Engineered in Vitro Models of Post-Implantation Human Development to Elucidate Mechanisms of Self-Organized Fate Specification During Embryogenesis

Engineered In vitro models of post-implantation human development to elucidate mechanisms of self-organized fate specification during embryogenesis

Mukul Tewary

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Institute of Biomaterials and Biomedical Engineering University of Toronto

Engineered In vitro models of post-implantation human development to elucidate mechanisms of self-organized fate specification during embryogenesis

Mukul Tewary

Doctor of Philosophy

Institute of Biomaterials and Biomedical Engineering University of Toronto

2018 Abstract

During embryogenesis, cells in different positions of the embryo acquire different fates in a seemingly autonomous process called ‘fate-patterning’. Fundamental studies have identified important signaling molecules (morphogens) that play crucial roles in coordinating developmental fate-patterning, examples include members of the transforming growth factor beta family – like bone morphogenetic proteins (BMPs), and Nodals. However, mechanistic understanding of how these morphogens coordinate fate-patterning remains unclear. Here we aim to apply bioengineering strategies to develop an in vitro model of developmental fate-patterning and employ it to interrogate the underlying mechanisms that govern this critical process.

We first developed a robust, high-throughput platform to enable geometric-confinement of adherent cell types and employed it to screen various BMP4 supplemented defined media to identify conditions that coaxed geometrically-confined human pluripotent stem cell (hPSC) colonies to undergo peri-gastrulation-associated fate-patterning. This screen resulted in identification of defined conditions that spatially segregated compartments in the differentiating hPSC colonies expressing fate markers of trophoblast-like, primitive-streak-like, endoderm-like, mesoderm-like, and ectoderm-like tissues. Using a combination of experimental and ii computational-modelling approaches, we identified a stepwise mechanism of reaction-diffusion and positional-information underlying the observed peri-gastrulation-like fate-patterning. Here, a

BMP4-Noggin reaction-diffusion network self-organized BMP signaling gradient, and this gradient patterned peri-gastrulation-associated fates in a manner consistent with positional- information. Furthermore, we found that Nodal signaling was necessary to induce the expression of the primitive-streak compartment – the precursor of gastrulation-derived fates. Interestingly, we also observed that Nodal signaling dissected gastrulation-associated and neurulation-associated gene expression profiles in differentiating hPSC lines. Specifically, in differentiating hPSCs, upregulation of Nodal signaling was observed in cells that upregulated a gene profile associated with gastrulation whereas absence of Nodal signaling correlated with upregulation of a neurulation-associated gene profile. We hypothesized that treatment of geometrically-confined hPSC colonies with BMP4 in the absence of Nodal signaling would induce fate patterning associated with neurulation. We observed experimental results consistent with this hypothesis and identified a conserved underlying mechanism of a stepwise model of reaction-diffusion and positional-information underlying the pre-neurulation-associated patterning as well. Taken together this work provides deep insight into how morphogens regulate early developmental stages of human embryogenesis – which have been previously inaccessible for experimentation.

iii

Acknowledgments

Over the course of this PhD project, I have been incredibly fortunate to be mentored by several amazing scientists who have shaped my progress both in this project, and as an early career scientist; and to whom I am deeply indebted. First and foremost, I would like to thank my supervisor, Dr. Peter Zandstra for not only providing me with the opportunity to be trained in one of the best stem cell bioengineering labs in the world, but also for the remarkable level of support and patience that he has displayed both for me, and for this project. Peter, thank you very much for being a shining example and role model of a great scientist and a great leader; and thank you for continuing to challenge and inspire me to be better in many different aspects of my life. I will forever cherish and look back fondly at my time spent in the Zandstra lab.

I would also like to thank my committee members – Drs. Janet Rossant, Penney Gilbert, and Aaron Wheeler. Thank you for investing a large amount of your time and effort to provide me with the support and constructive scientific critique I needed, and for remaining invested in my development and progress. Your contributions have been of immense value to how this project has evolved.

I would like to convey my deep gratitude to all my remarkably talented lab-mates – Dr. Celine Bauwens, Dr. Charles Yoon, Curtis Woodford, Dr. Elia Piccinini, Dr. Emanuel Nazareth, Jennifer Ma, Joel Ostblom, Dr. Laura Prochazka, Dr. Nafees Rahman, Dr. Nika Shakiba, Nimalan Thavandiran, visiting trainees – Dominika Dziedzicka, Dr. Hirokazu Akiyama, and other colleagues from various other labs whom I have been lucky enough to be around over the course of my graduate training. Being in the company of a group of such high caliber has always motivated me to strive to be a better scientist. In addition, as I have progressed through this project, I have made what I hope are lifelong friends.

Finally, I would like to thank my family – who have encouraged and supported me in ways that are far too numerous to list. My mother, Kumudini; my aunt, Anita; my sister, Priyanka – thank you all so very much.

Table of Contents Contents

Acknowledgments...... iv

Table of Contents ...... v

List of Tables ...... xi

List of Figures ...... xii

List of Abbreviations ...... xvii

Chapter 1 Introduction ...... 1

Introduction ...... 1

1.1 Why study fate patterning during embryogenesis? ...... 1

1.1.1 Self-organization in embryogenesis ...... 1

1.1.2 A fundamental question ...... 1

1.1.3 Applications in regenerative medicine...... 2

1.2 Signaling pathways in development ...... 2

1.2.1 TGF-beta pathway ...... 3

1.2.1.1 BMPs and Nodals ...... 3

1.2.1.2 Signaling via SMAD family ...... 4

1.2.1.3 Extracellular inhibitors of BMP and Nodal signaling ...... 5

1.2.2 Wnt pathway ...... 7

1.2.3 FGF pathway ...... 8

1.3 Brief introduction of early mammalian development ...... 9

1.3.1 Pre-implantation development in mice and humans ...... 9

1.3.2 Post-implantation embryonic development ...... 11

1.3.2.1 Gastrulation ...... 14

1.3.2.2 Fate patterning in ectoderm / getting set for Neurulation ...... 15

1.3.2.3 Markers of gastrulation and pre-neurulation ...... 17

1.4 Mechanisms of Developmental fate patterning ...... 18

1.4.1 Reaction-Diffusion ...... 18

1.4.2 Positional-Information ...... 20

1.5 Bioengineering technologies to control cellular environments ...... 22

1.6 In vitro models of early mammalian development ...... 26

1.6.1 Mouse ...... 27

1.6.2 Human ...... 29

1.7 Thesis motivation, hypothesis, and approach ...... 29

1.7.1 Motivation ...... 29

1.7.2 Hypothesis...... 30

1.7.3 Project summary ...... 31

Chapter 2 Development of a robust, high-throughput micro-patterning platform ...... 32

Development of a robust high-throughput micropatterning platform ...... 33

2.1 Abstract ...... 33

2.2 Introduction ...... 34

2.3 Results ...... 35

2.3.1 A high-throughput platform for screening studies of geometrically confined cell colonies ...... 35

2.4 Discussion ...... 40

2.4.1 Platform for high-content studies with geometrically confined colonies ...... 40

Chapter 3 A stepwise model of reaction-diffusion and positional-information governs self- organized human peri-gastrulation-like patterning ...... 42

A stepwise model of Reaction-Diffusion and Positional-Information governs self- organized human peri-gastrulation-like patterning ...... 43

3.1 Abstract ...... 43

3.2 Introduction ...... 44

3.3 Results ...... 46

3.3.1 A defined high throughput assay for induction of peri-gastrulation-like patterning in human pluripotent stem cell colonies ...... 46

3.3.2 Nodal signaling is necessary for BRA expression but does not induce peri- gastrulation-like patterning ...... 52

3.3.3 BMP4-NOGGIN interaction network regulates pSMAD1 gradient self- organization...... 54

3.3.4 pSMAD1 gradient formation is colony size and BMP4 concentration dependent ...... 63

3.3.5 Peri-gastrulation-like fates arise in a manner consistent with the PI paradigm .....66

3.3.6 A two-step process of RD and PI governs peri-gastrulation-like fate patterning ..70

3.4 Discussion ...... 79

3.4.1 Experimental models to study developmental fate patterning ...... 79

3.4.2 Relevance to recent studies for peri-gastrulation associated fate patterning ...... 79

3.4.3 Positional-Information ...... 83

3.4.4 Scaling of morphogen gradients during development ...... 83

3.4.5 Identity of fate compartments in the peri-gastrulation-like platform ...... 84

3.4.6 Conclusion ...... 85

3.5 Materials and Methods ...... 85

3.5.1 Human pluripotent stem cell culture ...... 85

3.5.2 Preparation of PEG plates to micro-pattern hPSC colonies...... 86

3.5.3 Cell seeding and induction of peri-gastrulation-like fate patterning ...... 86

3.5.4 Single cell data acquisition and analysis of immunofluorescence data ...... 87

3.5.5 Quantitative PCR analysis ...... 89

3.5.6 Statistics and data analysis ...... 91

3.5.7 siRNA transfection protocol ...... 91

Chapter 4 Nodal dissects peri-gastrulation-like and pre-neurulation-like fate patterning in geometrically confined human pluripotent stem cell colonies ...... 92

vii

Nodal dissects peri-gastrulation-like and pre-neurulation-like fate patterning in geometrically confined human pluripotent stem cell colonies ...... 93

4.1 Abstract ...... 93

4.2 Introduction ...... 94

4.3 Results ...... 96

4.3.1 Screening hPSC lines on PEG plates for peri-gastrulation-like patterning response yields variable induction of the primitive-streak-like compartment ...... 96

4.3.2 Undirected differentiation of hPSC lines yields gastrulation or neurulation associated gene expression correlated with upregulated or downregulated Nodal signaling ...... 97

4.3.3 Validation of variable gene expression in EB assay ...... 101

4.3.4 Nodal dissects gastrulation versus neurulation associated gene expression profiles in geometrically-confined hPSC colonies ...... 110

4.3.5 An RD network in BMP signaling can self-organize pSMAD1 activity independent of Nodal ...... 112

4.3.6 Nodal signaling contributes to the shape of the self-organized pSMAD1 gradient ...... 115

4.3.7 Pre-neurulation-like fate patterning arises in a manner consistent with PI ...... 120

4.3.8 A stepwise model of RD and PI governs pre-neurulation-like fate patterning ....128

4.3.9 PN and NN regions give rise to definitive ectodermal fates ...... 148

4.4 Discussion ...... 152

4.4.1 Nodal signaling in differentiating hPSCs ...... 152

4.4.2 Role of Nodal in self-organization of pSMAD1 gradient ...... 152

4.4.3 RD network in BMP signaling ...... 154

4.4.4 Variability in hPSC differentiation and assay to fingerprint lineage bias ...... 156

4.4.5 Conclusions ...... 156

4.5 Materials and Methods ...... 157

4.5.1 Human Pluripotent Stem Cell Culture ...... 157

4.5.2 Preparation of PEG plates ...... 158

viii

4.5.3 Comparison between PEG plates with μCP plates ...... 158

4.5.4 Peri-gastrulation-like and pre-neurulation-like fate patterning induction ...... 159

4.5.5 Embryoid body differentiation assay ...... 160

4.5.6 CA1 Nog-/- cell line generation ...... 160

4.5.7 Quantitative PCR analysis ...... 162

4.5.8 Immunofluorescent staining, and image analysis ...... 164

Chapter 5 Conclusions and Future Directions ...... 167

Conclusions and future directions ...... 168

5.1 Summary of results ...... 168

5.2 Limitations ...... 169

5.3 Impact ...... 171

5.3.1 In vitro platform of human peri-gastrulation and pre-neurulation ...... 171

5.3.2 Unified model of RD and PI ...... 172

5.4 Future Directions ...... 173

5.4.1 Robust alternative for organoid screening ...... 173

5.4.2 Identification of emergent fate subtypes ...... 173

5.4.3 Future steps towards recapitulating germ layer fate patterning ...... 174

5.4.4 Moving into the third dimension...... 175

5.4.5 Breaking symmetry in peri-gastrulation-like model ...... 176

5.4.6 Synthetic biology techniques for validation and development-by-design ...... 177

Chapter 6 Appendices ...... 179

Appendices ...... 180

6.1 Supplementary Information – Model description ...... 180

6.1.1 Summary of data in main text and motivation of the model...... 180

6.1.2 Two-component Reaction-Diffusion system ...... 181

6.1.3 Changing Variables:...... 182 ix

6.1.4 Dose dependence in BMP4 and NOGGIN production ...... 184

6.1.5 Initial conditions of BMP4 and NOGGIN distributions in micro-patterned colonies: ...... 184

6.1.5.1 BMP4 ...... 184

6.1.5.2 NOGGIN ...... 185

6.1.6 Boundary conditions for the BMP4-NOGGIN reaction-diffusion system in micro-patterned colonies ...... 185

6.1.7 Final Reaction-Diffusion PDE ...... 186

6.1.8 Parameter choices and parameter sensitivity ...... 186

6.2 Response of Wnt inhibition in peri-gastrulation-like platform ...... 189

References ...... 194

List of Tables

Table 1-1: Markers of early post-implantation mammalian development ...... 17

Table 3-1: Antibodies used in this study...... 89

Table 3-2: Primers used in this study ...... 90

Table 4-1: hPSC lines utilized in this study ...... 100

Table 4-2: Primers employed in this study ...... 162

Table 4-3: Antibodies employed in this study ...... 165

Table 6-1: Model parameters ...... 187

List of Figures

Figure 1-1: TGF-beta signaling overview...... 5

Figure 1-2: Simplified schematic representation of canonical Wnt signaling ...... 8

Figure 1-3: Simplified schematic of RTK signaling mediated activation of MAP Kinase ...... 9

Figure 1-4: General schematic overview of pre-implantation mammalian development ...... 11

Figure 1-5: Reaction-diffusion ...... 19

Figure 1-6: Positional Information...... 21

Figure 1-7: Micropatterning technologies for ‘niche engineering’...... 24

Figure 2-1: Development of Poly(ethylene glycol) based micro-patterning platform...... 37

Figure 2-2: Characterization of PEG plates...... 39

Figure 3-1: Defined peri-gastrulation-like patterning induction in differentiating hPSC colonies...... 48

Figure 3-2: Peri-gastrulation-like fate patterning in multiple basal medium conditions...... 49

Figure 3-3: Spatial trends of CDX2, BRA, and SOX2 observed in different basal media...... 50

Figure 3-4: Nanog does not co-localize with SOX2 expression at the center of differentiating hPSC colonies...... 51

Figure 3-5 Nodal signaling is required for primitive streak specification but does not induce differentiation and fate patterning in geometrically-confined hPSC colonies...... 53

Figure 3-6 pSMAD1 gradient self-organization in differentiating colonies suggests the presence of a BMP4-NOGGIN RD network...... 56

Figure 3-7: Basal expression of BMP inhibitors in hPSCs during routine culture...... 58

Figure 3-8: Initial condition for Noggin...... 59

xii

Figure 3-9: BMP4 and NOGGIN upregulation occur in a BMP4 induction dose-dependent manner...... 60

Figure 3-10: BMP4 induced upregulation of BMP4 and NOGGIN in tested medium conditions...... 61

Figure 3-11: Controls for NOGGIN and Scramble siRNA...... 62

Figure 3-12: BMP4-NOGGIN RD model predicts pSMAD1 gradient response to colony size and BMP4 dose perturbations...... 64

Figure 3-13: Quantified radial trends of pSMAD1 activity at 24 hours after induction with varying concentrations of BMP4...... 65

Figure 3-14: Fate patterning in hPSC colonies arises in a pSMAD1 threshold dependent manner...... 68

Figure 3-15: CDX2 and BRA expression in colonies arise as a function of BMP4 dose, and induction time...... 69

Figure 3-16: High BMP4 dose in induction media recapitulates stereotypic RD-like periodic patterns in 3mm diameter hPSC colonies...... 71

Figure 3-17: RD-like patterns noted in pSMAD1 activity in 3mm colonies when differentiated with high doses of BMP4...... 73

Figure 3-18: RD-like patterns noted in BRA fate acquisition in 3mm colonies differentiated in high doses of BMP4...... 75

Figure 3-19: Analysis pipeline for extracting dominant periods of theoretically predicted distribution of free BMP4 ligands and expression patterns in experimental colonies...... 75

Figure 3-20: Low BMP4 dose in induction media rescues fate patterning in 250µm diameter colonies...... 78

Figure 3-21: Mechanism of peri-gastrulation-like fate patterning in geometrically confined hPSC colonies...... 78

xiii

Figure 3-22: Dual inhibition model does not give rise to repetitive RD-like free BMP4 distribution...... 82

Figure 4-3: Variability in peri-gastrulation-like induction observed between test hPSC lines. ... 98

Figure 4-4: Starting populations of test hPSC lines show high expression of pluripotency associated proteins...... 99

Figure 4-5: Nodal dissects gastrulation and neurulation associated gene expression profiles. .. 103

Figure 4-6: Nodal expression dynamics in FBS mediated non-specific differentiation of hPSC embryoid bodies...... 105

Figure 4-7 Hierarchical clustering of Nodal and GDF3 is consistent with unsupervised K-means clustering...... 106

Figure 4-8 MIXL1 and EOMES dynamics during EB assay predict endoderm differentiation propensity of hPSC lines...... 107

Figure 4-9: Peri-gastrulation-like assay predicts endoderm differentiation bias of hPSC lines. 109

Figure 4-10: Modulation of Nodal signaling during BMP4 treatment of geometrically-confined hPSC colonies...... 111

Figure 4-11 Interaction network between BMP4-Noggin underlies self-organization of A stepwise model of reaction-diffusion and positional information patterns peri-neurulation- associated fates...... 113

Figure 4-13: pSMAD1 gradient formation is consistent with a BMP4-Noggin RD network mediated self-organization...... 114

Figure 4-14: Nodal signaling contributes to the formation of the pSMAD1 gradient...... 117

Figure 4-15 Nodal signaling contributes to the formation of the pSMAD1 gradient...... 119

Figure 4-16: Nodal inhibition during BMP4 treatment of hPSC colonies abrogates peri- gastrulation-associated fates and induces pre-neurulation-associated fates...... 122

xiv

Figure 4-17: SOX2 and GATA3 expression is consistent with a pSMAD1 dose-dependent fate patterning...... 124

Figure 4-18: Pre-neurulation-like fates arise in a manner consistent with positional-information...... 127

Figure 4-19: GATA3 expression arises as a function of BMP4 dose and induction time...... 127

Figure 4-20: Changing shapes does not affect outside-in spatial patterning...... 130

Figure 4-21: No spatial oscillations of pSMAD1 detected when large geometrically confined hPSC colonies are treated with 50ng/ml BMP4 and SB in SR medium...... 131

Figure 4-22: Negligible spatial oscillations of pSMAD1 detected when large geometrically confined hPSC colonies are treated with 200ng/ml BMP4 and SB in SR medium...... 133

Figure 4-23: Marginal spatial oscillations of pSMAD1 detected when large geometrically confined hPSC colonies are treated with 50ng/ml BMP4 and SB in N2B27 medium...... 136

Figure 4-24: Treatment of large geometrically confined hPSC colonies with 200ng/ml BMP4 and SB in N2B27 medium results in spatial oscillations of pSMAD1...... 137

Figure 4-25: Additional replicates of immunofluorescent images demonstrating oscillatory pSMAD1 expression in the center of large geometrically confined hPSC colonies treated with 200ng/ml BMP4 and SB in N2B27 medium...... 140

Figure 4-26: No spatial oscillations of pre-neurulation-like fates detected when large geometrically confined hPSC colonies are treated with 200ng/ml BMP4 and SB in SR medium...... 142

Figure 4-27: Minor spatial oscillations of pre-neurulation-like fates detected when large geometrically confined hPSC colonies are treated with 50ng/ml BMP4 and SB in N2B27 medium...... 144

Figure 4-28: hPSC colonies of 3mm diameter induce RD-like periodic patterns of pre- neurulation-like fates when treated with 200ng/ml of BMP4 and SB in N2B27 medium...... 145

Figure 4-29: RD-like spatial oscillations of pre-neurulation-like fates detected when large geometrically confined hPSC colonies are treated with 200ng/ml BMP4 and SB in N2B27 medium...... 147

Figure 4-30: Pre-neurulation-like platform can give rise to definitive fates associated with the differentiating ectoderm...... 149

Figure 4-31: Mechanism of Nodal dependent fate patterning in the geometrically confined hPSC colonies...... 151

Figure 6-1: Response of predicted gradient formation to perturbation of model parameters ..... 188

Figure 6-2: Gradient formation predicted with pre-defined mesh sizes in Comsol...... 189

Figure 6-3: Wnt inhibition response in peri-gastrulation-like platform...... 190

Figure 6-4: Immunofluorescent staining for BRA, CDX2, and β-catenin in control colonies ... 191

Figure 6-5: Immunofluorescent staining for BRA, CDX2, and β-catenin in geometrically confined hPSC colonies treated with BMP4 and Wnt inhibitors ...... 192

xvi

List of Abbreviations

2D – 2 dimensional

3D – 3 dimensional

μCP – Micro contact printing

AP-axis – Anterior Posterior Axis

AVE – Anterior Visceral Endoderm

BMP – Bone Morphogenetic Protein

BRA – Brachyury

DVE – Distal Visceral Endoderm

DV-axis – Dorsal-Ventral axis

EB – Embryoid body

ECM – Extracellular Matrix

EGF – Epidermal Growth Factor

EMT – Epithelial to Mesenchymal Transition

Epi – Epiblast

ESC – Embryonic Stem Cell

EXE – Extraembryonic Ectoderm

FGF – Fibroblast Growth Factor

FST – Follistatin

GDF – Growth and Differentiation Factor

GDNF – Glial cell Derived Neurotropic Factor

GSK3β – Glycogen synthase kinase 3 – β

ICM – Inner Cell Mass iPSC – Induced Pluripotent Stem Cell

NC – Neural Crest

NN – Non-neural xvii

NNE – Non-neural Ectoderm

NP – Neural Plate

NPB – Neural Plate Border

PE – Primitive Endoderm

PEG – Polyethylene Glycol

PI – Positional-Information

PSC – Pluripotent Stem Cell

PN – Pre-neural

RD – Reaction-Diffusion

SHH – Sonic Hedgehog

TAZ – transcriptional coactivator with PDZ-binding motif

TBO – Toluidine Blue O

TE – Trophectoderm

TFAP2α – Transcription factor AP-2α

TGFβ – Transforming Growth Factor β

VE – Visceral Endoderm

XPS – X-ray photoelectron spectroscopy

YAP – Yes-associate Protein

xviii Chapter 1 Introduction Introduction 1.1 Why study fate patterning during embryogenesis? 1.1.1 Self-organization in embryogenesis

When a sperm fertilizes an egg, the resultant cell (the zygote) is capable of generating all of the various cell types required for successful embryogenesis(Rossant & Tam 2004). Classical fate mapping studies in developing embryos of a variety of species have established that as the zygote divides and grows from a single cell to a multi-cellular cluster, cells in different spatial locations predictably specify to different embryonic cell types (or fates) of progressively lower developmental potency in a self-organized process called fate patterning. This raises a critical question – how does fate patterning occur?

1.1.2 A fundamental question

Fate patterning during embryogenesis facilitates the establishment of the embryonic axes, the formation of the future body plan, and subsequent generation of the entire organism with appropriate spatial allocation of the organs and appendages (Wolpert 1981). The question of how an apparently homogenous cluster of cells that arises after the first few divisions of the zygote can result in the formation of an orderly embryo appropriately ‘patterned’ is a fundamental question that has profound significance for basic science. The significance of this question is even more pronounced if one were to ask this question of a developing human embryo because it gets to the heart of the question of how humans come to be. Indeed, arguments and discussions relating to how an embryo patterns fates can be traced all the way to Aristotle (Green & Sharpe 2015; Tewary & Zandstra 2018). Fundamental developmental biology studies in model organisms like the Drosophila, Xenopus, and mouse have provided much insight into the mechanisms that regulate the self-organized emergence of the developing embryo. Notably, these studies have also demonstrated that while aspects of the self-organized formation of the embryo are conserved between species, some important inter-species differences also exist. Therefore, the question of how fate patterning occurs in human development has been somewhat difficult to interpret from studies performed in other organisms. Although studies performed directly using human embryos

would provide the most reliable interpretations of the underlying rules that govern human development, these are hindered by a lack of biological material and strong ethical concerns (Ruzo & Brivanlou 2017).

1.1.3 Applications in regenerative medicine

For over half a century, since Drs. James Till and Ernest McCulloch described their fundamental properties, stem cells have been touted for their potential to regenerate diseased tissues and organs. The advent of human embryonic stem cells (hESCs) (Thomson et al. 1998), and human induced pluripotent stem cells (hiPSCs) (Takahashi et al. 2007) – together referred to as human pluripotent stem cells (hPSCs) have catalyzed an incredible push to develop strategies for the purposes of cell therapy and regenerative medicine (Robinton & Daley 2012). The capacity of hPSCs to generate any cell type of the adult human not only highlights their utility in establishing cell therapies for devastating diseases and injuries, but also allows for them to be employed as a substrate from which scientists may derive disease models and tools for drug discovery (Robinton & Daley 2012). To capitalize on these characteristics of hPSCs to advance the field of regenerative medicine, scientists have turned to the lessons learned from the field of developmental biology to identify the differentiation protocols required to derive the appropriate cell types of interest (Murry & Keller 2008). The studies of how cell fate patterning during embryogenesis occurs in mammals, which have typically been performed in mice, have provided very valuable insight that has contributed to rapid progress in the field (Murry & Keller 2008).

1.2 Signaling pathways in development

During development, stem cells in progenitor tissues receive cues from their microenvironment in the form of both biochemical and biophysical signals which results in the controlled specification and differentiation of the cells giving rise to patterned fates. Biochemical signals sensed by stem cells can include autocrine signals that are secreted and subsequently absorbed by the same cell, as well as paracrine and juxtacrine signals that are received from neighbouring or distant cells. Biophysical signals on the other hand, are mediated by cell–cell contact, and interactions with the shape, topology and compliance of extracellular matrix (ECM) proteins. The cells sense specific cues either through receptors that bind ligands, or proteins (for example, integrins) that interact 2

with the neighbouring ECM. These cues are then transmitted through the cells via a cascade of molecular signal transduction and modification events that lead to the production of new signals or result in structural changes. We briefly discuss some of the key signaling pathways and the relevant signaling molecules that are involved in early human development below.

1.2.1 TGF-beta pathway

The transforming growth factor beta (TGFβ) signaling pathway is one of the most important regulators of embryogenesis and it plays critical roles in a variety of different evolutionarily conserved developmental stages (Wharton & Derynck 2009). The so called ‘super-family’ comprises over 30 different structurally related diffusible ligands which belong to subgroups like the TGFβs, activins/inhibins, bone morphogenetic proteins/growth and differentiation factors (BMPs/GDFs), Nodals/VG1, or glial cell derived neurotropic factors (GDNFs) (Weiss & Attisano 2013). The processing of the proteins of the TGFβ superfamily occurs in a manner such that the mature peptide is present on the carboxy-terminal region of the protein. These proteins are secreted from the cells as either homodimers (dimerized with themselves) or heterodimers (dimerized with other related members of the superfamily) that have been bonded via a covalent interaction between the cysteine residues of the monomers (Gilbert 2010).

1.2.1.1 BMPs and Nodals

BMPs and Nodals are of salient importance for post-implantation mammalian development, and consequently, play a central role in this study. BMPs are so named because they were discovered due to their ability to induce the formation of bone. However, it turned out that bone formation was just one of a very large number of functions they perform during development and homeostasis. BMPs are the largest TGFβ subfamily and play an important role in regulating cell proliferation, apoptosis, cell migration, and fate specification (Gilbert 2010). They are essential during embryogenesis, most notably playing a key role in the induction of mesodermal and cardiac tissues. BMP2 and BMP4 knockout mice are embryonic lethal showing severe developmental malformations (Wang et al. 2014). BMP2 mutants appear to have defects in the amnion, the chorion and the cardiac fields (Zhang & Bradley 1996), whereas BMP4 mutants lack mesoderm entirely indicating that BMP4 plays a crucial role during gastrulation in mice (Lawson et al. 1999). BMPs are also required for the specification of the epidermis during the fate patterning that occurs

in the developing ectoderm (Wang et al. 2014). Much as BMPs, Nodals are also known to play a central role in mammalian mesodermal development. In fact, the expression of Nodal during the onset of mesoderm specification occurs due to a BMP signaling (from the extraembryonic tissue) mediated activation of Wnt which in turn induces the expression of Nodal in the posterior end of the developing embryo at the onset of gastrulation. Whereas multiple Nodals have been identified in other model organisms, for instance - the zebrafish has three Nodals (cyclops, squint, and southpaw), and six Nodals have been identified in the frogs (xNr1-6) (Shen 2007), only one Nodal ligand has been detected in mammals. However, other TGFβ ligands, such as GDF3 for instance, may have parallel functions to Nodal as evidenced by the similarity in the phenotypes between GDF3 and Nodal mutant embryos (Chen 2005).

1.2.1.2 Signaling via SMAD family

The TGFβ signaling pathway initiates when the TGFβ ligand binds a type II TGFβ receptor which then recruits a type I receptor and phosphorylates a serine or a threonine thereby activating it. The activated type I receptors can phosphorylate and activate a family of transcription factor proteins called the receptor-SMADs. Nodal signaling also requires the function of proteins called Cripto and Criptic – co-factors belonging to the EGF-CFC family (Shen 2007). SMAD1/5/8 are activated by BMP signaling whereas SMAD2/3 are activated by TGFβs and Nodals. The phosphorylated and activated receptor SMADs are translocated to the nucleus through the function of a chaperone protein SMAD4 where they bind to transcriptional co-factors and activate the expression of the context-specific target genes. A schematic representing the overview of the TGFβ signaling cascade is shown below in Figure 1-1.

Figure 1-1: TGF-beta signaling overview A simplified schematic of the TGFβ signaling cascade. The ligand binds to the type II receptor which recruits and activates the type I receptor by phosphorylating a serine or a threonine. The activated receptor can phosphorylate the receptor SMADs resulting in the binding to a chaperone protein and the subsequent translocation of the transcription factors to the nucleus. Figure adapted from Motifolio images.

1.2.1.3 Extracellular inhibitors of BMP and Nodal signaling

The ligands belonging to the TGFβ superfamily almost exclusively signal through receptor binding. Given the critical roles that these ligands play in embryogenesis, their distribution and activity are tightly regulated in the embryo. This control is enforced by a variety of inhibitors that bind to the TGFβ ligands or receptors with higher affinities than the affinities characteristic to the ligand-receptor interaction and prevent the activation of the signaling pathway. The classical antagonist of Nodals are the Lefty proteins. In mammals, at least two Lefty genes have been identified – Lefty1, and Lefty2. Although it is unclear if Leftys inhibit Nodal signaling by binding the ligand or the receptor, loss of Leftys results in an extension in the range of Nodal signaling and an excessive amount of mesodermal progenitors establishing Lefty as a Nodal antagonist (Schier 2003). In addition to Leftys, Cerberus family proteins like Cerberus and Coco are also potent 5

inhibitors of Nodal signaling. These proteins are cysteine knots that bind to the ligands directly and inhibit their interaction with the TGFβ receptors. Notably, Cerberus and Coco are also competent inhibitors of BMP ligands. For instance, Cerberus has been shown to be involved in the spatial organization of the ureteric tree during mouse kidney development by antagonizing BMP2, and BMP4 (Mulloy & Rider 2015; Chi et al. 2011), and injections of Coco mRNA in developing mouse embryos phenocopies BMP overexpression by causing defects in gastrulation and mesodermal formation (Mulloy & Rider 2015). Another TGFβ family antagonist is Follistatin that was first identified and named due to its inhibitory effect of the Follicle Stimulating Hormone (FSH), but since then its function as a potent TFGβ family member inhibitor has also been identified. Although its inhibitory effect is the highest for Activin (KD ~ 0.03-0.3 nM), it is also known to strongly interact with the BMP ligands (KD ~ 1 nM) (Rider & Mulloy 2010) and prevent the association of BMP ligands to their cognate receptors.

While the Cerberus family proteins, which are downstream of Nodal signaling, can cross-talk with the BMP signaling pathway to inhibit BMP activity, classical antagonists of BMP signaling like Noggin and Chordin also have much evidence of being able to prevent the interaction of BMP ligands with their receptors to activate BMP signaling. In mammalian embryogenesis, Chordin is secreted by the node which assists with the establishment of the anterior-posterior axis, and Noggin is secreted by the notochord that assists with the establishment of the dorsal-ventral axis. In mice, embryos missing both Noggin and Chordin lack a forebrain, nose, and other facial structures (Bachiller et al. 2000) indicating their necessity in inhibiting BMP signaling to allow for the development of the anterior fates during embryogenesis.

The proteins mentioned above do not constitute an exhaustive list of antagonists of BMP signaling that can function in the extracellular environment. Examples of other proteins that have also been shown to antagonize BMP signaling include Twisted gastrulation, Differential screening-selected gene aberrative in neuroblastoma (DAN), Gremlin – which was previously referred to as ‘Down- regulation by v-mos’ (DRM), ‘Protein related to Dan and Cerberus’ (PRDC) – which is also referred to as Gremlin2 due to its homology with Gremlin(Church et al. 2015; Hung et al. 2012; Sudo et al. 2004). Interestingly, Gremlin can antagonize BMP4 ligands in the intracellular environment as well (Sun et al. 2006). For a more comprehensive discussion of molecules that are

able to inhibit BMP signaling, we direct the readers to some excellent reviews (Mulloy & Rider 2015; Rider & Mulloy 2010).

1.2.2 Wnt pathway

The name Wnt was derived from a fusion of the Drosophila wingless, and its vertebrate homolog integrated. Much like the TGFβ ligands, Wnt ligands are also secreted out of the cell. However, upon translation, Wnt ligands undergo a post-translational-modification whereby a protein called Porcupine appends a lipid molecule onto the ligand. This modification hinders the ability of the Wnt ligands to diffuse and likely results in an increase in their concentration at the cellular membrane.

The Wnt pathway is involved in a myriad of cellular functions, and the proteins involved in the signaling cascade of the pathway also perform multiple roles that regulate cellular behaviour. A comprehensive discussion of the Wnt pathway is out of the scope of this overview. Therefore, we will briefly discuss the current understanding of the ‘canonical’ Wnt pathway, which deals with the regulation of cytoplasmic versus nuclear localization of β-catenin – the chief protein/transcription factor of interest of the canonical Wnt pathway. Wnt ligands interact with transmembrane receptors that belong to the Frizzled family of proteins to activate the signaling cascade that results in the nuclear localization of β-catenin. As β-catenin performs numerous functions in the absence of Wnt signaling, it is constantly transcribed in cells. In the absence of Wnt ligands, a core protein complex referred to as the ‘β-catenin destruction complex’ which comprises GSK3β, APC, Axin among a few other proteins, results in the ubiquitination and degradation of β-catenin thereby regulating the levels of β-catenin in the cytoplasm and preventing its translocation to the nucleus. In the presence of Wnt ligands, binding of Frizzled receptors activates a cytoplasmic protein called Dishevelled, which phosphorylates GSK3β and disassembles the β-catenin destruction complex preventing the ubiquitination of β-catenin. Consequently, β-catenin accumulates in the cytoplasm subsequently resulting in its nuclear translocation where it displaces a transcriptional repressor called Groucho and induces the expression of the Wnt target genes. A simplified overview is shown in Figure 1-2.

Figure 1-2: Simplified schematic representation of canonical Wnt signaling A simplified overview of the canonical Wnt pathway. In the absence of Wnt ligands, Dishevelled remains inactivated and the β-catenin destruction complex comprising APC, Axin, and GSK3β promotes the degradation of β-catenin. The presence of Wnt ligands, on the other hand, results in the activation of Dishevelled, which inhibits GSK3β and compromises the function of the β- catenin destruction complex resulting in the accumulation and consequent nuclear translocation of β-catenin. In the nucleus, β-catenin displaces the Wnt repressor, Groucho, and activates transcription of Wnt targets. Figure adapted from Motifolio images.

1.2.3 FGF pathway

The fibroblast growth factor (FGF) family of paracrine signaling factors consists of over twenty proteins that share structural homology. Furthermore, they can generate hundreds of isoforms via RNA splicing or by varying their initial codons in a context specific manner. Of the various FGF proteins known, FGF2, FGF7, FGF8, FGF10 play critical roles in embryogenesis (Gilbert 2010). To activate the signaling pathway, FGF ligands bind a set of receptor tyrosine kinases (RTKs) called fibroblast growth factor receptors (FGFRs) which results in their dimerization, a subsequent conformational change, and phosphorylation of intracellular kinase domains. These phosphorylated sites can function as docking sites for adapter proteins like FGFR Substrate 2 (FRS2) which results in the activation of various signaling cascades (Turner & Grose 2010). Of the many potential signaling pathways that can be activated, the Mitogen Activated Protein Kinase (MAP-Kinase), and Phosphoinositide 3-Kinase (PI3K) are critical developmentally relevant signaling pathways associated with FGF signaling. FRS2 can phosphorylate and activate son of 8

sevenless (SOS) and growth factor receptor-bound 2 (GRB2) to activate RAS and the downstream MAP Kinase pathways (depicted in Figure 1-3). On the other hand, PI3K and its downstream AKT signaling pathways, which regulate anti-apoptotic signaling can be activated via an independent complex that includes GRB2-associated binding protein 1 (GAB1) (Turner & Grose 2010).

Figure 1-3: Simplified schematic of RTK signaling mediated activation of MAP Kinase Ligand binding of the FGFR receptor results in receptor dimerization and phosphorylation of the intracellular tyrosine domains which activates the receptor and allows adaptor proteins like GRB2 to bind and activate RAS. This initiates a signaling cascade activating MAP-kinase-kinase-kinase (also called RAF), which results in the activation of MAP-kinase-kinase (also called MEK), and the subsequent activation of MAP-Kinase. Activated MAPK can phosphorylate and activate transcription factors like CREBs and MYC which play crucial roles in biological functions. Figure adapted from Motofolio images. 1.3 Brief introduction of early mammalian development

1.3.1 Pre-implantation development in mice and humans

In mammals, the early developmental progression of the zygote as it forms a tissue structure referred to as the blastocyst, is markedly different from those studied in other model organisms. For instance, the cleavage (mitotic cell divisions that produce smaller nucleated daughter cells) frequency of developing mammalian embryos is among the slowest in the entire animal kingdom, occurring on average about once every 12-24 hours (Doronin et al. 2016). Another prominent difference is that the cell divisions during the early development stages in mammal are not 9

synchronized. Therefore, the progression of the number of cells in the early embryo do not go from 2 to 4 to 8 and so on in an exponential manner. Instead at many points during development, the early mammalian embryo can have an odd number of blastomeres (Goolam et al. 2016). Another point that differentiates mammalian development from other animals that develop rapidly is timing of activation of the genome. Unlike animals like the Xenopus for instance, where the proteins required for cleavage are maternally deposited, and present in the cytoplasm, the mammalian embryos rapidly activate their genomes to provide the requisite proteins. In mice, the genomic activation occurs at a 2-cell stage, and in humans it is thought to occur around 4-cell to 8-cell stage embryos (Lee et al. 2014).

As the early mammalian embryo undergoes cleavage divisions resulting in the increase of the number of blastomeres present, another important developmental event that is unique to mammalian development called compaction occurs around the 8-cell stage (Iwata et al. 2014; Fleming et al. 1993). In mice, during compaction, the blastomeres upregulate the expression of tight-junction associated proteins like E-Cadherin and Cingulin, flatten, and become tightly packed into a compact structure (Iwata et al. 2014; Fleming et al. 1993). Furthermore, opposed localization of cell polarity proteins like PAR3 and PAR1 can be observed at the apical, and the basolateral regions respectively of the cells on the periphery. Consequently, the outer layer of the embryo at the compaction stage becomes a polarized epithelium (Rossant & Tam 2009). Following compaction, in mice, the polarized outer layer upregulates a key differentiation-associated transcription factor called CDX2 and specifies toward the trophectodermal (TE) fate, whereas the inner cells maintain the expression of a pluripotency associated gene OCT4 and go on to form the inner cell mass (ICM) (Rossant & Tam 2009), which is the source from which embryonic stem cells (ESCs) are derived in both mice and humans (Thomson et al. 1998). The ICM then further specifies to give rise to the Epiblast (Epi) and the primitive endoderm (PE). The Epi comprises the pluripotent stem cells and gives rise to the fetus, whereas the PE contain the progenitors of the extraembryonic yolk sack. The embryonic stage where the TE, PE and Epi tissues have segregated is referred to as the blastocyst. Although the detailed studies of the fate choices and the underlying mechanisms regulating the lineage segregation have not been performed in the human blastocyst, it is hypothesized that somewhat similar events may regulate the morphogenetic reorganization of the human embryo owing to similar morphologies and tissue organization of the embryos of the two species at the early blastocyst stage. However, there appear to be some important differences 10

as well. For instance, in human embryos, the expression of the key TE associated marker – CDX2, is delayed in the TE tissues and is upregulated only after the TE has segregated from the ICM tissues in the early blastocyst (Rossant 2015). In addition, the expression of OCT4 remains broad at the blastocyst stage in human embryos and is co-localized with CDX2 in the TE region (Rossant 2015).

