/2014/679168

Freitas, V. M., do Amaral, J. B., Silva, T. A., Santos, E. S., Mangone, F. R., Pinheiro, J. de J., … Machado-Santelli, G. M. (2013). Decreased expression of ADAMTS-1 in human breast tumors stimulates migration and invasion. Molecular Cancer, 12(1). https://doi.org/10.1186/1476-4598-12-2

Furtado, M. B., Nim, H. T., Boyd, S. E., & Rosenthal, N. A. (2016, February 1). View from the heart: Cardiac fibroblasts in development, scarring and regeneration. Development (Cambridge), Vol. 143, pp. 387–397. https://doi.org/10.1242/dev.120576

Gade, P., & Kalvakolanu, D. V. (2012). Chromatin immunoprecipitation assay as a tool for analyzing transcription factor activity. Methods in Molecular Biology, 809, 85– 104. https://doi.org/10.1007/978-1-61779-376-9_6

Gambardella, L., Mcmanus, S. A., Moignard, V., Delaune, A., Andrews, S., Bernard, W. G., … Sinha, S. (2019). BNC1 regulates cell heterogeneity in human pluripotent stem cell derived-epicardium. (November). https://doi.org/10.1242/dev.174441

Gao, F., Zhang, J., Wang, X., Yang, J., Chen, D., Huff, V., & Liu, Y. X. (2014). Wt1 functions in ovarian follicle development by regulating granulosa cell differentiation. Human Molecular Genetics. https://doi.org/10.1093/hmg/ddt423

Gilbert, S. F. (2000). Developmental Biology, 6th Edition.

Gilles, C., Newgreen, D. F., Sato, H., & Thompson, E. W. (2013). Matrix Metalloproteases and Epithelial-to-Mesenchymal Transition: Implications for Carcinoma Metastasis.

Gitsels, L. A., Kulinskaya, E., & Steel, N. (2017). Survival prospects after acute myocardial infarction in the UK: A matched cohort study 1987-2011. BMJ Open. https://doi.org/10.1136/bmjopen-2016-013570

212 Sophie McManus - February 2020

Chapter 5: References

Gittenberger-De Groot, A. C., Vrancken Peeters, M. P. F. M., Bergwerff, M., Mentink, M. M. T., & Poelmann, R. E. (2000). Epicardial outgrowth inhibition leads to compensatory mesothelial outflow tract collar and abnormal cardiac septation and coronary formation. Circulation Research. https://doi.org/10.1161/01.RES.87.11.969

Gittenberger-de Groot, A C, Vrancken Peeters, M. P., Mentink, M. M., Gourdie, R. G., & Poelmann, R. E. (1998). Epicardium-derived cells contribute a novel population to the myocardial wall and the atrioventricular cushions. Circulation Research, 82(10), 1043–1052. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/9622157

Gittenberger-de Groot, Adriana C., Winter, E. M., Bartelings, M. M., Jose Goumans, M., DeRuiter, M. C., & Poelmann, R. E. (2012, July). The arterial and cardiac epicardium in development, disease and repair. Differentiation, Vol. 84, pp. 41–53. https://doi.org/10.1016/j.diff.2012.05.002

Gomez-Martinez, M., Schmitz, D., & Hergovich, A. (2013). Generation of Stable Human Cell Lines with Tetracycline-inducible (Tet-on) shRNA or cDNA Expression. Journal of Visualized Experiments. https://doi.org/10.3791/50171

Gonzalez-Rosa, J. M., Martin, V., Peralta, M., Torres, M., & Mercader, N. (2011). Extensive scar formation and regression during heart regeneration after cryoinjury in zebrafish. Development. https://doi.org/10.1242/dev.060897

González, A., Schelbert, E. B., Díez, J., & Butler, J. (2018). Myocardial Interstitial Fibrosis in Heart Failure: Biological and Translational Perspectives. Journal of the American College of Cardiology. https://doi.org/10.1016/j.jacc.2018.02.021

Granata, A., Serrano, F., Bernard, W. G., McNamara, M., Low, L., Sastry, P., & Sinha, S. (2017). An iPSC-derived vascular model of Marfan syndrome identifies key mediators of smooth muscle cell death. Nature Genetics, 49(1), 97–109. https://doi.org/10.1038/ng.3723

Groot, A. C. G. de, Winter, E. M., & Poelmann, R. E. (2010). Epicardium-derived cells (EPDCs) in development, cardiac disease and repair of ischemia. Journal of Cellular and Molecular Medicine, Vol. 14, pp. 1056–1060. https://doi.org/10.1111/j.1582- 4934.2010.01077.x

Sophie McManus – February 2020 213

BNC1 regulates human epicardial heterogeneity and function

Guadix, J. A., Orlova, V. V., Giacomelli, E., Bellin, M., Ribeiro, M. C., Mummery, C. L., … Passier, R. (2017). Human Pluripotent Stem Cell Differentiation into Functional Epicardial Progenitor Cells. Stem Cell Reports, 9(6), 1754–1764. https://doi.org/10.1016/j.stemcr.2017.10.023

Gui, T., Sun, Y., Shimokado, A., & Muragaki, Y. (2012). The Roles of Mitogen- Activated Protein Kinase Pathways in TGF- β -Induced Epithelial-Mesenchymal Transition . Journal of Signal Transduction, 2012, 1–10. https://doi.org/10.1155/2012/289243

Gunadasa-Rohling, M., Masters, M., Maguire, M. L., Smart, S. C., Schneider, J. E., & Riley, P. R. (2018). Magnetic Resonance Imaging of the Regenerating Neonatal Mouse Heart Circulation. Circulation, 138, 2439–2441. https://doi.org/10.1161/CIRCULATIONAHA.118.036086

Guo, J.-K. (2002). WT1 is a key regulator of podocyte function: reduced expression levels cause crescentic glomerulonephritis and mesangial sclerosis. Human Molecular Genetics. https://doi.org/10.1093/hmg/11.6.651

Haghverdi, L., Büttner, M., Wolf, F. A., Buettner, F., & Theis, F. J. (2016). Diffusion pseudotime robustly reconstructs lineage branching. Nature Methods, 13(10), 845– 848. https://doi.org/10.1038/nmeth.3971

Hami, D., Grimes, A. C., Tsai, H.-J., & Kirby, M. L. (2011). Zebrafish cardiac development requires a conserved secondary heart field. Development (Cambridge, England), 138(11), 2389–2398. https://doi.org/10.1242/dev.061473

Harvey, R. P. (2002). Patterning the vertebrate heart. Nature Reviews. Genetics, 3(7), 544–556. https://doi.org/10.1038/nrg843

Hatcher, C. J., Diman, N. Y. S.-G., Kim, M.-S., Pennisi, D., Song, Y., Goldstein, M. M., … Basson, C. T. (2004). A role for Tbx5 in proepicardial cell migration during cardiogenesis. Physiological Genomics. https://doi.org/10.1152/physiolgenomics.00060.2004

Hatcher, C. J., Goldstein, M. M., Mah, C. S., Susan Delia, C., & Basson, C. T. (2000). Identification and localization of TBX5 transcription factor during human cardiac morphogenesis. Developmental Dynamics, 219(1), 90–95.

214 Sophie McManus - February 2020

Chapter 5: References

https://doi.org/10.1002/1097-0177(200009)219:1<90::AID-DVDY1033>3.0.CO;2- L

Haubner, B. J., Schneider, J., Schweigmann, U., Schuetz, T., Dichtl, W., Velik-Salchner, C., … Penninger, J. M. (2016). Functional Recovery of a Human Neonatal Heart After Severe Myocardial Infarction. Circulation Research, 118(2), 216–221. https://doi.org/10.1161/CIRCRESAHA.115.307017

Hay, E. D. (1995). An overview of epithelio-mesenchymal transformation. Acta Anatomica, 154(1), 8–20. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8714286

Haynes, J., Srivastava, J., Madson, N., Wittmann, T., & Barber, D. L. (2011a). Dynamic actin remodeling during epithelial-mesenchymal transition depends on increased moesin expression. Molecular Biology of the Cell, 22(24), 4750–4764. https://doi.org/10.1091/mbc.E11-02-0119

Haynes, J., Srivastava, J., Madson, N., Wittmann, T., & Barber, D. L. (2011b). Dynamic actin remodeling during epithelial-mesenchymal transition depends on increased moesin expression. https://doi.org/10.1091/mbc.E11-02-0119

Heikinheimo, M., Scandrett, J. M., & Wilson, D. B. (1994). Localization of Transcription Factor GATA-4 to Regions of the Mouse Embryo Involved in Cardiac Development. Developmental Biology, 164(2), 361–373. https://doi.org/10.1006/dbio.1994.1206

Hirakow, R. (1992). Epicardial Formation in Staged Human Embryos. Kaibogaku Zasshi. Journal of Anatomy.

Hirose, T. (2006). PAR3 is essential for cyst-mediated epicardial development by establishing apical cortical domains. Development. https://doi.org/10.1242/dev.02294

Hong, K.-O., Lee, J.-I., Hong, S.-P., & Hong, S.-D. (2016). Thymosin β4 induces proliferation, invasion, and epithelial-to-mesenchymal transition of oral squamous cell carcinoma. Amino Acids, 48(1), 117–127. https://doi.org/10.1007/s00726-015- 2070-6

Hooper, J. D., Clements, J. A., Quigley, J. P., & Antalis, T. M. (2001, January 12). Type II transmembrane serine proteases. Insights into an emerging class of cell surface

Sophie McManus – February 2020 215

BNC1 regulates human epicardial heterogeneity and function

proteolytic enzymes. Journal of Biological Chemistry, Vol. 276, pp. 857–860. https://doi.org/10.1074/jbc.R000020200

Huang, H.-C., Hu, C.-H., Tang, M.-C., Wang, W.-S., Chen, P.-M., & Su, Y. (2007). Thymosin beta4 triggers an epithelial-mesenchymal transition in colorectal carcinoma by upregulating integrin-linked kinase. Oncogene, 26(19), 2781–2790. https://doi.org/10.1038/sj.onc.1210078

Huang, M., Wang, Y. P., Zhu, L. Q., Cai, Q., Li, H. H., & Yang, H. F. (2016). MAPK pathway mediates epithelial-mesenchymal transition induced by paraquat in alveolar epithelial cells. Environmental Toxicology, 31(11), 1407–1414. https://doi.org/10.1002/tox.22146

Humeres, C., & Frangogiannis, N. G. (2019, June 1). Fibroblasts in the Infarcted, Remodeling, and Failing Heart. JACC: Basic to Translational Science, Vol. 4, pp. 449–467. https://doi.org/10.1016/j.jacbts.2019.02.006

