An in Vitro Investigation of the Impact of PAX6 Haploinsufficiency on Human Corneal Stromal Cell Phenotype

Home , Aniridia, Corneal keratocyte, Ligand (biochemistry)

An in vitro investigation of the impact of PAX6 haploinsufficiency on human corneal stromal cell phenotype

Carla Sanchez Martinez

Supervisor: Professor Julie T Daniels

Institute of Ophthalmology

University College London

Thesis submitted to UCL for the degree of Doctor of Philosophy

September 2019

‘I, Carla Sanchez Martinez confirm that the work presented in this thesis is my own. Where information has been derived from other sources, I confirm that this has been indicated in the thesis.'

ABSTRACT

Aniridia related keratopathy (ARK), caused by PAX6 haploinsufficiency, leads to corneal opacification and sight loss. ARK is still an unmet clinical need, as treatments are not always available or successful. Therefore, it is essential to clarify the biological mechanisms behind ARK to develop new efficient clinical treatments.

It was hypothesized that aniridic corneal stromal cells are affected and contribute to the development of ARK. To test this hypothesis, stromal cells were isolated from central human aniridic corneas and compared to normal corneal stromal cells isolated from healthy donors. Cells were cultured in 2D and in 3D tissue equivalent culture models to identify ARK-associated features.

For the first time, aniridic and normal corneal stromal cells were successfully cultured and expanded in vitro. Unexpectedly, aniridic and normal corneal stromal cells showed similar phenotype in 2D. They presented similar stem cell-like characteristics when cultured in 2% FBS containing media and showed the potential to differentiate into keratocyte-like cells when differentiated in serum-free medium. Aniridic corneal stromal cells showed reduced proliferation capacity when cultured both in 2D and in 3D. Moreover, when cultured inside the 3D tissue equivalent, aniridic corneal stromal cells showed different morphology and distinct spatial distribution when compared to normal cells. RNA sequencing data showed major differences between aniridic and normal corneal stromal cells transcriptomes (with 153 differentially expressed genes being identified), with significant differences in expression levels of genes involved in several biological processes, including matrix remodelling. Real-Time qPCR and ELISA confirmed lower expression and secretion levels of MMP1 and MMP2 (proteins involved on matrix remodelling) by aniridic corneal stromal cells when compared to their normal counterparts.

In conclusion, it was demonstrated for the first time that aniridic corneal stromal cells can be successfully cultured in 2D and 3D environments. The differences observed between aniridic and normal corneal stromal cell phenotypes in vitro, suggest that aniridic corneal stromal cells have impaired functionality in vivo and therefore might contribute to the development of ARK.

IMPACT STATEMENT

The cornea is our window to the world and its transparency is essential for vision. Corneal transparency can be compromised by several congenital disorders, including aniridia, that is caused by mutations in the PAX6 gene and affects different parts of the eye. The most severe consequence of aniridia is the so-called aniridia related keratopathy (ARK). ARK is a rare condition that starts to affect individuals from their teenage years and can progress, leading to corneal opacification and sight loss. For these patients, ARK represents a significant clinical problem, as effective treatments are not always available.

Superficial keratectomy, penetrating keratoplasty and experimental cultured LESC therapy may be used for the treatment of ARK corneas with high levels of opacification. However, there are different reasons why these treatments are not always available or successful: there is a shortage of corneal donors for transplantation worldwide and aniridic patients present variable responses and high levels of scarring following surgery. In order to develop new therapeutic approaches, it is of crucial importance to better understand the biological mechanisms driving ARK.

For the first time, we showed that human ARK-derived corneal stromal cells can be successfully cultured and expanded in vitro, and that they present phenotypical differences to normal cells. These results are the platform for future studies to carry on the in vitro characterisation of this condition. Moreover, the results presented in this thesis provide further understanding on the mechanisms behind ARK. We provided additional evidence that suggests a stromal involvement during the progression of ARK, which adds to the current body of research on this condition.

Our results not only have an impact in academia but will also be of utmost interest for the patient community. The hypothesis that ARK derived corneal stromal stem cells might contribute to the development of ARK, opens a new window for using them as a novel therapeutic target. In a tangible example, an ongoing clinical trial in our laboratory aims to replace both the epithelium and the stroma in ARK patients with healthy cells cultured in a tissue engineering approach. This is a great advancement, and the results here presented provide evidence supporting the success of this reconstructive therapy.

Lastly, it is noteworthy that aniridia is a rare disease with a motivated and engaging patient network. We recently presented this work to the patient community, who are eager on updates on the most recent advances in aniridia research. In the near future, this work might also be used in platforms such as Aniridia Network UK to raise awareness for this condition, and to bridge the knowledge gap between the research community and the public. 7

In conclusion, our research will impact the research community in the short term, by going a step forward to the understanding of the mechanisms behind ARK. In the long term, it will hopefully have a clinical use, by providing the ground work to develop new therapeutic strategies.

ACKNOWLEDGEMENTS

The completion of the thesis here presented has been possible thanks to the support from a wide range of people, to whom I will be forever grateful.

To all the patients (and their families) who kindly donated their tissue, without whom this project would not have been possible, thank you. To Fundació La Caixa and Moorfields Eye Charity, for their financial support.

I would like to start by expressing my sincere thanks to my supervisor, Professor Julie Daniels, for her guidance and encouragement during the course of this project and for her help in correcting this thesis. I am also thankful to Dr. Victoria Tovell for introducing me to the aniridia project and for teaching me numerous techniques. I would also like to thank my secondary supervisor, Prof. Mike Cheetham, for his guidance during the upgrade process. Finally, I would like to thank Dr. Cerys Evans for her invaluable help during the genotyping of the aniridic samples.

I am extremely grateful to my lab colleagues for their help, support and friendship. I will miss those Friday morning coffees with the rounds of: “Anything fun happening this weekend?”. A special shout out goes to my portuguese twin, Rita Pinho, for being the best cell factory buddy and the greatest friend. I would also like to especially thank Dr. Tiago Ramos for being always super energetic and helping me in any possible way.

To my friends from the 3rd floor, also known as Tercer piso guay, you have made everything better. We have shared great experiences not only at IoO but also in Hawaii, Canada and Barcelona. I really hope that the list keeps getting longer because I am going to miss you all.

I also wish to thank my friends for their support during these four years, with a special mention to (soon to be Dr.) Roser Cañigueral for sharing the writing process during these past few weeks in our home away from home. I would also like to especially mention Dr. Javier Uceda, whose smile was and still is the greatest inspiration. His memory is a constant reminder to take joy in every single aspect of our lives.

Finally, this would not have been possible without the support from my family, the beques group, and especially my parents Anna Martinez and Eduard Sanchez. They have taken the best care of me, making sure I go back home often and never failing to ask when I would be coming back before I have even left. Gràcies per fer-me costat durant aquest viatge i per preocupar-vos de mi i de les meves cèl·lules.

Thank you,

TABLE OF CONTENTS

Abstract ...... 5 Impact statement ...... 7 Acknowledgements ...... 9 Table of contents ...... 11 List of figures ...... 17 List of tables ...... 21 Abbreviations ...... 23 CHAPTER 1: Introduction ...... 25 1.1 The ocular surface: The cornea ...... 25 Overview ...... 25 Corneal epithelium ...... 26 Limbal stem cell niche ...... 27 Corneal endothelium ...... 28 Corneal stroma ...... 29 1.2 Aniridia ...... 40 Overview ...... 40 Genetics ...... 40 PAX6 transcription factor ...... 45 Aniridia related keratopathy ...... 49 Hypothesis ...... 57 Aims and objectives ...... 57 CHAPTER 2: General materials and methods ...... 59 2.1 Isolation and culture of primary human corneal stromal cells ...... 59 Isolation protocol for corneal stromal cells ...... 59 Culture and maintenance of corneal stromal cells ...... 61 Cryopreservation and storage of corneal stromal cells ...... 61 2.2 Gene expression ...... 61 RNA extraction from cells in 2D ...... 61 Assessment of RNA purity and integrity ...... 62 Reverse transcription ...... 63 Real-time qPCR ...... 63 2.3 Western blot ...... 65

Protein extraction ...... 65 Protein quantification ...... 66 Electrophoresis ...... 66 Immunoblotting ...... 67 Western blot quantification ...... 68 2.4 Immunocytochemistry ...... 69 CHAPTER 3: Normal corneal and limbal stromal cells have similar phenotypical characteristics ...... 71 3.1 Introduction ...... 71 3.2 Hypotheses and aims ...... 75 3.3 Materials and methods ...... 75 Assessment of KDM and CSSCM as media to culture and expand normal corneal stromal cells ...... 75 Characterisation of normal corneal and limbal stromal cells cultured in CSSCM ...... 77 Differentiation of normal corneal and limbal stromal cells into keratocyte- like cells in KDM ...... 80 3.4 Results ...... 82 Assessment of KDM and CSSCM as media to culture and expand normal corneal stromal cells ...... 82 Normal corneal and limbal stromal cells present similar phenotype when cultured in CSSCM ...... 85 Normal corneal and limbal stromal cells have a similar capacity to differentiate into keratocytes-like cells in KDM ...... 91 3.5 Discussion ...... 109 Normal corneal stromal cells were successfully expanded in CSSCM . 109 Normal corneal and limbal stromal cells have similar phenotypes when cultured in CSSCM ...... 110 Normal corneal and limbal stromal cells share similar ability to differentiate into keratocyte-like cells ...... 111 Are there stem cells in the central cornea? ...... 114 CHAPTER 4: Aniridic and normal corneal stromal cells have similar phenotypical characteristics when cultured in 2D ...... 117 4.1 Introduction ...... 117 4.2 Hypotheses and aims ...... 119 4.3 Materials and methods ...... 119 Patient demographics ...... 119 Genotyping of PAX6 gene in aniridic corneal stromal cells ...... 121

Characterisation of aniridic and normal corneal stromal cells in CSSCM ...... 124 Differentiation of aniridic and normal corneal stromal cells into keratocyte- like cells in KDM ...... 126 Statistical analyses ...... 128 4.4 Results ...... 128 Genetic variants found in the aniridic samples ...... 128 Aniridic and normal corneal stromal cells present similar phenotype when cultured in CSSCM ...... 132 Aniridic and normal corneal stromal cells have similar capacity to differentiate into keratocytes-like cells in KDM ...... 141 4.5 Discussion ...... 159 Implications of obtaining ARK samples for corneal stromal cell isolation ...... 159 Pathogenicity of the PAX6 variants found in the aniridic samples ...... 160 Aniridic corneal stromal cells have lower proliferation capacity, but otherwise present the same phenotype as normal corneal stromal cells when cultured in CSSCM ...... 160 Aniridic and normal corneal stromal cells showed similar capacity to differentiate into keratocyte-like cells ...... 164 CHAPTER 5: Aniridic and normal corneal stromal cells present differences in proliferation, distribution, morphology and gene expression when cultured in RAFT TE ...... 167 5.1 Introduction ...... 167 5.2 Hypotheses and aims ...... 169 5.3 Materials and Methods ...... 170 Real Architecture for 3D Tissue ...... 170 Experiment design – Donor demographics ...... 170 Metabolic activity – PrestoBlue ® ...... 171 Real-Time qPCR ...... 171 Cell characterisation ...... 172 RNA sequencing ...... 174 Statistical analyses ...... 176 5.4 Results ...... 176 Aniridic corneal stromal cells have lower proliferation capacity than normal corneal stromal cells ...... 176 Aniridic and normal corneal stromal cells do not show differences in apoptosis ...... 178

Aniridic corneal stromal cells have a different spatial distribution than normal corneal stromal cells in RAFT TE ...... 180 Normal corneal stromal cells present higher alignment than aniridic corneal stromal cells ...... 183 Aniridic corneal stromal cells present a higher degree of circularity than normal corneal stromal cells in RAFT TE ...... 184 Aniridic corneal stromal cells show higher expression levels of tested CSSC markers than normal corneal stromal cells, and similar gene expression profiles for keratocyte and myofibroblast markers compared to normal corneal stromal cells ...... 185 Aniridic corneal stromal cells show differences in total gene expression profile compared to normal corneal stromal cells ...... 186 5.5 Discussion ...... 194 Aniridic and normal corneal stromal cells were successfully cultured inside the RAFT TE ...... 194 Differences were found in aniridic and normal corneal stromal cell phenotype when cultured in RAFT TE ...... 194 CHAPTER 6: Aniridic and normal corneal stromal cells express and secrete different levels of matrix metalloproteinases when cultured in RAFT TE ...... 205 6.1 Introduction ...... 205 6.2 Hypotheses and aims ...... 206 6.3 Materials and methods ...... 207 Experimental design ...... 207 RAFT TE Transparency ...... 207 Scanning electron microscopy ...... 207 Quantification of total protein – Bradford assay ...... 207 Quantification of matrix metalloproteinases secretion ...... 208 Inhibition of matrix metalloproteinases activity with Ilomastat ...... 210 Statistical analyses ...... 211 6.4 Results ...... 212 Extracellular matrix appearance is similar in RAFT TE populated with aniridic or normal corneal stromal cells ...... 212 Aniridic corneal stromal cells express and secrete lower levels of matrix metalloproteinases compared to normal corneal stromal cells ...... 215 Ilomastat successfully blocks the proteolytic activity of matrix metalloproteinases secreted by normal corneal stromal cells cultured in CSSCM but has no effect on cell alignment ...... 224 6.5 Discussion ...... 229

RAFT TE populated with aniridic and normal corneal stromal cells did not show differences in ECM transparency and appearance ...... 229 MMP1, MMP2, MMP3 and MMP14 govern the ECM remodelling process ...... 230 Aniridic corneal stromal cells express and secrete lower levels of MMP1 and MMP2 than normal corneal stromal cells ...... 233 Ilomastat successfully blocks matrix metalloproteinase activity but does not show an effect on cell alignment ...... 234 Ilomastat treatment resulted in a higher number of cell clusters inside the RAFT TE ...... 235 CHAPTER 7: General discussion ...... 239 7.1 Impact of expanding and culturing normal corneal stromal cells in RAFT TE ...... 239 7.2 Impact of expanding and culturing aniridic corneal stromal cells in RAFT TE ...... 240 7.3 Involvement of matrix metalloproteinases on corneal scarring ...... 242 7.4 Potential mechanisms by which aniridic and normal corneal stromal stem cells show phenotypical differences in vitro ...... 243 7.5 Do aniridic and normal corneal stromal cells support differently the epithelium? ...... 244 7.6 Genotype-phenotype correlation between the aniridic samples ...... 244 7.7 Limitations of this study ...... 245 7.8 Future work ...... 246 7.9 Concluding remarks ...... 247 Appendix ...... 249 References ...... 265

LIST OF FIGURES

Figure 1.1. The ocular surface ...... 25 Figure 1.2. Schematic representation of the structural layers of the cornea ...... 26 Figure 1.3. Schematic representation of the structure of the cornea and the limbus ... 28 Figure 1.4. Schematic representation of corneal epithelium and stromal wound healing ...... 39 Figure 1.5. Schematic representation of the paired box 6 gene (PAX6) and the Paired box protein Pax-6 (PAX6) ...... 41 Figure 1.6. Paired box protein Pax-6 structure...... 46 Figure 1.7. Aniridic cornea presenting ARK ...... 49 Figure 2.1. Corneal stromal cell isolation from the central cornea ...... 60 Figure 2.2. Schematic representation of the sandwich blot...... 67 Figure 3.1. Isolation of normal limbal stromal cells from the limbal rims ...... 78 Figure 3.2. Representative images of cells isolated from the central cornea and plated in KDM or CSSCM ...... 83 Figure 3.3. Growth curve of corneal stromal cells cultured in CSSCM...... 84 Figure 3.4. Representative images of NC and NL at P0 ...... 85 Figure 3.5. Representative images of NC and NL cultured in CSSCM ...... 86 Figure 3.6. Proliferation of NC and NL cultured in CSSCM over a period of 7 days…. 87 Figure 3.7. Relative gene expression levels of (A) putative CSSC markers (PAX6, CD90, CD73), (B) keratocyte markers (KERA, LUMI, ALDH1A1), and (C) proliferation (KI67) and apoptosis markers (TP53) in MSC, NC and NL...... 89 Figure 3.8. Protein production levels of putative CSSC markers (CD90 and CD73) by NC and NL ...... 90 Figure 3.9. Representative images of NC and NL cultured under different culture conditions...... 92 Figure 3.10. Relative gene expression levels of putative CSSC markers (CD90 and CD73) by NC and NL under different culture conditions ...... 94 Figure 3.11. Production levels of putative CSSC markers (CD90 and CD73) by NC and NL under different culture conditions ...... 96 Figure 3.12. Relative gene expression levels of keratocyte markers (LUM, KERA and ALDH1A1) in NC and NL under different culture conditions...... 98 Figure 3.13. Production levels of keratocyte markers (LUM and KERA) by NC and NL under different conditions ...... 100 Figure 3.14. Relative gene expression levels of the fibrosis marker αSMA (ACTA2) in NC and NL under different culture conditions...... 101 17

Figure 3.15. Production levels of the fibrosis marker αSMA by NC and NL under different culture conditions...... 103 Figure 3.16. Production and assembly of αSMA by NC and NL under different culture conditions...... 104 Figure 3.17. Relative gene expression levels of KI67 in NC and NL under different culture conditions ...... 105 Figure 3.18. Relative gene expression levels of apoptosis markers (TP53, BAX, and BCL2) in NC and NL under different culture conditions ...... 107 Figure 4.1. Representation of the transwell migration assay ...... 125 Figure 4.2. Representation of the last part of Exon 13 sequence in Aniridia 1 sample compared to a control sample ...... 129 Figure 4.3. Representation of a portion of Exon 8 sequence in Aniridia 2 compared to a control sample ...... 130 Figure 4.4. Representation of a portion of the Exon 7 sequence in Aniridia sample 3 compared to a control sample ...... 131 Figure 4.5. Representative images of AC and NCcultures at P0...... 132 Figure 4.6. Representative images of AC and NC cultured in CSSCM ...... 133 Figure 4.7. Proliferation of AC and NC in CSSCM over 7 days ...... 135 Figure 4.8. Relative gene expression levels of CSSC markers (PAX6, CD90, CD73), keratocyte markers (KERA, LUM, ALDH1A1), and proliferation and apoptosis markers (KI67, TP53, respectively) in MSC, AC and NC...... 137 Figure 4.9. Protein production levels of putative CSSC markers (CD90 and CD73) in AC and NC ...... 138 Figure 4.10. Graphic representation of the number of cells that migrated per well in each condition (aniridia and normal) ...... 139 Figure 4.11. Contraction of gels populated with NC, AC and without cells over a period of 7 days ...... 140 Figure 4.12. Representative images of AC and NC cultured under different conditions ...... 142 Figure 4.13. Relative gene expression levels of putative CSSC markers (CD90 and CD73) by AC and NC under different culture conditions...... 144 Figure 4.14. Production levels of putative CSSC markers (CD90 and CD73) by AC and NC under different culture conditions ...... 146 Figure 4.15. Relative gene expression levels of keratocyte markers (LUM, KERA and ALDH1A1) in AC and NL under different culture conditions...... 148 Figure 4.16. Production levels of keratocyte markers (LUM and KERA) by AC and NC under different culture conditions ...... 150 18

Figure 4.17. Relative gene expression levels of the fibrosis marker αSMA (ACTA2) in AC and NC under different culture conditions ...... 151 Figure 4.18. Production levels of the fibrosis marker αSMA by AC and NC under different culture conditions ...... 152 Figure 4.19. Production and assembly of the fibrosis marker αSMA by AC and NC . 154 Figure 4.20. Relative gene expression levels of KI67 (gene: MKI67) in AC and NC under different culture conditions...... 155 Figure 4.21. Relative gene expression levels of apoptosis markers (TP53, BAX and BCL2) in AC and NC under different culture conditions...... 157 Figure 5.1. Schematic representation of the RAFT TE manufacturing process ...... 170 Figure 5.2. Schematic representation of the protocol for Live/Dead analysis ...... 172 Figure 5.3. Proliferation capacity of AC and NC when cultured in RAFT TE for 3 weeks in CSSCM ...... 177 Figure 5.4. Relative gene expression levels of apoptosis markers (TP53, BAX and BCL2) by AC compared to the control NC...... 178 Figure 5.5. Representative images of Live/Dead analysis of AC and NC cultured in RAFT TE for 3 weeks in CSSCM...... 179 Figure 5.6. Distribution of AC and NC in RAFT TE cultured for 3 weeks in CSSCM . 181 Figure 5.7. Distribution of AC and NC on the RAFT TE surface at 3 weeks culture in CSSCM...... 182 Figure 5.8. Alignment of AC and NC between themselves when cultured for 3 weeks inside RAFT TE in CSSCM...... 183 Figure 5.9. Circularity of AC and NC cultured in RAFT TE for 3 weeks in CSSCM. .. 184 Figure 5.10. Relative gene expression levels of putative CSSC markers (CD90 and CD73), keratocyte markers (KERA, LUM, ALDH1A1) and the myofibroblast marker (αSMA) by AC and NC cultured in RAFT TE for 3 weeks in CSSCM...... 185 Figure 5.11. RNA sequencing results comparing AC against NC...... 187 Figure 5.12. RNA sequencing results comparing AC1 against NC...... 189 Figure 5.13. RNA sequencing results comparing AC2 and AC3 against NC...... 190 Figure 5.14. Gene ontology analysis of significant differentially expressed genes (DEGs) between AC and NC...... 191 Figure 5.15. Gene ontology analysis of significant differentially expressed genes (DEGs) between AC1 and NC ...... 192 Figure 5.16. Gene ontology analysis of significant differentially expressed genes (DEGs) between AC2/AC3 and NC...... 193

Figure 6.1. Transparency of the RAFT TE populated by AC, NC or no cells cultured for 3 weeks in CSSCM ...... 212 Figure 6.2. Representative images of collagen fibrils from the RAFT TE populated with aniridic and normal cells cultured for 3 weeks in CSSCM...... 214 Figure 6.3. Heatmaps comparing the expression levels of Matrix metalloproteinases by AC and NC...... 216 Figure 6.4. Heatmaps comparing the expression levels of tissue inhibitors of matrix metalloproteinases by AC and NC...... 217 Figure 6.5. Relative gene expression levels of MMP1 and MMP2 by AC, normalised to NC ...... 218 Figure 6.6. Concentration of pro-MMP1 released by AC and NC in RAFT TE for 3 weeks in CSSCM ...... 219 Figure 6.7. Concentration of pro-MMP1 released by AC1 and NC1, AC2 and NC2 and AC3 and NC3 cultured in the RAFT TE for three weeks in CSSCM ...... 220 Figure 6.8. Concentration of total MMP2 released by AC and NC cultured in RAFT TE for 3 weeks in CSSCM ...... 221 Figure 6.9. Secretion of total MMP2 by AC and NC cultured in the RAFT TE for 3 weeks in CSSCM ...... 223 Figure 6.10. Effect of different concentrations of Ilomastat on the contraction of non compressed hyper hydrated type I collagen gels populated by NC...... 225 Figure 6.11. Images of NC cultured in RAFT TE and supplemented with Ilomastat 0µM, 10µM and negative control ...... 226 Figure 6.12. Cell viability of NC in RAFT TE cultured with different concentrations of Ilomastat...... 227 Figure 6.13. The effect of Ilomastat on the number of cell clusters in RAFT TE...... 228 Figure A.8.1. Purity and integrity of RNA isolated from NL, NC and AC ...... 249 Figure A.8.2. Real-Time qPCR study to select an appropriate endogenous control . 250 Figure A.8.3. Negative controls for immunocytochemistry ...... 251 Figure A.8.4. Positive controls for immunocytochemistry ...... 252 Figure A.8.5. Quality and integrity of the RNA obtained from AC and NC populated RAFT TE ...... 255 Figure A.8.6. Transition process from long to short absorbers...... 256

LIST OF TABLES

Table 1.1. Classification and definition of the three main types of mutations in PAX6 . 43 Table 2.1. List of Taqman ™ assays used for Real Time qPCR ...... 65 Table 2.2 List of primary antibodies used for Western blot...... 68 Table 3.1. Donor information...... 75 Table 3.2. Keratocyte differentiation medium (KDM) composition ...... 76 Table 3.3 - Donor information. Age, gender, and storage time are shown ...... 77 Table 3.4. Culture conditions for NC and NL in CSSCM and KDM...... 81 Table 4.1. Donor information...... 120 Table 4.2. Information about clinical history of surgeries of aniridic patients ...... 121 Table 4.3. Primers used for the genotyping of PAX6 - RT-PCR ...... 123 Table 4.4. Culture conditions for AC and NC in CSSCM and KDM ...... 127 Table 4.5. Overview of the three mutations identified in the four aniridic samples used for this study ...... 131 Table 6.1. Time points at which different experiments were performed on the RAFT TE...... 207 Table 6.2. Dilution factors used for ELISA depending on MMP and time point...... 209 Table 6.3. Donor information...... 210

ABBREVIATIONS

A DMSO dimethyl sulfoxide

AC aniridic corneal stromal cells DNA deoxyribonucleic acid

ACTB β-actin DAPI 4',6-Diamidine-2'- phenylindole dihydrochloride ALDH1A1 aldehyde dehydrogenase 1 family member DPBS Dulbecco’s phosphate- A1 buffered saline

αSMA α-smooth muscle actin E

A.U. arbitrary units EGF epithelial growth factor

ARVO Association for Research in EC corneal epithelial stem cells Vision and Ophthalmology ECM extracellular matrix ARK aniridia-related keratopathy ELISA enzyme-linked B immunosorbent assay

BCL2 B-cell lymphoma 2 F

BCA bicinchoninic acid FBS foetal bovine serum

BSA bovine serum albumin FGF basic fibroblast growth factor

C G

CSSCM corneal stromal stem cell GAPDH glyceraldehyde 3- medium phosphate dehydrogenase

COL collagen GO gene ontology

CD cluster of differentiation H

Ct cycle threshold I

CTRL control IgG Immunoglobulin G CO carbon dioxide 2 J D K DEG differentially expressed genes KERA keratocan DMEM Dulbecco's Modified Eagle KDM keratocyte differentiation Medium medium

L Pax6 paired box protein Pax-6 (mouse) LUM lumican PDGF platelet derived growth factor M P/S penicillin-streptomycin MMP matrix metalloproteinases PCR polymerase chain reaction MSC mesenchymal stem cells R mRNA messenger ribonucleic acid RAFT real Architecture for 3D minimum essential medium (MEM) tissues N RNA ribonucleic acid NC normal corneal stromal cells RPL37A ribosomal Protein L37A NL normal limbal stromal cells S NCKiF normal corneal stromal SD spontaneous differentiation keratocyte-like cells SEM scanning electron microscope NCSD normal corneal stroma spontaneously differentiated cells T

NLKiF normal limbal stromal TC tissue culture keratocyte-like cells TE tissue equivalents NLSD normal limbal stroma TP53 tumour protein p53 spontaneously differentiated cells 2D two-dimensional O 3D three-dimensional OD optical density TBS tris-buffered saline P TPM transcripts per kilobase million PAX6 paired box 6 gene (human) TIMP tissue inhibitors of MMPs Pax6 paired box 6 gene (mouse) U PAX6 paired box protein Pax-6 UV ultraviolet (human)

CHAPTER 1: INTRODUCTION

The cornea is the clear central part of the eye, providing our window to the world. Its unique transparent properties are essential for vision. Currently, 285 million people are visually impaired worldwide, with corneal opacity being a major cause of blindness, as stated by the World Health Organisation. A small portion of the population affected by corneal opacities (1:40000 – 1:100000 of total population) suffer from a genetic disorder called aniridia [1, 2], which not only affects corneal transparency but can also involve iris hypoplasia, cataracts, and glaucoma, amongst other eye conditions [3]. Aniridia related keratopathy (ARK) is a progressive manifestation that threatens the integrity of the cornea in aniridic patients and the mechanisms involved are not yet fully understood. It is therefore of crucial importance to firstly understand the biology behind of the healthy and aniridic cornea to then work on finding possible treatments for ARK. The project presented herein aims to model the aniridic and healthy corneal stroma in vitro, recapitulating their biological microenvironment, and to identify any possible differences between the cells that populate those stromas.

1.1 The ocular surface: The cornea

The human ocular surface is composed of the central cornea, the peripheral conjunctiva, and the corneoscleral limbus (Figure 1.1). The limbus function is twofold; 1) it acts as a physical barrier, preventing the conjunctival epithelium from encroaching over the cornea, and 2) harbours stem cells for corneal regeneration [4-6].

Figure 1.1. The ocular surface. The cornea is the transparent surface on the front of the eye, surrounded by the limbus and the neighbouring sclera.

Overview

The cornea is a specialized transparent avascular structure that provides two-thirds of the eye’s refractive power. Its central part is approximately 500µm thick, and it is composed of five layers, the epithelium (outermost layer), the Bowman’s layer, the 25 stroma, the Descemet’s membrane and the endothelium (innermost layer) (Figure 1.2) [7]. The epithelium, the stroma, and the endothelium are cellular layers, each populated by a different type of cells, while the others are structural layers mainly made of collagen. For the purpose of this work, the stroma will be described in detail in the following sections.

Corneal transparency is partly granted by the avascular nature of the cornea. As it contains no blood vessels it is nourished by the diffusion of nutrients from the overlying aqueous humour that fills the anterior chamber and by the limbal rich vasculature [8].

Figure 1.2. The structural layers of the cornea. The cornea is composed of three cellular layers, namely epithelium, stroma, and endothelium, separated by the Bowman’s layer and the Descemet’s membrane, respectively. Haematoxylin and Eosin image of a cross section of the human cornea.

Corneal epithelium

The corneal epithelium is the outermost layer of the cornea. Since it is in contact with the external environment, the corneal epithelium serves as a first barrier against injury, protecting the eye from external insults. The ability of the epithelium to eliminate external 26

pathogens is provided by the epithelial cells [9] and the high density of tight junctions that connect them [10]. In the central part of the cornea, the epithelium is composed of six or seven layers of non-keratinized, stratified and squamous epithelial cells with a total thickness of approximately 52μm [11].

A directional process of cell differentiation takes place in the epithelium, with the cells in the lower layers being the least differentiated and the cells on the surface being in the latest stages of differentiation. The innermost layer of the epithelium consists of columnar basal cells that gradually proliferate and differentiate into wing cells, which migrate to the layers above. These continue to differentiate and migrate from the suprabasal layers to the surface of the epithelium. These cells in the last step of differentiation are called squamous cells and are in direct contact with the external environment [12]. Squamous cells are constantly shed into the tear pool, providing a clear necessity for the epithelium to regenerate its cells to maintain the homoeostasis and functionality of the entire layer.

Limbal stem cell niche

The capacity of the epithelium to self-renew is provided by a stem cell population situated at the limbus [13] (Figure 1.3). The limbus is the narrow border between the cornea and the conjunctiva, it starts at the termination of Bowman’s layer and is approximately 1.5mm wide. The limbus is composed of an epithelial, stromal, and endothelial layer, and Descemet’s membrane.

The corneal epithelial progenitor cells, commonly referred to as limbal epithelial stem cells (LESC) are situated in a very specialised niche within the limbal crypts of the Palisades of Vogt [14] (Figure 1.3). The limbal niche provides the LESC with optimal conditions to maintain their capacity for differentiation and self-renewal. LESC are situated near another reservoir of stem cells from the stroma [14], melanocytes [15] and blood vessels. Following asymmetric cell division, LESC daughter transient amplifying cells terminally differentiate into corneal epithelial cells (Figure 1.3) [16].

Figure 1.3. Schematic representation of the structure of the cornea and the limbus (not to scale). The limbus is populated by the limbal epithelial stem cells (LESC) and the corneal stromal stem cells (CSSC). The black arrow indicates the direction of stem cell differentiation from the limbus towards the central cornea. The central corneal epithelium consists of different layers of epithelial cells derived from the LESC population.

Corneal endothelium

The endothelium consists of a single layer of hexagonal shaped endothelial cells that do not proliferate in healthy conditions and rest on the Descemet’s membrane (Figure 1.2) [17]. Their main function is to secrete the collagen that constitutes Descemet’s membrane and to control the passage of fluid and nutrients from the aqueous humour to maintain the appropriate stromal hydration and homoeostasis preventing corneal oedema.

Corneal stroma

The stroma is the middle cellular layer of the cornea, situated between the Bowman’s layer and the Descemet’s membrane. It is an avascular structure with a very specialized organization (Figure 1.2). With 500μm, it is the thickest layer of the cornea, comprising nearly 90% of it [18]. It is populated by the keratocytes that reside within highly organised collagen type I lamellae. The following sections will focus on stroma biogenesis and mantaince, its transparent properties and wound healing processes.

1.1.5.1 Stroma during development – embryogenesis

The human eye starts to develop at the end of the 3rd week of gestation as a pair of two small optic grooves on each side of the forebrain. The optic grooves will eventually close, becoming the optic vesicles which thereafter invaginate to form a double-layered structure called the optic cup. The optic vesicles will contact the surface ectoderm triggering the necessary changes for further eye development, in which PAX6 plays a central role on initiating and controlling. Further differentiation and mechanical rearrangement of cells gives rise to the fully developed eye [19, 20].

Different germ layers form different areas of the eye: the neural tube ectoderm gives rise to the optic nerve, retina, iris, ciliary body epithelia, and some of the vitreous humour. Surface ectoderm gives rise to the lens, the corneal and conjunctival epithelium, the eyelids, and the lacrimal apparatus. The remaining ocular structures, including the corneal stroma, are formed from the mesenchyme.

Corneal stroma development and collagen fibrillogenesis was firstly studied using the embryonic chick as a developmental model [21, 22]. Later studies on mouse, rabbit, and, primate cornea showed that the mammalian cornea presents several differences but maintained the same general phases during its development [23, 24]. Stromal development consists of two stages. Firstly, the primary stromal collagenous matrix is secreted by ectodermal cells, which will later differentiate into corneal epithelium. The secreted matrix constitutes the foundations for the second stage, in which mesenchymal cells migrate from the periphery and infiltrate the stroma. A wave of mesenchymal cells derived from neural crest cells infiltrate the primary stroma and later become keratocytes. These cells show a rounded fibroblastic morphology during the migration phase and later change acquiring a flattened shape typical of keratocytes in the mature cornea. They will then continue to synthesise the bulk of the stromal matrix in a very organised manner [22].

The morphogenetics of the stroma is divided into three main non-aqueous constituents: cells, collagens, and proteoglycans.

1.1.5.1.1 Stromal cells

Studied for the first time in the chick cornea, the infiltration of neural crest cells in the stroma was shown to come in two waves [21, 22]. A first wave of cells migrate to the posterior stroma and acquire a flattened shape, which will condense to form the endothelium. A second wave of cells known as keratoblasts infiltrate the stroma, staying between the epithelium and the endothelium. These continue to proliferate and later differentiate into keratocytes. In rodents, the infiltration of neural crest-derived cells into the stroma exhibited one single influx, which differentiate in both endothelial cells and keratocytes [23]. Keratocytes present slower rates of proliferation and start to synthesise high levels of collagen type I, V and VI, and proteoglycans, shaping the stromal ECM [8, 23, 25]. ECM deposition by keratocytes creates space and generates cell-separating lamellae [26]. Keratocytes maintain strong associations with the assembled collagen bundles through a range of elongated cytoplasmic protrusions (~30µm). These are essential and dictate the orientation of the process of collagen assembly into lamellae [27]. Cells and collagen bundles distribute ortogonally across the cornea, presenting approximately 90 degree angles between each other [27]. They also maintain gap- junction mediated cell-cell connections with neighbouring keratocytes, creating a network across the corneal stroma [28].

1.1.5.1.2 Collagen

Collagen is synthesised, secreted and organised into fibrils in a tissue-specific manner. This proccess, also known as fibrillogenesis, is infuenced by a variety of factors, such as the types of collagen present and the interactions between these and proteoglycans, which will be explained in detail in the next section (1.1.5.1.3) [29]. Collagen fibrils in the cornea are narrow, of uniform diameter and are distributed in a specific range of distances from their adjacent fibrils.

One of the first studies to target the speciallised collagen fibrillogenesis in the cornea was done by Birk et al. on the chicken embryo [29]. They found that corneal collagen fibrils are heterotypic, i.e. composed of different types of collagen. Its main structural component is type I collagen, but unlike many other tissues composed of type I collagen, the cornea is enriched in collagen type V. These combination of different types of collagens, with a high ratio of collagen V, allows the collagen fibrils to be specially narrow (~25nm) and uniform. The regulation of the precise narrow collagen fibril diameter in the

cornea is modulated by the interaction of type V collagen with type I collagen [29]. The same results were found in the mammalian cornea, when heterozygous mice mutant for the Col5α1 gene showed an increased diameter of the collagen fibrils and a decrease in the amount of fibrillogenesis [30, 31]. The monotony of narrow fibril size is essential to achieve optical transparency [32].

Collagen fibrills are arranged in parallel and are surrounded by proteoglycans, which confer unique water-holding properties. Finally, collagen fibrils are organised in lamellae, which are arranged orthogonally, in a very organised structure that guarantees transparency by allowing the free passage of light [33].

1.1.5.1.3 Proteoglycans

Proteoglycans belong to a family of ECM proteins that play essential roles on corneal stroma development and maintenance [34]. Proteoglycan structure consists of a core protein with a variable number of sulfated glycosaminoglican side chains. The core protein interacts with the collagen fibrils, and the glycosaminoglycan side chains occupy the space between collagen fibrils. Thus, they regulate the hydrophylicity and the swelling properties of the cornea, the size of collagen fibrils and their spatial distribution [35, 36].

Extensive research on the proteoglycan composition of the cornea during development has been performed using the chick cornea as a developmental model. However, less extensive research has also focused on the mammalian cornea [34]. The secondary stroma formed during development already contains different glycosaminoglycans such as heparan sulfate, keratan sulfate and chondroitin sulphate, some of which specifically interact with regions of the collagen fibrils.

Four proteoglycans are present in the corneal stroma: lumican, keratocan, mimecan and decorin [37]. The first three present keratan sulfate glycosaminoglycan side chains, whereas the latter presents dermatan sulfate. From these, keratocan and lumican are are members of the small leucine-rich proteoglycan family, which present high number of Leucine aminoacid repetitions and constitute, after collagen, the most abundant biological material in the stroma [38]. They are the earliest markers of keratocyte differentiation from neural crest cells, as they start being expressed around 12 hours after neural crest cell invasion [25].

Lumican core protein is extensively expressed by mesenchymal cells in many connective tissues during embryonic development and adulthood. During mouse development, lumican starts to be expressed at around day 7 post-coitus in different tissues, including

31 in the intraocular lens, cartilage, heart, lungs, and cornea [39, 40]. However, further modifications of the lumican core protein with keratan sulphate were only observed in the murine cornea at around day 10 postnatally, after eye opening [41]. Lumican with kertan sulphate modifications is an essential component to maintain corneal transparency [42]. Lumican null mice have shown stromal structural defects, including collagen fibrils with increased diameter, wider fibril agreggates, unorganised lamellae, and reduced stromal thickness compared to their wild-type counterparts [42]. Taken together, these observations suggest that lumican is essential to maintain the normal corneal transparency and stromal structure.

Keratocan is expressed in the cornea and eyelids forming tissues during mouse embryonic development, however is more selectively expressed in the corneal tissue of the adult mouse [43]. In the developing mouse, keratocan is expressed by the periocular mesenchymal cells that migrate towards the central cornea and its expression is then gradually restricted to the keratocytes [39]. Mutations of the KERA gene in humans are associated with the cornea plana phenotype, which results in lower vision acuity due to the flattening of the cornea [44]. In mice, mutations of the Kera gene also result in thinner stroma with wider collagen fibrils, and less organised collagen lattice models than in the wild-type counterparts [45]. Despite having similar functions on the collagen fibrils arrangemnts, lumican and keratocan do not alleviate the phenotypes of each others null mice, suggesting that they bind to different regions of the collagen fibrils [45]. These results show that keratocan is also an essential component of the corneal stroma helping maintaining corneal hydration and appropriate collagen fibrillogenesis to guarantee its transparency.

As previously mentioned, corneal proteoglycans also regulate collagen fibril diameter [22, 36]. It was demonstrated by Birk and Lande that collagen fibrillogenesis in vitro is retarded by the presence of corneal proteoglycans, resulting in significantly thinner collagen fibrils, when compared to those in the absence of proteoglycans [46]. Interestingly, when comparing the effect of proteoglycans from the cornea and from the neighbouring fibrillar tissue - sclera, the first were three times more effective at slowing the rate of fibrillogenesis when compared to the latter [46]. Rada et al. identified the core proteins of decorin and lumican as the agents slowing down fibrillogenesis [36]. These findings were confirmed in mutant mice defficient for lumican and decorin that presented opacification of the cornea [42, 47].

Sumarizing, lumican and keratocan are the earliest markers of keratocyte differentiation and closely regulate the diameter of collagen fibrils on corneal stroma.

1.1.5.2 Theoretical basis of corneal transparency

The cornea provides two-thirds of the eye’s refractive power and its transparency is a matter of great interest, since it is essential to allow the free passage of light to the back of the eye. As previously mentioned, different factors regulate transparency and multiple studies have tried to tackle and identify them [33, 48, 49]. During the nineteenth century, a consensus was raised that the cornea needed to be optically homogeneous, with all its components having a uniform refractive index, to be transparent. However, one century ago, it was postulated that the transparency of the cornea is mainly regulated from the arrangement of the constituents of the stroma, since it accounts for 90% of its thickness [50]. In 1957, a novel theory was developed by Maurice in London, which proved previous theories of optical uniformity wrong [48]. In a thorough paper, Maurice firstly described the optical structure of the stroma, including its geometric form, dimensions, and refractive index. The arrangement proposed consisted of collagen fibrils of uniform diameter distributed on a perfect hexagonal lattice. Secondly, how the transmission of light across them and their interactions is the basis of its transparency. When the light hits the ordered stromal collagen fibrils, it produces secondary light rays interfering in a way that eliminates light scattering in all directions except forwards [48]. Later, it was discovered that the fibrils do not need to be positioned in perfect hexagonal lattice for these interactions to happen. It is sufficient for all the fibrils to have the same diameter and be positioned in a constrained and periodic distance between adjacent fibres for these interactions to happen [33, 50].

To explain corneal transparency, focus has historically been put on the collagen fibrils arrangement and the presence of cells within the stroma has been broadly ignored. This is usually justified as keratocytes are sparsely distributed and are thin and flat in the location of the passing light [48]. However, Jester and colleagues more recently postulated that keratocytes contain cytoplasmic molecules that act like crystallins. They have the ability to absorb UV light but also to change the refractive index of the cell cytoplasm to eliminate light scattering arising from the cells [49]. The crystalline candidates in the mammalian cornea are the aldehyde dehydrogenase 1 (ALDH1), the aldehyde dehydrogenase 3 (ALDH3), and the transketolase (TKT) [49].

Finally, as previously mentioned, it has been shown that proteoglycans also indirectly affect corneal transparency by regulating collagen fibril [42]. For instance, lumican deficient mice presented corneas with abnormally thick and disorganised collagen fibrils and showed a threefold increase in backscattered light [42].

1.1.5.3 Stroma during adulthood

Unlike the self-renewing epithelium, the maintenance of the stroma during adulthood is not based on a cycle of cell death and renewal. The stroma is populated by keratocytes, which reside between the collagen lamellae structures. They have an elongated and dendritic morphology, forming a network of connections between them. Keratocytes in healthy conditions are quiescent and only proliferate in the event of an injury or abnormal conditions [51]. In 1999 it was postulated that there is no cell turnover in the stroma, that keratocytes are quiescent, form a network by contact against each other, and are resistant to antigens and to stimuli [8]. Zieske et al. showed that keratocytes are uniformly distributed throughout the stroma, and there is less than 1 proliferating cell per mm3 [52].

Keratocytes obtain their nutrients from the aqueous humour, which are diffused through the endothelium. Signalling interactions between epithelium and stroma are also present during homeostasis. For instance, it has been hypothesise that low levels of IL-1 expression by epithelial cells cause negative apoptotic and chemotactic effects on the adjacent keratocytes [53].

Similar to the LESC population, a population of progenitor cells residing in the human limbal stroma has been recently described [54]. These cells are situated underneath the Palisades of Vogt in close proximity with LESC (Figure 1.3). Known as corneal stromal stem cells (CSSC) they were firstly identified and described in the human corneal stroma in 2005 by Du and colleagues [54]. Using human corneas from deceased donors they isolated a side population of cells with clonal growth ability that express ABCG2, a typical stem cell marker [54]. CSSC are broadly described as adult mesenchymal stem cells (MSC), that express the characteristic MSC markers (CD90 and CD73) when cultured in vitro [55]. CD90 and CD73 belong to a family of cell surface glycoproteins that are expressed in adult stem cells. CD90 has also been found to be expressed in fibroblasts, brain cells, and activated endothelial cells, where it plays a role in inflammation, wound healing, cell-cell and cell-matrix adhesion, and in growth and differentiation of stem cells [56]. CSSC also express PAX6, a transcription factor involved in the regulation of eye development (addressed in detail in section 1.2.3.2) [57]. CSSC are thought to be the progenitors of keratocytes, since when cultured in vitro in low mitogen media they upregulated the expression of keratan sulphate and ALDH3A1, and obtained a dendritic morphology, all characteristic features of keratocytes [57].

CSSC are thought to be derived from the neural crest cells during embryonic formation of the cornea [58]. During development they do not migrate to the central cornea but instead remain in the limbus in close proximity to the LESC, suggesting that there are

interactions between both layers of the cornea [59]. Together these results suggest that there is a population of stem cells that reside in the limbal stroma which may be the progenitors of the mature keratocytes.

1.1.5.4 Stroma during wound healing

Briefly, following injury, the keratocytes in the wounded area undergo apoptosis (programmed cell death). This is followed by an influx of keratocytes that have been activated and differentiated into myofibroblasts to the wounded region. These differentiated cells contract and secrete a fibrous matrix that differs from the healthy stromal extracellular matrix. It is not organised and results in scar tissue [60]. Corneal stroma scarring is difficult to solve without transplantation, and it is not only fruit of an event of injury but can also be caused by chronic conditions, such as aniridia and Lattice dystrophy. Aniridic patients easily develop high levels of stromal scarring in response to the recurrent corneal surgeries [61].

The process of corneal wound healing involves constant interactions between the epithelium and the stroma (extensively reviewed in [62]). Even though this project focuses on the corneal stroma, it is impossible to dissociate the events that occur in this layer from what occurs in the epithelium.

1.1.5.4.1 Keratocyte apoptosis

Keratocyte apoptosis is the first event after corneal injury, whether this affects the stroma or only the epithelium [62]. Loss of keratocytes after epithelial wounding was firstly described in mouse in 1988 by Nakayasu [63] and it was later attributed to apoptotic processes [64]. It is thought to have been originated as a mechanism to prevent the rapid spread of viral infections from the epithelium to the underlying stroma [65] therefore preventing further corneal inflammation and loss of transparency. This process has gained more attention in the past two decades with the increasing popularity of refractive surgery [65], as it triggers the same response [65]. The apoptosis is thought to be triggered by the release of unusually higher levels of Interleukin-1 (IL-1) by the epithelial cells in response to cell death or injury [64]. Apoptosis in keratocytes is mediated through the Fas system activated by IL-1 [66], and ultimately by the apoptotic regulatory molecules Bcl-2 and Bax [66]. Wilson et al. hypothesised that IL-1 reach lethal levels and causes apoptosis only on the nearby keratocytes and promote the release of hepatocyte (HGF) and keratinocyte (KGF) growth factors by the keratocytes situated farther from the wound [64]. Controlled regulation of the apoptotic response in keratocytes is essential to prevent states of chronic apoptosis and further complications such as keratoconus [67].

1.1.5.4.2 Keratocyte differentiation into proliferating and migrating fibroblasts

In the second phase of stromal wound healing (following keratocyte apoptosis), keratocytes adjacent to the area of cell apoptosis start proliferating and migrate to repopulate the wounded stroma [52]. Zieske et al. using rat models, showed the presence of proliferating Ki67-positive cells at the edges of the wound at 24 hours after injury, with increasing proliferative cell numbers peaking after 44 hours [52]. Ki67 levels did not return to normal basal levels until 7 days after wounding. As previously explained, keratocytes are mainly quiescent in the cornea, but become activated and acquire a fibroblastic migrative phenotype upon injury. Fini et al. showed that during differentiation and proliferation cell size and organelle content increases and cells acquire typical fibroblastic characteristics including fusiform shapes and multiple nucleoli [68]. In her study, Fini observed the migration of fibroblasts to the wounded area as early as 6 hours after injury [8]. Different factors have been suggested to activate the keratocytes, these include the previously mentioned IL-1, platelet-derived growth factor (PDGF), bone morphogenetic proteins 2 (BMP2) and 4 (BMP4), all released from the epithelium [69].

It is thought that during this phase, activated fibroblasts and epithelial cells secrete high levels of pro-inflammatory molecules, that create an influx of inflammatory cells into the stroma, including dendritic cells, macrophages, and lymphocytes [70]. To date, it is still not fully understood the mechanism by which inflammatory cells are recruited to the cornea. These inflammatory cells produce cytokines and chemokines that mediate the inflammatory response and prevent further damage from infections and environmental insults. Some of these molecules secreted by fibroblasts and immune cells are regulators not only of the inflammatory response but also of the remodelling process [71]. For instance, tumour necrosis factor α (TNF-α) secreted by macrophages has been shown to be essential to regulate the expression of pro-fibrotic molecules and thus, required to supress excessive scarring of the cornea during remodelling [71].

Keratocyte activation and differentiation into a repairing phenotype also involves the reorganisation of the actin cytoskeleton, with cells developing stress fibres and additional focal adhesion points [68]. These activated cells start expressing a new myriad of genes to repair and remodel the surrounding ECM. These include the cell-matrix adhesion integrinα5, fibronectin, fibronectin receptor, and matrix metalloproteinases (MMPs). Additionally, they stop expressing the associated-keratocyte proteins that allow to maintain corneal transparency, such as the crystallin ALDH1A1 [49, 68]. They also secrete higher amounts of chondroitin sulphate, a glycosaminoglycan upregulated during corneal stromal development [34]. In vitro, the repairing phenotype is easily promoted by culturing keratocytes in the presence of serum [8]. The activated fibroblasts acquire the 36

ability to express IL-1α, which works as an autocrine loop, initiating the expression of MMPs.

As previously mentioned, MMPs are rapidly expressed and activated by the repairing fibroblasts during tissue remodelling [72]. This process has been observed in many different tissues, including the cornea [73]. MMPs are secreted in a latent form (pro- MMPs) and need to be cleaved by either serine proteases or other MMPs to be catalytically active [74]. A wide range of growth factors and cytokines can transcriptionally activate expression of the MMPs, such as epidermal growth factor (EGF), HGF, fibroblast growth factor (FGF), vascular endothelial growth factor (VEGF), TNFα, transforming growth factor β (TGFβ), KGF, interleukins (such as IL-1), and interferons [8, 72]. MMP activity is also regulated by tissue inhibitors of MMPs (named TIMPs). MMPs allow cell migration by degrading the surrounding ECM or by modifying cellular adhesive properties. Interestingly, during this fibroblast migration process to the wounded site, Petroll et al. observed a high correlation between the pattern of fibroblast and collagen lamellae alignment, suggesting that intrastromal migration is guided by cell- matrix contact that facilitates wound repopulation [75].

Repairing fibroblasts in the cornea secrete high levels of MMPs, in order to remodel the surrounding ECM and to migrate towards the wounded area. As previously mentioned, IL-1 is one of the MMP regulators, which acts in an autocrine loop through the NFkβ pathway to promote the secretion of MMPs by fibroblasts [8]. In the healthy cornea, there is no expression of MMP1 and MMP3, however MMP2 is expressed at very low levels [8, 76]. MMP1, MMP2 and MMP3 are upregulated and activated in the repairing corneal stroma, allowing cell migration and subsequent wound contraction [77]. MMP1, also known as interstitial collagenase, acts mainly on cleaving collagen type I, and to less extent collagen type II and type III. It is the first MMP to act upon corneal wound healing, by making a single site-specific cleavage on the triple helix of collagen type I. The fragments of collagen released are thermally unstable and are denatured at 37°C. MMP2, also known as gelatinase A, has gelatinolytic activity, degrading the collagen fragments that result from MMP1 cleavage [77]. MMP3, also known as stromelysin 1, is mostly responsible for proteoglycan degradation. Abnormal MMP expression and secretion profiles contribute to many pathological ocular conditions, including diabetic retinopathy [78].

Repair tissue is slowly remodelled to the normal healthy stromal ECM, by the keratocytes and the repairing fibroblasts. However, few repairing fibroblasts can eventually

37 differentiate into myofibroblasts during the contraction phase of the corneal wound healing [8].

1.1.5.4.3 Myofibroblasts differentiation

In penetrating injuries that reach the stroma, some repairing fibroblasts can acquire a myofibroblastic-like phenotype, developing a putative contractile complex composed of intracellular F-actin stress fibres that contain alpha smooth muscle actin (αSMA), α- actinin, desmin, vimentin, myosin, and high-affinity fibronectin receptor [79]. Myofibroblasts acquire the capacity to contract the wound, and deposit unorganised ECM with different compositions than the normal healthy corneal stromal ECM. They secrete collagen type I and type III, the latter of which is typical from fibrotic tissue [80]. Moreover, in vitro investigations have shown that the matrix released by the myofibroblasts has lower levels of keratocan and lumican, proteoglycans with keratan sulphate, than the normal stromal ECM [8, 80]. Instead, they secrete higher levels of fibronectin and dermatan sulphate proteoglycans (namely biglycan), characteristic proteoglycans of fibrotic tissue [80].

Different factors regulate the process of myofibroblast activation, these include lysophosphatidic acid (LPA) and TGFβ. LPA mediates the cytoskeletal remodelling associated with myofibroblastic phenotype, by activating the GTP-binding proteins Ras and Rho [81]. Activated Rho acts by assembling focal adhesion proteins and stress fibres. TGFβ is also a modulator of the fibroblast-myofibroblast transition [79, 80]. Addition of TGFβ in rabbit corneal fibroblasts cultures in vitro induces further myofibroblast differentiation [79]. Moreover, PDGF and FGF2 are also regulators of the fibroblast-myofibroblast transition [82].

Apart from the differences in ECM composition secreted by myofibroblasts, the cellular changes of the activation of keratocytes and their differentiation into myofibroblasts are also a plausible reason for loss of corneal transparency. The change in cytoplasmic content (consequence of keratocyte-myofibroblast transition) might alter their refractive index and increase the light scattering [83]. In a recent study, Gardner et al. showed that in vitro quiescent bovine keratocytes had a refractive index of 1.381±0.004, which matches the surrounding stromal matrix. Interestingly, the refractive index dropped to 1.365±0.003 after their activation, leading to a mismatch with the stromal ECM contributing to light scattering and loss of transparency [83].

After matrix remodelling and wound closure, few of the myofibroblasts eventually die by apoptosis, but a small population differentiate into scar keratocytes, which still express

MMPs [8]. It has been shown that IL-1, together with TGFβ, are important in the regulation of myofibroblast survival. The decrease of TGFβ levels in the stroma, probably due to the restore of the basement membrane, triggers myofibroblast apoptosis through IL-1 activation [84]. Myofibroblast apoptosis and subsequent stroma repopulation by keratocytes is a key event to reduce the stromal opacity caused by the abnormal ECM secreted by the myofibroblasts [85].

Figure 1.4 shows a schematic representation of all these processes and their regulatory proteins.

Figure 1.4. Schematic representation of corneal epithelium and stromal wound healing. Upon injury, an increase in the levels of IL-1 leads to keratocyte apoptosis at the site of injury (1), this is followed by keratocyte proliferation, migration, and differentiation into repairing fibroblasts (2). The repairing fibroblasts secrete high levels of MMP1, MMP2, and MMP3 to remodel the stromal matrix (3). A small population of the repairing fibroblasts can differentiate into myofibroblasts, which release an abnormal ECM that leads to corneal scaring (4). The wounded cornea will show irregular ECM, higher opacity, which contributes to light scattering and visual loss (5).

1.2 Aniridia

Overview

The transparency of the cornea is compromised by several diseases including aniridia. Aniridia is a congenital disorder with a prevalence of 1:40000 to 1:100000 and no influence of gender or ethnic group known to date [1, 2]. In 90% of the patients, it is caused by a PAX6 haploinsufficiency. PAX6 is the master regulator of eye development and it is of crucial importance on the development of a fully functional eye [86, 87]. This is the reason behind the affectation of different parts of the eye in aniridic patients. Aniridia-like features can also be caused by mutations in FOXC1 [88], PITX2 [88], CYP1B1 [89], FOXD3 [90] and TRIM44 [91]. Aniridia can occur alone as an ocular disorder or together with an affectation of the brain, the kidney, and the genitourinary tract as part of the WAGR syndrome (Wilms tumour, Aniridia, Genitourinary anomalies and Retardation) [92].

The name aniridia (coming from the Greek “No-iris”) indicates that this condition is always characterised by a total or partial iris hypoplasia. However, patients can also suffer from cataracts, glaucoma, nystagmus, foveal hypoplasia, and corneal opacifications [93]. Although the iris hypoplasia is what gives the name to the disorder, it is not the most sight-threatening condition that patients with aniridia experience. Corneal opacification and vascularisation are significant causes of visual loss, and they are included under the term aniridia related keratopathy (ARK).

Genetics

1.2.2.1 Paired box 6 gene structure

The paired box 6 gene (PAX6) is located on the short arm of chromosome 11, at the 11p13 band [94]. It was cloned during the search for the gene responsible for the WAGR syndrome [95]. However, mutations in PAX6 alone are not responsible for WAGR syndrome, but only when they appear in conjunction with mutations in the neighbouring gene WT1 [96]. PAX6 is a highly conserved gene between species, with 100% identity in the coding region between human PAX6 and murine Pax6, and with the DNA binding domains being 90% identical with the homologous Drosophila gene Toy [97]. PAX6 in humans is composed of 13 exons that expand over 22kb (Figure 1.5). PAX6 has two alternative splicing sites, one of which creates a transcript with an additional 5a exon, consisting of 14 amino acids, which has different activation sites [98]. The predominant form of PAX6 in the eye is the one without the 5a exon.

Figure 1.5. Schematic representation of the paired box 6 gene (PAX6) and the Paired box protein Pax-6 (PAX6). PAX6 is located at the 11p13, and it is adjacent to WT1, BDNF, and ELP4 (not shown). PAX6 is transcribed from the reverse chain and has 13 exons over 22kb. It translates into PAX6 transcription factor that has four functional regions: paired domain (PRD), linker region, homeodomain (HD), and Proline, Serine, Threonine rich region (PST).

1.2.2.2 Described mutations in aniridia

Currently (last updated in 2018), in the Human PAX6 Allelic Variant Database, there are 491 unique DNA variants reported in a total of 1067 individuals. Two-thirds of PAX6 mutations are familial and follow an autosomal dominant inheritance, and the remaning are de novo mutations [99]. When the PAX6 mutations include deletions in the WT1 gene, there is a high risk that the person affected develops WAGR syndrome, presenting Wilm’s tumour, mental retardation, genitourinary problems, and aniridia [96].

Substitutions, deletions or insertions of DNA nucleobases are the causes for PAX6 mutations. Arising from these changes, there are mainly three different kinds of possible variants: Premature termination codons (PTC), C-terminal extensions (CTE) and missense mutations, as summarised in Table 1.1. Chromosomal rearrangements or deletions of an entire copy of PAX6 can also be the cause for aniridia, but they are less common than the previously mentioned [100].

The majority of the mutations recorded until now result in PTC and are caused by either non-sense or frame-shifts mutations. As suggested by its name, these types of mutations mean that the transcription is shortened over the expected location of the termination codon. PTC mutations in PAX6 are evenly distributed along the whole gene, except for

41 the 3’ distal region, where none have been described. Since there are no differences in the phenotypes regardless of the location of PTC mutation in the PAX6 gene, it is hypothesised that these are not translated into truncated proteins, but instead result in the degradation of the modified mRNA by the nonsense-mediated decay (NMD) [101].