A schematic representation of the sequence of morphogenetic reorganization that gives rise to the blastocyst with spatially segregated TE, PE, and Epi tissues in the mammalian embryo is depicted below in Figure 1-4.

Figure 1-4: General schematic overview of pre-implantation mammalian development An overview of pre-implantation mammalian development. The oocyte upon fertilization forms the zygote, which undergoes a series of cleavage divisions to give rise to the morula and eventually the blastocyst. The abbreviations used are: TE – Trophectoderm, ICM – Inner Cell Mass, PE – Primitive Endoderm, and Epi – Epiblast. Figure adapted from Motifolio images.

1.3.2 Post-implantation embryonic development

After forming the blastocyst that have the TE, PE, and Epi lineages segregated, the embryo implants into the uterus. Trophoblasts derived from the TE tissue are required for successful attachment to the uterine tissue and subsequent implantation of the blastocyst, which is an absolute requirement for the blastocyst to develop further. The Epi tissue in the blastocyst goes on to produce the embryo proper (Rossant & Tam 2004). The TE tissue, on the other hand, contributes to the formation of the extra-embryonic tissues like the fetal portion of the placenta called the chorion (Rossant & Tam 2009; Rossant & Tam 2004). The PE may contribute minimally to the embryo, but mostly gives rise to the yolk sac. Owing to the inaccessibility of the developing 11

embryo post-implantation, this stage has been dubbed the black box of human development. Much of what is known about the developmental events after the implantation stage, has been from embryonic specimens that were fortuitously discovered during medical procedures like hysterectomies, autopsies, and abortions. The reports from the studies on these embryos could mostly hypothesize and extrapolate cell lineages based on morphologies identified from the sectioned specimens. Studies in mice have provided far deeper insight into post-implantation mammalian embryonic development.

Following implantation, the embryo goes through a burst of proliferation in both the Epi and the extraembryonic tissues. The TE tissues proliferate and differentiate to give rise to the extraembryonic ectoderm (EXE). This proliferation and differentiation is mediated by FGF signaling from the Epi tissue. The TE cells near the Epi tissue are subjected to high levels of FGF signaling, remain proliferative, and provide a constant supply of cells for the formation of the EXE. As the cell numbers in this region become higher, some cells become displaced and migrate farther away from the TE-Epi boundary resulting in reduction of FGF signaling levels and consequently undergoing terminal differentiation (Tanaka et al. 1998). Concomitant with the proliferation in the extra-embryonic lineages, the Epi cells also increase in number rapidly during this developmental stage with the cell doubling time changing from once every 24 hours to once every 12 hours. The increase in cell division rate in the epiblast is also accompanied by morphogenetic reorganization of the tissue into an egg-shaped structure which is at times referred to as the ‘egg cylinder’. Previously, this change in morphology was thought to occur through programmed cell death at the center of the epiblast. However, recent studies have identified that this change in fact occurs through lumenogenesis instead where the epiblast undergoes morphogenetic reorganization and forms a rosette-like structure. The PE tissue which outlined the epiblast in the ICM, post- implantation also proliferates and completely envelops the epiblast and the EXE and is referred to as the visceral endoderm (VE) at this stage.

Symmetry breaking during development of many organisms occurs very early during embryogenesis. For instance, Xenopus development occurs in a manner such that the dorsal end of the embryo necessarily lies opposite the point of sperm entry (Gilbert 2010). Although how the asymmetries are perpetuated in human embryogenesis is unknown, the sperm entry point does not appear to dictate the organization of the mouse embryo (Rossant & Tam 2009). At the implantation

stage, however, the ICM and the initial thickening of the polar trophectoderm are visibly asymmetrically tilted (Rossant & Tam 2009). By embryonic day (E)-6, the alignment of this tilt consistently parallels the anterior posterior (AP) axis (the line extending from the head/mouth to the posterior end), however, the tilt surprisingly does not predict the axis polarity (Rossant & Tam 2009). The definitive clue predictive of the polarity of the AP axis can be found in the VE instead. At around E-5.25, a subset of VE cells at the distal end acquire an epithelial morphology that differs from the rest of the VE tissue that encapsulates the epiblast and the EXE resulting in a local thickening of the area (Gilbert 2010). This region is referred to as the distal visceral endoderm and expresses Nodal inhibitors like Cer1, Lefty1, Follistatin (FST). The DVE then migrates proximally, and the direction in which the tissue migrates becomes the anterior end of the AP axis, and by E-6, the (now) Anterior Visceral Endoderm (AVE) localizes at the anterior end and inhibits Nodal signaling mediated by the expression of Nodal inhibitors discussed above. In developing human embryos, a tissue that displays the characteristic thickening of the DVE/AVE, called the prechordal plate, is also observed. This tissue may be involved in establishing the AP axis during human development by expressing Nodal inhibitors much like the AVE. However detailed studies to characterize the function and nature of the prechordal plate have not been done.

Once the AP axis has been established in the post implantation embryo, the salient developmental event that occurs is called gastrulation, which has anecdotally been described as the most important stage in human development by the eminent scientist Lewis Wolpert. Gastrulation orchestrates the reorganization of the pluripotent Epi tissue into a tri-laminar structure of the three multipotent germ layers – the ectoderm, the mesoderm, and the endoderm. Broadly speaking, these germ layers give rise to different organ systems in the body. The ectoderm produces organs like the central nervous system, the epidermis, the eyes etc., the mesoderm is the progenitor of organs like the heart, the kidneys, the skeletal muscle, and the endoderm generates the intestinal tract, the pancreas, the liver etc. Closely following gastrulation, another critical developmental event called neurulation initiates which results in the morphogenetic reorganization of the developing ectoderm and spatially patterns the fates that give rise to the central nervous system, the epidermis, the neural crest, and the craniofacial placodes. Key stages that the embryo progresses through during gastrulation and neurulation are briefly described below.

1.3.2.1 Gastrulation

Gastrulation begins at the posterior end of the developing epiblast at around E-6.5 in mice and after day 14 in humans. Cells at the posterior end undergo an epithelial to mesenchymal transition (EMT) whereby they downregulate epithelial associated proteins like E-Cadherin and upregulate mesenchymal markers like Snai1 and start to migrate as single cells. These mesenchymal single cells ingress through the primitive streak and specify into the mesoderm and endoderm lineages, whereas the epiblast cells that escape the EMT transition specify to the ectodermal fates (Tam & Loebel 2007). Molecular analysis of the primitive streak in mice has identified a variety of different markers, some of which are globally present throughout the primitive streak (like Brachyury – BRA, MIXL1), whereas the expression of a few other markers is restricted to either the anterior (like FOXA2, and Goosecoid) or the posterior regions (like EVX1 and HOXB1) (Murry & Keller 2008). The epiblast cells at the very posterior end specify toward the extra embryonic mesoderm and contribute to the fetal tissues in the amnion (Kinder et al. 1999), and the cells from this region of the embryo also contribute to the formation of the germ cells (Ohinata et al. 2009). The cells marginally anterior of the regions that contribute to the extraembryonic and germ cell fates, ingress into the primitive streak and specify toward mesodermal lineages, whereas the cells that ingress toward the anterior end of the embryo specify toward the endodermal lineages (Tam & Loebel 2007; Rossant & Tam 2004; Murry & Keller 2008). Taken together, these morphogenetic changes and cell movements in the pluripotent epiblast result in the formation of a tri-laminar structure of the multipotent germ layers – the ectoderm, mesoderm, and endoderm spatially segregated.

From the point of view of the signaling pathways necessary, studies in a variety of systems have demonstrated that gastrulation begins due to the establishment of a BMP signaling gradient at the posterior end of the developing epiblast (Ohinata et al. 2009; Inomata et al. 2013). In the mammalian system, BMP signaling from the EXE is absolutely needed for the induction of the germ cell fate (Ohinata et al. 2009). BMP signaling from the EXE also induces the expression of Wnt in the epiblast cells which is imperative for the formation of the primitive streak and induction of the EMT in the epiblast cells (Liu et al. 1999). A signaling feedback in mice has been identified whereby BMP signals from the EXE induce Wnt activity in the epiblast cells which also upregulates Nodal expression locally, and Nodal signal in turn induces BMP signaling in the EXE tissue (Shen 2007). Much like Wnt, Nodal activity is also absolutely required for the formation of

the mesendodermal fates where embryos that lack Nodal signaling do not form the primitive streak (Schier 2003). Nodal signaling from the epiblast has also been implicated in the induction of the cavity in the developing extraembryonic compartment prior to the onset of gastrulation (Harrison et al. 2017), indicating that endogenous levels of Nodal are present prior to the activation of Nodal through the BMP – Wnt – Nodal access at the onset of gastrulation. Studies have also implicated a crucial requirement of FGF signaling in the activation of Wnt in the cells that undergo EMT in the epiblast. In Fgfr1-/- mutant mice, the induction of EMT is compromised, which prevents the cells from ingressing through the primitive streak and translocating under the epiblast to give rise to the mesendodermal cells (Ciruna & Rossant 2001). The key effector of Wnt signaling - β- catenin, closely interacts with proteins like E-cadherin that are present in epithelial cells. A likely explanation of the phenotype of the Fgfr1-/- mice is that the epiblast cells fail to downregulate E- CAD, which prevents the nuclear translocation of β-catenin and thereby prevents the activation of EMT genes like Snai1. Fgfr1-/- embryos also have reduced expression of the T-box genes Tbx6 and Brachyury (BRA), which affects the movement of the mesoderm (Ciruna & Rossant 2001; Tam & Loebel 2007). Taken together, the onset of gastrulation in the posterior end requires the activity of BMP, Wnt, Nodal, and FGF. High levels of these signals are mostly concentrated on the posterior end of the AP axis because of the function of the inhibitors of the signaling pathways being secreted by the AVE, though Wnt and Nodal signaling is present at low levels on the anterior region of the primitive streak.

1.3.2.2 Fate patterning in ectoderm / getting set for Neurulation

The process of neurulation begins after the three germ layers have been spatially segregated into a tri-laminar structure, and results in the restructuring of the overlaying ectoderm into the epidermis, the central nervous system, and the neural crest. In the anterior regions, the craniofacial placodes are also specified during this developmental stage and share a common progenitor tissue with the neural crest. The developing ectoderm is subjected to a BMP signaling gradient along the medial- lateral axis mirrored along either side of the center of the ectoderm. These BMP signaling gradients, together with the cues from the cells of the dorsal mesoderm underlying the ectodermal sheet, induce ectodermal cells located centrally to elongate, which distinguishes them from the more peripheral cells located in the germ layer (Plouhinec et al. 2017; Gilbert 2010). The elongated cells comprise the neural plate (NP) which develops to give rise to the neural tube – the progenitor

of the organism’s central nervous system (Nikolopoulou et al. 2017). The region on the peripheral edge of the ectoderm contains cells that do not display the elongated morphology and is referred to as the non-neural ectoderm (NNE) which gives rise to the epidermis (Plouhinec et al. 2017). The region that is flanked by the neural plate and the non-neural ectoderm is referred to as the neural plate border (NPB) and is a multipotent progenitor of the neural crest and the craniofacial placodes. The neural crest is at times referred to as the fourth germ layer as it contributes to a wide variety of embryonic tissues like the sensory and autonomic ganglia, the cartilage and bone of the face and the pigment cells of the skin (Melanocytes). The craniofacial placodes are key contributors to the development of the sensory systems (Plouhinec et al. 2017; Simoes-Costa & Bronner 2015). The BMP signaling gradient induces the expression of key NNE markers like GATA3 at the lateral end of the developing ectoderm whereas the pre-neural tissue maintains the expression of SOX2 and in time upregulates SOX1 and PAX6 which are markers of the definitive NP progenitors. Furthermore, markers like transcription factor AP2-alpha (TFAP2A) are expressed in the NNE, and maturing NPB region that marks the NC fate. The NPB region destined to give rise to the panplacodal competent tissues express markers like SIX1, and SIX3(Groves & LaBonne 2014).

Once the neural-plate, the neural-plate border, and the non-neural ectoderm have been specified by the BMP gradient across the developing ectoderm, the developmental stage of neurulation starts whereby the embryo gives rise to the mature ectodermal fates (the central nervous system, the epidermis, the neural crest, and the placodes). The neural plate becomes narrower and elongates along the anterior-posterior axis via an evolutionarily conserved mechanism called convergent extension. During convergent extension the neural plate tissue reduces the number of layers of cells present dramatically by intercalating the cells present in the tissue (Gilbert 2010). The process of neurulation is a complex and an intricate process, that results in the closure of the neural plate into the neural tube, and an epithelial to mesenchymal transition at the dorsal aspect of the closing tube that gives rise to the neural crest. We will not delve into the details of this developmental stage, but direct the readers to some excellent reviews that have covered the topic thoroughly (Nikolopoulou et al. 2017; Greene & Copp 2009; Greene & Copp 2014).

1.3.2.3 Markers of gastrulation and pre-neurulation

The markers of gastrulation and pre-neurulation developmental stages are provided in Table 1-1 below:

Table 1-1: Markers of early post-implantation mammalian development

GASTRULATION PRE_NEURULATION

Fate/Identity Marker Fate/Identity Marker

Epiblast OCT3/4 Neural Plate PAX6

Epiblast SOX2 Neural Plate Border PAX3

Epiblast NANOG Neural Plate Border PAX3

Primitive Streak Brachyury(BRA) Neural Plate Border TFAP2A

Primitive Streak MIXL1 Neural Plate Border SIX1

Mesoderm EOMES Non-Neural Ectoderm TFAP2A

Mesoderm KDR Non-Neural Ectoderm GATA3

Mesoderm HAND1 Non-Neural Ectoderm TROMA1

Mesoderm EVX1 Non-Neural Ectoderm DLX5

Mesoderm GATA4 Neural Crest SOX10

Endoderm SOX17 Neural Crest MYB

Endoderm FOXA2 Pre-Placodal Region SIX1

Endoderm GATA6 Pre-Placodal Region SIX3

1.4 Mechanisms of Developmental fate patterning

Due to the fundamental significance of the question, how embryos pattern fates in a self-organized manner has been studied with great interest for ages. Consequently, numerous theoretical models have been proposed that could explain the empirical and experimental observations. Of the various propositions, a select few have stood the tests of time, and of increasingly thorough experimental data which need to be explained to validate the proposed models. The three models that have undoubtedly contributed the most to our understanding of developmental biology are: Reaction- Diffusion, Positional Information, and Clock-and-Wavefront. The clock and wavefront model provides a conceptual framework that explains many aspects of the developing somites (Cooke & Zeeman 1976; Gomez et al. 2008). A discussion of this model is out of the scope of the introduction to this thesis but has been covered in detail in many excellent reviews (Pourquié 2004; Tajbakhsh & Spörle 1998). A brief introduction to the concepts of reaction-diffusion and positional- information is provided below.

1.4.1 Reaction-Diffusion

In his seminal paper ‘The chemical basis of morphogenesis’, Alan Turing posited the idea of reaction-diffusion (Turing 1952). In this theoretical paper, he focused on the idea of chemical morphogenesis rather than physical morphogenesis – meaning that in the proposed model, he ignored cell movements and changes in the shapes of tissues that occur during embryogenesis (Green & Sharpe 2015). In keeping with the idea of chemical morphogenesis, he introduced the concept of a morphogen - a chemical agent capable of inducing morphogenesis. He next sought to identify a theoretical framework that could predictably describe the spatiotemporal distribution of the spontaneous self-organization of this morphogen. He hypothesized the presence of an interaction network between the morphogen and its inhibitor – which have been referred to as the ‘activator, and inhibitor’ pair in later interpretations of the model (Gierer & Meinhardt 1972). Both these molecules were hypothesized to be extracellular, and diffusible, albeit had unequal characteristic diffusivities (inhibitor diffuses faster than activator). Next, he sought to describe a 18

set of differential equations that could describe the self-organized spatiotemporal evolution of the morphogen distribution depending on a set of initial, and boundary conditions, and worked out a remarkably elegant pair of partial differential equations (PDEs) that were able to generate complex spatial patterns of the activator. The proposed set of solutions, in fact, resulted in six different pattern types including travelling waves, oscillations and stable patterns of spots or stripes. These patterns have been reviewed in detail by Kondo et al (Kondo & Miura 2010). It is important to point out that in its inception, the reaction-diffusion model was aimed at proposing a paradigm by which a spatial pattern for two fates could arise in a self-organized manner. These fates were associated with the presence and the absence of the morphogen.

Figure 1-5: Reaction-diffusion In this model of self-organized developmental fate patterning, a morphogen is hypothesized to induce the expression of itself and its inhibitor. Here we represent ‘A’ as the morphogen, and ‘B’ as its inhibitor as depicted in the interaction cartoon. Assuming A, and B diffuse as differential diffusivities (DA, and DB), the spatiotemporal dynamics of the distribution of A can be mathematically solved for a given set of initial and boundary conditions and can generate a variety of complex patterns. The PDE formulation is described above where F, and G represent the production terms of A, and B; and dA, and dB are the degradation terms. An example of how RD- mediated self-organization can give rise to patterned fates (stripes shown as an example) is shown in the cartoon at the bottom.

The classical interpretations of the RD model as provided by Turing (Turing 1952), and Gierer and Meinhardt (Gierer & Meinhardt 1972) focused on RD interaction network that contained two nodes – the activator, and the inhibitor. From a qualitative perspective, the interaction network facilitated a critical requirement of the RD-paradigm: inducing a short-range activation, and a long-range inhibition of downstream signaling. From a quantitative, mathematical perspective, achieving the short-range activation and long-range inhibition based on a two-node network required the diffusivity of the inhibitor to be far greater (25 to 50-fold) than that of the activator (Figure 1-5). Indeed, the diffusivities of classical RD-like pairs have even been tested and shown to be consistent with the theoretical requirements like in the case of Nodal and Lefty(Muller et al. 2012). However, recent studies have identified that the interaction networks that can give rise to the classical RD-like periodic patterns can also be achieved by far more complex network topologies that contain multiple nodes, some of which may even be proteins that are non-diffusible (Raspopovic et al. 2014; Marcon et al. 2016).

1.4.2 Positional-Information

In a series of theoretical reviews in the late 1960s early 1970s, Lewis Wolpert proposed the concept of Positional Information (PI) (Wolpert 1969). Notably, Wolpert’s concept of PI was not intended to address the concept of symmetry breaking, or of how morphogens could self-organize in a developing embryo. Instead, Wolpert addressed the question of how developmental fates could pattern based on prior asymmetries present in the embryo. His proposition was that the prior asymmetries would result in an uneven distribution of morphogens across developing tissues. These asymmetrical distributions were suggested to be sufficiently gradual such that the cells within the developing tissue would be able to reliably detect the concentration of the morphogens in their immediate vicinity. Furthermore, the morphogen was thought to be capable of inducing different fates in a concentration dependent manner. For instance, high, intermediate, and low levels of the morphogen would induce three different fates in the progenitor cell type. Consequently, a graded distribution of the morphogen would be capable of inducing different fates within an initially homogenous field of progenitor tissue. This model has also been referred to as

the ‘French Flag model’ to describe the simplest case of a single morphogen diffusing over a field to pattern multiple fates (Figure 1-6).

Figure 1-6: Positional Information A developing tissue that is subjected to an asymmetrical distribution of a signaling molecule (morphogen) can sense different concentrations of the molecule. This results in a concentration dependent differential response within regions of the developing tissue resulting in differential fate acquisition. Here the letters A, B, and C represent different fates acquired in subsections of the developing tissue. The PI model is also referred to as the ‘French Flag Model’.

Notably, in the interpretation of PI described above, the fates are acquired seemingly under the control of a single morphogen. This does not necessarily need to be the case, however. Multiple morphogens can diffuse over the ‘field’ (a developing tissue) such that each position is subjected to different composite set of cues. As an example of this, one may consider the Bicoid and Caudal gradients in the developing Drosophila where the Bicoid gradient spreads so that its highest and the anterior end of the developing syncytium and the Caudal gradient becomes established such that the highest Caudal levels are found at the caudal/posterior end, as the name of the protein would suggest. Furthermore, the classical interpretation of PI described above does not account for time in the morphogen concentration dependent fate acquisition. Studies using the developing neural tube as a model system tested the involvement of time in the patterning of developmental fates and identified an integral involvement (Dessaud et al. 2008; Dessaud et al. 2007). These studies noted that fates corresponding to high doses of a morphogen manifested after long durations of exposure. Consequently, the PI-paradigm has been updated to include time as an integral component along with the concentration of the morphogen in imparting positional cues to cells located at different positions in developmental tissues.

1.5 Bioengineering technologies to control cellular environments

As evidenced from the controlled series of fate specifications and morphogenetic reorganization that occurs during embryogenesis, developmental tissues reside in highly controlled microenvironments. In the case of stem cells, these microenvironments are referred to as ‘niches’. We define microenvironment as the milieu that encapsulates biochemical and biomechanical cues sensed by developing tissues. Biochemical cues can include autocrine and paracrine signaling, and signals from the surrounding ECM depending on its composition, shape, compliance (stiffness) etc. contribute to the biophysical cues. These cues function in a concerted manner to guide the controlled development of tissues during embryogenesis.

In vitro studies aimed at understanding the underlying rules that control development employ stem cells as a model system to understand embryogenesis. Of note, these studies come with an important caveat. In sharp contrast to the high level of control present during embryonic development and in vivo in general, approaches to conduct in vitro experiments rely on techniques that suffer from poor regulation of cellular microenvironments. These studies typically employed ECM coated glass slides or tissue culture polystyrene plates onto which stem cells are seeded, exposed to test conditions, and evaluated for the output of interest. Experimental conditions such as these result in a highly heterogenous cellular microenvironments and have been known to result in dramatic variations in biological responses (Snijder et al. 2009; Nazareth et al. 2013). Consequently, bioengineers have sought to develop techniques that move closer toward approximating the in vivo environment that the standard tissue culture strategies and employ these technologies to understand the rules that govern stem cell responses. These technologies span a wider breadth of research interests than understanding embryogenesis. However, they can be very valuable in informing approaches employed for modelling developmental events in vitro. Therefore, we discuss bioengineering technologies that have been developed to probe general stem cell biology below.

Given the complex nature of cell-microenvironment interactions, bioengineers have taken the approach of studying the individual components of the microenvironment independently using niche engineering technologies. Niche engineering comprises a suite of bioengineering tools and technologies ranging from microfabrication, tissue engineering and biomaterials to polymer chemistry. These tools enable querying the design principles that regulate cell fate by perturbing 22

specific biological parameters. Examples of these parameters include the shape and size of a single cell, to controlled cell–cell communication with their neighbours, to cell fate responses to changing stiffness of the surrounding ECM.

Approaches to de-convolve the various potential interactions that cells can experience strive to separate individual stem cells from the rest of the population and environment such that the responses to the experimental conditions tested are devoid of any confounding factors. Micropatterning (Figure 1-7) represents one technology that readily enables control of stem cell interactions by defining colony geometry and environmental factors and has therefore been used by stem cell bioengineers for many years. One of the earliest studies that employed the use of micropatterning to probe stem cell fate choices investigated the effect of cell size on the differentiation of epidermal stem cells (Watt et al. 1988). The authors found that single epidermal stem cells or keratinocytes maintained stemness when patterned in large islands where the cells were able to spread, whereas cells confined to smaller micropatterned areas increased the expression of differentiation-associated proteins (Watt et al. 1988). A similar approach was subsequently taken to investigate the effect of cell size in differentiating human mesenchymal stromal cell (hMSCs) populations (McBeath et al. 2004). Differentiating hMSCs that were allowed to spread in large micropatterned ‘islands’ acquired an osteoblast fate whereas those that were restricted to smaller islands acquired an adipocyte cell fate, and RhoA activity was attributed to orchestrating these responses (McBeath et al. 2004). Later studies have demonstrated that the involvement of Yes-associate Protein (YAP) and the transcriptional coactivator with PDZ-binding motif (TAZ) as the nuclear effectors of mechano-transduction-associated cell fate decisions (Dupont et al. 2011). Taken together, these studies employed bioengineered platforms to reveal the critical importance of biomechanical signalling in maintaining and regulating cell fate outcomes in populations of progenitor cells.

Figure 1-7: Micropatterning technologies for ‘niche engineering’ Typical techniques to achieve micropatterned colonies. (i) Micro-contact printing. In this technique, a stamp is developed that has posts with the desired patterns. These stamps can be developed by generating molds which can be done through various techniques like 3D-printing, soft-lithography, etc. The posts are incubated with an ECM solution and the patterns are transferred to a tissue culture substrate. Additionally, incubating or ‘back-filling’ the substrate with polymers like polyethylene glycol (PEG) or Pluronic can prevent non-specific adhesion of cells (ii) Alternative techniques can employ the use of bio-inert polymers like PEG to passivate the entire culture substrate and then selectively induce the regions that facilitate ECM adsorption/immobilization to generate micropatterned culture surfaces.

A parallel question to the biomechanical regulation of stem cells is: how do stem cells respond to the compliance of the surrounding ECM? Identification of culture conditions to maintain muscle stem cells (MuSCs; also known as satellite cells) in vitro, which had been challenging for the field for a long period of time, represents a prominent example of how the compliance of the surrounding ECM can regulate stem cell fate. A recent study used polyethylene glycol (PEG) hydrogels with tunable stiffness to investigate the effect of substrate compliance in MuSC maintenance (Gilbert et al. 2010). The authors found that the stiffness of the substrate during in vitro culture directed the MuSCs to either self-renew or differentiate. Furthermore, they identified an intermediate level of stiffness which paralleled the stiffness of muscle tissues in vivo and was able to optimally promote self-renewal of the MuSCs. Notably, although substrate stiffness- induced effects have been demonstrated in MuSCs (Gilbert et al. 2010), it did not seem to have an

effect on the regulation of human keratinocyte cell fates (Trappmann et al. 2012). This finding suggests that biomechanical regulation mediated by substrate stiffness is specific to individual stem cell lineages. With advances being made in the field of bioengineered hydrogels, recent studies have also started to investigate the effects of matrix degradation and relaxation on stem cell fate (Chaudhuri et al. 2016; Madl et al. 2017), providing deeper insight into how dynamic biomechanical cues regulate stem cell behaviour.

Similar to biomechanical signals, biochemical signals also regulate stem cell fate. In the past, attempts to study the interactions between a specified number of cells has been hindered by our limited ability to control local signalling events. However, recent advances have been made to generate controlled cell–cell interactions using single cells that have been chemically functionalized with oligonucleotides to facilitate rapid reversible attachment to a substrate or cells coated with complementary DNA sequences (Chen et al. 2016). This technique is a valuable tool that enables careful study of juxtacrine signalling. Furthermore, in conjunction with computational models, micropatterning of mouse embryonic stem (mESCs) has allowed predictive control of the activation of endogenous JAK–STAT signalling (Peerani, Onishi, et al. 2009), providing an approach to investigate the dynamics of autocrine and paracrine signalling. Micropatterning has also been demonstrated to increase the efficiency of reversion of mouse epiblast stem cells (mEpiSCs) to mESC fates, an observation that occurs due to an enhanced responsiveness of the JAK–STAT pathway (Onishi et al. 2012). It has also been shown that increased local bone morphogenetic protein (BMP) activity, and the cross-talk between the BMP signalling pathway and the JAK–STAT pathway, underlie the increased responsiveness to the effector of JAK–STAT signalling, STAT3 (Onishi et al. 2014). This approach of investigating the niche parameters of mESCs has also been employed for human pluripotent stem cells (hPSCs), enabling identification of the effects of colony size on maintaining pluripotency or controlling differentiation trajectory (Lee et al. 2009; Peerani et al. 2007; Bauwens et al. 2008). Small colony sizes of hPSCs, which enable rapid fate switching in response to differentiation cues, can be employed to identify lineage decisions made by hPSCs in high-throughput platforms to provide deep insight into the early cell fate decisions made by hPSCs (Nazareth et al. 2013).

1.6 In vitro models of early mammalian development

Here, we define ‘early mammalian development’ as the chronologically foremost embryonic stages that occur during post-implantation mammalian development. Specifically, either gastrulation or at the latest, the onset of neurulation. There are excellent ex vivo models of the pre- somitic mesoderm development that can provide valuable information regarding the underlying mechanisms that pattern the mammalian somites (Hubaud et al. 2017; Tsiairis & Aulehla 2016). However, discussion of these models is beyond the scope of the introduction to this thesis and therefore, here, we will focus on the in vitro models that start from pluripotent stem cells and have been employed to recapitulate aspects of gastrulation and the onset of neurulation in mammalian systems.

Unequivocally, the ideal model to employ for studies aimed at mammalian embryonic development is the mammalian embryo. However, as mentioned in section 1.3, the fact that the mammalian embryo implants into the mother’s uterus to undergo the further stages of embryonic development after the establishment of the blastocyst. This makes studying the post-implantation development in embryos complicated. However, some progress has been made in terms of capturing post-implantation development by utilizing platforms that mimic the implantation event in embryos. Such a platform was first reported for mouse embryos where collagen coated polyacrylamide gels were employed to culture embryos (Bedzhov & Zernicka-Goetz 2014; Bedzhov et al. 2014). The cultured embryos attached to the collagen coated substrates and the trophoblast cells began to migrate distally from the ICM region. The ICM then appeared to acquire the stereotypical egg-cylinder morphology indicating that the embryo was able to recapitulate some aspects of post-implantation development in this in vitro platform. This platform has been of high value to the field of developmental biology. It has allowed studies into the mechanism by which the mouse epiblast (Epi) undergoes the morphological changes that are characteristic of that tissue. Specifically, the mouse Epi has a cavity at its center that was previously thought to arise due to programmed cell death at the Epi core. However, studies that employed the blastocyst culture platform revealed that the cavitation process occurred in the absence of any programmed cell death. Instead the cavitation arose due to lumenogenesis that occurred due to polarization of the epithelial cells of the epiblast into a rosette-like structure (Bedzhov & Zernicka-Goetz 2014). This platform has also been utilized to culture the embryos till the stage where gastrulation initiates

as marked by the formation of the anterior-posterior (AP) axis and the initiation of the specification of the primitive streak (Morris et al. 2012). Furthermore, it has also proved to mimic the implantation of human embryos and allow post-implantation morphogenetic restructuring of the human epiblast (Shahbazi et al. 2016; Deglincerti, Croft, et al. 2016). Overall, much progress has been made recently that can permit studying important developmental events that typically occur post-implantation in in vitro platforms. However, in the case of human embryos, global regulations require any studies to be halted after day 14 of embryonic development. This precedes the initiation of gastrulation. Therefore, all studies aimed at studying development at or post initiation of gastrulation need to be performed in in vitro platforms that employ the use of human pluripotent stem cells. We briefly discuss the recent progress been made in mammalian in vitro systems that recapitulate early post-implantation development below.

1.6.1 Mouse

One of the earliest studies that showed that mouse embryonic stem cells have the ability to self- organize in a manner that was reminiscent of post-implantation development was reported a decade ago (ten Berge et al. 2008). The authors of that study demonstrated that embryoid bodies (EBs) made with mouse Embryonic Stem Cells (mESCs) that were engineered to report the expression of a Wnt targets (Axin2 and TCF) would polarize Wnt activity within the EB. The region of local Wnt activation cells would undergo and epithelial to mesenchymal transition and specify into mesendodermal progenitors. When the authors sorted the cells in the EBs based on the reported expression (7TGP in this case), the cells expressing the Wnt reporter expressed genes associated with the posterior end of the gastrulating embryo whereas the cells that had the reporter absent expressed genes associated with the anterior end of the embryo. This indicated that EBs made of mESCs were able to self-organize in a manner that recapitulated the signaling and localized gene expression of the anterior-posterior (AP) axis. Future studies have built upon this idea and identified culture conditions that are able to spatially segregate fate markers associated with the AP axis in differentiating mESC EBs (Brink et al. 2014). Alternative 3D platforms that employ the use of fibrin gels to culture single mESC EBs for 2.5 days and then transfer the colonies to 2D culture on soft substrates have shown to induce the segregation of the germ (Poh et al. 2014), though no AP polarity is captured in models such as these. Another interesting system has been reported that much like the one from Poh et al, started from single mESCs and employed the use

of extracellular matrix (ECM) scaffold to generate a model of the neural tube (Meinhardt et al. 2014; Ranga et al. 2017). In contrast to Poh et al, instead of transferring to 2D substrates, Meinhart et al, and Ranga et al, cultured mESC aggregates till the EBs formed cyst-like structures that expressed ectodermal markers – specifically SOX1. Notably, Meinhardt et al also made a remarkable observation. They identified that a transient pulse of Retinoic Acid (RA) recapitulated aspects of the fate patterning observed along the dorsal-ventral axis of the developing neural tube. Specifically, they noted expression of FOXA2 – a marker of the floor plate localized at one end of the neural-tube-like cysts. This platform can provide a very valuable resource for studying the self- organized fate patterning of the developing neural tube in the absence of any supportive tissues like the notochord or the underlying mesenchyme. The interactions between different subpopulations in a developing embryo have been thought of as being critical for the appropriate organization and spatial allocation of developmental fates. Harrison et al opted to start from different established stem cell subtypes – specifically the trophoblast stem cells (TSCs) and embryonic stem cells (ESCs) and assembled them together in a 3D ECM (Matrigel) scaffold. Excitingly, they noted that in time the aggregate of the ESCs and the TSCs formed an organized structure that recapitulated many aspects of a developing embryo. They observed that the ESC compartment formed a rosette like structure with a lumen at the center, they also observed the formation of a lumen in the TSC compartment that depended on Nodal signaling from the developing ESCs. These ‘synthetic embryos’ notably lack any primitive endodermal tissue. Therefore, there was no anterior visceral endoderm (AVE) present which would induce the anterior-posterior axis. Interestingly, however, the authors noted the induced the expression of a primitive-streak-like tissue in a polarized manner in one end of the epiblast-like structure. During development, the extraembryonic ectoderm (the derivative of the embryonic trophoblast) expresses BMP signaling which induces Wnt and Nodal, on the posterior end of the AP axis which results in subsequent EMT and mesendodermal differentiation. These signals have been thought to localize at the posterior end due to the inhibitory function of the signals that emanate from the AVE like Leftys, Cerberus, FST, etc. The observations of Harrison et al suggest that the developing epiblast may have redundant mechanisms along with the function of the AVE to establish the AP axis in the developing embryo. These redundant mechanisms would also explain why the mESC EBs are able to polarize the expression of Wnt signaling at one end in the absence of any supporting cell types (ten Berge et al. 2008; Brink et al. 2014). Finally, some valuable

progress has also been made in starting 2D populations of mouse epiblast stem cells (mEpiSCs) – which can be derived in culture by treating mESCs with FGF2 (or basic FGF/bFGF) and ActivinA. Morgani et al have recently demonstrated that geometrically-confined mEpiSC colonies induced to differentiate in the presence of BMP4, and Wnt spatially segregated regions expressing markers of the germ layers (Morgani et al. 2017).

1.6.2 Human

Interestingly, contrary to the mESCs, where much progress has been reported in self-organization of post-implantation like tissue organization in 3D spheroids/clusters, little evidence suggests that this can be achieved when starting with 3D EBs of hPSCs. Similarly, mEpiSCs have not been induced to undergo gastrulation like organization in 3D EBs either. Given the differences between the naïve and primed pluripotent states of mESCs and mEpiSCs and the similarities between mEpiSCs and hPSCs – both of which are representative of the primed pluripotent states, it is interesting to speculate that some, as yet, uncharacterized properties of the epiblast-like state of pluripotent stem cells might explain the paucity of platforms available of 3D hPSC derived developmental models of post-implantation embryogenesis. Much like the platform reported by Morgani et al (Morgani et al. 2017), however, starting 2D populations of geometrically-confined hPSC colonies can be employed to undergo gastrulation-like fate patterning upon BMP4 treatment (Warmflash et al. 2014). Interestingly, however, studies have noted that culturing small hPSC clusters in an ECM rich environment induces the cells to undergo lumenogenesis much like the developing epiblast (Shahbazi et al. 2016). Strategies have been employed to utilize this hPSC derived cysts and subject them to differential boundary conditions of ECM presentation to develop in vitro models of human amniogenesis (Shao et al. 2016; Shao et al. 2017). Sophisticated approaches of incorporating controlled ECM signaling in the extracellular environment such as these can prove to be valuable to identify conditions that can be employed to study post- implantation human development in 3D starting populations.

1.7 Thesis motivation, hypothesis, and approach 1.7.1 Motivation

Understanding the underlying mechanisms that regulate the self-organized fate patterning that occurs during post-implantation human development can provide valuable information from a 29

basic science perspective and can also inform regenerative medicine strategies in the future. However, given the ethical concerns that cloud studies performed directly in embryos, these mechanisms need to be delineated using in vitro platforms that employ hPSCs as a model for the human epiblast, and engineer solutions that induce the appropriate signaling gradients necessary to initiate in vitro surrogates of post-implantation developmental events. This project aims to utilize bioengineering strategies to develop technologies that control hPSC colony microenvironments and utilize them to induce controlled signaling gradients of key morphogens involved in the initiation of gastrulation like BMPs.

1.7.2 Hypothesis

Advanced experimental techniques have started to verify that classical models of developmental fate patterning can explain the self-organized formation of morphogen signaling gradients. The demonstration of differential diffusivities of Nodal and Lefty underlying the formation of the Nodal signaling gradient in the developing zebrafish represents an archetypal example (Muller et al. 2012). Examples from in vitro studies that show the self-organization of Wnt signaling in mESC aggregates (ten Berge et al. 2008; Brink et al. 2014) are also consistent with the established RD network in Wnt and DKK1 (Sick et al. 2006). These signaling gradients can induce fate patterning in developmental tissues in a manner consistent with PI (Wolpert 1981). Importantly, PI can pattern developmental fates by consolidating multiple different signals, not just a single morphogen as is exemplified by the patterning of the gap genes in the developing Drosophila that occurs through the concerted function of Bicoid, Caudal, Hunchback, and Nanos gradients (Briscoe & Small 2015). Although hypothesized before (Green & Sharpe 2015), few studies have demonstrated a stepwise function of RD and PI in patterning developmental fates. The central hypothesis of this study is that in the case of hPSC colonies, a stepwise model of RD and PI can initiate developmental fate patterning. Specifically, signaling gradients that critically regulate the onset of post-implantation development in human embryos (like BMP4) can be enforced by RD, and this gradient of BMP activity can pattern developmental fates in a manner consistent with PI. In addition, changing the combinations of the complimentary morphogens (like Nodal) can result in different developmental fates in response to the signaling gradient of BMP activity.

1.7.3 Project summary

We develop a high-throughput micropatterning platform that employs deep-UV irradiation to generate robust micropatterned ECM islands that can geometrically confine a variety of different cell types (Chapter 2). We validate this platform by comparing early fate choices made by hPSCs cultured on this platform with hPSCs cultured on a platform previously using micro-contact printing (Nazareth et al. 2013). We employ our new platform to screen multiple media to identify defined conditions to induce gastrulation-associated fate patterning in geometrically confined hPSC colonies. We next identify the underlying mechanism regulating these observations as a stepwise process of RD and PI (Chapter 3). We next screen multiple different hPSC lines for their ability to induce gastrulation-associated fates in response to BMP4 treatment and note an endogenous Nodal signaling dependent switch in gastrulation-associated versus neurulation- associated gene expression profile. We exploit this knowledge to devise novel experimental conditions to capture pre-neurulation-like fate patterning in geometrically-confined hPSC colonies (Chapter 4). Finally, we suggest strategies whereby our findings can be extended to provide in depth information about the fate choices that occur in early human development and approaches that can provide immediate value to the field of regenerative medicine (Chapter 5).

Chapter 2 Development of a robust, high-throughput micro-patterning platform

A version of this chapter has been submitted for peer review by:

Mukul Tewary, Peter W. Zandstra

Attributions:

Mukul Tewary developed the platform. Dr. Peter Zandstra oversaw this work.

Development of a robust high-throughput micropatterning platform 2.1 Abstract

Herein, we present a robust, and scalable technology that permits micro-patterning of a variety of cell types in any wide range of colony shapes and sizes within high-throughput microtiter plates (96-wells). We validate this platform by comparing the differentiation capacity of human pluripotent stem cells (hPSCs) on this platform with the response observed in a previous standard called micro-contact printing (μCP). We observe that hPSCs patterned using both techniques differentiate is a similar manner (R2 >0.9). Furthermore, we demonstrate that our platform outperforms μCP in generating patterns of hPSCs as measured by two parameters – the number of colonies patterned per well of a 96-well plate and the number of cells present per colony identified. This platform represents a valuable tool that can reliably generate high-fidelity micro-patterns without compromising cellular responses like differentiation.