Ikeda, Y., Imai, Y., Kumagai, H., Nosaka, T., Morikawa, Y., Hisaoka, T., … Kitamura, T. (2004). Vasorin, a transforming growth factor beta-binding protein expressed in vascular smooth muscle cells, modulates the arterial response to injury in vivo. Proceedings of the National Academy of Sciences of the United States of America, 101(29), 10732–10737. https://doi.org/10.1073/pnas.0404117101

Ishii, Y., Garriock, R. J., Navetta, A. M., Coughlin, L. E., & Mikawa, T. (2010). BMP signals promote proepicardial protrusion necessary for recruitment of coronary vessel and epicardial progenitors to the heart. Developmental Cell. https://doi.org/10.1016/j.devcel.2010.07.017

Iuchi, S., & Green, H. (1997). Nuclear localization of basonuclin in human keratinocytes and the role of phosphorylation. Proceedings of the National Academy of Sciences of the United States of America, 94(15), 7948–7953. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/9223293

Ivanovitch, K., Temiño, S., & Torres, M. (2017). Live imaging of heart tube development in mouse reveals alternating phases of cardiac differentiation and morphogenesis. ELife. https://doi.org/10.7554/eLife.30668

Iyer, D., Gambardella, L., Bernard, W. G., Serrano, F., Mascetti, V. L., Pedersen, R. A.,

216 Sophie McManus - February 2020

Chapter 5: References

… Sinha, S. (2015). Robust derivation of epicardium and its differentiated smooth muscle cell progeny from human pluripotent stem cells. Development (Cambridge, England), 142(8), 1528–1541. https://doi.org/10.1242/dev.119271

Izdebska, M., Zielińska, W., Grzanka, D., & Gagat, M. (2018). The Role of Actin Dynamics and Actin-Binding Proteins Expression in Epithelial-to-Mesenchymal Transition and Its Association with Cancer Progression and Evaluation of Possible Therapeutic Targets. BioMed Research International, 2018, 1–13. https://doi.org/10.1155/2018/4578373

Jopling, C., Sleep, E., Raya, M., Martí, M., Raya, A., & Belmonte, J. C. I. (2010). Zebrafish heart regeneration occurs by cardiomyocyte dedifferentiation and proliferation. Nature. https://doi.org/10.1038/nature08899

K., L., S., S., S., A., H., A., M., A., T., E., … K., F. (2006). Intracoronary injection of mononuclear bone marrow cells in acute myocardial infarction. New England Journal of Medicine. https://doi.org/10.1056/NEJMoa055706

Katz, T. C., Singh, M. K., Degenhardt, K., Rivera-Feliciano, J., Johnson, R. L., Epstein, J. A., & Tabin, C. J. (2012). Distinct compartments of the proepicardial organ give rise to coronary vascular endothelial cells. Developmental Cell, 22(3), 639–650. https://doi.org/10.1016/j.devcel.2012.01.012

Kelly, R. G., Buckingham, M. E., & Moorman, A. F. (2014). Heart fields and cardiac morphogenesis. Cold Spring Harbor Perspectives in Medicine. https://doi.org/10.1101/cshperspect.a015750

Kelwick, R., Wagstaff, L., Decock, J., Roghi, C., Cooley, L. S., Robinson, S. D., … Edwards, D. R. (2015). Metalloproteinase-Dependent and -Independent processes contribute to inhibition of breast cancer cell migration, angiogenesis and liver metastasis by a disintegrin and metalloproteinase with thrombospondin motifs-15. International Journal of Cancer, 136(4), E14–E26. https://doi.org/10.1002/ijc.29129

Khurana, S., & George, S. P. (2008). Regulation of cell structure and function by actin- binding proteins: villin’s perspective. FEBS Letters, 582(14), 2128–2139. https://doi.org/10.1016/j.febslet.2008.02.040

Sophie McManus – February 2020 217

BNC1 regulates human epicardial heterogeneity and function

Kikuchi, K., Gupta, V., Wang, J., Holdway, J. E., Wills, A. A., Fang, Y., & Poss, K. D. (2011). tcf21+ epicardial cells adopt non-myocardial fates during zebrafish heart development and regeneration. Development (Cambridge, England), 138(14), 2895– 2902. https://doi.org/10.1242/dev.067041

Kikuchi, K., Holdway, J. E., Major, R. J., Blum, N., Dahn, R. D., Begemann, G., & Poss, K. D. (2011). Retinoic Acid Production by Endocardium and Epicardium Is an Injury Response Essential for Zebrafish Heart Regeneration. Developmental Cell. https://doi.org/10.1016/j.devcel.2011.01.010

Kikuchi, K., Holdway, J. E., Werdich, A. A., Anderson, R. M., Fang, Y., Egnaczyk, G. F., … Poss, K. D. (2010). Primary contribution to zebrafish heart regeneration by gata4+ cardiomyocytes. Nature. https://doi.org/10.1038/nature08804

Kim, E.-S., Kim, M.-S., & Moon, A. (2004). TGF-β-induced upregulation of MMP-2 and MMP-9 depends on p38 MAPK, but not ERK signaling in MCF10A human breast epithelial cells. International Journal of Oncology. https://doi.org/10.3892/ijo.25.5.1375

Kim, J.-Y., Jeong, H. S., Chung, T., Kim, M., Lee, J. H., Jung, W. H., & Koo, J. S. (2017). The value of phosphohistone H3 as a proliferation marker for evaluating invasive breast cancers: A comparative study with Ki67. Oncotarget. https://doi.org/10.18632/oncotarget.17775

Kim, J., Wu, Q., Zhang, Y., Wiens, K. M., Huang, Y., Rubin, N., … Lien, C.-L. (2010). PDGF signaling is required for epicardial function and blood vessel formation in regenerating zebrafish hearts. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.0915016107

Klaus, A., Saga, Y., Taketo, M. M., Tzahor, E., & Birchmeier, W. (2007). Distinct roles of Wnt/beta-catenin and Bmp signaling during early cardiogenesis. Proceedings of the National Academy of Sciences, 104(47), 18531–18536. https://doi.org/10.1073/pnas.0703113104

Klaus, Alexandra, Müller, M., Schulz, H., Saga, Y., Martin, J. F., & Birchmeier, W. (2012). Wnt/β-catenin and Bmp signals control distinct sets of transcription factors in cardiac progenitor cells. Proceedings of the National Academy of Sciences of the

218 Sophie McManus - February 2020

Chapter 5: References

United States of America, 109(27), 10921–10926. https://doi.org/10.1073/pnas.1121236109

Koh, P. W., Sinha, R., Barkal, A. A., Morganti, R. M., Chen, A., Weissman, I. L., … Loh, K. M. (2016). An atlas of transcriptional, chromatin accessibility, and surface marker changes in human mesoderm development. Scientific Data. https://doi.org/10.1038/sdata.2016.109

Komiyama, M., Ito, K., & Shimada, Y. (1987). Origin and development of the epicardium in the mouse embryo. Anatomy and Embryology, 176(2), 183–189. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/3619072

Krainock, M., Toubat, O., Danopoulos, S., Beckham, A., Warburton, D., & Kim, R. (2016). Epicardial Epithelial-to-Mesenchymal Transition in Heart Development and Disease. Journal of Clinical Medicine, 5(2). https://doi.org/10.3390/jcm5020027

Krampert, M., Kuenzle, S., Thai, S. N. M., Lee, N., Iruela-Arispe, M. L., & Werner, S. (2005). ADAMTS1 proteinase is up-regulated in wounded skin and regulates migration of fibroblasts and endothelial cells. Journal of Biological Chemistry, 280(25), 23844–23852. https://doi.org/10.1074/jbc.M412212200

Krumholz, H. M., Normand, S. L. T., & Wang, Y. (2019). Twenty-Year Trends in Outcomes for Older Adults With Acute Myocardial Infarction in the United States. JAMA Network Open. https://doi.org/10.1001/jamanetworkopen.2019.1938

Kwee, L., Baldwin, H. S., Shen, H. M., Stewart, C. L., Buck, C., Buck, C. A., & Labow, M. A. (1995). Defective development of the embryonic and extraembryonic circulatory systems in vascular cell adhesion molecule (VCAM-1) deficient mice. Development (Cambridge, England).

Lam, N. T., & Sadek, H. A. (2018, July 24). Neonatal heart regeneration comprehensive literature review. Circulation, Vol. 138, pp. 421–423. https://doi.org/10.1161/CIRCULATIONAHA.118.033648

Lamouille, S., Xu, J., & Derynck, R. (2014). Molecular mechanisms of epithelial- mesenchymal transition. Nature Reviews Molecular Cell Biology. https://doi.org/10.1038/nrm3758

Lange, S. S., & Vasquez, K. M. (2009). HMGB1: The Jack-of-all-trades protein is a

Sophie McManus – February 2020 219

BNC1 regulates human epicardial heterogeneity and function

master DNA repair mechanic. Molecular Carcinogenesis, 48(7), 571–580. https://doi.org/10.1002/mc.20544

Lavine, K. J., White, A. C., Park, C., Smith, C. S., Choi, K., Long, F., … Ornitz, D. M. (2006). Fibroblast growth factor signals regulate a wave of Hedgehog activation that is essential for coronary vascular development. Genes & Development, 20(12), 1651–1666. https://doi.org/10.1101/gad.1411406

Lavine, Kory J., Yu, K., White, A. C., Zhang, X., Smith, C., Partanen, J., & Ornitz, D. M. (2005). Endocardial and epicardial derived FGF signals regulate myocardial proliferation and differentiation in vivo. Developmental Cell. https://doi.org/10.1016/j.devcel.2004.12.002

Lavine, Kory J, Yu, K., White, A. C., Zhang, X., Smith, C., Partanen, J., & Ornitz, D. M. (2005). Endocardial and epicardial derived FGF signals regulate myocardial proliferation and differentiation in vivo. Developmental Cell, 8(1), 85–95. https://doi.org/10.1016/j.devcel.2004.12.002

Lawson, K. A., & Pedersen, R. A. (1992). Clonal analysis of cell fate during gastrulation and early neurulation in the mouse. Ciba Foundation Symposium, 165, 3–21; discussion 21-6. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/1516475

Lee, K.-H., Evans, S., Ruan, T. Y., & Lassar, A. B. (2004). SMAD-mediated modulation of YY1 activity regulates the BMP response and cardiac-specific expression of a GATA4/5/6-dependent chick Nkx2.5 enhancer. Development (Cambridge, England), 131(19), 4709–4723. https://doi.org/10.1242/dev.01344

Lepilina, A., Coon, A. N., Kikuchi, K., Holdway, J. E., Roberts, R. W., Burns, C. G., & Poss, K. D. (2006). A dynamic epicardial injury response supports progenitor cell activity during zebrafish heart regeneration. Cell, 127(3), 607–619. https://doi.org/10.1016/j.cell.2006.08.052

Lescroart, F., Chabab, S., Lin, X., Rulands, S., Paulissen, C., Rodolosse, A., … Blanpain, C. (2014). Early lineage restriction in temporally distinct populations of Mesp1 progenitors during mammalian heart development. Nature Cell Biology, 16(9), 829– 840. https://doi.org/10.1038/ncb3024

Li, J., Miao, L., Zhao, C., Shaikh Qureshi, W. M., Shieh, D., Guo, H., … Wu, M. (2017).