CTE are caused by frame-shifts or missense mutations that result in a run-on mutation, meaning that the translation is extended over the expected location of the termination codon. Hypothetically, longer proteins with an extended C-terminal region would arise from these mutations. NMD would not act in these cases, but as they cause the same aniridic phenotypes as the PTC mutations, there might be a loss-of-function of PAX6 with the addition of an extra peptide in the C-terminal region [102].

Finally, missense mutations that cause aniridia are mainly located in the Paired-box domain [102, 103]. It is believed that these mutations translate into a PAX6 protein with a different amino acid than the wild-type PAX6, which might cause a change in its structure and result in impaired function. However, this protein might still retain some of its functionality. There are also missense mutations described in the PAX6 gene that do not cause aniridia, but other phenotypes such as isolated foveal hypoplasia, optic nerve defects or iris hypoplasia [104].

Table 1.1. Classification and definition of the three main types of mutations in PAX6. The most common example for each of them and its frequency are shown. Table adapted from [105]

Type of Premature C-terminal Missense mutations mutation termination codons extensions (CTE) (PTC)

Definition Creates a stop codon Changes the stop Changes one amino in the open reading codon to an amino acid codon to another frame acid codon in the open reading frame

Most frequent p.(Arg240*) p.(*423Leufsext*1 p.(Gly51Arg) mutation 08)

Frequency 5.9% 3.1% 0.31%

Exon 9 13 5

Change in Arg(cga) > Ter(uga) Ter(uua) > Gly(gga) > Arg(aga) codon Leu(uua)

Reference Kondo-Saitoh, A [106] Chao, L.Y [107] Henderson, R.A[108]

1.2.2.3 Genotype-phenotype correlation

As previously mentioned, aniridic patients present a broad range of phenotypes, from mild partial iris hypoplasia to effects on lens and cornea, among others. The great variety of phenotypes hinders to elucidate the biological mechanisms behind aniridia. Some studies have tried to find a relation between the aniridia phenotype and genotype, such as the extent of affected iris or presence of ARK, with different PAX6 heterozygote variants. Although some claim to find a relation between particular types of mutations and their effect (reviewed in [103, 109]), most of the studies conclude that no apparent link exists [110, 111].

Nonetheless, it is clear that mutant homozygous and compound heterozygous patients for PAX6 (when two or more heterogeneous recessive alleles at a particular locus are found) have the most severe phenotypes and usually die during embryonic development. To date, five cases of compound heterozygosis have been described [112], however only in two of the cases the foetuses survived and were born with no [113] or small eyes [114] and other brain defects.

In 2000, Duncan et al. suggested that the most common PAX6 mutations (truncations in the C-terminal region), which result in PTC, yielded a negative effect and caused more severe phenotypes than the other variants [115]. An in vitro study demonstrated that truncated PAX6 in the C-terminal half had a dominant negative effect on the wild-type protein [116]. However, in 2005 Tzoulaki and colleagues reviewed and compared all the mutations described in PAX6 in association with their phenotypes and concluded that PTC mutations simply cause loss-of-function and create a null allele. They hypothesised that Non mediated decay (NMD) acts to degrade all mRNAs that differ from the wild type being the major mechanism by which PAX6 PTC mutated alleles do not produce protein. Therefore, aniridia would be a result of a true haploinsufficiency which would reject the idea of a link between the genotype and phenotype in these cases [102]. Liu and colleagues confirmed this hypothesis by finding that the PAX6 mRNA levels dropped about 50% in patients with PTC, independently of the type and location of the mutations, compared to healthy people [117]. If PTC mutations had a gain-of-function or a negative effect, it would be expected to see different effects depending on the region mutated, with more severe effects when the mutation was located in the DNA-binding domain. In line with this hypothesis, a study with 30 patients showed that despite having truncating mutations, high inter- and intra-familial phenotypic variability, from milder to severe affectations was observed [110].

In a recent study, Yokoi and colleagues clinically analysed five families and 16 sporadic cases of aniridia. They compared the mutations found in each instance with the phenotype of the eyes, but no obvious correlation was identified between the genotype and phenotype. They suggested that PAX6 acts in an extremely complicated network of proteins with many other factors influencing its function during eye development. However, they concluded that the PAX6 regulation, resulting from both paternal and maternal alleles, is similar in both eyes, which turns into the same phenotype in both eyes of each patient [111].

However, it is accepted that missense mutations are not as severe as PTC or CTE. In 2009, forty-three individuals with aniridia were clinically analysed, and their variants were identified. All mutations caused variable panocular malformation, including iris hypoplasia, foveal hypoplasia, nystagmus, cataracts, and corneal abnormalities. Missense mutations were found to cause milder phenotypes than PTC and CTE, which caused severe visual loss [103]. In these cases, NMD does not degrade all the mutated mRNAs, and the produced PAX6 proteins may retain some, but not all, functionality. When these mutations affect the DNA binding domain of PAX6, they result in more

severe phenotypes, including foveal hypoplasia, marked iris anomalies, and severe visual impairment [103].

PAX6 transcription factor

1.2.3.1 Protein structure

PAX6 gene encodes for the Paired box protein Pax-6 (PAX6), a highly conserved transcription factor mainly involved in developmental processes, namely of the eye, spinal cord, brain, and pancreas [118]. By binding to the specific DNA site, it regulates the rate of DNA transcription to mRNA. The protein consists of 422 amino acids with three different functional regions (Figure 1.5). The first of the DNA binding domains, which is the Paired domain (PD), is located at the N-terminal region. As shown in the X- ray structure (Figure 1.6A), the PD in itself comprises two DNA-binding subdomains, N- terminal and C-terminal (purple structures in Figure 1.6A) [119]. A 79 amino acids long linker joins the PD with the homeodomain (HD), which is the second DNA binding domain (Figure 1.6B). Finally, at the C-terminal region of PAX6, there is a Proline, Serine, and Threonine-rich domain, which has a transactivational function [113].

Figure 1.6. Paired box protein Pax-6 structure. (A) X-ray structure of the paired domain of PAX6 transcription factor binding to the DNA double-helix (PDB accession code: 6PAX, 2.50 Å resolution [119]). The protein backbone, with a high content of α helix structures (pink) and two β sheets (yellow), is binding to the DNA helix (red and blue). (B) X-ray structure of the Engrailed 2 homeodomain, homologous to the HD in PAX6 (PDB accession code: 1HDD, 2.8Å resolution) [120].

1.2.3.2 PAX6 role in eye and corneal development

PAX6 transcription factor is essential for the embryonic development of the eye, nose, pituitary gland, pancreas, and central nervous system [121-124]. PAX6 is usually referred to as the master control gene of eye development. In Drosophila melanogaster, its targeted expression (induction) was sufficient to induce the formation of ectopic eye structures in the antennae, legs, and wings [125]. In the eye, Pax6 is involved in the development of the cornea, lens, iris, ciliary body and retina [87]. Homozygous mice that have a null mutation in both the Pax6 alleles (Pax6-/-), do not develop eyes, or an

olfactory system, and die soon after birth, therefore they are not a suitable model to study the role of Pax6 in eye development [86]. Instead, the role of Pax6 on eye development has been mainly evaluated using mice chimeras with a combination of Pax6+/+ and Pax6- /-, and heterozygous animal models that have mutations in one of the copies of the gene, such as Pax6+/- mouse models [121, 126]. The next sections will focus on the importance of PAX6 during the development of the corneal epithelium and stroma.

It is well documented that Pax6 plays a role in the development of the corneal epithelium [126]. Li et al. showed that, in chick eye development, Pax6 is firstly produced in the head ectoderm that will form the lens placode, which in turn will give rise to the intraocular lens and corneal epithelium [127]. In mice embryos, the production of Pax6 is required between E10.5 and E16.5 for the correct development of the corneal epithelium [126]. However, it is known that some events of corneal epithelial development, such as differentiation, stratification, and migration, occur postnatally [128]. Pax6+/- epithelial cells showed abnormal patterns of clonal growth and migration compared to their wild-type counterparts [129], suggesting that Pax6 is also crucial for epithelium growth and centripetal migration during postnatal development. Interestingly, the cornea of Pax6- overexpressing mice developed normally during the embryonic stage. However, a failure on the cornea postnatal growth was observed, resulting in a micro cornea with fewer epithelial cell layers in these mice [130]. Altogether, these observations advocate for the need of a specific Pax6 dosage for the correct postnatal corneal development.

The studies of the role of Pax6 on eye development had focused primarily on the ectoderm-derived layers (including the corneal epithelium). However, there is strong evidence that Pax6 is also essential for the morphogenesis of tissues populated by neural crest-derived cells, such as the corneal stroma [131]. Supporting this hypothesis, Collinson et al. showed weak expression of Pax6 in corneal stroma between E10.5 and E16.5. Additionally, using mice chimeras for Pax6, they also found that functional Pax6 is an autonomous requirement for the development of the corneal stroma (i.e. expressed by the cells populating this tissue) [126]. Kanakubo et al. observed a thicker stroma with irregular lamellae in the adult Pax6+/- mouse engineered with GFP-labelled neural crest- derived cells, compared to the wild type [131]. However, they did not find expression of Pax6 in the GFP-labelled cells in the corneal stroma of Pax6+/- mice throughout embryonic development, suggesting that the regulation of Pax6 expression in other corneal layers (most likely in the corneal epithelium) is also critical for the correct development of the corneal stroma [131]. Whether Pax6 expression during stromal formation is an autonomous (necessarily expressed in the neural-crest derived cells) or

47 a non-autonomous requirement (expressed by the epithelium) is still unclear. However, there is a consensus that Pax6 is also a regulator of the development of this layer.

1.2.3.3 PAX6 role on corneal homeostasis

In the cornea, Pax6 is not only required during morphogenesis, but is also essential during the maintenance of this tissue in adulthood [20, 132]. Pax6 is expressed in both limbal and corneal epithelium (by the limbal and corneal epithelial cells, respectively) [133], and different studies have shown that knocking down Pax6 in these cells results in a disruption of corneal homeostasis [133-135]. Ouyang et al. showed that loss of PAX6 in primary human limbal epithelial stem cells downregulated the expression of corneal epithelial markers (keratin 3 and keratin 12), inducing differentiation towards a skin-like phenotype with upregulation of epidermis markers (keratin 1 and keratin 10) [133]. Similar observations were also made by Li et al. [136] and by a different team using the CRISPR-Cas9 system to knock down PAX6 in primary human central corneal epithelial cells [134]. They showed a downregulation of keratin 3, keratin 12, and ALDH3A1, a corneal crystalline, which also resulted in loss of corneal identity and a gain of skin-like phenotype [134]. Studies using the heterozygous Pax6+/- mouse model were also crucial in elucidating the involvement of Pax6 in corneal epithelium maintenance and wound healing [132, 137]. Pax6 is essential to maintain the correct cellular adhesion in the corneal epithelium, as Pax6+/- mice showed large gaps between epithelial cells and change in the appearance of the hemi desmosomes (structures involved on cell-matrix adhesion), which may impair cell adhesion [132]. Sivak et al. showed that Pax6 is highly expressed in the migrating front of epithelial cells after injury, regulating the expression of Mmp9 and allowing wound closure [138]. Consistent results linked the reduction of Pax6 dosage with wound healing delays, caused by defective calcium signalling in Pax6+/- epithelial cells [139]. All these studies corroborate the hypothesis that PAX6 is essential for maintaining corneal epithelium homeostasis throughout adulthood.

In the postnatal stroma, PAX6 is only expressed by the CSSC but not by the differentiated keratocytes [55, 57]. As explained in detail in section 1.1.5.1.1, CSSC account for 3-4% of the total cell population of the stroma and are stem cells located in the limbal stroma in close proximity to the LESC [54]. It has been demonstrated that the loss of PAX6 in CSSC results in their differentiation into keratocyte phenotypes [55, 57]. Despite being currently used as a marker for CSSC, there is not enough information available on the role of PAX6 in adult corneal stroma homeostasis.

Aniridia related keratopathy

1.2.4.1 Clinical features

As previously mentioned, aniridia related keratopathy (ARK) is a progressive manifestation that threatens the integrity of the cornea in aniridic patients.

Around 90% of aniridic patients suffer from ARK [140, 141]. Unlike the congenital iris hypoplasia and cataracts, which are affected since birth, aniridic patients are born with transparent corneas [142]. Although their corneas are normal during childhood, they gradually begin to lose their transparency in adolescence, worsening with ageing [140, 141]. Symptoms of ARK start with a thickening of the peripheral corneal epithelium and the appearance of a superficial vascularised pannus over the cornea (Figure 1.7). It was hypothesised by Nishida and Tseng that the pannus appearance was due to a limbal epithelial stem cell deficiency (LESCD) [3, 143]. However, currently it is not clear if this process is due to congenital abnormalities of the LESC, problems in their regulation, a decrease in their number overtime or a combination of different factors. More recent studies have shown that not only LESC, but also differentiated corneal epithelial cells and even corneal stromal cells might be affected [137, 144].

Early features of ARK show a weak epithelium with recurrent erosions, epithelial defects and opacity, which leads to pain, photophobia, and decreased visual acuity [142]. ARK patients at this stage already present severe visual impairment [142]. Following limbal damage, the corneal surface is encroached by the neighbouring conjunctival epithelium, which may result in chronic inflammation. Inflammation on the ocular surface causes painful symptoms such as photophobia, foreign body sensation and burning. Inflammation also leads to the formation of a superficial pannus, initially located on the periphery of the cornea, which eventually covers the central cornea (Figure 1.7). This is followed by the formation of a thicker pannus, resulting in stromal scarring, fibrosis, and vascularisation [140, 145].

Figure 1.7. Aniridic cornea presenting ARK. Severe corneal vascularisation and opacification is observed. Image adapted from Secker, G.A. and Daniels, J.T. [145]

Histological studies have shown the presence of vascularised structures, goblet and conjunctival cells in the epithelium of corneas with ARK [3]. Goblet cells are mucin producing cells that only reside scattered on conjunctival epithelium [146]. Consistent with the observations from Nishida et al. [3], a recent study using in vivo confocal microscopy also observed that patients with ARK have degraded palisades of Vogt, where the LESC reside [147]. These patients also presented increased corneal thickness, inflammatory cell invasion, and reduced density of sub basal nerves. Increased corneal thickness in ARK patients has been widely reported, with a study showing an average corneal thickness of 600µm, instead of the normal 500µm [148].

In addition, ARK patients develop severe stromal scarring and fibrosis overtime. They also respond very poorly to corneal transplants and surgery, which triggers scarring very easily and abnormal wound healing [149]. Currently, the mechanisms behind these processes are unknown. We, therefore, aim to enlighten some of these processes by assessing the behaviour of aniridic corneal stromal cells in vitro.

1.2.4.2 Diagnosis

There is not a clinical procedure to diagnose ARK as a condition. Instead, the clinical features are diagnosed by different methods and a clinical output is established. For instance, epithelial defects are assessed by fluorescein, and corneal ulcers and vascularisation by slit lamp examination. If corneal pannus is present, optical coherence tomography of the anterior segment can be used to investigate the depth of the opacity.

In cases of uncertainty, impression cytology of the corneal surface can also be performed. In this technique, a cellulose acetate filter is pressed onto the corneal surface under topical anaesthesia and subjected to histological and immunohistological analysis to assess if the surrounding conjunctival epithelial cells have migrate onto the corneal surface [150]. However, this procedure is highly invasive and should only be performed in very specific cases.

In vivo confocal microscopy, has emerged as a promising tool to assess the extent of corneal damage [151]. This technique does not require the removal of corneal epithelial cells but requires physical contact between the imaging probe and the ocular surface. It can distinguish between the bright well-defined membrane corneal epithelial cells from the hyper-reflective conjunctival epithelial cells and goblet cells [151].

1.2.4.3 Treatment

Given the wide variety of aniridia phenotypes, the best approach for treatment would be to target the genetic defects. However, this option is still utopic. Currently, there is no 50

specific treatment for aniridia, only treatments to alleviate symptoms and manifestations. It is advised to take simple measures, such as regular eye examination and use of correcting spectacles during young age. Due to iris underdevelopment, there is no regulatory mechanism by which the eye can regulate light entrance, so it is also recommended to use tinted or photochromic glasses to reduce light sensitivity [142]. Several therapeutic approaches have been used to meliorate some of aniridia- associated manifestations. These include lubricant drops, ointments solutions, and autologous serum drops (when artificial lubricants are not suitable) to target dry eye; contact lenses in more severe cases to maintain the ocular surface moist and hydrated [142].

Currently, a very promising Phase II clinical trial for aniridia is taking place in Canada and the United States using Ataluren, and it is expected to finish in 2022. Ataluren is a nonsense mutation suppressor that would manipulate the PAX6 dosage, thus the aniridic eyes would restore their normal amounts of this transcription factor [152]. However, as previously mentioned, aniridia is caused by a broad range of mutations, which makes this treatment only suitable for patients who have a nonsense mutation in one of the PAX6 alleles. Ataluren is an already approved drug in the European Union to treat Duchenne muscular dystrophy [153].

ARK-associated abnormalities on the cornea are difficult to treat. Aniridic patients easily trigger high levels of corneal scarring, so it is not advisable to treat ARK patients with anything other than lubricants and mucolytic solutions, unless the vision is completely compromised. When severe visual loss due to the corneal opacification is diagnosed, penetrating keratoplasty, superficial keratectomy, or LESC transplant may be used as therapeutic approaches [154]. However, none of these treatments are 100% successful. The responses from the patients are highly variable and usually do not completely solve the problem [149].

1.2.4.4 Study of Aniridia related keratopathy (ARK)

1.2.4.4.1 ARK in humans

In 1983, Margo and colleagues histologically studied eyes from aniridic children and found different abnormalities across the different parts of the eye [155]. Namely, in the cornea they found vascular pannus beneath the epithelium and, in one of the cases, severe stromal vascularisation. Where the pannus was present, there was a delicate fibro vascular tissue with few inflammatory cells replacing the Bowman’s layer. In some

51 cases, the vascular pannus was only found in the peripheral cornea rather than in the central cornea, therefore not causing ARK [155].

The first study of ARK in vivo with aniridic patients was published in 1995 and reported superficial corneal opacification and vascularisation. They also found that the palisades of Vogt, where the LESC usually reside, were absent at the limbus, and goblet cells, normally found in the conjunctiva, were present in the peripheral cornea. These findings provided evidence that LESCD is a feature of ARK [3]. One year later, another study was published with more evidence for a link between ARK and LESCD [156]. The authors studied sections from aniridic and healthy donor eyes and observed a reduced expression of typical corneal epithelial keratin 3 and 12, invasion of blood vessels from the neighbour conjunctiva to the central cornea, and limbal epithelial hyperplasia of the diseased tissue. The acquisition conjunctival characteristics in detriment of corneal epithelial phenotype are indicators of LESCD involvement.

Three patients with ARK were evaluated using in vivo confocal microscopy by Le and colleagues, who suggested that the morphological changes of the limbus and the cornea are highly dependent on the severity of the disease [157]. The patient with severe ARK did not have subbasal nerve cells, structured wing nor basal cells, and showed goblet cells present in the central cornea. No palisades of Vogt could either be found. The patient with moderate ARK also had abnormal structure of wing and basal epithelial cells, absence of nerves, and presented lower epithelial and stromal cell density. The third patient presented mild features, which included prominent nuclei in the basal cells, reduced nerve density, and abnormal palisades of Vogt. Another study published later in the same year gave further evidence supporting the same hypothesis [158]. The investigators also concluded that corneal abnormalities, such as absence of palisades of Vogt, increased corneal thickness, absence of nerves, presence of inflammatory cells in the cornea, and lower density of basal epithelial cells increase with the severity of ARK [158].

Several studies using human corneal sections have also demonstrated that the corneal epithelium is affected by ARK. In 2008, Li et al. conducted a histological study with patients who suffer from ocular surface diseases, including one patient with aniridia [159]. They used aniridic epithelial tissue to study PAX6 and keratin expression and observed a decrease on PAX6 and keratin 12 expression levels and an increase on keratin 10 levels, a marker of epidermis differentiation. Supporting the results from Li et al. the epithelium of five different ARK corneas showed decreased expression of PAX6 and increased expression of keratin 13, a marker of corneal conjunctivalisation [160]. Together, these results suggest that PAX6 regulation is essential on maintaining the 52

corneal phenotype and its loss of expression results in loss of corneal identity in detriment of different types of epithelia (skin [159] and conjunctiva [160]). Vicente et al. recently evaluated the corneal epithelium of five different ARK patients, submitted to different types of keratoplasty surgeries. Regardless of the type of surgery, all corneas showed irregular epithelium, disruption of epithelial basement membrane, loss of the orderly pattern of collagen lamellae, and absence of collagen type I expression [161]. Interestingly, KI67 expression was upregulated in the basal cells of the central corneal epithelium of ARK samples compared to the normal corneas, suggesting an ongoing cell proliferation perhaps associated with the continuous epithelial remodelling [161].

Recent corneal histological studies have shown that the stroma is also affected by ARK. Vicente et al. showed irregular pattern of collagen lamellae in all ARK corneal stromal sections, and absence of collagen type I in the anterior stroma [161]. An altered ECM composition was also observed in the ARK stroma, with increased expression of the fibrotic markers tenascin-C, fibronectin, vimentin, and α-SMA, helping to explain the loss of corneal transparency and vision in ARK patients [161]. The increased production of α- SMA suggests the presence of myofibroblasts and scarring. It should be noted that these studies used corneas at advanced stages of ARK, some of which underwent LESC transplantation, therefore it was not surprising to find high levels of stromal scarring [161]. Moreover, Vicente et al. were the first to study signalling pathways involved in the development of ARK. They found decreased detection of Notch1 signalling pathway, increased detection of Wnt/β-catenin, SHH, and mTOR signalling pathway in the whole epithelium and stroma of the ARK corneas, compared to the normal healthy corneas [160]. Differences in expression of markers from these signalling pathways was seen both in the epithelium and the anterior stroma, where the pannus was present, suggesting the possibility of a crosstalk between the two layers.

1.2.4.4.2 Pax6 haploinsufficiency in Small eye mouse model

In 1967, Roberts described for the first time a mutant mouse with smaller eyes than normal mice. The mutation was shown to be dominant and, due to the resulting phenotype, it was named Small eye (Sey) [162]. Later, in 1991, Hill and colleagues found that the gene mutated in Sey mice was also the transcription factor Pax6, homologous to the human PAX6 that is mutated in aniridic patients [163]. Heterozygous mice for Pax6 with the Sey mutation will be referred to as Pax6+/- (Figure 1.8). Since the discovery of the Sey mouse model and its link to the homologous PAX6 in humans, numerous studies have been conducted in an attempt to compare their corneas to those of the wild-type mice and ultimately to model human ARK.

Figure 1.8. Images of the eyes of a wild type mouse (Pax6+/+) and a Sey mouse (Pax6+/-). Image adapted from Secker, G.A. and Daniels J.T. [145].

The majority of studies using the Pax6+/- mouse model, focused on the corneal epithelium. In 2003 Davis et al. used Pax6+/- mouse model to show that Pax6 is not only important during the embryonic development of the cornea but also during its maintenance in adulthood. They observed abnormalities in the central corneal epithelium of the Pax6+/- mice, presenting aberrant cell adhesion with dysfunctional desmosomes (structures specialized on cell to cell adhesion) and lower levels of keratin12, a suggested marker for differentiated corneal epithelial cells [164].

Similarly, the Pax6+/- mouse was the key to understand that ARK is not solely due to a LESCD. Differentiated corneal epithelial cells of Pax6+/- heterozygous mice had abnormal migration, proliferation, and differentiation abilities, suggesting central corneal epithelium abnormalities in Pax6+/- corneas [129, 165]. Supporting results were found by Douvaras et al. who described a loss of corneal homoeostasis in the adult Pax6+/- mice. The basal to suprabasal cell movement in these mutants is quicker than in wild-type, which results in an excessive cell turnover and thus, high cell loss. This unstable homoeostasis results in a progressive corneal deterioration that resembles aniridic corneas [144]. Furthermore, two studies by Ou and colleagues point to a chronic wound state exacerbated by the oxidative stress in Pax6+/- corneas, which contributes to their disrupted homoeostasis. Unwounded corneal epithelial cells in the Pax6+/- mouse have abnormal morphology, cell junctions, and interaction with the cytoskeleton. They also show impaired production of actin, keratin 12, and desmoplakin. These cellular changes are similar to those observed during wound healing, suggesting that these corneas might be in a permanent wound state. Moreover, with ageing the stroma and epithelium become opaquer, resembling the events of ARK [166, 167]. Taken together, these

observations corroborated the hypothesis of epithelial impairment on Pax6+/- mice, regardless of the involvement of LESC.

In contrast, there have been limited studies addressing the corneal stroma of Pax6+/- mice. In 2003, Ramaesh and colleagues studied fetal, young, and adult corneas of Pax6+/- mice and observed that prenatal stroma appeared irregular, thicker, and hypercellular compared to wild-type [168]. Similar results were found in the adult Pax6+/- corneas, where the stroma also appeared thicker with irregular collagen alignment, particularly in the central cornea [168]. Taken together, these observations suggested that the stroma is also affected in ARK. Later, the same group showed that corneal would healing was impaired in Pax6+/-mice, including defects in both the stroma and the epithelium [137]. They compared the process of wound healing in Pax6+/- and wild-type mice, and found an increased apoptosis rate of corneal stromal cells in Pax6+/- mice during corneal wound healing, also suggesting a stromal involvement in impaired wound healing in ARK corneas [137]. Mort et al. also corroborated this hypothesis by showing highly vacuolated keratocytes (typical feature of apoptotic cells) in the Pax6+/- stroma compared to wild-type keratocytes [169]. They also reported higher number of nerve cells innervating the stroma of the Pax6+/- mice. These three studies suggested that the stroma is involved in the progressive deterioration of the cornea in aniridic patients, however further studies are required to unravel the mechanisms behind this processes.

Only one report has addressed the corneal endothelium in the Pax6+/- mouse model, however it presented scarce results. Being part of a wider study of the three corneal layers, Mort et al. reported that Pax6+/- endothelial cells had irregular shape and undefined edges compared to the typical hexagonal shape found in WT endothelium. Similarly to the stroma cells, Pax6+/- endothelial cells were also more vacuolated than wild-type and had a more irregular surface [169].

1.2.4.4.3 In vitro modelling of ARK with human cells

Aiming to understand the role of PAX6 in the human cornea, there have been several attempts to knock down PAX6, and its related genes, on human derived corneal cell lines [170, 171]. In one of these studies, PAX6 knockout in LESC induced cell differentiation towards a skin-like phenotype with increased expression of epidermis-specific keratins (keratin 10) [136], suggesting that PAX6 may have regulatory control on LESC lineage. In a recent fascinating study, Roux et al. created for the first time a cellular model of PAX6 haploinsufficiency by using the CRISPR/Cas9 system to produce a PAX6 nonsense mutation (c.687 G > T) [172]. Further characterisation of these cells in vitro showed reduced cell proliferation, migration and adhesion capacities, making this 55

CRISPR/Cas9 model suitable for studying ARK and testing novel therapeutic approaches [172].

In vitro culture of primary ARK-derived epithelial cells was firstly attempted by Li et al. but resulted unsuccessful, with limited and not conclusive results [159]. More recently, Latta et al. successfully isolated and cultured aniridic epithelial cells obtained from limbal biopsies of two ARK patients [173]. These cells confirmed the lower PAX6 protein levels and reduced expression of keratin 3 and 12 when compared to healthy cells, indicating an altered differentiation state of aniridic limbal epithelial cells [173]. In a follow up study, they compared these results with the expression profiles of PAX6-knockdown primary limbal epithelial cells, and confirmed the reduction of keratin 3 and 12 [135]. In addition, they observed a downregulation of ALDH1A1 both in the patients and in PAX6 knockdown epithelial cells, suggesting that PAX6 is a regulator of this protein.

However, to the best of our knowledge, there are no reported studies using stromal cells from aniridic patients.

HYPOTHESIS

“PAX6 haploinsufficient corneal stromal cells play a role in the development of ARK”

AIMS AND OBJECTIVES

The main aim of this thesis was to characterise aniridic and normal corneal stromal cells and to identify differences in their phenotype and functionality. To accomplish so, the work was divided into 4 different chapters with the following objectives:

1. To culture and characterise normal corneal stromal cells in 2D. Cells isolated from the central cornea of normal donors were expanded in vitro and prompted to differentiate towards a keratocyte-like phenotype. Their phenotype was compared to that of the well characterised limbal stromal cells. The studies in Chapter 3 investigate this aim.

2. To culture and characterise aniridic corneal stromal cells in 2D. Cells isolated from the stroma of ARK corneas were expanded in vitro, genotyped and prompted to differentiate towards a keratocyte-like phenotype. Their phenotype was compared to that of normal corneal stromal cells. The studies in Chapter 4 address this aim.

3. To compare the phenotype of aniridic and normal corneal stromal cells inside a collagen-based 3D tissue equivalent (RAFT TE). Cells were cultured inside a RAFT TE and were characterised regarding their proliferation, spatial distribution inside the matrix, morphology and gene expression. The studies in Chapter 5 investigate this aim.

4. To compare the expression and secretion levels of matrix metalloproteinases between aniridic and normal corneal stromal cells inside the RAFT TE. The studies in Chapter 6 assess this aim.

CHAPTER 2: GENERAL MATERIALS AND METHODS

2.1 Isolation and culture of primary human corneal stromal cells

Isolation protocol for corneal stromal cells

Whole corneas kept in organ culture media for a maximum of four weeks [174] were obtained from Moorfields Eye Bank and maintained in accordance with routine eye banking procedures. Tissues were used with ethics approval (10/H0106/57-11ETR10). Donor information is presented in each specific results chapter. Normal corneal stromal cells were harvested from whole corneas using a similar method to that described by Du et al. 2005 [54]. Briefly, whole corneas were washed 3 times for 10 minutes in DMEM/F12 (Gibco®) supplemented with 1% Penicillin-Streptomycin (P/S) (Gibco®), before removing the central corneal button with an 8mm trephine. The peripheral limbal rim was separated and used for isolation of limbal stromal cells. Corneal stromal cells were isolated from the central corneal button as shown in Figure 2.1. The corneal button was cut into four strips with a scalpel (Figure 2.1B). Using fine scissors and forceps, the anterior cornea (approximate depth of 250µm) containing the epithelium and the anterior stroma was separated from the rest of the cornea and cut into small pieces (roughly 1mm2) with a scalpel (Figure 2.1C and D). These were pooled together and digested overnight in 0.5mg/mL collagenase (Sigma-Aldrich) in DMEM/F12 at 37°C (Figure 2.1E). The remaining pieces of tissue were dissociated by repeated pipetting and the resulting solution was centrifuged at 1500xg for 5 minutes, resuspended in culture medium and plated in a 100mm2 TC-treated tissue culture plate (Corning®).

Figure 2.1. Corneal stromal cell isolation from the central cornea. First, the whole corneal button (A) was cut into four strips (B). The upper part of each strip, containing the epithelium and the anterior stroma, was then cut and separated from the posterior stroma with fine scissors (C, D). The strips obtained were cut into small pieces and incubated in collagenase overnight (E). Finally, the cells were plated on a culture dish and cultured in culture medium (F). Scale bar = 300μm

Corneal stromal stem cell medium (CSSCM) was used to culture cornea stromal cells, unless specified otherwise. Modified from the medium described by Jiang et al. 2002 [175], CSSCM consists of 200mL of MCDB (Sigma-Aldrich), 300mL of 1g/L low-glucose DMEM (Gibco®), supplemented with: 1% (V/V) P/S, 2% (V/V) foetal bovine serum (FBS) (Gibco®), 10ng/mL epidermal growth factor (Sigma-Aldrich), 10ng/mL platelet-derived growth factor (R&D Systems), 1mg/mL AlbuMaxI (Gibco®), 120μM L-ascorbic acid 2- phosphate (Sigma-Aldrich), 10nM dexamethasone (Sigma-Aldrich), 0.1 % (V/V) insulin- transferin-selenium (Gibco®) and 100ng/mL cholera toxin (Sigma-Aldrich).

Corneal stromal cells were separated from epithelial cells by selective trypsinisation with TrypLETM Express (Gibco®). The plates were washed once with sterile phosphate buffer saline DPBS (1X) (Gibco®) treated with 1mL of TrypLE Express for 1 minute at room temperature [176], gently tapped on the side and 9mL of CSSCM were added. The cell suspension was then seeded in a T75-cm2 tissue culture flask (NuncTM). Cells were kept in a humidified incubator at 37°C with 5%CO2 in air.

Culture and maintenance of corneal stromal cells

Cells from primary cultures were regularly monitored under a phase contrast microscope (Eclipse TS100, Nikon) and were passaged before reaching 60% confluence. To passage cells, the culture flasks were washed once with DPBS and treated with TrypLETM Express for 2 minutes before adding CSSCM. Cells in P1 were seeded on T75- cm2 tissue culture flasks and from P2 onwards cells were plated on T175-cm2 (NuncTM), T225-cm2 (NuncTM) or T500-cm2 (NuncTM). Cells were always passaged before reaching 60% confluence to prevent premature differentiation (triggered by cell confluence [174]) and were re-seeded again at an approximate density of 20% of surface area.

Cryopreservation and storage of corneal stromal cells

Corneal stromal cells from each donor were cryopreserved and stored in liquid nitrogen (-196ºC) at P3 or P4 until further use. Cells were passaged as previously explained (section 2.1.2) and the cell solution was centrifuged at 1500xg for 5 minutes. The supernatant was removed, and the resulting cell pellet resuspended in cryopreservation medium consisting of 70% CSSCM, 20% FBS and 10% dimethyl sulfoxide (DMSO) (Sigma-Aldrich). The new cell suspension (~200000cells/mL) was aliquoted into 1mL cryogenic vials (Corning®). From each T175-cm2 flask at 60% confluence enough cells for one vial were obtained. These were placed in a Mr. Frosty (Corning®) cryopreservation vessel in a -80°C freezer to attain a cooling rate of approximately -1°C per minute. Following temporary storage at -80°C for up to three days the cells were transferred to liquid nitrogen (-196°C).

Cells were recovered by fast thawing at 37°C of the vials. CSSCM was added dropwise to the cell solution and plated. The medium was replaced with fresh CSSCM after 24 hours.

2.2 Gene expression

RNA extraction from cells in 2D

RNA was extracted from cells cultured on tissue culture plastic (2D) using RNeasy micro kit (Qiagen) as per manufacturer’s instructions. The cell suspension was centrifuged for 5 minutes at 1000xg, the supernatant removed, and the resulting cell pellet resuspended in 350μL of RLT buffer (provided with the kit) supplemented with 10μL/mL of β- mercaptoethanol (Sigma-Aldrich). Cell lysates were stored at -80°C until all the lysates from the different conditions were obtained. For further processing, lysates were quickly thawed at 37°C. They were then loaded into a QIAshredder column (Qiagen) and

61 completely homogenised for 2 minutes by centrifugation at 13000xg. The homogenised cell lysate was mixed with an equal volume of 70% EtOH (Fisher Scientific) and placed in a RNeasy column (provided with the kit). By centrifuging for 30 seconds at 10000xg, the nucleic acid bound to the column polymer and the flow through was discarded. The column was then washed once with 350μL of the RW1 washing buffer (provided in the kit) and centrifuged for 30 seconds at 10000xg. DNA removal was achieved by using the RNase free DNase kit (Qiagen). Briefly, 70μL of RDD buffer supplemented with 10μL of DNase enzyme were loaded onto the column and incubated for 10 minutes at room temperature to degrade the genomic DNA. The column was then washed once more with 350μL of RW1 buffer and twice with 500μL of RPE buffer (provided with the kit). After air-drying by 5 min centrifugation at 13000xg, the RNA was eluted in 50μL of RNase free water (provided with the kit).

Assessment of RNA purity and integrity

The concentration and purity of the RNA was assessed using a Nanodrop spectrophotometer (ThermoFisher ScientificTM) capable of measuring the absorbance of small amounts (~1µL) of solutions across a wide range of wavelengths. RNA has a peak of absorbance at 260nm and by applying the Beer Lambert law; the nanodrop calculates its concentration. 260/280 and 260/230 absorbance ratios are indications of the RNA purity. The ideal 260/280 ratio of a pure RNA solution is ~2.0. Values lower than that are an indication of protein contamination in the sample. 260/230 ratios are indications of the presence of any remnants from the isolation process, such as phenols and guanidinium isothiocyanate.

The integrity of the RNA was assessed by running a control RT-PCR with a housekeeping gene (β -actin), to ensure that the target could be amplified, and the RNA was not degraded. The RT-PCR reaction was done in a two-step fashion. Firstly, total cDNA was synthesised as explained in the next section (2.2.3). Secondly, Go Taq® G2 Green Master Mix (Promega) was used for the amplification of the β-actin region as per manufacturer’s instructions. Briefly, a mixture of 10μL of Go Taq® Green Master Mix, 1μL of forward primer (Sigma-Aldrich), 1μL of reverse primer (10μM), 3μL of cDNA template (50-100ng/μL) and 5μL of molecular grade water (ThermoFisher ScientificTM) was prepared in a 0.2mL PCR reaction tube (Starlab) for each sample. The tubes were briefly centrifuged and placed into the thermal cycler (MasterCycler Gradient, Eppendorf). The running program was the following: 95°C for 2 minutes, 35 cycles of 94°C for 1 minute, 54°C for 1 minute, 72°C for 1 minute and finally 5 minutes at 72°C.

The sequence of the primers used was: β-actin forward: CCAACCGCGAGAAGATGA, β-actin reverse: CCAGAGGCGTACAGGGATAG.

When the RT-PCR reaction was finished, the products were visualized on an agarose gel. A 2% agarose gel was prepared by adding 2g of agarose (Promega) in 100mL of 1X TAE buffer (ThermoFisher ScientificTM). The mixture was boiled in the microwave and 5μL of SafeViewTM Classic (abm) was added to it before casting the gel. When the gel was set, 5μL of 100bp DNA ladder (Promega) and 15μL of sample were loaded and run for 45 minutes at 100V in running buffer (1X TAE with 5% (v/v) SafeViewTM Classic). The gel was then imaged in a chemiDocTM XRS (Bio-Rad).

Reverse transcription

All RNA samples from all donors and conditions that had to be compared by qPCR were brought to the same RNA concentration by diluting them with molecular grade water. cDNA was synthesised with the QuantiTect Reverse Transcription kit (Qiagen) as per manufacturer’s instructions. The first step of genomic DNA removal was performed by adding 2μL of gDNA wipe-out buffer to 12μL of RNA solution and incubating it at 42°C for 2 minutes. A mixture of 1μL RT primer mix, 4μL buffer and 1μL reverse transcriptase was added to the 14μL RNA solution, making a total volume of 20μL. The resulting solution was then incubated for 30 minutes at 42°C to activate the transcriptase and synthesise the cDNA, followed by 3 minutes at 95°C to stop the reaction. cDNA was then stored at -20°C until further use.

Real-time qPCR

The first step was to find a suitable endogenous control to use as reference for all the experiments. This control gene had to maintain constant expression across all samples regardless of the different experimental conditions. A TaqmanTM Array for Human Endogenous Control (Applied BiosystemsTM) was used for this purpose. The array had a 96-well plate format, consisting of 32 potential endogenous control genes in triplicate, where three different samples could be tested. Limbal, central and aniridic corneal stromal cells cultured at low density in CSSCM were tested. Each well contained dried Taqman ™ assay for a gene. A mix of 10μL of Gene Expression Master Mix (Applied BiosystemsTM) and 10μL of cDNA (5ng/μL) template was added to perform the qPCR reaction. The plate was then covered with MicroAmp® Optical Adhesive Film (Applied BiosystemsTM) and centrifuged for 1 minute at 1000xg to bring the solution to the bottom of the wells. The thermal cycling conditions were set in the QuantStudioTM 6 Flex Real-

Time PCR System (Applied BiosystemsTM) as following: 2 minutes at 50°C, 10 minutes at 95°C and 40 cycles of 15 seconds at 95°C and 1 minute at 60°C. qPCR was used to measure gene expression of the genes shown in Table 2.1, and was performed as follows. In a MicroAmp® Optical 96well Reaction Plate (Applied BiosystemsTM), 10µL of Gene Expression Master Mix, 4 µL nuclease-free water, 1 µL of Taqman ™ assay (Table 2.1, Applied BiosystemsTM) and 5 µL of template cDNA (5ng/µL) were added to each well. The plate was then covered with MicroAmp® Optical Adhesive Film and centrifuged for 1 minute at 1000g before placing it in the QuantStudioTM 6 Flex Real-Time PCR System. The thermal cycling conditions were set to 2 minutes at 50°C, 10 minutes at 95°C and 40 cycles of 15 seconds at 95°C and 1 minute at 60°C. qPCR data was stored and exported in an Excel file. Data was analysed using ΔΔCt method [177].

Table 2.1. List of Taqman ™ assays used for Real Time qPCR. Abbreviations: CD: cluster of differentiation, KERA: Keratocan; LUM: Lumican; ALDH1A1: aldehyde dehydrogenase 1 family member A1, αSMA: Alpha-smooth muscle actin; TP53: Tumour protein p53, BCL2: B-cell lymphoma, MMP: matrix metalloproteinases, RPL37A: Ribosomal Protein L37A.

Taqman ™ assay ID (Applied BiosystemsTM) Putative Corneal stromal stem cell markers PAX6 Hs01088114_m1 CD90 (gene: THY1) Hs00174816_m1 CD73 (gene: NT5E) Hs00159686_m1 Keratocyte markers KERA Hs00559942_m1 LUM Hs00929860_m1 ALDH1A1 Hs00946916_m1 Fibrosis markers αSMA (gene: ACTA2) Hs00909449_m1 Proliferation and apoptosis related markers KI67 (gene: MKI67) Hs01032443_m1 TP53 Hs01034249_m1 BAX Hs00180269_m1 BCL2 Hs00608023_m1 Matrix remodelling markers MMP1 Hs00899658_m1 MMP2 Hs01548727_m1 Endogenous control RPL37A Hs01102345_m1

2.3 Western blot

Protein extraction

Cells were extracted from a T175-cm2 tissue culture flask. After TrypLE ExpressTM dissociation the cell pellet was obtained and washed once with iced cold DPBS. The new suspension was centrifuged for 5 minutes at 1000xg and the supernatant was discarded. The remaining pellet was resuspended in RIPA Lysis and Extraction buffer (Thermo ScientificTM) with 1% (V/V) HaltTM Protease Inhibitor Cocktail (Thermo ScientificTM). The resulting solution was then incubated on ice on a rocker for 15 minutes and centrifuged

65 at 13000xg for 15 minutes. The supernatant was transferred to a new eppendorf (Eppendorf) and stored at -20°C until further use.

Protein quantification

A PierceTM BCA Protein Assay Kit (ThermoFisher ScientificTM) was used to quantify the total amount of protein in the cell lysates. The Bicinchroninic acid (BCA) assay relies on the capacity of the proteins to reduce the Cu2+ ion to Cu1+ in an alkaline environment combined with its colorimetric detection by bicinchroninic acid. In the first step, cupric ions form a blue chelate complex with amino acids (Biuret reaction). In the second step, the bicinchroninic acid reacts with this complex, forming a strong purple coloured product that has a peak of absorbance at 562nm. The higher concentration of protein present, the higher absorbance will be detected at 562nm.

A diluted series of Albumin (BSA, provided with the kit) standards were prepared in RIPA buffer with concentrations ranging from 2000μg/mL to 25μg/mL. A blank sample with 0μg/mL was also included. Samples were thawed at room temperature, briefly vortexed and centrifuged for 5 minutes at 13000xg. Working reagent was prepared by mixing 50 parts of BCA Reagent A with 1 part of BCA Reagent B. In a 96 well-plate (NuncTM), 25μL of standard or sample and 200μL of working reagent were added to each well. The plate was incubated for 30 minutes at 37°C and the absorbance was measured at 562nm under a microplate reader (Tecan Saphire).

Electrophoresis

All samples were diluted in water to the same concentration prior loading. 19.5μL of protein solution were added to 7.5μL of NuPAGE® LDS sample buffer (Invitrogen™) and supplemented with 3μL of NuPAGE® Sample reducing agent (Invitrogen™). The resulting solution was briefly vortexed, centrifuged, and boiled at 100oC for 10 minutes. While samples were boiling, 800mL of 1X running buffer was prepared from a stock of 20X NuPAGE® MOPS SDS running buffer (Life Technologies™). NuPAGE Bis-Tris 10% (W/V) polyacrylamide mini gels of 1.5mm thickness were used. The white strip of a gel was removed, and the gel was secured on a XCell SureLock™ Mini-Cell tank. 200mL of running buffer supplemented with 500µL of NuPAGE® Antioxidant (Invitrogen Invitrogen) were added to the inner chamber and the combs were carefully removed. The remaining running buffer was added to the outer chamber. Samples were briefly vortexed and centrifuged before being loaded to the gel. 10µL of PageRuler™Plus Prestained Protein Ladder (ThermoFisher ScientificTM) and 30µL of each sample were loaded. The gel was run at 200V for 50 minutes.

Immunoblotting

Wet (tank) transfer was used to transfer the proteins from the gels to the membranes. Transfer buffer was prepared by adding 40mL of 20X NuPAGE® Transfer Buffer stock solution (Life TechnologiesTM), 760mL of distilled water, and 120mL of methanol. A pre- cut InvitrolonTM PVDF membrane (Life TechnologiesTM) was separated from the filter papers that sandwiched it, and was activated in 100% ethanol for 2 minutes, washed in distilled water and placed in transfer buffer. The filter papers were directly placed in transfer buffer. Sponge pads (8) were also soaked in transfer buffer. After completion of electrophoresis the gel was removed from the cassette. The gel knife was inserted in the gaps of the 3 bonded sides of the gel case and pushed down and up, separating the two plates. The gel remained attached to only one of the plates and the wells were removed. The membrane was then placed on top of the gel and covered with a filter paper, always soaked in transfer buffer. The sandwich was removed from the plate and covered with another filter paper. The transfer sandwich was prepared as shown in Figure 2.2 inside the XCell II™ Blot Module and placed inside the tank.

Figure 2.2. Schematic representation of the sandwich blot, from the Cathode core (-) to the Anode core (+).

The Blot module and the outer chamber of the tank were filled with transfer buffer. The transfer was run for 90 minutes at 35V.

The necessary buffers were prepared for the next step of the Western blot. 10X TBS was prepared by adding 60.55g of Tris Base (Sigma-Aldrich) and 85.2g of NaCl (Sigma- Aldrich) to 500mL of deionised water. The pH of the resulting solution was adjusted to 7.6 with concentrated HCl (Fisher) and the volume was made up to 1L with deionised water. 1X TBS-T was prepared by adding 100 mL from the stock solution of 10X TBS, 900mL distilled water and 1mL Tween-20 (Sigma-Aldrich). The blocking solution was

67 prepared by adding 2.5g of skimmed milk to 50mL of 1X TBS-T to a final concentration of 5% (W/V).

The membrane was then blocked in blocking solution for 1 hour at room temperature on the rocker and incubated with the primary antibody in blocking solution at 4°C overnight (Table 2.2).

Table 2.2 List of primary antibodies used for Western blot. Abbreviations: CD: Cluster of differentiation, KERA: Keratocan; LUM: Lumican; αSMA: Alpha-smooth muscle actin; GAPDH: Glyceraldehyde 3-phosphate dehydrogenase.

Antibody Host Manufacturer Dilution Protein CD90 Rabbit Abcam (ab133350) 1/1000 30µg CD73 Rabbit Abcam (ab133582) 1/800 30µg KERA Rabbit Abcam (ab113115) 1/400 30µg LUM Rabbit Abcam (ab168348) 1/1000 30µg αSMA Rabbit Abcam (ab5694) 1/400 30µg GAPDH Rabbit Abcam (ab9485) 1/3000 30µg

The membrane was then washed three times with TBS-T for 5 minutes and incubated with the secondary antibody (Anti-rabbit IgG, HRP-linked Antibody) (Cell Signalling Technology) diluted 1:2000 in blocking solution for 1 hour at room temperature on the rocker. After washing three times with TBS-T for 5 minutes, the membrane was incubated for 5 minutes with Immobilon® Crescendo Western HRP Substrate (Merck Millipore), and visualised in a chemiDocTM XRS (Bio-Rad). The images were quantified using ImageLab 5.0 software.

Membranes were stained with three primary antibodies binding to proteins with different molecular weights: whether CD90 (18kDA), CD73 (63kDA) and αSMA (42kDA) or LUM (46kDA), KERA (40kDA) and GAPDH (36kDA). Membranes were stained for each antibody separately and no membrane striping or cutting up was necessary.

Western blot quantification

The immunoblots were scanned using a Chemidoc (chemiDoc ™ XRS+, Bio-Rad) and quantified using ImageLab 5.0 Software. Briefly, images were analysed using the Analysis Tool Box for lane and band detection. Individual bands were manually selected for better sensitivity. The region of interest around the band was selected and kept constant for all bands. The total volume intensity (arbitrary units) was assessed and compared against the total intensity of GAPDH band for the according sample.

2.4 Immunocytochemistry

Samples in 2D and 3D were fixed for 10 minutes and 30 minutes in 4% Paraformaldehyde in DPBS, respectively. They were then stained following a protocol described by Kureshi et al. with small modifications [55]. Briefly, samples were incubated for 10 minutes in 0.25% (V/V) Triton-X 100 in DPBS to permeabilise the cell membranes. They were then incubated for 1 hour in 5% (V/V) goat serum in DPBS to block and prevent unspecific staining. Samples were then incubated with primary antibody (See chapters accordingly) in 2% (V/V) goat serum in DPBS O/N at 4°C. The next day, they were washed 3 times in fresh DPBS for 5 minutes. Samples were then incubated for 1 hour at room temperature in the dark with a secondary antibody and Phalloidin 418 (Sigma-Aldrich) (1:1000) in DPBS. After 3 washes with DPBS, they were mounted in VECTASHIELD mounting medium with DAPI (Vector Laboratories) and visualised under a confocal microscope (LSM 700, Zeiss).

CHAPTER 3: NORMAL CORNEAL AND LIMBAL STROMAL CELLS HAVE SIMILAR PHENOTYPICAL CHARACTERISTICS

3.1 Introduction

The main aim of this thesis was to study corneal stromal cells isolated from aniridic patients. However, corneal surgery is not often recommended to aniridia patients, so this tissue is invaluable and difficult to obtain. Furthermore, only the central corneas from aniridic eyes (not the limbal region), are discarded when patients undergo surgery. Occasionally this redundant tissue became available for the research presented in this thesis. Therefore, it was necessary to first optimise the isolation and culture of corneal stromal cells from normal donor corneas because historically, these cells, as in vivo, proliferate very rarely in vitro [35, 79]. This chapter presents the novel culture and then characterisation of normal corneal stromal cells.

In the native cornea, the stroma is populated by quiescent keratocytes [22] that only proliferate and differentiate following injury [178] and disease [179]. Because of their non-proliferative nature, keratocytes have proven to be a challenging cell source for in vitro studies when trying to maintain their native phenotype [35, 79]. This can be achieved by isolating and culturing the keratocytes directly in serum-free medium conditions, but as one would expect, low cell numbers are usually obtained [35, 79]. Therefore, there have been many attempts to isolate and expand keratocytes in vitro by culturing them under different conditions [180-187]. There is a wide range of studies that, although all use cells isolated from the corneal stroma, describe different cell characteristics in vitro by modifying the culture conditions [35, 180-187]. Because of the myriad of phenotypes in vitro, authors refer to the corneal stromal cells by different names [22, 35, 79, 178-187], making their nomenclature sometimes confusing. The cells used in this chapter will be referred to as corneal stromal cells or limbal stromal cells, depending on their original anatomical location in vivo (cornea and limbus, respectively). In addition, they will be named normal indicating that were isolated from non-diseased donors. This is because in the next chapters, focus will be placed on aniridic corneal stromal cells.

Several studies have aimed to identify the optimal conditions to isolate and culture corneal stromal cells while maintaining the keratocyte phenotype [180, 182, 183, 188]. As previously mentioned, culturing corneal stromal cells directly in serum-free medium achieves this aim, but not enough numbers are obtained to perform extensive experiments [35, 79]. Thus, corneal stromal cells have been isolated and cultured for a

71 small period of time in the presence of serum, allowing their numbers to expand, before placing them in a specific keratocyte culture medium [180, 182, 183, 188]. Their keratocyte phenotype is then assessed by cell proliferation, morphology, expression of typical keratocyte markers (KERA, LUM, ALDH) and absence of myofibroblast markers (αSMA). Interestingly, after prior exposure to serum containing medium, corneal stromal cells could only partially recover their keratocyte phenotype when subsequently cultured under serum-free conditions [180, 183]. While the cells successfully regained quiescence, dendritic morphology, they only gained low expression of lumican and keratocan [180, 183], and no expression of ALDH [183]. Therefore, serum-free medium has been supplemented with different molecules and growth factors to better promote the keratocyte phenotype in vitro [180, 182]. For instance, L-Ascorbic acid 2 phosphate (A2P) stabilises collagen production, promotes higher expression and secretion of keratocan and lumican [180, 189]. Another example is insulin, which stimulates keratocyte proliferation over a period of 6 days while maintaining high expression of ALDH3A1 and keratocan, as well as dendritic morphology [184]. Finally, basic fibroblast growth factor (FGF) has been reported to increase cell proliferation and expression of keratan sulphates proteoglycans by keratocytes while maintaining the dendritic morphology and inhibiting the expression of αSMA [79, 186]. Therefore, in the work here presented a keratocyte differentiation medium (KDM) consisting of a serum-free medium containing A2P, L-glutamine supplements and FGF was used to prompt corneal and limbal stromal cells towards a keratocyte phenotype. This medium has been successfully used for the same purpose in previously published studies [54, 174].

Serum-containing medium is also widely used to culture corneal stromal cells in vitro but, perhaps confusingly, these are then usually referred to as corneal fibroblasts [35, 185]. Corneal fibroblasts may be defined as activated keratocytes, that in the presence of serum, proliferate and acquire an elongated spindle-shape morphology [35, 186], resembling the phenotype of activated keratocytes during wound healing in vivo [190]. High cell numbers were achieved by culturing corneal stromal cells in serum containing medium, however their phenotype significantly differs from the native keratocytes [185]. Corneal fibroblasts show downregulated expression of keratocyte markers, namely keratocan and ALDH3A1 [185]. Also, the proteoglycans secreted to the medium are different when compared to keratocytes, with a reduced expression of keratan sulphate and an increase in dermatan sulphates proteoglycans [185]. Interestingly, a study from 1996 showed that the presence of serum alone was sufficient to induce the expression of αSMA, a myofibroblast marker, in a small population of the fibroblasts [79]. However, most of the studies where corneal fibroblasts were cultured in the presence of serum,

did not report the expression of this marker [35, 185, 186, 191]. Instead, it is commonly accepted that a myofibroblast phenotype is easily promoted in vitro by supplementing the cell culture medium with TGFβ, regardless of the presence or absence of serum [79, 80, 180]. Together, these results suggest that different culture conditions are triggering agents to modulate cell behaviour.

Strikingly, a recent study reported that central corneal stromal cells cultured in 10% serum-containing medium could display features of mesenchymal stem cells (MSC) including expression of the MSC markers (CD90, CD73 and CD105), and the capability to differentiate into cells of apparently three different lineages (osteocytes, chondrocytes, and adipocytes) [187]. This data is surprising because the presence of corneal stromal stem cells (CSSC) had only been previously described in the limbus [54]. Although the mentioned study does not prove the presence of MSC in the central cornea, it does show the multifaceted capacities of central corneal stromal cells to acquire different phenotypes when cultured in vitro. Despite presenting stem cell-like properties, these cells acquired a fibroblastic morphology [187]. This is perhaps surprising because CSSC have been previously thought to present a typical small and square-like morphology (typical feature of stem cells). It was also thought that the loss of this morphology and acquisition of an elongation of the cell body was a result of cell differentiation [174].

Given the reported potential for stromal cells to change proliferative capability, in the present study it was attempted to culture corneal stromal cells using conditions based on those previously used to successfully propagate CSSC in vitro [54]. On the one hand, as previously mentioned, culturing central stromal cells in keratocyte differentiation medium (KDM) could maintain a relatively quiescent phenotype, similar to that found in vivo [79, 186]. However, these conditions may not provide sufficient cells with which to perform experiments. On the other hand, using culture medium supplemented with 10% serum generally promotes cells to become fibroblastic with limited potential to recover their native phenotype [183, 192]. In contrast, corneal stroma stem cell medium (CSSCM) has only 2% serum and is supplemented with epithelial growth factor (EGF) and platelet derived growth factor (PDGF) [174]. CSSC have been routinely isolated from the limbus and cultured in vitro in CSSCM, where they can divide extensively for up to 100 population doublings [54]. CSSC have been often described as MSC, expressing typical MSC markers (CD90, CD73) and PAX6, a marker of ocular development [57, 193]. They have the ability to differentiate into keratocytes in vivo, as well as in vitro [57]. For instance, CSSC cultured in 2D in KDM for 3 weeks acquired a dendritic morphology and upregulated expression of KERA, LUM and ALDH1A1, suggesting the acquisition of a keratocyte phenotype [174]. Hence, CSSCM can promote proliferation of CSSC in vitro 73 without inducing fibroblast and myofibroblast differentiation but maintaining their stem- like features instead [54, 57, 174]. With this in mind, it was hypothesised that CSSCM might also promote cell proliferation of central stromal cells without inducing fibroblast or myofibroblast differentiation.

As previously mentioned, after culture and expansion in CSSCM, CSSC can be differentiated into keratocyte-like cells in vitro by culturing them in KDM [174, 194]. KDM is absent of serum, which therefore reduces cell proliferation [35], supplemented with A2P, which promotes collagen production [189], and FGF. FGF is a growth factor that exerts its function through binding to transmembrane tyrosine kinase receptors and signalling mainly through RAS/MAP kinase pathway [195]. FGF has been shown to modulate proliferation, migration, differentiation and angiogenesis of different cell types in vitro [195], including keratocytes [186]. Specifically, FGF promoted keratocyte proliferation and increased mRNA levels of keratan sulphate proteoglycan core proteins (lumican and particularly, keratocan) [186]. In another study, FGF was capable of reverting the expression of αSMA by myofibroblasts that had previously been induced by serum and TGFβ [79]. Moreover, keratocytes derived from CSSC cultured on highly aligned fibrous substrate made from poly(ester urethane) urea with FGF-containing KDM secreted highly aligned collagen fibrils compared to those cultured on KDM depleted of FGF [196]. Hence, to better identify a putative mechanism for the regulation of CSSC differentiation into stromal cells with a keratocyte-like phenotype in vitro, two different conditions were studied; KDM with or without FGF.

In this chapter, normal corneal stromal cells were first isolated and cultured in KDM or CSSCM to find which of these culture media better supported their growth. The phenotype of normal corneal stromal cells when cultured in CSSCM was then studied and compared to normal limbal stromal cells (also referred in the literature as CSSC) cultured in the same conditions. Finally, both corneal and limbal stromal cells were prompted towards a keratocyte phenotype in KDM, to assess whether they successfully differentiated into keratocyte-like cells.

3.2 Hypotheses and aims

The hypotheses tested in this chapter are the following:

1. Corneal stromal cells isolated and cultured as keratocytes in KDM do not yield sufficient cell numbers from individual donors for experimentation.

2. CSSCM promotes corneal stromal cell proliferation in vitro.

3. Corneal and limbal stromal cells maintain different phenotypes even when cultured under the same conditions.

4. FGF is an essential regulator of corneal and limbal stromal cell differentiation into keratocytes in vitro.

Hence, the overarching aim of Chapter 3 was to determine to what extent normal corneal stromal cells, compared with limbal stromal cells (putative CSSC), share phenotypic characteristics and differentiation potential.