2.2 Introduction

In vitro models of development, both 2D and 3D, that employ the use of pluripotent stem cells (PSCs) have been known to benefit from geometric regulation of the starting stem cell population(Rahman et al. 2017; Warmflash et al. 2014; Etoc et al. 2016; Shao et al. 2017; Shao et al. 2016; Lancaster et al. 2012). In 2D models, enforcing geometric confinement of hPSC colonies enables intercellular communication to give rise to self-organized signaling gradients, capable of inducing developmentally relevant fate patterning(Etoc et al. 2016; Tewary et al. 2017). Numerous studies have started to employ geometric confinement to study developmental events, fate specification and transition in a variety of systems, underscoring the value of this methodology(Blin et al. 2017; Carlson-Stevermer et al. 2017; Morgani et al. 2017).

To further accelerate advancement, and potential translational impact, the field now needs robust, high-throughput platforms that enable geometric confinement of a variety of cell types to facilitate high-content studies. Microfabrication technologies like micro-contact printing (μCP)(Peerani, Bauwens, et al. 2009; Nazareth et al. 2013; Bauwens et al. 2008), or ones that utilize soft- lithography(Whitesides et al. 2001; Kane et al. 1999; Khademhosseini et al. 2004) have been previously used to enforce geometric confinement of hPSC colonies with some success. We have previously reported a high-throughput μCP platform that enables geometric confinement in 96- well microtiter plates(Nazareth et al. 2013). However, this technique requires a manual step of stamping the extracellular matrix (ECM) proteins to transfer the adhesive ‘islands’ onto the substrate of choice (glass, tissue culture polystyrene, etc.), which can result in variability in the patterning efficiency and fidelity between experiments, and between users. Although techniques that employ soft-lithography based protocols do not suffer from fidelity issues, they often require costly equipment and access to clean rooms, which is detrimental to their broad utility. Technologies that enable high fidelity transfer of the ECM ‘islands’, without requiring access to expensive equipment or clean rooms are of high value to the field of developmental biology. Techniques that employ the use of Deep UV (<200nm) light to photo-oxidize Polyethylene Glycol (PEG) coated substrates offer an attractive alternative to achieve this goal(Azioune et al. 2009; Azioune et al. 2010).

Here we report a robust, high-throughput platform (96-well microtiter plate) capable of enforcing geometric confinement of a variety of cell types. We validate the robustness of the platform by 34

comparing the patterning output with μCP and demonstrate improved response in both the following parameters – the number of colonies formed per well, and consistency in the number of cells patterned per well. This platform can be employed for high-content studies that rely on micropatterning the starting cell populations.

2.3 Results

2.3.1 A high-throughput platform for screening studies of geometrically confined cell colonies

Photo-oxidation of organic polymers like Polyethylene Glycol (PEG) – a widely reported bio-inert polymer(Knop et al. 2010), by Deep UV (<200nm) (DUV) light has been shown to upregulate carboxyl groups(Azioune et al. 2009; Azioune et al. 2010) which can be readily biofunctionalized with ECM proteins(Hermanson 2013). We employed this knowledge to develop a protocol to generate micropatterned, carboxyl-rich regions (Figure 2-1A). We confirmed that incubation with Poly-L-Lysine-grafted-Polyethylene Glycol (PLL-g-PEG) resulted in a PEGylated surface on plasma treated borosilicate glass coverslips by probing the carbon 1s (C1s) spectra profile using X-Ray Photoelectron Spectroscopy (XPS). Consistent with previous reports(Azioune et al. 2009), a peak indicating the presence of the C-O-C functional group present in PEG was detected at 286.6eV in the C1s spectrum on the PLL-g-PEG incubated glass coverslip, in addition to the peak at 285eV that was observed in the blank glass coverslip control (Figure 2-1B). DUV treatment of the PEGylated coverslips progressively reduced the peak at 286.6eV (Figure 2-1C) suggesting photo-oxidation mediated ablation of the PEG layer. However, we were unable to detect any carboxyl presence, which has been reported to occur at 289eV(Azioune et al. 2009). We hypothesized that the photo-oxidation of the PEG during DUV treatment reduced the polymer thickness below the detection limit of XPS, causing the absence of the carboxyl peak in the emission spectra. Given that biochemical assays circumvent the need of minimum polymer thickness to detect the presence of functional groups of interest, we opted to employ a previously reported assay based on the preferential affinity of Toluidine blue-O (TBO) to carboxyl functional groups (see Materials and Methods for assay description)(Rödiger et al. 2011) and asked if DUV treatment changed the amount of TBO adsorbed onto PEGylated coverslips. Indeed, DUV

treatment resulted in an increase in the amount of TBO adsorption on PEGylated coverslips, with relative levels increasing with exposure times up to 12 minutes after which the relative levels detected decreased (Figure 2-1D). These findings indicate that, consistent with previous reports (Théry 2010; Azioune et al. 2009), PLL-g-PEG incubation results in PEGylation of coverslips, and that the optimal exposure time to maximize the presence of carboxyl functional groups on the PEGylated coverslips in our experimental setup was 12 minutes. To produce 96-well microtiter plates for patterned cell-culture surfaces, PEGylated large coverslips (110mmx74mm) were photopatterned by DUV exposure through Quartz photo-masks for 12 minutes and assembled to bottomless 96-well plates (Figure 2-1Figure 2-1E). Carboxyl groups were activated using carbodiimide chemistry(Hermanson 2013) (Figure 2-1F) to enable covalent attachment to primary amines on ECM molecules. This “PEG plates” platform enabled robust geometrical-confinement of a variety of cell types in colonies of a variety of shapes and sizes (Figure 2-1G, Figure 2-2A- C).

Given the vital role that interactions between cells and the surrounding ECM play on cellular responses(Watt & Huck 2013), we next examined whether the covalent attachment of ECM molecules to the PEGylated coverslips in the PEG plates interfered with fate decisions of geometrically-confined hPSCs. We opted to employ a recently reported two-day assay using OCT4, and SOX2 expression as readouts to assess fate decisions in geometrically confined hPSC colonies(Nazareth et al. 2013) (Figure 2-1H), and directly compared fate acquisition of hPSCs on the PEG plates with μCP plates, a micro-patterning technique that does not require any chemical immobilization of ECM molecules. We observed a highly correlated (R2 > 0.9) differentiation response between μCP and PEG plates (Figure 2-1I,J). Furthermore, the PEG plates responded in a more reproducible manner than the μCP plates both in terms of the number of colonies achieved per well of a 96-well plate, and the number of cells attached per colony (Figure 2-2D-F). Taken together, these data demonstrate that the PEG plates enable robust geometric-confinement of cell colonies, and the differentiation response of hPSC colonies micro-patterned using the PEG plates differentiate in a highly correlated manner to those micro-patterned on μCP plates; making them a valuable platform for high-throughput screening studies for the bioengineering community.

Figure 2-1: Development of Poly(ethylene glycol) based micro-patterning platform.

A) Scheme of protocol of transferring carboxyl-rich micro-patterns onto glass coverslips. B-C) Carbon 1s (C1s) spectra acquired using X-Ray Photoelectron Spectroscopy. B) C1s spectra of glass coverslip incubated with PLL-g-PEG compared to blank glass coverslip. C) C1s spectra of PLL-g- PEG coated glass coverslips photo-exposed to Deep-UV light for different times of exposure. Dotted lines signify binding energies associated with untreated glass (285.0eV), or presence of PEG (286.6eV). D) Line plot representation of detected absorbance at 580nm wavelength of coverslips photo-oxidized for different times of exposure indicating the relative amounts of adsorbed Toluidine Blue-O (assay details in Materials and Methods). Data represented as mean (±s.d) for three technical replicates. The assay was performed once to identify optimal exposure times for our experimental setup. E) Overview of assembly procedure to produce 96-well microtiter plates with micro-patterned culture surface. F) Overview of carbodiimide based ECM protein immobilization scheme. G) Representative immunofluorescent images of micropatterned hPSCs colonies stained for OCT4, and SOX2. H) Overview of a previously described micro-patterning based hPSC differentiation assay(Nazareth et al. 2013) using OCT4 and SOX2 expression levels as indicators of early fate choices to compare PEG and μCP plates. I) Quantified compartments of early fate choices as defined in H), in both PEG, and μCP plates. The media conditions tested were ‘NS’ – Nutristem, Apel (vehicle for the following), ‘BMP’ (BMP4), ‘BA’ (BMP4+ActivinA), ‘FSB’ (bFGF+SB431542) (See Materials and Methods for concentration details). Data represented as mean (+s.d) of four independent replicates. The fate choice responses of hPSCs on both the plates were highly correlated (R2 >0.9). J) Representative immunofluorescent images of hPSC colonies stained for OCT4, and SOX2 in the different media conditions tested. Scale bars indicate 500μm.

Figure 2-2: Characterization of PEG plates. A-B) Representative images acquired on PEG plates for multiple cell types. A) Mouse embryonic fibroblasts (MEFs) stained for β-actin in green, and DAPI in blue. B) Hemogenic endothelial cells stained for VECAD in green and DAPI in blue. C-E) Comparison of patterning response on PEG plates vs μCP plates. C) Number of colonies identified per well between PEG and μCP plates. Each dot represents the number of colonies identified per well for 120 randomly chosen wells between the four replicates of PEG vs μCP plates. D) Number of cells identified per colony between PEG and μCP plates. Each dot represents the average number of cells per colony for

120 randomly chosen wells between the four replicates of PEG vs μCP plates. E) Representative images of hPSCs micropatterned in 96-well plates using PEG-based technique vs μCP.

2.4 Discussion 2.4.1 Platform for high-content studies with geometrically confined colonies

Stem cells have a remarkable, intrinsic ability to self-organize into complex, higher-order tissues. Numerous studies have exploited this capacity of stem cells to generate structures that resemble organs and developmentally-relevant tissues(Lancaster et al. 2012; Lancaster et al. 2017; Harrison et al. 2017; Eiraku et al. 2011; Gjorevski et al. 2016; Sato et al. 2009). In fact, an entire field of research has now been established that studies and develops strategies for generating self- organized organ-like 3-dimensional structures – the so-called ‘organoids’(Kretzschmar & Clevers 2016; Lancaster & Knoblich 2014). These organoids offer exciting possibilities as a substrate for screening studies of the organ or tissue of interest while maintaining aspects of the structure and organization of the native tissues which are widely accepted as being critical in eliciting outcomes that faithfully mimic in vivo responses. Although the field has taken dramatic strides towards making mature organoids, achieving a reproducible response between each organoid remains problematic. Furthermore, quantitative image analysis of immunofluorescent data from high- content organoid-based screening studies is currently challenging. An alternative approach for harnessing the potential of employing appropriately organized tissues in screening studies is to start with 2-dimensional cultures of the specific stem/progenitor cells and allow them to self- organize into tissue surrogates. Of late, numerous studies have employed this approach to derive developmentally-relevant organization in starting populations of PSCs(Tewary et al. 2017; Warmflash et al. 2014; Etoc et al. 2016; Morgani et al. 2017; Blin et al. 2017). Although this approach starts from 2-dimensional cultures, response between individual ‘organoid-like’ structures is far more reproducible. Furthermore, given that these organoid-like structures are secured in position as they undergo morphogenetic reorganization, they are far more amenable to high-content image analysis than their 3-dimensional counterparts. The high-throughput platform we report is highly robust and enables geometric confinement of a variety of cell types. Indeed, we have employed it to micro-pattern human PSCs, mouse PSCs, Retinal Pigmented Epithelial cells, human hemogenic, and pancreatic progenitors, human cardiomyocytes, mouse embryonic 40

fibroblasts (MEFs), among others. This platform is poised to be employed for high-throughput drug screens of organotypic surrogates of a variety of tissues.

Chapter 3 A stepwise model of reaction-diffusion and positional-information governs self-organized human peri-gastrulation-like patterning

A version of this chapter has been published by:

Mukul Tewary, Joel Ostblom, Laura Prochazka, Teresa Zulueta-Coarsara, Nika Shakiba, Rodrigo Fernandez-Gonzalez, Peter Zandstra

In Development (2017)

Attributions:

Image analysis was performed using the software written by Joel Ostblom. Teresa Zulueta- Coarsara analyzed the similarity between the theoretically predicted and experimentally observed periodic response associated with Reaction-Diffusion, with input from Dr. Rodrigo Fernandez- Gonzalez. Nika Shakiba assisted with qPCR analysis. Dr. Laura Prochazka assisted with knockdown experiments. All experiments and mathematical modelling were performed by Mukul Tewary. Dr. Peter Zandstra oversaw this work.

A stepwise model of Reaction-Diffusion and Positional-Information governs self-organized human peri-gastrulation-like patterning

3.1 Abstract

How position-dependent cell fate acquisition occurs during embryogenesis is a central question in developmental biology. To study this process, we developed a defined, high-throughput assay to induce peri-gastrulation-associated patterning in geometrically confined human pluripotent stem cell (hPSC) colonies. We observed that, upon BMP4 treatment, phosphorylated SMAD1 (pSMAD1) activity in the colonies organized into a radial gradient. We developed a reaction- diffusion (RD)-based computational model and observed that the self-organization of pSMAD1 signaling was consistent with the RD principle. Consequent fate acquisition occurred as a function of both pSMAD1 signaling strength and duration of induction, consistent with the positional- information (PI) paradigm. We propose that the self-organized peri-gastrulation-like fate patterning in BMP4-treated geometrically confined hPSC colonies arises via a stepwise model of RD followed by PI. This two-step model predicted experimental responses to perturbations of key parameters such as colony size and BMP4 dose. Furthermore, it also predicted experimental conditions that resulted in RD-like periodic patterning in large hPSC colonies, and rescued peri- gastrulation-like patterning in colony sizes previously thought to be reticent to this behaviour.

3.2 Introduction

During development, pluripotent stem cells (PSCs) in the epiblast are exposed to signaling gradients that initiate a sequence of fate specifications and cell movements resulting in spatial segregation of the germ layers in a developmentally conserved process called gastrulation (Rossant & Tam 2004; Rossant & Tam 2009; Morris et al. 2012; Tam et al. 2006; Ciruna & Rossant 2001). A number of model organisms have been used to study the molecular mechanisms that underpin the fate patterning that arises during this critical developmental checkpoint (Solnica-Krezel & Sepich 2012; Leptin 2005; Keller 2005; Nakaya & Sheng 2008; Tam & Behringer 1997; Tam & Loebel 2007; Myers et al. 2002). In all model organisms studied, transforming growth factor beta (TGFβ) superfamily members including bone morphogenetic proteins (BMPs) and Nodal, which signal though a family of mediator proteins called SMADs, play key roles in inducing gastrulation- specific pattern formation (Wu & Hill 2009). Although involvement of the TGFβ superfamily members is conserved between species during gastrulation, details of fate patterning in relation to cell movements, tissue structure, specific molecules involved etc., can vary between species (Wu & Hill 2009; Rossant 2015; Narasimha & Leptin 2000; Solnica-Krezel & Sepich 2012; Solnica- Krezel 2005). Therefore, it has been challenging to relate the regulatory mechanisms of pattern formation in previously studied developmental models to human gastrulation.

Since human embryos are not typically available for direct investigation, studying human gastrulation-associated fate patterning requires in vitro platforms that allow robust simulation and investigation of the signaling programs that initiate and drive gastrulation-like events. We and others have previously used micro-patterning technologies to control human (h)PSC colony geometry, demonstrating improved cell response consistency than is achieved in conventional adherent cultures (Nazareth et al. 2013; Peerani et al. 2007; Bauwens et al. 2008; Bauwens et al. 2011; Alom Ruiz & Chen 2007; Stevens et al. 2013; Ma et al. 2015). These studies highlight that control of endogenous signaling profiles, cell-cell contact, and mechanical forces are crucial to robustly regulate cell fate and spatial tissue organization. Recently, Warmflash et al. (Warmflash et al. 2014) used similar techniques to demonstrate that following BMP4 treatment, geometrically controlled hPSC colonies recapitulate many aspects of the peri-gastrulation-stage epiblast. Specifically, these colonies exhibited spatially patterned regions characteristic of primitive-streak-

like, mesoderm-like, endoderm-like, ectoderm-like and trophoblast-like tissues (Warmflash et al. 2014).

Two prominent biochemical models have influenced our understanding of cell fate patterning and morphogenesis that occurs during embryogenesis. The first is Reaction-Diffusion (RD), which describes the self-organization of homogenously distributed signaling molecules (morphogens) into complex, asymmetric patterns that provide spatial information to developing tissues (Turing 1952; Gierer & Meinhardt 1972). The second is Positional Information (PI), which describes how the asymmetric morphogen distributions across a developing tissue can be interpreted, and result in cell fate patterning (Wolpert 1981; Wolpert 1969). RD hypothesizes the presence of an interaction network of two molecules, an ‘activator’ which activates the expression of both molecules, and an ‘inhibitor’ which inhibits their expression. This interaction network, in conjunction with dissimilar activator and inhibitor diffusivities, is theoretically sufficient to self- organize asymmetrical morphogen distributions (Turing 1952; Gierer & Meinhardt 1972). The initial version of PI proposed a mechanism by which this asymmetric morphogen distribution could be translated into patterned developmental fates through a signaling threshold-based mechanism. Subsequent studies using the developing neural tube have demonstrated that fate patterning by PI is not just mediated by morphogen threshold levels, but is a function of both the morphogen concentration and exposure duration (Dessaud et al. 2007; Dessaud et al. 2008; Briscoe & Small 2015).

Here we demonstrate, using fully defined and scalable conditions, that geometrically-confined hPSC colonies organize into radially segregated regions that express markers characteristic of ectoderm-like, primitive streak-like, and trophoblast-like tissues. We show that upon BMP4 induction, an RD network regulated by BMP4 and NOGGIN (a cardinal BMP inhibitor) rapidly organizes nuclear localized phosphorylated (p)SMAD1 (effector of BMP signaling) into a gradient within the geometrically-confined colonies. The established gradient then patterns hPSC differentiation in a manner consistent with PI. We developed a computational model of a BMP4- NOGGIN RD system and demonstrate that, across a range of colony sizes and BMP4 doses, RD consistently predicts the formation of pSMAD1 signaling gradient and PI accurately predicts the patterned fate acquisition. A stepwise model of RD mediated self-organization of pSMAD1 signaling gradient followed by a PI mediated patterning of cell fates can predict the outcome of

previously unexplored experimental conditions. Specifically, we both predict and observe periodic fate patterning consistent with the cardinal RD paradigm. Furthermore, our model can identify conditions that rescue fate patterning in colonies previously deemed incapable of facilitating pattern formation (Warmflash et al. 2014). Taken together, our data support the concept that a stepwise process of RD and PI controls the peri-gastrulation-like patterns that develop in differentiating hPSC colonies.

3.3 Results

3.3.1 A defined high throughput assay for induction of peri-gastrulation-like patterning in human pluripotent stem cell colonies

Consistent with recent reports (Warmflash et al. 2014; Etoc et al. 2016), we observed radially segregated expression of gastrulation-associated markers CDX2, Brachyury (BRA), and SOX2 – representative of trophoblast-like, primitive-streak-like, and ectoderm-like tissues respectively – develop in geometrically-confined circular hPSC colonies differentiated in mouse embryonic fibroblast conditioned medium (CM) supplemented with BMP4 (Figure 3-1:A). Given the challenge of identifying key molecules regulating pattern formation in undefined CM, our first aim was to identify defined basal conditions that induce these patterns in hPSC colonies. Further, we optimized a previously described protocol that uses Deep Ultraviolet light (< 200 nm) mediated photo-oxidation of polyethylene glycol (PEG) coated slides (Azioune et al. 2009) to allow high- fidelity patterning of hPSC colonies and adapted this technique to produce 96-well microtiter plates (Materials and Methods).

Using our PEG plates, we performed a medium screen to identify defined conditions to induce patterning of gastrulation-associated fates in geometrically-confined hPSC colonies. The screen consisted of Nutristem (NS), mTeSR (MT), Essential-8 (E8), a Knockout Serum Replacement based medium (SR), and an N2B27 based medium – all supplemented with BMP4. Preliminary testing with N2B27 medium revealed that NODAL supplementation elicited a positive response in the induction of the BRA expressing region (data not shown), consistent with the importance of Nodal signaling in the specification of the primitive streak fate (Funa et al. 2015; Schier 2003; Brennan et al. 2001; Shen 2007). Therefore, we performed all N2B27 based experiments with 46

supplemented NODAL (100ng/ml). We found that, although segregated expression of CDX2 and SOX2 developed in all defined test media, BRA expression was not consistently observed in E8, NS, and SR media (Figure 3-2, Figure 3-3). Further, while spatial segregation of CDX2, BRA, and SOX2 occurred reproducibly in MT medium, colonies patterned in MT tended to lift off the plate by 48h. N2B27 medium robustly gave rise to differentiating hPSC colonies with regions expressing trophoblast-associated (CDX2) and primitive-streak-associated (BRA) markers, with concurrent spatial patterning of expression markers indicative of endoderm-like (SOX17) and mesoderm-like (EOMES) fates, in a manner indistinguishable from hPSC colonies differentiated in BMP4-supplemented CM (Figure 3-1A-C, Figure 3-2, Figure 3-3). Further, the regions at the center of the differentiating colonies expressed SOX2 (Figure 3-1B) but not NANOG (Figure 3-4), indicating a prospective ectoderm-like fate. Notably, the SNAIL-expressing mesenchymal cells appeared underneath the EPCAM-expressing epithelial layer in the regions expressing markers of the primitive streak (Figure 3-1D), which parallels the organization of the mesenchymal and epithelial cells at the onset of gastrulation. Also, this observation is consistent with what has previously been reported in BMP4-supplemented CM (Warmflash et al. 2014). Given these results, we define the fate patterning observed in BMP4 supplemented N2B27 medium as ‘peri- gastrulation-like’, and unless otherwise stated, this induction medium was used for all studies detailed below.

Figure 3-1: Defined peri-gastrulation-like patterning induction in differentiating hPSC colonies. Representative immunofluorescence images of: A) fate patterning of SOX2, BRA, and CDX2 in BMP4 supplemented CM as previously reported (Warmflash et al. 2014; Etoc et al. 2016); B) fate patterning in BMP4 supplemented N2B27 medium stained for SOX2, BRA, and CDX2, SOX17, and EOMES. C) Spatial trends for intensity of expression of SOX2, BRA, CDX2, and SOX17 in regions marked by white rectangle in B and C (Average trends of replicates shown in Figure 3-2, Figure 3-3). E) Representative image of SNAIL and EPCAM staining in micro- patterned colony differentiated in N2B27 shows that mesenchymal marker expressing cells in the primitive streak region are located underneath an epithelial layer. Scale bars represent 50µm.

Figure 3-2: Peri-gastrulation-like fate patterning in multiple basal medium conditions. Representative composite images and spatial expression average for SOX2, BRA and CDX2 staining in geometrically-confined hPSC colonies differentiated in BMP4 supplemented A) E8, B) Nutristem (NS), C) SR, D) mTeSR, E) N2B27, and F) CM. Scale bars represent 200µm.

Figure 3-3: Spatial trends of CDX2, BRA, and SOX2 observed in different basal media. Radial trends of A) BRA, B) SOX2, and C) CDX2 in BMP4 supplemented E8, NS, SR, MT, N2B27, and CM. Standard deviations shown in grey, and 95% confidence intervals shown in black.

Figure 3-4: Nanog does not co-localize with SOX2 expression at the center of differentiating hPSC colonies. A) Representative immunofluorescent images of colonies stained for DAPI, SOX2, and NANOG of geometrically confined hPSC colonies cultured in BMP4 supplemented N2B27. NANOG expression of 95 colonies shown as B) Average map, and C) line-plots of the average radial trend. Standard deviations shown in grey, and 95% confidence intervals shown in black. Data pooled from two experiments. Scale bars represent 200µm.

3.3.2 Nodal signaling is necessary for BRA expression but does not induce peri-gastrulation-like patterning

Nodal signaling is important in establishing the primitive streak fate (Hong et al. 2011; Funa et al. 2015). Given that our N2B27 medium contained NODAL, we first examined whether the spatially organized differentiation observed within the geometrically confined hPSC colonies required Nodal signaling. We found that selectively inhibiting Nodal signaling with SB-431542 (SB), an inhibitor of Alk4/5/7 receptors (Inman et al. 2002), in N2B27 medium supplemented with BMP4, abrogated the expression of BRA (Figure 3-5A-C) and significantly reduced the expression of CDX2 (Figure 3-5A). However, patterning of regions expressing SOX2 and residual CDX2 was still maintained (Figure 3-5C-E). These data indicate that Nodal signaling was necessary for the expression of BRA, and at least a subset of the CDX2, but did not induce organized fate patterning in the geometrically confined hPSC colonies. We then tested the necessity of BMP signaling to induce the peri-gastrulation-like fates within the geometrically confined hPSC colonies. To test this, we queried if the patterned fates within the hPSC colonies would arise in either N2B27 medium without supplemented BMP4, or in N2B27+BMP4 with supplementation of a small molecule BMP inhibitor LDN-193189 (LDN) (Cuny et al. 2008). We found that in both cases there was no observed peri-gastrulation-like patterning (Figure 3-5A-B). Furthermore, this patterning deficiency in N2B27 without supplemented BMP4 could not be rescued by further addition of either 100ng/ml or 200ng/ml of NODAL (Figure 3-5A-B). These data indicated that activation of BMP signaling was necessary to induce peri-gastrulation-like differentiation within the geometrically confined hPSC colonies. Taken together these data indicate that induction of peri- gastrulation-like fate patterning requires BMP signaling and the formation of BRA-expression region further requires Nodal activity.

Figure 3-5 Nodal signaling is required for primitive streak specification but does not induce differentiation and fate patterning in geometrically-confined hPSC colonies. A) Percentage of cells expressing BRA, SOX2, and CDX2 in N2B27, which includes 100ng/ml of NODAL (n=73), N2B27+BMP (n=100), N2B27+BMP+LDN (n=101), N2B27+BMP+SB (n=67), N2B27+100ng/ml Nodal (n=68), and N2B27+ 200ng/ml Nodal (n=65). The experiment was performed twice. **** (p<0.0001), NS (p>0.05). The p values were calculated using one-way 53

ANOVA (Dunnett’s post-hoc test). B) Representative images of colonies cultured in N2B27, BMP+LDN, Nodal (100), and Nodal (200) stained for BRA, SOX2, and CDX2. C) Representative immunofluorescent images of colonies cultured in BMP4, and BMP4+SB stained for BRA, SOX2, and CDX2. D) SOX2, BRA and CDX2 expression averages for 100 colonies cultured in BMP4, and 67 colonies cultured in BMP4+SB. E) Average radial trends from D) for BMP4 and BMP4+SB conditions. Standard deviations shown in grey, and 95% confidence intervals in black. Scale bars represent 200µm.

3.3.3 BMP4-NOGGIN interaction network regulates pSMAD1 gradient self- organization

Since we observed that BMP signaling was necessary for organized fate patterning within the geometrically confined hPSC colonies, we next examined how downstream effectors of BMP signaling were organized within the colonies. BMP ligands activate BMP receptors (BMPR1A, BMPR1B, BMPR2) leading to phosphorylation and nuclear translocation of SMAD1 followed by the transcription of context-specific target genes (Zhang & Li 2005). We used an antibody specific to pSMAD1 to measure the distribution of nuclear localized SMAD1 activity as a function of radial distance from colony centers at 1, 6, 18, and 24h following BMP4 induction. Robust analysis was enabled by overlaying pSMAD1 expression profiles of at least 100 colonies at each time point, yielding the average intensity of pSMAD1 as a function of colony radius. We noted that 1h after BMP4 induction, although the pSMAD1 activity appeared at all colony radii, the expression on average was slightly lower at the colony center then at the periphery (

Figure 3-6A), suggesting that centers of hPSC colonies might contain BMP inhibitors. This is consistent with previous reports that in the pluripotent state, hPSCs express secreted factors such as FOLLISTATIN, CERL, GDF3 etc. (Vallier et al. 2004; Yang et al. 2015; Besser 2004). Although these factors belong to the Activin/Nodal family, they are competent inhibitors of BMP signaling (Wu & Hill 2009). We observed relatively higher expression levels of these Activin/Nodal family member inhibitors of BMP signaling than the canonical BMP inhibitors like NOGGIN, and CHORDIN in hPSCs during regular culture conditions (Figure 3-7). The secretion of these BMP inhibitors in the geometrically-confined hPSC colony would produce a spatial distribution with the center of the colonies having relatively higher concentrations of BMP inhibitors (Supplement – Model description, Figure 3-8). In a colony with this expression profile

of non-canonical BMP inhibitors, BMP4 treatment would result in a small reduction of BMP signaling activity at the center of the colony, as we observed (

Figure 3-6A). Subsequently, over the first 24h of BMP4 treatment, we observed a rapid downregulation of pSMAD1 activity at the colony centers, and spontaneous organization into a radial gradient with the cells in the periphery exhibiting higher levels of pSMAD1 activity (

Figure 3-6A-C). To test whether the self-organization of pSMAD1 activity could be attributed to regulatory feedback of the BMP pathway, we measured the expression of both positive and negative feedback mediators of BMP signaling over the first 26h after BMP4 treatment. We found significant upregulation in the expression of both NOGGIN and BMP4 (

Figure 3-6D) in a dose dependent manner (Figure 3-9). Notably, BMP4 and NOGGIN upregulation was seen in all basal medium conditions tested (Figure 3-10), indicating that the upregulation was a response to the activation of BMP signaling and unrelated to the N2B27 medium. Furthermore, NOGGIN knockdown, using siRNA, significantly increased pSMAD1 activity in the central regions of the geometrically-confined colonies demonstrating a role for NOGGIN in the self- organization of the pSMAD1 gradient (Figure 3-6E-G, Figure 3-11), as has recently been proposed by others (Etoc et al. 2016). Interestingly, along with the NOGGIN mediated negative feedback, we noted the presence of a positive feedback loop in BMP signaling wherein BMP4 supplementation resulted in BMP4 expression (Figure 3-6D, Figure 3-10). Together, the presence of both a positive and negative feedback response mediated by BMP4 suggested the presence of a BMP4-NOGGIN RD network underlying the pSMAD1 radial self-organization in differentiating hPSC colonies (Figure 3-6H). We next set out to investigate properties of the RD framework as they relate to the peri-gastrulation-like patterning.

Figure 3-6 pSMAD1 gradient self-organization in differentiating colonies suggests the presence of a BMP4-NOGGIN RD network. A) Representative immunofluorescence images of colonies stained for pSMAD1 at 0, 1, 6, 18, and 24h after BMP induction. B) Average pSMAD1 intensity represented as overlays for 202, 56

193, 105, 100, 105 colonies for the respective induction times. Data were collected from two experiments. White arrowhead marks the region of relatively lower average pSMAD1 activation. C) The average radial trends of pSMAD1 activity at each induction duration. Standard deviations are shown in grey, and 95% confidence intervals in black. D) Temporal gene expression profiles for BMP signaling inhibitors (CHORDIN, NOGGIN, and GDF3) and BMP4 at 6, 12, 22, and 26h. Data shown as mean and standard deviation (S.D.) of three independent experiments. * p<0.05, ** p<0.01, *** p<0.001. E) Representative pSMAD1 immunofluorescence images for colonies treated with SCRAMBLE siRNA and NOGGIN siRNA after 24h of BMP4 treatment. F) Average pSMAD1 expression in colonies treated with SCRAMBLE siRNA, NOGGIN siRNA (64, and 62 colonies respectively). Data were collected from two experiments. G) Radial trends of pSMAD1 signaling distribution from (F) for SCRAMBLE, and NOGGIN siRNA conditions. Standard deviations are shown in grey, and 95% confidence intervals in black. **** p<0.0001, NS (p>0.05). The p values were calculated using Mann- Whitney U test. H) Overview of the implicated BMP4-NOGGIN RD network. Scale bars represent 200µm.

Figure 3-7: Basal expression of BMP inhibitors in hPSCs during routine culture. A) Overview of experimental setup – hPSCs were cultured in a variety of media conditions for 24 hours and gene expression was assessed for candidate inhibitors of BMP signaling. Expression of various inhibitors (NOG – NOGGIN, CHRD – CHORDIN, FST – FOLLISTATIN, GDF3, and CERL – CERBERUS-Like) of BMP signaling under basal conditions shown as ΔCt relative to GAPDH for CA1 cells cultured in B) N2B27, C) Nutristem (NS), D) Conditioned Medium (CM), E) Apel, F) Serum Replacement medium (SR), and G) mTeSR for 24 hours. Data are shown as mean (S.D) for three independent experiments

Figure 3-8: Initial condition for Noggin. A) A circular colony of cells (blue dots) modeled as a collection of sources of secreted molecules. B) Noggin is assumed to have an infinite sink at a certain distance from the colony periphery. The steady state diffusion profile of the secreted molecule is shown in C), and the expression profile within the colony along the diameter shown as a line plot D). E) The assumed initial condition of Noggin at the start of induction. ‘R’ represents the colony radius.

Figure 3-9: BMP4 and NOGGIN upregulation occur in a BMP4 induction dose-dependent manner. A) Experimental overview: Gene expression data gathered at 24 hours following induction at varying concentrations of BMP4. B) BMP4-induced expression of BMP4. C) BMP4 induced expression of NOGGIN. Data represented as mean and S.D. of three biological replicates. * p<0.05, ** p<0.01.

Figure 3-10: BMP4 induced upregulation of BMP4 and NOGGIN in tested medium conditions. Kinetic gene expression profiles for BMP4 and its cardinal inhibitors in response to BMP4 induced differentiation. Medium conditions tested include a Knockout serum – based medium (SR), a serum-free medium (SFI – for composition, please see Nazareth et al., Nature Methods 2013), Nutristem (NS), and Mouse Embryonic Fibroblast conditioned medium (CM). Data represented as mean and S.D of three biological replicates. * p<0.05, ** p<0.01, *** p<0.001.

Figure 3-11: Controls for NOGGIN and Scramble siRNA. A) Overview of experimental setup. hPSC cultures were treated with Scramble, and NOGGIN siRNA with BMP4 for 24h. B) NOGGIN gene expression for the Scramble and NOGGIN siRNA relative to the negative control shown as ΔΔCT. Data represented as mean (s.d.) of three biological replicates. The p values were calculated using Student’s t-test

3.3.4 pSMAD1 gradient formation is colony size and BMP4 concentration dependent

To simulate RD mediated self-organization of pSMAD1 activity, we developed a finite element model that predicts the spatio-temporal distribution of signaling-competent, free BMP4 ligands within the differentiating geometrically confined colonies using the RD-specific two-component, coupled, partial differentiation equation set (Turing 1952; Gierer & Meinhardt 1972) (See Appendix Chapter 6.1 – Supplement – Model Description). To determine if this model could accurately predict experimental data, we performed sweeps on two key model parameters: the initial concentration of BMP4 in the induction medium (BMPi), and the colony size. Our model predicted that reducing BMPi while maintaining the colony diameter at 1000µm would still lead to the formation of distribution gradients of free BMP4 within the colonies, but with lower ligand levels at the colony periphery (Figure 3-12A-B). Consistent with model predictions, the spatial distribution of the pSMAD1 concentration in 1000µm diameter colonies 24h after induction with varying BMPi (6.25ng/ml, 12.5ng/ml, 25ng/ml, and 50ng/ml) formed gradients, with lower BMPi conditions producing lower levels of peak pSMAD1 activity at colony peripheries (Figure 3-12C- E, Figure 3-13). We next queried the model to predict how reducing colony size, at a fixed BMPi dose, would affect the spatial distribution of free BMP4 ligands in the differentiating colonies. The model predicted that reducing the colony size would progressively increase the presence of free BMP4 ligands at the colony centers (Figure 3-12F-G). To test the model predictions experimentally, we differentiated colonies of varying sizes with constant BMPi dose (50ng/ml) and assessed pSMAD1 activity 24h post-induction. Consistent with the model predictions, we found that small colonies were unable to form regions with low pSMAD1 activity (Figure 3-12H- I). These findings are consistent with our hypothesis that the self-organization of pSMAD1 activity observed in differentiating hPSC colonies is governed by a BMP4-NOGGIN RD network.

Figure 3-12: BMP4-NOGGIN RD model predicts pSMAD1 gradient response to colony size and BMP4 dose perturbations. A) Predicted distribution of free BMP4 ligands in colonies of 1000µm diameter as a function of varying BMPi concentration (50ng/ml, 25ng/ml, 12.5ng/ml, and 6.25ng/ml). B) Line plots of predicted distributions. C) Representative immunofluorescence images of micro-patterned colonies stained for pSMAD1 24h after induction with varying BMPi conditions (50ng/ml,

25ng/ml, 12.5ng/ml, and 6.25ng/ml). D) pSMAD1 signaling distribution represented as a function of colony radius and E) averages of 143,119, 129, 163 colonies for respective conditions. Data were pooled from two experiments. F) Predicted distribution of free BMP4 ligands, following induction with 50ng/ml BMP4, as a function of colony diameter (700µm, 600µm, 500µm, 400µm, 300µm, and 200µm). G) Graphical depiction of predicted distributions in F. H) Representative immunofluorescence images of colonies stained for pSMAD1 24h after induction with 50ng/ml of BMP4. I) Average pSMAD1 expression of 87, 118, 178, 261, 437, and 373 colonies for respective conditions. J) pSMAD1 signaling distribution represented as a function of the colony radius. Data were collected from two experiments. Scale bars represent 200µm.

Figure 3-13: Quantified radial trends of pSMAD1 activity at 24 hours after induction with varying concentrations of BMP4. Radial trends of pSMAD1 activity were observed in varying BMPi concentrations (6.25 ng/ml, 12.5 ng/ml, 25 ng/ml, and 50ng/ml). Standard deviations shown in grey and 95% confidence intervals shown in black.

3.3.5 Peri-gastrulation-like fates arise in a manner consistent with the PI paradigm

Notably, the perturbations we tested above resulted in a change in pSMAD1 activity both at the periphery (Figure 3-12D) and at the center (Figure 3-12I) of the colonies. We reasoned that if fate acquisition in these colonies were a function of pSMAD1 activity thresholds, we would observe fate switches at different conditions. Accordingly, we tested the same conditions above – this time after 48h – and stained for the fate-associated markers SOX2, BRA, and CDX2. First, to perturb the pSMAD1 activity at the colony periphery we varied BMPi doses while keeping the colony size constant (Figure 3-14A). Consistent with a threshold-dependent fate acquisition model, we found a significant reduction of CDX2 expression in colonies induced at low BMPi conditions (Figure 3-14B-C). Furthermore, when we perturbed the pSMAD1 activity at the colony center by varying colony size at a constant BMPi dose (Figure 3-14D), we found that SOX2 expression disappeared in smaller colonies (≤ 300µm diameter) (Figure 3-14E-F). These data indicate that BMP signaling thresholds dictate fate acquisition in the differentiating geometrically-confined hPSC colonies.

The current understanding of PI, however, suggests that transcription factors associated with patterned fates arise not just as a function of the morphogen concentration, but also as a function of induction time; i.e. the fates associated with higher levels of morphogen concentration are induced after longer induction times (Briscoe & Small 2015; Green & Sharpe 2015; Dessaud et al. 2008; Dessaud et al. 2007). Therefore, we set out to investigate if fates in the differentiating hPSC colonies arose as a function of both the level of pSMAD1 activity and the time of induction. Accordingly, we tested four different BMPi doses (50ng/mL, 25ng/mL, 12.5ng/mL, and 6.25ng/mL), and analyzed the patterned fates that emerged at four different induction times (12h, 24h, 36h, and 48h). We found that the BRA and CDX2 fates did not arise consistently at either lower BMPi doses or at shorter durations of higher BMPi doses (Figure 3-14G-H, and Figure 3-15). This analysis suggests that the cell fate patterning in the hPSC colonies mediated by the pSMAD1 gradient follows PI, as reflected by the characteristic PI-like profile (Figure 3-14I). In summary, our findings demonstrate that the fate patterning within the differentiating hPSC colonies occurs via the cardinal PI model (Figure 3-14J).