220 Sophie McManus - February 2020

Chapter 5: References

CDC42 is required for epicardial and pro-epicardial development by mediating FGF receptor trafficking to the plasma membrane. Development, 144(9). Retrieved from http://dev.biologists.org/content/144/9/1635

Lie-Venema, H., Gittenberger-De Groot, A. C., Van Empel, L. J. P., Boot, M. J., Kerkdijk, H., De Kant, E., & DeRuiter, M. C. (2003). Ets-1 and Ets-2 transcription factors are essential for normal coronary and myocardial development in chicken embryos. Circulation Research, 92(7), 749–756. https://doi.org/10.1161/01.RES.0000066662.70010.DB

Lie-Venema, H., van den Akker, N. M. S., Bax, N. A. M., Winter, E. M., Maas, S., Kekarainen, T., … Gittenberger-de Groot, A. C. (2007). Origin, fate, and function of epicardium-derived cells (EPDCs) in normal and abnormal cardiac development. TheScientificWorldJournal, 7, 1777–1798. https://doi.org/10.1100/tsw.2007.294

Lints, T. J., Parsons, L. M., Hartley, L., Lyons, I., & Harvey, R. P. (1993). Nkx-2.5: a novel murine homeobox gene expressed in early heart progenitor cells and their myogenic descendants. Development (Cambridge, England), 119(2), 419–431. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/7904557

Liu, J., & Stainier, D. Y. R. (2010). Tbx5 and Bmp signaling are essential for proepicardium specification in zebrafish. Circulation Research, 106(12), 1818– 1828. https://doi.org/10.1161/CIRCRESAHA.110.217950

Liu, Q., Zhang, H., Tian, X., He, L., Huang, X., Tan, Z., … Zhou, B. Smooth muscle origin of postnatal 2nd CVP is pre-determined in early embryo. , 471 Biochemical and Biophysical Research Communications § (2016).

Loh, K. M. M., Chen, A., Koh, P. W. W., Deng, T. Z. Z., Sinha, R., Tsai, J. M. M., … Weissman, I. L. L. (2016). Mapping the Pairwise Choices Leading from Pluripotency to Human Bone, Heart, and Other Mesoderm Cell Types. Cell, 166(2), 451–467. https://doi.org/10.1016/j.cell.2016.06.011

Lu, J., Landerholm, T. E., Wei, J. S., Dong, X. R., Wu, S. P., Liu, X., … Majesky, M. W. (2001). Coronary smooth muscle differentiation from proepicardial cells requires rhoA-mediated actin reorganization and p160 rho-kinase activity. Developmental Biology, 240(2), 404–418. https://doi.org/10.1006/dbio.2001.0403

Sophie McManus – February 2020 221

BNC1 regulates human epicardial heterogeneity and function

Lu, P., Takai, K., Weaver, V. M., & Werb, Z. (2011). Extracellular Matrix degradation and remodeling in development and disease. Cold Spring Harbor Perspectives in Biology, 3(12). https://doi.org/10.1101/cshperspect.a005058

Lupu, I.-E., Redpath, A. N., & Smart, N. (2019). Spatiotemporal analysis reveals significant overlap of key proepicardial markers in the developing murine heart. BioRxiv, 666610. https://doi.org/10.1101/666610

Mahmoud, A. I., Kocabas, F., Muralidhar, S. A., Kimura, W., Koura, A. S., Thet, S., … Sadek, H. A. (2013). Meis1 regulates postnatal cardiomyocyte cell cycle arrest. Nature, 497(7448), 249–253. https://doi.org/10.1038/nature12054

Mahmoud, A. I., Porrello, E. R., Kimura, W., Olson, E. N., & Sadek, H. A. (2014). Surgical models for cardiac regeneration in neonatal mice. Nature Protocols, 9(2), 305–311. https://doi.org/10.1038/nprot.2014.021

Männer, J. (1993). Experimental study on the formation of the epicardium in chick embryos. Anatomy and Embryology, 187(3), 281–289. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/8470828

Marques, S. R., & Yelon, D. (2009). Differential requirement for BMP signaling in atrial and ventricular lineages establishes cardiac chamber proportionality. Developmental Biology, 328(2), 472–482. https://doi.org/10.1016/j.ydbio.2009.02.010

Martínez-Estrada, O. M., Lettice, L. A., Essafi, A., Guadix, J. A., Slight, J., Velecela, V., … Hastie, N. D. (2010). Wt1 is required for cardiovascular progenitor cell formation through transcriptional control of Snail and E-cadherin. Nature Genetics, 42(1), 89– 93. https://doi.org/10.1038/ng.494

Martinsen, B. J. (2005). REVIEWS-A PEER REVIEWED FORUM Reference Guide to the Stages of Chick Heart Embryology HH 1 (FIRST FEW HOURS OF INCUBATION, PRE-STREAK, “EMBRYONIC SHIELD”) Cardiac Progenitors. Developmental Dynamics, 233, 1217–1237. https://doi.org/10.1002/dvdy.20468

Marvin, M. J. (2001). Inhibition of Wnt activity induces heart formation from posterior mesoderm. Genes & Development, 15(3), 316–327. https://doi.org/10.1101/gad.855501

Masters, M., & Riley, P. R. (2014, November 1). The epicardium signals the way towards

222 Sophie McManus - February 2020

Chapter 5: References

heart regeneration. Stem Cell Research, Vol. 13, pp. 683–692. https://doi.org/10.1016/j.scr.2014.04.007

Maurer, M., Mary, J., Guillaud, L., Fender, M., Pelé, M., Bilzer, T., … Tiret, L. (2012). Centronuclear Myopathy in Labrador Retrievers: A Recent Founder Mutation in the PTPLA Gene Has Rapidly Disseminated Worldwide. PLoS ONE, 7(10). https://doi.org/10.1371/journal.pone.0046408

Mellgren, A. M., Smith, C. L., Olsen, G. S., Eskiocak, B., Zhou, B., Kazi, M. N., … Tallquist, M. D. (2008). Platelet-derived growth factor receptor beta signaling is required for efficient epicardial cell migration and development of two distinct coronary vascular smooth muscle cell populations. Circulation Research, 103(12), 1393–1401. https://doi.org/10.1161/CIRCRESAHA.108.176768

Mikawa, T., & Gourdie, R. G. (1996). Pericardial mesoderm generates a population of coronary smooth muscle cells migrating into the heart along with ingrowth of the epicardial organ. Developmental Biology. https://doi.org/10.1006/dbio.1996.0068

Mikawa T, G. R. (1996). Pericardial mesoderm generates a population of coronary smooth muscle cells migrating into the heart along with ingrowth of the epicardial organ. Dev Biol. https://doi.org/10.1006/dbio.1996.0068

Missinato, M. A., Tobita, K., Romano, N., Carroll, J. A., & Tsang, M. (2015). Extracellular component hyaluronic acid and its receptor Hmmr are required for epicardial EMT during heart regeneration. Cardiovascular Research, 107(4), 487– 498. https://doi.org/10.1093/cvr/cvv190

Mjaatvedt, C. H., Nakaoka, T., Moreno-Rodriguez, R., Norris, R. A., Kern, M. J., Eisenberg, C. A., … Markwald, R. R. (2001). The outflow tract of the heart is recruited from a novel heart-forming field. Developmental Biology. https://doi.org/10.1006/dbio.2001.0409

Moerkamp, A. T., Lodder, K., Van Herwaarden, T., Dronkers, E., Dingenouts, C. K. E., Tengström, F. C., … Smits, A. M. (2016). Human fetal and adult epicardial-derived cells: a novel model to study their activation. Stem Cell Research & Therapy, 7(174). https://doi.org/10.1186/s13287-016-0434-9

Moore, A. W., McInnes, L., Kreidberg, J., Hastie, N. D., & Schedl, A. (1999). YAC

Sophie McManus – February 2020 223

BNC1 regulates human epicardial heterogeneity and function

complementation shows a requirement for Wt1 in the development of epicardium, adrenal gland and throughout nephrogenesis. Development (Cambridge, England).

Moore, C. B., Guthrie, E. H., Huang, M. T. H., & Taxman, D. J. (2010). Short hairpin RNA (shRNA): design, delivery, and assessment of gene knockdown. Methods in Molecular Biology (Clifton, N.J.), 629, 141–158. https://doi.org/10.1007/978-1- 60761-657-3_10

Morris, M. R., Ricketts, C., Gentle, D., Abdulrahman, M., Clarke, N., Brown, M., … Maher, E. R. (2010). Identification of candidate tumour suppressor genes frequently methylated in renal cell carcinoma. Oncogene, 29(14), 2104–2117. https://doi.org/10.1038/onc.2009.493

Muhammad, E., Reish, O., Ohno, Y., Scheetz, T., DeLuca, A., Searby, C., … Parvari, R. (2013). Congenital myopathy is caused by mutation of HACD1. Human Molecular Genetics. https://doi.org/10.1093/hmg/ddt380

Müller, U., Bossy, B., Venstrom, K., & Reichardt, L. F. (1995). Integrin α8β1 promotes attachment, cell spreading, and neurite outgrowth on fibronectin. Molecular Biology of the Cell, 6(4), 433–448. https://doi.org/10.1091/mbc.6.4.433

Murry, C. E., Murry, C. E., Soonpaa, M. H., Soonpaa, M. H., Reinecke, H., Reinecke, H., … Field, L. J. (2004). Haematopoietic stem cells do not transdifferentiate into cardiac myocytes in myocardial infarcts. Nature. https://doi.org/10.1038/nature02446

Nahirney, P. C., Mikawa, T., & Fischman, D. A. (2003). Evidence for an extracellular matrix bridge guiding proepicardial cell migration to the myocardium of chick embryos. Developmental Dynamics. https://doi.org/10.1002/dvdy.10335