3.3 Materials and methods

Assessment of KDM and CSSCM as media to culture and expand normal corneal stromal cells

3.3.1.1 Corneal stromal cell isolation

To determine the optimal culture conditions, three central corneas from normal donors were kindly gifted by Fondazione Banca Degli Occhi del Veneto-Onlus with the appropriate research consent and used according to the ethics (10/H0106/57-11ETR10). Donor details can be found in Table 3.1. For the detailed protocol of cell isolation see section 2.1.1 in Chapter 2. After isolation, half of the cells from each cornea was resuspended in 5mL KDM and the other half was resuspended in 5mL CSSCM.

Table 3.1. Donor information. Age, gender, retrieval time post-mortem and storage time are shown.

Tissue ID Age Gender Retrieval time post- Storage time mortem (hours) (days) (years)

Donor 1 61 Male 16h 25m 17

Donor 2 72 Male 10h 0m 19

Donor 3 70 Male 20h 32m 15

3.3.1.2 Corneal stromal cell culture in Keratocyte differentiation medium (KDM)

KDM composition can be found in Table 3.2. Three T25-cm2 flasks (NuncTM) were coated by adding 0.5mL of fibronectin FNC Coating mix®, (AthenaES) for 30 seconds at room temperature. FNC was removed and the cell suspensions in KDM were added. The flasks were then placed in a tissue culture incubator at 37ºC and 5% CO2 in air. The cells were kept in culture for 22 days and medium was refreshed three times a week. Cells were daily visualised under a phase contrast microscope to monitor their confluence levels and images were taken every 7 days.

Table 3.2. KDM composition adapted from Kureshi et al. [174]. Abbreviations FGF: basic fibroblast growth factor, KDM: keratocyte differentiation medium.

Component KDM-FGF

Advanced DMEM Low glucose (Gibco®) 500mL

Recombinant human FGF (Gibco®) 10ng/mL

L-Ascorbic Acid 2-phosphate sesquimagnesium salt hydrate 0,1mM (Sigma-Aldrich)

GlutaMAXTM 1 (100X) (Gibco®) 1X

Penicillin-Streptomycin (Gibco®) 100ug/mL

3.3.1.3 Corneal stromal cell culture in Corneal stromal stem cell medium (CSSCM)

Cell suspensions in CSSCM were seeded in three pre-coated T25-cm2 flasks and placed in a tissue culture incubator at 37 ºC and 5% CO2 in air. Medium was refreshed three times a week and cells were daily observed under a phase-contrast microscope to monitor their confluence. Cells were maintained and passaged as explained in section 2.1.2 of Chapter 2.

3.3.1.4 Total cell number and population doubling

To measure the ability of corneal stromal cells to expand when cultured in KDM or in CSSCM, cells were counted after each trypsinisation. After the first trypsinisation all cells were seeded into a T75-cm2 flask for Passage 1. For Passage 2 all cells were transferred into a T175-cm2. In the subsequent passages only 40000 cells were then seeded in a T175-cm2 flask and the rest discarded. This process was repeated until Passage 9 for cells cultured in CSSCM. Cells cultured in KDM showed high levels of cell death.

To represent the results, a graph with the total cell number over time was plotted. The total cell number was calculated by extrapolation.

To calculate the population doubling time, the cell number data points from the linear growth phase were used, and applied to the following equation:

Equation 1 – Population doubling time. tf = last time point, ti = initial time point, xf = cell number at last time point, xi = cell number at initial time point ln (2) 푃퐷푇 = (푡푓 − 푡푖) × 푥 ln ( 푓) 푥푖

Characterisation of normal corneal and limbal stromal cells cultured in CSSCM

3.3.2.1 Donor information

Table 3.3 - Donor information. Age, gender, and storage time are shown. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells.

Tissue ID Age (years) Eye Storage time (days)

NC1 54 Right 26

NC2 46 Right 29

NC3 84 Right 19

NC4 63 Left 20

NL1 63 Left 30

NL2 36 Right 20

NL3 84 Right 28

NL4 46 Left 29

3.3.2.2 Isolation of normal limbal stromal cells

Normal limbal stromal cells (NL) were obtained from limbal rims as described by Du et al [54] with few modifications. The whole corneas were washed and trephined as described in section 2.1.1. Limbal rims were then processed for isolation of NL as shown in Figure 3.1. Briefly, the remaining scleral tissue was removed, leaving a 2mm wide limbal ring. A thin layer of limbal surface of approximately 100μm depth was cut away from the remaining stroma using fine scissors and cut into small pieces (roughly 1mm2). These were further processed, cultured and passaged as in section 2.1.2. 77

Figure 3.1. Isolation of NL from the limbal rims. The limbal rim (A) was cut in half and the sclera removed (B). The anterior part of each strip, containing the epithelium and the anterior stroma, was then cut and separated from the posterior stroma with fine scissors (C, D). The strips were then cut into small pieces and incubated in collagenase overnight (E). Finally, the cells were plated on a culture dish and cultured in CSSCM in tissue culture incubator at 37ºC and 5% CO2. (F). Scale bar = 300μm

3.3.2.3 Cell proliferation - Total DNA

Total DNA quantitation kit (Sigma-Aldrich) was used to measure cell proliferation. This assay uses bisBenzamide H 33258 (Hoechst 33258), a fluorescent dye that is excited at 360 nm and emits a peak of fluorescence at 460nm. When bound to the AT rich regions of the double-stranded DNA, the fluorescence emitted is enhanced.

Cells were seeded at a density of 3000 cells/well on 12 well-plates (NuncTM) and cultured in a tissue culture incubator at 37°C and 5% CO2 in air. Three technical replicates for three biological replicates were prepared for each cell population (NC, NL). Total DNA was assessed at day 3, 5, and 7 according to manufacturer’s instructions. Briefly, cells were washed once with 1mL of DPBS and 800μL of sterile water were added to each well. Cells were then lysed with a series of three 60min:60min freeze-thaw cycles at - 80°C and 30°C. A standard curve was prepared with different DNA concentrations ranging from 10μg/mL to 0.165μg/mL. Standards (100μL) or samples (100μL) were added to a black 96 well-plate with flat bottom in duplicate (Corning®). Three extra wells were filled with 100μL of sample buffer to be used as a blank. Hoechst dye (100μL of a 4 μg/mL solution) was added to each well and fluorescence was read under a microplate reader (Tecan Saphire). The excitation wavelength was set at 360nm and emission at 78

460nm. The results were obtained in an Excel file, where the standard curve was drawn and the DNA concentrations were calculated.

At each time point, pictures were taken under a phase contrast microscope (Eclipse TS100, Nikon).

3.3.2.4 Real time qPCR and Western blot of NC and NL

The expression of putative CSSC markers (PAX6, CD90, CD73), keratocyte markers (KERA, LUM, ALDH1A1), proliferation (KI67), and apoptosis (TP53) markers were measured in NC and NL cultured in CSSCM.

Positive controls used in Real time qPCR included corneal epithelial stem cells (ECs) for PAX6, MSC for CD90 and CD73 and fresh central cornea for ALDH1A1, KERA and LUM.

Gene expression in NC, NL and MSC was normalized to ECs levels of expression. Statistical analysis could only be performed between NL and NC, as MSC and ECs data was obtained from only one biological replicate.

3.3.2.4.1 RNA extraction from MSC and ECs

The same preparation used to isolate limbal stromal (section 3.3.2.2) was also used to culture corneal epithelial cells. After plating the cell suspension, some epithelial colonies appeared in P0. These were left to grow in CSSCM for 7-10 days and RNA was extracted as explained in section 2.2.1. An RNA extract from MSC was kindly donated from Rita Pinho (using the same before mentioned protocol).

3.3.2.4.2 RNA extraction from whole corneas

The fresh central cornea was placed in RNA later (Qiagen) and processed as fresh as possible. Briefly, it was cut into small pieces (1mm2), placed in a 1,5mL eppendorf tube with 350μL of lysis buffer from the RNeasy mini kit and kept on ice. The sample was then sonicated for 1 minute at an amplitude of 23μm, until the tissue had been slightly dissociated. The tissue was left on a shaker on ice for 10 minutes to remove the foam produced during sonication. To help with the tissue homogenization, a disposable pellet pestle was used to grind the remaining corneal tissue after sonication in the same Eppendorf tube. The sample was centrifuged for 5 minutes at 13000xg and the supernatant was added to a Qiashredder column. From this point onwards, the RNA extraction was performed as explained in section 2.2.2.

Differentiation of normal corneal and limbal stromal cells into keratocyte-like cells in KDM

3.3.3.1 Cell culture conditions

To characterise the potential for differentiation of the NC and NL cells, these were differentiated towards a keratocyte-like phenotype was attempted using an adaptation of the protocol described by Du et al. [54]. NC and NL were passaged and counted before being seeded at a density of 1000cells/cm2 in a FNC coated T175-cm2 flask (NuncTM) in CSSCM and in 4-well chamber slides (VWR). Three hours later when the cells were attached, the flasks and slides were washed once with DPBS and then the medium was changed to KDM. Cells were kept for 3 weeks in tissue culture incubator at 37ºC and 5%

CO2 in air and the medium refreshed three times a week. To assess the importance of FGF for the guided differentiation of NC and NL into keratocytes-like cells; these were either cultured in KDM supplemented with FGF (10ng/mL) or in FGF-depleted KDM. For each donor, three flasks and one chamber slide were seeded for keratocyte differentiation in KDM-FGF+ and three more in KDM-FGF-.

As a control, NC and NL cells were also cultured to confluence in CSSCM to allow spontaneous cell density-mediated differentiation [174]. NC and NL were seeded at a density of 1000cells/cm2 in a T175-cm2 flask in CSSCM. Cells were cultured without being passaged for 5 days, and the medium refreshed at day 2.

NC and NL seeded at a density of 1000cells/cm2 in a T175-cm2 flask in CSSCM for only 24 hours will be addressed as basal conditions.

The following nomenclature and abbreviations will be used in the Results section.

+/- i = 1,2,3 refers to the time point in weeks when cells were collected, and F refers for the presence or absence of FGF in KDM. A schematic representation of the cell culture conditions is found in Table 3.4.

Normal corneal stromal cells – NC

+ - Normal corneal stromal keratocyte-like cells – NCKiF , NCKiF

Normal corneal stroma spontaneously differentiated cells – NCSD

Normal limbal stromal cells - NL,

+ - Normal limbal stromal keratocyte-like cells NLKiF , NLKiF

Normal limbal stroma spontaneously differentiated cells – NLSD

Table 3.4. – Culture conditions for NC and NL in CSSCM and KDM. Abbreviations CSSCM: corneal stromal stem cell medium, KDM: keratocyte differentiation medium, NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCKiF: Normal corneal stromal keratocyte-like cells, NCSD: Normal corneal stroma spontaneously differentiated cells, NLKiF: Normal limbal stromal keratocyte-like cells, NLSD: Normal limbal stroma spontaneously differentiated cells.

Cell culture Time (days)

medium 1 … 5 ... 7 … 14 … 21

CSSCM NC NCSD

NL NLSD

KDM FGF+ NCK1F+ NCK2F+ NCK3F+

NLK1F+ NLK2F+ NLK3F+

FGF- NCK1F- NCK2F- NCK3F-

NLK1F- NLK2F- NLK3F-

3.3.3.2 Gene expression

A total of 175000 NC and NL were used for RNA extraction. The whole flask of NCK, NLK, NCSD and NLSD was used to isolate RNA from these conditions. Photographs were taken of the cells in culture before being trypsinised and counted. Further cell lysis, RNA extraction and Real time qPCR were performed as explained in section2.2. qPCR was performed to assess the changes in gene expression when NC and NL were cultured in different media. Firstly, the expression of CSSC markers (CD90, CD73, and PAX6) was studied. Secondly, to assess the ability of NL and NC to successfully differentiate into keratocytes, expression levels of keratocyte markers (KERA, LUM, and ALDH1A1) were studied. Expression of αSMA was also measured to confirm that the cells were not differentiating into myofibroblasts. Finally, expression of proliferation (MKI67) and apoptosis (TP53, BAX, BCL2) markers was also assessed.

3.3.3.3 Protein production

The same culture settings stated in section 3.3.3.2. were used for protein extraction. Further cells lysis, protein quantification and western blotting were performed as explained in section 2.3.

Western blotting was performed to assess the changes in protein expression when NC and NL were cultured in different media. The production of CSSC markers (CD90 and CD73), keratocyte markers (KERA and LUM) and myofibroblast marker (αSMA) was studied.

3.3.3.4 Immunocytochemistry

A total of 4 chamber slides for each central (NC, NCK3F-, NCK3F+, NCSD) and limbal samples (NL, NLK3F-, NLK3F+, NLSD) were used for immunocytochemistry. The culture medium was aspirated, and the chamber slides were washed once with DPBS and fixed for 10 minutes in 4% (V/V) PFA in DPBS. Chambers were then washed 3 times with fresh DPBS for 5 minutes and stained as explained in section 2.4)

The assembly of αSMA was assessed with the polyclonal rabbit anti-αSMA primary antibody (Table 2.2) at a concentration of 1:100. A normal rabbit polyclonal IgG control (Cell signalling) was used as a negative control at the same concentration as the primary antibody. Secondary antibody used was Alexa Fluor ® 555 goat anti-rabbit (Life Technologies) at a concentration 1:500. A positive control for αSMA expression was prepared using a similar protocol described by Funderburgh et al. [80]. Briefly, NC were cultured for 3 days in DMEM containing 10% (V/V) FBS, 1% (V/V) P/S, and 10ng/mL of TGFβ to differentiate into myofibroblasts. All reagents were all provided by Sigma- Aldrich, unless otherwise specified. These were then stained as previously explained.

3.4 Results

Assessment of KDM and CSSCM as media to culture and expand normal corneal stromal cells

When cultured in CSSCM, a higher number of corneal stromal cells attached to the tissue culture plate 1 day after seeding, when compared to KDM (Figure 3.2). A higher number of dead floating cells could also be seen in the KDM flasks. The density of cells in CSSCM rapidly increased over the following days and at day 3 cells had reached confluence and had to be passaged. On the other hand, cells in KDM did not divide and were not able to be passaged. At day 15 density of cells was still higher in CSSCM than in KDM (Figure 3.2). The morphology of corneal stromal cells is similar when these are cultured in KDM or in CSSCM. A high percentage of cells show a small cellular body and high nucleus to cytoplasm ratio (Figure 3.2, black arrows). In KDM some cells present a dendritic morphology with small protrusions from the cell body (Figure 3.2, white arrow).

Cells cultured in CSSCM grew exponentially and reached the linear growth phase at day 22 (Figure 3.3). At that time the total number of cells ranged between 1x109 – 4x109. A variability between the growth rate of cells isolated from the different donors can be appreciated (Figure 3.3). The calculated cell doubling time was 24 (±5) hours. Cells cultured in KDM did not divide since the time of plating, so a growth curve could not be presented.

Figure 3.2. Representative images of cells isolated from the central cornea and plated in KDM or CSSCM at day 1 and day 15. Cells with dendritic morphology (white arrows) and with high nucleus to cytoplasm ratio (black arrows) were observed. Scale bars = 300µm. Abbreviations CSSCM: corneal stromal stem cell medium, KDM: keratocyte differentiation medium. 83

4.010 09 Donor 1 Donor 2 3.010 09 Donor 3 Median 2.010 09

1.010 09 Number of cells of Number

0.0 0 5 10 15 20 25 Time (days)

8.010 6 Donor 1 Donor 2 6.010 6 Donor 3 Median 4.010 6

2.010 6 Number of cells of Number

0 5 10 15 Time (days)

Figure 3.3. Growth curve of corneal stromal cells cultured in CSSCM. Each colour represents a biological donor and the black line represents the median. Cell doubling time was 24 (±5) hours. Corneal stromal cells did not grow in KDM. Abbreviations CSSCM: corneal stromal stem cell medium, KDM: keratocyte differentiation medium.

Summary of results:

• Corneal stromal cells do not proliferate when cultured in KDM.

• Between 1x109-4x109 corneal stromal cells were obtained after 22 days of corneal stromal cell culture in CSSCM. The population doubling time was 24 (±5) hours).

Normal corneal and limbal stromal cells present similar phenotype when cultured in CSSCM

3.4.2.1 NC and NL are successfully isolated and expanded in vitro

NC and NL were successfully isolated from corneal buttons and limbal rims, respectively. Cells attached to the culture plates and started proliferating within 24 hours after plating. Multiple healthy epithelial colonies were observed in the plates with cells isolated from the limbal rims (Figure 3.4, black arrow), whereas only few and small colonies were appreciated in the plates from corneal stroma isolations (Figure 3.4, red arrow). Stromal cells were present in both cultures (Figure 3.4, white arrows). Stromal cells were successfully trypsinised from the plates, leaving behind the epithelial colonies, as the former detach easier from the tissue culture surface.

Figure 3.4. Representative images of NC and NL at P0 cultured in CSSCM. Black and red arrows point to corneal epithelial cell colonies and white arrows point to stromal cells. Scale bars = 300µm

At passage 0, defined as cells being plated immediately after isolation, stromal cells cultures appeared to be a mixed population, with cells presenting a wide variety of morphologies (Figure 3.4). However, as the cells were passaged, a morphologically uniform population prevailed (Figure 3.5). Between passages 3 and 5, most cells presented a small, almost square-like cell body with a high nucleus to cytoplasm ratio (Figure 3.5, white arrows). A few cells cultured in the same conditions acquired a more elongated morphology (Figure 3.5, black arrows). No differences in cell morphology were observed between NC and NL nor between any of the donors (Figure 3.5).

Figure 3.5. Representative images of NC and NL cultured in CSSCM. Images were taken between passages 3 and 5. Most cells were small with a square-like morphology and with high nucleus to cytoplasm ratio (white arrows). A small portion of the cells presented more elongated bodies (black arrows). Images were taken under a phase contrast microscope. n=3. Scale bars =300µm. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells.

3.4.2.2 NC and NL have similar proliferation capacity when cultured in 2D in CSSCM

NC and NL were cultured in CSSCM over 7 days in a 12 well-plate and observed under a phase contrast microscope (Figure 3.6). At day 3, cells were relatively sparse and most of them maintained the small square-like morphology described in section 3.4.2.1. As

cells divided and became more confluent, they elongated and became more spindle shaped (Figure 3.6A).

Quantification of total DNA at day 3, day 5 and day 7, confirmed that there were no significant differences in the proliferation capacity between NC and NL (Figure 3.6B, p>0.05). Both NC and NL multiplied by a factor of 6 the amount of DNA present in the wells between day 3 and day 7 (p<0.001). NC and NL did not show contact inhibition from day 5 to day 7, with the amount of DNA doubling in that time period (Figure 3.6).

8000 B) *** NC 6000 NL

4000

2000 Total DNA (ng) TotalDNA

0 3 5 7 Time (days)

Figure 3.6. Proliferation of NC and NL cultured in CSSCM over a period of 7 days. (A) Representative images of the cells in culture at 3, 5 and 7 days. Scale bars= 300µm. (B) Quantification of cell proliferation measured by total DNA. Data shows the mean ± SD, n=3. Two-way repeated measurements ANOVA was used for statistical analysis. ***<0.001. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells.

3.4.2.3 NC and NL show similar profile of gene expression in tested markers

There were no significant differences in PAX6, CD90 and CD73 gene expression levels between NC and NL (Figure 3.7A, p>0.05). Expression levels of PAX6 in NC and NL were very low, with Real Time qPCR Ct values appearing after 34-36 cycles (Table A.8.1). NC and NL had higher expression values of CD90 and CD73 than PAX6. Noteworthy, NC, NL, and MSC showed similar gene expression levels of the MSC associated makers (CD90 and CD73).

There were no significant differences in KERA, LUM and ALDH1A1 gene expression levels between NC and NL (Figure 3.7B, p>0.05). KERA was not expressed in any of the conditions (Black arrow in Figure 3.7B). LUM and ALDH1A1 were expressed in lower levels in NC and NL than in ECs.

No statistically significant differences were found in KI67 (proliferation marker) and TP53 (apoptosis marker) gene expression levels between NC and NL (Figure 3.7C, p>0.05).

A) 100

10 MSC 1 NC 0.1 NL 0.01 0.001 0.0001 0.00001

Relative gene expression Relative PAX6 CD90 CD73 (Normalised to epithelial cells) epithelial to (Normalised

B) 10 MSC 1 NC NL 0.1

0.01

0.001 

0.0001 Relative gene expression Relative KERA LUMI ALDH1A1 (Normalised to epithelial cells) epithelial to (Normalised C) 10 MSC NC NL

0.1 Relative gene expression Relative KI67 TP53 (Normalised to epithelial cells) epithelial to (Normalised

Figure 3.7. Relative gene expression levels of (A) putative CSSC markers (PAX6, CD90, CD73), (B) keratocyte markers (KERA, LUMI, ALDH1A1), and (C) proliferation (KI67) and apoptosis markers (TP53) in MSC, NC and NL. Data is represented as the median of the obtained data points for each condition, n=1 for ECs and MSC, n=3 biological replicates for NC and NL. Dashed line represents the gene expression of ECs, to which the other conditions are compared to. Arrow represents gene expression not detected. Abbreviations MSC: Mesenchymal stem cells NC: Normal corneal stromal cells, NL: Normal limbal stromal cells.

3.4.2.4 NC and NL show similar profile of protein production in tested markers

To confirm and compare the results obtained in the previous section, the production of selected proteins was tested using Western blotting. The putative CSSC markers that were highly expressed at the mRNA level, CD90 and CD73, were chosen. There was a visible trend of lower CD90 and CD73 expression by NL, although not statistically significant (Figure 3.8, p>0.05). AA) 2.0 NC NL 1.5

1.0

0.5

0.0

CD90 CD73 B) (A.U.) expression GAPDH to Relative

Figure 3.8. Protein production levels of putative CSSC markers (CD90 and CD73) by NC and NL. (A) Densitometry quantification of each protein normalized against the expression of GAPDH. Data is represented as the median of n=3 biological replicates. Mann Whitney test, p>0.05 (not significant). (B) Western blots representative of the 3 donors. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells A.U. arbitrary units.

Summary of results:

• NC and NL were successfully isolated and cultured in CSSCM.

• NC and NL did not show differences in morphology when cultured in CSSCM.

• NC and NL showed similar proliferation capacity in culture.

• There were no significant differences in gene and protein expression of tested putative CSSC, keratocyte and proliferation markers between NC and NL cultured in CSSCM.

• When cultured in CSSCM NC and NL showed a gene expression profile similar to MSC.

Normal corneal and limbal stromal cells have a similar capacity to differentiate into keratocytes-like cells in KDM

3.4.3.1 NC and NL change their morphology during keratocyte differentiation

NC and NL cultured sparsely in CSSCM prior to any differentiation process had the same appearance as previously described in section 3.4.2.1. Cells left in CSSCM for further 5 days without being passaged (NCSD and NLSD), started proliferating, reaching 90% confluence and changed their morphology (Figure 3.9A). Moreover, they lost their square-like shape and acquired an elongated, polarized body, forming unorganized networks.

As shown in Figure 3.9B, NCK and NLK also became elongated. However, the number of cells was much lower when compared to NCSD and NLSD. Interestingly, there was an apparent difference in cell behaviour between NCK and NLK when they were treated in the presence or absence of FGF. NCK and NLK cultured without FGF attained higher numbers of cells, especially at week 2 and 3, and presented thinner bodies than NCK and NLK cultured in FGF, respectively. NCK and NLK cultured in FGF depleted KDM showed high cell alignment at week 3. A proportion of NCK and NLK cultured in FGF presented a bright body, at week 1 and 2. These cells had broader bodies compared to cells treated in the absence of FGF.

No apparent differences in cell morphology between central and limbal cells in any of the studied conditions could be appreciated.

Figure 3.9. Representative images of NC and NL cultured under different conditions. (A) Images of cells cultured in CSSCM at day 1 and day 5. (B) Images of cells differentiating into keratocyte-like cells during a period of 3 weeks using KDM supplemented or depleted of FGF. Scale bars = 300 µm. Abbreviations CSSCM: corneal stromal stem cell medium, KDM: keratocyte differentiation medium, NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD: Normal corneal stroma spontaneously differentiated cells, NLSD: Normal limbal stroma spontaneously differentiated cells, NCKiF: Normal corneal stromal keratocyte-like cells, NLKiF: Normal limbal stromal keratocyte-like cells.

3.4.3.2 NC and NL change their gene expression and protein production profiles under conditions shown to promote keratocyte differentiation

NC and NL seeded at a density of 1000cells/cm2 in a T175-cm2 flask in CSSCM for only 24 hours will be addressed as basal conditions (i.e. the phenotype that was used to start the keratocyte or spontaneous differentiation).

3.4.3.2.1 Gene expression profiling of CSSC markers during directed differentiation of NC and NL towards a keratocyte phenotype

Expression of CD90 was upregulated during keratocyte differentiation

As shown in Figure 3.10A, when prompted to differentiate towards a keratocyte phenotype, normal corneal stromal cells (NCK) showed a significant increase (p<0.05) in CD90 expression. This remained true regardless of the presence or absence of FGF in the differentiating medium KDM (Figure 3.10A). The same trends were found for normal limbal stromal cells (NLK; Figure 3.10B).

However, neither corneal nor limbal stromal cells allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in CD90 expression.

Expression of CD73 did not change during keratocyte differentiation

As shown in Figure 3.10C, expression levels of CD73 did not significantly change when normal corneal stromal cells were differentiated towards a keratocyte phenotype (NCK). This remained true regardless of the presence or absence of FGF. In contrast, CD73 was upregulated when limbal corneal stromal cells were differentiated in KDM in the presence of FGF for 3 weeks but did not significantly change in the other KDM conditions (NLK; Figure 3.10D). However, the trend of CD73 expression fold-change was very similar between NCK and NLK. Noteworthy, the Fold-change of CD73 (approximately 2X-Fold change) upregulation was much lower than that of CD90 (approximately 10X- Fold change).

Additionally, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate in CSSCM (NCSD, NLSD) showed any change in expression levels of CD73 (Figure 3.10C, Figure 3.10D).

Expression of PAX6 was very low across all conditions during keratocyte differentiation

Levels of PAX6 expression were very low in all conditions. Ct values for NC and NL were found around 30-35 cycles (Table A.8.2). In all keratocyte like (NCK, NLK) and spontaneously differentiated (NCSD, NLSD) cells the expression of PAX6 was even lower than in the basal conditions or not detected. This could suggest that PAX6 expression was downregulated in these conditions, but as the expression at the basal condition was so low, no graphs were prepared, and statistical analysis could not be done.

B) A) CD90 CD90 100 100 ** * * ** * * * ** 10 10

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLSD No FGF FGF No FGF FGF

C) D) CD73 CD73 100 100

10 10 *

1 1

(normalised to NC) to (normalised

(normalised to NL) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

Figure 3.10. Relative gene expression levels of putative CSSC markers (CD90 and CD73) by corneal (A, C) and limbal (B, D) stromal cells under different culture conditions as assessed by Real-Time qPCR. Results are shown for NC and NL cultured in KDM supplemented or depleted of FGF at different time points (NCKi and NLKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (NCSD and NLSD, respectively). Dashed line represents the gene expression for the basal conditions NC and NL. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the initial condition. *p<0.05, **p<0.01. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

3.4.3.2.2 Protein production profiling of CSSC markers during directed differentiation of NC and NL towards a keratocyte phenotype

Production of CD90 was upregulated during keratocyte differentiation

Western blot data confirmed the results obtained by Real Time qPCR. As shown in Figure 3.11A, when prompted to a keratocyte phenotype, corneal stromal cells (NCK) significantly upregulated (p<0.05) production of CD90 protein. This was only true when NCK were cultured in KDM without FGF. The exact same trends were observed when limbal stromal cells (NLK, Figure 3.11B). Interestingly, there was no significant upregulation of CD90 when NCK or NLK were cultured in KDM the presence of FGF, although real time qPCR showed differently. As shown in the WB membranes, an upregulation of CD90 production could also be observed when NCK and NLK were cultured in the presence of FGF. However, in the densitometry quantification, this upregulation was not significant.

In addition, there were no significant differences in CD90 production when corneal and limbal stromal cells were allowed to spontaneously differentiate in CSSCM (NCSD, NLSD, respectively). (Figure 3.11A, Figure 3.11B, p<0.05).

Production of CD73 did not change during keratocyte differentiation

As shown in Figure 3.11C, production of CD73 did not change significantly when corneal stromal cells were differentiated towards a keratocyte phenotype (NCK). This remained true regardless of the presence or absence of FGF. The same trends were shown for limbal stromal cells (NLK; Figure 3.11D). No changes in CD73 production were observed when normal or limbal stromal cells were allowed to spontaneously differentiate in CSSCM (NCSD, NLSD) (Figure 3.11C, Figure 3.11D).

A) CD90 B) CD90 3 * 3 *

2 2

1 1

0 0 Relative expression GAPDH to Relative NC expression GAPDH to Relative NL NCSD NLSD NCK3F- NCK3F+ NLK3F- NLK3F+ CD90 (17KDa)

GAPDH (36KDa)

C) CD73 D) CD73 3 3

2 2

1 1

0 0 Relative expression GAPDH to Relative NC expression GAPDH to Relative NL NCSD NLSD NCK3F- NCK3F+ NLK3F- NLK3F+ CD73 (63KDa)

GAPDH (36KDa)

Figure 3.11. Production levels of putative CSSC markers (CD90 and CD73) by corneal (A, C) and limbal (B, D) stromal cells under different culture conditions as assessed by Western blot. Results are shown for the basal condition NC and NL, for NC and NL cultured for 3 weeks in KDM supplemented (NCK3F+, NLK3F+, respectively) or depleted of FGF (NCK3F-, NLK3F-, respectively), and for NC and NL cultured for 5 days in CSSCM (NCSD and NLSD, respectively). Graphs showing the densitometry quantification normalised to GAPDH. Data is represented as the median of n=3 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

3.4.3.2.3 Gene expression profiling of keratocyte markers during directed differentiation of NC and NL towards a keratocyte phenotype

Expression of LUM was upregulated during keratocyte differentiation

As shown in Figure 3.12A, when prompted to differentiate towards a keratocyte phenotype, corneal stromal cells (NCK) showed a significant increase (over 1000X Fold increase, p<0.05) in LUM expression. This remained true whether NCK were differentiated in the presence or absence of FGF (Figure 3.12A). The same trends were found for limbal stromal cells (NLK; Figure 3.12B). A very clear trend shows that LUM gene expression gradually increased with time in NCK and NLK (Figure 3.12).

However, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in LUM expression.

Expression of KERA was upregulated during keratocyte differentiation

As shown in Figure 3.12C, when prompted to differentiate towards a keratocyte phenotype, corneal stromal cells (NCK) significantly upregulated KERA expression (over 100X Fold increase, p<0.05). This remained true both when NCK were differentiated in KDM in the presence or absence of FGF. The same trends were found for limbal stromal cells (NLK; Figure 3.12D). A very clear trend shows that KERA gene expression gradually increased with time in NCK and NLK (Figure 3.12C, Figure 3.12D).

However, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in KERA expression.

Expression of ALDH1A1 did not significantly change during keratocyte differentiation

As shown in Figure 3.12E, expression levels of ALDH1A1 did not significantly change when corneal stromal cells (NCK) were differentiated towards a keratocyte phenotype. The same trends were found for limbal stromal cells (NLK; Figure 3.12D). However, the same trend of upregulation seen in LUM and KERA could also appreciated for ALDH1A1 in both NCK and NLK, particularly at week 2 and 3.

Additionally, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate in CSSCM (NCSD, NLSD) showed any change in expression levels of ALDH1A1 (Figure 3.12E, Figure 3.12F).

LUM LUM 100000 A) 100000 B) 10000 * ** * ** 10000 * ** * ** 1000 1000 100 100 10 10 1

1 NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

C) KERA D) KERA 1000 * * * 1000 * * ** * 100 100

10 10

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

E) ALDH1A1 F) ALDH1A1 1000 1000

100 100

10 10

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

Figure 3.12. Relative gene expression levels of keratocyte markers (LUM, KERA and ALDH1A1) in corneal (A, C, E) and limbal (B, D, F) stromal cells under different culture conditions as assessed by Real-Time qPCR. Results are shown for NC and NL cultured in KDM supplemented or depleted of FGF at different time points (NCKi and NLKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (NCSD and NLSD, respectively). Dashed line represents the gene expression for the basal conditions NC and NL. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the initial condition. *p<0.05, **p<0.01. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

3.4.3.2.4 Protein production profiling of keratocyte markers during directed differentiation of NC and NL towards a keratocyte phenotype

Production of LUM was not upregulated during keratocyte differentiation

Interestingly, the LUM protein production was not consistent with the Real-time qPCR results. As shown in Figure 3.13A, when prompted towards a keratocyte phenotype, corneal stromal cells (NCK) did not significantly change LUM production, compared to the basal conditions. This remained true regardless of the presence or absence of FGF. The same trend was observed in limbal stromal cells (NLK, Figure 3.13B).

Seemingly, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in LUM production.

Production of KERA was not significantly upregulated during keratocyte differentiation

Corneal stromal cells prompted to differentiate to a keratocyte phenotype (NCK), showed an upregulation of KERA production, as indicated by more intense bands in the western blot membranes (Figure 3.13A). However, this upregulation was not statistically significant (p>0.05). The same trend could be observed in limbal stromal cells (Figure 3.13B).

However, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in KERA production.

A) LUM B) LUM 2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0 Relative expression GAPDH to Relative NC expression GAPDH to Relative NL NCSD NLSD NCK3F- NLK3F- NCK3F+ NLK3F+ LUM (46KDa) GAPDH (36 KDa)

C) D) KERA KERA 2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0 Relative expression GAPDH to Relative NC expression GAPDH to Relative NL NCSD NLSD NCK3F- NLK3F- NCK3F+ NLK3F+ KERA (41 kDa) GAPDH (36kDa)

Figure 3.13. Production levels of keratocyte markers (LUM and KERA) by corneal (A, C) and limbal (B, D) stromal cells under different culture conditions as assessed by Western blot. Results are shown for the basal condition NC and NL, for NC and NL cultured for 3 weeks in KDM supplemented (NCK3F+, NLK3F+, respectively) or depleted of FGF (NCK3F-, NLK3F-, respectively), and for NC and NL cultured for 5 days in CSSCM (NCSD and NLSD, respectively). Graphs showing the densitometry quantification normalised to GAPDH. Data is represented as the median of n=3 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

100

3.4.3.2.5 Gene expression and protein production profiling of myofibroblast markers during directed differentiation of NC and NL towards a keratocyte phenotype

Expression of αSMA is upregulated during keratocyte differentiation

As shown in Figure 3.14A, when prompted to differentiate towards a keratocyte phenotype, corneal stromal cells (NCK) showed a significant increase (around 10X-Fold change p<0.05) in αSMA expression compared to the basal condition. This was true regardless of the presence or absence of FGF. The same trends were found for limbal stromal cells (NLK; Figure 3.14B).

In contrast, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in αSMA expression (Figure 3.14A, Figure 3.14B). B) A) SMA SMA 100 100 * * * * * 10 10

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

Figure 3.14. Relative gene expression levels of the fibrosis marker αSMA (ACTA2) in corneal (A) and limbal (B) stromal cells under different culture conditions as assessed by Real-Time qPCR. Results are shown for NC and NL cultured in KDM supplemented or depleted of FGF at different time points (NCKi and NLKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (NCSD and NLSD, respectively). Dashed line represents the gene expression for the basal conditions NC and NL. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the initial condition. *p<0.05. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

101

Production of αSMA was significantly upregulated during keratocyte differentiation

Results of αSMA protein production obtained by Western blot were consistent with those of Real Time qPCR. Corneal stromal cells that differentiated to a keratocyte phenotype (NCK), showed a significant upregulation of αSMA production (Figure 3.15A). This was true only when NCK were differentiated in KDM in the absence of FGF. The same trend could be observed in limbal stromal cells (Figure 3.15B).

When central and limbal stromal cells were allowed to spontaneously differentiate in CSSCM (NCSD, NLSD) there was no significant differences observed, compared to the basal conditions (Figure 3.15A, Figure 3.15B).

Immunofluorescence showed that αSMA was produced by the majority of central and limbal corneal stromal cells, indicated by diffused red staining in Figure 3.16. However, only a small proportion of cells that were prompted towards a keratocyte phenotype (NCK and NLK), had assembled αSMA into stress fibres. This remained true regardless of the presence or absence of FGF. IgG negative controls did not show staining (Appendix, Figure A.8.3). No differences were observed between limbal and central corneal stromal cells regarding their production and assembly of αSMA.

102

A) B) SMA SMA 3 * 3 *

2 2

1 1

0 0 Relative expression GAPDH to Relative NC expression GAPDH to Relative NL NCSD NLSD NCK3F- NLK3F- NCK3F+ NLK3F+ αSMA (42 kDa) GAPDH

(36 KDa)

Figure 3.15. Production levels of the fibrosis marker αSMA by corneal (A) and limbal (B) stromal cells under different culture conditions as assessed by Western blot. Results are shown for the basal condition NC and NL, for NC and NL cultured for 3 weeks in KDM supplemented (NCK3F+, NLK3F+, respectively) or depleted of FGF (NCK3F-, NLK3F-, respectively), and for NC and NL cultured for 5 days in CSSCM (NCSD and NLSD, respectively). Graphs showing the densitometry quantification normalised to GAPDH. Data is represented as the median of n=3 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

103

Corneal Limbal

Figure 3.16. Production and assembly of αSMA by corneal and limbal stromal cells, as assessed by immunofluorescence. Red: Anti-αSMA, Green: Phalloidin, Blue: DAPI. Scale bars: 300µm. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCK3F Normal corneal stromal keratocyte-like cells, NLK3F: Normal limbal stromal keratocyte- like cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

104

3.4.3.2.6 Gene expression profiling of the proliferation marker KI67 during directed differentiation of NC and NL towards a keratocyte phenotype

Expression of KI67 is downregulated during keratocyte differentiation

As shown in Figure 3.17C, the expression levels of KI67 were significantly downregulated when corneal stromal cells were differentiated towards a keratocyte phenotype (NCK) (100X- and 1000X-Fold Decrease, p<0.05). This remained true regardless of the presence or absence of FGF. The same trends could be observed for limbal stromal cells (NLK; Figure 3.17B).

KI67 expression did not significantly change when corneal or limbal stromal cells were allowed to spontaneously differentiate for 5 days in CSSCM (NCSD, NLSD).

A) KI67 B) KI67 10 10

1 1 ** * * 0.1 * * 0.1 0.01 0.01

0.001 0.001

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.0001 0.0001

NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLSD No FGF FGF No FGF FGF

Figure 3.17. Relative gene expression levels of KI67 in corneal (A) and limbal (B) stromal cells under different culture conditions as assessed by Real-Time qPCR. Results are shown for NC and NL cultured in KDM supplemented or depleted of FGF at different time points (NCKi and NLKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (NCSD and NLSD, respectively). Dashed line represents the gene expression for the basal conditions NC and NL. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the initial condition. *p<0.05, **p<0.01. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

The protein expression levels of KI67 was not assessed to confirm the Real-Time qPCR results as a decline in cell proliferation could clearly be seen just by direct observation under the microscope. This is reflected in Figure 3.9B in all NCK and NLK conditions, compared to NCSD and NLSD.

105

3.4.3.2.7 The expression profiling of apoptosis markers during directed differentiation of NC and NL towards a keratocyte phenotype

Expression of TP53 was significantly upregulated during keratocyte differentiation

As shown in Figure 3.18A, when prompted to differentiate towards a keratocyte phenotype, corneal stromal cells (NCK) showed a significant increase (over 5X Fold increase, p<0.05) in TP53 expression. This remained true whether NCK were differentiated in the presence or absence of FGF (Figure 3.18A). The same trends were found for limbal corneal stromal cells (NLK; Figure 3.18B).

However, neither corneal nor limbal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (NCSD, NLSD) showed an increase in TP53 expression (Figure 3.18A, Figure 3.18B).

Expression of BAX did not change during keratocyte differentiation

The expression levels of BAX did not change across any of the conditions in corneal nor limbal stromal cells (p>0.05, Figure 3.18C and Figure 3.18D).

Expression of BCL2 did not change during keratocyte differentiation

BCL2 gene expression levels were significantly downregulated when corneal stromal cells were differentiated towards a keratocyte phenotype in the presence of FGF (NCK, Figure 3.18A). However, when limbal corneal stromal cells were differentiated towards a keratocyte phenotype BCL2 was not significantly downregulated (Figure 3.18B).

In addition, BCL2 expression did not significantly change when normal or limbal corneal stromal cells were allowed to spontaneously differentiate in CSSCM (NCSD, NLSD).

106

A) TP53 B) TP53 10 * * ** 10 * * *

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

C) BAX D) BAX 10 10

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

E) BCL2 F) BCL2 10 10

1 1

(normalised to NL) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD NLK1 NLK2 NLK3 NLK1 NLK2 NLK3 NLSD No FGF FGF No FGF FGF

Figure 3.18. Relative gene expression levels of apoptosis markers (TP53, BAX, and BCL2) in corneal (A, C, E) and limbal (B, D, F) stromal cells under different culture conditions as assessed by Real-Time qPCR. Results are shown for NC and NL cultured in KDM supplemented or depleted of FGF at different time points (NCKi and NLKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (NCSD and NLSD, respectively). Dashed line represents the gene expression for the basal conditions NC and NL. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the initial condition. *p<0.05, **p<0.01. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

107

Summary of results:

• Corneal and limbal stromal cells cultured in KDM and allowed to spontaneously differentiate in CSSCM acquired an elongated spindle shaped morphology.

• Corneal and limbal stromal cells cultured in KDM proliferated less than when cultured in CSSCM.

• The morphology of corneal and limbal stromal cells was different after differentiation in KDM for 3 weeks, regardless of the presence or absence of FGF.

• Corneal and limbal stroma cells cultured in KDM have different gene expression profiles than when cultured in CSSCM:

o Expression levels of CD90, KERA, LUM, αSMA and TP53 were upregulated.

o Expression levels of KI67 were downregulated.

• Corneal and limbal stroma cells cultured in KDM without FGF have a different protein expression profile than when cultured in CSSCM:

o Expression levels of CD90, αSMA were upregulated.

• Corneal and limbal stroma cells cultured confluent in CSSCM have a similar expression profile for the genes tested to when cultured sparsely in the same medium.

108

3.5 Discussion

Data presented in this chapter showed that corneal stromal cells were successfully expanded in CSSCM whereas they remained quiescent in KDM. Unexpectedly, corneal and limbal stromal cells cultured in CSSCM presented similar phenotype and both could be differentiated into keratocyte-like cells.

Normal corneal stromal cells were successfully expanded in CSSCM

Due to tissue shortage, the first aim of this chapter was to determine if it would be possible to sufficiently expand the number of stromal cells isolated from central corneas. Supporting the 1st hypothesis, KDM was revealed not to be a suitable culture medium to do so, as these cultures showed no proliferation, consistent with previously published results [35, 79]. KDM is a serum-free medium suitable for keratocytes culture if one aims to maintain their native phenotype [194]. In contrast, and supporting the 2nd hypothesis, the results here presented showed that corneal stromal cells cultured in CSSCM proliferate and expand in number (approximately 22 population doublings while still in the exponential phase of growth). Similar results have been reported in limbal stromal stem cells cultured in CSSCM, which have been grown for up to 100 population doublings [54].

Corneal stromal cells showed a small square-like morphology both when cultured in CSSCM and in KDM, which resembles the stem cell morphology found in corneal stromal stem cells (CSSC) cultured in CSSCM [174]. A small cell population cultured in KDM presented a dendritic morphology with small protrusions, representative of cultured keratocytes [185]. However, these cells did not form any network, a feature seen in keratocytes in vitro [35, 185] and in vivo [197], likely due to the low cell density at the time of seeding and the subsequent low proliferative potential. Corneal stromal cells cultured in CSSCM attached to the tissue culture plate, with some cells presenting spindle shaped cell bodies, similar to corneal fibroblasts cultured in 10% FBS-containing media [79, 183]. After passaging and expanding the cells in CSSCM, the majority lost the fibroblastic morphology and acquired a small square-like body. These observations suggest that CSSCM is a suitable medium to expand corneal stromal cells in vitro. In conclusion, results here supported the hypotheses that KDM was not a suitable culture medium to yield enough cell numbers and that CSSCM promoted corneal stromal cell proliferation.

109

Normal corneal and limbal stromal cells have similar phenotypes when cultured in CSSCM

Unexpectedly, the results here presented rejected the 3rd hypothesis that corneal and limbal corneal stromal cells present different phenotype even when cultured in the same conditions. Namely, there were no significant differences between the phenotypes of central corneal (NC) and limbal (NL) stromal cells cultured in CSSCM at low density. Different populations of cells are thought to reside in the central cornea and the limbus in vivo. Corneal stromal stem cells (CSSC) have been reported to be present in the limbus (here referred to as NL) [54, 57] while the central corneal stroma is populated by keratocytes [22]. Interestingly, both NC and NL cultured in CSSCM presented the same small square-like morphology with high nucleus to cytoplasm ratio, which is a feature of CSSC [174]. Previously, CSSC were isolated from the limbal stroma as a side population that expressed stem cell related markers (PAX6, CD90, CD73) and also exhibited multipotent differentiation potential [54, 174]. Despite not being expressed in keratocytes, PAX6 is a marker of eye development and therefore it identifies ocular progenitor cells [57]. In the data here presented, Real-Time qPCR results showed that NC and NL had very low PAX6 expression. However, it has been previously reported that PAX6 expression in CSSC decreases with cell passaging [198] (and personal communication by Martha Funderburgh, ARVO Meeting 2019). In this project, gene expression of NC and NL was tested at passages 5-7 and therefore PAX6 expression could have already decreased to undetectable values. CD90 and CD73 belong to the list of genes whose expression is a minimal requirement for cells to be considered as MSC [199]. CD90 and CD73 were expressed by NC and NL at a similar level as control MSC suggesting that these two cell populations may share MSC features under the conditions tested.

NC and NL did not express KERA and had low expression of LUM and ALDH1A1. Supporting studies also showed low expression of typical keratocyte markers in CSSC in vitro [54]. The expression of KI67 was higher in NC and NL than in MSC and ECs. However, only one MSC and ECs cell line was used for this study so conclusions cannot be drawn from this comparison. Medium supplemented with FBS and ITS has previously been shown to promote proliferation of corneal stromal cells [180], which supports the results found herein. Nevertheless, it is interesting that NC and NL have similar proliferation rates, both demonstrated by total DNA and by KI67 expression. High proliferation is typical of stem cells when cultured in vitro [54], and as previously explained a stem cell population has been described in the limbus (where NL are derived from), but not in the central cornea (where NC are derived from). NC and NL presenting 110

the same proliferation capacity does not support the hypothesis that NC and NL behave differently when cultured in CSSCM.

In a recent study, Sidney and colleagues isolated and cultured keratocytes (from central cornea) in M199 supplemented with 20% FBS, a medium used to culture MSC. They also found an increase in cell proliferation, upregulation of CD90 and CD73 and a downregulation of KERA [181]. This is consistent with the results presented herein and with the hypothesis that central corneal stromal cells can be expanded in vitro when cultured in a suitable media. Cells cultured in such conditions express the putative MSC markers and do not express typical keratocyte markers. This shift in phenotype suggested that corneal stromal cells acquire stem cell-like properties when cultured in CSSM and that they present a very plastic behaviour in vitro. Therefore, the next step was to assess if they can be reverted to a keratocyte phenotype.

Normal corneal and limbal stromal cells share similar ability to differentiate into keratocyte-like cells

Unexpectedly, and not supporting our hypothesis, both central and limbal stromal cells had the same ability to differentiate into keratocyte-like cells. When cultured in KDM or left to reach confluence in CSSCM, NC and NL acquired an elongated morphology. Other studies suggest that cells can acquire either a dendritic [183] or a spindle-shaped [180, 181] morphology when cultured in the serum-free medium supplemented with FGF-3, TGFβ3 and retinoic acid. In the presence of PDGF-BB, cells have shown an even longer spindle shaped morphology [180]. The latter results are consistent with the data presented here, in which cells acquire an elongated morphology when cultured to confluence in CSSCM that contains PDGF. Moreover, there were apparent differences in cell morphology and density between NCK and NLK when cultured with FGF or without FGF. Controversially, two published studies suggested that serum-free medium supplemented with FGF promotes a more fibroblastic morphology [180, 184], whereas another claimed that FGF does not alter cell morphology [79]. In the results here presented, KDM without FGF promoted a more elongated fibroblastic morphology. Although, the main differences arose from cell density. Unexpectedly, the microscopy images showed that normal and limbal stromal cells differentiated towards a keratocyte phenotype in the absence of FGF had higher cell density than when FGF was present suggesting that the former proliferated more. However, no differences in KI67 gene expression were observed when cells were cultured in the presence or absence of FGF. Interestingly, FGF has been previously shown to promote proliferation of keratocytes

111

[184, 186], nevertheless the results here presented suggest that in these conditions it did not affect cell proliferation.

It has been previously reported that when CSSC differentiate into keratocytes, they lose their progenitor markers expression, namely PAX6, CD90 and CD73 [57]. PAX6 was expressed at a very low level in the basal conditions when corneal and limbal stromal cells were cultured in CSSCM (NC and NL, respectively). Therefore, no statistical analysis comparing NC and NL with NCK and NLK cultured in KDM could be performed. Unexpectedly, CD90 was upregulated in NCK and NLK at all time points and its expression increased overtime. Similar results have been reported in recent studies, in which corneal stromal cells have been expanded in serum containing medium and later differentiated into keratocytes-like cells [180, 181]. Instead of being downregulated, CD90 maintained its expression across all serum-free treated cells, suggesting that they only achieved a partial keratocyte differentiation [181]. Apart from being a MSC marker, CD90 has also been shown to be expressed in corneal fibroblasts and myofibroblasts [200], which suggests that some cells might be differentiating into myofibroblasts in our cultures. Similar results were shown for CD73, which was not downregulated when NC and NL were differentiated into keratocyte-like cells, instead, its expression did not change across conditions. At the mRNA level, CD73 was significantly upregulated in NLK3 but the same was not shown at the protein level. Fernandez-Perez and Ahearne have also shown higher CD73 gene expression in serum-deprived cells when compared to those cultured in serum and PDGF containing media [180], although the differences did not appear statistically significant (p>0.05).

Keratocan gene and protein expression was highly upregulated in NCK and NLK, both in the presence and absence of FGF, suggesting that cells were differentiating into keratocytes [57, 185]. Keratocan was not upregulated in NCSD and NLSD, consistent with results that show no expression of keratocan in cells cultured in serum [180]. Lumican on the other hand, was upregulated at the mRNA level but the same was not observed at the protein level. Lumican regulates the expression of keratocan at the transcriptional level [201]. Blocking lumican at the mRNA level decreased the expression of keratocan, and adding a lumican-expressing mini gene was able to restore the expression of keratocan [201]. In the present study, lumican mRNA was upregulated in NCK and NLK, which could be acting to activate the expression of keratocan, hence this could be why the mRNA levels were maintained at high levels. However, lumican protein might have already been synthesised and secreted into the medium by NCK and NLK, lowering the levels of intracellular lumican to be detected in the Western blot. Although a trend could be observed, ALDH1A1 was not significantly upregulated in NCK and NLK, 112

consistent with results showed by Barryhill et al in which ALDH expression could not be restored after expansion of corneal stromal cells in serum [183]. Upregulation of lumican and keratocan suggest that NCK and NLK were acquiring a keratocyte phenotype.

Unexpectedly, gene and protein expression of αSMA was also upregulated in NCK and NLK when compared to NC and NL. Published results differ regarding the expression of αSMA in cells cultured serum-free or in the presence of serum. Beales et al reported that cells cultured in serum-free medium downregulate αSMA and serum-containing medium promoted αSMA upregulation [79], whereas Fernandez-Perez et al showed the opposite [180]. In both studies FGF was shown to inhibit the expression of αSMA [79, 180]. The same results were not observed here at the mRNA level. No differences in αSMA gene expression between NCK and NLK cultured with or without FGF were appreciated. However, at the protein level, expression of αSMA was only significantly upregulated when corneal and limbal stromal cells were differentiated in KDM without FGF, consistent with studies showing that FGF inhibits αSMA expression [79].

It is likely that the NCK and NLK cultures represented a heterogeneous population of cells with regards to their different stages of differentiation (keratocyte-like cells, fibroblasts and myofibroblasts). Thus, the keratocyte-like cells upregulated the expression of keratocan and lumican and downregulated MKI67, while other cells that were progressing towards a fibroblast or myofibroblast phenotype upregulated αSMA and CD90. This is consistent with other published studies in which a small percentage of cells in serum-free medium express αSMA, while the majority remain negative [181]. Immunocytochemistry results here presented supported this hypothesis, by showing that only a small proportion of the corneal stromal cells prompted towards a keratocyte phenotype (NCK and NLK) were actually assembling the αSMA into stress fibres. These cells would have possibly differentiated into myofibroblasts, while the remaining would still retain their keratocyte-like phenotype. These results support the idea that corneal stromal cells cultured for 3 weeks in KDM might constitute a population of cells at different stages of differentiation.

Both NCK and NLK cultured in KDM with or without FGF upregulated at the transcriptional level CD90, KERA, LUM and αSMA and downregulated KI67. At the protein level, NCK and NLK cultured in FGF depleted KDM upregulated significantly CD90 and αSMA, even though an upregulation of KERA was also clearly seen by looking at the Western blot images. On the contrary, NCK and NLK cultured in KDM with FGF did not upregulate any marker significantly, but a higher expression of CD90 and KERA was also appreciated, when compared to NC and NL. This suggests that there were not

113 major differences between the ability of cells treated in the presence or absence of this growth factor to differentiate into keratocytes, due to the upregulation of KERA and downregulation of MKI67 in both conditions. However, there might be a difference in the differentiation of fibroblasts and myofibroblasts, since CD90 and αSMA were more upregulated in NCK and NLK cultured in FGF depleted KDM. This is consistent with previous results that showed a reduction in myofibroblast differentiation when cells are cultured in FGF [79]. These results support the hypothesis that FGF is an essential regulator of the keratocyte differentiation process.

In the present work, despite cells being seeded at the same density for treatment with KDM supplemented or depleted of FGF, the cell density after 1 week of culture appeared different. It has already been reported that cell density is another modulator of cell phenotype [202]. Masur and colleagues showed that fibroblasts seeded at low density easily differentiated into myofibroblasts, increasing their expression of αSMA and TGFβ. However, when these enriched myofibroblasts cultures were seeded again at high density the expression of αSMA was lost within 3 days [202]. It is difficult to directly compare our data with that in the published literature, regarding the influence of FGF on cell behaviour, because many other confounding factors such as cell density and the presence of serum vary between studies.

Are there stem cells in the central cornea?

Similar to results shown here, in which corneal stromal cells share some CSSC-like characteristics, such as expression of CD90, CD73 and the ability to differentiate into keratocytes, recent reports have described a population of stem cells in the central corneal stroma. Furthermore, these cells have been attributed with MSC properties [58, 181, 187, 203]. Some of these studies claim that their results support the hypothesis that there is a population of MSC also located in the central corneal stroma and not only in the limbus [58, 187]. However, during the last decade, there has been a call for caution when using the term mesenchymal stem cells. The term MSC was coined more than 25 years ago to describe a population of cells isolated from bone marrow or periosteum that could be expanded in vitro while maintaining the ability to differentiate into a variety of mesodermal lineages. However, the term MSC has been exponentially used in the regenerative medicine field, regardless of cells presenting this multipotential capacity in vitro [204]. The fact that MSC include the term “stem cells” has allowed business to exploit MSC in some questionable stem-cell treatments. Caplan suggested changing the term mesenchymal stem cells to medicinal stem cells, implying that they have regenerative potential in vivo [205]. Whether the cells used herein should be called MSC

114

or not will depend on further in vitro characterisation and on which application they will be used.

Nevertheless, there exists a discussion on how the culture medium may be able to change the phenotype of the cells in vitro. It could be that by isolating differentiated keratocytes and culturing them in CSSCM, which was first described for CSSC, results in driving the cells back to a more undifferentiated phenotype in vitro. This ability of a culture medium, different to the one used here, could change the phenotype of corneal or limbal stromal cells was already discussed by Sidney and colleagues [206]. They also showed that by changing the culture medium, they could promote stem-cell like characteristics in corneal stromal cells, concluding that the culture conditions can easily modulate the corneal stromal cells phenotype [206].

Regardless of what is the nature of the cells isolated, whether they were stem cells in vivo or whether the in vitro culture conditions afforded them a degree of plasticity, the aim of this study was fulfilled. A novel protocol was developed to expand corneal stromal cells in vitro without becoming fibroblastic, while maintaining the ability to differentiate into keratocyte-like cells. Importantly, the results indicated that enough cells for subsequent in vitro studies could be obtained. In the next chapter a similar study was performed to determine if the method would be suitable for the culture of corneal stromal cells from aniridia patient tissue.

115

116

CHAPTER 4: ANIRIDIC AND NORMAL CORNEAL STROMAL CELLS HAVE SIMILAR PHENOTYPICAL CHARACTERISTICS WHEN CULTURED IN 2D

4.1 Introduction

The first studies that attempted to identify the mechanisms behind ARK development attributed this condition to a limbal epithelial stem cell deficiency (LESCD) [3, 143]. These reported the absence of palisades of Vogt in ARK limbus, where the limbal epithelial stem cells reside, as the possible cause for ARK [3]. It was then suggested that in LESCD the limbus cannot functionally serve as the barrier to segregate the cornea from the surrounding conjunctiva. This is followed by an incursion of the neighbouring conjunctiva onto the cornea, resulting in corneal opacification and vascularisation [3, 143]. Thereafter, other published evidence has suggested that LESCD is not the solely cause of ARK. Dysregulation of the differentiated corneal epithelium or of the stroma have also been hypothesised to play a role in the progress of this condition [129, 144, 161, 166, 207, 208].

In a recent review, Ihnatko and colleagues integrated an updated version of the genetic, morphologic, and molecular findings in human and animal model ARK corneas, with the studies published from 2003 to 2015 [209]. However, most of these studies focused on the involvement of the epithelium in ARK. In this layer, even though 60-70% of the normal Pax6 protein levels are kept unaltered, severe forms of ARK arise from heterozygous Pax6+/- gene transcription in mouse models [139]. Many other manifestations affect the corneal epithelium of ARK corneas, including altered protein profile expression, transparency, cell adhesion, migration, and morphology [129, 165]. However, for the purpose of this thesis, the focus will be placed on the evidence that supports the involvement of the stromal cells in the development of ARK.

Scarce evidence supports the hypothesis that the stroma might also be affected in ARK, and those that do are mainly based on in vivo or ex vivo observations in human corneas [161, 207, 210] and in murine models [169, 208, 211]. The aniridic stroma appears to be thicker and less regular than the normal corneal stroma, observations made in both humans and Pax6+/- mice corneas [161, 210, 211]. Additionally, the anterior stromal pannus in human ARK corneas presents strucutral changes compared to the healthy stroma, with modified ECM composition [161]. On the one hand, a downregulation of collagen type I was observed, which would help explain the loss of transparency in ARK corneas. On the other hand, an upregulation of fibronectin, vimentin and αSMA, proteins 117 involved in corneal wound healing and scarring, was also observed [161]. These results suggested that the anterior stromal pannus in ARK migh present alterations in the wound healing capacity, remaining in a state of on-going wound healing and fibrosis [161]. Supporting results showed higher levels of aniridic stromal cell apoptosis in an ex vivo model of wound healing in adult mice corneas [208]. Finally, keratocytes presented large intracellular vacuoles in Pax6+/- mice corneas [169], what may be a sign of cell death [212]. These results show that the stroma is affected during ARK. However, studies on the human aniridic cornea focus mainly on the dysregulation of ECM, rather than on the cells themselves [161, 210]. The presented study aims to fill this gap by assessing the behaviour of human aniridic corneal stromal cells in vitro.

To the best of our knowledge, there are no reported in vitro studies investigating the properties of primary corneal stromal cells extracted and cultured from patients with ARK. Therefore, it is currently unknown if they differ in any way from their non-diseased counterparts. The only evidence of cultured corneal stromal cells with a PAX6 mutation is a conference abstract in which Secker and colleagues reported the successful culture of murine Pax6+/- corneal fibroblasts [213]. However, these fibroblasts were cultured in DMEM supplemented with 10% FBS, and as discussed previously, the effect of the culture medium modulates cell phenotype. In addition, murine corneas differ significantly from human [214]. For example, the corneal stroma in mouse only contributes to 70% of the total corneal thickness, compared to 90% in humans, suggesting that the events happening in mouse model are not always comparable to those of humans [214].