Figure 3-14: Fate patterning in hPSC colonies arises in a pSMAD1 threshold dependent manner. A) Overview of the experimental setup: varying BMPi while maintaining colony size constant perturbs pSMAD1 signaling at the colony periphery. T1 and T2 represent putative thresholds that determine fate patterning. B) Quantification of cells expressing CDX2 in 1000µm colonies induced to differentiate at varying BMPi concentrations (6.25ng/ml, 12.5ng/ml, 25ng/ml, and 50ng/ml) (p value calculated using the Kruskal Wallis test. Number of colonies are 136, 168, 169, and 156 for the respective conditions. Results pooled from two separate experiments. C) Representative immunofluorescence images of SOX2, BRA, and CDX2 expression in geometrically confined 1000µm diameter colonies differentiated in 6.25ng/ml, 12.5ng/ml, 25ng/ml and 50ng/ml of BMP4. D) Overview of the experimental setup: colony size is varied while maintaining constant BMPi perturbs the level of pSMAD1 signaling in the colony center. T1 and T2 represent putative thresholds that determine fate patterning. E) Quantification of cells expressing SOX2 in colonies of varying diameters (700µm, 600µm, 500µm, 400µm, 300µm, 200µm) differentiated in BMPi = 50ng/ml (p value calculated using the Kruskal Wallis test). Number of colonies analyzed from two separate experiments were 144, 160, 279, 466, 789, and 1607 for the respective conditions. F) Representative immunofluorescent images of SOX2, BRA, and CDX2 expression in geometrically confined colonies of varying diameters (700µm, 600µm, 500µm, 400µm, 300µm, and 200µm). G) Representative immunofluorescence images of SOX2, BRA, and CDX2 stained 1000µm diameter colonies differentiated at varying BMPi concentrations (50ng/ml, 25ng/ml, 12.5ng/ml, and 6.25ng/ml) and induction times (12, 24, 36, and 48h). H) Average percentage of cells expressing SOX2, BRA, and CDX2 in hPSC colonies. Each condition had over 140 colonies. Data were pooled from two experiments. The p values were calculated for the concentration of 50ng/ml and induction time of 48h using the Kruskal Wallis test. I) Overview of peri-gastrulation-like fate acquisition arising as a function of both morphogen concentration and induction time. J) Representation of fate acquisition arising in a manner consistent with the cardinal Positional Information paradigm. Scale bars represent 200µm.

Figure 3-15: CDX2 and BRA expression in colonies arise as a function of BMP4 dose, and induction time. Percentage of cells expressing SOX2, BRA, and CDX2 in colonies induced to differentiate at varying concentrations of BMP4 (6.25 ng/ml, 12.5 ng/ml, 25 ng/ml, and 50 ng/ml) and induction times (12 hours, 24 hours, 36 hours, and 48 hours). Each condition had over 140 colonies. Data pooled from two experiments.

3.3.6 A two-step process of RD and PI governs peri-gastrulation-like fate patterning

Our data indicate that (1) the formation of the pSMAD1 gradient follows an RD mechanism, and (2) the subsequent fate acquisition follows a PI mechanism, suggesting a two-step process of biological fate patterning in geometrically-confined hPSC colonies. Recognizing that the classic RD models have periodic peaks of signaling activity, and patterned fates (Green & Sharpe 2015; Turing 1952; Kondo & Miura 2010), we queried our mathematical model to identify conditions that would result in a periodic distribution of free BMP ligands. Our model predicted that an increase in the colony size alone would be insufficient to induce this periodic distribution (Figure 3-16A). Consistent with this prediction, we did not detect a noticeable periodic response in either pSMAD1 levels at 24h post differentiation with BMP4 (Figure 3-16B, Figure 3-17A), or BRA expression after 48h of BMP4 treatment with 50ng/ml (Figure 3-16C, Figure 3-18A) when we increased colony diameter from 1000μm to 3mm. However, our model predicted that a concomitant increase in the BMP4 dose along with an increase in the colony size would result in the induction of a periodic response of BMP activity (Figure 3-16D). Consistent with this prediction, we found that differentiating hPSC colonies of 3mm diameter with a BMP4 dose of 200ng/ml resulted in periodicity in both pSMAD1 at 24h (Figure 3-16E, Figure 3-17B), and BRA expression at 48h post induction (Figure 3-16F, Figure 3-18B). We quantitatively compared the theoretical RD-like periodicity with the experimentally observed behavior to assess the predictive ability of the model. We measured the profile of the predicted BMP activity radially from the center of the colony every 30 degrees, for the condition where a 3mm colony was induced to differentiate in presence of 200ng/ml of BMP4, and quantified the dominant periods using a 1D Fourier transform. The same analysis was performed for pSMAD1 and BRA expression patterns (Figure 3-19) (Materials and Methods). The means and distributions of the theoretically predicted, and experimentally observed periods were not significantly different from each other (Mann- Whitney U: p>0.5 and >0.2, and Kolmogorov-Smirnov: p =0.3781 and 0.1452 for pSMAD1 and BRA respectively) (Figure 3-16G-I), validating the ability of the RD model to predict the experimental RD-like periodic response.

Figure 3-16: High BMP4 dose in induction media recapitulates stereotypic RD-like periodic patterns in 3mm diameter hPSC colonies.

A) Model predictions for spatial profile of BMP activity for colonies of 3mm diameter at BMPi = 50ng/ml. Experimental data showing representative immunofluorescent images and spatial expression profiles for B) pSMAD1, and C) BRA. D) Model predictions for spatial profile of BMP activity for 3mm diameter colonies differentiated at a higher dose of BMPi (200ng/ml). Experimental data showing representative immunofluorescent images and spatial expression profiles for E) pSMAD1, and F) BRA. Histogram of dominant periods identified in the G) computational model of RD-like BMP pattern formation, and experimentally identified periods for H) pSMAD1 (n=28 colonies pooled from 3 experiments), and I) BRA (n = 47 colonies pooled from 3 experiments) expression. The p-values in red were calculated using Mann-Whitney U test, and in blue were calculated using Kolmogorov-Smirnov test. Periods greater than 1 indicate no periodicity. Scale bars represent 1mm.

Figure 3-17: RD-like patterns noted in pSMAD1 activity in 3mm colonies when differentiated with high doses of BMP4.

Representative immunofluorescent images of 3mm diameter colonies stained for pSMAD1 for a BMP4 dose of A) 50ng/ml, and B) 200ng/ml. White arrowheads denote representative areas of high pSMAD1 activity indicative of RD-like patterns. Scale bars represent 1mm.

Figure 3-18: RD-like patterns noted in BRA fate acquisition in 3mm colonies differentiated in high doses of BMP4. Representative immunofluorescent images of 3mm diameter colonies stained for BRA for a BMP4 dose of A) 50ng/ml, and B) 200ng/ml. Scale bar represents 1mm.

Figure 3-19: Analysis pipeline for extracting dominant periods of theoretically predicted distribution of free BMP4 ligands and expression patterns in experimental colonies. A) Model prediction, and the inverted image (B) shown in grey scale. Profiles extracted every 30 degrees. C) Profile observed along red line in B. Acquired (D), and thresholded (E) (to remove background noise) images for pSMAD1, and BRA. Profiles extracted every 30 degrees for multiple colonies (n = 28 for pSMAD1, and n=47 for BRA). F) Identified periodic profile along the red lines in (E) for pSMAD1 and BRA.

As a final validation, we considered the recently reported claims that ‘edge-sensing’ underlies the observed pattern formation (Etoc et al. 2016). Edge-sensing is a mechanism in which fates are patterned sequentially with the trophectoderm-like fate at the colony edge, followed by the primitive-streak-like region, and the ectoderm-like region positioned at the colony center. Consequently, small colonies of 250µm diameter, that don’t have sufficient space from the colony periphery to pattern the primitive-streak-like, and ectoderm-like regions, would be incapable of fate patterning – only allowing the induction of the trophectoderm-like region (Warmflash et al. 2014). Since our interpretation of the data mechanistically implicated a stepwise coordination of RD and PI, our interpretation of the inability of 250µm diameter colonies to induce all three fates differs from the edge-sensing explanation. Specifically, we argue that at a BMPi dose of 50ng/ml (the dose tested in the previous reports), RD mediated organization of free BMP4 ligands within colonies of 250µm diameter are sustained at high levels throughout the colony (Figure 3-12F-I), and past the thresholds that would induce the primitive-streak-like and ectoderm-like fates after a 48h induction (Figure 3-14E-F) as per PI. To demonstrate this claim, we queried our model to test if perturbing BMPi could induce the organization of BMP activity at appropriate levels to rescue the fate patterning of all three lineages. Our model predicted that reducing BMPi would reduce the levels of free BMP4 ligands throughout the colony whereby the BRA, and SOX2 expressing regions might be rescued (Figure 3-20A-B). We tested the expression profiles of pSMAD1 activity at 24h post induction with varying BMPi conditions (50ng/ml, 25ng/ml, 12.5ng/ml, and 6.25ng/ml) of over 500 per dose condition. Consistent with our model predictions, we observed a reduction of pSMAD1 levels overall, but especially in the center of the colonies (Figure 3-20C-E). These data suggested that lower BMPi conditions might rescue the formation of the BRA, and SOX2 expressing regions in colonies of 250µm diameter in accordance with PI. Indeed, when we tested the fate expression within 250µm diameter colonies after 48h of induction with varying BMPi doses, we observed a significant reduction of CDX2 expression at reduced doses of BMP4, and a concomitant increase of BRA and SOX2 expression within the colonies (Figure 3-20F-G). Importantly, at a BMPi dose of 12.5ng/ml, patterning of all three fates is rescued in 250µm diameter colonies (Figure 3-20G).

Figure 3-20: Low BMP4 dose in induction media rescues fate patterning in 250µm diameter colonies. A) Model predictions for gradient formation of free BMP4 ligands distribution in 250µm diameter colonies in response to varying BMPi doses. B) Line plot representation of free BMP4 ligands as a function of the colony radius. C) Representative immunofluorescence images of colonies stained for pSMAD1 24h after induction with varying doses of BMP4. D) Average pSMAD1 expression for 976, 638, 475, and 689 colonies for respective conditions. E) Signaling distribution represented as a function of the colony radius. The results were collected from two experiments. F) Quantification of percentage of cells in each colony expressing CDX2, BRA, and SOX2 in 250µm diameter colonies induced to differentiate at varying BMPi concentrations (6.25ng/ml, 12.5ng/ml, 25ng/ml, and 50ng/ml), p-value calculated using the Kruskal-Wallis test. Number of colonies are 1024, 1102 1134, and 1135 for the respective conditions. Results pooled from two separate experiments. G) Representative immunofluorescence images of SOX2, BRA, and CDX2 expression in geometrically confined 250µm diameter colonies differentiated in 6.25ng/ml, 12.5ng/ml, 25ng/ml and 50ng/ml of BMP4. Scale bars represent 200µm.

Taken together, these data are consistent with our hypothesis that the fate acquisition in the geometrically confined hPSC colonies, in response to BMP4 mediated differentiation, arises via a two-step process of RD and PI (Figure 3-21).

Figure 3-21: Mechanism of peri-gastrulation-like fate patterning in geometrically confined hPSC colonies. A BMP4-NOGGIN RD network induces a radial and periodic pSMAD1 activity gradient in geometrically confined hPSC colonies. The fate acquisition due to the pSMAD1 gradient follows the classical PI paradigm.

3.4 Discussion 3.4.1 Experimental models to study developmental fate patterning

Of the various proposed models of biological pattern formation, RD and PI have emerged as the dominantly accepted mechanisms to describe tissue organization during early development. RD describes a mechanism by which a symmetric morphogen distribution across a developing tissue self-organizes into a signaling gradient, and PI explains how this gradient is interpreted by the tissue resulting in patterned cell fates. Although both RD (Economou et al. 2012; Sick et al. 2006; Raspopovic et al. 2014), and PI (Gregor et al. 2007; Gregor et al. 2008; Houchmandzadeh et al. 2002; Jaeger et al. 2004; Green & Smith 1990; Chen et al. 2012) have been widely studied independently of each other, few developmental model systems allow for studying both aspects of biological pattern formation. Here, we introduced such a model which is a defined in vitro platform of developmentally relevant fate patterning using hPSCs to enable studies probing the mechanistic underpinnings behind both RD and PI concurrently.

3.4.2 Relevance to recent studies for peri-gastrulation associated fate patterning

Our study closely follows two recent reports where BMP4 supplemented CM induced similar fate patterning in geometrically-confined hPSC colonies as we observed in our work (Warmflash et al. 2014; Etoc et al. 2016). Although we use defined conditions, our observations were remarkably similar to those reports - underscoring the robustness of their findings. Similar to our claims, Etoc et al. (Etoc et al. 2016) also proposed that BMP4 induced fate patterning in hPSC colonies occurs due to a self-organized pSMAD1 signaling gradient, and suggested that this self-organization is regulated by coordination of two different mechanisms. First – a negative feedback control of BMP activity mediated by BMP-signaling induced NOGGIN expression; and second – a density- dependent reduction in sensitivity to BMP4 ligands over the entire colony except for the periphery due to re-localization of BMP receptors from apical to basolateral regions (a mechanism called ‘edge-sensing’), rendering them inaccessible and therefore incapable of inducing the BMP signaling pathway. Our interpretation of both our own data and that reported by Etoc et al., however, differs in a few important aspects outlined below.

Although their model of dual negative feedback on BMP signaling mediated by NOGGIN and receptor inaccessibility (Figure 3-22A) predicts the formation of a radial signaling gradient, it does not describe a classical RD system. This is because it lacks positive feedback of the morphogen (Turing 1952; Kondo & Miura 2010; Murray 2008; Gierer & Meinhardt 1972), and would therefore never give rise to periodic RD-like patterns (simplified model shown in Figure 3-22B,C). In contrast, we demonstrate that the presence of BMP4 in our cultures, regardless of the basal medium used, elicits both a positive, and a negative feedback response mediated by the upregulation of BMP4, and NOGGIN respectively. BMP4 and NOGGIN form a stereotypical activator-inhibitor pair (Gierer & Meinhardt 1972), as BMP4 can induce short range activation of BMP signaling (Zhang & Li 2005), while NOGGIN is a highly-diffusible molecule that can result in long-range inhibition of BMP signaling (Smith & Harland 1992; Inomata et al. 2013). Importantly, we explicitly demonstrate the induction of an RD-like response in larger colonies upon high dose induction of BMP4, both in pSMAD1 activity at 24h after induction and in the fate acquisition as shown by BRA expression 48h after induction (Figure 3-16, Figure 3-17, Figure 3-18).

Etoc et al. demonstrate that as cell density increases, the BMP receptors re-localize to basolateral regions of the hPSCs. They also demonstrate that these lateralized receptors are incapable of inducing the BMP signaling pathway in hPSCs when BMP4 ligands are presented on the apical side (Etoc et al. 2016). They claim that this mechanism is one of the drivers of the self-organized formation of the pSMAD1 gradient, given that the cell density of the entire colony increases with time. However, they only investigated this at a 500µm colony diameter. Here, we explored a range of colony sizes, and in 1000µm diameter colonies we generally observed regions of varying cell densities arise 48h after BMP4 induction (Figure 3-2E-F, Figure 3-5C-D). This finding is somewhat inconsistent with the mechanism proposed by Etoc et al. that the self-organization of a pSMAD1 signaling gradient is induced by density dependent changes that block BMP signaling. An alternate interpretation to the one put forth by Etoc et al., is that the density dependent receptor re-localization is in fact a response to the RD-mediated gradient formation, rather than its cause. This interpretation is directly supported by the appearance of RD-like periodic patterns in both pSMAD1, and BRA expression (Figure 3-16, Figure 3-17) in colonies of 3mm diameter where many sporadic spots of higher density relative to the adjacent regions can be observed. In this case, differentiating hPSCs either migrate down a pSMAD1 gradient, or proliferate at different rates 80

depending on the pSMAD1 activity they are exposed to, resulting in regions of varying cell densities. Further work is required to determine which of these possibilities is dominant.

Although our interpretations differ on how ‘edge-sensing’ mediated receptor availability relates to pSMAD1 gradient formation, the ‘edge-sensing’ model proposed by Etoc el al. (Etoc et al. 2016) nicely parallels our assumed boundary condition for the RD model. For our mathematical RD model, we assumed that the distribution of free BMP4 ligands across the colony evolves over time and space with a fixed concentration of BMP4 at the colony periphery that is equal to the concentration in the bulk medium, a boundary condition that changes depending on the BMP4 concentration of the induction medium. In a recent study, Warmflash, et al., claim that the fates are patterned inward from the edge of the geometrically-confined colony (Warmflash et al. 2014). Consequently, they contend that 250μm diameter smaller colonies are incapable of inducing the fate patterning observed in 1000μm diameter colonies, since the smaller colonies do not have enough space to pattern the primitive-streak-like and ectoderm-like regions. However, our interpretation of the ‘edge-sensing’ model identifies conditions that rescue fate patterning of all three fates in 250μm diameter colonies. Specifically, we demonstrate that in conditions with lower doses of BMP4 in the induction medium, smaller colonies maintain regions that express primitive- streak-like, and ectoderm-like fates after 48h of peri-gastrulation-like induction.

Figure 3-22: Dual inhibition model does not give rise to repetitive RD-like free BMP4 distribution. A) Proposed dual-inhibition model of gradient formation in differentiating hPSC colonies by Etoc et al (Etoc et al. 2016). B) Simplified mathematical representation of a generic dual 82

inhibition model. C) Gradient formation of free BMP4 ligands as predicted by the model. D) Varying doses and colony sizes demonstrates the inability of the dual-inhibition model to generate a periodic Turing-like response in the free BMP4 distribution.

3.4.3 Positional-Information

We specifically tested and demonstrated that, consistent with the PI paradigm, in the differentiating hPSC colonies fate acquisition arises as a function of both pSMAD1 concentration and induction duration. Notably, although the fate acquisition mediated by the pSMAD1 gradient follows the PI paradigm, the primitive-streak-like region is not specified in the absence of Nodal signaling. This underscores that patterning of developing tissues depends on coordination of multiple signaling pathways, reminiscent of quintessential models of PI mediated fate patterning such as the gap genes in the Drosophila which arise from a collaborative stimulation of multiple signals that include Bicoid, Caudal, Hunchback etc. (Briscoe & Small 2015). Consequently, we propose that although pSMAD1 activity patterns fates according to PI, it does so in conjunction with other transcription factors, for instance SMAD2 which is the effector of Nodal signaling.

3.4.4 Scaling of morphogen gradients during development

Morphogen gradients formed in developing embryos are thought to scale with size. For instance, the scaling of the dorsal-ventral axis in the developing Xenopus is robust to dramatic manipulations. In the case where the Xenopus embryo undergoes resection of the ventral half, it still develops into a smaller but proportionally patterned larva (Ben-Zvi et al. 2011). This characteristic has been attributed to the function of the Spemann’s organizer (Inomata et al. 2013; Ben-Zvi et al. 2011), underscoring that the robustness of the induction and the stabilization of the morphogen gradients that pattern a developing embryo is the result of a concerted effort by multiple tissues. Our data show that fate organization in differentiating micro-patterned hPSC colonies across varying sizes is not robust to a specific BMP4 concentration. Nevertheless, modulation of the BMP4 induction dose can recapitulate the appropriate patterning. This highlights that robustness to changes in the size of the developing epiblast is attainable, and emphasizes the importance of regulation from other tissues like the vertebrate organizer (the primitive node) in

conferring robustness to a developing human embryo. The mechanism that regulates peri- gastrulation-like pattern formation in the geometrically-confined hPSC colonies is one of multiple layers of complexity that would be present in a developing human embryo.

3.4.5 Identity of fate compartments in the peri-gastrulation-like platform

It has been widely reported that BMP4 treatment of hPSCs gives rise to a population that expresses CDX2. However, the identity of this population has been a subject of debate for nearly two decades. Some groups identify this population as trophectoderm(Xu et al. 2002; Warmflash et al. 2014; Li et al. 2013), while other claim that this population is mesodermal in fate (Bernardo et al. 2011; Mendjan et al. 2014; Loh et al. 2016). Interestingly, inhibiting Nodal signaling in the differentiating geometrically-confined colonies results in the abrogation of the primitive-streak- like population and a significant reduction in CDX2 expression. This suggests that at least a subset of the CDX2 positive population within the patterned peri-gastrulation-like fates require Nodal signaling and is mesodermal in fate.

Additionally, SOX17 expression in differentiating hPSCs has been widely used as a bona fide marker of the endoderm fate(Warmflash et al. 2014; Blauwkamp et al. 2012; Tsakiridis et al. 2014; Green et al. 2011). However, recent studies have shown that SOX17 is important in the induction of the primordial germ cell (PGC) fate in development (Irie et al. 2015; Kobayashi et al. 2017). Although in our report, consistent with the current literature, we define the SOX17 expressing population as endoderm, we have not ruled out the possibility that this population could be the precursor that subsequently gives rise to PGCs.

Of note, the boundary between the ectoderm-like and the primitive-streak-like regions in our platform appears to have cells that express both SOX2, and BRA. We propose two different possibilities that can result in this observation. First, these cells may reflect the population transitioning from a PSC state to a primitive-streak identity which would be upregulating BRA, and downregulating SOX2. Although SOX2 expression in this population would be low, it might still be detectably via immunofluorescence indicating a presence of a SOX2+ BRA+ phenotype, which would stabilize in a SOX2-BRA+ identity. Alternatively, this population might parallel the presumptive neuromesodermal-progenitor-like (NMP-like) population that reside in the node- 84

streak border, caudal lateral epiblast, and the chordoneural hinge sections in the posterior end of the elongating embryo (Gouti et al. 2014; Turner et al. 2014). However, NMPs have been proposed to be present relatively later during gastrulation (Henrique et al. 2015), and whether the SOX2- BRA double position population in our peri-gastrulation-like model is, in fact, NMP-like requires further investigation.

The fully defined platform we describe here facilitates investigation of the true identity of all these populations in a developmentally appropriate in vitro model system.

3.4.6 Conclusion

In conclusion, we report a defined, in vitro model of self-organized human peri-gastrulation-like biological fate patterning and demonstrate the mechanistic underpinning as a stepwise model of RD, and PI paradigms. This in vitro model can be employed to investigate the identity of cell populations that arise out of differentiating hPSCs, especially the populations whose identity has been a subject of debate, while maintaining an appropriate developmental context. We further report that fate acquisition that occurs in a manner consistent with PI can require multiple signaling pathways working in concert. Finally, our data implicates the coordination of multiple tissues to induce morphogen gradients that can scale and maintain robustness to perturbations that may occur in the developing embryo. Consequently, our work not only provides deep insight into one of the earliest stages of human embryonic development, but also into general mechanisms involved in patterning of biological form.

3.5 Materials and Methods 3.5.1 Human pluripotent stem cell culture

CA1 hPSCs (generously provided by Andras Nagy, Samuel Lunenfeld Research Institute) were cultured on Geltrex (Life Technologies, diluted 1:50) coated 6-well tissue culture plates using mTeSR1TM medium (StemCell Technologies) as per manufacturer’s instructions. The cells were passaged at a ratio of 1:12 using ReleSRTM (StemCell Technologies) per manufacturer’s

instructions. For the first 24h after passage, the cells were cultured in ROCK inhibitor Y-27632 to increase cell viability. The medium was changed every day and passaged every 4 to 5 days or when the cells reached 75-80% confluence.

3.5.2 Preparation of PEG plates to micro-pattern hPSC colonies

Custom sized (110mmx74mm) Nexterion-D Borosilicate thin glass coverslips (SCHOTT) were activated in a plasma cleaner (Herrick Plasma) for 3 minutes at 700 mTorr, and incubated with 1 ml of Poly-L-Lysine grafted Polyethylene Glycol (PLL-g-PEG(5KD), SUSOS,) at a concentration of 1 mg/ml at 37°C overnight. The glass slides were then rinsed with ddH2O and dried. The desired patterns were transferred to the surface of the PEG-coated side of the coverslip by photo-oxidizing select regions of the substrate using Deep UV exposure for 10 minutes through a Quartz photomask in a UV-Ozone cleaner (Jelight). Bottomless 96-well plates were plasma treated for 3 minutes at 700 mTorr and the patterned slides were glued to the bottomless plates to produce micro-titer plates with patterned cell culture surfaces. Prior to seeding cells onto the plates, the wells were activated with N-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (Sigma) and N- Hydroxysuccinimide (Sigma) for 20 minutes. The plates were thoroughly washed three times with ddH2O, and incubated with Geltrex (diluted 1:150) for 4h at room temperature on an orbital shaker. After incubation, the plate was washed with Phosphate Buffered Saline (PBS) at least three times to get rid of any passively adsorbed extracellular matrix (ECM) and seeded with cells to develop micro-patterned hPSC colonies.

3.5.3 Cell seeding and induction of peri-gastrulation-like fate patterning

To seed cells onto ECM-immobilized PEG-UV 96-well plates, a single cells suspension of the CA1 line was generated by incubation in 1ml of TryplE (Invitrogen) per well for 3 minutes at 37°C. The TryplE was blocked using in equal volume DMEM + 20% KnockOut Serum Replacement (SR) (Invitrogen) and the cells were dissociated by pipetting to generate a single cell suspension. The cells were centrifuged and re-suspended at a concentration of 1 x 106 cells/ml in SR medium supplemented with 20ng/ml bFGF (R&D) and 10µM ROCK inhibitor Y-27632. SR

medium consists of 74% DMEM, 1% Penicillin/Streptomycin, 1% non-essential amino acids, 0.1mM β-mercaptoethanol, 1% Glutamax, 2% B27 minus retinoic acid, and 20% SR (all Invitrogen). Wells were seeded in the PEG-patterned 96 well plates at a density of 80,000 cells/well and incubated for 2h at 37°C. After 2h, the medium was changed to SR without ROCKi. An alternative seeding process that has also been tested and provides good results is described in a recent protocol (Deglincerti, Etoc, et al. 2016). When confluent colonies were observed (12-16h after seeding), the peri-gastrulation-like induction was initiated in N2B27 medium supplemented with BMP4 (R&D) (BMP4 dose depended on experimental design). The peri-gastrulation-like pattern formation assay was typically performed with 1000µm diameter colonies 48h following induction, although additional time points and colonies sizes were tested as described in the Results section for specific experiments. N2B27 medium consists of 93% DMEM, 1% Penicillin/Streptomycin, 1% non-essential amino acids, 0.1mM β-mercaptoethanol, 1% Glutamax, 1% N2 Supplement, 2% B27-retinoic acid supplement (all Invitrogen) supplemented with 100ng/ml Nodal (R&D), and 10ng/ml bFGF (R&D).

3.5.4 Single cell data acquisition and analysis of immunofluorescence data

The patterned plates were fixed with 3.7% paraformaldehyde for 20 minutes, rinsed three times with PBS and then permeabilized with 100% methanol for 3 minutes. After permeabilization, the patterned colonies were blocked using 10% Fetal Bovine Serum (Invitrogen) in PBS overnight at 4ºC. Primary antibodies were incubated at 4ºC overnight (antibody sources and concentrations shown in Table 3-1). The following day the primary antibodies were removed, and the plates were washed three times with PBS followed by incubation with the secondary antibodies and DAPI nuclear antibody at room temperature for 1h. Single cell data was acquired by scanning the plates using the Cellomics Arrayscan VTI platform using the TargetActivation.V4 bioassay algorithm. This algorithm utilizes the expression intensity in the DAPI channel to identify individual nuclei in all fields imaged and acquires the associated intensity of proteins of interest localized within the identified region. The single cell data was exported into Context Explorer (CE), a custom software developed in-house for image analysis (Ostblom, et al., unpublished). In CE, cell colonies are identified through the DBSCAN algorithm as implemented in Python’s Scikit-learn package. Within a colony, each cell is assigned x- and y-coordinates relative to the colony centroid. To 87

create the colony overlay plots, cells from multiple colonies are grouped in hexagonal bins per their relative x- and y-coordinates. These positional bins are color-coded to represent the average protein expression level of all cells within a bin. The color map range is normalized to the lowest and highest expressing hexagonal bins. Spatial expression trends within colonies are also visualized as line plots, where cells are grouped by the Euclidean distance between a cell and the centroid of the colony. For each colony, the average expression value of all cells within a distance bin is computed. The line plots describe the mean expression value, standard deviation and 95% confidence interval (CI) between colonies as indicated in figure legends. The line plots depicting radial trends of proteins of interest in individual colonies were acquired through the sections of the colonies depicted (shown as white rectangles in the figures) in Fiji (ImageJ). The plot profile extracted was then run through a Savitzky-Golay smoothing filter in Matlab and represented as a function of radial distance.

The quantification of RD-like patterns in differentiating, geometrically-confined hPSC colonies of 3mm diameters was performed using SIESTA, an image analysis platform (Fernandez- Gonzalez & Zallen 2011; Leung & Fernandez-Gonzalez 2015), and custom scripts written in MATLAB (Mathworks) using the DIPImage toolbox (TU Delft, The Netherlands). To quantify the periodicity of free BMP4 ligands in the computer model, or pSMAD1 and BRA signals in colonies, we used the Fourier transform. To avoid quantifying background noise, we thresholded the BRA and pSMAD1 signals, and the pixels with intensity lower than the mean of the image plus one standard deviation were set to zero. We measured the signal intensity in the colonies along radii separated by an angle of 30°. The signals were detrended by subtracting the mean pixel value, and smoothened using a Gaussian of σ = 3 pixels for the colonies and σ = 20 for the model, consistent with the different spatial resolution of experimental and model images. The edge of the BRA colonies was excluded from the analysis to avoid artefacts generated by accumulation of migrating cells. Finally, we used the Fourier transform to decompose each signal into its constituent frequencies and computed a characteristic period, defined as the inverse of the dominant frequency. To be able to compare periods between colonies and the model, the length of each radius was normalized to 1. Periods greater than one indicated lack of periodicity.

Table 3-1: Antibodies used in this study

Antibody Target Company and catalog ID Dilution

CDX2 Abcam (ab15258), Cedarlane Abcam (1:50), Cedarlane (MU392A-UC) (1:400)

BRA R&D (AF2085) 1:500

SOX2 Cell Signaling (3579S), R&D Cell Signaling (1:200), R&D (MAB2018) (1:500)

SOX17 R&D (AF1924) 1:500

EOMES Abcam (ab23345) 1:500

EPCAM R&D (SC026) Kit 1:10

SNAIL R&D (SC026) Kit 1:10 pSMAD1 Cell Signaling (9516S) 1:200

3.5.5 Quantitative PCR analysis

RNA extraction was performed using Qiagen RNAeasy miniprep columns according to the manufacturer’s protocol, and the cDNA was generated using Superscript III reverse transcriptase (Invitrogen) according to the manufacturer’s instructions. The generated cDNA was mixed with primers for the genes of interest and SYBR green mix (Roche, Sigma) and the samples were run on an Applied Biosystems QuantStudio 6 flex real-time PCR machine. Relative expression of described genes was determined by the delta–delta cycle threshold (ΔΔCt) method with the

expression of GAPDH as an internal reference. Primer sequences used are provided below in Table 3-2.

Table 3-2: Primers used in this study

Primer target Sequence or catalog ID

BMP4 (fwd) ATGATTCCTGGTAACCGAATGC

BMP4 (rev) CCCCGTCTCAGGTATCAAACT

NOGGIN (fwd) NM_005450.4 (GeneCopoeia Cat # HQP054071)

NOGGIN (rev) NM_005450.4 (GeneCopoeia Cat # HQP054071)

CHORDIN (fwd) NM_001304473.1 (GeneCopoeia Cat # HQP067561)

CHORDIN (rev) NM_001304473.1 (GeneCopoeia Cat # HQP067561)

FOLLISTATIN (fwd) NM_013409.1 (GeneCopoeia Cat # HQP000565)

FOLLISTATIN (rev) NM_013409.1 (GeneCopoeia Cat # HQP000565)

GDF3 (fwd) GTACTTCGCTTTCTCCCAGAC

GDF3 (rev) GCCAATGTCAACTGTTCCCTT

CERL (fwd) CTTCTCAGGGGGTCATCTTG

CERL (rev) TCCCAAAGCAAAGGTTGTTC

3.5.6 Statistics and data analysis

All gene expression results were expressed as mean (+S.D.). Statistical tests for gene expression results were performed using two-tailed Student’s t-test assuming unequal variance between datasets. The statistical significance for fate acquisition results were calculated either using one- way ANOVA (Dunnett’s post-hoc test) or the Kruskal-Wallis test as described in the legends. The calculations for Student’s t-tests were performed in Excel. The one-way ANOVA, and the Kruskal-Wallis test calculations were performed in Prism. To evaluate sample means of periodic distributions predicted by the RD-model with the distributions experimentally observed, we used a non-parametric Mann–Whitney U test; and we used the Kolmogorov–Smirnov test was used to compare sample distributions

3.5.7 siRNA transfection protocol

All siRNA transfection was performed with CA1 hPSCs seeded on either PEG plates as described above, or 24 well plates for the qPCR control experiments to validate the siRNA specificity. Noggin specific siRNAs – NOG Silencer (Thermo Fisher), and siRNA NOG (Santa Cruz Biotechnology), mixed at 1:1 ratio, at a total concentration of 40nM; or scrambled control siRNA (Santa Cruz Biotechnology) at a concentration of 40nM. Transfection was performed using EditProTM Stem Transfection Reagent (MTI-Globalstem). Specifically, 1μl (per 24 well) or 0.2μl (per 96 well) EditProTM Stem Transfection Reagent was diluted with 50μl or 10μl Opti-MEM I reduced serum medium (Thermo Fischer) respectively. Diluted reagent was incubated with siRNA for 15 minutes at RT and subsequently added to cells containing 500μl (24 well) and 100μl (96 well) culture medium without Penicillin/Streptomycin. Medium was replaced with N2B27+BMP4 18 hours after transfection. Cells were analyzed 24 hours after transfection by qPCR (24 well) for NOGGIN gene expression, or by microscopy (96 well) for pSMAD1 spatial trends.

Chapter 4 Nodal dissects peri-gastrulation-like and pre-neurulation- like fate patterning in geometrically confined human pluripotent stem cell colonies

A version of this chapter has been submitted for peer review by:

Mukul Tewary, Dominika Dziedzicka, Joel Östblom, Laura Prochazka, Nika Shakiba, Curtis Woodford, Nafees Rahman, Elia Piccinini, Christopher Demers, Davide Danovi, Mieke Geens, Fiona M. Watt, Peter W. Zandstra

Attributions:

Dominika Dziedzicka assisted with experimentation and gene expression analysis associated with the involvement of Nodal in the formation of the SMAD1 gradient and the use of the peri- gastrulation-like assay for identifying the lineage biad of hPSC lines. Joel Ostblom wrote the software employed to analyze the high-content images. Dr. Laura Prochazka assisted with knockdown experiments. Nika Shakiba assisted with qPCR analysis. Curtis Woodford, and Elia Piccinini assisted with the differentiation toward definitive and mature endodermal fates. Nafees Rahman assisted with demonstration of the applicability of the platform in a variety of other cell types. Drs. Christopher Demers, Davide, Danovi and Mieke Geens provided insight for the manuscript. Dr. Fiona Watt provided support for a portion of the experiments. The remainder of all the experiments were performed by Mukul Tewary. Dr. Peter Zandstra oversaw this work.

Nodal dissects peri-gastrulation-like and pre- neurulation-like fate patterning in geometrically confined human pluripotent stem cell colonies

4.1 Abstract

In vitro models of post-implantation human development are valuable to the fields of regenerative medicine, and developmental biology. We established a robust, high-throughput micro-patterning platform and screened multiple human pluripotent stem cell (hPSC) lines in geometrically- confined colonies, for their ability to undergo peri-gastrulation-like fate patterning in response to BMP4 treatment; and observed significant variability. Further, we observed that in differentiating hPSC lines, upregulation of Nodal expression corresponded to expression of gastrulation- associated gene profile, while downregulation of Nodal expression corresponded to expression of neurulation-associated gene profile. Given these observations, we hypothesized that inhibiting Nodal while differentiating geometrically-confined hPSC colonies in response to BMP4, would induce peri-neurulation-like fate patterning, and observed experimental results consistent with this hypothesis. Mechanistically, we demonstrate a reaction-diffusion mediated self-organization of phosphorylated-SMAD1 signaling, and a positional-information mediated patterning of neurulation-associated fates. Our work identifies Nodal signaling dependent switch in gastrulation versus neurulation-associated fate patterning in hPSC colonies and hints towards possible conserved mechanisms of self-organized fate specification of the differentiating epiblast and ectoderm tissues.

4.2 Introduction

Following implantation, human embryos undergo a dramatic transformation mediated by tissue growth, cell movements, morphogenesis, and fate specifications resulting in the self-organized formation of the future body plan(Hertig et al. 1956). Post-implantation development to the neurula stage embryo is orchestrated by two vital developmentally conserved events called gastrulation and neurulation. Gastrulation refers to the stage that segregates the pluripotent epiblast into the three multipotent germ layers, namely – the ectoderm, the mesoderm, and the endoderm(Tam & Loebel 2007; Tam et al. 2006; Rossant & Tam 2004). Closely following gastrulation, the ectoderm undergoes further fate specification resulting in the patterned neural plate, neural plate border, and non-neural ectoderm regions thereby setting the stage for the onset of neurulation(Nikolopoulou et al. 2017; Greene & Copp 2014; Greene & Copp 2009; Groves & LaBonne 2014; Simões-costa & Bronner 2013). As neurulation proceeds, morphogenetic changes in these tissues result in the formation of the neural tube, the neural crest, and the epithelium respectively. Initiation of the morphogenetic restructuring of the epiblast and the ectoderm occurs due to self-organized gradients of signaling molecules called morphogens, and morphogens belonging to the transforming growth factor beta (TGFβ) superfamily, such as bone morphogenetic proteins (BMPs) and Nodal, play vital roles in these developmental stages.

Two biochemical models, Reaction-Diffusion (RD) and Positional-Information (PI), have influenced our mechanistic understanding of self-organized fate specification during embryogenesis. The RD model describes how a homogenously distributed morphogen can self- organize into a signaling gradient in a developing tissue. The classical version of the model hypothesizes the presence of an interaction network between the morphogen and its inhibitor, both of which are diffusible albeit with differential diffusivities(Turing 1952; Gierer & Meinhardt 1972; Green & Sharpe 2015). Recent interpretations of RD have proposed that higher order (>2 molecules) network topologies underlie this self-organization(Marcon et al. 2016). The PI model describes fate patterning in a developing tissue as a result of an asymmetric morphogen distribution. The classical version of this paradigm hypothesized that the cells in the developing tissue sense the morphogen concentration in their immediate vicinity and acquire fates according to a threshold model(Wolpert 1969; Wolpert 1981). Recent studies have updated this interpretation of the PI model. Current interpretation of PI suggests that fates are acquired as a function of both the morphogen concentration and time of induction(Dessaud et al. 2007; Dessaud et al. 2008).

Studying post-implantation developmental events, like gastrulation and neurulation, directly in human embryos would unequivocally provide the most reliable interpretations of human development. However, while human embryos have recently been cultured in vitro past the implantation stage(Deglincerti, Croft, et al. 2016; Shahbazi et al. 2016), ethical concerns preclude their maintenance beyond 14 days – prior to the onset of gastrulation. Therefore, human embryos are not currently available for direct investigation of the mechanisms underpinning post- implantation embryonic development. Nevertheless, recent studies have demonstrated the remarkable ability of stem cells to self-organize into structures in vitro that mimic aspects of post- implantation human development, when provided the biophysical and biochemical cues that mimic the microenvironment of the associated stage of embryogenesis(Kretzschmar & Clevers 2016; Lancaster et al. 2012; Tewary et al. 2017; Warmflash et al. 2014; Etoc et al. 2016). Consequently, bioengineered in vitro experimental models that use stem cells to recapitulate development have been gaining significant interest. We and others have used human pluripotent stem cells (hPSCs) to demonstrate that BMP4 treatment of geometrically confined hPSC colonies recapitulate numerous aspects of human peri-gastrulation-like self-organized fate patterning(Warmflash et al. 2014; Etoc et al. 2016; Tewary et al. 2017). We recently reported that this peri-gastrulation- associated fate patterning was mediated by a stepwise process of RD followed by PI(Tewary et al. 2017).

Here, we report a high-throughput platform to produce microtiter plates that allow robust patterning of a variety of cell types. Employing this platform, we tested the response of a variety of hPSC lines to a previously reported peri-gastrulation-like assay and observed significant variability in the induction of the Brachyury (BRA) expressing region between the lines. To probe the emergent differentiation trajectories of hPSC lines, we generated embryoid bodies (EBs) from a panel of test lines and cultured them in conditions unsupportive of pluripotency, assessing differentiation-associated emergent gene expression profiles daily. Our findings revealed a switch- like response in the upregulation gene expression profiles either associated with gastrulation or with neurulation. This switch in emergent gene expression showed a strong trend with the temporal dynamics of Nodal signaling – where the hPSC lines that exhibited higher levels of a gastrulation- associated gene expression profile also exhibited upregulated Nodal signaling, and those that exhibited a higher neurulation-associated gene expression profile exhibited downregulated Nodal signaling. We further report that geometrically-confined hPSC colonies induced to differentiate in

95 the presence of BMP4 and a Nodal inhibitor undergo an RD-mediated self-organization of pSMAD1 activity and PI-mediated fate patterning into compartments that express markers of the differentiating ectodermal progenitors. We further demonstrate the ability of these progenitor regions to induce marker expression of the definitive fates of the respective compartments.