Nieto, M. A., Huang, R. Y. Y. J., Jackson, R. A. A., & Thiery, J. P. P. (2016, June 30). EMT: 2016. Cell, Vol. 166, pp. 21–45. https://doi.org/10.1016/j.cell.2016.06.028

Nishimoto, S. K., & Nishimoto, M. (2014). Matrix gla protein binds to fibronectin and enhances cell attachment and spreading on fibronectin. International Journal of Cell Biology, 2014. https://doi.org/10.1155/2014/807013

Orlic, D., Kajstura, J., Chimenti, S., Jakoniuk, I., Anderson, S. M., Li, B., … Anversa, P. (2001). Bone marrow cells regenerate infarcted myocardium. Nature, 410(6829),

224 Sophie McManus - February 2020

Chapter 5: References

701–705. https://doi.org/10.1038/35070587

Ouderkirk, J. L., & Krendel, M. (2014, August 1). Non-muscle myosins in tumor progression, cancer cell invasion, and metastasis. Cytoskeleton, Vol. 71, pp. 447– 463. https://doi.org/10.1002/cm.21187

Ozkan, J. (2019, April 1). Piero Anversa and cardiomyocyte regeneration. European Heart Journal, Vol. 40, pp. 1036–1037. https://doi.org/10.1093/eurheartj/ehz096

Pangeni, R. P., Channathodiyil, P., Huen, D. S., Eagles, L. W., Johal, B. K., Pasha, D., … Morris, M. R. (2015). The GALNT9, BNC1 and CCDC8 genes are frequently epigenetically dysregulated in breast tumours that metastasise to the brain. Clinical Epigenetics, 7(1), 57. https://doi.org/10.1186/s13148-015-0089-x

Park, M., & Yoon, Y. S. (2018, November 1). Cardiac regeneration with human pluripotent stem cell-derived cardiomyocytes. Korean Circulation Journal, Vol. 48, pp. 974–988. https://doi.org/10.4070/kcj.2018.0312

Peralta, M., Steed, E., Harlepp, S., González-Rosa, J. M., Monduc, F., Ariza-Cosano, A., … Mercader, N. (2013). Heartbeat-driven pericardiac fluid forces contribute to epicardium morphogenesis. Current Biology : CB, 23(18), 1726–1735. https://doi.org/10.1016/j.cub.2013.07.005

Pérez-Pomares, J. M., Phelps, A., Sedmerova, M., Carmona, R., González-Iriarte, M., Muoz-Chápuli, R., & Wessels, A. (2002). Experimental studies on the spatiotemporal expression of WT1 and RALDH2 in the embryonic avian heart: A model for the regulation of myocardial and valvuloseptal development by epicardially derived cells (EPDCs). Developmental Biology. https://doi.org/10.1006/dbio.2002.0706

Pérez-Pomares, José M., Phelps, A., Sedmerova, M., & Wessels, A. (2003). Epicardial- like cells on the distal arterial end of the cardiac outflow tract do not derive from the proepicardium but are derivatives of the cephalic pericardium. Developmental Dynamics. https://doi.org/10.1002/dvdy.10284

Pérez-Pomares, José María, Macías, D., García-Garrido, L., & Muñoz-Chápuli, R. (1998). The origin of the subepicardial mesenchyme in the avian embryo: An immunohistochemical and quail-chick chimera study. Developmental Biology.

Sophie McManus – February 2020 225

BNC1 regulates human epicardial heterogeneity and function

https://doi.org/10.1006/dbio.1998.8949

Pestonjamasp, K. N., Pope, R. K., Wulfkuhle, J. D., & Luna, E. J. (1997). Supervillin (p205): A novel membrane-associated, F-actin-binding protein in the villin/gelsolin superfamily. Journal of Cell Biology, 139(5), 1255–1269. https://doi.org/10.1083/jcb.139.5.1255

Phillips, M. D., Mukhopadhyay, M., Poscablo, C., & Westphal, H. (2011). Dkk1 and Dkk2 regulate epicardial specification during mouse heart development. International Journal of Cardiology. https://doi.org/10.1016/j.ijcard.2010.04.007

Pinto, A. R., Ilinykh, A., Ivey, M. J., Kuwabara, J. T., D’Antoni, M. L., Debuque, R., … Tallquist, M. D. (2016). Revisiting Cardiac Cellular CompositionNovelty and Significance. Circulation Research, 118(3), 400–409. https://doi.org/10.1161/CIRCRESAHA.115.307778

Planas-Rigol, E., Terrades-Garcia, N., Corbera-Bellalta, M., Lozano, E., Alba, M. A., Segarra, M., … Cid, M. C. (2017). Endothelin-1 promotes vascular smooth muscle cell migration across the artery wall: a mechanism contributing to vascular remodelling and intimal hyperplasia in giant-cell arteritis. Annals of the Rheumatic Diseases, 76(9), 1624–1634. https://doi.org/10.1136/annrheumdis-2016-210792

Plavicki, J. S., Hofsteen, P., Yue, M. S., Lanham, K. A., Peterson, R. E., & Heideman, W. (2014). Multiple modes of proepicardial cell migration require heartbeat. BMC Developmental Biology, 14(1), 18. https://doi.org/10.1186/1471-213X-14-18

Polizzotti, B. D., Ganapathy, B., Haubner, B. J., Penninger, J. M., & Kühn, B. (2016). A cryoinjury model in neonatal mice for cardiac translational and regeneration research. Nature Protocols. https://doi.org/10.1038/nprot.2016.031

Porrello, E. R., Mahmoud, A. I., Simpson, E., Johnson, B. A., Grinsfelder, D., Canseco, D., … Sadek, H. A. (2013). Regulation of neonatal and adult mammalian heart regeneration by the miR-15 family. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.1208863110

Porrello, Enzo R, Mahmoud, A. I., Simpson, E., Hill, J. A., Richardson, J. A., Olson, E. N., & Sadek, H. A. (2011). Transient regenerative potential of the neonatal mouse heart. Science (New York, N.Y.), 331(6020), 1078–1080.

226 Sophie McManus - February 2020

Chapter 5: References

https://doi.org/10.1126/science.1200708

Poss, K. D., Wilson, L. G., & Keating, M. T. (2002). Heart regeneration in zebrafish. Science. https://doi.org/10.1126/science.1077857

Poveda, J., Sanz, A. B., Fernandez-Fernandez, B., Carrasco, S., Ruiz-Ortega, M., Cannata-Ortiz, P., … Sanchez-Niño, M. D. (2017). MXRA5 is a TGF-β1-regulated human protein with anti-inflammatory and anti-fibrotic properties. Journal of Cellular and Molecular Medicine, 21(1), 154–164. https://doi.org/10.1111/jcmm.12953

Quaggin, S. E., Schwartz, L., Cui, S., Igarashi, P., Deimling, J., Post, M., & Rossant, J. (1999). The basic-helix-loop-helix protein pod1 is critically important for kidney and lung organogenesis. Development (Cambridge, England).

Ramjee, V., Li, D., Manderfield, L. J., Liu, F., Engleka, K. A., Aghajanian, H., … Epstein, J. A. (2017). Epicardial YAP/TAZ orchestrate an immunosuppressive response following myocardial infarction. Journal of Clinical Investigation, 127(3), 899–911. https://doi.org/10.1172/JCI88759

Reddy, K. B., Krueger, J. S., Kondapaka, S. B., & Diglio, C. A. (1999). Mitogen-activated protein kinase (MAPK) regulates the expression of progelatinase B (MMP-9) in breast epithelial cells. International Journal of Cancer, 82(2), 268–273. https://doi.org/10.1002/(SICI)1097-0215(19990719)82:2<268::AID- IJC18>3.0.CO;2-4

Reid, J. E., & Wernisch, L. (2016). Pseudotime estimation: Deconfounding single cell time series. Bioinformatics, 32(19), 2973–2980. https://doi.org/10.1093/bioinformatics/btw372

Reifers, F., Walsh E., Leger S., Stainier D, W. M. (2000). Induction and differentiation of the zebrafish heart requires fibroblast growth factor 8 (fgf8/acerebellar). Development, (127), 225–235.

Reiter, J. F., Alexander, J., Rodaway, A., Yelon, D., Patient, R., Holder, N., & Stainier, D. Y. (1999). Gata5 is required for the development of the heart and endoderm in zebrafish. Genes & Development, 13(22), 2983–2995. https://doi.org/10.1101/gad.13.22.2983

Sophie McManus – February 2020 227

BNC1 regulates human epicardial heterogeneity and function

Rhee, D. Y., Zhao, X.-Q., Francis, R. J. B., Huang, G. Y., Mably, J. D., & Lo, C. W. (2009). Connexin 43 regulates epicardial cell polarity and migration in coronary vascular development. Development (Cambridge, England), 136(18), 3185–3193. https://doi.org/10.1242/dev.032334

Riley, P. R., & Smart, N. (2009). Thymosin beta4 induces epicardium-derived neovascularization in the adult heart. Biochemical Society Transactions, 37(Pt 6), 1218–1220. https://doi.org/10.1042/BST0371218

Risebro, C. A., Vieira, J. M., Klotz, L., Riley, P. R., Dingenouts, C. K. E., Tengström, F. C., … Pedersen, R. (2015). Characterisation of the human embryonic and foetal epicardium during heart development. Development, 142(21), 3630–3636. https://doi.org/10.1242/dev.127621

Risebro, Catherine A, Vieira, J. M., Klotz, L., & Riley, P. R. (2015). Characterisation of the human embryonic and foetal epicardium during heart development. https://doi.org/10.1242/dev.127621

Risebro, Catherine A, Vieira, J. M., Klotz, L., Riley, P. R., Balmer, G. M., Bollini, S., … Duffy, H. S. (2015). Characterisation of the human embryonic and foetal epicardium during heart development. Development (Cambridge, England), 142(21), 3630– 3636. https://doi.org/10.1242/dev.127621

Robb, L., Mifsud, L., Hartley, L., Biben, C., Copeland, N. G., Gilbert, D. J., … Harvey, R. P. (1998). epicardin: A novel basic helix-loop-helix transcription factor gene expressed in epicardium, branchial arch myoblasts, and mesenchyme of developing lung, gut, kidney, and gonads. Developmental Dynamics. https://doi.org/10.1002/(SICI)1097-0177(199809)213:1<105::AID- AJA10>3.0.CO;2-1

Robertson, I. B., Horiguchi, M., Zilberberg, L., Dabovic, B., Hadjiolova, K., & Rifkin, D. B. (2015, September 1). Latent TGF-β-binding proteins. Matrix Biology, Vol. 47, pp. 44–53. https://doi.org/10.1016/j.matbio.2015.05.005

Rodgers, L. S., Lalani, S., Runyan, R. B., & Camenisch, T. D. (2008). Differential growth and multicellular villi direct proepicardial translocation to the developing mouse heart. Developmental Dynamics. https://doi.org/10.1002/dvdy.21378

228 Sophie McManus - February 2020

Chapter 5: References

Row, R. H., Pegg, A., Kinney, B. A., Farr, G. H., Maves, L., Lowell, S., … Martin, B. L. (2018). BMP and FGF signaling interact to pattern mesoderm by controlling basic helix-loop-helix transcription factor activity. ELife, 7. https://doi.org/10.7554/eLife.31018

Rudat, C., & Kispert, A. (2012). Wt1 and epicardial fate mapping. Circulation Research. https://doi.org/10.1161/CIRCRESAHA.112.273946

Saga, Y., Miyagawa-Tomita, S., Takagi, A., Kitajima, S., Miyazaki, J. i, & Inoue, T. (1999). MesP1 is expressed in the heart precursor cells and required for the formation of a single heart tube. Development (Cambridge, England).