In the previous chapter, CSSCM was successfully used to culture and expand normal central corneal stromal cells in vitro. Additionally, cells cultured in CSSCM presented some stem-like characteristics and maintained their ability to differentiate back into keratocyte-like cells. The next step was to establish whether the same results could be achieved using aniridic corneal stromal cells. Therefore, in this chapter human aniridic corneal stromal cells were extracted, culture, and expanded in CSSCM. These cells were then further characterised and compared to their non-diseased counterparts. Taking into account the changes observed elsewhere on aniridic epithelium and stroma [129, 161, 165, 208, 211], differences between aniridic and normal corneal stromal cells regardig their proliferation and migration capacities, differentiation potential, and apoptosis were assessed.

118

4.2 Hypotheses and aims

The hypotheses tested in this chapter were the following:

1. “Aniridic and normal corneal stromal cells have different phenotypes when cultured in CSSCM in vitro.” 2. “Aniridic corneal stromal cells differentiate less readily into keratocyte-like cells in vitro, compared to normal corneal stromal cells.”

3. “FGF is an essential regulator of the normal and aniridic corneal stromal cell differentiation into keratocyte-like cells in vitro.”

Hence, the overarching aim of Chapter 4 was to characterise aniridic corneal stromal cells cultured in CSSCM and compare their phenotype, differentiation potential, and apoptosis rate with those of normal corneal stromal cells.

4.3 Materials and methods

Patient demographics

Cadaveric donor corneas and live patient aniridic samples were obtained from Moorfields Eye Hospital (UK) with ethics approval (10/H0106/57-11ETR10) and appropriate research consent. Cadaveric donor corneas were stored in organ culture medium until dissection to isolate corneal stromal cells. Aniridic samples were promptly collected and stored temporarily in DPBS and processed as soon as received.

In this study, four aniridic samples and four control normal corneal buttons were used. Detailed donor information can be found on Table 4.1.

119

Table 4.1. Donor information. Age, eye of retrieval, type of surgery, storage time, and sample ID is shown. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells, NA: not applicable.

Condition Age Storage Eye Type of surgery Sample ID (years) (days)

54 Left Superficial keratectomy 1 AC1

63 - Superficial keratectomy 1 AC2

Aniridia Keratoprosthesis – 56 Left 2 AC3 corneal button

Keratoprosthesis – 58 - 1 AC4 corneal button

54 Right NA 26 NC1

46 Right NA 29 NC3 Normal 84 Right NA 19 NC2

63 Left NA 20 NC4

Patient clinical history was obtained from Moorfields Eye Hospital, Table 4.2. Two patients had previous corneal graft transplantation, and published evidence showed that donor stromal cells can remain in the graft for up to 30 years [215]. To obtain more information on the cells populating the cornea of the aniridic patients at the time of sample collection and to confirm that the cells being used were patient specific, genotyping was performed both from patient blood and their isolated corneal cells. Aniridic samples were processed and used as soon as received from surgery. Aniridic corneal stromal cells (AC) were isolated, cultured, and passaged as described in section 2.1 in Chapter 2.

120

Table 4.2. Information about clinical history of surgeries of aniridic patients. Graft date, tissue type, eye grafted and graft gender are included. No information for donor A4 could be obtained. Abbreviations NA: not applicable, NK: not known.

Patient Graft date Tissue type Eye grafted Graft gender

08/05/2013 Cornea Right Male A1 23/12/2015 Cornea Left Male

A2 NA None NA NA

12/04/2000 Cornea Unknown Male

27/03/2002 Sclera Unknown Male

A3 02/03/2005 Cornea Left Female

21/01/2009 Cornea Left Male

04/09/2013 Cornea Left Female

A4 NK NK NK NK

Genotyping of PAX6 gene in aniridic corneal stromal cells

4.3.2.1 DNA extraction of aniridic corneal stromal cells

Aniridic corneal stromal cells cultured in CSSCM, were trypsinised as in section 2.1.2, centrifuged and the pellet was resuspended in 150μL of lysis buffer, provided in the kit. Cultured stromal cell lysates from each aniridic sample were stored at -20°C until DNA was extracted. Genomic DNA purification was performed using the Wizard® SV Genomic DNA Purification System (Promega) as per manufacturer instructions. Briefly, the lysate was transferred to a Wizard SV Minicolumn assembly and spun for 3 minutes at 13000xg. The flow-through was discarded and 650μL of Column Wash Solution was added to the Minicolumn, centrifuged for 1 minute at 13000xg and the flow-through discarded. This washing step was repeated three times for a total of four washes. After the last wash, the Minicolumn was centrifuged for 2 minutes at 13000xg to dry the binding matrix. The Minicolumn was placed into a new, labelled 1.5mL microcentrifuge tube and 250μL of Nuclease-Free Water were added on top and incubated for 10 minutes. The Minicolumn/tube assembly was placed into the centrifuge and spun for 1 minute at 13000xg to elute the DNA. To remove copurified RNA, 2μL of RNase A Solution was added to the solution and incubated for 5 minutes, before storing the DNA at -20°C.

121

Concentration and purity of DNA were then analysed with the Nanodrop System (ThermoFisher Scientific). Nucleic acids have a peak of absorbance at 260 nm, from which, using the Beer’s Lambert law, their concentration can be calculated. Values of the ratio of absorbance at 260/280 and 260/230 indicate if any present contaminants absorb at 280 nm (proteins) or 230 nm (phenols and salts) and are an indication of the purity of the DNA.

4.3.2.2 RT- PCR

The coding exons of the PAX6 gene (exon 4- exon 13) were each amplified as single PCR products. Additionally, exon 2 and the conserved Simo enhancer element were also studied given that some aniridia causing mutations had been previously found in this location [216]. Primers used for the PCR reactions can be seen in Table 4.3 [217]. All regions were amplified with GoTaq ® Polymerase (Promega) except exon 4 that was amplified with Kapa Robust polymerase (Sigma-Aldrich).

For amplification with GoTaq ® Polymerase, the following mixture was prepared: 6.25μL of 2X GoTaq ® Green Master Mix, 0.5μL of forward primer (10μM), 0.5μL of reverse primer (10μM), 4.25μL of water and 1μL of template DNA (50-100ng/μL). The resulting mixture was then placed in the Thermocycler and programmed as follows: 95°C for 2 minutes, 35 cycles of 95°C for 30 seconds, annealing temperature for 30 seconds and 72°C for 30 seconds, and finally 72°C for 5 minutes.

For amplification with KAPA Robust polymerase, the following mixture was prepared: 6.25μL of 2X KapaRobustTM, 0.625μL of forward primer (10μM), 0.625μL of reverse primer (10μM), 0.625μL DMSO, 3.375μL of water and 1μL of template DNA (50- 100ng/μL). The resulting mixture was then placed in the Thermocycler and programmed as follows: 95°C for 5 minutes, 35 cycles of 95°C for 15 seconds, annealing temperature for 15 seconds and 72°C for 30 seconds, and finally 72°C for 10 minutes.

122

Table 4.3. Primers used for RT-PCR [217]. Sequence, annealing temperature and amplicon size is shown. Abbreviations bp: base pair.

Annealing Amplicon size Primer name Primer sequence temperature (°C) (bp) PAX6-Simo-f tgtttacgatgtagaatagaagattgc 60 695 PAX6-Simo-r agaagaagaatttagttttatgggact PAX6-Exon2-f 60 PAX6-Exon2-r PAX6-Exon4-f tgtaggggaaacaga 55 491 PAX6-Exon4-r atcgagaagagccaagcaaa PAX6-Exon5-f ctggtggtcctgttgtcctt 60 446 PAX6-Exon5-r atgaagagagggcgttgaga PAX6-Exon6-f gggctacaaatgtaattttaagaa 60 509 PAX6-Exon6-r agagagggtgggaggaggta PAX6-Exon7-f gagctgagatgggtgactg 63 300 PAX6-Exon7-r gagagtaggggacaggcaaa PAX6-Exon8-f tcaggtaactaacatcgca 55 719 PAX6-Exon8-r gttgactgtacttggaagaa PAX6-Exon9-f aggtgggaaccagtttgatg 60 311 PAX6-Exon9-r catggcagcagagcatttag PAX6-Exon10-f gctaaatgctctgctgccat 60 329 PAX6-Exon10-r agagtgagagtcagagcccg PAX6-Exon11-f ccgggctctgactctcact 60 221 PAX6-Exon11-r gccactcctcacttctctgg PAX6-Exon12-f gaggcttgatacataggc 55 452 PAX6-Exon12-r ccataagaccaggagatt PAX6-Exon13-f tccatgtctgtttctcaaagg 60 220 PAX6-Exon13-r tcaactgttgtgtccccatag

4.3.2.3 Gel electrophoresis of PCR products

A 2% agarose gel was prepared by adding 3g of agarose powder to a conical flask with 150mL of 1X TAE. The mixture was microwaved for periods of 30 seconds until the powder was dissolved. When the mixture was slightly cooled 7.5μL of SafeView were added before pouring into a gel mould. When the gel was set after approximately 30

123 minutes, the gel was loaded with 7.5μL of 2-log ladder and 4μL of PCR product. The gel was then run for 45 minutes at 100V.

4.3.2.4 Purification of PCR products

A volume of 90μL of water was added onto the 8μL remaining PCR product. The diluted samples were then loaded into a 96-well Millipore plate (Merck Millipore) and placed on a vacuum pump (Merck Millipore). Once dry, the membranes were washed with 40μL of water. The dried PCR products were resuspended in 25μL of water; the plate was sealed and shaken for 10 minutes using a vortex. The purified PCR products were placed in labelled 0,5mL eppendorf.

4.3.2.5 Sanger sequencing

PCR products were sent to Source BioScience © (Nottingham, UK) to be sequenced by the Sanger Sequencing Service and results were received within a working day in .ab1 and .seq formats.

4.3.2.6 Sequence analysis

Sequences were analysed with Benching, a free-access online data management platform and images were finalised with BioEdit. Sequences were aligned with the Clustal Omega or MAFFT algorithms to the human genome, and variants found were annotated against the NM_000280 PAX6 transcript following the HGVS sequence variant nomenclature (http://varnomen.hgvs.org).

Characterisation of aniridic and normal corneal stromal cells in CSSCM

4.3.3.1 Gene expression – Real time qPCR

RNA extraction, assessment of RNA purity and integrity, reverse transcription and Real- time qPCR were all performed as described in section 3.3.2.4 in Chapter 3.

As explained in the previous chapter, MSC were used as positive controls for CD90 and CD73. In the same way, ECs were used as positive controls for PAX6. These two types of cells were included in the Real-Time qPCR experiments, to compare their phenotype against AC and NC.

4.3.3.2 Protein production - Western blot

Protein extraction, quantification, electrophoresis, immunoblotting, and densitometry quantification were all performed as described in section 3.3.2.4 in Chapter 3.

124

4.3.3.3 Cell migration towards a chemoattractant

Corning® Transwell® polycarbonate membrane cell culture inserts (Corning® Costar) of 6.5mm diameter with an 8μm membrane pore size were used to assess corneal stromal cells migration. Cells have the ability to migrate through the porous membrane towards a chemoattractant (chemotaxis). Chemotaxis towards cytokines in vivo or foetal bovine serum, the latter was used in the present study, has been shown elsewhere [218].

As previously described by Justus et al [218], cells were seeded on top of the inserts at a density of 30120 cells/cm2 (9000 cells/insert) in serum-free medium DMEM (High glucose, GlutaMAXTM, Gibco®) supplemented with 1%P/S (Gibco®) and incubated for 4 hours at 37°C to allow to attach. CSSCM supplemented with 10% FBS was then added to the lower chamber and cells were let to migrate for 24 hours through the porous membrane as shown in Figure 4.1.

After washing once with DPBS, cells were fixed with 100% iced cold methanol for 10 minutes and stained with Harris’ haematoxylin for 30 minutes followed by 3 more washes with DPBS. Cells remaining on top of the membrane that had not migrated were removed by scraping with a cotton bud. Five viewpoints of the membranes were imaged under a contrast phase microscope (Eclipse TS100, Nikon). Three aniridic and three normal donors were used for this experiment, with three technical replicates respectively. All the inserts were imaged in the same way, obtaining pictures from the same five regions (top, bottom, right, left, middle). Images were analysed using Fiji ImageJ 1.49s software to allow counting of the migrating cells.

4h 24h

Figure 4.1. Representation of the transwell migration assay. Cells were first seeded on the upper chamber of the transwell and then allowed to migrate through the porous membrane towards a chemoattractant (FBS) added to the lower chamber.

4.3.3.4 Collagen gel contraction assay

Collagen gel contraction assay was performed to assess the differences in contraction capacity between the AC and NC. This is a simplified model to study the contractile profile of the cells that populate the gels [219]. The assay was performed with a few 125 modifications of a published protocol [220]. Briefly, a mix of 1 mL of rat-tail type 1 collagen (First Link), 120μL of minimum essential medium (MEM) 10X (Gibco®), 100μL of sodium hydroxide 1X (Sigma-Aldrich) and 120μL of cell solution (25000 cells/gel) was prepared (sufficient for 8 gels) and kept on ice for 30 minutes. An extra volume of gel solution prepared without cells was prepared to act as a control.

Of the mixture, 150μL were added to each well of a 48 well plate to make 4 gels per condition (aniridia, normal and cell-free negative control) and left to polymerise at 37ºC for 30 minutes. Fresh CSSCM was then added to all the wells, and the edges of the gels gently released with a pipette tip to allow matrix contraction. Images of all the gels were taken with a digital camera (Canon) against a white light source (Hancocks) at the following time points: day 0, 1, 2, 3, 4 and 7. Images were then analysed using Fiji ImageJ 1.49s software, and the area of the gels measured and plotted as a percentage of day 0.

Differentiation of aniridic and normal corneal stromal cells into keratocyte-like cells in KDM

AC and NC cells were cultured in KDM to characterise their potential to differentiate into keratocyte-like cells in vitro. Briefly, AC and NC were seeded (1000cells/ cm2) into T175- cm2 flasks and 4-well chamber slides and cultured in KDM depleted or supplemented with FGF for a period of 1, 2 and 3 weeks. At these time points, the RNA and protein were extracted for further analysis.

As a control, AC and NC cells were cultured to confluence in CSSCM to promote cell density-induced differentiation without a specific guided differentiation protocol [221]. Briefly, AC and NC were also seeded (1000cells/ cm2) into a T175-cm2 flask and 4-well chamber slide and cultured in CSSCM for 5 days. Medium was refreshed at day 2.

AC and NC seeded into (1000cells/ cm2) into a T175-cm2 flask and cultured in CSSCM for only 24 hours will be addressed as basal conditions.

A detailed protocol of the different cell culture conditions can be found in section 3.3.3.1 in Chapter 3.

126

The following nomenclature and abbreviations will be used in the Results section. i = 1,2,3 refers to the time point in weeks when cells were collected, and F+/- refers for the presence or absence of FGF in KDM. A schematic representation of the cell culture conditions is found in Table 4.4.

Aniridic corneal stromal cells - AC

+ - Aniridic corneal stromal keratocyte-like cells ACKiF , ACKiF

Aniridic corneal stroma spontaneously differentiated cells – ACSD

Normal corneal stromal cells – NC

+ - Normal corneal stromal keratocyte-like cells – NCKiF , NCKiF

Normal corneal stroma spontaneously differentiated cells – NCSD

Table 4.4. Culture conditions for AC and NC in CSSCM and KDM. Abbreviations CSSCM: corneal stromal stem cell medium, KDM: keratocyte differentiation medium, AC: Aniridic corneal stromal cells, NC: Normal corneal stromal cells, ACKiF: Aniridic corneal stromal keratocyte-like cells, ACSD: Aniridic corneal stroma spontaneously differentiated cells, NCKiF: Normal corneal stromal keratocyte-like cells, NCSD: Normal corneal stroma spontaneously differentiated cells.

Cell culture Time (days)

medium 1 … 5 ... 7 … 14 … 21

CSSCM AC ACSD

NC NCSD

KDM FGF+ ACK1F+ ACK2F+ ACK3F+

NCK1F+ NCK2F+ NCK3F+

FGF- ACK1F- ACK2F- ACK3F-

NCK1F- NCK2F- NCK3F-

To characterise AC and NC cultured in different conditions, a myriad of markers was assessed by Real Time qPCR, Western blot, and immunocytochemistry. These included CSSC (CD90, CD73, PAX6), keratocyte (KERA, LUM, ALDH1A1), myofibroblast (αSMA), proliferation (KI67) and apoptosis markers (TP53, BAX, BCL2).

4.3.4.1 Gene expression – Real-Time qPCR

Real-time qPCR was performed as described in section 3.3.3.2 in Chapter 3.

127

4.3.4.2 Western blot

Western blot was performed as described in section 3.3.3.3 in Chapter 3.

4.3.4.3 Immunocytochemistry

Immunocytochemistry was performed as described in section 3.3.3.4 in Chapter 3.

Statistical analyses

Three biological repeats were used for each condition (aniridia and normal) for all of the experiments presented in this chapter, except for Real Time qPCR in which 4 biological repeats could be used. This was due to donor tissue availability at the time of the experiments.

4.4 Results

Genetic variants found in the aniridic samples

Two different types of mutations were found in the four aniridic samples here used (two run-on and two PTC mutations). Each pair (blood: corneal cell) showed the same PAX6 mutation, confirming that isolated cells were patient specific and not derived from the previously transplanted donor graft.

A run-on mutation was identified in the stop codon of Exon 13 for Aniridia 1 cells (AC1) (Figure 4.2). The codon TAA of the DNA is a stop codon that would be transcribed into UAA in the mRNA sequence and results in stopping protein translation. In this sample, the TAA codon is mutated into TTA that translates into the amino acid Leucine instead of a stop codon. This results in prolongation of protein synthesis after its presumed stopping point.

128

Aniridia sample 1

Patient sample: Het NM_000280: c. 1268A>T; p.*423Leuext*?

Figure 4.2. Representation of the last part of Exon 13 sequence in Aniridia 1 sample compared to a control sample. In one of the copies of the PAX6 gene of Aniridia sample 1, the A nucleotide is mutated into a T, and the codon translates into a Leucine instead of a stop codon. Imaging and genotyping of Aniridia sample 1 was performed in house by Dr Alice Davidson.

In the sequence obtained from the genotyping of Aniridia 2 cells (AC2), a premature stop mutation was identified in Exon 8 in one of the copies of the PAX6 gene (Figure 4.3). In healthy individuals the codon at that position is GGA, which translates to Glycine, however in this patient it was mutated to a stop codon TGA. This results in a premature termination of the translation of the protein.

129

Aniridia sample 2

Patient sample: Het NM_000280: c.580G>T; p.(Gly194*)

Control sample: Wild-type sequence

Figure 4.3. Representation of a portion of Exon 8 sequence in Aniridia 2 compared to a control sample. The codon GGA that would translate into a glycine is mutated to a stop codon TGA resulting in a premature stop of protein translation.

In the sequence obtained from Aniridia 3 cells (AC3), a deletion of two nucleotides in Exon 7 in one of the two copies of the PAX6 gene was seen. The nucleotides 375 and 376 (G and A, respectively) were deleted, so where healthy individuals would have the codon AGA (Arginine), patient 3 has the codon AGT that would translate into a Serine. Since the two nucleotides are deleted, a frameshift of the rest of the sequence will occur, and the protein continues to be synthesised with incorrect amino acids until a premature stop codon is found at Exon 7 (Figure 4.4).

130

Aniridia sample 3

Patient sample: Het NM_000280: c.375_376del; p.(Arg125Serfs*7)

Control sample: Wild-type sequence

Figure 4.4. Representation of a portion of the Exon 7 sequence in Aniridia sample 3 compared to a control sample. The nucleotides 375 and 376 in Aniridia sample 3 were deleted, changing the codon AGA (arginine) into an AGT codon (serine). This results in a frameshift of the protein that would eventually have a premature stop in Exon 7.

Aniridia 4 cells (AC4) presented the same Run-on mutation as AC1, and a graphic representation can be found in Figure 4.2.

To summarize, three different mutations were found amongst the four aniridic samples provided (Table 4.5).

Table 4.5. Overview of the three mutations identified in the four aniridic samples used for this study. Names of the mutations are annotated against the Ensembl entry NM_000280, (c. annotated against cDNA, p. annotated against protein).

Sample Mutation Exon Type of Cause of the mutation mutation

Aniridia 1 c. 1268A>T; p.*423Leuext*? 13 Run on mutation Substitution

Aniridia 2 c.580G>T; p.(Gly194*) 8 Premature stop Substitution

Aniridia 3 c.375_376del; 7 Premature stop Deletion p.(Arg125Serfs*7)

Aniridia 4 c. 1268A>T; p.*423Leuext*? 13 Run on mutation Substitution

131

Aniridic and normal corneal stromal cells present similar phenotype when cultured in CSSCM

4.4.2.1 AC and NC can be successfully isolated and expanded

Stromal cells were successfully isolated from aniridic and normal corneal buttons. Within 24 hours of plating, attached cells were readily observed. Few small epithelial cell colonies were observed in the normal cultures. However, no epithelial cell colonies were seen in the aniridic cultures. Corneal stromal cultures appeared initially to be a mixed population with cells presenting different morphologies: cuboidal (white arrows, Figure 4.5), elongated and dendritic (Figure 4.5).

Figure 4.5. Representative images of cultures at P0 isolated from aniridic and normal central corneas. White arrows point to square-like shaped stromal cells. Scale bars = 300 µm

As cells were passaged, a more uniform population prevailed. Most of the cells in culture after passage 3 showed a small square-like morphology with a high nucleus-cytoplasm ratio, as shown in Figure 4.6 (white arrows). A small proportion of cells acquired a more elongated morphology (black arrows, Figure 4.6). No evident differences in morphology were seen between aniridic (AC) and normal (NC) stromal cells or between any of the aniridic samples when cells were cultured in CSSCM.

132

Figure 4.6. Representative images of AC and NC cultured in CSSCM. Images were taken between passages 3 and 5. Most cells were small with almost square-like morphology with high nucleus to cytoplasm ratio (white arrows). A small portion of the cells presented with more elongated bodies (black arrows). Images were taken under a phase contrast microscope. n=3. Scale bars =300µm. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

133

4.4.2.2 AC have lower proliferation capacity than NC when cultured in CSSCM

AC and NC were cultured over a period of 7 days in CSSCM (Figure 4.7). At day 3, cells showed 30-50% confluence and the majority maintained the small square-like morphology described in section 3.4.2.1. As the cells divided and became more confluent, their morphology changed, acquiring an elongated, spindle shaped appearance (Figure 4.7A). AC density at day 7 was lower than NC, suggesting that the former proliferated less. A few AC appeared to have wider cell bodies and formed an unorganised network, possibly resembling a more differentiated phenotype.

Quantification of the total DNA harvested at days 3, 5, and 7, confirmed that AC had lower proliferation capacity than NC (Figure 4.7B). While DNA amount significantly increased on NC from day 3 to day 7 (p<0.001), AC samples did not show any significant differences over time. The initial cell seeding density was identical between AC and NC, however at day 5 and 7 there was a significant difference in the amount of DNA between AC and NC (p<0.001), suggesting lower AC cell numbers.

134

B) 8000 *** *** AC 6000 NC *** 4000

2000 Total DNA (ng) TotalDNA

0 3 5 7 Time (days)

Figure 4.7. Proliferation of AC and NC in CSSCM over 7 days. (A) Representative images of cells in culture at day 3, 5 and 7. Scale bars= 300µm. (B) Cell proliferation assessed by total DNA quantification. Data shows the mean ± standard deviation, n=3. Repeated measurements ANOVA were used for statistical analysis. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

4.4.2.3 AC and NC show a similar pattern of tested gene expression.

No significant differences in PAX6, CD90 and CD73 gene expression levels were found between AC and NC (Figure 4.8A, p>0.05). Expression levels of PAX6 in AC and NC were found to be very low, with Ct values appearing after 34-36 cycles (Appendix, Table A.8.1). Interestingly, in one of the aniridic cell lines, AC1, there was no detected

135 expression of PAX6 (Table A.8.1). CD90 and CD73 gene expression levels were similar between AC and NC, as well as in MSC (p>0.05).

No significant differences in KERA, LUM and ALDH1A1 gene expression levels were found between AC and NC (Figure 4.8B, p>0.05). KERA gene expression was not detected in any of the conditions (Black arrow in Figure 4.8B). Although not significant, ALDH1A1 was less expressed in AC than in NC.

No statistically significant differences in KI67 (proliferation marker) and TP53 (apoptosis marker) gene expression levels were found between AC and NC (Figure 4.8C, p>0.05) when cultured at low density in CSSCM.

136

A) 100 MSC 10 AC 1 NC 0.1 0.01

Fold change Fold 0.001 0.0001 0.00001

(Normalised to epithelial cells) to epithelial (Normalised PAX6 CD90 CD73

10 B) MSC 1 AC 0.1 NC

0.01

0.001 Fold change Fold 0.0001 

0.00001

(Normalised to epithelial cells) to epithelial (Normalised KERA LUM ALDH1A1

C) 10 MSC AC NC

1 Fold change Fold

0.1

(Normalised to epithelial cells) to epithelial (Normalised KI67 TP53

Figure 4.8. Relative gene expression levels of (A) CSSC markers (PAX6, CD90, CD73), (B) keratocyte markers (KERA, LUM, ALDH1A1), and (C) proliferation and apoptosis markers (KI67, TP53, respectively) in MSC, AC and NC. Data is represented as the median of the obtained data points for each condition, n=1 for ECs and MSC, n=3 biological replicates for AC and NC. Dashed line represents the gene expression of ECs, to which the other conditions are compared to. Arrow represents gene expression not detected. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells, MSC: mesenchymal stem cells.

137

4.4.2.4 AC and NC show a similar pattern of tested protein production

CD90 and CD73, the putative CSSC markers expressed at mRNA level, were chosen to confirm their protein production levels by Western blot. Similar to the Real Time qPCR results, no statistically significant differences in CD90 and CD73 protein production between AC and NC were found (Figure 4.9, p>0.05). When the data from the donors was plotted separately, high donor to donor variability was observed, mostly in CD90 production profile (Figure 4.9A). AA) 2.5 AC 2.0 NC

1.5

1.0

0.5

0.0

CD90 CD73 Relative expression GAPDH Relative to

Figure 4.9. Protein production levels of putative CSSC markers (CD90 and CD73) in AC and NC. (A) Densitometry quantification of each protein normalized against the expression of GAPDH. Data is represented as the mean of n=3 biological replicates. Mann Whitney test, p>0.05 (not significant). (B) Western blot representative of the 3 donors. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

138

4.4.2.5 AC and NC show similar migration capacity

When data from the three aniridic and normal donors were plotted together, the number of cells that migrated through the membrane towards the chemoattractant was not significantly different between the two groups (p>0.05). An average of 300 normal cells migrated through the porous membrane compared to a mean of 350 aniridic cells (Figure 4.10A). Individual analysis of the results for each of the aniridic donors as shown in Figure 4.10B, demonstrated a higher biological variability as evidenced by the higher discrepancy in the number of migrating cells. When comparing individually the results from the three different aniridic donors with normal, there were no significant differences between NC and AC1 and AC3 (p>0.05). However, AC2 cells migrated more than NC cells (p<0.05, Figure 4.10B).

A) B)

400 600 *

300 400 200 200

100

Migration (# cells) Migration (# Migration (# cells) Migration (# 0 0 AC NC AC1 AC2 AC3 NC

Figure 4.10. Graphic representation of the number of cells that migrated per well in each condition. (A) Data is represented as the mean of n=3 biological replicates ± standard deviation. (B) AC is separated by donor; data is represented as the mean of each donor separately ± standard deviation (n=3 technical replicates for each donor) NC is represented as in (A). Kruskal- Wallis test followed by Dunn’s post hoc test was performed *p<0.05. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

4.4.2.6 AC and NC have similar ability to contract a collagen gel

Gels populated by AC or NC were allowed to contract over a period of 7 days (Figure 4.11A). The gels without cells that were used as negative controls did not contract, whereas the gels populated with AC and NC contracted approximately 80% over this time period. No significant differences between gel contraction rates of aniridic and normal cells were appreciated (Figure 4.11B).

Results were also plotted for each aniridic donor separately (Figure 4.11C). No significant differences were found in the contraction ability of any the three aniridic donors compared between themselves nor with the normal cells. 139

100 B) No cells 80 AC NC 60

Contraction(%) 20

0 0 2 4 6 8 Time (days)

100 C) No cells 80 AC1 AC2 60 AC3 NC 40

Contraction(%) 20

0 0 2 4 6 8 Time (days)

Figure 4.11. Contraction of gels populated with NC, AC and without cells over a period of 7 days. (A) Representative pictures of gels at 6 different time points. Dashed line marks the area of the gel. (B) Contraction is indicated as a % of the starting area from time 0, i.e. 0% not contracted and 100% complete contraction. The data represents the mean of all three donors pooled for each condition (n=3 technical replicates for each donor). Error bars represent the standard deviation. p>0.05 between AC and NC at each time point. (C) Contraction of gels populated with NC, AC and without cells measured over a period of 7 days. NC is represented as in A). For AC, each line represents the mean of each donor separately (n=3 technical replicates for each donor). Error bars represent the standard deviation. p>0.05 between AC1, AC2, AC3 and NC. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells,

140

Summary of results:

When cultured in CSSCM:

• AC and NC were successfully isolated and cultured.

• AC and NC did not show differences in morphology.

• AC showed lower proliferation capacity than NC.

• There were no significant differences in gene and protein expression of putative CSSC, keratocyte and proliferation markers between AC and NC.

• AC and NC showed a gene expression profile similar to MSC.

• AC had similar average migration capacity to NC, with a high donor to donor variability.

• AC and NC had similar ability to contract collagen gels.

Aniridic and normal corneal stromal cells have similar capacity to differentiate into keratocytes-like cells in KDM

Images and graphics containing data obtained for normal corneal stromal cells are also included in this chapter for ease of results interpretation – aniridic versus normal corneal stromal cells. However, their data and trends were thoroughly explained in the previous chapter.

4.4.3.1 AC and NC change their morphology during keratocyte differentiation

AC and NC cultured sparsely in CSSCM prior to any differentiation process had the same appearance as previously described in section 3.4.2.1. When cells were left in CSSCM for further 5 days without being passaged (ACSD and NCSD), they proliferated and changed morphology to a more elongated, spindle shape (Figure 4.12A). There was a visible difference in cell confluence between NCSD and ACSD, with the latter condition showing lower levels of cell density.

As shown in Figure 4.12B, ACK and NCK also became elongated over time. However, the number of cells was much lower when compared to ACSD and NCSD. Interestingly, there were apparent differences in cell behaviour between ACK and NCK when they were treated in the presence or absence of FGF. AC and NC prompted to differentiated towards a keratocyte phenotype in KDM with no FGF presented higher numbers of cells, particularly at week 2 and 3, and presented thinner bodies when compared to ACK and NCK cultured in the presence of FGF. A proportion of ACK and NCK cultured with FGF

141 presented a bright body at week 1 and 2, with thicker bodies when compared to those cultured in KDM without FGF. A)

Figure 4.12. Representative images of AC and NC cultured under different conditions. (A) Pictures of cells cultured in CSSCM at day 1 and day 5. (B) Images of cells differentiating into keratocyte-like cells during a period of 3 weeks using KDM supplemented or depleted of FGF. Scale bars = 300µm. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells, ACKiF aniridic corneal stromal keratocyte-like cells, NCKiF normal corneal stromal keratocyte-like cells.

142

4.4.3.2 AC and NC change their gene expression and protein production patterns under different culture conditions to promote keratocyte differentiation

AC and NC seeded at a density of 1000cells/cm2 in a T175-cm2 flask in CSSCM for only 24 hours were addressed as basal conditions (i.e. the phenotype that was used to start the keratocyte or spontaneous differentiation).

4.4.3.2.1 Gene expression profiling of putative CSSC markers during directed differentiation of AC and NC towards a keratocyte phenotype

Expression of CD90 was upregulated during keratocyte differentiation

As shown in Figure 4.13A, when prompted to differentiate towards a keratocyte phenotype in KDM supplemented with FGF, aniridic corneal stromal cells (ACK) showed a significant increase (p<0.05) in CD90 expression. However, this upregulation was not significant when ACK were differentiated in KDM supplemented with FGF (Figure 4.13A). The same trends were found for normal corneal stromal cells (NCK; Figure 4.13B). However, NCK significantly upregulated CD90 both when differentiated in the presence and absence of FGF.

In contrast, neither aniridic nor normal corneal stromal cells allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in CD90 expression.

Expression of CD73 did not change during keratocyte differentiation

As shown in Figure 4.13C, expression levels of CD73 did not significantly change when aniridic corneal stromal cells were differentiated towards a keratocyte phenotype (ACK). This remained true regardless of the presence or absence of FGF. The same trends were found for normal corneal stromal cells (NCK; Figure 4.13D).

Additionally, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate in CSSCM (ACSD, NCSD) showed any change in expression levels of CD73 (Figure 3.10C, Figure 4.13D).

Expression of PAX6 was very low across all conditions during keratocyte differentiation

PAX6 expression was very low expressed by all the different cell phenotypes. Ct values of PAX6 in undifferentiated NC cultured in CSSCM (basal conditions) were found around 34-35 cycles in three of the donors and higher than 40 in the other one. Ct values of PAX6 in undifferentiated AC were found around 37 in three of the donors and higher than 40 in the other one (Table A.8.2). When AC and NC were prompted to a keratocyte 143 phenotype (ACK, NCK) or spontaneously differentiated in CSSCM (ACSD, NCSD), expression of PAX6 was even lower than in the basal conditions or not detected (Ct values higher than 40), suggesting that PAX6 expression might be downregulated when cells are cultured in KDM. However, since the expression at the basal conditions was so low, no graphs were prepared, and statistical analysis could not be done.

A) B) CD90 CD90 100 100 * * ** * * ** 10 10

1 1

(normalised to AC) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

C) D) CD73 CD73 100 100

10 10

(normalised to NC) to (normalised (normalised to AC) to (normalised

Relative expression level expression Relative 0.1 Relative expression level expression Relative 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

Figure 4.13. Relative gene expression levels of putative CSSC markers (CD90 and CD73) by aniridic (A, C) and normal corneal (B, D) stromal cells under different culture conditions. Results are shown for AC and NC cultured in KDM supplemented or depleted of FGF at different time points (ACKi and NCKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (ACSD and NCSD, respectively). Dashed line represents the gene expression for the basal condition AC or NC. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition (i.e. the phenotype from which they were differentiated). *p<0.05, **p<0.01. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

144

4.4.3.2.2 Protein production profiling of putative CSSC markers during directed differentiation of AC and NC towards a keratocyte phenotype

Production of CD90 was upregulated during keratocyte differentiation

Western blot data confirmed the results obtained by Real-Time qPCR. As shown in Figure 4.14A, when prompted to a keratocyte phenotype, aniridic corneal stromal cells (ACK) significantly upregulated (p<0.05) production of CD90 protein. In the same way than in the mRNA expression, this was only true when ACK were cultured in KDM without FGF. The same trends were observed when normal corneal stromal cells (NCK) were differentiated towards a keratocyte phenotype. NCK significantly upregulated (p<0.05) CD90 production when cultured in KDM for 3 weeks without FGF (Figure 4.14B). Interestingly, there was no significant upregulation of CD90 when NCK were cultured in the presence of FGF, although real time qPCR showed differently. As shown in the WB membranes, an upregulation of CD90 production could also be observed when ACK and NCK were cultured in the presence of FGF. However, in the densitometry quantification, this upregulation was not significant.

In addition, there were no significant differences in CD90 production when aniridic and normal corneal stromal cells were allowed to spontaneously differentiate in CSSCM (ACSD, NCSD, respectively).

Production of CD73 did not change during keratocyte differentiation

As shown in Figure 4.14C, production of CD73 did not change significantly when aniridic corneal stromal cells were differentiated towards a keratocyte phenotype (ACK). This remained true regardless of the presence or absence of FGF. The same trends were shown for normal corneal stromal cells (NCK; Figure 4.14D).

However, when comparing aniridic corneal stromal cells that were allowed to spontaneously differentiate in CSSCM (ACSD) with the basal condition (AC), a significant downregulation (p<0.05) of CD73 was observed (Figure 4.14C).

145

A) B) CD90 CD90 4 * 4 * 3 3

2 2

1 1

0 0 Relative expression GAPDH to Relative Relative expression GAPDH to Relative AC NC ACSD NCSD ACKF- ACKF+ NCKF- NCKF+ CD90 (17kDa)

GAPDH (36 kDa)

C) D) CD73 CD73 3 * 3

2 2

1 1

0 0 Relative expression GAPDH to Relative

Relative expression GAPDH to Relative NC AC NCSD ACSD ACKF- NCK3F- NCK3F+ CD73 ACKF+ (63kDa)

GAPDH (36kDa)

Figure 4.14. Production levels of putative CSSC markers (CD90 and CD73) by aniridic (A, C) and normal (B, D) corneal stromal cells under different culture conditions as assessed by Western blot. Results are shown for the basal condition AC and NC, for AC and NC cultured for 3 weeks in KDM supplemented (ACK3F+, NCK3F+, respectively) or depleted of FGF (ACK3F- , NCK3F-, respectively), and for AC and NC cultured for 5 days in CSSCM (ACSD and NCSD, respectively). The graphs show the densitometry quantification for each marker normalised to GAPDH. Data is represented as the mean of n=3 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

146

4.4.3.2.3 Gene expression profiling of keratocyte markers during directed differentiation of AC and NC towards a keratocyte phenotype

Expression of LUM was upregulated during keratocyte differentiation

As shown in Figure 4.15A, when prompted to differentiate towards a keratocyte phenotype, aniridic corneal stromal cells (ACK) showed a significant increase (over 1000X Fold increase, p<0.05) in LUM expression. This remained true whether ACK were differentiated in the presence or absence of FGF (Figure 4.15A). The same trends were found for normal corneal stromal cells (NCK; Figure 4.15B). A very clear trend shows that LUM gene expression gradually increased with time in ACK and NCK (Figure 4.15).

However, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in LUM expression.

Expression of KERA was upregulated during keratocyte differentiation

As shown in Figure 4.15C, when prompted to differentiate towards a keratocyte phenotype, aniridic corneal stromal cells (ACK) did not show a significant increase in KERA expression. However, a trend could be appreciated in which KERA expression increased over time in ACK. In contrast, normal corneal stromal cells differentiated towards a keratocyte phenotype (NCK) showed a significant upregulation (100X-Fold increase, p<0.05) of KERA expression (Figure 4.15B). This remained true both when NCK were differentiated in KDM in the presence or absence of FGF.

However, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in KERA expression.

Expression of ALDH1A1 did not significantly change during keratocyte differentiation

As shown in Figure 4.15E, expression levels of ALDH1A1 did not significantly change when aniridic corneal stromal cells (ACK) were differentiated towards a keratocyte phenotype. The same trends were found for normal corneal stromal cells (NCK; Figure 4.15D). However, the same trend of upregulation seen in LUM and KERA could also appreciated for ALDH1A1 in both ACK and NCK, particularly at week 2 and 3.

147

LUM LUM A) 100000 B) ** 100000 10000 * ** * 10000 * ** * ** 1000 1000 100 100 10 10 1

(normalised to AC) to (normalised 1

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

KERA KERA C) 10000 D) 10000 1000 1000 * * *

100 100

10 10

1 1

(normalised to NC) to (normalised

(normalised to AC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

ALDH1A1 ALDH1A1 E) 1000 F) 1000

100 100

10 10

1 1

(normalised to NC) to (normalised

(normalised to AC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

Figure 4.15. Relative gene expression levels of keratocyte markers (LUM, KERA and ALDH1A1) in aniridic (A, C, E) and normal (B, D, F) corneal stromal cells under different culture conditions. Results are shown for AC and NC cultured in KDM supplemented or depleted of FGF at different time points (ACKi and NCKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (ACSD and NCSD, respectively). Dashed line represents the gene expression for the basal condition AC or NC. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05, **p<0.01. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

148

4.4.3.2.4 Protein production profiling of keratocyte markers during directed differentiation of AC and NC towards a keratocyte phenotype

Production of LUM was not upregulated during keratocyte differentiation

Surprisingly, the LUM protein production profile did not follow the same trend observed at RNA level. As shown in Figure 4.16A, when prompted towards a keratocyte phenotype, aniridic corneal stromal cells (ACK) seemingly downregulated LUM, compared to the basal conditions AC. However, these differences were not statistically significant in the densitometry (p>0.05) (Figure 4.16A). In normal cells, a similar trend could be observed (Figure 4.16B). No significant changes in production of LUM could be observed when normal corneal stromal cells were differentiated to a keratocyte phenotype (NCK; Figure 4.16B).

Seemingly, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in LUM production.

Production of KERA was not significantly upregulated during keratocyte differentiation

Aniridic corneal stromal cells that differentiated to a keratocyte phenotype (ACK), showed an upregulation of KERA production, as indicated by more intense bands in the western blot membranes (Figure 4.16A). However, this upregulation was not statistically significant (p>0.05). The same trend could be observed in normal corneal stromal cells (Figure 4.16B).

However, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in KERA production.

149

A) LUM B) LUM 2.5 2.5

2.0 2.0

1.5 1.5

1.0 1.0

0.5 0.5

0.0 0.0 Relative expression GAPDH to Relative AC expression GAPDH to Relative NC ACSD NCSD ACK3F- NCK3F- ACK3F+ NCK3F+ LUM (46 kDa) GAPDH

(36 kDa)

KERA C) KERA D) 2.5 2.5 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0

0.0 Relative expression GAPDH to Relative

Relative expression GAPDH to Relative NC AC NCSD ACKF- ACSD NCK3F- ACKF+ NCK3F+ KERA (41kDa) GAPDH (36 kDa)

Figure 4.16. Production levels of keratocyte markers (LUM and KERA) by aniridic (A and C) and normal (B and D) corneal stromal cells under different culture conditions as assessed by Western blot. Results are shown for the basal condition AC and NC, for AC and NC cultured for 3 weeks in KDM supplemented (ACK3F+, NCK3F+, respectively) or depleted of FGF (ACK3F-, NCK3F-, respectively), and for AC and NC cultured for 5 days in CSSCM (ACSD and NCSD, respectively). Graphs showing the densitometry quantification normalised to GAPDH. Data is represented as the mean of n=3 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition.

150

4.4.3.2.5 Gene expression and protein production profiling of myofibroblast markers during directed differentiation of AC and NC towards a keratocyte phenotype

Expression of αSMA is upregulated during keratocyte differentiation

As shown in Figure 4.17A, when prompted to differentiate towards a keratocyte phenotype in KDM supplemented with FGF, aniridic corneal stromal cells (ACK) showed a significant increase (p<0.05) in αSMA expression. However, this upregulation was not significant when ACK were differentiated in KDM supplemented with FGF (Figure 4.17A). The same trends were found for normal corneal stromal cells (NCK; Figure 4.17B). However, NCK significantly upregulated αSMA both when differentiated in the presence and absence of FGF.

In contrast, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in αSMA expression.

A) SMA B) SMA 100 100 * * * * 10 10

1 1

(normalised to NC) to (normalised

(normalised to AC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

Figure 4.17. Relative gene expression levels of the fibrosis marker αSMA (ACTA2) in aniridic (A) and normal (B) corneal stromal cells under different culture conditions. Results are shown for AC and NC cultured in KDM supplemented or depleted of FGF at different time points (ACKi and NCKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (ACSD and NCSD, respectively). Dashed line represents the gene expression for the basal condition AC or NC. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells,

151

Production of αSMA was significantly upregulated during keratocyte differentiation

Results obtained by Western blot were consistent with Real Time qPCR results. Aniridic corneal stromal cells that differentiated to a keratocyte phenotype (ACK), showed an upregulation of αSMA production, as indicated by more intense bands in the western blot membranes (Figure 4.18A). However, this upregulation was not statistically significant (p>0.05). The same trend could be observed in normal corneal stromal cells (Figure 4.18B), but in this case the upregulation of αSMA was significant when normal corneal stromal cells were differentiated to a keratocyte phenotype in KDM without FGF (Figure 4.18B).

When aniridic and normal corneal stromal cells were allowed to spontaneously differentiate in CSSCM (ACSD, NCSD) there was no significant differences observed, compared to the basal conditions (Figure 4.18).

A) SMA B) SMA 3 3 *

2 2

1 1

0 0 Relative expression GAPDH to Relative AC expression GAPDH to Relative NC ACSD NCSD ACK3F- NCK3F- ACK3F+ NCK3F+ αSMA (42 kDa) GAPDH (36 kDa) Figure 4.18. Production levels of the fibrosis marker αSMA by aniridic (A) and normal (B) corneal stromal cells under different culture conditions as assessed by Western blot. Results are shown for the basal condition AC and NC, for AC and NC cultured for 3 weeks in KDM supplemented (ACK3F+, NCK3F+, respectively) or depleted of FGF (ACK3F-, NCK3F-, respectively), and for AC and NC cultured for 5 days in CSSCM (ACSD and NCSD, respectively). Graphs showing the densitometry quantification normalised to GAPDH. Data is represented as the mean of n=3 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

152

Immunofluorescence showed that αSMA was produced by the majority of aniridic and normal corneal stromal cells, indicated by diffused red staining in Figure 4.19. However, only a small proportion of cells that were prompted towards a keratocyte phenotype (ACK and NCK), showed assembly of αSMA into stress fibres. This remained true regardless of the presence or absence of FGF. Interestingly, although densitometry of Western blot did not show a significant upregulation of αSMA in aniridic cells differentiated to keratocytes (ACK), immunofluorescence showed a proportion of cells producing and assembling αSMA into stress fibres. IgG negative controls did not show staining (Appendix, Figure A.8.3).

153

Figure 4.19 - Production and assembly of αSMA by aniridic and normal corneal stromal cells, as assessed by immunofluorescence. Red: Anti-αSMA, Green: Phalloidin, Blue: DAPI. Scale bars: 300µm. Abbreviations NC: AC: Aniridic corneal stromal cells, Normal corneal stromal cells, ACK3F: Aniridic corneal stromal keratocyte-like cells, NCK3F: Normal corneal stromal keratocyte-like cells, ACSD: Aniridic corneal stroma spontaneously differentiated cells, NCSD: Normal corneal stroma spontaneously differentiated cells.

4.4.3.2.6 Gene expression profiling of proliferation marker KI67 during directed differentiation of AC and NC towards a keratocyte phenotype

Expression of KI67 is downregulated during keratocyte differentiation

As shown in Figure 4.20C, the expression levels of KI67 were significantly downregulated (100X- and 1000X-Fold Decrease, p<0.05) when aniridic corneal stromal cells were differentiated towards a keratocyte phenotype (ACK). This remained true

154

regardless of the presence or absence of FGF. The same trends could be observed for normal corneal stromal cells (NCK; Figure 4.20B).

KI67 expression did not significantly change when aniridic or normal corneal stromal cells were allowed to spontaneously differentiate for 5 days in CSSCM (ACSD, NCSD). Interestingly, ACSD showed significantly lower KI67 expression than NCSD, confirming that aniridic cells had lower proliferation capacity when grown for 5 days without passaging in CSSCM (to induce their spontaneous differentiation), Figure 4.20C.

A) KI67 B) KI67 10 10

1 1 0.1 * * ** 0.1 * * 0.01 0.01

0.001 0.001

(normalised to AC) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.0001 level expression Relative 0.0001

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

C) KI67 10 **

1 (normalised to NCSD) to (normalised

Relative expression level expression Relative 0.1

NCSD ACSD

Figure 4.20. Relative gene expression levels of KI67 (gene: MKI67) in aniridic (A and C) and normal (B and C) corneal stromal cells under different culture conditions. Results are shown for AC and NC cultured in KDM supplemented or depleted of FGF at different time points (ACKi and NCKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (ACSD and NCSD, respectively). Dashed line represents the gene expression for the basal conditions AC (A) and NC (B) and NCSD (C). Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05, **p<0.01. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

155

KI67 protein expression levels were not assessed by Western blot as a decline in cell proliferation in all ACK and NCK conditions could clearly be seen by observing them under the microscope (Figure 4.12B).

4.4.3.2.7 Gene expression profiling of apoptosis markers during directed differentiation of AC and NC towards a keratocyte phenotype

Expression of TP53 was significantly upregulated during keratocyte differentiation

As shown in Figure 4.21A, when prompted to differentiate towards a keratocyte phenotype, aniridic corneal stromal cells (ACK) showed a significant increase (over 5X Fold increase, p<0.05) in TP53 expression. This remained true whether ACK were differentiated in the presence or absence of FGF (Figure 4.21A). The same trends were found for normal corneal stromal cells (NCK; Figure 4.21B).

However, neither aniridic nor normal corneal stromal cells that were allowed to spontaneously differentiate when cultured in CSSCM for 5 days (ACSD, NCSD) showed an increase in TP53 expression.

Expression of BAX did not change during keratocyte differentiation

The expression levels of BAX did not change across any of the conditions in aniridic nor normal corneal stromal cells (p>0.05, Figure 4.21C and Figure 4.21D).

Expression of BCL2 did not change during keratocyte differentiation

BCL2 gene expression levels were significantly downregulated (4X-Fold decrease, p<0.05) when aniridic corneal stromal cells were allowed to spontaneously differentiate in CSSCM (ACSD) when compared to the basal conditions (Figure 4.21E).

In addition, BCL2 was also downregulated when normal corneal stromal cells were differentiated towards a keratocyte phenotype (NCK1) in the presence of FGF (5X-Fold decrease, p<0.05).

156

A) TP53 B) TP53 10 * * * 10 * * **

1 1

(normalised to NC) to (normalised

(normalised to AC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

C) BAX D) BAX 10 10

1 1

(normalised to NC) to (normalised

(normalised to AC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

E) BCL2 F) BCL2 10 10

1 1 *

(normalised to NC) to (normalised

(normalised to AC) to (normalised Relative expression level expression Relative Relative expression level expression Relative 0.1 0.1

ACK1 ACK2 ACK3 ACK1 ACK2 ACK3 ACSD NCK1 NCK2 NCK3 NCK1 NCK2 NCK3 NCSD No FGF FGF No FGF FGF

Figure 4.21. Relative gene expression levels of apoptosis markers (TP53, BAX and BCL2) in aniridic (A, C, and E) and normal (B, D, and F) corneal stromal cells under different culture conditions. Results are shown for AC and NC cultured in KDM supplemented or depleted of FGF at different time points (ACKi and NCKi, respectively. i= 1, 2 or 3 weeks) and in CSSCM after 5 days (ACSD and NCSD, respectively). Dashed line represents the gene expression for the basal condition AC or NC. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Kruskal-Wallis test followed by Dunn’s post hoc test was performed. *indicates degree of significance against the basal condition. *p<0.05, **p<0.01. Abbreviations AC: aniridic corneal stromal cells, NC: normal corneal stromal cells.

157

Summary of results:

• ACK, NCK, ACSD and NCSD acquired an elongated spindle shaped morphology over time.

• ACK and NCK have lower proliferative capacity than AC, NC, ACSD and NCSD.

• ACSD have lower proliferative capacity than NCSD.

• ACK and NCK acquired an elongated spindle shaped morphology after differentiation for 3 weeks in the presence or absence of FGF.

• ACK and NCK have different gene expression patterns when compared to their basal conditions (i.e. the phenotype from which they were differentiated). AC and NC, respectively:

o Gene expression levels of CD90, KERA, LUM, αSMA and TP53 are upregulated in ACK and NCK.

o Gene expression levels of KI67 was downregulated in ACK and NCK.

• ACK and NCK have different protein production patterns when compared to their basal conditions from AC and NC, respectively:

o CD90, KERA, and αSMA protein levels are upregulated in ACK and NCK.

• ACSD and NCSD have gene expression profiles similar to AC and NC.

158

4.5 Discussion

In the previous chapter, it was shown that CSSCM is a suitable medium to culture and expand central corneal stromal cells. These exhibited stem-like characteristics and maintain their ability to differentiate back into keratocyte-like cells. Corneal stromal cells from ARK patients have never been isolated and cultured in vitro previously. The aim of this chapter was primarily to establish whether corneal stromal cells isolated from ARK corneas could be successfully cultured in vitro in CSSCM and; secondly to assess whether aniridic corneal stromal cells can be also differentiated towards a keratocyte- like phenotype.

Implications of obtaining ARK samples for corneal stromal cell isolation

Aniridic cells were effectively isolated and cultured from fresh post-surgical samples. For in vitro modelling of ARK, CSSC isolated from the limbus would have been the optimal cell source to study this disorder, since at least in the healthy cornea they express PAX6 [57]. However, it was not possible to safely obtain a limbal biopsy because ARK patients are prone to aberrant wound healing and scarring during surgery and manipulation [154]. Hence, redundant central buttons from aniridic patients, for whom surgery had become necessary, were utilised for this project. This is the reason why so few human ARK tissue samples are available for research, despite Moorfields Eye Hospital having the largest patient cohort in the UK and possibly Europe. To the best of our knowledge, there has not been any previous study using human aniridic corneal stromal cells, which brings novelty and originality to the project.

Some of the aniridic patients whose corneas were used in this study had already had previous surgeries and transplants. A concern for this study was that normal cells from previous donor grafts might have been still populating the tissue samples harvested from the aniridic patients. Upon transplantation donor epithelial cells are promptly replaced by the host cells within days [215]. However, very different results have been observed regarding allogenic stromal and endothelial cells. These have been found in the graft even 30 years after transplantation. However, there is high variability in the occurrence of stromal cells remaining in the graft, with the rate ranging from 4 to 95%, 30 years postransplanation [215]. If donor cells had been populating the grafts used for this study, cells without the PAX6 mutation could have been used inadvertently. However, genotyping performed from the aniridic corneal cells in culture revealed that they had the PAX6 mutation, matching the one found in the patients’ blood, suggesting that the cells used were patient specific. 159

Pathogenicity of the PAX6 variants found in the aniridic samples

Three PAX6 variants were found in the open reading frame of the gene in all four aniridic patients used in the present work. AC1 and AC4 had a CTE mutation and AC2 and AC3 had PTC mutations (substitution and frame-shift, respectively). The three different mutations found here have been already described and related to aniridic phenotypes in the literature, which adds evidence that they are pathogenic mutations. Back in 1999, Baum and colleagues published a mutation report where they described for the first time the run-on mutation found in AC1 and AC4 [222]. It was a change on the stop codon, caused either by an insertion of a T or a transversion from an A to a T, producing an extended protein. Since then, the same mutation has been described in more articles [223]. In 2006, Hever and colleagues published an article with new mutations they had found in their patient screenings. They also described the mutation found in AC2, in which a G is mutated to a T, resulting in an amino acid change from Glycine to a premature termination codon. They suggest that non-mediated decay system will act in this case and probably there will not be protein synthesis [224]. Interestingly, this is the same mutation that was described in the Sey mouse model (Gly208*; NP_001231127.1) [163]. Finally in 1998, Love and colleagues described the mutation found in AS3 as the deletion of two bases in exon 7, an A and a G at position 735-736, produced a stop codon seven codons later, truncating the protein in the paired domain [225].

Hingorani et al. previously attempted to find a correlation between genotype and phenotype in aniridia, but their study included in a wide range of clinical observations in a cohort of 43 patients [226]. Correlations between the mutation variant and stromal cell behaviour in vitro would be of the upmost interest to be established. However, insufficient ARK donor samples are available to yet have representative numbers of each mutation.

Aniridic corneal stromal cells have lower proliferation capacity, but otherwise present the same phenotype as normal corneal stromal cells when cultured in CSSCM

In the results here presented, aniridic corneal stromal cells were successfully isolated and cultured in vitro for the first time. The propagation of these cells opens new and exciting possibilities for modelling ARK in vitro by using primary cells. Here, cells were cultured in CSSCM for a number of different reasons, as described in detail in Chapter 3. Keratocytes in the native corneal stroma do not express PAX6, on the other hand, their originating progenitor cells in the limbal stroma do express this marker [57]. Since aniridia is caused by PAX6 mutations, culturing the cells isolated from the central stroma in CSSCM might drive them towards more primitive states resembling those of CSSC 160

cells, where PAX6 is involved in modulating the cell phenotype [57]. Alternatively, aniridic keratocytes, derived from CSSC with PAX6 mutations, could somehow retain functional changes that were pre-determined earlier in their PAX6 expressing development stage.

Aniridic corneal tissue affected by ARK and removed during surgery has usually undergone conjunctivalisation. This is a result of an impairment in limbal function and leads to a migration of conjunctival epithelium and its blood vessels onto the cornea [227]. It is therefore plausible that stromal cells from the conjunctiva were also isolated and expanded together with the corneal stromal cells. However, currently there is no known way to avoid this. Similarly, epithelial cells may have also been present in the primary cultures, however, these were easily be removed using selective trypsinisation [221].

4.5.3.1 AC and NC show similar morphologies and expression profiles of putative CSSC, keratocyte, and apoptosis markers

Similar to NC, aniridic cells presented a small square-like shape and similar passaging frequency requirements when cultured in CSSCM at low density. Both NC and AC showed very low expression of PAX6 and, strikingly, in some aniridic donors PAX6 was even undetectable. However, AC and NC were used for testing gene expression at passage 5 (to obtain sufficient cell numbers for experiments), and it is known that the expression of PAX6 is lost with passaging [198]. Interestingly, AC and NC expressed similar levels of the putative CSSC markers, CD90 and CD73, both at mRNA and protein level, suggesting that PAX6 is not involved in regulating the expression of these two markers. When NC and AC were cultured in CSSCM at low density, no differences in expression were observed in the remaining tested markers (LUM, KERA, ALDH1A1, KI67, and TP53), suggesting that AC and NC have a similar capacity to expand and acquire the stem cell-like properties reported for CSSC.

4.5.3.2 AC have lower proliferation capacity than NC

Aniridic corneal stromal cells had lower proliferation capacity than normal cells when cultured for 7 days in CSSCM. These results are consistent with numerous articles published that support the involvement of PAX6 in the regulation of the cell cycle and proliferation [130, 144, 165, 228]. Ramaesh and Douvaras showed that in the Pax6+/- mouse model, corneal epithelial cells presented increased proliferation and higher cell turnover compared to wild-type controls, resulting in abnormal corneal homeostasis [144, 165]. In a different approach, overexpression of Pax6 resulted in increased [130] or reduced [228] proliferation in genetically engineered mouse or rabbit corneal epithelial

161 cells, respectively. The contradiction in these results is maybe due to different Pax6 dosage (haploinsufficiency vs overexpression), different animal models, and experimental settings. However, they all suggest that PAX6 plays a role in determining the corneal epithelia dynamics [130, 228]. PAX6 has also been shown to modulate proliferation in other cell types such as retinoblastoma [229], retinal progenitors [230], cerebral cortical cells [231] and colorectal cancer cells [232]. Interestingly, Hsieh and colleagues found highly variable PAX6 expression levels during different stages of the cell cycle and in different stages of retinal cell differentiation, suggesting that the dosage of PAX6 and its involvement in deciding cell dynamics is a complex process [230]. Importantly, it is noteworthy that epithelial cells have a different proliferation profile than stromal cells in the native cornea. Peripheral corneal epithelial cells constantly proliferate to repopulate the central cornea when the cells are shed by desquamation [233], whereas stromal cells remain quiescent until the event of an injury [22, 51]. This may help to explain why results obtained from corneal stromal cells might not directly correlate with epithelial cell behaviour. There are only a restricted number of studies that look into the numbers and proliferation of stromal cells in Pax6 haploinsufficient models [157, 211]. Ramaesh et al. in described a hypercellular stroma prenatally in the Pax6+/- mouse model [211]. Others showed lower density of stromal cells in human aniridic corneas in vivo [157]. Results published herein are, to the best of our knowledge, the first specific evidence of PAX6 involvement in regulating the proliferation of human corneal stromal cells, in vitro. However, in these experimental settings the cells are in a proliferative state, growing in the presence of serum in vitro, and therefore the results do not necessarily reflect how keratocytes would behave in the native stroma. However, the results here present are a first indication that AC have some functionality impairment compared to NC.