4.3 Results

4.3.1 Screening hPSC lines on PEG plates for peri-gastrulation-like patterning response yields variable induction of the primitive-streak- like compartment

Recent studies have reported that BMP4 treatment of geometrically confined hPSC colonies result in self-organized fate patterning of gastrulation-associated markers(Warmflash et al. 2014; Etoc et al. 2016; Tewary et al. 2017). Notably, these studies demonstrated that the differentiating geometrically-confined hPSC colonies gave rise to a Brachyury (BRA) expressing compartment, representing a primitive-streak-like identity. Given that lineage-specific differentiation potential between hPSC lines is known to vary widely(Ortmann & Vallier 2017; Nazareth et al. 2013; Keller et al. 2018), we hypothesized that different hPSC lines would induce the primitive-streak-like compartment at different efficiencies. We employed our high-throughput PEG plates to evaluate the response of BMP4-treatment of geometrically-confined hPSC colonies (1mm in diameter) in a screen of the following five hPSC lines: H9-1, H9-2, HES2, MEL1, and HES3-1. The induction medium employed for this screen, and all subsequent experiments (unless otherwise stated) was a Knockout Serum Replacement based medium supplemented with BMP4 and bFGF (see Materials and Methods for composition). Although all hPSC lines tested expressed high levels of pluripotency markers at the start of the differentiation culture (Figure 4-2), induction of BRA expression levels varied drastically between hPSC lines at 48h after BMP4 treatment (Figure 4-1). Notably, although the MEL1, and HES3-1 lines were unable to induce the expression of BRA, they did differentiate as indicated by the reduction of SOX2 expression relative to the starting population (Figure 4-1, Figure 4-2). These data indicate that although all hPSC lines tested differentiated upon BMP4 treatment, induction of the primitive-streak-like compartment, as indicated by BRA expression in the peri-gastrulation-like model, varied considerably.

4.3.2 Undirected differentiation of hPSC lines yields gastrulation or neurulation associated gene expression correlated with upregulated or downregulated Nodal signaling

We hypothesized that differences in regulation of key signaling pathways controlling mesendodermal induction between the tested hPSC lines underlay the variation in BRA expression observed in the peri-gastrulation-like patterning. To test this hypothesis, we employed a recently reported approach that addressed a similar question in mouse epiblast stem cell (mEpiSC) lines(Kojima et al. 2014). In their study, Kojima et al made embryoid bodies (EBs) out of various mEpiSC lines, allowed them to spontaneously differentiate in culture conditions unsupportive of pluripotency, and assayed for the expression of differentiation associated genes to compare the transcriptional and functional profiles between the lines(Kojima et al. 2014). Employing a similar approach, we generated EBs from nine hPSC lines – H9-3, H1, H7, HES3-2 added to the previous panel (complete list of lines and their respective culture conditions shown in Table 4-1) – and cultured them in conditions unsupportive of pluripotency for three days and analyzed differentiation marker gene expression levels daily (henceforth – ‘EB assay’) (Figure 4-3A). We observed strong variation in expression dynamics of differentiation associated genes between the test hPSC lines (Figure 4-3Bi). To simplify data interpretation, we used unsupervised K-means

Figure 4-1: Variability in peri-gastrulation-like induction observed between test hPSC lines. A-B) Quantified expression of BRA (A), and SOX2 (B) observed within different hPSC lines tested. Number of colonies are 252, 245, 327, 288, and 304 for GKH9, 7TGP, HES2, PDX1-GFP, and MIXL1-GFP respectively. Each data point represents individual colonies identified. Data pooled from two experiments. B) Representative immunofluorescent images for BRA, SOX2, and CDX2 for the test hPSC lines. Scale bar represents 200μm

Figure 4-2: Starting populations of test hPSC lines show high expression of pluripotency associated proteins. FACS plots of OCT4, SOX2, and NANOG of starting populations of H9-1, H9-2, MEL1, and HES3- 1.

Table 4-1: hPSC lines utilized in this study * The wnt reported line was generated using a puromicine selection cassette. We did not perform any selection and consequently, the wnt activity reporter was undetectable in any of our experiments

hPSC lines Parental Reporter? Culture Culture Source Source Conditions medium

H1 H1 No Feeders KOSR based WiCell

H7 H7 No Feeders KOSR based WiCell

H9-1 H9 Yes* Feeder Free mTeSR Dr. Sean Palecek

H9-2 H9 No Feeder Free mTeSR Dr. Gordon Keller

H9-3 H9 No Feeders KOSR based WiCell

HES2 HES2 No Feeders KOSR based Dr. Gordon Keller

HES3-1 HES3 Yes – MIXL1 Feeders KOSR based Dr. Andrew Elefanty

HES3-2 HES3 Yes – RUNX1 Feeders KOSR based Dr. Andrew Elefanty

MEL1 MEL1 Yes – RUNX1 Feeders KOSR based Dr. Gordon Keller

CA1 CA1 No Feeder free mTeSR Dr. Andras Nagy

100

clustering to segregate the hPSC lines into ‘Strong’, ‘Intermediate’, and ‘Weak’ expressers for each gene tested. Interestingly, this analysis revealed a switch-like response in the test lines where some lines upregulated expression of genes associated with gastrulation while others upregulated expression of neurulation-associated genes (Figure 4-3Bii). In a recent study, Funa et al showed that wnt signaling mediated differentiation of hPSCs results in fate acquisition that is dependent on Nodal signaling(Funa et al. 2015). Specifically, the authors demonstrated that presence of Nodal signaling during wnt mediated differentiation of hPSCs resulted in the acquisition of a primitive streak fate, whereas the absence of Nodal signaling during wnt mediated differentiation resulted in the induction of the neural crest fate(Funa et al. 2015). Given that the primitive streak is a gastrulation-associated fate, neural crest arises during neurulation, and the fact that our data demonstrated a gastrulation versus neurulation switch in differentiating hPSC lines, we hypothesized that Nodal signaling could be dissecting these gene expression profiles in our EB assay. Consistent with this hypothesis, the dynamics of Nodal and GDF3 (a Nodal target) in the differentiating hPSC lines showed a strong trend indicative of their upregulation linked with the induction of gastrulation-associated genes and their downregulation linked with the induction of neurulation-associated genes (Figure 4-3Ci). Furthermore, clustering the hPSC lines with reference to the dynamics of Nodal and GDF3 by either unsupervised K-means clustering (Figure 4-3Cii, Figure 4-4), or by hierarchical clustering based on Euclidian distance (Figure 4-5), indicated that upregulation of Nodal and GDF3 coincided with gastrulation-associated gene expression, whereas their downregulation corresponded with neurulation-associated gene expression.

4.3.3 Validation of variable gene expression in EB assay

We next sought to validate the gene expression level differences observed in the differentiating hPSC lines in the EB assay by asking if the variation translated to cell fate acquisition during directed differentiation. Given the key role that MIXL1 plays in the induction of definitive endoderm(Hart et al. 2002), Kojima et al investigated the temporal dynamics of Mixl1 in their EB

101

102

Figure 4-3: Nodal dissects gastrulation and neurulation associated gene expression profiles. A) Overview of experimental setup for Embryoid Body (EB) assay. EBs were made from each test hPSC line and allowed to spontaneously differentiate in presence of Fetal Bovine Serum (FBS) for three days. B) Observed gene expression dynamics of test cell lines when differentiated as EBs in FBS. i) Observed gene expression for a panel of differentiation associated genes (shown under ‘Gastrulation’ and ‘Neurulation’ groups) along with POU5F1 and NANOG. Data shown as heatmap of mean expression of each day from three biological replicates (s.d. not shown), represented as log2(Fold Change) relative to the D0 sample of respective hPSC line. ‘Pluri’ indicates the pluripotency associated genes. ii) Heatmap representation of (B(i)) with the panel of hPSC lines clustered into three groups of ‘Strong’, ‘Intermediate’, and ‘Weak’ responders for each gene using unsupervised K-means clustering. ‘Pluri’ indicates the pluripotency associated genes. C) Nodal dynamics during EB assay. i) Observed gene expression of Nodal and a Nodal signaling target (GDF3). Data shown as heatmap of mean expression of each day from three biological replicates (expression levels for individual replicates shown in Fig. S3B), represented as log2(Fold Change) relative to the D0 sample of the respective hPSC line. ii) Heatmap representation of (C(i)) with the panel of hPSC lines clustered into three groups of ‘Strong’, ‘Intermediate’, and ‘Weak’ responders for Nodal, and GDF3 using unsupervised K-Means clustering. D) Effect of modulation of Nodal during previously reported peri-gastrulation-like assay using geometrically-confined colonies of the ‘CA1’ hPSC line(Tewary et al. 2017). i) Overview of experimental setup. Geometrically- confined colonies of CA1s were induced to differentiate for three days, with either a two-day pulse of BMP4 and Nodal, and just Nodal for the third day; or a two-day pulse of BMP4 and an inhibitor of Nodal signaling (SB431542 – ‘SB’), and just SB for the third day. The vehicle employed in this experiment was SR medium (see Materials and Methods for composition). ii) Heatmap representation of a panel of differentiation genes associated with either gastrulation, or neurulation. Dark blue represents higher levels of expression, whereas light blue represents lower levels of expression. Data shown as mean of three biological replicates. Expression levels of individual replicates shown in Fig. S7.

assay with mEpiSCs, and demonstrated that Mixl1 expression dynamics were able to predict the endodermal differentiation bias of the mEpiSC lines(Kojima et al. 2014). Importantly, much like MIXL1, EOMES is also known to play an important role in the endoderm specification(Bjornson et al. 2005; Teo et al. 2011). Consistent with this idea, the ‘Strong’, ‘Intermediate’, and ‘Weak’ responders of the panel of hPSCs for both MIXL1, and EOMES contained the identical hPSC line cohorts (Figure 4-3Bii, Figure 4-6), suggesting the likelihood of parallel functions of both these genes in differentiating hPSCs. To validate the gene expression profiles observed in our EB assay, we asked if the observed temporal dynamics of these genes that critically regulate endoderm specification were able to predict the propensity of the hPSC lines to differentiate toward the definitive endodermal fates. Consequently, we differentiated the panel of hPSCs toward definitive endoderm using an established protocol (Figure 4-6B), and consistent with the findings of Kojima et al(Kojima et al. 2014), the temporal dynamics of endoderm specifiers (MIXL1 and EOMES in

103 our case) in our EB assay closely matched the propensity of the hPSC lines to induce SOX17 expression upon directed differentiation toward the definitive endoderm fate (Figure 4-6C,D). The expression dynamics of MIXL1 and EOMES in the EB assay were also able to predict the induction efficiency of mature endodermal fates. Specifically, lines from the MIXL1/EOMES- Strong cluster outperformed candidate lines from the MIXL1/EOMES-Weak cluster in the induction of pancreatic progenitors as marked by the co-expression of PDX1 and NKX6.1 (Figure 4-6E). These data provide protein level phenotypic validation of the variable gene expression observed in our EB assay.

Given that geometrically confined hPSC colonies are able to induce organized fate patterning(Tewary et al. 2017; Warmflash et al. 2014; Etoc et al. 2016), we asked if subjecting geometrically-confined hPSC colonies to defined endodermal differentiation conditions could be used as an assay to predict the differentiation propensity of hPSC lines. We selected three hPSC lines – H9-1, HES3-2, and HES3-1 – to represent each MIXL1/EOMES induction compartment defined in the EB assay (Figure 4-3Bii, Figure 4-6A), and differentiated them as geometrically- confined colonies in defined endodermal induction conditions (Figure 4-7A). Interestingly, we found that the relative efficiency of SOX17 and FOXA2 double-positive expression under these experimental conditions closely matched the endodermal lineage-bias of the lines as predicted by the EB assay (Figure 4-6A-D, Figure 4-7B,C). Taken together, these data validate the differential gene expression observed between the panel of hPSC lines by demonstrating congruence between MIXL1 and EOMES temporal dynamics and endoderm lineage bias of hPSC lines and provide proof-of-concept data that the defined differentiation protocols in geometrically-confined hPSC colonies can be used to fingerprint the lineage bias of hPSCs.

104

Figure 4-4: Nodal expression dynamics in FBS mediated non-specific differentiation of hPSC embryoid bodies. Temporal dynamics of Nodal for the test hPSC lines shown for the three clusters of Nodal- Strong, Nodal-Intermediate, and Nodal-weak (Figure 4-3Bii). Each data point represents the detected expression level for a biological replicate. Bar plots represent mean ± s.d.

105

Figure 4-5 Hierarchical clustering of Nodal and GDF3 is consistent with unsupervised K-means clustering. A) Hierarchical clustering of the Nodal expression in the test hPSC lines based on Euclidian distance reveals similar clusters as the ones from unsupervised K-means clustering. B) Hierarchical clustering of the Nodal target (GDF3) expression in the test hPSC lines based on Euclidian distance reveals similar clusters as the ones from unsupervised K-means clustering of Nodal signaling.

106

Figure 4-6 MIXL1 and EOMES dynamics during EB assay predict endoderm differentiation propensity of hPSC lines.

107

A) Panel of hPSC lines clustered into three groups of ‘Strong’, ‘Medium’, and ‘Weak’ responders for (i) MIXL1, and (ii) EOMES from Figure 4-3Bii. The expression levels of MIXL1 and EOMES in the pluripotent state (Day 0) shown in the boxes adjacent to the heatmaps. B) Overview of the protocol for directed differentiation toward definitive endoderm. The cells were treated with Wnt3a from 0h-24h, and Wnt3a+ActivinA from 24h-72h. C-D) Efficiency of SOX17 induction in the test hPSCs using the protocol in B). C) Black dash denotes the mean of three independent replicates represented by the dots. D) FACS plots for individual replicates from C). E) FACS plots showing the efficiency of induction of pancreatic progenitors as indicated by the expression of PDX1, and NKX6.1 for hPSC lines in ‘Strong’ and ‘Weak’ clusters from A). The differentiation was performed using a previously described protocol(Nostro et al. 2015). The data are from one biological replicate.

108

Figure 4-7: Peri-gastrulation-like assay predicts endoderm differentiation bias of hPSC lines. A) Overview of assay for predicting endodermal differentiation bias. Geometrically-confined hPSC lines were treated with Wnt+ActivinA for 48hours prior to fixation and staining. B) Quantified fraction of endodermal cells, defined as double positive for SOX17, and FOXA2, detected within the geometrically-confined hPSC colonies. The hPSC lines chosen were one each from the ‘Strong’, ‘Medium’, and ‘Weak’ clusters of MIXL1, and EOMES from Figure 4-6A. Each data point represents an individual identified colony. Data pooled from two different experiments and represented as mean ± s.d.; p-values calculated using one-way ANOVA (Kruskal-Wallis test). C) Representative immunofluorescent images of colonies from B) stained for DAPI, SOX17, and FOXA2. Scale bar represents 500μm

109

4.3.4 Nodal dissects gastrulation versus neurulation associated gene expression profiles in geometrically-confined hPSC colonies

Our data, thus far, showed that in conditions that do not support pluripotency, differentiating embryoid bodies made from hPSC lines assume a transcriptional state associated with either gastrulation or neurulation, and endogenous Nodal dynamics correlated with this switch. However, whether the differential Nodal dynamics caused the switch in the acquired transcriptional state or if the association was merely correlative remained unclear. Given that BMP4 treatment has been previously reported to induce gastrulation-associated fate patterning(Warmflash et al. 2014; Etoc et al. 2016; Tewary et al. 2017), and that we have previously demonstrated the robust response of peri-gastrulation-like fate acquisition in the CA1 hPSC line(Tewary et al. 2017), we revisited the peri-gastrulation-like model in hPSC colonies to test if Nodal signaling had a direct effect in regulating this switch. We asked if inducing geometrically confined colonies of the CA1 line to differentiate in response to BMP4 either in the presence or absence of a small molecule inhibitor of Alk4/5/7 receptors (SB431542, hereafter ‘SB’) which antagonizes Nodal signaling (Figure 4-3Di) recapitulated the observed switch in emergent gene expression. Excitingly, after a three- day induction, we observed that colonies grown in the presence of SB upregulated genes associated with neurulation whereas those grown in the absence of SB upregulated genes associated with gastrulation (Figure 4-3Dii, Figure 4-8). These results are consistent with our hypothesis that Nodal signaling dissects gastrulation and neurulation-associated gene expression profiles in differentiating hPSCs. Given that differentiating hPSCs in the absence of Nodal signalling upregulated neurulation associated genes, we next set to investigate if BMP4 treatment of geometrically confined hPSC colonies in presence of SB gave rise to the neurulation-associated spatially patterned fate allocation.

110

Figure 4-8: Modulation of Nodal signaling during BMP4 treatment of geometrically-confined hPSC colonies. Response of modulation of Nodal signaling in expression of gastrulation versus neurulation associated genes in BMP4 treated geometrically-confined CA1 colonies (assay details in Figure 4-3Di). Individual data points represent biological replicates. Data shown as mean ± s.d., and p-values calculated using Mann-Whitney U test.

111

4.3.5 An RD network in BMP signaling can self-organize pSMAD1 activity independent of Nodal

In a recent study, we demonstrated that the peri-gastrulation-like fate patterning in geometrically confined hPSC colonies occurs via a stepwise process of RD and PI where a BMP4-Noggin RD network self-organizes a phosphorylated SMAD1 (pSMAD1) signaling gradient within the colonies, resulting in the peri-gastrulation-like fates being patterned in a manner consistent with the PI paradigm(Tewary et al. 2017). We set out to investigate if a conserved mechanism would give rise to neurulation-associated fate patterning. As a first step, we asked if a BMP4-Noggin RD network governed pSMAD1 self-organization within the geometrically-confined hPSC colonies treated with BMP4 and SB. Consistent with the presence of a BMP4-Noggin RD network(Tewary et al. 2017), we observed an upregulation of both BMP4 and Noggin upon BMP4 treatment of hPSCs in the presence of SB (Figure 4-9A). We next asked if BMP4 treatment of geometrically- confined hPSC colonies in the presence of SB would result in the self-organized gradient of nuclear localized pSMAD1. Indeed, pSMAD1 activity within the colonies rapidly self-organized into a radial gradient under these experimental conditions (Figure 4-9B-D). In our previous study using the peri-gastrulation-like model, we showed that a BMP4-Noggin RD computational model predicts the experimentally observed responses of a pSMAD1 self-organized gradient at the periphery and the center of the colonies to perturbations to the BMP4 dose in the induction medium and size of the geometrically-confined hPSC colony(Tewary et al. 2017). Specifically, we showed that reducing the BMP4 dose while maintaining the colony size reduces the levels of pSMAD1 at the periphery, and reducing the colony size while maintaining a constant BMP4 dose in the induction medium results in an increase of pSMAD1 levels at the center of the colonies(Tewary et al. 2017). We reasoned that a conserved mechanism underlying the pSMAD1 self-organization would result in identical responses to these perturbations. Consistent with our anticipated results, reducing the BMP4 dose in the induction medium while maintaining the colony size resulted in a reduction of the detected immunofluorescent levels of nuclear localized pSMAD1 at the colony periphery (Figure 4-10A-C). Furthermore, reducing the colony size while maintaining the BMP4 dose in the induction medium increased the detected immunofluorescent levels of nuclear localization of pSMAD1 at the colony centers (Figure 4-10D-E). Taken together, these data demonstrate that in absence of Nodal signaling, pSMAD1 activity in the geometrically-confined hPSC colonies self-organizes into a signaling gradient and suggest that a BMP4-Noggin RD

112 system governs this observation (Figure 4-9H) – consistent with our previous study(Tewary et al. 2017).

Figure 4-9 Interaction network between BMP4-Noggin underlies self-organization of A stepwise model of reaction-diffusion and positional information patterns peri-neurulation- associated fates. A) Temporal gene expression for BMP4 and Noggin at 4h, 14h, 20h, and 24h after BMP4 treatment. Data shown as mean ± s.d. of three independent experiments. The p-values shown were calculated using Kruskal-Wallis test. B) Representative immunofluorescent images of geometrically confined hPSC colonies of 500μm in diameter stained for pSMAD1 after different times (0h, 6h, 12h, 18h, and 24h) of BMP4 exposure. Scale bar represents 200μm. C) Average pSMAD1 intensity represented as overlays of 231, 241, 222, 238, and 228 colonies for respective induction times. Data pooled from two experiments. D) The average radial trends of pSMAD1 at each duration shown as line plots. Standard deviations shown in grey, and 95% confidence intervals shown in black.

113

Figure 4-10: pSMAD1 gradient formation is consistent with a BMP4-Noggin RD network mediated self-organization. A) (i-ii) Radial gradient formed in colonies of 500μm diameter treated with varying doses of BMP4 in induction medium (3.125ng/ml, 6.25ng/ml, 12.5ng/ml, and 25ng/ml) represented as line plots. (i) The gradients shown individually. Data pooled from two experiments, and 114

represent 299, 293, 302, and 343 colonies for the respective doses. Standard deviations shown in grey and 95% confidence intervals shown in black. (ii) Line plots shown in one graph for comparison of pSMAD1 levels at colony periphery. B) Average pSMAD1 expression levels shown as overlay of the detected colonies (numbers mentioned in A). C) Representative immunofluorescent images of pSMAD1 for respective conditions. Scale bars represent 200μm. D) Average pSMAD1 expression of 987, 528, 280, 182, 107, and 89 colonies for varying colony sizes (200μm, 300μm 400μm, 500μm, 600μm, and 700μm) treated with 25ng/ml of BMP4 in induction medium. Data pooled from two experiments. Standard deviations shown in grey and 95% confidence intervals shown in black. Scale bars represent 200μm.

4.3.6 Nodal signaling contributes to the shape of the self-organized pSMAD1 gradient

Our data indicate that the pSMAD1 signaling gradient self-organizes via an RD network present in the BMP signaling pathway where Noggin functions as a key inhibitor (Figure 4-9). Given that Nodal signaling targets include multiple BMP antagonists such as Cer1, GDF3, Follistatin (FST), etc.(Mulloy & Rider 2015), we asked if Nodal signaling contributed to the formation of the pSMAD1 signaling gradient in BMP4-treated geometrically confined hPSC colonies. To probe the role of Nodal in the observed pSMAD1 self-organization, we compared the formation of the pSMAD1 gradient in BMP4 treated geometrically confined hPSC colonies of 500μm diameter where the induction media either did or did not contain SB (Figure 4-11A). The results from this experiment provided two notable observations. First, in the vehicle control, we observed prominent spatial oscillations of pSMAD1 expression (Figure 4-11B-D). This observation is in agreement with the proposition that the self-organization of pSMAD1 arises via an RD mechanism which can result in spatial oscillations of morphogen activity, and provides further support for the RD hypothesis(Tewary et al. 2017). The second notable observation was that the pSMAD1 gradient formed between these conditions were dramatically different (Figure 4-11B-D). A prominent change was the level of immunofluorescence detected at the colony periphery; in the presence of SB, the level of pSMAD1 fluorescence at the colony periphery was significantly higher than the levels detected in the vehicle control (Figure 4-11B-D). This observation was also noted when the colonies in the Vehicle and SB media were treated with a higher dose (50ng/ml) of BMP4 (Figure 4-12A-C). Taken together these data indicate that Nodal signaling can contribute to the formation of the pSMAD1 signaling gradient.

115

Since an RD network results in the morphogen gradient as a consequence of the expression of both activators and inhibitors of the morphogen (Turing 1952; Gierer & Meinhardt 1972), we hypothesized that the likely reasons for this observation could be due to either a change in the amount of activator (change in BMP4 levels) or the amount of inhibitor (change in the level of BMP antagonists) present in the system. In support of both of these possibilities, when we tested gene expression of activators and inhibitors after 24h of Vehicle versus SB treatment on hPSCs which either allowed Nodal expression or dramatically downregulated it (Figure 4-12D), SB treatment provoked an increased positive feedback as indicated by increased detected levels of BMP4 transcripts (Figure 4-11E); and a reduced negative feedback as indicated by significantly reduced transcript levels of BMP antagonists like CERL, GDF3, and FST (Figure 4-11F). Taken together, these data suggest that Nodal signaling can contribute to the RD-mediated self- organization of the pSMAD1 signaling gradient; however, the pSMAD1 signaling gradient forms in the absence of Nodal signaling as well. Having established that pSMAD1 activity in the geometrically confined hPSC colonies treated with BMP4 and SB self-organizes into a signaling gradient, we next focused on investigating if this gradient induced the expression of fates associated with the differentiating ectoderm.

116

Figure 4-11: Nodal signaling contributes to the formation of the pSMAD1 gradient. A) (i-ii) Overview of experimental setup. (i) Geometrically confined hPSC colonies were treated with BMP4 for 24h. (ii) Media tested. ‘Vehicle’ indicated SR medium (see Materials and Methods for composition) supplemented with BMP4 and bFGF. ‘SB’ indicated vehicle 117

supplemented with 10μM SB431542. B-D) Inhibition of Nodal signaling results in a significant change in the pSMAD1 self-organized gradient formation. B) Average pSMAD1 intensity represented as overlays of colonies pooled from two experiments. C) (i-ii) The average radial trends of pSMAD1 shown as line plots. (i) Line plots shown individually for SB and NODAL conditions. Standard deviations shown in grey, and 95% confidence intervals shown in black. (ii) Line plots represented in the same graph. D) Representative immunofluorescence images of 500μm diameter hPSC colonies stained for pSMAD1 after 24h of BMP4 treatment in ‘Vehicle’ and ‘SB’ conditions (average response shown in B). Scalebar represents 200μm. White arrows indicate regions where second peak of pSMAD1 appears. White triangles indicate regions of discernable pSMAD1 levels that appear to be lower than the levels at the colony periphery. E) Gene expression for BMP4 after 24h of treatment with either ‘Vehicle’ of ‘SB’ media (described in A(ii)) of hPSCs. Data shown as mean + s.d. (n=3, technical replicates, independent wells). The p-value was calculated using two-sided t-test. F) Gene expression of Activin-Nodal pathway associated targets that are known antagonists of BMP signaling (CERL, GDF3, Follistatin – ‘FST’). The data represented as mean ± s.d. of hPSCs from three independent wells. The experiment was performed once. The p-values were calculated using two-sided Student’s t-test.

118

Figure 4-12 Nodal signaling contributes to the formation of the pSMAD1 gradient. A) Representative immunofluorescence images of 500μm diameter hPSC colonies stained for pSMAD1 after 24h of BMP4 treatment. Scalebar represents 200μm. B) Average pSMAD1 intensity pooled from two experiments. C) (i-ii) The average radial trends of pSMAD1 shown as line plots. (i) Line plots shown individually for SB and NODAL conditions. Standard deviations shown in grey, and 95% confidence intervals shown in black. (ii) Line plots represented in the same graph. D) (i-ii) SB supplementation in the induction medium robustly inhibits Nodal signaling. (i) Geometrically confined hPSC colonies were treated with BMP4 for 24h. Media tested were: ‘Vehicle’ indicated SR medium (see Materials and Methods for composition) supplemented with BMP4 and bFGF. ‘SB’ indicated vehicle supplemented with 10μM SB431542. (ii) Gene expression Nodal and Lefty-A (a Nodal target) after 24h of treatment with

119

either ‘Vehicle’ of ‘SB’ media. Data shown as mean ± s.d. (n=3, technical replicates, independent wells). The p-value was calculated using two-sided Student’s t-test.

4.3.7 Pre-neurulation-like fate patterning arises in a manner consistent with PI

In the presence of Nodal signaling, BMP4 treatment of geometrically-confined hPSC colonies results in self-organized pSMAD1 gradient and a spatially patterned acquisition of gastrulation associated fates (Tewary et al. 2017). Although we observe the formation of the pSMAD1 gradient when geometrically-confined hPSC colonies are treated with BMP4 and SB (Figure 4-9, Figure 4-11), we did not observe expression of key gastrulation associated markers like BRA, EOMES, SOX17, and GATA6 (Figure 4-13A-B). These observations are consistent with the need of Nodal in inducing gastrulation-associated fates (Tewary et al. 2017). Since we observed that differentiating hPSCs in the absence of Nodal signaling upregulate a neurulation-associated gene profile (Figure 4-3D), we asked if BMP4 and SB treatment of geometrically confined hPSC colonies resulted in the fate patterning associated with the differentiating ectoderm. After the germ layers segregate from the epiblast, a BMP signaling gradient along the medial-lateral axis in the developing ectoderm patterns the early pre-neural (PN) tissue at the medial end, and non-neural (NN) tissue at the lateral end appropriately arranging the tissue for the onset of neurulation(Groves & LaBonne 2014). The PN tissue gives rise to the neural plate (NP), which later folds to form the neural tube (Nikolopoulou et al. 2017; Greene & Copp 2009), and the NN tissue gives rise to the non-neural ectoderm (NNE) and the neural plate border (NPB)(Groves & LaBonne 2014). The NNE subsequently specifies to generate the epidermis and the NPB is a multipotent tissue that produces the neural crest (NC) and the craniofacial placodes in the anterior ectoderm(Groves & LaBonne 2014; Plouhinec et al. 2017; Simoes-Costa & Bronner 2015; Pieper et al. 2012; Simões- costa & Bronner 2013). The early PN region maintains the expression of SOX2 which is present in the epiblast, and the early NN regions induce expression of markers like GATA3(Groves & LaBonne 2014). Furthermore, markers like transcription factor AP2-alpha (TFAP2A) mark the NN, NNE, and maturing NPB region that marks the NC fate; and SIX1 are expressed in the maturing NPB region which marks panplacodal competent tissues(Groves & LaBonne 2014). Consistent with our observation that BMP4 treatment in the absence of Nodal signaling upregulated genes associated with neurulation, we observed spatially segregated expression of

120

SOX2 (PN), and GATA3 (NN) with concomitant expression of TFAP2A (NN, NPB), and SIX1 (panplacodal competent NPB) (Figure 4-13C). We define this fate patterning as ‘pre-neurulation- like’ and using SOX2, and GATA3 as the markers of the PN, and NN tissues, we set out of test if the fate patterning arose in a manner consistent with the positional information (PI) paradigm.

Given that the PI paradigm posits that developmental fates arise due to thresholds of morphogen levels, and we asked if the perturbations of pSMAD1 levels at the colony periphery (Figure 4-10A- C) and the colony center (Figure 4-10D-E) resulted in pSMAD1 threshold mediated changes in expression of GATA3 (NN), and SOX2 (PN) fates respectively. Consistent with the idea of a pSMAD1 threshold dependent patterning of the PN and the NN tissues marked by GATA3, and SOX2, we find that reducing the pSMAD1 levels at the colony periphery (Figure 4-14A) significantly reduced the GATA3 expression at the colony periphery (Figure 4-14B-C) and increasing the pSMAD1 levels at the colony center (Figure 4-14D) dramatically reduced the SOX2 expression (Figure 4-14E-F). These data indicate that thresholds of pSMAD1 regulated the patterning of the SOX2 and GATA3 within the geometrically confined hPSC colonies. However, the formalization of the PI paradigm has been updated to include time as a critical parameter that patterns the developmental cell fates. Specifically, fate patterning mediated by PI is known to arise as a function of the morphogen concentration and time of induction(Dessaud et al. 2008; Dessaud et al. 2007; Briscoe & Small 2015; Tewary et al. 2017). Consequently, we tested four different doses of BMP4 (3.125ng/ml, 6.25ng/ml, 12.5ng/ml, and 25ng/ml) in the induction medium for four different induction times (12h, 24h, 36h, and 48h) and measured the levels of SOX2 and GATA3 detected. We observed that the fate patterning of GATA3 arose as a function of both the concentration of BMP4 in the induction medium and the time of induction (Figure 4-15A-B, C(i), Figure 4-16) indicating that the patterning within the geometrically confined colonies arises in a manner consistent with PI (Figure 4-16C(ii)).

121

Figure 4-13: Nodal inhibition during BMP4 treatment of hPSC colonies abrogates peri- gastrulation-associated fates and induces pre-neurulation-associated fates.

122

A) (i-ii) Overview of experimental setup. (i) Geometrically confined hPSC colonies were treated with BMP4 for 48h. (ii) Media tested. ‘Vehicle’ indicated SR medium (see Materials and Methods for composition) supplemented with BMP4 and bFGF. ‘SB’ indicated vehicle supplemented with 10μM SB431542. B) (i-ii) Response of gastrulation-associated fate patterning in Vehicle and SB conditions. (i) Representative immunofluorescent images for BRA, GATA6, EOMES, and SOX17 for Vehicle and SB conditions. White triangle represents non- specific background staining for EOMES. Scale bar represents 200μm. (ii) Quantified expression of gastrulation-associated fates. Each data point represents an identified colony. The total number of colonies were (373,248), (325, 317), (81,72), and (506, 329) for BRA, GATA6, EOMES, and SOX17 respectively for (Vehicle and SB treatments). The data are pooled from two experiments except for EOMES, which was performed once. C) (i-ii) Pre-neurulation-like fate patterning observed in the presence of SB. (i) Representative immunofluorescent images of TFAP2A, SIX1, OTX2, and co-stained image of SOX2, and GATA3. Scale bar represents 200μm. (ii) Average radial expression intensity of the pre-neurulation-associated fates represented as line plots. Standard deviation shown in grey, and 95% confidence intervals shown in black.

123

Figure 4-14: SOX2 and GATA3 expression is consistent with a pSMAD1 dose-dependent fate patterning. A) Overview of experimental setup. Perturbing BMP4 dose in induction medium while maintaining colony size varies pSMAD1 concentration levels at the colony periphery (see Fig S9A-C for details). B) Percentage of cells in each identified colony expressing GATA3 when colonies of 500μm in diameter were treated with varying doses of BMP4 (3.125ng/ml, 6.25ng/ml, 12.5ng/ml, and 25ng/ml) in induction medium. Each data point represents an

124

identified colony. The total number of colonies were 131, 208, 215, and 244 for the respective doses. Data pooled from two experiments. Bars represent mean ± s.d. The p-value was calculated using Kruskal-Wallis test. C) Representative immunofluorescent images of colonies stained for SOX2 and GATA3. Scale bar represents 200μm. D) Overview of experimental setup. Perturbing the colony size while maintaining the BMP4 dose constant in the induction medium varies the pSMAD1 levels at the colony center (see Fig. S9D-E for details). E) Percentage of cells in each identified colony expressing SOX2 when colonies of varying sizes (200μm, 300μm, 400μm, 500μm, 600μm, and 700μm in diameter) were treated with 25ng/ml of BMP4 in induction medium. Each data point represents an identified colony. The total number of colonies were 932, 439, 256, 175, 122, and 45 for the respective sizes. Data pooled from two experiments. Bars represent mean ± s.d. The p-value was calculated using Kruskal-Wallis test. F) Representative immunofluorescent images of colonies stained for SOX2 and GATA3. Scale bar represents 200μm.

125

126

Figure 4-15: Pre-neurulation-like fates arise in a manner consistent with positional- information. A) Representative immunofluorescence images of 500μm diameter colonies stained for SOX2, and GATA3 after different doses (6.25ng/ml, 12.5ng/ml, 25ng/ml, and 50ng/ml) and times of BMP4 treatment. Scale bar represents 200μm. B) Mean expression levels of SOX2 and GATA 3 represented as heat maps. Darker shades represent higher expression levels and lighter shades represent lower levels of expression (for detailed data see Figure 4-16). C) (i-ii) Model of GATA3 patterning. (i) Overview of (B) where GATA3 is expressed as a function of BMP4 dose and induction time. (ii) Fate patterning of GATA3 consistent with positional information. ‘T’ indicates the presumptive threshold of fate switch to GATA3.

Figure 4-16: GATA3 expression arises as a function of BMP4 dose and induction time. Percentage of cells expressing SOX2, and GATA in 500μm colonies induced to differentiate at varying concentrations of BMP4 (3.125 ng/ml, 6.25ng/ml, 12.5 ng/ml, and 25 ng/ml) and induction times (12 hours, 24 hours, 36 hours, and 48 hours). Each data point represents an 127

identified colony, and each condition had over 100 colonies. Data pooled from two experiments. Bars represent mean ± s.d.

4.3.8 A stepwise model of RD and PI governs pre-neurulation-like fate patterning

Thus far, our data indicate that the pSMAD1 gradient was enforced outside-in within the geometrically confined hPSC colonies via a BMP4-Noggin RD network, and the pre-neurulation- like fates arose in a manner consistent with PI. In agreement with this idea, perturbing the shapes of the geometrically confined hPSC colonies did not result in fate patterning that deviated from the expected results (Figure 4-17). However, a strong test of this overall model is asking if large colonies are able to generate stereotypical RD-like periodic signaling and fate profile. We previously reported that treatment of large geometrically confined hPSC colonies (3mm) with high doses of BMP4 would give rise to spatial oscillations of BMP activity as indicated by pSMAD1 staining, and patterned gastrulation-associated fates(Tewary et al. 2017). Interestingly, when we tested the response of pSMAD1 spatial signaling dynamics in 3mm diameter colonies after BMP4 and SB treatment for 24h, we did not observe any obvious oscillations at either 50ng/ml (Figure 4-18) or 200ng/ml (Figure 4-19) BMP4 dose. Of note, the medium used for differentiating these geometrically confined hPSC colonies contained Knockout Serum Replacement (SR). An ingredient of SR called AlbumaxII is known to contain lipid associated proteins that have been shown to have an effect on hPSC biology – mechanisms of which are currently unclear(Garcia- Gonzalo & Belmonte 2008; Blauwkamp et al. 2012). We asked if using medium devoid of SR would rescue the expected spatial oscillations of pSMAD1 consistent with the RD paradigm(Tewary et al. 2017). Indeed, when we tested N2B27 medium which does not contain any AlbumaxII or SR (see Materials and Methods for composition), a 24h BMP4 and SB treatment of hPSC colonies of 3mm diameter resulted in rudimentary spatial oscillations of PSMAD1 activity at a BMP4 dose of 50ng/ml (Figure 4-20) and prominent spatial oscillations of pSMAD1 activity at a dose of 200ng/ml (Figure 4-21, Figure 4-22). When we tested the emergent fates after 48h of BMP4 and SB treatment, we did not note any oscillations of SOX2 and GATA3 in SR medium at BMP4 doses of either 50ng/ml (data not shown) or 200ng/ml (Figure 4-23); but in an N2B27 basal medium supplemented with SB, rudimentary oscillations were observed at 50ng/ml of BMP4 (Figure 4-24), and robust oscillations were noted at 200ng/ml of BMP4 (Figure 4-25,

128

Figure 4-26). These observations of spatial oscillations of the pre-neurulation-like fates is consistent with the observations of the spatial profile of pSMAD1 signaling at the respective basal media and BMP4 doses. These data provide strong evidence that fate patterning arises in a stepwise model of RD and PI, although we note that undefined components present in the induction medium contribute to deviations from the expected results.

129

Figure 4-17: Changing shapes does not affect outside-in spatial patterning. Representative images of various shapes of geometrically-confined hPSC colonies treated with BMP4 and SB in SR medium. Varying colony shapes does not result in any deviation from anticipated fate patterning. The experiment was performed once. Scale bar represents 200μm.

130

Figure 4-18: No spatial oscillations of pSMAD1 detected when large geometrically confined hPSC colonies are treated with 50ng/ml BMP4 and SB in SR medium. A-B) No discernable spatial oscillations of pSMAD1 expression detected with geometrically confined hPSC colonies of 3mm diameter were treated with 50ng/ml of BMP4 and SB for 24h 131

in SR medium. A) Stitched images of the entire colony stained for pSMAD1 shown in greyscale for ease of visibility. B) Enlarged fields that are indicated by white squares in A. White arrows indicate regions that contain cells with positive pSMAD1 expression. Scale bar represents 1mm.

132

Figure 4-19: Negligible spatial oscillations of pSMAD1 detected when large geometrically confined hPSC colonies are treated with 200ng/ml BMP4 and SB in SR medium.

133

A-B) Negligible spatial oscillations of pSMAD1 expression detected with geometrically confined hPSC colonies of 3mm diameter were treated with 200ng/ml of BMP4 and SB for 24h in SR medium. A) Stitched images of the entire colony stained for pSMAD1 shown in greyscale for ease of visibility. B) Enlarged fields that are indicated by white squares in A. White arrows at the colony periphery indicate regions that contain cells with positive pSMAD1 expression. White arrows with accompanying question marks indicate regions that possibly show expression of pSMAD1; however, the staining in these regions is inconclusive. Scale bar represents 1mm.