Samatov, T. R., Wicklein, D., & Tonevitsky, A. G. (2016, August 1). L1CAM: Cell adhesion and more. Progress in Histochemistry and Cytochemistry, Vol. 51, pp. 25– 32. https://doi.org/10.1016/j.proghi.2016.05.001

Sanford, L. P., Ormsby, I., Gittenberger-de Groot, A. C., Sariola, H., Friedman, R., Boivin, G. P., … Doetschman, T. (1997). TGFbeta2 knockout mice have multiple developmental defects that are non-overlapping with other TGFbeta knockout phenotypes. Development (Cambridge, England).

Santini, M. P., Forte, E., Harvey, R. P., & Kovacic, J. C. (2016). Developmental origin and lineage plasticity of endogenous cardiac stem cells. https://doi.org/10.1242/dev.111591

Sawai, M., Uchida, Y., Ohno, Y., Miyamoto, M., Nishioka, C., Itohara, S., … Kihara, A. (2017). The 3-hydroxyacyl-CoA dehydratases HACD1 and HACD2 exhibit functional redundancy and are active in a wide range of fatty acid elongation pathways. Journal of Biological Chemistry. https://doi.org/10.1074/jbc.M117.803171

Sazonova, O., Zhao, Y., Nürnberg, S., Miller, C., Pjanic, M., Castano, V. G., … Quertermous, T. (2015). Characterization of TCF21 Downstream Target Regions Identifies a Transcriptional Network Linking Multiple Independent Coronary Artery Disease Loci. PLoS Genetics, 11(5), e1005202. https://doi.org/10.1371/journal.pgen.1005202

Scarpa, E., Szabó, A., Bibonne, A., Theveneau, E., Parsons, M., & Mayor, R. (2015).

Sophie McManus – February 2020 229

BNC1 regulates human epicardial heterogeneity and function

Cadherin Switch during EMT in Neural Crest Cells Leads to Contact Inhibition of Locomotion via Repolarization of Forces. Developmental Cell, 34(4), 421–434. https://doi.org/10.1016/j.devcel.2015.06.012

Schlueter, J., & Brand, T. (2013). Subpopulation of proepicardial cells is derived from the somatic mesoderm in the chick embryo. Circulation Research. https://doi.org/10.1161/CIRCRESAHA.113.301347

Schnabel, K., Wu, C. C., Kurth, T., & Weidinger, G. (2011). Regeneration of cryoinjury induced necrotic heart lesions in zebrafish is associated with epicardial activation and cardiomyocyte proliferation. PLoS ONE. https://doi.org/10.1371/journal.pone.0018503

Schnapp, L. M., Hatch, N., Ramos, D. M., Klimanskaya, I. V., Sheppard, D., & Pytela, R. (1995). The human integrin α8β1 functions as a receptor for tenascin, fibronectin, and vitronectin. Journal of Biological Chemistry, 270(39), 23196–23202. https://doi.org/10.1074/jbc.270.39.23196

Schultheiss, T. M., Burch, J. B., & Lassar, A. B. (1997). A role for bone morphogenetic proteins in the induction of cardiac myogenesis. Genes & Development, 11(4), 451– 462. https://doi.org/10.1101/gad.11.4.451

Schultheiss, T. M., Xydas, S., & Lassar, A. B. (1995). Induction of avian cardiac myogenesis by anterior endoderm. Development (Cambridge, England).

Segura, A. M., Frazier, O. H., & Buja, L. M. (2014). Fibrosis and heart failure. Heart Failure Reviews. https://doi.org/10.1007/s10741-012-9365-4

Serluca, F. C. (2008). Development of the proepicardial organ in the zebrafish. Developmental Biology, 315(1), 18–27. https://doi.org/10.1016/j.ydbio.2007.10.007

Shankar, J., & Nabi, I. R. (2015). Actin cytoskeleton regulation of epithelial mesenchymal transition in metastatic cancer cells. PloS One, 10(3), e0119954. https://doi.org/10.1371/journal.pone.0119954

Shi, L. U., Zhou, S. S., Chen, W. B., & Xu, L. (2017). Functions of endothelin-1 in apoptosis and migration in hepatocellular carcinoma. Experimental and Therapeutic Medicine, 13(6), 3116–3122. https://doi.org/10.3892/etm.2017.4314

230 Sophie McManus - February 2020

Chapter 5: References

Shi, Z., & Rockey, D. C. (2017). Upregulation of the actin cytoskeleton via myocardin leads to increased expression of type 1 collagen. Laboratory Investigation, 97(12), 1412–1426. https://doi.org/10.1038/labinvest.2017.96

Sissman, N. J. (1970). Developmental landmarks in cardiac morphogenesis: Comparative chronology. The American Journal of Cardiology, 25(2), 141–148. https://doi.org/10.1016/0002-9149(70)90575-8

Smart, N., Bollini, S., Dubé, K. N., Vieira, J. M., Zhou, B., Davidson, S., … Riley, P. R. (2011). De novo cardiomyocytes from within the activated adult heart after injury. Nature, 474(7353), 640–644. https://doi.org/10.1038/nature10188

Smart, N., Risebro, C. A., Melville, A. A. D., Moses, K., Schwartz, R. J., Chien, K. R., & Riley, P. R. (2007). Thymosin β4 induces adult epicardial progenitor mobilization and neovascularization. Nature. https://doi.org/10.1038/nature05383

Smith, C. L., Baek, S. T., Sung, C. Y., & Tallquist, M. D. (2011). Epicardial-derived cell epithelial-to-mesenchymal transition and fate specification require PDGF receptor signaling. Circulation Research. https://doi.org/10.1161/CIRCRESAHA.110.235531

Smits, A. M., Dronkers, E., & Goumans, M. J. (2018). The epicardium as a source of multipotent adult cardiac progenitor cells: Their origin, role and fate. Pharmacological Research. https://doi.org/10.1016/j.phrs.2017.07.020

Solanas, G., Porta-de-la-Riva, M., Agusti, C., Casagolda, D., Sánchez-Aguilera, F., Larriba, M. J., … Baulida, J. (2008). E-cadherin controls β-catenin and NF-κB transcriptional activity in mesenchymal gene expression. Journal of Cell Science, 121(13), 2224–2234. https://doi.org/10.1242/jcs.021667

Song, K., Li, Q., Jiang, Z. Z., Guo, C. W., & Li, P. (2011). Heparan sulfate D- glucosaminyl 3-O-sulfotransferase-3B1, a novel epithelial-mesenchymal transition inducer in pancreatic cancer. Cancer Biology and Therapy, 12(5), 388–398. https://doi.org/10.4161/cbt.12.5.15957

Sridurongrit, S., Larsson, J., Schwartz, R., Ruiz-Lozano, P., & Kaartinen, V. (2008). Signaling via the Tgf-β type I receptor Alk5 in heart development. Developmental Biology. https://doi.org/10.1016/j.ydbio.2008.07.038

Sophie McManus – February 2020 231

BNC1 regulates human epicardial heterogeneity and function

Steele, A. N., Macarthur, J. W., & Woo, Y. J. (2017). Stem Cell Therapy: Healing or Hype? Circulation Research. https://doi.org/10.1161/CIRCRESAHA.117.310584

Stevens, S. M., von Gise, A., VanDusen, N., Zhou, B., & Pu, W. T. (2016). Epicardium is required for cardiac seeding by yolk sac macrophages, precursors of resident macrophages of the adult heart. Developmental Biology, 413(2), 153–159. https://doi.org/10.1016/j.ydbio.2016.03.014

Stoll, S. W., Rittié, L., Johnson, J. L., & Elder, J. T. (2012). Heparin-binding EGF-like growth factor promotes epithelial-mesenchymal transition in human keratinocytes. Journal of Investigative Dermatology, 132(9), 2148–2157. https://doi.org/10.1038/jid.2012.78

Sun, B. O., Fang, Y., Li, Z., Chen, Z., & Xiang, J. (2015). Role of cellular cytoskeleton in epithelial-mesenchymal transition process during cancer progression. Biomedical Reports, 3(5), 603–610. https://doi.org/10.3892/br.2015.494

Sweetman, D., Wagstaff, L., Cooper, O., Weijer, C., & Münsterberg, A. (2008). The migration of paraxial and lateral plate mesoderm cells emerging from the late primitive streak is controlled by different Wnt signals. BMC Developmental Biology. https://doi.org/10.1186/1471-213X-8-63

Takeichi, M., Nimura, K., Mori, M., Nakagami, H., & Kaneda, Y. (2013). The Transcription Factors Tbx18 and Wt1 Control the Epicardial Epithelial- Mesenchymal Transition through Bi-Directional Regulation of Slug in Murine Primary Epicardial Cells. PLoS ONE, 8(2), e57829. https://doi.org/10.1371/journal.pone.0057829

Takizawa, N., Ikebe, R., Ikebe, M., & Luna, E. J. (2007). Supervillin slows cell spreading by facilitating myosin II activation at the cell periphery. Journal of Cell Science, 120(21), 3792–3803. https://doi.org/10.1242/jcs.008219

Tam, P. P., Parameswaran, M., Kinder, S. J., & Weinberger, R. P. (1997). The allocation of epiblast cells to the embryonic heart and other mesodermal lineages: the role of ingression and tissue movement during gastrulation. Development (Cambridge, England).

Tandon, P., Miteva, Y. V, Kuchenbrod, L. M., Cristea, I. M., & Conlon, F. L. (2013).