4.5.3.3 AC and NC show similar migration capacities

Aniridic and normal corneal stromal cells did not present significant differences in their migration capacity, as measured by the chemotactic transwell assay. These results are perhaps surprising as there has been extensive research showing that Pax6 haploinsufficient corneal epithelial cells have delayed migration ability [129, 130, 139, 234]. Collinson and colleagues found that the centripetal migration of corneal epithelial cells was diverted and delayed in an in vivo wound healing model of the Pax6+/- mouse model [129]. In a later article, they suggested that the centripetal migration of the epithelim is directed by contact mediated cues from the underlying stroma. In the Pax6+/- mice, epithelial cells fail to read these cues and so the migration is impaired [234]. However, peripherial epithelial cells in healthy corneas are constantly migrating 162

centripetally to replace cells desquamated in the central cornea, and maintain the epithelium homeostasis [233]. On the contrary, stromal cells only migrate in the event of an injury that affects this layer [62]. It is possible that using a different chemoatractant or an alternative colony dispersion assay, i.e. where the extent of cell migration is away from a previously confined confluent area, may have revealed different results [235]. Interestingly, when the results were evaluated for each aniridic donor separately, AC2 migrated significantly more than NC (p<0.05). No further comment can be made due to the low number of donor tissues, however these results are an indication that different PAX6 mutations might have an effect on cell migration.

4.5.3.4 AC and NC show similar collagen contraction capacities

Corneas presenting ARK usually develop scarring at some point during the lifespan of the aniridic patients [61]. The collagen matrix contraction assay has been previously used as a measure of cell ability to re-organise ECM (as an in vitro surrogate of the contractile phase of scarring) [236]. Surprisingly, when collagen contraction capacity was assessed for AC and NC in vitro there were no significant differences between them nor between the three aniridic donors with different PAX6 variants. These results might suggest that there are no differences in scarring abilities between the two cell populations, in the conditions tested. However, these results are unexpected because ARK corneas are prone to scarring [154], and previous studies have shown a high presence of myofibroblasts in their stroma [161]. It would be expected that a higher number of contractile myofibroblasts would increase the rate of collagen gel contraction.

However, the results of the migration and collagen contraction revealed high donor to donor variability. This may have contributed to the lack of significant differences found between aniridic and normal cells when the results for all the donors in each condition are pooled together. This variability could be due to multiple factors, such as the different mutations in the aniridic cells, intrinsinc donor to donor variation which is often seen in cultures derived from ‘normal’ subjects, different cell cycle stage or differentiation state of the stromal cells, and experimental settings. The latter were controlled as far as possible to maximise experimental reproducibility.

In conclusion, AC and NC cultured sparsely in 2D in CSSCM did not show differences in morphology, gene expression, migration and contraction capacity. However, AC presented lower proliferation capacity when grown to confluece for a period of one week in CSSCM . These last results supported the hypothesis that AC and NC have a different phenotype when cultured in vitro in CSSCM.

163

Aniridic and normal corneal stromal cells showed similar capacity to differentiate into keratocyte-like cells

As discussed in the previous chapter, NC cultured in KDM for 3 weeks, with or without FGF, showed upregulation of keratocyte markers. Interestingly, aniridic corneal stromal cells showed the same ability, suggesting that they may be capable of reverting back to a more quiescent keratocyte-like phenotype (as opposed to the presumed wound healing phenotype of the pannus). This similarity does not support the hypothesis that aniridic cells differentiate less readily to a keratocyte phenotype. These results are perhaps surprising because some previously published evidence pointed to differences in the aniridic stromal cells when compared to normal corneal stromal cells in vivo (keratocyte phenotype) [161]. In a recent study, Vicente et al. showed abnormalities in the structure of the anterior stroma with a presence of a subepithelial pannus in all tested aniridic samples [161]. In this region, keratocytes were strongly labelled with antibodies against vimentin and αSMA, markers that are upregulated in myofibroblasts during wound healing [237] and consistent with the high stromal scarring observed in aniridic patients [238]. However, the posterior stroma of the same aniridic samples was unaffected and presented the same structure and staining than the normal stroma [161]. This suggest that cells extracted from different regions of aniridic stroma have different behaviours with distinct protein expression profiles. In their study, the size of the subepithelial pannus varied between aniridic donors [161], which suggests that depending on the severity of the phenotype, the isolated corneal stromal cells might present different phenotypes. Despite being extracted from the anterior stroma, the aniridic cells used in the present study showed the same phenotypical characteristics than their normal non- diseased counterparts. When cultured in CSSCM, aniridic cells showed stem-cell like properties, and when differentiated for 3 weeks in KDM they showed a healthy keratocyte-like phenotype; high expression of keratocyte markers and expression of αSMA to similar levels of those of healthy keratocytes. Taken together, these observations suggest that (i) aniridic cells cultured in CSSCM can acquire a CSSC phenotype and (ii) can revert to a keratocyte-like phenotype when cultured in KDM for three weeks.

Aniridic cells cultured in KDM did not significantly upregulate ALDH1A1 expression. Nevertheless, normal cells neither upregulated this marker when they were prompted to differentiate towards a keratocyte phenotype. Interestingly, previously published results showed a downregulation of this marker in aniridic corneal epithelial cells [135]. Since ALDH1A1 is a corneal crystalline that helps maintain the transparency of the cornea [49],

164

its downregulation in ARK epithelium might help understanding the corneal opacification on aniridic patients [135].

Both aniridic and normal stromal cells signifcantly upregulated TP53 when cultured in KDM for 2 and 3 weeks, suggesting that keratocyte-like cells are more apoptotic. However, TP53 is a gene involved in many other different cellular processes and the upregulation of this marker alone is not sufficient to make this affirmation [239]. Cell apoptosis is regulated by TP53 in conjunction with other genes, such as BAX and BCL. The first is an proapoptotic and the second an antiapoptotic gene [240]. The trends of BAX (upregulation) and BCL2 (downregulation) were mirrored in keratocyte-like cells, suggesting that the cells cultured in KDM are more apoptotic. Nevertheless, the upregulation of BAX or downregulation of BCL2 was not statistically significant in the majority of the conditions in either aniridia or normal. On the other hand, it is noteworthy that the aniridic stromal cells grown for 5 days in CSSCM (ACSD) had a significant downregulation of the anti-apoptotic marker BCL2, suggesting that they might also be more susceptible to apoptosis. These results are consistent with the study by Ramaesh and colleagues in which a higher and abnormal apoptosis rate was observed in aniridic stromal cells during wound healing in a Pax6+/- mouse model.

Interestingly, there were not major differences between when AC were prompted towards a keratocyte phenotye in KDM (ACK) with FGF or without FGF. The only difference was a significantly higher production of CD90 when ACK were cultured in the absence of FGF. However, this was not accompanied with a significant increase in production of αSMA, which would suggest a shift towards a myofibroblast phenotype. These results do not support the hypothesis that FGF is an essential growth factor to drive the differentiation of AC towards a keratocyt phenotype.

Results presented here were obtained from cells cultured in 2D, however, as it has previously been demonstrated, spatiotemporal features of a 2D environment profoundly affect cell behaviour [241]. In the native cornea, stromal cells produce high amounts of ECM [242]. Matrix deposition in 2D may be limited by the availabitility of space which may lead to hypertrophic phenotyopes as demonstrated in chondrocytes [243]. Therefore, the data presented in this chapter will be followed by an investigation of aniridic stromal cell behaviuor in 3D.

In conclusion, aniridic corneal stromal cells were sucessfully isolated from patient samples and when cultured sparsely in CSSCM did not present differences compared to the normal controls. They also presented the ability to differenciate into keratocyte-like cells in vitro, feature that was extensibly discussed in the previous chapter. However,

165 when cultured confluent in CSSCM, aniridic cells had lower proliferation capacity and showed downregulation of an antiapoptotic marker. This evidence, consistent with other published studies [244], supports the hypothesis that PAX6 might be involved in the regulation of cell proliferation and apoptosis in human corneal stromal cells.

166

CHAPTER 5: ANIRIDIC AND NORMAL CORNEAL STROMAL CELLS PRESENT DIFFERENCES IN PROLIFERATION, DISTRIBUTION AND GENE EXPRESSION WHEN CULTURED INSIDE THE RAFT TE

5.1 Introduction

In vivo, cells are found in complex settings including extracellular matrix proteins and autocrine, paracrine and endocrine signalling factors that provide essential information about their surroundings. It is well recognised that the extracellular microenvironment influences cell phenotype and controls a wide range of biological processes (reviewed in [245]). For instance, ECM modulates patterns of cell migration [246]. Cells within a matrix need to overcome its resistance to be able to migrate, either by changing cell shape or by remodelling the matrix [246]. ECM also regulates gene expression, as shown in rabbit keratocytes that upregulated the keratocyte marker ALDH1A1 and downregulated the fibrotic marker αSMA when cultured in a 3D collagen sponge in vitro, compared to a 2D culture [241]. During development, the mechanical interactions between corneal stromal cells and the early deposited stromal ECM strongly regulate keratocyte behaviour, modulating cell alignment and morphology [247]. Similar results have also been observed in vitro, where corneal fibroblast spreading, morphology and alignment was influenced by the mechanical properties of the surrounding ECM [248].

Experiments performed in vitro in two-dimensional (2D) tissue culture can provide some insight into cell behaviour. However, such 2D conditions may have limited physiological relevance regarding human cell function [249]. 3D tissue equivalents can be constructed from a wide variety of different materials with tuneable features. These include a range of polymers, copolymers solutions, and collagens gels [250]. From these, collagen gels have many advantageous properties as they are biocompatible, are naturally remodelled by the cells, cause low immunogenic response and are inexpensive to obtain [251]. Moreover, type I collagen gels are suitable to use in cornea research since this protein represents the principal structural element in vivo [21, 22]. Non-compressed hyper- hydrated collagen gels are a suitable platform to culture corneal stromal cells, however due to their high water content, they are inherently weak, and their mechanical properties highly differ from those of the cornea [252, 253]. Brown et al. developed a method for using a 3D compressed type I collagen gel with strong mechanical properties as a substrate for cell culture [251]. From then on, these compressed collagen gels have been extensively used in our laboratory to culture different types of cells, including corneal 167 endothelial cells [254], limbal epithelial stem cells [253], limbal fibroblasts [191, 255] and CSSC [55, 256]. The majority of studies using the various corneal stromal cell phenotypes (limbal fibroblasts and CSSC) aim to assess their supportive role in the coculture of epithelial cells [55, 191]. In the limbal niche, CSSC are in close proximity to the LESC, therefore these two cell types have been cocultured in the compressed collagen gels to study their interactions [55]. However, Shojaati et al. recently showed that compressed collagen gels are suitable delivery vehicles to culture CSSC alone without epithelial cells. In their study they transplanted the CSSC populated tissue equivalent into a mouse corneal wound healing model, reducing visible scarring and suppressing the expression of fibrotic markers [256]. A brief characterisation of CSSC cultured in the compressed collagen gels for 13 days showed that they maintained their expression of the CSSC-associated markers CD73 and CD90, but did not express the myofibroblast marker αSMA [55, 191]. These data is encouraging because expression of αSMA would suggest a shift towards a scarring phenotype [79], which is not desired if the final goal is to transplant the CSSC [256]. Taken together, these studies showed that limbal stromal cells can be successfully embedded within the compressed collagen gel and remained viable for at least 2 to 3 weeks, making it a suitable matrix to culture and study cells in vitro. Hence in this chapter, aniridic and normal corneal stromal cells were further characterised within a compressed collagen type I gel, in an attempt to mimic aspects of the natural in vivo microenvironment.

The preparation process of making compressed collagen gels has evolved over time. The first compressed collagen gels produced by Levis et al. in 2010 consisted of neutralised hyper-hydrated collagen gels compressed between nylon mesh backed with filter paper. Water content inside the hydrogel was reduced by applying a load on top of the gels [253]. Since then, the compression process has been modified to improve reproducibility and tunability of the compressed collagen gels, and to make the process more compatible with Good Manufacturing Practice (features required to meet the regulations to be used for human therapy). In 2013, a new standardised approach was developed by Levis and colleagues in collaboration with TAP Biosystems (now Sartorius). The hyper-hydrated collagen gels were confined inside tissue culture plate wells and sterile absorbers were place on top to compress them by removal of water [257]. The compressed collagen gels fabricated using this new process were termed Real Architecture for 3D Tissues (RAFT), here referred as RAFT tissue equivalents (RAFT TE). RAFT TE may be suitable for two different applications; (i) to be used as a matrix to support cell therapy, and (ii) for disease modelling in vitro (the focus of this thesis).

168

To the best of our knowledge, RAFT TE have not previously been used to culture stromal cells isolated from central cornea (normal or diseased). Furthermore, human aniridic corneal stromal cells have never been cultured in RAFT TE either, to enable their properties to be studied. During this project, RAFT TE were used for the first time as a simple model to mimic the stromal environment of normal versus aniridic cornea. As it was unknown whether aniridic corneal stromal cells could even be propagated in vitro, the first step was to culture and characterise cells in 2D, as described in the previous chapter. Encouraging results showed that aniridic corneal stromal cells could be expanded and cultured in vitro. Differences in proliferation were found between aniridic and normal corneal stromal cells, but surprisingly, the other phenotypical characteristics tested did not show significant differences between both cell types. In this chapter aniridic cells were further characterised in RAFT TE and compared to normal controls. CSSC, keratocyte and fibrosis markers studied in the previous chapter were assessed, to find possible differential expression driven by the 3D environment. Also, because of the published evidence supporting differences in apoptosis and proliferation in aniridic corneal stromal cells in vivo, these parameters were studied in the cells in RAFT [169, 208, 211]. Additionally, a comparison of aniridic and normal stromal cell transcriptomes was performed using RNA sequencing analysis. As far as we are aware, this is the first time that next generation sequencing has been used to study human aniridic corneal stromal cells.

5.2 Hypotheses and aims

The hypotheses tested in this chapter were the following:

1. Viable aniridic and normal corneal stromal cells can be cultured and maintained inside RAFT TE. 2. Aniridic and normal corneal stromal cells have different phenotypical characteristics when cultured inside the RAFT TE.

3. Aniridic and normal corneal stromal cells present differences in their transcriptomes.

Hence, the overarching aim of Chapter 5 was to culture and characterise aniridic and normal corneal stromal cells inside the RAFT TE, by assessing morphology, rates of proliferation, susceptibility to apoptosis, and gene expression profile.

169

5.3 Materials and Methods

Real Architecture for 3D Tissue

RAFT tissue equivalents (RAFT TE) were prepared to culture stromal cells in an attempt to mimic their residing conditions in the native tissue. Preparation of the gels was performed following a protocol described by Levis et al. with modifications [258] (Figure 5.1). Briefly, a mixture containing 0,9mL MEM 10X, 7,4mL rat-tail collagen type 1 (2.05mg/mL in 0,6% acetic acid, First link, UK), 0,546mL neutralising solution (TAP Biosystems) and 0,377mL cells suspended in CSSCM was prepared by mixing according to the order indicated. A total number of 140000 cells/gel was used. The mixture was prepared and then left on ice for 30 minutes to allow elimination of air bubbles. A total volume of 2.4mL was added to each well of a 24 well plate (Greiner Bio-One) and the plate was incubated at 37°C for 30 minutes inside a plate heater (TAP Biosystems) to allow collagen polymerisation. UV pre-sterilised absorbers were then added on top of the gels and left for 15 minutes at 37°C to compress the gels through removal of water. When the absorbers were removed, 1.5mL of CSSCM was added.

Figure 5.1. Schematic representation of the RAFT TE manufacturing process. From left to right, a mixture of medium, collagen, neutralising solution and cells was added to the wells and left to polymerise for 30 min at 37°C. The gel was then compressed by removing the excess of water with a sterile absorber. When the absorber was removed, fresh medium was added to the wells. Abbreviations: RAFT TE: Real Architecture for 3D tissue equivalent.

Experiment design – Donor demographics

Aniridic and normal corneal stromal cells from eight different donors (four for each condition) were cultured inside the RAFT TE for three weeks. Donor information can be

170

found in Chapter 4, Table 3.1. A total of three RAFT TE were prepared for each donor. Additional RAFT TE without cells were used as negative controls.

Metabolic activity – PrestoBlue ®

At the end of a 3-week time course, stromal cell metabolic activity within RAFT TE was measured by the PrestoBlue® assay (Thermo Fisher Scientific) as per manufacturer’s instructions. Briefly, 250μL PrestoBlue® reagent diluted in 10% (V/V) CSSCM was added to each RAFT TE containing well. The same volume of PrestoBlue® reagent was added to two extra wells without RAFT TE and used as a blank. The plate was then incubated for 15 minutes at 37°C in the dark. Two technical replicates were measured from each RAFT TE and blank well, by transferring 100μL/well of the solution into two wells of a black flat bottom 96-well plate (Nunc™). Fluorescence was red in a microplate reader (Excitation wavelength: 560nm, Emission wavelength: 590nm).

Real-Time qPCR

5.3.4.1 RNA isolation from cells inside the RAFT TE

At the end point (3 weeks), the culture medium was removed from the wells and 500µL of TRIzol® reagent (Invitrogen™) were added directly on top of the RAFT TE. The resulting solution was incubated at room temperature for 30 minutes and occasionally pipetted up and down to dissolve the RAFT TE. The solution was then transferred into a 1,5mL Eppendorf and any remains of the RAFT TE were dissolved by vortex. Chloroform (100µL) was added to the previous resulting solution and mixed by strongly handshaking the eppendorf for 15 seconds. The resulting solution was then centrifuged at 11000 rpm for 5 minutes to allow its separation in three different phases. The clear phase on top, containing the RNA, the pink phase at the bottom containing the proteins and the interphase containing the DNA and cell waste. The maximum possible volume (roughly 500µL) was taken from the aqueous phase, without disturbing the other phases, and transferred into a labelled new eppendorf. Isopropanol (half volume of the obtained clear phase) was added and mixed by inverting 3 times. The resulting solution was incubated overnight at -20°C to allow the RNA to precipitate. On the following day, the solution was added into a RNeasy Mini Spin Columns and processed as described in Chapter 2 section 2.2.1.

5.3.4.2 cDNA synthesis and Real Time qPCR

Further steps for cDNA synthesis and Real Time qPCR were performed as previously explained in section 2.2.3 in Chapter 2. Primers used are shown in Table 2.1.

171

Cell characterisation

5.3.5.1 Live/Dead viability analysis

After 3 weeks of culture the RAFT TE were washed once with DPBS to remove any serum remaining and a straight strip was cut in the centre of the RAFT TE to be used for Live/Dead analysis. The two remaining pieces of the RAFT TE were fixed for later immunocytochemistry.

The LIVE/DEAD® Viability/Cytotoxicity Kit (InvitrogenTM) relies on plasma membrane integrity and esterase activity of the cells to determine their viability. It is based on the detection of either the dye calcein or ethidium homodimer. Calcein is retained intracellularly if the cells are alive, enhancing its green fluorescence, whereas ethidium homodimer enters the cells if the membrane is compromised and enhances its red fluorescence when bound to nucleic acids (information retrieved from manufacturer’s datasheet).

Figure 5.2. Schematic representation of the protocol for Live/Dead analysis. A strip of the RAFT TE was cut with a scalpel to be tested. A small piece was cut and incubated for 30 minutes in 70% methanol and used as a positive control for cell death. Both pieces were then stained and imaged. White squares represent areas that were imaged: for each stripe 9 pictures at 3 different depths were taken, from both edges and central part of the RAFT TE. Abbreviations: RAFT TE: Real Architecture for 3D tissue equivalent.

A dead cell control was prepared by fixing the cells for 30 minutes with 70% methanol. The Live/Dead assay reagents were removed from the freezer and warmed at room temperature before use. The working solution was prepared by adding 20μL of EthD-1 and 5μL of Calcein to 10mL of DPBS. In a 12 well plate, 500μL of working solution was added to each well with a RAFT TE strap or the dead cell control and incubated for 45 minutes at room temperature in the dark. The resulting samples were then transferred

172

and mounted on microscope slides with Live/Dead working solution and the coverslips were secured with nail polish.

The samples were imaged under the Zeiss LSM 710 confocal microscope. Images were taken at three different depths, (top, middle and bottom) at three different areas of the RAFT TE, namely the two sides and the middle (Figure 5.2).

5.3.5.2 Immunocytochemistry

After three weeks of culture, the RAFT TE were washed once with DPBS and fixed by adding 400µL of 4% (V/V) PFA in DPBS. They were then washed 3 times for 5 minutes in DPBS, and finally stored in DPBS at 4°C until further use.

After fixation, the RAFT TE were stained following a protocol described by Kureshi et al. with small modifications [55]. Briefly, RAFT TE were incubated for 10 minutes in 0.25% (V/V) Triton-X 100 in DPBS to permeabilise the cell membranes. They were then incubated for 1 hour in 5% (V/V) goat serum in DPBS to block and prevent unspecific staining. Samples were then incubated for 1 hour at room temperature in the dark with Phalloidin 418 (Sigma-Aldrich) (1:1000) in DPBS. After 3 washes with DPBS, they were mounted in Vectashield mounting medium with DAPI (Vector Laboratories) and visualised under a confocal microscope (Zeiss LSM 700).

5.3.5.3 Quantification of cell alignment and circularity

Three biological donors were used for AC and NC, respectively, to calculate cell alignment and circularity. For each RAFT TE, five different image fields were obtained under the confocal microscope. The confocal images were processed in Fiji ImageJ 1.49s software. Shape descriptors (area, perimeter and position) were calculated from different image fields of each RAFT TE. Briefly, all cells in the imaged field were manually traced around their membrane borders, and the following parameters were recorded: area, perimeter and orientation angle. To obtain the orientation angle of each cell, ImageJ fits the cell into an oval and records the position angle of its longest axis.

Cell circularity was calculated using the following formula:

Equation 2 – Cell circularity. 퐴푟푒푎 퐶푖푟푐푢푙푎푟푖푡푦 = 4 휋 × 푃푒푟푖푚푒푡푒푟2

Circularity = 1 for a perfect circle. Circularity < 1 for elongated bodies.

173

To calculate cell alignment, a line was drawn on the direction in which the majority of cells were aligned, and its angle was recorded and compared to the position angles for each cell. These were sorted into bins of 10° displacements. The percentage of cells that fell into each of these bins was plotted. Cell with 0° would be perfectly aligned, and 90° would be perpendicular. Cell populations with a random orientation should have an average 45° angle [259].

5.3.5.4 Scanning Electron Microscopy

I would like to acknowledge Miss. Rita Pinho for her help during the preparation, processing, and analysis of SEM samples. At 3 weeks, RAFT TE were fixed with 4% (V/V) PFA in DPBS for 30 minutes and washed three times with DPBS for five minutes. They were then placed in a 1:1 water and osmium solution with rotation for 2 hours to enhance the contrast of cells, as osmium binds preferently to unsaturated bonds in the lipid bilayers. Subsequently, they were washed three times with water and dehydrated in a series of ethanol concentrations (V/V) in distilled water (50%, 70%, 90%, and 100%) for 10 minutes in each, followed by a final wash in 100% ethanol for 10 minutes. The samples were then dried using the critical point drying method by which they are placed in 100% ethanol inside the appropriate holder and go through a series of liquid CO2 immersion to remove the ethanol and dry. Finally, the samples were gold sputter coated (5nm, Bal-Tec) and imaged under a field-emission scanning electron microscope (Zeiss) at 3kV or 5kV.

RNA sequencing

RNA was isolated from aniridic (AC1, AC2, AC3) and normal (NC1, NC2, NC3) corneal stromal cells cultured in RAFT TE for 3 weeks as previously described in section 5.3.4.1. Sample quality was also assessed using Nanodrop. The samples were then sent to GENEWIZ Ltd for RNA sequencing. Sample requirements were the following: RNA concentration ≥ 50ng/μL and Quantity ≥ 2μg. Sample quality control was assessed again at GENEWIZ Ltd, by measuring the RNA integrity number (RIN).

RNA sequencing was performed by GENEWIZ Ltd. The first step of library preparation was performed via polyA selection, to enrich the mRNA. This was followed by cDNA synthesis and sequencing. Sequencing configuration was as follows, HiSeq, 2x150bp configuration, single index, per lane. Data quality assessments were then performed to examine the distribution of read counts in the libraries from the different samples and to assess the similarity between the different samples. Sequence reads were trimmed to remove possible adapter sequences and nucleotides with poor quality using Trimmomatic v.0.36. The trimmed reads were then mapped to the Homo sapiens 174

GRCh38 reference genome available on ENSEMBL using the STAR aligner v.2.5.2b, which incorporates splice junctions to help align the entire read sequence. Unique gene hit counts that fall within exon regions were calculated by using featureCounts from the Subread package v.1.5.2. The data extracted from the gene hit counts was used for downstream differential expression analysis between conditions. The comparison of gene expression was performed using DESeq2, and the Wald test was used to generate p-values and log2 fold changes. Genes with an adjusted p-value < 0.05 and absolute log2 fold change > 1 were called as differentially expressed genes, for each of the comparisons. Finally, significantly differentially expressed genes were clustered by their gene ontology and the enrichment of gene ontology terms was tested using Fisher exact test (GeneSCF v1.1-p2). The goa_human GO list was used to cluster the set of genes based on their biological processes. Finally, to estimate the expression levels of alternatively spliced transcripts, the splice variant hit counts were extracted from the RNA-seq reads mapped to genome, using DEXSeq.

GENEWIZ Ltd provided the following documents:

- Raw data o Raw data in fastq format - Aligned data o Alignment file in bam format - Hit-counts o Unique gene hit counts for each sample - Differential gene expression analysis results o Raw gene hit counts of all different samples used for comparison o Normalised gene hit counts of all samples used for comparison o Rlog transformed gene hit counts of all samples used for comparison o Log2 fold change and adjusted p-value table o Log2 fold change and adjusted p-value table for significant differentially expressed genes (DEGs) - GO enrichment analysis results o GO enrichment analysis table - Differential splicing analysis results o DEXSeqReport GENEWIZ Ltd the documents presented above from three different comparisons of the RNA sequencing data: (i) NC (all donors) compared against AC (all donors), (ii) NC (all donors) against AC1 (donor with CTE mutation), and (iii) NC (all donors) against AC2/AC3 (donors with PTC mutation). 175

The graphs for RNA sequencing data visualisation used herein were made using RStudio [260]. Scripts used to create the Volcano plots, Venn diagrams, Heatmaps, and Gene Ontology graphs can be found in the Appendix.

A list of the 50 most significant DEGs can be found in the Appendix (Table A.8.4, Table A.8.5, and Table A.8.6), for the three different comparisons.

Statistical analyses

Excel (Microsoft) was used to organise and study the data and Graphpad Prism (GraphPad) was used to perform statistical analysis and to plot the graphs.

Three technical replicates (three independent RAFT TE populated by cells) were used for each biological replicate (n=3), giving a total sample size of 9. Only in Real Time qPCR data, an extra biological replicate was used (total n=4), as a new donor became available before performing these experiments.

After confirming that the dataset followed a normal distribution, one-way ANOVA was used to analyse metabolic activity data measured by PrestoBlue®, where n=3 technical replicates for each n=3 biological replicates were used. This was followed by Bonferroni’s multiple comparisons test.

After confirming that the datasets followed a normal distribution, unpaired student t-test was used to analyse gene expression values of all genes, when comparing all AC pooled together against NC. When results were shown for all different aniridic samples individually, non-parametric Mann-Whitney test was performed as the normality of the data set could not be assessed, due to the small sample size (n=3).

Unpaired student t-test was also used to analyse cell circularity data, after confirming that the data set followed a normal distribution.

Statistical analysis used in the RNA sequencing results are specified in section 5.3.6.

5.4 Results

Aniridic corneal stromal cells have lower proliferation capacity than normal corneal stromal cells

Both AC and NC showed significantly higher metabolic activity than the acellular RAFT TE (p<0.001) (Figure 5.3A). Interestingly, AC showed significantly lower metabolic activity than NC when cultured inside RAFT TE (p<0.001) (Figure 5.3A).

Additionally, proliferation capacity was assessed by studying the gene expression levels of the proliferation marker KI67 by qPCR. AC showed significantly downregulated levels

176

of KI67 expression compared to NC (7X Fold decrease, p<0.001), suggesting lower proliferation capacity (Figure 5.3B). However, the results appeared to vary depending on the aniridic donor, therefore KI67 relative gene expression was also plotted individually for each aniridic donor (Figure 5.3C). Interestingly, AC1 and AC4 showed significant lower levels of Ki67 compared to NC (220-320-Fold Decrease, respectively p<0.05), whereas AC2 and AC3 did not show significant differences.

A) *** 1500 ***

1000 ***

500

Metabolic activity Metabolic (% fluorescence A.U) fluorescence (% 0 No cells AC NC

B) KI67 C) KI67 10 *** 1

1 0.1 0.1 0.01 * *

0.01

(normalised to NC) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.001 level expression Relative 0.001 AC NC AC1 AC2 AC3 AC4

Figure 5.3. Proliferation capacity of AC and NC when cultured in RAFT TE for 3 weeks in CSSCM. A) PrestoBlue® results normalised to a RAFT TE without cells (100%). B) Relative gene expression of KI67 in AC, compared to NC. C) Relative gene expression of KI67 in the different aniridic donors plotted individually compared to NC. Dashed line represents gene expression in NC. Data is represented as the (A) mean ± standard deviation, (B) Median with 5-95 percentile n=12, (C) Median of n=3 technical replicates for each donor. Statistical analysis performed were (A) One-way ANOVA, (B) student t-test and (C) Mann-Whitney test against NC. *p<0.05, ***p<0.001. Abbreviations: A.U. arbitrary units, NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

177

Aniridic and normal corneal stromal cells do not show differences in apoptosis

AC and NC did not show significant differences in apoptosis. The expression levels of the tested apoptotic genes, namely TP53, BAX and BCL2, was similar between AC and NC (Figure 5.4).

1 (normalised to NC) to (normalised

Relative expression level expression Relative 0.1 TP53 BAX BCL2

Figure 5.4. Relative gene expression levels of apoptosis markers (TP53, BAX and BCL2) by AC compared to the control NC. Dashed line represents the gene expression for the control NC. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Student t-test was performed for AC against NC for each marker. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

Further investigation of apoptosis was performed using the Live/Dead cell assay. No apparent differences between AC and NC were observed (Figure 5.5). AC and NC cultured for 3 weeks in the RAFT TE and fed with CSSCM, showed a majority of cells stained green marker, suggesting that these were alive. A small proportion of red cells can be seen in some of the image fields, but no apparent differences can be observed between AC and NC, nor between any of the RAFT TE areas or slices (Figure 5.5). A positive control of dead cells in RAFT TE (treated with methanol) only showed red marker staining, ensuring that the assay was working correctly.

178

Figure 5.5. Representative images of Live/Dead analysis of AC and NC cultured in the RAFT TE for 3 weeks in CSSCM. Each panel of 9 pictures represents the top, middle and bottom slices (rows) of three different areas or the RAFT TE (columns). Cells stained for green marker suggest alive cells and the red suggest dead cells. Scale bars= 300µm. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

179

Aniridic corneal stromal cells have a different spatial distribution than normal corneal stromal cells in RAFT TE

Each panel of 6 images (A and B) in Figure 5.6, shows images of cells taken at 6 different depths of the RAFT TE, as indicated with white font. Specific depths were obtained with the Z-stack tool on the confocal microscope. AC and NC created different levels of organisation across all depths of the RAFT TE. Despite being evenly seeded inside the RAFT TE during the manufacturing process, it was evident that some of the cells migrated towards the surface (Figure 5.6). Higher number of cells can be observed on the top slice of the RAFT TE, both in AC and NC.

RAFT TE were also imaged under the scanning electron microscope to assess how cells distributed themselves on the surface (Figure 5.7). Interestingly, the distribution of AC and NC inside and on the surface of the RAFT TE appeared to be different after 3 weeks in culture. NC appeared to align to each other (Figure 5.6, Figure 5.7 white arrows) whereas AC, particularly AC1 and AC2, appeared to be sparse and randomly distributed along the RAFT TE surface (Figure 5.6, Figure 5.7). Cell alignment was also observed between AC3 (Figure 5.7). Morphology appeared more consistent across NC populating the RAFT TE. In contrast, AC presented a wider variety of morphologies, some with an elongated and thin shape (Red triangle, Figure 5.7) while others displayed larger cell bodies (white triangle, Figure 5.7).

From the images presented in Figure 5.6 and Figure 5.7, a clear difference in cell number populating the RAFT TE could also be observed, with NC showing a higher cell density compared to AC, confirming the previous proliferation results. In Figure 5.7, NC were seen to form a monolayer on the surface of the RAFT TE. However, it must be pointed out that the cells inside the RAFT TE are not homogenously distributed throughout the whole area of the gel, with higher cell number being presented on the top layers of the RAFT TE, Figure 5.6 and Figure 5.7.

180

Figure 5.6. Distribution of AC (A) and NC (B) inside the RAFT TE after three weeks in culture in CSSCM. Images of AC1 and NC1, representative of all donors. Images were taken under the confocal microscope with the Z-Stack tool. Each panel of 6 images shows slices of the RAFT TE at the depths indicated in white font. Blue stains for DAPI (nuclei) and red stains for Phalloidin (actin filaments). Scale bar = 250 µm. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

181

Figure 5.7. Distribution of AC and NC on the RAFT TE surface at 3 weeks culture in CSSCM. Scale bars are 100μm. White arrows indicate the direction of cell alignment. White triangle indicates cells with wide cell bodies. Red triangles indicate elongated and thin cells. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

182

Normal corneal stromal cells present higher alignment than aniridic corneal stromal cells

When cultured inside RAFT TE for 3 weeks in CSSCM, AC appeared less uniformly aligned than NC (Figure 5.8). In each image field a line was drawn in the direction where the majority of the cells were aligned (Figure 5.8A). Around 70% of NC aligned into the same direction (<10° angular displacement). In contrast, only around 40% of AC aligned into the same direction, and around 25% with an angular displacement of 10 to 20 degrees.

80 B) Normal Aniridia 60

20 Number of(%) cells Number 0 0 10 20 30 40 50 60 70 80 Angle (degrees)

Figure 5.8. Alignment of AC and NC between themselves when cultured for 3 weeks inside RAFT TE in CSSCM. (A) Representative images of different fields in RAFT TE populated with AC and NC. Arrows indicate the direction of cell alignment. Scale bars= 250µm. (B) Percentage of AC and NC that align into the same 10° angular displacement range. Data is represented as number of aligned cells as a percentage of the total number of cells. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells. 183

Aniridic corneal stromal cells present a higher degree of circularity than normal corneal stromal cells in RAFT TE

To further characterize differences in cell morphology inside the RAFT TE circularity of AC and NC was also quantified. Circularity is defined as a ratio of the cell short axis over its long axis with a value close to 0 indicating an elongated cell whilst a value of 1 suggests a perfectly spherical cell. AC presented a wider and higher circular morphology compared to NC (Figure 5.9A). Quantification of cell circularity confirmed the observations; AC circularity had an average score of 0.30, significantly higher than the one observed for NC which was around 0.1 (p<0.001) (Figure 5.9B).

B) 0.5 ***

0.4

0.3

0.2

0.1 Circularity(A.U.) 0.0 AC NC

Figure 5.9. Circularity of AC and NC cultured inside RAFT TE for 3 weeks in CSSCM. A) Representative images of AC and NC surrounded by a yellow stroke delimiting their perimeter. Scale bar = 300µm. B) Quantification of cell circularity. Five different fields were imaged in each RAFT TE for a total of n=3 biological replicates. Perimeter and area of the cells was measured using ImageJ tools to calculate cell circularity. Results are represented as the mean ± standard deviation. ***p<0.001. Abbreviations: A.U. arbitrary units, NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

184

Aniridic corneal stromal cells show higher expression levels of tested CSSC markers than NC, and similar gene expression profiles for keratocyte and myofibroblast markers compared to normal corneal stromal cells

AC expressed significantly higher levels of the CSSC markers CD90 and CD73 (2- and 1.5-Fold-Increase, respectively), compared to NC (p<0.05). PAX6 expression was not found in neither AC nor NC after 3 weeks in culture inside the RAFT TE in CSSCM (Figure 5.10A).

Moreover, AC and NC showed similar gene expression levels of keratocyte markers (LUM and ALDH1A1) and the myofibroblast marker αSMA. KERA expression was not detected in either AC or NC (Figure 5.10B, Figure 5.10C).

A) CSSC markers 10 * *

1 (normalised to NC) to (normalised

Relative expression level expression Relative 0.1 PAX6 CD90 CD73

B) Keratocyte markers C) SMA 10 10

1 1

(normalised to NC) to (normalised

(normalised to NC) to (normalised Relative expression level expression Relative 0.1 level expression Relative 0.1 KERA LUM ALDH1A1 SMA

Figure 5.10. Relative gene expression levels of (A) putative CSSC markers (CD90 and CD73), (B) keratocyte markers (KERA, LUM, ALDH1A1) and (C) the myofibroblast marker (αSMA) by AC and NC cultured in RAFT TE for 3 weeks in CSSCM. Dashed line represents the gene expression for the control condition, NC. Data is represented as median ± 5-95 percentile, n=4 biological replicates. Student T-test was performed. *indicates degree of significance against the control condition NC. *p<0.05. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells, KERA: keratocan, LUM: lumican, αSMA: alpha smooth muscle actin 185

Aniridic corneal stromal cells show differences in total gene expression profile compared to normal corneal stromal cells

As shown by RNA sequencing, AC significantly upregulated 42 genes and downregulated 111 compared to NC, when cultured inside the RAFT TE in CSSCM for 3 weeks (Figure 5.11A). These results were obtained by comparing the data from all three aniridic donors (AC1, AC2, AC3) against all the three normal donors (NC1, NC2, NC3).

A total of 153 genes were statistically differentially expressed (fold change higher than 2 and adjusted p value lower than 0.05) between AC and NC (Figure 5.11 A). The Volcano plots shows a quick visual identification of genes with large-magnitude changes, which were also statistically significant, shown as red dots (downregulated) and blue dots (upregulated) (Figure 5.11B). The top of the plot (higher statistical significance) and the extreme left or right (strongly down and upregulated, respectively) were the two regions of interest. The top five differentially expressed genes (DEGs) were GABRA2, TMEM200B, FGF10, CLCA2, and CHI3L1; the first four were found to be downregulated in AC, with CHI3L1 upregulated in AC. The top 30 DEGs were plotted in a heatmap and clustered according to their similarities in expression levels (Figure 5.11C). The dendrogram on the left-hand side of the heatmap enabled identification of genes that were similarly expressed across donors. The dendrogram on top shows which cells isolated from different donors were more similar between each other in terms of gene expression levels. As expected, the three NC donors were clustered together and AC2 and AC3 were also clustered together. Interestingly, AC1, which has a different mutation to AC2 and AC3, was not grouped within the same cluster as the other aniridic donors. Therefore, all RNA sequencing results were also plotted by comparing AC1 against NC (Figure 5.12) and AC2/3 against NC, in an attempt to find any mutation specific differences (Figure 5.13).

186

A) B)

Figure 5.11. RNA sequencing results comparing AC against NC. (A) Venn diagram showing the number of differentially expressed genes (padj<0.05). (B) Volcano plot displaying differentially expressed genes between AC and NC. The Y-axis corresponds to the mean expression value of log 10 (p-value), and the X-axis displays the log2 fold change, AC vs. NC. (C) Heatmap showing the top 30 DEGs with lower padj value. The expression level is represented as normalised read counts per gene (TPM). Samples were hierarchically clustered using Spearman’s rank correlation. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells. 187

When AC1 alone was compared against NC, the number of statistically DEGs was higher than when comparing all aniridic donors against NC. AC1 showed 284 significantly upregulated genes and 428 significantly downregulated genes (Figure 5.12A). The difference in the number of statistically DEGs can also be visualised in the Volcano plots in Figure 5.11B and Figure 5.12B. In addition to the difference in the number of DEGs, differences in their expression levels were also appreciated. The y axis scale on the later graph reaches 40, which indicates a p-adj value of around 10-40, whereas in the comparison of all aniridic donors pooled together, the highest p-adj values reach approximately 10-10. Taken together, these observations suggested that both DEG number and expression levels were different, with greater differences being observed on AC1 versus NC. The top 30 DEGs with lowest p-adj values were selected and plotted in the heat map of Figure 5.12C. Out of the 30 DEGs, only 2 common genes, CHI3L1 and SCUBE1, were found in both comparisons (AC versus NC and AC1 versus NC) The remaining 28 genes were different between the two comparisons.

On the other hand, when AC2 and AC3 were compared together against NC, the number of statistically DEGs was lower than when comparing all aniridic donors together against NC. AC2 and AC3 showed 18 significantly upregulated genes and 67 significantly downregulated genes (Figure 5.13A). Consistently, the volcano plot shown in Figure 5.13B shows a reduced number of coloured dots, representing statistically DEGs. The top 30 DEGs with lowest p-adj values were selected and plotted in Figure 5.13C. From the top 30 DEGs, 15 were commonly found between this comparison (AC2/AC3 versus NC) and the comparison with all donors pooled together (AC versus NC).

188

A) B)

Figure 5.12. RNA sequencing results comparing AC1 against NC. (A) Venn diagram showing the number of differentially expressed genes (padj<0.05). (B) Volcano plot displaying differentially expressed genes between AC and NC. The Y-axis corresponds to the mean expression value of log 10 (p-value) and the X-axis displays the log2 fold change, AC vs. NC. (C) Heatmap showing the top 30 DEGs with lower padj value. The expression level is represented as normalised read counts per gene (TPM). Samples were hierarchically clustered using Spearman’s rank correlation. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells. 189

A) B)

Figure 5.13. RNA sequencing results comparing AC2 and AC3 against NC. (A) Venn diagram showing the number of differentially expressed genes (padj<0.05). (B) Volcano plot displaying differentially expressed genes between AC and NC. The Y-axis corresponds to the mean expression value of log 10 (p-value), and the X-axis displays the log2 fold change, AC vs. NC. (C) Heatmap showing the top 30 DEGs with lower padj value. The expression level is represented as normalised read counts per gene (TPM). Samples were hierarchically clustered using Spearman’s rank correlation. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

190

To classify the DEGs between AC and NC by biological processes, Genewiz provided excel tables with the results of gene ontology (GO) enrichment analysis. From the excel tables, graphs were created in Rstudio. Results revealed 474 significantly overrepresented GO terms (padj<0.05) between AC and NC. These were organised according to their significance. The top 30 most significant were plotted in Figure 5.14.

Figure 5.14. Gene ontology analysis of significant differentially expressed genes (DEGs) between AC and NC. Assigned gene ontology terms were used to classify functions of DEGs based on biological process. The plots show significantly enriched GO terms, ordered from the most to less significant, top to bottom. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

The same GO analysis was performed to compare the DEGs between AC1 and NC. As expected, due to the higher number of DEGs, there was also a higher number of statistically significant overrepresented GO terms. Specifically, there were 491 significant GO terms. The top 30 more significantly overrepresented GO terms are shown in Figure 5.15. Out of 30, 10 identified GO terms were found in both comparisons AC1 versus NC and AC versus NC. Namely: angiogenesis (GO:0001525), inflammatory response

191

(GO:0006954), positive regulation of peptidyl-tyrosine phosphorylation (GO:0050731), induction of positive chemotaxis (GO:0050930), activation of protein kinase activity (GO:0032147), positive regulation of vascular endothelial growth factor signalling pathway (GO:0030949), extracellular matrix organisation (GO:0030198), positive regulation of Ras protein signal transduction (GO:0046579), wound healing (GO:0042060) and cellular metabolic process (GO:0044267). These were the most significant GO terms found in common between the both comparisons, however there were many others in common that were still significant and not represented in Figure 5.14 and Figure 5.15.

Figure 5.15. Gene ontology analysis of significant differentially expressed genes (DEGs) between AC1 and NC. Assigned gene ontology terms were used to classify functions of DEGs based on biological process. The plots show significantly enriched GO terms, ordered from more to less significant, from top to bottom. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

GO enrichment analysis was also performed to compare the DEGs between AC2/AC3 versus NC. As expected, due to the lower number of DEGs in this comparison, there was also a lower number of statistically significant overrepresented GO terms. Indeed, only 192

397 significant GO terms were found. The top 30 more significantly overrepresented GO terms are shown in Figure 5.16. Out of 30, there were 8 GO terms commonly overrepresented in both comparisons AC2/AC3 versus NC and AC versus NC. Namely, angiogenesis (GO: 0001525), inflammatory response (GO: 0006954), negative regulation of transcription from RNA polymerase II promoter (GO: 0000122), response to hypoxia (GO: 0001666), extracellular matrix organisation (GO: 0030198), negative regulation of apoptotic process (GO: 0043066), positive regulation of transcription from RNA polymerase II promoter (GO: 0045944) and cellular protein metabolic process (GO: 0044267).

Figure 5.16. Gene ontology analysis of significant differentially expressed genes (DEGs) between AC2/AC3 and NC. Assigned gene ontology terms were used to classify functions of DEGs based on biological process. The plots show significantly enriched GO terms, ordered from more to less significant, from top to bottom. Abbreviations: NC: normal corneal stromal cells, AC: aniridic corneal stromal cells.

193

5.5 Discussion

Aniridic and normal corneal stromal cells were successfully cultured inside the RAFT TE

In Chapter 4, it was shown that aniridic and normal stromal cells share phenotypic characteristics when cultured in 2D. Only proliferation capacity was found to be different between the two types of cells. However, as previously mentioned, it has been shown that 2D culture approaches are less physiologically relevant and therefore are not the most suitable and adequate setting to study the biology and functionality of human cells [249]. By culturing cells inside the RAFT TE, they are provided with an environment that may recapitulate some of the features of their native tissue.

Significantly higher metabolic activity levels were found in the RAFT TE populated with either AC or NC compared to the acellular control, this was the first indication that the cells were growing and surviving. Cell viability and elevated cell number was also confirmed by imaging the cells that had been in culture inside the RAFT TE for 3 weeks. These results support the hypothesis that AC and NC can be cultured and maintained in RAFT TE in a viable state. On the one hand, the successful culture of NC in RAFT TE was partially predicted based on the similarities between NC and NL in 2D culture (Chapter 3) and published evidence showing high viability of CSSC [256] and fibroblasts [191] in RAFT TE. On the other hand, the successful culture of AC in RAFT TE was an exciting finding, since culturing these diseased cells has not been previously attempted in the literature. Taken together these results suggested that RAFT TE populated with aniridic corneal stromal cells may be a useful surrogate disease model for the aniridic corneal stroma.

Differences were found in aniridic and normal corneal stromal cell phenotype when cultured in RAFT TE

To fully characterize aniridic and normal stromal cells inside the RAFT TE, their viability, morphology and gene expression was assessed. Interestingly, some findings supported the hypothesis that AC and NC have different phenotype when cultured in RAFT TE. A detailed discussion about the impact of the different results can be found in the next sections.

5.5.2.1 AC proliferated less but showed similar levels of apoptosis, compared to NC, when cultured in RAFT TE

Supporting the hypothesis that AC and NC present different phenotypes when cultured in RAFT TE, PrestoBlue® assay showed that AC had lower metabolic activity than NC. 194

Assessing cell proliferation of corneal stromal cells cultured inside the RAFT TE by quantification of total DNA content could not be performed due to the cells being trapped inside the dense collagen matrix. This assay would require a collagenase digestion step to release the cells from the matrix which would have been disruptive, leaving no possibility to use the same RAFT TE for any other assay. (This was a key consideration throughout the thesis owing to the very limited availability of patient cells). Instead, cell proliferation was assessed by measuring the relative gene expression levels of KI67. KI67 was shown to be significantly downregulated in AC, compared to NC, confirming a lower proliferation capacity. Specifically, it was found that that AC1 and AC4 (with C- terminal extensions) showed the lowest expression of Ki67 and therefore the lowest proliferative capacity. To the best of our knowledge no correlations had ever been made between the different types of PAX6 mutations and cell metabolic activity. In addition, lower numbers of AC compared to NC were observed in all images taken of the RAFT TE under the confocal microscope and SEM. These results further corroborate the lower proliferation capacity found in AC when cultured in 2D. The possible impact of lower proliferation in AC was critically discussed in section 4.5.3.2 in Chapter 4.

Surprisingly, there were no apparent differences in apoptosis between AC and NC when cultured in RAFT TE, as shown by the similar levels of apoptotic genes expression (TP53, BAX and BCL2) and similar staining in the Live/Dead assay. These results do not support previously published data showing higher levels of stromal cell apoptosis in wounded Pax6+/- corneas compared to the wild type [208]. However, it is complicated to compare an in vivo study with the in vitro results presented herein. From the results here presented, it can be concluded that AC and NC did not present differences in apoptosis rates when cultured in RAFT TE in CSSCM. However, further experiments would be needed to confirm whether, for example, a wound healing environment would trigger these processes, as observed in the Pax6+/- mouse cornea [208].

Several studies have addressed the role of PAX6 on cell proliferation and apoptosis, but none of them use corneal stromal cells. For instance, Meng et al. showed that PAX6 knockdown in retinoblastoma cells promoted cell proliferation and inhibited apoptosis [244]. Supporting these, Ouyang et al. showed that Pax6 overexpression suppresses cell proliferation in corneal epithelial cells [261]. Other studies found contradictory results, showing that high expression of PAX6 is found in proliferating lens epithelial cells [262] and that Pax6 overexpression leads to increased proliferation of ductal epithelial cells in mouse pancreas [263]. These studies all suggest that either an abnormally low or high level of PAX6 expression can be associated with altered rates of cell apoptosis and proliferation. 195

5.5.2.2 AC align less and are more circular than NC when cultured in RAFT TE

Stromal cells present a highly organised distribution in the native cornea, orienting themselves orthogonally within the collagen layers [264]. Previously published evidence showed that when cultured inside 3D collagen gels, corneal fibroblasts maintained their native phenotype and this characteristic distribution [265]. Corneal fibroblasts started orienting perpendicular to one another after 3 days of culture inside the 3D collagen genes and, by day 7 they had started to create orthogonal layers [265]. In the results here presented, both AC and NC also created different layers, although higher cell densities were found in the surface of the RAFT TE, as observed in the set of stacks imaged under the confocal microscope.

In the same images, it could be seen that the NC appeared to spatially distribute themselves inside the RAFT TE in a different way than AC. NC were highly aligned, which was confirmed by quantifying the number of cells aligned in the same direction. In contrast, there was a general lack of alignment between the AC inside the RAFT TE. As stated in the results, there were some parts of the RAFT TE in which AC aligned more than in others, but in general, the alignment was less than in NC.

In a recent study, Kivanany et al. described in detail the cell morphology and arrangement stages that stromal cells undergo during corneal wound healing [266]. They imaged the stromal cells in a rabbit cornea after photorefractive keratectomy and found that by 21 days, there was maximum stromal haze and two distinct patterns of cells were observed. On the one hand, myofibroblasts were identified in the anterior section of the ablated surface, presenting stress fibres, broader cell bodies and were randomly arranged. On the other hand, posterior to the photoablated region, keratocytes had differentiated into fibroblasts of elongated morphology and which had organised in parallel groups, co-aligned with the collagen fibrils. The region populated by randomly organised myofibroblasts was the source of stromal haze whereas the region populated with aligned fibroblasts was regenerating the tissue [266]. Interestingly, in the results presented herein, NC were highly organised and aligned between themselves. To the contrary, AC organised randomly, suggesting that aniridic cells might have acquired/were acquiring a myofibroblastic phenotype resemblance.

As previously mentioned, myofibroblasts present a broader cell body than fibroblasts. Thus, to assess whether aniridic cells were acquiring a myofibroblastic phenotype, cell circularity was addressed. From both data sets (images collected with confocal microscope and SEM), it is observed that NC had a more elongated morphology than AC at the end of the three weeks of being cultured inside the RAFT TE. Higher circularity

196

rates of AC were also confirmed by quantification. Taken together, these observations suggest that NC resemble keratocytes, presenting elongated cell bodies with long protrusions. In contrast, AC displayed various morphologies, with a high proportion of larger cells presenting prominent actin filaments. The latter morphology has been shown elsewhere to be typical of myofibroblasts [267]. It has been previously reported that murine Pax6+/- corneal epithelial cells have delayed and abnormal differentiation [129]. If this trend is extrapolated to the corneal stromal cells, aniridic keratocytes could be aberrant and might more readily differentiate into myofibroblasts. This would aid to understand the high levels of scarring observed in ARK following insult [61, 140].

5.5.2.3 AC and NC presented different gene expression profiles

To identify an underlying explanation at the transcriptional level for the differences seen in cell distribution and morphology between AC and NC cultured in the RAFT TE, gene expression in AC and NC was assessed. As an initial step, the markers previously tested in AC and NC in Chapter 2 were studied. Namely, CSSC (PAX6, CD90 and CD73), keratocyte (KERA, LUM and ALDH1A1) and myofibroblast (αSMA) markers.

PAX6. Corneal stromal cells (AC and NC) did not express PAX6 when cultured in RAFT TE for 3 weeks. This finding is perhaps not surprising, because when AC and NC were firstly expanded in 2D in CSSCM, prior to culture in RAFT TE, PAX6 expression was already lost (discussed in Chapter 4). This initial expansion is of crucial importance to obtain the required cell number to perform the experiments but may also be the cause of PAX6 expression loss. Confirming these results, it has been shown elsewhere that PAX6 expression significantly decreases with cell passaging [198].

A possible explanation for the low expression of PAX6 when corneal stromal cells are cultured in CSSCM could be that this downregulation is necessary to modulate cell proliferation. Li and Lu have shown in corneal epithelial cells that PAX6 downregulation is required for EFG mediated proliferation [268]. Through the Erk signalling pathway, EGF upregulated CCCTF binding factor, which binds to PAX6 promoter and inhibits its transcription [268]. Moreover, they found that PAX6 overexpression resulted in a reduced EGF-mediated cell proliferation. Although the mentioned study was performed with epithelial cells, proliferation of stromal cells, such as dermal fibroblasts, can also be mediated by EGF [269, 270]. In the experiments presented herein, corneal stromal cells were cultured in CSSCM, which contains EGF, so the same mechanisms might be regulating cell proliferation. To summarize these observations, downregulation of PAX6 might modulate EGF-mediated cell proliferation which may help to understand its low expression when corneal stromal cells are cultured and expanded in CSSCM. Moreover, 197 only 4% of the freshly isolated stromal cells from the cornea showed PAX6 expression, suggesting that it is a small population of cells that express this marker in vivo [57]. These cells probably remain quiescent until required to proliferate and differentiate into keratocytes, losing their PAX6 expression [57, 174].

CD90 and CD73. Although AC showed an upregulation of CD90 and CD73 gene expression compared to NC, biological relevance remains to be further investigated.

KERA. KERA was not expressed by AC and NC when cultured for 3 weeks inside RAFT TE in CSSCM. These results are in accordance with those found in 2D (Chapter 4), in which KERA was only upregulated when cells were cultured in KDM. Previously published evidence showed that a specific serum-free keratocyte differentiation medium is required for corneal stromal cells to express KERA in a 3D environment [196, 271- 273]. For example, Wu et al have shown that CSSC cultured on a highly aligned fibrous substrate made of poly(ester urethane) urea (PEUU) for 3 weeks in KDM expressed high levels of keratocan at both mRNA and protein level [196]. In a different approach, Yoshida et al. observed an upregulation of KERA when keratocytes were cultured as spheroids in serum-free medium for 2 weeks. In contrast, no KERA upregulation was found when spheroids were cultured in the presence of serum [272, 273]. Keratocytes cultured in compressed collagen gels and fed with serum free media express higher levels of KERA when treated with retinoic acid [274]. Taken together, these observations suggest that, despite being cultured in a mimic of their native environment, the mechanical and physical properties of the RAFT TE are not sufficient to induce the expression of KERA by corneal stromal cells when cultured in a serum-containing medium.

LUM. LUM was expressed at similar levels by AC and NC. Contrarily to KERA, LUM has been previously shown to be expressed by corneal stromal cells when these were cultured both in the presence and absence of serum [192, 275], supporting the results found herein.

ALDH1A1. Expression of ALDH1A1 was not statistically significant between AC and NC when culture in RAFT TE. Interestingly, previous evidence showed lower abundance of ALDH1A1 in protein extracts of keratocytes isolated from injured and hazy corneas, when compared to keratocytes from transparent and healthy rabbit corneas [276]. Taking these results into account, and the fact that PAX6 has been shown to regulate the expression of ALDH1A1 [135], a downregulation of this marker in AC would have been expected. However, it has been previously shown that serum-containing medium results in a significant reduction of ALDH1A1 expression by corneal stromal cells [277]. 198

αSMA. Based on the results observed in the confocal images, where AC presented low cell alignment and wide cell bodies (typical feature of myofibroblasts [266]), it was surprising that expression of αSMA was similar between AC and NC. αSMA is a protein involved on cell motility, structure and integrity, and is commonly used as a marker of myofibroblast formation [278]. Increase in the expression of αSMA in corneal stromal cells is normally associated with loss of keratocyte-characteristics and acquisition of a more myofibroblastic nature [279]. Similar expression of αSMA by AC and NC, might suggested that they have similar keratocyte-fibroblast transition capacity. However, it is not possible to make such affirmation based only on αSMA transcription levels. In fully differentiated myofibroblast, αSMA is also produced at the protein level and assembled into stress fibres [191]. In a published RAFT TE study with limbal fibroblasts, Massie et al. did not see any production and assembly of αSMA after two weeks in culture [191], suggesting that the ECM in RAFT TE might not induce myofibroblast differentiation. Interestingly, Vicente et al. showed higher production of αSMA at the protein level in human aniridic corneal stroma ex vivo [161]. Further characterisation at the protein level and stress fibre assembly would be needed to assess if there are differences between AC and NC in myofibroblast differentiation capacity in RAFT TE.

The expression of tested CSSC, keratocyte and myofibroblast markers was not sufficient to characterise and compare AC and NC in RAFT TE. Therefore, since the differences and similarities between these was not sufficient to account for the differences seen in AC and NC phenotypes (alignment and morphology), a thorough gene expression profiling was performed using RNA sequencing.

In any case, it would be interesting to assess the 7 tested markers previously mentioned again when attempting to differentiate AC and NC towards a keratocyte phenotype in RAFT TE. In the present work, the differentiation in RAFT TE towards keratocyte-like cells was not pursued due to the low number of samples, and the large number of cells needed for making that RAFT TE.

5.5.2.4 AC and NC present different gene profiling, as assessed by RNA sequencing

For the first time RNA sequencing was used to investigate the transcriptome of aniridic cells. High-through put sequencing has been increasingly used to study the active transcriptomes of tissues, cells, and models in the field of ophthalmology and vision research (reviewed by Owen et al. [280]). However, there are still only a limited number of published studies performing RNA sequencing in cornea and corneal cells [133, 135, 281-283]. Moreover, most of these were performed in either endothelium/endothelial 199 cells or epithelium/epithelial cells. Examples of cell sources used in these studies are endothelial cells from Fuch’s dystrophy patients [282], corneal epithelial tissue from keratoconus and myopia patients [281], limbal epithelial stem cells with knocked-down PAX6 [133], and limbal epithelial cells from aniridic patients biopsies [135]. Only one study from 2013 includes stromal cells in the analysis to compare them against endothelial cells, with the aim of identifying specific endothelial cell markers not expressed in the stroma [283]. Another recent study of the active transcriptome in keratoconus patients used the full-thickness corneal button including the three layers as the source for RNA sequencing, however it is not possible to distinguish between these [284]. In conclusion, even though high-through put sequencing is being increasingly used, there have not been previous reports studying the transcriptome of the corneal stroma or cells derived from this tissue, bringing novelty to the results here presented.