134

135

Figure 4-20: Marginal spatial oscillations of pSMAD1 detected when large geometrically confined hPSC colonies are treated with 50ng/ml BMP4 and SB in N2B27 medium. A-B) Marginal spatial oscillations of pSMAD1 expression detected with geometrically confined hPSC colonies of 3mm diameter were treated with 50ng/ml of BMP4 and SB for 24h in N2B27 medium. A) Stitched images of the entire colony stained for pSMAD1 shown in greyscale for ease of visibility. B) Enlarged fields that are indicated by white squares in A. White arrows indicate regions that contain cells with positive pSMAD1 expression. Scale bar represents 1mm.

136

Figure 4-21: Treatment of large geometrically confined hPSC colonies with 200ng/ml BMP4 and SB in N2B27 medium results in spatial oscillations of pSMAD1.

137

A-B) Spatial oscillations of pSMAD1 expression detected with geometrically confined hPSC colonies of 3mm diameter were treated with 200ng/ml of BMP4 and SB for 24h in N2B27 medium. A) Stitched images of the entire colony stained for pSMAD1 shown in greyscale for ease of visibility. B) Enlarged fields that are indicated by white squares in A. White arrows indicate regions that contain cells with positive pSMAD1 expression. Scale bar represents 1mm.

138

139

Figure 4-22: Additional replicates of immunofluorescent images demonstrating oscillatory pSMAD1 expression in the center of large geometrically confined hPSC colonies treated with 200ng/ml BMP4 and SB in N2B27 medium. Additional representative images of 3mm diameter geometrically confined hPSC colonies treated with 200ng/ml BMP4 and SB in N2B27 medium for 24h and stained for pSMAD1. Zoomed-in images of fields contained within white squares shown adjacent to the stitched images. White arrows indicate regions that contain cells with positive pSMAD1 expression. Scale bar represents 1mm.

140

141

Figure 4-23: No spatial oscillations of pre-neurulation-like fates detected when large geometrically confined hPSC colonies are treated with 200ng/ml BMP4 and SB in SR medium. A-B) Treatment of geometrically confined-hPSC colonies with 200ng/ml of BMP4 and SB for 48h results in RD-like periodic spatial oscillations of SOX2 and GATA3 expression. i) Representative stitched images of 3mm diameter hPSC colonies differentiated with 200ng/ml of BMP4 for 48h. Scale bar represents 1mm. ii) Zoomed section outlined by the white square in (i). The experiment was repeated two times.

142

143

Figure 4-24: Minor spatial oscillations of pre-neurulation-like fates detected when large geometrically confined hPSC colonies are treated with 50ng/ml BMP4 and SB in N2B27 medium. A-B) Treatment of geometrically confined-hPSC colonies with 200ng/ml of BMP4 and SB for 48h results in RD-like periodic spatial oscillations of SOX2 and GATA3 expression. i) Representative stitched images of 3mm diameter hPSC colonies differentiated with 200ng/ml of BMP4 for 48h. Scale bar represents 1mm. ii) Zoomed section outlined by the white square in (i). The experiment was repeated two times.

144

Figure 4-25: hPSC colonies of 3mm diameter induce RD-like periodic patterns of pre- neurulation-like fates when treated with 200ng/ml of BMP4 and SB in N2B27 medium.

145

A-B) Treatment of geometrically confined-hPSC colonies with 200ng/ml of BMP4 and SB for 48h results in RD-like periodic spatial oscillations of SOX2 and GATA3 expression. i) Representative stitched images of 3mm diameter hPSC colonies differentiated with 200ng/ml of BMP4 for 48h. Scale bar represents 1mm. ii) Zoomed section outlined by the white square in (i). White arrows indicate regions of high GATA3 and low SOX2 expression indicative of PI mediated fate patterning due to presumptive localized pSMAD1 expression. The experiment was repeated three times. Additional images shown in Figure 4-26.

146

Figure 4-26: RD-like spatial oscillations of pre-neurulation-like fates detected when large geometrically confined hPSC colonies are treated with 200ng/ml BMP4 and SB in N2B27 medium.

147

Representative immunofluorescent images of geometrically confined hPSC colonies of 3mm diameter stained for SOX2, and GATA3. The colonies were treated with 200ng/ml of BMP4 and SB for 48h. Scale bar represents 1mm.

4.3.9 PN and NN regions give rise to definitive ectodermal fates

As a final validation that BMP4 and SB treatment induced pre-neurulation-associated fates in the differentiating geometrically confined hPSC colonies, we asked if the patterned fates of the early PN and NN tissues were capable of inducing marker expression of definitive ectodermal fates like the NP, the NC, and the NNE. During embryogenesis, the NNE specifies toward the lateral end of the medial-lateral axis due to sustained levels of high BMP signaling in the ectoderm; the NC fate is specified at regions of intermediate BMP levels that activate wnt signaling; and the NP is specified at the medial end where the tissue is subject to low/no BMP signaling. To test the competence of the pre-neurulation-like patterned colonies to give rise to these fates, we treated the colonies with BMP4 and SB for 24h, then tested three different treatments. Specifically, we either treated the colonies with for a further 48h with BMP and SB and stained for keratins using a pan- keratin antibody and DLX5 (markers of NNE); or CHIR99021 (‘CHIR’ – a wnt agonist) and SB and stained for SOX10 (a marker of the NC fate); or for a period of 72h with Noggin and SB and stained for PAX6 (Figure 4-27A). Consistent with our expected results, we observed that sustained BMP4 and SB treatment resulted in robust expression of DLX5 and showed clear staining of a pan-keratin antibody, indicating acquisition of an NNE identity (Figure 4-27B). Furthermore, robust SOX10 staining was observed in colonies treated with CHIR and SB (Figure 4-27C); and the colonies treated with Noggin and SB expressed PAX6 – a bona fide marker of the NP (Figure 4-27D).

Taken together, our data are consistent with our hypothesis that a RD network in BMP signalling self-organizes the pSMAD1 gradient in geometrically confined hPSC colonies and Nodal signalling dissects peri-gastrulation-associated and pre-neurulation-associated fates that arise within these colonies in a manner consistent with PI (Figure 4-28).

148

Figure 4-27: Pre-neurulation-like platform can give rise to definitive fates associated with the differentiating ectoderm.

149

A) Overview of the experimental setup. Geometrically confined hPSC colonies were treated with BMP4 for 24h, and then treated with one of the following conditions: SB and Noggin for 72h and subsequently stained for PAX6; SB and CHIR99021 (CHIR) for 48h and stained for SOX10; SB and BMP4 for 48h and stained for DLX5 and TROMA1. B) Expression of NP marker (PAX6) in colonies differentiated with BMP4 for 24h and SB+Noggin for 72h. (i) Quantified expression observed for PAX6 observed in the treated and control conditions. The number of colonies were 76 for control, and 606 for treated. (ii) Immunofluorescent images of representative colonies stained for PAX6. C) Expression of NC marker (SOX10) in colonies differentiated with BMP4 for 24h and SB+CHIR for 48h. (i) Quantified expression observed for SOX10 observed in the treated and control conditions. The number of colonies were 286 for control, and 493 for treated. (ii) Immunofluorescent images of representative colonies stained for SOX10. D) Expression of NNE markers (DLX5 and TROMA1) in colonies differentiated with BMP4 for 24h and SB+BMP4 for 48h. (i) Quantified expression observed for DLX5 and TROMA1 observed in the treated and control conditions. The number of colonies were 163 for control, and 376 for treated. (ii) Immunofluorescent images of representative colonies stained for DLX5 and TROMA1. For B(i), C(i), and D(i), each data point represents an identified colony, and bars represent mean ± s.d. The data were pooled from two experiments, and the p-values were measured using Mann-Whitney U test.

150

Figure 4-28: Mechanism of Nodal dependent fate patterning in the geometrically confined hPSC colonies. A) An RD network in BMP signaling which comprises BMP ligands, NODAL, and BMP antagonists self-organizes the pSMAD1 gradient within the geometrically confined hPSC colonies. BMP antagonists downstream of Nodal signaling can contribute to the self- organization of the pSMAD1 gradient. B) In the presence of Nodal signaling, the fate patterning recapitulates the peri-gastrulation-like stage of human development. C) In the absence of Nodal signaling, the fate patterning recapitulates pre-neurulation-like stage of human development. In both instances, the fate patterning arises in a manner consistent with positional information. 151

4.4 Discussion

4.4.1 Nodal signaling in differentiating hPSCs

We report that differentiating hPSCs upregulate a gastrulation-associated expression profile when endogenous Nodal signaling is active and in the case where Nodal signaling is downregulated, the same differentiation pulse (either FBS, or BMP4 treatment) upregulates a neurulation-associated gene expression profile. Notably, in a recent study, Funa et al employed activation of wnt signaling in hPSCs and identified a dissection of the primitive streak and the neural crest fates. Through a chromatin-immunoprecipitation sequencing (CHIP-seq) study they identified that β-catenin is able to directly regulate both the primitive streak, and the neural crest genes. However, expression of the primitive streak genes requires β-catenin to form a physical complex with SMAD2/3 – the effectors of Nodal signaling. Furthermore, upon the formation of the complex, the expression of genes associated with the neural crest fate were inhibited. The mechanism by which SMAD2/3 can prevent β-catenin mediated activation of the neural crest genes remains unclear. We hypothesize that a similar mechanism with BMP signaling could explain much of our data. Much like β-catenin, SMAD1 may activate a peri-gastrulation-associated gene profile in the presence of SMAD2/3 whereas in the absence of SMAD2/3, SMAD1 may activate pre-neurulation-associated genes. Future studies that employ a similar approach to Funa et al by performing CHIP-seq studies to identify the binding dynamics of SMAD1 in the presence and absence of Nodal signaling can provide valuable insights toward a mechanistic understanding of how SMAD1 regulates the expression of the peri-gastrulation-associated and pre-neurulation-associated gene profiles.

4.4.2 Role of Nodal in self-organization of pSMAD1 gradient

In experimental conditions permissive of Nodal signaling, BMP4 treatment of the geometrically confined hPSC colonies results in a gradient that downregulates sharply (Figure 4-12B-C)(Tewary et al. 2017; Etoc et al. 2016; Warmflash et al. 2014). This has led some to speculate that in BMP4 treated geometrically confined hPSC colonies, BMP signaling is active exclusively at the colony periphery and inactive everywhere else – a spatial profile that can be modelled as a step-function along the colony radius(Siggia & Warmflash 2017). These authors claim that this apparent step-

152 like response in the pSMAD1 activity occurs due to not just the function of BMP inhibitors like Noggin(Tewary et al. 2017; Etoc et al. 2016), but also BMP receptors re-localizing due to increased cell-density and transitioning from being present apically to becoming localized at basolateral regions, and therefore not being accessible for BMP ligands to activate pSMAD1 everywhere in the colony except for the periphery(Etoc et al. 2016). Importantly, we report that when geometrically confined hPSC colonies of 500μm diameter are treated with a dose of 25ng/ml of BMP4 for 24h, prominent spatial oscillations of pSMAD1 activity are observed (Figure 4-11D). These results directly contradict the claim that BMP activity as measured by nuclear localized pSMAD1 can be modelled as a step-function and provide strong credibility to the hypothesis that an RD network in BMP signaling governs the pSMAD1 self-organization. Furthermore, our interpretation of the observed sharpness in drop of pSMAD1 signaling in experimental conditions permissive of Nodal signaling is different in an important aspect. We contend that the sharp nature of the spatial profile of the pSMAD1 gradient is being mis-interpreted as a purported ‘step- function’ because the contribution of Nodal signaling to the formation of the pSMAD1 gradient has thus-far been ignored in studies that employ the peri-gastrulation-like platform.

Intermediate levels of BMP signaling upregulate Nodal expression, which in turn results in the expression of BMP antagonists like CERL and GDF3 that belong to the Nodal pathway. In support of this interpretation, we demonstrate in this study that in experimental conditions where Nodal signaling, and by extension, the activity of CERL and GDF3 is inhibited by SB supplementation, the self-organized gradient of pSMAD1 becomes significantly less sharp. Indeed, after 24h of BMP4 and SB supplementation, we observe cells with detectable, but lower levels of pSMAD1 as compared to the cells at the periphery located close to the colony center (Figure 4-11A-D, Figure 4-12A-C). Given the fact that a model proposing the function of Noggin and purported density- dependent ‘re-localization’ of BMP receptors does not accommodate the Nodal signaling mediated effects on the formation of the pSMAD1 signaling gradient, these results contradict their interpretation. Importantly, data reported by the same group directly supports our interpretation of the involvement of Nodal signaling in the formation of the pSMAD1 signaling gradient. In their study, Warmflash et al reported that specific downregulation of CERL and LEFTY1 by siRNA treatment – an experimental condition that does not interfere with either Noggin or the colony density, dramatically interfered with the spatial organization of the gastrulation-associated fate patterning(Warmflash et al. 2014). Specifically, the authors reported that in colonies where CERL

153 and LEFTY1 were downregulated, the gastrulation-associated fate linked to high pSMAD1 levels (CDX2) took over the entire colony implying that pSMAD1 levels would have been sustained at high levels throughout the colony(Warmflash et al. 2014), further validating the involvement of Nodal signaling in the self-organization of the pSMAD1 signaling gradient.

The involvement of Nodal signaling in organizing the pSMAD1 signaling gradient also brings up an alternative interpretation of the observation that higher colony densities result in BMP signaling inhibition instead of the purported inaccessibility of BMP receptors(Etoc et al. 2016; Siggia & Warmflash 2017). Nodal signaling has been shown to have a community effect – where dense cell populations result in a strong upregulation of Nodal activity(Gurdon 1988; Nemashkalo et al. 2017; Peerani et al. 2007). Importantly, the community effect in Nodal signaling has specifically been proposed to be present during the onset of gastrulation in amniotes(Voiculescu et al. 2014). Consequently, another possibility for why colonies of high densities result in the attenuated BMP signaling at the center could be due to the community effect mediated upregulation of Nodal signaling and consequent expression and function of BMP antagonists belonging to the Nodal family.

4.4.3 RD network in BMP signaling

In a recent study, we proposed that the pSMAD1 signaling gradient self-organizes under the regulation of a BMP4-Noggin RD system. Our data in this study indicates the necessity of updating the proposed model to include Nodal signaling which activates downstream of BMP as it induces multiple other antagonists of BMP signaling thereby playing a role in the pSMAD1 gradient formation. In experimental conditions where Nodal signaling is inhibited due to SB supplementation, Noggin, Follistatin (Figure 4-12F), Gremlin family proteins(Etoc et al. 2016), and possibly others can play a role; and in conditions permissive of Nodal signaling, CERL and GDF3 and possibly others can further antagonize BMP signaling. Importantly, the pSMAD1 gradient formed in the presence and absence of Nodal signaling is manifestly different (Figure 4-11, Figure 4-12). Taken together, the topology of the RD network in BMP signaling needs to incorporate the role played by Nodal and potentially multiple BMP4 antagonists in addition to Noggin. A deeper and more comprehensive understanding of the RD network in the BMP pathway

154 requires further careful studies and computational platforms that enable studying multiple nodes in RD networks will be very valuable(Marcon et al. 2016).

In our study, we report different aspects that can result in variability in experimental results when studying the stereotypic RD-like periodic response in BMP signaling in the hPSC context. One such source of variability is the level of endogenous Nodal signaling between different hPSC lines (Figure 4-3C). Given the role that Nodal signaling plays in the formation of the pSMAD1 gradient (Figure 4-11, Figure 4-12), and the critical role it plays in ensuring the peri-gastrulation-associated fate patterning(Tewary et al. 2017), the variability in endogenous levels of Nodal signaling can cause inconsistent responses between different cell lines and culture conditions. Indeed, a recent study has shown drastically different responses in endogenous Nodal activation within the same hPSC line when cultured under different conditions for routine maintenance(Findlay & Postovit 2018). Secondly, we observed that when geometrically confined hPSC colonies of 3mm diameter were treated with a high dose of BMP4 in SR medium, the stereotypical RD-like periodic spatial oscillations of either pSMAD1 or the pre-neurulation-like fates were not readily observed. However, changing the medium to an N2B27 based medium was able to recapitulate the periodic response in both morphogen signaling and fate patterning. A key component of SR medium is Knockout Serum Replacement (KSR) which is known to contain lipid associated proteins like lysophosphatidic acid (LPA), and although the mechanism of action remains unclear, molecules like LPA have been shown to have an inhibitory effect on hPSC differentiation(Garcia-Gonzalo & Belmonte 2008; Blauwkamp et al. 2012). Given the above caveats associated with in vitro experiments, studies aimed at studying the details of the RD network in BMP signaling in the hPSC context – especially those directed toward investigating the specifics of the spatial oscillations of morphogen activity and fate patterning, would benefit from removing the salient sources of variability between hPSC lines. Employing basal medium like N2B27 which is devoid of components like AlbmaxII and LPA as the basal medium, avoiding undefined media like those conditioned on MEFs, and removing Nodal signaling from their system by supplementation of SB would provide experimental conditions better suited for these studies.

155

4.4.4 Variability in hPSC differentiation and assay to fingerprint lineage bias

Since hPSCs represent an endless source from which somatic cells may be derived, they are highly valuable for cell therapy and personalized medicine. To capitalize on this promise, many groups and initiatives have banked large numbers of human induced (hi)PSCs for future use in regenerative medicine applications(Ortmann & Vallier 2017). Importantly, it is widely recognized that different hiPSC lines – even if derived from identical genetic and tissue backgrounds – significantly vary in their ability to induce certain cell fates(Nazareth et al. 2013; Ortmann & Vallier 2017; Keller et al. 2018). Our study specifically demonstrated the marked variation in a relatively small panel of hPSC lines in the ability to induce fates and a gene profile associated with gastrulation. Taken together, these facts highlight a need of an assay to quantify lineage bias of a pool of hPSCs from which an ideal line would be chosen to produce cells of the requisite fate. Although some assays currently attempt to provide a solution to this need(Lensch et al. 2007; Müller et al. 2011; Tsankov et al. 2015), they are either qualitative, or prohibitively expensive. We demonstrate that Wnt3a and ActivinA treatment of geometrically confined hPSC colonies results in variable endoderm induction efficiencies that parallel their predicted propensity both from directed differentiation toward definitive endoderm, and as indicated by the MIXL1 and EOMES temporal dynamics in the EB assay (much like with mEpiSCs as shown by Kojima et al(Kojima et al. 2014)). Consequently, we both corroborate the approach taken by Kojima et al in the human system and provide proof-of-concept data that indicates that morphogen treatment of geometrically confined hPSC colonies in defined conditions might represent a rapid, quantitative, and an inexpensive solution to the need of finger-printing hPSCs (and hiPSCs).

4.4.5 Conclusions

In conclusion, we report the production of a high-throughput microtiter plate that enables robust geometric confinement of a variety of cell types. We employ this platform to screen hPSC lines for their ability to induce gastrulation-associated fate patterning and observe a Nodal signaling- dependent response in the efficiency of gastrulation-associated fate induction. We also report a proof-of-principle study that suggests the utility of geometrically confined hPSC colonies differentiated in defined conditions as an assay to fingerprint lineage bias of hPSCs. Further, we identify that differentiating hPSCs upregulate a neurulation associated gene profile in the absence

156 of Nodal signaling and exploit this knowledge to identify experimental conditions that induce pre- neurulation-like fate patterning in geometrically confined hPSC colonies. Finally, we report that much like in peri-gastrulation-like fate patterning of geometrically confined hPSC colonies, a stepwise model of reaction-diffusion and positional-information underpins the observed pre- neurulation-like fate patterning in the absence of Nodal signaling, hinting at possible conservation of mechanisms underlying the self-organized fate specification in differentiating epiblast and ectoderm in the human system.

4.5 Materials and Methods 4.5.1 Human Pluripotent Stem Cell Culture

CA1 human embryonic stem cell line was provided by Dr. Andras Nagy (Samuel Lunenfeld Research Institute). H9-1 was provided by Dr. Sean Palecek (University of Wisconsin – Madison). H9-2, HES2, and MEL1 (PDX1-GFP) were provided by Dr. Gordon Keller (McEwen Centre for Regenerative Medicine/University Health Network). HES3-1, and HES3-2 were provided by Dr. Andrew Elefanty (Monash University). H1, H7, H9-3 were acquired from WiCell Research Institute. For routine maintenance, CA1, H9-1, and H9-2 were cultured on Geltrex (Life Technologies, diluted 1:50) coated 6-well tissue culture plates using mTeSR1 medium (StemCell Technologies) as per manufacturer’s instructions. The cells were passaged at a ratio of 1:12 using ReleSR (StemCell Technologies) per manufacturer’s instructions. For the first 24h after passage, the cells were cultured in ROCK inhibitor Y-27632 to increase cell viability. The medium was changed every day and passaged every 4-5 days or when the cells reached 75-80% confluence. For routine maintenance, H1, H7, H9-3, HES3-1, HES3-2, MEL1, HES2 were cultured on feeder layers of irradiated MEFs in Dulbecco’s Modified Eagle’s Medium (DMEM) (Invitrogen), 1% Penicillin/Streptomycin, 1% non- essential amino acids, 0.1mM β-mercaptoethanol, 1% Glutamax, 2% B27 minus retinoic acid, 20% KnockOut serum replacement (referred to as ‘SR’ medium) and supplemented with 20 ng ml−1 FGF-2 (PeproTech). H1, H7, and H9-3 cells were passaged 1:6 every 4–5 days and were disassociated into small clumps using 0.1% collagenase IV (Invitrogen). HES3-1, HES3-2 were passaged 1:24 every 4-5 days and dissociated using TryplE Express (Invitrogen). All cell lines were confirmed negative for mycoplasma contamination.

157

4.5.2 Preparation of PEG plates

Platform set up, and XPS studies were performed using 22mmx22mm borosilicate coverslips (Fisher Scientific), and the 96-well plate platform was developed using custom sized (110mmx74mm) Nexterion-D Borosilicate thin glass coverslips (SCHOTT). The glass coverslips were activated in a plasma cleaner (Herrick Plasma) for 3 minutes at 700 mTorr and incubated with 1 ml of Poly-L-Lysine-grafted-Polyethylene Glycol (PLL-g-PEG(5KD), SUSOS,) at a concentration of 1 mg/ml at 37°C overnight. The glass slides were then rinsed with ddH2O and dried. The desired patterns were transferred to the surface of the PEG-coated side of the coverslip by photo-oxidizing select regions of the substrate using Deep UV exposure for 10 minutes through a Quartz photomask in a UV-Ozone cleaner (Jelight). Bottomless 96-well plates were plasma treated for 3 minutes at 700 mTorr and the patterned slides were glued to the bottomless plates to produce micro-titer plates with patterned cell culture surfaces. Adhesives validated for biocompatibility standards ISO10993, and USP Class VI were utilized for the assembly of the plates. Prior to seeding cells onto the plates, the wells were activated with N-(3- Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (Sigma) and N-Hydroxysuccinimide (Sigma) for 20 minutes. The plates were thoroughly washed three times with ddH2O, and incubated with Geltrex (diluted 1:150) for 4h at room temperature on an orbital shaker. After incubation, the plate was washed with Phosphate Buffered Saline (PBS) at least three times to get rid of any passively adsorbed extracellular matrix (ECM) and seeded with cells to develop micro- patterned hPSC colonies.

4.5.3 Comparison between PEG plates with μCP plates

PEG plates (as described above) and μCP plates (as reported previously(Nazareth et al. 2013)) were generated with patterned islands of 200μm in diameter with 500μm separation between adjacent colonies. A single cell suspension of CA1s was generated by incubating in 1ml of TryplE (Invitrogen) per well for 3 minutes at 37°C. The TryplE was blocked using in equal volume SR medium (see ‘Human pluripotent stem cell culture’ section above for composition) and the cells were dissociated by pipetting to generate a single cell suspension. The cells were centrifuged into a pellet and the supernatant aspirated to remove any residual TryplE. A single cell suspension was then generated in SR medium supplemented with 10μl of ROCKi and 20ng/ml of bFGF at a cell density of 500,000 cells/ml and 100μl of the suspension was plated onto the PEG and μCP plates for a period of 2-3h till robust cell attachment was observed. The cells were then left to make a 158 confluent colony overnight (~12h). Once confluent colonies were observed, the differentiation was performed using 100μl per well of the following inductive conditions in Apel (Stem Cell Technologies) basal media for 48h – bFGF (40ng/ml) +SB431542 (10μM) to induce differentiation into ectodermal fates, BMP4 (10ng/ml) +ActivinA (100ng/ml) to induce mesendodermal differentiation, BMP4 (40ng/ml) to induce extra-embryonic/’other’ fates, and as controls, Nutristem, and basal Apel media were used. After 48h, the colonies were fixed, and stained for OCT4 and SOX2. The relative percentages of the colonies that were positive for the two markers were used to identify the early fates induced within the colonies. A detailed description of the assay has been previously reported(Nazareth et al. 2013).

4.5.4 Peri-gastrulation-like and pre-neurulation-like fate patterning induction

Except for the experiments where we demonstrated the spatial oscillations of pSMAD1 and the pre-neurulation-like fates in 3mm diameter colonies, all fate patterning studies were performed in SR medium (see ‘Human pluripotent stem cell culture’ section above for composition) supplemented with 100ng/ml of bFGF. The studies where we demonstrate the spatial oscillations of the morpohogen activity and fate patterning were performed in N2B27 medium. N2B27 medium was composed of 93% Dulbecco’s Modified Eagle’s Medium (DMEM) (Invitrogen), 1% Penicillin/Streptomycin, 1% non- essential amino acids, 0.1mM β-mercaptoethanol, 1% Glutamax, 1% N2 Supplement, 2% B27 Supplement minus retinoic acid.

The hPSC lines that were cultured in feeder-dependent techniques for routine maintenance were first feeder depleted by passaging the cells at 1:3 on geltrex and cultured on Nutristem. To seed cells onto ECM-immobilized PEG-UV 96-well plates, a single cell suspension of the hPSC lines was generated as described above. The cells were centrifuged and re-suspended at a concentration of 1 x 106 cells/ml in SR medium supplemented with 20ng/ml bFGF (R&D) and 10µM ROCK inhibitor Y-27632. Wells were seeded in the PEG-patterned 96 well plates at a density of 60,000 cells/well for plates with colonies of 500μm diameter, 80,000 cells/well for colonies of 1mm diameter, and at 120,000 cells/well for plates with colonies of 3mm diameter and incubated for 2- 3h at 37°C. After 2-3h, the medium was changed to SR without ROCKi. When confluent colonies were observed (12-18h after seeding), the peri-gastrulation-like induction or pre-neurulation-like induction was initiated as follows. A) Peri-gastrulation-like induction (Figure 4-1) was performed in SR medium supplemented with 100ng/ml of bFGF (R&D) and 50ng/ml of BMP4. B) Unless

159 otherwise stated, pre-neurulation-like induction with 500μm colonies was performed with SR medium (see ‘Human pluripotent stem cell culture’ section above for composition) supplemented with 100ng/ml of bFGF with 25ng/ml of BMP4, and 10μM SB431542 (‘SB’). C) Endoderm fingerprinting assay (Fig. S6) was performed with N2B27 medium supplemented with 25ng/ml of Wnt3A and 50ng/ml of ActivinA. D) RD-like periodic pattern induction of pSMAD1 activity, and pre-neurulation-like fates was tested in both SR, and N2B27 mediums. In the case of SR, the medium was supplemented with 10μM SB, 100ng/ml of bFGF, and either 50ng/ml or 200ng/ml of BMP4. In the case of N2B27, the medium was supplemented with 10μM SB, 10ng/ml of bFGF, and either 50ng/ml of 200ng/ml of BMP4.

4.5.5 Embryoid body differentiation assay

The differentiation media for the EB assay contained 76% DMEM, 20% Fetal Bovine Serum (FBS), 1% Penicillin/Streptomycin, 1% non- essential amino acids, 0.1mM β-mercaptoethanol, 1% Glutamax, (all Invitrogen). A large volume of the medium was prepared with a single batch of FBS and frozen at -80C and was used to differentiate the EBs made from all the hPSC lines tested. EB formation from the hPSC lines was achieved by generating a single cell suspension (as described in the section above) directly in the differentiation media supplemented with ROCKi for the first day. The cell suspension was then plated on 24-well microwell plates (Aggrewell - 400μm, Stem Cell Technologies). The seeding density was chosen to allow generation of size-controlled EBs (~500cells/EB) for all hPSC lines. The media was carefully replaced with differentiation media without ROCKi 24hours after seeding to ensure that the EBs were not disturbed. EBs were harvested from the Aggrewell plates each day by adding 1ml of DMEM into the wells and pipetting till the EBs lifted off from the microwells, and frozen as a pellet at -80C till gene expression was assessed using qPCR.

4.5.6 CA1 Nog-/- cell line generation

CA1 Noggin knock out lines were generated using a CRISPR/Cas9 mediated donor-free dual knock-out using a previously described strategy(Liu et al. 2016). The sgRNA design was performed with CRISPRko Azimuth 2.0 (Broad Institute) using human Noggin (NCBI ID9241) as entry data and SpCas9 for the nuclease. The software ranks sgRNAs with high on-target activities and low off-target activities in a combined rank(Doench et al. 2016). We chose sgRNA1 (5’- CTGTACGCGTGGAACGACCT-3’) and sgRNA2 (5’–CAAAGGGCTAGAGTTCTCCG-3’)

160 with a combined rank of 4 and 1, respectively. sgRNA1 & 2 can be used individually or applied together to produce Noggin knock-out. Latter leads to a deletion of a DNA fragment of 112 bp and to a predetermined stop codon (Error! Reference source not found.A).

Transfection and evaluation of cutting efficiency: We first evaluated the cutting efficiency of SpCas9 for each individual gRNA on a population level. For this, we seeded CAI hESC into 24- well plates such that they are 50-60 % confluent at the day of transfection (approx. 24h after seeding). CmgRNA were generated by mixing 1μM AltR CRISPR crRNA (IDT, custom oligo entry) with 1μM AltR CRISPR tacrRNA (IDT, Cat. 1073189), annealed at 95°C for 5 min and cooled down at room temperature. GeneArtTMPlatinumTM Cas9 Nuclease (Invitrogen, B25641) was diluted to 1μM using Opti-MEM (Thermo Fisher Scientific, 31985062). Cas9 and cmgRNA were mixed at a concentration of 0.3μM each in 25μl of OptiMEM. After incubation at room temperature for 5 minutes, 1μl of EditProTM Stem (MTI Globalstem) diluted in 25μl of Opti- MEM was added to the Cas9/cmgRNA complex and incubated for 15 minutes at room temperature. Before adding the reagent –Cas9/cmgRNA mix to the cells, medium was replaced with 500μl / 24 well of fresh mTeSR. Medium was replaced 24h after transfection. 48h after transfection, cells were harvested by incubation in Gentle Cell Dissociation Reagent (STEMCELL Technologies, 07174) for 7 min. Dissociation reagent was removed and cells were resuspended in cultivation medium, pipetted to single cells and spin down for 5min at 200g. Cells were resuspended in 25μl of Cell Lysis Buffer mixed with 1μl Protein Degrader, both from the GeneArtTM Genomic Cleavage Detection Kit (Invitrogen, A24372). Cells were lysed at 68°C for 15min, 95°C for 10min and kept on ice. PCR was performed using Phusion High Fidelity DNA Polymerase (NEB, M0530) according to manufactures protocol using 2μl of the cell lysate. Primer for the PCR were the following: (fwd) 5’CTACGACCCAGGCTTCATGGC’3, (rev) 5’GACGGCTTGCACACCATGC3’. PCR product of un-transfected and transfected samples were analyzed on 2.5% MetaPhore Agarose Gel (Lonza) PCR products were analyzed using GeneArt Genomic Cleavage Detection Kit (Invitrogen, A24372) according to manufacturer’s protocol. The cleaved and un-cleaved samples were loaded on 2.5% MetaPhore Agarose Gel (Lonza) and the bands were analyzed using ImageJ. Percentage of gene modification was calculated as described in a previous report (Error! Reference source not found.B)(Guschin et al. 2010). Additionally, PCR p roducts were send for Sanger Sequencing. Chromatograms were analyzed using TIDE(Brinkman et al. 2014).

161

Cell line generation: The cell line was generated using gRNA1 & 2 mixed with Cas9 at 0.3μM each. The transfection was proceeded with exact same protocol as described above using 6 x 24 wells. After three days, cells reached confluency and were seeded to 6 well plates at sufficiently low densities to achieve clonal growth from single cells. Approximately 7 days after seeding, single clones were picked and transferred to 96 well plates. 24 clones were expanded for 2 passages and PCR was performed on cell lysates as described above (Error! Reference source not found.C). P CR products were send for Sanger Sequencing and aligned to (NCBI ID9241) and to untransfected wildtype sequence (Error! Reference source not found.C). Clones with clear loss of function mutations in both alleles (C1 and C7) were further characterized for their pluripotency marker expression (Error! Reference source not found.D).

4.5.7 Quantitative PCR analysis

RNA extraction for all gene expression analysis studies was performed using Qiagen RNAeasy miniprep columns according to the manufacturer’s protocol, and the cDNA was generated using Superscript III reverse transcriptase (Invitrogen) as per the manufacturer’s instructions. The generated cDNA was mixed with primers for the genes of interest and SYBR green mix (Roche, Sigma) and the samples were run on an Applied Biosystems QuantStudio 6 flex real-time PCR machine. The relative expression of genes of interest was determined by the delta–delta cycle threshold (∆∆Ct) method with the expression of GAPDH as an internal reference. Primer sequences used are provided in Supplementary Information (Table 4-2).

Table 4-2: Primers employed in this study

Gene Name Forward Reverse

GAPDH GTTTACATGTTCCAATATGATTCCAC TGGAAGATGGTGATGGGATT

POU5F1 AGCGATCAAGCAGCGACTAT AGAGTGGTGACGGAGACAGG

NANOG ACCTTCCAATGTGGAGCAAC GAGAATTTGGCTGGAACTGC

MIXL1 CAGAACAGGCGTGCCAAGTC TTCCAGGAGCACAGTGGTTGA

162

T CCTTGCTCACACCTGCAGTAG GGCCAACTGCATCATCTCCA

EOMES ACCCCCTTCCATCAAATCTC CCATGCCTTTTGAGGTGTCT

KDR GCATGGAAGAGGATTCTGGA CTGATTCCTGCTGTGTTGTCA

MYB CATTTGATCCGCATCCCCTG TCAAAAGTTCAGTGCTGGCC

HAND1 GCCTAGCCACCAGCTACATC ATCCGCCTTCTTGAGTTCAG

MESP1 CCCAAGTGACAAGGGACAAC TCTTCCAGGAAAGGCAGTCT

PAX6 GTGTCCAACGGATGTGTGAG AGACCCCCTCGGACAGTAAT

OTX2 GCCAATCCTTGGTTGAATCTTAGG CAATCAGTCACACAATTCACACAGC

SOX1 CAGGCCATGGATGAAGGA CTTAATTGCTGGGGAATTG

TAL1 ACTTGCCTTCCTAAGCCTGT CATTCACTCGCCAGCATGAA

RUNX1 CAATTTGCCTCTGTGTGCCT ATAGGGTAGGGTCTCAGCCT

SOX17 TTCGTGTGCAAGCCTGAGATG GTCGGACACCACCGAGGAA

FOXA2 ACCACTACGCCTTCAACCAC GGGGTAGTGCATCACCTGTT

SOX9 AGGAAGCTCGCGGACCAGTAC GGTGGTCCTTCTTGTGCTGCAC

TFAP2A AGGTCAATCTCCCTACACGAG GGAGTAAGGATCTTGCGACTGG

DLX5 TTCCAAGCTCCGTTCCAGAC GAATCGGTAGCTGAAGACTCG

GATA4 TCCAAACCAGAAAACGGAAG AAGACCAGGCTGTTCCAAGA

GATA6 TCCACTCGTGTCTGCTTTTG TCCTAGTCCTGGCTTCTGGA

GATA3 TCCTGTGCGAACTGTCAGAC TCGGTTTCTGGTCTGGATGC

163

SIX1 CCTCACAACCACCCCAAACT AGTGGAAATTTTCGGCGCAC

SIX3 CTCCTCCTCCATCCCCAGAA GTGGTAGATGGTGGTTGGGG

EVX1 GACCAGATGCGTCGTTACCG GTGGTTTCCGGCAGGTTTAG

NODAL TGAGCCAACAAGAGGATCTG TGGAAAATCTCAATGGCAAG

GDF3 GTACTTCGCTTTCTCCCAGAC GCCAATGTCAACTGTTCCCTT

BMP4 ATGATTCCTGGTAACCGAATGC CCCCGTCTCAGGTATCAAACT

NOGGIN GAAGCTGCGGAGGAAGTTAC TACAGCACGGGGCAGAAT

CERL CTTCTCAGGGGGTCATCTTG TCCCAAAGCAAAGGTTGTTC

LEFTY-A CTGGACCTCAGGGACTATGG CACACACTCGTAAGCCAGGA

4.5.8 Immunofluorescent staining, and image analysis

After the peri-gastrulation-like or the pre-neurulation-like induction was completed, the plates were fixed with 3.7% paraformaldehyde for 20 min, rinsed three times with PBS and then permeabilized with 100% methanol for 3 min. After permeabilization, the patterned colonies were blocked using 10% fetal bovine serum (Invitrogen) in PBS overnight at 4°C. Primary antibodies were incubated at 4°C overnight (antibody sources and concentrations are shown in Table 4-3). The following day, the primary antibodies were removed, and the plates were washed three times with PBS followed by incubation with the secondary antibodies and DAPI nuclear antibody at room temperature for 1 h. Single-cell data were acquired by scanning the plates using the Cellomics Arrayscan VTI platform using the ‘TargetActivation.V4’ bioassay algorithm. This algorithm utilizes the expression intensity in the DAPI channel to identify individual nuclei in all fields imaged and acquires the associated intensity of proteins of interest localized within the identified region. As previously described(Tewary et al. 2017), single-cell data extracted from fluorescent images were exported into our custom built software, ContextExplorer (Ostblom et al,

164 unpublished), which classifies cells into colonies via the DBSCAN algorithm. Cartesian coordinates relative to the colony centroid are computed for every cell within a colony. Hexagonal binning is used to group cells from multiple colonies according to their relative location within a colony. Average protein expression of cells within a bin is represented by the color map, which is normalized to the lowest and highest expressing hexagonal bins. In the line plots of spatial expression trends, cells are grouped in annular bins according to the Euclidean distance between a cell and the colony centroid. For each colony, the mean expression of all cells within an annular bin is computed. The average of all the colony means is displayed in the line plot together with the standard deviation and the 95% confidence interval (CI).

Table 4-3: Antibodies employed in this study

Protein Company (cat#) Concentration

OCT3/4 BD Biosciences (561556) 1/500

NANOG Cell Signaling (4903S) 1/200

Brachyury R&D (AF2085) 1/500

SOX2 R&D (AF2018, MAB2018) 1/500 beta-Actin Cell Signaling (3700S) 1/200

VECAD Cedarlane (160840) 1/500

SOX17 R&D (AF1924) 1/500

FOXA2 Abnova (H00003170-M12) 1/250 pSMAD1 Cell Signaling (9516S) 1/100

GATA3 Abcam (AB199428) 1/250

165

EOMES Abcam (AB23345) 1/500

TFAP2A Abcam (AB52222) 1/200

SIX1 Abcam (AB211359) 1/200

OTX2 EMD Millipore (ab9566) 1/500

GATA6 R&D (AF1700) 1/500

PAX3 R&D (MAB2457) 1/500

PAX6 R&D(AF8150) 1/250

DLX5 Abcam (AB64827) 1/500

Pan Keratin Abcam (AB8068) 1/100

SOX10 R&D (MAB2864) 1/200

166 Chapter 5 Conclusions and Future Directions

Conclusions and future directions 5.1 Summary of results

Over the course of this project, we developed a robust platform that enables geometric confinement of a variety of different cell types. We employed this platform to screen media and identify defined conditions to induce geometrically-confined hPSC colonies – an in vitro surrogate of the human epiblast – to undergo self-organized differentiation wherein the emergent fates pattern in a manner reminiscent of a gastrulating epiblast. We queried the underlying principles that regulated this self- organized fate patterning event and observed experimental results consistent with classical theoretical models of developmental fate patterning – RD, and PI. In our system, global treatment of the geometrically confined hPSC colonies with a morphogen crucially involved in human gastrulation – BMP4, resulted in upregulation of both itself and its inhibitor (Noggin). We demonstrated that Noggin plays an important role in the self-organization of the signaling gradient of BMP activity. Furthermore, we develop a computational model of an RD mediated self- organization of free BMP4 ligands within the geometrically-confined hPSC colonies and demonstrated consistency in the response to perturbations to key parameters between the model predictions and experimental data as measured by spatial localization of pSMAD1. We next demonstrated that the pSMAD1 signaling gradient patterned fates in a manner consistent with PI, where the patterned fates manifest as a function of the morphogen dose and induction time. Together, we proposed a stepwise model of RD and PI underpinning the self-organized fate patterning within the differentiating hPSC colonies. In further justification of this model, we identified conditions that rescue fate patterning within colonies that had previously thought to be incapable of facilitating fate patterning, and we also validate the model with demonstration of a stereotypic RD-like periodic response in BMP signaling activity (pSMAD1) and patterned fates (BRA) in large hPSC colonies.