232 Sophie McManus - February 2020

Chapter 5: References

Tcf21 regulates the specification and maturation of proepicardial cells. Development (Cambridge, England), 140(11), 2409–2421. https://doi.org/10.1242/dev.093385

Tang, D. D., & Gerlach, B. D. (2017, April 8). The roles and regulation of the actin cytoskeleton, intermediate filaments and microtubules in smooth muscle cell migration. Respiratory Research, Vol. 18. https://doi.org/10.1186/s12931-017- 0544-7

Tang, J., Wang, J., Huang, K., Ye, Y., Su, T., Qiao, L., … Cheng, K. (2018). Cardiac cell–integrated microneedle patch for treating myocardial infarction. Science Advances, 4(11). https://doi.org/10.1126/sciadv.aat9365

Tao, G., Miller, L. J., & Lincoln, J. (2013). Snai1 is important for avian epicardial cell transformation and motility. Developmental Dynamics, 242(6), 699–708. https://doi.org/10.1002/dvdy.23967

Tiyaboonchai, A., Mac, H., Shamsedeen, R., Mills, J. A., Kishore, S., French, D. L., & Gadue, P. (2014). Utilization of the AAVS1 safe harbor locus for hematopoietic specific transgene expression and gene knockdown in human ES cells. Stem Cell Research, 12(3), 630–637. https://doi.org/10.1016/j.scr.2014.02.004

Tomanek, R. J., Hansen, H. K., & Christensen, L. P. (2008). Temporally expressed PDGF and FGF-2 regulate embryonic coronary artery formation and growth. Arteriosclerosis, Thrombosis, and Vascular Biology, 28(7), 1237–1243. https://doi.org/10.1161/ATVBAHA.108.166454

Tomic-Canic, M., Ojeh, N., Liu, S., Stone, R. C., Chen, V., Garzon, K. I., & Pastar, I. (2016). Epithelial-mesenchymal transition in tissue repair and fibrosis. Cell and Tissue Research. https://doi.org/10.1007/s00441-016-2464-0

Trapnell, C. (2015, October 1). Defining cell types and states with single-cell genomics. Genome Research, Vol. 25, pp. 1491–1498. https://doi.org/10.1101/gr.190595.115

Trembley, M. A., Velasquez, L. S., de Mesy Bentley, K. L., & Small, E. M. (2015). Myocardin-related transcription factors control the motility of epicardium-derived cells and the maturation of coronary vessels. Development (Cambridge, England), 142(1), 21–30. https://doi.org/10.1242/dev.116418

Trembley, M. A., Velasquez, L. S., de Mesy Bentley, K. L., Small, E. M., Aguilera, K.

Sophie McManus – February 2020 233

BNC1 regulates human epicardial heterogeneity and function

Y., Brekken, R. A., … Thannickal, V. J. (2015). Myocardin-related transcription factors control the motility of epicardium-derived cells and the maturation of coronary vessels. Development (Cambridge, England), 142(1), 21–30. https://doi.org/10.1242/dev.116418

Tseng, H., & Green, H. (1992). Basonuclin: A keratinocyte protein with multiple paired zinc fingers. Cell Biology, 89, 10311–10315. Retrieved from http://www.pnas.org/content/89/21/10311.full.pdf

Tsubakihara, Y., & Moustakas, A. (2018, November 20). Epithelial-mesenchymal transition and metastasis under the control of transforming growth factor β. International Journal of Molecular Sciences, Vol. 19. https://doi.org/10.3390/ijms19113672

Ueno, S., Weidinger, G., Osugi, T., Kohn, A. D., Golob, J. L., Pabon, L., … Murry, C. E. (2007). Biphasic role for Wnt/beta-catenin signaling in cardiac specification in zebrafish and embryonic stem cells. Proceedings of the National Academy of Sciences. https://doi.org/10.1073/pnas.0702859104 van Berlo, J. H., Kanisicak, O., Maillet, M., Vagnozzi, R. J., Karch, J., Lin, S.-C. J., … Molkentin, J. D. (2014). c-kit+ cells minimally contribute cardiomyocytes to the heart. Nature, 509(7500), 337–341. https://doi.org/10.1038/nature13309 van der Linde D, Konings E E M, Slager M A, Witsenburg M, Helbing W A, Takkenberg, J. J. M., Roos-Hesselink, J. W., van der Linde, D., Konings, E. E. M., Slager, M. A., Witsenburg, M., … Roos-Hesselink, J. W. (2011). Birth Prevalence of Congenital Heart Disease Worldwide A Systematic Review and Meta-Analysis. Journal of the American College of Cardiology. https://doi.org/10.1016/j.jacc.2011.08.025 van Wijk, B., Gunst, Q. D., Moorman, A. F. M., & van den Hoff, M. J. B. (2012). Cardiac Regeneration from Activated Epicardium. PLoS ONE, 7(9), e44692. https://doi.org/10.1371/journal.pone.0044692

Vanhoutteghem, A., Delhomme, B., Hervé, F., Nondier, I., Petit, J.-M., Araki, M., … Djian, P. (2016). The importance of basonuclin 2 in adult mice and its relation to basonuclin 1. Mechanisms of Development, 140, 53–73. https://doi.org/10.1016/j.mod.2016.02.002

234 Sophie McManus - February 2020

Chapter 5: References

Vanhoutteghem, A., & Djian, P. (2006). Basonuclins 1 and 2, whose genes share a common origin, are proteins with widely different properties and functions. Proceedings of the National Academy of Sciences of the United States of America, 103(33), 12423–12428. https://doi.org/10.1073/pnas.0605086103

Vega-Hernández, M., Kovacs, A., De Langhe, S., & Ornitz, D. M. (2011). FGF10/FGFR2b signaling is essential for cardiac fibroblast development and growth of the myocardium. Development (Cambridge, England), 138(15), 3331–3340. https://doi.org/10.1242/dev.064410

Venugopal, S., Chen, M., Liao, W., Er, S. Y., Wong, W. S. F., & Ge, R. (2015). Isthmin is a novel vascular permeability inducer that functions through cell-surface GRP78- mediated Src activation. Cardiovascular Research, 107(1), 131–142. https://doi.org/10.1093/cvr/cvv142

Vicente-Steijn, R., Scherptong, R. W. C., Kruithof, B. P. T., Duim, S. N., Goumans, M. J. T. H., Wisse, L. J., … Anderson, R. (2015). Regional differences in WT-1 and Tcf21 expression during ventricular development: implications for myocardial compaction. PLOS ONE, 10(9), e0136025. https://doi.org/10.1371/journal.pone.0136025

Villarejo, A., Cortés-Cabrera, Á., Molina-Ortíz, P., Portillo, F., & Cano, A. (2014). Differential Role of Snail1 and Snail2 Zinc Fingers in E-cadherin Repression and Epithelial to Mesenchymal Transition. Journal of Biological Chemistry, 289(2), 930–941. https://doi.org/10.1074/jbc.M113.528026

Vincent, S. D., & Buckingham, M. E. (2010). How to Make a Heart. https://doi.org/10.1016/S0070-2153(10)90001-X

Viragh, S., & Challice, C. E. (1981). The origin of the epicardium and the embryonic myocardial circulation in the mouse. The Anatomical Record. https://doi.org/10.1002/ar.1092010117 von Gise, A., & Pu, W. T. (2012). Endocardial and epicardial epithelial to mesenchymal transitions in heart development and disease. Circulation Research, 110(12), 1628– 1645. https://doi.org/10.1161/CIRCRESAHA.111.259960

Wagstaff, L., Kelwick, R., Decock, J., Arnold, H., Pennington, C., Jaworski, D., &

Sophie McManus – February 2020 235

BNC1 regulates human epicardial heterogeneity and function

Edwards, D. (2010). ADAMTS15 metalloproteinase inhibits breast cancer cell migration. Breast Cancer Research, 12(S1). https://doi.org/10.1186/bcr2512

Waldo, K. L., Kumiski, D. H., Wallis, K. T., Stadt, H. A., Hutson, M. R., Platt, D. H., & Kirby, M. L. (2001). Conotruncal myocardium arises from a secondary heart field. Development (Cambridge, England).

Walker, M. G., & Volkmuth, W. (2002). Cell adhesion and matrix remodeling genes identified by co-expression analysis. Gene Function & Disease, 3(34), 109–112. https://doi.org/10.1002/gnfd.200290000

Wang, Jinhu, Cao, J., Dickson, A. L., & Poss, K. D. (2015). Epicardial regeneration is guided by cardiac outflow tract and Hedgehog signalling. Nature, 522(7555), 226– 230. https://doi.org/10.1038/nature14325

Wang, Jinhu, Karra, R., Dickson, A. L., & Poss, K. D. (2013). Fibronectin is deposited by injury-activated epicardial cells and is necessary for zebrafish heart regeneration. Developmental Biology, 382(2), 427–435. https://doi.org/10.1016/j.ydbio.2013.08.012

Wang, Jinhu, Panáková, D., Kikuchi, K., Holdway, J. E., Gemberling, M., Burris, J. S., … Poss, K. D. (2011). The regenerative capacity of zebrafish reverses cardiac failure caused by genetic cardiomyocyte depletion. Development. https://doi.org/10.1242/dev.068601

Wang, Junwen, Zhang, S., Schultz, R. M., & Tseng, H. (2006). Search for basonuclin target genes. Biochemical and Biophysical Research Communications, 348(4), 1261–1271. https://doi.org/10.1016/j.bbrc.2006.07.198

Wang, Y., Shi, J., Chai, K., Ying, X., & Zhou, B. P. (2013). The Role of Snail in EMT and Tumorigenesis. Current Cancer Drug Targets, 13(9), 963–972. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/24168186

Wang, Zhaoning, Cui, M., Shah, A. M., Ye, W., Tan, W., Min, Y.-L., … Olson, E. N. (2019). Mechanistic basis of neonatal heart regeneration revealed by transcriptome and histone modification profiling. Proceedings of the National Academy of Sciences, 201905824. https://doi.org/10.1073/pnas.1905824116

Wang, Zhi, Yu, D., Wang, M., Wang, Q., Kouznetsova, J., Yang, R., … Gong, D. W.