Supporting the hypotheses that AC and NC have different transcriptomes, an elevated number of significantly differentially expressed genes (DEGs) was found between these two populations of cells. It would be very interesting to compare this dataset with the literature. One other report is currently available for RNA sequencing in ARK samples. However, this study focussed on gene expression in epithelial cells only [135], and therefore might not be relevant and comparable.

In the study here presented, RNA sequencing was performed on corneal stromal cells from 3 aniridic samples (AC1, AC2, AC3) and from 3 normal donors (NC1, NC2, NC3), due to sample availability at the time of the experiment. From the 3 aniridic samples used, AC1 had a run-on mutation and AC2 and AC3 had nonsense mutations. Results presented in the previous chapter had shown that AC1 presented a different phenotype than the other two, with lower cell proliferation and lower cell alignment. For this reason, RNA sequencing datasets were compared in 3 different ways: all aniridic samples pooled together against all normal samples (AC vs NC), the aniridic sample with a run-on mutation against all normal samples (AC1 vs NC) and the two aniridic samples with nonsense mutations against all normal (AC2/3 vs NC).

Interestingly, when the two different mutations were compared to each other striking differences were observed. AC1 vs NC had the highest number of significant DEGs suggesting that AC1 was the donor with a more different transcriptome compared to NC. These results correlated with the differences in behaviour of cells with different mutation cultured in the RAFT TE. A larger sample size would be necessary to draw any further conclusions on whether the different genotypes caused variable phenotypes.

200

CHI3L1 was abundantly upregulated in all AC samples. This gene has been associated with chronic inflammation, stress, and tissue remodelling [285]. CHI3L1 actively participates in the pathogenesis of chronic and acute inflammation in the liver, lungs, and intestine. ARK patients also develop chronic inflammation on the ocular surface, which could be related to the expression of this marker. On a different level, CHI3L1 suppresses the cytokine-induced production of MMPs in skin fibroblasts and chondrocytes, producing an anti-catabolic effect preserving the ECM during tissue remodelling [286]. Abnormal MMP expression has also been associated with aniridia, with Pax6+/- corneal epithelial cells producing lower MMP9 activity [137]. In the cornea, CHI3L1 is synthesised by epithelial cells and plays a role in response to fungal infection [287]. In 2011, Davis and Piatigorsky showed that overexpression of Pax6 in the murine cornea led to Chi3l1 upregulation, suggesting that Chi3l1 might be regulated by Pax6. Corneal overexpression of Pax6, and therefore Chi3l1, resulted in altered epithelium morphology, vascularisation and inflammation, all hallmarks of aniridic corneas [288].

SCUBE1 was downregulated in all AC vs NC comparisons. This gene has only been recently cloned and mapped as a novel family of EGF-related proteins. Published studies suggest its involvement in supporting cell adhesion [289] but there is no evidence about its role in the cornea. However, there is a study that showed a downregulation of Scube1 in Pax6+/- in cerebral cortex tissue using a microarray, suggesting that this marker may be regulated by Pax6 [290].

It is extremely challenging to make assumptions about the role of singular genes, differentially expressed by AC and NC, on ARK development. RNA sequencing takes a snapshot, in this case, at 3 weeks of cultured AC and NC in the RAFT TE, however gene expression is plastic and changes overtime. Further gene ontology analysis was performed and focused on identifying which biological processes might be regulated by the differentially expressed genes.

5.5.2.4.1 What biological pathways might be governing the differences observed between AC and NC when cultured in RAFT TE?

Gene ontology analysis of the DEGs between aniridic and normal samples was performed to identify potential processes involved in the differences seen in cell behaviour. From the gene ontology enrichment analysis, a wide repertoire of biological processes appeared to be significantly different between AC and NC. Of interest for the development of ARK, those related to angiogenesis, cell adhesion, wound healing and ECM remodelling were highlighted.

201

Angiogenesis, whose related genes are differentially expressed between AC and NC, is a well-reported feature of ARK. It might be expected then than this biological process is aberrantly regulated in AC. In the healthy ocular surface, the blood vessels do not encroach beyond the limbal area, where they provide a source of nutrients and are essential for the maintenance of the limbal niche [14]. In ARK, blood vessels can start infiltrating the superficial peripheral cornea during patients’ first decade of life and can variably end up covering the central cornea [3, 161]. It would be interesting to pursue this investigation further and identify which mechanisms are driving neovascularisation in the aniridic stroma.

Cell-cell and cell-matrix adhesion are other biological process that are possibly involved in the development of ARK. The role of Pax6 in cell-cell adhesion during corneal epithelial development and maintenance has previously been shown on Pax6+/- mice. Their corneas showed large gaps between corneal epithelial cells and abnormal appearance of desmosomes, specialized structures involved on cell-cell adhesion [132]. Additionally, cell layers of the Pax6+/- corneal epithelium were not packed as tightly as in the wild type. Despite being focused mainly on the epithelium, the stroma of these mice also appeared to be disorganised [132]. The role of Pax6 in adhesion has also been well studied in other parts of the eye, such as in the optic vesicle where it affects their cell adhesive properties [291], and in the brain [292-294]. Most of these studies, investigate cell-matrix adhesion and specifically how it affects cell migration rather than cell-cell adhesion. Neural crest cells in homozygous Pax6+/- embryos did not migrate successfully from the anterior midbrain to the nasal rudiments, suggesting that Pax6 is involved in conducting their migration [293]. Taken together, these observations suggest that aniridic cells may have impaired cell-cell, cell-matrix, and cell migration capacities.

Wound healing was also shown to be another biological process differently affected in AC when compared to NC. These results are not surprising given that ARK corneas tend to respond negatively to insults and surgeries, with impaired wound healing capacity usually developing into hazy scarring [295]. A state of chronic corneal wound healing caused by mutations in PAX6 has also been previously suggested [160, 166]. Although RAFT TE was not designed as a model for wound healing, cells are cultured on these TE in an active state which may model this process to some extent. Considering the results obtained by RNA sequencing together with previously published evidence, where ARK corneas were strongly positively labelled for αSMA [161], it was surprising that αSMA was not upregulated in AC (discussed previously in section 4.5.4). Real-Time PCR and RNA sequencing data showed consistent similar expression levels of αSMA between the two cell types, suggesting that their differences on wound healing and cell 202

motility arise from other pathways. Tenascin-C, a key player on corneal epithelial wound healing [161], showed no differences between AC and NC. Therefore, further studies are needed to properly assess the mechanisms behind the altered corneal wound healing in ARK.

The differences between AC and NC in extracellular matrix organisation and remodelling are of special interest in the present work. The lack of cell alignment in AC may be due to an impaired ability of these cells to break the surrounding extracellular matrix, therefore promoting their constraint inside the collagenous matrices. It has also been shown that PAX6 regulates the expression of matrix metalloproteinases in corneal epithelial cells, a protein family responsible for degrading the collagen matrix during cell remodelling processes [138]. The lack of matrix remodelling capacity may be one of the causes of corneal opacification often observed in aniridic patients [137, 161]. Therefore, subsequent experiments were performed, in the next chapter, to study this process in NC- and AC-populated RAFT TE.

In conclusion, processes that may be related to the phenotypical features of ARK such as angiogenesis, cell migration and adhesion, wound healing and matrix remodelling appeared to be significantly different between AC and NC. For the first time, RNA sequencing was performed to characterize aniridic cells, and a new window of possibilities is now open to study in detail any of these processes and the mechanisms behind them. The aim of the next chapter was to explore in more detail the process of extracellular matrix remodelling by assessing the expression of matrix metalloproteinases.

203

204

CHAPTER 6: ANIRIDIC AND NORMAL CORNEAL STROMAL CELLS EXPRESS AND SECRETE DIFFERENT LEVELS OF MATRIX METALLOPROTEINASES WHEN CULTURED IN RAFT TE

6.1 Introduction

In the previous chapter, alignment of AC within RAFT TE was found to be impaired compared to NC. Furthermore, RNA sequencing suggested that AC expressed altered gene profiles, which could influence AC interactions with the RAFT TE extracellular matrix, and therefore cell alignment. Hence, the present chapter investigated potential differences in matrix remodelling capacity between AC and NC.

Previous evidence points to an aberrant wound healing process in ARK corneas, affecting both the epithelium and the stroma. Firstly, Ramaesh et al. showed that in Pax6+/- mouse corneas, epithelial cells proliferated at higher rates than their wild type counterparts. However, their differentiation appears to take longer [165, 296]. Secondly, Sivak and Ramaesh demonstrated that Pax6 regulates the expression of Mmp9, possibly affecting the reepithelialisation rates and wound closure in Pax6 haploinsufficient mice [297]. Thirdly, despite the scarce evidence, it has been shown that corneal stromal wound healing might also be abnormal in ARK conditions, with increased keratocyte apoptosis already observed in Pax6+/- mice [208]. It is possible then, that the scarring observed in ARK patients, might be due to altered stromal wound healing capacities. However, the specific mechanisms involved in stromal remodelling of the ARK cornea remain unclear.

The first group of experiments conducted in this chapter focused on studying potential differences in the capacity for matrix remodelling between aniridic and normal corneal stromal cells by assessing their matrix metalloproteinases (MMPs) expression levels. MMPs belong to a broad family of calcium-dependent zinc-containing proteases involved in extracellular matrix degradation, cleavage of cell surface receptors, and cytokine inactivation. MMPs play major roles in cell migration, differentiation, and proliferation [298]. In tissues, MMPs are regulated at different levels: transcription, activation and proteolytic inhibition. Tissue inhibitors of MMPs (TIMPs) are responsible for the latter, by binding to the MMPs active site to repress their function [299]. Therefore, TIMPs are also regulators of matrix and tissue remodelling. Here, MMP1 and MMP2 were chosen for study since they are two of the most highly expressed MMPs by stromal cells during

205 matrix remodelling [300]. They have also been previously shown to be produced by stromal cells in collagen type I gels used as an in vitro model of stromal contraction [77].

The second group of experiments was conducted to further investigate whether there existed a correlation between levels of MMP production and cell alignment in the RAFT TE. As indicated by RNA sequencing data in the previous chapter, matrix remodelling capacity might be altered in AC, and this alteration might be the cause for lower cell alignment. It has been shown elsewhere that cell migration in a space constrained 3D collagen type I gels is modulated by the capacity of the cells to remodel the matrix around them [301]. Namely, production of MMPs was necessary to allow the cells to migrate [301]. Perhaps a similar process is necessary for the cells in the RAFT TE to migrate and align between themselves. Here, to assess whether low levels of MMPs were not allowing the cells to align in the RAFT TE Ilomastat was employed. Ilomastat (HONHCOCH2CH (i-Bu) CO-L-Trp-NHMe; GM6001, Galardin) is a synthetic broad- spectrum MMP inhibitor. It acts by reversibly making a complex with the active site zinc of the MMPs inhibiting their activity. Ilomastat has already been successfully used by Daniels et al. to inhibit the activity of MMPs produced by humans Tenon’s capsule fibroblasts in free floating, non-compressed type I collagen gel[302]. Ilomastat reduced fibroblast-mediated collagen contraction and inhibited the production of new collagen fibres, without affecting cell viability [302], suggesting that Ilomastat is an adequate agent to block MMPs in fibroblast-populated collagen gels. Owing to the limited availability of AC, the concept had to be tested using NC.

6.2 Hypotheses and aims

The hypotheses tested in this chapter were the following:

1. Aniridic and normal corneal stromal cells cultured in the RAFT TE present differences capacities in matrix remodelling, specifically in MMP1 and MMP2 expression and production. 2. Reduced MMP activity accounts for the reduced cell alignment observed in AC cultured in the RAFT TE.

Hence, the overarching aim of Chapter 6 was to confirm lower production of MMPs by AC compared to NC and to assess whether this affects cell alignment in the RAFT TE.

206

6.3 Materials and methods

Experimental design

The same materials used for Chapter 5 were also used in the experiments presented in this chapter. RAFT TE populated by AC and NC were prepared as described in section 5.3.1 of Chapter 5. The experiments here presented were performed at the time points indicated in Table 6.1.

Table 6.1. Time points at which different experiments were performed on the RAFT TE.

Time

1 week 2 weeks 3 weeks

Transparency ✔

SEM ✔

Real Time qPCR ✔

Bradford assay ✔ ✔ ✔

Zymogram ✔ ✔ ✔

ELISA (MMP1, 2, 9) ✔ ✔ ✔

RAFT TE Transparency

The RAFT TE-containing wells were washed three times with DPBS before adding 250μL of DPBS in each of them, including to an empty well filled with DPBS only and used as blank. The absorbance, ranging from 350nm to 900nm, was measured in a microplate reader (Sapphire™). Assigning the absorbance of the blank well as 100% transparency, the transparency of the RAFT TE was then represented as a percentage of this measurement. The results plotted correspond to the measurements taken at 550nm.

Scanning electron microscopy

Visualisation of samples in the SEM was performed as previously explained in section 5.3.5.4 in Chapter 5.

Quantification of total protein – Bradford assay

In order to quantify the concentration of total protein secreted by the cells, the Coomassie Plus (Bradford) Assay kit (Thermo Fisher ScientificTM) was used. This colorimetric 207 method is based on the ability of the Coomassie dye to shift its absorption maximum from 465nm to 595nm when it binds to protein. The BCA method could not here be used as the supernatant contained phenol red, which is incompatible with the assay. Since the concentration of phenol red is less than 0.5 mg/mL, the Bradford assay could be employed.

At three different time points (day 7, 14, and 21) RAFT TE were washed with 1mL of DPBS and 1mL of serum free DMEM (High glucose, GlutaMAXTM, Gibco®) supplemented with 1% (V/V) P/S was added to each RAFT TE containing well. After 24 hours, the supernatant was removed and centrifuged for 15 minutes at 5000xg to remove cell debris. The supernatant was divided into aliquots that were stored immediately at -80°C until further use in Bradford assay and ELISA. Fresh CSSCM was then added to the RAFT TE containing wells.

The supernatants from all AC and NC from all time points were thawed at the end of the experiment and were analysed simultaneously. Firstly, the standard curve with BSA protein controls had to be optimised to cover the range of concentrations of the supernatants (10-250μg/mL). The assay was performed as per manufacturer’s instructions but scaling up the volumes to 50μL of standard or unknown sample and 250μL of dye in a 96 well plate. The mixture was allowed to react for 10 minutes and the absorbance was measured at 595nm in a microplate reader (Sapphire™). The absorbance of the blank (serum free DMEM supplemented with 1% P/S) was subtracted from all other readings and the standard curve was used to calculate the concentration from all the samples.

Quantification of matrix metalloproteinases secretion

6.3.5.1 Zymography

Secretion of matrix metalloproteases (MMP) was semi-quantitatively assessed with zymography as described by Massie et al [191]. This electrophoretic method is based on the proteolytic activity of the MMPs and their capacity to degrade the substrate of which the gel is composed of. Samples from the conditioned medium (5μL) at week 1, 2, and 3 were mixed with 5μL of sample buffer (1:1 ratio), loaded into NovexTM 10% Zymogram gelatin protein gels and run at 125V for 90 minutes in Novex™ Tris-Glycine SDS Running Buffer. Gels were then renatured for 30 minutes on a shaker in Novex™ Zymogram Renaturing Buffer and developed for 30 more minutes in Novex™ Zymogram Developing Buffer. Subsequently, gels were placed in new developing buffer and incubated overnight at 37 °C. On the following day, the gels were stained with (%V/V): 0.01% Coomassie blue (Bio-Rad) in 10% acetic acid/40% ethanol in water for 30 minutes, rinsed once with 208

ddH2O and destained using 7.5% acetic acid/10% ethanol for 1 hour. Gels were visualised and image against a white field. All reagents were purchased from Novex ™ unless otherwise specified.

6.3.5.2 ELISA

The concentration of MMP1, MMP2 and MMP9 in cell culture supernatants was quantified by enzyme-linked immunosorbent assay (ELISA, R&D Systems) as per manufacturer’s instructions. ELISA were performed using the sandwich method and assessed the concentration of pro-MMP1, total MMP2, and total MMP9. Briefly, 100µL of Assay Diluent was added to the wells, followed by addition of 100 µL of samples and standards. The plates were then sealed and incubated for 2 hours at room temperature on a shaker. Wells were then washed four times by adding Wash Buffer (400 µL) and removing it by tapping against tissue paper for the complete removal of liquid. Human Pro-MMP1, MMP2, or MMP9 conjugate (100 µL) was added to the wells and incubated for a further 2 hours at room temperature on a shaker. The substrate colouring reagent (200 µL) was then added, so it reacted with the conjugate antibodies that had attached and the reaction was stopped after 20 minutes by adding 50 µL of Stop Solution. Absorbance at 450nm was measured in a microplate reader (Sapphire) and corrected against absorbance at 540nm. Different dilutions of the cell supernatants were necessary depending on the MMP type, Table 6.2. All absorbance measures were corrected by the absorbance at concentration 0 ng/mL, and a standard curve was plotted with absorbance against the known concentrations of MMP. The standard curve was used to calculate the concentration of MMPs in each of the supernatants.

Table 6.2. Dilution factors used for ELISA depending on MMP and time point.

Time

1 week 2 weeks 3 weeks

MMP1 1 1 1:4

MMP2 1 1:10 1:20

MMP9 1 1 1

209

Inhibition of matrix metalloproteinases activity with Ilomastat

6.3.6.1 Donor demographics

Three normal donors were used to isolate corneal stromal cells as explained in section 2.1.1 in Chapter 2. Donor information can be found in Table 6.3.

Table 6.3. Donor information. Age, eye of retrieval, storage time, and sample ID is shown.

Condition Age (years) Eye Storage (days)

63 Left 20

Normal 82 Right 23

46 Right 30

6.3.6.2 Cell culture in non-compressed gels and RAFT TE in the presence of Ilomastat

Cells were isolated and expanded in CSSCM as previously explained. They were then cultured for 7 days in non-compressed collagen gels (section 4.3.3.4 in Chapter 4) and RAFT TE (section 5.3.1 in Chapter 5) with different concentrations of Ilomastat. Ilomastat was provided as a stock solution in DMSO (1mg/mL – 2.6mM) and was diluted accordingly in CSSCM into working concentrations of 10nM, 1μM and 10μM. Three technical replicates were cultured for each donor. Cultures were fed with freshly prepared working solutions every other day.

Cell viability was assessed using PrestoBlue® Reagent at day 7 as previously explained (section 5.3.3 in Chapter 5).

Contraction of non-compressed untethered collagen gels was assessed overtime as described in section 4.3.3.4 in Chapter 4.

RAFT TE with normal corneal stromal cells (NC) cultured in 10μM Ilomastat in CSSCM were visualised under a Cytosmart live microscope (Lonza) for the 7-day period. At the end point, the RAFT TE were fixed and stained with Phalloidin as explained in section 2.4 to visualise the cells under the confocal microscope (LSM710, Zeiss). For an overview of cell distribution in the RAFT TE a Leica DMRB Fluorescence microscope was used (Leica) at the lowest magnification (2X). The number of cell clusters that had developed inside the RAFT TE were manually counted in five different image fields (top, bottom, right, left and middle of the RAFT TE).

210

Statistical analyses

For the first set of experiments to compare MMP expression and production between AC and NC, three biological repeats were used for each condition (aniridia and normal) for all of the experiments presented in this chapter, except for Real Time qPCR in which 4 biological repeats could be used. This was due to donor tissue availability at the time of the experiments.

For the MMP secretion quantification by ELISA, each aniridic donor was coupled with a normal donor to perform the experiment at the same time. Specifically, the ELISAs were done using the following pairs: AC1 and NC1, AC2 and NC2, AC3 and NC3. This was due the high numbers of cells needed for culture in RAFT TE experiments. Thus, it would not be feasible to culture cells from the six donors at the same time and obtain enough cell numbers for all the assays to be performed at once. In each experiment three technical replicates were performed. Post cell culture experiments (Bradford and ELISA) were performed on all conditioned media samples collected from the three RAFT TE experiments. They were assayed at the same time and in the same plate to allow comparison. Results are represented either by pooling together the results of all biological replicates or by separating each of them to better understand any sources of data variability. The mean was calculated in each case and represented with the standard deviation. The non-parametric equivalent of the T-test for independent samples, the Mann–Whitney U test, was used to analyse the data from aniridia and normal when all the donors were pooled together. The non-parametric equivalent of the One-way ANOVA, Kruskal-Wallis test was used to analyse the data from aniridic and normal samples when three or more donors were represented separately. Finally, two- way repeated measurements analysis of variance (ANOVA) was used to analyse data from AC and NC over the three time points in the RAFT TE experiments.

For the second set of experiments in which Ilomastat was used to block the activity of MMPs three biological donors were used. Four technical replicates were added for each condition: CSSCM, different concentrations of Ilomastat, and negative control. One-way ANOVA, was used to analyse the data from NC cultured in non-compressed gels or RAFT TE in the presence of different concentrations of Ilomastat.

211

6.4 Results

Extracellular matrix appearance is similar in RAFT TE populated with aniridic or normal corneal stromal cells

As observed in Figure 6.1, no significant differences were observed in the transparency of the cellular (AC and NC) and acellular RAFT TE at the endpoint (3 weeks). The transparency of the RAFT TE depends on its thickness, the presence of cells, and their capacity to remodel the matrix.

150

100

50 Transparency

(% of RAFT without cells) of RAFTwithout (% 0 AC NC No cells

Figure 6.1. Transparency of the RAFT TE populated by aniridic, normal or no corneal stromal cells, measured at the end of the three weeks. Transparency is indicated as a % of the RAFT without cells (100%) that is used as a negative control. Values over 100% would represent more transparent RAFT TE and values under 100% would suggest less transparent RAFT TE. Data represents the mean of all three donors pooled for each condition (n=3 technical replicates for each donor). Error bars represent the standard deviation. Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

High magnification SEM pictures were taken and used to visualise the ECM surrounding the cells in the RAFT TE at the end of the three weeks in culture. No apparent differences were seen between the RAFT TE populated with AC or NC in the size or organisation of the fibrils, nor with the acellular RAFT TE (Figure 6.2).

The six top images in Figure 6.2 show the distribution of the collagen fibrils that formed the matrix on AC and NC populated RAFT TE. Several cells were seen interacting with the collagen matrix underneath them. The collagen fibrils were randomly distributed and had variable diameters. In certain areas, they aggregated to form bundles of collagen fibrils. However, no regular pattern (alignment) was observed.

212

An overview of how the acellular RAFT TE appeared at the end of the three weeks in culture can also be seen at the bottom of Figure 6.2. The surface of the acellular RAFT TE was not uniform, with collagen randomly organised and creating some bundles. Fibrils of collagen in the bottom left image resembled those seen in the cell populated RAFT TE.

213

Figure 6.2. Representative images of collagen fibrils from the RAFT TE populated with aniridic and normal cells cultured for 3 weeks in CSSCM. Cell bodies are indicated with white triangles. Scale bars represent 2μm. Abbreviations A: aniridic corneal stromal cells, N: normal corneal stromal cells.

214

Aniridic corneal stromal cells express and secrete lower levels of matrix metalloproteinases compared to normal corneal stromal cells

Figure 6.3 represents an overview of the relative MMPs gene expression by AC and NC obtained by RNA sequencing. AC showed lower expression of MMPs, indicated by the red colour, especially of MMP1, MMP2, MMP3, MMP8, MMP9 and MMP10 when compared to NC (Figure 6.3A). MMP9 and MMP10 showed significantly reduced expression in AC compared to NC, as obtained by the statistical analysis following RNA sequencing. Interestingly, NC3 presented a different pattern compared to the other two normal samples, with a generally lower expression level of MMPs.

A different visualisation of the MMP expression by AC and NC can be observed in Figure 6.3B. This graph shows the differential expression levels of the different MMPs within each sample (i.e. most expressed MMPs are indicated in blue and least expressed MMPs are indicated by red). MMP2 and MMP14 were the most highly expressed across all samples. MMP1 and MMP3 were also highly expressed in normal cells, but at lower levels than MMP2 and MMP14.

215

A) B)

Figure 6.3. Heatmaps comparing the expression levels of Matrix metalloproteinases (MMPs) by AC and NC. (A) Results are scaled by row, indicating for each MMP if AC or NC had higher expression. (B) Results are scaled by column, indicating within each sample, which MMP was most and least expressed. The expression level is represented as normalised read counts per gene (TPM). Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

No visible differences in the expression levels of tissue inhibitors of matrix metalloproteinases (TIMPs) were appreciated between AC and NC as observed in Figure 6.4A. TIMP1 was the most highly expressed inhibitor across all samples. Interestingly, NC2 expressed higher levels of TIMP3 compared to the rest of the samples.

216

A) B)

Figure 6.4. Heatmaps comparing the expression levels of tissue inhibitors of matrix metalloproteinases (TIMPs) by AC and NC. (A) Results are scaled by row, indicating for each TIMP if AC or NC had higher expression. (B) Results are scaled by column, indicating within each sample, which TIMP was most and least expressed. The expression level is represented as normalised read counts per gene (TPM). Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

6.4.2.1 AC express lower levels of MMP1 and MMP2

MMP1 and MMP2 expression levels were significantly downregulated (10 to 100-fold decrease, p<0.001, and 2 to 5-fold decrease, p<0.01, respectively) in AC, compared to NC when cells were cultured in RAFT TE for 3 weeks in CSSCM (Figure 6.5A).

Interestingly, when results were plotted for each aniridic donor separately, AC1 and AC4 (with C-terminal extension mutations) had the lowest expression levels of MMP1 (100- Fold decrease, p<0.05), followed by AC2 which also significantly downregulated MMP1 (10-Fold decrease, p<0.05). In contrast, AC3 expressed similar expression levels of MMP1 to NC (Figure 6.5B). A similar trend was shown for MMP2, but this was less downregulated than MMP1 (approximately 2-fold decrease) (Figure 6.5C).

217

10 A) *** **

0.1

0.01

(normalised to RPL37A) to (normalised Relative expression level expression Relative

0.001 MMP1 MMP2

MMP2 B) MMP1 C) 1 1 * * * *

0.1

0.01

(normalised to RPL37A) to (normalised

Relative expression level expression Relative Relative expression level expression Relative

0.001 0.1 AC1 AC2 AC3 AC4 AC1 AC2 AC3 AC4

Figure 6.5. Relative gene expression levels of MMP1 and MMP2 by AC, normalised to NC. Results are shown for MMP1 and MMP2 for all AC donors pooled together (A) and for all AC donors plotted separately (B, C). Dashed line represents MMP expression by NC. Data is represented as the median with 5-95 percentile n=12 (A), median of n=3 technical replicates for each donor (B, C). Statistical analysis performed were student t-test (A) and Mann-Whitney test (B, C) against NC. *p<0.05, **p<0.01, ***p<0.001. Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

6.4.2.2 AC secrete lower levels of MMP1 and MMP2

Differences were found between the amount of total protein released by AC and NC. Therefore, the relative concentrations of MMPs released from AC and NC were normalised to the concentration of total protein obtained by the Bradford assay.

The concentration of MMP1 in the supernatant was quantified by ELISA. The levels of released pro-MMP1 by NC increased over time (p<0.01 from week 1 to week 3). In contrast, the concentration of pro-MMP1 secreted by AC decreased from week 1 to week 2 (although not statistically significant, p>0.05) and then stabilised until week 3 (Figure 6.6). When the pro-MMP1 expression values of all donors were pooled together for each condition, it was observed that the concentration of pro-MMP1 secreted into the

218

supernatant of NC appeared higher than in AC. However, statistical analysis showed that the difference was not significant (p>0.05) (Figure 6.6).

0.25 ** AC 0.20 NC

0.15

0.10

0.05

0.00

7 14 21 Pro-MMP1 concentration (ng/ug) concentration Pro-MMP1 Time (days)

Figure 6.6. Concentration of pro-MMP1 released by AC and NC in RAFT TE over three weeks in culture, represented as nanograms of pro-MMP1 per µg of total protein. Data represents the mean ± standard deviation of n=3 donors pooled for each condition (n=3 technical replicates for each donor). Significance of results were assessed by a two-way repeated measurements ANOVA. **p<0.01. Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

In order to better understand the differences between the three donors, the results from all normal and aniridic donors are shown separately in Figure 6.7. Ideally, all results would be plotted with the same scale for Y axis for all the samples, however, variability between different donors compromised this. A high variability not only between aniridic samples but also between normal samples can be observed. Specifically, AC1 secreted lower concentrations of pro-MMP1 compared to AC2 and AC3. Similarly, NC1 also released lower concentrations of pro-MMP1 compared to the other two normal samples. A possible explanation for this is arise from the fact that ELISA for each aniridic sample (AC) and its control sample (NC) was performed at different times. This raises the possibility of different conditions influencing the data, such as exposure to room temperature or quality of the medium. This is the reason for Figure 6.7 and Figure 6.9 being plotted for each aniridic sample separately with its paired control (NC).

When the results from each individual pair aniridic/normal samples were analysed separately interesting trends are observed. The concentration of pro-MMP1 released by NC1 increased over time. In contrast, the concentration of pro-MMP1 released by AC1

219

did not increase and maintained the same low values from the first week onwards (Figure 6.7A).

The secretion levels of pro-MMP1 by AC2 at week 1 appeared to be significantly higher than those of NC2 (p<0.001, Figure 6.7B). However, from this time point onwards the secretion levels of pro-MMP1 by AC2 cells decreased radically and stabilised at a low level, whereas the levels of pro-MMP1 released by NC2 increased significantly to higher levels (p<0.05). Finally, the same observations as for AC2 and NC2 were observed in AC3 and NC3. Briefly, the secretion levels of pro-MMP1 by AC3 were higher than those of NC3 at week 1 (p<0.001) but this trend was mirrored at week 2 and 3 (p<0.001) (Figure 6.7C).

AC1 A) *** B) AC2 0.03 NC1 0.08 NC2 ***

*** 0.06 * 0.02 *** *** 0.04 0.01 0.02

0.00 0.00

7 14 21 7 14 21

Pro-MMP1 concentration (ng/ug) concentration Pro-MMP1 Pro-MMP1 concentration (ng/ug) concentration Pro-MMP1 Time (days) Time (days)

C) AC3 0.30 NC3 *** *** *** 0.20

0.10

0.00

7 14 21 Pro-MMP1 concentration (ng/ug) concentration Pro-MMP1 Time (days)

Figure 6.7. Concentration of pro-MMP1 released by (A) AC1 and NC1, (B) AC2 and NC2 and (C) AC3 and NC3 in the RAFT TE over three weeks in culture. Protein concentration is represented as ng of MMP1 per µg of total protein. Data represents the mean ± standard deviation of n=1 donor for each condition (n=3 technical replicates for each donor). Significance of results were assessed by a two-way repeated measurements ANOVA. * p<0.05, ***p<0.001. Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

220

The concentration of total MMP2 released by the cells cultured in RAFT TE was also quantified by ELISA using the supernatant collected from the three time-points. When the results of all three NC donors were pooled together, it was evident that the secretion levels of total MMP2 increased over time to values of 1.5ng of MMP2 per µg of total protein (p<0.001 between week 1 and week3) (Figure 6.8). In contrast, the secretion levels of total MMP by AC (when the results for the three were pooled together) appeared to not significantly vary over time (p>0.05)

When comparing the secretion of total MMP2 by AC against NC in each of the time points, no significant differences were found in secretion levels at week 1 (p>0.05). However, MMP2 release levels appears to be significantly higher by NC at week 2 and

week 3 when compared to AC for the same periods (Figure 6.8).

g)  2.5 *** *** AC 2.0 NC 1.5 * 1.0

0.5

0.0 7 14 21 Total MMP2 concentration (ng/ concentration MMP2 Total Time (days)

Figure 6.8. Concentration of total MMP2 released by AC and NC in RAFT TE over three weeks in culture, represented as nanograms of total MMP2 per µg of total protein. Data represents the mean ± standard deviation of n=3 donors pooled for each condition (n=3 technical replicates for each donor). Significance of results were assessed by a two-way repeated measurements ANOVA. * p<0.05, ***p<0.001. Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

The amount of total MMP2 released by NC1 cells increased over time, on the other hand AC1 cells maintained their release levels of total MMP2 stable and significantly lower over the three time points (p<0.01, Figure 6.9A).

The same trend was observed in NC2, which also significantly increased their secretion levels of total MMP2. AC2 cells also maintained lower levels of total MMP2 release when compared to their paired normal cells (p<0.001, Figure 6.9B). 221

The levels of total MMP2 released by NC3 increased over the three time points, these were slightly increased by AC3 from week 1 to week 2 but then remained unaltered during the last week. Only differences in total MMP2 secretion levels between AC3 and NC3 at week 3 were found to be significantly different (p<0.01, Figure 6.9C).

Zymography results showed two clearly distinguished bands in every AC and NC lane, at 55 and 70kDa. These bands corresponded to the molecular weights of matrix metalloproteinase 2 (MMP2), in its pro (72kDa) and active (64kDa) forms.

There appeared to be a difference in the production levels of MMP2 between AC and NC especially at week 3 (Figure 6.9D). All three aniridic samples did not seem to increase their MMP2 secretion during the 3-week time frame. For instance, AC1 showed a very light band that did not change size or intensity regardless of the time point. Similarly, AC2 did not show striking changes in band intensity. Only AC3 showed differences over time with a slightly bigger and brighter band being seen at the end of the experiment, when compared to the previous two time points (Figure 6.9D). In contrast, MMP2 levels secreted by NC appeared to increase over time, with faint and small bands at week 1, increasing in intensity over the three weeks, reaching stronger bands at the end of the experiment. Interestingly, secretion levels by NC2 appeared to be high at week 1 compared to the other two NC samples (Figure 6.9D).

222

A) AC1 1.0 *** NC1 0.8

0.6 **

0.4

0.2

0.0 7 14 21 Total MMP2 concentration (ng/ug) concentration MMP2 Total Time (days)

B) C) AC3 2.5 AC2 2.5 NC2 *** NC3 ** 2.0 2.0

1.5 *** 1.5

1.0 1.0

0.5 0.5

0.0 0.0 7 14 21 7 14 21 Total MMP2 concentration (ng/ug) concentration MMP2 Total Time (days) (ng/ug) concentration MMP2 Total Time (days) D)

Figure 6.9. Secretion of total MMP2 by AC and NC in the RAFT TE over three weeks in culture as assessed by ELISA (A, B, C) and Zymography (D). (A) MMP2 concentration released by AC1 and NC1, (B) AC2 and NC2 and (C) AC3 and NC3, represented as nanograms of total MMP2 per µg of total protein. Data represents the mean ± standard deviation of n=1 donor for each condition (n=3 technical replicates for each donor). Significance of results were assessed by a two-way repeated measurements ANOVA. ** p<0.01, ***p<0.001 (D) Zymogram of conditioned medium obtained from RAFT TE at the following time points: 1 week and 2 weeks, 3 weeks. First lane = protein ladder, B = Blank (conditioned medium from the RAFT TE without cells). A: aniridia. C: normal. Abbreviations AC: aniridic stromal corneal cells, NC: normal stromal corneal cells.

223

Ilomastat successfully blocks the proteolytic activity of matrix metalloproteinases secreted by normal corneal stromal cells cultured in CSSCM but has no effect on cell alignment

6.4.3.1 Ilomastat blocks the proteolytic activity of MMPs secreted by NC in non- compressed collagen gels.

NC in non-compressed collagen gels were cultured with different concentrations of Ilomastat to assess the effect of blocking the MMP in contraction. Concentrations of 1μM and 10μM Ilomastat significantly reduced the contraction rates of gels populated with NC for 7 days (Figure 6.10A and Figure 6.10B). A dose-dependent response was found, with high concentrations of Ilomastat promoting lower contraction rates. In a positive control where 5% (V/V) DMSO was included to kill the cells, no gel contraction was appreciated through the time course. A negative control with (0.5% (V/V) DMSO in cell culture medium) was also included and did not affect the contraction rates of NC-populated gels. Taken together, these results showed that Ilomastat at concentrations of 1 µM and 10µM reduces contraction of NC-populated non-compressed collagen gels by blocking MMP activity.

Cell viability was not affected by any of the concentrations of Ilomastat. Viability of the positive control was significantly reduced as indirectly indicated by high rates of cell death (Figure 6.10C).

224

A) B) 100 80 ** 0 nM 10 nM 60 ** *** 1 M *** 10 M 40 ** -ve ctrl

% Contraction % +ve ctrl 20 *** *** *** 0 *** 0 1 2 3 4 5 6 7 8 Time (days) C) 50000

40000

30000

20000

10000

Metabolic activity (OD) activity Metabolic *** 0

M M   0 nM 1 10 nM 10 -ve ctrl +ve ctrl

Figure 6.10. Effect of different concentrations of Ilomastat on the contraction of untethered type I collagen gels populated by NC. (A) Representative pictures of gels at 3 different time points. Negative control: 0.5% (V/V) DMSO. Positive control: 5% (V/V) DMSO. (B) Contraction is indicated as a percentage of the initial area at time 0, i.e. 0% represents gels not contracted and 100% gels completely contracted. (C) Cell viability in untethered collagen gels using different concentrations of Ilomastat. (B, C) Data is represented as the mean ± standard deviation of n=3 biological replicates (n=3 technical replicates each). Significance of results were assessed by a One-way ANOVA. **p<0.01, ***p<0.001.

6.4.3.2 Ilomastat does not affect NC alignment and viability when cultured in RAFT TE

RAFT TE populated with NC and cultured in the highest tested concentration of Ilomastat (lowest contraction rates) was chosen to be imaged under the confocal microscope. Culture with Ilomastat did not affect the capacity of NC to align with each other (Figure 6.11). It was noted that the distribution of NC inside RAFT TE was not uniform across

225 the whole gel and that some areas presented random cell alignment (Figure 6.11). NC cultured in the presence of 0.5% DMSO as a negative control also showed high levels of cell alignment. Interestingly, cell clusters were observed when culturing the cells with 10µM Ilomastat (red triangles). To summarize, these results showed that Ilomastat did not affect the ability of NC cells to align between themselves in RAFT TE.

0 µM Ilomastat 10 µM Ilomastat -ve ctrl

Figure 6.11. Images of NC cultured in RAFT TE and supplemented with Ilomastat 0µM, 10µM and negative control. Top row scale bars= 300µm. Bottom row scale bars= 100µm. White triangles indicate areas presenting high cell alignment. Red triangle indicates areas with cell clusters.

Similar to the results obtained in non-compressed collagen gels, Ilomastat did not affect cell viability in RAFT TE (Figure 6.12). Optical density levels measured by PrestoBlue® were similar across all conditions at around 40000 A.U.

226

50000

40000

30000

20000

10000

Metabolic activity (OD) activity Metabolic 0

M M   0 nM 1 10 nM 10 -ve ctrl

Figure 6.12. Cell viability of NC in RAFT TE cultured with different concentrations of Ilomastat. Data is represented as the mean ± standard deviation of n=3 biological replicates (n=3 technical replicates each). Significance of results was assessed by a One-way ANOVA. Abbreviations OD: optical density.

Interestingly, cell clusters were previously observed when NC were culture in 10 µM Ilomastat in RAFT TE, therefore these were quantified. NC cultured in RAFT TE and supplemented with 10µM Ilomastat showed an average of 27 clusters per each image field (Figure 6.13). In contrast NC cultured in RAFT TE in CSSCM (0µM Ilomastat) and the negative control samples had an average of 1 cluster per each image field (Figure 6.13). This difference in the number of cell clusters across samples was statistically significant (p<0.001).

227

A) 0 µM Ilomastat 10 µM Ilomastat -ve ctrl

B) 50 *** *** 40

10 Number of clusters of Number

0 0 M 10 M -ve ctrl

Figure 6.13. The effect of Ilomastat on the number of cell clusters in RAFT TE. (A) Representative pictures of NC cultured in RAFT TE with 0 µM, 10µM Ilomastat and 0.5% (V/V) DMSO. (B) Number of clusters per image field. Data is represented as the mean ± standard deviation of n=3 biological replicates (n=3 technical replicates each). Significance of results were assessed by a One-way ANOVA analysis.

228

6.5 Discussion

In the previous chapter, differences in cell alignment were observed between aniridic (AC) and normal (NC) corneal stromal cells when cultured inside RAFT TE. In addition, RNA sequencing data revealed possible abnormalities in the capacity of AC to remodel the matrix. This chapter was aimed to confirm these possible differences in matrix remodelling and to possibly link them to the observed low alignment in AC. A short ECM characterisation of AC and NC-populated RAFT TE showed that there were no differences in the appearance of the ECM. However, a study of the MMPs, proteolytic enzymes involved in ECM remodelling, confirmed that AC express and secrete lower levels of MMP1 and MMP2. Finally, it was interesting to assess whether there was a link between the lower expression and secretion of MMPs by AC and their lower ability to align themselves. This was achieved by blocking the activity of the MMPs released by NC in RAFT TE, but it did not show a significant decrease in cell alignment. These results are novel because although there have been precedents showing a downregulation of MMP9 in the aniridic corneal epithelium [208], no previous evidence had identified differences in MMP production in the stroma of ARK corneas.

RAFT TE populated with aniridic and normal corneal stromal cells did not show differences in ECM transparency and appearance

Transparency is an essential property of the cornea, being gradually lost in patients with ARK [140]. To measure this loss of transparency in patients over time would require frequent clinical examination (which may be intolerable in ARK patients with photophobia). RAFT TE, populated with AC, was therefore tested as an in vitro model of transparency loss. However, no significant differences in transparency of RAFT TE populated with AC or NC were found. An opportunity for longer-term matrix remodelling by the cells inside the RAFT TE may be required to see a difference. Alternatively, RAFT TE might not provide a sufficiently complex environment in which to study this important characteristic of native human cornea. This is because the organisation of the collagen fibres in RAFT TE populated by AC and NC did not appear to be different. They presented a wide range of diameters from very narrow to wide ones, which may be aggregates of single fibres. Their appearance was similar that of collagen fibres produced by human limbal epithelial cell on plastic compressed collagen gels [253].

Taken together these results suggest that both AC and NC do not alter the transparency of the RAFT TE in the time course studied in vitro, and that collagen fibres of similar properties constitute the AC and NC-populated RAFT TE. These results did not support

229 the hypotheses that AC and NC remodelled the matrix in different ways. However, longer experiments might be necessary to see an effect on the remodelled ECM.

MMP1, MMP2, MMP3 and MMP14 govern the ECM remodelling process

MMPs are zinc-dependent endopeptidases whose substrates in the cornea include components of the basement membrane and stroma. MMP activity is necessary for wound healing regulation, tissue remodelling and allowing cell migration through the extracellular matrix [300]. MMPs are secreted in a latent form (pro-MMP) that needs to be activated before they can start degrading their target substrates.

Even though culturing corneal stromal cells in the RAFT TE was not explicitly designed as a model of corneal wound healing, cells were grown in CSSCM supplemented with 2% of serum, and PDGF and were highly proliferative. Therefore, this model may mimic aspects of stromal wound healing conditions in vivo, when keratocytes are activated, by the signalling of PDGF among other factors [69, 303]. It is expected that in this activated state stromal cells remodel and migrate through the collagenous matrix. In addition, collagen fibres that compose the RAFT TE are randomly organised, and stromal cells might also remodel the matrix to create collagen and cellular alignment, as in the native corneal stroma [75, 304]. To do so, the stromal cells might secrete MMPs. In support of this theory, different studies have previously demonstrated that MMPs are involved in the remodelling and contraction of fibroblast-populated type I collagen gels [191, 302]. Briefly, Daniels et al. studied non-compressed collagen gels with different mechanical properties to the compressed collagen gel (RAFT TE) used in the present chapter [302]. They observed expression and secretion of MMP1, MMP2 and MMP3 by fibroblasts populating the gels during cell-mediated contraction. Moreover, Massie et al. also observed the expression and production of MMPs by limbal fibroblasts cultured in the RAFT TE [191].

In the present study, the MMPs that were highly expressed at the mRNA levels by AC and NC were MMP1, MMP2, MMP3 and MMP14, as shown by the RNA sequencing data. There are precedents showing that PAX6 regulates the expression of MMPs [232, 305, 306]. However, this has been only demonstrated in MMP9, which is expressed by the corneal epithelial cells but not the stromal cells [305]. Interestingly, MMP9 was significantly downregulated in AC compared to NC, as showed by RNA sequencing results. A discussion about the possible role of these MMPs released by AC and NC during their culture inside RAFT TE is presented in the next paragraphs.

230

The main substrate of MMP1 is type I collagen gel, the structural component of RAFT TE. Therefore, the presence of MMP1 at both mRNA and protein expression levels in the supernatant collected from AC and NC cultures was expected. MMP1 cleaves the collagen triple helix in a site-specific manner, which results in collagen fragments that are thermally unstable around 37°C, making them susceptible to further degradation by other MMPs.

MMP2 was also expressed and produced at high levels by AC and NC cultured in the RAFT TE. Interestingly, Massie et al. also found MMP2 to be secreted by both normal and ocular mucous pemphigoid fibroblasts in RAFT TE [191]. They suggested that the secretion of MMP2 was associated to an attempt from the fibroblasts to remodel the ECM to achieve a closer resemblance to the native. However, the secretion of MMP2 was reduced by the addition of epithelial cells on top of the RAFT, perhaps by suppressing the activated fibroblast phenotype.

MMP14 was highly expressed in both AC and NC cultured in the RAFT TE. MMP14 is a type I transmembrane protein that is activated intracellularly and relocated to the cell surface when active. It has collagenolytic activity on collagens I, II and III, and the ability to activate pro-MMP2, controlling the balance between ECM synthesis and degradation [307]. Agreeing with the results presented herein, Mmp14 expression has been observed in keratocytes of normal and, at higher levels, wounded rat corneas [308]. Despite not being a focus in this project, recent published evidence points to the crucial importance of MMP14 in corneal angiogenesis [309-313]. For example, Mmp14 has a role in promoting neovascularisation of the cornea of mouse models by upregulating the expression of VEGF [313]. Interestingly, MMP14 released by keratocytes also presents anti-angiogenic properties by preventing the proliferation of calf pulmonary artery endothelial cells [309].

There have been precedents that show alterations in MMP9 in the ARK corneal epithelial cells [297, 305, 314]. Interestingly, RNA sequencing analysis showed that MMP9 was significantly downregulated in AC compared to NC. However, it was expressed at very low levels, and further analysis by ELISA showed that no MMP9 protein was detected in the AC nor NC-populated RAFT TE supernatant. Similar results have already been shown, in which MMP9 is expressed and produced at high levels by epithelial cells and not by stromal cells [191]. In a study by Massie et al. MMP9 expression was only detected in epithelial cells cultured together with limbal fibroblasts in the RAFT TE, but not if the later were cultured alone [191].

231

Remarkably, there is extensive evidence that the production of MMP9 is regulated by the PAX6 transcription factor [305, 314, 315]. In a study using Pax6+/- mice, the regeneration of the epithelium was abnormal and the expression of Mmp9 in the cornea was significantly lower than in the wild-type mice [314]. MMP9 is a typical extracellular matrix proteinase secreted by the corneal epithelial cells and involved primarily in corneal epithelial wound healing and regeneration [316-318]. Although its expression has also been found in the corneal stroma in pathogenic conditions, such as primary Sjorgen’s syndrome [319] and cancer [320], very low expression was found in AC or NC cultured in RAFT TE. This suggests that MMP9 might be residually expressed by AC and NC in the RAFT TE, but it might be less relevant to events happening in the stroma of ARK patients nor in healthy individuals. Noteworthy, despite not being from the scope of this study, MMP9 is reported to activate latent TGFβ in epithelial cells [321]. Consistent with these results, inhibition of MMP9 in lung fibroblasts reduced active TGFβ and suppressed 3D collagen gel contraction [322]. TGFβ can also modulate the release of MMP9 [323]. Kobayashi et al. have shown that MMP9 production by unstimulated fibroblasts is low. Together, these observations suggest that the MMP9 TGFβ feedback loop participates on matrix remodelling, and its activation might be a triggering signal for corneal scarring. In the presented study, there was no addition of TGFβ, and MMP9 was observed in residual levels. Further studies would be needed to assess whether these processes are affected by the PAX6 haploinsufficiency.

In the normal cornea, there are undetectable or very low levels of latent MMP secretion. In the event of an injury, MMP secretion is upregulated and MMPs are activated [317]. Specifically, MMP1, MMP2, and MMP9 are essential for wound healing in the anterior cornea, as demonstrated in rabbits by Mulholland and colleagues [317]. They showed that these three proteins are key mediators of epithelial cell migration and participate in the remodelling of corneal stroma and in the reformation of epithelial basement membrane [317]. In vivo, stromal cells in health reside in a very well organised matrix that provides them with important cues (physical and chemical) that regulate their behaviour. These include the perfect collagen lattice organisation, with periodic distance between adjacent collagen fibres. These stromal cells have critical roles on matrix organisation and remodelling by releasing MMPs. Despite being a well-suited model of the native stroma as previously explained [324], the collagen fibres of the RAFT TE lack organisation, being randomly distributed. These states of random disorganised matrix may resemble those of wound healing and scar tissue and might act as triggering signals for the secretion of MMP by the stromal cells for matrix remodelling. Therefore, these

232

cells might remodel the RAFT TE matrix, in an attempt of the cells to recapitulate the collagen fibril organisation of healthy stroma.

Aniridic corneal stromal cells express and secrete lower levels of MMP1 and MMP2 than normal corneal stromal cells

Results obtained from the present work supported the hypotheses that AC and NC might have different matrix remodelling capacities, by expressing and producing lower levels of MMP1 and MMP2. Real-time qPCR showed lower expression of MMP1 and MMP2 in AC compared to NC after three weeks in culture inside the RAFT TE. Consistently, zymography showed that NC secreted higher levels of total MMP2 than AC. However, zymography is only a semi-quantitative assay and further quantification by ELISA was required to compare the amount of MMPs released by the two cell populations. ELISA showed lower secretion of MMP1 and MMP2 by AC compared to NC. As previously explained, there might be a difference in proliferation and/or apoptosis between NC and AC, resulting in higher numbers of NC populating the RAFT TE at week 3, which could also account for the difference in MMP release. To overcome this issue, the concentration of MMP was standardised to the concentration of total protein released by the cells, to elucidate the amount of secreted MMP as a proportion of total protein.

The secretion of pro-MMP1 by NC and AC cultured inside the RAFT TE was detected. MMP1 is the first enzyme to be synthesised by stromal cells following corneal injury [325] and it is part of their long-term response, in which they slowly remodel the collagen of the stroma [300]. It is synthesised in its latent form pro-MMP1 and it needs to be activated to allow stromal remodelling. Using rabbits’ corneas as models, Mmp1 was localised in the stroma and basal epithelium after anterior keratectomy and it was activated only when both epithelial and stroma were injured [317]. Thus far, there has been no evidence of a link between PAX6 expression and pro-MMP1 secretion. In the present work a clear difference in pro-MMP1 secretion between AC and NC was observed. Briefly, AC2 and AC3 samples secreted higher levels of pro-MMP1 than their paired controls (NC) at week 1. Thereafter a significant decrease (p<0.05) over the next two weeks was observed. Secretion levels of pro-MMP1 by AC1 samples were steadily low during the three weeks of culture. In contrast, the amount of pro-MMP1 secreted by NC increased over the three- week observation period (p<0.01 from week 1 to week 3).

The secretion of both pro (72kDa) and active (64kDa) MMP2 by NC and AC cultured inside the RAFT TE was detected by zymography. MMP2 has already been detected in all corneal layers. In the stroma, it is secreted by stromal fibroblasts when activated after injury and it is involved in the long-term response of wound repair [326]. Azar and 233 colleagues found high levels of Mmp2 in the stroma of rabbits corneas after keratectomy [327]. Corroborating observations were made by Mulholland and colleagues, who localised Mmp2 using immunohistochemistry in the stroma of rabbit corneas after keratectomy. It was present whether the epithelium was removed or not during keratectomy, suggesting that Mmp2 plays a role in early basement membrane assembly, with stromal remodelling being partly independent of epithelial injury [317]. There has been evidence that PAX6 might regulate the expression of MMP2. A study with colorectal cancer cells showed that, following PAX6 upregulation, there was an overexpression of MMP2, suggesting that PAX6 is involved in the regulation of MMP secretion [232]. In a computation approach, Urrutia et al. showed that binding sites for PAX6 were found in the regulatory region of MMP2. They also showed that overexpression of PAX6 resulted in an upregulation of MMP2 in breast cancer cells [306]. However, none of this evidence studied corneal stromal cells. In accordance with the mentioned observations [232], the herein results show lower levels of MMP2 secreted by AC compared to NC. Specifically, AC2 and AC3 secreted lower levels of total MMP2. AC1 also secreted very low levels of total MMP2 during the three weeks in culture. In contrast, the concentration of MMP2 released by normal cells increased over the three weeks in the RAFT TE culture. These observations suggest that different mutations on the PAX6 gene may alter MMP secretion, however a bigger sample size would be necessary to confirm this hypothesis.

Taken together these results suggest that, following injury, aniridic cells in the native stroma might have impaired MMP secretion profiles. Moreover, aniridic cells may not able to secrete the necessary amounts of MMP1, to cleave the type I collagen, and of MMP2, to continue to remodel the extracellular matrix. The altered MMP secretion profiles may therefore result in altered wound healing capacity.

Ilomastat successfully blocks matrix metalloproteinase activity but does not show an effect on cell alignment

In the work here presented, a first set of experiments was performed with NC cultured in hyper hydrated, non-compressed collagen gels, to assess whether Ilomastat was a suitable compound to block MMPs on corneal stromal cells cultured in CSSCM. Daniels et al. had previously showed that treatment with Ilomastat resulted in reduced proteolytic activity of MMPs released by Tenon’s capsule fibroblasts on hyper hydrated collagen gels, preventing their contraction [77]. Here, it was shown that Ilomastat was also able to block the activity of MMPs secreted by normal corneal stromal cells in the non- compressed hyper hydrated collagen type I gel. The Ilomastat effect was shown to be dose-dependent, in which higher concentrations of the MMP inhibitor caused slower

234

rates of gel contraction. This study indirectly confirmed that Ilomastat could be used to inhibit the activity of MMPs secreted by NC inside the RAFT TE. To study whether this inhibition caused lower scores of cell alignment similar to those seen in aniridic samples, Ilomastat at the highest concentration (10µM) was used to inhibit the activity of MMPs secreted by NC inside the RAFT TE.

Unexpectedly, blocking the activity of MMPs did not affect the ability of NC cultured in the RAFT TE to align. Wolf et al. showed that cell migration in 3D collagen gels results from a balance between MMP activity and physical space constraints [301]. On the one hand, when the ECM pores were too small, MMP activity was required for cells to migrate through them. On the other hand, when the pores were big enough and did not constitute space confinement, cells migration was independent of MMP activity. It was hypothesised that in a compressed collagen gel (RAFT TE), MMP activity was required for cells to move and create alignment. However, the results found did not support this hypothesis, because cell alignment was observed even when blocking MMPs with high concentrations of Ilomastat. Thus, it is possible that corneal stromal cells are able to squeeze through the RAFT TE collagen fibres without the need of MMP1 and MMP2. These results suggest that alterations in MMP-regulated matrix remodelling are not the cause of low cell alignment observed in aniridic cells cultured in the RAFT TE.

Ilomastat treatment resulted in a higher number of cell clusters inside the RAFT TE

Interestingly, cell clusters were observed when NC were cultured in 10µM Ilomastat in RAFT TE and subsequent quantification showed a significant increase in their number in these conditions when compared to the controls. The cell clustering might be a result of cell proliferation and impaired ability to remodel the surrounding cell matrix to allow cell migration. As discussed in the previous section, some of the cells are able to migrate through the RAFT TE and show alignment. However, some cells might be constrained in space and rely on MMP activity to degrade the surrounding ECM to migrate [301]. If Ilomastat is successfully blocking MMPs activity, this would result in inability for cell migration and consequent cell clustering arising from continued proliferation.

Cell clustering is an important event during tissue morphogenesis, homeostasis and pathological disorders (reviewed in da Rocha-Azevedo and Grinnell [328]). For instance, in scarring and fibrosis clusters of contractile fibroblasts are seen in hypertrophic wounds [328]. Also in pulmonary fibrosis, clusters of myofibroblasts have been observed [329]. Cell clustering has been mimicked in vitro by culturing cells in nonadhesive substrates or in 3D scaffolds [330, 331]. On these substrates, cell dispersion from clusters and 235 migration can be induced by the addition of PDGF [330, 331]. In the results here presented, NC were cultured in RAFT TE and fed with CSSCM, which contains PDGF, so migration could be driven by similar mechanisms. Interestingly, it was demonstrated by Stringa et al. on vascular smooth muscle cells, that PDGF-induced cell migration on collagen relies on the degradation of this material. They showed that upon culture on degraded fragments of collagen, PDGF induced cell motility, whilst cells cultured on inert fibres of collagen type I did not show PDGF induced cell motility [331]. Rocha-Azevedo et al. unravelled more details for this mechanism by showing that the proteolytic activity of MMP2 is necessary during PDGF-induced cluster dispersion [330]. In their studies, MMP2 was detected at high levels in cultures of human foreskin fibroblasts undergoing cluster dispersion compared to cultures maintained in clusters states, suggesting that MMP2 plays a key role during this process. They also showed that MMP2 silencing supresses cluster dispersion when cells are cultured in the presence of PDGF [330]. These data might be relevant to the findings presented herein since it could help explain the mechanisms driving increased cell clustering in the presence of Ilomastat. Cell proliferation and cell migration of NC in RAFT TE might be partially driven by PDGF, which in its turn may require MMP2 activity. Therefore, blocking the activity of MMP2 with Ilomastat, might show a reduced cell migration ability in the RAFT TE.

Cell migration has been previously related to PAX6 function. For instance, embryonic neural crest cells from Pax6+/- rats cannot migrate to their destination in the nasal rudiments, suggesting that PAX6 is essential for the correct migration of neural crest cells from the brain [332]. Similar observations have been described using murine Pax6+/- corneal epithelial cells, which lose their ability to move centripetally from the limbus to the central cornea, as they would do in health [129, 234]. Thus far, no evidence has been published regarding the impaired migration of PAX6 haploinsufficient corneal stromal cells. As previously explained, during corneal wound healing PDGF is released by the epithelial cells to promote the proliferation, differentiation, and the migration of fibroblasts to the wounded site [69, 303]. If aniridic cells are not able to migrate to the injury site, they might become activated and start to secrete matrix in inappropriate locations, resulting in the typical high levels of scarring seen in aniridic patients. This may help to explain the reasons behind the exaggerated wound healing response observed in aniridic corneas.

Taken together, these results showed that blocking MMP activity did not have any effect on cell alignment, however it possibly impaired cell migration ability resulting in higher levels of clustering inside the RAFT TE. By extrapolation, corneal stromal wound healing

236

in ARK conditions might be altered through the process of matrix remodelling and cell migration due to altered MMP activity.

237

238

CHAPTER 7: GENERAL DISCUSSION

7.1 Impact of expanding and culturing normal corneal stromal cells in RAFT TE

The results presented in Chapter 3 of this thesis showed that central and limbal corneal stromal cells have similar phenotypes in vitro. Until recently, it was broadly accepted that different populations of cells reside in the central cornea and in the limbus. The central cornea was populated by the terminally differentiated keratocytes [8, 49, 51], while residing in the limbal stroma there were their progenitors, a population of stem cells (the so called CSSC) [54, 57, 221]. However, some studies have started questioning this theory [333]. It has been shown that keratocytes retain neural crest progenitor (their embryonic origin) properties, and when injected into a new environment they change their phenotype, proliferating and contributing to the development of different tissues, suggesting that they have a certain degree of multipotency [333]. Interestingly, there are reports attributing these phenotypical changes to the culture conditions, and demonstrating how different culture media can modify cell phenotype [181, 203]. Other studies directly suggest that there is a neural crest-derived population of multipotent stem cells in the adult cornea [58, 187]. The data shown in this thesis supports the hypothesis that cells isolated from the central corneal stroma maintain a certain degree of plasticity and when cultured in adequate culture conditions they acquire a CSSC-like phenotype. Further studies would be needed to assess whether corneal stromal cells can also differentiate into different lineages (adipocytes, chondrocytes, osteoblasts), as it has been demonstrated elsewhere with CSSC, and therefore confirm that they have stem cell characteristics [334].