We next asked whether geometrically-confined colonies generated from different hPSC lines exhibited a consistent peri-gastrulation-like response when induced to differentiate in the presence of BMP4 and identified dramatically different responses in BRA expression between the test hPSC lines. Whereas some of the lines induced BRA robustly, others did so at intermediate levels, and in a subset of the lines BRA expression was undetected. This observation led us to hypothesize that signaling pathways critically involved in the specification of the primitive-streak-like fate

168 were likely differentially regulated within the different lines. To test this hypothesis, we employed an approach that has previously been demonstrated to provide valuable insight into the transcriptional profiles of mEpiSCs. Specifically, we generated EBs from various hPSC lines and cultured them in conditions unsupportive of pluripotency, thereby allowing them to differentiate in a manner regulated by their endogenous signaling profiles. Under these experimental conditions, we observed that the subset of lines that robustly upregulated BRA in BMP4 treated geometrically- confined colonies, upregulated a gene expression profile consistent with gastrulation. On the other hand, the subset of hPSC lines that induced the expression of BRA poorly in the BMP4 treated geometrically-confined colonies upregulated a gene expression profile associated with neurulation. Consistent with previous studies, this switch correlated with Nodal expression where the upregulation of gastrulation-associated gene expression was associated with upregulation of Nodal, and a neurulation-associated gene expression profiled was associated with the absence of Nodal signaling. These observations led us to hypothesize that treating geometrically-confined hPSC colonies with BMP4 in the presence of a small molecule inhibitor of Nodal signaling would induce fate patterning that paralleled the patterning observed in the differentiating ectoderm. We observed experimental results consistent with this hypothesis where the geometrically-confined hPSC colonies patterned pre-neural and non-neural fates associated with the initial ectoderm- specific patterning events. Furthermore, much like the mechanism that induced the geometrically- confined hPSC colonies to undergo gastrulation-associated fate patterning, we observed a stepwise activation of RD and PI governing the patterning of pre-neurulation-like fates.

Taken together, we employed in vitro experiments in conjunction with bioengineering strategies ranging from micro-fabrication to computational modelling to identify conditions that induce fate patterning associated with a post-implantation developmental stage of human embryogenesis, and delineate the underlying principles regulating this observation. Given that hPSCs represent the in vitro surrogate of human epiblast cells, and that the emergent fates in our platform are consistent with those that arise during post-implantation human development, the mechanisms we report may parallel those that regulate self-organized fate patterning during post-implantation human development.

5.2 Limitations

169

Although we developed a platform that induces gradients of activity of key morphogens in an approximated model of the developing human epiblast (geometrically confined hPSC colonies) and results in developmentally-relevant fate patterning, there are important caveats to consider. Firstly, although the human epiblast can be approximated as a 2D monolayer of pluripotent stem cells, they are housed in a complex 3D environment that have interactions with the underlying PE, the surrounding amnion, and trophoblast. Moreover, the epiblast develops in the blastocyst which are known to be far more compliant (soft) than glass coverslips which were employed in this study. In addition, the extraembryonic tissues surrounding the developing epiblast play critical roles in enforcing the signaling gradients that induce the morphogenetic re-organization of the epiblast. These additional layers of complexity that regulate the epiblast microenvironment in vivo have not been accounted for in our in vitro model. The absence of these levels of control is likely responsible for the second caveat – the patterning disparity between in vivo gastrulation and our platform. Specifically, during development, gastrulation segregates the three germ layers in a manner that results in the formation of a tri-laminar tissue structure. In contrast to this layered organization, the in vitro platform we report results in a concentric segregation of markers expressing the germ layer fates along the same 2D plane. Therefore, as we note in the discussion section in Chapter 3, the overall model we identify using our platform is likely one of multiple layers of complexity that contribute to the initiation and guided progression of post-implantation human embryogenesis. Consequently, the interpretations from our platform should be mapped to embryogenesis in a cautious manner as the reality in development is undoubtedly far more complex. Testing the predictions derived from this platform in model organisms could provide far more reliable results. Alternatively, developing synthetic networks to test these models also represent an attractive approach for validating the predictions derived from these models (discussed in detail in Future Directions subsection below).

In addition to the caveats noted above that deal with the fate patterning platform, it is important to note that the computational model we employed in this study heavily relied on training the output to the response of pSMAD1 self-organization in 1000μm diameter colonies induced to differentiate with 50ng/ml of BMP4 (Appendix – Supplemental Information – Model description). Two different aspects of the mathematical model – the production functions of BMP4 and Noggin, and the diffusivities of BMP4 and Noggin, were predicted from fitting the experimental data and have not been directly measured. In the case of the diffusivities of BMP4 and Noggin, although these

170 predicted values lie in realistic ranges as determined in previous studies, direct measurement in our system was not performed. Measuring these quantities using techniques like fluorescence recovery after photobleaching would provide more credibility to the model. The production functions of BMP4 and Noggin proteins – ‘F’ and ‘G’ in computational model (Appendix – Supplemental Information – Model description), were approximated to be linear. Therefore, this model only represents the RD response in the region where the production terms behave like linear functions, i.e. close to steady state. Mapping the production of BMP4 and Noggin, in response to BMP4 and Noggin molecules present in the extracellular environment to production functions in the RD model is currently extremely difficult owing to the absence of detailed information regarding the signal transduction pathway, and the transcriptional dynamics of BMP signaling effectors. As detailed knowledge of these complex regulations become available, the production functions can be comprehensively defined. Notably, the fact that the computational model is valid only for solutions close to steady state does not invalidate our use of the model to study RD-like behaviour in our system. This is because in systems where morphogen activity self-organizes via RD, although the rudimentary oscillations that arise close to steady state (t=0) become progressively enhanced in time, the patterned distributions of the morphogen peaks (and valleys) do not change. Consequently, a model that provides a predictive response of morphogen distribution – even if close to steady-state, can be valuable in predicting the spatial patterning at times far from t=0. Given that we limited the comparison between the computational predictions and the experimental data to the radial periods of pSMAD1 activity and the emergent BRA fates – i.e. we exclusively focused on the spatial patterning of the morphogen activity, the fact that we employ linear approximations of the production terms does not adversely affect our conclusions.

5.3 Impact 5.3.1 In vitro platform of human peri-gastrulation and pre-neurulation

Fundamental studies in other model organisms have provided much insight into the mechanistic understanding of developmental fate patterning. However, given the species-specific differences, extrapolating how human embryos develop can be problematic (Rossant 2015). Therefore, studies aimed at understanding human development would be better served to employ the use of human embryos. Studies aimed at investigating pre-implantation human development can be performed on blastocysts accessed through in vitro fertilization clinics (Fogarty et al. 2017), and human embryos that are acquired from terminations can provide valuable information on development of

171 organs and mature tissues (Belle et al. 2017). However, culturing blastocysts past 14 days is currently prohibited due to ethical concerns, and samples from terminations are typically accessible after the initial stages of development have already manifested, i.e. after the embryo has progressed through the stages of gastrulation and neurulation. Consequently, access to human embryos during the developmental stages of gastrulation and neurulation is especially difficult. The platform that we report establishes the bridge between accessible stages of human embryogenesis and enables studies specific to fate patterning associated with gastrulation and the onset of neurulation in the human system.

5.3.2 Unified model of RD and PI

Of the various models of biological pattern formation that have been proposed, RD and PI have emerged as the dominantly accepted mechanisms to describe tissue organization during early development. However, due to apparent inconsistencies between the two models, RD and PI have historically been considered mutually exclusive. For instance, the RD hypothesis aimed to explain the mechanism of symmetry breaking during development(Kondo & Miura 2010; Turing 1952) whereas PI explicitly relied on the prior presence of polarities and asymmetries in the developing tissue(Briscoe & Small 2015; Wolpert 1969). Furthermore, the RD model implies a very close and bijective correspondence of the patterned fates with the morphogen distribution(Turing 1952; Kondo & Miura 2010; Murray 2008) whereas PI requires a sufficiently graded distribution to pattern multiple fates, needing a stage of interpretation of the morphogen concertation(Briscoe & Small 2015; Wolpert 1969; Wolpert 1981; Green & Sharpe 2015). Consequently, fate acquisition during development have predominantly been attributed to either RD(Sick et al. 2006; Muller et al. 2012; Economou et al. 2012; Raspopovic et al. 2014) or PI(Gregor et al. 2008; Gregor et al. 2007; Houchmandzadeh et al. 2002; Jaeger et al. 2004; Green & Smith 1990; Chen et al. 2012) mechanisms. Although biological fate patterning occurring by both RD and PI working in concert has recently been hypothesized(Green & Sharpe 2015), little evidence of this idea has been demonstrated. Our study provides evidence of RD mediated self-organization of a morphogen gradient (reported by pSMAD1 spatial patterning) which proceeds to pattern fates in a time- and dose-dependent manner consistent with the PI paradigm. It is interesting to speculate if the regulatory network topologies that underlie other established mechanisms of developmental fate patterning like the clock and wavefront model for instance, share any similarities with that of

172

RD+PI, suggesting that evolution may have converged to an overarching rule to pattern different developmental tissues.

5.4 Future Directions 5.4.1 Robust alternative for organoid screening

As discussed in Chapter 4, the micro-patterning platform that we have developed can harness the self-organizing nature of stem and progenitor cell types to give rise to spatially organized colonies with the structure and organization reminiscent of the developmental tissues. Given that adult (multipotent) stem cells can build and maintain mature organs, this platform may also be able to exploit the self-organizing nature of adult stem cells to generate in vitro surrogates of mature organs, though this remains to be demonstrated. Furthermore, the fact that this system starts from 2D colonies allows for the emergent tissue surrogates to be readily amenable to immunofluorescence-based assays.

The important role that tissue organization plays in the biological function of tissues and organs is well established (Lancaster & Knoblich 2014). Therefore high-throughput in vitro screens to study organ-specific behaviour will be benefitted by choosing in vitro models that recapitulate the appropriate tissue organization. The 3D organoid field offers a valuable opportunity to achieve this goal (Lancaster & Knoblich 2014; Kretzschmar & Clevers 2016). However, current studies that employ 3D organoids to perform high-throughput screens struggle with getting high-content output data due to technical challenges of acquiring immunofluorescent data from 3D tissue structures. The platform developed in this project represents a valuable alternative to harness the power of high-throughput screens in structures that start from 2D monolayers and recapitulate organization reminiscent of in vivo tissues, much like their 3D counterparts (organoids). We specifically demonstrate the ability to generate organized structures reminiscent of two different stages of embryogenesis, but this approach can be employed for a variety of different tissue-types and organ progenitors.

5.4.2 Identification of emergent fate subtypes

As discussed in chapter 3, controversies exist in the field of stem cell biology regarding the identity of the fates that emerge out of hPSCs. For instance, for many years now the expression of SOX17 in differentiating hPSCs has been accepted as marking the definitive endoderm compartment.

173

However recent evidence suggests that SOX17 also plays a critical role in the specification of the germ cell fate. In addition, the expression of CDX2 in hPSCs treated with BMP4 has been a source of controversy for over a decade where some groups identify this population as trophoblasts, whereas others consider it to be mesoderm. Given that this platform can enable the specification of these identities while maintaining an appropriate developmental context, live cell reporters for populations that express these markers can allow for enrichment of these populations and in-depth analyses of their gene and protein expression profiles and their developmental potency. Such studies can provide valuable contributions by helping settle long-standing debates in the field of stem cell biology.

5.4.3 Future steps towards recapitulating germ layer fate patterning

Although we started with exploring conditions to recapitulate the spatial patterning observed during gastrulation during human development (Chapter 3), the variation between hPSC lines led us to identify experimental conditions that induces pre-neurulation-like fate patterning (Chapter 4). The pre-neurulation-like fate patterning occurs during the differentiation of the developing ectoderm. Consequently, one may ask – can the fate patterning of the other two germ layers, i.e. the developing mesoderm and endoderm, also be recapitulated? In vitro models of fate patterning in the other germ layers would be a valuable step toward acquiring a holistic understanding of how human development proceeds through the segregation of the germ layers and specification of the organ progenitors. Furthermore, models such as these can especially inform the current ongoing debate on the patterning of the mesoderm. The dominant model of mesoderm patterning during development suggests that a BRA-expressing, multipotent primitive streak population arises during the onset of gastrulation and can specify into different lineages depending on the signaling levels of key morphogens like BMP, Wnt, FGF, and Nodal present at different positions in the gastrulating embryo. This model has recently been challenged, and some groups have claimed that instead of a multipotent progenitor that can specify to all mesodermal subtypes, BRA expression in the primitive streak can be subdivided into groups that have more restricted potency in their ability to induce mesodermal fates (Mendjan et al. 2014; Scialdone et al. 2016). An in vitro model of mesoderm development can prove valuable for settling this debate.

The induction of the ectodermal fates out of hPSCs can be readily achieved by BMP4 treatment in the background of Nodal inhibition (Chapter 4) in starting populations of geometrically-confined

174 hPSCs. Arguably, in these conditions the starting population is uniform, although some have noted slight variations in the intensity of proteins associated with pluripotency along the colony radius (Warmflash et al. 2014). Achieving the goal of patterned mesoderm and endoderm specific cell identities in geometrically-confined colonies pose interesting challenges – primarily that of attaining a homogenous starting population. Identifying the appropriate experimental conditions to achieve fate patterning of the mesoderm and endoderm lineages deserves careful consideration.

As development of the ectoderm progresses following the initial regionalization of the neural plate, the neural plate border, and the non-neural ectoderm, complex morphogenetic reorganization that includes folding, and elongation of the tissue result in the formation of mature ectodermal tissues like the neural tube, the neural crest, and the epidermis. The approach of having a fixed hPSC colony geometry that is likely not amenable to such morphogenetic changes like folding and elongation, may not be suited to enable a faithful recapitulation of the morphogenetic changes that occur in subsequent stages of embryogenesis. Strategies to recapitulate morphogenesis characteristic of these deeper developmental stages of the ectoderm would likely require other creative approaches and represent an exciting opportunity for bioengineers to develop solutions that build on the pre-neurulation-like platform.

5.4.4 Moving into the third dimension

As discussed in the ‘Limitations’ subsection of this chapter, the platform we report approximates the human epiblast as a 2D monolayer tissue. Although this approximation did prove to be a valuable initial step to extract important mechanistic information specific to self-organization of morphogen signaling and signaling pathways involved in instructing fate patterning during human development, next iterations should build upon this platform to develop models that more closely parallel the tissue organization observed in vivo. This can be achieved by constructing models of the human epiblast in a more realistic, 3D environment. This proposition is slightly problematic as it requires engineering solutions to construct 3D models while maintaining the 2D epithelial nature of the developing epiblast. We propose two different ways that this may be achieved.

The first approach can exploit the fact that when clumps of hPSCs are cultured in an ECM rich environment, they develop into cyst-like structures; evidence suggests that this morphogenetic reorganization shares the lumenogenesis that occurs in the developing human epiblast (Shahbazi et al. 2016). This observation also suggests the critical importance that ECM signaling can play in

175 the morphogenetic reorganization that comes into play in 3D culture of hPSCs. Given that morphogens like BMP4 are able to self-organize their signaling activity, titrating the dose of BMP4 in the induction media might identify a concentration (likely low levels of the ligands in the induction medium) whereby the BMP signaling activity might be restricted to a subset of the cells in the hPSC cyst-like structure. Such an approach could conceivably induce a similar organized differentiation as has been reported in our 2D platform, and result in the spatial segregation of the gastrulation-associated fates in a manner that more closely resembles in vivo development. Another approach that has already shown much progress in fact, is to exploit the crucial role that ECM signaling plays in morphogenesis. Recent studies have demonstrated that engineering environments with differential ECM signaling at opposing ends of an hPSC cyst resulted in morphogenetic organization with similar characteristics to amniogenesis (Shao et al. 2016; Shao et al. 2017). Guiding platforms such as these to undergo organized differentiation, maybe by employing a low-dose treatment with BMP4 could further guide the in vitro developmental process in a more realistic 3D environment.

5.4.5 Breaking symmetry in peri-gastrulation-like model

A salient difference between the reported peri-gastrulation-like platform and the gastrulation stage epiblast is that absence of the extraembryonic lineages. After implantation, a subset of the visceral endodermal tissue at the distal end, known as the distal visceral endoderm (DVE) acquires an epithelial morphology and begins to express antagonists like Lefty, Cer1, DKK1 etc. The DVE migrates to what becomes established as the anterior end of the epiblast, at which stage it is referred to as the Anterior Visceral Endoderm (AVE) ensuring the anterior end has low/no activity of TGFβ and Wnt signals. The extraembryonic ectoderm (EXE) that span the proximal region of the developing epiblast, secretes BMP signals, which become concentrated on the posterior end due to the inhibitory effect of signals emerging from the AVE (Ohinata et al. 2009; Tam & Loebel 2007). Taken together, this results in an asymmetrical distribution of the morphogen signaling activity and predictably induces the primitive streak at the posterior end. Although the necessity of these extraembryonic tissues in initiating the polarized expression of the primitive streak in in vitro platforms may not be absolutely required as evidenced by the polarized induction of primitive-streak-like cells in mouse EBs (ten Berge et al. 2008), developing a more complete model of the gastrulation stage embryo would be of interest to the community. Furthermore, a platform with a clear AP axis would enable further engineered approaches to guide deeper

176 developmental stages far more readily than colonies that have a radially symmetric signaling gradient as is seen in our system.

Inducing an asymmetric signaling gradient can be enforced by existing microfluidic techniques, whereby hPSC cells cultured in a chamber between two parallel laminar flow streams emanating from reservoirs that act as either a source of a morphogen or its sink (or a source of the antagonist of the morphogen). Such approaches have been employed with much success in modelling the dorsal-ventral (DV) axis patterning via a Sonic Hedgehog (SHH) gradient (Demers et al. 2016). Enforcing signaling gradients of key morphogens like BMP4, Nodal, Wnt, etc. using this strategy can enable development of an engineered in vitro platform with an asymmetric distribution of signaling molecules. Other approaches that can also break symmetry within these colonies can employ techniques like optogenetics to induce fates of tissues like the AVE at strategically chosen locations in a field of hPSCs to enable the self-organized formation of an asymmetrical signaling gradient. However, currently no reliable protocols exist that enable trans-differentiation of hPSCs to human AVE-like cells which presents a roadblock and would need to be addressed to explore this avenue. Other approaches can include using recently developed technologies that employ the interactions between single-stranded DNA fragments and their complementary sequences to rapidly enforce patterned cell-cell interactions (Chen et al. 2016; Todhunter et al. 2015). These authors developed techniques to allow cells to present single stranded DNA fragments on their cell walls and demonstrated that when different cell types presented complementary sequences, the DNA interaction rapidly enforced a controlled cell-cell interaction. This approach can be employed to selectively immobilize AVE cells (sourced from developing mouse embryos, for instance) to the differentiating hPSC colony to enforce an asymmetric signaling gradient of BMP and Nodal.

5.4.6 Synthetic biology techniques for validation and development-by- design

The models that are proposed regarding the underlying principles that regulate fate patterning and morphogenesis during embryonic development have been proposed through studies are that mostly analytical. That is, models proposed are based on the analysis of responses to experimental perturbations in developing embryos. Over many years and numerous studies, the developmental biology community distills the principles that are thought to underly these embryological events. It is worth noting that the complexity that can be involved in orchestrating these events can rapidly become intractable and very difficult to understand – this complexity arises at multiple levels,

177 ranging from gene regulatory networks within each progenitor cell, to the micro-environment and cell-ECM interactions, to long range communication with cells. Consequently, the models proposed tend to be simple and aim to encapsulate the core characteristics of the real model that regulates the developmental event of interest. Although the concerted effort of the field identifies principles based on multiple studies, and these principles are therefore reliable, direct verification of the sufficiency of the proposed principles in orchestrating the relevant embryological events is problematic (Davies 2017). Here, the emergent field of synthetic biology can provide valuable support. If a complex morphogenetic or a fate patterning event is thought to occur via a simple condensed model, and if the condensed model is sufficient to induce the developmental events then engineering cells in a manner such that the proposed model holds true, and asking if the developmental events are induced is a strong validation of the model (Davies 2017). Notably, the technical tools required to engineer a synthetic network that allows for an RD mediated self- organization of a morphogen and a PI mediated interpretation of the distribution are available (Lim 2010; Johnson et al. 2017), and validation of the stepwise model of RD and PI resulting in fate self-organized fate patterning is an important next step.

178

Chapter 6 Appendices

179

Appendices 6.1 Supplementary Information – Model description

6.1.1 Summary of data in main text and motivation of the model

We observed that when geometrically-confined human pluripotent stem cell (hPSC) colonies were induced to differentiate in media supplemented with BMP4, pSMAD1 activity within the colonies self-organized into a radial gradient. We hypothesized that regulators of the BMP signaling pathway were controlling this observation and investigated the expression of key genes involved in the BMP signaling. We found that upon induction with BMP4, hPSCs upregulated the expression of both BMP4, and NOGGIN. Furthermore, siRNA mediated knockdown of NOGGIN significantly interfered with the formation of the pSMAD1 gradient. These observations implicated a possible ‘Activator/Inhibitor’ model of reaction-diffusion (RD) mediated self- organization of signaling activity. The affinity of NOGGIN and BMP4 (Zimmerman et al. 1996) to be sufficiently high (Kd ~ 19 pM) to prevent the BMP ligands’ interaction with their receptor -9 -10 (Kd 10 – 10 M). Furthermore, NOGGIN is a highly diffusible molecule (Inomata et al. 2013)that is capable to inducing a long-range inhibition of BMP signaling. Therefore, they make a stereotypical activator/inhibitor pair.

To test the assertion that a BMP4-NOGGIN RD system was self-organizing the pSMAD1 gradient, we set out to develop an RD model specific to our tissue geometry, initial, and boundary conditions to query testable predictions of signaling gradient formation that may arise in our geometrically confined hPSC colonies. Notably, the purpose of this model was not to attain detailed information of the response of the signal transduction within the differentiating cells, or identify the gene regulatory network that would allow differential induction of fates in response to different signaling levels. Instead, we set out to identify the simplest possible RD model that could produce a BMP signaling gradient for a 1000µm diameter colony when differentiated with 50ng/ml of BMP4 in the induction media – reminiscent of the pSMAD1 signaling patterns observed in our experimental data under those specific conditions. We then validated this model based on the predictions to perturbations of two key experimental parameters (Figure 3-12– main text); and then employed this model to made testable predictions of the differentiation behavior of the hPSC colonies beyond the conditions under which the model was build (Figure 3-16, Figure 3-20- main text). The RD model described below is a simplified, and an idealized model. Nevertheless, this

180 model generates predictions of morphogen (BMP4) distribution in response to perturbations of experimental conditions which is constantly shown to be consistent with both pSMAD1 gradient formation and associated hPSC differentiation.

6.1.2 Two-component Reaction-Diffusion system

We set out to develop a mathematical model of the concentration profiles of BMP4 and NOGGIN molecules as a function of space and time in our micro-patterned colonies, based on a reaction- diffusion (RD) system as described by Alan Turing (Turing 1952), J D Murray (Murray 2008), and Gierer and Meinhardt (Meinhardt 2015). The partial differentiation equation (PDE) set can be described as follows:

휕푏푚푝 = 퐹(푏푚푝, 푛표푔) − 푑 푏푚푝 + 퐷 ∇2푏푚푝 푑푡 퐵푀푃 퐵푀푃

(1)

휕푛표푔 = 퐺(푏푚푝, 푛표푔) − 푑 푛표푔 + 퐷 ∇2푛표푔 푑푡 푁푂퐺 푁푂퐺

Here, bmp, and nog are functions of both space, and time. They represent the local concentrations of BMP4 and NOGGIN molecules at a particular point in our micro-patterned colonies at a given point in time. F (bmp, nog), and G (bmp, nog) represent the non-linear functions which describe the production rates of BMP4, and NOGGIN. The degradation rates of the molecules are given by dBMP, and dNOG; and DBMP, and DNOG represent the diffusivities of the molecules.

Assuming the production terms of BMP4, and NOGGIN can be approximated by linear functions (close to the steady state), as has previously been done by Turing(Turing 1952), and Kondo and Miura(Kondo & Miura 2010). The nature of the production terms is such that as the values of F(bmp,nog), and G(bmp,nog) increase, the system transitions away from steady state, increasing the error and preventing the convergence of the solutions(Murray 2008; Kondo & Miura 2010). Attempts to restrict the values for the morphogen near steady state, to enable convergence, have either used non-linear functions that saturate at increasing values (e.g. the Hill function(Sick et al. 2006)) or have enforced a range in which the linear approximation of the reaction function is

181 confined(Kondo & Miura 2010). Since we used linear production functions (2) in our model (1), we chose the latter strategy and restricted the reaction functions to a defined range(Kondo & Miura 2010). The production terms are represented by:

퐹(푏푚푝, 푛표푔) = 0 ≤ 푎퐵푀푃푏푚푝 + 푏퐵푀푃푛표푔 + 푐퐵푀푃 ≤ 1

(2)

퐺(푏푚푝, 푛표푔) = 0 ≤ 푎푁푂퐺푏푚푝 + 푏푁푂퐺푛표푔 + 푐푁푂퐺 ≤ 5

6.1.3 Changing Variables:

To circumvent the issue of intractability of the number of BMP4 and NOGGIN molecules in the circular region of interest modelled by our PDE solutions, we chose to change the variables bmp, and nog into normalized, dimensionless variables which we represent as bmp*, and nog*.

Quantities, and assumptions of note – All experiments were performed in micro-patterned 96-well plates with a volume of a 100µl of induction media per well. The culture surface area of each well is 0.3165cm2. The molecular weight of BMP4 is 34KDa. We assumed that the two-dimensional colony ‘surface’ across which the BMP4 and NOGGIN distributions are predicted in the PDE set could be approximately represented by a 1µm height from the colony surface. Therefore, the equivalent ‘surface concentration’ of one colony of 1mm diameter when 100ul of induction media containing 1ng/ml of BMP4 in SI units is 1.77x1010 molecules/m2. We opted to change the variables (bmp, nog) by normalizing the entire PDE set by 1.77x1010 molecules/m2.

The linearized PDE set from (1), and (2) together are of the following form:

휕푏푚푝 = 푎 푏푚푝 − 푏 푛표푔 + 푐 − 푑 푏푚푝 + 퐷 ∇2푏푚푝 휕푡 퐵푀푃 퐵푀푃 퐵푀푃 퐵푀푃 퐵푀푃

(3)

182

휕푛표푔 = 푎 푏푚푝 − 푏 푛표푔 + 푐 − 푑 푛표푔 + 퐷 ∇2푛표푔 휕푡 푁푂퐺 푁푂퐺 푁푂퐺 푁푂퐺 푁푂퐺

The SI units for the parameters, and variables in (3) are shown below:

푚표푙푒푐푢푙푒푠 푚표푙푒푐푢푙푒푠 푏푚푝 [ ] 푛표푔 [ ] 푚2 푚2

1 1 푎퐵푀푃 [ ] 푎 [ ] 푠 푁푂퐺 푠

1 1 푏 [ ] 푏 [ ] 퐵푀푃 푠 푁푂퐺 푠

푚표푙푒푐푢푙푒푠 푚표푙푒푐푢푙푒푠 푐 [ ] 푐 [ ] 퐵푀푃 푚2 × 푠 푁푂퐺 푚2 × 푠

1 1 푑 [ ] 푑 [ ] 퐵푀푃 푠 푁푂퐺 푠

푚2 푚2 퐷 [ ] 퐷 [ ] 퐵푀푃 푠 푁푂퐺 푠

10 2 Dividing throughout by 1.77x10 molecules/m , we changed the variables bmp, nog, cBMP, and cNOG as follows:

푚표푙푒푐푢푙푒푠 푚표푙푒푐푢푙푒푠 푚표푙푒푐푢푙푒푠 푚표푙푒푐푢푙푒푠 bmp [ ] = bmp* [ ] x1.77x1010 [ ] nog [ ] = nog* [ ] x1.77x1010 [ ] 푚2 푚2 푚2 푚2

푚표푙푒푐푢푙푒푠 ∗ 1 10 푚표푙푒푐푢푙푒푠 푚표푙푒푐푢푙푒푠 ∗ 1 10 푚표푙푒푐푢푙푒푠 푐 [ ] = 푐 [ ] x1.77x10 [ ] 푐 [ ] = 푐 [ ] x1.77x10 [ ] 퐵푀푃 푚2푋푠 퐵푀푃 푠 푚2 푁푂퐺 푚2푋푠 푁푂퐺 푠 푚2

183

Taken together, the new set of partial differential equations with the changed variables (bmp*, nog*, c*BMP, and c*NOG) is given below:

∗ 휕푏푚푝 ∗ ∗ ∗ ∗ = 푎퐵푀푃푏푚푝 + 푏퐵푀푃푛표푔 + 푐 퐵푀푃 − 푑퐵푀푃푏푚푝 휕푡 2 ∗ + 퐷퐵푀푃∇ 푏푚푝

(4)

휕푛표푔∗ = 푎 푏푚푝∗ + 푏 푛표푔∗ + 푐∗ − 푑 푛표푔∗ 휕푡 푁푂퐺 푁푂퐺 푁푂퐺 푁푂퐺 2 ∗ + 퐷퐵푀푃∇ 푛표푔

6.1.4 Dose dependence in BMP4 and NOGGIN production

We observed a BMPi dose dependent production of both BMP4 and NOGGIN (Figure 3-9). To incorporate this response into our model we chose the following expressions for aBMP, and aNOG.

푎퐵푀푃 = 훼퐵푀푃(1 + 퐵푀푃푖 ∗ 훾퐵푀푃)

(6)

푎푁푂퐺 = 훼푁푂퐺(1 + 퐵푀푃푖 ∗ 훾푁푂퐺)

6.1.5 Initial conditions of BMP4 and NOGGIN distributions in micro- patterned colonies:

6.1.5.1 BMP4

BMP4 is added in the differentiation medium and presented to the colonies at a uniform dose. Therefore, we considered the initial concentration of BMP4 to be a constant value given by BMPi.

184

6.1.5.2 NOGGIN

The initial conditions for NOGGIN are more nuanced. Although NOGGIN is produced in response to BMP signaling in the cardinal ‘Activator-Inhibitor’ paradigm, BMP4 inhibition in the initial condition (at time t=0), can be achieved by a variety of different molecules (e.g. FOLLISTATIN (FST), CHORDIN, GDF3, and CERBERUS-Like (CERL) among others) in addition to NOGGIN(Wu & Hill 2009). Since we observed elevated basal expression of BMP signaling inhibitors like FST, GDF3, and CERL relative to NOGGIN in hPSCs during basal culture conditions (Figure 3-7B-G), we opted to consider the spatial profile of a ‘generic BMP inhibitor’ as the initial condition for the RD paradigm.

To identify the specific spatial profile of a generic BMP inhibitor, we developed a simplified model of a passive diffusion-driven profile that would arise in a circular hPSC colony where each cell is a source of the secreted molecule. Over the course of the formation of a confluent hPSC colony, we assumed that the expression profile of a ‘generic BMP inhibitor’ would reach a steady state. To approximate this steady state spatial profile, we considered each cell (a point source of the inhibitor), evenly distributed within the colony (Figure 3-8A), and assumed an infinite sink at a large distance from the colony (Figure 3-8B). Simulation of a steady state response revealed a spatial profile that could broadly be approximated as an elliptical paraboloid (Figure 3-8C-D). Accordingly, we considered the initial condition of NOGGIN, which at t=0 can be replaced by the effective contribution of all BMP inhibitors being expressed by the hPSCs in the micro-patterned colony, to be an elliptical paraboloid function (Figure 3-8E). Notably the peak concentration chosen for the NOGGIN initial condition is arbitrary – and the patterning of free BMP4 distribution was found to be robust to the choice of the peak value.

6.1.6 Boundary conditions for the BMP4-NOGGIN reaction-diffusion system in micro-patterned colonies

We assumed that the cells at the radial edge of the micro-patterned colonies were always subjected to the same concentration of BMP4 that is in the bulk medium, which parallels the ‘edge-sensing’ model that has been proposed by previous studies(Etoc et al. 2016), and that there was no flux at the boundary for NOGGIN.

185

6.1.7 Final Reaction-Diffusion PDE

The final two-component PDE can be written as shown below:

휕푏푚푝∗ = 푎 푏푚푝∗ + 푏 푛표푔∗ + 푐∗ − 푑 푏푚푝∗ 휕푡 퐵푀푃 퐵푀푃 퐵푀푃 퐵푀푃 2 ∗ + 퐷퐵푀푃∇ 푏푚푝

∗ 휕푛표푔 ∗ ∗ ∗ ∗ = 푎푁푂퐺푏푚푝 + 푏푁푂퐺푛표푔 + 푐 푁푂퐺 − 푑푁푂퐺푛표푔 휕푡 2 ∗ + 퐷퐵푀푃∇ 푛표푔

Initial Conditions Boundary Conditions

bmp*(t=0) = BMPi bmp*(boundary) = BMPi

푥 2 푦 2 푑(푛표푔 ∗ (푏표푢푛푑푎푟푦)) nog*(t=0) = {0 ≤ − ( ) − ( ) + 1} = 0 푅 푅 푑푡

In the initial condition for Noggin, R represents the colony radius.

6.1.8 Parameter choices and parameter sensitivity

Importantly, the coefficients kaBMP, bBMP, c*BMP, kaNOG, bNOG, and c*NOG in the production terms for BMP4, and NOGGIN do not correspond to experimentally determined parameter values. The values for these parameters were chosen as per Kondo et al(Kondo & Miura 2010) – we first chose a parameter set that resulted in oscillating values of BMP4, and NOGGIN, and then chose the values of the diffusion coefficients such that DNOG > DBMP4 (Table 6-1).

186

The diffusivity values for both NOGGIN, and BMP4 are in realistic ranges (Raspopovic et al. 2014; Sick et al. 2006; Inomata et al. 2013). For instance, Inomata et al. calculated the diffusivity of NOGGIN in Xenopus embryos to be 37±6.6 μm2/s, which is remarkably close to the predicted diffusivity of Noggin in our system. However, the exact values of the diffusivities in our system have not been measured.

The value for bNOG was chosen to be zero since NOGGIN does not have any receptors and is therefore, unable to repress its own expression.

Table 6-1: Model parameters

훼BMP 0.01 [1/s] Kondo et al.

γBMP

bBMP 0.01[1/s] Kondo et al.

c*BMP 0.003 [1/s] Kondo et al.

dBMP 0.003 [1/s] Kondo et al.

DBMP 5 [µm2/s] -

훼NOG 0.008 [1/s] Kondo et al.

γNOG

bNOG 0 -

c*NOG -0.015 [1/s] Kondo et al.

dNOG 0.009 [1s] Kondo et al.

DNOG 25 [µm2/s] -

187

In Figure 6-1, we show the response of the model to predicting the spatial profile of free BMP4 molecules within the differentiating hPSC colony to varying the above parameters to provide a sense of the sensitivity of the model output to the model parameters. The sensitivity of the model to perturbing the mesh definition is shown in Figure 6-2.

Figure 6-1: Response of predicted gradient formation to perturbation of model parameters

188

Figure 6-2: Gradient formation predicted with pre-defined mesh sizes in Comsol. A) ‘Normal’, B) ‘Fine’, C) ‘Coarse’, D) ‘Finer’, E) ‘Coarser’, F) ‘Extra Fine’, and G) ‘Extra Coarse’.

6.2 Response of Wnt inhibition in peri-gastrulation-like platform

The peri-gastrulation-like platform can provide valuable insight into the signaling requirements that regulate the patterning of gastrulation-associated fates as evidenced by its ability to identify the necessity of Nodal signaling in the induction of the primitive streak compartment. We asked if we could employ this platform to identify if any other signaling pathways that have been implicated as playing important roles in the induction of gastrulation associated fates either in

189 other model organisms or from in vitro studies in hPSCs. To this end, we focused on Wnt as its role in gastrulation and the induction of mesendodermal fates has long been established (Logan & Nusse 2004; van Amerongen & Nusse 2009; Cadigan & Nusse 1997). We asked if inhibiting Wnt signaling in the peri-gastrulation-like induction of geometrically-confined hPSCs would result in aberrant fate patterning or absence of fate compartments as was observed in the case of Nodal inhibition (Figure 3-5). Notably, pharmacological inhibitors or Wnt can enable two different variants of Wnt inhibition. Small molecules like IWP2, and IWP4 can prevent the secretion of Wnt ligands (Chen et al. 2009); and molecules like XAV can inhibit Wnt signaling by promoting degradation of β-catenin by stabilization of a key component of the β-catenin destruction complex – Axin (Huang et al. 2009). We inhibited Wnt signaling in the peri-gastrulation-like platform via both these avenues and asked if this perturbation resulted in the alteration of the expression of either BRA or CDX2. Interestingly, we observed a marked reduction of CDX2 expression upon inhibition or Wnt and a complete abrogation of the BRA expressing compartment (Figure 6-3).

Figure 6-3: Wnt inhibition response in peri-gastrulation-like platform The expression of BRA and CDX2 in response to Wnt inhibition by either inhibiting ligand secretion (IWP2, IWP4) or promoting β-catenin degradation (XAV). The X-axis lists the conditions tested along with the control (N2B27) which represents the control peri-gastrulation- like condition devoid of any perturbations. The Y-axis represents the percentage of each colony that expressed the respective protein. Each dot represents an identified colony. The number of colonies shown are 53 (N2B27), 141 (XAV), 119 (IWP2), and 118 (IWP4). The data were pooled from two experiments.

190

Consistent with these quantified results, whereas the immunofluorescent staining of the control condition showed robust expression of BRA, and CDX2 and nuclear localization of β-catenin in the region expressing the primitive streak compartment (Figure 6-4), the BRA expression in the conditions where Wnt was inhibited was lost, and CDX2 expression appeared downregulated (Figure 6-5). Furthermore, we were unable to observe clear nuclear localization of β-catenin in the colonies (Figure 6-5).

Figure 6-4: Immunofluorescent staining for BRA, CDX2, and β-catenin in control colonies Control colonies (indicated as ‘N2B27’) stained for BRA, CDX2, and β-catenin. The square dashed-box toward the top right corner represents a chosen region of interest (ROI) within the control colonies and the rectangular dashed-box shows immunofluorescnt images stained for β- catenin acquired from individual z-planes.

These data have three different implications. Firstly, these indicate an absolute necessity of Wnt signaling in the induction of BRA and a key role in the induction of CDX2. Secondly, the Wnt signaling required is mediated by Wnt ligands as inhibiting the secretion of Wnt ligands by IWPs abrogates the expression of BRA and drastically impairs the expression of CDX2. Finally, given the fact that Wnt signaling is known to induce mesendodermal lineages out of the hPSC state, and that inhibiting Wnt also downregulated CDX2 expression, these data suggest that at least a subset of the CDX2-positive emergent population from BMP4 treated hPSCs is likely mesodermal in fate. These data can inform the debate around the true identity of the CDX2 positive population that emerges from hPSCs upon BMP4 treatment.

191

Figure 6-5: Immunofluorescent staining for BRA, CDX2, and β-catenin in geometrically confined hPSC colonies treated with BMP4 and Wnt inhibitors Immunofluorescent images of BRA, CDX2, and β-catenin for colonies induced to undergo peri- gastrulation-like patterning in the presence of Wnt inhibitors that either inhibit Wnt ligand

192 secretion (IWP2, IWP4) or promote degradation of β-catenin by stabilizing Axin. The square dashed-box toward the top right corner represents a chosen region of interest (ROI) within the control colonies and the rectangular dashed-box shows immunofluorescnt images stained for β- catenin acquired from individual z-planes.

193

References

Alom Ruiz, S. & Chen, C.S., 2007. Microcontact printing: A tool to pattern. Soft Matter, 3(2), pp.168–177. van Amerongen, R. & Nusse, R., 2009. Towards an integrated view of Wnt signaling in development. Development (Cambridge, England), 136(19), pp.3205–14. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19736321 [Accessed August 9, 2013].