236 Sophie McManus - February 2020

Chapter 5: References

(2015). Elabela-Apelin Receptor Signaling Pathway is Functional in Mammalian Systems. Scientific Reports, 5. https://doi.org/10.1038/srep08170

Watt, A. J., Battle, M. A., Li, J., & Duncan, S. A. (2004). GATA4 is essential for formation of the proepicardium and regulates cardiogenesis. Proceedings of the National Academy of Sciences of the United States of America, 101(34), 12573– 12578. https://doi.org/10.1073/pnas.0400752101

Wei, B., & Jin, J. P. (2016, May 10). TNNT1, TNNT2, and TNNT3: Isoform genes, regulation, and structure-function relationships. Gene, Vol. 582, pp. 1–13. https://doi.org/10.1016/j.gene.2016.01.006

Wei, K., Serpooshan, V., Hurtado, C., Diez-Cunado, M., Zhao, M., Maruyama, S., … Ruiz-Lozano, P. (2015). Epicardial FSTL1 reconstitution regenerates the adult mammalian heart. Nature. https://doi.org/10.1038/nature15372

Weinberger, M., Simões, F. C., Patient, R., Sauka-Spengler, T., & Riley, P. R. (2018). Functional heterogeneity within the developing zebrafish epicardium. BioRxiv. https://doi.org/10.1101/460394

Wessels, A., van den Hoff, M. J. B., Adamo, R. F., Phelps, A. L., Lockhart, M. M., Sauls, K., … Burch, J. B. E. (2012). Epicardially derived fibroblasts preferentially contribute to the parietal leaflets of the atrioventricular valves in the murine heart. Developmental Biology, 366(2), 111–124. https://doi.org/10.1016/j.ydbio.2012.04.020

Wheelock, M. J., Shintani, Y., Maeda, M., Fukumoto, Y., & Johnson, K. R. (2008). Cadherin switching. Journal of Cell Science, 121(Pt 6), 727–735. https://doi.org/10.1242/jcs.000455

Wilm, B. (2005). The serosal mesothelium is a major source of smooth muscle cells of the gut vasculature. Development. https://doi.org/10.1242/dev.02141

Winter, E M, Grauss, R. W., Hogers, B., van Tuyn, J., van der Geest, R., Lie-Venema, H., … Gittenberger-de Groot, A. C. (2007). Preservation of left ventricular function and attenuation of remodeling after transplantation of human epicardium-derived cells into the infarcted mouse heart. Circulation, 116(8), 917–927. https://doi.org/10.1161/CIRCULATIONAHA.106.668178

Sophie McManus – February 2020 237

BNC1 regulates human epicardial heterogeneity and function

Winter, Elizabeth M., Van Oorschot, A. A. M., Hogers, B., Van Der Graaf, L. M., Doevendans, P. A., Poelmann, R. E., … Goumans, M. J. (2009). A new direction for cardiac regeneration therapy: Application of synergistically acting epicardium- derived cells and cardiomyocyte progenitor cells. Circulation: Heart Failure, 2(6), 643–653. https://doi.org/10.1161/CIRCHEARTFAILURE.108.843722

Witty, A. D., Mihic, A., Tam, R. Y., Fisher, S. A., Mikryukov, A., Shoichet, M. S., … Keller, G. (2014). Generation of the epicardial lineage from human pluripotent stem cells. Nature Biotechnology, 32(10), 1026–1035. https://doi.org/10.1038/nbt.3002

Wu, J., Jackson-Weaver, O., & Xu, J. (2018). The TGFβ superfamily in cardiac dysfunction. Acta Biochimica et Biophysica Sinica. https://doi.org/10.1093/abbs/gmy007

Wu, M., Smith, C. L., Hall, J. A., Lee, I., Luby-Phelps, K., & Tallquist, M. D. (2010). Epicardial Spindle Orientation Controls Cell Entry into the Myocardium. Developmental Cell, 19(1), 114–125. https://doi.org/10.1016/j.devcel.2010.06.011

Wulfkuhle, J. D., Donina, I. E., Stark, N. H., Pope, R. K., Pestonjamasp, K. N., Niswonger, M. L., & Luna, E. J. (1999). Domain analysis of supervillin, an F-actin bundling plasma membrane protein with functional nuclear localization signals. Journal of Cell Science, 112 ( Pt 13), 2125–2136. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/10362542

Xin, M., Kim, Y., Sutherland, L. B., Murakami, M., Qi, X., McAnally, J., … Olson, E. N. (2013). Hippo pathway effector Yap promotes cardiac regeneration. Proceedings of the National Academy of Sciences of the United States of America, 110(34), 13839–13844. https://doi.org/10.1073/pnas.1313192110

Xiong, Q., Hill, K. L., Li, Q., Suntharalingam, P., Mansoor, A., Wang, X., … Zhang, J. (2011). A fibrin patch-based enhanced delivery of human embryonic stem cell- derived vascular cell transplantation in a porcine model of postinfarction left ventricular remodeling. Stem Cells, 29(2), 367–375. https://doi.org/10.1002/stem.580

Xu, J., Lamouille, S., & Derynck, R. (2009). TGF-β-induced epithelial to mesenchymal transition. Cell Research, 19(2), 156–172. https://doi.org/10.1038/cr.2009.5

238 Sophie McManus - February 2020

Chapter 5: References

Yabluchanskiy, A., Ma, Y., Iyer, R. P., Hall, M. E., & Lindsey, M. L. (2013, November 1). Matrix metalloproteinase-9: Many shades of function in cardiovascular disease. Physiology, Vol. 28, pp. 391–403. https://doi.org/10.1152/physiol.00029.2013

Yagi, H., Yotsumoto, F., & Miyamoto, S. (2008). Heparin-binding epidermal growth factor-like growth factor promotes transcoelomic metastasis in ovarian cancer through epithelial-mesenchymal transition. Molecular Cancer Therapeutics, 7(10), 3441–3451. https://doi.org/10.1158/1535-7163.MCT-08-0417

Ye, L., Chang, Y. H., Xiong, Q., Zhang, P., Zhang, L., Somasundaram, P., … Zhang, J. (2014). Cardiac repair in a porcine model of acute myocardial infarction with human induced pluripotent stem cell-derived cardiovascular cells. Cell Stem Cell, 15(6), 750–761. https://doi.org/10.1016/j.stem.2014.11.009

Yu, H. B., Johnson, R., Kunarso, G., & Stanton, L. W. (2011). Coassembly of REST and its cofactors at sites of gene repression in embryonic stem cells. Genome Research, 21(8), 1284–1293. https://doi.org/10.1101/gr.114488.110

Yutzey, K. E. (2017, February 17). Cardiomyocyte proliferation. Circulation Research, Vol. 120, pp. 627–629. https://doi.org/10.1161/CIRCRESAHA.116.310058

Zamora, M., Männer, J., & Ruiz-Lozano, P. (2007). Epicardium-derived progenitor cells require beta-catenin for coronary artery formation. Proceedings of the National Academy of Sciences of the United States of America, 104(46), 18109–18114. https://doi.org/10.1073/pnas.0702415104

Zhang, D., Liu, Y., Zhang, Z., Lv, P., Liu, Y., Li, J., … Huang, H. (2018). Basonuclin 1 deficiency is a cause of primary ovarian insufficiency. Human Molecular Genetics, 27(21), 3787–3800. https://doi.org/10.1093/hmg/ddy261

Zhang, J., Zhu, W., Radisic, M., & Vunjak-Novakovic, G. (2018). Can we engineer a human cardiac patch for therapy? Circulation Research, Vol. 123, pp. 244–265. https://doi.org/10.1161/CIRCRESAHA.118.311213

Zhang, X., Chou, W., Haig-Ladewig, L., Zeng, W., Cao, W., Gerton, G., … Tseng, H. (2012). BNC1 is required for maintaining mouse spermatogenesis. Genesis (New York, N.Y. : 2000), 50(7), 517–524. https://doi.org/10.1002/dvg.22014

Zhang, X., & Tseng, H. (2007). Basonuclin-null mutation impairs homeostasis and

Sophie McManus – February 2020 239

BNC1 regulates human epicardial heterogeneity and function

wound repair in mouse corneal epithelium. PloS One, 2(10), e1087. https://doi.org/10.1371/journal.pone.0001087

Zhang, Y., Chen, M., Venugopa, S. L., Zhou, Y., Xiang, W., Li, Y. H., … Ge, R. (2011). Isthmin exerts pro-survival and death-promoting effect on endothelial cells through alphavbeta5 integrin depending on its physical state. Cell Death and Disease, 2(5). https://doi.org/10.1038/cddis.2011.37

Zhang, Z., Jiang, H., Wang, Y., & Shi, M. (2018). Heparan sulfate D-glucosamine 3-O- sulfotransferase 3B1 is a novel regulator of transforming growth factor-beta- mediated epithelial-to-mesenchymal transition and regulated by MIR-218 in nonsmall cell lung cancer. Journal of Cancer Research and Therapeutics, 14(1), 24– 29. https://doi.org/10.4103/jcrt.JCRT_659_17

Zhao, J., Cao, H., Tian, L., Huo, W., Zhai, K., Wang, P., … Ma, Y. (2016). Efficient Differentiation of TBX18 + /WT1 + Epicardial-Like Cells from Human Pluripotent Stem Cells Using Small Molecular Compounds . Stem Cells and Development. https://doi.org/10.1089/scd.2016.0208

Zhou, B., Gise, A. von, Ma, Q., Rivera-Feliciano, J., & Pu, W. T. (2008). Nkx2-5- and Isl1-expressing cardiac progenitors contribute to proepicardium. Biochemical and Biophysical Research Communications. https://doi.org/10.1016/j.bbrc.2008.08.044

Zhou, B., Honor, L. B., He, H., Ma, Q., Oh, J.-H., Butterfield, C., … Pu, W. T. (2011). Adult mouse epicardium modulates myocardial injury by secreting paracrine factors. The Journal of Clinical Investigation, 121(5), 1894–1904. https://doi.org/10.1172/JCI45529

Zhou, B., Honor, L. B., Ma, Q., Oh, J.-H., Lin, R.-Z., Melero-Martin, J. M., … Pu, W. T. (2012). Thymosin beta 4 treatment after myocardial infarction does not reprogram epicardial cells into cardiomyocytes. Journal of Molecular and Cellular Cardiology, 52(1), 43–47. https://doi.org/10.1016/j.yjmcc.2011.08.020

Zhou, B., Ma, Q., Rajagopal, S., Wu, S. M., Domian, I., Rivera-Feliciano, J., … Pu, W. T. (2008). Epicardial progenitors contribute to the cardiomyocyte lineage in the developing heart. Nature, 454(7200), 109–113. https://doi.org/10.1038/nature07060

Zhou, B., & Pu, W. T. (2011). Epicardial epithelial-to-mesenchymal transition in injured

240 Sophie McManus - February 2020

Chapter 5: References

heart Images in Cellular/Molecular Medicine. J. Cell. Mol. Med, 15(12), 2781–2783. https://doi.org/10.1111/j.1582-4934.2011.01450.x