These results, together with the successful culture of corneal stromal cells inside the RAFT TE, are novel findings that open new possibilities for using central corneal stromal cells in corneal regenerative strategies. RAFT TE has been previously shown to be a successful vehicle to deliver LESC for ocular reconstruction in patients with LESC deficiency [253, 257]. In the past few years, there has been growing interest in coculturing LESC with limbal stromal cells on RAFT TE, in a closer attempt to recreate the limbal niche for corneal repair [55, 253]. Early studies used limbal fibroblasts (cultured in medium supplemented with 10% serum) cocultured with LESC, which remained viable and supported the growth of the epithelium [191, 253, 255]. However, Wu et al. showed that fibroblasts did not secrete corneal-like ECM when they were driven to differentiate back into keratocyte-like cells in serum free medium [192]. Moreover,

239 fibroblasts do not support the pre-expansion of LESC in vitro, prior to transfer into RAFT TE [191]. In addition, coculturing LESC with fibroblasts, relies on the use of animal derived feeder cells, which is not favoured for translation into clinical application [55]. On the other hand, CSSC can differentiate into keratocyte-like cells and secrete a specialised matrix consisting of a multilayer collagen lamellae organised orthogonally, in a similar pattern to the stroma-specific matrix [192]. Moreover, the secreted matrix is rich in the associated stromal ECM components, such as keratocan, lumican, and keratan sulphate [192]. CSSC have been recently cocultured with LESC in RAFT TE, supporting their growth while maintaining stem cell-like characteristics [55]. In a different approach, CSSC have also been cultured as a single cell population in RAFT TE to be used as a stem cell therapy for corneal scarring [256]. CSSC embedded in the RAFT TE remarkably prevented corneal scarring and suppressed gene expression of the fibrotic marker Col3a1, in a murine wound-healing cornea model [256]. Additionally, CSSC remained viable for more than 1 week and were successfully cryopreserved, working towards an off-the-shelf product [256]. Taken together, these observations suggested that (i) CSSC are appropriate to culture with LESC in RAFT TE, and (ii) can also be cultured on their own, showing great clinical therapeutic potential.

Here, for the first time, it was shown that central corneal stromal cells present phenotypical similarities with CSSC when cultured in vitro, and that they can be successfully cultured in RAFT TE. CSSC have shown anti-inflammatory and anti-scarring properties, crucial features to be used for therapeutic purposes [176, 335, 336]. If further experiments confirm that corneal stromal cells share the same anti-inflammatory and anti-scarring properties, they might constitute a suitable cell source for cell-based therapies for corneal regeneration. These would be advantageous as the central cornea could be also used for therapeutic processes even if it did not meet the requirements for corneal transplant.

7.2 Impact of expanding and culturing aniridic corneal stromal cells in RAFT TE

To the best of our knowledge, this is the first extensive study using primary aniridic corneal stromal cells. To date, the observations suggesting an altered stroma in ARK have been part of broader studies. These mostly target the whole cornea of ARK patients or Pax6+/- mouse model, and mainly focus on the altered epithelium [137, 160, 161, 211]. However, none of these studies have isolated corneal stromal cells to assess their phenotype in vitro. Here, aniridic corneal stromal cells were successfully expanded in vitro and enough numbers were obtained to perform different characterisation 240

experiments at both mRNA and protein level. Additionally, aniridic corneal stromal cells from four different donors were successfully banked in liquid nitrogen and stored until further use. This was an invaluable achievement to study different phenotypical features of these cells, as it allowed preservation of the majority of cells while preliminary experiments were being performed and avoided excessive cell passaging.

Interestingly, aniridic corneal stromal cells showed lower proliferation than normal corneal stromal cells, but otherwise did not show any other phenotypical differences when cultured in 2D. Both cell populations acquired CSSC-like characteristics when cultured in CSSCM and had similar abilities to differentiate into keratocyte-like cells when cultured in KDM. These results reconfirmed the plasticity of cells isolated from the central corneal stroma under the conditions studied. Further characterisation of these cells was performed in a 3D environment (RAFT TE) to better mimic their surrounding ECM in the native stroma. RAFT TE may be a useful model for ARK in vitro, where aniridic and normal stromal cells successfully remained viable and showed differences in phenotype. Lower proliferation of aniridic corneal stromal cells was further confirmed in the RAFT TE. In addition, lower degrees of cell alignment and rounder morphologies (higher cell circularity) were observed in aniridic corneal stromal cells compared to normal corneal stromal cells. Broader cell body and random organization are typical features of myofibroblasts during corneal wound healing [266, 337]. Moreover, it has also been suggested in an equine model that fibroblasts proliferate more than myofibroblasts in vitro [338]. The results presented here, included lower proliferation, more circular bodies and random organization suggested that aniridic corneal stromal cells might be differentiating into myofibroblast-like cells while normal corneal stromal cells maintain a more fibroblastic-like phenotype. It has previously been reported that stromal scarring is an important event in the progression of ARK [61, 140] and higher levels of αSMA have been observed in the subepithelial pannus of ARK compared to normal corneas, suggesting a presence of myofibroblasts [161]. However, when assessing the expression of αSMA, no significant differences were observed between the two types of cells.

When culturing corneal stromal cells in medium containing 10% serum, cells are known to be in an activated state, possibly resembling that of repair fibroblasts during wound healing [186, 339]. In the results presented here, aniridic and normal stromal cells were cultured in CSSCM which contains 2% serum, so a similar phenotype might be induced. However, RAFT TE was not specifically designed as a model of wound healing so the conditions might not be suitable to study this process. In the future, promoting a wound healing environment by adding factors such as TGFβ might be a good option to further 241 study the differentiation of aniridic cells into a scarring phenotype [82, 340]. Another interesting option would be to culture corneal epithelial cells on the surface of the RAFT TE (with corneal stromal cells embedded on the collagen gel) and create an area of epithelial debridement [341]. During reepithelialisation of the debridement area, the migrating epithelium might then secrete factors that would also trigger the process of wound healing in the stromal cells [342].

To further characterise the phenotype of aniridic and normal corneal stromal cells inside the RAFT TE, RNA sequencing was performed. Differences in their transcriptome were observed, with differentially expressed genes involved in several biological processes, including matrix remodelling. Further experiments confirmed lower expression and secretion profiles of MMP1 and MMP2 by aniridic corneal stromal cells, proteins with central roles on matrix remodelling [73]. These results support other studies that have previously identified PAX6 as a regulator of another MMP (MMP9) [297].

It was hypothesised that the lower expression of MMP could account for the impaired cell alignment observed in aniridic corneal stromal cells. However, further experiments showed that reduced MMP activity (blockage induced by Ilomastat) did not affect cell alignment when normal cells at least were cultured in the RAFT TE. Cell mobility within a matrix depends on the space confinement, the capacity of the cells to break the matrix and cell ability to change its body shape [301]. Here, MMP activity was blocked, suggesting that the type I collagen fibrils around the cells may not have been cleaved. Therefore, for the cells to align to each other as seen on the results here presented, they might have squeezed between the collagen fibres.

7.3 Involvement of matrix metalloproteinases on corneal scarring

Interestingly, when MMP activity was blocked, higher number of cell clusters were observed in the RAFT TE. Cell clusters might be a result of cell proliferation and impaired ability of cells to remodel the surrounding cell matrix to allow cell migration [331]. If aniridic cells have reduced capacity to express and secrete MMP, they might have impaired migration within the ECM. Matrix remodelling, migration and proliferation are all processes essential during the normal course of corneal wound healing (reviewed in [343]). Numerous studies have suggested that aniridic corneas might present abnormal wound healing processes [130, 161, 166, 208]. In the stroma, higher levels of cell apoptosis have been observed in the Pax6+/- mouse [208]. Ramaesh et. al therefore hypothesised that an increase in stromal apoptosis after wounding combined with a deficient and delayed ECM metabolism might be the cause for the aggravated corneal changes observed in ARK [208]. The results presented here, showed lower production 242

levels of MMP by aniridic corneal stromal cells, further supporting this hypothesis. However, further studies are needed to enlighten how the reduction in MMP activity affect the process of wound healing in ARK.

7.4 Potential mechanisms by which aniridic and normal corneal stromal stem cells show phenotypical differences in vitro are needed to enlighten how the reduction in MMP activity affect the process of wound healing in ARK.

It is a matter of great interest how, although both expressing residual levels of PAX6, the stromal cells from normal and aniridic corneas show phenotypical differences when cultured in vitro. Further experiments would be needed to assess the mechanism by which these differences are triggered and maintained. Nevertheless, potential explanations for this phenomenon are summarised in the following paragraphs.

As previously mentioned, keratocytes populating the stroma of the adult cornea do not express PAX6, contrary to their progenitors [57]. By culturing aniridic and normal corneal stromal cells in conditions previously defined for their progenitors, the so-called CSSC, they might be shifting their phenotype towards a more progenitor state, recapitulating events that might happen in vivo when cells expressed PAX6. The results presented here in the first chapter, in which central stromal cells and those isolated from the limbal stroma, where CSSC reside, behave similarly in vitro, support this statement. It has been previously shown elsewhere that even low expression levels of transcription factors can produce an effect on the cell phenotype. For example, residual levels of the reprogramming factors Oct4, Klf4, Sox2 and c-myc prevent the differentiation of induced pluripotent stem cells towards other lineages, maintaining their stemness [344]. However, in the cited publication they observe the residual expression of these markers by endpoint RT-PCR, and thus cannot be compared to the concentration of PAX6 shown in the present work (Ct values ~35). It is interesting, though, to note that Aniridia 1 sample showed the lowest expression of PAX6 (Ct values undetermined) and showed the most striking differences in phenotype compared to the normal cells, suggesting a possible PAX6 dose response. It also needs to be noted that PAX6 expression is lost as the cells are passaged. Therefore, the cells cultured herein might have expressed higher levels of PAX6 in previous passages and may still retain the difference in phenotypes that this caused.

Another plausible explanation is that, although PAX6 is not expressed in the adult corneal stroma, the levels of PAX6 exhibited during development in the aniridic corneal

243 stroma still have an effect during the adult life. PAX6 necessary during stromal development might act by modulating the transcription of other genes that are essential for normal adult stromal homeostasis. A possible mechanism for this is the PAX6 epigenetic regulation of its downstream genes. This capacity has been previously shown in the developing neocortex where PAX6 drives changes in the chromatin remodelling complexes, modulating gene expression during cortical cell differentiation [345]. Additionally, there is also evidence that PAX6 binds to methyltransferases and regulates methylation at the promoters and enhancers, regulating downstream genes, for example Plekha1 gene at the lens and retinal pigmented epithelium [346]. In this scenario, certain genes downstream of PAX6 might be differently regulated at the epigenetic level, and these differences might still be conserved and noticeable during adulthood.

7.5 Do aniridic and normal corneal stromal cells support differently the epithelium?

One of the prevalent hypotheses regarding the driving mechanisms of ARK is a dysregulation of the limbal epithelial niche. In the limbal niche the stroma is populated by the corneal stromal stem cells, which together with other factors and cell types, support the LESC homeostasis [55, 111]. Differences in the AC and NC phenotype might indicate that these cells have different capacity to support the epithelium, with higher impairment of AC to support LESC homeostasis It is plausible that LESC/CSSC or epithelial cells/keratocyte interactions are disrupted in ARK and this contributes to the development of this condition. It would be of great interest to further assess these interactions by culturing epithelial cells on the surface of the RAFT TE in a mimicry of the in vivo environment.

7.6 Genotype-phenotype correlation between the aniridic samples

High variability in the phenotype between aniridic corneal stromal cells isolated from different aniridic samples was observed. However, these results were not surprising, because the aniridic patient cohort also exhibits a high degree of heterogeneity [111]. A wide range of phenotypes affecting different parts of the eye have been observed in aniridic patients [347]. In the cornea, this broad variability ranges from no keratopathy being observed to severe visual loss due to ARK [226]. Even within the same family, patients sometimes show different phenotypes, suggesting that a genotype-phenotype correlation in aniridia is complicated to identify [111]. PAX6 can additionally be regulated at epigenetic level, resulting in different gene expression, and therefore phenotype, even within the same family [348]. In addition, PAX6 interacts with a broad network of genes,

244

suggesting that the aniridic phenotypes might also depend on the regulation and variants of upstream and downstream genes [209]. It would be of the utmost interest to obtain the clinical phenotype of the aniridic eyes used for this study, in order to try to establish a link between severity of ARK and the phenotype of the cells in vitro.

Surprisingly, in the results here presented a correlation could be observed between the PAX6 mutation and the cell phenotype. AC1 and AC4 had C-terminal extension mutations (CTE) whereas AC2 and AC3 had premature termination codons (PTC). Interestingly, AC1 presented a more different transcriptome (high number of DEG) compared to normal cells than AC2 and AC3. In addition, AC1 and AC4 had lower expression of MMP1 and MMP2, compared to AC2 and AC3. Moreover, the same correlation was observed regarding KI67 expression, with cells that had CTE mutations showing slower proliferation rates. These results are in agreement with previous hypotheses that CTE mutations have a more severe phenotype [209, 226, 349]. In the future, increasing the number of samples tested within each mutation group would be of utmost interest to validate this correlation.

7.7 Limitations of this study

The first and biggest limitation for the study here presented was the low number of biological replicates available. Aniridic patients presenting ARK phenotypes are not advised to undergo surgery unless their vision is severely threatened, since their ocular tissue often responds aggressively to insult and manipulation [149]. Occasionally, when a patient underwent surgery at the Moorfields Eye Hospital, the redundant tissue from their corneas became available for this project. However, the numbers were very scarce. Although not at the same level, control tissue from normal cadaveric donors was also hard to obtain, being only available when it did not meet the necessary requirements to be used for transplant.

To overcome this limitation, it would be useful to create immortalised cell lines from the available primary cells. Although some features of cell lines do not always replicate those of primary cells, there would be enough cell numbers to perform more experiments and thoroughly characterize this condition [350].

Another very interesting option would be to make induced pluripotent stem cells (iPSC) from aniridic patients. iPSC are a suitable cell source for banking and expansion in vitro, and have the advantage that can be differentiated into different cell types. Following up from the previous section, a coculture of corneal epithelial and stromal cells in RAFT TE would be a suitable model to further study the interactions between aniridic corneal

245 stroma and epithelium and identify what interactions between these two cell types govern the processes observed in ARK. Stromal cells can be embedded within the matrix as shown in this thesis and epithelial cells can be seeded on top of the RAFT TE, as a way of mimicking these two corneal layers [55]. To do this, an epithelial cell source with a PAX6 mutation would be needed. Here, when culturing primary cells from ARK corneas, there was also growth of epithelial cell colonies. Two published studies have reported the culture of human primary epithelial cells from two patients [173, 351], but as found here with stromal cells, the number of cells obtained is limited. To date, there are no published studies that have reprogrammed iPSC from different patients. If this were to be successfully achieved, iPSC lines could be differentiated into both epithelial [352, 353] and stromal [354-356] cells. During the course of this thesis, two attempts of reprogramming the aniridic central stromal cells failed. This may be because they are too sensitive to survive the electroporation step of the transfection process. In the future, cells from urine or skin could be reprogrammed, using protocols that are well described and more robust [357, 358].

Secondly, there are also some limitations of using the normal (healthy) tissue. Aniridic cells were isolated from fresh tissue that had been stored for a maximum of two days post-extraction. In contrast, normal cells were isolated from tissue stored in organ culture for up to four weeks, since its primary allocation was for corneal transplantation rather than research. It is a routine procedure that donor tissue is stored in the eye banks for up to 4 weeks to assess whether it is suitable to use for transplantation. If not suitable, then the tissue is donated to research. Tissue stored in organ culture had to be used in this project because finding fresh tissue directly donated for research is of utmost difficulty. Moreover, when available, this tissue is from diseased donors that were in general older than the aniridic patients, a factor that could contribute to differences. However, it has been reported that there are no significant differences between donor age and the clinical outcome when donor corneas are transplanted [359], suggesting that age probably does not contribute to the phenotype of the cells isolated.

7.8 Future work

Possible future lines of study arising from the results presented herein have been discussed. However, they are summarised below for reference:

- To differentiate aniridic and normal corneal stromal cells towards a keratocyte- like phenotype in RAFT TE. - To further study the impact of low levels of MMP expression and production by aniridic cells on wound healing/scarring model in RAFT TE. 246

o The effect of TGFβ on these wound healing models would enlighten about the higher scarring potential of aniridic stromal cells. - To make immortalised cell lines from the aniridic corneal stromal cells to further characterise them and therefore develop new clinical approaches for aniridic treatment. - To reprogram patient cells into iPSC and differentiate them into epithelium

7.9 Concluding remarks

In conclusion, it was demonstrated that aniridic corneal stromal cells can be cultured in vitro both in 2D and 3D environments. Moreover, RAFT TE was shown to be a useful model to study ARK in vitro. Characterisation of aniridic corneal stromal cells and comparison with their non-diseased counterparts showed that some phenotypical features of the former cells are altered. Namely, differences were found in proliferation capacity, cell alignment inside the RAFT TE and expression and production of matrix metalloproteinases. This work is novel and contributes to the evidence suggesting that aniridic corneal stromal cells also contribute to the development of ARK. The results presented here are the platform for future studies to carry on the characterisation of aniridic corneal stromal cells. Understanding the biological mechanisms behind ARK is the first step to the future development of therapeutic strategies.

247

248

APPENDIX

8.1 RNA quality, purity, and integrity of NC, NL and AC in 2D

The quality, purity, and integrity of the RNA isolated from NL, NC and AC was assessed. Namely, RNA was isolated from cells cultured in CSSCM at day 1 (CC) and at day 5 (SD) and from cells cultured in KDM at 3 different time points.

All samples presented a ratio of absorbance at 260/280 around 1.8-2, which is considered pure for RNA (Figure A.8.1). All samples amplified β-actin in a control RT- PCR, which confirmed the good integrity of the RNA.

Figure A.8.1. Purity and integrity of RNA isolated from NL, NC and AC cultured under different conditions: CC, KDM with FGF at week1, 2, 3, KDM without FGF at week 1, 2, 3 and SD. (A) Values of 260/280 absorbance measured in the nanodrop as an indication of the purity of RNA. Values between 1.8-2.0 are considered pure for RNA. (B) RT-PCR of β-actin as an indication of RNA integrity. Abbreviations NCC: Normal corneal stromal cells, ACC: Aniridic corneal stromal cells, NL: Normal limbal stromal cells, KDM: keratocyte differentiation medium.

249

8.2 Choosing an appropriate endogenous control

RPL37A was chosen as the endogenous control to be used in qPCR experiments. Different housekeeping genes that have been commonly used as endogenous control were tested in a Taqman array. Those that had Ct values around 20 and remained constant between the three samples tested were selected (orange arrows in Figure A.8.2). From these RPL37A was chosen and compared to ACTB and 18S (routinely used as housekeeping genes in-house). From the three, RPL37A was shown to be the most reliable with less oscillation in Ct values across all the three donors for NL, NC and AC.

Figure A.8.2. Real-Time qPCR study to select an appropriate endogenous control. (A) Taqman array of 32 different genes commonly used as endogenous control. Only one sample per condition NL, NC and AC were used in the array. Orange arrows show the most appropriate genes, which have CT values around 20 that remain close between the three samples. (B) qPCR of the gene selected from the array, RPL37A and two other commonly used genes, B-actin and 18S. All three lines/donors from each condition were used in this qPCR. Abbreviations NCC: Normal corneal stromal cells, ACC: Aniridic corneal stromal cells, NL: Normal limbal stromal cells, EPI: epithelial cells, MSC: mesenchymal stem cells.

250

8.3 Immunocytochemistry controls

Cells were cultured with IgG for negative immunocytochemistry controls, Figure A.8.3. Additionally, NC were cultured as described by Funderburgh et al. [80] to induce their αSMA expression Figure A.8.4.

Figure A.8.3 - Negative controls for immunocytochemistry. Samples stained with IgG controls showed no staining in either central or limbal corneal stromal cells. Red: Anti-αSMA, Green: Phalloidin, Blue: DAPI. Scale bars= 300µm. Abbreviations NC: Normal corneal stromal cells, NL: Normal limbal stromal cells, NCK3F Normal corneal stromal keratocyte-like cells, NLK3F: Normal limbal stromal keratocyte-like cells, NCSD Normal corneal stroma spontaneously differentiated cells, NLSD Normal limbal stroma spontaneously differentiated cells.

251

Figure A.8.4. Positive controls for immunocytochemistry. NC were cultured in DMEM supplemented with 10% (V/V) FBS and 10ng/µL of TGFβ and showed positive staining of αSMA. Interestingly, not all cells assembled the αSMA stress fibres. Red: Anti-αSMA, Green: Phalloidin, Blue: DAPI.

252

8.4 PAX6 expression under different conditions

PAX6 expression was found very low across all samples in a result of the high Ct values detected. (Table A.8.1, Table A.8.2, Table A.8.3).

Table A.8.1- Ct values determined for PAX6 gene in different cell types cultured in CSSCM. Table shows the average of 3 technical replicates per donor. Abbreviations EC: epithelial cells, MSC: mesenchymal stem cells, NL: Normal limbal stromal cells, NC: Normal corneal stromal cells, AC: Aniridic corneal stromal cells.

Donor ECs MSC NL NC AC

1 22.082 27.887 35.518 35.185 >40

2 - - 35.065 34.620 36.437

3 - - 34.811 33.723 34.768

Table A.8.2- Ct values determined for PAX6 gene in different cell types under different culture conditions. Table shows the average of 3 technical replicates per donor. Abbreviations NL: Normal limbal stromal cells, NC: Normal corneal stromal cells, SD: spontaneous differentiation.

NC1 NC2 NC3 NC4 NL1 NL2 NL3 NL4

NC/NL 35.489 35.645 34.758 >40 36.276 36.502 35.606 36.962

KDM1F- 36.777 37.801 36.671 >40 36.442 35.031 >40 37.156

KDM1F+ 36.717 36.793 34.630 >40 37.145 35.545 37.522 36.997

KDM2F- 36.634 37.526 36.604 >40 36.791 34.704 37.624 36.900

KDM2F+ 37.528 37.972 34.604 >40 36.728 35.545 37.265 36.372

KDM3F- 36.834 >40 36.131 >40 37.084 33.845 37.552 >40

KDM3F+ 37.635 38.396 34.793 >40 36.256 35.080 37.445 36.337

SD 36.618 36.755 35.660 >40 >40 36.932 37.580 36.141

253

Table A.8.3 Ct values determined for PAX6 gene in different cell types under different culture conditions. Table shows the average of 3 technical replicates per donor. Abbreviations AC: Aniridic corneal stromal cells, NC: Normal corneal stromal cells, SD: spontaneous differentiation.

AC1 AC2 AC3 AC4 NC1 NC2 NC3 NC4

AC >40 36,42 36,90 36,14 NC 35,49 35,65 34,76 >40

ACK1F- >40 >40 >40 36,31 NCK1F- 36,78 37,80 36,67 >40

ACK1F+ >40 37,24 >40 36,16 NCK1F+ 36,72 36,79 34,63 >40

ACK2F- >40 >40 >40 35,94 NCK2F- 36,63 37,53 36,60 >40

ACK2F+ >40 36,62 >40 35,72 NCK2F+ 37,53 37,97 34,60 >40

ACK3F- >40 >40 >40 36,32 NCK3F- 36,83 >40 36,13 >40

ACK3F+ >40 >40 >40 35,39 NCK3F+ 37,64 38,39 34,79 >40

ACSD >40 37,04 >40 37,26 NCSD 36,62 36,75 35,66 >40

254

8.5 RNA quality, purity, and integrity of NC, NL and AC in 3D

RNA quality and integrity isolated from AC and NC populating RAFT TE after 3 weeks in culture was assessed. RNA isolated from all samples had a ratio ~2 and amplified β- actin successfully (Figure A.8.5).

Figure A.8.5. Quality and integrity of the RNA obtained from AC and NC populated RAFT TE. (A) Ration of 260/280 absorbance obtained with the Nanodrop. (B) Control RT-PCR for β- actin to measure RNA integrity. Abbreviations: NC: normal corneal stromal cells, AC: normal corneal stromal cells.

255

8.6 Transition into different absorbers for the manufacturing of RAFT TE

Unfortunately, the company selling the long absorbers to manufacture RAFT TE discontinued their sales. Instead, only shorter absorbers available. A set of experiments was conducted to assess which protocol would be suitable to obtain the most similar results (Figure A.8.6). It was concluded that the use of short absorbers on the same volume used (2,4mL) gave the most similar results to the original protocol.

*** A) B) 60 ***

Transparency % of PBS at 550nm ofat PBS% 0

2,4 mL long 2,4 mL short 1,2 mL short Type of RAFT

Figure A.8.6. Transition process from long absorbers (-plungers) to short absorbers. (A) Schematic representation of the initial protocol (left) and the two different options to replace it. (B) Transparency of the different RAFT TE prepared with different volumes and absorbers. (C) Images of cells inside RAFT TE prepared with different protocols.

256

8.7 RNA sequencing data for the 50 most significant DEGs

A list of the 50 most significant DEGs between AC and NC can be found in Table A.8.4.

Table A.8.4. List of the fifty most significant DEGs obtained from the comparison AC against NC (all donors pooled together). ID: Gene ID, log2FoldChange: Log2 fold change of the normalised mean hit counts, P-adj value: The Benjamini-Hochberg adjusted p-value. For each sample (AC1, AC2, AC3, NC1, NC2, NC3): The normalised hit counts for the gene.

ID log2F P-adj AC1 AC2 AC3 NC1 NC2 NC3 Gene.na oldCh value me ange ENSG00000151834 -2.20 9.45E-12 157.78 122.39 141.49 502.77 775.54 659.59 GABRA2 ENSG00000253304 -2.11 8.92E-10 77.45 56.13 83.66 292.62 359.50 282.55 TMEM200B ENSG00000133048 4.21 3.09E-09 55869.43 19206.14 12166.74 1801.80 1896.72 1003.23 CHI3L1 ENSG00000070193 -4.36 2.1E-07 3.82 1.84 7.38 117.93 72.54 66.82 FGF10 ENSG00000137975 -3.65 2.91E-07 104.23 177.60 414.62 4154.26 1745.24 2860.78 CLCA2 ENSG00000261115 -2.66 7.85E-07 20.08 18.40 29.53 120.59 189.89 115.50 TMEM178B ENSG00000178773 1.26 9.17E-07 2304.51 1877.27 2076.80 828.19 1011.30 777.00 CPNE7 ENSG00000163485 2.17 1.63E-06 578.52 479.44 323.58 112.61 73.61 120.27 ADORA1 ENSG00000002587 -4.80 1.65E-06 2.87 1.84 2.46 87.78 73.61 37.23 HS3ST1 ENSG00000104722 -3.00 8.5E-06 181.68 162.88 313.73 1425.84 2966.69 870.55 NEFM ENSG00000170153 -1.99 8.5E-06 146.30 242.94 259.60 917.75 1026.24 630.00 RNF150 ENSG00000209082 1.04 8.5E-06 2121.87 1687.70 2004.21 908.00 909.96 1010.86 MT-TL1 ENSG00000133055 4.11 9.67E-06 384.40 103.07 99.66 16.85 7.47 9.55 MYBPH ENSG00000139211 -2.75 1.02E-05 27.73 20.25 17.22 111.73 219.76 109.77 AMIGO2 ENSG00000159307 -4.06 1.35E-05 16.26 112.27 164.86 1900.23 2009.80 989.86 SCUBE1 ENSG00000277586 -2.30 2.31E-05 165.43 246.62 126.72 548.88 1193.72 906.82 NEFL ENSG00000164330 2.10 9.64E-05 699.96 684.65 306.35 117.05 129.08 149.86 EBF1 ENSG00000143995 2.05 9.92E-05 135.78 187.73 211.62 35.47 59.74 34.36 MEIS1 ENSG00000186281 -1.48 0.000156 85.10 111.35 78.74 300.60 224.02 245.32 GPAT2 ENSG00000187479 -2.88 0.000162 50.68 66.26 49.21 757.25 237.89 231.95 C11orf96 ENSG00000238113 -2.24 0.000211 22.95 15.64 35.68 125.91 93.88 126.95 LINC01410 ENSG00000231389 -2.21 0.000213 32.51 24.85 39.37 135.67 205.89 103.09 HLA-DPA1 ENSG00000279508 2.18 0.000229 183.60 197.85 285.44 27.49 61.87 58.23 NA ENSG00000167191 -3.55 0.000296 19.12 11.96 52.90 348.48 500.32 126.95 GPRC5B ENSG00000173114 -1.97 0.000327 105.18 131.59 93.51 348.48 646.46 303.55 LRRN3 ENSG00000225684 7.78 0.000327 55.46 157.36 14.76 0.00 0.00 0.95 FAM225B ENSG00000164116 3.56 0.000536 65.02 473.92 553.65 24.83 35.20 32.45 GUCY1A1 ENSG00000128513 -1.11 0.000551 695.18 703.98 680.37 1825.74 1441.21 1211.32 POT1 ENSG00000138944 -3.40 0.000551 35.38 16.56 87.35 688.09 195.22 575.59 SHISAL1 ENSG00000154864 -4.44 0.00059 5.74 5.52 2.46 101.97 19.20 179.45 PIEZO2 ENSG00000157168 -3.96 0.00059 43.99 162.88 365.41 1576.58 6052.87 1255.23 NRG1 ENSG00000170381 -4.19 0.000612 2.87 1.84 1.23 47.00 41.60 21.95 SEMA3E ENSG00000122378 -2.81 0.000635 48.77 55.21 148.87 902.67 550.45 315.95 FAM213A ENSG00000145911 -5.03 0.000784 9.56 10.12 35.68 300.60 1454.01 42.95 N4BP3 ENSG00000232815 -2.10 0.000784 12.43 15.64 14.76 59.41 52.27 71.59 DUX4L50

257

ENSG00000071282 2.02 0.000798 490.54 328.52 200.54 79.80 68.27 104.05 LMCD1 ENSG00000160191 3.12 0.000826 255.31 106.75 87.35 7.98 16.00 27.68 PDE9A ENSG00000110852 -1.58 0.00085 1351.15 1188.02 1060.54 3107.93 5157.85 2509.50 CLEC2B ENSG00000120708 -1.50 0.00117 25770.30 23876.31 38712.35 88491.40 101171.65 59434.83 TGFBI ENSG00000092969 -2.12 0.001209 255.31 538.33 773.88 3020.15 1748.44 2055.14 TGFB2 ENSG00000152270 -2.99 0.001295 133.87 410.42 923.98 5310.53 3624.89 2693.73 PDE3B ENSG00000088882 -3.00 0.001561 21.04 122.39 70.13 846.81 547.25 316.91 CPXM1 ENSG00000181652 -4.92 0.001627 6.69 11.96 13.53 141.87 810.75 19.09 ATG9B ENSG00000106688 1.26 0.001886 3263.60 2609.77 1956.23 1076.47 1265.19 927.82 SLC1A1 ENSG00000149968 -2.55 0.001894 2468.98 2312.54 3207.47 14125.35 27544.07 5162.19 MMP3 ENSG00000064692 -1.97 0.00191 25.82 28.53 51.67 142.76 102.41 165.14 SNCAIP ENSG00000110427 2.62 0.002479 227.58 260.43 188.24 23.94 70.41 16.23 KIAA1549L ENSG00000074317 -6.77 0.00305 3.82 0.92 1.23 29.26 625.13 2.86 SNCB ENSG00000259207 -4.13 0.003352 21.04 18.40 12.30 145.42 726.47 37.23 ITGB3

A list of the 50 most significant DEGs between AC and NC can be found in Table A.8.5.

Table A.8.5. List of the fifty most significant DEGs obtained from the comparison AC1 against NC. ID: Gene ID, log2FoldChange: Log2 fold change of the normalised mean hit counts, P-adj value: The Benjamini-Hochberg adjusted p-value. For each sample (AC1, NC1, NC2, NC3): The normalised hit counts for the gene.

ID log2FoldChange P-adj value AC1 NC1 NC2 NC3 Gene.name ENSG00000138207 5.36 7.85E-45 6124.39 151.55 133.28 162.82 RBP4 ENSG00000228560 8.63 3.48E-34 1827.94 5.51 2.20 5.92 AL513323.1 ENSG00000164120 6.43 6.98E-31 4926.16 78.99 45.16 46.38 HPGD ENSG00000182870 4.87 1.84E-28 3965.80 134.09 159.71 113.48 GALNT9 ENSG00000164761 -6.55 5.83E-26 304.99 35111.64 34065.36 16821.93 TNFRSF11B ENSG00000173376 -3.80 2.74E-23 2735.00 34867.33 43760.55 35297.73 NDNF ENSG00000158859 3.39 6.85E-21 4071.41 388.51 408.65 370.05 ADAMTS4 ENSG00000159307 -6.65 8.82E-20 16.78 1968.25 2075.18 1023.31 SCUBE1 ENSG00000240342 -5.07 1.08E-19 312.88 7080.37 13145.02 11273.17 RPS2P5 ENSG00000215030 3.79 2.12E-19 3928.29 276.46 327.14 246.70 RPL13P12 ENSG00000133048 5.15 2.36E-19 57667.95 1866.30 1958.42 1037.12 CHI3L1 ENSG00000282826 -7.00 2.36E-19 7.90 1164.60 1170.87 706.55 FRG1CP ENSG00000120659 -5.79 2.87E-17 17.77 1036.94 1239.16 673.98 TNFSF11 ENSG00000154175 5.33 3.52E-17 2274.07 48.68 39.65 80.92 ABI3BP ENSG00000237973 3.87 3.52E-17 13918.79 1108.58 743.50 1011.47 MTCO1P12 ENSG00000198805 3.27 1.10E-16 11637.82 1340.03 1236.96 1040.08 PNP ENSG00000188856 -6.67 2.21E-16 10.86 1267.47 1398.87 650.30 RPSAP47 ENSG00000047648 -5.16 2.39E-16 71.06 3595.75 1830.65 2219.31 ARHGAP6 ENSG00000115461 -6.05 4.35E-16 3751.62 252842.55 365836.65 125397.22 IGFBP5 ENSG00000152270 -4.86 7.71E-16 138.18 5500.63 3742.82 2784.74 PDE3B ENSG00000050555 -3.27 7.91E-15 114.49 1158.17 989.13 1175.28 LAMC3 ENSG00000188910 7.26 3.27E-14 405.66 2.76 2.20 2.96 GJB3 258

ENSG00000121068 -2.53 7.85E-14 1914.80 10520.91 11432.22 11199.16 TBX2 ENSG00000088826 -3.25 1.17E-13 400.73 3763.83 4478.60 3224.85 SMOX ENSG00000154856 -5.42 1.26E-13 421.45 23930.36 9148.86 21223.04 APCDD1 ENSG00000205978 -4.11 1.30E-13 34.55 719.15 525.40 543.73 NYNRIN ENSG00000149451 -3.53 2.24E-13 116.47 1347.37 1106.98 1578.88 ADAM33 ENSG00000154678 -4.43 6.16E-13 36.52 1000.20 587.09 769.70 PDE1C ENSG00000165912 -3.47 1.72E-12 77.97 925.80 957.18 713.45 PACSIN3 ENSG00000225670 6.11 1.72E-12 482.65 4.59 5.51 10.85 CADM3-AS1 ENSG00000116667 -2.64 2.11E-12 503.37 3321.14 2857.23 3264.32 C1orf21 ENSG00000104321 -3.37 3.14E-12 2340.20 18643.73 24733.65 29180.57 TRPA1 ENSG00000137975 -4.81 7.32E-12 107.58 4302.96 1802.02 2957.43 CLCA2 ENSG00000185633 -4.11 4.02E-11 1085.71 18021.02 25914.43 12446.47 NDUFA4L2 ENSG00000172935 -4.00 4.68E-11 178.65 2556.06 3924.56 2082.14 MRGPRF ENSG00000213088 6.30 6.28E-11 2613.60 22.04 15.42 62.17 ACKR1 ENSG00000268388 -4.02 7.56E-11 361.25 8196.30 5180.24 4173.16 FENDRR ENSG00000170891 -3.24 1.07E-10 89.82 1010.30 798.57 736.15 CYTL1 ENSG00000146966 -7.51 1.62E-10 15.79 2153.78 5548.14 904.89 DENND2A ENSG00000196872 4.90 1.88E-10 327.69 10.10 12.12 10.85 KIAA1211L ENSG00000149531 -5.77 5.13E-10 5.92 412.39 323.83 231.90 FRG1BP ENSG00000142973 10.12 1.13E-09 397.76 0.00 0.00 0.99 CYP4B1 ENSG00000137463 -2.15 1.67E-09 864.62 3688.52 3881.60 3945.21 MGARP ENSG00000077942 -3.00 1.81E-09 11983.27 85361.26 80847.26 122388.48 FBLN1 ENSG00000160862 7.08 2.17E-09 272.41 1.84 3.30 0.99 AZGP1 ENSG00000171517 -5.53 2.62E-09 14.81 987.34 340.36 724.31 LPAR3 ENSG00000169583 4.71 5.38E-09 302.02 13.78 9.91 10.85 CLIC3 ENSG00000244586 -2.80 5.83E-09 201.35 1523.72 1548.68 1142.71 WNT5A-AS1 ENSG00000133055 5.08 7.92E-09 396.78 17.45 7.71 9.87 MYBPH

A list of the 50 most significant DEGs between AC and NC can be found in Table A.8.6.

Table A.8.6. List of the fifty most significant DEGs obtained from the comparison AC1 against NC. ID: Gene ID, log2FoldChange: Log2 fold change of the normalised mean hit counts, P-adj value: The Benjamini-Hochberg adjusted p-value. For each sample (AC1, NC1, NC2, NC3): The normalised hit counts for the gene.

ID log2Fold P-adj value AC2 AC3 NC1 NC2 NC3 Gene.name Change ENSG00000164116 4.07 1.10E-32 471.62 549.34 24.54 34.77 32.16 GUCY1A1 ENSG00000159307 -3.55 3.61E-12 111.72 163.58 1878.34 1984.87 980.86 SCUBE1 ENSG00000133048 3.33 3.94E-12 19113.10 12072.09 1781.05 1873.19 994.10 CHI3L1 ENSG00000151834 -2.29 1.14E-09 121.80 140.39 496.98 765.92 653.59 GABRA2 ENSG00000253304 -2.16 2.94E-07 55.86 83.01 289.25 355.04 279.97 TMEM200B ENSG00000137975 -3.30 3.58E-06 176.74 411.40 4106.41 1723.59 2834.74 CLCA2 ENSG00000198535 -2.22 1.89E-05 65.94 112.31 359.37 509.91 372.67 C2CD4A ENSG00000260244 4.15 1.89E-05 87.00 108.65 1.75 7.37 7.57 AC104083.1

259

ENSG00000133055 3.16 3.18E-05 102.57 98.88 16.65 7.37 9.46 MYBPH ENSG00000170153 -1.76 4.51E-05 241.76 257.58 907.18 1013.50 624.27 RNF150 ENSG00000070193 -4.28 5.58E-05 1.83 7.32 116.57 71.64 66.21 FGF10 ENSG00000139211 -2.96 0.000122 20.15 17.09 110.44 217.03 108.77 AMIGO2 ENSG00000143995 2.22 0.000122 186.82 209.97 35.06 59.00 34.05 MEIS1 ENSG00000169432 -1.37 0.000122 502.76 726.35 1533.00 1541.33 1693.09 SCN9A ENSG00000178773 1.19 0.000122 1868.18 2060.64 818.65 998.75 769.93 CPNE7 ENSG00000261115 -2.57 0.000178 18.32 29.30 119.20 187.53 114.45 TMEM178B ENSG00000002587 -4.96 0.000204 1.83 2.44 86.77 72.69 36.89 HS3ST1 ENSG00000119630 -2.07 0.000209 361.73 390.64 1084.23 1413.85 2249.25 PGF ENSG00000152270 -2.53 0.000301 408.43 916.79 5249.36 3579.92 2669.22 PDE3B ENSG00000132854 -3.61 0.000316 80.59 59.82 1585.59 685.85 310.24 KANK4 ENSG00000163485 1.98 0.000316 477.12 321.06 111.32 72.69 119.18 ADORA1 ENSG00000167851 -22.31 0.000316 0.00 0.00 0.88 262.33 0.00 CD300A ENSG00000163703 -1.11 0.000499 1076.95 1173.15 2706.63 2050.18 2547.20 CRELD1 ENSG00000134247 1.97 0.000514 3279.38 6495.66 1001.84 1317.98 1422.57 PTGFRN ENSG00000197380 -1.33 0.000548 144.69 148.93 339.21 436.16 332.00 DACT3 ENSG00000092969 -1.79 0.000683 535.73 767.86 2985.36 1726.75 2036.44 TGFB2 ENSG00000050555 -1.49 0.000947 289.38 466.33 1105.27 946.08 1126.52 LAMC3 ENSG00000226210 -1.23 0.001147 140.11 146.49 382.15 326.60 296.05 AC215219.1 ENSG00000146151 -3.84 0.001377 54.03 14.65 931.72 291.83 264.84 HMGCLL1 ENSG00000088882 -2.56 0.0015 121.80 69.58 837.06 540.46 314.03 CPXM1 ENSG00000185565 -2.40 0.001586 43.04 72.02 353.23 174.89 376.45 LSAMP ENSG00000104722 -2.87 0.00172 162.09 311.29 1409.41 2929.89 862.62 NEFM ENSG00000279508 2.31 0.002505 196.89 283.22 27.17 61.11 57.70 NA ENSG00000225684 7.97 0.003089 156.60 14.65 0.00 0.00 0.95 FAM225B ENSG00000277586 -2.23 0.003184 245.43 125.74 542.55 1178.91 898.57 NEFL ENSG00000157168 -3.48 0.004024 162.09 362.57 1558.42 5977.77 1243.81 NRG1 ENSG00000162496 -1.85 0.004675 76.01 40.29 182.31 248.63 202.41 DHRS3 ENSG00000254122 -1.13 0.00548 318.69 379.66 894.91 724.83 664.94 PCDHGB7 ENSG00000164418 -2.39 0.00612 18.32 12.21 55.22 99.03 88.91 GRIK2 ENSG00000158747 -1.11 0.006664 3779.40 3683.03 9792.27 6582.50 7753.21 NBL1 ENSG00000231389 -2.21 0.006664 24.73 39.06 134.10 203.33 102.15 HLA-DPA1 ENSG00000186281 -1.42 0.006838 110.81 78.13 297.13 221.24 243.09 GPAT2 ENSG00000110852 -1.67 0.008191 1182.26 1052.29 3072.14 5093.86 2486.66 CLEC2B ENSG00000170381 -4.55 0.00825 1.83 1.22 46.45 41.09 21.75 SEMA3E ENSG00000235863 -1.38 0.00825 84.25 87.89 242.79 254.96 174.98 B3GALT4 ENSG00000238113 -2.19 0.00825 15.57 35.40 124.46 92.71 125.80 LINC01410 ENSG00000164330 1.92 0.008287 681.34 303.97 115.70 127.48 148.50 EBF1 ENSG00000154864 -4.62 0.008968 5.49 2.44 100.80 18.96 177.82 PIEZO2 ENSG00000187479 -2.82 0.008968 65.94 48.83 748.53 234.94 229.84 C11orf96

260

8.8 Rscripts used for plotting the RNA sequencing data

Volcano plot

## Load libraries for plotting library(ggplot2) library(ggrepel)

## Changes the the directory setwd('/Users/victoriatovell/Documents/PhD/RNAseq/Normal vs diseased/deseq2')

## Read data from excel into a matrix like format with the same header names as in the file data <- read.table("Differential_expression_analysis_table.txt",header = T)

## Removes NA values from the adjusted p-value and creates a new variable ## is.na is a function that returns true if the value is NA ## in this case we want to keep the values that are different from NA, e hence !is.na data_cleaned <- data[!is.na (data$padj),]

## Creates a data frame with our variables of interest ## a data.frame is required for ggplot FINAL <- data.frame(data_cleaned$Name, data_cleaned$log2FoldChange, data_cleaned$padj)

## Change the names of the columns otherwise the column names would be data_cleaned.Name, ## data_cleaned$log2FoldChange and data_cleaned.padj, respectively colnames(FINAL) <- c("Name", "log2FoldChange", "padj")

## function to get a list of colors for the points in the graph ## if the adjusted p-value is higher than 0.05 the color is black. If it is lower than 0.05 ## and the genes are upregulated is blue, if the genes are downregulated is orange. clr <- ifelse(FINAL$padj>0.05, "black", ifelse(FINAL$log2FoldChange>0, "dodgerblue3","orangered2"))

## Select the subset to annotate with names selected <- FINAL[-log10(FINAL$padj)>5,]

## Creates a png file with this plot ## In this case we are considering a width=1200px, height=1200px and a resolution of 300 dpi setwd("/Users/victoriatovell/Documents/PhD/RNAseq/Figures/Figures_thesis") png(file="Volcano1.png",width=1200,height=1200)

## plotting- Calls the ggplot function ggplot() + ## Tells ggplot that we what to plot points where the data is in the variable FINAL ## and the x value is log2FolChange of FINAL and the y value is the -log10 of the ## adjusted p-value of the FINAL. We also give a vector (the clr variable set above) of the different colors for these points geom_point(data= FINAL, mapping= aes(log2FoldChange, -log10(padj)), colour=clr, size=3) + ## limit the x axis from -10 to 10. Needed to be changed accordingly xlim(c(-10,10)) + ## plot the names of the selected genes and repels (more or less) names ## that fall above each other. it used the variable "selected" ## and used the log2FoldChange and -log10 of the adjusted p-value as the exact ## location for plotting the name of the genes geom_text_repel(data= selected, mapping= aes(log2FoldChange, -log10(padj), label = Name), size=10) + 261

## Changes the name of the x label xlab("Log2 Fold Change") + ## Changes the Name of the y label ylab("-Log10(adjusted pvalue)") + ## sets the background white and removes gridlines theme_classic(base_size = 20) + theme (axis.title.x = element_text(size= 40), axis.title.y = element_text(size= 40), axis.text.x = element_text(face="bold", size=40), axis.text.y = element_text(face="bold", size=40)) ## Tells R that we are done with this plot dev.off()

Venn Diagram

# install.packages('VennDiagram') library(VennDiagram)

## lty - outline of cirlces ## fill - colour ## alpha - colour transparency ## cat.pos - position of category titles, by degree from the middle of the circle ## cat.dist - distance of the category titles from the edge of the circle grid.newpage() venn.plot <- draw.pairwise.venn(area1 = 42, area2 = 111, cross.area = 0, category = c("Upregulated","Downregulated"), lty = rep("blank", 2), fill = c("dodgerblue3","orangered2"), alpha = rep(0.7, 2), fontface= c(2,2,2),fontfamily= c('sans','sans','sans'), cex= c(4,1.5,4), cat.pos = c(0, 180), cat.cex= c(3,3), cat.fontfamily= c('sans','sans'), cat.dist= c(0.05,0.05),euler.d = TRUE ,sep.dist = 0.03, rotation.degree = 20, cex.prop = TRUE) # Writing to file tiff(filename = "Pairwise_Venn_diagram.tiff", compression = "lzw"); grid.draw(venn.plot); dev.off(); Heat Map library(gplots) library(dplyr) setwd("/Users/victoriatovell/Documents/PhD/RNAseq/Normal vs diseased/deseq2")

# Reads the file from excel, selects those genes with a p-adj<0,05, orders them from most to least significant, removes the genes that are not annotated DEGs_AN <- read.csv("Significant-DEGs.csv") DEGs_AN_significant <- subset(DEGs_AN, DEGs_AN[,4] < 0.05) DEGs_AN_significant_sorted <- DEGs_AN_significant[order(DEGs_AN_significant[,4]),] DEGs_AN_significant_sorted_rows <- DEGs_AN_significant_sorted[c(1:22,24:31),] DEGs_AN_significant_columns <- DEGs_AN_significant_sorted_rows[,5:10]

# Creates a matrix DEGs_AN_significant_matrix <- as.matrix(DEGs_AN_significant_columns)

# Creates a vector with the colors for the heatmap hmcols<-colorRampPalette(c("orangered3",'orangered1','white','dodgerblue1','dodgerblue3'))(40)

262

par(oma=c(1,1,1,1)) setwd("/Users/victoriatovell/Documents/PhD/RNAseq/Figures/Figures_thesis")

# Writing to file - plot png(file="Heatmap_30_A.png",width=1200,height=1004) heatmap.2(DEGs_AN_significant_matrix, lhei=c(3, 10), dendrogram = 'both',scale = 'row' , labCol= c('AC1','AC2','AC3','NC1','NC2','NC3'), cexCol = 4, cexRow= 2, labRow = DEGs_AN_significant_sorted_rows$Gene.name, ,col= hmcols, trace = 'none', srtCol=45, margins= c(8,14), key.title ='NA',keysize = 1, key.par = list(cex=1.5), density.info = 'none') dev.off()

Gene ontology bar graph library(gplots) library(dplyr) library(ggplot2) library(scales) setwd("/Users/victoriatovell/Documents/PhD/RNAseq/Normal vs diseased/GO")

# Reads from the excel file, sorts the GO from low to high p-value, chooses the 30 most significant GO GO_AN <- read.csv("GO_analysis_log10.csv") GO_AN_sorted <- GO_AN[order(GO_AN[,7]),] GO_AN_30 <- GO_AN_sorted[1:30,]

# Saves into a png - plot setwd("/Users/victoriatovell/Documents/PhD/RNAseq/Figures/Figures_thesis") png(file="GO_AN.png",width=1200,height=1004) ggplot(data=GO_AN_30, aes(x=reorder(Process_name, LOG10Padj),y=LOG10Padj)) + geom_bar(stat="identity", width = 0.5, colour= 'brown3', fill='brown3') + theme_bw() + theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank(), panel.background = element_rect(colour = "black", size=1), axis.title.y=element_blank(), axis.text.y = element_text(colour = c('black','black','black','black','black','black','black','black','black','black','brown3','black','black','black','black', 'black','black','brown3','black','black','black','black','black','black','black','brown3','black','black','black','brown 3'), face = c('plain','plain','plain', 'plain', 'plain','plain','plain', 'plain','plain','plain', 'bold','plain','plain', 'plain','plain','plain', 'plain','bold','plain', 'plain','plain','plain', 'plain','plain','plain', 'bold','plain','plain', 'plain','bold')), text=element_text(size=26), axis.line = element_line(colour = "black")) + coord_flip() + labs(y ="-Log10(Padj)") dev.off()

263

264

REFERENCES

1. Hu, D.N., Prevalence and mode of inheritance of major genetic eye diseases in China. J Med Genet, 1987. 24(10): p. 584-8. 2. Eden, U., et al., Epidemiology of aniridia in Sweden and Norway. Acta Ophthalmol, 2008. 86(7): p. 727-9. 3. Nishida, K., et al., Ocular surface abnormalities in aniridia. Am J Ophthalmol, 1995. 120(3): p. 368-75. 4. Chen, J.J. and S.C. Tseng, Abnormal corneal epithelial wound healing in partial-thickness removal of limbal epithelium. Invest Ophthalmol Vis Sci, 1991. 32(8): p. 2219-33. 5. Ahmad, S., F. Figueiredo, and M. Lako, Corneal epithelial stem cells: characterization, culture and transplantation. Regen Med, 2006. 1(1): p. 29-44. 6. Kolli, S., et al., Successful Application of Ex Vivo Expanded Human Autologous Oral Mucosal Epithelium for the Treatment of Total Bilateral Limbal Stem Cell Deficiency. STEM CELLS, 2014. 32(8): p. 2135-2146. 7. Forrester, J.V., et al, The eye, basic sciences in practice. 3rd edition ed, ed. G.F. Saunders. Vol. 92. 2002: Clin. Exp. Optom. 8. Fini, M.E., Keratocyte and fibroblast phenotypes in the repairing cornea. Prog Retin Eye Res, 1999. 18(4): p. 529-51. 9. McDermott, A.M., The role of antimicrobial peptides at the ocular surface. Ophthalmic Res, 2009. 41(2): p. 60-75. 10. Ban, Y., et al., Tight junction-related protein expression and distribution in human corneal epithelium. Exp Eye Res, 2003. 76(6): p. 663-9. 11. Tao, A., et al., Topographic thickness of Bowman's layer determined by ultra-high resolution spectral domain-optical coherence tomography. Invest Ophthalmol Vis Sci, 2011. 52(6): p. 3901-7. 12. Beebe, D.C. and B.R. Masters, Cell lineage and the differentiation of corneal epithelial cells. Invest Ophthalmol Vis Sci, 1996. 37(9): p. 1815-25. 13. Graves, B., Certain Clinical Features of the Normal Limbus. Br J Ophthalmol, 1934. 18(7): p. 369-87. 14. Shortt, A.J., et al., Characterization of the limbal epithelial stem cell niche: novel imaging techniques permit in vivo observation and targeted biopsy of limbal epithelial stem cells. Stem Cells, 2007. 25(6): p. 1402-9. 15. Dziasko, M.A., S.J. Tuft, and J.T. Daniels, Limbal melanocytes support limbal epithelial stem cells in 2D and 3D microenvironments. Exp Eye Res, 2015. 138: p. 70-9. 16. Pellegrini, G., et al., Location and clonal analysis of stem cells and their differentiated progeny in the human ocular surface. J Cell Biol, 1999. 145(4): p. 769-82. 17. Bourne, W.M., Biology of the corneal endothelium in health and disease. Eye (Lond), 2003. 17(8): p. 912-8. 18. Reinstein, D.Z., et al., Stromal thickness in the normal cornea: three-dimensional display with artemis very high-frequency digital ultrasound. J Refract Surg, 2009. 25(9): p. 776- 86. 19. Davis, J.A. and R.R. Reed, Role of Olf-1 and Pax-6 Transcription Factors in Neurodevelopment. The Journal of Neuroscience, 1996. 16(16): p. 5082-5094. 20. Koroma, B.M., J.M. Yang, and O.H. Sundin, The Pax-6 homeobox gene is expressed throughout the corneal and conjunctival epithelia. Invest Ophthalmol Vis Sci, 1997. 38(1): p. 108-20. 21. Trelstad, R.L. and A.J. Coulombre, Morphogenesis of the collagenous stroma in the chick cornea. J Cell Biol, 1971. 50(3): p. 840-58. 22. Hay, E.D., Development of the Vertebrate Cornea, in International Review of Cytology, G.H. Bourne, J.F. Danielli, and K.W. Jeon, Editors. 1980, Academic Press. p. 263-322. 265

23. Cintron, C., H. Covington, and C.L. Kublin, Morphogenesis of rabbit corneal stroma. Invest Ophthalmol Vis Sci, 1983. 24(5): p. 543-56. 24. Song, J., et al., Neonatal corneal stromal development in the normal and lumican- deficient mouse. Invest Ophthalmol Vis Sci, 2003. 44(2): p. 548-57. 25. Funderburgh, J.L., B. Caterson, and G.W. Conrad, Keratan sulfate proteoglycan during embryonic development of the chicken cornea. Dev Biol, 1986. 116(2): p. 267-77. 26. Poole, C.A., N.H. Brookes, and G.M. Clover, Keratocyte networks visualised in the living cornea using vital dyes. J Cell Sci, 1993. 106 ( Pt 2): p. 685-91. 27. Young, R.D., et al., Three-dimensional aspects of matrix assembly by cells in the developing cornea. Proceedings of the National Academy of Sciences, 2014. 111(2): p. 687. 28. Watsky, M.A., Keratocyte gap junctional communication in normal and wounded rabbit corneas and human corneas. Invest Ophthalmol Vis Sci, 1995. 36(13): p. 2568-76. 29. Birk, D.E., et al., Collagen fibrillogenesis in vitro: interaction of types I and V collagen regulates fibril diameter. J Cell Sci, 1990. 95 ( Pt 4): p. 649-57. 30. Wenstrup, R.J., et al., Murine model of the Ehlers-Danlos syndrome. col5a1 haploinsufficiency disrupts collagen fibril assembly at multiple stages. J Biol Chem, 2006. 281(18): p. 12888-95. 31. Segev, F., et al., Structural abnormalities of the cornea and lid resulting from collagen V mutations. Invest Ophthalmol Vis Sci, 2006. 47(2): p. 565-73. 32. Hay, E.D., Development of the vertebrate cornea. Int Rev Cytol, 1980. 63: p. 263-322. 33. Benedek, G.B., Theory of transparency of the eye. Appl Opt, 1971. 10(3): p. 459-73. 34. Quantock, A.J. and R.D. Young, Development of the corneal stroma, and the collagen– proteoglycan associations that help define its structure and function. Developmental Dynamics, 2008. 237(10): p. 2607-2621. 35. Beales, M.P., et al., Proteoglycan synthesis by bovine keratocytes and corneal fibroblasts: maintenance of the keratocyte phenotype in culture. Invest Ophthalmol Vis Sci, 1999. 40(8): p. 1658-63. 36. Rada, J.A., P.K. Cornuet, and J.R. Hassell, Regulation of corneal collagen fibrillogenesis in vitro by corneal proteoglycan (lumican and decorin) core proteins. Exp Eye Res, 1993. 56(6): p. 635-48. 37. Blochberger, T.C., et al., cDNA to chick lumican (corneal keratan sulfate proteoglycan) reveals homology to the small interstitial proteoglycan gene family and expression in muscle and intestine. J Biol Chem, 1992. 267(1): p. 347-52. 38. Funderburgh, J.L., et al., Unique glycosylation of three keratan sulfate proteoglycan isoforms. J Biol Chem, 1991. 266(22): p. 14226-31. 39. Kao, W.W. and C.Y. Liu, Roles of lumican and keratocan on corneal transparency. Glycoconj J, 2002. 19(4-5): p. 275-85. 40. Ying, S., et al., Characterization and expression of the mouse lumican gene. J Biol Chem, 1997. 272(48): p. 30306-13. 41. Funderburgh, J.L., et al., Sequence, molecular properties, and chromosomal mapping of mouse lumican. Invest Ophthalmol Vis Sci, 1995. 36(11): p. 2296-303. 42. Chakravarti, S., et al., Corneal opacity in lumican-null mice: defects in collagen fibril structure and packing in the posterior stroma. Invest Ophthalmol Vis Sci, 2000. 41(11): p. 3365-73. 43. Liu, C.Y., et al., The cloning of mouse keratocan cDNA and genomic DNA and the characterization of its expression during eye development. J Biol Chem, 1998. 273(35): p. 22584-8. 44. Pellegata, N.S., et al., Mutations in KERA, encoding keratocan, cause cornea plana. Nat Genet, 2000. 25(1): p. 91-5.