Azioune, A. et al., 2010. Chapter 8 – Protein Micropatterns: A Direct Printing Protocol Using Deep UVs. In Methods in Cell Biology. pp. 133–146. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0091679X10970088.

Azioune, A. et al., 2009. Simple and rapid process for single cell micro-patterning. Lab on a chip, 9(11), pp.1640–1642.

Bachiller, D. et al., 2000. The organizer factors Chordin and Noggin are required for mouse forebrain development. Nature, 403(6770), pp.658–661.

Bauwens, C.L. et al., 2008. Control of human embryonic stem cell colony and aggregate size heterogeneity influences differentiation trajectories. Stem cells, 26(9), pp.2300–2310.

Bauwens, C.L. et al., 2011. Geometric Control of Cardiomyogenic Induction in Human Pluripotent Stem Cells. , 17(15–16), pp.1901–1909.

Bedzhov, I. et al., 2014. In vitro culture of mouse blastocysts beyond the implantation stages. Nature Protocols, 9(12), pp.2732–2739. Available at: http://dx.doi.org/10.1038/nprot.2014.186.

Bedzhov, I. & Zernicka-Goetz, M., 2014. Self-organizing properties of mouse pluripotent cells initiate morphogenesis upon implantation. Cell, 156(5), pp.1032–1044. Available at: http://dx.doi.org/10.1016/j.cell.2014.01.023.

Belle, M. et al., 2017. Tridimensional Visualization and Analysis of Early Human Development. Cell, 169(1), p.161–173.e12.

194

Ben-Zvi, D., Shilo, B.Z. & Barkai, N., 2011. Scaling of morphogen gradients. Current Opinion in Genetics and Development, 21(6), pp.704–710. ten Berge, D. et al., 2008. Wnt signaling mediates self-organization and axis formation in embryoid bodies. Cell stem cell, 3(5), pp.508–18. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2683270&tool=pmcentrez&ren dertype=abstract [Accessed October 20, 2013].

Bernardo, A.S. et al., 2011. BRACHYURY and CDX2 mediate BMP-induced differentiation of human and mouse pluripotent stem cells into embryonic and extraembryonic lineages. Cell Stem Cell, 9(2), pp.144–155.

Besser, D., 2004. Expression of nodal, lefty-A, and lefty-B in undifferentiated human embryonic stem cells requires activation of Smad2/3. Journal of Biological Chemistry, 279(43), pp.45076–45084.

Bjornson, C.R.R. et al., 2005. Eomesodermin is a localized maternal determinant required for endoderm induction in zebrafish. Developmental Cell, 9(4), pp.523–533.

Blauwkamp, T. a et al., 2012. Endogenous Wnt signalling in human embryonic stem cells generates an equilibrium of distinct lineage-specified progenitors. Nature communications, 3, p.1070.

Blin, G. et al., 2017. Geometrical confinement guides Brachyury self-patterning in embryonic stem cells. BioRxiv.

Brennan, J. et al., 2001. Nodal signalling in the epiblast patterns the early mouse embryo. Nature, 411(6840), pp.965–9.

Brink, S.C. Van Den et al., 2014. Symmetry breaking, germ layer specification and axial organization in aggregates of mouse ES cells. In press, pp.4231–4242.

Brinkman, E.K. et al., 2014. Easy quantitative assessment of genome editing by sequence trace decomposition. Nucleic Acids Research, 42(22), pp.1–8.

Briscoe, J. & Small, S., 2015. Morphogen rules: design principles of gradient-mediated embryo

195

patterning. Development, 142(23), pp.3996–4009.

Cadigan, K.M. & Nusse, R., 1997. Wnt signaling: a common theme in animal development. Genes & Development, 11(24), pp.3286–3305. Available at: http://www.genesdev.org/cgi/doi/10.1101/gad.11.24.3286 [Accessed August 9, 2013].

Carlson-Stevermer, J. et al., 2017. Micropatterned Substrates To Promote And Dissect Reprogramming Of Human Somatic Cells. bioRxiv.

Chaudhuri, O. et al., 2016. Hydrogels with tunable stress relaxation regulate stem cell fate and activity. Nature Materials, 15(3), pp.326–334.

Chen, B. et al., 2009. Small molecule-mediated disruption of Wnt-dependent signaling in tissue regeneration and cancer. Nature Chemical Biology, 5(2), pp.100–107.

Chen, C., 2005. The Vg1-related protein Gdf3 acts in a Nodal signaling pathway in the pre- gastrulation mouse embryo. Development, 133(2), pp.319–329. Available at: http://dev.biologists.org/cgi/doi/10.1242/dev.02210.

Chen, H. et al., 2012. A system of repressor gradients spatially organizes the boundaries of bicoid-dependent target genes. Cell, 149(3), pp.618–629.

Chen, S. et al., 2016. Interrogating cellular fate decisions with high-throughput arrays of multiplexed cellular communities. Nature Communications, 7, pp.1–8. Available at: http://dx.doi.org/10.1038/ncomms10309.

Chi, L. et al., 2011. A secreted BMP antagonist, Cer1, fine tunes the spatial organization of the ureteric bud tree during mouse kidney development. PLoS ONE, 6(11).

Church, R.H. et al., 2015. Gremlin1 preferentially binds to bone morphogenetic protein-2 (BMP- 2) and BMP-4 over BMP-7. Biochemical Journal, 466(1), p.55 LP-68. Available at: http://www.biochemj.org/content/466/1/55.abstract.

Ciruna, B. & Rossant, J., 2001. Mesoderm Cell Fate Specification and Morphogenetic Movement at the Primitive Streak. Dev Cell., 1(1), pp.37–49.

Cooke, J. & Zeeman, E.C., 1976. A clock and wavefront model for control of the number of

196

repeated structures during animal morphogenesis. Journal of Theoretical Biology, 58(2), pp.455–476.

Cuny, G.D. et al., 2008. Structure–activity relationship study of bone morphogenetic protein (BMP) signaling inhibitors. Bioorganic & Medicinal Chemistry Letters, 18(15), pp.4388– 4392.

Davies, J., 2017. Using synthetic biology to explore principles of development. Development, 144(7), pp.1146–1158. Available at: http://dev.biologists.org/lookup/doi/10.1242/dev.144196.

Deglincerti, A., Etoc, F., et al., 2016. Self-organization of human embryonic stem cells on micropatterns. Nature Protocols, 11(11), pp.2223–2232.

Deglincerti, A., Croft, G.F., et al., 2016. Self-organization of the in vitro attached human embryo. Nature, 533(7602), pp.251–254.

Demers, C.J. et al., 2016. Development-on-chip: in vitro neural tube patterning with a microfluidic device. Development, 143(11), pp.1884–1892. Available at: http://dev.biologists.org/lookup/doi/10.1242/dev.126847.

Dessaud, E. et al., 2007. Interpretation of the sonic hedgehog morphogen gradient by a temporal adaptation mechanism. Nature, 450(7170), pp.717–720.

Dessaud, E., McMahon, A.P. & Briscoe, J., 2008. Pattern formation in the vertebrate neural tube: a sonic hedgehog morphogen-regulated transcriptional network. Development, 135(15), pp.2489–2503.

Doench, J.G. et al., 2016. Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nature Biotechnology, 34(2), pp.184–191. Available at: http://dx.doi.org/10.1038/nbt.3437.

Doronin, Y.K. et al., 2016. Cleavage of human embryos: Options and diversity. Acta Naturae, 8(3), pp.88–96.

Dupont, S. et al., 2011. Role of YAP/TAZ in mechanotransduction. Nature, 474(7350), pp.179–

197

83. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21654799%5Cnhttp://dx.doi.org/10.1038/nature1013 7.

Economou, A.D. et al., 2012. Periodic stripe formation by a Turing mechanism operating at growth zones in the mammalian palate. Nature genetics, 44(3), pp.348–351.

Eiraku, M. et al., 2011. Self-organizing optic-cup morphogenesis in three-dimensional culture. Nature, 472(7341), pp.51–56. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21475194.

Etoc, F. et al., 2016. A Balance between Secreted Inhibitors and Edge Sensing Controls Gastruloid Self-Organization. Developmental Cell, 39(3), pp.302–315.

Fernandez-Gonzalez, R. & Zallen, J.A., 2011. Oscillatory behaviors and hierarchical assembly of contractile structures in intercalating cells. Physical Biology, 8(4), p.45005.

Findlay, S.D. & Postovit, L.-M., 2018. Comprehensive characterization of transcript diversity at the human NODAL locus. BioRxiv, pp.1–26.

Fleming, T.P. et al., 1993. Localisation of tight junction protein cingulin is temporally and spatially regulated during early mouse development. Development (Cambridge, England), 117(3), pp.1135–44. Available at: http://www.ncbi.nlm.nih.gov/pubmed/8325238.

Fogarty, N.M.E. et al., 2017. Genome editing reveals a role for OCT4 in human embryogenesis. Nature, 550(7674), pp.67–73.

Funa, N.S. et al., 2015. β-Catenin Regulates Primitive Streak Induction through Collaborative Interactions with SMAD2/SMAD3 and OCT4. Cell Stem Cell, 16(6), pp.639–652.

Garcia-Gonzalo, F.R. & Belmonte, J.C.I., 2008. Albumin-associated lipids regulate human embryonic stem cell self-renewal. PLoS ONE, 3(1), pp.1–10.

Gierer, A. & Meinhardt, H., 1972. A theory of biological pattern formation. Kybernetik, 12(1), pp.30–39.

Gilbert, P.M. et al., 2010. Substrate Elasticity Regulates Skeletal Muscle Stem Cell. Science,

198

1078(2010), pp.1078–1081. Available at: http://www.ncbi.nlm.nih.gov/pubmed/20647425%5Cnhttp://www.sciencemag.org/cgi/doi/1 0.1126/science.1191035.

Gilbert, S., 2010. Developmental Biology ninth., Sunderland, MA: Sinauer.

Gjorevski, N. et al., 2016. Designer matrices for intestinal stem cell and organoid culture. Nature, 539(7630), pp.560–564. Available at: http://www.nature.com/doifinder/10.1038/nature20168.

Gomez, C. et al., 2008. Control of segment number in vertebrate embryos. Nature, 454(17).

Goolam, M. et al., 2016. Heterogeneity in Oct4 and Sox2 Targets Biases Cell Fate in 4-Cell Mouse Embryos. Cell, 165(1), pp.61–74. Available at: http://dx.doi.org/10.1016/j.cell.2016.01.047.

Gouti, M. et al., 2014. In Vitro Generation of Neuromesodermal Progenitors Reveals Distinct Roles for Wnt Signalling in the Specification of Spinal Cord and Paraxial Mesoderm Identity. , 12(8).

Green, J.B.A. & Smith, J.C., 1990. Graded changes in dose of a Xenopus activin A homologue elicit stepwise transitions in embryonic cell fate. Nature, 347, pp.391–394.

Green, J.B. a & Sharpe, J., 2015. Positional information and reaction-diffusion: two big ideas in developmental biology combine. Development, 142(7), pp.1203–1211.

Green, M.D. et al., 2011. Generation of anterior foregut endoderm from human embryonic and induced pluripotent stem cells. Nature biotechnology, 29(3), pp.267–272.

Greene, N.D.E. & Copp, A.J., 2009. Development of the vertebrate central nervous system: formation of the neural tube. Prenatal diagnosis, 29, pp.303–311.

Greene, N.D.E. & Copp, A.J., 2014. Neural Tube Defects. Annual Review of Neuroscience, 37(1), pp.221–242. Available at: http://www.annualreviews.org/doi/10.1146/annurev-neuro- 062012-170354.

Gregor, T. et al., 2007. Probing the Limits to Positional Information. Cell, 130(1), pp.153–164.

199

Gregor, T., McGregor, A.P. & Wieschaus, E.F., 2008. Shape and function of the Bicoid morphogen gradient in dipteran species with different sized embryos. Developmental Biology, 316(2), pp.350–358.

Groves, A.K. & LaBonne, C., 2014. Setting appropriate boundaries: Fate, patterning and competence at the neural plate border. Developmental Biology, 389(1), pp.39–49. Available at: http://dx.doi.org/10.1016/j.ydbio.2013.11.027.

Gurdon, J.B., 1988. A community effect in animal development. Nature, 336(6201), pp.772– 774. Available at: http://www.nature.com/doifinder/10.1038/336772a0.

Guschin, D.Y. et al., 2010. A Rapid and General Assay for Monitoring Endogenous Gene Modificatio. Methods in Molecular Biology, 649, pp.247–256. Available at: http://link.springer.com/10.1007/978-1-60761-753-2.

Harrison, S.E. et al., 2017. Assembly of embryonic and extraembryonic stem cells to mimic embryogenesis in vitro. Science, 356(6334), p.eaal1810. Available at: http://www.sciencemag.org/lookup/doi/10.1126/science.aal1810.

Hart, A.H. et al., 2002. Mixl1 is required for axial mesendoderm morphogenesis and patterning in the murine embryo. Development (Cambridge, England), 129(15), pp.3597–608. Available at: http://www.ncbi.nlm.nih.gov/pubmed/12117810.

Henrique, D. et al., 2015. Neuromesodermal progenitors and the making of the spinal cord. Development (Cambridge, England), 142(17), pp.2864–2875.

Hermanson, G.T., 2013. Zero-Length Crosslinkers. Bioconjugate Techniques, pp.259–273. Available at: http://linkinghub.elsevier.com/retrieve/pii/B9780123822390000042.

Hertig, A.T., Rock, J. & Adams, E.C., 1956. A description of 34 human ova within the first 17 days of development. American Journal of Anatomy, 98(3), pp.435–493.

Hong, S.-K. et al., 2011. Embryonic mesoderm and endoderm induction requires the actions of non-embryonic Nodal-related ligands and Mxtx2. Development (Cambridge, England), 138(4), pp.787–95.

200

Houchmandzadeh, B., Wieschaus, E. & Leibler, S., 2002. Establishment of developmental precision and proportions in the early Drosophila embryo. Nature Letters, 415, pp.798–802.

Huang, S.A. et al., 2009. Tankyrase inhibition stabilizes axin and antagonizes Wnt signalling. Nature, 461(7264), pp.614–620. Available at: http://dx.doi.org/10.1038/nature08356.

Hubaud, A. et al., 2017. Excitable Dynamics and Yap-Dependent Mechanical Cues Drive the Segmentation Clock. Cell, 171(3), p.668–682.e11.

Hung, W.-T. et al., 2012. DAN (NBL1) Specifically Antagonizes BMP2 and BMP4 and Modulates the Actions of GDF9, BMP2, and BMP4 in the Rat Ovary1. Biology of Reproduction, 86(5), pp.1–9. Available at: https://academic.oup.com/biolreprod/article- lookup/doi/10.1095/biolreprod.111.096172.

Inman, G.J. et al., 2002. SB-431542 is a potent and specific inhibitor of transforming growth factor-beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4, ALK5, and ALK7. Molecular pharmacology, 62(1), pp.65–74.

Inomata, H. et al., 2013. Scaling of Dorsal-Ventral Patterning by Embryo Size-Dependent Degradation of Spemann’s Organizer Signals. Cell, 153(6), pp.1296–1311.

Irie, N. et al., 2015. SOX17 is a critical specifier of human primordial germ cell fate. Cell, 160(1–2), pp.253–268.

Iwata, K. et al., 2014. Analysis of compaction initiation in human embryos by using time-lapse cinematography. Journal of Assisted Reproduction and Genetics, 31(4), pp.421–426.

Jaeger, J. et al., 2004. Dynamic control of positional information in the early Drosophila embryo. Nature, 430(6997), pp.368–371.

Johnson, M.B., March, A.R. & Morsut, L., 2017. Engineering multicellular systems: Using synthetic biology to control tissue self-organization. Current Opinion in Biomedical Engineering, 4(August), pp.163–173. Available at: http://linkinghub.elsevier.com/retrieve/pii/S2468451117300375.

Kane, R.S. et al., 1999. Patterning proteins and cells using soft lithography. Biomaterials, 20(23–

201

24), pp.2363–76. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10614942.

Keller, A. et al., 2018. Genetic and epigenetic factors which modulate differentiation propensity in human pluripotent stem cells. Human Reproduction Update, (February), pp.1–14. Available at: http://academic.oup.com/humupd/advance- article/doi/10.1093/humupd/dmx042/4825062.

Keller, R., 2005. Cell migration during gastrulation. Current opinion in cell biology, 17(5), pp.533–541.

Khademhosseini, A. et al., 2004. A Softlithographic Approach To Fabricate Patterned Microfluidic Channels. Analytical Chemisty, 76(13), pp.5783–5789.

Kinder, S.J. et al., 1999. The orderly allocation of mesodermal cells to the extraembryonic structures and the anteroposterior axis during gastrulation of the mouse embryo. Development, 126(21), pp.4691–4701. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10518487.

Knop, K. et al., 2010. Poly(ethylene glycol) in drug delivery: Pros and cons as well as potential alternatives. Angewandte Chemie - International Edition, 49(36), pp.6288–6308.

Kobayashi, T. et al., 2017. Principles of early human development and germ cell program from conserved model systems. Nature Publishing Group.

Kojima, Y. et al., 2014. The transcriptional and functional properties of mouse epiblast stem cells resemble the anterior primitive streak. Cell Stem Cell, 14(1), pp.107–120.

Kondo, S. & Miura, T., 2010. Reaction-Diffusion Model as a Framework for Understanding Biological Pattern Formation. Science, 329(5999), pp.1616–1620.

Kretzschmar, K. & Clevers, H., 2016. Organoids: Modeling Development and the Stem Cell Niche in a Dish. Developmental Cell, 38(6), pp.590–600. Available at: http://dx.doi.org/10.1016/j.devcel.2016.08.014.

Lancaster, M.A. et al., 2017. Guided self-organization and cortical plate formation in human brain organoids. Nature Biotechnology, 35(7), pp.659–666.

202

Lancaster, M.A. & Knoblich, J.A., 2014. Organogenesisin a dish: Modeling development and disease using organoid technologies. Science, 345(6194).

Lancaster, M. a et al., 2012. Cerebral organoids model human brain development and microcephaly. Nature, 501(1), pp.373–9. Available at: http://www.ncbi.nlm.nih.gov/pubmed/22813947%5Cnhttp://www.ncbi.nlm.nih.gov/pubmed /23940280%5Cnhttp://www.ncbi.nlm.nih.gov/pubmed/22704498%5Cnhttp://www.pubmed central.nih.gov/articlerender.fcgi?artid=3899231&tool=pmcentrez&rendertype=abstract%5 Cnhttp://www.ncbi.nl.

Lawson, K.A. et al., 1999. Bmp4 is required for the generation of primordial germ cells in the mouse embryo. Genes & development, pp.424–436.

Lee, L.H. et al., 2009. Micropatterning of human embryonic stem cells dissects the mesoderm and endoderm lineages. Stem Cell Research, 2(2), pp.155–162.

Lee, M.T., Bonneau, A.R. & Giraldez, A.J., 2014. Zygotic genome activation during the maternal-to-zygotic transition. Annu Rev Cell Dev Biol. 2014, (30), pp.581–613.

Lensch, M.W. et al., 2007. Teratoma Formation Assays with Human Embryonic Stem Cells: A Rationale for One Type of Human-Animal Chimera. Cell Stem Cell, 1(3), pp.253–258.

Leptin, M., 2005. Gastrulation movements: the logic and the nuts and bolts. Developmental cell, 8(3), pp.305–320.

Leung, C.Y.B. & Fernandez-Gonzalez, R., 2015. Quantitative Image Analysis of Cell Behavior and Molecular Dynamics During Tissue Morphogenesis. Methods in Molecular Biology, 1189.

Li, Y. et al., 2013. BMP4-directed trophoblast differentiation of human embryonic stem cells is mediated through a ΔNp63+ cytotrophoblast stem cell state. Development (Cambridge, England), 140(19), pp.3965–76. Available at: http://www.ncbi.nlm.nih.gov/pubmed/24004950.

Lim, W.A., 2010. Designing customized cell signalling circuits. Nature Reviews Molecular Cell Biology, 11(6), pp.393–403. Available at: http://dx.doi.org/10.1038/nrm2904.

203

Liu, P. et al., 1999. Requirement for Wnt3 in vertebrate axis formation. Nature genetics, 22(4), pp.361–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/10431240.

Liu, Z. et al., 2016. Efficient CRISPR/Cas9-Mediated Versatile, Predictable, and Donor-Free Gene Knockout in Human Pluripotent Stem Cells. Stem Cell Reports, 7(3), pp.496–507. Available at: http://dx.doi.org/10.1016/j.stemcr.2016.07.021.

Logan, C.Y. & Nusse, R., 2004. The Wnt signaling pathway in development and disease. Annual review of cell and developmental biology, 20, pp.781–810. Available at: http://www.ncbi.nlm.nih.gov/pubmed/15473860 [Accessed August 6, 2013].

Loh, K.M.M. et al., 2016. Mapping the Pairwise Choices Leading from Pluripotency to Human Bone, Heart, and Other Mesoderm Cell Types. Cell, 166(2), pp.451–468.

Ma, Z. et al., 2015. Self-organizing human cardiac microchambers mediated by geometric confinement. Nature communications, 6(7413).

Madl, C.M. et al., 2017. Maintenance of neural progenitor cell stemness in 3D hydrogels requires matrix remodelling. Nature Materials, 16(December). Available at: http://www.nature.com/doifinder/10.1038/nmat5020.

Marcon, L. et al., 2016. High-throughput mathematical analysis identifies turing networks for patterning with equally diffusing signals. eLife, 5(APRIL2016), pp.1–60.

McBeath, R. et al., 2004. Cell shape, cytoskeletal tension, and RhoA regulate stemm cell lineage commitment. Developmental Cell, 6, pp.483–495.

Meinhardt, A. et al., 2014. 3D reconstitution of the patterned neural tube from embryonic stem cells. Stem Cell Reports, 3(6), pp.987–999.

Meinhardt, H., 2015. Models for patterning primary embryonic body axes: The role of space and time. Seminars in Cell and Developmental Biology, 42, pp.103–117. Available at: http://dx.doi.org/10.1016/j.semcdb.2015.06.005.

Mendjan, S. et al., 2014. NANOG and CDX2 Pattern Distinct Subtypes of Human Mesoderm during Exit from Pluripotency. Cell Stem Cell, 15(3), pp.310–325.

204

Morgani, S.M. et al., 2017. Micropattern differentiation of mouse pluripotent stem cells recapitulates embryo regionalized fates and patterning. , pp.1–35. Available at: http://dx.doi.org/10.1101/236562.

Morris, S.A. et al., 2012. Dynamics of anterior-posterior axis formation in the developing mouse embryo. Nature communications, 3(673).

Müller, F.-J. et al., 2011. A bioinformatic assay for pluripotency in human cells. Nature Methods, 8(4), pp.315–317. Available at: http://www.nature.com/doifinder/10.1038/nmeth.1580.

Muller, P. et al., 2012. Differential Diffusivity of Nodal and Lefty Underlies a Reaction- Diffusion Patterning System. Science, 336(6082), pp.721–724.

Mulloy, B. & Rider, C.C., 2015. The Bone Morphogenetic Proteins and Their Antagonists 1st ed., Elsevier Inc. Available at: http://dx.doi.org/10.1016/bs.vh.2015.06.004.

Murray, J.D., 2008. Mathematical Biology II - Spatial Models and Biomedical Applications Third. Springer, ed.,

Murry, C.E. & Keller, G., 2008. Differentiation of embryonic stem cells to clinically relevant populations: lessons from embryonic development. Cell, 132(4), pp.661–80. Available at: http://www.ncbi.nlm.nih.gov/pubmed/18295582 [Accessed November 7, 2013].

Myers, D.C., Sepich, D.S. & Solnica-Krezel, L., 2002. Convergence and extension in vertebrate gastrulae: cell movements according to or in search of identity? Trends in genetics, 18(9), pp.447–455.

Nakaya, Y. & Sheng, G., 2008. Epithelial to mesenchymal transition during gastrulation: an embryological view. Development, 50(9), pp.755–766.

Narasimha, M. & Leptin, M., 2000. Cell movements during gastrulation: come in and be induced. Trends in cell biology, 10(5), pp.169–72.

Nazareth, E.J.P. et al., 2013. High-throughput fingerprinting of human pluripotent stem cell fate responses and lineage bias. Nature methods, 10(12), pp.1225–1231.

205

Nemashkalo, A. et al., 2017. Morphogen and community effects determine cell fates in response to BMP4 signaling in human embryonic stem cells. BioRxiv.

Nikolopoulou, E. et al., 2017. Neural tube closure: cellular, molecular and biomechanical mechanisms. Development, 144(4), pp.552–566. Available at: http://dev.biologists.org/lookup/doi/10.1242/dev.145904.

Nostro, M.C. et al., 2015. Efficient generation of NKX6-1+ pancreatic progenitors from multiple human pluripotent stem cell lines. Stem Cell Reports, 4(4), pp.591–604. Available at: http://dx.doi.org/10.1016/j.stemcr.2015.02.017.

Ohinata, Y. et al., 2009. A Signaling Principle for the Specification of the Germ Cell Lineage in Mice. Cell, 137(3), pp.571–584. Available at: http://dx.doi.org/10.1016/j.cell.2009.03.014.

Onishi, K. et al., 2014. Local BMP-SMAD1 signaling increases LIF receptor-dependent STAT3 responsiveness and primed-to-naive mouse pluripotent stem cell conversion frequency. Stem Cell Reports, 3(1), pp.156–168. Available at: http://dx.doi.org/10.1016/j.stemcr.2014.04.019.

Onishi, K. et al., 2012. Microenvironment-mediated reversion of epiblast stem cells by reactivation of repressed JAK–STAT signaling. Integrative Biology, 4(11), p.1367.

Ortmann, D. & Vallier, L., 2017. Variability of human pluripotent stem cell lines. Current Opinion in Genetics & Development, 46, pp.179–185. Available at: http://linkinghub.elsevier.com/retrieve/pii/S0959437X16301861.

Peerani, R., Onishi, K., et al., 2009. Manipulation of signaling thresholds in “engineered stem cell niches” identifies design criteria for pluripotent stem cell screens. PloS one, 4(7), p.e6438. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2713412&tool=pmcentrez&ren dertype=abstract [Accessed November 14, 2013].

Peerani, R. et al., 2007. Niche-mediated control of human embryonic stem cell self-renewal and differentiation. The EMBO journal, 26(22), pp.4744–4755.

Peerani, R., Bauwens, C., et al., 2009. Stem Cells in Regenerative Medicine J. Audet & W. L.

206

Stanford, eds. , 482, pp.21–33. Available at: http://www.springerlink.com/index/10.1007/978-1-59745-060-7 [Accessed November 7, 2013].

Pieper, M. et al., 2012. Differential distribution of competence for panplacodal and neural crest induction to non-neural and neural ectoderm. Development, 139(6), pp.1175–1187. Available at: http://dev.biologists.org/cgi/doi/10.1242/dev.074468.

Plouhinec, J.L. et al., 2017. A molecular atlas of the developing ectoderm defines neural, neural crest, placode, and nonneural progenitor identity in vertebrates,

Poh, Y. et al., 2014. Generation of organized germ layers from a single mouse embryonic stem cell. Nature Communications, 5(May), pp.1–12. Available at: http://dx.doi.org/10.1038/ncomms5000.

Pourquié, O., 2004. The chick embryo: A leading model in somitogenesis studies. Mechanisms of Development, 121(9), pp.1069–1079.

Rahman, N. et al., 2017. Engineering the haemogenic niche mitigates endogenous inhibitory signals and controls pluripotent stem cell-derived blood emergence. Nature Communications, 8(May), p.15380. Available at: http://www.nature.com/doifinder/10.1038/ncomms15380.

Ranga, A. et al., 2017. Neural tube morphogenesis in synthetic 3D microenvironments. Proceedings of the National Academy of Sciences, 114(15), pp.E3163–E3163. Available at: http://www.pnas.org/lookup/doi/10.1073/pnas.1703993114.

Raspopovic, J. et al., 2014. Digit patterning is controlled by a Bmp-Sox9-Wnt Turing network modulated by morphogen gradients. Science, 345(6196), pp.566–570.

Rider, C.C. & Mulloy, B., 2010. Bone morphogenetic protein and growth differentiation factor cytokine families and their protein antagonists. Biochemical Journal, 429(1), pp.1–12. Available at: http://biochemj.org/lookup/doi/10.1042/BJ20100305.

Robinton, D.A. & Daley, G.Q., 2012. The promise of induced pluripotent stem cells in research and therapy. Nature, 481(7381), pp.295–305. Available at:

207

http://www.nature.com/doifinder/10.1038/nature10761.

Rödiger, S. et al., 2011. Fluorescence dye adsorption assay to quantify carboxyl groups on the surface of poly(methyl methacrylate) microbeads. Analytical chemistry, 83(9), pp.3379– 3385.

Rossant, J., 2015. Mouse and human blastocyst-derived stem cells: vive les differences. Development, 142(1), pp.9–12.

Rossant, J. & Tam, P.P.L., 2009. Blastocyst lineage formation, early embryonic asymmetries and axis patterning in the mouse. Development (Cambridge, England), 136(5), pp.701–13.

Rossant, J. & Tam, P.P.L., 2004. Emerging Asymmetry and Embryonic Patterning in Early Mouse Development. Developmental cell, 7, pp.155–164.

Ruzo, A. & Brivanlou, A.H., 2017. At Last: Gene Editing in Human Embryos to Understand Human Development. Cell Stem Cell, 21(5), pp.564–565. Available at: https://doi.org/10.1016/j.stem.2017.10.008.

Sato, T. et al., 2009. Single Lgr5 stem cells build crypt-villus structures in vitro without a mesenchymal niche. Nature, 459(7244), pp.262–5. Available at: http://www.ncbi.nlm.nih.gov/pubmed/19329995.

Schier, A.F., 2003. Nodal signaling in vertebrate development. Annual review of cell and developmental biology, 19, pp.589–621.

Scialdone, A. et al., 2016. Resolving early mesoderm diversification through single-cell expression profiling. Nature, 535(7611), pp.289–293. Available at: http://dx.doi.org/10.1038/nature18633.

Shahbazi, M.N. et al., 2016. Self-organization of the human embryo in the absence of maternal tissues. Nature Cell Biology, (February).

Shao, Y. et al., 2017. A pluripotent stem cell-based model for post-implantation human amniotic sac development. Nature Communications, 8(1), p.208. Available at: http://www.nature.com/articles/s41467-017-00236-w.

208

Shao, Y. et al., 2016. Self-organized amniogenesis by human pluripotent stem cells in a biomimetic implantation-like niche. Nature Materials, 16(4), pp.419–425. Available at: http://www.nature.com/doifinder/10.1038/nmat4829.

Shen, M.M., 2007. Nodal signaling: developmental roles and regulation. Development (Cambridge, England), 134(6), pp.1023–34.

Sick, S. et al., 2006. WNT and DKK Determine Hair Follicle Spacing through a Reaction- Diffusion Mechanism. Science, 314(5804), pp.1447–1450.

Siggia, E.D. & Warmflash, A., 2017. Modeling mammalian gastrulation with embryonic stem cells. Available at: http://arxiv.org/abs/1712.03335.

Simoes-Costa, M. & Bronner, M.E., 2015. Establishing neural crest identity: a gene regulatory recipe. Development, 142(2), pp.242–257. Available at: http://dev.biologists.org/cgi/doi/10.1242/dev.105445.

Simões-costa, M. & Bronner, M.E., 2013. Insights into neural crest development and evolution from genomic analysis Insights into neural crest development and evolution from genomic analysis. , pp.1069–1080.

Smith, W.C. & Harland, R.M., 1992. Expression cloning of noggin, a new dorsalizing factor localized to the Spemann organizer in Xenopus embryos. Cell, 70(5), pp.829–840.

Snijder, B. et al., 2009. Population context determines cell-to-cell variability in endocytosis and virus infection. Nature, 461(7263), pp.520–523.

Solnica-Krezel, L., 2005. Conserved patterns of cell movements during vertebrate gastrulation. Current biology : CB, 15(6), pp.R213-28.

Solnica-Krezel, L. & Sepich, D.S., 2012. Gastrulation: making and shaping germ layers. Annual review of cell and developmental biology, 28, pp.687–717.

Stevens, K.R. et al., 2013. InVERT molding for scalable control of tissue microarchitecture. Nature communications, 4(1847).

Sudo, S. et al., 2004. Protein related to DAN and cerberus is a bone morphogenetic protein

209

antagonist that participates in ovarian paracrine regulation. Journal of Biological Chemistry, 279(22), pp.23134–23141.

Sun, J. et al., 2006. BMP4 activation and secretion are negatively regulated by an intracellular Gremlin-BMP4 interaction. Journal of Biological Chemistry, 281(39), pp.29349–29356.

Tajbakhsh, S. & Spörle, R., 1998. Somite development: Constructing the vertebrate body. Cell, 92(1), pp.9–16.

Takahashi, K. et al., 2007. Induction of Pluripotent Stem Cells from Adult Human Fibroblasts by Defined Factors. Cell, 131(5), pp.861–872.

Tam, P.P. & Behringer, R.R., 1997. Mouse gastrulation: the formation of a mammalian body plan. Mechanisms of development, 68(1–2), pp.3–25.

Tam, P.P.L. & Loebel, D.A.F., 2007. Gene function in mouse embryogenesis: get set for gastrulation. Nature reviews. Genetics, 8(5), pp.368–81.

Tam, P.P.L., Loebel, D.A.F. & Tanaka, S.S., 2006. Building the mouse gastrula: signals, asymmetry and lineages. Current opinion in genetics & development, 16(4), pp.419–425.

Tanaka, S. et al., 1998. Promotion to trophoblast stem cell proliferation by FGF4. Science, 282(5396), pp.2072–2075.

Teo, A.K.K. et al., 2011. Pluripotency factors regulate definitive endoderm specification through eomesodermin. Genes and Development, 25(3), pp.238–250.

Tewary, M. et al., 2017. A stepwise model of Reaction-Diffusion and Positional-Information governs self-organized human peri-gastrulation-like patterning. Development, p.dev.149658. Available at: http://dev.biologists.org/lookup/doi/10.1242/dev.149658.

Tewary, M. & Zandstra, P., 2018. Mechanics-guided developmental fate patterning. Nature Materials, 17(July), pp.571–572.

Théry, M., 2010. Micropatterning as a tool to decipher cell morphogenesis and functions. Journal of cell science, 123(Pt 24), pp.4201–13. Available at: http://www.ncbi.nlm.nih.gov/pubmed/21123618 [Accessed November 7, 2013].

210

Thomson, J.A. et al., 1998. Embryonic Stem Cell Lines Derived from Human Blastocysts. Science, 282(5391), pp.1145–1147.

Todhunter, M.E. et al., 2015. Programmed synthesis of three-dimensional tissues. Nature Methods, 12(10), pp.975–981.

Trappmann, B. et al., 2012. Extracellular-matrix tethering regulates stem-cell fate. Nature Materials, 11(7), pp.642–649. Available at: http://dx.doi.org/10.1038/nmat3339.

Tsakiridis, A. et al., 2014. Distinct Wnt-driven primitive streak-like populations reflect in vivo lineage precursors. Development (Cambridge, England), 141(6), pp.1209–21.

Tsankov, A.M. et al., 2015. A qPCR ScoreCard quantifies the differentiation potential of human pluripotent stem cells. Nature Biotechnology, 33(11), pp.1182–1192. Available at: http://www.nature.com/doifinder/10.1038/nbt.3387.

Tsiairis, C.D. & Aulehla, A., 2016. Self-Organization of Embryonic Genetic Oscillators into Spatiotemporal Wave Patterns. Cell, 164(4), pp.656–667. Available at: http://dx.doi.org/10.1016/j.cell.2016.01.028.

Turing, A., 1952. The chemical basis of morphogenesis. Bulletin of Mathematical Biology, 237(641), pp.153–197.

Turner, D.A. et al., 2014. Wnt / β -catenin and FGF signalling direct the specification and maintenance of a neuromesodermal axial progenitor in ensembles of mouse embryonic stem cells. Development, pp.4243–4253.

Turner, N. & Grose, R., 2010. Fibroblast growth factor signalling: from development to cancer. Nature Reviews Cancer, 10, pp.116–129.

Vallier, L., Reynolds, D. & Pedersen, R.A., 2004. Nodal inhibits differentiation of human embryonic stem cells along the neuroectodermal default pathway. Developmental Biology, 275(2), pp.403–421.

Voiculescu, O. et al., 2014. Local cell interactions and self-amplifying individual cell ingression drive amniote gastrulation. eLife, 2014(3), pp.1–26.

211

Wang, R.N. et al., 2014. Bone Morphogenetic Protein (BMP) signaling in development and human diseases. Genes and Diseases, 1(1), pp.87–105. Available at: http://dx.doi.org/10.1016/j.gendis.2014.07.005.

Warmflash, A. et al., 2014. A method to recapitulate early embryonic spatial patterning in human embryonic stem cells. Nature methods, 11(8), pp.847–854.

Watt, F.M. & Huck, W.T.S., 2013. Role of the extracellular matrix in regulating stem cell fate. Nature Reviews Molecular Cell Biology, 14(8), pp.467–473. Available at: http://dx.doi.org/10.1038/nrm3620.

Watt, F.M., Jordan, P.W. & O’Neill, C.H., 1988. Cell shape controls terminal differentiation of human epidermal keratinocytes. Proceedings of the National Academy of Sciences of the United States of America, 85(15), pp.5576–80. Available at: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=281801&tool=pmcentrez&rend ertype=abstract.

Weiss, A. & Attisano, L., 2013. The TGFbeta superfamily signaling pathway. Wiley Interdisciplinary Reviews: Developmental Biology, 2(1), pp.47–63.

Wharton, K. & Derynck, R., 2009. TGF family signaling: novel insights in development and disease. Development, 136(22), pp.3691–3697. Available at: http://dev.biologists.org/cgi/doi/10.1242/dev.040584.

Whitesides, G.M. et al., 2001. Soft lithography in biology and biochemistry. Annual review of biomedical engineering, 3, pp.335–73. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11447067.

Wolpert, L., 1981. Positional information and pattern formation. Phil. Trans. R. Soc. Lond. B, 295, p.441–450.

Wolpert, L., 1969. Positional information and the spatial pattern of cellular differentiation. Journal of theoretical biology, 25(1), pp.1–47.

Wu, M.Y. & Hill, C.S., 2009. TGF-b Superfamily Signaling in Embryonic Development and Homeostasis. Developmental Cell, 16(3), pp.329–343.

212

Xu, R.-H. et al., 2002. BMP4 initiates human embryonic stem cell differentiation to trophoblast. Nature Biotechnology, 20(12), pp.1261–1264.

Yang, Y. et al., 2015. Heightened potency of human pluripotent stem cell lines created by transient BMP4 exposure. Proceedings of the National Academy of Sciences, 112(18), pp.E2337–E2346.

Zhang, H. & Bradley, A., 1996. Mice deficient for BMP2 are nonviable and have defects in amnion/chorion and cardiac development. Development (Cambridge, England), 122(10), pp.2977–2986.

Zhang, J. & Li, L., 2005. BMP signaling and stem cell regulation. Developmental Biology, 284(1), pp.1–11.

Zimmerman, L.B., De Jesús-Escobar, J.M. & Harland, R.M., 1996. The Spemann organizer signal noggin binds and inactivates bone morphogenetic protein 4. Cell, 86(4), pp.599–606.

213

Chapter 1 – Some of the figures in this chapter were adapted from Motifolio images. Some portions of this chapter have been published in Nature Materials (Tewary, and Zandstra, 2018) accepted for publication in Nature Reviews Genetics (Tewary, Shakiba, and Zandstra, 2018). Other portions of this chapter are in preparation for submission for publication.

Chapter 2 – This chapter is in preparation for submission for publication

Chapter 3 – A stepwise model of reaction-diffusion and positional information governs self- organized human peri-gastrulation-like patterning

A version of this chapter was published in Development (Tewary et al., 2017). Co-authors include Joel Ostblom, Laura Prochazka, Teresa Zulueta-Coarasa, Nika Shakiba, Rodrigo Fernandez-Gonzalez, Peter Zandstra.

Chapter 4 – Nodal dissects peri-gastrulation-like and pre-neurulation-like fate patterning in geometrically confined human pluripotent stem cell colonies

A version of this chapter is under peer review. Co-authors include Mukul Tewary, Dominika Dziedzicka, Joel Ostblom, Laura Prochazka, Nika Shakiba, Curtis Woodford, Nafees Rahman, Elia Piccinini, Davide Danovi, Mieke Geens, Fiona M. Watt, Peter W. Zandstra

Chapter 5 – Some portions of this chapter have been published in Nature Materials (Tewary and Zandstra, 2018), Nature Reviews Genetics (Tewary, Shakiba, and Zandstra, 2018) and others are in preparation for submission for publication.