Zhu, W., Zhang, E., Zhao, M., Chong, Z., Fan, C., Tang, Y., … Zhang, J. (2018). Regenerative potential of neonatal porcine hearts. Circulation, 138(24), 2809–2816. https://doi.org/10.1161/CIRCULATIONAHA.118.034886

Zimmermann, W. H., Melnychenko, I., Wasmeier, G., Didié, M., Naito, H., Nixdorff, U., … Eschenhagen, T. (2006). Engineered heart tissue grafts improve systolic and diastolic function in infarcted rat hearts. Nature Medicine, 12(4), 452–458. https://doi.org/10.1038/nm1394

Sophie McManus – February 2020 241

BNC1 regulates human epicardial heterogeneity and function

6. APPENDIX

6.1 Carnegie stages in human embryos compared to gestational age (days) and mouse embryonic stage (days)

Carnegie Stage Human GA (days) Mouse GA (days)

1 1 E1.0

2 2-3 E2.0

3 4-5 E3.0

4 5-6 E4.5

5 7-12 E5.0

6 13-15 E6.0

7 15-17 E7.0

8 17-19 E8.0

9 20 E9.0

10 22 E9.5

11 24 E10.0

12 28 E10.5

13 30 E11.0

14 33 E11.5

15 36 E12.0

16 40 E12.5

17 42 E13.0

18 44 E13.5

19 48 E14.0

20 52 E14.5

21 54 E15.0

242 Sophie McManus - February 2020

Chapter 6: Appendix

22 55 E15.5

23 58 E16.0

Table S1. A table to show an approximate comparison between human and mouse gestational age and Carnegie embryo staging. This comparison is taken from the website UNSW Embryology; Hill, M.A. (2019, October 4) Embryology Carnegie Stage Comparison.

6.2 Gene Ontology enrichment plot for each hPSC-epi subpopulation

Figure S1. Predicted tissue and cellular specificities of BNC1high and TCF21high cells. Results of Gene Ontology over-representation and gene expression differential analyses. Each bubble represents an over-represented GO term, the disk size being proportional to the relative enrichment. The vertical axis presents the significance of the enrichment, while the horizontal axis indicates if the term-enrichment is mostly due to genes over- expressed in BNC1high cells (negative z-scores) or in TCF21high cells (positive z- scores). Bubble colours show the mean difference of expression, for all the genes

Sophie McManus – February 2020 243

BNC1 regulates human epicardial heterogeneity and function

annotated by the GO term, between BNC1high cells (turquoise) and TCF21high cells (magenta). Figure generated by Dr Nicolas Le Novère and provided by Dr Laure Gambardella (Gambardella et al., 2019)..

6.3 RNAscope reveals heterogeneous BNC1 expression in human foetal epicardium

Figure S2. RNAscope for BNC1 in human heart. Arrowheads indicate cells with notably high BNC1 mRNA levels. Experiment performed by Irina Pshenichnaya.

244 Sophie McManus - February 2020

Chapter 6: Appendix

6.4 Double-staining for WT1 and BNC1 in human foetal epicardial explant shows both WT1/BNC1 double-positive and WT1/BNC1 single-positive cells

Figure S3. WT1 and BNC1 are both detected by immunofluorescence in epicardial explant cultures from embryonic human heart at 8 weeks GA. The pink arrowhead on each panel points toward a single BNC1-positive cell, while the blue ones points towards a BNC1 WT1-double negative cell. The other cells displayed on the images are all double- positive for WT1 and BNC1, and are indicated by asterisks. Scale bar is 20 µm. Figure from (Gambardella et al., 2019).

Sophie McManus – February 2020 245

BNC1 regulates human epicardial heterogeneity and function

6.5 PODXL and WT1 are co-expressed in epicardial cells of the human foetal heart

Figure S4. There is apparent co-expression of PODXL and WT1 in some epicardial cells in human foetal heart, shown by immunohistochemistry. Immunohistochemistry for WT1 (green) reveals that some epicardial cells are WT1-negative (empty arrowheads), while some WT1-positive cells are apparent within the myocardium. PODXL immunohistochemistry also reveals some PODXL-negative cells within the epicardial layer, and some cells within myocardium that appear to express PODXL. Combined immunohistochemistry for WT1 and PODXL shows some overlap in expression for each protein. CS, Carnegie Stage. Epi, epicardium. Myo, myocardium. Scale bars 50um.

246 Sophie McManus - February 2020

Chapter 6: Appendix

6.6 Targeting rates for each sOPTiKD vector in h9 lipofection

Targeting Total Not HET HOM HOM % % HOM % HOM vector clones targeted + R.I. targeted targeted + no R.I.

sOPTIkd_ 7 1 2 4 2 86 57 29 B2M (‘A’)

sOPTIkd_ 6 0 1 5 2 100 83 50 BNC1.1 (‘B’)

sOPTIkd_ 9 1 0 8 7 89 89 11 BNC1.2 (‘C’)

sOPTIkd_ 5 2 0 3 3 60 60 0 BNC1.4 (‘D’)

sOPTIkd_ 12 0 2 10 8 100 83 20 BNC1.5 (‘E’)

Table S2. Summary table of genotyping results for siKD clones. The above table summarises relative success rate of targeting for each siKD vector, in particular the number of homozygous clones generated that were free from off-target vector integration (R.I., random integration).

Sophie McManus – February 2020 247

BNC1 regulates human epicardial heterogeneity and function

6.7 Different BNC1 sOPTiKD vectors mediate a similar relationship between BNC1 reduction, WT1 reduction and TCF21 increase, at mRNA and protein level

Figure S5. BNC1 reduction mediates WT1 reduction and TCF21 increase in different clones via different sOPTiKD vectors. (A) Both clones ‘D’ and ‘E’ have reduced BNC1

248 Sophie McManus - February 2020

Chapter 6: Appendix

mRNA in the presence of tetracycline, concomitant with reduction in WT1 and increase in TCF21. (B) Immunocytochemistry reveals that TET-cultured 1Di and 1Ei hPSC-epi cells both have reduced WT1 (green) and increased TCF21 (red). Scale bars 50um.

Sophie McManus – February 2020 249

BNC1 regulates human epicardial heterogeneity and function

6.8 Incubation of control sOPTiKD hPSC-epi cells with tetracycline does not mediate cell death

Figure S6. Control B2M vector siKD hPSC-epi cultured under the presence of tetracycline does not exhibit cell death. Top panel shows brightfield-imaged hPSC-epi

250 Sophie McManus - February 2020

Chapter 6: Appendix

without TET, while in the bottom panel cells have been differentiated from FB50 with TET. Scale bars are 1000um.

6.9 BNC1 knockdown and selected EMT gene expression by qPCR

Figure S7. BNC1 knockdown from d5 of hPSC-epi differentiation causes no significant difference in expression of genes associated with EMT by qPCR, compared to controls.

Sophie McManus – February 2020 251

BNC1 regulates human epicardial heterogeneity and function

Each gene’s mRNA is normalised to d7 hPSC-epi; EMT was induced at d8, and samples were analysed at d1, d2 and d3 of TGFβ treatment. n=3 epi-EMT experiments for each graph. No significant difference was seen between no TET and TET-treated cells for any of the genes shown.

252 Sophie McManus - February 2020

Chapter 6: Appendix

APPENDIX II

The following is the ‘ReadMe’ file detailing the workflow for the analysis of bulk RNA sequencing data performed by Dr Nicolas Le Novère.

Content of the report archive

#############################

* README.txt this file

* fastqc/

Directory containing the results of FastQC analyses of the source FASTQ files. The HTML files can be read in any web browser.

More information on FastQC can be found at: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/

* trimmed/

Directory containing FASTQ files after processing by fastp, a software that remove bad reads and trim remaining adapter sequences. The cleaning process is documented in HTML files that can be read in any web browser.

More information on Fastp can be found at: https://academic.oup.com/bioinformatics/article/34/17/i884/5093234 https://github.com/OpenGene/fastp

* bam/

Directory containing the reads aligned to the human genome GRCh38, using HISAT2. The mapped reads are in the BAM files, and short reports on the mapping process can be found in the .hisat2 files (pure text files).

More information on HISAT2 can be found at: https://ccb.jhu.edu/software/hisat2/index.shtml

* data/

Sophie McManus – February 2020 253

BNC1 regulates human epicardial heterogeneity and function

Directory where to find the counts and associated documents.

* RNAseqQC.txt

* RNAseqQC.png

Quality control of the RNAseq results, analysing where the reads are located. Performed with the software SeqMonk. More information on SeqMonk can be found at: https://www.bioinformatics.babraham.ac.uk/projects/seqmonk/

Procedure: The RNAseq quantitation pipeline of SeqMonk was used. All mRNA isoforms were merged, and matched to the Gene names. Based on visual inspection, the libraries were found opposite-strand specific, and counting accordingly. The libraries looked very clean and no correction for DNA contamination was applied. Two quantitations were performed: Raw reads and corrected for transcript length.

The counts were loaded in R, and counts for probes matched to the same feature (i.e. gene) were aggregated. Counts wer correction for library size.

** LibrarySize.png

Image showing the number of reads mapped for each sample.

** samples.csv

File containing the total number of reads (among other things) for each samples

** Counts.csv

File containing the raw read counts for each gene.

** CPM.csv

File containing the read counts corrected for the library size (counts per million reads)

** RPK.csv

File containing the read counts corrected for the transcript length (reads per kilobase)

** RPKM.csv

File containing the reads counts corrected for the transcript length and the library size (reads per kilobase per million reads)

254 Sophie McManus - February 2020

Chapter 6: Appendix

* PCA

Principal component analysis of the sample according to the following procedure:

-> remove genes that do not have at least 10 CPM in at least one sample

-> remove genes that do not vary by at least twofold across all samples

-> Normalise gene expressions using DESeq2's rlog procedure

-> PCA using the prcomp R package

The images show the 3 main components Vs each other, in 2D or 3D. The file

PCA-rot.csv presents the gene loadings for all components.

* DESeq2

Differential gene expression analyses ran with DESeq2. More information on DESeq2 is available at https://bioconductor.org/packages/release/bioc/vignettes/DESeq2/inst/doc/DESeq2.html

DESeq-D1_D7.csv, DESeq-NO_D1.csv, DESeq-NO_D7.csv

Complete results, encompassing all the genes included in the analyses

DESeq-D1_D7-p0.01.csv, DESeq-NO_D1-p0.01.csv, DESeq-NO_D7-p0.01.csv

Annotated list of differentially expressed genes, with an adjusted p-value (FDR) inferior of 0.01.

Sophie McManus – February 2020 255