266

45. Liu, C.Y., et al., Keratocan-deficient mice display alterations in corneal structure. J Biol Chem, 2003. 278(24): p. 21672-7. 46. Birk, D.E. and M.A. Lande, Corneal and scleral collagen fiber formation in vitro. Biochim Biophys Acta, 1981. 670(3): p. 362-9. 47. Zhang, G., et al., Genetic evidence for the coordinated regulation of collagen fibrillogenesis in the cornea by decorin and biglycan. J Biol Chem, 2009. 284(13): p. 8888- 97. 48. Maurice, D.M., The structure and transparency of the cornea. J Physiol, 1957. 136(2): p. 263-86. 49. Jester, J.V., et al., The cellular basis of corneal transparency: evidence for 'corneal crystallins'. J Cell Sci, 1999. 112 ( Pt 5): p. 613-22. 50. Meek, K.M. and C. Knupp, Corneal structure and transparency. Prog Retin Eye Res, 2015. 49: p. 1-16. 51. Muller, L.J., L. Pels, and G.F. Vrensen, Novel aspects of the ultrastructural organization of human corneal keratocytes. Invest Ophthalmol Vis Sci, 1995. 36(13): p. 2557-67. 52. Zieske, J.D., S.R. Guimaraes, and A.E. Hutcheon, Kinetics of keratocyte proliferation in response to epithelial debridement. Exp Eye Res, 2001. 72(1): p. 33-9. 53. Wilson, S.E. and W.J. Kim, Keratocyte apoptosis: implications on corneal wound healing, tissue organization, and disease. Invest Ophthalmol Vis Sci, 1998. 39(2): p. 220-6. 54. Du, Y., et al., Multipotent stem cells in human corneal stroma. Stem Cells, 2005. 23(9): p. 1266-75. 55. Kureshi, A.K., et al., Human corneal stromal stem cells support limbal epithelial cells cultured on RAFT tissue equivalents. Sci Rep, 2015. 5: p. 16186. 56. Kisselbach, L., et al., CD90 Expression on human primary cells and elimination of contaminating fibroblasts from cell cultures. Cytotechnology, 2009. 59(1): p. 31-44. 57. Funderburgh, M.L., et al., PAX6 expression identifies progenitor cells for corneal keratocytes. Faseb j, 2005. 19(10): p. 1371-3. 58. Yoshida, S., et al., Isolation of multipotent neural crest-derived stem cells from the adult mouse cornea. Stem Cells, 2006. 24(12): p. 2714-22. 59. Pinnamaneni, N. and J.L. Funderburgh, Concise review: Stem cells in the corneal stroma. Stem Cells, 2012. 30(6): p. 1059-63. 60. Carlson, E.C., et al., Altered KSPG expression by keratocytes following corneal injury. Mol Vis, 2003. 9: p. 615-23. 61. Mayer, K.L., et al., Keratopathy in congenital aniridia. Ocul Surf, 2003. 1(2): p. 74-9. 62. Zieske, J.D., Extracellular matrix and wound healing. Curr Opin Ophthalmol, 2001. 12(4): p. 237-41. 63. Nakayasu, K., Stromal changes following removal of epithelium in rat cornea. Jpn J Ophthalmol, 1988. 32(2): p. 113-25. 64. Wilson, S.E., et al., Epithelial injury induces keratocyte apoptosis: hypothesized role for the interleukin-1 system in the modulation of corneal tissue organization and wound healing. Exp Eye Res, 1996. 62(4): p. 325-7. 65. Wilson, S.E., Role of apoptosis in wound healing in the cornea. Cornea, 2000. 19(3 Suppl): p. S7-12. 66. Mohan, R.R., et al., Apoptosis in the cornea: further characterization of Fas/Fas ligand system. Exp Eye Res, 1997. 65(4): p. 575-89. 67. Kim, W.J., et al., Keratocyte apoptosis associated with keratoconus. Exp Eye Res, 1999. 69(5): p. 475-81. 68. Fini, M.E. and B.M. Stramer, How the cornea heals: cornea-specific repair mechanisms affecting surgical outcomes. Cornea, 2005. 24(8 Suppl): p. S2-s11.

267

69. Kim, W.J., et al., Effect of PDGF, IL-1alpha, and BMP2/4 on corneal fibroblast chemotaxis: expression of the platelet-derived growth factor system in the cornea. Invest Ophthalmol Vis Sci, 1999. 40(7): p. 1364-72. 70. Spandau, U.H., et al., High expression of chemokines Gro-alpha (CXCL-1), IL-8 (CXCL-8), and MCP-1 (CCL-2) in inflamed human corneas in vivo. Arch Ophthalmol, 2003. 121(6): p. 825-31. 71. Saika, S., et al., Loss of tumor necrosis factor alpha potentiates transforming growth factor beta-mediated pathogenic tissue response during wound healing. Am J Pathol, 2006. 168(6): p. 1848-60. 72. Caley, M.P., V.L. Martins, and E.A. O'Toole, Metalloproteinases and Wound Healing. Adv Wound Care (New Rochelle), 2015. 4(4): p. 225-234. 73. Mulholland, B., S.J. Tuft, and P.T. Khaw, Matrix metalloproteinase distribution during early corneal wound healing. Eye (Lond), 2005. 19(5): p. 584-8. 74. Van Wart, H.E. and H. Birkedal-Hansen, The cysteine switch: a principle of regulation of metalloproteinase activity with potential applicability to the entire matrix metalloproteinase gene family. Proc Natl Acad Sci U S A, 1990. 87(14): p. 5578-82. 75. Petroll, W.M., et al., Corneal Fibroblast Migration Patterns During Intrastromal Wound Healing Correlate With ECM Structure and Alignment. Invest Ophthalmol Vis Sci, 2015. 56(12): p. 7352-61. 76. Matsubara, M., et al., Differential roles for two gelatinolytic enzymes of the matrix metalloproteinase family in the remodelling cornea. Dev Biol, 1991. 147(2): p. 425-39. 77. Daniels, J.T., et al., Matrix metalloproteinase inhibition modulates fibroblast-mediated matrix contraction and collagen production in vitro. Invest Ophthalmol Vis Sci, 2003. 44(3): p. 1104-10. 78. Kowluru, R.A., Q. Zhong, and J.M. Santos, Matrix metalloproteinases in diabetic retinopathy: potential role of MMP-9. Expert Opin Investig Drugs, 2012. 21(6): p. 797- 805. 79. Jester, J.V., et al., Induction of alpha-smooth muscle actin expression and myofibroblast transformation in cultured corneal keratocytes. Cornea, 1996. 15(5): p. 505-16. 80. Funderburgh, J.L., et al., Proteoglycan expression during transforming growth factor beta -induced keratocyte-myofibroblast transdifferentiation. J Biol Chem, 2001. 276(47): p. 44173-8. 81. Scharenberg, M.A., et al., TGF-beta-induced differentiation into myofibroblasts involves specific regulation of two MKL1 isoforms. J Cell Sci, 2014. 127(Pt 5): p. 1079-91. 82. Singh, V., et al., Transforming growth factor β and platelet-derived growth factor modulation of myofibroblast development from corneal fibroblasts in vitro. Exp Eye Res, 2014. 120: p. 152-60. 83. Gardner, S.J., et al., Measuring the Refractive Index of Bovine Corneal Stromal Cells Using Quantitative Phase Imaging. Biophys J, 2015. 109(8): p. 1592-9. 84. Kaur, H., et al., Corneal myofibroblast viability: opposing effects of IL-1 and TGF beta1. Exp Eye Res, 2009. 89(2): p. 152-8. 85. Wilson, S.E. and A. Esposito, Focus on molecules: interleukin-1: a master regulator of the corneal response to injury. Exp Eye Res, 2009. 89(2): p. 124-5. 86. Hogan, B.L., et al., Small eyes (Sey): a homozygous lethal mutation on chromosome 2 which affects the differentiation of both lens and nasal placodes in the mouse. J Embryol Exp Morphol, 1986. 97: p. 95-110. 87. Grindley, J.C., D.R. Davidson, and R.E. Hill, The role of Pax-6 in eye and nasal development. Development, 1995. 121(5): p. 1433-42. 88. Sadagopan, K.A., et al., Anirdia-like phenotype caused by 6p25 dosage aberrations. Am J Med Genet A, 2015. 167a(3): p. 524-8.

268

89. Alzuhairy, S., et al., A novel CYP1B1 mutation with congenital glaucoma and total aniridia. Ophthalmic Genet, 2015. 36(1): p. 89-91. 90. Kloss, B.A., et al., Analysis of FOXD3 sequence variation in human ocular disease. Mol Vis, 2012. 18: p. 1740-9. 91. Zhang, X., et al., Variants in TRIM44 Cause Aniridia by Impairing PAX6 Expression. Hum Mutat, 2015. 36(12): p. 1164-7. 92. Fischbach, B.V., et al., WAGR syndrome: a clinical review of 54 cases. Pediatrics, 2005. 116(4): p. 984-8. 93. Nelson, L.B., et al., Aniridia. A review. Surv Ophthalmol, 1984. 28(6): p. 621-42. 94. Ton, C.C., et al., Positional cloning and characterization of a paired box- and homeobox- containing gene from the aniridia region. Cell, 1991. 67(6): p. 1059-74. 95. Ton, C.C.T., et al., Positional Cloning and Characterization of a Paired Box-Containing and Homeobox-Containing Gene from the Aniridia Region. Cell, 1991. 67(6): p. 1059- 1074. 96. Royer-Pokora, B., et al., Twenty-four new cases of WT1 germline mutations and review of the literature: genotype/phenotype correlations for Wilms tumor development. American Journal of Medical Genetics Part A, 2004. 127A(3): p. 249-57. 97. Czerny, T., et al., twin of eyeless, a second Pax-6 gene of Drosophila, acts upstream of eyeless in the control of eye development. Mol Cell, 1999. 3(3): p. 297-307. 98. Epstein, J.A., et al., Two independent and interactive DNA-binding subdomains of the Pax6 paired domain are regulated by alternative splicing. Genes Dev, 1994. 8(17): p. 2022-34. 99. Prosser, J. and V. van Heyningen, PAX6 mutations reviewed. Hum Mutat, 1998. 11(2): p. 93-108. 100. Fantes, J., et al., Aniridia-Associated Cytogenetic Rearrangements Suggest That a Position Effect May Cause the Mutant Phenotype. Hum Mol Genet, 1995. 4(3): p. 415- 422. 101. Vincent, M.C., et al., Screening for PAX6 gene mutations is consistent with haploinsufficiency as the main mechanism leading to various ocular defects. European Journal of Human Genetics, 2003. 11(2): p. 163-169. 102. Tzoulaki, I., I.M.S. White, and I.M. Hanson, PAX6 mutations: genotype-phenotype correlations. BMC Genet, 2005. 6. 103. Hingorani, M., et al., Detailed Ophthalmologic Evaluation of 43 Individuals with PAX6 Mutations. Invest Ophthalmol Vis Sci, 2009. 50(6): p. 2581-2590. 104. Hanson, I., et al., Missense mutations in the most ancient residues of the PAX6 paired domain underlie a spectrum of human congenital eye malformations. Hum Mol Genet, 1999. 8(2): p. 165-172. 105. Sanchez-Martinez, C., Phenotypic and functional differences between aniridic and healthy corneal stromal cells, in Institute of Ophthalmology. 2015, University College London. 106. Kondo-Saitoh, A., et al., Two nonsense mutations of PAX6 in two Japanese aniridia families: case report and review of the literature. Eur J Ophthalmol, 2000. 10(2): p. 167- 72. 107. Chao, L.Y., et al., Missense mutations in the DNA-binding region and termination codon in PAX6. Hum Mutat, 2003. 21(2): p. 138-45. 108. Henderson, R.A., et al., Inherited PAX6, NF1 and OTX2 mutations in a child with microphthalmia and aniridia. Eur J Hum Genet, 2007. 15(8): p. 898-901. 109. Gupta, S.K., et al., Genotype/phenotype correlations in aniridia. Am J Ophthalmol, 1998. 126(2): p. 203-210. 110. Dubey, S.K., et al., Mutational analysis and genotype-phenotype correlations in southern Indian patients with sporadic and familial aniridia. Mol Vis, 2015. 21: p. 88-97. 269

111. Yokoi, T., et al., Genotype-phenotype correlation of PAX6 gene mutations in aniridia. Hum Genome Var, 2016. 3: p. 15052. 112. Schmidt-Sidor, B., et al., Malformations of the brain in two fetuses with a compound heterozygosity for two PAX6 mutations. Folia Neuropathologica, 2009. 47(4): p. 371-382. 113. Glaser, T., et al., Pax6 Gene Dosage Effect in a Family with Congenital Cataracts, Aniridia, Anophthalmia and Central-Nervous-System Defects. Nat Genet, 1994. 7(4): p. 463-471. 114. Solomon, B.D., et al., Compound Heterozygosity for Mutations in PAX6 in a Patient With Complex Brain Anomaly, Neonatal Diabetes Mellitus, and Microophthalmia. American Journal of Medical Genetics Part A, 2009. 149A(11): p. 2543-2546. 115. Duncan, M.K., et al., Truncated forms of Pax-6 disrupt lens morphology in transgenic mice. Invest Ophthalmol Vis Sci, 2000. 41(2): p. 464-473. 116. Singh, S., et al., Truncation mutations in the transactivation region of PAX6 result in dominant-negative mutants. Journal of Biological Chemistry, 1998. 273(34): p. 21531- 21541. 117. Liu, Q., et al., A novel PAX6 deletion in a Chinese family with congenital aniridia. Gene, 2015. 563(1): p. 41-4. 118. Simpson, T.I. and D.J. Price, Pax6; a pleiotropic player in development. Bioessays, 2002. 24(11): p. 1041-1051. 119. Xu, H.E., et al., Crystal structure of the human Pax6 paired domain-DNA complex reveals specific roles for the linker region and carboxy-terminal subdomain in DNA binding. Genes Dev, 1999. 13(10): p. 1263-75. 120. Kissinger, C.R., et al., Crystal-Structure of an Engrailed Homeodomain-DNA Complex at 2.8-a Resolution - a Framework for Understanding Homeodomain-DNA Interactions. Cell, 1990. 63(3): p. 579-590. 121. Hanson, I. and V. Van Heyningen, Pax6: more than meets the eye. Trends Genet, 1995. 11(7): p. 268-72. 122. Kioussi, C., et al., Pax6 is essential for establishing ventral-dorsal cell boundaries in pituitary gland development. Proc Natl Acad Sci U S A, 1999. 96(25): p. 14378-82. 123. Dohrmann, C., P. Gruss, and L. Lemaire, Pax genes and the differentiation of hormone- producing endocrine cells in the pancreas. Mech Dev, 2000. 92(1): p. 47-54. 124. Osumi, N., et al., Concise Review: Pax6 Transcription Factor Contributes to both Embryonic and Adult Neurogenesis as a Multifunctional Regulator. STEM CELLS, 2008. 26(7): p. 1663-1672. 125. Halder, G., P. Callaerts, and W.J. Gehring, Induction of ectopic eyes by targeted expression of the eyeless gene in Drosophila. Science, 1995. 267(5205): p. 1788-92. 126. Collinson, J.M., et al., The roles of Pax6 in the cornea, retina, and olfactory epithelium of the developing mouse embryo. Dev Biol, 2003. 255(2): p. 303-12. 127. Li, H.S., et al., Pax-6 is first expressed in a region of ectoderm anterior to the early neural plate: implications for stepwise determination of the lens. Dev Biol, 1994. 162(1): p. 181- 94. 128. Collinson, J.M., et al., Clonal analysis of patterns of growth, stem cell activity, and cell movement during the development and maintenance of the murine corneal epithelium. Dev Dyn, 2002. 224(4): p. 432-40. 129. Collinson, J.M., et al., Corneal development, limbal stem cell function, and corneal epithelial cell migration in the Pax6(+/-) mouse. Invest Ophthalmol Vis Sci, 2004. 45(4): p. 1101-8. 130. Dorà, N., et al., PAX6 dosage effects on corneal development, growth, and wound healing. Dev Dyn, 2008. 237(5): p. 1295-306. 131. Kanakubo, S., et al., Abnormal migration and distribution of neural crest cells in Pax6 heterozygous mutant eye, a model for human eye diseases. Genes Cells, 2006. 11(8): p. 919-33.

270

132. Davis, J., et al., Requirement for Pax6 in corneal morphogenesis: a role in adhesion. J Cell Sci, 2003. 116(Pt 11): p. 2157-67. 133. Ouyang, H., et al., WNT7A and PAX6 define corneal epithelium homeostasis and pathogenesis. Nature, 2014. 511(7509): p. 358-61. 134. Kitazawa, K., et al., PAX6 regulates human corneal epithelium cell identity. Exp Eye Res, 2017. 154: p. 30-38. 135. Latta, L., et al., Expression of retinoic acid signaling components ADH7 and ALDH1A1 is reduced in aniridia limbal epithelial cells and a siRNA primary cell based aniridia model. Exp Eye Res, 2019. 179: p. 8-17. 136. Li, G., et al., Transcription Factor PAX6 (Paired Box 6) Controls Limbal Stem Cell Lineage in Development and Disease. J Biol Chem, 2015. 290(33): p. 20448-54. 137. Ramaesh, T., et al., Increased apoptosis and abnormal wound-healing responses in the heterozygous Pax6+/- mouse cornea. Invest Ophthalmol Vis Sci, 2006. 47(5): p. 1911-7. 138. Sivak, J.M., et al., Transcription Factors Pax6 and AP-2alpha Interact To Coordinate Corneal Epithelial Repair by Controlling Expression of Matrix Metalloproteinase Gelatinase B. Mol Cell Biol, 2004. 24(1): p. 245-57. 139. Leiper, L.J., et al., The roles of calcium signaling and ERK1/2 phosphorylation in a Pax6+/- mouse model of epithelial wound-healing delay. BMC Biol, 2006. 4: p. 27. 140. Mackman, G., F.S. Brightbill, and J.M. Optiz, Corneal changes in aniridia. Am J Ophthalmol, 1979. 87(4): p. 497-502. 141. Eden, U., R. Riise, and K. Tornqvist, Corneal Involvement in Congenital Aniridia. Cornea, 2010. 29(10): p. 1096-1102. 142. Rama, P., Viganò, M., & Knutsson, K. A., The ocular surface in aniridia, in In Aniridia: Recent Developments in Scientific and Clinical Research. 2015, Springer International Publishing. p. (pp. 45-52)

143. Tseng, S.C., Concept and application of limbal stem cells. Eye (Lond), 1989. 3 ( Pt 2): p. 141-57. 144. Douvaras, P., et al., Increased corneal epithelial turnover contributes to abnormal homeostasis in the Pax6(+/-) mouse model of aniridia. PLoS One, 2013. 8(8): p. e71117. 145. Secker, G.A. and J.T. Daniels, Corneal epithelial stem cells: Deficiency and regulation. Stem Cell Reviews, 2008. 4(3): p. 159-168. 146. Gipson, I.K. and T. Inatomi, Mucin genes expressed by the ocular surface Epithelium. Progress in Retinal and Eye Research, 1997. 16(1): p. 81-98. 147. Lagali, N., et al., In Vivo Morphology of the Limbal Palisades of Vogt Correlates With Progressive Stem Cell Deficiency in Aniridia-Related Keratopathy. Invest Ophthalmol Vis Sci, 2013. 54(8): p. 5333-5342. 148. Brandt, J.D., L.A. Casuso, and D.L. Budenz, Markedly increased central corneal thickness: an unrecognized finding in congenital aniridia. Am J Ophthalmol, 2004. 137(2): p. 348- 50. 149. Hassanaly, S.I., J.C. Talajic, and M. Harissi-Dagher, Outcomes following Boston type 1 keratoprosthesis implantation in aniridia patients at the University of Montreal. Am J Ophthalmol, 2014. 158(2): p. 270-276.e1. 150. Egbert, P.R., S. Lauber, and D.M. Maurice, A simple conjunctival biopsy. Am J Ophthalmol, 1977. 84(6): p. 798-801. 151. Dua, H.S., et al., The role of limbal stem cells in corneal epithelial maintenance: testing the dogma. Ophthalmology, 2009. 116(5): p. 856-63. 152. Gregory-Evans, C.Y., et al., Postnatal manipulation of Pax6 dosage reverses congenital tissue malformation defects. J Clin Invest, 2014. 124(1): p. 111-6.

271

153. Namgoong, J.H. and C. Bertoni, Clinical potential of ataluren in the treatment of Duchenne muscular dystrophy. Degenerative Neurological and Neuromuscular Disease, 2016. 6: p. 37-48. 154. Seitz, B., B. Kasmann-Kellner, and A. Viestenz, [Stage-related therapy of congenital aniridia]. Ophthalmologe, 2014. 111(12): p. 1164-71. 155. Margo, C.E., Congenital aniridia: a histopathologic study of the anterior segment in children. J Pediatr Ophthalmol Strabismus, 1983. 20(5): p. 192-8. 156. Tseng, S.C. and D.Q. Li, Comparison of protein kinase C subtype expression between normal and aniridic human ocular surfaces: implications for limbal stem cell dysfunction in aniridia. Cornea, 1996. 15(2): p. 168-78. 157. Le, Q., S.X. Deng, and J. Xu, In vivo confocal microscopy of congenital aniridia-associated keratopathy. Eye (Lond), 2013. 27(6): p. 763-6. 158. Lagali, N., et al., In vivo morphology of the limbal palisades of vogt correlates with progressive stem cell deficiency in aniridia-related keratopathy. Invest Ophthalmol Vis Sci, 2013. 54(8): p. 5333-42. 159. Li, W., et al., Down-regulation of Pax6 is associated with abnormal differentiation of corneal epithelial cells in severe ocular surface diseases. Journal of Pathology, 2008. 214(1): p. 114-122. 160. Vicente, A., B. Byström, and F. Pedrosa Domellöf, Altered Signaling Pathways in Aniridia- Related KeratopathySignaling Pathways in Aniridia-Related Keratopathy. Investigative Ophthalmology & Visual Science, 2018. 59(13): p. 5531-5541. 161. Vicente, A., et al., Aniridia-related keratopathy: Structural changes in naïve and transplanted corneal buttons. PLoS One, 2018. 13(6): p. e0198822. 162. Roberts, R.C., Small Eyes - a New Dominant Eye Mutant in Mouse. Genetical Research, 1967. 9(1): p. 121-&. 163. Hill, R.E., et al., Mouse small eye results from mutations in a paired-like homeobox- containing gene. Nature, 1991. 354(6354): p. 522-5. 164. Davis, J., et al., Requirement for Pax6 in corneal morphogenesis: a role in adhesion. Journal of Cell Science, 2003. 116(Pt 11): p. 2157-67. 165. Ramaesh, T., et al., Developmental and cellular factors underlying corneal epithelial dysgenesis in the Pax6+/- mouse model of aniridia. Exp Eye Res, 2005. 81(2): p. 224-35. 166. Ou, J., et al., Chronic wound state exacerbated by oxidative stress in Pax6+/- aniridia- related keratopathy. J Pathol, 2008. 215(4): p. 421-30. 167. Ou, J., C. Lowes, and J.M. Collinson, Cytoskeletal and cell adhesion defects in wounded and Pax6+/- corneal epithelia. Invest Ophthalmol Vis Sci, 2010. 51(3): p. 1415-23. 168. Ramaesh, T., et al., Corneal abnormalities in Pax6(+/-) small eye mice mimic human aniridia-related keratopathy. Invest Ophthalmol Vis Sci, 2003. 44(5): p. 1871-1878. 169. Mort, R.L., et al., Effects of aberrant Pax6 gene dosage on mouse corneal pathophysiology and corneal epithelial homeostasis. PLoS One, 2011. 6(12): p. e28895. 170. Sasamoto, Y., et al., PAX6 Isoforms, along with Reprogramming Factors, Differentially Regulate the Induction of Cornea-specific Genes. Sci Rep, 2016. 6: p. 20807. 171. Zhang, J., et al., Fibroblast growth factor receptor 2 (FGFR2) is required for corneal epithelial cell proliferation and differentiation during embryonic development. PLoS One, 2015. 10(1): p. e0117089. 172. Roux, L.N., et al., Modeling of Aniridia-Related Keratopathy by CRISPR/Cas9 Genome Editing of Human Limbal Epithelial Cells and Rescue by Recombinant PAX6 Protein. Stem Cells, 2018. 36(9): p. 1421-1429. 173. Latta, L., et al., Human aniridia limbal epithelial cells lack expression of keratins K3 and K12. Exp Eye Res, 2018. 167: p. 100-109.

272

174. Kureshi, A.K., J.L. Funderburgh, and J.T. Daniels, Human corneal stromal stem cells exhibit survival capacity following isolation from stored organ-culture corneas. Invest Ophthalmol Vis Sci, 2014. 55(11): p. 7583-8. 175. Jiang, Y., et al., Pluripotency of mesenchymal stem cells derived from adult marrow. Nature, 2002. 418(6893): p. 41-9. 176. Basu, S., et al., Human limbal biopsy-derived stromal stem cells prevent corneal scarring. Sci Transl Med, 2014. 6(266): p. 266ra172. 177. Livak, K.J. and T.D. Schmittgen, Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods, 2001. 25(4): p. 402-8. 178. West-Mays, J.A. and D.J. Dwivedi, The keratocyte: corneal stromal cell with variable repair phenotypes. Int J Biochem Cell Biol, 2006. 38(10): p. 1625-31. 179. Wilson, S.E., et al., Herpes simplex virus type-1 infection of corneal epithelial cells induces apoptosis of the underlying keratocytes. Exp Eye Res, 1997. 64(5): p. 775-9. 180. Fernandez-Perez, J. and M. Ahearne, Influence of Biochemical Cues in Human Corneal Stromal Cell Phenotype. Curr Eye Res, 2019. 44(2): p. 135-146. 181. Sidney, L.E. and A. Hopkinson, Corneal keratocyte transition to mesenchymal stem cell phenotype and reversal using serum-free medium supplemented with fibroblast growth factor-2, transforming growth factor-beta3 and retinoic acid. J Tissue Eng Regen Med, 2018. 12(1): p. e203-e215. 182. Lynch, A.P., F. O'Sullivan, and M. Ahearne, The effect of growth factor supplementation on corneal stromal cell phenotype in vitro using a serum-free media. Exp Eye Res, 2016. 151: p. 26-37. 183. Berryhill, B.L., et al., Partial restoration of the keratocyte phenotype to bovine keratocytes made fibroblastic by serum. Invest Ophthalmol Vis Sci, 2002. 43(11): p. 3416-21. 184. Musselmann, K., et al., Maintenance of the keratocyte phenotype during cell proliferation stimulated by insulin. J Biol Chem, 2005. 280(38): p. 32634-9. 185. Funderburgh, J.L., M.M. Mann, and M.L. Funderburgh, Keratocyte phenotype mediates proteoglycan structure: a role for fibroblasts in corneal fibrosis. J Biol Chem, 2003. 278(46): p. 45629-37. 186. Long, C.J., et al., Fibroblast growth factor-2 promotes keratan sulfate proteoglycan expression by keratocytes in vitro. J Biol Chem, 2000. 275(18): p. 13918-23. 187. Vereb, Z., et al., Role of Human Corneal Stroma-Derived Mesenchymal-Like Stem Cells in Corneal Immunity and Wound Healing. Sci Rep, 2016. 6: p. 26227. 188. Foster, J.W., R.M. Gouveia, and C.J. Connon, Low-glucose enhances keratocyte- characteristic phenotype from corneal stromal cells in serum-free conditions. Sci Rep, 2015. 5: p. 10839. 189. Musselmann, K., et al., Stimulation of collagen synthesis by insulin and proteoglycan accumulation by ascorbate in bovine keratocytes in vitro. Invest Ophthalmol Vis Sci, 2006. 47(12): p. 5260-6. 190. Jester, J.V., et al., Temporal, 3-dimensional, cellular anatomy of corneal wound tissue. J Anat, 1995. 186 ( Pt 2): p. 301-11. 191. Massie, I., S.B. Dale, and J.T. Daniels, Limbal Fibroblasts Maintain Normal Phenotype in 3D RAFT Tissue Equivalents Suggesting Potential for Safe Clinical Use in Treatment of Ocular Surface Failure. Tissue Eng Part C Methods, 2015. 21(6): p. 576-84. 192. Wu, J., et al., Corneal stromal stem cells versus corneal fibroblasts in generating structurally appropriate corneal stromal tissue. Exp Eye Res, 2014. 120: p. 71-81. 193. Polisetty, N., et al., Mesenchymal cells from limbal stroma of human eye. Mol Vis, 2008. 14: p. 431-42. 194. Du, Y., et al., Secretion and organization of a cornea-like tissue in vitro by stem cells from human corneal stroma. Invest Ophthalmol Vis Sci, 2007. 48(11): p. 5038-45. 273

195. Yun, Y.R., et al., Fibroblast growth factors: biology, function, and application for tissue regeneration. J Tissue Eng, 2010. 2010: p. 218142. 196. Wu, J., et al., Bioengineering organized, multilamellar human corneal stromal tissue by growth factor supplementation on highly aligned synthetic substrates. Tissue Eng Part A, 2013. 19(17-18): p. 2063-75. 197. Jester, J.V., et al., Corneal keratocytes: in situ and in vitro organization of cytoskeletal contractile proteins. Invest Ophthalmol Vis Sci, 1994. 35(2): p. 730-43. 198. Chen, S.Y., et al., Pax 6 Controls Neural Crest Potential of Limbal Niche Cells to Support Self-Renewal of Limbal Epithelial Stem Cells. Sci Rep, 2019. 9(1): p. 9763. 199. Dominici, M., et al., Minimal criteria for defining multipotent mesenchymal stromal cells. The International Society for Cellular Therapy position statement. Cytotherapy, 2006. 8(4): p. 315-7. 200. Pei, Y., D.M. Sherry, and A.M. McDermott, Thy-1 distinguishes human corneal fibroblasts and myofibroblasts from keratocytes. Exp Eye Res, 2004. 79(5): p. 705-12. 201. Carlson, E.C., et al., Keratocan, a cornea-specific keratan sulfate proteoglycan, is regulated by lumican. J Biol Chem, 2005. 280(27): p. 25541-7. 202. Masur, S.K., et al., Myofibroblasts differentiate from fibroblasts when plated at low density. Proc Natl Acad Sci U S A, 1996. 93(9): p. 4219-23. 203. Hashmani, K., et al., Characterization of corneal stromal stem cells with the potential for epithelial transdifferentiation. Stem Cell Res Ther, 2013. 4(3): p. 75. 204. Caplan, A.I., Mesenchymal Stem Cells: Time to Change the Name! Stem Cells Transl Med, 2017. 6(6): p. 1445-1451. 205. Caplan, A.I., What's in a name? Tissue Eng Part A, 2010. 16(8): p. 2415-7. 206. Sidney, L.E., et al., Effect of culture medium on propagation and phenotype of corneal stroma-derived stem cells. Cytotherapy, 2015. 17(12): p. 1706-22. 207. Vicente, A., B. Bystrom, and F. Pedrosa Domellof, Altered Signaling Pathways in Aniridia- Related Keratopathy. Invest Ophthalmol Vis Sci, 2018. 59(13): p. 5531-5541. 208. Ramaesh, T., et al., Increased Apoptosis and Abnormal Wound-Healing Responses in the Heterozygous Pax6 +/− Mouse Cornea. Investigative Ophthalmology & Visual Science, 2006. 47(5): p. 1911-1917. 209. Ihnatko, R., et al., Congenital Aniridia and the Ocular Surface. Ocul Surf, 2016. 14(2): p. 196-206. 210. Bausili, M.M., et al., Histopathology Findings of Corneal Buttons in Congenital Aniridia Patients. Ophthalmic Res, 2016. 56(4): p. 202-206. 211. Ramaesh, T., et al., Corneal abnormalities in Pax6+/- small eye mice mimic human aniridia-related keratopathy. Invest Ophthalmol Vis Sci, 2003. 44(5): p. 1871-8. 212. Shubin, A.V., et al., Cytoplasmic vacuolization in cell death and survival. Oncotarget, 2016. 7(34): p. 55863-55889. 213. G. A. Secker, Q.P.S., A. J. Shortt, P. T. Khaw, R. R. Ali, J. T. Daniels, Pax6 Deficient Fibroblasts Display an Exaggerated Wound Healing Response, in ARVO. 2007, Invest. Ophthalmol. Vis. Sci. 2007;48(13):3502. 214. Henriksson, J.T., A.M. McDermott, and J.P. Bergmanson, Dimensions and morphology of the cornea in three strains of mice. Invest Ophthalmol Vis Sci, 2009. 50(8): p. 3648-54. 215. Lagali, N., et al., Survival of donor-derived cells in human corneal transplants. Invest Ophthalmol Vis Sci, 2009. 50(6): p. 2673-8. 216. Bhatia, S., et al., Disruption of autoregulatory feedback by a mutation in a remote, ultraconserved PAX6 enhancer causes aniridia. American Journal of Human Genetics, 2013. 93(6): p. 1126-34. 217. Weisschuh, N., B. Wissinger, and E. Gramer, A splice site mutation in the PAX6 gene which induces exon skipping causes autosomal dominant inherited aniridia. Mol Vis, 2012. 18: p. 751-7.

274

218. Justus, C.R., et al., In vitro cell migration and invasion assays. J Vis Exp, 2014(88). 219. Grinnell, F., Fibroblast-collagen-matrix contraction: growth-factor signalling and mechanical loading. Trends Cell Biol, 2000. 10(9): p. 362-5. 220. Dahlmann-Noor, A.H., et al., Dynamic protrusive cell behaviour generates force and drives early matrix contraction by fibroblasts. Exp Cell Res, 2007. 313(20): p. 4158-69. 221. Kureshi, A.K., J.L. Funderburgh, and J.T. Daniels, Human Corneal Stromal Stem Cells Exhibit Survival Capacity Following Isolation From Stored Organ–Culture CorneasSurvival Capacity of Human CSSCs. Investigative Ophthalmology & Visual Science, 2014. 55(11): p. 7583-7588. 222. Baum, L., et al., Run-on mutation and three novel nonsense mutations identified in the PAX6 gene in patients with aniridia. Hum Mutat, 1999. 14(3): p. 272-3. 223. Singh, S., et al., Missense mutation at the C-terminus of PAX6 negatively modulates homeodomain function. Hum Mol Genet, 2001. 10(9): p. 911-918. 224. Hever, A.M., K.A. Williamson, and V. van Heyningen, Developmental malformations of the eye: the role of PAX6, SOX2 and OTX2. Clin Genet, 2006. 69(6): p. 459-70. 225. Love, J., et al., A new set of primers for mutation analysis of the human PAX6 gene. Hum Mutat, 1998. 12(2): p. 128-34. 226. Hingorani, M., et al., Detailed Ophthalmologic Evaluation of 43 Individuals with PAX6 Mutations. Investigative Ophthalmology & Visual Science, 2009. 50(6): p. 2581-2590. 227. Lim, H.T., D.H. Kim, and H. Kim, PAX6 aniridia syndrome: clinics, genetics, and therapeutics. Curr Opin Ophthalmol, 2017. 28(5): p. 436-447. 228. Ouyang, J., et al., Pax6 overexpression suppresses cell proliferation and retards the cell cycle in corneal epithelial cells. Invest Ophthalmol Vis Sci, 2006. 47(6): p. 2397-407. 229. Bai, S.W., et al., Pax6 regulates proliferation and apoptosis of human retinoblastoma cells. Invest Ophthalmol Vis Sci, 2011. 52(7): p. 4560-70. 230. Hsieh, Y.W. and X.J. Yang, Dynamic Pax6 expression during the neurogenic cell cycle influences proliferation and cell fate choices of retinal progenitors. Neural Dev, 2009. 4: p. 32. 231. Quinn, J.C., et al., Pax6 controls cerebral cortical cell number by regulating exit from the cell cycle and specifies cortical cell identity by a cell autonomous mechanism. Dev Biol, 2007. 302(1): p. 50-65. 232. Li, Y., et al., PAX6, a novel target of microRNA-7, promotes cellular proliferation and invasion in human colorectal cancer cells. Dig Dis Sci, 2014. 59(3): p. 598-606. 233. Lehrer, M.S., T.T. Sun, and R.M. Lavker, Strategies of epithelial repair: modulation of stem cell and transit amplifying cell proliferation. J Cell Sci, 1998. 111 ( Pt 19): p. 2867- 75. 234. Walczysko, P., A.M. Rajnicek, and J.M. Collinson, Contact-mediated control of radial migration of corneal epithelial cells. Mol Vis, 2016. 22: p. 990-1004. 235. Pilcher, B.K., et al., The activity of collagenase-1 is required for keratinocyte migration on a type I collagen matrix. Journal of Cell Biology, 1997. 137(6): p. 1445-1457. 236. Mishima, H., et al., SPARC from corneal epithelial cells modulates collagen contraction by keratocytes. Invest Ophthalmol Vis Sci, 1998. 39(13): p. 2547-53. 237. Chaurasia, S.S., et al., Dynamics of the expression of intermediate filaments vimentin and desmin during myofibroblast differentiation after corneal injury. Exp Eye Res, 2009. 89(2): p. 133-9. 238. Samant, M., et al., Congenital aniridia: etiology, manifestations and management. Expert Rev Ophthalmol, 2016. 11(2): p. 135-144. 239. Fridman, J.S. and S.W. Lowe, Control of apoptosis by p53. Oncogene, 2003. 22(56): p. 9030-9040. 240. Adams, J.M. and S. Cory, The Bcl-2 apoptotic switch in cancer development and therapy. Oncogene, 2007. 26(9): p. 1324-37. 275

241. Thompson, R.E., et al., Three-dimensional cell culture environment promotes partial recovery of the native corneal keratocyte phenotype from a subcultured population. Tissue Eng Part A, 2013. 19(13-14): p. 1564-72. 242. Chen, S., M.J. Mienaltowski, and D.E. Birk, Regulation of corneal stroma extracellular matrix assembly. Exp Eye Res, 2015. 133: p. 69-80. 243. Ahmed, M., et al., A combinatorial approach towards the design of nanofibrous scaffolds for chondrogenesis. Sci Rep, 2015. 5: p. 14804. 244. Meng, B., Y. Wang, and B. Li, Suppression of PAX6 promotes cell proliferation and inhibits apoptosis in human retinoblastoma cells. Int J Mol Med, 2014. 34(2): p. 399-408. 245. Adams, J.C. and F.M. Watt, Regulation of development and differentiation by the extracellular matrix. Development, 1993. 117(4): p. 1183-98. 246. Friedl, P., E.-B.J.C. Bröcker, and M.L.S. CMLS, The biology of cell locomotion within three- dimensional extracellular matrix. 2000. 57(1): p. 41-64. 247. Bard, J.B. and K. Higginson, Fibroblast-collagen interactions in the formation of the secondary stroma of the chick cornea. J Cell Biol, 1977. 74(3): p. 816-27. 248. Karamichos, D., N. Lakshman, and W.M. Petroll, Regulation of Corneal Fibroblast Morphology and Collagen Reorganization by Extracellular Matrix Mechanical Properties. Investigative Ophthalmology & Visual Science, 2007. 48(11): p. 5030-5037. 249. Duval, K., et al., Modeling Physiological Events in 2D vs. 3D Cell Culture. 2017. 32(4): p. 266-277. 250. Wei, Q., et al., Synthetic Polymers for Biomedical Applications. International Journal of Biomaterials, 2018. 2018: p. 2. 251. Brown, R.A., et al., Ultrarapid Engineering of Biomimetic Materials and Tissues: Fabrication of Nano- and Microstructures by Plastic Compression. Advanced Functional Materials, 2005. 15(11): p. 1762-1770. 252. Wilson, S.L., Y. Yang, and A.J. El Haj, Corneal stromal cell plasticity: in vitro regulation of cell phenotype through cell-cell interactions in a three-dimensional model. Tissue Eng Part A, 2014. 20(1-2): p. 225-38. 253. Levis, H.J., R.A. Brown, and J.T. Daniels, Plastic compressed collagen as a biomimetic substrate for human limbal epithelial cell culture. Biomaterials, 2010. 31(30): p. 7726- 7737. 254. Levis, H.J., et al., Plastic Compressed Collagen as a Novel Carrier for Expanded Human Corneal Endothelial Cells for Transplantation. PLoS One, 2012. 7(11). 255. Levis, H.J. and J.T. Daniels, Recreating the Human Limbal Epithelial Stem Cell Niche with Bioengineered Limbal Crypts. Curr Eye Res, 2016. 41(9): p. 1153-60. 256. Shojaati, G., et al., Compressed Collagen Enhances Stem Cell Therapy for Corneal Scarring. Stem Cells Transl Med, 2018. 7(6): p. 487-494. 257. Levis, H.J., et al., Plastic Compressed Collagen Constructs for Ocular Cell Culture and Transplantation: A New and Improved Technique of Confined Fluid Loss. Curr Eye Res, 2013. 38(1): p. 41-52. 258. Levis, H.J., et al., Rapid tissue engineering of biomimetic human corneal limbal crypts with 3D niche architecture. Biomaterials, 2013. 34(35): p. 8860-8. 259. Peterbauer, T., et al., Dynamics of Spreading and Alignment of Cells Cultured In Vitro on a Grooved Polymer Surface %J Journal of Nanomaterials. 2011. 2011: p. 10. 260. ., R.T., RStudio: Integrated Development for R. RStudio, Inc., Boston, MA URL http://www.rstudio.com/. 2015. 261. Ouyang, J., et al., Pax6 Overexpression Suppresses Cell Proliferation and Retards the Cell Cycle in Corneal Epithelial Cells. Investigative Ophthalmology & Visual Science, 2006. 47(6): p. 2397-2407. 262. Duncan, M.K., et al., Dual roles for Pax-6: a transcriptional repressor of lens fiber cell- specific beta-crystallin genes. Mol Cell Biol, 1998. 18(9): p. 5579-86.

276

263. Yamaoka, T., et al., Diabetes and pancreatic tumours in transgenic mice expressing Pa x 6. Diabetologia, 2000. 43(3): p. 332-9. 264. Hassell, J.R. and D.E. Birk, The molecular basis of corneal transparency. Exp Eye Res, 2010. 91(3): p. 326-35. 265. Doane, K.J. and D.E. Birk, Fibroblasts retain their tissue phenotype when grown in three- dimensional collagen gels. Exp Cell Res, 1991. 195(2): p. 432-42. 266. Kivanany, P.B., et al., Assessment of Corneal Stromal Remodeling and Regeneration after Photorefractive Keratectomy. Sci Rep, 2018. 8(1): p. 12580. 267. Kim, A., et al., Growth factor regulation of corneal keratocyte differentiation and migration in compressed collagen matrices. Invest Ophthalmol Vis Sci, 2010. 51(2): p. 864-75. 268. Li, T. and L. Lu, Epidermal growth factor-induced proliferation requires down-regulation of Pax6 in corneal epithelial cells. J Biol Chem, 2005. 280(13): p. 12988-95. 269. Yu, A., et al., Effect of EGF and bFGF on fibroblast proliferation and angiogenic cytokine production from cultured dermal substitutes. J Biomater Sci Polym Ed, 2012. 23(10): p. 1315-24. 270. Lembach, K.J., Induction of human fibroblast proliferation by epidermal growth factor (EGF): enhancement by an EGF-binding arginine esterase and by ascorbate. Proc Natl Acad Sci U S A, 1976. 73(1): p. 183-7. 271. Wu, J., et al., The engineering of organized human corneal tissue through the spatial guidance of corneal stromal stem cells. Biomaterials, 2012. 33(5): p. 1343-52. 272. Yoshida, S., et al., Serum-free spheroid culture of mouse corneal keratocytes. Invest Ophthalmol Vis Sci, 2005. 46(5): p. 1653-8. 273. Byun, Y.S., et al., Keratocytes derived from spheroid culture of corneal stromal cells resemble tissue resident keratocytes. PLoS One, 2014. 9(11): p. e112781. 274. Abidin, F.Z., R.M. Gouveia, and C.J. Connon, Application of retinoic acid improves form and function of tissue engineered corneal construct. Organogenesis, 2015. 11(3): p. 122- 36. 275. Cui, Z., et al., Cell-laden and orthogonal-multilayer tissue-engineered corneal stroma induced by a mechanical collagen microenvironment and transplantation in a rabbit model. Acta Biomater, 2018. 75: p. 183-199. 276. Jester, J.V., et al., Corneal keratocytes: phenotypic and species differences in abundant protein expression and in vitro light-scattering. Invest Ophthalmol Vis Sci, 2005. 46(7): p. 2369-78. 277. Karring, H., et al., Proteomic analysis of the soluble fraction from human corneal fibroblasts with reference to ocular transparency. Mol Cell Proteomics, 2004. 3(7): p. 660-74. 278. Nagamoto, T., G. Eguchi, and D.C. Beebe, Alpha-smooth muscle actin expression in cultured lens epithelial cells. Invest Ophthalmol Vis Sci, 2000. 41(5): p. 1122-9. 279. Hinz, B. and G. Gabbiani, Mechanisms of force generation and transmission by myofibroblasts. Curr Opin Biotechnol, 2003. 14(5): p. 538-46. 280. Owen, N. and M. Moosajee, RNA-sequencing in ophthalmology research: considerations for experimental design and analysis. Ther Adv Ophthalmol, 2019. 11: p. 2515841419835460. 281. You, J., et al., RNA-Seq analysis and comparison of corneal epithelium in keratoconus and myopia patients. Sci Rep, 2018. 8(1): p. 389. 282. Winkler, N.S., et al., Fuchs' Endothelial Corneal Dystrophy in Patients With Myotonic Dystrophy, Type 1. Invest Ophthalmol Vis Sci, 2018. 59(7): p. 3053-3057. 283. Chng, Z., et al., High throughput gene expression analysis identifies reliable expression markers of human corneal endothelial cells. PLoS One, 2013. 8(7): p. e67546.

277

284. Khaled, M.L., et al., Differential Expression of Coding and Long Noncoding RNAs in Keratoconus-Affected CorneasDifferential Expression in Keratoconic Corneas. Investigative Ophthalmology & Visual Science, 2018. 59(7): p. 2717-2728. 285. Eurich, K., et al., Potential role of chitinase 3-like-1 in inflammation-associated carcinogenic changes of epithelial cells. World J Gastroenterol, 2009. 15(42): p. 5249- 59. 286. Ling, H. and A.D. Recklies, The chitinase 3-like protein human cartilage glycoprotein 39 inhibits cellular responses to the inflammatory cytokines interleukin-1 and tumour necrosis factor-alpha. Biochem J, 2004. 380(Pt 3): p. 651-9. 287. Gao, N. and F.S. Yu, Chitinase 3-Like 1 Promotes Candida albicans Killing and Preserves Corneal Structure and Function by Controlling Host Antifungal Responses. Infect Immun, 2015. 83(10): p. 4154-64. 288. Davis, J. and J. Piatigorsky, Overexpression of Pax6 in mouse cornea directly alters corneal epithelial cells: changes in immune function, vascularization, and differentiation. Invest Ophthalmol Vis Sci, 2011. 52(7): p. 4158-68. 289. Tu, C.F., et al., Localization and characterization of a novel secreted protein SCUBE1 in human platelets. Cardiovasc Res, 2006. 71(3): p. 486-95. 290. Walcher, T., et al., Functional dissection of the paired domain of Pax6 reveals molecular mechanisms of coordinating neurogenesis and proliferation. Development, 2013. 140(5): p. 1123-36. 291. Collinson, J.M., R.E. Hill, and J.D. West, Different roles for Pax6 in the optic vesicle and facial epithelium mediate early morphogenesis of the murine eye. Development, 2000. 127(5): p. 945-56. 292. Chapouton, P., A. Gartner, and M. Gotz, The role of Pax6 in restricting cell migration between developing cortex and basal ganglia. 1999. 126(24): p. 5569-5579. 293. Matsuo, T., et al., A mutation in the Pax-6 gene in rat small eye is associated with impaired migration of midbrain crest cells. Nat Genet, 1993. 3(4): p. 299-304. 294. Brunjes, P.C., M. Fisher, and R. Grainger, The small-eye mutation results in abnormalities in the lateral cortical migratory stream. Brain Res Dev Brain Res, 1998. 110(1): p. 121-5. 295. Gonzalez-Salinas, R. and F. Pena-Guani, Combined cataract phacoemulsification and aniridia endocapsular rings implantation in a patient with bilateral congenital aniridia and cataract: A case report. Saudi J Ophthalmol, 2016. 30(1): p. 64-7. 296. Ramaesh, K., et al., Evolving concepts on the pathogenic mechanisms of aniridia related keratopathy. Int J Biochem Cell Biol, 2005. 37(3): p. 547-57. 297. Sivak, J.M., et al., Pax-6 expression and activity are induced in the reepithelializing cornea and control activity of the transcriptional promoter for matrix metalloproteinase gelatinase B. Dev Biol, 2000. 222(1): p. 41-54. 298. Verma, R.P. and C. Hansch, Matrix metalloproteinases (MMPs): chemical-biological functions and (Q)SARs. Bioorg Med Chem, 2007. 15(6): p. 2223-68. 299. Brew, K. and H. Nagase, The tissue inhibitors of metalloproteinases (TIMPs): an ancient family with structural and functional diversity. Biochim Biophys Acta, 2010. 1803(1): p. 55-71. 300. Sivak, J.M. and M.E. Fini, MMPs in the eye: emerging roles for matrix metalloproteinases in ocular physiology. Progress in Retinal and Eye Research, 2002. 21(1): p. 1-14. 301. Wolf, K., et al., Physical limits of cell migration: control by ECM space and nuclear deformation and tuning by proteolysis and traction force. J Cell Biol, 2013. 201(7): p. 1069-84. 302. Daniels, J.T., et al., Matrix Metalloproteinase Inhibition Modulates Fibroblast-Mediated Matrix Contraction and Collagen Production In Vitro. Investigative Ophthalmology & Visual Science, 2003. 44(3): p. 1104-1110.

278

303. Kamiyama, K., et al., Effects of PDGF on the migration of rabbit corneal fibroblasts and epithelial cells. Cornea, 1998. 17(3): p. 315-25. 304. Mukhey, D., et al., Controlling human corneal stromal stem cell contraction to mediate rapid cell and matrix organization of real architecture for 3-dimensional tissue equivalents. Acta Biomater, 2018. 67: p. 229-237. 305. Ramaesh, T., et al., Increased apoptosis and abnormal wound-healing responses in the heterozygous Pax6(+/-) mouse cornea. Invest Ophthalmol Vis Sci, 2006. 47(5): p. 1911- 1917. 306. Urrutia, G., et al., PAX6 Promoter Methylation Correlates with MDA-MB-231 Cell Migration, and Expression of MMP2 and MMP9. Asian Pac J Cancer Prev, 2018. 19(10): p. 2859-2866. 307. Ohuchi, E., et al., Membrane type 1 matrix metalloproteinase digests interstitial collagens and other extracellular matrix macromolecules. J Biol Chem, 1997. 272(4): p. 2446-51. 308. Ye, H.Q., et al., Differential expression of MT1-MMP (MMP-14) and collagenase III (MMP-13) genes in normal and wounded rat corneas. Invest Ophthalmol Vis Sci, 2000. 41(10): p. 2894-9. 309. Azar, D.T., et al., Effect of MT1-MMP deficiency and overexpression in corneal keratocytes on vascular endothelial cell migration and proliferation. Curr Eye Res, 2008. 33(11): p. 954-62. 310. Han, K.Y., et al., MT1-MMP modulates bFGF-induced VEGF-A expression in corneal fibroblasts. Protein Pept Lett, 2012. 19(12): p. 1334-9. 311. Chang, J.H., et al., Matrix metalloproteinase 14 modulates signal transduction and angiogenesis in the cornea. Surv Ophthalmol, 2016. 61(4): p. 478-97. 312. Du, H.T. and P. Liu, Matrix metalloproteinase 14 participates in corneal lymphangiogenesis through the VEGF-C/VEGFR-3 signaling pathway. Exp Ther Med, 2016. 12(4): p. 2120-2128. 313. Zhou, Z., et al., Impaired endochondral ossification and angiogenesis in mice deficient in membrane-type matrix metalloproteinase I. Proc Natl Acad Sci U S A, 2000. 97(8): p. 4052-7. 314. Sivak, J.M., et al., Transcription factors Pax6 and AP-2 alpha interact to coordinate corneal epithelial repair by controlling expression of matrix metalloproteinase gelatinase B. Molecular and Cellular Biology, 2004. 24(1): p. 245-257. 315. Sivak, J.M., et al., Pax-6 expression and activity are induced in the reepithelializing cornea and control activity of the transcriptional promoter for matrix metalloproteinase gelatinase B. Developmental Biology, 2000. 222(1): p. 41-54. 316. Fini, M.E., et al., Role of matrix metalloproteinases in failure to re-epithelialize after corneal injury. American Journal of Pathology, 1996. 149(4): p. 1287-1302. 317. Mulholland, B., S.J. Tuft, and P.T. Khaw, Matrix metalloproteinase distribution during early corneal wound healing. Eye, 2005. 19(5): p. 584-588. 318. Mohan, R., et al., Matrix metalloproteinase gelatinase B (MMP-9) coordinates and effects epithelial regeneration. Journal of Biological Chemistry, 2002. 277(3): p. 2065- 2072. 319. Brejchova, K., et al., Matrix metalloproteinases in recurrent corneal melting associated with primary Sjorgen's syndrome. Mol Vis, 2009. 15(253): p. 2364-2372. 320. Stuelten, C.H., et al., Breast cancer cells induce stromal fibroblasts to express MMP-9 via secretion of TNF-alpha and TGF-beta. Journal of Cell Science, 2005. 118(10): p. 2143- 2153. 321. Perng, D.W., et al., Matrix metalloprotease-9 induces transforming growth factor- beta(1) production in airway epithelium via activation of epidermal growth factor receptors. Life Sci, 2011. 89(5-6): p. 204-12. 279

322. Kobayashi, T., et al., Matrix metalloproteinase-9 activates TGF-beta and stimulates fibroblast contraction of collagen gels. Am J Physiol Lung Cell Mol Physiol, 2014. 306(11): p. L1006-15. 323. Overall, C.M., J.L. Wrana, and J. Sodek, Independent regulation of collagenase, 72-kDa progelatinase, and metalloendoproteinase inhibitor expression in human fibroblasts by transforming growth factor-beta. J Biol Chem, 1989. 264(3): p. 1860-9. 324. Cheema, U., et al., Oxygen diffusion through collagen scaffolds at defined densities: implications for cell survival in tissue models. Journal of Tissue Engineering and Regenerative Medicine, 2012. 6(1): p. 77-84. 325. Girard, M.T., et al., Stromal fibroblasts synthesize collagenase and stromelysin during long-term tissue remodeling. Journal of Cell Science, 1993. 104 ( Pt 4): p. 1001-11. 326. Matsubara, M., et al., Differential roles for two gelatinolytic enzymes of the matrix metalloproteinase family in the remodelling cornea. Developmental Biology, 1991. 147(2): p. 425-39. 327. Azar, D.T., et al., Gelatinase B and A expression after laser in situ keratomileusis and photorefractive keratectomy. Archives of Ophthalmology, 1998. 116(9): p. 1206-1208. 328. da Rocha-Azevedo, B. and F. Grinnell, Fibroblast morphogenesis on 3D collagen matrices: the balance between cell clustering and cell migration. Exp Cell Res, 2013. 319(16): p. 2440-6. 329. Kuhn, C. and J.A. McDonald, The roles of the myofibroblast in idiopathic pulmonary fibrosis. Ultrastructural and immunohistochemical features of sites of active extracellular matrix synthesis. Am J Pathol, 1991. 138(5): p. 1257-65. 330. da Rocha-Azevedo, B., C.H. Ho, and F. Grinnell, PDGF‑stimulated dispersal of cell clusters and disruption of fibronectin matrix on three-dimensional collagen matrices requires matrix metalloproteinase-2. Mol Biol Cell, 2015. 26(6): p. 1098-105. 331. Stringa, E., et al., Collagen degradation and platelet-derived growth factor stimulate the migration of vascular smooth muscle cells. J Cell Sci, 2000. 113 ( Pt 11): p. 2055-64. 332. Matsuo, T., et al., A Mutation in the Pax-6 Gene in Rat Small-Eye Is Associated with Impaired Migration of Midbrain Crest Cells. Nat Genet, 1993. 3(4): p. 299-304. 333. Lwigale, P.Y., P.A. Cressy, and M. Bronner-Fraser, Corneal keratocytes retain neural crest progenitor cell properties. Dev Biol, 2005. 288(1): p. 284-93. 334. Dos Santos, A., et al., Differentiation Capacity of Human Mesenchymal Stem Cells into Keratocyte Lineage. Invest Ophthalmol Vis Sci, 2019. 60(8): p. 3013-3023. 335. Hertsenberg, A.J., et al., Corneal stromal stem cells reduce corneal scarring by mediating neutrophil infiltration after wounding. PLoS One, 2017. 12(3): p. e0171712. 336. Orozco Morales, M.L., et al., Anti-inflammatory potential of human corneal stroma- derived stem cells determined by a novel in vitro corneal epithelial injury model. World J Stem Cells, 2019. 11(2): p. 84-99. 337. Jester, J.V., et al., Myofibroblast Differentiation of Normal Human Keratocytes and hTERT, Extended-Life Human Corneal Fibroblasts. Investigative Ophthalmology & Visual Science, 2003. 44(5): p. 1850-1858. 338. Buss, D.G., et al., Isolation and cultivation of equine corneal keratocytes, fibroblasts and myofibroblasts. Vet Ophthalmol, 2010. 13(1): p. 37-42. 339. Jester, J.V., et al., Temporal, 3-dimensional, cellular anatomy of corneal wound tissue. J Anat, 1995. 186 ( Pt 2)(Pt 2): p. 301-11. 340. Harvey, S.A.K., et al., Responses of Cultured Human Keratocytes and Myofibroblasts to Ethyl Pyruvate: A Microarray Analysis of Gene Expression. Investigative Ophthalmology & Visual Science, 2010. 51(6): p. 2917-2927. 341. Massie, I., H.J. Levis, and J.T. Daniels, Response of human limbal epithelial cells to wounding on 3D RAFT tissue equivalents: effect of airlifting and human limbal fibroblasts. Exp Eye Res, 2014. 127: p. 196-205.

280

342. Wilson, S.E., J.J. Liu, and R.R. Mohan, Stromal-epithelial interactions in the cornea. Prog Retin Eye Res, 1999. 18(3): p. 293-309. 343. Ljubimov, A.V. and M. Saghizadeh, Progress in corneal wound healing. Prog Retin Eye Res, 2015. 49: p. 17-45. 344. Ramos-Mejia, V., et al., Residual expression of the reprogramming factors prevents differentiation of iPSC generated from human fibroblasts and cord blood CD34+ progenitors. PLoS One, 2012. 7(4): p. e35824. 345. Elsen, G.E., et al., The Epigenetic Factor Landscape of Developing Neocortex Is Regulated by Transcription Factors Pax6→ Tbr2→ Tbr1. Front Neurosci, 2018. 12: p. 571. 346. Sun, J., et al., Pax6 associates with H3K4-specific histone methyltransferases Mll1, Mll2, and Set1a and regulates H3K4 methylation at promoters and enhancers. Epigenetics Chromatin, 2016. 9(1): p. 37. 347. Souzeau, E., et al., PAX6 molecular analysis and genotype-phenotype correlations in families with aniridia from Australasia and Southeast Asia. Mol Vis, 2018. 24: p. 261- 273. 348. Ooki, A., et al., Epigenetically regulated PAX6 drives cancer cells toward a stem-like state via GLI-SOX2 signaling axis in lung adenocarcinoma. Oncogene, 2018. 37(45): p. 5967- 5981. 349. Lee, H.J. and K.A. Colby, A review of the clinical and genetic aspects of aniridia. Semin Ophthalmol, 2013. 28(5-6): p. 306-12. 350. Kaur, G. and J.M. Dufour, Cell lines: Valuable tools or useless artifacts. Spermatogenesis, 2012. 2(1): p. 1-5. 351. Li, W., et al., Down-regulation of Pax6 is associated with abnormal differentiation of corneal epithelial cells in severe ocular surface diseases. The Journal of Pathology, 2008. 214(1): p. 114-122. 352. Hayashi, R., et al., Generation of corneal epithelial cells from induced pluripotent stem cells derived from human dermal fibroblast and corneal limbal epithelium. PLoS One, 2012. 7(9): p. e45435. 353. Mikhailova, A., et al., Small-molecule induction promotes corneal epithelial cell differentiation from human induced pluripotent stem cells. Stem Cell Reports, 2014. 2(2): p. 219-31. 354. Chambers, S.M., et al., Dual-SMAD Inhibition/WNT Activation-Based Methods to Induce Neural Crest and Derivatives from Human Pluripotent Stem Cells. Methods Mol Biol, 2016. 1307: p. 329-43. 355. Naylor, R.W., et al., Derivation of Corneal Keratocyte-Like Cells from Human Induced Pluripotent Stem Cells. PLoS One, 2016. 11(10): p. e0165464. 356. Joseph, R., O.P. Srivastava, and R.R. Pfister, Modeling Keratoconus Using Induced Pluripotent Stem Cells. Invest Ophthalmol Vis Sci, 2016. 57(8): p. 3685-97. 357. Takahashi, K. and S. Yamanaka, Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell, 2006. 126(4): p. 663-76. 358. Zhou, T., et al., Generation of induced pluripotent stem cells from urine. J Am Soc Nephrol, 2011. 22(7): p. 1221-8. 359. gal, R.L., et al., The effect of donor age on corneal transplantation outcome results of the cornea donor study. Ophthalmology, 2008. 115(4): p. 620-626.e6.

281